Photosynthesis is the main primary production process fueling life on Earth. It requires light as energy source, inorganic carbon to be fixed, and a source of electrons (Hamilton et al., 2016). While oxygenic photosynthesis is currently the dominant process for fixing inorganic carbon into organic matter, this has changed during the history of life (Olson and Blankenship, 2004). Very different conditions existed on early Earth (Canfield et al., 2006), as the chemical environment (e.g., the availability of electron acceptors and donors for biogeochemical processes) and, consequently, the favored biogeochemical processes, differed in the ancient biosphere from those currently prevailing. Even though rates of marine primary production were much lower in these primitive seas than in modern oceans, the most active ecosystems were probably driven by cycling of H₂ and Fe(II) (Canfield et al., 2006), though other photo- and chemolithoautotrophic processes could also have contributed [e.g., non-photosynthetic Fe(II) and sulfide oxidation]. Still, the early appearance of photosynthesis enhanced biological productivity by orders of magnitude compared to metabolisms based on acetogenesis and methanogenesis as the ancestral forms of carbon and energy metabolisms (Sleep and Bird, 2008; Sousa et al., 2013). Some modern ecosystems (the so-called “analogs,” Burns et al., 2009) still show determinant similar features (i.e., redox conditions, iron and sulfur chemistry) to those predominating on ancient Earth, and offer opportunities to study the processes that sustained microbial life in the primitive biosphere. Photoferrotrophs, which use Fe(II) as an electron donor and light as an energy source for inorganic carbon fixation (Widdel et al., 1993), as well as their biogeochemical role in modern and ancient ferruginous systems, are particularly addressed in this review.

Once liquid water appeared on Earth around 4.3 Ga ago (Mojzsis et al., 2001), life could emerge and develop in the oceans of the Archean (4 to 2.5 Ga ago) not later than around 3.8 Ga ago (Mojzsis et al., 1996) as evidenced from biologically fractionated organic carbon (Nutman et al., 2016), or even earlier (Abramov and Mojzsis, 2009). The energy required for inorganic carbon fixation was available from sunlight, and from the oxidation of inorganic chemical substrates by chemolithoautotrophs using inorganic electron acceptors. Both photo- and chemolithoautotrophy were, consequently, an option for primary production on the primitive Earth (Canfield et al., 2006). The advent of photosynthesis, however, relieved life from its dependence on the co-occurrence of reduced and oxidized inorganic compounds, as required for chemolithoautotrophy. Indeed, their co-availability is often limited by a high chemical reactivity; i.e., if oxidized and reduced substances readily react chemically, their availability for microbially mediated energy-gaining redox processes diminishes. In this context, light offers the advantage of being an energy source that is not dependent on such co-availability, therefore extending the potential niches for life colonization, although in water columns its availability is also limited to the surface photic layers.

Photosynthesis appeared in the early Archean (Olson, 2006; Fischer et al., 2016), and photosynthetic microbial mats populated benthic environments of oceans shores by around 3.4 Ga ago (Tice and Lowe, 2004), though stromatolites formed around 3.7 Ga ago have also been recently reported (Nutman et al., 2016). It is generally recognized that anoxygenic photosynthesis evolved before the more complex cyanobacterial-type oxygenic photosynthesis (Xiong et al., 2000; Xiong, 2006; Schopf, 2011; Gupta, 2013). The latter is thought to have appeared around 2.75 Ga ago (Buick, 2008), but isotope-based and other data suggest that it may have been present even much earlier (Rosing and Frei, 2004; Planavsky et al., 2014). Some studies, for instance, provide evidence that oxygen was already present in the late Archean (ca. 2.7 Ga) surface environment, supporting oxidative elemental cycling (Anbar et al., 2007; Kendall et al., 2010; Stolper et al., 2010; Crowe et al., 2013; Lalonde and Konhauser, 2015; Frei et al., 2016). These oxygen traces or the consequences of these oxidative processes could have resulted from abiotic reactions such as CO₂ photodissociation (Lu et al., 2014) and H₂O₂ disproportionation (Haqq-Misra et al., 2011), but also from biological processes like benthic oxygenic photosynthesis (Lalonde and Konhauser, 2015). Atmospheric oxygen accumulation, however, did not occur at a global scale before the Great Oxidation Event around 2.33 Ga ago (Bekker et al., 2004; Kaufman et al., 2008; Konhauser et al., 2009; Lyons et al., 2014; Luo et al., 2016). Instead multiple evidences suggest that photoferrotrophy could have been a relevant photoautotrophic process in the ancient biosphere.

During most of Earth history the ocean was anoxic, and its chemistry was influenced (Figure 1) by the activity of microbial life (Blake et al., 2010). Reduced iron [Fe(II)], of hydrothermal origin (Klein, 2005), from biologically processed continental sources (Li et al., 2015), and/or released by tectono-magmatic events (Konhauser et al., 2007a), dominated the mesophilic (Hren et al., 2009; Blake et al., 2010; Poulton and Canfield, 2011), sulfur-poor (Shen et al., 2003; Crowe et al., 2014b) Archean ocean chemistry (Figures 1, 2). From these ferruginous conditions a transition toward a more sulfidic (euxinic) ocean occurred from the late Archean to the Mesoproterozoic (Poulton et al., 2004; Canfield et al., 2008; Reinhard et al., 2009), though sulfide was likely spatially confined to parts of the ocean (Reinhard et al., 2013). Modern analogs of these ancient euxinic environments still exist, such as the Black Sea (e.g., Overmann et al., 1992; Meyer and Kump, 2008), or sulfide-rich meromictic (Noguerola et al., 2016) and holomictic (e.g., Camacho and Vicente, 1998; Camacho et al., 2000, 2001; Camacho, 2006) lacustrine basins. In this sulfide-richer environment of the Proterozoic, sulfide-driven anoxygenic photosynthesis acquired a more relevant role as a primary production process (Kharecha et al., 2005; Johnston et al., 2009). However, ferruginous conditions probably persisted in some zones of the oceans throughout the Proterozoic (Poulton and Canfield, 2011) and even transiently dominated again deep-water chemistry in the Neoproterozoic (Canfield et al., 2008; Planavsky et al., 2009, 2011), though these deep layers were probably aphotic. Later on, during the Phanerozoic (0.54 Ga ago), the ocean became fully oxygenated (Holland, 2006; Hamilton et al., 2016; Figure 1). In any case, during the first stages of life, reduced iron dominated ocean chemistry, and microorganisms with iron-based metabolisms were likely important biogeochemical actors on the primitive Earth (Poulton and Canfield, 2011; Kendall et al., 2012; Llirós et al., 2015).

In contrast to euxinic conditions, which can be identified in the geological record by a set of proxies (trace metals, isotopes, and lipid biomarkers), the identification and mineralogical characterization of sedimentary iron enrichments are used to reveal ferruginous deposition conditions (Poulton and Canfield, 2011). Banded iron formations (BIF) are rocks of alternating layers rich in silica with layers rich in iron oxides (Posth et al., 2008; Walter, 2011; Frei et al., 2016). BIF occur in geological records of several periods of the Archean (4 Ga to 2.5 Ga) and the Proterozoic (2.5 Ga – 0.54 Ga), especially between the Neoarchean (late Archean) and Early Paleoproterozoic – the Siderian – (2.7–2.4 Ga, Pecoits et al., 2015). Both chemical and biological processes have been proposed as possible mechanisms for BIF genesis in the different time periods (Figure 2). These include (Cairns-Smith, 1978; Johnson et al., 2008; Posth et al., 2013): (i) Fe(II) photo-oxidation by UV light (François, 1986; Klein, 2005), or by photochemically produced H₂O₂, although the significance of these processes seems rather low (Konhauser et al., 2007a; Pecoits et al., 2015); (ii) chemical oxidation by dissolved oxygen of photosynthetic origin (Holland, 2006; Kump, 2008), and, hypothetically, could even include; (iii) abiotic oxidation of Fe(II) by microbially produced oxidized compounds such as nitrogen oxides (e.g., Klueglein and Kappler, 2013), though the sufficient availability of these oxides to support a visible contribution of this mechanism is yet to be shown. However, (iv) direct oxidation by microbial processes (Garrels et al., 1973; Hartman, 1984; Li et al., 2011; Posth et al., 2013; Czaja et al., 2013; Chan et al., 2016), mainly iron-oxidation by anoxygenic phototrophs using Fe(II) as electron donor, seems an attractive mechanism for early BIF formation, particularly for the time before oxygenic photosynthesis appeared on Earth (Ehrenreich and Widdel, 1994a; Konhauser et al., 2002, 2005, 2007a,b, 2011; Kappler and Newman, 2004; Kappler et al., 2005; Croal et al., 2009; Fru et al., 2013; Li et al., 2013; Eickhoff et al., 2014; Sun et al., 2015).

Iron is the most abundant metal on Earth, and among the most abundant elements in the Earth’s crust (Kendall et al., 2012). Under circumneutral pH conditions Fe(II), particularly in its ionic form (Fe²⁺), it is rapidly oxidized by oxygen to highly insoluble Fe(III) oxides. As a consequence iron can become a limiting nutrient for primary producers in vast areas of the well-oxygenated modern ocean, where iron concentration is typically <1 nM (Martin and Fitzwater, 1988). Under anoxic reduced conditions Fe(II) is the dominant iron form. Ionic Fe²⁺ is several orders of magnitude more soluble than Fe(III) (Stumm and Morgan, 1995; Cornell and Schwertmann, 2003) and can reach high concentrations in ferruginous water columns (Table 1) and in the pore water of freshwater sediments (e.g., Bravo et al., 2015). In sulfur-rich environments iron reacts with dissolved sulfide forming highly insoluble iron-sulfide minerals (e.g., Zopfi et al., 2004).

The biological significance of iron is based on its redox behavior, with Fe(III) and Fe(II) as main oxidation states, making it a suitable electron donor or acceptor, respectively, for different dissimilatory processes (Ehrenreich and Widdel, 1994a; Straub et al., 2001, 2004; Lovley et al., 2004; Weber et al., 2006) that greatly influence the iron cycle on Earth. These microbial processes include: (i) anaerobic ferric iron reduction with organic matter and H₂, perhaps the first respiratory process on Earth (e.g., Vargas et al., 1998; Lovley et al., 2004); (ii) anaerobic nitrate-reducing Fe(II)-oxidation (e.g., Straub et al., 2004; Weber et al., 2006); (iii) aerobic chemolithoautotrophic microorganisms that oxidize Fe(II) with oxygen (e.g., Emerson and Weiss, 2004; Chan et al., 2016) and; (iv) Fe(II) photosynthesis, where photoferroautotrophic microorganisms use light energy (photo-) and the reducing power of Fe(II) (-ferro-) to fix inorganic carbon (-autotrophic), according to Widdel et al. (1993):

Anoxygenic photosynthesis with Fe(II) as electron donor using exclusively photosystem I has been proposed as being the earliest type of photosynthetic process (Des Marais, 2000; Xiong et al., 2000; Raymond et al., 2003). While there is currently no supporting evidence from genomic studies (Frigaard and Bryant, 2008; Gupta, 2013) that photoferrotrophy evolved prior to anoxygenic photosynthesis with other electron donors, such as sulfide or hydrogen, the ancient environmental conditions suggest that Fe(II)-photosynthesis was an important metabolic process in the iron-rich Archean oceans, prior to the appearance of oxygenic photosynthesis. Molecular phylogenetic analyses of enzymes involved in bacteriochlorophyll biosynthesis support the existence of anoxygenic phototrophs prior to oxygenic phototrophs (Xiong et al., 2000; Xiong, 2006; Gupta, 2013), although horizontal gene transfer between photosynthetic organisms complicates the phylogeny (Raymond et al., 2003). Before demonstrating its occurrence, photoferrotrophy was hypothesized as a possible autotrophic metabolism for different types of phototrophic prokaryotes (Olson and Blankenship, 2004; Olson, 2006). Some authors suggested the possibility that both cyanobacteria (e.g., Cohen, 1984, 1989; Cohen et al., 1986; Pierson and Olson, 1989; Pierson et al., 1999) and green sulfur bacteria (Garcia-Gil et al., 1990) could perform anoxygenic photosynthesis with Fe(II) as electron donor. In spite of their capacity for anoxygenic photosynthesis (Cohen et al., 1975, 1986), and its crucial role on Earth biogeochemistry, no evidence for the occurrence of photoferrotrophy has been provided so far for modern cyanobacteria (Trouwborst et al., 2007; Swanner et al., 2015; Hamilton et al., 2016).

Anoxygenic photosynthesis using Fe(II) as electron donor was unequivocally demonstrated, for the first time, in purple bacteria, some of which were able to grow either photoautotrophically and/or photoheterotrophically (Widdel et al., 1993; Ehrenreich and Widdel, 1994b). In the meantime, photoferrotrophy has been detected in bacteria of different phylogenetic groups and either from freshwater or marine origin, including representatives of purple sulfur bacteria (PSB, Gammaproteobacteria), purple non-sulfur bacteria (PNSB, Alphaproteobacteria), and green sulfur bacteria (GSB, Chlorobi). Even though most of these anoxygenic phototrophs are able to use several types of electron donors commonly found in anaerobic environments (e.g., H₂S and H₂; Bryant and Frigaard, 2006; Camacho, 2009), there are some specific strains that have been demonstrated to perform Fe(II)-dependent anoxygenic photosynthesis (Croal et al., 2009). These include strains phylogenetically related to PSB, such as the freshwater Thiodictyon sp. (Croal et al., 2004; Hegler et al., 2008), and the marine species Rhodovulum iodosum and Rhodovulum robiginosum (Straub et al., 1999; Wu et al., 2014); freshwater PNSB such as Rhodobacter ferrooxidans sp. strain SW2 (Ehrenreich and Widdel, 1994b; Hegler et al., 2008), Rhodomicrobium vannielii (Heising and Schink, 1998), and Rhodopseudomonas palustris (Jiao et al., 2005; Jiao and Newman, 2007); as well as a freshwater species of GSB, Chlorobium ferrooxidans (Heising et al., 1999). Although the PNSB Rhodobacter capsulatus is also capable of oxidizing Fe(II) in presence of light, this is not a real photoautotrophic process (Caiazza et al., 2007; Kopf and Newman, 2012) and can be considered as a Fe(II) detoxification mechanism (Poulain and Newman, 2009).

The best studied photoferrotrophs are purple non-sulfur bacteria (e.g., Widdel et al., 1993; Eickhoff et al., 2013; Wu et al., 2014). When comparing PSB with PNSB, rates of Fe(II) photooxidation are influenced by the response to light (e.g., light saturation) of each photoferrotrophic species. This was shown by Kappler et al. (2005), comparing the iron photoxidation of the purple sulfur bacteria Thiodictyon sp. and the purple non-sulfur bacteria Rhodobacter ferrooxidans. R. ferrooxidans, with lower saturation intensity, is able to efficiently oxidize iron at rates of about 32 pmol Fe(II) h^-1 cell^-1 at 20°C, circumneutral pH and under light saturation conditions (Hegler et al., 2008). Photoferrotrophs display mechanisms preventing cell encrustation with oxidized iron (Miot et al., 2009; Saraiva et al., 2012; Wu et al., 2014).

Presumably, photoferrotrophs use a periplasmic c-type cytochrome for cyclic electron flow and other iron-oxidoreductases for linear electron transfer (Bird et al., 2011). In Rhodopseudomonas palustris, the pio (phototrophic iron oxidation) operon is essential for phototrophic Fe(II) oxidation, as it encodes, apart from a membrane transport protein, for small high-potential redox proteins such as PioC and cytochrome c₂ (Jiao and Newman, 2007). The latter is essential for cyclic electron flow, whereas PioC linearly transfers electrons from iron (Bird et al., 2014). Operons related to phototrophic iron oxidation are also described for other purple non-sulfur bacteria, such as the foxEYZ operon from R. ferrooxidans SW2 and R. capsulatus, also coding for proteins of similar functions as those encoded by the pio operon of R. palustris (Croal et al., 2007; Saraiva et al., 2012).

Green sulfur bacteria are among the most ancient photosynthetic organisms (Gupta, 2013; Gauger et al., 2016). Although it has been hypothesized that several Chlorobium species could perform Fe(II)-dependent anoxygenic photosynthesis in lakes (e.g., Garcia-Gil et al., 1990), Chlorobium ferrooxidans (both strains KoFox and KB) is so far the only GSB species whose photoferrotrophic ability has been demonstrated. While all other GSB species can use various sulfur compounds as electron donors, genomic analyses revealed that C. ferrooxidans has apparently lost most genes involved in oxidation of sulfur compounds (Frigaard and Bryant, 2008). Heising et al. (1999) showed that strain KoFox was able to grow photoferroautotrophically in co-culture with Geospirillum sp. strain KoFum; whose fermentation of fumarate to organic acids enhanced Fe(II) oxidation by KoFox. Among other possible explanations, acidification of the medium by organic acids could be responsible for this enhancement as this could increase the solubility and hence the bioavailability of iron. Growth of photoferrotrophic KoFox is stimulated by the presence of silica, which is possibly due to its influence on iron chemistry, mediating encrustation patterns and cell–mineral interactions and reducing iron toxicity (Posth et al., 2010; Gauger et al., 2016). On the other hand, Chlorobium ferrooxidans strain KoFox is able to oxidize Fe(II) at very low light intensities, with saturation at <50 lux compared to saturations of 400 lux for R. ferrooxidans and 800 lux for Thiodictyon (Hegler et al., 2008). The capacity of GSB to present spectral modifications in their pigments (Borrego et al., 1997; Chew et al., 2007) and grow under a very dim light (e.g., <0.0005% of surface irradiance, Overmann et al., 1992; Manske et al., 2005), enable GSB to thrive in deep zones of anoxic ferruginous basins where Fe(II) is available and light availability is still sufficient for a “frugal” photosynthesis (Ormerod et al., 1993), making these organisms good candidates for performing photoferrotrophy in current Archean ocean analogs.

Although the possibility of manganese-oxidizing photo-synthesis has also been proposed (Olson, 1970), even as a possible previous step to the advent of oxygenic photosynthesis (Johnson et al., 2013; Fischer et al., 2016), so far no phototrophic Mn(II)-oxidizing bacterium has been found and supporting evidences for manganese-oxidizing photosynthesis are “highly tenuous” (Jones and Crowe, 2013). Even chlorinic photosynthesis –biologically mediated photolytic oxidation of aqueous chloride to form halocarbon or dihalogen products, coupled with CO₂ assimilation-, has been proposed as a potential metabolism on exoplanets under conditions that may approximate to the terrestrial Archean (Haas, 2010).

Apart from the information preserved in the geological record, knowledge about the metabolisms supporting life on ancient Earth, e.g., anoxygenic photosynthesis, can be gained by studying environments that are considered as “modern analogs” of different states of the ancient ocean (Burns et al., 2009). Most meromictic lakes and other permanently stratified water bodies are euxinic, i.e., anoxic and sulfidic, below the chemocline (Zadereev et al., 2017). Sulfidic basins of stratified lakes and closed seas might resemble the sulfidic ocean of the Paleoproterozoic and Mesoproterozoic. Contrastingly, sulfide poor anoxic layers of iron-rich stratified (often meromictic) lakes are the most similar modern environments to the iron-rich Archean oceans, where photoferrotrophy could have played a crucial role in governing the biogeochemistry and providing energy to drive microbial growth and evolution (Llirós et al., 2015). Ferruginous water columns are rare, largely unexplored ecosystems of which only freshwater representatives exist today because of the high sulfate concentrations in the modern ocean. In recent years, a few lakes have been described and studied as possible Archean ocean analogs providing insights on how ancient photoferrotrophs could flourish in the Archean ocean and on their possible role in its biogeochemistry. Koeksoy et al. (2016) reviewed the biogeochemical characteristics of some of these environments and paid special attention to iron chemistry and the usefulness of such systems to interpret Precambrian BIF deposition.

According to the current knowledge, the most appropriate Archean ocean analogs are natural iron-rich meromictic lakes. Meromictic lakes show a permanent stratification whose bottom water (monimolimnion), which has higher concentrations of dissolved salts (Imboden and Wüest, 1995), does not mix with the overlaying waters due to the water density gradient and other factors linked to climate, lake morphometry and water flow (Boehrer et al., 2009). In iron-rich meromictic lakes the high monimolimnetic iron concentrations contribute to the water column stability, which drove Kjensmo (1967) in his seminal manuscripts to use the term “iron-meromixis”. However, the strongest density gradient in these lakes is commonly owed to dissolved compounds other than iron (e.g., calcium bicarbonate, Rodrigo et al., 2001). In stratified iron-rich lakes the interface between oxic and anoxic water bodies, the oxycline, is accompanied by a steep gradient of iron forms, the so-called “ferrocline” (Bravidor et al., 2015). In addition to iron meromictic lakes, temporally stratified lakes have also been proposed as possible study sites as ancient ocean analogs. Specifically, these are the holomictic man-made gravel Lake Vechten in the Netherlands (Koeksoy et al., 2016), as well as dimictic iron-rich Boreal Shield lakes (lakes L227 and L442 from the Experimental Lake Area in Canada) where molecular microbial and stable isotope data suggest that these lakes may be also good candidates to be studied as analogs of ancient oceans (Schiff et al., 2016).

In Table 1 we summarized the basic characteristics of most of the currently described meromictic ferruginous lakes. Mining lakes were excluded because of the acidic pH and strongly different biogeochemical conditions (Boehrer et al., 2009). The best studied ferruginous systems are the lakes Pavin (France), Matano (Indonesia), La Cruz (Spain) and the Kabuno Bay of Lake Kivu (Democratic Republic of the Congo). Even though most of these lakes show maximum iron concentrations largely overpassing those estimated for the Archean ferruginous oceans (Table 1), iron concentrations at the photic anoxic zones of the lakes are lower, likely with higher resemblances to those estimated for the Archean ocean (Holland, 2004; Canfield, 2005; Canfield et al., 2005). However, although it can be stated that iron was present at relatively high concentrations, there is a high uncertainty in the estimation of the iron concentrations in the Archean ocean, so only categorical comparisons can be made with some confidence. On the other hand, some of these iron-rich lakes are also known to accumulate high concentrations of methane and CO₂ in the bottom waters (Crowe et al., 2011), which might cause catastrophic limnic eruptions, similar to what occurred in the African Lakes Nyos and Lagos (Zhang and Kling, 2006; Katsev et al., 2017).

Our knowledge on photoferrotrophy and its past and current significance has tremendously increased in the last decade. We now have evidence from pure cultures as well as from environmental studies that this metabolism could have been a significant process in the early times of our planet, but also that it remains as a relict example of how this primitive Earth may have been functioning during hundreds of millions years. These advances likely place photoferrotrophy among the oldest photosynthetic processes on Earth, probably playing an important biogeochemical role during most of the Archean Eon. It has also been demonstrated that oceans evolved from predominantly reduced to the current oxidized conditions. Interestingly, these primordial anoxic ferruginous conditions still prevail in environments such as the iron-rich meromictic lakes. The study of these modern analogs of primitive oceans has revealed that in many of them GSB related to Chlorobium ferrooxidans are the main photoferrotrophs, but also that the current ecological and biogeochemical significance of photoferrotrophy is minor. Moreover, even when photoferrotrophs are abundant in an anoxic, iron rich, natural environment, sulfur driven anoxygenic photosynthesis still plays an important or dominant role despite the typically very low sulfide concentrations in ferruginous systems. This, in addition to the fact that bacterial anoxygenic photosynthesis is restricted to small compartments of the modern biosphere, implies that the current ecological relevance of Fe(II)-driven anoxygenic photosynthesis is circumstantial. Exploring photoferrotrophy, however, provides insights into how photosynthesis appeared on Earth and how it progressively shaped the biogeochemistry of our planet.

AC, JZ, and XW conceived and designed the work. All coauthors provided information and data to the manuscript. AC made the extensive critical review and wrote the manuscript. XW and AP made the figures. JZ, XW, and AP critically revised the manuscript and provided parts of the text. AC prepared and submitted the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.