Formation of Elemental Sulfur Nodules

There are three primary theories for the microbially mediated formation of sulfur nodules; reduction of sulfate [38], oxidation of sulfide [14,16,18–21] or some form of disproportionation—with the caveat being that these processes must terminate at S⁰ [22,23]. Most previous studies have concluded that the reoxidation of sulfide, terminating at S⁰, is the mechanism behind the formation of sulfur nodules [14,39]. We use our data to conclude instead that elemental sulfur nodules in the Lake Lisan were formed by reduction of sulfate through bacterial sulfate reduction during gypsum diagenesis, terminating at elemental sulfur.

We have several reasons for making this conclusion. One of the primarily reasons is that there is no pyrite found within the formation [12,36]. We would expect, given that we have significant concentrations of iron within the gypsum close to the nodule, if sulfide was formed during bacterial sulfate reduction, then some pyrite should be present. This is because of the high reactivity of sulfide towards iron. It has been suggested that the Lisan Formation sediments, titanomagnetite that was initially laid down is replaced with greigite (Fe₃S₄) during anoxic burial diagenesis in the presence of bacterial sulfate reduction and that this greigite could represent the ‘missing sulfide’ that was oxidized to make elemental sulfur [37]. In the currently wet Dead Sea sediments, greigite has been found, although it is completely absent in the Lisan Formation, because this greigite was quantitatively oxidized when the Lisan Formation was exposed to the air [37]. However, greigite itself is not generated from aqueous sulfide; greigite is generated as a precursor to pyrite in settings where bacterial sulfate reduction is terminated prematurely and does not go all the way to sulfide [40]. The presence of griegite in the Lisan Formation does not, therefore, require that sulfide was ever produced during bacterial sulfate reduction. Indeed, the past-presence of greigite in the Lisan formation may be further evidence that bacterial sulfate reduction during microbial diagenesis did not go all the way to sulfide. Alternatively, we would need to invoke the production of sulfide that exists but separated from all local iron; this seems implausible. This leads us to suggest that our data is more consistent with bacterial sulfate reduction terminating at elemental sulfur rather than sulfide oxidation. As discussed in the introduction, this has implications for both the biochemical pathways of sulfate reduction, which impinges on the sulfur isotope composition of the nodules.

The fact that the sulfur nodules have a uniform isotope composition suggests that they formed in an open system with constant replenishment of sulfate to feed bacterial sulfate reduction. During closed system bacterial sulfate reduction, the continual formation of some reduced sulfur product into some mineral phase results in progressive enrichment in the δ³⁴S of the residual sulfate (Figure 3a). This increase in δ³⁴S of the residual sulfate in turn can manifest in a progressive δ³⁴S across the accreting and growing reduced sulfur mineral phase, typically pyrite. This can result in pyrite crystals that vary by 30-50‰ across a single grain. We can contrast this with open system conditions where the sulfate feeding bacterial sulfate reduction is maintained at a constant δ³⁴S because the system is ‘open’ and the aqueous sulfate is continually replenished (Figure 3b). In this case the δ³⁴S across the growing crystal is constant and reflects the condition of formation and the sulfur isotope partitioning during the reduction of sulfate and formation of the mineral.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. A schematic of the sulfur isotope composition during bacterial sulfate reduction in closed versus open systems.

In closed systems the sulfur isotopes in the reduced product (pyrite or elemental sulfur, for example) may have significant isotope variability as they are growing from a pool that is evolving isotopically with time. In contrast in an open system with constant replenishment of the source of sulfate the isotope composition of the reduced product would not be expected to vary. We use our data to conclude that the sulfur nodules in the Lake Lisan formed in an open system. The symbol ε_s is the sulfur isotope fractionation during sulfate reduction.

https://doi.org/10.1371/journal.pone.0075883.g003

What is the sulfur isotope fractionation during the formation of the sulfur nodules in the Lisan Formation? While the absolute difference between the gypsum and the sulfur nodules is 27 to 29‰, it is likely that during the formation of the sulfur nodules, the sulfur isotope fractionation was larger than this. Dissolving gypsum in the Lisan Formation during aqueous phase burial diagenesis would have produced initial pore fluid sulfate with a δ³⁴S of 18 to 19‰ because there is no initial isotope fractionation during gypsum dissolution. For each mol of this pore fluid sulfate that is reduced, another mol of gypsum would dissolve into the pore fluid to maintain gypsum saturation; this is similar to gypsum beds undergoing aqueous diagenesis today in Guatemalan lakes [14,41]. The pore fluid sulfate would then be the electron acceptor used during bacterial sulfate reduction, and ³²S (and ¹⁶O_SO4) would have been preferentially reduced from this pool. Similar to bacterial sulfate reduction in marine pore fluids, this creates a sulfur isotope gradient in the aqueous sulfate where the sulfate closest to the nodule is isotopically ‘heaviest’ and there is a diffusive gradient to the sulfate further away from the nodule. This suggests that the δ³⁴S of the pore fluid sulfate could have been significantly higher than the precursor gypsum; in Lake Peten Itza in Guatemala the pore fluid sulfate δ³⁴S is 47 to 50‰ from a precursor gypsum of 18 to 19‰ [14]. A schematic showing this is given in Figure 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Our hypothesized model for the geochemical environment leading to the formation of the sulfur nodules.

Our precursor gypsum is +19‰ and the elemental sulfur is -8‰; these are the two ‘knowns’ as we are able to measure them today. In theory the pore fluid sulfate would be significantly heavier than the gypsum since this would be the mobile pool from which sulfate is reduced to form the sulfur nodule. Thus the sulfur isotope fractionation could be much larger than the 28‰ difference between the gypsum and the nodule.

https://doi.org/10.1371/journal.pone.0075883.g004

The nodules’ spherical nature and the appearance of a redox boundary in iron and manganese concentrations near this redox boundary hint that there was an in situ bio-film surrounding an initial nucleation point, although this is not unequivocal. If this was the case, once all the gypsum at the initiation point was consumed, and if it was energetically favorable under the conditions to continue, the bio-film would have expanded, allowing the nodule to gain mass in the centre. The manganese and iron concentration profiles could therefore be read as a snapshot of the redox conditions of the pore fluids at and approaching the nodule-gypsum interface. The increased level of Fe near the nodule hints at the possibility of a dissolved iron pool near the nodule. Part of the Fe(II) could have been oxidized during desiccation, producing the rust colored rims seen in hand specimen (Figure 2). By this token, the low manganese concentrations near the nodule are suggestive of consumption of manganese at the nodule-gypsum interface. We will fit these observations together with a modified model for bacterial sulfate reduction in the next section.

A modified thiosulfate shunt in unique environments

Our ultimate objective is to use the isotope and major element data to understand the conditions under which bacterial sulfate reduction may result in elemental sulfur as the end product. We suggest that the formation of elemental sulfur in the Lisan Formation results from a modification of the final biochemical step during bacterial sulfate reduction: the six-electron reduction of sulfite to sulfide by dissimilatory sulfite reductase (Dsr). This hypothesis invokes the unique environmental conditions that likely existed in the Lake Lisan sediments during burial-diagenesis: high salinity, supersaturated sulfate in pore fluids, low sulfate reduction rates and low electron potential (not much organic carbon).

The recently published crystal structure of dissimilatory sulfite reductase demonstrates that it consists of two proteins, the DsrA and DsrB dimer and DsrC [42] (Figure 5). Sulfite is reduced in the DsrAB-C complex in sequential two-electron transfers; first to S²⁺, then to S⁰ (Figure 5, pathway 1). If either S²⁺ or S⁰ are released from DsrA-B-C, and there is excess intracellular sulfite, then nucleophilic attack by this sulfite on the S²⁺ and S⁰ intermediates will produce S₂O₆^2- (trithionate) and S₂O₃^2- (thiosulfate) respectively (Figure 5, pathway 2). Thus intracellular thionate formation is promoted through DsrC inhibition (termination of sulfate reduction before the final two-electron reduction), or excess sulfite supplied to DsrAB relative to the capacity of DsrC to remove the reduced intermediates S^2- and S⁰. This is supported by previous work: the accumulation of thionate compounds, particularly thiosulfate, has been observed during bacterial growth both on sulfite and sulfate, as well as when electron supply to DsrC is slow (low electron transfer into the cell by the membrane complex [42-44]. Of the two thionates that can be formed during DsrC inhibition, thiosulfate is much more common [43].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. A diagram of the crystal proteins and the proposed pathway to make elemental sulfur.

Sulfite is formed during the first two-electron reduction of sulfate within a microbial cell (see Figure 1). Sulfite can be further reduced to sulfide through three two-electron reductions (pathway 1, sulfur species in red boxes). During this, sulfite binds to the DrsA-B and DrsC proteins, where it is sequentially reduced. It has been hypothesized that the DrsC protein plays a particularly important role in the terminal two-electron reduction of elemental sulfur to sulfide. Excess sulfite (blue box) in the cell can further attack S²⁺ and S⁰ while they are bound in the DsrA-B-C complex, forming thiosulfates (green boxes). We propose that these thiosulfates are disproportionated (or oxidized) forming elemental sulfur (yellow circles) and sulfite again, allowing for large amounts of elemental sulfur to accumulate (pathway 3). The specific conditions that permit this to happen are the supersaturated sulfate in the pore fluids coupled with low electron donor.

https://doi.org/10.1371/journal.pone.0075883.g005

What happens once thiosulfate (or generic thionates) are produced within the cell? Thiosulfate can be subsequently reduced by iron as follows: 4Fe²⁺ + S₂O₃^2- + 6H⁺ 4Fe³⁺ + 2S⁰ + 3H₂O. Alternatively, thiosulfate can undergo intracellular disproportionation. A pathway connecting the two species was first suggested by Jørgensen (1990) [45]: S₂O₃^2- S⁰ + SO₃^2-. This latter chemical reaction (disproportionation) also allows recycling of SO₃^2- and therefore repetition of this process to yield higher quantities of S⁰. Although not explicitly written in the former chemical reaction, thiosulfate dispropotionation requires an external electron donor as well, which could be Fe(II) [45]. Thamdrup et al. (1993) [46] suggest a close coupling between thiosulfate and sulfite in marine sediments; thiosulfate concentrations covary with sulfite concentrations when either is found, although neither are particularly abundant in marine sediments. Through the chemical reactions above, the net result of thiosulfate production through DsrC inhibition is the production of elemental sulfur. It might be expected that during bacterial sulfate reduction if excess thionates are produced then elemental sulfur could be an end product.

High production of thionate intermediates during bacterial sulfate reduction has been linked to very low rates of sulfate reduction, likely through the limited supply of electrons to the DsrC dimer preventing the final two-electron reduction [47]. It is possible, therefore, that if our hypothesis for the formation of elemental sulfur nodules is correct, that the rates of sulfate reduction could be very low during the formation of sulfur nodules. Brandt et al. (2001) [7] studied bacteria in the Great Salt Lake, USA, and concluded that very extreme salinities (12% in Great Salt Lake) helped result in low in situ sulfate reduction rates. Recent evidence, however, has suggested that hypersaline conditions in the Dead Sea have not impeded the microbially mediated sulfur cycle (Avrahamov, pers comm). There are many natural environments where sulfate reduction rates are exceptionally low, and sulfur nodules are comparatively rare. Therefore, while low sulfate reduction rates can aid in the formation of sulfur nodules through incomplete reduction to H₂S via the thionate pathway, it is likely that other factors are also at play in the Lisan Formation.

We hypothesize that the pore fluids that are supersaturated with respect to gypsum may play a role in the modified microbial metabolism and the resulting formation of elemental sulfur. Specifically, a high and sustained supply of aqueous sulfate (it is effectively an infinite reservoir because the pore fluids would have been saturated with gypsum) relative to a small supply of electron donor may poise the system such that intracellular sulfite concentrations are exceptionally high but the terminal reduction cannot occur and therefore the system produces high levels of elemental sulfur. Given the importance of intercellular sulfite on the oxygen isotope evolution of extracellular sulfate [48], this hypothesis could be tested using pore fluid measurements of δ¹⁸O_SO4 in modern gypsum bearing sediments where elemental sulfur is forming. Similarly, this intracellular branching point is also key for non-zero Δ³³S produced during different sulfur related microbial metabolisms [49]. Measurements of Δ³³S on the nodules could help resolve this speculation of the incomplete termination of bacterial sulfate reduction.

Our invoking the trithionate pathway for bacterial sulfate reduction to form elemental sulfur is consistent with the large sulfur isotope fractionation required by our data (Figure 4). Recent pure culture experiments have shown that at very low sulfate reduction rates and small supply of electrons, sulfur isotope fractionation can often exceed the 46‰ from the straight sulfite reduction classically proposed [28] (Figure 1). Therefore it is possible that at these slow rates of sulfate reduction, the thionates produced from the aborted trithionate pathway could be as much as 50‰ lighter than the precursor sulfite. During the further reduction of thiosulfate either through straight reduction or disproportionation, sulfur isotopes may be partitioned further. Therefore the large isotope difference between gypsum and nodule is consistent with the mechanism we propose. It is however, not inconsistent with the alternative hypothesis that sulfate was reduced all the way to sulfide and then reoxidized to elemental sulfur; the sulfur isotope fractionation during standard bacterial sulfate reduction can, in certain unique environments, be greater than 46‰ and there is very little isotope fractionation on sulfide oxidation back to elemental sulfur.

Given the changes in the iron and manganese concentrations we measure in the gypsum (Figure 2), we favor redox gradients within the aqueous pore fluids surrounding nodule formation that are captured in the trace element data within the gypsum. Our data suggest that at the rim of the nodule, iron is being produced and manganese is being consumed. Iron recycling at the rim itself would generate diffusive gradients of iron within the paleo-pore fluids and would draw iron towards the forming nodule. Specifically we favor the use of iron as an external electron donor coupled to either chemical reaction. Iron oxidation coupled to thiosulfate reduction or disproportionation could be occurring on a bio-film surrounding the area of nodule formation. This reduced iron (Fe(II)) could then be reoxidized through manganese reduction, driving a decrease in the manganese concentrations right next to the nodule. Reduced manganese (Mn(II)) is soluble and would diffuse away from the gypsum-nodule interface, creating the iron/manganese profiles similar to what is measured in modern pore fluids [1]. We measure the total iron and total manganese, therefore redox speciation is difficult to determine. However, the geochemical gradients are, at a minimum, suggestive of redox changes at the nodule-gypsum boundary.

While we prefer the explanation that bacterial sulfate reduction is terminated at elemental sulfur, our data is not wholly inconsistent with full bacterial sulfate reduction followed by partial reoxidation to elemental sulfur. The primary logic against this explanation is that we would need to explain the complete absence of pyrite. We believe that our data showing the very large sulfur isotope fractionation between the gypsum and nodule, coupled with the redox changes near the gypsum-nodule boundary, is more consistent with sulfate reduction terminating at elemental sulfur. It is possible also that some of our data are better explained by a second stage of diagenesis, post-nodule formation. For example, the red staining around the rim could be from the erosion of iron oxide, causing redistribution of some iron oxides to create the ‘halos’ seen (Figures 2 and 4). This could be through simple transportation in pore fluids during desiccation or reduction to water-soluble Fe²⁺ followed by re-oxidation. If the latter is true, the iron (III) reduction could be coupled to the oxidizing of some eroded S⁰.

The slight decrease in the δ³⁴S and δ¹⁸O towards the nodule may also be due to a second phase of diagenesis, post-nodule formation (for example, redistribution of small amounts of elemental sulfur into the gypsum or oxidation of small amounts of elemental sulfur to sulfate during lithification and desiccation). The average δ¹⁸O of the Lake Lisan was estimated through modelling to be around ~7‰ [50] so if this oxidation proceeded not in the presence of atmospheric oxygen, the resulting sulfate would be close to 7‰ as well (all oxygen atoms come from water). Elemental sulfur oxidation could be coupled to manganese or iron reduction if the redox boundary no longer existed post-nodule formation.

Sulfate reducing bacteria have broad ecological tolerances, and likely were some of the earliest metabolisms on the planet. If our mechanism for elemental sulfur formation is correct, then this is a modern example of how sulfate-reducing bacteria may have their biochemical pathway modified in response to the chemical conditions in unique environments (super saturated sulfate concentrations, low electron donor). There are also testable predictions from this work for sulfur and oxygen isotopes in modern sulfur nodule forming environments, as well as for the less abundant sulfur isotopes, which are sensitive to changes in sulfur flow through the microbial cell.