Groundwater was sampled using peristaltic pumps from 20 monitoring wells within a shallow (20–30 ft), unconfined alluvial aquifer adjoining the Colorado River located 0.3 miles east of Rifle, Colorado (USA) (Figure 1). Groundwater flows in a south-southwesterly direction and discharges into the Colorado River. The Holocene-age alluvial sediments consist of sandy gravel and gravelly sand containing silts and clays, characteristic of many alluvial aquifers. Distributed throughout the aquifer are also lenses of naturally reduced sediments containing reduced minerals and high concentrations of organic carbon originating from buried plant material (Campbell et al., 2012; Janot et al., 2015). Groundwater within the aquifer is typically suboxic (<1 mg L⁻¹) and contains spatially varying concentrations of reduced chemical species such as Fe(II) (10–50 μM; Williams et al., 2011). Leaching of U from former stockpiles of ore and mill tailings stored at the site resulted in a persistent plume of groundwater with elevated U concentrations (>100 μg L⁻¹; Zachara et al., 2013).

The major plume region has been described in prior studies conducted at the Rifle field site (Zachara et al., 2013). For the purpose of this study the boundaries of the plume region were defined where groundwater U concentrations exceeded 100 μg L⁻¹. An elevated groundwater U concentration (171 μg L⁻¹) was measured in well 656 in the eastern portion of the site, but groundwater in this region is geochemically distinct from the major plume with significantly lower groundwater DOC, DIC, and sulfate concentrations. Accordingly, well 656 is excluded from assignment to the major contaminant plume that is located at the center of the field site. Eleven monitoring wells within the uranium contaminated plume and nine monitoring wells outside the plume region were sampled in this study (Figure 1). Each well was approximately 6 m deep. Wells directly associated with regions impacted during prior experimental augmentation (Anderson et al., 2003) were excluded from this study. Further site details have been described elsewhere (Anderson et al., 2003; Vrionis et al., 2005; Zachara et al., 2013).

Replicate groundwater samples (<50 mL) for geochemical analyses and enumeration of cells and viruses were collected from purged (12 L, ca. 1–1.5 well volumes) wells at a 5 m depth using a peristaltic pump. Groundwater was filtered through 0.45 μm PVDF filters to remove sediment particles for cell enumeration, and through 0.1 μm PVDF filters to remove cells for virus enumeration. In order to reduce background noise and improve filter clarity, samples for virus enumeration were treated with a nuclease, DNase I (10 U mL⁻¹; Danovaro and Middelboe, 2010; Carreira et al., 2015). In this study, viruses are operationally defined as DNase resistant, SYBR Green I fluorescent particles smaller than 0.1 μm and larger than 20 nm. Using this operational definition, defective viruses, gene transfer agents, and other unknown particles may plausibly be included. A maximum cutoff size of 0.1 μm was selected to prevent inclusion of cells smaller than 0.2 μm as was identified by Luef and colleagues at the Rifle site (Luef et al., 2015). To avoid including viruses larger than 0.1 μm in the cellular fraction, a minimum cutoff size of 0.2 μm was used for cell enumeration. A comparison of the sub 0.2 μm fraction and the sub 0.1 μm fraction showed that there was no statistically signficant difference (Figure S1). Thus viruses and cells that are between 0.1 μm to 0.2 μm were not enumerated in this study. Aliquots (1 mL) were preserved for both virus and cell counts by adding glutaraldehyde to a final concentration of 0.5% and incubating 15–30 min at 4°C prior to freezing in liquid N₂ (Brussaard, 2009). Samples were packed in dry ice and shipped to the University of Nebraska-Lincoln for storage at −80°C prior to enumeration.

Duplicate samples were thawed for enumeration by epifluorescence microscopy. Viruses were collected on Anodisc filters (0.02 μm), while cells were collected on black polycarbonate filters (0.2 μm). Between 0.5 and 1 mL of sample was passed through each filter. SYBR Green I (400x dilution from original stock) was used to stain the filters (15 min) for enumeration by epifluorescence microscopy. We note that because SYBR Green I binds to dsDNA more efficiently than ssDNA and RNA, the total virus count may be underestimated. After staining and drying, filters were mounted on slides with an anti-fading solution (50% glycerol, 50% phosphate buffered saline, 0.1% p-phenylenediamine). Background fluoresence in groundwater samples was minimal and did not interfere with enumeration (Figure S2). At least 10 fields or 200 particles were enumerated per filter (Patel et al., 2007). For each sample, duplicate field replicates were enumerated for cells and viruses. Blanks of TE buffer were routinely checked to confirm the lack of viral and microbial contaminants on filters.

Groundwater dissolved oxygen (DO) concentrations were measured in situ deploying multi-parameter sondes (YSI Inc., OH) into the well. Groundwater samples were directly filtered (0.45 μm PTFE) into glass vials for analysis of DOC/DIC while samples for anion analysis were filtered directly into HDPE vials. Vials were capped, leaving no headspace, and stored at 4°C prior to analyses (Williams et al., 2011). Aqueous anions (sulfate and nitrate) were measured by ion chromatography (ICS-2100 equipped with AS18 column, Dionex, CA; Kantor et al., 2013). DOC/DIC was measured by combustion catalytic oxidation and NDIR method using a Shimadzu Total Organic Carbon Analyzer (TOC-VCSH; Shimadzu, Corp.). Measurements of DOC will also include cell and viral biomass due to the cutoff used for defining the aqueous fraction (<0.45 μm), however cell and viral biomass do not make a significant portion of any DOC measurement in this study. Using upper limits of 149 fg of C per cell (Vrede et al., 2002) and 10⁶ cells mL⁻¹, no more than 0.15 mg L⁻¹ can come from cellular biomass, which constitutes only a minor fraction of the DOC measured in this study. Viruses, being orders of magnitude smaller than cells, are a negligible component of total measured DOC.

Spatial interpolation (ArcGIS, Desktop Release 10.1, Environmental Systems Research Institute, Redlands, CA) was used to geographically represent the distribution of viruses and cells as well as geochemical parameters (DOC, DIC, sulfate, dissolved Fe, dissolved Mn, nitrate, and pH) across the alluvial aquifer (Nolan and Weber, 2015). Spline interpolation method (from the ArcGIS Spatial Analyst extension) was selected from among three interpolation methods (kriging, inverse distance weighting, spline) as it resulted in the lowest residual error (Akkala et al., 2010).

Correlation analyses and statistical comparisons were conducted in GraphPad Prism 5.0.3 (GraphPad Software). Significance level was defined at p < 0.05. Analyses involving cell and virus abundances were conducted on log-transformed values. All correlations between all measured parameters were calculated by the Spearman rank correlation method. Statistical comparisons between parameters within the major plume region and outside were conducted by t-test.

Groundwater virus abundance ranged from 8.0 × 10⁴ to 1.0 × 10⁶ viruses mL⁻¹ (Figures 1A,D) and exceeded cell abundance (range 6.0 × 10⁴ to 6.1 × 10⁵ cells ml⁻¹; Figures 1B,E) in all 20 monitoring wells. These total abundance values of viruses enumerated in groundwater were similar to results obtained from other groundwater sites (Kyle et al., 2008; Roudnew et al., 2012). Virus abundance in this shallow aquifer was strongly correlated to cell abundance (ρ = 0.73, p < 0.001) (Figure 2C) and is consistent with prior studies comparing virus and cell abundance in aquatic and sedimentary environments including lakes (Maranger and Bird, 1995; Bettarel et al., 2006; de Araújo and Godinho, 2009; Barros et al., 2010), marine waters (Alonso et al., 2001; Pereira et al., 2009), marine surface sediments (Danovaro and Serresi, 2000), marine subsurface sediment (Bird et al., 2001; Engelhardt et al., 2014), and deep granitic groundwater (Kyle et al., 2008). The virus-to-microbial cell ratio (VMR) in this shallow aquifer ranged from 1.1 to 8.1 and averaged 3.0 ± 1.6 (mean ± S.D., n = 20; Figures 1C,F). VMR range observed in groundwater collected from the Rifle aquifer is consistent with another shallow aquifer ranging from 0.4 to 6.1 (Roudnew et al., 2012) but is slightly less than has been observed in deep aquifers (average VMR of 12; Kyle et al., 2008). However, the VMR was notably lower than was measured in water collected from the Colorado River (VMR = 33; Table S1). It should be noted that the VMR in the Colorado River reported in this study is consistent with results from other river systems (Mathias et al., 1995; Jiao et al., 2006; Luef et al., 2007). This result is not surprising as total virus abundance in the river water sample exceeds virus abundance in groundwater. The lower VMR in groundwater relative to the river water sample could be a result of adsorption of viruses to the aquifer alluvium. Free (planktonic) viruses in the groundwater were enumerated in this study so any viruses produced that were adsorbed onto the aquifer alluvium (including clays and reactive minerals) would have been excluded resulting in a lower total virus abundance in the groundwater. The adsorption of viruses to minerals such as clays and iron oxides (Hewson and Fuhrman, 2003; You et al., 2005; Kernegger et al., 2009; Nieto-Juarez and Kohn, 2013) is recognized to reduce planktonic virus abundance. Previous studies have found a correlation between viral abundance and microbial activity in marine waters (Corinaldesi et al., 2003). In subsurface sediments, stimulation of microbial activity was found to result in an increase in VMR (Pan et al., 2014). Thus, the differences in viral abundance/VMR may indicate differences in microbial activity between groundwater and surface water. Together these are plausible reasons that may explain the difference in the abundance of viruses between the river water and groundwater samples.

A correlation between microbial cell abundance, virus abundance, and groundwater geochemistry was observed within the aquifer. Microbial cell abundance in groundwater was significantly higher (Unpaired t-test, p < 0.01) within the uranium plume (2.0 × 10⁵ ± 1.4 × 10⁵ cells mL⁻¹, mean ± S.D., n = 11) compared to cell abundance in groundwater outside of the plume (8.6 × 10⁴ ± 3.1 × 10⁴ cells mL⁻¹, mean ± S.D., n = 9; Table 1, Figures 1B,E). Cell abundance had a strong positive correlation to groundwater DOC (Spearman's ρ = 0.54, p < 0.05), U (Spearman's ρ = 0.51, p < 0.05), and sulfate concentrations (Spearman's ρ = 0.47, p < 0.05). However, it should be noted that groundwater DOC concentrations were also higher within the uranium plume (3.45 ± 0.88 mg L⁻¹; mean ± S.D., n = 11), relative to the concentrations measured outside (2.36 ± 0.60 mg L⁻¹; mean ± S.D., n = 9) of the uranium plume (Table 1, Figure 3A). Elevated concentrations of DOC may arise from buried organic carbon identified at the site which have been demonstrated to be responsible for persistence of the U plume (Campbell et al., 2012; Janot et al., 2015; Boye et al., 2017). Groundwater DOC and U concentrations were also correlated with virus abundance (Spearman's ρ = 0.46, p < 0.05 and ρ = 0.46, p < 0.05, respectively; Figures 2A,B, Table 2). Similar to the distribution of cells in the aquifer, virus abundance in the uranium plume region was statistically higher (Unpaired t-test, p < 0.01), 5.2 × 10⁵ ± 2.4 × 10⁵ viruses mL⁻¹ (mean ± S.D., n = 11), relative to the abundance of viruses outside of the region, 2.5 × 10⁵ ± 1.4 × 10⁵ viruses mL⁻¹ (mean ± S.D., n = 9; Table 1, Figures 1A,D). While virus and cell abundance in groundwater were positively correlated to groundwater DOC concentration, VMR did not exhibit a statistically significant correlation to any geochemical factor (Table 2, Figures 2A,B).

While there is the potential for hyporheic intrusion of DOC from the Colorado River, elevated DOC concentrations were not observed in samples collected from wells located near the river, nor were other proxies for river water incursion into the aquifer, such as low electrical conductivity (data not shown). While we did not test lability of the DOC, correlations with cell (Spearman's ρ = 0.46, p < 0.05) and virus abundance (Spearman's ρ = 0.54, p < 0.05) strongly suggest that DOC was sufficiently bioavailable to stimulate microbial activity. Microbial activity in groundwater is often stimulated by inputs of DOC (Baker et al., 2000; Sobczak and Findlay, 2002; Findlay et al., 2003; Foulquier et al., 2011; Li et al., 2012). The presence of bioavailable DOC and available electron acceptors may thus provide sufficient energy for stimulation of microbial respiration. Because viruses are reliant on metabolically active hosts for replication, microbial host energy availability favors the production of viruses. Host cell metabolic activity and growth rate has been directly demonstrated to increase virus adsorption rate and decrease the period of time between viral infection and lysis of the host for lytic viruses (Hadas et al., 1997). Lysogenic bacteriophage have also been demonstrated to respond to host cell metabolic activity; control of the lytic and lysogenic pathway is controlled by levels of cAMP, with high energy conditions favoring lysis (Hong et al., 1971; Rolfe et al., 1973). In addition, chronic infections, in which viruses are released without lysis of the host, also produce greater numbers of viruses under higher energy conditions (Brown and Dowell, 1968). This was recently demonstrated in a series of alluvial aquifer sediment microcosms where additions of an energy source, acetate, and electron acceptor, nitrate, not only resulted in the oxidation of organic carbon, but also significant virus production (Pan et al., 2014). This result is consistent with prior studies conducted in surface aquatic environments in which correlation between virus abundance and DOC has been observed (Laybourn-Parry et al., 2001, 2013; Auguet et al., 2005; Holmfeldt et al., 2010; Säwström and Pollard, 2012). Inputs of organic carbon increase microbial activity (Peter et al., 2012), and organic-rich regions are also inferred to have elevated microbial activities (Campbell et al., 2012). As a result elevated microbial activity would result in enhanced virus production and hence higher virus abundance within the plume. Viral production and expression of virus-related genes following acetate biostimulation has been demonstrated previously at the Rifle site (Holmes et al., 2015). Given that stimulated microbial activity will result in the production of viruses (Pan et al., 2014), the elevated abundance of viruses in the plume correlated with organic carbon suggests that there is ongoing microbial activity and virus production in the aquifer.

The consumption of organic carbon in this aquifer has been linked to the reduction of molecular oxygen, nitrate, iron, manganese, uranium, and sulfate as well as fermentation reactions (Wrighton et al., 2012; Anantharaman et al., 2016). While suboxic conditions (DO < 1 mg L⁻¹) predominated throughout most of the plume, oxic conditions were identified in two wells within the center of the plume region: LQ105 (4.3 mg L⁻¹) and U01 (2.6 mg L⁻¹) (Table S2). Dissolved Fe and Mn concentrations suggested the presence of reduced Fe (Fe(II)) and Mn (Mn(II)) and were also substantially lower within these two wells (Table S2, Figure S3). Thus lower Fe and Mn concentrations may be due to oxidative precipitation of Fe or Mn oxide minerals or lack of metal reduction due to oxic conditions in the groundwater. The highest dissolved Fe and Mn concentrations were found in wells located along the central portion of the site closest to the Colorado River (744, 310, 743, JB01, LQ107, 304) (Figure S3). Groundwater U concentrations ranged from 26.5 μg L⁻¹ to 7.4 mg L⁻¹, largely localized to the center of the site (Figure 3B). Nitrate was also low to undetectable in most of the wells throughout the floodplain with the exception of the two oxic monitoring wells LQ105 and U01 (5.89 mg L⁻¹ and 8.25 mg L⁻¹, respectively) (Table S2, Figure 3C). This may reflect operative nitrification processes or lack of denitrification due to oxic conditions in these wells. These geochemical conditions indicate a substantially different redox environment, potentially due to intrusion of dissolved oxygen or nitrate at the capillary fringe (Williams and Oostrom, 2000). The oxic monitoring wells U01 and LQ105 also contained some of the highest cell and virus abundances in the major plume region (Figure 1). This may be expected because O₂ mediated respiration is expected to support greater cell abundance and hence greater virus abundance. Sulfate concentrations averaged 630.2 ± 134.5 mg L⁻¹ (mean ± S.D., n = 11) within the plume and 451.5 ± 125.8 mg L⁻¹ outside of the contaminant plume (mean ± S.D., n = 9) (Unpaired t-test P < 0.05; Table 1, Figure 3D). In addition, DIC was also higher within the plume (69.8 ± 9.9 mg L⁻¹, mean ± S.D., n = 11) compared to outside (56.9 ± 12.5 mg L⁻¹, mean ± S.D., n = 9) (Table 1, Figure 3E). Sulfate could also serve as a potential electron acceptor and has been implicated in the generation and precipitation of reduced sulfur phases such as framboidal pyrite, mackinwite, and greigite (Qafoku et al., 2009; Janot et al., 2015) within the aquifer. Elevated virus abundance in the plume region is consistent with prior results demonstrating that bacterial sulfate reduction rates were correlated with viral abundance and distribution in estuarine sediments (Middelboe et al., 2003). As such microbial metabolisms could thus be supported by the dynamic changes in redox conditions that are associated with the influx of oxidants into a carbon-rich reduced system.

Here we demonstrate that viruses, cells, and DOC are enriched in the major plume region of the Rifle aquifer. The widespread distribution of viruses at the Rifle aquifer is consistent with prior suggestions that viruses may play a potential role in influencing biogeochemical cycling at the site (Wrighton et al., 2014; Holmes et al., 2015). Interactions between organic carbon, cells, and viruses may be important in riparian aquifers across the upper Colorado River basin where buried organic matter plays an important role in mediating biogeochemical cycles and metal/radionuclide sequestration. Organic carbon availability can promote microbial activity (Baker et al., 2000; Sobczak and Findlay, 2002; Findlay et al., 2003; Foulquier et al., 2011; Li et al., 2012) and, in turn, virus production, which is reflected by elevated virus abundances. Thus, organic rich sediments may potentially represent regions of increased viral activity as a response to higher microbial metabolic activity.

Within the Rifle aquifer, virus mediated cell lysis has been suggested to contribute biologically available organic carbon or to suppress certain taxa responsible for biogeochemically important reactions at the site (Wrighton et al., 2014; Holmes et al., 2015). In addition to lysis, lysogeny is another possible life cycle. Recent proposed models of lysogeny have suggested that at high and low host cell densities lysogenic life cycles may be favored, however with the cell densities encountered in this study (10⁵–10⁶ cells/mL), lytic kill-the-winner dynamics are suggested to be favored (Knowles et al., 2016). Lysis of active members of the microbial community and liberation of organic carbon can both influence biogeochemical cycling. Together these impacts have biogeochemical implications for the long-persisting groundwater U plume located in the Rifle aquifer and other similar aquifers within the upper Colorado River basin. The increase in bioavailable carbon would accelerate cell turnover rates and organic carbon liberation (Middelboe et al., 1996; Noble and Fuhrman, 1999; Middelboe and Lyck, 2002; Eissler et al., 2003), creating labile particulate and dissolved organic carbon thereby providing a source of biologically available carbon (Xu et al., 2013, 2014) subsequently controlling the community from the bottom up. The liberated carbon can increase respiration rates of heterotrophic bacteria (Middelboe and Lyck, 2002), which may include U-reducing bacteria responsible for the precipitation and immobilization of U(IV) (Anderson et al., 2003; Chang et al., 2005; Williams et al., 2011). Alternatively, viruses may also depress rates of biogeochemical transformations by infecting and lysing the organisms responsible for important organic carbon dependent subsurface processes such as denitrification (Burt et al., 1999) or metal reduction (Holmes et al., 2015). Further research is necessary to explore the role that viruses play mediating microbial processes underpinning reactions responsible for the fate and transport of metals/radionuclides impacting groundwater quality.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.