All samples (biocrusts and plants) were collected near “Henryk Arctowski” (Figure 1A) Polish Station in Admiralty Bay, King George Island (62°09′S; 50°28′W). This area is characterized by a mean annual temperature of −2.8°C and a mean rainfall of 700 mm, mostly as snow during the winter season or rain during the summer season (Kejna et al., 2013). Biocrusts were sampled on three different sites (site 1: 62°9′47.11″S, 58°27′32.07″W; site 2: 62°9′51.79″S, 58°27′46.18″W; site 3: 62°9′33.24″S, 58°28′14.66″W). Sampling sites were selected based on the observation of C. quitensis plants growing in apparent association with BSCs to maximize BSC diversity naturally co-occurring with plants in maritime Antarctica (Figure 1B). Seven samples of biological soil crusts were obtained from three different sites. Each of these seven samples was pooled into a single sample per site (BSC-1, BSC-2, and BSC-3, respectively). To avoid crust surface break-up during sampling, crusts were moistened (Song et al., 2017) and a 9-cm Petri dish was gently pushed over the substrate, as described by Jung et al. (2018). After sampling, the obtained BSC disks (about 5 mm thick) were air dried at room temperature for 2–3 days. Then, Petri dishes containing dried BSC samples were sealed with parafilm (GmbH, Wertheim, Germany), transported, and stored at Universidad de Talca at 4°C for further analysis. In addition, 21 bare soil samples (with no apparent organic cover) were collected with a sterile shovel, 1 m away from a BSC collection point, and stored in sterile 50-ml screwcap tubes. Colobanthus quitensis plants were sampled as part of the 56th Antarctic Scientific Expedition (ECA-56) of the Instituto Antártico Chileno (INACH) during the growing season of 2019–2020. Thirty individuals were collected, along with soil attached to the roots, and were put in a plastic box (120 × 70 × 50 cm). All plants were well watered until their arrival 1 week after to the Instituto de Ciencias Biológicas at Universidad de Talca (Chile).

Biological Soil Crust (BSC) prokaryotic characterization was conducted using the methodology described by Abed et al. (2019). For each sampling site, samples were pooled before metabarcoding analysis. Therefore, three different pools of biocrust samples were obtained from different sites. For each pool, total DNA was extracted in a laminar flow cabinet to prevent environmental contamination from 0.25 g of crust biomass using the PowerSoil DNA kit (MoBio Laboratories Inc., Carlsbad, United States) following manufacturer instructions. DNA sample quality was assessed using 1% agarose gel electrophoresis, while DNA concentration and purity were estimated by spectrophotometry (Nanodrop Technologies, Wilmington, United States) at 260 nm and OD260/280 ratio > 1.8. All DNA samples were sent to MACROGEN (Seoul, South Korea) to produce amplicon libraries for the V4 hypervariable region of the 16S rRNA gene, by using 16S amplicon library preparation protocol and sequencing (Illumina, San Diego, United States). Briefly, an amplicon library was constructed by PCR amplification using single-indexed primers flanked by Illumina standard adapter sequences. According to this methodology, two PCR steps were necessary; the first step was for the identification of bacteria, the targeted gene region was amplified from the crust DNA samples using the Illumina overhang adapter sequences attached to locus-specific primers S-D-Bact-0341-b-S-17 (forward primer, 5′-CCTACGGGNGGCWGCAG-3′) and S-D-Bact-0785-a-A-21 (reverse primer, 5′-GACTACHVGGGTATCTAATCC-3′). These are considered the most suitable primer pairs to conduct PCR-based soil and plant-associated bacterial microbiome diversity studies (Klindworth et al., 2013; Thijs et al., 2017). Then, a second amplification step attached unique indexing primers to the PCR product to identify multiplexed samples after sequencing steps (Kraler et al., 2016). Finally, libraries were sequenced on an Illumina MiSeq (Paired End, 2 × 300bp) at Macrogen Inc. (Seoul, Korea).

16S Illumina metabarcoding libraries (N = 3) were analyzed using the open-source software pipeline Quantitative Insights Into Microbial Ecology package (QIIME2, 2021.4¹) (Bolyen et al., 2019). This pipeline performed quality filtering and primer removal of Illumina amplicon sequencing data using DADA2 (Callahan et al., 2016). Then, QIIME2 assigned Illumina short reads into amplicon sequence variant (ASV) tables. The overall adequacy of reads generated by each sample was estimated through a rarefaction curve using QIIME2 at a depth of 20,000 reads. This ASV table was used as a reference to conduct taxonomic assignments using the Silva 138 99% 16S rRNA gene reference database classifier for prokaryotes.² Finally, the number of reads assigned to each taxon for bacterial BSCs was pooled (by adding) among samples. This approach allowed us to explore, from a global perspective, the taxonomic composition of BSC microbial communities (Simões-Silva et al., 2020).

Due to variability in terms of BSCs microorganism composition, the following analyses were conducted using a BSC collected from site 3 (BSC-3) (Figure 1C), as it showed a higher relative proportion of Cyanobacteria compared with BSC-1 and BSC-2. Hence, to evaluate the effect of BSC-3 crust samples on soil properties, we prepared twenty 250-ml pots (16 × 13 cm), which were filled with a mixture of sterile sand, native soil, and peat in a 4:4:1 proportion, that has been used on previous studies involving Antarctic vascular plants (Barrera et al., 2020). In half of the pots (N = 10), we randomly placed four 1-cm² BSC-3 fragments per pot soil surface, equidistantly from each other, in a cross-shaped fashion, to reduce any potential bias caused by fragmentation and/or sample distribution. As a control condition, in the other 10 pots, we used the same soil mixture but without BSC fragments (bare soil, BS). All pots were maintained in a growth chamber under controlled conditions, at a photosynthetic photon flux density (PPFD) of 350 μmol photons m^–2s^–1 with a photoperiod of 20/4 h light/dark, 75% relative humidity, and at 4°C. All pots were watered with 20 ml of water once per week, mimicking current water availability in Maritime Antarctica during a typical growing season (Barrera et al., 2020; Hereme et al., 2020).

Pots were kept for 90 days in the growth chambers. During that time, we evaluated the role of BSCs on physical–chemical properties of the soil by measuring soil moisture and the activity of three enzymes (β-glucosidase, urease, and dehydrogenase). Soil moisture is related to C and N biogeochemical cycles, and enzyme activity to the abundance of microorganisms (Sinsabaugh et al., 2009). Soil moisture was measured by following the methodology described in Benavent-González et al. (2018). Seven days after the first watering, we randomly selected pots with or without BSC (N = 3 per condition) and took 20 g of soil samples, which were weighed on a digital scale (Boeco BBL-52; 0.01 g precision). Then, each sample was dried on a stove at 110°C for 24 h, and weighed again, to assess the available water content.

Enzymatic activity was measured at 0, 30, 60, and 90 days, at 25°C and following the method proposed by Bell et al. (2014). β-Glucosidase activity was measured using 0.5 g of soil (N = 3 per treatment) adding 0.5 ml of 4-nitrophenyl-β-D-glucopyranoside 50 mM (PNG) as an enzymatic substrate. Results were expressed as micrograms of p-nitrophenol (PNP) per gram of PNG produced per hour (g PNG g^–1 PNP h^–1). For urease activity, we used 1 g of soil in a 0.64% v/v solution of urea to determine the amount of NH₄⁺ produced. This was done through colorimetric methods, measuring absorbance with a spectrophotometer at 525 nm. Results were expressed as micrograms of N-NH₄ per gram per hour. Dehydrogenase activity was determined also using 1 g of soil per pot. To each sample, we added 0.2 ml of a 0.4% v/v solution of 2(p-iodophenyl)-3-(p-nitrophenyl)-5-phenyl tetrazolium chloride (INT) as substrate. Results were expressed as micrograms of reduced iodonitrotetrazolium formazan per hour (INTF g^–1h^–1). All measurements considered three technical replicates and a negative control (blank, with no BSC). Finally, soil nutrient levels (N, P, K, Ca, Mg, and Na) were quantified in three 10-g samples per condition (BSC/BS). Analyses were conducted at 0, 30, 60, and 90 days. Nutrient measurements were done in the Centro Tecnológico de Suelos y Cultivos at Universidad de Talca, Talca, Chile, as described by Song et al. (2017).

To evaluate the effect of BSC on the ecophysiological performance of C. quitensis, we conducted a manipulative experiment using 100 individuals, which were obtained through vegetative propagation in the laboratory from 30 field-collected plants using the methodology described in Zuñiga et al. (2009). All plants were maintained in 50-ml pots containing a 4:4:1 mixture of sterile sand, native soil, and peat, respectively, and were kept well-watered until they reached ∼2 cm in diameter. Then, these plants were randomly transplanted to the 20 pots with or without BSC, whose soil properties were previously characterized (5 plants/pot). Plants were transplanted equidistantly from each other and maintained in a growth chamber with the same environmental conditions described earlier. After 60 days, we measured the maximal photochemical efficiency of the PSII (Fv/Fm), leaf nutrient content, biomass, and flower number per plant.

Fv/Fm has been described as a good proxy for the overall physiological status of the plant (Molina-Montenegro et al., 2012a,b). This parameter was measured in 10 randomly selected C. quitensis individuals (one per pot) in each experimental condition (with or without BSC) using a portable fluorimeter (Hansatech FMS 2; Hansatech Instruments Ltd, Norfolk, United Kingdom). Prior to each measurement, the target leaf was kept in the dark for 30 min, ensuring that the data correspond to the maximal PSII efficiency (Torres-Díaz et al., 2016). Leaf nutrient content was determined in three samples (three individuals per sample) to reach the minimum weight required for this analysis. Thus, we obtained N, K, P, Ca, Mg, Mn, Zn, Cu, Fe, and B content at 0, 30, and 60 days after the beginning of the experiment. Nutrient analysis was done at the Centro Tecnológico de Suelos y Cultivos at Universidad de Talca (Chile), based on the method proposed by Song et al. (2017). To estimate the effect of the BSCs on the plant foliar biomass, we extracted and weighed at the end of the experiment the aerial part of all C. quitensis individuals. Fresh biomass measurements were made using a digital scale (Boeco BBL-52; precision 0.01 g). Finally, plant fitness was measured after 60 days, as the flower number per plant, in 25 individuals per experimental group (with or without BSC), which were randomly selected at the beginning of the experiment.

Since data were not normally distributed (enzymatic activities, nutrient concentrations), we used the Wilcoxon rank sum test (also known as Mann–Whitney U test) to assess the effect of the BSCs on the physical–chemical characteristics of the soil. On each time point, the nine monitored variables—soil water content, three enzymatic activities (β-glucosidase, urease, and dehydrogenase), and six nutrient concentrations (N, P, K, Ca, Mg, Na)—were tested independently between groups. Similarly, for each of the four ecophysiological traits of C. quitensis (i.e., Fv/Fm, and the foliar concentrations of N, P, and K), the effect of the BSCs was also analyzed with the same non-parametric method. In addition, after verifying the parametric assumptions of normality and homoscedasticity, the overall effect of the BSC on the mean fitness response of C. quitensis in terms of flower production and foliar fresh biomass was estimated by independent Student tests (t-tests) for each variable, comparing plants grown with and without the influence of BSC. All analyses were performed on the R environment and language for statistical computing v4.0.2 (R-Core Team, 2020).

All sequence data raw libraries were submitted to NCBI Sequence Read Archive (SRA) and are available under the Bioproject accession number PRJNA765698. Representative ASV sequences and ASV feature tables for both 16S sequencing were deposited at Figshare platform (available at doi: 10.6084/m9.figshare.1668081).

High-throughput sequencing of 16S rRNA gene amplicons from pooled samples generated 2,042,990 reads in total, representing more than 600 megabases. After DADA2 analysis, all reads were clustered into 641 ASVs; furthermore, for all samples, rarefaction curves reached a stable asymptote, suggesting the availability of sufficient reads for identification of all the bacteria present in these samples. Regarding the taxonomic analysis, 20 different bacterial phyla were found among all sites (BSC-1, BSC-2, BSC-3). Overall, the most abundant phyla were Proteobacteria (42.3%), Cyanobacteria (16.09%), Actinobacteria (12.46%), Bacterioidetes (8.76%), and Chloroflexi (4.88%). Our results show that the same phyla were found among all BSCs, but with different relative abundances (Figure 1C). As we observed a higher relative abundance of Cyanobacteria in BSC-3, we selected this sample for further manipulative studies (see Materials and methods section). In this particular sample, a dominance of Cyanobacteria (43.44%) followed by Proteobacteria (25.69%), Actinobacteria (12.12%), and Bacterioidetes (6.41%) was observed (Figures 1C,D). Finally, a total of 252 taxa (either species/genus/family/phyla levels) were found among all analyzed samples, but only 41 were shared between BSCs, while 52, 40, and 63 taxa were found exclusively in BSC-1, BSC-2, and BSC-3, respectively.

Soil moisture, measured at the end of each watering cycle (every 7 days), was higher in soils with BSCs (26%) compared with soils without BSC (with BS, 13%). Similarly, we observed differences in enzymatic activity (for all tested enzymes) between treatments, with higher activity in BSC soils compared with BS, particularly after 90 days (Figures 2A–C). Dehydrogenase activity, associated with abundance and metabolic activity of microorganisms, increased significantly after 60 days in BSC soils, peaking at 0.30 μg INTF g^–1 h^–1 after 90 days from the beginning of the experiment. In contrast, BS soils reached only 0.02 μg INTF g^–1 h^–1 after 90 days. For the β-glucosidase activity, there were statistically significant differences between BSC and BS soils after 30, 60, and 90 days. Moreover, for BSC this enzymatic activity was an order of magnitude greater than BS at the end of the experiment (BSC = 11.30 μg PNP g^–1 h^–1; BS = 1.99 μg PNP g^–1 h^–1). Finally, urease activity (related to NH₄⁺ availability in the soil) was higher in BSC compared with BS after 30 days. These differences also increased after 60 and 90 days from the beginning of the experiment, the time at which measured enzymatic activity was 71 μg N-NH₄ g^–1 h^–1 in BSC and 3.08 μg N-NH₄ g^–1h^–1 in BS, respectively.

Soil nutrient content was higher in BSC soils compared with BS, but only for N and K (Figures 2D,E). N increased significantly in BSC since day 30 compared with BS, reaching 130 mg/kg at the end of the experiment (day 90), while BS only reached 68 mg/kg in the same time (Figure 2D). For K, differences between treatments (BSC and BS) were evident only after 90 days, the time at which BSC soils reached 1,017 mg/kg versus only 843 mg/kg on BS (Figure 2E). No differences in P, Ca, Mg, or Mn content were found between treatments, except for a P burst after 60 days in BSC (Figures 2F–I).

In C. quitensis grown with BSCs, leaf N, P, and K was higher toward the end of the experiment (day 60, Figures 3A–C) compared with individuals grown in BS. In the case of N, we observed a notorious increase in BSC plants after 30 days (Figure 3A). In terms of physiological performance, plants growing in association with BSCs had a higher Fv/Fm compared with BS plants after 60 and 90 days from the beginning of the experiment (Figure 4A). Regarding the number of flowers, we did not find statistical differences between treatments (Figure 4B). However, BSC plants had a slightly higher, although not significant, flower number than BS (55 vs. 33, respectively). In the case of fitness-related traits, we observed a significant increase for C. quitensis plants growing in presence of BSC. Plants with BSC have a higher final biomass (0.98 g) compared with plants grown in bare soil (0.77 g), while leaf biomass was also higher in BSC compared with BS (Figure 4C).