environments, Mo is soluble and readily bioavailable (e.g., MoO 2- 4 ). Mo-nitrogenase has the highest catalytic activity (Eady, 1996). The Fe-only and V-based alternative nitrogenases are also important in natural environments, especially when Mo concentration is low (McRose et al., 2017). V-nitrogenase has been observed to increase N 2 fixation rate in V-amended forest soils (Bellenger et al., 2014).

Alternative nitrogenases were once believed to have been important under the reducing conditions of early Earth, because of the low solubility of Mo-bearing minerals in anoxic oceans (Anbar & Knoll, 2002). Indeed, Mo isotope data from 3.0 Gyr iron formations suggest a low level of molybdate ions in seawater (Johnson et al., 2021). After the Great Oxidation Event (GOE), through oxidative weathering, Mo would have been mobilized from minerals and rocks as soluble MoO 2- 4 , thus making the metal increasingly available for biological reactions (Jelen et al., 2016). Consistent with this scenario, Mo-based nitrogenase was once believed to have evolved after the GOE around 2.3-2.4 Gyr ago (Scott et al., 2008). However, new evidence argues for a Mo-dominant nitrogenase in the mid-Archaean oceans, as early as 3.2 Gyr ago, even long before oxidative weathering became important. For example, tight distribution of nitrogen isotope fractionation of ~0‰ ± 1.5‰ in sedimentary rocks (Stueken et al., 2015) supports the presence of such Mo-nitrogenase as early as the Mesoarchean age (3.2-2.8 Gyr). Furthermore, recent studies, employing phylogenetic analysis, ancestral sequence inference and structural homology modelling, suggest that Mo-based nitrogenase appears to be ancestral to the Fe or V versions (Boyd & Peters, 2013;Garcia et al., 2020;Parsons et al., 2021).

The emergence of the Mo-based nitrogenase Nif before the GOE raises an apparent paradox: dissolved Mo levels in the ocean were only a few nM (Johnson et al., 2021), yet geochemical and biological evidence suggests a widespread occurrence and function of Mobased nitrogenase in Archaean oceans (Parsons et al., 2021). This apparent paradox is based on the untested assumption that only soluble Mo is bioavailable for use in Mo-based nitrogenase. Most studies investigating trace metal bioavailability consider soluble metals as the only bioavailable form, and there is little understanding of other sources of trace metals for various biogeochemical pathways including nitrogen fixation. We hypothesize that mineral-associated trace metals may be bioavailable and Mo-nitrogenase may have been synthesized on early Earth using mineral-associated Mo.

There are hints to support this hypothesis. In our previous study, Mo-based metabolic processes were detected in hot spring across temperature and pH gradients, even though soluble Mo was only measured at certain sites with circumneutral pH (Srivastava et al., 2018). This observation hints that there may be other sources of Mo.

Furthermore, a previous study (Liermann et al., 2005) observed that a model N 2 -fixing bacterium Azotobacter vinelandii was able to extract Mo from Fe-and Mo-enriched silicate glass for growth in the absence of soluble Mo. Apparently, A. vinelandii secreted a high-affinity ligand aminochelin ("molybdophore") to extract Mo from the glass (Liermann et al., 2005). However, the authors neither used natural Mo-or Fecontaining minerals nor measured the rates of N 2 fixation. Similarly, Knapp et al. (2007) demonstrated that the aerobic methanotroph Methylosinus trichosporium was able to extract Cu from Cu-doped iron oxides and borosilicate glass. A fluorescent chromopeptide, called methanobactin, mediated the release of Cu from the solids and allowed the methanotroph to express pMMO gene, however, they did not measure the rates of methane oxidation (Knapp et al., 2007).

The objective of this study was to test the bioavailability of naturally occurring mineral-associated trace metals and their impact on N 2 fixation rate, a first step toward understanding the ancestral origin of Mo-based nitrogenase. Experiments were performed where mineralassociated Mo, V, and Fe were used as the sole sources of trace metals for N 2 fixation by A. vinelandii. Although the V and Fe bioavailability study does not directly address the origin of Mo-based nitrogenase, it would offer a meaningful comparison in testing bioavailability of different mineral-associated trace metals. The N 2 fixation rate was measured using the acetylene reduction assay (ARA). Trace metal mobilization and siderophore production were analyzed, along with microscopic imaging of cell-mineral associations, to determine the potential microbial mechanisms of extracting these mineral-associated cofactors.

Two types of mineral-associated Mo, V, and Fe were prepared including adsorbed and structural forms. The reason for using these two types was to better understand the difference in bioavailability between sorbed and structurally bound metals. In the first type, metal cofactors were adsorbed onto minerals: (1) Mo sorption to ferrihydrite; (2) V sorption to goethite; (3) Fe sorption to montmorillonite.

The use of ferrihydrite for Mo was because there is significant adsorption of molybdate on ferrihydrite at pH 6.8 (Gustafsson, 2003) and at pH 7.7 (Goldberg et al., 1996), the optimal growth pH range of A. vinelandii. Ferrihydrite is believed to have been formed during the Archaean through the photoferrotrophic oxidation of soluble Fe(II) in the marine photic zone (Konhauser et al., 2005). Goethite was used for V adsorption because similar to Mo, there is significant V adsorption to goethite at pH 6-8 (Peacock & Sherman, 2004). Goethite is also believed to have been present during the Archaean (Angerer & Hagemann, 2010). The reason for using montmorillonite for Fe sorption was because this clay mineral contains a minimal amount of Fe so that the bioavailability of sorbed Fe can be determined.

Montmorillonite is the most common member of the smectite family.

In addition to these sorbed forms, a second set of minerals included structurally bound metal cofactors. The minerals used for Mo, V, and Fe were molybdenite (MoS 2 ), cavansite (Ca(VO)Si 4 O 10 •4H 2 O), and ferrihydrite (Fe 3+ 2 O 3 •0.5H 2 O), respectively. Molybdenite was available in the Archaean (Hazen et al., 2014). Cavansite was used as a model V-bearing mineral.

Ferrihydrite was synthesized by adapting the method from Schwertmann and Cornell (2007). Briefly, 50 mL of 0.06 M FeCl 3 •6H 2 O was prepared in a 250 mL flask. The solution was vigorously stirred while 1 N NaOH was slowly added to raise the pH to 6.8. Molybdate was sorbed to freshly prepared ferrihydrite using a previous method (Gustafsson, 2003). Briefly, a stock solution of Na 2 MoO 4 (corresponding to 50 μM MoO 2- 4 ) was mixed with ferrihydrite suspension in 50 mL polypropylene centrifuge tubes in a solution of 0.01 M ionic strength (as NaNO 3 ). HNO 3 and NaOH were used to adjust the pH to circumneutral.

After 24 h equilibrium, the sample was centrifuged for 10 min at 13,000 g, and the pellet was collected and resuspended in an equal volume of Burk medium. This centrifugation-resuspension process was repeated three times to wash the Mo-ferrihydrite pellet and to eliminate any potential effect of NaNO 3 . A previous study (Gustafsson, 2003) showed that at pH 6.8 most of the MoO 2- 4 (~97%) should be sorbed to ferrihydrite. To confirm Mo sorption, the supernatants from all wash steps were analyzed. by inductively coupled plasma-optical emission spectrometry (ICP-OES; Agilent Technologies) to measure the amount of remaining aqueous Mo concentration. The detection limit of the ICP-OES instrument was <0.6, <0.7 and <0.5 ng/mL for Fe, Mo, and V, respectively. The result confirmed nearly complete adsorption of Mo to ferrihydrite, equivalent to 9.6 mg Mo/g ferrihydrite (hereafter termed Mo-ferrihydrite).

A vanadium stock solution (100 ppm) was prepared by dissolving V 2 O 5 solid in deionized water. V was sorbed onto goethite (Sigma-Aldrich; Cat. # 71063) following a previous method (Peacock & Sherman, 2004). Briefly, 7.5 mL of the V stock solution of 25 ppm at pH 6 was added to 0.1 g goethite in 22.5 mL of 0.1 mM NaNO 3 solution. The suspension was shaken continuously for 144 h followed by repeats of centrifugation and resuspension. The supernatants were analyzed with ICP-OES and the result showed 5.71 mg V/g goethite (hereafter termed V-goethite), which fell within the range of 4.9 mg/g (at pH 8) -6.5 mg/g (at pH 6) reported by Peacock and Sherman (2004).

Sorption experiment was performed by mixing 100 mM aqueous Fe 3+ (FeCl 3 •6H 2 O in 0.1 N HCl) with a 10 g/L montmorillonite SWy-2 (hereafter referred to as SWy-2) suspension (Bhattacharyya & Gupta, 2008). SWy-2 was purchased from the Source Clays Repository of the Clay Minerals Society (Purdue University, IN). This mixture was shaken for a period of 10 h followed by three repeats of centrifugation and resuspension. The supernatants were analyzed using the 1,10-phenanthroline method (Amonette & Templeton, 1998) for dissolved Fe using a Genesys 10 S UV-Vis spectrometer (Thermo Scientific). The result showed 0.56 mg Fe/g SWy-2 (hereafter termed Fe-SWy-2).

The trace metal containing minerals included molybdenite (MoS 2 ), cavansite [Ca(VO)Si 4 O 10 •4H 2 O], and ferrihydrite (Fe 3+ 2 O 3 •0.5 H 2 O). Molybdenite was acquired from the Limper Geology Museum of Miami University. Cavansite was purchased from an online source (Exquisite Crystals: https://www.exqui sitec rysta ls.com). Ferrihydrite was synthesized as described above (Schwertmann & Cornell, 2007).

Molybdenite was broken into small particles using a sterile scalpel and tweezer, while cavansite was broken using a mortar and pestle.

The mineral particles were >10 μm as confirmed by SEM images.

Azotobacter vinelandii DJ (ATCC BAA-1303) was routinely cultured in a modified liquid Burk medium of the following composition: KH 2 PO 4 , 0.2 g; K 2 HPO 4 , 0.8 g; MgSO 4 •7H 2 O, 0.2 g; CaCl 2 •2H 2 O, 0.09 g; sucrose, 20 g at 30°C in a 150-rpm rotary shaker (Bishop et al., 1982). A. vinelandii was first revived in Burk medium with ammonium acetate (NH 4 OAc) as the nitrogen source. Subsequently, to promote diazotrophy, A. vinelandii was cultured in Burk medium with soluble Mo, V, and Fe but without NH 4 OAc. To favor the use of Modependent nitrogenase (Mo-diazotrophy), 2 μM Mo (as Na 2 MoO 4 ) and 52 μM Fe (as FeSO 4 •7H 2 O) were added to the Burk medium (hereafter termed Nif medium). In Nif medium, Fe was still required because it is part of the Mo-and V-dependent nitrogenases. This amount of Fe was considered replete (McRose et al., 2017). To favor V-diazotrophy (V-dependent nitrogenase), V 2 O 5 (final conc. 100 μM) was added to Burk medium along with 52 μM Fe (Vnf medium). Anf medium (to favor Fe-dependent nitrogenase or Fe-diazotrophy) contained 52 μM Fe only. A. vinelandii was cultured under each diazotrophy condition for 24-48 h at 30°C in a 150-rpm rotary shaker to achieve an OD 620 of ~1 [equivalent to 1.16 (±0.16) × 10 8 cells/mL] (Wichard et al., 2009). After cell growth, the suspension was centrifuged for 10 min at 6000 g, and the pellet was collected and resuspended in an equal volume of Burk medium but without any trace metals. This centrifugation-resuspension process was repeated three times to eliminate any carryover of aqueous metals from initial growth. This inoculum was used to determine the ability of mineralassociated trace metals as cofactors for nitrogenase activity under the respective diazotrophic condition.

To test the bioavailability of solid source trace metals, Mo-ferrihydrite or molybdenite replaced aqueous Mo (up to 2 μM aqueous concentration assuming full dissolution) in Nif medium, V-goethite or cavansite replaced aqueous V (up to 150 μM) in Vnf medium, and Fe-SWy-2 or ferrihydrite replaced aqueous Fe (up to 52 μM) in Anf medium. The corresponding mass concentrations of Mo-ferrihydrite, V-goethite, and Fe-SWy-2 were 0.02, 1.35, and 5.20 g/L, respectively, while those of molybdenite, cavansite, and ferrihydrite were 0.32, 67, and 9 mg/L, respectively. The 30-mL serum vials contained 10 mL medium and were sealed using butyl rubber septa. The medium and the headspace were each purged with He for 30 min, followed by sterilization at a temperature of 121°C and a pressure of 103 kPa for 30 min. Finally, 1 mL of A. vinelandii inoculum (OD 620 of ~1) grown in the corresponding aqueous metal medium was added with a syringe needle to achieve a final OD 620 of ~0.1. Positive controls contained aqueous forms of Mo, V, and Fe (i.e., Nif, Vnf and Anf medium, respectively). In an abiotic control, 1 mL cell suspension was replaced by the same amount of Burk medium. All experiments were conducted in triplicates.

The bioavailability of mineral-associated metals was assessed by measuring the rate of N 2 fixation with the Acetylene Reduction Assay (ARA). The ARA is widely used to measure the rate of N 2 -fixation because nitrogenases are able to similarly reduce acetylene to ethylene (C 2 H 4 ) and N 2 to ammonia (Dilworth, 1966). We assume that ARA is similarly effective for measuring the rate of N 2 fixation using nitrogenases synthesized from both forms of trace metals (adsorbed and structurally bound). The assay was initiated by injecting 2 mL of high purity (99.6%) (Airgas) acetylene into the headspace of the serum vials after removing 2 mL He (i.e., ~10% acetylene and ~90% helium in the headspace). The amount of ethylene produced from acetylene reduction was determined in a 250 μL subsample of the headspace.

Subsamples were collected every 15 min using a gas-tight syringe.

The syringe was flushed for four times with He gas. The ARA was conducted for a period of 4 h, because a longer time would exhaust dissolved O 2 in the medium and inactivate A. vinelandii cells. Ethylene concentration in the subsamples was measured using gas chromatography (Varian 3300 Gas Chromatography) equipped with an 80/100 Porapak Q column and flame ionization detector using nitrogen as the carrier gas. The temperature of the injector and detector was maintained at 100°C and the column at 50°C. Ethylene concentration was determined by comparing to a freshly prepared standard using ultra high purity ethylene gas (99.9%) (Airgas). A calibration curve was pre-established relating the peak area to ethylene (nmol) amount.

Ethylene amount at the end of 4 h was calculated as follows:

where ethylene amount

The rate of nitrogen fixation was calculated as follows:

In the N 2 -fixation rate measurement, Burk medium itself may have contained some trace metals that would affect N 2 fixation. In or 72 h (for metal release from abiotic dissolution of molybdenite, canvansite or ferrihydrite), supernatant samples were collected by centrifugation at 8000 g for 10 min followed by syringe filtration (0.22 μm polypropylene sterile syringe filter) (ThermoFisher). The samples were acidified (2% v/v HNO 3 ) and analyzed by ICP-OES.

To confirm if any trace metals passively released from abiotic desorption or mineral dissolution were sufficient to fix N 2 , aliquots of the above supernatant solutions were used for N 2 fixation rate measurements.

To test whether A. vinelandii required a direct contact to produce siderophores and other metabolites as a mechanism of extracting mineral-associated trace metals (Ahmed & Holmstrom, 2015), the minerals were placed inside a dialysis bag (~0.01 μm pore size, molecular weight cut-off (MWCO) 12,000-14,000 Da; ThermoFisher), while A. vinelandii cells were in bulk solution. This pore size of the bag would not allow passage of any cells or mineral particles but should not hinder passage of secondary metabolites with a molecular weight <1500 Da (Rai et al., 2020).

Molybdenite, cavansite, and ferrihydrite may not be readily available to support N 2 fixation, likely due to their crystalline structures. Thus, in certain experiments, A. vinelandii cells were pre-acclimated to these minerals for 24-96 h as the sole sources of Mo, V, and Fe in Nif, Vnf, and Anf media, respectively, followed by ARA experiments.

Because wild-type A. vinelandii has a generation time of 4 h under diazotrophic condition, the incubation period of 24-96 h would constitute at least 6-24 generations (Mus et al., 2017).

Azotobacter vinelandii is known to produce siderophores as a strategy to increase the bioavailability of mineral-associated trace metals (Liermann et al., 2005). To quantify various siderophore production in response to Mo-, V-, and Fe-diazotrophy, A. vinelandii was

incubated in separate experiments. Different from the ARA measurement (4 h culture starting with an OD of 0.1), in this experiment a lower inoculum (OD ~0.005) was used and siderophore production was analyzed during exponential and early stationary phases after 24 and 48 h incubation, respectively (final ODs = 1.1-1.7). The longer incubation time and higher biological growth usually lead to more siderophore production. Siderophore samples were syringe filtered (0.22 μm) and analyzed with LC-MS (Ultimate 3000 UPLC/ ISQ EC; Thermo Fisher Scientific) as described previously (Baars et al., 2016). The final siderophore concentrations were normalized to the optical density (OD) of A. vinelandii.

Scanning electron microscopy (SEM) was performed for some selective experimental groups to visualize cell-mineral associations. A. vinelandii was grown for 48 h in Nif and Vnf media with molybdenite and cavansite as the sole sources of Mo and V, respectively. Subsequently, cell-mineral suspensions were fixed using 2.5% glutaraldehyde and 2% paraformaldehyde in 0.05 M sodium cacodylate buffer followed by stepwise dehydration in increasing concentrations of ethanol (25%, 50%, 75%, 95%, and 100%) and critical point drying (CPD) (Dong et al., 2003). Finally, the samples were mounted on SEM stubs and sputter-coated with gold before imaging using a Zeiss Supra 35 VP Field Emission Scanning Electron Microscope with an accelerating voltage of 7 kV and a working distance of 9.5 mm.

The background levels of Mo and V in Burk medium were found to be below detection limit (BDL) of the ICP-OES method (Table S1).

However, there was a small amount of background Fe in Burk medium (4.04 μM). In the absence of bacteria, the amount of desorption of Mo from Mo-ferrihydrite, V from V-goethite, and Fe from Fe-SWy-2 was 0.02, 0.37, and 0.70 μM, respectively. There were certain levels of aqueous Fe because the desorption experiments were conducted in Nif, Vnf, and Anf media, where 52 μM aqueous Fe was added except the case of Fe-SWy-2. However, the measured Fe concentrations in Mo-ferrihydrite (in Nif medium) and V-goethite (in Vnf medium) were lower than 52 μM, likely because of adsorption.

Without bacteria, aqueous Mo from dissolution of molybdenite, V from cavansite, and Fe from ferrihydrite in their respective medium (Nif, Vnf, and Anf, respectively) were all BDL (Table S1). Apparently, molybdenite and cavansite did not sorb any added aqueous Fe, and ferrihydrite did not release any Fe from its structure. pH decreased slightly from ~7.2 to ~6.9 during all these abiotic experiments.

In the presence of A. vinelandii cells, positive controls (with aqueous trace metals) showed time-course consumption. In particular, aqueous Mo and Fe concentrations decreased from 2.05 to 0.39 μM and from 52.99 to 1.08 μM, respectively (Table 1).

However, in V-diazotrophy, it was the aqueous Fe, not the V, that was actually consumed. When both A. vinelandii cells and minerals were present, aqueous concentrations of metals all increased. For example, Mo release from molybdenite was BDL in abiotic control (Table S1), but after 24-h incubation with A. vinelandii, it increased to 0.35 μM (Table 1). Likewise, the amount of aqueous V released from V-goethite increased from 0.37 μM due to abiotic desorption to 2.61 μM in biotic treatment. The amount of aqueous V from canvansite dissolution increased from BDL in abiotic control to 2.25 μM in the presence of microbial activity. A similar case was observed for ferrihydrite. pH decreased slightly from ~7.3 to ~6.9 in all biotic experiments.

Aqueous trace metal concentrations in the supernatant of the three diazotrophy conditions.

Sample name Fe (μM) Mo (μM) V (μM) Aq. Mo 0 52.99 2.05 BDL Aq. Mo 24 1.08 0.39 0.13 Mo-ferrihydrite 0 53.56 BDL BDL Mo-ferrihydrite 24 0.53 BDL BDL Mo-ferrihydrite DB 0 53.40 BDL BDL Mo-ferrihydrite DB 24 BDL BDL BDL Molybdenite 0 53.24 BDL BDL Molybdenite 24 62.54 0.35 BDL Aq. V 0 52.33 0.02 132.88 Aq. V 24 0.62 BDL 156.62 V-goethite 0 52.06 BDL BDL V-goethite 24 1.03 BDL 2.61 V-goethite DB 0 51.76 BDL BDL V-goethite DB 24 0.28 BDL 2.03 Cavansite 0 53.56 BDL BDL Cavansite 24 1.18 BDL 2.25 Aq. Fe 0 52.17 0.01 BDL Aq. Fe 24 h 1.40 BDL BDL Fe-Swy2 0 BDL BDL BDL Fe-Swy2 24 h 0.77 BDL BDL Fe-Swy2 DB 0 BDL BDL BDL Fe-Swy2 DB 24 0.56 BDL BDL Ferrihydrite 0 BDL BDL BDL Ferrihydrite 24 0.53 BDL 0.02 Note: Error (+/-): <2%. Abbreviations for the different sources follow Figures 1-3. Abbreviations for sampling times are: T0 (0 h) and T24 (24 h). The total concentrations of Mo, V and Fe (either in aqueous or sorbed form) are 2 μM Mo, 150 μM V, and 52 μM Fe, respectively.

Starting aqueous concentration of Fe 2+ in Mo-and V-diazotrophic conditions was 52 μM (in the form of FeSO 4 ).

Abbreviation: BDL, below detection limit.

The N 2 fixation rate is usually normalized to biomass (Bellenger et al., 2011), however, over the short measurement duration used in this study (4 h), biomass did not change (OD 620 within the range of 0.10-0.11). Therefore, the N 2 fixation rate was expressed as both cumulative ethylene production over 4 h (nmol) and ethylene production rate (nmol/h).

As expected, A. vinelandii was able to use aqueous forms of Mo, V, and Fe for N 2 fixation. Relative to aqueous Mo, the ethylene production rate with aqueous V and Fe was 71% and 76% lower, respectively (Figure S1). Using abiotic leachates in the Nif, Vnf, and Anf media, neither the total amount nor the rate of ethylene production was higher than that in the presence of 52 μM aqueous Fe alone, suggesting that any aqueous Mo, V, and Fe passively released from the minerals were not sufficient to stimulate N 2 fixation. Interestingly, A. vinelandii utilized mineral-adsorbed trace metals for N 2 fixation but with reduced rates. By the end of 4 h and with an inoculum OD 620 of ~0.1, the average rate of ethylene production 28 nmol/h. When aqueous Mo was used, with a total of 111 nmol of ethylene production (Figure 1). However, when Mo-sorbed ferrihydrite was used, the average rate was only 11 nmol/h, 59% lower than that when aqueous Mo was used, with a total of 46 nmol of ethylene produced. Similarly, the average rates of ethylene production with V-sorbed goethite and Fe-sorbed SWy-2 were 5 and 4 nmol/h, respectively, which was 37% and 42% lower than their aqueous equivalents (Figures 2 and 3). Likewise, the amounts of ethylene production with V-sorbed goethite and Fe-sorbed SWy-2 were

E 1 Amount of ethylene produced by the end of 4 h (1st y-axis) and the average ethylene production rate (2nd y-axis) under Mo-diazotrophy conditions. The relative amount of nitrogen fixation with aq. Mo was taken as 100% for the purpose of comparison with other Mo sources. Abbreviations for the different sources are aqueous Mo (Aq. Mo), Mo-Fh (Mo-ferrihydrite), Mo-Fh DB (Mo-ferrihydrite in dialysis bag), Moly. 24, 48, 72, and 96 h (molybdenite with pre-incubated inoculum for 24, 48, 72, and 96 h). 0 5 10 15 20 25 30 0 20 40 60 80 100 120 Aq. Mo Sorbed Mo Sorbed Mo DB MoS2 1 Day MoS2 2 Day MoS2 3 Day MoS2 4 Day AbioƟc Control C 2 H 4 ProducƟon Rate (nmol C 2 H 4 /hr. ) CumulaƟve C 2 H 4 Produced in 4 hours (nmoles)

F I G U R E 2 Amount of ethylene produced by the end of 4 h (1st y-axis) and the average ethylene production rate (2nd y-axis) under Vdiazotrophy condition. The relative amount of nitrogen fixation with aq. V was taken as 100% for the purpose of comparison with other V sources. Abbreviations for different sources of V are aqueous V (Aq. V), V-Goe (V-Goethite), V-Goe DB (V-goethite in dialysis bag), Cav 24, 48, 72, and 96 h (cavansite with pre-incubated inoculum for 24, 48, 72, and 96 h).

0 1 2 3 4 5 6 7 8 9 0 5 10 15 20 25 30 35 Aq. V V-Goe V-Goe DB Cav 24 hr Cav 48 hr Cav 72 hr Cav 96 hr AbioƟc Control C 2 H 4 ProducƟon Rate (nmol C 2 H 4 /hr. ) CumulaƟve C 2 H 4 Produced in 4 hours (nmoles)

V-diazotrophy 32 and 15 nmol, respectively (Figures 2 and 3). The average rate of ethylene production with Mo-ferrihydrite (11 nmol/h) was higher than that with aqueous Fe (7 nmol/h). The total amount of ethylene production by the end of 4 h with Mo-ferrihydrite (46 nmol) was also higher than that with aqueous Fe (27 nmol), even when both were present in the Nif medium, suggesting that adsorbed Mo (Mo-ferrihydrite) was a preferred metal cofactor over aqueous Fe.

When A. vinelandii inoculum was incubated with molybdenite and cavansite as the sole sources for Mo-and V-diazotrophy, respectively, a negligible amount of ethylene was produced after 4 h (data not shown). Therefore, a pre-acclimated culture was used to measure the N 2 fixation rate. The average rate of ethylene production by the inoculum that was pre-acclimated with molybdenite for 24 h was 2 nmol/h (Figure 1) with a total of 7 nmol of ethylene produced (Figure 1). The rate and total amount of ethylene production doubled when the pre-acclimation time increased to 48 h. However, a further increase in pre-acclimation time decreased the rate of ethylene production, possibly due to over-growth of culture, decrease of nutrient supply, and accumulation of toxic waste products (Mukhtar et al., 2018). When compared with aqueous Mo, even with the optimal pre-acclimation time (48 h), the rate of ethylene production with molybdenite was still 86% lower, with a total of 16 nmol of ethylene produced (Figure 1). Similarly, in case of cavansite, the average rate of ethylene production with pre-acclimated inoculum for 24-72 h was 0.2-0.5 nmol/h (Figure 2) with only 1-2 nmol of ethylene produced (Figure 2). When compared to aqueous V, even with the optimal pre-incubation time (72 h), the rate of ethylene production with cavansite was 94% lower with a total of only 2 nmol of ethylene produced (Figure 2).

Ferrihydrite was a better source of Fe for Fe-diazotrophy compared with molybdenite and cavansite for Mo-and V-diazotrophy, respectively, because the average rate of ethylene produced was only 54% lower than that with aqueous Fe. After 4 h, the average rate of ethylene production using inoculum that was pre-acclimated with ferrihydrite for 24 h was 3 nmol/h (Figure 3), with a total of 12 nmol of ethylene production (Figure 3). These results indicate that of the three structural trace metal sources tested, ferrihydrite was the most effective for Fe-diazotrophy by A. vinelandii, consistent with more bioavailable nature of ferrihydrite.

When Mo-ferrihydrite, V-goethite, and Fe-SWy-2 were enclosed inside a dialysis bag, the average ethylene production rate was only 1 nmol/h with Mo-ferrihydrite (Figure 1a), 1 nmol/h with V-goethite (Figure 2), and 2 nmol/h with Fe-SWy-2 (Figure 3). These rates were 96%, 82%, and 75% lower than their aqueous equivalents. The total amount of ethylene production was 6, 5 and 7 nmol for these three treatments, respectively (Figures 123), again significantly lower than their aqueous equivalents.

Azotobacter vinelandii produces three classes of siderophores:

the weak α-hydroxycarboxylate siderophore vibrioferrin, catechol siderophores (2,3-dihydroxybenzoic acid, aminochelin, azotochelin, protochelin), and azotobactins (Baars et al., 2016).

Among these, aminochelin (Liermann et al., 2005) and azotochelin (Duhme et al., 1998) are known to bind Mo. These siderophores were quantified in all experimental incubations (Table S2). Unexpectedly, there was no significant correlation between Fe availability and production of sideophores. The highest F I G U R E 3 Amount of ethylene produced by the end of 4 h (1st y-axis) and the average ethylene production rate (2nd y-axis) under Fediazotrophy condition. The relative amount of nitrogen fixation with aq. Fe was taken as 100% for the purpose of comparison with other Fe sources. Abbreviations for the different sources are aqueous Fe (Aq. Fe), Fe-SWy-2 (Fe-Montmorillonite SWy-2), Fe-SWy-2 DB (Fe-Montmorillonite SWy-2 in dialysis bag) and ferrihydrite (Fh).

concentrations were detected in Mo-diazotrophy (Figure 4), followed by V-(Figure 5) and Fe-diazotrophy (Figure 6).

Interestingly, the highest siderophore concentrations were produced in cultures where cofactors were supplied in the forms of minerals. By contrast, siderophore concentrations were much lower with aqueous metals or when the minerals were enclosed inside a dialysis bag. Generally, the siderophore pool consisted of vibrioferrin, aminochelin and azotochelin, while azotobactin and protochelin were negligible.

The highest siderophore production was detected in the 24 h molybdenite culture (Figure 4), with a dominant amount of vibrioferrin (up to ~178 μM per OD), and smaller amounts of aminochelin and azotochelin (~35 μM each). This condition had the highest dissolved Fe concentration, suggesting that siderophore production may have been induced by factors other than a low level of dissolved iron.

The second highest vibrioferrin production was observed in Mosorbed ferrihydrite culture (~105 μM). However, when Mo-sorbed ferrihydrite was supplied in a dialysis bag, only small amounts of siderophores were detected (up to 2 μM), although cells reached similar optical densities. These data suggest that a physical contact between cells and mineral surface may be required to enhance the production of siderophores.

Spherical A. vinelandii cells attached to molybdenite (Figure 7) and cavansite (Figure 8) surfaces when these minerals served as the sole sources of Mo and V in the medium. These images were consistent with increased aqueous concentrations of Mo and V after 24 h of incubation (Table 1), suggesting that A. vinelandii were able to extract these trace metals from minerals by physical attachment.

4.1 | Biological N 2 fixation using mineralassociated trace metals

Because aqueous Fe (~52 μM) was added to Nif and Vnf media in the presence of mineral-associated Mo and V, there may be ambiguity of the actual metals used to for N 2 fixation (aqueous Fe or mineral-associated Mo, V, or Fe). However, multiple lines of evidence strongly suggest that mineral-associated Mo, V, or Fe were bioavailable, in addition to or as alternative to aqueous Fe.

First, the average ethylene production rate with aqueous Fe (<7 nmol/h) was much lower than that with Mo-ferrihydrite (11 nmol/h). Likewise, the total amount of ethylene produced after 4 h with aqueous Fe (27 nmol) was much lower than that with Mo-ferrihydrite (46 nmol). These results suggest that at least some sorbed Mo was utilized for nitrogenase synthesis. It was likely that both Fe-and Monitrogenases may have been responsible for the amount of ethylene production. In either case, Mo-nitrogenase should have contributed to N 2 fixation.

Second, in case of molybdenite, the average rate and amount of ethylene production were much lower (2-4 nmol/h and a total of 7-16 nmol, respectively) than that with aqueous Fe alone (7 nmol/h and a total of 27 nmol, respectively). Aqueous Fe did not appear to sorb to molybdenite surface (Table S1). Furthermore, unlike aqueous

Fe-diazotrophy where its concentration was consumed after 24 h, its concentration in the molybdenite experiment actually significantly increased (Table 1). The only explanation was the use of a

Mo-based nitrogenase for the observed N 2 -fixation activity, even in the presence of ample aqueous Fe (Table S1). The source of aqueous their absence) (Table 1; Table S1). Molybdenite may have also contained some Fe and released it to solution upon biotic dissolution, thus accounting for its concentration increase after 24 h.

Third, a previous study showed that A. vinelandii produces a highaffinity ligand ("molybdophore"), aminochelin, to extract Mo from a silicate glass under Mo-limited and Fe-replete conditions (Liermann et al., 2005). Our data are consistent showing higher amounts of aminochelin in the presence of Mo-ferrihydrite and molybdenite than in the presence of aqueous Mo and ferrihydrite (Figures 4 and 6).

A similar case was observed for V-diazotrophy. The average rates and amounts of ethylene production were much lower with mineralassociated V (5 nmol/h and a total of 20 nmol with V-goethite, and <1 nmol/h and a total of 1-2 nmol for cavansite) than that with aqueous Fe (7 nmol/h and a total of 27 nmol), even in the presence of added aqueous Fe (Table S1). A higher concentration of vibrioferrin was detected in the presence of these V minerals (Figure 5) than under any Fe-diazotrophy condition (Figure 6). High production of vibrioferrin has been previously observed under conditions of "mild" Fe limitation (Baars et al., 2016;Zhang et al., 2019), but its role in V complexation is unclear. Nonetheless, it cannot be ruled out that A. vinelandii might have employed vibrioferrin to obtain V from minerals. Indeed, aqueous V concentration significantly increased from 0.37 to 2.61 μM in the V-goethite treatment and from BDL to 2.25 μM in the canvansite treatment (Table 1; Table S1), most likely due to active extraction of V from these minerals by A. vinelandii cells.

F I G U R E 5 Siderophore production under V-diazotrophy condition using different sources of V at 24 and 48 h. Abbreviations for the different sources follow Figure 2. et al., 2019). The results of this study are particularly relevant to life on land, because unlike the oceans where there are certain levels of dissolved Mo [a few nM in Archean oceans (Johnson et al., 2021) and ~107 nM in modern oceans (Collier, 1985)], terrestrial environ-