Resident bacterial populations mediating biogeochemical dynamics in a Uranium roll front deposit

Background Uranium-mineralized sandy aquifer, planned for a mining by in situ recovery (U ISR), harbors a reservoir of bacterial life that may influence the biogeochemical cycles surrounding the Uranium roll front deposits. Since microorganisms are likely to play an important role at all stages of U ISR, a better knowledge of the resident bacteria before any ISR actuations is essential to face environmental quality assessment. The focus here was made on the characterization of the resident microbiome of an aquifer surrounding uranium roll-front deposit that is part of an ISR facility project at Zoovch Ovoo (Mongolia). Water samples were collected following the natural redox zonation inherited in the aquifer, including the native mineralized orebody, and both the upstream and downstream compartments. Results An imposed chemical zonation for all sensitive redox elements through the roll-front system was observed in all water samples. High-throughput sequencing showed that the bacterial community structure was shaped by the redox gradient and the oxygen availability. Several interesting bacteria were observed including sulphate-reducing (e.g. Desulfovibrio, Nitrospira), iron-reducing (e.g. Gallionella, Sideroxydans) iron-oxidizing (e.g. Rhodobacter, Albidiferax, Ferribacterium), and nitrate-reducing bacteria (e.g. Pseudomonas, Aquabacterium), which may be also involved in metal reduction (e.g. Desulfovibrio, Ferribacterium, Pseudomonas, Albidiferax, Caulobacter, Zooglea). The taxa residing in each aquifer compartment followed a strong redox zonation differentiation, although as a whole the population of each water section seems to define an ecologically functional ecosystem containing suitable microorganisms that are probably prone to promote the remediation of the acidified aquifer by natural attenuation. Co-occurrence patterns confirmed a strong correlations among the bacterial genera suggesting either a shared and preferred environmental conditions or the performance of functions similar or

The use of each approach actually depends on the localization of the ore and also on the uranium content that must be between 0.1% and 0.2% [2]. From low grade ores (<0.1%), mining of uranium is based on utilizing In Situ Recovery (U ISR), one of the most economical strategies that accounts for almost half of the world's uranium production [3,4]. This strategy consists in recovering uranium by circulating acidic or alkaline extraction solution into the mineralized aquifer, using injection and recovery wells, through which the uranium-bearing solution is pumped to the surface for further processing [5]. The type of solution used as leaching reactant is mainly dependent on the mineralogical (carbonate content) and geochemical properties of the host rock. Such mining technique is mainly applied to sandstone-hosted roll-front uranium deposits localized in confined aquifers preferably below the water table [6]. The ore deposits occur within a redox barrier in a diffuse boundary delineating a reducing domain enriched in organic matter and sulfides on the down-gradient side (downstream compartment), and an oxidizing domain where sediments are usually oxidized on the up-gradient side (upstream compartment). Uranium being insoluble under reducing conditions, precipitates in this environment as uraninite and coffinite [6][7][8].
ISR is considered beneficial for many advantages involving no impact on the landscape, no de-watering of groundwater system above or around the deposit and a minimum distortion of the hydrological/environmental system [3]. However, the main challenges with this technology are 1) the incapacity to recover all of the U in the ore zone, and most importantly, 2) the complications in restoring groundwater to baseline conditions and the mined host rock to chemically reducing conditions able to immobilize any residual U or other contaminants. During ISR mining, acidification of groundwater could lead to mineral dissolution and heavy metal/radionuclide mobilization, which have the potential to leach into the environment contaminating adjacent aquifers [9,10]. These potential aquifer contaminations call for the urgent need to develop in situ remediation technologies for the removal of uranium released [11,12]. As the main focus is succeeding in the reestablishment of water pre-existing conditions, restoration of the mined aquifer following ISR mining is typically conducted by using groundwater sweep, reverse osmosis, reducing agents (such as H 2 S), bioremediation (stimulating metal reducing bacteria) and/or natural attenuation (NA) [5,[13][14][15]. The latter can be envisaged as a possible remediation strategy, and in fact, while considered as the prime groundwater restoration technique, it is often chosen as a cost-and labour-efficient strategy in many ISR operations [3]. NA is commonly used to reduce the concentration of dissolved contaminants by different processes including physical (e.g.: dilution), chemical (e.g.: precipitation), physicochemical (e.g.: sorption, ion exchange), and microbial (e.g.: reduction, biomineralization) processes [10,13].
One point of U ISR that has remained largely overlooked, and yet might help mitigate some of the issues mentioned above as well as improve ISR efficiency, is the monitoring of the naturally occurring aquifer microbiome. It is well known that bacteria are key organisms in many elemental biogeochemical cycles on earth (e.g. carbon, nitrogen, and sulphur cycles), and strongly influence the mobility of a wide range of metals (e.g., iron, manganese, copper and U) [16]. Under ISR conditions, the presence of key bacteria could partially govern the return to the pristine original biogeochemical conditions. Indeed, they could reduce and re-precipitate contaminants to an insoluble form through natural metabolic pathways while increasing their proliferation by using several amended electron donors [17,18]. Hence, the microbiome of the native aquifer may very likely play an important role at all stages of U ISR. Bioremediation may thus be considered as an alternative for biorestoration of post-ISR aquifer due mainly to several advantages such as shorter restoration periods, lower costs and high potential for a more effective restoration [15]. In fact, this strategy has been tested in many aquifer systems with some success [12,17,19]. Many field studies have demonstrated that bacterial mediation is often required to enzymatically catalyze uranium reduction in natural environmental systems even in the presence of Fe(II) and sulfide (i.e., abiotic uranium reduction is negligible).
From this concept, bioremediation may help the system to return to pre-ISR redox conditions characterized by a very low dissolved oxygen that thermodynamically favours low contaminant concentrations. However, data on water chemistry and bacterial community composition in roll-front context is relatively rare in the literature [20,21].
Water chemistry in ISR-prone aquifers is mainly governed by classical water/rock equilibria (solubility, sorption), while redox conditions are induced by biogeochemical reactions involving both water/rock equilibria and bacterial activities, especially when redox contrasts are noticed and the water temperature is below 100 °C. Therefore it is necessary to accurately determine hydrological, chemical, and especially, bacterial composition relative to original groundwater and their potential under roll-front relevant conditions to influence the constituent concentrations before ISR actuations. All of these point to the importance of assessing the presence of bacterial communities, prior mining operations, in order to verify to what extent the hydrochemistry is governed by such key bacterial players, leading in turn to the verification of natural attenuation or bioremediation hypotheses.
In this study, a deep 16S rRNA gene sequencing was conducted to characterize the bacterial communities present within groundwater collected from a uranium roll-front deposit planned to be mined by acidic ISR in Mongolia (Zoovch Ovoo) (Fig. 1). Beside the description of abundance, diversity, distribution, and co-occurrence of bacteria present in the aquifer zones, accurate analyses of the water geochemical characteristics were performed. While distinct abiotic features may be sufficiently recovered by groundwater monitoring, this study points out the importance of naturally occurring microorganisms and their potential to neutralize the acid employed at the ISR site. Thus, the complete description of the groundwater bacterial structure helps to improve the understanding of the natural attenuation process, and to predict the efficacy of remediation strategies.
Finally, this will also enable a better understanding of the influence of the acidification process on the water bacterial community and consequently on the biogeochemical cycling, allowing thus an effective evaluation and prevention of environmental risks at such U ISR sites.

Water chemistry
The electrical balance was assessed for every sample. All of them exhibit values lower than 5% aiming at using the total analyses for comparison and discussion of the results.
Chemical analyses of the different water samples involved in this study are represented in Table 1; complete water chemistry analyses are presented in Table SI-1 (Additional file 1).
Concentrations of several trace elements (PO 4 , Al, Co, Cu, Ni, Pb, V, Zn, Re, W) are below detection limits in all studied samples (Table SI-1) and are therefore not discussed here.
All the waters exhibit Na + K-Cl-SO 4 facies, with a relatively high electrical conductivity EC (ranging from 3.79 to 6.81 mS.cm -1 ) and circumneutral pH values.
Two grids of lecture regarding the water chemistry can be proposed and superimposed: i) the one governed by the contrast in salinity, mainly the Na and Cl concentrations, with higher values observed downstream the roll front [22,23]. This trend may be explained by the hydrogeology of the aquifer typical of an arid endorheic basin, as described earlier by Grizard et al. [24]; ii) the second is driven by the roll front chemistry inducing a redox contrast as usually proposed in such geochemical context, even if such literature is often limited by the lack of data [25]. Here, a focus will be made on this second aspect. As the salinity gradient tends to fade the redox contrast (Fig. 2), a correction using EC was applied to the reported concentrations (see Supporting information S1).
As expected in such geochemical system, a strong contrast in Eh was observed according (1)) and, on the other hand, the affinity of cations towards clay minerals through sorption processes [27][28][29].
Concentrations of major anions (Cl, SO 4 ) are therefore also varying in the same way with respect to the electrical balance.

Richness and distribution of the bacterial communities
Alpha-diversity analysis revealed no significant differences in bacterial richness of the water samples regardless of the metrics used (  (Fig. S1).
Beta-diversity revealed significant differences in the bacterial community structure and abundance. Water samples formed three distinct clusters in accordance with the redox zonation on the principal coordinate analysis (PCoA = Multidimensional scaling, MDS) plot of the bacterial community composition (using OTU abundance) of all tested wells (Fig. 3).
Accordingly, the analysis of the oxidized (PZOV_0013, and PZOV_0001), the mineralized (MOZO_0007, MOZO_0009, and PTZO_0001) and the reduced (PZOV_0024, PZOV_0021, and PZOV_0022) samples indicated that they were much more similar to one another in terms of bacterial abundance, which revealed significant clustering of replicates by water sample as well as by water physico-chemical properties (redox, O 2 , etc).

Bacterial diversity analyses and statistics
To further identify the similarity in abundance between the bacterial communities, a heatmap was performed based on the relative abundance of the genera with an average abundance of > 0.5% in at least one sample, which were defined as dominants. Abundance Similarity of percentages analysis (SIMPER) was used to determine the relative contribution of each individual genus to the dissimilarity between the three water clusters.
The average Bray-Curtis dissimilarity and the contribution of each genus to the total dissimilarity between communities in the different zonation waters (Mineralized, Oxidized, and Reduced) were calculated. The top major genera responsible for the microbial community difference (>98% contribution to cumulative dissimilarity) was summarized in To further investigate the taxonomic distribution and significantly differential dominant clades in the water microbiomes, LDA effect Size (LefSe) method was performed to compare and detect the abundance of groups at each taxonomic level and determine taxa involving significant differences among the different water samples (upstream, orebody, and downstream compartments). Fig. 6 depicts cladogram that visualize all detected significant taxa from domain to genus level in each water sample, being the reduced (PZOV_0024, PZOV_0021, and PZOV_0024) and mineralized (MOZO_0009, and PTZO_0001) waters with more significant taxa than the others (Fig. S3).
Correlations between bacterial genera and groundwater chemistry Co-occurrence and co-exclusion patterns among bacterial taxa and environmental parameters The co-occurrence and co-exclusion patterns among the bacterial communities and the environmental factors were explored using network analysis based on strong correlations  [31]. Interesting is also the metabolization by Pseudomonas of chemical pollutants in the environment, resulting in their suitability for bioremediation processes of many compounds including heavy metals and radionuclides (e.g. uranium) [32][33][34], polycyclic aromatic hydrocarbons, organic solvents (e.g., toluene), carbon tetrachloride, carbazole, and a variety of simple aromatic organic compounds [35,36].
Rhodobacter, an iron-oxidizing bacterium who is able to utilize Fe(II) as an electron donor for iron oxidation in the periplasm [37], was found among the dominant bacteria in these waters. In turn, the produced biogenic Fe(III) mineral phases would play a major role in the reduction of U(VI) under anoxic conditions [38]. Besides, the presence here of members of the family Alteromonadaceae, collecting diverse sets of Gammaproteobacteria, mostly marine in origin and requiring sodium to grow [39], was enhanced but not unexpected due to the high concentrations of Na and Cl characterizing these water samples. Lastly, members of the family Sphingomonadaceae also accounted among the dominant bacteria in the upstream compartments. These organisms are free living and widely distributed in nature [40], being aerobic with a strictly respiratory type of metabolism where oxygen is used as terminal electron acceptor and providing in this section of the aquifer their known ability to degrade some aromatic compounds, in addition to their known ability for biomineralization of U(VI) as U phosphate mineral phases [41], all of which makes them of interest to environmental remediation [42].
The mineralized sections of the aquifer encompass the roll-front ore deposits, which are Rhodobacter) are known to use periplasmic nitrate reductase as the major player for energy generation process through the mechanism of dissimilatory aerobic nitrate reduction [37,45]. Furthermore, although in low abundance, many iron-reducing bacteria were detected in these water samples including Albidiferax, Ferribacterium, Pedomicrobium, and Desulfovibrio (some of them are also metal-reducing bacteria), while members of the genera Aquabacterium, Rhodobacter, and Gallionella dominated the iron oxidizers group. In fact, neutrophilic Fe-oxidizing bacteria are rather associated with subsurface groundwater where reducing Fe 2+ -containing water flows from anoxic to oxic conditions. Here, the low O 2 concentrations retard abiotic Fe oxidation, allowing thus certain specialized Fe oxidizers (e.g., Gallionella, and Rhodobacter) to grow [46]. In these mineralized samples, Fe, nitrate, sulphate, and heavy metals/radionuclides such as Mn, Mo, Se, and U were present with considerable concentrations in comparison to the rest of the samples. These elements are used to be multiple electron acceptors, as NO 3 − , Mn(IV), Fe(III), U(VI), and SO 4 2− in many natural environments are used by microbes typically in sequence of energy yield, playing an important role in the biogeochemistry of the water.
For example, according to Elias et al. [47], Desulfovibrio vulgaris showed utilization of Fe(III) first, followed by U(VI), and finally sulphate in a competition experiment [47]. In the field, nitrate has been shown to be reduced prior to the U(VI) reduction, which often occurs simultaneously with Fe(III) reduction and prior to sulphate reduction [48,49]. Methylotenera are methylotroph organisms that can oxidize methylamine (one-carbon compound) as a single source of carbon, energy and nitrogen, while Hidrogenophaga is another interesting microorganism for its capacity to oxidize hydrogen as an energy source and to utilise CO 2 as a carbon source. In this genus, firstly, the oxidative carbohydrate metabolism is either achieved with oxygen as a terminal electron acceptor or by heterotrophic denitrification of nitrate [53], which is likely to occur in these water Lutibacter that is an interesting microorganism characterized by the capacity to adapt to environments with different oxygen affinities enabling them to thrive in microaerophilic to aerophilic conditions, which is further indicated by the production of diverse ferredoxin utilizing enzymes [55]. According to Le Moine-Bauer et al [55], some members of this genus display the complete pathway for denitrification in addition to oxygen respiration and nitrate reduction to nitrite, which were confirmed under microaerobic conditions.
Interestingly, these bacteria are characterized by sulfide:quinone oxidoreductase (SQR) involved in the sulphur metabolism, which may play an important role in sulphide detoxification in an environment with high sulphide concentration. Besides, strong activity for alkaline and acid phosphatase is characteristic of these bacteria, which is a plus considering their possible application in bioremediation of radionuclides such as uranium [55]. Some interesting bacteria such as Planctomyces, Pirellula, Algoriphagus, and unclassified members of the family Saprospiraceae seem to widely co-occur in these samples. The first two belong to the phylum Planctomycetes,whichare in general intriguing because they are the only free-living bacteria known to lack peptidoglycan in their cell walls; they are instead stabilized by the protein sacculus with disulfide bonds [56].
Planctomyces, a marine bacterium found in various habitats around the world (e.g. freshwater, saltwater, acid bog water), owns a relatively large genome that is thought to be necessary for the adaptation to changing environmental conditions [56]. In Planctomyces,as well as in the rest of the planctomycetes, "anammox" metabolism, a process in which ammonia is oxidized by nitrate to nitrogen gas yielding energy under anaerobic conditions, is mainly conducted. The other genus of this order, Pirellula, is in addition an interesting microorganism since it has two orders of magnitude more sulfatase genes, used in sulphur scavenging and for the effective assessment of sulphated compounds as an energy source [57]. In the water column, Pirellula are able to survive under low oxygen conditions, since anoxic conditions force them to switch to heterolactic acid fermentation or pathways involving formaldehyde conversion, if not to support growth then at least to allow basic metabolism maintenance [57]. The genus Algoriphagus, a marine psychrotolerant and moderately halotolerant bacteria, requires nearly seawater concentration of NaCl for their growth [58]. Thus their high detection here is not surprising since high concentrations of Na and Cl were characterizing this water samples.
Lastly suggest that some of the co-occurrence patterns may reveal associations of bacteria performing functions similar or complementary to each other, while others may co-occur due to shared and preferred environmental conditions. Thus, the co-occurrence and coexclusion patterns will improve our understanding of ecological niches and help to gain clues on ecologically bound microorganisms under these ISR relevant conditions.
Finally, our study demonstrates that 1) NA as a part of ISR remediation strategy requires a proof that the aquifer down-gradient of mining operations is able to reduce U and that 2) the in situ microbial communities through their metabolic versatility are prone to promote such metal reductions predicting and improving ISR efficiency. The next important step in our investigations will be to determine the microbial metabolic activities associated with the distinct aquifer compartments. Studying microbial processes enhanced or induced by ISR mining may significantly contribute to better predict remediation efficacy and even prevent operational failure. Future work should focus on monitoring and assessing the effects of acidification on the diversity and activity of these microorganisms, which is critical for the correct evaluation of the potential and efficacy of either natural attenuation or active bioremediation strategies, and definitely for the determination and mitigation of the environmental risks at the ISR sites.

Conclusions
In summary, this study was aimed to highlight how pre-mining distribution of the bacterial communities resident in the different groundwater compartments surrounding roll-front deposit can provide insights not only into the removal of U from groundwater, but also direct implications for the optimization of the ISR process. In terms of microbial communities, such background data and understanding of the natural system are a prerequisite to predicting microbial assistance in ISR operations as well as in further developing bioremediation strategies. In the mineralized zones, bacteria would be impacted by environmental influences in the oxic/anoxic boundary resulting from the acid solution injections. Significant population change would probably occur upon such intervention at the ISR mining site, with a large decrease in both bacterial cell number and diversity, as was seen in a previous study performed in Kazakhstan ISR mine after acid injection [20]. Although, more acidophilic bacteria and other mixotrophs similar to those abundant in an acid mine drainage, and detected in these pristine waters with low abundance (e.g. Acidovorax),, would be stimulated and the redox-zonation would be enhanced with additional facultative anaerobic and acidophilic microorganisms, present now with a low abundance. These conditions may be suitable for the natural attenuation of ISR mining through the reduction of metals/metalloids, sulphate, and other elements. The elucidation of the overall bacterial diversity and the chemistry of these groundwater underline the first step in the vast swath of U ISR context that remains to be explored for the correct understanding of the natural attenuation process.

Site description and sample collection
The studied site is located in the Gobi desert of Mongolia, Zoovch Ovoo in Sainshand region, whose resource is estimated in excess of 50.000 tons of uranium and where sandstone is the main host rock of the uranium found at this site (Fig. 1). Since 2015, this region of Mongolia is known with recoverable uranium resources producing, through ISR mining, about 2 % of the world uranium resources. The Zoovch Ovoo deposits are mineralized roll-front extended on approx. 10 km long and consist of a massive subtabular sandstone deposits, which are irregular, elongated lenticular bodies parallel to the depositional trend [60].
Groundwater was sampled from August to November 2016 with a special emphasis on the localization of each piezometer related to the U orebody (see Fig. 1). Two sampling campaigns were performed: 1) for the geochemical analysis and 2) for the bacterial diversity study. As the hydraulic gradient is very low (~ m/y) [24], we made the assumption that both sampling campaigns are representative of the same in situ conditions. Sampling was performed accordingly to the best standards. Hence the samples were taken after properly purging the volume of piezometer several times in order to 1) ensure representativeness of the water samples, but also to 2) benefit from this operation Water chemistry measurements
Clustered and annotated OTUs were finally analyzed in Explicet 2.10.5 [67]        Taxonomic distribution of water bacterial community at genus level Heatmap based on the relative abundance of the genera with an average abundance of > 0.1 % in at least one sample.

Figure 6
LEfSe analysis of microbial abundance with significant difference from phyla to genus level detected in the water samples collected from the different compartments of the roll-front deposit.

Figure 7
A canonical correspondence analysis plot reveals the relationships between the target bacterial genera and the geochemical parameters.

Figure 8
Network analysis revealing the co-occurrence patterns among bacterial taxa and environmental parameters. The nodes are colored according to modularity class.
A connection represents a strong correlations based on Pearson's correlation coefficient (ρ of >0.6). The size of each node is proportional to the number of connections, i.e., the degree.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.