The root system of Pappostipa frigida as a hotspot of active bacterial communities in the Andean steppe of the Atacama desert

Abstract

Background

In arid ecosystems, soil microorganisms experience prolonged drought and nutrient limitation, resulting in widespread dormancy and a decoupling between microbial presence and activity. The rhizosphere is a resource-rich microenvironment where plants, inputs may alleviate these constraints and regulate microbial metabolic states. However, it remains unclear whether root system effects in extreme environments reflect taxonomic turnover or the activation of dormant soil bacteria. Here, we investigated whether the rhizosphere-surrounding soil (RSS) of Pappostipa frigida, a dominant perennial grass of the Andean steppe of the Atacama Desert, reshapes soil bacterial communities primarily by promoting the activation of dormant taxa rather than by recruiting distinct bacterial populations, and how this process influences microbial diversity, activity, and functional potential.

Results

We employed a combination of soil physicochemical analyses, RNA/DNA metabarcoding, shotgun metagenomic sequencing and culture-based assays to compare bulk soil (BS) and RSS bacterial communities. The active bacterial community in the RSS exhibited significantly higher diversity and Shannon index than those in BS, whereas total (DNA-based) communities showed no significant differences between soil compartments. Taxonomic structure was primarily shaped by soil compartment rather than nucleic acid type, and active bacteria in the RSS showed a stronger correlation with total bacterial populations than those in the BS. Notably, 65% of putatively dormant bacterial taxa in BS were detected as active in RSS, and 75% were recoverable in culture. Additionally, 24% were members of PGP bacterial families. Functionally, the bacterial communities of the BS were enriched in sporulation genes, whereas active bacterial communities in the RSS showed higher abundances of genes associated with osmotic stress tolerance, siderophore synthesis, and resuscitation-promoting factors.

Conclusions

Our results indicate that the root system of Pappostipa frigida functions as a localized hotspot of microbial activity in an extreme arid environment by promoting the activation of dormant members of the soil microbial seed bank rather than by recruiting distinct taxa. By integrating microbial activity, functional traits, and culturability, this study highlights the central role of plant-associated microenvironments in regulating microbial life-history strategies in drylands and provides a mechanistic framework for understanding plant-microbe interactions under chronic environmental stress.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40793-026-00867-x.

Background

Arid and semi-arid ecosystems cover approximately 45% of the Earth’s terrestrial surface and are among the most climate-sensitive biomes worldwide [1, 2]. In these environments, soil microorganisms play a central role in sustaining ecosystem processes such as nutrient cycling, soil structure maintenance, and plant productivity [3, 4]. However, chronic water limitation, high salinity, strong temperature fluctuations, and nutrient scarcity impose severe constraints on microbial growth and activity, leading to a decoupling between microbial presence and microbial function [5, 6].

A defining feature of microbial life in dryland soils is the prevalence of dormancy. It is estimated that only a small fraction of soil microorganisms is metabolically active at any given time, while the majority persist in inactive or low-activity states that contribute to a microbial seed bank [7, 8]. Dormancy allows microorganisms to survive prolonged periods of environmental stress while maintaining diversity and functional potential at the community level [9, 10]. Transitions between dormant and active states are governed by environmental cues such as moisture pulses and nutrient availability, and mediated by molecular mechanisms including sporulation, osmoprotectant synthesis, toxin–antitoxin systems, and resuscitation-promoting factors [11–13]. Despite their recognized importance, the spatial contexts in which these transitions occur in natural arid ecosystems remain poorly understood.

Perennial plants represent key biological structures in drylands, often creating localized patches of enhanced biological activity known as fertility islands [14]. Through root exudation, litter deposition, and modification of soil physicochemical properties, plant root systems generate microenvironments that differ markedly from the surrounding bulk soil in terms of nutrient availability, pH, salinity, and moisture [15]. These rhizosphere-associated conditions can alleviate resource limitation and physiological stress for microorganisms, potentially promoting microbial activity even under harsh climatic conditions [16]. As a result, the rhizosphere-surrounding soil (RSS) may function not only as a zone of microbial enrichment but also as a regulator of microbial life-history strategies, including the activation of dormant populations.

Most rhizosphere studies in arid systems have focused on taxonomic shifts between bulk soil and root-associated communities [15, 17–19]. However, changes in community composition do not necessarily reflect changes in microbial activity or functional contribution, particularly in ecosystems where dormancy is widespread and relic DNA can obscure signals of living communities [20]. Distinguishing between recruitment of distinct taxa and physiological activation of shared taxa is therefore critical for understanding how plants influence soil microbiomes under extreme environmental constraints. In this context, RNA-based approaches provide a valuable proxy for microbial activity, as RNA is rapidly degraded and predominantly synthesized by metabolically active cells [7, 21].

The Andean steppe of the Atacama Desert constitutes an extreme natural laboratory for examining these processes. This high-elevation ecosystem is characterized by chronic aridity, high salinity, strong diurnal temperature oscillations, and a short, unpredictable rainy season that triggers brief pulses of biological activity [22, 23]. Pappostipa frigida, a dominant perennial grass endemic to this region, persists across a wide altitudinal and precipitation gradient and plays a key role in structuring plant communities [17, 23]. Previous studies have shown that the rhizosphere-surrounding soil (RSS) of P. frigida differs from adjacent bulk soil (BS) in nutrient availability, metal content, metabolite composition, and bacterial community structure [24, 25]. Yet, whether these conditions primarily reshape microbial communities through taxonomic replacement or through modulation of microbial physiological states remains unresolved.

In this study, we investigate how the RSS of P. frigida influences the activity and functional organization of soil bacterial communities in the Andean Atacama steppe. We focused on the period immediately after the rainy season. Under these conditions, we hypothesize that, in this extreme arid ecosystem, the root system of P. frigida acts as a functional ecological filter that reshapes microbial communities primarily by modulating bacterial physiological states rather than by recruiting distinct taxa. Specifically, we propose that the RSS selectively reactivates members of the soil microbial seed bank by favoring populations equipped with traits that facilitate dormancy exit, stress tolerance, and resource acquisition.

To test this hypothesis, we integrate soil physicochemical characterization, shotgun metagenomics, DNA- and RNA-based 16 S rRNA metabarcoding, absolute quantification, and culture-based assays. We first define abiotic differences between BS and the RSS to establish the environmental context in which microbial communities develop. We then compare the functional potential of bacterial assemblages, focusing on genes associated with dormancy, resuscitation, osmotic stress tolerance, and siderophore-mediated iron acquisition. Next, we assess how these functional patterns are reflected in the diversity and structure of total and active bacterial communities. Finally, we evaluate whether bacterial populations that appear inactive in BS or RSS can be reactivated under RSS conditions or in vitro, thereby testing the role of the P. frigida root system as a driver of microbial state transitions rather than taxonomic turnover.

Methods

Sampling site and sampling procedure

The sampling site was located at 4000 m above sea level in the Andes of the Atacama Desert, close to the border between Chile and Argentina (22°55’01.8’’S 67°53’58.0’’W) (Fig. 1A). In this place, the soils correspond to the Entisol soil order [26] and they are dominated by Andean grasslands composed mainly of perennial grasses, including P. frigida [22, 23, 27], one of the plants that had the highest coverage, persistence, and elevation distribution range (3600–4500 m above sea level) [23, 28, 29] (Fig. 1A, B and C). Soil sampling was conducted in April 2022, after the rainy period (January and March). We demarcated a 10 m² area in which samples were taken from five randomly selected P. frigida plants, each approximately 50 cm tall and with root systems about 15 cm deep. Following the protocol described by Fernández-Gómez et al. [17], we defined two compartments according to their proximity to the roots: the rhizosphere-surrounding soil (RSS), which is soil attached to the roots, and the bulk soil (BS), which is plant-free soil. For each plant, the roots were gently shaken at least three times to remove soil that had adhered due to compaction rather than interaction with the roots. The roots were then vigorously shaken at least five additional times to collect soil that was more tightly bound, resulting in the RSS fraction (25 g) [17, 24]. In addition, BS samples (50 g) were collected at 10 to 15 cm depth from the ground and at least 1 m away from each sampled plant. They were deliberately taken from visually plant-free patches, avoiding the immediate vicinity of other plants. All soil samples (n = 10) were immediately frozen in liquid nitrogen and then kept on dry ice until their arrival at the laboratory, where they were stored at − 80 °C. Another ten 10 g samples of BS and RSS were stored at 4 °C for absolute DNA quantification and for determining soil pH, conductivity, macro- (K, P and Ca) and micro-nutrients (Cu, Fe, Mn and Zn). Additionally, 300 g of a pool of collected BS samples were stored at 4 °C to prepare Soil Extract Media (SEM).

Fig. 1.

Sampling site and soil properties. A Vegetation floors in the northern part of Chile and the distribution of Pappostipa frigida. The red triangle indicates the designated sampling area for P. frigida. B Representative photograph of the sampling area dominated by P. frigida; a 50 cm scale bar is shown, corresponding approximately to the diameter of an adult plant. C Field photograph of an individual P. frigida plant with exposed root system and surrounding soil; a 50 cm scale bar is included to provide size reference and allow comparison with panel B. D Principal component analysis (PCA) of soil properties. The points represent the soil samples and the arrows indicate the direction and contribution to the principal components of each chemical property. BS: bulk soil, RSS: rhizosphere-surrounding soil, WC: gravimetric water content, EC: electrical conductivity

Soil properties

To determine the soluble fraction of metal composition of the soil, individual BS and RSS samples (n = 5 per compartment) were processed following Fernández-Gómez et al. [17]. Briefly, 1 g of soil was resuspended in 1 mL of distilled water and homogenized for 2 h at room temperature. After mixing, the samples were centrifuged at 11,440 g for 10 min in a Hettich Universal 32R. The soluble fraction was recovered and metals were measured by total reflection X-Ray fluorescence (TXRF) in a Bruker S2 PICOFOX following a previously published protocol [30] and adding gallium as an internal standard element. Soil pH was measured on individual samples by mixing 2 g of soil with 5 mL of Milli-Q water and homogenizing for 2 h in a S1000 Gyrotwister at 60 rpm, followed by 2 h of sedimentation. pH was measured in triplicate for each sample in a Thermo Scientific Orion 3 star benchtop pH meter [17]. Electrical conductivity was also determined on individual samples by diluting 5 g of soil in 25 mL of double-distilled water, incubating for 2.5 h at room temperature under continuous rotation. The samples were then centrifuged at 6000 × g for 10 min, and the supernatant was recovered. Ten mL of the supernatant was used to measure conductivity using a Hanna bench conductivity meter HI 4321. All physicochemical measurements were performed in duplicate per sample. Gravimetric water content was determined for each individual sample by calculating the ratio between water weight (fresh soil weight minus dry soil weight) and dry soil weight after drying for 48 h at 60 °C until constant weight was reached.

Differences between the metal soil composition and pH were evaluated using the Wilcoxon rank-sum test corrected for multiple comparisons with the Benjamini-Hochberg method [31]. In addition, a principal component analysis (PCA) was performed using R version 4.2.3 on the centered and scaled physicochemical properties to visualize their influence on the soil samples.

Soil RNA and DNA extraction

To simultaneously isolate RNA and DNA from soil samples, the PowerSoil total RNA kit (Qiagen) and PowerSoil DNA elution kit (Qiagen) was used following the manufacturer’s recommendations with minor modifications. Briefly, 3 g of soil were used as an initial sample, and the mechanical lysis was performed with a mixture of glass beads and bead solution that was processed for three pulses of 30 s at 6 m/s in a FastPrep-24 (MP Biomedicals). Nucleic acids were quantified by fluorometry using the Qubit dsDNA BR Assay Kit (Invitrogen) for DNA samples and Qubit RNA BR Assay Kit for RNA samples. Integrity of the DNA samples was corroborated in a 1% agarose gel, while an agarose/formaldehyde gel was used for the RNA samples. Negative extraction controls were included by performing DNA/RNA extractions in parallel using only the buffers provided in the extraction kit, without biological material. All extracts were subsequently subjected to PCR amplification of the V1-V3 region of the bacterial 16 S rRNA gene, using primers 28 F (5′-GAGTTTGATCNTGGCTCAG-3′) and 519R (5′-GTNTTACNGCGGCKGCTG-3′), in order to assess potential reagent or procedural contamination. The DNA and RNA samples were stored at -20 °C and − 80 °C, respectively. DNA samples were used for shotgun metagenomics and 16 S rRNA gene metabarcoding, whereas RNA samples were used only for 16 S rRNA metabarcoding. To obtain cDNA, RNA was pretreated with DNase I (RNase-free) (New England Biolabs) and reverse transcribed using the High-Capacity RNA-to-cDNA kit (Applied Biosystems).

Shotgun metagenomic sequencing

Library preparation and shotgun metagenomic sequencing were performed on ten individual DNA samples (five BS and five RSS samples) using the Illumina NovaSeq platform (2 × 250 bp) at Molecular Research LP (Mr. DNA Lab), following the manufacturer’s guidelines. Illumina reads were quality-filtered and trimmed using fastp v0.23 (–cut_front –cut_right –length_required 30) [32].

In addition, for each soil compartment (BS and RSS), DNA from the five corresponding samples was pooled and sequenced using Oxford Nanopore Technologies (ONT) to improve assembly contiguity. Genomic DNA was fragmented to an N50 of ~ 8000 bp using a g-TUBE (Covaris) and processed with the Ligation Sequencing Kit SQK-LSK109 (Oxford Nanopore Technologies) according to the manufacturer’s protocol. Sequencing was carried out on a MinION device (version 1B) with an R9.4.1 flow cell, and basecalling was performed using Guppy v6.5.7. Nanopore reads were retained if they had a minimum mean Q-score of 7 and a minimum read length of 1000 bp.

For each soil compartment, a compartment-specific co-assembly was performed using metaSPAdes v4.1.0 in hybrid mode [33, 34]. Specifically, Illumina short reads from the five individual samples of a given compartment were co-assembled together with Nanopore long reads generated from pooled DNA of the same compartment. This strategy resulted in two independent metagenomic assemblies, one for BS and one for RSS; reads from the two compartments were never combined at any stage. To obtain replicate-level quantitative information, Illumina reads from each individual sample were mapped back to the corresponding compartment-specific assembly using BWA-MEM v0.7 [35]. Contig abundance in each sample was then estimated with CoverM [36] using the mean coverage approach. This workflow allowed gene and contig abundances to be compared across biological replicates while using a single non-redundant reference assembly per soil compartment. Taxonomic classification of contigs was performed using Kraken v2.1.3 (--confidence 0.05) [37] against the GTDB r226 database [38]. Coding sequences (CDSs) were predicted using Pyrodigal v2.1 [39], a Python library binding to Prodigal [40]. Functional annotation of the proteins, including KEGG KO assignment [41], was performed using eggNOG-mapper v2.1.9 [42]. Taxonomic affiliation of proteins was determined by similarity searches against the GTDB r226 protein database [38] using DIAMOND v2.1.21 with the --max-target-seqs 1 option [43]. Finally, the abundance of each selected KEGG KO (Table S1) was calculated as the sum of the abundances of the contigs containing the corresponding CDSs, and a differential abundance analysis was carried out using LinDA [44] from the R package MicrobiomeStat v1.2 in proportion mode.

Culture-based assays of soil bacteria

Soil Extract Media (SEM, a minimal nutrient medium) [51] and the rich medium Luria-Bertani (LB), which was supplemented with NaCl (LB 10% NaCl), were used to obtain the culturable microbiota from the BS and RSS samples, and to determine whether dormant bacteria can reactivate and grow in these media. In this context, LB was not used as a proxy for natural soil conditions, but rather as a selective and highly reproducible reference medium to test the viability and resuscitation potential of bacteria that appear inactive in situ. Moreover, it is expected that LB combined with high salinity, will favor the recovery of stress-tolerant taxa. SEM was prepared with a combination of 300 g of soil and 300 mL of distilled water homogenized by stirring for 4 h. Subsequently, the mixture was centrifuged at 9000 rpm for 10 min, the supernatant was recovered, and 1.5% agar was added, autoclaved, and plated on petri dishes. LB 10% NaCl (10 g of tryptone, 100 g of NaCl, 5 g of yeast extract, 15 g of agar and 1 L of distilled water) was sterilized by autoclaving and plated on petri dishes. For culture assays, a suspension of 1 g of each soil sample in 1 mL of 1X Phosphate-Buffered Saline (PBS 1X) was used as inoculum. The mixture was homogenized by shaking in a Revolver Rotator (LabNet) for 2 h and then decanted by gravity for 30 min at room temperature. After that, 100 µL of supernatant were inoculated into SEM and LB 10% NaCl plates and distributed with an L-shaped cell spreader. Five samples of each soil type were plated on both SEM and LB media, producing a total of 20 plates. These were then incubated at 30 °C for 72 h. After this time, the plates were scraped to recover all bacterial growth. Bacteria were suspended in 200 µL of sterile PBS, homogenized by vortexing and used for DNA isolation using the DNeasy Blood & Tissue kit (Qiagen), following manufacturer’s recommendations. The 20 DNA samples obtained were quantified by fluorometry using Qubit dsDNA BR Assay Kits (Invitrogen) and their integrity was corroborated in a 1% agarose gel. The extracted DNA was used for metabarcoding of the 16 S rRNA gene.

16S rRNA metabarcoding

DNA and cDNA from soil samples and DNA from culture-based assays were prepared for sequencing by Molecular Research LP (Mr. DNA Lab). DNA and cDNA samples were amplified using primers 28F (5’-GAGTTTGATCNTGGCTCAG-3’) and 519R (5’-GTNTTACNGCGGCKGCTG-3’), targeting the V1-V3 region of the bacterial 16 S rRNA gene, and the HotStarTaq Plus Master Mix Kit (Qiagen) for sequencing on an Illumina MiSeq platform. The reads were processed using Qiime 2 v2022.2 plugins [45]. The barcodes and primers were removed from the demultiplexed reads using the Cutadapt [46] trim-paired plugin. ASVs (Amplicon Sequence Variants) were generated using the DADA2 [47] denoise-single plugin and were assigned a taxonomy using the classify-sklearn plugin with a naive Bayes classifier trained on the V1-V3 region of the 16 S rRNA sequences from the GTDB r226 database [38] downloaded using the RESCRIPt plugin [48]. To exclude ASVs for which there was no evidence of activity in a sample or potential reactivation under different conditions, only ASVs that met at least one of the following criteria were retained: (i) the ASV was detected in at least one DNA-derived soil sample and in at least one RNA-derived soil sample (indicating in situ activity), or (ii) the ASV was detected in at least one soil sample (DNA or RNA) and in at least one culture-based sample (indicating viability and potential for reactivation). Additionally, the KEGG Orthology gene groups (KOs) (Table S1) [41] associated with each ASV were predicted using PICRUSt2 [49]. ASV relative abundance was calculated as the proportion of total counts assigned to that ASV in each sample, whereas genus relative abundance was calculated as the sum of the corresponding ASV relative abundances.

To assess bacterial species with plant growth-promoting (PGP) traits, a curated list of PGP bacterial families (PGPB) was generated by searching for the term ‘PGPB’ alongside ‘soil microbiome’ in the literature (see the supplementary material for Table S6).

For diversity analyses, bacterial ASVs samples were rarefied to 58,000 counts (even depth without replacement) using the feature-table rarefy plugin of Qiime 2. The R package vegan v2.6-4 [50] was used to calculate the alpha diversity indices (richness and Shannon index). The statistical significance of the comparisons of alpha diversity indices between groups was evaluated using the Wilcoxon rank-sum test corrected with the Benjamini-Hochberg method [31]. For beta diversity analysis, a phylogenetic tree was constructed using FastTree v2.1.11 [51] with the gtr model based on an ASV alignment generated with Infernal v1.1.4 [52] (using the cmsearch command with gathering cutoffs) and the Rfam [53] bacterial small subunit ribosomal RNA covariance model (RF00177). Weighted UniFrac distances were then calculated from this tree using the R package phyloseq v1.42 [54].

Digital PCR analysis of soil bacterial communities

Absolute DNA quantification was performed, in the 10 soil samples, using the QuantStudio 3D Digital PCR system (Thermo Fisher Scientifics, USA), following the manufacturer’s protocols. A commercial TaqMan^® probe targeting the 16 S rRNA gene was used. For quantification, 2 µL of Absolute Q™ DNA Digital PCR Master Mix (5X), 0.5 µL of the TaqMan^® probe-labelled target primer, 1 µL of DNA and 5.5 µL of nuclease-free water were mixed. Subsequently, 9 µL of the reaction mixture was loaded into the QuantStudio™ 3D Digital PCR MAP with the addition of 15 µL of elution buffer. The results were analyzed using the QuantStudio 3D Analysis Suite software. Concentrations of target DNA sequences, together with their Poisson-based 95% confidence intervals, were calculated using QuantStudio 3D Analysis Suite Cloud.

Results

Abiotic properties of the RSS and the BS

A general description of the BS and RSS abiotic properties was performed through analysis of their physicochemical characteristics and the composition of their soluble metal fractions (Table S2). The comparative analysis demonstrated that, with the exception of soluble iron, which did not exhibit a significant difference between the compartments, there was a substantial increase in electrical conductivity and in the contents of K, P, Ca, Cu, Mn, and Zn in the RSS compared to BS (Wilcoxon rank-sum test; p ≤ 0.05). In contrast, a significant decrease in pH and water content was observed in the RSS compartment (Wilcoxon rank-sum test; p ≤ 0.05). The principal component analysis (Fig. 1D) indicated that these variables were highly correlated with the first principal component, which explained 72.4% of the variance and separated the BS samples from the RSS samples. These results indicate that the RSS is physicochemically distinct from bulk soil, characterized by higher electrical conductivity and chemical elements availability but lower pH and water content, defining a differentiated abiotic context for microbial communities.

Functional analysis of soil metagenomes

Shotgun metagenomic sequencing was performed using Illumina and Nanopore technologies. The number of Illumina reads obtained from the five BS and five RSS samples was 3.1 ± 0.7 and 3.5 ± 0.4 Gbp per sample, respectively. The total number of reads obtained from Nanopore sequencing was 9.0 Gbp for the pooled BS samples and 7.3 Gbp for the pooled RSS samples. Metagenome assembly of BS using reads from both strategies yielded 5.5 × 10⁶ contigs and a total length of 2.8 Gb, with an N₅₀ length of 1086 bp. For RSS, the metagenome consisted of 6.9 × 10⁶ contigs with a total length of 3.3 Gb and an N₅₀ length of 902 bp. A total of 4.9 × 10⁶ and 5.8 × 10⁶ prokaryotic proteins were annotated within the assembled contigs of BS and RSS, respectively.

In the BS and RSS metagenomes, we explored the patterns exhibited by traits reported as relevant to the survival of microorganisms under arid and semi-arid conditions (Fig. 2, Table S2). In order to do that, we measured the relative abundance of genes associated with dormancy mechanisms, salt tolerance and iron cycling based on siderophore synthesis in BS and RSS microbial communities.

Fig. 2.

Functional potential of the soil samples. Differences in abundance of the KEGG KOs between bulk soil (BS) and rhizosphere-surrounding soil (RSS). Only KEGG KOs with an absolute log₂ fold change (FC) greater than 1 and a p-value less than 0.05 are displayed

In particular, genes implicated in sporulation exhibited a significantly higher abundance in BS (Fig. 2; |log₂FC| ≥ 1, p ≤ 0.05). With respect to dormancy-related toxin-antitoxin genes, only relE showed increased relative abundance in BS. In contrast, four genes encoding dormancy resuscitation factors (rpfA, rpfB, rpfC and rpfE) were found to be significantly more abundant in RSS (Fig. 2; |log₂FC| ≥ 1, p ≤ 0.05), suggesting that bacteria inhabiting the P. frigida RSS have different strategies for entering dormancy and transit to resuscitation compared to bacteria living in bulk soils. Furthermore, a significant increase in the abundance of genes implicated in salinity tolerance was observed in RSS, including Na⁺ efflux-related genes (mnhB, mnhC, mnhF; |log₂FC| ≥ 1, p ≤ 0.05) and organic solute biosynthesis (codA, ggpS, mltD; |log₂FC| ≥ 1, p ≤ 0.05).

Considering that siderophores facilitate the resuscitation of stressed and viable but non-cultivable cells [55, 56], and also increase soil iron availability under saline conditions [57], the relative abundance of genes implicated in the synthesis of siderophores was also ascertained. The results indicated that 11 genes encoding for enzymes involved in the synthesis of isochorismate (entC), 2,3-dihydroxybenzoate (entB), mycobactin (mbt genes), pyochelin (pchE) and yersiniabactin (irp2) showed a significant increase of their relative abundance in RSS (Fig. 2; |log₂FC| ≥ 1, p ≤ 0.05). The results demonstrated that, at the community level, the majority of the tested genes associated with resuscitation mechanisms, salt tolerance, and siderophore production that exhibited significant differences were more abundant in the RSS compartment. In contrast, the BS compartment exhibited a greater representation of genes associated with sporulation.

Analysis of total and active communities

Given the significant differences in traits associated with the survival of microorganisms between BS and RSS, our next step was to compare the diversity, composition, and taxonomic structure of total (DNA-based) and active (RNA-based) bacterial communities in BS and RSS. To that end, we generated amplicon sequence variants (ASVs) from the amplification of 16 S rRNA sequences derived from DNA and RNA extractions. The primary benefit of employing ASVs in this scenario is that they facilitate the efficient identification of shared individuals across diverse samples. Following the definition given by Jones and Lennon [9], the ASVs of 16 S rRNA genes from DNA extractions were considered to be a representation of the total bacterial community, including active and dormant individuals. However, we also have to take into account relic DNA from dead individuals [20]. Meanwhile, the ASVs of 16 S rRNAs from RNA extractions were considered to be representative of active members of the soil bacterial community. Since our focus was on active and dormant bacteria, we excluded ASVs for which there was no evidence of activity in a sample or potential reactivation under different edaphic or culture conditions. On average, this resulted in a 11.12% ± 2.05% reduction in counts per sample in the total BS community and a 3.80% ± 1.64% reduction in the total RSS community. After filtering, a total of 4626 and 3783 ASVs (Fig. 3A, Table S4) were identified in the DNA and RNA samples, which were classified into 810 and 718 genera (Table S5), respectively. Of the 4626 ASVs detected by DNA, 1418 (30%) were exclusive to RSS and 518 (11%) to BS. Meanwhile, of the 3783 ASVs detected by RNA, 2275 (60%) were exclusive of RSS, whereas only 48 (1%) were exclusive of BS.

Fig. 3.

Diversity analyses of the total (DNA) and active (RNA) bacteria present in the bulk soil (BS) and the rhizosphere-surrounding soil (RSS). A Venn diagram showing the ASVs detected in each sample type. B A comparison of alpha diversity metrics between soil types for the total and active communities. C A comparison of alpha diversity metrics between the total and active communities of each soil type. D Complete-linkage hierarchical clustering constructed from weighted UniFrac distances between samples. ns: not significant, * p < 0.05

Rarefaction curves demonstrated that at a depth of 58,000 counts, the alpha diversity metrics approached a plateau, with only negligible increases observed beyond this point (Fig. S1). Using this cut-off, we observed a considerable difference in alpha diversity metrics between total and active communities when comparing BS and RSS (Fig. 3B). While the total communities (DNA) in BS and RSS showed no significant differences in richness or Shannon index, with median richness values of 1632 and 1918, and median Shannon index values of 5.75 and 6.00, respectively, the active community (RNA) in BS had significantly lower richness and Shannon index than in RSS (p ≤ 0.05), with median richness values of 600 and 1958, and median Shannon index values of 3.52 and 6.32, respectively. Additionally, when we compared the total and active communities within each soil type, we found that the active community of BS had significantly lower richness and Shannon index values than the total community, but this was not the case for RSS (Fig. 3C).

To assess differences in beta diversity, weighted UniFrac distances were calculated and group differences were tested using PERMANOVA, followed by complete-linkage hierarchical clustering. The 20 samples (10 DNA-derived and 10 RNA-derived) clustered into two main groups (Fig. 3D), separating bacterial communities from the RSS and BS. This pattern indicates greater similarity among communities from the same soil compartment, regardless of whether they were derived from DNA or RNA. Consistently, PERMANOVA attributed a larger proportion of variance to soil compartment (R² = 0.32) than to nucleic acid type (R² = 0.09). Although no significant differences were detected between DNA- and RNA-derived communities in RSS, DNA-derived samples in BS formed a tight cluster that differed significantly from RNA-derived samples (PERMANOVA, p ≤ 0.05; see Fig. 3D).

Of the 4672 ASVs and 818 genera detected in any of the soil sample types, 898 ASVs (19%) (Fig. 3A, Table S4) and 317 genera (39%) (Table S5) were shared across all sample types. On the other hand, considering both active and total ASVs, a total of 982 ASVs (21%) and 113 genera (14%) were found exclusively in RSS communities, while 315 ASVs (6.7%) and 40 genera (4.9%) were found exclusively in active and total BS communities. Interestingly, of the 2180 ASVs detected by DNA but not by RNA in BS, 1386 (64%) were found in the active RSS community (Fig. 3A). This suggests that the physical and nutritional features of root systems could influence the composition of the rhizosphere by regulating the activities of shared microbial members between BS and RSS. This fraction accounted for 11.1% of the active RSS community’s relative abundance (Table S4). The observed dissimilarity between active and total BS communities was also evident when comparing the number of genera per phylum, since most phyla had fewer genera detected in the active community (n = 375) than in the total community of BS (n = 657) (Fig. S2), including Actinomycetota (110 in DNA-BS, 81 in RNA-BS), Pseudomonadota (139 in DNA-BS, 104 in RNA-BS) and Acidobacteriota (62 in DNA-BS, 42 in RNA-BS) (Table S5). In contrast, in RSS, the number of detected genera for each phylum tends to be more similar between total and active communities. This is evident in Actinomycetota (134 in DNA-RSS, 127 in RNA-RSS), Pseudomonadota (176 in DNA-RSS, 171 in RNA-RSS) and Acidobacteriota (66 in DNA-RSS, 65 in RNA-RSS). When considering both DNA and RNA, it was found that relatively few genera were exclusive to BS (n = 40) or RSS (n = 113), and together they accounted for less than 1% of the mean relative abundance in their respective sample types (see Table S5 for details). Therefore, under these conditions, the majority of the ASVs (Fig. 3A) and genera detected in BS were also present in RSS, albeit with differences in their abundances, as suggested by the beta diversity analysis (Fig. 3D).

Overall, these results indicate that while BS and RSS share most bacterial taxa, they differ markedly in the proportion of metabolically active members, with the RSS supporting a more diverse and coherent active community largely derived from taxa inactive in bulk soil.

Relationship between the relative abundances of the active and total bacterial communities in the BS and the RSS

By comparing DNA- and RNA-based abundance profiles at the genus level, we evaluated whether microbial activity in each compartment reflects the structure of the total community or is decoupled from it, thereby providing insight into the degree of metabolic activation and dormancy within each soil environment. In the total bacterial community of the BS, we detected genera that were not present in the RNA samples, regardless of their relative abundance in the DNA-derived samples (Fig. 4A, arrow). In contrast, in the RSS (Fig. 4B), the few genera not present in the RNA samples corresponded mainly to those with the lowest abundance in the DNA-derived community. In addition, a Spearman’s rank correlation test revealed a significant monotonic positive association (p ≤ 0.05) between the abundance of each genus in the total and active communities for both BS and RSS. However, RSS had a considerably higher ρ value (0.74) than BS (0.28) (Fig. S3). Furthermore, it was noted that the predominant phyla (Actinomycetota, Pseudomonadota and Acidobacteriota) were present across the full spectrum of abundances in the active and total BS and RSS communities (Fig. 4A and B). However, the phylum Bacillota was an exception, with most of its highly abundant genera not detectable in the active community of the BS or showing a relatively low abundance in the RSS (Fig. 4A and B, green dots). The observed differences in the taxonomic composition and proportion of the total and active bacterial populations of the RSS compared to those of the BS were accompanied by a significant increase in total bacterial abundance as measured by digital PCR, which increased from 1845.6 (± 120.7) copies/ng of DNA in the BS compartment to 2657.6 (± 96.2) copies/ng of DNA in the RSS compartment (Table S2).

Fig. 4.

Relationship between total and active bacterial communities in bulk soil (BS) and rhizosphere-surrounding soil (RSS). Genera are ordered along the x-axis according to their mean relative abundance in DNA-derived libraries, from the most abundant (left) to the least abundant (right). DNA-derived relative abundances (representing the total community) are shown as continuous lines, whereas RNA-derived relative abundances (representing the active community) are shown as dots. Panels A and B display genera colored by phylum in BS and RSS, respectively, whereas panels C and D show the same genera colored according to their mean relative abundance in the corresponding culture-based communities. Black arrows indicate genera detected in DNA libraries but lacking detectable RNA signal, consistent with inactive or low-activity populations

Together, these results indicate that bacterial activity in RSS closely mirrors the structure of the total community, whereas in BS microbial activity is strongly decoupled from community composition, consistent with widespread dormancy. The higher correspondence between total and active communities, together with increased bacterial abundance in RSS, supports the view that rhizosphere-associated conditions promote metabolic activation of soil bacteria rather than taxonomic turnover.

Reactivation of putatively dormant bacteria from BS and RSS by in vitro assays

To assess whether bacterial populations lacking detectable activity in situ could be reactivated under favorable conditions, we combined culture-based assays with 16 S rRNA gene amplicon sequencing. BBS and RSS samples were incubated for 72 h in Soil Extract Medium (SEM) and in LB supplemented with 10% NaCl, a condition previously shown to select for Atacama-adapted microbiota [51, 67–69]. RNA-derived communities from the opposite soil compartment were used as an additional reference to evaluate whether inactivity was context-dependent. Hereafter, we define putatively dormant bacteria as ASVs detected in DNA but not RNA from a given soil compartment that were either recovered in culture or detected as active in the other soil type.

At the genus level, most putatively dormant taxa identified in BS and RSS (Fig. 4C–D, black arrows) were recovered in culture, and their relative abundances in culture tended to scale with their abundance in the total (DNA-based) soil community. This pattern is consistent with the presence of a viable microbial reservoir whose resuscitation potential remains despite low or undetectable activity in situ. At the ASV level, the size of the putatively dormant fraction differed markedly between soil compartments. BS contained 2127 putatively dormant ASVs, whereas only 394 were identified in RSS, indicating a greater resuscitation potential in the BS bacterial community. Of the BS putatively dormant ASVs, 769 (36%) were detected in SEM, 1367 (64%) in LB 10% NaCl, and 1597 (75%) in at least one of the two culture media (Fig. 5A; Table S4). In addition, 1386 ASVs (65%) were detected as active in RSS, indicating that a large fraction of the inactive BS community can transition to an active state under rhizosphere-associated conditions. In contrast, although most RSS putatively dormant ASVs were culturable (377 ASVs, 96%), only a small subset (30 ASVs) was detected as active in BS (Fig. 5B; Table S4), suggesting stronger physiological constraints on microbial activation in bulk soil.

Fig. 5.

Reactivation of putatively dormant bacterial ASVs. Venn diagrams show the overlap between putatively dormant ASVs (defined as ASVs detected in DNA but not RNA in a given soil compartment) and ASVs detected as active in the opposite soil compartment (RNA-based) or recovered in culture-based assays. A Putatively dormant ASVs from bulk soil (BS) shared with the active rhizosphere-surrounding soil community (RNA-RSS) and/or with BS cultures grown in Soil Extract Medium (SEM) or in LB supplemented with 10% NaCl. B Putatively dormant ASVs from rhizosphere-surrounding soil (RSS) shared with the active bulk soil community (RNA-BS) and/or with RSS-derived cultures. Numbers indicate ASV counts

Given that only ~ 40% of the BS putatively dormant ASVs were shared between RSS active communities and cultures, we explored whether these fractions differed in functional potential. We examined functional traits predicted using PICRUSt2 based on KEGG orthologs that were previously identified as differentially abundant in the shotgun metagenomes, noting that resuscitation-promoting factor (rpf) genes are not represented in the PICRUSt2 database. At the genus level, culture-based communities derived from BS were dominated by Bacillota taxa carrying sporulation-related genes, with Priestia reaching an average relative abundance of 34%. These taxa also contributed substantially to genes associated with ion homeostasis (mnh). In contrast, Pseudomonas_E was the main contributor to mltD and relE in culture-based communities (Fig. S4). The active RSS community, however, was dominated by Actinomycetota genera, including Mycobacterium, Actinomycetospora, Pseudonocardia, and Blastococcus, predicted to carry genes linked to osmotic stress tolerance and siderophore biosynthesis (mnh, mltD, entBC, mbt). The most abundant genus contributing to relE in the active RSS community was Lacunisphaera, whereas genera with predicted sporulation-related genes were present at lower relative abundance and were not affiliated with Bacillota.

Given that plant root systems can act as unique ‘resource islands’ and ‘diversity hotspots’ [19], enriching and selecting specific microbial groups with distinct metabolic traits (e.g. PGP) compared to the surrounding bulk soil [18], we investigated whether putative dormant ASVs were associated with bacterial families containing genera that are widely recognised in the literature as plant growth-promoting bacteria (PGPB). Of the putatively dormant ASVs identified in BS, 575 belonged to 28 families containing PGP-associated genera (Tables S4 and S6). These families are characterized by traits such as nutrient solubilisation, phytohormone production and stress mitigation. Of these ASVs, 352 were detected as active in RSS, suggesting that a significant proportion of potentially beneficial bacteria can transition from an inactive state in BS to an active state in the RSS. Within this subset, the five most abundant ASVs in the active RSS community were affiliated with three families: one ASV classified as Acetobacteraceae (0.10% of the average relative abundance), two classified as Acidobacteriaceae (0.18%) and two classified as Caulobacteraceae (0.18%) (Table S4). Overall, these results suggest that culture-based reactivation preferentially enriches stress-tolerant, sporulating taxa, whereas RSS-mediated activation favors metabolically flexible lineages associated with active stress adaptation and plant growth-promoting potential.

To corroborate that the taxa highlighted above were present in the corresponding soils and harbored the predicted functional traits, we examined shotgun metagenomic contigs using taxonomic classification and protein searches against the GTDB r226 database. Of the 5,520,285 contigs in BS and the 6,909,481 contigs in RSS, 2,189,320 and 2,813,784, respectively, were taxonomically classified to at least the phylum level. Of these, 4813 were assigned to Bacillota in BS, including six specifically classified as Priestia. Within Bacillota, 34 sporulation-related genes and four mnh genes were detected (see Table S7). Additionally, 217 contigs were classified as Pseudomonas_E, though no relE or mltD proteins were identified within this genus. According to the STRING database v12.0 [58], many Pseudomonas species possess relE homologues and all contain mltD gene. Consistent with this, the genome of Pseudomonas sp. 31-12 (NCBI RefSeq assembly GCF_003151075.1) contains both relE (NCBI GenBank: AWM89625.1) and mltD (NCBI GenBank: AWM91966.1), and one BS contig exhibited 94% query coverage and 78.65% identity with this genome, according to the NCBI BLAST results [59]. Therefore, the absence of these genes in the assembly may reflect the low abundance of this genus in the soil rather than their true absence. Overall, these results confirm the presence of these taxa in BS, which is consistent with their detection in the corresponding culture-based assays. In RSS, 1,445,014 contigs were classified as Actinomycetota, including 69,072 that were assigned to Mycobacterium, 4188 to Actinomycetospora, 18,742 to Pseudonocardia and 10,155 to Blastococcus. Within this phylum, genes associated with osmotic stress tolerance and iron acquisition were detected, including 479 mnh, 1031 mltD, 33 entB, 156 entC and 420 mbt genes (see Table S7). Additionally, 187 contigs were classified as Lacunisphaera. Although no relE gene was recovered, homologs of relE have previously been reported for this genus (NCBI RefSeq: WP_237023476.1), which suggests that limited coverage may again explain this absence. Finally, several families commonly associated with plant growth-promoting bacteria were also prevalent in RSS metagenomes, including Acetobacteraceae (58,925 contigs), Acidobacteriaceae (51,297 contigs) and Caulobacteraceae (23,495 contigs). Together, these results confirm the presence of these taxa in RSS and support their potential contribution to rhizosphere-associated functional processes.

Discussion

Functional capacities of the bacterial communities of the BS and the RSS: dormancy, salt tolerance and siderophore synthesis traits

Shotgun metagenomic analysis revealed significant differences in abundance and functional potential of bacterial communities inhabiting BS and RSS. The genes that demonstrated higher consistency in their abundance in RSS were those implicated in the synthesis of siderophores. Siderophores play a key role in regulating iron bioavailability in soils and can also contribute to the suppression of phytopathogens, thereby influencing plant-microbe interactions [60]. Some of these genes include entC and entB, which are involved in the production of isochorismate and 2,3-dihydroxybenzoate, precursors of multiple siderophores [61]. At the genus level, Actinomycetospora was the most abundant taxon in the active RSS community predicted to carry these genes, consistent with reports of its occurrence in rhizobacterial communities of the Andean Altiplano [62]. Furthermore, a greater abundance of several mbt genes involved in the synthesis of mycobactin was observed in RSS compared to BS. This siderophore is typically produced by Mycobacterium species, with some species having been isolated from plants and their rhizosphere [63–65]. In our study, Mycobacterium was the most abundant genus carrying mbt genes within the active RSS community, further supporting the idea that the root system favors metabolically active lineages equipped with specialized iron-scavenging strategies.

Research has indicated that bacteria in the Atacama Desert soil are able to survive by adjusting their metabolism within a given range of salinity fluctuations. This enables them to respond to salt stress [66–70]. Consistently, we found that the genes mnh, encoding components of an Na+/H+ antiporter, which are a common mechanism used by bacteria to maintain intracellular sodium homeostasis, were more abundant in the RSS compartment [71, 72]. As part of the bacterial osmoadaptation strategy, genes involved in synthesis of osmoprotectants were significantly enriched in RSS. Among these, codA, which encodes choline oxidase. This enzyme is found in certain soil bacteria, such as Arthrobacter pascens, and it catalyzes the conversion of choline to glycine betaine [73, 74], an important osmoprotectant in bacteria [75]. Additional genes include ggps, mltD, and mpgs, which are involved in the biosynthesis of the organic solutes glucosylglycerol, mannitol, and mannosylglycerate, respectively [76–78], compounds that accumulate intracellularly in response to osmotic stress [79]. Several Actinomycetota genera highly abundant in the active RSS community, such as Actinomycetospora, Pseudonocardia, and Blastococcus, were predicted to carry these genes. Notably, Pseudonocardia and Blastococcus have previously been isolated from Atacama soils, characterized as drought tolerant, and proposed as indicators of aridity in rhizosphere environments [80, 81].

In contrast to the functional profile of RSS, BS communities exhibited a higher abundance of genes associated with dormancy, particularly sporulation. Dormancy is a widespread strategy among soil microorganisms, and it has been estimated that only about 1% of the soil microbial biomass is metabolically active [7]. The remaining viable biomass serves as a reservoir of dormant microorganisms that plays a crucial role in population persistence, biodiversity maintenance, and ecosystem functioning [10, 11, 21]. In this work, we found higher abundance of genes associated with sporulation in the BS, which is consistent with meta-analysis showing that strategies based on spore formation were more abundant in bulk dryland soils [15]. The formation of endospores is a mechanism unique to the Bacillota phylum that allows its members to survive adverse conditions [82]. We found that some spore-forming Bacillota genera, such as Priestia and Paenibacillus_E, were inactive in the BS but grew extensively in the culture media. Despite their capacity to rapidly respond to favorable conditions, these spore-forming taxa generally exhibited low activity levels in the active RSS community. This suggests that sporulation alone does not provide a competitive advantage in the RSS of an established plant. However, it is possible that spore-forming bacteria play a more significant role in responding to changes in resource availability at the earlier stages of plant development, a hypothesis that warrants further investigation [83]. In addition, we found that resuscitation-promoting factor genes were more abundant in RSS than in BS. These factors, present in the phyla Actinomycetota and Bacillota, facilitate peptidoglycan hydrolysis, and promote the resuscitation of dormant bacteria with thick cell walls that are particularly resistant to environmental stress [84]. Actinomycetota was among the most abundant phyla in our soils and has been repeatedly reported as a dominant and active group in arid environments worldwide [85]. In our study, Actinomycetota genera in RSS exhibited activity levels that closely matched their abundance in the total community.

Diversity of the total and active bacterial communities of BS and RSS

Given the pivotal role of active bacteria in driving soil biogeochemical processes and interacting with plant roots [86], it is necessary to assess not solely the alterations in the total community, but also the degree of activity among the community members. In particular, unfavorable conditions, such as prolonged drought in arid regions [87], have been shown to result in an increase in dormant bacteria [9, 10]. In addition, such conditions can also lead to cell death and the generation of relic DNA, which can be detected in metagenomic experiments [20, 88]. In the present study, RNA sequencing was employed as an indicator of bacterial activity, since RNA is exclusively synthesized by actively growing cells and rapidly degrades [89]. The sampling was conducted following a brief period of rain, a time frame that maximized the probability of identifying active bacteria in both the BS and the RSS. This approach allowed us to assess whether the nutritional contribution of plant litter and exudates enhance the diversity and activity of soil bacteria in the rhizosphere context.

Our results show that the active bacterial community from RSS samples exhibited a significantly higher diversity (as measured by richness and Shannon’s index) compared to BS samples. In contrast, total bacterial communities did not differ significantly in alpha diversity between RSS and BS. Similar patterns have been reported in temperate grassland soils, where the active fraction of bulk soil communities tends to show lower richness than the total community, albeit with smaller effect sizes than those observed here [90]. The most probable cause of the observed differences in active community diversity between BS and RSS is the distinct environmental conditions found in each location, particularly nutrient availability. Consistent with this interpretation, RSS soils exhibited higher concentrations of nutrients, which may support a larger proportion of metabolically active bacteria and contribute to both increased diversity and biomass [91–93]. This explains the higher correlation between RNA- and DNA-derived relative abundances in RSS (Spearman’s ρ of 0.74) compared to BS (Spearman’s ρ of 0.28), as well as the significant increase in absolute DNA quantified in the RSS. In this regard, a previous report indicates that the RNA: DNA ratio is significantly lower in Calcisol, a soil with relatively limited nutrient content, compared to Retisol and Luvisol [94]. It is noteworthy that the addition of glucose to these soils resulted in an increase in the RNA: DNA ratio of Calcisol to levels that exceeded those observed in other soil types, which remained unaltered.

Given that the majority of ASVs identified in the DNA libraries were shared between BS and RSS, and a large fraction of ASVs undetected in RNA libraries from BS were detected as active in RSS, it is highly probable that most of these individuals were in a dormant or deceased state in BS. The present findings indicate that the stimulation of a bacterial community in response to elevated nutrient availability within the RSS compartment may constitute a viable strategy for the cessation of dormancy in soils with comparatively limited nutrient content [7]. Indeed, under the conditions of this study, it was found that of the 2127 putatively dormant ASVs at the BS, 1386 were detected as active ASVs at the RSS, indicating that a substantial fraction of the BS microbial community retains the capacity to become metabolically active under rhizosphere-associated conditions. In this context, root exudation and litter inputs associated with Pappostipa frigida may provide local nutritional cues that facilitate the transition of dormant bacteria into active states, effectively recruiting them into the rhizosphere-associated microbiome. A previous study [34] revealed that a subset of the ASVs exhibited a positive response to enhanced iron availability, increasing their abundance and thereby reducing the disparities between the total bacterial communities present in BS and RSS.

Patterns of beta diversity further support the prevalence of dormancy in BS relative to RSS. According to the microbial dormancy model proposed by Jones and Lennon [9], systems characterized by high dormancy and strong environmental constraints are expected to show marked divergence between total and active community structure, as observed in BS. The presence of a single cluster mixing total and active bacterial populations in the RSS suggests that, in contrast to the BS, most bacteria remain metabolically active in the RSS, possibly due to the reactivation of dormant individuals recruited from the BS. This pattern contrasts with observations from less extreme environments, such as paddy soils, littoral zones, and agricultural systems, where active and total rhizosphere communities often remain distinct but resemble their respective BS counterparts [95–97]. The convergence of total and active communities in the RSS of P. frigida likely reflects the comparatively harsh conditions of the Andean steppe, where localized nutrient enrichment associated with plant roots may represent one of the few opportunities for sustained microbial activity.

Therefore, it can be argued that the results of the present study are consistent with the model in which plant-associated fertility islands facilitate bacterial survival during drought periods [16] and the transition of inactive bacteria into active states [98], thereby promoting the establishment of these microbial populations, which would otherwise remain inactive in the oligotrophic environment of the BS [83, 94]. Under this framework, the active bacterial community associated with the root system of P. frigida appears to derive largely from the same bacterial pool present in the surrounding bulk soil. The data concerning the taxonomic composition of the active fraction of BS and RSS of P. frigida suggests that the root system activates a fraction of the inactive microbial community in BS, which has the capacity to respond to changes in resource availability. In turn, they can modify the resource conditions of plant root systems. In fact, 24% of the putatively dormant ASVs in BS belonged to families that promote plant growth. Of this subset, 70% showed activity in RSS. The most abundant ASVs in RSS belonged to the Acetobacteraceae and Acidobacteriaceae families, which have been associated with traits such as nutrient solubilisation [99, 100] that could support plant survival in nutrient-poor conditions that could be particularly relevant for supporting plant survival under nutrient-poor conditions. Together, these results support the view that the root system acts as a selective recruitment mechanism, triggering the activation of dormant soil bacteria with the capacity to rapidly exploit resource inputs and potentially modify the nutritional environment of the rhizosphere [101]. Alternatively, the root system may enable bacterial activation during drought periods due to its more favorable conditions [18]. One probable explanation for the presence of hotspots of active bacteria may be the increased release of root exudates, including mucilage. This can enhance the root system’s water-holding capacity, prevent water loss, improve water storage, and optimize water and nutrient uptake [102, 103]. These mechanisms could be particularly useful for coping with prolonged drought periods in arid environments. In turn, these bacterial communities provide their plant growth-promoting capabilities to the host plant [19, 101, 104].

Culture-based evidence for microbial activation in the RSS

Culture-based assays provided independent support for the viability of bacteria lacking detectable activity in situ and helped contextualize activity and functional patterns observed across soil compartments. Bacteria that were inactive in BS or in RSS were able to grow under laboratory conditions designed to mimic local soil chemistry [68, 105]. Approximately 54% of the putatively dormant ASVs in BS that grew in culture were also detected as active in RSS, consistent with plant-associated conditions promoting microbial reactivation. These results align with the higher nutrient availability, increased bacterial abundance, and enrichment of stress-adaptation and resuscitation-related functional traits observed in RSS, suggesting that the root system environment alleviates constraints that limit microbial activity in bulk soil. However, differences between culture-based communities and the active RSS community, such as the overrepresentation of spore-forming Bacillota in cultures, indicate that culturability alone does not predict RSS dominance. Instead, additional factors, including long-term adaptation, metabolic flexibility, and competitive interactions, likely shape which taxa successfully persist and remain active in the RSS. Together, these results highlight the importance of integrating culture-based, molecular, and functional approaches to disentangle the mechanisms governing microbial activation and persistence in arid environments.

Conclusions

In this study, we report clear distinctions between bacterial communities in BS and in the RSS of P. frigida. These distinctions include community structure, functional potential, and the balance between total and metabolically active populations. Our results suggest that the RSS of the P. frigida acts as an oasis for soil bacteria, enabling the development of a community with a high level of metabolic activity under the extreme conditions of the Andean steppe. A significant proportion of soil bacterial taxa exhibit a similar shift in life strategy: in the BS, they adopt an inactive or dormant state to survive the hostile environment, and in the RSS, they resume metabolism and vegetative growth in response to root exudates during colonization. These findings underscore the importance of analyzing microbial community dynamics in harsher soil environments compared to typical croplands. Linking taxonomic composition with physiological traits and spatial context provides a more mechanistic understanding of soil microbial seed bank dynamics and their contribution to ecosystem functioning. Future studies should assess the extent to which these patterns are conserved across other dominant plant species of the Andean steppe and evaluate how seasonal variability, particularly prolonged drought and episodic rainfall, modulates microbial activity and dormancy in plant-associated soils.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

D.E.P.: Conceptualization, Formal analysis, Methodology, Writing—original draft. C.S.: Conceptualization, Writing—original draft. A.G.: Investigation, Methodology. P.A.: Investigation. C.A.-C.: Investigation. C.A.-N.: Formal analysis. M.G.: Conceptualization, Funding acquisition, Methodology, Supervision, Project administration, Writing—original draft. V.C.: Conceptualization, Funding acquisition, Methodology, Project administration, Writing—original draft.

Funding

This study was funded by ANID FONDECYT Grants 1241424 to M.G., 1211893 to V.C., 3240385 to C.S. and the Millennium Science Initiative Program: ICN2021_044.

Data availability

All sequencing data generated in this study have been submitted to the Sequence Read Archive (SRA) of the National Center for Biotechnology Information (NCBI) under BioProject accession number PRJNA1219082.

Declarations

Competing interests

The authors declare no competing interests.