Keywords

1 Introduction

Organisms living below the euphotic zone in marine environments are dependent on the organic matter supplied to this environment. Here, the quantity and quality of organic carbon is in most cases the limiting factor for heterotrophic growth. In some cases, also oxygen becomes limiting for metabolic processes in deeper layers, mostly as an effect of reduced water circulation. These oxygen minimum zones (OMZ) are well-known phenomena in oceanography. It is assumed that the extension of hypoxic zones will increase especially in coastal marine waters due to both eutrophication and global warming [1] with severe effects on these ecosystems. The supply of both oxygen and organic carbon to layers below the euphotic zone of most marine systems is essential to maintain lives of higher organisms, as nearly all living organisms beyond the prokaryotic level depend on oxygen as terminal electron acceptor generating energy by respiration of organic material.

OMZs in the open ocean like in the Arabian Sea or in the central Pacific are under recent conditions not exhausting the total oxygen and nitrate inventory, as the organic input from the oligotrophic surface layer is comparatively low. Hypoxia in coastal waters (oxygen concentration below 2 mL L−1 or <90 μM) has a higher capacity of exhausting these resources of oxidants, and the system might turn to sulfate as electron acceptor because of higher biomasses, metabolic rates, concentrations of organics, and turnover of nutrients.

This shift is common within sediments, particularly in coastal areas [2]. Due to the reduced mobility of sedimentary pore-waters and the constant oxidation in surface sediments, this process remains largely unnoticed, since in most cases only the depth of the oxic–anoxic interface moves upward under reduced supply of electron acceptors. Most of the produced H2S in sediments will be buried as pyrite in anoxic sediments [3] as long as iron is in good supply, but certain amounts of it will diffuse upward. Depending on the organic load of the sediment and the reduced flux of electron acceptors from supernatant water, which is already free of oxygen, the export of reduced compounds from sediments can push the redoxcline into the open water. This leads to higher H2S concentrations in the overlying water than sulfate reduction processes within this water body could provide.

Pelagic redox gradients differ from benthic gradients, as they allow particulate phases to be sinking into deeper layers and therefore are selectively separating elements. In a lot of coastal and marginal seas, pelagic redox zones are unstable on a regular seasonal basis or triggered by hydrographic events. This sometimes leads to the transport of oxygen-free or even sulfidic water into the surface layer and provokes undesired effects on higher organisms including humans. The occurrence of hydrogen sulfide in the water column documents an advanced stage in the development of hypoxia due to a larger imbalance in the fluxes of electron acceptors and electron donors. The depths, where such pelagic chemoclines [defined as the first appearance of H2S (e.g., [4, 5])] are observed, depend on the equilibrium between the transport of electron acceptors and organic matter.

Although the term “suboxic” is widely used in oceanographic literature, the exact boundaries in terms of oxygen concentrations are rather diffuse. Values for the Black Sea range from 4.5 μM O2 (e.g., [6]) to 15 μM O2 [7]. More often a value of 10 μM is used [8]. Canfield and Thamdrup [9] argue for a stricter definition of this term, which is the base for a differentiation of biogeochemical processes. We advocate the value of about 10 μM O2 thereby integrating water layers with enhanced nitrification from upward flux of ammonia, so that the major coupled nitrogen conversion processes are included. Here, we define the upper border of a pelagic redox zone as the layer, where denitrification is not detectable (which otherwise is often marked by a secondary nitrite peak) and nitrification is the dominant chemolithotrophic process. This usually is the case at oxygen concentrations of around 10 μM. The lower boundary is within the sulfidic layer at a concentration of about 10–15 μM hydrogen sulfide. This sulfide concentration seems to be the upper limit for a peak of enhanced carbon dioxide fixation [10–12], which is supposed to have a functional relation to the redox gradient. The absolute vertical extension of the pelagic redox gradient varies between the different ecosystems and shows a seasonal variability. This is due to the different strength of vertical mixing that merges sulfide from the deeper and oxygen from the surface layers.

Seasonal hypoxia often occurs at ocean margins dominated by large rivers, in estuaries and in brackish marginal seas [13]. Since decades, pelagic chemoclines have been quite well investigated in marine areas such as the Black Sea [14–16], the Cariaco Basin [17–19], the deep basins of the Baltic Sea [20–22], and several fjords like Framvaren [23, 24]. One of the largest hypoxic areas, besides the central Black Sea, is found in the Baltic [25, 26]. Here, changing areas of hypoxic bottom water are accompanied by often perennial anoxic bottom waters in the central basins.

Pelagic redox gradients are much more affected by lateral transport processes than their sedimentary images. Dependent on the basin morphology of different marine areas with pelagic redox zones, there are distinct depths at which lateral intrusions of water can occur, which then cause fluctuations in the depth of the oxycline or instabilities above the redox zone. Examples of these lateral intrusions are reported as inflow events into the western Black Sea [27, 28], in the Gotland Basin of the Baltic Sea [29], and in the Cariaco Basin [30].

A unique property of oxic/sulfidic pelagic redox zones is a pronounced peak in dark carbon dioxide fixation which was found in any of the investigated gradients. First, Sorokin [14] measured carbon dioxide fixation in the redox zone of the Black Sea and attributed it to bacterial chemosynthesis fueled by oxidation of reduced sulfur compounds. It took about 15 years until this feature was observed in Cariaco Basin [31]. Earlier investigations by the same authors showed the occurrence of sulfide- and thiosulfate-oxidizing bacteria at the redox zones of the Black Sea and the Cariaco Basin, and also their potential to use thiosulfate for dark fixation of carbon dioxide [32]. But up to now any attempt to explain the high rates of carbon dioxide fixation quantitatively failed. We therefore want to review and discuss the existing data and hypotheses concerning this process focusing on three topics:

  1. (a)

    Microbial communities across pelagic redox gradients

  2. (b)

    Flux of electron acceptors and donors as source of energy

  3. (c)

    Observed carbon dioxide fixation and quantitative explanation

2 Microbial Communities Across Pelagic Redox Gradients

Although in most pelagic redox zones gradients of temperature and salinity provide the density gradient, which inhibits vertical mixing, these gradients itself are rarely so strong that they exert direct effects on biological diversity and activity. The resulting gradients of the different electron acceptors and donors are much more directly influencing the abundance of bacteria across these water layers. There is, however, not a general pattern in bacterial biomass and diversity. Ho et al. [33] reported a significant increase in bacterial numbers at the suboxic/anoxic interface which at least in some cases showed more than twice the number of cells compared to the layers above or below. Similar conditions were later reported by [34, 35]. Bacterial abundances in the pelagic redox zone of the Black Sea are much more variable compared to the layers above and below. Bird and Karl [36] could only find an “absence of a general enrichment in total microbial or bacterial biomass across the oxic-to-anoxic boundary”. Sorokin et al. [37] reported a stock of bacterial numbers and biomass within the redox zone, which was nearly twice as high as that in the water layer between redox and thermocline (40–100 m). Sometimes, the differences in bacterial numbers between these layers do not seem to be so high [38–40]. In pelagic redox zones of the Cariaco Basin and the Black Sea, bacterial cell numbers between 0.3 and 0.5 × 106 mL−1 were most often found, although these numbers range from less than 0.2 to more than 1 × 106 mL−1 for the Black Sea and the Cariaco Basin, respectively. Bacterial numbers are about twice as high for Baltic Sea redox gradients (mean between 0.8 and 1.2 × 106 mL−1, ranging from 0.5 to more than 2 × 106 mL−1) (e.g., [41–44]).

During transition from oxic-to-anoxic conditions in the Gotland Basin, also higher cell volumes of bacteria with increasing depths were reported [37, 41]. Especially in the Black Sea, the occurrence of filaments at the redox zone was reported [39], but this was already observed in samples from sulfidic water layers by Kriss [45].

Estimates of bacterial production in the different redox zones were published infrequent and also performed with quite different methods. From the values provided in Table 1, we conclude that the earlier estimations [37, 47] were probably too high. Growth rates of about 0.1 day−1 seem to be in the mean range. In the Baltic Sea, the bacterial production and the growth rates were higher as the values reported more recently from the Black Sea and Cariaco Basin (Table 1; [46]).

Table 1 Bacterial growth estimations around redox zones of the Black Sea, Cariaco Basin, and Baltic Sea
Full size table

The DOC concentration is slightly decreasing with depth in the Black Sea [48], the Cariaco Basin [33], and the Baltic Sea [49]. Since most of the prokaryotic organisms are heterotrophs, they gain decreasing amounts of energy over the redox gradient by respiration of the same amount of dissolved organic matter, because of the decreasing energy gain by using different electron acceptors. This combination should result in at least slightly decreasing bacterial productivity with depth.

Within all three pelagic oxic–anoxic transition zones, mentioned above, pronounced shifts in the composition of bacterial communities are observed [4, 19, 50, 51]. Within the prokaryotic organisms, Eubacteria are dominant. Also common for all of these areas are higher proportions of Epsilonproteobacteria at the redox zone, accounting for up to 27% and 20% of the DAPI counts for the Cariaco Basin and the Black Sea, respectively [50]. For the Baltic Sea redox zone, up to 16% of the DAPI counts were related to the Epsilonproteobacteria [5]. For the Cariaco Basin and the Black Sea, comparable, but low abundances of Gammaproteobacteria of around 2–4% of the DAPI counts were reported [50], whereas in both regions peaks of around 8% were detected just above the chemocline. Using fingerprinting techniques, Labrenz et al. [51] could detect bands from Alpha-, Gamma-, Delta-, and Epsilonproteobacteria from the suboxic to the sulfidic layer at the central Baltic redox zone.

In many marine systems, also Crenarchaeota are at least partly involved in ammonia oxidation. Crenarchaeota dominate the variable proportions of archaea in all basins [50, 51]. Especially in the suboxic part of the redox zone, these Crenarchaeota seem to be mainly composed of nitrifying organisms related to the Candidatus “Nitrosopumilus maritimus” [52, 53]. These nitrifying Crenarchaeota seem to be well adapted to low oxygen conditions in marine environments [54–56].

Their contribution to ammonia oxidation was proven for the Black Sea [6, 52] and indicated for the Cariaco Basin [50]. In the suboxic part of the Baltic Sea redox zone, a peak of transcripts of amoA from Crenarchaeota just above the chemocline supports the importance of these organisms versus ammonia-oxidizing bacteria [53]. These Crenarchaeota were recently detected in nearly all, especially suboxic, marine waters [54, 57, 58].

The existence of chemolithoautotrophic bacteria in pelagic redox zones is acknowledged since decades. Already Kriss [45] described the filamentous bacteria found in sulfidic waters as autotrophs oxidizing hydrogen sulfide. He, however, assumed that they used the energy from radioactive disintegrations. First measurements of chemosynthetic production by autotrophs at the pelagic redox zone of the Black Sea were reported by Sorokin [14], who addressed it to thio-oxidizing bacteria. Tuttle and Jannasch [59] reported the isolation of bacteria from the Black Sea oxidizing different reduced sulfur compounds, capable for autotrophic growth as well as for surviving under anoxic conditions. They later repeated that with material from the pelagic redox zone of the Cariaco Basin [32].

Although at least since the late 1980s a peak of elevated dark carbon dioxide fixation at the redox zone of the Baltic Sea is known [60, 61]. Until the work of Labrenz et al. [62] resulted in enrichment cultures, there are no attempts known to isolate these chemolithoautotrophic bacteria.

Since only a few metazoans are adapted to very low oxygen concentrations or even anoxic waters, protozoa seem to be the only organisms left to fulfill the function of grazers, preying on either other protozoa, flagellates, or prokaryotes [7]. Their abundance decreases with increasing concentrations of H2S. In the central Baltic redox zone, ciliates instead of heterotrophic nanoflagellates seem to be the major grazers of bacteria within the suboxic part of the redox zone showing a peak at the oxic–anoxic interface [63]. Already Detmer et al. [61] reported a peak of ciliates at the chemocline followed by a drastic decrease further downward in the sulfidic water. This results in a decrease in grazing pressure on the prokaryotic populations over the pelagic redox zone, which might at least compensate the effect of reduced bacterial productivity on the standing stock of prokaryotes. A more pronounced increase in prokaryote abundance below the redox zone might be limited by increased mortality due to bacteriophages [46]. This might be also supported by a higher ratio of virus-like particles to bacteria as found in the anoxic part of the Cariaco Basin [34].

3 Flux of Electron Acceptors and Donors as Source of Energy

A shift in the composition of bacterial communities across oxic–anoxic transition zones is not surprising since the main electron acceptor oxygen and is gradually replaced by other oxidants. The next energetically favorable electron acceptor is nitrate, followed by oxidized iron, manganese, and, finally, sulfate. This microbial redox sequence based on the energy gained by the different electron acceptors describes principally the temporal succession of the exhausted electron acceptors and their spatial disappearance [2]. While such cascades in sediments happen often within several millimeters or centimeters, we have to assume spacial scales of decimeters for strong gradients like in several fjords [24] to several meters like in the Cariaco Basin [11]. The availability of the electron acceptors will cause a transition from aerobic via facultative anaerobic to anaerobic microbial metabolisms. Following this succession, not only the energy gain for bacterial growth by alternative electron acceptors will be reduced drastically, but also alternative bacterial metabolism becomes dominant as reflected by the shift in the community structure.

Below we will focus our attention on compounds that are known to fuel chemolithoautotrophic growth of microorganisms. A number of different bacteria are specialized to gain their energy by oxidation of inorganic or one-carbon compounds and assimilating carbon dioxide [64]. The number of compounds that could react as electron acceptors and/or as electron donors for microbial carbon dioxide fixation is relatively limited. A list of the most probable compounds used by chemolithotrophic microorganisms is shown in Table 2. We will focus our attention on oxygen and nitrate as the most important electron acceptors. Less-oxidized nitrogen compounds can be detected in significantly lower concentrations during different redox transitions as intermediates and can therefore be included in the nitrate inventory. Particulate manganese oxides are important due to their different transport behavior. In contrast to dissolved substances that are dependent on enhanced eddy diffusion for their vertical distribution, these manganese particles are transported much faster by sinking [22, 65]. The highest energy gain can be achieved by chemolithoautotrophic oxidation of hydrogen or methane by oxygen or even nitrate [64], but both compounds can only be found close to the detection limit (hydrogen) or in the nMol range at pelagic chemoclines ([66], Heyer, personal communication). Therefore, the contribution of these oxidation pathways can be considered a minor contribution to the observed carbon dioxide fixation. According to the measured concentration gradients and calculated fluxes, this leaves only ammonia, sulfide, and manganese to contribute substantially to chemolithotrophy. Other more oxidized sulfur compounds such as elemental sulfur and thiosulfate can be included in the sulfide budget since they remain most probably intermediates within the redox zone. According to the vertical profiles, a vertical transport of these compounds from sediments to the pelagic redox zone in the Baltic Sea [67] or in the Cariaco Basin [68] is rather unlikely.

Table 2 Inorganic electron acceptors and/or donors which could enable chemolithoautotrophic bacterial growth
Full size table

Biological turnover in the redox gradient is largely dependent on the downward flux of the most oxidized and upward flux of the most reduced chemical species into the pelagic redox zone. Intermediates that can be used as either electron donor or acceptor are mainly involved in cyclic processes within the system and not relevant for import or export.

The successive disappearance of the different electron acceptors with depth following the decreasing amount of energy gained by their use is well known from sediments. Like in sediments, the decline of oxygen with depth in pelagic redox zones is as well followed by a decline of nitrate. Denitrification already starts under suboxic conditions. For oceanic denitrification, oxygen concentrations above around 2 μM seem to be inhibiting [69, 70]. In the suboxic water of the Gotland Basin, denitrification was detectable at oxygen concentrations below 9 μM [71]. But already Brettar and Rheinheimer [21] reported that heterotrophic denitrification in the suboxic water column was hard to measure and assumed that the main nitrogen loss process is coupled to oxidation of reduced sulfur compounds. They concluded that heterotrophic denitrification is limited especially by the quality of available organic matter [21, 72]. The importance of chemolithotrophic denitrification as the main loss factor for nitrogen at pelagic redox zones of the Baltic Sea was also stressed by Hannig et al. [73].

This lack of electron acceptors leads to a decrease in denitrification capacity of the ecosystem. In sedimentary environments with high organic loads, nitrate penetrates from the bottom water into the sediment where it is reduced at the depth where oxygen becomes limited. Even ammonia produced during remineralization will be oxidized within the sediment before it will be released as nitrate or denitrified to dinitrogen. In most pelagic redox zones, the lack of available organic material for heterotrophic denitrification is also the limiting factor for microbial heterotrophic activities. This might, at least in the central Baltic, explain why there are no drastic changes in bacterial numbers and growth rates over the whole redox zone, which should be expected by the drastic changes of energy yield in the use of different electron acceptors. It as well implies the importance of chemolithoautotrophy for life at pelagic redox zones.

Many organisms are known to oxidize manganese and in theory it should be sufficient to support chemolithoautotrophic growth. Real proof is lacking, but as genes for CO2 fixation were found in a manganese oxidizing bacterium [74], this pathway seems very likely. The energy gain will, however, be relatively low. Under oxic conditions, these microorganisms should be outcompeted by others using oxygen more efficiently for growth, although the indicated lack of degradable organic matter (lacking heterotrophic denitrification) would impede both denitrifiers and other heterotrophic bacteria. There are other possible competitors for oxygen-nitrifying bacteria, manganese oxidizers, and other chemolithotrophs which oxidize reduced sulfur compounds, as they are not in need of organic substrates (Table 3). In Black Sea studies, a coexistence of reduced sulfur and oxygen is often negated [81]. In this case, there should be differences in manganese oxidation rates, as the concurrent process of sulfur oxidation is omitted. Rates of Mn oxidation should be higher in the Black Sea, as compared to the Baltic Sea and the Cariaco Basin, where the overlap between oxygen and reduced sulfur compounds is the standard situation, and therefore more favorable options for the exploitation of oxygen are present.

Table 3 Examples of redox reactions which could fuel chemolithoautotrophy within different layers of the redox zone
Full size table

According to our present knowledge of chemolithoautotrophic bacteria and the distribution of electron acceptors, we would expect their main activity around the oxic–anoxic interface. Since oxidation pathways with nitrogen species provide nearly the same energy yield as oxygen-based processes and due to the fact that these compounds are also dissolved, they should be depleted within a relative short distance after the disappearance of oxygen. The only possible electron acceptors which could enter deeper layers by sedimentation can be oxidized iron and manganese particles.

4 Observed Carbon Dioxide Fixation and Quantitative Explanation

The gradient system of oxidized and reduced inorganic compounds within redox zones seems to be a perfect place for chemolithotrophic organisms. Most of them use either oxygen or oxidized nitrogen compounds as terminal electron acceptors oxidizing reduced compounds. Therefore, we can expect their peak in activity at the oxic–anoxic interface until oxidized nitrogen compounds disappear. There are few reports from different pelagic redox zones that support this hypothesis [31, 61, 82]. But more often the maximum of the carbon dioxide fixation is located further down in the anoxic part of the redox zone [11, 12, 14, 83]. This vertical structure as well as the amount of carbon dioxide fixation was explained quite differently. At first, Jørgensen et al. [10] recognized the difficulty to explain the values of chemolithotrophic carbon dioxide fixation quantitatively. Although they mentioned the possible contribution of anoxygenic phototrophic bacteria, they ruled out that they could contribute a substantial part. Since in most of these pelagic redox zones light is of minor importance (except shallow fjords), we might rule out the importance of photosynthesis as an important process in the major marine pelagic redox zones of the Cariaco Basin, the Black Sea and the Baltic Sea. Even if there are some reports about photosynthetic activity of anaerobic bacteria at the chemocline of the Black Sea [84–86], their contribution to the measured carbon fixation (e.g., [10, 14, 37]) can only be of minor importance.

There are still two more options to explain dark CO2 fixation. Heterotrophic bacteria might take up carbon dioxide by anaplerotic reactions. It is assumed that this could account for up to 8% of bacterial carbon fixation [87, 88]. Under anaerobic conditions, this percentage seems even to be slightly higher [89]. As in all pelagic redox zones, rates of dark CO2 fixation are not coupled to heterotrophic bacterial production and often show comparatively low values in the suboxic part, where the heterotrophic bacterial production is higher than in the anoxic part of the redox zone, and anaplerotic CO2 fixation cannot explain the observed rates.

The only remaining choice of probable agents for this process is then narrowed down to chemolithoautotrophic organisms. Isolation of chemolithoautotrophic bacteria from the respective layers can proof the presence of this type of organisms but not their quantitative role. First evidence for the importance of these bacteria came from the Baltic Sea. Jost et al. [44] demonstrated the existence of about 20–40% of chemolithoautotrophic bacteria within the redox zone fixing between 3.5 and more than 20 fg C per cell per day. According to their calculations, 0.8 fg C should be the upper limit, which could be fixed by heterotrophic bacteria within these samples. At least one group of chemolithoautotrophic bacteria could be easily discerned from others by relatively high 90° light side scatter signals in their flow cytometric signature. This unique cluster of bacteria just appeared at the lower transition of the chemocline to sulfidic waters [44]. Casamayor et al. [90] observed such a shift in the flow cytometric signature of phototrophic bacteria when they accumulated internal elemental sulfur. Such a morphological signature may also explain the appearance of the optically active bacterial cluster below the chemoclines of the Baltic Sea, which was also observed in the Black Sea (unpublished results). The existence of chemolithoautotrophic bacteria was also detected by MICRO-CARD-FISH analyses at redox zones of the Baltic Sea and Black Sea [43]. This investigation verified not only the importance of chemolithoautotrophic bacteria, but also showed that their activity was nearly 100% restricted to Eubacteria below the chemoclines. As already suggested by Lin et al. [19], it could be proven that Epsilonproteobacteria play a major role in CO2 fixation at pelagic redox zones. Between 70 and nearly 100% of the detected chemolithoautotrophic bacteria could be identified as Epsilonproteobacteria by MICRO-CARD-FISH probes [43]. Studies by Glaubitz et al. [82, 91] render it likely that the remaining chemolithoautotrophs belong to Gammaproteobacteria both in the Baltic and in the Black Sea. Although their contribution to the estimated CO2 fixation was less than half of the amount fixed by Epsilonproteobacteria, the chemolithoautotrophic Gammaproteobacteria were more diverse than the Epsilonproteobacteria [82]. At least one group of these Gammaproteobacteria is closely related to a symbiont, and it is also known that sulfide-oxidizing bacteria appear as ectosymbionts of protozoa in sulfidic environments [92–95]. The observed transfer of the CO2 fixation by bacteria to protozoa could be related not only to unspecified grazing but also to grazing of own ectosymbionts [91]. The closely related Epsilonproteobacteria belong to Sulfurimonas sp. and proved them to be versatile key players within pelagic redox zones by microdiversity analyses [5, 43, 82, 91]. These approaches for direct detection and identification of chemolithoautotrophic organisms were used in layers showing enhanced CO2 fixation. For these layers, a clear dominance of the oxidation of reduced sulfur compounds to fuel chemolithoautotrophy could be demonstrated. Here, it can only be speculated, whether this is also the case for the upper suboxic part of the redox zone. In this environment, at least the presence of Epsilonproteobacteria was demonstrated [5], and according to Kamyshny et al. [67] the presence of reduced sulfur compounds is as well proven.

In the suboxic part of the redox zone, alternative reactions allow chemolithoautotrophy. The importance of nitrifying bacteria within suboxic environments is well known [96]. Ammonia oxidation is restricted to at least suboxic environments. In the Baltic Sea, this process is indicated by the appearance of a nitrate peak below the halocline [51, 97], which should be related to highest ammonia oxidation activities. In this layer, CO2 fixation rates are, however, near the detection limit (Jost, unpublished results) and can be assigned only to the activity of chemolithoautotrophic ammonia oxidizers [12]. This view will probably be modified by new findings that Crenarchaeota are as well involved in the marine nitrogen cycle. It is assumed that the marine Crenarchaeota are also autotrophs [54, 98], and for the Black Sea it was suggested that both Gammaproteobacteria and Crenarchaeota contribute to nitrification [6]. Because of different pathways of carbon fixation in bacterial and archaeal metabolism [99], their fixation rate per oxidation equivalent might slightly differ. Even if these organisms were just partly autotrophic or mixotrophic, as was assumed from crenarchaeotal uptake of amino acids [100, 101], their respective metabolism cannot change the overall budget within the suboxic layers of pelagic redox zones substantially, as CO2 fixation rates are generally low. The calculated rates of carbon dioxide fixation by chemolithoautotrophic ammonia-oxidizing bacteria are not only in good agreement with the low rates measured in the suboxic part of the pelagic redox zone of the Baltic Sea, but can also explain the disappearance of ammonia around the chemocline [12].

The contribution of anammox bacteria (oxidation of NH4 with NO2 to N2 and H2O) to CO2 fixation seems as well to be of minor importance, as their energy gain is in the same range as that of aerobic ammonia oxidizers [102]. Since these bacteria use nitrate or nitrite under anoxic conditions with an energy yield as low as nitrifying bacteria, they cannot contribute more to CO2 fixation than the nitrifiers. Earlier, it was assumed that anammox would be restricted to anoxic and non-sulfidic environments, since it was completely inhibited already at oxygen concentrations as low as 2 μM [103]. But then it was found that anammox still proceeds under low oxygen concentrations [104], though under suboxic conditions anammox rates are lower than under anoxic conditions [105]. Lam et al. [6] postulated a short circuit between nitrification and anammox for the suboxic to upper anoxic layers of the redox zone of the Black Sea since they could not detect any denitrification. This implies that anammox bacteria could outcompete denitrifying bacteria. Since the energy yield of the anammox bacteria per mol nitrite/nitrate is about ten times less than that of heterotrophic denitrifiers or even chemolithoautotrophic denitrifiers using reduced sulfur compounds, this is not very probable in a concurrent situation. In case it occurs, it indicates a lack of suitable organic matter or reduced sulfur compounds. Under these conditions, the highest rates of CO2 fixation should then be found in the deeper part of the suboxic layer and within the upper part of the anoxic layer. But here enhanced fixation rates of chemolithoautotrophic denitrifiers are not probable because of lacking nitrite/nitrate. Such a scenario seems, however, quite possible in an environment like the Black Sea, where an extended anoxic layer without detectable hydrogen sulfide seems to be existing at least temporarily [106].

In the Baltic Sea anammox could only be detected occasionally, rates were in the range of 0.05–0.005 μmol N L−1 day−1 [73]. This was attributed to higher Mn(IV) concentrations, which developed after intense salt water inflow led to a short period of fully oxygenated bottom water in the Gotland Basin. In a newly establishing pelagic redox zone extending from the sediment, relatively large amounts of sinking manganese oxide and its potential to oxidize sulfide seemed to be able to establish a temporarily zone, free of oxygen and H2S. It should, however, be noted that these anammox measurements are based on the addition of nitrate to anoxic water samples without detectable in situ nitrate concentrations. Therefore, induction or at least an enhancement of the measured rates must be assumed.

The basic process that fuels all chemolithoautotrophic activities is basically driven by the stepwise oxidation of sulfide with oxygen over the total redox gradient. But all calculations to explain the measured carbon dioxide fixation by sulfide oxidation failed so far [10–12, 24]. First, there is a lack of sulfide reaching the zone of enhanced carbon dioxide fixation by upward transport in all of the investigated systems, and second, the peak of the fixation zone is often within the sulfidic part of the redox zone without detectable amounts of oxygen and oxidized nitrogen, which are the most important electron acceptors for chemolithoautotrophic sulfur-oxidizing bacteria.

Although even in suboxic parts of pelagic redox zones sulfate-reducing bacteria were detected in the Baltic Sea [51], the Black Sea [50, 107], and the Cariaco Basin [35, 50], it seems very unlikely that these bacteria supplied enough, if any, sulfide to produce the required amount of sulfide to fuel the process. Albert et al. [108] measured production rates of up to 3.5 nmol H2S L−1 day−1 below the pelagic redox zone of the Black Sea, but also deeper than the zone of elevated carbon dioxide fixation. Within this zone, Jørgensen et al. [10] measured values between 3 and 36 nmol H2S L−1 day−1. Since these rates were two orders of magnitude lower than measured sulfide oxidation rates within the same layer, sulfate reduction cannot provide an important source for sulfide [10]. Jost et al. [12] suggested a recycling of sulfide from only partly oxidized sulfur compounds like elemental sulfur or thiosulfate. Although the calculated turnover time of the sulfide pool seems with about 1 day comparatively low, this still needs to be verified. The most critical point in this argumentation is to explain the origin of the organic matter for the heterotrophic bacteria using thiosulfate or elemental sulfur as electron acceptors, as the quality and amount of organic matter in layers only several meters above are not even sufficient to fuel heterotrophic denitrification.

Since oxidized nitrogen compounds can only explain a small part of the observed carbon dioxide fixation by sulfur oxidizers, the question for alternative electron acceptors within the sulfidic environment remains open. Most likely seems to be the involvement of particulate iron and manganese [10–12]. But the amount of oxidized metals which could reach these anoxic layers would also by far be not sufficient to explain the observed carbon dioxide fixation rates [12]. It is well known that humic substances could mediate the electron transfer between particulate metal oxides and bacteria [109]. It has also been shown that humic substances could be used as electron acceptors for the oxidation of organic substances [110]. It is, however, unknown whether these substances could also be used to oxidize reduced sulfur compounds by chemolithoautotrophic bacteria. But even if this would be possible, these electron acceptors would needed to be recycled, e.g., by metal oxides, as otherwise their capacity to accept electrons would degrade quickly.

At the moment, we cannot provide a satisfying explanation for the extent of the carbon dioxide fixation rates especially in the anoxic part of the redox zones. Lateral intrusions were often assumed to supply the redox gradients with additional oxygen [5, 11, 28]. But even if this amount would be large enough, the capacity of the dissolved metal inventory would not be sufficient to transport it as equivalents into the anoxic part, which is characterized by enhanced carbon dioxide fixation.

Although a suite of precautionary measures were taken to eliminate oxygen contamination of the incubation samples by all groups engaged in these experiments, an ultimate guarantee cannot be given. Even the observed progression of the carbon dioxide fixation maximum in the anoxic layer could be explained by assuming that every sample might have been contaminated with about the same amount of oxygen. Then, depending on the amount of reduced sulfur compounds, an increasing amount of carbon dioxide may be fixed, as long as enough microbes capable of oxidizing reduced sulfur compounds are present. In deeper layers with higher sulfide concentrations, their cell numbers will either decline or the bacteria might need longer lag phases to reestablish their physiology. Due to the fact that intense care has been taken, to keep oxygen out of the experiments, this scenario is not very probable but cannot be ruled out completely.

All published results showing enhanced carbon dioxide fixation within anoxic environments without nitrite/nitrate are based on incubations of water samples which have been transferred through oxic environments before incubation with 14C- or 13C-bicarbonate solution. Incubation-independent measurements like the estimation of stable isotopes within fatty acids show significant changes only at the layer around the chemocline and not in the sulfidic part [91]. At present, the trade-off of these arguments leaves us to conclude that there is at least the capacity for anoxic carbon dioxide fixation. The rates might, however, be overestimated by contaminations during incubation procedures. We therefore advocate the improvement of techniques toward in situ rate measurements and the integration of in situ molecular approaches like metatranscriptomics or -proteomics regarding carbon dioxide fixing key enzymes.