Unraveling biogeochemical phosphorus dynamics in hyperarid Mars‐analogue soils using stable oxygen isotopes in phosphate

With annual precipitation less than 20 mm and extreme UV intensity, the Atacama Desert in northern Chile has long been utilized as an analogue for recent Mars. In these hyperarid environments, water and biomass are extremely limited, and thus, it becomes difficult to generate a full picture of biogeochemical phosphate‐water dynamics. To address this problem, we sampled soils from five Atacama study sites and conducted three main analyses—stable oxygen isotopes in phosphate, enzyme pathway predictions, and cell culture experiments. We found that high sedimentation rates decrease the relative size of the organic phosphorus pool, which appears to hinder extremophiles. Phosphoenzyme and pathway prediction analyses imply that inorganic pyrophosphatase is the most likely catalytic agent to cycle P in these environments, and this process will rapidly overtake other P utilization strategies. In these soils, the biogenic δ18O signatures of the soil phosphate (δ18OPO4) can slowly overprint lithogenic δ18OPO4 values over a timescale of tens to hundreds of millions of years when annual precipitation is more than 10 mm. The δ18OPO4 of calcium‐bound phosphate minerals seems to preserve the δ18O signature of the water used for biogeochemical P cycling, pointing toward sporadic rainfall and gypsum hydration water as key moisture sources. Where precipitation is less than 2 mm, biological cycling is restricted and bedrock δ18OPO4 values are preserved. This study demonstrates the utility of δ18OPO4 values as indicative of biogeochemical cycling and hydrodynamics in an extremely dry Mars‐analogue environment.


today.
The Atacama Desert is one of the driest and oldest temperate deserts (Dunai, Lopez, & Juez-Larre, 2005;Hartley, Chong, Houston, & Mather, 2005) that reaches the extreme limit of life (Navarro-Gonzalez et al., 2003). This desert occupies around 105,000 km 2 of northern Chile, and has remained extremely dry for the past 150 Myr (Hartley et al., 2005), with annual precipitation at the hyperarid core less than 2 mm (Houston, 2006) for more than 10 Myr (Sun, Bao, Reich, & Hemming, 2018). Alongside aridity, extremes in UV irradiation are similar to high levels of UV delivered to exposed Martian surfaces, approximately 150 kWh/m 2 UV-A (315-400 nm) and 5 kWh/ m 2 UV-B (280-315 nm) (Cordero et al., 2018). UV irradiation causes organic degradation, resulting in an oligotrophic terrain, and the lack of precipitation reduces leaching and runoff, meaning phosphate minerals can be preserved. This location is therefore ideal to investigate both the dynamics of P cycling in relation to other key nutrients (C, N) in hostile environments, and to potentially identify long-lived biomarkers of life in an environment similar to that found on Mars.
The biogeochemical cycling of P is less well-studied than C, N, and S in modern terrestrial soils. This is because P usually remains in the form of orthophosphate (PO 4 3-, HPO 4 2-, or H 2 PO 4 -) and only has one stable isotope, 31 P. These facts render it relatively difficult to track P in biosystems and understand its reaction processes. Recently, however, methods have used phosphate-bound stable oxygen isotopes (hereafter noted as δ 18 O PO4 ) as a tracer for the detection of biological activity (Blake, Chang, & Lepland, 2010;Blake, O'Neil, & Surkov, 2005;Jaisi & Blake, 2014;Stout, Joshi, Kana, & Jaisi, 2014;Tamburini, Pfahler, von Sperber, Frossard, & Bernasconi, 2014), since phosphorus is normally associated with oxygen in both abiotic and organic forms.
Stable oxygen isotope values are reported in parts per thousand (‰) using the standard δ notation: where R sample is the 18  We further define an isotopic fractionation factor, α, as follows: where R product and R reactant are the 18 O/ 16 O ratios in the product and reactant, respectively. The isotopic enrichment factor, ε p-r is then defined as follows: The principle of this method for tracing P cycling using the δ 18 O PO4 signature relies on the relative stability of the P-O bond under abiotic processing. Abiotic P transformations (including phosphate precipitation, UV photolysis and decay of organophosphorus, and solubilization) do not introduce significant oxygen isotope fractionations in phosphate (Figure 1), resulting in an isotopic signature which is theoretically extremely long-lived (Dahms & Boyer, 1973;Jaisi, Blake, & Kukkadapu, 2010;Kolodny, Luz, & Navon, 1983;Tudge, 1960).
However, where biological uptake and processing of P occurs the P-O bond can be broken, enabling isotopic exchange of phosphate O with surrounding O sources (normally soil or cellular water). Hence, deviations in δ 18 O PO4 from the source mineral signature can potentially be utilized as a biosignature for life (Blake, Alt, & Martini, 2001).
where Eδ 18 O PO4 is the 18 O/ 16 O ratio of phosphate at equilibration with water catalyzed by inorganic pyrophosphatase, T is temperature in °C, and δ 18 O H2O is the δ 18 O of ambient water.
We can therefore assess the extent to which abiotic or biotic processes have impacted P by measuring the δ 18 O PO4 signature of soils.
To investigate deeper into the P cycle, we extract different forms of P using well-developed chemical extractions (Olsen, 1954) and analyze δ 18 O PO4 for individual pools of phosphate. By measuring the δ 18 O PO4 of each extract, we can identify the extent to which each pool has been subjected to biological utilization. In this study, we focused on four extracted pools of soil phosphate, operationally defined as (a) resin-extractable P, (b) microbial P, (c) NaOH-extractable Al/Fe oxide-bound P and organic P, and (d) HCl-extractable calcium-bound P (Ca-P).
In normal soil environments, resin-extractable P is turned over on timescales from seconds to minutes, oxide-bound P is recycled on weekly to monthly timescales, and Ca-bound P over years to millennia (Helfenstein et al., 2020;Helfenstein et al., 2018). The least bioavailable P pool is therefore normally Ca-bound P. In an alkaline environment like the Atacama Desert (Crits-Christoph et al., 2013;Makhalanyane et al., 2015;Shen, Zerkle, Stueken, & Claire, 2019) or on Mars (pH 7.2-8.2; (Hecht et al., 2009;Smith et al., 2009)), where resin-extractable P (bioavailable P) is high, it would tend to be sorbed by calcite (Reddy, Wetzel, & Kadlec, 2005) and incorporated through precipitation into Ca-bound P mineral phases (Kreuzeder, 2011), making it less bioavailable (Oburger, Jones, & Wenzel, 2011;Urrutia et al., 2014). However, where this Ca-bound P is a major source of P within a biological system it can be reversibly solubilized by the excretion of acids by specialized microbial communities, breaking down Ca-P and once again making it bioavailable. Alternatively, Ca-bound P can dissolve abiotically over an extremely prolonged period under alkaline conditions (Guidry & Mackenzie, 2003) such as those found in Atacama soils. In non-arid soil samples, the concentration of Ca-P is usually low and hard to detect (Amelung et al., 2018). However, the relative pool size increases significantly in desert environments (Angert, Weiner, Mazeh, & Sternberg, 2012;Helfenstein et al., 2018).
Where exchange is rapid, or P cycling long enough, biogenic P becomes the main component of the Ca-P pool; however, this can take thousands of years (Tamburini et al., 2012). Previous studies have shown that in low precipitation environments, the exchange rate between resin-extractable P and Ca-P is slowed and pedogenic P may persist in the Ca-P pool for hundreds of thousands of years . In the case of the Atacama Desert where precipitation is extremely scarce (<20 mm/year, an order of magnitude lower than the samples studied by Helfenstein et al., 2018), this turnover time may be expected to be further inhibited, or P cycling completely retarded, unless biological communities can access less conventional forms of water to support nutrient cycling.
For example, organisms under hyperarid conditions are able to utilize hydration water from gypsum minerals to supplement the extremely scarce rain water (Palacio, Azorin, Montserrat-Marti, & Ferrio, 2014).
This research aims to investigate the microbial P utilization in the Atacama Desert and to examine the extent to which biological processing can overprint a mineral-dominated pool, even under the harshest of conditions. Previous studies using δ 18 O PO4 in Marsanalogous or arid terrestrial environments all focused on relatively water-rich locations. These studies suggested that δ 18 O PO4 values are sensitive to levels of enzymatic activity and microbial metabolism wherever liquid water might be present (Blake et al., 2001;Blake et al., 2005;. To improve interpretation of δ 18 O PO4 analyses, we performed soil metagenomic analyses and cell culture experiments: metagenomic analyses can provide information about the existence and relative abundance of P-processing enzymes and pathways to support isotopic findings, and cell culture experiments can indicate how microbes grow in soils of atypical C:N:P ratios and how they respond to bioavailable P. Therefore, we combine this novel δ 18 O PO4 proxy for phosphorus cycling with enzyme pathway analyses and cell culture experiments to trace P dynamics in soils spanning an aridity gradient from hyperarid to arid environments.

| Study site and soil characterizations
The Atacama Desert is bounded by two high mountain ranges (the Coastal Cordillera and Altiplano/Andes Mountains) (Garreaud, Molina, & Farias, 2010;Garreaud, Vuille, Compagnucci, & Marengo, 2009;Houston & Hartley, 2003;Rech et al., 2010;Veblen, Young, & Orme, 2015), which prevent Pacific and Atlantic Ocean moisture from reaching the hyperarid core. Modern hyperaridity of the Atacama Desert is also a result of subtropical atmospheric subsidence (Takahashi & Battisti, 2007), coupled with a temperature inversion caused by the interaction of the constant strong Pacific anticyclone (Trewartha, 1961) and coastal cold upwelling of the Humboldt ocean current . For this study, five sites were sampled on a latitudinal gradient from 22°S to 28°S ( Figure 2). This sampling strategy was selected to provide an aridity gradient from hyperarid in the north to arid in the south (    (Schenk et al. 1999;Tapia et al., 2018).
Detailed sampling methods are described in Shen et al. (2019). In short, 3 pits were dug to 10-20 cm depth at each site, where ~ 200 g of material was removed for geochemical analyses. From each study site, one of the pits was further sampled for δ 18 O PO4 analyses (~500 g of sediment), DNA extraction (~50 g), and cell spreading experiments (~100 g) using sterile techniques. Soils of an additional TZ-5 pit that was surrounded by dead dried shrub, similar to all TZ-4 pits, were sampled for δ 18 O PO4 analyses. Samples were stored at 4°C to avoid any change in the microbial community structure and minimize microbial activity.
Mean annual precipitation and mean temperature (Table 1)  Regolith P is referred to total P (TP). Carbonate from each sampling pit was determined in triplicate using the decarbonation method during the preparation step for TOC and TON measurements.

| Sequential phosphate extraction
Sequential extractions were undertaken on each soil sample to isolate different P pools of potential interest for δ 18 O PO4 analysis and P concentrations (Hedley, Stewart, & Chauhan, 1982). Each bulk extraction started with large volumes of soil (~100 g) in an attempt to ensure enough P was obtained for isotope purification and analysis, following Tamburini et al. (2010). Alongside this bulk method, we extracted a subsample of the soil using the same soil:solution ratios for concentration analysis. The concentrations of phosphate from each of the four extractable pools were quantified on an Aquahem 250 analyzer (Thermo Fisher Scientific) using a molybdenum blue reaction (Murphy & Riley, 1962).
The stages of extraction followed Tamburini et al. (2018); briefly these were: • Resin-extractable P (resin-P), extracted from approximately 600 g of soil in 3.6 L of 18.2 μΩ ultrapure water with 3 pre-conditioned 125 mm x 125 mm anion-exchange resin strips (VWR International Ltd). It should be noted that as these soils were extremely dry, there was a possibility that a water-based extraction would promote some rapid microbial response and release of P captured in this extraction. This is an unavoidable laboratory effect, but was mitigated as best possible by shaking at 4°C to limit microbial activity within the extract .
• Microbial P, extracted from 100 g of fresh soil with 12.5 ml of hexanol in 1 L of 18.2 μΩ ultrapure water with one 125 mm x 125 mm pre-conditioned anion-exchange resin strip. This extraction captured the mixture of inorganic (resin-P as above) and microbial phosphate (Granger et al., 2018;Kouno, Tuchiya, & Ando, 1995;Mclaughlin, Alston, & Martin, 1986). Microbial P was calculated by subtracting resin-P from hexanol-extractable P.
• Oxide-bound and organic P pools were extracted using a NaOH/ ethylenediaminetetraacetic acid (EDTA) mix (6.5 mg/ml NaOH and 12.0 mg/ml EDTA), added to hexanol extracted samples.

| δ 18 O PO4 isotope analyses
As we aimed to identify a long-lived biosignature of life in the Atacama sediments, we focused our δ 18 O analyses on the most stable P pool, Ca-P. Phosphate in these extracts was converted to Ag 3 PO 4 using a slightly adapted purification protocol described by Tamburini et al (2010). Briefly, phosphate was precipitated as am-

| Metagenomic extraction, sequencing, and enzyme/pathway predictions
To understand the microbial pathways in P dynamics, we performed soil metagenomic analyses. All implementations for gene experiments were either filter-sterilized, autoclaved, or UV-irradiated to prevent any external contamination; water was molecular biology grade and nuclease-free. DNA of MES, PONR-2, Yungay, TZ-4, and TZ-5 was extracted with one negative control using the  (Caspi et al., 2016;Caspi et al., 2018).
Here, we focused only on data pertaining to biogeochemical P dynamics in the sampled soils, as it was beyond the scope of this study to provide a full metagenomic dataset. Enzymes and metabolic pathways related to microbial phosphate dynamics were selected from all records of predicted pathways. The abundance of phosphate-involved pathway copy numbers was plotted with OriginPro 2019 (OriginLab Corporation).

| Phosphate amendments and cell culture experiments
To explore the microbial growth in these extremely dry soils of atypical C:N:P ratios, we performed cell culture experiments and phosphate amendments. About 10 g of soil from each sampling site was amended with 4.5 ml sterilized 10% monosodium phosphate (NaH 2 PO 4 ) and 4.5 ml sterilized ultrapure water (as a control) and left for 4 days at room temperature (Shen et al., 2019). Soils were then refrigerated at 4°C. Duplicate amended soils and original soils without amendment were suspended in sterilized ultrapure water by an applicable dilution factor and spread on Luria-Bertani (LB) agar and plate count agar plates, and left at 21°C. Colonies were counted after 20 days of growth (Bagaley, 2006). Colony-forming units (CFUs) were calculated by multiplying the number of colonies formed on agar plates by the correspondent dilution factor, and a factor 1.45 to account for the addition of 4.5 ml solution to 10 g of soil. CFUs on phosphate amendments were further transformed into logarithmic scale, and normalized as the difference relative to water only amendments.

| Nutrients, mafic elements, and δ 18 O PO4
We saw no correlation between TOC and TP at the hyperarid core, and the TOC and TP of arid sites were both higher than hyperarid sites ( Figure 5a). The P/N ratio anticorrelated with TOC, with 4 of the 5 sites showing a distinctive relationship. However, site TZ-5 (the wettest site) displayed a different decreasing trend from the other four sites (Figure 5b). P/N significantly correlated (R 2 = 0.88*** and p = .000) with SiO 2 /Al 2 O 3 (used to approximate quartz to clay ratios (Broadhurst & Loring, 1970;Olivares et al. 2017;Xiao, Porter, An, Kumai, & Yoshikawa, 1995)) ( Figure 5c), which was generally taken as a proxy of sedimentation rates. Additionally, sites with lower levels of total P commonly had larger grain sizes (Table 1).
Mg and Fe were two common mafic elements, the richness of which indicated the abundance of igneous rock sources (Barker, 1978). MgO in the sediments of sampling sites ranged from 0.9 to 1.4% in hyperarid sites, and was higher in arid sites, from 2.2 to 3.3%.

| Distribution of different P pools
Compared to the hyperarid sites (MES, PONR-2, and Yungay, <2 mm/ year), arid sites (TZ-4 and TZ-5, 15-20 mm/year) had a greater pool size of microbial P and NaOH-P (including Al/Fe oxide-bound P and organic P), but a relatively smaller pool size of resin-P and HCl-P (usually calcium-bound P, hereafter noted as Ca-P) (Figure 4). Microbial P positively correlated with NaOH-P (R 2 = 0.88*, p = .022). In addition, sites with smaller microbial P and NaOH-P pools generally had larger grain sizes (Table 1). Plots of microbial P and total P clustered together within hyperarid and arid sites (Figure 6a). The δ 18 O PO4 values of soil Ca-P from the hyperarid sites (MES, PONR-2, and Yungay) varied between 7.7 and 13.8‰, and from arid sites  between 19.5 and 25.3‰ (Table 2). The contents of TOC, TON, and microbial P correlated with the δ 18 O PO4 of Ca-P (Figure 6b-d).

| Enzymes and pathways in microbial phosphate metabolisms
The overall structure of the relative abundance of phosphate path-    (Table 3).

| Microbial growth with phosphate amendments
Cultivable heterotrophic micro-organisms from the hyperarid sites and the drier arid site TZ-4 universally did not prefer growing on excessive bioavailable inorganic phosphate. However, the cultivable microbial communities of site TZ-5 did seem to display a preference for growth on phosphate amended plates (Tables 4 and 5).

F I G U R E 4 (a)
Absolute pool sizes and (b and c) relative pool sizes of resin-extractable P, microbial P, NaOH-extractable P, HClextractable P (Ca-P), and the residual P in (b) hyperarid sites and (c) arid sites

| Phosphorus distribution and microbial P cycling in Atacama sediments
In hyperarid environments such as the core of the Atacama Desert, P loss induced by leaching is negligible; hence, phosphate can be preserved to high levels in sediments. Release of inorganic P in desert minerals can be stimulated via chemical weathering (Sheldon, 1982) and by various microbial solubilization mechanisms, usually includ- Although Ca-P can dissolve abiotically, this process is negligible when pH is more than 8 (Guidry & Mackenzie, 2003) (Table 1).
At these high levels of phosphorus (with C:P of 67:1 on average) in spite of a large portion of Ca-P, micro-organisms in Atacama soils might be less phosphorus-limited than in normal terrestrial soils, where C:P is typically 186:1 on average (Cleveland & Liptzin, 2007;Reed & Wood, 2016). We also see that the addition of inorganic P to cell culture experiments makes no positive impact on the colony-forming units (CFU) for all of the sites with rainfall below ~ 15 mm/year (MES, PONR-2, Yungay, TZ-4; Table 5).
However at TZ-5, where annual precipitation levels can be 20 mm or more, addition of phosphate to the agar plates enhances growth.
Therefore, it appears that P can once again become a limiting nutrient at higher precipitation levels, even though the C:N:P ratio (258:1:5) still suggests a remarkably greater P reserve than N compared to normal terrestrial soils (186:13:1) (Cleveland & Liptzin, 2007;Reed & Wood, 2016). When amended with excess inorganic phosphate, growth of microbial colonies from TZ-5 increases significantly, by about 2 orders of magnitude (Table 5). However, the phosphate salts used in this amendment are directly available for biochemical use, quite different than the major pool of P at TZ-5, Various P sources can be accessed by microorganisms via immobilization, mineralization, and biological solubilization (Gyaneshwar, Kumar, & Parekh, 1998;Illmer & Schinner, 1992;Nahas, 1996) ( Figure 1). It appears that these processes must occur to some extent for our sites, as at higher TP levels, we observe that microbial P pool size is also greater (Figure 6a). This trend is also consistent with the transition from hyperarid to arid sites recorded by the relationship between TOC and TP ( Figure 5a). These correlations indicate that when rainfall exceeds ~ 10 mm/year, some TP can be desorbed by biological solubilization and made bioavailable by native microorganisms in situ. Accordingly, the microbial communities could be utilizing mineral P at TZ-5, but at a much slower pace than the resin-P which was added in the cell culture experiments, hence the hundred-fold increase in colony development (Table 5). This reasoning highlights the possibility for resin-P limitation even where mineral P may be more plentiful than other key nutrients such as nitrogen, and at a site where mineral P breakdown appears to be dominated by biological solubilization.
Thus at low N/P (namely high P/N) ratios, the nitrate-driven promotion of microbial P utilization is reduced, which is further unfavorable to microbial communities.
Along the hyperarid to arid gradient, the proportion of resin-P from Atacama soils decreases ( Figure 4) as it is presumably immobilized by the more abundant microbial communities in arid sites (Figure 1), although the concentrations of resin-P from Atacama soils are approaching the detection limit for our method (Crain, McLaren, Brunner, & Darrouzet-Nardi, 2018;Lester, Satomi, & Ponce, 2007).
Similarly, the relative sizes of microbial P and NaOH-P pools increase in arid sites (Figure 4). Besides the impact of rainfall on microbial P and NaOH-P (as containing organic P) pools, sedimentation potentially plays an important role in diluting these P pools, since the higher microbial P and NaOH-P levels are associated with smaller grain size (Table 1) and higher SiO 2 /Al 2 O 3 (quartz to clay ratios) (Figure 5c) in our study sites. Usually, larger grain sizes cause a decline in the sorption rates of phosphate in organic form (Meng, Yao, & Yu, 2014).
Similarly, we recently postulated that bulk N concentrations, largely supplied as nitrate by atmospheric deposition (Ewing et al., 2007;Michalski et al., 2004), could be diluted by higher sedimentation rates and thus more limited to microorganisms (Shen et al., 2019).
High sedimentation rates thus promote higher P/N ratios and a small relative pool size of microbial P and NaOH-P, and thus are hostile to these extremophilic micro-organisms.

| Phosphoenzymes and phosphate pathways
Several phosphoenzymes can drive exchange reactions between the stable oxygen isotopes of phosphate and surrounding water. The most studied phosphoenzymes are inorganic pyrophosphatase, alkaline and acid phosphatases, DNase, RNase, 5'-nucleotidase, and phytases, which are all found to be primary phosphoenzymes in Atacama microbiomes (Table 3) Among these pathways, inorganic pyrophosphatase plays a role in methylphosphonate degradation and any hydrolysis of pyrophosphate into two phosphate molecules; it is also a highly conserved phosphoenzyme that ubiquitously exists across all organisms (Knowles, 1980) (Figure 1). DNase and RNase break the phosphodiester bonds of corresponding nucleic acids, and alkaline phosphatase is responsible in completing DNA and RNA degradation by hydrolyzing nucleotides into nucleosides and single phosphate molecules (Liang & Blake, 2006 (Figure 1). Alkaline phosphatase is also involved in glycolysis II (from fructose 6-phosphate) (Cho, Seo, Kim, Jung, & Park, 2012). Pathways associated with 5'-nucleotidase and acid phosphatase are found to be involved in nucleotide degradation (Passariello et al., 2006). Pathways of phytases, such as phytate degradation I, were not identified in our sequence data, and the abundance of phytases is significantly lower than other phosphoenzymes. These data suggest that phytases are not the dominant drivers of stable oxygen isotope exchange between soil phosphate and the surrounding water.

| Phosphate δ 18 O
The dominant pool of P was Ca-bound (HCl-extractable) phosphate at all of the sites sampled (Figure 4). This pool either represents (a) the primary mineral P preserved in bedrock, or (b) P formed during early diagenesis as a result of the high pH of the soils, which promotes conversion of resin-P to Ca-P minerals (Oburger et al., 2011;Urrutia et al., 2014). To identify any biosignatures in soil δ 18 O PO4 , it is important to first differentiate the stable oxygen isotope fingerprints of biogenic phosphoenzyme-driven water-phosphate equili- the ambient water, as mainly directed by the activity of phosphoenzymes, usually inorganic pyrophosphatase (Shemesh, Kolodny, & Luz, 1983;Shemesh et al., 1988;Tamburini et al., 2014). Accordingly, the δ 18 O PO4 values are always greater in post-Cretaceous sedimentary rocks than in igneous rocks, and can be further enriched by biological activity (Figure 1).
The most hyperarid sites we sampled (with precipitation < 2 mm/ year; MES, PONR-2, and Yungay) all have low δ 18 O PO4 values in the Ca-P (7.9, 13.6, and 11.2‰, respectively periods, which is partially older than the solely Quaternary sedimentary rock source to the PONR-2 and Yungay sites. As a result, the δ 18 O PO4 value of MES soils is the smallest (Table 2). This parent bedrock isotopic endmember is thought to have remained stable for millennia due to the hyperarid nature of the environment and the minimal isotope fractionation associated with abiotic cycling of P ( Figure 1) (Azua-Bustos et al., 2015).
Arid sites (TZ-4 and TZ-5), on the other hand, have a higher abundance of magnesium and iron contents than hyperarid sites (Table 2 and Figure  Only hydrolysis by inorganic pyrophosphatase results in a positive δ 18 O PO4 (Figure 1). The results of enzyme and pathway prediction analyses imply that the hydrolysis by inorganic pyrophosphatase is the most likely reaction to cycle P in these environments, or that this process will rapidly overprint other P utilization strategies, in agreement with previous studies (Blake, 1998;Blake et al., 2005;Jaisi, Kukkadapu, Stout, Varga, & Blake, 2011;Tamburini et al., 2014).
The results of enzyme and pathway predictions imply that the hydrolysis by inorganic pyrophosphatase dominates the biogeochemical P dynamics in Atacama soils ( Figure 1). Therefore, if we assume that isotopic equilibrium has been reached at the arid sites (TZ-4 and TZ-5), we can re-arrange the microbial P turnover equation by inorganic pyrophosphatase (Eq. (4)) to establish the primary moisture source for the biological communities surviving and actively cycling P in this environment. Using the average temperature recorded from the nearest meteorological stations (Table 1) (Table 3 and Figures 1 and 7), or (c) the water sources for life are distinct from each other even on a small spatial scale. Further investigation is required to ascertain the driving force behind these small differences in calculated source water δ 18 O values at site TZ-5.
However, observed differences of ~ 5‰ between the calculated isotopic signature of source water at TZ-4 and TZ-5 point to a stronger driving force. One clear difference between the sites is their altitudes, where TZ-5 is at almost 1,000 m lower elevation than TZ-4 (Table 1). When considering this difference, theoretical water isotope values follow the expected trend of decreasing rainwater δ 18 O H2O with altitude (R. Aravena et al., 1999;Jordan et al., 2019).
Again this finding requires further investigation, but if correct, this suggests that rainwater does indeed dominate water resources at these arid sites, even under hostile conditions of < 20 mm/yr rainfall (Orlando et al., 2010;Schulze-Makuch et al., 2018). How exactly this rainwater is accessed by the biological community is still uncertain. It can be assumed that a proportion is available directly after a rain event but that there must also be a critical slow release water reservoir, or regular additional water (e.g., fog and groundwater), which reflects rainfall isotope values and is utilized between the infrequent rainfall events. We suggest that gypsum hydration water is the most likely reserve of bioavailable moisture between rainfall events in these environments (Palacio et al., 2014).
The fractionation of δ 18 O during uptake into gypsum is minimal but slightly positive, potentially explaining the theoretical positive isotope value of + 2.04‰ at TZ-5 which is higher than all theoretical rainwater and fog sources outlined above. The uptake of water into the gypsum mineral phase should also preserve the altitude-driven difference between rainfall at TZ-4 and TZ-5 as suggested by our theoretical δ 18 O H2O calculations.

| CON CLUS IONS
Here, we demonstrate the utility of the δ 18 O PO4 proxy in detecting biological phosphorus cycling in a Mars-analogue system. In the Atacama Desert, at hyperarid sites with annual rainfall < 2 mm,

CO N FLI C T O F I NTE R E S T
We have no conflict of interest to declare.