Understanding the Genetic Diversity of Bacteria Isolated from Across the Atacama Desert

Even the harshest landscapes can be home to microbial life. Microbes have been identified as far as humans have been able to sample, from the deep depths of the sea floor to the Arctic to near the centers of active volcanoes [1,2,3]. The conventional wisdom has been that archaea, not bacteria, are usually found in the harshest environments since they have a number of key adaptations to survive in extreme conditions [4]. Understanding how bacteria survive stressful conditions is essential to many areas, including agriculture, biotechnology, and human health.

Bacteria capable of living in harsh conditions are called extremophiles, known for encoding adaptations to help them survive and thrive. Examples include structural changes in the DNA itself, such as G-quadruplex formation (G4), and/or high GC content, pigment production, biofilm formation, production of unique secondary metabolites, and antibiotic resistance mechanisms, among others. Some bacteria produce pigments that protect them from damage by UV rays, such as melanin or carotenoids [5,6]. These pigments provide a physical barrier against UV damage as well as protect cells from reactive oxygen species (ROS) and free radicals [5,6,7]. In addition, many bacteria produce antimicrobial compounds to help ward off competing species and/or be involved in signaling, quorum sensing, impacting virulence, and in DNA repair [8,9].

Many bacteria live in nature as biofilms, which are multicellular communities bound together by a self-produced matrix [10]. Many species can come together to form these matrix-bound communities [11]. The matrix is made up of sugars, proteins, DNA, etc., known to be beneficial to the community [10]. Benefits include providing a medium for concentrating and sharing nutrients, adding a protective barrier to protect the community from antimicrobials, maintaining hydration, and protecting cells from UV radiation [10,12,13,14]. In addition, some bacteria within the biofilm produce compounds that inhibit the growth of other organisms to protect their community and resources [15]. On the other hand, bacterial cells also develop resistance mechanisms to help them withstand inhibitory compounds produced by others [16,17].

Notably, the DNA sequence itself can be an adaptation. DNA is made up of nucleotide pairings, where adenine (A) pairs with thymine (T) and guanine (G) pairs with cytosine (C) [18]. DNA that is higher in GC content may have stronger DNA stability due to more favorable nucleotide stacking patterns [19]. It has also been shown in plants that individuals that live in cold or dry climates are more likely to have a high GC content [20]. Sections of DNA that contain several runs of guanines are sometimes capable of forming G-quadruplexes (G4’s), which are a non-helical secondary structure of DNA [21]. G4’s forms in a wide variety of organisms, including bacteria, and serves a variety of functions. While investigating bacterial G4 functions is a newly popular field, it has been shown that G4’s in bacteria often concentrate near promoter regions [22]. In particular, RecA-dependent DNA repair pathways in Deinococcus radiodurans and Deinococcus geothermalis, known for being able to withstand lethal doses of radiation, are partially regulated by G4’s [21].

One environment that is a high priority for scientists studying extremophiles is the Atacama Desert, located in northern Chile, South America. The Atacama Desert is recognized as one of the most ancient, driest, and most UV-irradiated deserts on Earth [23,24]. The desert spans approximately 128,000 km² and sits at 2000–5000 m above sea level, limiting available oxygen and causing high levels of UV radiation [25,26]. Additionally, the desert lies between two mountain ranges, the Andes Mountains and the Chilean Coast mountain range, which severely limit precipitation in the area [23,24]. In fact, NASA has used the Atacama Desert as a testing ground for rovers and as a proxy for Mars [27]. For these reasons, the Atacama Desert remains an important site for studying how microorganisms can adapt to harsh environments. We have surveyed and sampled a variety of locations across the Atacama Desert to understand how eubacteria live there and the adaptations they have developed to survive this unique environment [28].

In July 2018, we performed global 16S sequencing on soil samples across 18 locations with the goal of understanding what bacterial species could live across the Atacama Desert. To enhance this work, in 2022 and 2024, we isolated and whole genome sequenced eight bacteria from a variety of locations throughout the desert. We selected environmental strains that produce pigments in different conditions. Pigments are visually obvious when produced by a bacterial colony, and thus pigment formation is a great starting point when looking for well-adapted bacterial candidates to sequence. Organisms may produce pigments for a variety of reasons, including ultra-violet (UV) protection, antioxidant purposes, biofilm control, and sequestering molecules, to name a few [6,29,30]. Using our genomic sequences, in addition to identifying the genes involved in pigment production, we can also identify other genes involved in mechanisms beneficial for survival in inhospitable environments, including production of antimicrobial and bioactive compounds, key biofilm formation genes, and cell to cell communication genes.

The knowledge gained from this survey will enable an evolutionary understanding of key survival strategies in bacteria, including pigment production, bioactive compound production, and biofilm formation. Comparing these data to similar data from known model bacteria provides insight into how distinctive these genes are across environmental species. Insight into how bacteria handle stress is relevant in many fields, including agriculture, biotechnology, and health.

Environmental samples were collected in May and June 2018 (Figure 1). Pictures of each location were reported by Reverdy Pearson in 2024 [28]. Each day had little to no cloud cover, no precipitation, temperatures ~16 °C at midday, and 10–20% humidity. At each location, soils were collected from a depth of 1–5 cm using sterile tools. Samples were stored in sterile 2 mL microcentrifuge tubes. The date and exact coordinates were recorded at each sampling location (Table S1). Samples were kept at ambient temperature, ranging from 1–21 °C, while in the field for up to 10 h. A fraction of some of the samples (~1 g) was kept in a solution in Zymo DNA/RNA shield (Zymo Research, Irvine, CA, USA) to preserve nucleic acids, and kept at room temperature as indicated by the manufacturer. Once brought to the laboratory, 20% (v/v) glycerol was added to the leftovers of all samples (~1 g). Samples were stored at −80 °C until thawed for further analysis.

Approximately 1 g of soil sample was added to the UBiome microbiome sequencing kit tubes (UBiome, San Francisco, CA, USA). Samples were processed and analyzed with the ZymoBIOMICS^® Service: Targeted Metagenomic Sequencing (Zymo Research, Irvine, CA, USA).

The ZymoBIOMICS^® DNA Miniprep Kit (Zymo Research, Irvine, CA, USA) was used for DNA extraction. Bacterial and archaeal 16S ribosomal RNA gene targeted sequencing was performed using the Quick-16S^TM NGS Library Prep Kit (Zymo Research, Irvine, CA, USA). The V3–V4 region of the 16S rRNA genes was amplified using primers 341f (CCTACGGGDGGCWGCAG, CCTAYGGGGYGCWGCAG, 17 bp) and 806r (GACTACNVGGGTMTCTAATCC, 24 bp) custom-designed by Zymo Research to provide the best coverage of the 16S rRNA gene while maintaining high sensitivity (Zymo Research, Irvine, CA, USA). The sequencing library was prepared using a Zymo-developed library preparation process in which PCR reactions were performed in real-time PCR machines to control cycles and limit PCR chimera formation. The final PCR products were quantified with qPCR fluorescence readings and pooled together based on equal molarity. The final pooled library was cleaned up with the Select-a-Size DNA Clean & Concentrator^TM (Zymo Research, Irvine, CA, USA), then quantified with TapeStation^® and Qubit^®.

The final library was sequenced on Illumina^® MiSeq^TM with a v3 reagent kit (600 cycles). The sequencing was performed with >10% PhiX spike-in. Amplification was compared to a negative control standard.

Unique amplicon sequences, also termed amplicon sequence variants (ASV’s), were inferred from raw reads using the DADA2 pipeline [31]. Chimeric sequences were also removed with the DADA2 pipeline. Taxonomy was assigned based on the determined ASV’s using Uclust from Qiime v.1.9.1 with the Greengenes 16S database as the reference. The Visualization and Analysis of Microbial Population Structure (VAMPS) program was used for all subsequent diversity, abundance, and cluster plot analyses [32]. Alpha diversity, or richness, was quantified as the number of unique sequences found in the sample. Phyla and genera relative abundance were calculated as the percent number of sequences out of the total number of sequences in the sample. The cluster plots were calculated based on Morisita Horn parameters at the genus level.

To study how specific bacteria can handle stress, we started by isolating a variety of bacterial colonies from the desert soil. Samples were collected in July 2022 and July 2024. Weather conditions were similar across both years of sampling, with little to no cloud cover each day, no precipitation, and 10–20% humidity on average. Details are outlined in Table S2. At each sampling location, soils and other environmental samples were collected from between 1–5 cm depth using sterile tools (Figure 2). Samples were stored in sterile 2 mL microcentrifuge tubes. The date and exact coordinates were recorded at each sampling location (Table S2). Samples were kept at ambient temperature in the field, ranging from 1 °C to 20 °C. Within one week, samples were brought to the laboratory, and bacterial colonies were isolated after plating 100 µL of a water/sample slurry on Reasoner’s 2A (R2A) agar medium and incubating the plates at 28 °C for 48–72 h. R2A medium was selected because of its minimal nature, since sampling was done in areas that have low visible macro biodiversity, and nutrient availability is also likely low in these environments. This isolation yielded bacteria of many morphologies and colors. Each isolate was further purified by streaking and stocked in 20% (v/v) glycerol in water and stored at −80 °C.

To test biofilm formation in isolates that were phenotypically like Bacillus subtilis, strains were grown for 4 h at 37 °C to OD600 = 1.0 in LB broth in shaking conditions. A volume of 2 µL of culture was spotted on LBGM agar plates-biofilm-inducing plates where LB is supplemented with 1% glycerol and 0.1 mM MnSO₄ [33]. Colony biofilms were grown at 30 °C for 72 h before imaging.

A selection of pigment-producing isolates from a variety of desert environments became candidates for whole genome sequencing.

The genomes of three isolates, 0102A, 0209A, and 0909A, were sequenced by Illumina MiSeq (Illumina) as described in Cavanaugh et al., 2024. Details about sequencing and assembly can be found in the publication [34].

For isolates 1020B, 0516A, 0819A, Iso1_2024, and Iso2_2024, genomic DNA was isolated from 2-mL R2B cultures using Promega’s Wizard Genomic DNA Purification Kit (Promega, Madison, WI, USA), following the provided instructions for Gram-positive bacteria. Purified DNA was quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

Purified gDNA was sequenced by Oxford Nanopore long-read sequencing (Oxford Nanopore Technologies, Oxford, UK) by Plasmidsaurus. Libraries were prepared using the V14 Ligation Sequencing Kit, and sequencing was performed using the MinION R10.4.1 flow cell (Oxford Nanopore Technologies). Before assembling the genomes, the 5% lowest quality reads were removed by Fitlong v0.2.1. [35]. Reads were downsampled to 250 Mb by Fitlong, and a rough assembly was generated with Miniasm v0.3. [36]. Reads were further downsampled to 100× coverage when coverage was over 100×. Quality reads were assembled using Flye v2.9.1 [37] and cleaned using Medaka v1.8.0 (Oxford Nanopore Technologies).

A phylogenetic tree was generated using the autoMLST2 web server using denovo mode [38]. The fasta files from each genome sequence obtained were uploaded to the webserver, and autoMLST2 compares the uploaded genomes to those in the Genome Taxonomy Database (GTDB) [39]. autoMLST2 was run using default parameters.

Biosynthetic Gene Cluster (BGC) analysis was performed on all genomes using antiSMASH v8.0.2 using a relaxed detection strictness [40]. FASTA files were used, and KnownClusterBlast, TFBS analysis, activesitefinder, SubClusterBlast, and RREFinder analyses were selected from the “extra features” menu. Secondary metabolite-producing regions found in each genome were recorded.

For individual identification of carotenoid synthesis genes, DIAMOND v2.0.15 alignment software was utilized to align translated protein sequences against the RefSeq Bacterial Protein Database (REFSEQBP) [41]. Additionally, 0819A’s genome was aligned against all known Kocuria rhizophila proteins (KRAP), Iso2_2024’s genome was aligned against all known Kocuria turfanensis proteins (KTAP), and 1020B’s genome was aligned against all known Kocuria oceani proteins (KOAP). For REFSEQBP alignments, alignments were performed using very relaxed parameters, compared to the more stringent parameters used for the KTAP, KRAP, and KOAP alignments. For each strain, the top 15,000 best aligned proteins were selected and sorted for carotenoid biosynthesis genes. A pipeline was created using Bash and Python languages in the Linux environment, creating scripts to download databases, converting them to DIAMOND format, performing the DIAMOND alignments, assigning protein names to taxonomic IDs, arraying the data into an interpretable format, and back calculating genome positions of hits.

Genome GC content and G4 quadruplex (G4) content were assessed by the G4 Hunter web application, available through DNA Analyser [42]. Genome FASTA files were uploaded to G4 Hunter and analyzed using a window size of 25 nt and a threshold of 1.2. The estimated number of G4’s, G4 frequency per thousand base pairs (kbp), GC content, and genome size were recorded.

To assess the microbial community composition of the samples, Illumina 16S rRNA sequencing was used directly on soil samples from the 18 sampling locations and analyzed for its Archaea and Bacteria composition. Importantly, the soil samples from Ojos de Salar and Baltinache did not provide reliable data when compared to the negative control and could not be analyzed any further. This indicates that the soil composition did not permit the identification of microbial life with the test used, or that the soil contained inhibitors of the processing used for the determination. Community composition was successful in the remaining 16 soil samples.

The focus of our search was on bacterial diversity. To infer the amount of diversity in each soil sample, ASV’s were assigned species identifications based on similarities with the Greengenes 16S database. The alpha diversity, a measurement of richness, ranged from 184 to 652 estimated species (Figure 3A). The two dominant phyla across all locations were Actinobacteria at 35.9% (brown bars) and Proteobacteria (dark blue bars) at 26.7% (Figure 3B). Our results are consistent with other studies that have investigated the bacterial content in the Atacama Desert [43,44,45]. Other studies have also found Bacteroidetes and Cyanobacteria to be the dominant phyla, however, they were sampling water, while we sampled soil [46,47,48]. Bacteroidetes is the third most abundant phylum across all the samples at 11.3% abundance (Figure 3B).

To compare bacterial diversity across the different sampling locations, we generated a genus level cluster plot (Figure 3C). The salt flat, Tebenquiche, was the outlier group, and the other salt flat, Chaxa, was in another outlier grouping with Talabre. The clustering of Apacheta and Miscanti is most likely due to their high altitude and shrub-containing characteristics. Salar de Tara and Aguas Calientes are closely located and near high altitude lakes (Figure 1). Rio Grande and Valle de Arcoiris are two dry and barren locations (Table S1). The clustering based on location similarities highlights that the genera are most likely selected by the environmental conditions.

Though our main interest was to find eubacteria, we were curious to learn about the percentage of Archaea in our samples, as they are well known extremophiles [49,50] and likely to be in the most extreme samples. In fact, we found that our levels of 1.1% (Figure 4) are comparable to levels of Archaea found in the other hyper-arid desert soils [43,49].

Bacterial growth from environmental samples on R2A medium yielded many colonies (Figure S1). After bacteria were isolated from environmental samples on R2A medium, several (~30–50%) appeared to produce pigments. Isolates 1020B, 0819A, and Isolate 2 displayed strong yellow and orange pigmentation after growing on R2A medium for 3 days (Figure 5C–E). Two isolates also appeared to secrete blue and green pigments (Figure 5A,B).

Three isolates appeared to have similar colony morphology to Bacillus subtilis-0102A, 0209A, and 0909A. It is known that B. subtilis produces a pigment called pulcherrimin late in the biofilm formation cycle, regulating biofilm aging and disassembly [29]. When these isolates were grown on LBGM biofilm inducing medium, they produced a red-brown pigment that resembled pulcherrimin (Figure 5F–H). These eight isolates were selected to be subjected to whole genome sequencing (WGS) (Figure 5).

After sequencing, all resulting genome FASTA files were aligned using autoMLST2 [38]. autoMLST2 uses multiple loci across the genome, focusing on housekeeping genes to provide higher resolution, to identify the most closely related species for each sequence by generating a multi-locus species tree (Figure S2). autoMLST2 automatically identified related genome sequences to create the species tree de novo and was able to identify the identities of most isolates to the species level. Iso1_2024 was identified as Pseudomonas aeruginosa, 0516A was identified as Pseudomonas rhodesiae, 0819A was identified as Kocuria rhizophila, 1020B was identified as Kocuria oceani, Iso2_2024 was identified as Kocuria turfanensis, and 0909A was identified as Bacillus mojavensis (Table 1). All identifications were made with a high level of confidence, according to the Mash distance [51], estimated average nucleotide identity (ANI), and p-value. The top hits for 0102A and 0209A identification were “Bacillus subtilis group”, rather than being matched directly with a species. The top three most similar species matches are listed in Table 2. Both strains are most closely related to Bacillus inaquosorum (Table 2).

All genome FASTA files were analyzed by antiSMASH v8.0.2. [40]. antiSMASH identified 5–20 biosynthetic gene clusters (BGC’s) in each genome (Table 3). Potential pigment-producing BGC’s were identified in each genome that was sequenced (Table 4). Pulcherrimin clusters were identified in all Bacillus sp. genomes that were sequenced, 0102A, 0209A, and 0909A. Carotenoid clusters were identified in the three Kocuria sp. that were sequenced, 0102B, 0819A, and Iso2_2024. Pyoverdine SXM-1 and aryl polyene clusters were identified in 0516A, both of which may contribute to its green color. Pf-5 pyoverdine and pyocyanine clusters were both identified in Iso1_2024. These produce yellow-green and blue pigments, respectively.

Biosynthetic gene cluster analysis by antiSMASH identified that each Kocuria strain that was sequenced produced carotenoid pigments (Table 4) [40]. To understand what types of carotenoids each strain produces, DIAMOND v2.0.15 software was used to identify individual carotenoid genes [41]. It was determined that strains 0819A and 1020B produce yellow pigments that are like decaprenoxanthin and that strain Iso2_2024 produces an orange carotenoid closely related to bacterioruberin (Figure 6). Predicted colors of intermediates and products are depicted in small squares to the left of each pathway in Figure 6, which are colored either cream, yellow, orange, or red. The confidence in the enzyme identification is shown by placing enzyme names in empty white rectangles (high confidence) or grey rectangles (medium confidence). Confidence level was determined based on a combination of qualitative and quantitative assessments, factoring in percent identity matches, annotation name of the aligned protein, genome positioning relative to other genes in the pathway, and supporting literature from the Kocuria genus and species-specific carotenoid production.

antiSMASH also identified antibiotic and/or bioactive compound BGC’s in each genome that was sequenced (Table 5) [40]. The Bacillus sp. isolates were predicted to produce a wide range of antibiotic compounds, including fengycin, bacillaene, macrolactin H, bacilysin, subtilosin A, bacillibactin, zwittermicin A, plipastatin, staphylococcin C55, and surfactins. The Kocuria sp. were each found to produce siderophores. The identified Pseudomonas sp. produces different varieties of bioactive compounds. 0516A, P. oceani, was shown to produce pyoverdine SXM-1, pyochelin, a betalactone, and ambactin. Iso1_2024, P. aeruginosa, was shown to produce pseudopaline, pyochelin, azetidomonamide, Pf-5 pyoverdine, L-2-amino-4-methoxy-trans-3-butenoic acid, hydrogen cyanide, pyocyanine, and a homoserine lactone.

Each genome from this study, along with similar strains from other environments, was analyzed by the G4 Hunter online server [42,52,53]. The Kocuria strains from this study had the highest G-quadruplex (G4) frequencies, above 4 occurrences per thousand base pairs (kb), or 11,606–19,373 total G4’s (Table 6). This was slightly higher than K. rhizophila strain 7_17 (GenBank ID: CP124833.1), which contained 10,368 G4’s at a frequency of 3.8 per 1 kb. The Pseudomonas sp. have comparatively lower G4 content. Strains 0516A and Iso1_2024 have a G4 frequency of 1.5 per 1 kb with a total of 9072 or 9632 total G4’s, respectively. These were slightly lower than G4 concentration in P. aeruginosa clinical strain AR_0230, which contained 10,336 G4s at a frequency of 1.5 per 1 kb [52]. The Bacillus strains have the lowest estimated G4 content and are each estimated to have a 0.4 per 1 kb frequency, or 1462–1693 G4’s. This is consistent with estimated values for B. subtilis strain 3610 [53]. These differences in frequency align with the GC content of each genome. The Kocuria species had GC contents over 70%, the Pseudomonas strains had GC percents in the 60%’s, and the Bacillus strains all had GC percents in the 40%’s.

In this study, we used sequencing technologies and bioinformatics to assess microbial diversity and identify genes that are beneficial for bacterial survival in inhospitable environments. Using the wide-range method of Illumina 16S rRNA sequencing, we identified the dominant phyla to be Actinobacteria and Proteobacteria in Atacama soil (Figure 3B). This agrees with other studies using similar approaches in the Atacama [47,54]. Locations that have more plant life had a greater richness (Figure 3A). Hierbas Buenas has a larger abundance of low grass and bushes, and accordingly, had the largest estimated richness at 652 species (Figure 3A). Conversely, the locations with the driest and most barren soil had the lowest richness as seen at Laguna Chita, Ckari Mirador, Tebenquiche, Salar de Tara, and Valle de Arcoiris with an estimated 184, 194, 227, 240, and 265 species, respectively (Figure 3A). As expected, some of these locations also had the highest levels of Archaea (Figure 4), which are well known extremophiles [49,50].

The correlation of “plant-rich” areas containing higher bacterial richness is expected due to the strong symbiosis and known importance that the bacterial rhizobiome has for plant growth and survival. Overall, this richness is low compared to other environments. Sea water has an estimated 10,000 species richness, the temperate forest and grasslands soil has a few thousand species, and Arctic soils have been shown to have a species richness between 950 and 1250 species [55,56,57]. As predicted, our results suggest that the extreme environment has a significantly lower number of unique species detected using the direct sequencing method. It is important to note that 16S amplicon sequencing is not sufficient to determine species. Rather, it provides an estimate for diversity.

Eight isolates that produce pigments were selected for whole-genome sequence (Figure 5). Using autoMLST2, three Kocuria strains were identified, which are closely related to K. rhizophila, K. turfanensis, and K. oceani (Table 1) [38]. Two Pseudomonas strains were identified, which are closely related to P. aeruginosa and P. rhodesiae. Three Bacillus strains were identified, two of which are closely related to B. inaquosorum and one of which is related to B. mojavensis (Table 1 and Table 2). Previously, both B. inaquosorum isolates were identified as B. vallismortis using autoMLST v1 [58]. B. inaquosorum was previously considered a subspecies of B. subtilis [59], which is very closely related to B. vallismortis [60]. autoMLST2 improves upon autoMLST1 by including new databases of sequence data to create alignments. This added information slightly altered the identification of these bacteria. Our findings agree with previous findings [61,62,63,64] including the finding that bacteria were more prominent than archaea [43]. In addition, it is notable that some of these species may not be necessarily native to the Atacama, but brought in by the wind [65].

Potential pathways for producing the observed pigments were identified bioinformatically (Figure 5 and Table 4). The types of pigments identified have different roles in species survival. Carotenoids often exhibit strong antioxidant activity and confer UV resistance [6]. They are a class of tetraterpenoids made through the assembly of 8 isoprene units to make a 40 carbon (C₄₀) backbone. A major differentiating characteristic is their conjugated double bond system, which absorbs light between 400 and 550 nm in the UV-Vis range. Aryl polyene has a very similar structure and functionality to carotenoids [30]. They serve as antioxidants and have a yellow/green color. Pulcherrimin is a red-brown pigment produced by B. subtilis and other bacteria and yeast that controls biofilm aging and disassembly [29]. It has iron chelating activity and protects the cells from oxidative stress by lowering the levels of reactive oxidative species (ROS). Pyoverdine and pyocyanine are blue/green pigments that are secreted into the environment. These pigments are siderophores, secreted to chelate mainly iron and other metal ions from the environment for cellular processes [66]. Pyocyanin also acts as an antimicrobial [67,68]. These pigments also act as signaling molecules for virulence factors and are quorum sensors [69]. In the low nutrient R2 media, there is abundant pigment production, visible within 24 h after plating, since metals are in low abundance in the minimal medium. Pulcherrimin and pyoverdine pigments also play roles in biofilm formation in Bacillus and Pseudomonas sp., respectively [29,70]. Azetidomonamide has also been shown to control biofilm formation in Pseudomonas [71]. Biofilm has been shown to play a role in the survival of bacteria by providing a hydrated coating (matrix) and providing a physical barrier against UV [10].

Potentially bioactive compounds were identified in every species that was sequenced (Table 5). Production of antibacterial compounds allows for bacterial communities to ward off other species and protect their limited resources in a low-nutrient environment. Bioactive compounds were especially abundant in the Bacilli sp. (Table 5). This is not surprising, since several antibiotics were discovered from B. subtilis and related species [8,72,73] Fengycin, bacillaene, bacilysin, subtilosin A, bacillibactin, plipastatin, and surfactin, which were identified in the B. mojavensis and B. inaquosorum strains, are known to be produced by B. subtilis group organisms (Table 5) [72,73,74,75,76,77]. Notably, though well known to be produced by Bacillus sp., some of these have low similarity to well-known compounds, suggesting they may have a similar core but with differences (Table 5). Unexpectedly, it was predicted that B. inaquosorum strains may produce zwittermicin A, known to be produced by Bacillus cereus, which is not part of the B. subtilis group. It is also unexpected that strain 0209A, B. inaquosorum, is predicted to produce staphylococcin C55, known to be produced by Staphylococcus aureus strain C55 [78]. Each Kocuria strain was shown to produce siderophores, essential for iron uptake, which hold potential for antimicrobial activity [79]. P. oceani and P. aeruginosa strains were also shown to produce several siderophores, including pyoverdine, pyochelin, and polycinine [66]. Iso1_2024, P. aeruginosa, is capable of producing pseudopaline, an iron chelator for zinc uptake [66] as well as two additional bioactive compounds: L-2-amino-4-methoxy-trans-3-butenoic acid, an antimicrobial, and hydrogen cyanide, which is a generally toxic compound that helps suppress plant fungal diseases [80,81]. 0516A, P. oceani, produces two additional antimicrobial compounds: one beta lactone and ambactin (Table 5). Closely related Pseudomonas species are known to produce beta-lactone antibiotics, such as P. fluorescens, which produces obafluorin [82]. Ambactin is infrequently described in literature, but has been identified in other genera, including Streptomyces and Ochrobactrum [83,84]. It has been shown to have anti-MRSA activities in some species [84]. All other compounds produced by 0516A and Iso1_2024 are known to be produced by Pseudomonas sp. [66,80,81,82]. It is remarkable that many of the predicted antibiotics or other bioactive compounds have low similarity with known compounds, suggesting these are likely to be novel. This is one of the reasons studying these bacteria is advantageous. The same is true for some of the predicted pigments. It will be beneficial to isolate these pigments and characterize their activities in the future.

Several strains had very high GC content and a high frequency of G4 formation (Table 6). Increased GC content may increase the stability of DNA since guanines and cytosines form more favorable nucleotide stacking patterns than adenine and thiamine [19]. Areas that are GC-rich are sometimes capable of forming G4 quadruplexes, a non-helical DNA secondary structure [21]. It has been shown that in some bacteria, G4’s regulate RecA-dependent DNA repair pathways [22]. This shows that G4 formation could play a role in helping organisms survive in a highly irradiated environment. While the strains from the Atacama did not have significantly different GC content or G4 count from similar strains from less extreme environments (Table 6), these features may aid in maintaining DNA stability. Future work will include a more in-depth analysis of G4 placement in these genomes as opposed to the high-level overview provided here.

This survey presents a small dataset that enables an evolutionary glimpse into bacterial survival strategies found throughout the Atacama Desert. This study does have limitations. For example, while 16S rRNA sequencing gives us an idea of what bacteria are present, it does not provide enough detail to resolve bacterial identities to the species level accurately. For genome-level data, we are working to sequence and characterize more isolates from our sampling trips. However, this study focuses on a culturable subset of pigment-producing isolates. This makes the research of interest since the pigment can be isolated and not just predicted. Future work will include the functional validation of predicted bioactive compounds.

Using the genomic sequences of the selected isolates, we identified genes and possible pathways beneficial for bacterial survival and fending off competitors in inhospitable environments. The isolates chosen for WGS were selected for their pigment-production properties, and we learned that though many of these are carried out by known enzymes and pathways, some are not. This suggests that these bacteria and the pigments they produce are an untapped source of novel enzymatic activities. The same is true for other compounds. For example, we identified and predicted the production of antimicrobial and bioactive compounds, key biofilm formation genes, molecules used for nutrient sequestering, and high GC content. Moreover, we also got a glimpse of possible gene horizontal transfer as we detected bioactive compounds normally found in other species in Bacillus sp., suggesting that extreme conditions may push genetic exchange for full adaptation. There is obviously much to be done, but the knowledge gained from this survey will enable further evolutionary understanding of key survival strategies in bacteria. Insights into how bacteria handle stress are relevant in many fields, including agriculture, biotechnology, and biomedical sciences.

The following supporting information can be found at: https://www.sciepublish.com/article/pii/795. Figure S1: An example of a plate from which bacteria were isolated for WGS; Figure S2: A phylogenetic tree generated by autoMLST2; Table S1: Locations, coordinates, and elevations of sampling locations for 16S sequencing; Table S2: Locations, coordinates, and elevations of sampling locations where bacteria where isolated from for WGS.

We would like to thank all Northeastern University (NU) students who participated in Dialogue of Civilizations (DOC) trips to Chile (2018, 2022, 2024). These students helped perform sample collection and bacterial isolation. Thank you to Joey Lehman Morris for co-teaching these DOC courses and participating in sample collection. Thank you also to our OneSeed team, including Sofia Mardones and Guillermo Maluenda, for guiding us to sampling locations. A special thanks also to the Riquelme Lab for supporting our team while in Chile.

Conceptualization: V.G.-C., C.R., Y.C., A.R.P. and N.T.C.; Methodology: V.G.-C., N.T.C., A.R.P. and E.I.; Formal Analysis: N.T.C., A.R.P., E.I., E.A., A.H. and M.C.F.; Investigation: N.T.C., A.R.P., E.I., E.A., A.H., N.T., M.C.F. and M.T.; Writing—Original Draft: N.T.C. and A.R.P.; Writing—Review & Editing: N.T.C., A.R.P. and V.G.-C.; Supervision: Y.C., C.R. and V.G.-C.; Project Administration: V.G.-C. and Y.C.; Funding Acquisition: V.G.-C. and Y.C.

Not applicable.

Not applicable.

WGS projects have been deposited into NCBI. Each sample is deposited under the following BioSample numbers: 0102A is under SAMN39706723, 0209A is under SAMN39706724, 0909A is under SAMN39706725, 0516A is under SAMN50557742, 1020B is under SAMN50557744, 0819A is under SAMN50557743, Iso1_2024 is under SAMN50557745, and Iso2_2024 is under SAMN50557746. Additionally, we assessed the G4 and GC content of three strains which are publicly available on NCBI. We assessed Bacillus subtilis with an NCBI ID of CP020102.1, Pseudomonas aeruginosa with an NCBI ID of NZ_CP027174.1, and Kocuria rhizophila with an NCBI ID of CP124833.1.

This work was funded by the Northeastern University (NU) Global Experience Office. E.I., E.A., A.H., N.T., M.C.F. and M.T. were funded by NU’s PEAK Undergraduate Research Fellowships. A.R.P. was funded by the NU Provost Dissertation Completion Fellowship. N.T.C. was funded by the National Science Foundation (NSF) Graduate Research Fellowship Program (1938052). V.G.-C. was funded by NuSci, a grant from HHMI. Y.C. was supported by NSF grant MCB1651732.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.