The greatest diversity of species on Earth is found in the tropics. However, the arid and semi-arid zones of the American tropics are poorly known. Our work metagenomic analysis describes, for the first time, the influence that seasonality, vegetation, and physical-chemical properties have on the microbial taxonomical and functional diversity, in a semi-arid zone of La Guajira. Our results suggest that two of the main drivers of microbial communities in these semi-arid environments are fertility islands and the rainy season.
Comparison of the chemical properties of the samples
We found 15 physicochemical parameters such as sand, EC, OM, OC, P, TN, S, CEC, B, Ca, Mg, K, Mn, Zn, and K-sat with statistically higher values in D than W (Table 1; Supplementary Table 2). These differences could be explained by a shutdown of microbial activity during the periods of drought that predominate throughout the year. In arid and semi-arid ecosystems, biological processes such as nutrient cycling are often limited by moisture availability, so a decrease in moisture can have a strong negative impact on microorganism processes, which are generally shut down, allowing nutrient accumulation in soils to occur during warm and dry periods [19]. On the contrary, in W, nutrients could decrease because moisture increases microbial activity, excessive plant growth increases nutrient consumption [20], and nutrients can be lost by leaching [21].
The correspondence analysis shows differences between VD, VW, CD, and CW (Fig. 2A), but VD had significantly higher concentrations of EC, OM, OC, TN, S, CEC, B, Ca, Mn, and Zn when compared to CD, VW, and CW. This is expected because plants on fertile islands allow greater OM accumulation compared to bare soils, while at the same time, there is low nutrient consumption due to low biological activity and loss of vegetation during D. In contrast, Fe was higher in both VD and CD because Fe2+ and Pi form insoluble salts that allow their accumulation in soils during the dry season [22]. The physical-chemical parameters that explain the differences between these four conditions were Ca, Mg, Mn, S, EC, CEC, OC, and OM (Fig. 2B and 2C). Similar to Liu et al (2021) [23], who reported no differences between the physical and chemical parameters of different nurse trees in the same area, we found no significant differences between different nurse trees in the same season (Supplementary Table 2).
Comparison of the microbiomes
Microbial community composition was estimated by reconstructing the 16S rRNA gene (size range between 955–1550 bp). The Bacteria domain (97.35%) was the most abundant, followed by Eukarya (2.11%) and Archaea (0.54%). Fourteen bacterial phyla were identified; among them, the most abundant were Proteobacteria (38.28%), Actinobacteriota (32.90%), Bacteroidota (9.62%), Acidobacteriota (5.79%), and Gemmatimonadota (4.94%) (Fig. 3) (Supplementary Table 4). It is not surprising to find the first four phyla in these soils as they correspond to cosmopolitan groups reported in high abundances in deserts and different environments from hot and dry to wet and cold [24]. Additionally, members of Gemmatinomonadota are characterized by withstanding stress conditions [25]. Actinobacteriota is the dominant phylum in CD (49.92%); in contrast, the phylum Proteobacteria prevails in VD (47.80%). This suggests that vegetation could play an important role in microbial community structure due to organic matter input, nutrient availability, and moisture retention, although humidity is also an important factor [26]. The distribution of these phyla in semi-arid areas reflects their ability to withstand and sustain themselves in nutrient-poor soils, also it may suggest that these bacteria play an ecological role in semi-arid environments. Proteobacteria play a key role in different biogeochemical cycles, several members of this phylum have the ability to metabolize different sources of leaf litter carbon from the canopy [27], and Actinobacteriota have sporulation capacity, which enhances their ability to withstand harsh oligotrophic environments, such as sandy soils with low organic matter characteristic of the absence of vegetation [26] (Fig. 3). Taxonomic composition represented in a PCoA separated samples from seasons D and W (Fig. 4A). Each season grouped samples with vegetation and separated them from unvegetated controls (Fig. 4B and 4C), but microorganisms were not grouped by nurse tree. The Chao1 and Shannon indices (Fig. 5) also support the differences between seasons and vegetation, which showed differences between D and W, and the presence of vegetation between C and V.
For all tree species, the associated microbial community had a similar taxonomic composition and there were no unique genera in these trees. This is congruent with the research of Li and collaborators (2021), showing that no matter the nurse tree taxa (legumes or not legumes), there are similar effects on the microbiome of the soil under the canopy. We found 96.67% similarity in abundances. Only five genera (Novosphingobium, Flavisolibacter, Stenotropobacter, Longilinea, and Anaeromyxobacter) showed significant differences among trees during the wet season. This high similarity indicates that the composition of the organic matter and physical protection provided by the tree canopy does not present major differences [3], which could be related to the lack of significant differences in most physicochemical parameters of the soil among trees.
Our results show an increase in taxonomic abundances and functional potential during the wet season, as water in arid ecosystems is the main limiting factor for microbial diversity [3]. Generally, during the dry season, the abundance of bacteria decreases markedly. Only some phyla remain dormant in drought conditions, but others cannot survive [28]. Moisture in the wet season activates microorganisms, increasing microbial abundance and diversity, even after years of drought [29]. Rainfall promotes significant changes in the landscape as short-lived plants cover the soil, influencing the microbial community through their inputs of organic matter and rhizosphere exudates. However, in our samples, in the presence of vegetation, the phyla Proteobacteria, Actinobacteriota, and Bacteroidota were more common during the dry season, particularly in exposed surface soils. This may be caused by the chemoheterotrophs that are dominant in these phyla, which maintain mild transcriptional and metabolic activity, mainly using their energy for maintenance rather than growth [30].
We identified 151 genera in taxonomic analyses. Sphingomonas, Azospira, Solirubrobacter, Flavisolibacter, Geodermatophilus, Massilia, Microvirga, Blastococcus, Rubrobacter, and Nocardioides were the top 10 predominant genera; these represented about 65% of the counts. However, 20% of the counts were unclassified genera. Fifty-six genera were shared in all the conditions (VD, VW, CD, and CW). The highest abundances were recorded in VW with 106 genera and the least in CD with 28 (Fig. 6). The success of these genera lies in their ability to adapt to the limiting conditions of arid and semi-arid zones. For instance, the ability to withstand drought and high radiation conditions has been reported for genera such as Flavisolibacter [23] and Geodermatophilus [31]. Further, these genera can metabolize a variety of organic and inorganic compounds, as reported for Sphingomonas [32], Solirubrobacter [33], Flavisolibacter [34], Geodermatophilus [35], and Azospira [36]. Likewise, they can produce secondary metabolites [32] and colonize plants in dry environments, as has been reported for Sphingomonas [37] and Geodermatophilus [35].
We found unique genera in all the conditions as follows: 56 in VW, 21 in CW, 13 in VD, and 5 in CD. These unique genera represent the rare prokaryotic fraction of these soils [10]. Most of the unique genera belonged to Proteobacteria (31), followed by Actinobacteriota (12) (Supplementary Table 5). The presence of unique genera suggests high specialization for the utilization of energy sources and the ability to live in specific environmental conditions. Unique genera in VD have a high metabolic versatility in using organic and inorganic compounds [14], plus the ability to form resistance structures to withstand dry environments with high radiation [38]. Unique genera in CD, on the other hand, commonly metabolize simple carbon sources [23] to survive in environments with low availability of nutrients [39] and use different strategies of resistance to withstand dry environments with high radiation [40]. The unique genera in VW have periods of dormancy during the dry season while the rains begin [41] and can promote plant growth [42]. The wet season can favor the occurrence of aerobic genera [43] with the ability to break down cellulose [44].
Physical-chemical correlation with taxonomic composition
Pearson's correlations were calculated to identify key environmental drivers of microbial composition, and only significant values (p < 0.05, r > 0.6) are shown (Fig. 7). Overall, the phylum Bacteroidota presented a positive correlation with most environmental parameters. Meier et al. (2021) [45] describes Bacteroidota as a group with heterotrophic genomic potential, so they could be chemosynthetic primary producers, using atmospheric trace gases as the main energy source. Phyla such as, Chloroflexi, Crenarchaeota, Myxococcota, and Verrucomicrobiota, were negatively correlated with EC, OM, OC, TN, Ca, and Mn; we found that these phyla were more abundant in VW, where these physical-chemical parameters decreased because humidity increases microbial metabolic activity [20]. The genera Flavisolibacter, Massilia, Pseudonocardia, and Unclassified Chitinophagacaceae showed a higher positive correlation to TN, OM, and OC; these genera were the most abundant in VD, likely due to the high nutrient input from vegetation. Bacteria activity, though, is expected to be low due to the lack of moisture [19]. It is widely recognized that organic matter is one of the main modulating factors of microbial activity and diversity in the soil; however, the role of microorganisms in litter and wood degradation in arid environments is very limited [46]. Sulfur positive correlation includes Sphingomonas, Microvirga, and Massilia; it makes sense because some Sphingomonas species encoded the genes to promote phosphate and sulfur assimilation [47] (Lombardino et al., 2022), also exist evidence that some Microvirga species encoded genes related to sulfur metabolism (Jiménez-Gómez et al., 2019) and the Sox complex, important to the oxidation of reduced sulfur species (Meier et al., 2021). Anaeromyxobacter, Uncultured Pyrinomonadaceae, and Uncultured Nitrosomonadaceae negatively correlated with all parameters except sand (Fig. 7).
Microbial function potential
The functional categories with the highest gene abundances were metabolism (29.3% of the total predicted genes), genetic information processing (8.4%), environmental information processing (6.7%), and cellular processes (5.5%). Functional characterization of the metagenomes showed significant differences in these functional categories at the season and vegetation level (Supplementary Table 6). Comparing W with D, there were 130 modules with significant differences from the 21 pathways reviewed; the highest abundances were found in W with 88 modules. The remaining 42 had higher abundances in D (Supplementary Table 6). This may be because microbial functional potential appears to be largely determined by microbial community composition. The taxonomic and functional profiles of soil microbial communities in arid and semi-arid environments are strongly influenced by rainfall [29, 48, 49]. During the dry season, the energetic input is very low to cover the energetic needs of the microbial community. Only bacteria that can generate dormancy structures and some genera tolerant to desiccation survive in these conditions; this decreases the abundance and diversity of microbes during this season and, therefore, their functional potential [30]. In contrast, vegetation showed higher abundances of KEGG modules than bare soils, with 61 modules in D, of which 43 corresponded to VD and 21 modules in W with 19 in VW. The metabolism category presented the greatest differences in pathways (10) and modules (75) (Supplementary Table 6). This could be since plant litter are the main source of energy in oligotrophic soils and determine the ecological dynamics of the soil, the physicochemical parameters, and, therefore, the microbial structure (Berg et al., 2015; Hu et al., 2019; Chukwuneme et al., 2021).
Analyzing the abundances of genes belonging to the energy metabolism pathway, we found significant differences between W and D in the modules KO00920 sulfur metabolism, KO00720 carbon fixation pathways, KO00190 oxidative phosphorylation, KO00910 nitrogen metabolism, KO00680 methane metabolism, and KO00196 photosynthesis - antenna proteins. Consistent with the overall functional analyzes, modules KO00920 sulfur metabolism, KO00910 nitrogen metabolism, and KO00680 methane metabolism showed higher gene abundance in W (Supplementary Fig. 1).
Biogeochemical cycles
Methane metabolism. In KEGG orthology, we identified 163 KO related to methane metabolism and modules of methanogenesis, coenzyme M biosynthesis, methane oxidation, formaldehyde assimilation, and acetyl-CoA pathways. These modules were mainly influenced by seasonality. The periods of flooding and drying could generate microniches with different availability of oxygen that under anaerobic conditions promote methanogenesis and aerobic conditions for the methane oxidation. Our data suggest the absence of complete methanogenesis since the mcrA - methyl CoA reductase gene and methanogenic taxonomic groups were not detected. In the soils, methane oxidation prevailed despite the lack of key genes (pmoA - particulate methane monooxygenase). Methane could be used as a carbon source through the assimilation of formaldehyde by methanotrophs such as Methylobacter and Chloroflexus that use the formate hydrogenase to oxidize formaldehyde to CO2 (Ward et al., 2019).
Nitrogen metabolism.
We identified 42 KO related to nitrogen metabolism in KEGG orthology. Dissimilatory nitrate reduction (M00530), denitrification (M00529), and assimilatory nitrate reduction (M00531) presented the highest abundances without seasonal influence. These processes occur under anaerobic conditions and can increase with soil moisture [50]. Therefore, these bacteria must be protected from wet season oxygen tensions by colonizing anaerobic soil spaces, producing exopolysaccharides, and altering the composition of membrane lipids and cell wall components [51]. Dissimilatory nitrate reduction presented significantly higher abundances in vegetation than in control during the dry season (Fig. 9). This is consistent with increases in nitrate reductase activity due to increased organic matter [50], and the addition of N [52]. These nutritional conditions were higher in the fertile islands than in bare soil.
There was no evidence of genes related to nitrogen fixation, given their very low abundances in most samples (Fig. 9). On the contrary, despite the limitations of nitrogen in the ecosystem, denitrification prevailed. Lynch et al. (2014) [53] reported predominance of denitrification and nitrate reduction over N fixation due to a strong selection of the bacterial community of desert soils to assimilate ammonia and nitrate and catabolize trace gases such as H2, CO, and C1 organic compounds. Organic matter is the main source of nitrogen in desert soils [5], because the fixation of N is an energetically expensive process and is highly regulated by the availability of oxygen, nitrogen, carbon, and trace metals such as iron, molybdenum, or vanadium [54]. However, legume symbiont diazotrophs such as Mesorhizobium and Bradyrhizobium were detected in low abundances. This could be associated with the fact that diazotrophs can develop in free life before colonizing leguminous species that can provide the conditions to fix N and that the samples analyzed are not rhizospheric soil. Diazotrophs such as Mesorhizobium, Rhizobium, and Ensifer predominate in nodules of leguminous plants with the potential to promote plant growth under the limited conditions of arid and semi-arid zones [55]. Azospira reduces N sources such as nitrate [56] and N2O under microaerobic or anaerobic conditions from easily degradable substrates such as acetate [36]. Azospira has genes for denitrification such as narG (NO3- reductase), nirK (NO2- reductase), norB (NO reductase), and nosZ (N2O reductase) [36], and can reduces N sources such as nitrate [56]. Although Azospira can fix nitrogen freely under microaerophilic [57] and nitrogen-scarce conditions [58], the functional potential for nitrogen fixation was not found in our samples. However, knowledge of diazotrophs in deserts and their role in these ecosystems is limited (Suleiman et al., 2019).
Sulfur metabolism
We detected 18 KOs related to sulfur metabolism. The KOs related to the sulfate assimilatory reduction module (M00176), were found in higher abundances; followed by KOs corresponding to thiosulfate oxidation (M00595). The higher abundance of KOs related to sulfate reduction in fertile islands may be related to the accumulation of organic sulfur in the soil due to OM input [44] since this form of sulfur must be mineralized to remain in its plant-available form or mobilized by leaching [59]. Genes and genera related to sulfate reduction had high abundances in D while those of the thiosulfate oxidation in W. This could be associated with: the available S in D was higher than in W, the presence of anoxic niches related to the seasonality and the ability of cyanobacteria, microalgae, fungi, lichens, mosses, and some prokaryotes to secrete polysaccharides and some mucous substances that limiting the available oxygen [60]. There is no evidence of dissimilatory sulfite reductase due to the absence of genes for the alpha dsrA subunits, beta dsrB [EC: 1.8.99.5], and adenylyl sulfate reductase, subunit B [EC: 1.8.99.2]. In addition, abundant genera such as Rubrobacter are associated with sulfate reduction by adenyl sulfate kinases and phosphoadenyl sulfate reductases in deserts (Miralles et al., 2021). On the other hand, thiosulfate oxidation by the SOX complex (KEGG module M00595) was favored by humid conditions (Fig. 9), possibly due to the decrease in available S and organic matter in the wet season. Also, some genera associated with S oxidation were favored by humidity, such as Geodermatophilus [61].
Although metagenomic studies in arid environments have increased in recent years, expanding the knowledge of the taxonomic diversity and functional potential of microorganisms, the knowledge of these microbial communities is still very limited. New insight into microbial adaptation and survival strategies in these environments is related to the contribution of organic matter to the soil and its influence on the composition of the microbial communities they harbor. As expected, our results confirm that one of the main drivers of bacterial communities in these semi-arid environments are fertile islands and the rainy season. However, little is known about how leaf litter influences bacterial community composition in these soils and how the microbial community changes in these environments under different precipitation events. We originally hypothesized that the species of the three nurse trees could be an important driving force of desert island diversity. However, we found no major differences among microbial populations associated with the trees in the same season. This could be related to all three trees being legumes of the same size, the lack of differences in the physical-chemical parameters in the fertile islands, and not rhizospheric samples. Similar results were found by Li et al. (2021)[6]; they found no differences between trees in the same area, not even between legumes and non-legumes, and that most physical-chemical parameters were not related to microbial communities, except for organic matter, carbon and nitrates. Likewise, the canopy cover offered by artificial plants in arid environments does not induce significant changes in the diversity of bacteria [3]. Metatranscriptome and metaproteome analysis will help us better understand how factors such as of fertile islands and precipitation regimes influence microbial communities and its gene expression. This could contribute to the potential knowledge of these microorganisms, applied in agriculture to curb deforestation, propose food security management plans and combat the desertification processes so common in arid areas.