Vegetation type, not the legacy of warming, modifies the response of microbial functional genes and greenhouse gas fluxes to drought in Oro-Arctic and alpine regions

Abstract Climate warming and summer droughts alter soil microbial activity, affecting greenhouse gas (GHG) emissions in Arctic and alpine regions. However, the long-term effects of warming, and implications for future microbial resilience, are poorly understood. Using one alpine and three Arctic soils subjected to in situ long-term experimental warming, we simulated drought in laboratory incubations to test how microbial functional-gene abundance affects fluxes in three GHGs: carbon dioxide, methane, and nitrous oxide. We found that responses of functional gene abundances to drought and warming are strongly associated with vegetation type and soil carbon. Our sites ranged from a wet, forb dominated, soil carbon-rich systems to a drier, soil carbon-poor alpine site. Resilience of functional gene abundances, and in turn methane and carbon dioxide fluxes, was lower in the wetter, carbon-rich systems. However, we did not detect an effect of drought or warming on nitrous oxide fluxes. All gene–GHG relationships were modified by vegetation type, with stronger effects being observed in wetter, forb-rich soils. These results suggest that impacts of warming and drought on GHG emissions are linked to a complex set of microbial gene abundances and may be habitat-specific.


Introduction
High latitude and alpine environments are experiencing climate c hange mor e r a pidl y and se v er el y than temper ate ar eas (Gr abherr et al. 2010, Singh et al. 2010, Palomo 2017, Rantanen et al. 2022 ).Combined effects of warming and changes in precipitation patterns are likely to have profound effects on soil microbial communities in these areas (Olefeldt et al. 2013 ), potentially resulting in large shifts in microbial community composition and functioning.With the release of physiological constraints created by low temper atur es and waterlogging, there is potential for soil microbial activity to increase in response to simultaneous warming and drought (Sheik et al. 2011, Seo et al. 2015, Commane et al. 2017 ).As Arctic and alpine tundra are globally important carbon (C) stores, an y incr ease in micr obial activity could r esult in significant incr eases in gr eenhouse gas (GHG) emissions, further acceler ating climate change (Seo et al. 2015 ).Additionally, warming will result in shortened snow cover duration and increased permafrost melt, and encr oac hment of tr ees and alter ed v egetation (Ha gedorn et al. 2019 ), which has been shown to enhance soil C loss due to incr eased micr obial decomposition (Hartley et al. 2012, García Criado et al. 2020 ).GHGs ar e gener ated thr ough micr obial tr ansformation of soil C and nitrogen (N) pools, which leads to the emission of carbon dioxide (CO 2 ), methane (CH 4 ), and nitrous oxide (N 2 O).Various guilds of micr obes hav e r oles in the emissions of these gases, and in converting them to less potent gases in terms of global warming potential (Fig. 1 ; Martins et al. 2017, Lafuente et al. 2020 ).We can identify specific genes that dir ectl y code for enzymes whic h degr ade or synthesize GHGs.Ho w e v er, while we hav e a r elativ el y good understanding of the role of microbial functional genes involved in GHG emissions, little is known about how they respond to warming and drought either individually or in combination, or how these r esponses r egulate GHG emissions (Li et al. Figure 1.Expectations of functional gene response to long-term warming and drought.A tilde infers no expected effect.Colours to provide contrast.Stronger expected effects on gene expression are shown as larger arrows.We also include the pathway of the enzyme that the gene encodes for completeness.Further expectation of effect of vegetation type is mentioned, though not emphasized due to lack of literature.(A) pmoA is expected to have a low or negligible response to warming (Keuschnig et al. 2022 ), and an increase under drought, leading to a reduction in CH 4 emissions (Olefeldt et al. 2013 ).(B) nosZII genes ar e thought to incr ease under dr ought and warming, although e vidence is curr entl y limited (Xu et al. 2020a ).(C) expected effects of drought and warming on acdS expression.There is currently limited evidence for responses to changes in climatic conditions, but since the gene is expressed by plant-associated microbes, we anticipate a strong effect of vegetation type (Nikolic et al. 2011, Liu et al. 2019a ).2020 , Zhang et al. 2022 ).The identity of these gases is also critical as they hav e differ ent global warming effects: CO 2 has a weak effect on climate change but is very persistent in the atmosphere, while CH 4 and N 2 O are short-li ved, the y are more potent (Shine et al. 2005 ).As such, the abundance of functional genes that code for GHG reducing enzymes could be critical to assessing future impacts of individual gases.
While CO 2 fluxes are likely to be governed by a complex mixture of plant and microbial activity, CH 4 emissions are largely micr obiall y mediated: methanogenic microbes undertake methanogenesis using organic sugars as a substrate, while methanotrophs br eak the extr emel y str ong C-H bond in the CH 4 via expression of functional genes, pr edominantl y pmoA (although other genes play important roles including mmoX and mxaF ), under aerobic conditions (Singh et al. 2010 ).The pmoA genes encode the enzyme particulate CH 4 monooxygenase, which catalyzes the conversion of CH 4 to methanol (Liu et al. 2015 ).While warming might increase the emissions of CH 4 from soils, the temperature sensitivity of the pr ocesses under pinning methanogenesis and methanotrophy is lo w er in tundra soils than in temperate soils (Tveit et al. 2015 ).Ther efor e, at low temper atur es, CH 4 emissions ar e high, but with warming, the accompanying increase in CH 4 in tundra is smaller compared with temperate soils, which could be due to incr eased expr ession of pmoA and consequent consumption of CH 4 (Fig. 1 A ;Tveit et al. 2015 ).There is evidence from pad d y fields, under extr emel y cold and waterlogged soils, that CH 4 emissions are str ongl y dependent on soil organic matter content, while in drier, warmer soils the CH 4 emissions are linked to the abundance of pmoA (Xu et al. 2020b ).Ho w e v er, the bulk of r esearc h on functional gene responses to climate changes has been carried out on highly organic soils, and little is known about more mineral-based soils (K eusc hnig et al. 2022 ), whic h also dominates man y tundr a and alpine regions (Walker et al. 2005 ).
The denitrifying genes nosZI and nosZII have a parallel role to pmoA , producing gene products which convert N 2 O to N 2 gas, and thus decreasing the global warming potential of the soil (Orellana et al. 2014 ).Much work has been carried out looking at nosZ genes in w etlands, sho wing that functional genes associated with N processing tend to increase with warming (Zhang et al. 2015, Wang et al. 2018 ).While nosZI is sensitive to soil moisture, there is no evidence of an impact of warming.For nosZII , the only clear effect is a negative relationship with N 2 O emissions, and unclear effects of drought and warming (Xu et al. 2020a ).Overall, there is little information about an interaction between warming and drought for both pmoA and nosZ genes, and while climate change potentially could mean increases in water from snowmelt and permafrost thaw, IPCC projections also indicate regional variation in high latitudes (IPCC 2023 ).There could equally be an increase in the frequency and amplitude of summer or late season dr ought, r esulting in a drop in the water table (Li et al. 2019), and/or an increase in e v a potr anspir ation due to higher plant pr oductivity (K eusc hnig et al. 2022 ) leading to drier soils .T he consequences for functional gene abundances and the link with GHG emissions are therefore unclear, as is the potential for r ecov ery fr om extended dr ought.
The impact of warming and drought on GHG fluxes is also likely influenced by vegetation, because the varying quality and quantity of litter inputs and root exudates dictate microbial substrate av ailability.Further, v egetation types are likely to influence soil abiotic pr operties suc h as por osity and a ggr egate stability, whic h will in turn affect diffusion rates of gases, soil moisture conditions, and microclimate (Szyma ński 2017 ).Another important question is whether microbial stress genes have the potential to reduce GHG fluxes through encoding enzymes that reduce plant stress.The 1-aminocyclopropane-1-carboxylate (ACC) deaminase gene acdS is released by plant-associated rhizobacteria and bacterial endophytes and is known to aid stress relief in plants (Nikolic et al. 2011, Liu et al. 2019a ).The mechanism for relieving stress in plants is by breaking down ACC, a precursor of ethylene, to ammonium and α-ketobutyrate (Gupta and Pandey 2019 ; Fig. 1 ).To our knowledge, no studies to date have considered the abundance of acdS in response to warming, with the majority looking at stresses such as salinity (Orhan et al. 2016, Liu et al. 2019a ) or heavy metals (Loganathan et al. 2015 ).Ho w e v er, it is likel y that both the plants and rhizobacteria will increase activity in response to warming and drought, and that this will mean increased expression of acdS .
T he o v er arc hing aim of this study was to test whether longterm experimental warming modifies the functional response of microbial communities to drought and determine the consequences for GHG emissions.We tested this with four soil samples taken from two established International Tundra EXperiment sites (ITEX; Henry andMolau 1997 , Henry et al. 2022 ) subject to long term experimental warming: a high latitude Oro-Arctic site in Sweden, Latnjajaure Field Station (three vegetation types with matched pairs of warmed and ambient plots), and an alpine site in Val Bercla, Switzerland (one vegetation type with warmed and ambient plots).Soils taken from experimental plots were subject to an experimental drought in the laboratory, and the impact of the drought on functional gene abundance and GHG flux immediately after the drought period (i.e.resistance; note that an incr ease r elativ e to the control also constitutes low resistance) and its r ecov ery bac k to w ar ds the contr ol v alues (i.e.r esilience) was assessed over a period of 56 days after r e w etting.First, w e hypothesized that due to release of the cold and wet constraints on microbes, long-term warming will increase functional gene abundances involved in GHG emissions, but that this might further increase gene abundance in response to dr ought, r educing their resistance.Ho w ever, w e expect that this legacy of warming will increase the resilience of the functional genes, resulting in more r a pid r ecov ery of gene abundance after dr ought in warmed soils compared to ambient conditions (Xu et al. 2020b ).Second, we expected that GHG emissions will increase in the warming legacy plots r elativ e to ambient plots due to incr eases in kinetic ener gy, despite concomitant increases in GHG mitigating genes.Furthermore, the combination of drought and warming will increase GHG emissions, as the conditions become dryer and have higher oxygen a vailability.T hir d, w e hypothesized that the effect of longterm warming and subsequent drought on functional gene abundances and GHG fluxes ar e moder ated by v egetation type, with effects being stronger for soils of wetter, more forb or gr aminoid-ric h plant communities than drier, shrub-dominated communities.

Experimental set up
The long-term warming design used in this study consisted of open top chambers (OTCs; Hollister et al. 2022 ), which had a basal area of 1 m 2 and passively warmed the interior air and soil (average warming 1-3 • C; Marion et al. 1997 ).We used soils from an Oro-Arctic environment (Latnjajaure field station), and an alpine environment (Val Bercla field station), with four soil types that repr esent a gr adient of both soil moisture and soil C and N, as well as contr asting v egetation types (Table 1 ).At Latnjajaur e field station, Sweden (68 • 22N 18 • 13E, 981 m a.s .l, a v er a ge January and Jul y temper atur e −8.9 • C and 8.6 • C, r espectiv el y, mean pr ecipitation 846 mm, ranging between 600 and 1100mm yr −1 ) the year-round experiment was set up in 1993-1994 across diverse vegetation types dominated by different plant species .T he experiment consisted of a full factorial design, with warming as one treatment and vegetation type as another.Each treatment was replicated five times per plant community (Scharn et al. 2022 ).Here, we collected soils from thr ee v egetation types (thr ee le v els of the v egetation tr eatment), forming a moisture gradient, where the phenology, germination and growth of the dominant plant species have been shown to be highl y r esponsiv e to the warming treatment (Scharn et al. 2021(Scharn et al. , 2022 ) ), namely: a w et meado w envir onment c har acterized by Bistorta vivipara , Carex bigelowii , Calamagrostis stricta , and Poa arctica , a dry heath environment characterized by Betula nana , Cassiope tetragona , Salix herbacea , and Empetrum nigrum , and a tussock tundr a site, c har acterized by Eriophorum v aginatum , Phyllodoce caerulea , Vaccinium vitis-idaea , and S. herbacea (Henry and Molau 1997, Molau 1997, Scharn et al. 2021, 2022 ).In total, 30 soil samples were obtained at Latnjajaure, including soils taken from warmed and ambient plots and thr ee v egetation types, eac h r eplicated fiv e times (Table 1 ).
At the alpine site Val Bercla, Switzerland (46 • 29N, 9 • 35E, 2490 m a.s.l, 20% NNW facing slope, av er a ge January and July temperature −6.4 • C and 9.7 • C, r espectiv el y, mean pr ecipitation 1605 mm), the experiment was set up in 1994 on loamy sand (Stenström et al. 1997 ).At this site, the vegetation was more homogeneous and dominated by the forb Saxifraga oppositifolia , so it comprises one le v el of the vegetation treatment.Saxifraga oppositifolia has been shown to be r esponsiv e to warming at the northerly edge of its range (Wook e y et al. 1993 ).A total of 18 soil samples were collected from warmed and ambient plots (av er a ge OTC warming 1.0 • C), with each treatment being replicated nine times.
Soils to 10 cm deep were collected in July 2018 and shipped to The University of Manchester, United Kingdom on cold blocks [see Figure S2 ( Supporting Information ) for starting soil moisture content].Once in Manchester, the soils were sieved through 2 mm mesh and the permanent wilt point (PWP) was measured for each sample (based on Saxton and Rawls 2006 ).We calculated gravimetric soil moisture content for each soil sample by drying soils at 80 • C for 48 hours and determining the proportional water loss.Preliminary soil moisture tests on the soils sho w ed that in our experimental soils, only Latnjajaure dry heath soils were significantly drier in warmed than ambient soils, all others wer e r elativ el y unc hanged ( Figur e S2 , Supporting Information ).Eac h of the 48 samples were divided into four equal parts and placed in 50 ml Falcon tubes .T his meant we had a matched pair of samples (one drought, one control) that would be destructiv el y harvested at two time points ( Figure S1 , Supporting Information ; days 0 and 56).The reason for having a separate Falcon tube to be harvested at each time point was to prevent disturbance to the micr obial comm unities due for later harv est.The soils wer e k e pt in a gr owth c hamber in the dark at 9 • C, the av er a ge Jul y temper atur e for both sites.Soils in the drought treatment were monitored as they dried down to the PWP, with the control maintained at 60% water holding capacity (WHC).When soil had r eac hed PWP (9.17% for Latnjajaure, 6.93% for Val Bercla), the drought treatment commenced.All soils were maintained at either WHC (control) or PWP (drought) for 5 weeks, at which half of the microcosms were destructiv el y harv ested (day 0) to measur e r esistance of functional genes and GHG fluxes to drought.The Day 0 measurements were used to calculate resistance metrics (see the section 'Statistical anal ysis').The r emaining dr oughted micr ocosms wer e r e wetted to WHC and then WHC was maintained in the remaining microcosms, with weekly gas flux measures and harvests before the final tubes were harvested on da y 56.T hese weekl y harv est measur es wer e used to calculate resilience (see the section 'Statistical analysis') on a w eek-b y-w eek basis.

GHG measurements
GHGs were sampled on each harvest day, immediately prior to harvest.The Falcon tubes were sealed using a suba-seal and P ar afilm.

Sta tistical anal ysis
To ask if long-term warming increases functional gene abundances in response to drought, and if vegetation type is associated with a significant difference in gene abundance, we first standardized the functional genes pmoA , nosZII , and acdS by calculating the ratio with 16S rRNA genes.We built linear mixed effects models with experimental r eplicate (i.e.matc hed dr ought-contr ol pairs) nested within origin field treatment (i.e.experimental warming or ambient temper atur e) as r andom factors, and thr ee-way interactions between the fixed factors of timepoint, vegetation type, warming, and drought (Bates et al. 2013 ).Including a three-way interaction between our fixed effects allo w ed us to test if longterm warming increases gene abundance in response to drought, and if vegetation type modifies this response.We ran these models with functional gene abundances, and GHG fluxes as the response variables in turn.We used likelihood ratio deletion tests to determine whether the random effects improved model fit using the MuMIn pac ka ge in R to find both marginal and conditional R 2 values (Barton 2023 , Nakagawa and Schielzeth 2013 ).Marginal R 2 values calculate fixed effects only, while conditional effects consider both fixed and random effects.Where models were simpli-fied these values would be identical.We used Tuk e y's Honest Significant Difference to identify specific treatment effects (vegetation, drought, and warming).We also built these models for GHG emissions.
To ask if warming affects the resistance or resilience of functional gene abundances to drought, we calculated resistance and resilience of gene abundance after drought using the metrics of Orwin and Wardle ( 2004 ).Gene abundances and r atios wer e first standardized to the mean, then resistance was defined as the magnitude of the change in the variable caused by the treatment, and was calculated as , where D 0 is the difference between the well-watered control (C 0 ) and the drought (P 0 ) on Day 0, the last day of the drought treatment.Resilience by day 56 was calculated as RL , where D x is the difference between the well-watered control and the drought treatment on day 56.Drought against control pairs were compared for ambient soils, then warmed soils, so it was possible to detect whether resistance and resilience to drought were changed if there was a legacy of warming.The values are bounded between 0 and 1, where 0 is a 100% change from the control, 1 is full resistance or resilience (i.e.no change from the control), and negati ve n umbers denote an increase of the values relative to the control, indicating an overshoot of the response variable.95 % confidence intervals were calculated to explain whether there was a difference between the resistance and resilience of the gene abundances under drought and warming vs the day 0 contr ol.Wher e the error bars overlapped with 1, there was no difference between the treatment in question and the day 0 control.
To test the second hypothesis (and partially third), that GHG emissions will be associated with functional microbial genes, and that this link will be modified by prior warming and a subsequent drought, we fit linear mixed effects models testing the response of each GHG to each experimental treatment.We used day 56 only (8 weeks of being maintained at field capacity) and included warming, dr ought, v egetation type and the genes we expected to play roles in the flux.For each GHG, we repeated the model described above including each single gene in turn.We included all possible three-way interaction terms between vegetation type, drought, warming, and the single gene (for 16S rRNA gene and ITS) or gene ratio (for pmoA , nosZII , and acdS ), to reduce the potential for Type I error caused by multiple testing.Four-way interactions are shown to be difficult to inter pr et and hav e low explanatory po w er (Leuzinger et al. 2011 ).For each model, we included experimental r eplicate (i.e.matc hed dr ought-contr ol pairs) nested within origin field treatment (i.e.experimental warming or ambient temperature) as nested random factors .T hese were simplified for random effects as before.
For CO 2 flux (r espir ation) and CH 4 , we fit linear mixed effects models with vegetation type, warming and drought as factor le v el explanatory variables, and one gene for each model: 16S rRNA genes, ITS, pmoA :16S rRNA, and acdS :16S rRNA genes.For N 2 O pmoA :16S rRNA genes were substituted for nosZII :16S rRNA genes.The model was initially simplified by testing the importance of the nested random effects to conserve degrees of freedom using likelihood ratio deletion tests, but was not simplified further.We fit all models using lmer in R version 4.0.3(Bates et al. 2013 , R Development Team 2020 ).All code can be found in the supplementary material.

Treatment effects on functional genes
When we analyzed the response of the functional genes between days 0 and 56, we found no significant effect of drought at all, although ther e wer e significant effects of warming and v arying r esponses under different vegetation types (Fig. 2 ).For the raw gene abundances, we only observed a timepoint and vegetation effect (Fig. 2 A; Timepoint F 1,123 = 67.52,P < .001,Vegetation type F 3,123 = 3.55, P = .023),where on day 0 there was a decrease in 16S rRNA gene fr om wetter, mor e C-ric h soils (Latnjajaure w et meado w) to drier, less C-rich soils (Val Bercla alpine).The raw values for the genes in day 56 sho w ed an increase in 16S rRNA gene compared with day 0 for all soils, and a decrease in all gene ratios, but no strong or consistent treatment effects (Fig. 2 ).
When we calculated resistance and resilience, we found more nuanced responses to the drought.The legacy of soil warming did not affect the resistance of functional gene abundance to drought on day 0 (Fig. 3 A).None of the genes we examined were resistant to dr ought, e videnced by a lac k of ov erla p of confidence interv als with 1.Only the 16S rRNA gene showed a significant response to w arming, which w as mor e r esistant to dr ought in warmed than ambient tussoc k tundr a tr eatment plots (F 1,8 = 12.18, P = .008).No genes sho w ed resilience to drought, as gene abundances did not r ecov er to the le v el of the control after 56 days r ecov ery time (as there was no overlap of the bars with 1; Fig. 3 B).In the Latnjajaure w et meado w, the r esilience v alues of all fiv e genes wer e near zer o, indicating no impr ov ement ov er the r ecov ery period.For the other v egetation types, r esilience of gene abundances wer e mostl y negative, indicating that the abundances were higher on day 56 relative to the control, and thus signifying a loss of resilience .T here was a significant effect of warming on nosZII gene resilience in tussoc k tundr a, wher e w armed soils sho w ed evidence of gene abundances returning to the control values .T here was no recovery in ambient temper atur es, as r esilience still sho w ed increased abundance r elativ e to the control on day 56, so there was no true recovery.In the alpine vegetation type, warmed soils showed higher r esilience postdr ought than soils of the ambient plots, as shown by a significant effect of warming on resilience of pmoA genes.

Treatment effects on GHG fluxes
CO 2 gas fluxes sho w ed a significant interaction between timepoint, drought and vegetation type, but no effect of warming (Fig. 4 A; F 3,163 = 4.51, P = .005).This meant all fluxes were higher on day 56 than day 0, and while there was a weak drought effect in most v egetation types, ther e was a significant decrease in droughted soils from the wet meadow on day 56, as shown by post hoc tests.For CH 4 fluxes, there was no significant effect of drought, but we observed a three-way interaction between timepoint, warming and vegetation type (Fig. 4 B; F 3,163 = 3.13, P = .028).On day 0, all vegetation types sho w ed lo w er CH 4 in warmed soils except the wet meadow.On day 56, warmed soils led to higher CH 4 fluxes than ambient, again apart from the wet meadow.For N 2 O, we observed a time effect, where fluxes were slightly higher (A) (B) Figure 3. Resistance and resilience metrics of functional gene abundances.The warmed and ambient treatments are calculated as the droughted abundance against the control abundance for their respective warming treatment.It is then apparent if warming alters the level of gene abundance change in response to drought.Here, there is little evidence of resistance or resilience in any gene.(A) Resistance of microbial functional genes at the end of the drought, comparing the effect sizes of drought on ambient (solid) and warmed (dashed) soils.(B) Resilience of microbial functional genes 56 days after the end of the drought.Significance stars represent a significant difference between warmed and unwarmed tr eatments.Err or bars depict confidence intervals at 95%.If the error bar crosses 0, this indicates a 100% change in value from the control, meaning no resistance.If the values are negativ e, the dr ought tr eatment yielded an incr ease in the functional gene compar ed with the contr ol.Significance stars denote a significant difference between warmed and unwarmed soil ( * P < .05,* * P < .01,and * * * P < .001). on day 56 than day 0, and an interaction between drought and warming, but no vegetation effect (Fig. 4 C; Time F 1,63 = 4.48, P = .036,Drought × Warming F 1,163 = 3.93, P = .049).

GHGs are linked to functional gene abundances
Following the 56-day r ecov ery period, a set of complex interactions between functional genes and the experimental treatments influenced CO 2 emissions.16S rRNA bacterial gene abundance impacted CO 2 flux via a three-way interaction with warming and drought, and this varied across vegetation types (Fig. 5 ).Warming alone resulted in both positive and negative effects on bacterial associations with CO 2 flux, while drought alone increased CO 2 with increased 16S rRNA gene in all vegetation types .T he mixed r esults her e meant that soils that were both warmed and subjected to drought generally had a lo w er CO 2 flux, which w as positiv el y associated with bacterial abundance in all vegetation types except dry heath.
pmoA :16S rRNA gene abundance ratios was positiv el y associated with CO 2 fluxes, with a significant effect of drought in vegetation types from wetter soils (Fig. 6 A and B).We observed no effect of prior warming.In wet meadow and tussock tundra, increased pmoA :16S rRNA gene r atio decr eased CO 2 in contr ol soils while under drought, pmoA did not influence CO 2 flux (Fig. 6 A and  C).In dry heath soils, pmoA :16S rRNA gene positiv el y corr elated with CO 2 in control soils, and negatively correlated with CO 2 in droughted soils (Fig. 6 B).In alpine soils, pmoA :16S rRNA gene ratio was associated with slightly increased CO 2 flux in control soils, and str ongl y incr eased CO 2 in dr oughted soils.
We did not detect significant effects of prior warming or drought on the relationship between pmoA: 16S rRNA gene and CH 4 flux, although there was a significant interaction between vegetation type and pmoA: 16S rRNA gene (F 3,67 = 3.11, P = .032).Here, we found that for all vegetation types except wet meadow, CH 4 emissions, and ultimately CH 4 uptake, wer e r educed with an increase in the pmoA: 16S rRNA gene ratio.In the wet meadow, warming, dr ought, and v egetation type did not influence CH 4 flux.
For N 2 O fluxes, there was a consistent relationship between the 16S rRNA gene, warming and vegetation type, although we observed no effect of drought (Fig. 7 ).Bacterial gene abundance was negativ el y associated with N 2 O flux, although the interaction between vegetation type and warming meant that for wet meadow and tussoc k tundr a, warming incr eased N 2 O emissions, while for dry heath and alpine, warming decreased N 2 O (Fig. 7 B  and D).There was no association between N 2 O fluxes and the nosZII gene, the acdS gene, or the ITS fungal rRNA gene abundance ( Table S2, Supporting Information ).

Discussion
In this study, we used soil from a series of long-term open-top chamber experiments to test whether long-term warming mediated the resistance and resilience of microbial functional gene abundances involved in mitigating GHG emissions to extreme drought.We further asked whether this has net consequences for GHG fluxes.Our first hypothesis was that long term warming will increase functional gene abundances involved in GHG emissions, and that these would have lo w er r esistance, but gr eater r esilience to drought than genes from unwarmed soils.While our findings did not support this hypothesis, they provide some support for our second hypothesis that GHG emissions are higher under warming and drought, although responses varied with the vegetation type the soils were derived from.We also found complex interactions betw een w arming and drought affecting GHG emissions, sho wing that the introduction of strong climate perturbations may disrupt the link between functional genes and GHG emissions.In particular, we found strong support for the third hypothesis, and here sho w ed that vegetation type has a large modulating role in warming and drought effects on both functional genes and GHGs, with forb-rich, wetter soils being more affected by long-term warming and subsequent drought.
We hypothesized that because of the cold, wet conditions in our study sites, leading to metabolic constraints on microbes, longterm warming would increase microbial functional gene abundances, and that this would be further increased by drought.This would lead to a loss of resistance to drought.We further expected a r a pid r ecov ery of gene abundances fr om dr ought in soils with a legacy of warming.In our study sites, temper atur es ar e low, and in tussoc k tundr a and w et meado w, w aterlogging is common (Molau 2010, Scharn et al., 2021 ).These tw o factors tend to slo w microbial growth and activity (Freeman et al. 2001 ).Howe v er, in our study, similarly to that of Jeanbille et al. ( 2021), we did not observe any change in gene abundance in response to drought, while warm- ing effects occurred in ITS, pmoA and acdS .ITS gene abundance was increased under warming in the wet meadow, but decreased b y w arming in tussock tundra.The growth and extracellular signalling r equir ed for metabolic activity ar e ener geticall y costl y and occur more slowly in cold soils (Solly et al. 2017 ).Ho w ever, there has been some question over whether fungal growth in response to warming occurs due to the decline in soil moisture that often occurs in tandem with warming (Rudgers et al. 2014 ).Here, w e sho w that this is not the case: responses of fungi to warming and drought were independent, but there is a legacy effect from the dominant plant species and/or soil type .T his is consistent with findings from Rudgers et al. ( 2014 ), who sho w ed that in lo w er moisture soil conditions, arbuscular mycorrhizal fungi colonized plants mor e r eadil y, but not in soils that are solely warmed.In our study, drought was imposed on bare soil under laboratory conditions in the absence of plants, so any shift derived from the vegetation type would have occurred through the legacy impact of the different plant communities on the soil properties through factors such as residues.
The gene that encodes the CH 4 -oxidizing enzyme CH 4 monooxygenase, pmoA , incr eased immediatel y after dr ought acr oss all v egetation types except for the alpine.While some previous studies have recorded an increase in pmoA abundance after experimental warming (Yue et al. 2015 ), a recent meta-analysis found little evidence of this gene responding to warming across a range of ecosystem types (Liu et al. 2019b ).pmoA is one of a number of genes that encode enzymes that consume CH 4 , including mmoX .Ho w e v er, mmoX is found in a low number of bacterial strains and is often not detected in sufficient numbers to be able to infer any effect of climate change drivers (Gupta et al. 2012, Unger et al. 2021 ).Others hav e observ ed incr eased activity of methanotrophic bacteria as soils dry because drought enables CH 4 and oxygen to be diffused into the soil (Ran et al. 2017 ).Under dr ought, micr obes often incr ease in activity, particularly when soils were previously waterlogged, and so an increase in abundance of pmoA gene could be expected.In a similar system (tussoc k tundr a in Latnjajaur e), K eusc hnig et al. ( 2022) sho w ed that methanogens were negatively impacted by drought, leading to a net decrease in CH 4 pr oduction.The or ganic C content of the soil has been suggested to only weakly effect CH 4 emissions (Xu et al. 2020b ).T hus , as we did see a change in pmoA abundance acr oss v egetation type, it is likel y to be linked with water availability and soil drainage properties, influenced by vegetation type.
We further expected that a legacy of long-term warming would reduce the resistance but increase the resilience of the microbial community to a severe drought, and that we could use abundance of functional genes in the soil as a proxy for this, as gene expression may be up-or down-regulated in response to abiotic conditions.Our results indicate that a legacy of warming was acting on microbial gene abundance, but it did not clearly influence the response of gene expression to drought.In general, we observed an increase in pmoA and acdS genes after drought, as expected, as pmoA is associated with aerobic conditions (Knief 2015 ), while acdS is presumed to be expressed by microbes in response to stress (Singh et al. 2015 ).Further, we observ ed str onger effects of warming on microbial drought resistance in the Oro-Arctic soils compared with the alpine soils.It is possible that this is due to the larger warming effect of the O TCs , but even then, the effect of warming on pmoA and acdS was inconsistent.Many studies have observ ed r elativ el y w eak effects of w arming on the soil microbial community and associated functions (Li et al. 2020, Sarneel et al. 2020 ).A common finding is that simulated drought and warming only affected the microbial community structure when both tr eatments wer e in combination (Zhang et al. 2015 ).By day 56, we expected that they would have reduced back to control levels, but it a ppear ed that after 7 weeks of r e wetting the functional genes had increased in abundance r elativ e to the control.This was irrespective of the number or type of global change factors applied, suggesting that both warming and drought leads to a release of abiotic constraints on microbial gene expression.This effect is likely to be linked with pulses of microbial activity and subse- quent mineralization of N and C, known as the 'Birch effect' (Birch 1958 ).
Our second hypothesis was that GHG emissions would be linked with functional gene abundance, and that long-term w arming w ould incr ease net GHG flux thr ough incr eased micr obial activity.We further anticipated that drought would increase GHG flux further through exposing decomposable substrates and increasing oxygen availability.In Oro-Arctic tundra and alpine regions, there is potential for changes in abiotic conditions to result in r a pid incr eases in GHGs thr ough the decomposition of pr e viously inaccessible organic matter (Voigt et al. 2017 ).Drying and warming of soils that are often waterlogged or frozen can result in strong and long-lasting changes in net GHG emissions (Günther et al. 2020 ).In our study, we saw that vegetation type , genes , warming and drought were all important factors contributing to changes in GHG emissions, with the most important factor being vegetation type.Interestingly, one suggestion is that warming increases the competition between plants and fungi for N in these alpine or Oro-Arctic soils, which may mask the response of fungi (Xiong et al. 2014 ).Given that for most genes there was little effect of the warming and drought treatments, we can infer that these treatments impact emissions through a more complex set of mechanisms than release of constraints caused by low temper atur es.It is likel y that warming and dr ought has increased accessibility of formerly una vailable substrates , leading to increased activity and metabolism through release of the enzymatic latc h (Fr eeman et al. 2001 ).This is particularly true for CO 2 , which in our study had complex interactions with the treatments and bacteria.Ho w e v er, the str ong positiv e effect of bacterial 16S rRNA gene abundance on GHG fluxes suggests that while bacterial abundances were not altered by the treatments, bacterial activity was.In particular, the warming effect compounded the drought effect, leading to a reduction in CO 2 and loss of the bacterial signal in the data.By removing the plants in our study, we have provided str ong e vidence of a significant r elationship between bacterial abundance and CO 2 emissions under abiotic stress.N 2 O emissions occur through a primarily anaerobic pathway (Prosser et al. 2020 ), and so by droughting and warming the soil, leading to aeration, it is possible that the pathway was interrupted in our study.
Our third hypothesis was that vegetation type, and the associated edaphic characteristics, are likely to result in different gene and GHG responses to drought and w arming.Specifically, w e expected soils from wetter more forb-rich areas to be more responsive to the climate change treatments.Wetter soils were more C rich, and we demonstrated a simple gradient of soil moisture, C and N from Latnjajaure wet meadow, at the wet C rich end, to Val Bercla alpine, at the dry, C poor end.We expected to observe decreasing effect sizes as soils became drier and C poor.Contrary to this hypothesis, we found that the gene that reduced N 2 O, nosZII , was only affected by vegetation type, with a clear increase in the gene in drier alpine ar eas compar ed to the wetter forb-rich areas and not by warming and drought.We also found that CO 2 fluxes wer e str ongl y influenced by vegetation type, being higher in the dry heath than the alpine, which could be due to the differing inputs by plant species (Kuzyakov and Gavric hk ov a 2010 ).Dwarf shrubs such as those dominant in the dry heath (e.g. C. tetragona ) are known to add more polyphenolic compounds to the soil than other plant types, which can inhibit microbial activity (Ward et al. 2013 ).While the drought was ongoing (day 0), droughted soils from dry heath communities had lo w er CO 2 fluxes than soils from grasses or forb-based communities.After rewetting, the dry heath soils had higher fluxes than the other vegetation types, which could indicate that these are less stable and more susceptible to C losses during climate perturbations.Ho w e v er, the other GHGs (CH 4 and N 2 O) were highly inconsistent across vegetation type.T his ma y be because there is no clear direct link between plants and CH 4 or N 2 O emissions, which are mainly driven by specialized microbial activities.Our study is an in vitro look at the response of gene abundances to abiotic stress, and as such the common explanations for observed trends , e .g. changes in plant metabolism, root exudation and litter decomposition, do not a ppl y her e (Hammerl et al. 2019 ).Because our experiment took place in microcosms, the effects seen would be due to a legacy of the plant species, and the soil type, which could take the form of partially decomposed plant tissues, root exudates, or potentially changes in soil structure.

Conclusions
Our findings suggest that across Oro-Arctic and alpine sites, longterm increases in temperature, combined with extreme drought e v ents, could lead to changes in functional gene abundances, and that these r esponses ar e closel y linked with v egetation type.Specifically, our findings suggest that wetter ecosystems, with forb-dominated vegetation and C-rich soil, are most vulnerable to individual and combined effects of long-term warming and dr ought.Pr e vious r esearc h on highl y or ganic soils in Or o-Arctic and alpine soils points to strong and long-lasting responses of micr obial comm unities, c har acterized by functional gene abundances , to climate change .Our study, which was carried out on mor e miner al soils shows a similar r esult.Soil micr obial comm unities in very cold or waterlogged soils are not resilient to changes in soil temper atur e or water av ailability.Howe v er, in most cases, the effect on GHG fluxes may be less than expected.Taken together, our results indicate that while individual microbial functional genes may be resistant to warming and drying, there could be shifts in GHG emissions through altered microbial gene expression.

Figure 6 .
Figure 6.Flux of CO 2 influenced by an interaction between vegetation type, pmoA : 16S bacterial ratio and drought (F 3,67 = 3.13, P = .031,R 2 = 0.44).Data presented is measured on day 56 of recovery from drought.(A) Latnjajaure wet meadow, (B) Latnjajaure dry heath, (C) Latnjajaure tussock tundra, and (D) Val Bercla alpine.Values shown here are predicted values from the mixed effects models, and density graphs are presented for each axis for clarity.

Figure 7 .
Figure 7. N 2 O gas flux affected by a main effect of 16S gene abundance (F 1,67 = 7.13, P = .010),and an interaction between vegetation type and warming (F 3,67 = 3.48, P = .021,conditional R 2 = 0.51, marginal R 2 = 0.30).Data presented is measured on day 56 of recovery from drought.(A) Latnjajaure wet meadow, (B) Latnjajaure dry heath, (C) Latnjajaure tussock tundra, and (D) Val Bercla alpine.Values shown here are predicted values from the mixed effects models, and density graphs are presented for each axis for clarity.
DN A w as extracted from each soil harvest using the DNeasy Pow-erSoil Kit (Qia gen, Manc hester, UK), following the manufacturer's instructions.As the soils ranged in organic matter content, 0.25 g of soil was extracted, as recommended for highly organic soils.DNA concentration and quality were assessed using a NanoDrop 2000 (Thermo Fisher Scientific), and was PCR amplified with 16S rRN A gene primers.Amplified DN A w as shipped on ice blocks to the Hawkesbury Institute for the Environment, Western Sydney, Austr alia, wher e they were kept at −80 • C until further analysis.Each sample was diluted to 2 ng μl −1 , and quantitative PCR (qPCR) r eactions wer e carried out on fiv e genes: 16S rRNA for general bacteria, ITS for fungi, acdS for ACC deaminase (as a pr oxy for str ess), nosZII for N 2 O reductase, and pmoA for CH 4 monooxygenase .T he primers were diluted to 1:30.Each 10 μl qPCR reaction consisted of 1 μl of template (2 ng), 2.5 μl water, 0.75 μl of each primer, and 5 μl of mastermix (Light Cycler ® 480 Probes Master, Roche Diagnostics Ltd.).The pmoA qPCR r equir ed 2 μl of template (4 ng).A calibr ation curve was included for each gene from 10 9 to 10 4 gene copies.The samples were analyzed in duplicate on a 384-well plate on a Light Cycler [see TableS1( Supporting Information ) for details of primers and optimum annealing temper atur es].The amplification of the samples was analyzed using Light Cycler software.