Heat wave‐induced microbial thermal trait adaptation and its reversal in the Subarctic

Climate change predictions suggest that arctic and subarctic ecosystems will be particularly affected by rising temperatures and extreme weather events, including severe heat waves. Temperature is one of the most important environmental factors controlling and regulating microbial decomposition in soils; therefore, it is critical to understand its impact on soil microorganisms and their feedback to climate warming. We conducted a warming experiment in a subarctic birch forest in North Sweden to test the effects of summer heat waves on the thermal trait distributions that define the temperature dependences for microbial growth and respiration. We also determined the microbial temperature dependences 10 and 12 months after the heat wave simulation had ended to investigate the persistence of the thermal trait shifts. As a result of warming, the bacterial growth temperature dependence shifted to become warm‐adapted, with a similar trend for fungal growth. For respiration, there was no shift in the temperature dependence. The shifts in thermal traits were not accompanied by changes in α‐ or β‐diversity of the microbial community. Warming increased the fungal‐to‐bacterial growth ratio by 33% and decreased the microbial carbon use efficiency by 35%, and both these effects were caused by the reduction in moisture the warming treatments caused, while there was no evidence that substrate depletion had altered microbial processes. The warm‐shifted bacterial thermal traits were partially restored within one winter but only fully recovered to match ambient conditions after 1 year. To conclude, a summer heat wave in the Subarctic resulted in (i) shifts in microbial thermal trait distributions; (ii) lower microbial process rates caused by decreased moisture, not substrate depletion; and (iii) no detectable link between the microbial thermal trait shifts and community composition changes.

and subarctic regions are expected to experience a disproportional amount of weather extremes with rates of temperature increase twice that of other terrestrial ecosystems (Brown et al., 2017;Dobricic et al., 2020;IPCC, 2021).Microorganisms regulate the balance between soil C release and sequestration by controlling C losses via respiration and gains through forming SOM with long residence times via microbial growth (Bardgett et al., 2008;Camenzind et al., 2023;Liang et al., 2017).We know that microbial processes are strongly dependent on and regulated by environmental factors such as temperature (Schipper et al., 2014).Therefore, determining the temperature relationships of microbial respiration and growth can help us understand and anticipate how soil C cycling will feedback to climate change (Frey et al., 2013;García-Palacios et al., 2021;Lloyd & Taylor, 1994;Sierra et al., 2015).
Most ecosystem models use temperature relationships as rate modifiers to determine how temperature regulates microbial rates (Davidson & Janssens, 2006;Sierra et al., 2015).Direct and indirect temperature effects both regulate microbial process rates.
The direct effect of temperature on microbial rates is the biochemical temperature dependence of microbial processes driven by enzyme-catalyzed biochemical reactions (Schipper et al., 2014).
Microbial rates are zero at the theoretical minimum temperature (T min ) and increase with warmer temperatures until maximal rates are reached at the optimal temperature (T opt ).At temperatures exceeding the T opt , rates decrease rapidly until they again reach zero at the theoretical maximum temperature (T max ) (Noll et al., 2020).
The discrimination between direct and indirect temperature effects is challenging yet essential since most ecosystem models rely on the direct temperature dependence (Sierra et al., 2015).
To minimize the indirect effects of temperature, basal microbial process rates (e.g., growth, respiration) can be determined using short-term incubation assays under stable laboratory conditions (Bradford et al., 2008;Fang et al., 2005;Kirschbaum, 1995Kirschbaum, , 2006;;Rinnan et al., 2009).In assessments along climate gradients, the thermal traits that define microbial temperature relationships are cold-shifted in cool ecosystems, and warm-shifted in hot ecosystems (Cruz-Paredes et al., 2023;Nottingham et al., 2019;Rinnan et al., 2009).In laboratory studies, however, microbial temperature relationships could only be experimentally shifted by exposure to warm temperatures at or exceeding the T opt , with no evidence for a similar shift toward cold temperature exposure within 1-2 months of controlled temperatures (Bárcenas-Moreno et al., 2009;Birgander et al., 2013).In addition, in experiments using chronic warming in the field, microbial temperature relationships became warm-shifted (Nottingham et al., 2022;Rousk et al., 2012), while winter warming did not affect the microbial temperature dependences (Birgander et al., 2018).Taken together, it seems that the thermal traits that define microbial temperature dependences are adjusted to the local climate, and can respond to long-term, constant warming, but changes in mean annual temperature induced by milder cool season temperatures (or laboratory exposure to cool temperatures) leave thermal traits unaffected.From this, we arrive at the prediction that it must then be the warm season that most strongly affects soil microbial thermal traits.
Here, we investigated how experimental warming during summer could shift microbial community trait distributions for temperature and how such changes will persist and affect microbial process rates in situ.To do this, we conducted a warming experiment simulating a summer heat wave in the Subarctic in North Sweden.We hypothesized (i) that the microbial temperature relationships would become warm-shifted, increasing T min by 0.2-1.0°Cper 1°C increase in temperature (Bååth, 2018;Rousk et al., 2012); (ii) that warmshifted microbial temperature relationships would also increase the temperature sensitivity (Q 10 ) of microbial rates (Bååth, 2018); (iii) that the shifts in thermal traits would be linked with changes in microbial community composition; (iv) that removing the warming treatment would result in a gradual recovery of microbial temperature relationships to the ambient state; and (v) that basal microbial rates would decrease with warming due to indirect effects, including a reduction in soil moisture or substrate depletion.In June 2020, four randomly assigned blocks were established, each consisted of two control and two warmed plots.The blocks were separated by more than 10 m to ensure spatial independence.

| MATERIAL S AND ME THODS
Prior to the statistical analyses, we combined field technical replicates into means per block, ending with one control and one warming treatment per block.The warming treatments started in mid-June and ran until mid-August, covering the growing season of 2020.The plots were heated from aboveground with IR heaters (PAS 2, 650 W, Backer BHV AB, Sweden) 1.2 m from the soil surface.Data loggers (TMS-4 29 cm, TOMST®, Czech Republic) were installed before the start of the experiment in each plot.The in situ soil (−8 cm), surface (0 cm), air (+15 cm) temperatures, and volumetric soil moisture (between 0 and −14 cm) were monitored with 15 min resolution.In mid-August 2020, after 2 months of warming treatment, one sample per experimental plot was composited from three randomized cores (5 cm diameter), covering only the O-horizon down to 5 cm depth.
After the removal of stones and visible roots, the soil samples were sieved (4 mm mesh size) and kept cold (at 4°C) until the laboratory assessments (within 2 days of sampling).In mid-August 2020, the IR heaters were removed after 2 months of warming, and the plots were left without warming treatment under ambient conditions.
Then, after one winter (in June 2021) and one entire annual cycle (in August 2021), the plots were sampled again to investigate if any warming legacy remained in the microbial community temperature relationships.

| Soil physiochemistry and field observations
We measured gravimetric soil moisture (105°C for 24 h) and SOM content through loss on ignition (550°C for 12 h).Soil pH and electrical conductivity (EC) were determined in a 1:5 (w:v) soil: water extraction.Soil total carbon (TC) and total nitrogen (TN) were measured using Dumas dry combustion by a C/N 144 elemental analyzer (VarioMAX CN, Elementar, Germany).The normalized difference vegetation index (NDVI) as a proxy for plant productivity, the fraction of photosynthetically active radiation (fPAR), and the leaf area index (LAI; the projected area of leaves over the unit of measured ground area; m 2 m −2 ) were determined with an NDVI meter (SpectroSense2+, Skye, UK).

| Microbial growth, respiration, and carbon use efficiency (CUE)
Bacterial growth was determined by radioactively labeled 3 H-Leucine (Leu) incorporation into extracted bacteria from 0.5 g of fresh soils according to the homogenization/centrifugation technique (Bååth, 1994;Bååth et al., 2001).After centrifugation, each sample was divided into 10 subsamples and incubated at 10 different screening temperatures (Section 2.4; Table S1).The amount of Leu incorporated into bacteria (pmol Leu incorporated g −1 SOM h −1 ) was used to estimate bacterial growth.The bacterial incorporation rates were converted from Leu to thymidine using a conversion factor determined by Cruz-Paredes et al. (2021).Then, the bacterial C production (μg C g −1 SOM h −1 ) was estimated according to Soares and Rousk (2019).
Fungal growth was measured in 0.5 g of fresh soils using 14 C-acetate (Ace)-in-ergosterol incorporation method (Bååth, 2001;Rousk et al., 2009).Individual soil samples were incubated at 10 different screening temperatures (Section 2.4; Table S1).The amount of incorporated 14 C Ace into the extracted ergosterol (pmol Ace incorporated g −1 SOM h −1 ) was used as a proxy for fungal growth.The rate of 14 C Ace incorporation was converted into C units (μg C g −1 SOM h −1 ) according to Soares and Rousk (2019).
Microbial respiration was determined using 0.5 g of fresh soil in 20 mL glass vials, purged with pressurized air, sealed with crimp caps, and followed by incubation (Section 2.4; Table S1).CO 2 concentrations were measured by a gas chromatograph equipped with a methanizer and flame ionization detector.Empty vials were purged with pressurized air in parallel, and the CO 2 concentration in the air was subtracted from the measured CO 2 production of the soil samples.
Bacterial growth, fungal growth, and respiration measurements allowed us to determine the CUE, the ratio of total C production (bacterial growth + fungal growth) to total C use (bacterial growth + fungal growth + respiration) according to Equation (1).

| Microbial temperature relationships
The temperature relationships of bacterial growth, fungal growth, and respiration were determined by using short-term laboratory assays.Subsamples of each soil sample were incubated in water baths set to 10 different screening temperatures from 0 to 45°C in 5°C intervals, ending with one separate sample for each temperature.Samples for bacterial growth, fungal growth, and respiration were all run in parallel.The incubation times were adjusted to different temperatures to yield a similar level of microbial activity at the selected temperature, thus minimizing differences in C availability between incubations (Table S1).For bacterial growth, the incubation was 2 h at 20-25°C, and adjusted to the lower (e.g., 4 h at 15°C) and higher (e.g., 1 h at 30°C) temperatures.For fungal growth, we used 4 h at 20-25°C and adapted the time to the other temperatures (e.g., 8 h at 15°C, 2 h at 30°C).For respiration, it was 18 h at 20-25°C and adjusted to the other temperatures (e.g., 30 h at 15°C, 6 h at 30°C).By keeping these incubation time frames, time for a change in microbial temperature relationships, substrate availability, or other environmental conditions besides the direct effect of temperature should all be minimized (Kirschbaum, 2006;Rousk & Bååth, 2011).

| Indirect effects of warming
To test whether moisture or substrate availability caused any changes in microbial rates in soils as an indirect effect of warming, we adjusted the water content of the soil samples (both control and warmed) to optimal moisture, to 50% water holding capacity (WHC), and left them at room temperature for 1 week.We then measured bacterial growth, fungal growth, and respiration at 20°C (1) CUE = (Bacterial growth + Fungal growth) ((Bacterial growth + Fungal growth) + Respiration) standardized incubation temperature.For this assessment, soil samples at the end of the warming treatment (August 2020) were used.

| Microbial community composition
We analyzed microbial communities by extracting DNA from 250 mg of freeze-dried soils using MoBio Power Soil Pro Kit (MoBio, Carlsbad, USA) following the protocol provided by the manufacturer.
For this, we used soils that were sampled after 2 months of warming treatments in August 2020.DNA concentration was determined using a Nanodrop spectrophotometer system (Thermo Scientific, Wilmington, DE, USA), and DNA extracts were sent to BGI for amplicon sequencing according to their standard protocol (as provided by BGI Tech Solutions; www.bgi.com).Samples entered the library preparation process as follows, 30 ng DNA sample and fusion primer were used to configure PCR reaction system.The PCR reaction parameters were set as a standard for PCR amplification.Agencourt AMPure XP beads were then used to purify DNA and dissolve in elution buffer.The library was qualified by the Agilent Technologies 2100 bioanalyzer.The qualified libraries were sequenced pair-end on the Hiseq 2500.For the bacterial community, the V3-V4 region of the 16S region was amplified using the primers 341F (5′-CCTAY GGG RBG CAS CAG-3′) and 806R (5′-GGACT ACN NGG GTA TCTAAT-3′) (Herlemann et al., 2011).For the fungal community, the ITS1-ITS2 region was amplified using the primers ITS1 (5′-CTTGG TCA TTT AGA GGA AGTAA-3′) (Gardes & Bruns, 1993) and ITS2 (5′-GCTGC GTT CTT CAT CGATGC-3′) (White et al., 1990).

| Data analysis
The microbial temperature relationships were modeled in two steps.
The square root relationship was first determined.We plotted the square root of microbial growth and respiration (y-axis) against temperature (x-axis), resulting in a linear relationship (between 0 and

20°C) according to Equation (2).
where R is the rate of bacterial growth (Leu incorporation), fungal growth (Ace incorporation), or respiration; a is the slope parameter; T is the screening temperature (°C), and T min is the minimum temperature (the x-axis intercept).Second, the entire temperature range (between 0 and 45°C) of microbial rates was modeled.Here, the full function of the square root model was used for bacterial and fungal growth according to Ratkowsky et al. (1983), reported in Equation (3).
where b is the slope parameter, and T max is the maximum temperature (the second x-axis intercept).With the derivative of Equation ( 3), the optimum temperature (T opt ) was determined as a local maximum.
Around the T opt , the growth relationship curves start to bend until it reaches the x-axis, and with that, the maximum temperature (T max ) can be determined.Low T min, T opt , and T max indicate a community with a better capacity to grow at cold temperatures and higher values indicate a community with a better ability to grow at warm temperatures.At temperatures above the T opt , the microbial growth and respiration are uncoupled since the respiration rates do not decrease within the studied temperature range in the short assays used (see discussion in Birgander et al., 2013).For this reason, Equation (2) was used for respiration (between 0 and 20°C), while Equation (3) for bacterial and fungal growth was applied for the entire temperature interval (between 0 and 45°C).
The temperature sensitivity (Q 10 )-the factor by which microbial rates increase by 10°C increase in temperature-was determined between 5 and 15°C for bacterial growth, fungal growth, and respiration according to Equation (4).
where Q 10 is the change in bacterial growth, fungal growth, and respiration rate.All the parameters are the same as reported in Equation (2).
Before conducting the statistical analyses, we combined the field technical replicates into means per block, resulting in one control and one warming treatment for each block.One-way ANOVA analyses were performed to test for differences between treatments (control and warming) in T min , T opt , T max , Q 10 , soil properties, temperature, and moisture effects (Tables 1-3; Table S2).Treatments were used as fixed effect and blocks as random effect.We used one-way repeated measures ANOVA to test the effect of the 2-month warming treatment on the change in temperature and moisture with treatment as fixed and block as a random effect.
(2) To understand how increased temperature shaped the microbial thermal trait distribution in the soil samples (0-5 cm) we used the two levels where the temperature was measured (at surface and at 8 cm depth) (Table S2).However, relating the T min shifts to the surface temperature is misleading for multiple reasons: (i) when the surface heats, it also dries, reducing microbial process rates, turnover times, and contribution to ecosystem processes; (ii) a dry soil surface will also insulate lower levels of the soil profile from the IR heaters, and as a result below the top layer of the soil the warming effect is reduced, resulting in a more stable temperature throughout the soil profile (Harte et al., 1995).Therefore, the closest estimate that we can use for the sampled top 5 cm of the soil profile is the soil temperature measurement at −8 cm depth.
All sequence data were processed using DADA2 and the DADA2 ITS Pipeline Workflow (1.8) (Callahan et al., 2016) to determine the amplicon sequence variants (ASVs).Briefly, demultiplexed pairedend reads for bacteria and fungi were quality checked and trimmed using standard filtering parameters.Forward and reverse reads were merged to obtain the full denoised sequences and chimeras were removed.Taxonomy was assigned using the Silva reference  Holmes, 2013).We calculated Bray-Curtis dissimilarity matrix and visualized the differences in the bacterial and fungal communities with a PCoA ordination.Differences between the control and warming treatments were calculated with PERMANOVA analysis using the function adonis() from the vegan package in R (Oksanen, 2015).
Treatments were used as fixed effect and blocks as random effect.
We performed a differential abundance test using the function trans diff() from the microeco package in R (Liu et al., 2021) to find taxa with significantly different relative abundance across control and warming treatments.
We calculated the running mean of 4 h (the mean of 16 consecutive measurements) for soil, surface and air temperature and soil moisture.Then, we used the function envfit() from the vegan package
The aboveground IR heating resulted in a high temperature increase on the surface, and the temperature effect size then decreased with depth throughout the soil profile.The warming treatment increased the 2-month mean temperature from 7.6°C ± 0.2 to 10.4°C ± 0.5 in soil (−8 cm) (p = .01),from 11.5°C ± 0.1 to 23.6°C ± 1.5 on the surface (p = .005),and from 12.7°C ± 0.2 to 26.3°C ± 2.0 in the air (p = .008)(Figure 1; Figure S1, Table S2).The warmest 4 h increased from 10.3 ± 0.3 to 14.0 ± 0.6 in soil (p = .01),and tended to increase
In June 2021, following 10 months of ambient conditions, the elevated T min of bacterial growth still tended to be higher than control (Δ T min = 1.2) although the difference was not statistically distinguishable anymore.After 12 months without active warming treatment, the Δ T min had fully dissipated (Δ T min = 0.0).For fungal growth, the Δ T min decreased from 2.2°C (August 2020) to 0.8°C (June 2021) (Figure 3; Table S3).

| Indirect warming effects
Bacterial growth decreased from 7.3 ± 0.7 to 3.2 ± 0.8 μg C g −1 SOM h −1 (p = .009),and fungal growth from 1.1 ± 0.1 to 0.6 ± 0.1 μg C g −1 SOM h −1 (p = .002)(Table 4).For respiration, the warming treatment Normalized bacterial growth, (c) normalized fungal growth, and (e) normalized respiration temperature dependence curves; and the determined T min (minimum temperature) of (b) bacterial growth, (d) fungal growth, (f) respiration temperature dependence.In panels (a) and (c), fitted curves are based on the Ratkowsky model (Equation 3).Below the T opt , square root transformed rates showed a linear response to the screening temperature.Unlike bacterial and fungal growth, the respiration rate did not reach a T opt or T max in the studied incubation temperature interval.Therefore, the simplified square root relationship was used in panel (e) (Equation 2; see Section 2).In panels (a), (c), and (e) all the replicates are shown for control (n = 8) and warming (n = 8) treatment with curves fitted to the mean values.In panels (b), (d), and (f) the data derived from the fitted curves of the individual replicates were used to estimate the mean ± 1SE; Control (n = 4); Warming (n = 4) (see Section 3).Asterisk symbols represent significant differences between control and warming treatments (p < .05).

| Microbial carbon use efficiency
Bacterial growth, fungal growth, and respiration results were all converted into C units to be able to estimate the microbial CUE.Upon warming, microbial growth rates decreased more than respiration rates, resulting in decreased CUE from 0.20 ± 0.01 to 0.13 ± 0.02 (p = .01)(at 20°C) (Table 4; Figure 4c).When we adjusted the water content of the soil samples to 50% WHC, treatment differences for CUE were eliminated (Figure S2d; Table S4).

| Fungal-to-bacterial growth ratio and microbial community composition
Upon warming, fungi became more dominant in the warmed plots, and the fungal-to-bacterial growth ratio increased from 0.15 ± 0.01 to 0.20 ± 0.02 (p = .04).After we adjusted the water content to 50% WHC, treatment differences were eliminated (Table S4; Figure S2E).
We found no significant differences in the α-diversity, evaluated as Shannon index of bacterial or fungal communities.Additionally, we evaluated the β-diversity with Bray-Curtis dissimilarities.
Similarly, we did not find significant differences between treatments (Figure S3).We correlated the microbial community composition with measured environmental factors and soil physiochemical properties.We found that the differences in the bacterial community composition were linked to the coldest 4 h (R 2 = .77,p = .001)and warmest 4 h (R 2 = .63,p = .01)period measured in soil in situ and to soil water content (R 2 = .48,p = .04)determined in the sampled soils at the end of the warming experiment in August 2020 (Figure S3A vectors).For fungi, we found that the coldest 4 h (R 2 = .45,p = .03),the driest 4 h (R 2 = .49,p = .03),and the 2-month mean soil moisture (R 2 = .47,p = .04)measured in situ explained the variation in the community composition (Figure S3b vectors).Mantel tests showed no significant effect of the measured soil and environmental properties on bacterial or fungal β-diversity.
With the differential abundance test we found that 55 bacteria taxa had significantly different relative abundance in warming and control treatments (Table S5a).These taxa were distributed among 12 different phyla.For fungi, we found that eight taxa within the two phyla Ascomycota and Basidiomycota had significantly different relative abundance between the warming and control treatments (Table S5b).

| Heat wave-induced microbial thermal trait adaptation and its reversal
During the year of this study, a period of unprecedented heat waves was documented in arctic Siberia, including the record-breaking 38°C-the highest temperature recorded in the last 135 years (Overland & Wang, 2021).According to IPCC (2021), human-induced climate change will intensify the occurrence of such heat waves.However, so far, only a few field studies have investigated how extreme temperatures affect soil microbial processes (Anjileli et al., 2021;Li et al., 2020), and there are no reports on how microbial temperature relationships respond to heat waves.This highlights the need to understand how rare weather extremes affect the role of microbes in maintaining ecosystem functions.To fulfill this need, we tested if drastic temperature increase in the warmest part of the year shapes the microbial thermal trait distributions.The 2 months of experimental summer heat wave simulation resulted in raising the temperature by 13.6°C in the air (+15 cm), 12.1°C at the surface (0 cm), and 2.8°C in soil (−8 cm), closely matching the temperature recorded in arctic Siberia in the year of the study (Overland & Wang, 2021).As a result of this warming treatment, the communitylevel bacterial growth T min increased by 2.1°C.A similar trend was seen for fungal growth, with a tendency for 2.2°C increase in T min .To understand how increased temperature shaped the microbial thermal trait distribution in the soil samples, we related the T min to temperature increase in soil.As predicted, for bacterial growth, the T min increased by 0.7°C per 1°C increase in soil (−8 cm) and by 0.8°C per 1°C for fungal growth (−8 cm).Previous warming experiments that linked the temperature relationships with chronic increases in temperature have found limited increase in T min for bacterial growth between 0.2 and 0.3°C per 1°C increase in temperature (Nottingham et al., 2019;Rousk et al., 2012).In laboratory incubation studies (Bárcenas-Moreno et al., 2009;Birgander et al., 2013), and a winter warming field experiment (Birgander et al., 2018), temperatures lower than the T opt had no effects on the microbial temperature relationships within months or seasons.However, in laboratory studies, the T min increased within days or weeks by 0.8 per 1°C increase when soil samples were incubated at temperatures higher than the T opt (Bárcenas-Moreno et al., 2009;Birgander et al., 2013).
These observations suggest that the microbial thermal trait shifts occur in the warm part of the year when temperatures exceed the T opt .Therefore, using temperature increases averaged over the year (e.g., MAT) that includes the colder seasons might be a poor predictor for possible shifts in microbial thermal trait distributions.
Instead, a more appropriate predictor can be specifically the spells of high temperatures, including heat waves in the warmest part of the year.In this experiment, the warming effect size built up in the soil over the course of the 2-month warming period (Figure 1a).Although the warming effect remained stable 2 weeks after the warming had started, the environmental temperature rose steadily over the 2 months, resulting in the highest soil temperature being recorded in the last 2 weeks of the experiment (marked with a black line in Figure 1a).We believe that the warmest 2-week period exerted most of the selective pressure on the microbial community, resulting in the observed thermal trait distribution shifts.This highlights that to TA B L E 4 Bacterial growth, fungal growth, and respiration in carbon (C) unit.improve the understanding of how soil microbial communities run soil OM cycling under climate forcing, the speed with which microorganisms can shift their temperature traits to heat waves is required.
We also aimed to understand how microbial processes are affected as a legacy of rare events such as heat waves.Our experiment is the first warming experiment where not only the shifts of the microbial temperature relationships were studied but also the warm-shifted temperature relationships' subsequent dissipation after the warming had ended.We found support for our prediction that by removing the warming treatment, the microbial growth temperature relationships will gradually adjust to the ambient state.
The increased bacterial growth Δ T min , had been reduced from 2.1 to 1.2°C, but the remaining difference in the thermal traits suggested that most of the trait change still affected the microbial functions even 10 months after the warming treatment had ended (Figure 3; Table S3).Only after another summer season, altogether covering a whole year at ambient conditions, had the shift vanished, suggesting that traits again were in balance with environmental conditions.This implied that the bacterial growth T min differences decreased only marginally during the long cold winter to subsequently decrease more abruptly in the short but warmer summer.In another study, Nottingham et al. (2021) translocated soil cores in a tropical forest along a temperature elevation gradient (between 6.5 and 26.4°C MAT) and assessed the bacterial growth temperature dependence shifts to warmer and colder temperatures.Their results suggest that after 2 years, almost 80% of the differences in bacterial thermal trait distributions between the translocated soil cores and the native environment had disappeared.The adjustment of the temperature traits appeared to be faster in warmer than colder soils which can be explained by accelerated microbial turnover at higher temperatures (Hagerty et al., 2014).For fungal growth, the Δ T min had been reduced from 2.2 to 0.8 within 10 months after the warming had ended.However, further research is needed to confirm these patterns and evaluate how fast the fungal thermal trait distribution can shift and for how long the legacy of shifted traits can determine SOM decomposition.Additionally, we here studied only one heat wave "pulse" perturbation.It is likely that the rate of microbial thermal trait responses to repeated heat waves progressively accelerates (Jentsch & White, 2019), but testing this prediction will require a specific study.

| Indirect controls of temperature
Upon heat wave simulation, both bacterial and fungal growth rates decreased (Figure 4a,b).These decreased rates translated into a 33% higher fungal-to-bacterial growth ratio (Figure 4d).The more drastic drop in microbial growth than respiration in the warmed plots resulted in 35% lower CUE (Figure 4c; Table 4).We aimed to delineate between the indirect controls of increased temperature, such as moisture reduction or substrate availability.To achieve that, one possible approach is to adjust either of these factors to eliminate their contribution to differences between treatments.We adjusted the water content to optimal level (50% WHC) for both control and warmed soils.After 1-week incubation to stabilize microbial rates, this adjustment eliminated treatment differences for bacterial growth, fungal growth, and respiration along with fungalto-bacterial growth ratio, and CUE (Figure S2; Table S4).With that, we could confirm our prediction that decreased moisture, as an indirect effect of warming, was the main driver for the observed differences in the measured rates.In a laboratory experiment where soil samples from the same subarctic system were incubated at four decreasing moisture levels (from 50% to 10% WHC), both bacterial and fungal growth rates dropped with soil drying (Cruz-Paredes et al., 2021).However, fungi could maintain higher growth rate than bacteria at reduced moisture levels, and the differences between groups were more pronounced at lower moistures resulting in an increased fungal-to-bacterial growth ratio.These results suggest that fungal growth is more resistant to drought, indicating that fungaldominated soil systems could cope better with decreased soil moisture during heat waves (Yuste et al., 2011).Also, the results highlight the need to study interactions between multiple factors affected by climate warming, including heat waves and drought.Since the differences between field treatments were eliminated by adjusting moisture in our study, we could also confirm that substrate depletion did not control microbial process rates.

| Representing microbial temperature relationships in soil C models
We predicted that with warm-shifted microbial temperature relationships, the Q 10 values of microbial rates would also increase (Bååth, 2018).The shift in the microbial temperature relationships translated into a tendency of increase in Q 10(5-15°C) for bacterial growth from 3.2 to 3.8 (Table 3).This was similar to an increase (from 4.0 to 4.4) observed in 3 years of continuous 5°C-warming in a temperate hardwood forest (Rousk et al., 2012).Current ecosystem models (e.g., CENTURY, Roth C) use a constant Q 10 (Q 10 = 2) for all microbial processes in all climatic conditions (Davidson & Janssens, 2006;García-Palacios et al., 2021;Hamdi et al., 2013).We here showed (i) that a summer heat wave can result in warm-shifted microbial trait distributions with a transiently elevated Q 10 , and (ii) that functional shifts that can persist up to 1 year, influencing how environmental temperatures force decomposer processes (Figure 3; Table 3).We propose that the representation of heat waves in cold environments is needed to understand how ecosystems exchange C with the atmosphere.

| Response of microbial community composition to extreme warming
Soil temperature (de Vries et al., 2018;Deslippe et al., 2012;Oliverio et al., 2017), moisture (Banerjee et al., 2016;Leizeaga et al., 2021), and pH (Barnett et al., 2020) have been identified as important factors shaping the microbial β-diversity.We found that bacterial β-diversity correlated with the minimum and maximum soil temperature (coldest and warmest 4 h period) and with the water content at the sampling time (August 2020) (Table 1; Table S2), while fungal β-diversity correlated with the minimum soil temperature and moisture (coldest and driest 4 h period) (Table S2).The correlations between bacterial and fungal β-diversity and extreme environmental conditions suggest a shift toward warm-and drought-responsive microbial communities, consistent with recent experimental tests (Oliverio et al., 2017).To further explore the microbial community responses to warming, we determined signature taxa (Liu et al., 2021).We identified 55 bacterial and eight fungal temperature-responsive taxa at ASV level.Most Actinobacteria ASVs increased in the relative abundance with warming, including a thermophilic taxa Acidothermus and a heat-tolerant, aerobic Thermoleophilia (Hu et al., 2019).This finding is consistent with previously reported results from a laboratory incubation experiment (Oliverio et al., 2017), and a 2-year field warming experiment in the tropics (Nottingham et al., 2022).
One mechanism that can affect the temperature sensitivity of microbial growth is thermal adaptation through evolutionary change (Bennett et al., 1990).However, that often requires substantial time.
For instance, E. coli even after 2000 generations at optimal growth temperature showed only minor shifts in thermal traits in response to strong selection (Mongold et al., 1996).In natural environments, soil bacterial turnover time is considerably slower, ranging between 15 and 20 days at 10°C (Bååth, 1998).This means that the estimated bacterial turnover in this field experiment is only a few generations during the 2 months of warming treatment, especially given the substantial decrease in soil moisture in the warmed plots.Therefore, it is unlikely that genetic changes in the microbial community induced the shifts in the thermal trait distributions.Species sorting is another mechanism that can explain the shifts in the community-level microbial temperature relationships, where taxa with traits that confer competitive advantages under high-temperature conditions outcompete others and become dominant.Evidence for thermal trait shifts via species sorting has been reported from laboratory experiments on both alpine and arctic soils exposed to new thermal regimes (Donhauser et al., 2020;Rijkers et al., 2022).However, in our study, the observed change in T min could not be clearly assigned to community composition differences since the warming treatments did not result in a significant shift in bacterial nor fungal β-diversity.
That raises a question: What is the possible explanation for the measured differences in T min that could not be detected in β-diversity?
The heat wave simulation significantly reduced soil moisture, which may have dampened warming-induced changes in community composition.For example, a 2-year warming experiment resulted in detectable warm-shifted microbial community compositions but only when soil moisture was controlled (Sheik et al., 2011).Following these lines, the link between the increased T min and the microbial β-diversity is likely to be stronger when indirect effects of warming on the microbial community composition are controlled.To test this in future work, soil moisture levels could be maintained at control levels in a section of the warmed plots.

| CON CLUS ION
As a result of an experimental heat wave simulation in the Subarctic, the bacterial growth temperature dependences shifted to become warm-tolerant within one growing season, with a similar trend for fungi.Unexpectedly, the shifts in microbial temperature relationships could not be explained by differences in α-or β-diversity of the microbial community.After removing the warming treatments, the warmshifted bacterial temperature dependences fully recovered to match ambient conditions only after 1 year.As an indirect effect of warming, lower soil moisture decreased all basal microbial rates, resulting in 33% higher fungal-to-bacterial growth ratio and 35% lower CUE.This suggested that the microbial responses to summer heat waves would exacerbate C losses from subarctic soils.However, when the soil moisture was adjusted, the observed differences in microbial rates were eliminated, ruling out any contribution by a reduction in the quality of C. In summary, a summer heat wave in the Subarctic resulted in (i) shifts in microbial thermal trait distributions within 2 months that then lasted up to 1 year; (ii) decreased microbial process rates due to lower moisture, not substrate depletion; and (iii) no clear link between the warm-shifted thermal trait distributions and differences in community composition.

2. 1 |
Study site, field experiment, and soil sampling Soil samples were collected in a subarctic birch forest in North Sweden (68°21′16.7″N 18°49′08.7″E) close to the Abisko Scientific Research Station.The field site is approximately 200 km north of the Arctic Circle, 385 m above sea level.The soil in the sampled area is a Histosol, rich in OM, formed on base-rich schist (WRB-IUSS, 2014), where the organic horizon is ca.8-25 cm deep.Abisko has a subarctic climate, characterized by long, cold winters and short, cool summers, with a mean annual air temperature (MAT) of 0.4°C, mean annual soil temperature of 2.1°C (at 5 cm depth), and mean annual precipitation (MAP) of 352 mm (20year mean between 2000 and 2020) (Abisko Scientific Research Station, 2021).The growing season is limited to the frost-free period between June and August.The vegetation on the experimental site is dominated by deciduous shrubs (Betula pubescens, Rubus chamaemorus, Salix lapponum, Salix glauca, Vaccinium uliginosum, V. myrtillus, V. vitis-idaea), evergreens (Cassiope tetragona, Empetrum hermaphroditum), ferns (Equisetum scirpoides), and bryophytes (Pleurozium schreberi, Sphagnum spp., Hylocomium splendens).
database (v138) for bacteria and UNITE database (04.02.2020) for fungi.We calculated diversity metrics for bacteria and fungi with the ASVs obtained from the microbial communities' analyses.We calculated the Shannon index (α-diversity) for bacterial and fungal communities in each sample.Moreover, ASVs were filtered for β-diversity analysis by only keeping ASVs with at least five counts.Then, samples were transformed to even sampling depth with the function transform_sample_counts() from the phyloseq package (McMurdie &

(
Oksanen, 2015) to correlate indices for the bacterial and fungal temperature relationships, temperature, and moisture data and other measured environmental and soil properties with the differences in the bacterial and fungal community composition.Mantel tests with the different environmental variables and soil physiochemical properties were performed to address community structure drivers.All statistical analyses were done in R version 4.0.3(R Core Team, 2020).
from 25.3 ± 1.0 to 35.3 ± 3.0 on the surface (p = .08),and increased from 30.3 ± 0.5 to 40.1 ± 2.0 in air (p = .03).The 2-month mean volumetric moisture was 0.34 ± 0.08 in the control and 0.22 ± 0.05 in the warmed plots (p = .08).The one-way repeated measures ANOVA showed significant temperature increase in the soil (p < .001), on the surface (p < .001), in the air (p < .001),and significant decrease in soil volumetric moisture (p < .001)(between June 18, 2020 and August 11, 2020).F I G U R E 1 (a) Soil temperature (measured at −8 cm), (b) soil volumetric moisture (between 0 and −14 cm), (c) surface temperature (0 cm), and (d) air temperature (15 cm) measured between June 18, 2020 and August 11, 2020.The blue color indicates the temperature and moisture measured in control (mean values; n = 8), and the red color indicates measurements in warmed (mean values; n = 8) plots.The thick line in the x-axis of panel (a) indicates the warmest 2-week period throughout the experiment.
Temperature indices: T min (minimum temperature), T opt (optimum temperature), and T max (maximum temperature) of bacterial growth, fungal growth, and respiration at three different sampling times (August 2020, June 2021, and August 2021).
TA B L E 2Note: Data represent mean ± 1SE; Control (n = 4); Warming (n = 4).See Materials and Methods for further description.Abbreviation: NM, not measured.*Significantdifferencesbetweencontrol and warming treatments (p < .05).TA B L E 3 Q 10 values between 5 and 15°C for bacterial growth, fungal growth, and respiration rates at three different sampling times(August 2020, June 2021, and August 2021).*Significant differences between control and warming treatments (p < .05).