Adaptation of soil microbial growth to temperature: Using a tropical elevation gradient to predict future changes

Abstract Terrestrial biogeochemical feedbacks to the climate are strongly modulated by the temperature response of soil microorganisms. Tropical forests, in particular, exert a major influence on global climate because they are the most productive terrestrial ecosystem. We used an elevation gradient across tropical forest in the Andes (a gradient of 20°C mean annual temperature, MAT), to test whether soil bacterial and fungal community growth responses are adapted to long‐term temperature differences. We evaluated the temperature dependency of soil bacterial and fungal growth using the leucine‐ and acetate‐incorporation methods, respectively, and determined indices for the temperature response of growth: Q 10 (temperature sensitivity over a given 10oC range) and T min (the minimum temperature for growth). For both bacterial and fungal communities, increased MAT (decreased elevation) resulted in increases in Q 10 and T min of growth. Across a MAT range from 6°C to 26°C, the Q 10 and T min varied for bacterial growth (Q 10–20 = 2.4 to 3.5; T min = −8°C to −1.5°C) and fungal growth (Q 10–20 = 2.6 to 3.6; T min = −6°C to −1°C). Thus, bacteria and fungi did not differ significantly in their growth temperature responses with changes in MAT. Our findings indicate that across natural temperature gradients, each increase in MAT by 1°C results in increases in T min of microbial growth by approximately 0.3°C and Q 10–20 by 0.05, consistent with long‐term temperature adaptation of soil microbial communities. A 2°C warming would increase microbial activity across a MAT gradient of 6°C to 26°C by 28% to 15%, respectively, and temperature adaptation of microbial communities would further increase activity by 1.2% to 0.3%. The impact of warming on microbial activity, and the related impact on soil carbon cycling, is thus greater in regions with lower MAT. These results can be used to predict future changes in the temperature response of microbial activity over different levels of warming and over large temperature ranges, extending to tropical regions.


| INTRODUCTION
Soil microorganisms regulate terrestrial biogeochemical cycles, and their response to temperature is a critical factor in regulating feedbacks associated with climate warming .
Models have demonstrated that the nature of temperature-adaptive responses in soil microbial physiology, community composition, enzyme function or growth, may have major influences on atmospheric CO 2 accumulation in the 21st century (Wieder, Bonan, & Allison, 2013). Of all terrestrial ecosystems, tropical forests exert the largest influence on global climate because they are the most productive and have the highest respiration rates (Beer et al., 2010;Pan et al., 2011), in addition to containing the highest biomass of soil microorganisms (Serna-Chavez, Fierer, & Bodegom, 2013). It is surprising, therefore, that we have a limited understanding of the temperature response of soil microbial communities in these ecosystems.
Research on the temperature response of soil organic matter cycling has been extensive, albeit concentrated outside the tropics, but a consensus remains elusive (Conant et al., 2011;Karhu et al., 2014;Kirschbaum, 2006). The focus of this work has often been on the temperature response of respiration, in the context of its potential impact as a positive feedback on climate warming  and the potential for a temperature-adaptive response of the microbial community in affecting this feedback (Bradford et al., 2008;Karhu et al., 2014). Such an adaptation response has been defined as a change in microbial community composition, physiology or enzyme function, which has a net result of metabolism being better-optimized to a given temperature (Bárcenas-Moreno, Gomez-Brandon, Bradford, 2013). The lack of consensus among studies on the temperature response of microbial activity arises partly because respiration, a commonly measured index of microbial activity, has an "apparent" temperature sensitivity that is influenced by multiple environmental variables ("indirect effects") that vary among soils . It will also be affected indirectly by factors other than the temperature regime, such as substrate availability and moisture . To isolate the direct effect of temperature, one must estimate the "intrinsic" temperature sensitivities of specific processes such as carbon-assimilation, enzyme activities and growth, which are independent of these indirect effects. These intrinsic temperature sensitivities can be assessed in controlled short-term incubation experiments (Kirschbaum, 2006), providing standard reproducible information on microbial temperature responses which are then comparable across biomes.
Another reason for the lack of consensus is due to the different ways in which temperature responses are modelled. The temperature sensitivity of microbial processes, such as growth and respiration, has been described using various metrics. A commonly used parameter is the Q 10 value, which represents the ratio of a process at (T + 10°C)/T, where T = standard reference temperature. However, comparison of Q 10 among studies requires careful consideration of the differences in temperature range and reference temperatures used for its calculation, because Q 10 is not constant over a given temperature range. The Q 10 of respiration and microbial growth is higher when determined at lower temperatures (Bååth, 2018;Kirschbaum, 2006;Lloyd & Taylor, 1994) and models often incorporate a higher Q 10 for lower temperatures (Del Grosso et al., 2005;Jenkinson, Adams, & Wild, 1991). This dependency of Q 10 on the measurement temperature range makes it difficult to compare Q 10 among studies that use different temperature ranges in their calculations.
The measurement of Q 10 over a large temperature range, assuming constant Q 10 , can also introduce problems in predicting the effects of temperature on respiration and growth (Bååth, 2018). Temperature dependency is also often modelled using an Arrhenius relation- energy, R = universal gas constant, and T = absolute temperature.
Here, the activation energy (E a ) determines the temperature sensitivity. However, since there is a close relationship between E a and Q 10 within the range of temperatures normally found in soils (Raven & Geider, 1988), E a has the same problems in interpretation and determination as Q 10 .
An alternative approach that can be used to characterize respiration and growth, while below its temperature optimum (T opt ), is the use of the square root relationship: A 0.5 = a × (T-T min ), where A is activity (e.g., growth or respiration), T min the apparent minimum temperature for activity (°C), and a is a slope parameter related to absolute activity (Ratkowsky, Olley, Mcmeekin, & Ball, 1982) (Figure 1).
The square root relationship, also called the Ratkowsky equation, has been widely used to model the rate of bacterial growth in water (Bell & Ahlgren, 1987;Li & Dickie, 1987) and bacterial and fungal growth in soil (Dıaz-Raviña, Frostegård, & Bååth, 1994;Pietikäinen, Pettersson, & Bååth, 2005;Rinnan, Rousk, Yergeau, Kowalchuk, & Bååth, 2009;van Gestel, Reischke, & Bååth, 2013). It has also been shown to be an adequate representation of the temperature responses of respiration and decomposition (Bååth, 2018;Kätterer, Reichstein, Andren, & Lomander, 1998;Pietikäinen et al., 2005), and it has been shown that T min increases and decreases following community temperature adaptation to the thermic environment (Bååth, 2018;Bárcenas-Moreno et al., 2009). Because it is independent of the temperature range at which it is calculated, it can be compared more easily among studies than E a or Q 10 . By determining T min following the square root model, we can obtain information on the temperature responses of microbial growth or respiration, which can be related with other process rates such as enzyme kinetics (E a , Q 10 ). The T min metric, therefore, provides information on the community-level adaptation to temperature, which can be easily compared across biomes and can be used to predict future effects of climate change.
Biogeographic variation in the T min for microbial growth and respiration in soils from different ecosystems has been tentatively estimated as −10 to −15°C in Arctic/Antarctic regions, −5 and −10°C in temperate regions and 0 to −5°C in tropical regions (Pietikäinen et al., 2005;Rinnan et al., 2009;van Gestel et al., 2013). Furthermore, by combining several studies on the effect of mean annual temperature (MAT) on T min for bacterial growth, it has been predicted that a 1°C increase in MAT would result in an increased T min of between 0.2 and 0.3°C (Rousk, Frey, & Bååth, 2012). Bååth (2018) also predicted, using a tentative global envelope of T min in soils, that a 1°C increase in MAT would result in an increased T min of around 0.3°C. However, the temperature sensitivity of soil microbial growth has been studied across only a limited MAT range. For example, Rinnan et al. (2009) studied soils ranging from −4°C to +9°C, while Rousk et al. (2012) compared a MAT of 7°C with an artificially heated treatment with a MAT of 12°C. To be able to predict changes in temperature sensitivity over a major part of the global MAT variation (i.e., −15 to 30°C), a larger range needs to be tested, of course also including tropical regions.
The temperature sensitivity of bacterial growth in soil has been reasonably well studied (Rinnan et al., 2009;Rousk et al., 2012;van Gestel et al., 2013), but this is not the case for fungal growth. Only two earlier studies, in soils and only from the temperate zone, have compared T min for fungi and bacteria (Birgander, Olsson, & Rousk, 2018;Pietikäinen et al., 2005). Similar but slightly lower T min for fungal compared to bacterial growth was found, suggesting fungi to be better adapted to low temperature conditions. However, more studies are needed to test this further, covering a larger variation in MAT.
Elevation gradients on mountainsides have been used to understand plant biogeography by ecologists since the 18th century (Linnaeus, 1781; von Humboldt & Bonpland, 1805), but more recently they have been used as powerful tools to understand how climate change affects plant and microbial ecology Sundqvist, Sanders, & Wardle, 2013), by revealing the long-term temperature acclimation or adaptive changes in plant physiology, soil processes and soil microbial composition (Giardina, Litton, Crow, & Asner, 2014;Girardin et al., 2014;Nottingham, Fierer, et al., 2018). Here, we used a 3.5 km elevation gradient in Peru to explore the long-term temperature adaptation of bacterial and fungal growth to a 20°C gradient in MAT. We tested the following hypotheses: (a) Increasing MAT (i.e., decreasing elevation) will increase the temperature optima for bacterial growth, resulting in higher T min , temperature sensitivity index (growth at 35°C/0°C) in lower elevation sites with higher MAT; (b) T min will increase around 0.2 to 0.4 per degree Celsius increase in MAT, equivalent to around 0.05 higher Q 10-20 per degree increase in MAT; and (c) based on earlier results in temperatezone studies (Pietikäinen et al., 2005), we hypothesize a lower temperature sensitivity index for fungi than bacteria. Since our results cover a gradient in MAT from 6 to 26°C, our data can be used to predict future changes in the temperature sensitivity of microbial growth and activity in soil over a large range of MAT.  Information Table S1).

| Study sites
Mean annual temperature (MAT) decreases with increasing elevation across the transect (dropping from 26 to 6°C; Supporting Information Figure S1). There is little variation in seasonal temperature across the gradient, with mean daily air temperature differing only by about 4°C between warmest and coolest month, irrespective of elevation, although diurnal variation can increase this range slightly . Mean annual precipitation (MAP) is consistently high and does not vary consistently with elevation, ranging from 1,506 to 5,302 mm/year among the sites . F I G U R E 1 A comparison of the temperature sensitivity of growth for two hypothetical microbial communities, plotted with a square root transformation. Community A (full line, red) is low temperature-adapted (T min = −7.3°C), and community B (dashed line, black) is high temperature-adapted (T min = −4.3°C). Three indices of temperature sensitivity are shown. T min was determined by linear regression (thick lines) using measurements at 20 and 4°C (thin stippled vertical lines) and extrapolation, where T min for community A < T min for community B. Q 10-20 (see Methods) was calculated using the same regression, where Q 10-20 for community A <Q 10-20 for community B. A temperature sensitivity index (log 35/0) was determined by the log of the ratio of growth at 35°C and 0°C (thin vertical lines), that is log [(A/a) 2 ] for community A and log [(B/b) 2 ] for community B. The temperature sensitivity index log 35/0 for community A <log 35/0 for community B [Colour figure can be viewed at wileyonlinelibrary.com] elsewhere (Nottingham, Fierer, et al., 2018;Fyllas et al., 2017;Whitaker et al., 2014;van de Weg, Meir, Grace, & Atkin, 2009).

| Soil sampling and analyses
For all sites, soil samples were collected during November 2011.
These ecosystems are highly aseasonal, with no significant intra-annual variation in mean monthly temperature  and no evidence of seasonal soil or plant moisture constraints (van de Weg et al., 2014;Zimmermann, Meir, Bird, Malhi, & Ccahuana, 2010); therefore, the comparison of soil properties for these sites at a single time point was approximated as representative of patterns likely to be found throughout the year. Furthermore, temperature seasonality has earlier been shown to have no or very little effect on the temperature-growth response of bacterial communities, even in soils with a large amplitude in temperature over the year (Birgander et al., 2018;van Gestel et al., 2013). We collected soil from four corner subplots and a central subplot, within each of the 1 ha permanent study plots at each elevation site, with soil from these subplots

| Bacterial and fungal growth
Temperature sensitivity of microbial growth was determined by measuring instantaneous growth of bacteria and fungi at different temperatures, as earlier used by Pietikäinen et al. (2005). Bacterial growth was estimated using the leucine (Leu) incorporation method, while fungal growth was estimated using the acetate-in-ergosterol (Ac-in-erg)-incorporation method (Bååth, 2001;Rousk & Bååth, 2011). Since many samples were processed (14 soils × 5 replicates = 70 soil samples), microbial growth was measured for all soils at one temperature on separate days. This experimental design was suitable to determine relative changes in temperature sensitivity with differences in MAT between sites.
The growth rate of bacteria was estimated using the leucine incorporation method, following Bååth, Pettersson, and Söderberg (2001). Briefly, soil samples (1 g fresh weight) were vortexed with 20 ml distilled H 2 O for 3 min. and then centrifuged at 1,000 g for 10 min. Aliquots (1.5 ml) of the resulting suspension were transferred to 2 ml tubes and 2 µl [ 3 H]Leu (37 MBq/ml and 5.74 TBq/ mmol) combined with unlabelled Leu, resulting in 275 nM Leu in the bacterial suspensions. After incubation at the desired temperature (in water baths), the reaction was terminated with 75 µl 100% trichloroacetic acid. Incubation time was modified according to the incubation temperature to compensate for lower incorporation at low temperatures (Pietikäinen et al., 2005), with 24 hr for 0 and 4°C, and 2 hr for 20 and 35°C. Washing of the samples and measurement of incorporated radioactivity was performed following Bååth et al. (2001).
The growth rate of fungi was estimated using the acetate-in-ergosterol-incorporation method (Newell & Fallon, 1991) adapted for soil (Bååth, 2001), with modifications. Briefly, soil samples (1 g fresh weight) were transferred to test tubes to which 20 µl [1-14 C] acetic acid (sodium salt; 7.4 MBq/ml and 2.04 GBq/mmol) with unlabelled sodium acetate, and 1.5 ml distilled H 2 O, resulting in a final acetate concentration of 220 µM. The resulting soil slurry was then incubated in the dark (for twice as long as the corresponding samples used for bacterial growth), after which 1 ml formalin (5%) was added to terminate the reaction. Ergosterol was then extracted, separated and quantified using high-performance liquid chromatography and a UV detector (282 nm). The ergosterol peak was collected, and the amount of incorporated radioactivity was determined.

| Calculation of T min , temperature sensitivity index and Q 10
T min was calculated using growth at 4°C and 20°C, assuming a straight-line relationship for the squared growth rates versus temperature ( Figure 1), according to the Ratkowsky equation (Ratkowsky et al., 1982): where T min (°C) is the apparent minimum temperature for growth, T is the measurement temperature (in our case 4 and 20°C), and a is a slope parameter related to the absolute growth rate. Since T min will always be determined by extrapolation, an alternative temperature sensitivity index, log 35/0 was defined as the log ratio of growth at 35°C and 0°C (Figure 1). A similar ratio was suggested by Bárcenas-Moreno et al. (2009) as a rapid and sensitive way to study changes in temperature sensitivity and has been shown to correlate with T min (Rinnan et al., 2009). We chose a large temperature range for the temperature sensitivity index (i.e., 35°C and 0°C), to accommodate the variation expected by the large range in MAT between sites.
Means and SE were calculated for each site (n = 5 replicate soil samples per site). Regressions against temperature for the different growth indices were then made using mean values per site (n = 14), since they were the independent samples. For indices of fungal growth, sample size (n) was 13 because one site (TC, high elevation grassland) had activity values which were too low to be able to calculate temperature sensitivity.
We calculated Q 10 for the 10 to 20°C range (Q 10-20 ) using the mean T min values for each site, according to the equation: where R = the temperature range (for Q 10-20 , R = 10) and T L is the lowest temperature in the range (e.g., for Q 10-20 T L = 10) (Bååth, 2018). To calculate Q 10 at MAT±5°C (Q MAT±5 ), we modified the equation where T L = (MAT−5), with MAT from Supporting Information Table S1. To estimate activity with an increase in MAT of 2°C (representing the lowest end of the range of predicted global MAT increase by 2100; 2-6°C; IPCC, 2013), we used R = 2 and T L = MAT.
We further modified Equation (2) to account for temperature adaptation-according to our finding that thermal adaptation of growth led to an increase in T min by 0.6°C per 2°C increase in MAT-by replacing T min with (T min + 0.6).

| Soil properties
Increased elevation was highly negatively correlated with a decrease in MAT (R 2 = 0.99, p < 0.001; Supporting Information Figure S1), such that elevation and MAT were interchangeable as explanatory variables. Increased elevation was associated with an increase in total C (R 2 = 0.46, p = 0.008) and total N (R 2 = 0.51, p = 0.004); small but nonsignificant increases in total P (R 2 = 0.18, p = 0.13) and extractable P (R 2 = 0.22, p = 0.09). Soil pH did not vary with elevation (R 2 = 0.001, p = 0.90) (Supporting Information Table S1). Further detail on soil and microbial community properties for these sites, including in organic horizons, is provided elsewhere Whitaker et al., 2014).

| Bacterial growth
Higher varies with the temperature range used for its calculation (Bååth, 2018;Kirschbaum, 2000), we calculated Q 10 only for one intermediate range of temperatures (between 10 and 20°C). Q 10-20 increased from approximately 2.4 at a MAT of 6°C to almost 3.5 at a MAT of 26°C (R 2 = 0.93, Figure 2c). This translates to an increase in Q 10-20 of 0.055 ± 0.004 per 1 degree of increase in MAT for this temperature range.

| Fungal growth
The temperature sensitivity of fungal growth was affected by site

| Fungal/bacterial relationships
There was a significant decrease in the log ratio of bacterial to fungal growth with increased MAT (R 2 = 0.53, p < 0.01, Figure 4a), with the ratios approximately five times lower in soil from low elevation (high MAT) compared to soil from high elevation (low MAT). The T min for fungal and bacterial growth was linearly related with no significant difference from a 1:1 line (R 2 = 0.47, p < 0.05, Figure 4b), indicating that fungal and bacterial community responses were similar over the gradient in MAT studied here (6-26°C). This similarity was further indicated by nonsignificant differences between bacteria and fungi for changes in T min and Q 10-20 per 1 degree of increase in MAT.

| Predicting future changes
We compared Q 10 for the three temperature intervals 5-15°C, where T min does not respond to warming, the increase in microbial activity with warming was largest in soils from high elevation and low MAT. For example, microbial activity was predicted to increase by 27% in the four sites with lowest MAT and by 15% in the sites with highest MAT. When we accounted for an adaptation response, whereby T min increased by 0.6°C per 2°C increase in MAT (Figures 2   and 3a), these predicted increases in microbial activity with warming were slightly higher, with the largest increases for soils from high elevation (low MAT). For example, the predicted increase in "temperature-adapted" soil microbial activity with a 2°C warming was 28% in the four sites with lowest MAT (an increase of 1.2% points compared to no adaptation), while at a MAT of 26°C, the predicted increase was only 15% (an increase of 0.3% points compared to no adaptation; Figure 5c). The thin line at a value of 2 for Q 10 indicates an approximately upper limit for Q 10 for most enzyme activities (V max ) for these sites (Nottingham et al., 2016). (c) Predicted increases in growth with a 2°C increase in MAT calculated using T min from Figure 2a. "No adaptation" was calculated using Equation (2), while "adaptation" was calculated assuming a 0.6°C increase in T min 4 | DISCUSSION

| Adaptation to MAT along the gradient
Our main finding, along the 3.5 km tropical elevation gradient with a 20°C change in MAT, was that the temperature response of microbial growth (T min ) is determined by MAT. Our results suggest that an increase in MAT by 1°C will result in an increased T min by approximately 0.3°C (and Q 10-20 by 0.05 units) for bacteria and fungi. This outcome is consistent with our second hypothesis. This also provides further evidence for long-term temperature adaptation of soil microbial growth (hypothesis 1) and the first information for both bacteria and fungi across such a large MAT range and in the tropical biome.
Our key finding that the long-term temperature adaptation of

| Comparing temperature effects on bacterial and fungal growth
Our results showed that bacterial growth and fungal growth respond similarly to temperature differences, contrary to our third hypothesis that fungi would be better adapted to lower temperatures (

| Application of T min and the square root equation for respiration and growth
In studies of respiration along the same elevation transect, Q 10-20 of soil respiration varied between 2.1 and 6.9 (Zimmermann, Meir, Bird, Malhi, & Ccahuana, 2009), which is equivalent to a T min variation of −12.3 to +3.9°C, assuming a square root relationship (equation 4 in Bååth, 2018). This is a similar range, albeit slightly larger, to that found for microbial growth. However, the variation in Q 10 for respiration was not related to elevation (Zimmermann et al., 2009), suggesting no temperature adaptation of respiration. However, only four sites were studied in Zimmermann et al. (2009), with large variations in estimates of Q 10 of soil respiration, which likely reflected different temperature responses of soil and root-derived respiration (Zimmermann et al., 2010). Bååth (2018) argued, using a compilation of a large number of respiration studies (Hamdi, Moyano, Sall, Bernoux, & Chevallier, 2013) and models used to predict respiration (Del Grosso et al., 2005;Jenkinson et al., 1991;Kätterer & Andren, 2001;Kirschbaum, 2000;Lloyd & Taylor, 1994;Svensson et al., 2008), that T min (and Q 10 ) for microbial growth and respiration should be very similar in soils globally, and that both could be described by the square root equation. Although more precise data on respiration-temperature relationships for the present elevation gradient are needed, we suggest that our data on the temperature sensitivity of soil microbial growth will also be relevant for respiration. Specifically, we hypothesize that our result showing that an increase in MAT by 1°C results in an increased T min by approximately 0.3°C (and Q 10-20 by 0.05 units) over the gradient of MAT between 6 and 26°C may also be applicable for soil heterotrophic respiration.
The application of the square root model (using T min ) was suggested as a simple method to quantify the growth or activity response of the microbial community to the temperature regime (Bååth, 2018). This approach is particularly useful because, unlike Q 10 , T min values are not dependent on the temperature range used for their calculation. This was clear in the present study. Thus, the use of one Q 10 value for the range of temperatures found in the present elevation gradient would result in bias when predicting growth and respiration. In contrast, T min as descriptor of temperature adaptation of the community and the square root model to estimate the temperature response can be used to calculate Q 10 for any temperature interval (Bååth, 2018). Thus, T min can be used to calculate a standardized Q 10 in each of the sites studied, for example, Q 10-20 in Figure 2c, but also to calculate a Q 10 related to the MAT at each specific site (Figure 5b).

| Comparing temperature sensitivity of growth and enzyme activity
Before microorganisms can use soil organic matter for growth and respiration, macromolecules must be degraded by extracellular enzymes. The strong temperature-adaptive responses of microbial growth we found across this elevation and MAT gradient (Figures 2 and 3c) occurred despite a largely insensitive temperature response of enzyme activities reported in a previous study of the same gradient (Nottingham et al., 2016). This previous study of enzyme temperature sensitivity found no elevation patterns in the Q 10 of the maximum enzymatic catalytic rate (V max ) for 5 out of 7 soil enzymes, with only small increases in Q 10 for V max with increased elevation for 2 enzymes, ß-glucosidase and ß-xylanase (Nottingham et al., 2016). Studies from other sites are consistent with a general insensitivity of Q 10 of enzymatic V max to temperature. Nine years of experimental soil warming in a temperate forest increased enzymatic V max but did not affect its Q 10 response (Schindlbacher, Schnecker, Takriti, Borken, & Wanek, 2015); while a cross-latitudinal study found no differences in the Q 10 of V max for five hydrolytic enzymes, although a relationship was observed between MAT and the Q 10 of the halfsaturation constant (K m ) of ß-glucosidase (German, Marcelo, Stone, & Allison, 2012). Thus, the temperature responses of growth do not appear to be principally the result of differences in enzyme function, and there appear to be differences both in the applicable model and in the adaptation responses, for enzyme activity, growth and respiration. Enzymatic activity usually follows a strict Arrhenius relationship with temperature (Davidson, Janssens, & Luo, 2006;German et al., 2012), with very little increase in Q 10 at decreasing temperature.
Furthermore, Q 10 values of enzyme activity are often ≤2, irrespective of temperature (Nottingham et al., 2016;Schindlbacher et al., 2015). A comparison of the temperature sensitivities (Q 10 ) of MATrelevant microbial growth and enzyme activities (Figure 5b), indicates that enzyme activities have a lower Q 10 over the whole range of MAT, with the discrepancy increasing at lower temperatures. Overall, these results reinforce the need to understand intrinsic temperature responses of discrete biochemical processes-microbial growth, respiration and enzymatic activity-which together determine the temperature response of the overall C balance (Conant et al., 2011;. Thus, using this model based on Q MAT±5 , we can conceptualize the microbial growth response to warming as the result of two counteracting effects: the direct temperature effect according to the Q 10 trajectory at a fixed T min and the adaptation effect in changing the T min (and thereby altering the Q 10 trajectory; Figure 6). Using data from Figure 2a for bacterial growth, a soil with MAT of 6°C will have a T min of −8°C and Q 10 will vary with temperature according to the −8°C trajectory: decreasing with increasing temperature. However, we can include our results for long-term temperature adaptation of microbial growth, 0.3°C increase in T min per degree increase in MAT (Figures 2 and 3). By including temperature adaptation in this model, +6°C warming will increase T min by around 2°C and alter the Q 10 trajectory to one where microbial growth is slightly higher at the new temperature regime (dashed red arrow; Figure 6). suggesting that with the same predicted increase in MAT, the relative effect will be stronger at lower temperatures. The relatively greater impact at lower MAT may be further exacerbated because greater warming is predicted in higher-latitude ecosystems (IPCC, 2013), although significant impacts in tropical regions could occur if MAT exceeds thermal optima for activity (T opt ).

|
Because the elevation gradient under study here is largely aseasonal in temperature , we could use MAT and one single Q MAT±5 to characterize the temperature adaptation of microbial growth. This will not be the case in ecosystems with large seasonal temperature variation, including deserts (van Gestel et al., 2013) or temperate and continental climates, where Q 10 will vary seasonally (Bååth, 2018). However, by determining T min and using the square root equation, it is straightforward to model the effect of seasonal temperature variation, as shown by Rousk et al. (2012). Similar to our study, their data also suggested that the effect of temperature adaptation was minor compared to the effect of seasonal temperature variation, in determining the Q 10 of microbial activity (cf Figures 5c, 6).
Our results demonstrate consistent patterns of temperature adaptation (T min ) in growth across a large temperature range. Our results also show how T min can be used as a single descriptor of temperature adaption of the microbial community, which together with the square root equation (Ratkowsky et al., 1982) can be used to predict temperature effects on microbial growth. However, in order to fully understand climate warming impacts on microbial communities and the carbon balance, further studies are required on the responses of microbial carbon use efficiency (carbon uptake, growth and respiration) (Bradford, 2013) and on the intrinsic temperature responses of, and the interactions between, different physical, biological and chemical components of the soil carbon cycle (Conant et al., 2011;. A major outstanding question is also whether these microbial growth responses to longterm temperature differences observed along a tropical mountain elevation gradient, will shift either through acclimation or adaptation in response to short-term climatic change.

ACKNOWLEDG EMENTS
This study was led with support from the UK Natural Environment  F I G U R E 6 Conceptual figure on the effect of increasing MAT on Q 10 , calculated using the square root relationship. A soil with MAT of 6°C, and a T min of −8°C, has Q 10~3 (left black arrow). By increasing MAT by 6°C to 12°C (right black arrow), there is a decrease in Q 10 (~2.7) but an increase in T min (by 2°C to −6°C). The new T min trajectory results in a decrease in Q 10 when calculated at the new temperature (Q MAT+2 ; red dashed arrow) but a higher Q 10 if calculated across a fixed temperature range (e.g., Q 0-10 ). The trajectories were calculated using the square root relationship for T min of −6 and −8°C (see Bååth, 2018) [Colour figure can be viewed at wileyonlinelibrary.com]