Temporal dynamics of microbial transcription in wetted hyperarid desert soils

Abstract Rainfall is rare in hyperarid deserts but, when it occurs, it triggers large biological responses essential for the long-term maintenance of the ecosystem. In drylands, microbes play major roles in nutrient cycling, but their responses to short-lived opportunity windows are poorly understood. Due to its ephemeral nature, mRNA is ideally suited to study microbiome dynamics upon abrupt changes in the environment. We analyzed microbial community transcriptomes after simulated rainfall in a Namib Desert soil over 7 days. Using total mRNA from dry and watered plots we infer short-term functional responses in the microbiome. A rapid two-phase cycle of activation and return to basal state was completed in a short period. Motility systems activated immediately, whereas competition–toxicity increased in parallel to predator taxa and the drying of soils. Carbon fixation systems were downregulated, and reactivated upon return to a near-dry state. The chaperone HSP20 was markedly regulated by watering across all major bacteria, suggesting a particularly important role in adaptation to desiccated ecosystems. We show that transcriptomes provide consistent and high resolution information on microbiome processes in a low-biomass environment, revealing shared patterns across taxa. We propose a structured dispersal–predation dynamic as a central driver of desert microbial responses to rainfall.


Introduction
Arid lands cover approximately one-third of the terrestrial surface (Laity 2009 ) and ar e curr entl y expanding through desertification processes (Huang et al. 2016 ).(Hyper)Aridity, simultaneously caused by very low precipitation and high potential e v a potr anspir ation r ates, se v er el y limits biomass pr oduction, and leads to soil nutrient depletion (Delgado-Baquerizo et al. 2013, Maestre et al. 2015 ), resulting in habitat fragmentation where productivity is concentrated in sheltered 'islands' (Schlesinger et al. 1995 , Pointing andBelnap 2012 ).Microbial biomass is likewise constrained, sometimes forming macroscopic structures such as biological soil crusts (BSCs) and hypoliths.Open soils outside of these privileged micr oenvir onments ar e mor e extr eme and subjected to an intense desiccation stress which limits biological activity (Lebre et al. 2017 ).Despite this, micr oor ganisms in these unshelter ed ar eas ar e r esponsiv e to r ar e pulses of w ater (Gar cia-Pichel and Pringault 2001, Austin et al. 2004, Collins et al. 2014, Armstrong et al. 2016, Štoví ček et al. 2017 ) and may be important for long-term soil fertility and postrainfall grass germination (Delgado-Baquerizo et al. 2016 ).
In this environment, the expected reduced activity-or dormancy-of microbes during the long dry periods and the pr eserv ation of legacy DNA (Austin et al. 2004, Lennon et al. 2018 ), can be confounding factors for the study of microbiomes.Tr anscriptomics thus r epr esents a po w erful tool b y providing a direct insight into microbial activity due to the ephemeral nature of mRNA, going beyond community structure or genetic potential anal yses (Mor an et al. 2013, Rajee v et al. 2013, Štoví ček et al. 2017, Ste v en et al. 2018, León-Sobrino et al. 2019 ).
Given the importance of rare and stochastic water pulses in driving ecosystem functions in exposed desert soils (e .g. P ointing and Belnap 2012, Armstrong et al. 2016 ), we aimed to identify the temporal dynamics of microbial community responses to wetting in these water-deficient desert soils.We hypothesize that the activation of key cellular functions is a factor in the adaptation of desert soil microbiomes to these extreme conditions.A r epr esentativ e gr av el plain soil fr om the centr al Namib Desert was subject to an artificial rainfall pulse of 30 l/m 2 , sufficient to stimulate plant germination (Seely and Pallett 2008 ).The short-term responses of the soil microbiome were monitored by analyzing gene function through mRNA sequencing.As a result of this analysis, we propose a structured water response model with differentiated phases and trophic interactions, which may serve as a basis for impr ov ed, function-oriented anal yses of micr obiomes in arid ecosystems.

Materials and methods
Subsurface soils were collected from two adjacent 3.5 m × 3.5 m plots ( ∼10 m apart) in the central Namib Desert gravel plains (23 • 33 18 S, 15 • 3 20 E, or −23.505 • , 15.056 • ).The contr ol plot r emained dry, while the experimental plot was manually watered with 30 l/m 2 , the a ppr oximate av er a ge annual r ain r eceiv ed in this region (25 mm) (Eckardt et al. 2013 ), at T 0 (26 April 2017 10 a.m.WAT/UTC + 1), using a synthetic 'Namib Desert rain' solution, prepar ed fr om ultr a pur e DNA-fr ee water supplemented with a defined salt mixture (F rossar d et al. 2015 ).Plots were subdivided into 0.5 m × 0.5 m quadrats and r andoml y sampled, in triplicate, at specified times after water addition ( Figure S1 , Supporting Information ).A total of 20 g samples of surface (0-5 cm) soil were collected at 10 min; 1, 3, and 7 h; and 1, 3, and 7 days after simulated r ainfall, pr eserv ed immediatel y in RNAlater solution (Sigma-Aldrich, St. Louis MO, USA) and subsequently frozen at −20 • C prior to RNA extraction.Additional 200 g soil samples were collected from the same locations into WhirlPak™ bags (Nasco, Fort Atkinson WI, USA) and frozen for subsequent soil chemistry analysis.
RN A w as extracted follo wing pr otocols described pr e viousl y (León-Sobrino et al. 2019 ) from triplicate soil samples.Briefly, RNA was purified using TRIzol and treated with DNase I, followed by precipitation in 2 M LiCl to ensure complete elimination of genomic material.In order to mitigate the effect of soil chemistry heterogeneity, the two biological replicates most similar to the av er a ge composition of all sampled soils were selected for RNA extr action.Str anded, rRNA-depleted libr aries wer e pr epar ed with the ScriptSeq Complete Gold Kit (Epidemiology) (Illumina, San Diego, USA) and 150 bp paired-end sequences were read on a HiSeq4000 platform (Illumina).
Sequencing outputs were processed using the BBtools suite v. 38.26 (Bushnell et al. 2017 ) ( https:// sourceforge.net/projects/ bbmap/).Read ends below a quality Phred value of 20 were trimmed using BBDuk ; rRNA and human RNA sequences were identified and r emov ed using SILV A v .111 (July 2012) (Quast et al. 2012 ) and 5SRNA (Szymanski et al. 2016 ) databases and a curated human genome r efer ence assembl y hG19 ( https://driv e .google .com/ file/ d/ 0B3llHR93L14wd0pSSnFULUlhcUk ) (bbmap 2018 ) following recommended protocols.Optical duplicates generated by the patterned sequencing flowcell wer e r emov ed using the Clumpify function from the BBtools suite, setting the distance cutoff to 2500 pixels.Transcript assembly was performed using trans-Abyss v.2.0.1 (Robertson et al. 2010 ) for eac h libr ary.Individual assemblies wer e mer ged with the same software ( transab yss-mer ge function) to generate a reference metatranscriptome.
Reads were aligned to the reference assembly using BBmap (Bushnell et al. 2017 ) and reads for annotated regions in each libr ary wer e counted using Featur eCounts v. 1.6.3(Liao et al. 2014 ).The assembled counts matrix was a ggr egated along functional and/or taxonomic categories as r equir ed for each analysis.
Differ ential tr anscription along the time series was anal ysed with the DESeq2 v. 1.14 pac ka ge in R (Love et al. 2014 ), comparing data from the control (dry) soil samples with those from the experimental (wetted) samples from the same point in the timeseries.Tr anscripts that significantl y div er ged in abundance fr om the control at any given time were considered upregulated in response to watering (adjusted P -value ≤ .05for the likelihood ratio test).Transcripts per million (TPM) of ribosomal protein genes as a fraction of the total for their respective taxonomic group (Rp: T) w ere emplo y ed to estimate absolute activity of each group along the time course, following the premise that ribosome densities in a cell relate to metabolic activity and growth rates (Bremer andDennis 1996 , Bosdriesz et al. 2015 ).
Vir al contigs wer e identified fr om assembled tr anscriptomes using VirSorter v. 1.0.3 (Roux et al. 2015 ) and the virome database on the iVirus platform hosted by CyVerse (Bolduc et al. 2017 ).Only contigs > 1 kb, and classified as categories 1, 2, 4, and 5 wer e consider ed (pha ges and pr opha ges identified with the 'pr etty sure' and 'quite sure' qualification).To calculate the r elativ e abundances of the different viral contigs in each transcriptome, quality filter ed metatr anscriptomic r eads wer e ma pped bac k to the viral contigs with Bowtie2 v. 2.2.6, using the default parameters (Langmead and Salzberg 2012 ).The output SAM files were converted into BAM files, sorted and indexed, using SAMtools (Li et al. 2009 ).
ORFs were predicted within putative viral contigs using Prodigal (Hyatt et al. 2010 ).TPM were employed to normalize the final 'viral O TU' (vO TU) values for each sample.Predicted protein sequences were clustered with proteins from viruses in the NCBI Vir alRefSeq-pr okaryotes-v85 based on all-v ersus-all BLASTp search with an E value of 1 × 10 −3 , and clusters were defined with the Markov clustering algorithm and processed using vConTACT2 (Bin Jang et al. 2019 ).The stringency of the similarity scor e was e v aluated thr ough 1000 r andomizations by perm uting protein clusters or singletons (proteins without significant shared similarity to other protein sequences) within pairs of sequences having a significance score ≤ 1 (negative control).Subsequently, pairs of sequences with a similarity score > 1 were clustered into VCs with the Markov clustering algorithm using an inflation value of 2. The resulting gene-sharing network from vConTACT2 classification was visualized with Cytoscape software v. 3.7.0(Smoot et al. 2011 ).Refer ence sequences fr om RefSeq database that coclustered with the putative viral sequences were used to predict viral taxonomy.

Site characteristics and taxonomic analysis
The sample site (Fig. 1 A and B) is a c har acteristic Namib Desert calcr ete gr av el plain (Gombeer et al. 2015 ) with high sand composition (92 ± 1.6%) and very low organic carbon content (0.04%) ( Table S1 , Supporting Information ).Soil composition was relativ el y homogeneous in all sampled sectors and between the sample and control sites ( Table S1 , Supporting Information ).Gravimetric water content measurements showed that more than half of the water content in the surface soils (0-5 cm) was lost 24 h after the simulated 30 mm rainfall.After 3 da ys , the water content was similar to that of the dry control (Fig. 1 C).
Tr anscript r ead assembl y yielded a consensus metatr anscriptome of 208.95 Mb in which 378 802 coding regions were annotated, including 372 044 predicted protein-coding genes.On average, for the 24 sequenced libraries, 61.7% of reads could be aligned back to contigs.Functions were predicted for 29.9% of the proteincoding genes using the KEGG database (Kanehisa et al. 2016 ), and 51.6% using the CDD.56.8% of contigs wer e taxonomicall y classified at phylum le v el, and 56.3% down to family level.
A functional and taxonomic analysis of the transcriptional profiles r e v ealed lar ge differ ences in gene expr ession le v els between treatment and control soil transcriptomes within 10 min after watering (Fig. 2 ; Figure S2 , Supporting Information ).Communities from dry soils were characterized by stable (i.e.largely unchanged) tr anscription pr ofiles ov er the 7-day experimental period.In contr ast, micr obial comm unities in the water ed soils underwent an abrupt change in gene expression (Fig. 2 ) that pr ogr essiv el y r eturned to the basal state within the 7 days of the experiment.
In both watered and dry soils, Actinomycetota and Pseudomonadota were the most transcribed phyla (40% and 16% in avera ge, r espectiv el y, see Table S2 , Supporting Information ).Strikingl y, tr anscripts fr om Nitr ososphaeria, a class of ammonia oxidizing Thaumar chaeota, w ere the thir d most abundant (av er a ge 14% of classified TPM in all samples).Pr otist tr anscripts (Oligohymenophorea class and Dictyostellales), which were rare in dry soil samples ( < 0.5%), increased to > 4% within 7 h of soil watering.Conv ersel y, tr anscripts of fungal origin were more common in dry soils (1% of classified TPM) than in watered soils.
The production of ribosomes in each taxonomic group was estimated by calculating the ratio of ribosomal protein transcripts with respect to the total (Rp: T) (Bremer andDennis 1996 , Bosdriesz et al. 2015 ), and used as an indicator of ov er all metabolic activity.This ratio was stable for each taxon in the dry control samples, typically < 4% TPM (Fig. 3 ).In contrast, after water addition, Rp: T increased in all the major taxa, peaking at c har acteristic times over the course of the experiment and returning to basal values by the end of the 7 day period.Actinomycetia, Alphapr oteobacteria, and Chlor oflexia wer e r epr esentativ es of 'earl yactiv ation' gr oups, with Rp: T-v alues peaking within the first hour.Gemmatimonadetes experienced the largest relative increase in Rp: T, despite being a minor component of the soil microbiome ( < 1% TPM).A delayed rise in activity, r eac hing the highest values 7 h after the w atering, w as evident for Delta-proteobacteria, protists and Bacter oidota, especiall y those belonging to the Cytopha gia class, which constituted the principal 'late-activation' group.The Pezizomycetes, the most transcribed fungal class, also responded late to water addition.Intermediate patterns, with maximal Rp: Tvalues at 3 h after water addition, wer e observ ed for taxa such as Planctomycetia.Thaumarc haeal Rp: T-r atios sho w ed only modest changes after watering and a relatively even Rp: T throughout the experiment.

Temporal changes in core cellular functions
A gener al r eduction of str ess r esistance gene tr anscripts (e.g.groEL , dnaK , clp , and tre ) was observed immediately after watering.Trehalose synthesis genes ( tre ), which drive solute accumulation under conditions of water stress (Lebre et al. 2017 ), declined in r elativ e abundance in water ed soils.Tr anscripts for c ha per ones groEL and dnaK , and the clp pr otease involv ed in protein misfolding contr ol wer e also r educed.The most conspicuous c hange in cor e str ess r esistance systems w as, ho w e v er, that of the heat-shock protein HSP20 genes (KO13993, CD223149, or CD278440).Genes for this ATP-independent c ha per one experienced the lar gest and most consistent transcript reduction across all significant (average ≥ 1% of transcripts) bacterial taxa after watering (Fig. 4 ).
Motility r elated tr anscripts wer e r a pidl y affected b y soil w etting.Fla gellar gene expr ession was upr egulated within the first hour in the dominant Actinomycetota and Alpha-proteobacteria taxa (Fig. 4 ) and in other taxa such as Planctomycetota and Bacillota.Expression of type IV pilus genes , in volved in gliding motility, was also r a pidl y upr egulated after water addition, most notabl y in Actinomycetota.Sim ultaneousl y, hallmark c hemotaxis genes mcp and motB were upregulated during the initial period in Actinomycetota, Proteobacteria, Bacteroidota, and Planctomycetota.A significant upregulation of Eukaryotic adhesion and cytoskeletal components, myosin, actin, and tubulin, was evident 3-7 h after soil wetting (Fig. 4 ), but was limited to nonfungal taxa, particularly in Oligohymenophorea (ciliate) and Dictyosteliales (slime mould).
Another highly significant increase in transcriptional activity observed 3-7 h after water addition involved interbacterial competition and predation genes .T hese included type VI secretion systems (T6SS) and vibriolysin genes from the order Myxococcales (Delta-proteobacteria), and T6SS and serralysin genes from Alpha-pr oteobacteria.Heightened T6SS tr anscription was also observed in several other bacterial taxa, such as Planctomycetota, Gemmatimonadota, and Gamma-proteobacteria.

Carbon utilization
Micr obial photosynthetic pr ocesses ar e limited in hyperarid soils, but are strongly activated after wetting (Warren-Rhodes et al. 2006, Tracy et al. 2010, Gwizdala et al. 2021 ).Ho w e v er, sur prisingl y, tr anscript data did not show a significant r elativ e incr ease in ov er all c y anobacterial tr anscripts acr oss the 7-day dur ation of the experiment.Cyanobacterial Rp: T-ratios remained low throughout (Fig. 3 ; Table S2 , Supporting Information ), suggesting that c y anobacterial metabolism w as lar gel y unaffected by water addition in these soils.
Core carbon-fixation genes, including RuBisCO and carboxylase genes from chemoautotrophic pathwa ys , consistently sho w ed r educed r elativ e tr anscription in wetted soils .T haumarchaeal carbon fixation was seemingly also affected by watering, reflected in the transient inhibition of 4-h ydroxybutyryl-CoA deh ydr atase fr om the 3-hydr oxypr opionate/4-hydr oxybutyr ate cycle following inundation.Ov er all, we observ ed an inhibition of both photosynthetic and nonphotosynthetic carbon fixation mechanisms immediately after soil saturation ( Figure S3A , Supporting Information ).Conv ersel y, tr anscription of carbon fixation genes r ecov er ed by the end of the 7-day experimental period, after soil water content had returned to basal le v els (Fig. 4 ; Figur e S3A , Supporting Information ).
Dr amatic incr eases in CO 2 emissions fr om ne wl y wetted desert soils, attributed to degradation of dissolved organic matter, have been widely reported (Austin et al. 2004 ).Soil chemistry analyses, ho w e v er, sho w ed no significant reduction in soil Total Organic Carbon pools ( Table S1 , Supporting Information ), possibly due to the very low organic carbon levels present in these soils.In our transcript data, indicators of biomass degradation, including carbohydrate and peptide transport systems (e.g.rbsB , xylF , and livK , dppA ) significantl y incr eased immediatel y after water addition.The Bacteroidota phylum, dominated by the Cytophagales, sho w ed dr amatic upr egulation of subtilisin pr otease tr anscripts and components of the protein and carbohydrate import machinery ( ragA / susC CD274948; susD CD185760).

Nitrogen and phosphate utilization
Transcript data suggested that Thaumarchaea were the main drivers of nitrogen cycling in the soil.Ammonia monooxygenase and NO-forming nitrite reductase ( amo and nirK ) transcripts, originating lar gel y fr om Thaumarc haea, wer e among the most abundant ov er all.These declined in r elativ e abundance after soil watering (Fig. 5 A; Figure S3B , Supporting Information ).All other transcripts implicated in ammonification, such as nitrate and nitrite reductases narGH and nirAD (mainly transcribed by Nitrospirota and Actinom ycetota, respecti vely), and nitric o xide reductase ( norC ), were also reduced upon watering.Conversely, peptide tr ansporter tr anscripts ( livKH , dppA ) significantl y incr eased in r esponse to water addition for se v er al taxa (Actinomycetota, Alpha-, Beta-, and Delta-proteobacteria), suggesting that or ganic nitr ogen dominated N acquisition processes.
Changes in phosphorus acquisition pathways were also identified.Phosphate acquisition after water addition was dominated by the proteobacterial ugpB gene from the sn -gl ycer ol 3-phosphate (G3P) tr ansport system, particularl y in Alpha-pr oteobacteria, follo w ed b y the pst inor ganic phosphate tr ansporter (Fig. 5 B).By

Analysis of viral transcripts
A total of 68 contigs were characterized as being of viral origin using VirSorter ( Figure S5 , Supporting Information ).In both dry and watered soil samples, viral contigs were significantly lo w er than bacterial contigs (ANOVA, P < .01),r epr esenting an av er a ge of 0.12 ± 0.07% TPM of viral protein transcripts.Read numbers remained low in all dry soil samples.An initial increase in viral RNAs ( ∼2.2-fold at 10 min) was followed by a 12-fold decline over 7 h ( Figure S4 , Supporting Information ).A second increase of viral read numbers ( ∼6-fold) occurred between 7 h and 7 da ys .
To investigate the diversity of the 'active' viral population, we used a genome-based network analysis of the shared protein content with the prokaryotic viral genomes (RefSeq v85).This analysis grouped 35 viral contigs into viral clusters (VCs) ( Figure S5 , Supporting Information ).In the network, 10 VCs containing viral contigs from our study were predicted, seven of which did not belong to VCs with RefSeq virus genomes but instead clustered together into novel VCs, and three of which could be assigned taxonomy at the family level ( Figure S5 , Supporting Information ) as members of the Caudovirales ( Siphoviridae and Leviviridae ).

Discussion
Studies of the microbial ecology of desert eda phic nic hes hav e tended to focus on biological 'hotspots': hypoliths, soil crusts, and soils in the immediate vicinity of plants (e .g. P ointing and Belna p 2012 , Mar asco et al. 2018 , Ramond et al. 2018 ).While the use of high-throughput sequencing techniques (Crits-Christoph et al. 2013, Fierer et al. 2007, 2012, Jordaan et al. 2020, Vikram et al. 2016, Marasco et al. 2022 ) has greatly expanded of knowledge of the microbial ecology of desert soil niches, most studies have used meta genomics.Total RNA sequencing (metatr anscriptomics) is, ther efor e, a v aluable tool for monitoring micr obial functionality at a high temporal resolution (Moran et al. 2013 ), particularly since mRNA is only generated by active organisms and is ephemeral, leaving little to no legacy signal (León-Sobrino et al. 2019, Rajeev et al. 2013, Ste v en et al. 2018 ).
Water e v ents in deserts may trigger different biological responses depending on their intensity and duration (Schwinning andSala 2004 , F rossar d et al. 2015 ).Whereas certain processes respond to small moisture events, especially on the immediate surface of soils and r oc ks, other ecological r esponses and nic hes r equir e a larger pulse to be activated (Pointing andBelnap 2012 , Schwinning andSala 2004 ).Grass germination, for example, is triggered after around 20 mm rainfall (Seely and Pallett 2008 ).For this work, focused on subsurface soils, we chose the central hyperarid zone of the Namib Desert, which receives water approximatel y equall y fr om r ain and fog (Ec kardt et al. 2013 ).This 'neutral' location would optimize the r esponsiv eness of the soil microbiome to the water input (F rossar d et al. 2015 ) and serve as a baseline r efer ence for either r ain-or fog-sha ped soil micr obiomes.We decided to saturate the soil with 30 l/m 2 of water to ensure a complete biological activation and homogeneous sample conditions in the first centimetres of soil, rather than an arbitrary point along the gradient of possible precipitation events.Although the most fr equent r ains in deserts ar e < 5 mm (Pointing and Belnap 2012 )

The HSP20 chaperone is important for microbial life in desiccated soils
One of the projected effects of water addition was an a ppar ent reduction in cellular stress as suggested by the immediate downregulation in transcription of stress-resistance genes, such as trehalose biosynthesis genes or c ha per ones .T he most conspicuous of these changes was the abrupt reduction in tr anscription, acr oss all major bacterial taxa, of the small ATP-independent heat-shock protein HSP20.This chaperone has been c har acterized as a broadspectrum bacterial stress resistance mechanism (Bepperling et al. 2012 , Haslbeck andVierling 2015 ).Functional characterization of this pr otein r emains limited, but our data suggest that this protein is specifically involved in desiccation stress adaptation in many bacteria.

Desert soil microbes are sequentially activated after a water event
Data from the control (unwatered) soil site confirmed the presence of a diverse and functionally active microbial community in desiccated hyperarid desert soils (Gunnigle et al. 2017, Jordaan et al. 2020, León-Sobrino et al. 2019, Sc hulze-Makuc h et al. 2018 ).Our data also suggest a le v el of remarkable 'metabolic readiness', with a dramatic increase in transcription associated with previousl y inactiv e or undetected taxa occurred within 10 min after water addition.We note that transcriptional response rates may be e v en faster, giv en that 10 min was the first sampling time-point.In pol yextr eme hyper arid desert soils, wher e most micr oor ganisms remain in a state of metabolic dormancy (Bär et al. 2002, Lebr e et al. 2017 ), suc h a r a pid and opportunistic response to the sudden availability of water is clearly an ada ptativ e adv anta ge to access and utilize more fa vourable , and newly a vailable , ecological substrates and niches.Water addition led to a general increase in the relative abundance of ribosomal protein transcripts (Rp: T), which we interpret as an increase in cellular activity (Bosdriesz et al. 2015 , Bremer andDennis 1996 ).Cellular activity le v els r eturned to basal (contr ol) le v els within 7 da ys , in parallel with the desiccation of the soil samples .T he c har acteristic tempor al patterns after watering and contrasting stability of control samples suggest that Rp: T is indeed a regulated factor in bacterial cells, supporting its use as a global activity indicator.Our data are consistent with the paradigm that desert ecosystems and their indigenous microbiota are both resilient and water-pulse responsive (e.g.Belnap et al. 2005, Noy-Meir 1973, Armstrong et al. 2016 ).
Inter estingl y, v arious gr oups of taxa r eac hed maxim um Rp: Tv alues at differ ent times, suggesting a contr olled pattern of functionality reminiscent of a stepwise model where ecosystem functions gr aduall y e volv e as a function of the dur ation and intensity of the water pulses (Schwinning andSala 2004 , Placella et al. 2012 ).The most immediate microbial response was characterized by transcription of genes implicated in the motility apparatus (type IV pili in Alpha-proteobacteria and flagella in Acti-nomycetota).Accordingl y, c hemotaxis genes fr om the same taxa wer e upr egulated, although in a less dramatic manner.It has been pr e viousl y noted that one of the main impacts of water inundation is increased soil particle connectivity, providing access to ne w nic hes and solubilized nutrients (Schimel 2018 ).Actinomycetota and Alpha-proteobacteria have been reported as the dominant active taxa in desiccated soils (León-Sobrino et al. 2019 ), possibl y uniquel y positioned to access new and mor e favour able niches during periods of interconnection associated with the water-saturated state.
A significant, but delayed, transcriptional activation was observed in the nonfungal microbial eukaryotes (e.g.protists) and Delta-proteobacteria; i.e. 3-7 h after water addition.The former wer e mostl y c har acterized by structur al gene tr anscripts fr om the cytoskeleton, a generic indication of ov er all cellular activity (cell motility and/or cell division).Upregulated Delta-proteobacterial tr anscripts wer e pr edominantl y deriv ed fr om the Myxococcales, an order of well-known predatory bacteria (Jurke vitc h and Da vido v 2007 , Shimkets et al. 2006 ).The dr amatic incr ease in protist and myxobacterial activity is str ongl y suggestiv e of predatory behaviour (Thiery and Kaimer 2020 ), possibly triggered by increases in prey abundance (i.e.Actinomycetota and Alpha pr oteobacteria populations) r ather than just by soil r ehydr ation.
T his study pro vides , to our knowledge , the first temporal, rather than spatial, assessment of phages in a desert edaphic environment (Hwang et al. 2021, Zablocki et al. 2016, 2017 ).In parallel with the r a pid c hanges in bacterial metatr anscriptomic patterns, the phage population responded within 10 min after water addition, follo w ed b y a shar p decr ease.We hypothesize that, mirr oring the initial burst of motility in some microbial taxa, there might be a readiness for the rapid generation of virions and expansion to new hosts.

Fungal and cyanobacterial soil populations are not significantly activated by water
Among the edaphic taxa that were essentially nonresponsive to w ater addition, w e particularly identified the Cy anobacteria and most fungi, with almost none of their genes being differ entiall y upregulated at a statistically significant level.This is a surprising result: we anticipated that water addition would trigger a significant and r a pid incr ease in primary pr oduction mark ers link ed to c y anobacterial and photosynthetic activity, as pr e viousl y observ ed in BSCs and hypolithic comm unities (Angel and Conrad 2013, Oren et al. 2017, Pringault and Garcia-Pic hel 2004, Rajee v et al. 2013, Ste v en et al. 2018, Warr en-Rhodes et al. 2006 ).The almost complete absence of water input-related activation of c y anobacterial functionality suggests that primary pr oductivity in hyper arid soils may not be driven b y c y anobacteria and is consistent with pr e vious observ ations showing that hypolithons (and maybe other cryptic communities) are the foundation of productivity after rain events in the Namib Desert (Ramond et al. 2018 ).

Water pulses shift microbial C, N, and P nutrient utiliza tion pa tterns
A peak of r espir ation during water-triggered blooms in a wellknown phenomenon (Austin et al. 2004, Placella et al. 2012 ).We also expected a significant increase in primary productivity, since photosynthetic processes are highly sensitive to water (e.g.Br oc k 1975, Ste v en et al. 2018, Warr en-Rhodes et al. 2006 ).Ho w e v er, markers for photosynthetic and c hemoautotr ophic carbon fixation (the latter being active in desiccated periods; León-Sobrino et al. 2019, Sghaier et al. 2016 ), were either not activ ated or significantl y suppr essed.T hus , the carbon balance during these water pulses appears to be almost entir el y negativ e in the bulk soil.Our measurements of the organic carbon content of soils indicated very low amounts well below 0.1% wt. ( Table S1 , Supporting Information ), to fuel this activity bloom.Open soil communities may be dependent on carbon input from alternative sources, such as sporadic vegetation growth, productive cryptic nic hes suc h as hypolithons and BSCs (Armstr ong et al. 2016, Ramond et al. 2018 ) and/or little-known autotr ophic pr ocesses suc h as trace gas chemotrophy (e.g.Greening andGrinter 2022 , Jordaan et al. 2020 ).
Nitrogen cycling genes, particularly those involved in inorganic nitr ogen acquisition (i.e.nitr ate, via nitr ate r eductases), wer e not significantl y upr egulated after water addition.We note that compositional effects-since we measur e r elativ e, r ather, than absolute, abundances-might be responsible for the apparent reduction in N cycling transcripts .T his was nonetheless surprising, as activ e N miner alization, nitrification and N loss pr ocesses ar e of-ten increased in arid soils in response to rainfall (Austin et al. 2004, Belnap et al. 2005, Ramond et al. 2022 ), due to both biological activity and solubilization of nitrate, forming substantial r eserv es of under gr ound N essentiall y in the form of nitr ate (Gr aham et al. 2008, Walv oor d 2003 ).A metabolic switch to nitrogen acquisition fr om or ganic substr ates was str ongl y suggested by the upregulation of peptide transporter genes, mirroring the situation observed for carbon acquisition.It is most pr obabl y linked to the important release of N-rich compounds (e.g.proteins) following the intense death of soil microbial biomass via osmolysis (i.e.betw een a thir d to half of it; Belnap et al. 2005 ).The transient reduction in autotrophic C and N fixation after watering may be explained in terms of ener gy efficiency, wher e the sudden availability of 'energy-rich' substrates provides a favoured heter otr ophic r esource ov er ener geticall y expensiv e autotr ophic pr ocesses (Fuc hs 2011 ).
The addition of water triggered an upregulation of genes involved in inorganic phosphate transport and a downregulation of those implicated in organic phosphonate acquisition.We speculate that the solubilization of inorganic phosphate from soil particles displaces phosphonates as the pr eferr ed P source (Sc howanek and Verstraete 1990 ). Noticeable exceptions were the Alphapr oteobacterial taxa, whic h a ppar entl y favour or ganic G3P as a pr eferr ed P source, both in desiccated soils and after wetting (León-Sobrino et al. 2019 ).

A conceptual response model of desert soil edaphic microbial communities to water
From a composite analysis of our metatranscriptomic data, we propose a rainfall response model for desert soil microbiomes (Fig. 6 ).Immediately after soil wetting ( ≤ 10 min), some bacterial taxa (particularly Actinomycetota and Alpha-proteobacteria, that show significant c hemoautotr ophic ca pacity in dry soils; León-Sobrino et al. 2019 ) reduce autotrophic carbon fixation activities, acti vate cellular uptak e mec hanisms and enga ge in dispersal using both natatory (flagella) and gliding (type IV pili) mechanisms, pr esumabl y in order to colonize ne w nic hes and access new substr ate r esource pools .T her e is a concomitant incr ease in activ e pha ge particles, whic h ma y add to the turno v er of or ganic matter and the pr edatory pr essur e on microbial populations upon the initial r a pid wav e of dispersal.
Following this r a pid dispersal burst, ∼3-7 h after water addition, predatory and saprophytic microbial taxa are activated.These predators include eukaryotes, especially ciliates (Oligohymenophorea class), Dictyostellid amoebae, Delta-proteobacteria (myxobacteria), and Bacteroidota (Cytophagia class).Simultaneousl y, and pr esumabl y in r esponse to the activ ation of pr edators, se v er al bacterial groups upregulate the transcription of defensive systems, most notably T6SS.
In the final stage of the wetting-drying cycle ( ∼7 days after water addition), when soils are effectiv el y dehydr ated to pr e watering le v els, autotr ophic carbon fixation processes and nitrogen cycling ar e r eactiv ated in the bacterial community, along with certain desiccation str ess r esistance mec hanisms, most especiall y HSP20.

Conclusion
In this 7-day in situ metatr anscriptomics study, we pr ofiled the dominant microbial processes induced by precipitation on a hyperarid desert soil.Overall, RNA analysis proved to be a robust tool for micr obiome pr ofiling fr om low biomass environmental samples .T he dynamic and function-targeted nature of this mRNA- dependent analyses allo w ed us to ca ptur e short-term v ariations in microbiome structure and function and offers a valuable complementary analysis tool for environmental microbiomics.
The transcriptomes described a cyclical pattern of community functionality starting immediately after water addition, and returning to the basal state following soil drying.We show robust evidence of short-term temporal succession and, by implication, tightl y r egulated pr ocesses.Shar ed functional r esponses across taxa suggest that some functions are important for adaptation to these ecosystems.In particular, we observed a dispersionpredation dynamic and a strong shift to w ar ds a heterotrophic lifestyle upon watering.Some c har acteristic 'dry genes' were also documented, particularly those involved in chemoautotrophic carbon fixation (in accordance with pr e vious r eports by León-Sobrino et al. 2019 ), and also HSP20, which might be a k e y c ha perone for adaptation to desiccation stress in many bacterial taxa.T hese function-o v er-taxonomy observ ations may be conserv ed in micr obiomes fr om other locations sharing similar environmental conditions , and we en vision that this study can serve to inspire future work in that direction.

Figure 1 .
Figure 1.Sample site and environmental conditions of the sampled soils.(A) Map of the Namib Desert gravel plain location.Modified from European Space Agency, ESA/Envisat CC BY-SA 3.0 IGO.(B) View of a re presentati ve portion of the gravel plain sampling site.(C) Environmental conditions over the experimental period: Air temper atur e, humidity, and wind were recorded by the nearby Gobabeb meteorological station [Southern African Science Service Centre for Climate Change and Adaptive Land Management (SASSCAL), station 8893].

Figure 2 .
Figure 2. Principal components analysis of transcriptomes according to KEGG Ortholog functional annotations across soil samples.

Figure 3 .
Figure 3. Ribosomal protein gene transcripts as a fraction of the total (Rp: T) among the most transcriptionally active microbial classes in the Namib soil community ( > 1% TPM average in dry or watered samples).Left panel shows early response classes whose Rp: T-ratios peak within the first hour after watering.Right panel includes classes with Rp: T maxima 3 and 7 h after watering ( middle and late response taxa, respectively).

Figure 4 .
Figure 4. Tempor al tr anscription patterns of the 30 most variable genes among those with significantly differential transcription along the time series.Transcript data was aggregated along taxonomic (class) as well as functional (KEGG Orthologs) groups for the analysis.Values were normalized using the Variance Stabilizing Transformation ( DESeq R package).
, heavy r ain e v ents > 20 mm also occur in natur e (Ec kardt et al. 2013 , Armstrong et al. 2016 ).

Figure 5 .
Figure 5. (A) Nitrogen cycling and (B) phosphate assimilation system transcription in the subsurface soil microbial community.Average TPM measur ements ar e pr ovided for the dry plot samples (or ange arr o ws) and w etted samples up to 7 h postw atering (gr een arr ows).Asterisks indicate significant differ entiall y tr anscribed KOs at comm unity le v el at an y point along the time series.P anel (B) figur e modified fr om León-Sobrino et al. ( 2019 ).

Figure 6 .
Figure 6.Response model of microbial communities to water events in hyperarid desert soils.