Connectivity of bacterial assemblages along the Loa River in the Atacama Desert, Chile

Site description, sampling and environmental variables

The Loa River basin (Fig. 1) originates at the foot of the Miño Volcano (5651 m a.s.l), located in the Antofagasta Region of northern Chile). It is the longest river in the country (440 km) and the catchment extends across an area of 33,570 km². Crosses the Atacama Desert from east to west all the way to the Pacific Ocean (see Fig. 1), representing the main source for industrial mining (e.g., the Chuquicamata copper mine), agricultural, and domestic water supply (Valdés-Pineda et al., 2014).

The extremely arid conditions in the region favor high evaporation rates and high salinity. As such, all rivers in the region are temporal or endorrheic except for the Loa River, which is permanently exorheic, with flow varying seasonally in the range of 0.5–3 m³/s. The river drains a catchment rich in metalloids, including arsenic (concentrations of up to 27 mg As L⁻¹ are attributed to natural and human sources, notably the contribution from El Tatio Geothermal field) as well as copper, boron, chloride, sulfate and other elements (Romero et al., 2003; Bugueño et al., 2014).

The main tributaries of the Loa are the San Pedro, El Salado, and San Salvador rivers, which originate in the high Andes (Pell et al., 2013). As such, the Loa basin has a rain-dominated hydrologic regime, with flows typically increasing during the Austral summer (January and February) reflecting increased precipitation during the so-called Altiplanic Winter, when flows can be up to four orders of magnitude above baseflows (Jordan et al., 2015). However, this aquatic system is a critical water resource in a water-scarce region with an intense water usage pattern (Bugueño et al., 2014).

We sampled at eight representative sites distributed systematically along the Loa River (Fig. 1). Sites were selected to address issues that affect all sections of the river, such as pollution (from agriculture, mining, urban inputs). Sites were also selected that are identified within the Chilean National water network database for monitoring, established by the General Water Authority (DGA; its acronym in Spanish) which allowed us to cover a wide range of spatial heterogeneity that offers a unique opportunity to investigate how biotic and abiotic factors modulate the structure of bacterial communities in different sections of the river.

We sampled three sites from the upstream sector of the catchment during the austral spring of 2014: Lequena (U1), Output Reservoir Conchi (U11), and the river before the junction with the Salado River (U3), three in the middle sector (La Finca (M6), the river after the junction with the San Salvador River (M12), and the origin of the San Salvador River (M13), and two downstream (the river before Quillagua (L8) and the river mouth (L9)). Sampling sites U1, U3, and U11 have a mean flow of 1.2 m³/s and are characterized by the dominance of aquatic vegetation (algae, shrubs, and grasses) within the stream and on the two banks of the river. Supporting bofedales (local name for wetlands) and unique peatlands (Villagrán & Castro, 1997). Mean flow at the sampling sites M6, M12, and M13 are 0.25 m³/s. Human activities in this sector are associated with mining (e.g., the Chuquicamata, Radomiro Tomic, and El Abra mines), agriculture, and the chemical industry (Valdés-Pineda et al., 2014).

The hydromorphology of the Lower sector of the Loa catchment (L8 and L9) has been significantly altered since the early 1900s, following to the construction of three dams, Sloman, Santa Fe and Santa Teresa. These have not only regulated flow, but have also affected the natural sediment flux from the Andes to the Pacific Ocean (Aravena & Suzuki, 1990; Bugueño et al., 2014). The Quillagua oasis (L8) is a natural wetland with abundant aquatic life and was the main agricultural area in the Loa (Pell et al., 2013). Site L9 is the point where the river enters the Pacific Ocean but the water flow is too low (0.44–0.60 m³/s) considering the total available annual recharge of the Loa surface. The groundwater hydrological basin is 6.4 m³/s, which is attributed to evapotranspiration and consumptive water use (Romero et al., 2003; Jordan et al., 2015), thereby there is no major estuarine zone.

At each of eight sites (Fig. 1), we collected surface water (approximately depth from the surface 5 cm) and sediment samples. At each sampling site, three 1 L replicates of river water were collected manually into sterilized amber Boston bottles from the riverbank. These samples were pre-filtered firstly through 3.0 µm pore size polycarbonate filters (GSWP, Millipore) and then through 0.22 µm pore size to collect particle-attached and free-living bacteria, respectively, with an electric vacuum pump, connected to a portable electric generator in the field. Each filter was cut into half immediately following filtration in conical sterile tubes and both fractions were analyzed. The filtration time was no longer than 30 min, and the filters were replaced when the filtration speed slowed down due to the clogging.

Sediment samples were collected using a clean plastic scoop. A sterile spatula was used to transfer an aliquot of the scooped sample into a sterile 50 mL Falcon™ tube. All samples were transported on ice (4 °C) to the laboratory and stored at −20 °C prior to DNA extraction. At each sampling site we determined water temperature, salinity, conductivity, pH, redox potential, dissolved oxygen (DO) and chlorophyll concentrations in situ with a CTD Sea Bird 19 Plus multiparameter probe. Site locations and physical data are presented in Table 1.

DNA extraction, PCR amplification and sequencing

DNA was extracted from individual water and sediment samples using the DNA Isolation Kit PowerSoil ™ (MoBio Laboratories, CA. USA), according to the manufacturer’s instructions. Quantity and purity were tested in a NanoDrop 2000 (Thermo Scientific, Wilmington, DE, USA) and quality was checked by 0.8% p/v agarose gel electrophoresis. A mixture of DNA was made from the true replicate sample, resulting in 16 samples, divided into two sets of 8 samples for each substrate type (water/sediment) that were finally sequenced.

The hypervariable V1–V3 regions of bacterial 16S rRNA genomic sequences were PCR-amplified with universal primers (27 F mod 5′—AGR GTT TGA TCM TGG CTC AG—3′; 519 R mod bio 5′—GTN TTA CNG CGG CKG CTG—3′), with a barcode attached to the forward primer, through 30 cycles of PCR using the HotStarTaq Plus Master Mix Kit (Qiagen, USA). The PCR conditions were as follows: 94 °C for 3 min, followed by 28 cycles of 94 °C for 30 s, 53 °C for 40 s and 72 °C for 1 min, after which a final elongation step at 72 °C for 5 min was performed. After amplification, PCR products were checked in 2% agarose gel electrophoresis. Multiple samples of the same original sample were pooled together in equal proportions based on their molecular weight and DNA concentrations. Pooled samples were purified using calibrated Ampure XP beads. Then the pooled and purified PCR product was used to prepare a DNA library by following Illumina TruSeq DNA library preparation protocol. The read length was 300 bp, and the reads were both forward and reverse. Sequencing was performed at MR DNA (http://www.mrdnalab.com, Shallowater, TX, USA) on a MiSeq following the manufacturer’s guidelines. Sequencing results were obtained as fasta-qual and pipeline files.

Bioinformatic and statistical analysis

Physicochemical parameters of the Loa River

The physical and chemical characteristics of the samples are shown in Table 1. Although at all sampling sites conductivity was high for a river Conductivity values increased along the river from 3300 m a.s.l to the mouth. In the Upper river section (U1, U11 and U3), the samples show the lowest conductivity (1217 to 3874 µS/cm). While in the Middle-Low section (M6, M12, L8 and L9), the value increased from 5931 µS/cm to 18541 µS/cm.

Dissolved oxygen concentrations varied considerably (58.5 –105.1 mg/L) between sites, with higher altitude sites (>1200 m) showing lower concentrations (58.5 to 87.2 mg/L) and highest values being observed from sites nearest to the river mouth (103.4 to 105.1 mg/L).

Chlorophyll concentrations were related to the characteristics of each sampling site, with highest concentrations being recorded found at sites U1, U3 and L8 stations (11.33 mg/L, 38.84 g/L and 25.73 mg/L, respectively). These sites were characterized by abundant aquatic vegetation made up of instream algae and macrophytes and riparian vegetation. pH values ranged from 7.7 and 9.1 among the eight sampling sites, the highest value was recorded at station M6 and the lowest value at U3. On average, water temperature increased with distance from source from 11.25 °C to 19.28 °C, except for station U11 where temperatures averaged 7.46 °C. It should be noted that this parameter depends on the time of day when the sample was taken.

Spatial distribution of bacterial assemblages along the Loa river

After quality filtering and discarding singletons, chimeras and chloroplast sequences, a total of 473266 reads were obtained, with a mean of 29579 sequences per sample with an average length of 452 bp. These represented a total of 7935 bacterial ASVs with 99% similarity, for a total of 16 samples sequenced from the water column and sediment, sampling at eight sites located along the Loa River (Fig. 1).

Phylogenetic characterization identified 50 phyla, 129 classes, 319 orders, 543 families and 1056 genera at both substrates. Based on the phylogenetic classification, ASVs were assigned to Proteobacteria, Bacteroidetes, Epsilonbacteraeota, Planctomycetes, Firmicutes, Chloroflexi and Actinobacteria, as the most abundant taxa (relative abundance >1%), that made up 87% of the diversity in datasets.

The microbial communities differed between the water column and sediment samples. The water column contained taxa groups related to Gammaproteobacteria (Burkholderiaceae family with a relative abundance of 19.4%), Campylobacteria (Arcobacteraceae - 5.7%) and Actinobacteria, represented by Sporichthyaceae family (1.9%, especially in the U11 sampling site; Fig. 2A). Meanwhile, Bacteroidia (Flavobacteriaceae), Deltaproteobacteria (Desulfobulbaceae) and Planctomycetacia (Pirellulaceae), were most abundant taxa in the sediment samples, accounted for 10.4%, 5.3% and 3.6% of total ASVs, respectively (Figs. 2–3B).

Relative abundance of Arcobacteraceae and Marinomonadaceae families increased in the water column of the mid-sector of the Loa catchment, specifically in M6, M12 and M13 sampling sites, with a strong differentiation between the upstream and downstream. Similarly, Burkholderiacea, Rhodobacteraceae and Coxiellaceae were the major families in a downstream site (L8 and L9), which included agricultural lands and the desert habitat where the river enters the Pacific Ocean.

Furthermore, water bacterial communities from the Loa River, with the diversity identified at the genus-level, comprised mainly of Arcobacter (4.9%), Hydrogenophaga (4.3%), Marinomonas (3.2%) and Flavobacterium (3.1%) members. Meanwhile, Fusibacter, Halanaerobium, Actibacter and Thiobacillus exhibited a higher abundance in the sediment (1.0%). The results also revealed that the less abundant bacteria in both substrates (<0.7% of the total sequences) were unevenly distributed. However, the high percentage of unclassified ASVs at genus level in the sediment samples did not allow providing detailed information about the substrates at this level.

Relationship between α- and β-diversity at different sampling sections of the Loa River

Our results showed that α-diversity was slightly higher in sediment samples (compared to water) for a variety of metrics, including observed number, Chao1, and Shannon (Table 2). Sites located in the upper part of the catchment displayed the highest bacterial diversity. Conversely, sites in the middle sector of the catchment, the lowest α-diversity was observed in water samples. However, among the whole bacterial community, different sampling sectors did not vary significantly (p > 0.05). ANOSIM confirmed that the similarity of microbial communities was greater in the different sampling sectors (Global R: 0.07; P > 0.1) than in the substrate type (Global R: 0.561; P = 0.001).

The NMDS plots also showed marked differences in bacterial community structures, with a clear separation of water samples from sediment samples (Fig. 4A). PERMANOVA was then performed, and the results further corroborated that bacterial communities in the water column were significantly different from those in the sediment (p=0.001). Considering the relative abundance of ASVs in some sediment samples from the middle and upper sectors dominated by greater communities, as evidenced by their higher evenness or homogeneity (Inverse Simpson), it was suggested that sediment samples captured most of the bacterial taxa, resulting in higher diversity.

This observation was supported by the distribution gradient of the core, defined as those common to most samples in this study and shared bacterial families, which illustrated the extensive overlap between the upper and middle sections (15.5% ASVs) and less overlap between upper-lower and middle-lower sections with ASVs of 6.6% and 6.4%, respectively (Fig. 4B). The upper section had the largest number of unique families (see Fig. 4), followed by the middle and lower sections. This is reflected in the placement of bacterial communities in the upper sector, incorporating 76 ASVs (13.9%) until 47 (8.7%) and 45 (8.4%) unique ASVs of bacterial assemblages in the middle and lower sections of the river, respectively (Fig. 4B).

From 543 ASVs identified at family level, over 220 (representing 40.5%) were cosmopolitan to the three sampling sections of the river and distributed within a clear preference pattern by the sediment matrix. That is, about 162 families (29.8%) were unique ASVs in the sediment, while only 87 (16.0%) were specific to the column water (Fig. 4C).

Next, SIMPER analysis attributes 75.0% of dissimilarity recorded mainly to Arcobacteraceae, Burkholderiaceae and Flavobacteriaceae, which generated a contribution greater than or equal to 0.9%. Members of Desulfobulbaceae, Marinomonadaceae, Pirellulaceae, Anaerolineaceae, Pseudomonadaceae, Halanaerobiaceae, Alteromonadaceae, Flavobacteriaceae, Prolixibacteraceae, Spirochaetaceae and Hydrogenophilaceae contributed more than 0.5% for the community dissimilarity.

Influential physicochemical factors on bacterial communities along the river

CCA (Fig. 5) highlighted the differences in profiles obtained from samples of water and sediment and their relationships with different physiochemical parameters. It is important to note that the relation of a particular parameter with the community composition and the samples is provided by the length of a physicochemical parameter arrow in the ordination plot. The cumulative percentage of the variance of the species-environment indicates that the first and second canonical axis explained 51.6% and 22.8% of this variation, respectively. The remaining axes accounted for less than 16% of the total variance and therefore were not considered. It was determined that the first axis was positively correlated with all environmental variables, except with salinity (Fig. 5).

CCA analysis showed that ORP, chlorophyll and salinity were the main variables explaining variation in bacterial ASV composition. Sphingomonadaceae, Opitutaceae, Thiovulaceae, Flavobacteriaceae, Cellvibrionaceae, Saprospiraceae, Thiotrichaceae, Marinomonadaceae, Pseudomonadaceae, Devosiaceae, Bacillaceae were positively correlated with ORP, chlorophyll and DO mainly in the upper and middle section of the river. Meanwhile, Paludibacteraceae, Hyphomicrobiaceae, Rhodobacteraceae, Anaerolineaceae, Desulfobacteraceae, Desulfobulbaceae, Melioribacteraceae and Bacteroidetes_vadinHA17 families were strongly and negatively correlated with salinity in sediment samples across the different sectors.

Bacterial networks interactions

We inferred a co-occurrence network based on correlation relationships and p-values, adjusted with the Fruchterman-Reingold layout algorithm which allowed the visual identification of a clustered topology, expressed as a set of “closely nodes” (that is, taxa richness with average path length 1843) connected predominantly by blue edges (average degree 90447; Fig. 6A). These relationships between taxa shaped the network as a modular structure of bacterial communities (modularity index of 0.901; see Table 3).

Based on the modularity class, the network analysis of the total bacterial community contained 3,966 nodes and 358,713 edges. Fourteen groups of interactive genera were observed (Fig. 6A). The inner associations among each subcommunity indicated 3966 strongly connected components, in contrast with just 43 weakly connected components.

At the community level, core ASVs affiliated to the network including Flavobacteriaceae (9.3%), Rhodobacteraceae (5.9%), Arcobacteraceae (5.0%), Pirellulaceae (3.9%), Desulfobacteraceae (1.5%), Saprospiraceae (1.2%), Sphingomonadaceae (1.1%), Thiotrichaceae (1.0%) and Pseudomonadaceae (1.1%) were dominant families in both sediment and water samples community. The network also displayed taxa with less relative abundance and high values BC (see Table 3) such as Intrasporangiaceae (0.01%), Melioribacteraceae (0.1%), Erysipelotrichaceae (0.1%), Bacillaceae (0.1%), Devosiaceae (0.2%), Christensenellaceae (0.2%), Cellvibrionaceae (0.2%), Paludibacteraceae (0.3%), Micavibrionaceae (0.4%), Opitutaceae (0.4%), Bacteroidetes_vadinHA17 (0.5%), Hyphomicrobiaceae (0.6%) and Thiovulaceae (0.7%).

However, despite the bacterial communities varied among sites probably because the habitat heterogeneity, the patterns of presence/absence network showed that mainstem communities were spatially homogeneous, revealing a considerable overlap of ASVs among the different sampling sites. Indicating that the bacterial assemblages in the river present a well-defined core and strongly interconnected through the sectors but differ significantly (p = 0.001) between sediment and water samples (Fig. 6B).

Shift of bacterial assemblages with the substrate types along the river

We detected a total of 50.5% known bacterial phyla in the Loa River (50 out of 99 phyla reported) (Parks et al., 2018). This diversity is comparable to the Liu Xi River Reservoir in China, where 48 phyla were found (Wang et al., 2012). In contrast, the bacterial community of the Upper Mississippi River included 32 phyla (Staley et al., 2013). However, Payne et al. (2017) identified 37 phyla of free-living bacteria and 49 phyla in the particle-associated fraction collected from the Mississippi River.

Such diversity suggests that the Loa River could provide an “extreme local reservoir” that helps harbor the major prokaryotic diversity among the driest terrestrial biome of the world (Bull et al., 2016), playing a significant role in shaping bacterial communities and their dispersal into the stream, which are considered relevant to understanding the coupling of biogeochemical cycles and energy fluxes between the riverine and Pacific Ocean ecosystem (Velimirov et al., 2011; Aufdenkampe et al., 2011).

The results of this study also revealed that bacterial diversity gradually decreased from the headwaters to the lower reaches, especially in the water column where the Shannon index decreased (upstream: 5.95; middle: 3.92; lower: 3.85). Nevertheless, these changes were not significant (p > 0.05) between different sampling sectors. Therefore, it was assumed that α-diversity in the Loa River remained relatively constant across the catchment, with similar distribution trends in diversity. This trend has previously been observed in the Amazon River, where microbial communities were shown did not vary spatially over the 675 river kilometers (Doherty et al., 2017).

In contrast, significant differences (PERMANOVA, p = 0.001) in the bacterial assemblages were observed between water and sediment samples, exhibiting higher relative abundance in sediment than water samples. It is noteworthy that although the connection between the microbial community structure and spatial variation in rivers has been widely investigated (Böckelmann et al., 2000; Read et al., 2015; Adhikari et al., 2019), water and sediments together have rarely been explored (Lu et al., 2016).

In this respect, our results are consistent with previous studies in the Fenghe River (Lu et al., 2016), the Yellow River estuary (Wei et al., 2016), and the Santa Ana River (Ibekwe, Ma & Murinda, 2016), where greater prokaryotic diversity was observed in sediments compared to water, and bacterial communities in sediments were clearly distinct from those of the overlying water. These studies also reported that Proteobacteria phylum included more than 50% of the total number of bacterial sequences. However, a previous study on sediments of the Camarones River located approximately 250 km further from the north in the Atacama Desert, identified and isolated Alphaproteobacteria, Betaproteobacteria and Gammaproteobacteria in both water and sediments (Azua-Bustos & González-Silva, 2014). To our knowledge, the current study represents the first report of microbial community structure in the Loa River.

It is not surprising to find that the sediment entrapped the majority of the bacterial communities because this matrix, commonly documented as a relatively stable habitat, serves as a reservoir of nutrients and organic matter (Staley et al., 2015; Wei et al., 2016). Moreover, stream sediments in desert environments are habitats that likely promote unique patterns of microbial diversity (Zeglin et al., 2011), thus playing an important role in the formation of microbial mats that can help microorganisms assemble and resist environmental changes (Read et al., 2015; Lu et al., 2016).

Furthermore, patterns observed in this study are very similar to those reported in previous studies, which bacterial communities were dominated by Proteobacteria (46.3%), followed by Bacteroidetes, Epsilonbacteraeota, Actinobacteria, Patescibacteria, Planctomycetes, Firmicutes, Verrucomicrobia, Chloroflexi and Acidobacteria, accounting more than 93% of bacterial diversity. Such phyla are regarded as typical freshwater bacteria and are commonly found in large/urban rivers around the world (e.g., De Oliveira & Margis, 2015; Read et al., 2015; Adhikari et al., 2019).

Chen et al. (2018) and Hou et al. (2019) investigated the bacterial communities in riparian sediments and found that Proteobacteria, Firmicutes, Bacteroidetes, Acidobacteria, Actinobacteria, Chloroflexi, Verrucomicrobia, and Planctomycetes were predominant, comprising of the ranged of 85% to 93% of the total sequence data along a large-scale longitudinal gradient. In addition, comparative studies of the community structure of aquatic bacteria in urban rivers yielded similar results, indicating that the most abundant genes belonged to Proteobacteria, Bacteroidetes, Actinobacteria, Acidobacteria, Firmicutes, Chloroflexi and Verrucomicrobia (Kolmakova et al., 2014; Doherty et al., 2017).

Factors shaping the bacterial assemblages

The overall findings showed that bacterial assemblages of the Loa river were correlated with changes in physicochemical properties, including salinity, DO concentration, ORP, and chlorophyll. This is agreement with previous studies from several river systems reporting how shifts in the diversity and distribution of the bacterial community composition were strongly predicted by environmental variables (Ortega-Retuerta et al., 2013; Zhang et al., 2019).

Our results also showed the possible impacts of the anthropogenic activities like discharge on physicochemical variables measured, which in turn, could explain the dissimilarities observed in bacterial assemblages at upstream and downstream observed (Kolmakova et al., 2014; De Oliveira & Margis, 2015; Read et al., 2015), especially in the middle to the downstream section of the river, where a decline in diversity (Drury, Rosi-Marshall & Kelly, 2013; Jani et al., 2018), as well as distinct groupings of bacterial families that are often associated with nutrient pollution (e.g., Arcobacteraceae, Flavobacteriaceae, Rhodobacteraceae, Burkholderiaceae, Sphingomonadaceae and Pirellulaceae) were observed. For example, Arcobacteraceae including Campylobacterales, was particularly dominant in untreated and river samples obtained from a wastewater discharge area (Bai et al., 2014). Similarly, Rhodobacteraceae and Pirellulaceae were among the most dominant groups in the most urbanized sites along a highly impacted stream (Roberto, Van & Leff, 2018).

Banerjee, Schlaeppi & van der Heijden (2018) reported that various members of Burkholderiaceae, more specifically of Burkholderiales order, largely distributed in the water column of the upper and lower sections of the river, were consistently identified as keystone taxa in different ecosystems, such as desert shrubland, Antarctic lakes, contaminated soil, sediments of Amazonian lakes and other rivers. The elevated abundance of Flavobacteriaceae members may reflect the high inputs of organic matter in the middle sector of the river because this family is mainly particle-associated and strongly dependent on the availability of organic matter (Newton et al., 2011). Likewise, Desulfobacteraceae, has been reported as family typically associated with regular organic loads (Saravanakumar et al., 2012).

In the present study, the middle areas of the river experienced higher degrees of discharge associated with mining effluents (Chuquicamata and El Abra mines), as well as agricultural and urban wastewater discharges (Romero et al., 2003; Valdés-Pineda et al., 2014). In this context, a marked increase in Sporichthyaceae from the Actinobacteria phylum obtained from water samples at site U11 was observed. The family Sporichthyaceae includes species adapted to both impacts associated with urban wastewater (Newton & McLellan, 2015) and waters contaminated by acid mine drainage (Ettamimi et al., 2019). This can support the notion that local conditions in different sections of the Loa River could influence the structure of the microbial community (Payne et al., 2017).

Co-occurrence network of the bacterial assemblages in the Loa River

Desert bacterial communities are recognized as being unstable and vulnerable to change (Wang et al., 2019). However, the increased of water and nutrient availability afforded by the water supply from a river, provides increased scope for interactions between different species (Zeglin et al., 2011). Therefore, with regard to the study of bacterial assemblages, our co-occurrence analysis provided insights into structural variation and interactions between bacterial communities in the water and sediment along the Loa River.

The network analysis identified common ASVs as strongly interconnected mainstream community assemblages along the river that were spatially homogeneous. Among total of 543 ASVs identified at the family level, over 220 (40.5%) were ubiquitous in all three river sections and distributed with a clear preference for the sediment matrix. The core bacterial assemblage, present throughout the study area (Staley et al., 2013; De Oliveira & Margis, 2015; Savio et al., 2015), was dominated by members of the Proteobacteria, including mainly Sphingomonadaceae, Hyphomicrobiaceae and Rhodobacteraceae and Bacteroidetes (Paludibacteraceae, Flavobacteriaceae and Saprospiraceae).

In addition, recent studies of fluvial microbial co-occurrence networks identify gatekeepers affiliated with Saprospiraceae and Sphingomonadaceae (Widder et al., 2014). Sphingomonadaceae are strict aerobic chemoheterotrophs, frequently reported to be found in soils and eutrophic waters and on corals and plant surfaces (Vaz-Moreira, Nunes & Manaia, 2011). Moreover, this family has been found in habitats that contain high proportions of xenobiotic and recalcitrant aromatic compounds in various studies reported (Zhou et al., 2016).

The core communities, found in the Loa River were also comprised of numerically inconspicuous families including Intrasporangiaceae, Opitutaceae, Hyphomicrobiaceae, Melioribacteraceae, Bacillaceae, Devosiaceae, Paludibacteraceae, Cellvibrionaceae, Micavibrionaceae. The major ecological role of these taxa in different aquatic ecosystems has remained unclear. Therefore, their characterization, along with a spatial disturbance gradient, may help us understand the bacterial community functioning in the Loa basin. Nevertheless, keystone taxa are increasingly recognized to regulate the functioning in terrestrial and aquatic environments (Jousset et al., 2017).

Our findings concur with those of previous studies in which keystone species that are typically not the most abundant/dominant species, were identified (Zhao et al., 2016; Hu et al., 2017). Thus, it is important to take into account that the abundance of a species is not the best determinant of its contribution to the community (Lynch & Neufeld, 2015).

The Loa River did not exhibit clear fragmentation between different sampling sections. The higher bacterial diversity reported in this study likely increases network interconnections and the robustness of the community at a larger scale. A possible explanation for these observations is that the flashy nature of the hydrological profile of the Loa River system, where large water flow occurs during rain-dominated hydrological regimes in December to-March, may drive the homogenization of microbial assemblages in the river system (Hu et al., 2017). These findings are important as they show that the strong connectivity of the bacterial assemblages in a desert river has been maintained by major hydrologic modifications since the early 1900s to support mining and other human activities (Bugueño et al., 2014). Such connectivity could enhance the survival and mutually beneficial interactions between microbes (Graham et al., 2017), in depauperate environments such as desert rivers, where nutrients are limiting.

This is a first attempt to characterize the microbial community of the Loa River. We are aware that future work must include a temporal component to improve understanding of the temporal microbial dynamics in multiple hydrological periods. In addition, we recognize the need to include other specific components of the microbiome, namely archaea and eukaryotes, whose interactions can modulate the entire environmental, biological function under natural conditions, and the need to examine the influence of a wider range of physicochemical factors such as nutrients and heavy metals on microbial community. This would provide insights into the likely ecological consequences of the ongoing deterioration of this important aquatic system, as well as the global ecological risk assessment of desert environments in the context of anthropogenic changes.