Dissolved Organic Matter in Continental Hydro-Geothermal Systems: Insights from Two Hot Springs of the East African Rift Valley

: Little is known about the quantity and quality of dissolved organic matter (DOM) in waters from continental geothermal systems, with only a few reports available from the Yellowstone US National Park. In this study, we explored the chemodiversity of DOM in water samples collected from two geothermal hot springs from the Kenyan East African Rift Valley, a region extremely rich in fumaroles, geysers, and spouting springs, located in close proximity to volcanic lakes. The DOM characterization included in-depth assessments performed by negative electrospray ionization Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR-MS). Reduced, saturated and little aromatic DOM compounds were dominant in the hot spring waters collected from either the Ol Njorowa gorge (ON) or the south shore of the soda-saline Lake Elementaita (ELM). Oxygen-poor and sulfur-bearing DOM molecules prevailed in ON, probably reﬂecting abiotic sulfurization from sulﬁde-rich geoﬂuids. Nitrogen-bearing aliphatic and protein-like molecules were abundant in ELM, possibly perfusing through the organic-rich sediments of the adjacent Lake Elementaita. Notably, the heat-altered DOM of ancient autochthonous derivation could represent an overlooked source of aliphatic organic carbon for connected lentic environments, with a potential direct impact on nutrient cycling in lakes that receive geothermal water inputs.


Introduction
Hydrothermal systems are considered active biogeochemical reactors with unique microbial communities and firm candidates for the origin of life [1]. Recent investigations of the carbon cycle in aquatic ecosystems have focused on the role of hydrothermal systems in the synthesis, processing and transformation of dissolved organic matter (DOM), the main carbon and energy source for the abundant heterotrophic fraction of the aquatic microbial community. From a chemical perspective, DOM is a highly complex mixture of thousands of organic molecules which strictly interact with underwater light, selectively absorbing the radiance spectra and minerals through absorption-desorption processes [2,3]. The structural and compositional diversity is known to reflect the DOM origin, persistence and bioavailability for the microbiota [4].
Key literature studies have looked at how the marine DOM is reworked and degraded when recirculating through sub-aquatic hydrothermal vents at oceanic ridges [5], in deep water volcanic mud [6], and in shallow hydrothermal systems [7]. In geothermal fluids under high temperatures, it was reported that oxygen-rich DOM molecules can be selectively removed with the loss of CO 2 , carboxyl and hydroxyl groups [8]. Notably, continental geothermal environments were overlooked and, to our best knowledge, a detailed DOM description is available only at Yellowstone Hot Springs [9].
The aims of this study were to (i) provide new data about the chemodiversity of DOM in continental hot springs from a very active geological region, and (ii) compare the DOM properties between continental and marine hydrothermal systems.
We targeted the Kenyan central Rift Valley, a region with a relevant volcanic activity [10], where dozens of hot springs [11] border highly productive saline-alkaline lakes [12]. The DOM properties were analyzed by negative electrospray ionization Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) for a molecular-level description of the chemodiversity of DOM extracted through a solid-phase extraction procedure (SPE-DOM).

Geological Setting and Sampling Sites
The sampling sites were located in proximity to Lake Naivasha (East African Rift Valley, Kenya). The target area has a high geothermal gradient (>200 • C per km) associated with recent volcanic activity in the area [10]. The groundwater is generally warm (>25 • C), alkaline, and rich in Na + , K + , CO 3 2− , with a moderately high concentration of F-anions, which prevent its direct use as a source of drinking water [13,14]. The geology is characterized by comendites, basalts, trachytes and pyroclastics of Pleistocene age, overlain by Holocene volcanites. Soils are moderately alkaline, formed by lacustrine and fluvial sediments [15]. The bedrock is dissected by numerous faults and fractures trending NW-SE, N-S, and E-W, thereby enhancing the permeability of the basin by allowing groundwater to flow through the fractured network [16]. Water samples (approx. 0.5 L at each site) were collected from two hot springs located at the Ol Njorowa gorge (ON, 0 • 53 48.62" S; 36 • 19 8.60" E, 1860 m a.s.l.) within the Olkaria volcanic high-temperature geothermal field (the largest African geothermal energy field), and at the southern shore of saline-soda Lake Elementaita (ELM, 0 • 28 22.42" S, 36 • 15 29.32" E, 1776 m a.s.l.). The sample ON was collected from a waterfall flowing over bedrock with patches of a green biofilm. The water temperature was >50 • C (i.e., aquifer temperatures are higher than 150 • C within the geothermal field [16]). The sample ELM was collected from a shallow pool (1 m × 1 m; depth = 0.2 m) on the lakeshore. Water temperature was 36.6 • C and the residence time in the pool was of a few minutes only, because it drained rapidly (approx. 0.3 l/s) towards the lake.
For comparative purposes, an additional four samples were collected from the mixolimnion (from the surface down to 3 m depth) of the Lake Sonachi (SO, 0 • 46 57.68" S; 36 • 16" E, 1891 m a.s.l.), a small (0.18 km 2 ) meromictic saline-alkaline lake with high DOM concentrations [17]. The Lake Sonachi is located in between the two hot springs (34 km from ELM, 15 km from ON) ( Figure 1).

Sample Preparation
Filtered samples (300 mL onto combusted GF/F Whatman filters) were acidified to pH 2 (by addition of HCl) and stored in pre-combusted glass bottles for all DOM analyses. Non-acidified aliquots (100 mL) were stored in polyethylene bottles and used for geochemical characterization. Dissolved gases (i.e., CO2 and CH4) were sampled in a pre-evacuated 250 mL glass flask, equipped with a Teflon stopcock [18].

Inorganic Solutes and Gas Analysis
The ion chromatography was used to measure the concentrations of Cl − , SO4 2− , F − (761 IC-Metrohm AG ® ), Na + , and Ca 2+ (861 Advanced Compact IC-Metrohm AG ® ). Trace elements (Mn, Fe, Li and As) were assessed using the PerkinElmer ® Optima 8000 ICP-OES. Elemental S 2− was assessed as SO4 2− after oxidation with H2O2 of CdS resulting from the reaction of ΣS 2− with a Cd-NH3 solution [19]. Acidimetric titration (HCl 0.01 N) was used to estimate the concentrations of HCO3 − and CO3 2− . Dissolved CO2 in the headspace of the sampling flask was measured with a 15A Shimadzu ® gas chromatograph, equipped with a thermal conductivity detector. CH4 was analyzed with a 14A Shimadzu ® gas chromatograph equipped with a flame ionization detector [18].

Quali-Quantitative Characterization of Dissolved Organic Matter
Dissolved organic carbon (DOC) in filtered and acidified samples (triplicates) was measured by oxidative combustion and infrared analysis (TOC-VCSH analyzer, Shimadzu®, Kyoto, Japan). Considering the instrumental detection limit of 0.3 ± 0.1 ppm, two calibration curves with EDTA standard solutions were used (from 0 to 4 ppm for ELM and ON; from 0 to 20 ppm with 1:10 dilution factor for SO).

Sample Preparation
Filtered samples (300 mL onto combusted GF/F Whatman filters) were acidified to pH 2 (by addition of HCl) and stored in pre-combusted glass bottles for all DOM analyses. Non-acidified aliquots (100 mL) were stored in polyethylene bottles and used for geochemical characterization. Dissolved gases (i.e., CO 2 and CH 4 ) were sampled in a pre-evacuated 250 mL glass flask, equipped with a Teflon stopcock [18].

Inorganic Solutes and Gas Analysis
The ion chromatography was used to measure the concentrations of Cl − , SO 4 2− , Dissolved CO 2 in the headspace of the sampling flask was measured with a 15A Shimadzu ® gas chromatograph, equipped with a thermal conductivity detector. CH 4 was analyzed with a 14 A Shimadzu ® gas chromatograph equipped with a flame ionization detector [18].

Quali-Quantitative Characterization of Dissolved Organic Matter
Dissolved organic carbon (DOC) in filtered and acidified samples (triplicates) was measured by oxidative combustion and infrared analysis (TOC-VCSH analyzer, Shimadzu®, Kyoto, Japan). Considering the instrumental detection limit of 0.3 ± 0.1 ppm, two calibration curves with EDTA standard solutions were used (from 0 to 4 ppm for ELM and ON; from 0 to 20 ppm with 1:10 dilution factor for SO). The DOM quality was assessed by the Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR-MS). The two hot spring samples (200 mL filtered and acidified) were processed through 100 mg styrene-divinyl-polymere (PPL) solid-phase cartridges (Agilent ® , Waldbronn, Germany). The protocol by Dittmar et al. [20] was adopted to concentrate, desalt, and elute the samples and for preconditioning and drying the cartridges. A blank was treated as a sample to remove the interference of cartridges. The SPE-DOM was eluted with 5 mL methanol and kept frozen until analysis [21]. The solid phase extracted DOM (SPE-DOM) extracts were diluted 1:1 (v/v) with ultrapure water and analyzed with a FT-ICR mass spectrometer equipped with a dynamically harmonized analyzer cell (Solari XR, Bruker Daltonics Inc. ® , Billerica, MA, USA) and a 12 T refrigerated actively shielded superconducting magnet instrument (Bruker Biospin ® , Wissembourg, France) located at the ProVIS Centre for Chemical Microscopy (Helmholtz Centre for Environmental Research). Samples were infused by an autosampler at a flow rate of 4 µL/min. For each spectrum, 256 scans (ion accumulation time: 15 ms) were co-added in the mass range 150-3000 m/z with a 4 Mword time domain. Mass spectra were internally calibrated with a list of peaks (m/z 250-550, n > 110) commonly present in natural organic matter; the mass accuracy after internal linear calibration was better than 0.1 ppm. Peaks were considered when the signal/noise (S/N) ratio was >4.
Formulas were assigned to peaks in the mass range 150-700 m/z, allowing for elemental compositions C 1-60 H 1-120 O 1-40 N 0-2 S 0-1 with an error range of ±0.5 ppm, as suggested by Lechtenfeld et al. (2014) [22]. Relative peak intensities were calculated based on the summed intensities of all the assigned peaks in each sample. A relative signal intensity of 0.01% was chosen as the dataset detection limit to compensate for the variability in sensitivity between different measurements. A molecular mass peak formula is referred to as "molecule" in this article although one molecular formula can represent several different structural isomers. Only formulas falling within the range −10 ≤ DBE -O ≤ +10 (DBE: double bound equivalents) were considered for further data evaluation, as discussed elsewhere [23]. Following the molecular formula identification process, each sample was assessed according to the following descriptors: number of detected molecules (richness), molecular mass, H/C, O/C, N/C, S/C, DBE, modified aromaticity index (AImod) [24], nominal oxidation state of carbon (NOSC) [25]. Except for richness, all the descriptors were calculated for each component and as intensity weighted averages. The ratios H/C, DBE and AImod provided information about the degree of saturation of a molecule (i.e., the number of C-C double bonds) whereas the O/C ratio and NOSC were used as descriptors of the (average) carbon oxidation state [25]. According to the values of the different DOM indicators, DOM molecular classes were divided into: carboxyl-rich alicyclic like molecules (CRAM-like), Aliphatics, Aromatics and Condensed Aromatics (CA), saturated molecules, and nitrogen-rich aliphatic compounds (Table 1). CRAM-like, aliphatic and aromatic molecules were subdivided into O-poor and O-rich molecules if the O/C was <0.5 or ≥0.5, respectively. Table 1. Molecular categories adopted in this study for describing the molecular diversity of the solid phase extracted DOM (SPE-DOM).

Molecular Class Compositional Constrains Source
Carboxyl-rich alicyclic like molecules (CRAM-like)

Geochemical Context and DOC Concentration
Both springs had pH > 8, reflecting moderate alkalinity, and a high electrical conductivity (EC = 2.9 mS/cm in ON, EC = 3.2 mS/cm in ELM). Waters from ON were richer in sulfate, manganese, iron, lithium and arsenic compared to those from ELM, in which high concentrations of CO 2 and CH 4 were also detected ( Table 2). The concentrations of dissolved organic carbon (DOC) were 1.67 mg C/L (ON) and 0.97 mg C/L (ELM), far lower than those measured at Lake Sonachi (97.5 ± 6.8 mg C/L).

SPE-DOM
Upon the assignment of formulae, FT-ICR-MS detected 3608, 3892 and 4054 ± 281 SPE-DOM molecules at ELM, ON, and SO, respectively (Supplementary Table S1). A total of 2058 molecules were shared between the three waters (ON ∩ ELM ∩ SO). About 16-17% of the SPE-DOM molecules were exclusive to one of two hot springs. These distinctive compounds contributed in both cases to about 9% of the total signal intensity. About 13-14% of the molecules were detected in the two hot springs, but were missing in SO samples (ON ∩ ELM). The most dissimilar samples were SO and ELM, which shared only 7% of molecules (Table 3). In terms of relative abundance, the most abundant compositional group in ELM comprised O-poor CRAM-like molecules (35.2%), followed by O-poor aliphatic ones (23%). O-rich aromatics (0.5%) and CA (0.6%) compounds were less abundant. An outstanding feature of SPE-DOM from ELM was the high occurrence of N-rich aliphatic compounds (12% of relative abundance) in relation to ON and SO (1.1% and 2.6%, respectively) ( Figure 2a). The presence of several O-poor CHOS (1 < H/C < 1.5) molecules in ON and many O-poor aliphatic (H/C > 1.5) CHNO molecules was also noted in ELM (Figure 2g). The SPE-DOM from ELM was also visibly more reduced (NOCSw = −0.6; O/C = 0.38), more highly saturated (DBEw = 6.82 and H/Cw = 1.42) and less aromatic (AImod = 0.2) than that from SO. At ON, the relative abundance of aromatics (5.4%) and CA (1.4%) was relevant while the O-rich aliphatics were low (0.6%) (Figure 2b). Overall, SPE-DOM from ON was slightly more reduced (NOCSw = −0.46; O/C = 0.38), more aromatic (AImod = 0.277), and more unsaturated (DBEw = 8.09; H/C = 1.27) than SO (Table 4). Among heteroatomic molecules, the relative abundance of CHNO molecules at ELM (23.4%) was higher than that of SO (18.4%) and ON (10.7%). The relative abundance of S-containing molecules at ON (15.1%) was remarkably greater than at SO (3.5%) and ELM (3.7%) (Figure 2j-l). Furthermore, the relative signal intensity of these S-bearing molecules tended to increase with the increasing number of hydrogen atoms, suggesting the presence of sulfur at a reduced oxidation state in the ON sample (Supplementary Figure S1). At ELM, the exclusive SPE-DOM molecules showed a high relative intensity of heteroatoms (CHNO, 62%; CHOS, 11%; CHNOS, 4%) (Figure 2j). Almost all the ELM exclusive molecules fell within the category of aliphatics (70.4%); meanwhile, the contribution of CRAM-like, aromatic, and CA molecules, dropped to 4%, 1.4%, and 0.4%, respectively. At ON, the exclusive SPE-DOM was strongly enriched in S-bearing molecules (CHOS, 75.5% and CHNOS, 3.8%, Figure 2k). In a compositional context, the frequency of CRAM-like and O-poor aliphatics strongly decreased, whereas the frequency of CA strikingly increased (Table 4). Table 4. Molecular characterization of SPE-DOM from the two hot springs (ON, ELM) and Lake Sonachi (SO). Molecular class criteria and SPE-DOM descriptors are defined in Table 1. Percentage values are weighted against relative abundance. All the values are signal intensity weighted. CA: condensed and aromatic.

Discussion
The chemodiversity of DOM in geothermal fluids is likely to provide new insights into prebiotic processes occurring in fluid-dependent aquatic environments [27]. Most of the current knowledge is grounded in evidence gathered from deep marine hot springs, while continental hot springs/fumaroles have been largely disregarded [9].
The DOC concentrations estimated at ELM and ON were within the same range of those reported from Yellowstone [28], suggesting that the two samples were not in contact with the lixiviates derived from terrestrial vegetation and soils, nor with the waters of Lake Elementaita that showed extremely high DOC concentrations [29].
Mass spectrometry evidenced that the SPE-DOM from ELM and ON was strongly reduced with an O/Cw ratio in the range of that reported for the Yellowstone hot springs (O/C < 0.4) [9], which was lower than the values typically reported for freshwaters (O/C > 0.45) [30,31]. On the one hand, these results are in agreement with the hypothesis that heat should induce DOM dehydration and the loss of CO 2 , carboxyl, and hydroxyl groups [8]. On the other hand, a different pattern was reported for SPE-DOM sampled at deep marine hot vents with an O/Cw ratio higher than 0.4 (Figure 3 and references therein).
Mass spectrometry evidenced that the SPE-DOM from ELM and ON was strongly reduced with an O/Cw ratio in the range of that reported for the Yellowstone hot springs (O/C < 0.4) [9], which was lower than the values typically reported for freshwaters (O/C > 0.45) [30,31]. On the one hand, these results are in agreement with the hypothesis that heat should induce DOM dehydration and the loss of CO2, carboxyl, and hydroxyl groups [8]. On the other hand, a different pattern was reported for SPE-DOM sampled at deep marine hot vents with an O/Cw ratio higher than 0.4 (Figure 3 and references therein).
The interaction between the deep (recalcitrant) marine SPE-DOM within hydrothermal systems did not appear to alter substantially the O/C ratio. Yet, an analysis of the unique SPE-DOM described in shallow marine vents [7] revealed that their molecules were clearly more reduced with an O/Cw ratio almost identical to that measured in ELM and ON, or that estimated in marine SPE-DOM exposed to experimental heat alteration [5,8]. Based on the few case studies reported, a generalized tendency towards a more reduced SPE-DOM seems to characterize continental hot springs in opposition to offshore deep hydrothermal vents.
The interaction between the deep (recalcitrant) marine SPE-DOM within hydrothermal systems did not appear to alter substantially the O/C ratio. Yet, an analysis of the unique SPE-DOM described in shallow marine vents [7] revealed that their molecules were clearly more reduced with an O/Cw ratio almost identical to that measured in ELM and ON, or that estimated in marine SPE-DOM exposed to experimental heat alteration [5,8]. Based on the few case studies reported, a generalized tendency towards a more reduced SPE-DOM seems to characterize continental hot springs in opposition to offshore deep hydrothermal vents.
Heat is expected to alter significantly the saturation degree (H/C) of SPE-DOM, along with progressive dehydration, the accumulation of aromatic rings, and a reduction in methyl (CH 3 − ) groups [8]. Heat-altered SPE-DOM should be more unsaturated (i.e., lower H/C ratio and high DBE). However, changes in the H/Cw ratio are likely more subtle than those in O/Cw (Figure 3). Previous experiments have evidenced little (or almost nonexistent) change towards more unsaturated molecules [5,8] (Figure 3). In those few reports that provide specific information about the H/Cw ratio of these molecules, the change patterns were unclear. For instance, the decrease in the H/Cw ratio is evident exclusively in SPE-DOM molecules found at the Dominica vent (DOMI) only [7]. Higher values were found in this study (either in ON or more noticeably in ELM) and in the SPE-DOM molecules reported in Iceland (ICEL) [7]. In the SPE-DOM from the Yellowstone hot springs, changes in the H/C ratio were likely inconsistent, but the magnitude was not described [9].
The H/Cw value of ELM is the highest reported so far for SPE-DOM in hydrothermal waters. The ELM hot spring is the only site, among those included in Figure 3, located in close proximity to a lake. Saline-alkaline soda lakes from the East African rift are characterized by very high primary production (mainly driven by Arthrospira spp. and other cyanobacteria) [32,33]. Their decay and settling induce the accumulation of large amounts of organic matter at the sediment-water interface [34]. Moreover, these lakes can undergo remarkable water level oscillations, and the nearby hot springs are frequently submerged. Cohen and Nielsen [35] reported that the first 12.5 m of lacustrine sediments, generated over the last 10,000 years, from Lake Elementaita consisted of mud with an organic content of 10-20%. Consequently, ELM hot springs may perfuse through organic-rich sediments reintroducing buried organic carbon (mainly aliphatic) into the lacustrine system.
The presence of high amounts of dissolved CH 4 measured at ELM provides a clue to the presence of an active subsurface prokaryotic community. The dissolved CH 4 was one order of magnitude higher than that of CO 2 , and several orders of magnitude higher than the values typically reported in hot spring/fumarole fluids [36].
Moreover, it is noteworthy that N-bearing molecules represent almost 60% of those detected in ELM. Although little is known about the occurrence of N-rich molecules in hydrothermal fluids, N-rich molecules are frequent in experimentally heat-altered SPE-DOM [5,8]. High concentrations of amino acids have been reported in hydrothermal fluids from mid ocean ridges [37]. It cannot be excluded that aliphatic N-rich compounds are metabolites produced by interstitial microbiota [38] that thrive in organic-rich sediments within the lakebed. FT-ICR-MS analyses might have underestimated the abundance of aliphatic substances at the ELM hot spring. Raeke et al. [21] reported that PPL sorbents were less efficient in retaining O-poor (O/C < 0.4) and saturated (i.e., H/C > 1.5) molecules. Therefore, even the contribution of aliphatics in the ELM hot spring might have been underestimated.
At ON, the most prominent property of SPE-DOM was the remarkable abundance of sulfur-bearing molecules. The S/Cw ratio detected in this sample, estimated using all the molecules assigned as well as by using uniquely assigned S-bearing molecules, averaged 0.012 and 0.055, respectively. These values were similar to those reported for shallow hydrothermal vents characterized by strong evidence of abiotic sulfurization [7], suggesting that the ON hydrothermal system DOM might be affected by sulfurization together with dehydration. Likewise, the hydrothermal waters from the nearby Olkaria geothermal system are rich in sulfide (from 2 and 18 mg L −1 ) [16]. This information corroborates the evidence of H 2 S-rich geothermal fluids, together with high temperature and pressure, might facilitate DOM sulfurization.

Conclusions
The occurrence of hot springs across the eastern African rift valley provides the opportunity to improve our understanding on how DOM properties may change under high temperature and pressure. Our study revealed the distinctive compositional footprints of heat-modified solid phase-extracted DOM. The signature from two hot springs indicated the accumulation of small-sized O-poor DOM as a consequence of dehydration, the loss of CO 2 , and the degradation of carboxyl (COOH) and alcohol/hydroxyl (OH − ) groups. Moreover, we highlighted the potential impact of hydrothermal fluids in introducing heat-altered DOM of ancient autochthonous derivation in lakes with adjacent hot springs. A better understanding of the origin, age, and fate of such a disregarded DOM source will complement the current knowledge on carbon cycling in lakes from geothermal systems.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4441/12/12/3512/s1, Table S1: Complete dataset of FT-ICR-MS analyses. Figure S1: Ol Njorowa Gorge. Distribution of CHO (red dots) and CHOS (black dots) intensity peaks in respect to the number of hydrogen atoms. Distributions are for molecules containing 16 (panel a) and 20 (panel b) carbon atoms respectively. Plots show that the distribution of signal intensities of the CHOS molecules is biased towards larger numbers of hydrogen atoms indicating the presence of sulfur at a reduced oxidation state.