Comparative proteomics of Geobacter sulfurreducens PCAT in response to acetate, formate and/or hydrogen as electron donor

Summary Geobacter sulfurreducens is a model bacterium to study the degradation of organic compounds coupled to the reduction of Fe(III). The response of G. sulfurreducens to the electron donors acetate, formate, hydrogen and a mixture of all three with Fe(III) citrate as electron acceptor was studied using comparative physiological and proteomic approaches. Variations in the supplied electron donors resulted in differential abundance of proteins involved in the citric acid cycle (CAC), gluconeogenesis, electron transport, and hydrogenases and formate dehydrogenase. Our results provided new insights into the electron donor metabolism of G. sulfurreducens. Remarkably, formate was the preferred electron donor compared to acetate, hydrogen, or acetate plus hydrogen. When hydrogen was the electron donor, formate was formed, which was associated with a high abundance of formate dehydrogenase. Notably, abundant proteins of two CO2 fixation pathways (acetyl‐CoA pathway and the reversed oxidative CAC) corroborated chemolithoautotrophic growth of G. sulfurreducens with formate or hydrogen and CO2, and provided novel insight into chemolithoautotrophic growth of G. sulfurreducens.


Introduction
Geobacter species play an important role in organic matter degradation coupled to Fe(III) reduction, and are wellknown for making electrical contacts with external electron acceptors, and also with other microorganisms (Lovley et al., 1987;Lovley, 2011Lovley, , 2017. Geobacter species are applied in bioelectrochemical systems owing to their capacity to transfer electrons to and from electrodes, and their remarkable respiratory versatility (Bond et al., 2002;Holmes et al., 2004). Moreover, some Geobacter species can degrade organic pollutants such as benzene and phenol, and transform Co(III), Fe(III), Mn(IV), U(VI) and V(V) oxides of which some are toxic or radioactive (North et al., 2004;Ortiz-Bernad et al., 2004;Schleinitz et al., 2009;Zhang et al., 2014).
Geobacter sulfurreducens PCA has served as a model microorganism in different studies where it is often grown with acetate as the electron donor and carbon source and Fe(III) citrate as the electron acceptor (Segura et al., 2008;Yang et al., 2010;Speers and Reguera, 2012). G. sulfurreducens degrades acetate via the citric acid cycle (CAC), whereas its carbon assimilation starts from acetyl-CoA via pyruvate through gluconeogenesis (Galushko and Schink, 2000;Yang et al., 2010). In contrast, less consistent information is available about the growth of G. sulfurreducens on formate and/or hydrogen. Although G. sulfurreducens is known to reduce Fe(III) with formate or hydrogen as electron donor, addition of low amounts of acetate was necessary for carbon assimilation Speers and Reguera, 2012). Former genomic analysis of G. sulfurreducens has predicted formate catabolism by a periplasmic membrane-bound formate dehydrogenase (Fdh) . Formate assimilation was predicted to be mediated by a pyruvate formate lyase (Pfl), which needs acetyl-CoA and formate to produce pyruvate and CoA (Segura et al., 2008). The acetyl-CoA needed for the assimilation derived from the acetate (0.1 mM) supplemented to the cultures (Speers and Reguera, 2012). Hydrogen oxidation by G. sulfurreducens is catalysed by Hyb, a periplasmically oriented respiratory uptake hydrogenase . When traces of acetate ($1 mM) are available as carbon source, the energy conservation from hydrogen oxidation decreases the demand for acetyl-CoA oxidation through the CAC, and more acetate is channelled towards pyruvate synthesis for anabolic reactions (Yang et al., 2010). A recent study revealed chemolithoautotrophic growth of G. sulfurreducens on formate or hydrogen and CO 2 , underpinning the role of the reversed oxidative CAC in CO 2 reduction using a bidirectional citrate synthase (Zhang et al., 2020). Citrate synthase of G. sulfurreducens was previously proposed to function only in oxidative CAC (Galushko and Schink, 2000). However, the carbon metabolism of G. sulfurreducens when grown with hydrogen and/or formate with CO 2 or acetate as carbon source has not been fully understood.
The aim of this study was to gain insight in metabolism of G. sulfurreducens in response to a variety of electron donors. We performed physiological studies using either acetate, formate, hydrogen or different combinations of these electron donors with Fe(III) citrate as electron acceptor. Acetate was never supplied as carbon source to the cultures growing with formate or hydrogen and Fe(III) citrate was always used as electron acceptor. Comparative proteomic analysis was used to understand the response of G. sulfurreducens to individual electron donors or a mixture of all three.

Results and discussion
Growth phenotypes using different electron donors G. sulfurreducens oxidized acetate, formate, hydrogen, and different combinations of these substrates coupled to the reduction of Fe(III) to Fe(II) (Fig. 1). Acetate was depleted within 3 days (Fig. 1A), whereas formate was degraded within 19 days (Fig. 1C). Hydrogen with CO 2 as the sole carbon source was depleted within 30 days (Fig. 1G).
When hydrogen was used in combination with acetate ( Fig. 1B), formate (Fig. 1D) and acetate plus formate (Fig. 1F), all substrates were used throughout the growth period. However, in the presence of hydrogen, consumption of acetate and/or formate was slower ( Fig. 1 B, D and F) than the corresponding cultures in absence of hydrogen ( Fig. 1 A, C and E). This points to a negative impact of hydrogen on consumption of acetate and formate. Hydrogen is known to induce the expression of HgtR (GSU3364), a repressor which downregulates CAC enzymes such as citrate synthase (GltA), which in turn lowers acetate oxidation in the CAC (Ueki and Lovley, 2010). HgtR was not detected in our results. Surprisingly, when acetate and formate were used concurrently, formate was the preferred electron donor over acetate (Fig. 1E). As such, the electrons released from formate consumption (16.25 × 2 = 32.5) after 6 days is more than those from acetate consumption (3.46 × 8 = 27.7). This is also the case in Fig. 1F after 6 days. Formate was depleted faster in the presence of acetate (in 10 days, Fig. 1E) than when formate was the sole electron donor (in 19 day, Fig. 1C). This preferential consumption of formate over acetate by G. sulfurreducens has not been described before. A possible explanation could be hydrogen formation from formate, and subsequent downregulation of the CAC enzymes by hydrogen (Ueki and Lovley, 2010).
In the presence of acetate ( Fig. 1 E, B and F), total Fe(II) production was more than the corresponding cultures without acetate (Fig. 1 C, G and D). Therefore, acetate was used as an electron donor throughout all the conditions. For instance, in the mix condition when  A comparison of electron donor consumption and Fe(II) production at the end of each batch culture showed that the electron recovery was not 100% (Table 1). This indicates that a part of the electrons released was used for biomass production.

Proteomic analyses
The genome of G. sulfurreducens PCA is 3.8 Mb in size with 3432 open reading frames and contains 3430 protein coding genes (Butler et al., 2012). From the four growth conditions studied, a total of 1584 proteins were detected (46.2% of the total proteins) with two or more peptides of which at least one is unique (Supplementary file 1). Among these, 1204 proteins (76% of the detected proteins) were shared in all the studied conditions (Supporting Information Fig. S1). The total proteins identified in any of the individual four growth conditions ranged from 1297 (growth with formate) to 1526 (growth with acetate). Principal component analysis (PCA) revealed that the protein abundance patterns were reproducible among the triplicates of a given growth condition (Supporting Information Fig. S2A, B) except for one replicate of the mix condition (Mix_2) which contained all three electron donors. This was likely due to a difference in sample preparation e.g. that the cells were not lysed completely. Therefore, this replicate was not considered for further analysis (heatmap, PCA and Venn diagram).
The central metabolic network of G. sulfurreducens in response to the electron donor changes Acetate has been the main substrate used in studies of the metabolism of G. sulfurreducens. In contrast, little attention has been given to formate and/or hydrogen as growth substrates for G. sulfurreducens. Even when formate (Speers and Reguera, 2012) or hydrogen  was used as electron donor, acetate was always added as the carbon source.
In the central metabolic network of G. sulfurreducens, catabolic pathways consist of acetate uptake, acetate activation, the CAC, conversion of pyruvate to acetyl-CoA, and anabolic pathways consist of phosphoenolpyruvate (PEP) synthase and anaplerotic reactions (using PEP carboxykinase and pyruvate carboxylase; Segura et al., 2008;Fig. 2).
Acetate uptake by G. sulfurreducens is mediated by four sodium/solute symporter proteins (AplA; GSU1068, AplB; GSU1070, AplC;GSU2352 and AplD; GSU0518). Three of these proteins, AplA, AplB and AplC, are in a tight phylogenetic cluster (Group I) and presence of at least two of them is necessary for acetate uptake (Risso et al., 2008;Mahadevan et al., 2011). The fourth protein, AplD, is phylogenetically distinct (Group II) and its function is independent of acetate availability (Risso et al., 2008). When growth is limited by acetate availability, Group I proteins have shown to be highly transcribed (Risso et al., 2008).
Interestingly, when we compared the abundance of AplA, AplB and AplC between the different growth conditions to the acetate condition, we found significantly higher presence of these proteins under formate (12.8, 19.1, 9.7 fold, respectively) and hydrogen conditions (9.7, 15.3, 1.6 fold, respectively) as opposed to the mix condition (6.2, 0.9, 0.4 fold, respectively; Fig. 3). In line with the former studies (Risso et al., 2008), AplD showed almost similar abundance in formate (0.1), hydrogen (0.9), and mix condition (−4.4) compared to the acetate condition (Fig. 3). Following uptake, acetate is activated to acetyl-CoA for oxidation through the CAC and for gluconeogenesis. Acetate activating proteins are succinyl CoA:acetate CoA-transferase (Ato-1; GSU0490, Ato-2; GSU0174), acetate kinase (AckA; GSU2707), and phosphotransacetylase (Pta; GSU2706; Segura et al., 2008;Fig. 2, red arrows). In all the conditions with or without acetate, these proteins were present, but clearly with higher abundance of Ato-1&2 in the acetate condition, i.e., Ato-1 (1.8 fold vs Hy, 9.9 fold vs Fo, 8.9 fold vs Mix) and Ato-2 (30.6 fold vs Hy, 25.5 fold vs Fo, 17.3 fold vs Mix), AckA in the hydrogen (1.6 fold vs Ac, 4.3 fold vs Fo, 11 fold vs Mix) and Pta in the formate condition (0.7 fold vs Ac, 2.2 fold vs Hy, 11.4 fold vs Mix; Fig. 3). In the acetate condition, the higher abundance of Ato-1&2 over AckA/Pta may indicate that the CAC is the main route for acetate conversion. An additional acetate activation pathway via AckA/Pta is necessary to provide enough acetyl-CoA for pyruvate synthesis, and to fulfil gluconeogenic and anaplerotic demands. (Fig. 2, red arrows; Segura et al., 2008).
Another key enzyme of the CAC is succinyl-CoA synthetase that reversibly catalyses succinate to succinyl-CoA in the CAC (Fig. 2). Interestingly, the four subunits of this enzyme (SucA; GSU2449, SucB; GSU2448, SucC; GSU1058, SucD; GSU1059) were more abundant in the hydrogen (2.5, 3.8, 4.8, 6.8 fold, respectively), formate (−2.3, 5.7, 1.2, 4.8 fold, respectively) compared to the acetate containing conditions (Fig. 3). In contrast, Ato-1 and Ato-2 were more abundant in the acetate condition (Fig. 3). Notably, Ato-1&2, which activate acetate, have an additional function of converting succinyl-CoA to succinate (Fig. 2, red arrows). Therefore, rather than succinyl-CoA synthetase, Ato-1&2 seem to be the main producers of succinate in the acetate condition. In line with this, a former study showed that when acetate was used by G. sulfurreducens, succinyl-CoA synthetase activity was reduced and succinate was formed in a transferase reaction with acetate as the CoA-acceptor (Galushko and Schink, 2000). Another study showed that a G. sulfurreducens mutant lacking Ato activity could not grow with acetate alone, and grew only when hydrogen was added as electron donor (Segura et al., 2008). These findings indicate that when acetate alone is used, Ato-1&2 fulfil the function of succinyl-CoA synthetase. In formate-adapted G. sulfurreducens, deletion of the succinyl-CoA synthetase genes prevented growth with formate as substrate (Zhang et al., 2020), which indicated that the absence of succinyl-CoA synthetase disrupts carbon assimilation via roCAC in chemolithoautotrophic growth by G. sulfurreducens. GSU0514 is a transcriptional regulator of the IclR family involved in the repression of the transcription of sucC and sucD (Summers et al., 2012). A GSU0514-dependent regulatory mechanism on succinyl-CoA synthetase was shown when G. sulfurreducens was adapted to acetate, formate or lactate (Summers et al., 2012;Zhang et al., 2020). However, in our study both GSU0514 (0.87 vs Ac, 2.5 vs Fo, 6.6 vs Mix) and succinyl-CoA synthetase were more abundant in the hydrogen condition. This challenges the proposed regulatory role of GSU0514 when hydrogen was used as electron donor and merits further investigation.
Another notable finding of our study was the low abundance of proteins involved in acetate metabolism in the mix condition (Fig. 3). This indicates that in the presence of hydrogen and formate, acetate was mostly used as a carbon source for cell synthesis, which is in line with the former reports on reduced acetate oxidation in the CAC in the presence of hydrogen (Ueki and Lovley, 2010).
In the route of converting malate to oxaloacetate via pyruvate (anaplerosis), we detected pyruvate carboxylase (Pyc; GSU2428) and the NAD-dependent malic enzyme (MaeB; GSU1700; Fig. 3; Fig. 2, purple arrows). This pathway could potentially be an alternative for the malate dehydrogenase (Mdh), but at the cost of an extra ATP by Pyc (Segura et al., 2008). We detected both malate dehydrogenase and malic enzyme in all four growth conditions (Fig. 3). However, malate dehydrogenase (Mdh; GSU1466) showed significantly higher abundance in the acetate condition (7 fold vs Hy, 6.4 fold vs Fo, 14.6 fold vs Mix; Fig. 3). Similarly, malic enzyme (MaeB) was also higher in the acetate condition (0.8 fold vs Hy,4.2 fold vs Fo, 1 fold vs Mix) (Fig. 3). This indicates that (i) even in presence of Mdh, the anaplerotic pathway with an extra ATP cost is still necessary and (ii) more oxaloacetate via anaplerosis needs to be routed to the citrate synthase reaction to replenish CAC intermediates or to the gluconeogenesis for PEP synthase. A previous study reported that Mdh-deficient G. sulfurreducens strain was able to grow with hydrogen as the electron donor, but not with acetate (Segura et al., 2008) which indicates MaeB is mostly functional in anabolism.
We further evaluated pyruvate oxidation to acetyl-CoA with the three predicted reactions (Fig. 2, blue arrows) under the studied growth conditions. Having these three pathways in G. sulfurreducens seem to give a versatile approach to capture energy and carbon . The pyruvate: ferredoxin/flavodoxin oxidoreductase (Por; GSU0097) encoding a putative homodimeric type enzyme was abundant in the acetate condition (2.8 fold vs Hy, 4.1 fold vs Fo, 2.7 fold vs Mix). The alternative reactions using pyruvate dehydrogenase (PdhA; GSU2443, PdhB; GSU2436) were present in all the conditions but were the highest in the hydrogen condition i.e. PdhA (3.6 fold vs Ac, 2.7 fold vs Fo, 3.9 fold vs Mix) and PdhB (6.6 fold vs Ac, 1.8 fold vs Fo, 2.9 fold vs Mix), and were the lowest in acetate condition (Fig. 3). Pyruvate formate-lyase enzyme (Pfl; GSU2102) was not detected in any of the conditions, and does not seem to be functional under the studied growth conditions. Acetyl-CoA to pyruvate conversion by pyruvate oxidoreductase (Por) has also been proposed to initiate gluconeogenesis in G. sulfurreducens grown with Fe(III) (Yang et al., 2010). According to a previous study, a POR mutant strain did not grow with acetate as the carbon source and electron donor but grew with acetate and pyruvate (Segura et al., 2008). This indicated that Por has an essential role in G. sulfurreducens to assimilate acetate. In line with this, in our study high abundance of Por was detected in the acetate condition (Fig. 3).
One key step in the gluconeogenesis is the synthesis of phosphoenolpyruvate (PEP). PEP synthesis via pyruvate can occur through three pathways with different energetic demands (Fig. 2, green, and green dashed arrows); pyruvate phosphate dikinase (PpdK; GSU0580), PEP synthase (PpsA; GSU0803) and anaplerotic reactions by pyruvate carboxylase (Pyc; GSU2428) and phosphoenolpyruvate carboxykinase, GTP-dependent (PckA; GSU3385). Variation in electron acceptors (Fe(III) or fumarate) has been reported to influence the contribution of different pathways in PEP synthesis (Yang et al., 2010). For instance, in Fe(III)-grown cultures, PEP was mainly generated from pyruvate via PpsA/PpdK compared to fumarate-grown cultures where PEP synthesis from oxaloacetate by PckA was the major route. The Ppdk pathway is energetically more favourable than either the PckA or the PpsA pathway. This is because the Ppdk reaction produces diphosphate, which is hydrolyzed to translocate protons across the cell membrane, resulting in an energetic advantage (Segura et al., 2008). This is in general favourable for the growth with Fe(III) citrate as electron acceptor because of the already low energy yields compared to fumarate as the electron acceptor. In our study using Fe(III), the abundance of PpdK, PpsA, and Pyc and PckA varied with different electron donors (Fig. 3). The higher abundance of PpdK was found in the acetate containing condition (1.3 fold vs Hy, 8.3 fold vs Fo, 0.0 fold vs Mix), PpsA in the hydrogen condition (6.1 fold vs Ac, 11.7 fold vs Fo, 4.9 fold vs Mix), and Pyc (from anaplerotic pathway) in the hydrogen condition (1.1 fold vs Ac, 4.8 fold vs Fo, − 0.2 fold vs Mix) and PckA in the acetate condition (2.3 fold vs Hy, 6.4 fold vs Fo, 2.2 fold vs Mix; Fig. 3). This high contribution of PckA to gluconeogenesis in the acetate cultures seems to decreases the oxaloacetate level. This seems necessary as the conversion of malate to oxaloacetate by malate dehydrogenase is thermodynamically unfavourable (standard free energy change is +29.7 kJ/mol; Segura et al., 2008).
It was proposed that the electrons from hydrogen directly enter the menaquinone pool, and that this avoids the energy demand via the CAC (Segura et al., 2008). However, presence of the CAC enzymes in the hydrogen culture (although with reduced activities compared to the acetate culture) contribute to anabolism via roCAC. Our results in the hydrogen condition show that gluconeogenesis may initiate from pyruvate by Ppsa or/and oxaloacetate by PckA. Possibly decreased energy conservation via the CAC in the presence of hydrogen may make intermediates such as oxaloacetate more available for anabolic purposes such as amino acid biosynthesis (Yang et al., 2010;Zhang et al., 2020). Acetyl-CoA may replenish from phosphoenolpyruvate via Pyk, and subsequently from pyruvate by Pdh or Por. Pyruvate kinase (Pyk; GSU3331) that converts phosphoenolpyruvate (PEP) to pyruvate was more abundant in the hydrogen condition (1.1 fold vs Ac, 1.4 fold vs Fo, 5.1 fold vs Mix; Fig. 3). Overall, formate-grown cells in our study showed the lowest abundance of enzymes involved in the CAC and PEP synthesis pathways compared to the other conditions.
Hydrogenases and formate dehydrogenases in response to electron donor changes G. sulfurreducens has been shown to produce and consume hydrogen (Caccavo et al., 1994;Coppi et al., 2004). Four [NiFe]-hydrogenases (Hya, Hyb, Hox and Mvh cluster), and three other hydrogenases clusters (Hyp, Ehr and Hdr clusters) have been identified in the genome of G. sulfurreducens Coppi et al., 2004).
Previous studies on the role of the different hydrogenases in G. sulfurreducens have established the importance of Hyb as the only respiratory uptake hydrogenase required for growth of G. sulfurreducens with hydrogen as the electron donor and either fumarate, AQDS or Fe(III)-citrate as the electron acceptor . Moreover, Hyb is the only hydrogenase present in the G. sulfurreducens genome that is absent from the genome of a related species, G. metallireducens, which is unable to grow using hydrogen as electron donor (Lovley et al., 1993). As a membrane-bound respiratory hydrogenase with a periplasmic-oriented active site, Hyb was suggested to transfer electrons from hydrogen to the menaquinone pool, where they can be redistributed to various reductases . Hyb is encoded by the HybSABLP (GSU0782, GSU0783, GSU0784, GSU0785 and GSU0786) operon of which we detected four subunits (GSU0782-85) in all growth conditions (Fig. 4). Remarkably, Hyb subunits were more abundant in formate grown cells (Fig. 4). For instance, the large subunit of the uptake hydrogenase (HybL; GSU0785) showed the highest abundance in the formate condition (4.7 fold vs Hy, 25.6 fold vs Ac, 0.7 fold vs Mix). The unexpected high abundance of Hyb in the formate condition may indicate hydrogen formation from formate by G. sulfurreducens, which to our knowledge, has gone unnoticed in previous reports. Formate-based hydrogen formation has been reported in diverse microorganisms including Acetobacterium woodii (Schuchmann and Müller, 2013), enterobacteria (Sawers, 2005), Clostridiaceae or Archaea such as Methanococcus (Calusinska et al., 2010), and the thermophilic bacterium Thermococcus onnurineus (Kim et al., 2010;Lim et al., 2014).
The other three hydrogenases of G. sulfurreducens, Hya, Hox and Mvh have not been suggested to play a direct role in hydrogen consumption, but in hydrogen production Tremblay and Lovley, 2012). Hya (GSU0120-23) might be involved in converting excess reducing equivalents in the menaquinone pool to hydrogen (Tamagnini et al., 2002;Coppi et al., 2004). We only detected the large subunit, HyaL (GSU0122), that showed significantly the highest abundance in the formate condition (2.6 fold vs Hy, 15.6 fold vs Ac, 14.7 fold vs Mix). This may further indicate hydrogen formation from formate.
Although G. sulfurreducens is capable of hydrogen production, hydrogen-evolving Ech hydrogenases are absent in its genome. Instead, it contains a cluster encoding a multimeric membrane-bound Ech hydrogenase related (Ehr) complex. This gene cluster is most similar in composition to the operons encoding Ech hydrogenases. We detected the high abundance of all subunits of the Ehr complex (GSU0739-45) in the formate condition compared to the hydrogen-containing conditions. The physiological role of the Ehr complexes has never been investigated in G. sulfurreducens. However, an evolutionary link between the Ehr complexes and Ech hydrogenases that are associated with formate hydrogen lyases (Fhl) was suggested . As Fhl is absent in the genome of G. sulfurreducens, the increased abundance of Ehr subunits in the formate condition in our results may support the functional link between the Ehr complex and formate dehydrogenases that allow production of hydrogen from formate.
The genome of G. sulfurreducens contains a cluster encoding a periplasmically oriented membrane-bound formate dehydrogenase . Of the four subunits of formate dehydrogenase (GSU0777-80), we detected three subunits among all the conditions (Fig. 4), but we did not find FdhD/mobA-2 (GSU0780). The catalytic subunit of formate dehydrogenase, FdnG (GSU0777), was significantly more abundant in the hydrogen condition (5.3 fold vs Fo, 38.8 fold vs Ac, 0.01 fold vs Mix; Fig. 4). In contrast, in the formate condition, FdnH (GSU0778; 1.3 fold vs Hy, 25.9 fold vs Ac, 5.3 fold vs Mix) and FdnI (GSU0779; 14.3 fold vs Hy, 11.1 fold vs Ac, 12.7 fold vs Mix) were more abundant (Fig. 4). The observed difference in the abundance of FdhG, FdhH and FdhI in the studied growth conditions (Fig. 4) may suggest different roles for these subunits of formate dehydrogenases. For instance, FdhI is a membraneassociated domain of CbcL (c-and b-type cytochrome for low potential) that can function as a MQ oxidoreductase (Reguera and Kashefi, 2019).

Ac_1
Ac_2 Ac_3    Transient trace formate concentrations (0.1-0.3 mM) were detected in the different replicates of the hydrogen condition (data not shown) and the high abundance of two important subunits of formate dehydrogenase (FdnG and FdnH) indicate formate formation in the hydrogen condition. The production of formate from hydrogen and vice versa and the absence of formate hydrogen lyase (Fhl) in G. sulfurreducens genome might indicate the formation of formate hydrogen lyase complex to fulfil the role of the missing Fhl. This has also been suggested for anaerobes such as sulfate reducers and methanogens (Lupa et al., 2008;Martins and Pereira, 2013;Martins et al., 2015).

Variability of electron transport proteins in response to electron donor changes
During the oxidation of acetate, formate or/nd hydrogen, electrons should be transported to the electron acceptor using specialized respiratory chains. In all the cases and according to the detected proteins in this study, produced electrons flow to menaquinone (MQ), c-type cytochromes of the inner membrane (MacA, Cbc complexes; putative menaquinol oxidoreductase), periplasmic c-type cytochromes (PpcA, PpcB), outer membrane cytochromes (OmcB and OmcS), and finally to Fe(III). From oxidation of acetate using the CAC, electrons and protons are carried to menaquinone in the form of NADH. The oxidation of formate and hydrogen likely directly contributes to the proton motive force and electron flow to the MQ pool in the inner membrane.
We noted that the electron transport proteins show variation in response to the electron donor changes (Fig. 5). For instance, Nuo-1 cluster (GSU0338-351) which encodes one of the two NADH dehydrogenase complexes clearly showed the highest abundance in the acetate condition and lowest in the formate condition (Supporting Information, Fig. S3), indicating efficient electron transfer from NADH produced by the CAC to menaquinone, which in turn transfers electrons to the terminal electron acceptor Fe(III). Proton pumping by NADH dehydrogenase complexes generates the transmembrane proton gradient that drives ATP synthesis by the ATPase complex. In all the studied conditions, we detected eight subunits of ATP synthase (GSU0108-114 and GSU0333) with generally higher abundance in the formate and mix conditions (Supporting Information Fig. S3).
MacA (GSU0466), an inner membrane c-type cytochrome was reported to be abundant during growth on Fe(III) oxide (Lloyd et al., 2003;Butler et al., 2004;Kim and Lovley, 2008). We detected MacA during growth on Fe(III) citrate with higher abundance in the formate condition (2.1 fold vs Hy, 4.4 fold vs Ac, 6.4 fold vs Mix; Fig. 5). This may indicate cytoplasmic oxidation of formate and extracellular electron transport from the inner membrane.
The genome of G. sulfurreducens was predicted to encode five putative menaquinol oxidoreductase protein complexes (Cbc1, Cbc3, Cbc4, Cbc5, Cbc6;Butler et al., 2010;Aklujkar et al., 2013). These proteins combine electron transfer to form a proton gradient across the inner membrane via either a Q loop or a Q cycle (Trumpower, 1990). According to former transcriptomic and proteomics data, Cbc-like gene clusters in G. sulfurreducens are differentially expressed in response to electron acceptor availability or growth in syntrophic co-culture with other strains (Mehta et al., 2005;Leang et al., 2010). Some putative structural subunits of Cbc1, Cbc3, Cbc4 and Cbc5 complexes were detected in our study with dynamic responses towards various electron donors. Generally, the highest abundance was found in the formate condition and the lowest in the acetate condition which may indicate a direct flow of electron from formate to the menaquinone pool (Supporting Information Fig. S3). Although Cbc complexes show a high degree of conservation among Geobacter genomes (Butler et al., 2010) and interact with other electron transfer proteins, these complexes still remain ill-defined.
In order to transfer electrons across the periplasm from electron donors such as acetate that are metabolized in the cytoplasm, Cbc proteins transfer electrons to one of several small periplasmic cytochromes (Ppc). Of the five closely related periplasmic c-type cytochromes (PpcA-E) of G. sulfurreducens , PpcA, PpcB, PpcD and PpcE have been detected in the former proteomic study with acetate and Fe(III) citrate (Ding et al., 2008). In contrast, we detected only PpcB (GSU0364) and PpcA (GSU0612). PpcB was more abundant in the formate condition (6.5 fold vs Hy, 8.2 fold vs Ac, 6.2 fold vs Mix) and PpcA in the hydrogen condition (2.8 fold vs Fo, 4.6 fold vs Ac, 8.5 fold vs Mix) (Fig. 5). In contrast, a former study showed that PpcA is not required for Fe(III) reduction with hydrogen as the electron donor when a PpcA-mutant of G. sulfurreducens reduced Fe(III) with hydrogen (Lloyd et al., 2003). This suggests that the hydrogenases involved in hydrogen oxidation coupled to Fe(III) reduction are in the periplasm and can fulfil the function of PpcA (Lloyd et al., 2003). In our study, the presence of other carriers in the periplasm such as formate dehydrogenases may also fulfil the function of the missing Ppcs.
From outer membrane cytochromes, OmcS (GSU2504) as the terminal Fe(III) reductase (Mehta et al., 2005;Qian et al., 2011), is essential for Fe(III) oxide reduction by G. sulfurreducens Kim et al., 2005;Mehta et al., 2005;Leang et al., 2010). The omcS gene has been reported to be upregulated during growth on Fe(III) oxide but not in cultures grown on Fe(III) citrate (Mehta et al., 2005), or electrodes (Holmes et al., 2006), nor in a coculture of G. sulfurreducens and G. metallireducens where direct interspecies electron transfer (DIET) occurs (Summers et al., 2010). In contrast, we detected OmcS during growth on Fe(III) citrate and with significantly high abundance in the formate condition (28.3 fold vs Hy, 13.3 fold vs Ac, 18.9 fold vs Mix; Fig. 5). Although OmcS was considered essential for DIET, but not for hydrogen interspecies transfer (HIT; Shrestha et al., 2013), the increased abundance of OmcS in the formate condition in our study may point towards a role of OmcS in formate interspecies electron transfer (FIT). This hypothesis can be tested in the future, e.g., by mutants of G. sulfurreducens that lack omcS in co-cultures mediating FIT. Another outer membrane cytochromes, OmcB (GSU2737) has an important function in electron transport to Fe(III) Mehta et al., 2005;Leang et al., 2010). The localization of OmcB suggests that electron transport to the outer membrane is necessary for outside reduction of Fe(III) by G. sulfurreducens. Accordingly, deletion of the omcB gene inhibited reduction of Fe(III) citrate and Fe(III) oxide . In line with this, in our study with Fe(III) citrate, OmcB was detected in all the conditions. The high abundancy of OmcB in the formate condition (1.3 vs Ac, 1.6 vs Hy, 10.8 vs Mix fold; Fig. 5) is in line with its poor functionality in DIET in the coculture of G. metallireducens-G. sulfurreducens (Summers et al., 2010) and may imply OmcB function in IIET (Indirect interspecies electron transfer).
Despite the limited range of C-type cytochromes which have been discussed here, our results showed variability in abundance of other cytochromes in G. sulfurreducens (Fig. 5) in response to electron donor changes in presence of Fe(III) citrate. Although, this emphasizes the  importance of cytochromes for Fe(III) reduction in G. sulfurreducens, future research is necessary to understand the roles of these cytochromes in electron transport mechanisms. The high abundance of the studied electron transport proteins in the formate condition in G. sulfurreducens, may support the ease and preference of formate for conversion over acetate and hydrogen. The results also suggest that portions of the route of electron transfer to Fe(III) citrate differ based on the electron donor type.
Chemolithoautotrophic growth of G. sulfurreducens G. sulfurreducens has been considered a heterotrophic bacterium as it lacks a complete set of enzymes to grow with CO 2 or other C1 compounds such as formate as sole source of carbon (Mahadevan et al., 2006;Feist et al., 2014). In contrast, potential autotrophic growth was predicted from genomic analysis via the reductive CAC or the acetyl-CoA pathway . In line with these genomic predictions, a recent study (Zhang et al., 2020) and our findings corroborate chemolithoautotrophic growth of G. sulfurreducens on formate or hydrogen plus CO 2 with Fe(III) citrate as electron acceptor. G. sulfurreducens cannot utilize citrate as electron donor or carbon source (Galushko and Schink, 2000;Esteve-Núñez et al., 2005;Coppi et al., 2007). Citrate was not likely used as the carbon source as its concentration was always stable throughout the experiment, and we also did not detect a citrate transporter in our proteomic results as predicted from the genome of G. sulfurreducens . Similar arguments for the lack of citrate consumption was also presented in a recent study where chemolithoautotrophic growth of G. sulfurreducens on Fe(III) citrate and formate or hydrogen plus CO 2 was shown (Zhang et al., 2020). Further proof for the lack of citrate consumption in the presence of acetate by G. sulfurreducens was provided by carbon labelling ([U-13C6]Fe(III)citrate) showing that Fe(III) citrate was exclusively used as an electron acceptor (Yang et al., 2010).
Another possibility for autotrophic growth is via acetyl-CoA pathway . However, G. sulfurreducens was reported to lack the acetyl-CoA pathway due to lack of genes encoding a bifunctional enzyme carbon monoxide dehydrogenase/acetyl-Co synthase (CODH/ACS) and formyl tetrahydrofolate synthetase (FTS; Galushko and Schink, 2000;Geelhoed et al., 2016). In contrast, our results imply the occurrence of some reactions of the acetyl-CoA pathway in the hydrogen and formate conditions without the addition of acetate as a carbon source. First, we found high abundance of formate dehydrogenase in formate and hydrogen conditions (Fig. 4), which is a typical enzyme of the CO/acetyl-CoA pathway forming acetyl-CoA for subsequent assimilation into biomass. Second, we found a high abundance of hydrogenase complexes of Ech homologue (GSU0739-GSU0745) and Hox (GSU2718-GSU2722) in formate and hydrogen conditions ( Fig. 4) that may be involved in reverse electron transport and fulfil the FTS function by driving CO 2 reduction to a formyl group with subsequent attachment to the C1 carrier . Third, formate was formed from hydrogen and CO 2 . Fourth, except rcoM (GSU2099), we detected significantly higher abundance of all subunits of monofunctional CODH (CooS; GSU2098, CooC; GSU2097, CooF; GSU2096, FNOR; GSU2095) and high abundance of two other proteins (FolD1; GSU0215, MetF-2; GSU2974) in the hydrogen and formate conditions compared to the acetate-containing conditions (Fig. 3). However, the corrinoid enzyme (GSU2386) and methyltransferase (GSU2387) which may interact with CODH during acetyl-CoA formation were not found in our proteome results. Future experiments will be necessary to verify autotrophic growth using the acetyl-CoA pathway, for example using labelled ( 13 C) bicarbonate to track carbon flow (Zhang et al., 2020), and/or using G. sulfurreducens lacking genes of the acetyl-CoA pathway when grown with hydrogen or formate.

Conclusions
The differential abundance of proteins was studied for deciphering metabolic response of G. sulfurreducens to different electron donors for Fe(III) reduction. The growth phenotypes and the proteomics results revealed that the physiological state of G. sulfurreducens is altered in response to variation of electron donors. A notable finding was the utilization of formate as a preferred electron donor for G. sulfurreducens over acetate and/or hydrogen, whereas most of the previous physiological and genetic studies of this bacterium have used acetate as the preferred electron donor. Moreover, interconversion of formate and hydrogen occurred. Another remarkable finding was chemolithoautotrophic growth of G. sulfurreducens with formate or hydrogen and CO 2 as the sole carbon source. Our results present an important framework for future dedicated studies to verify the actual role of underlying proteins in the central metabolism including hydrogen and formate metabolism and electron transfer mechanism of G. sulfurreducens.

Materials and methods
Microorganism and cultivation G. sulfurreducens strain PCA T (DSM 12127) was grown under strict anoxic conditions at 35 C in 120 ml serum bottles with 50 ml of bicarbonate-buffered medium as described previously (Stams et al., 1993) with N 2 /CO 2 (80:20, v/v) as the headspace. The medium was reduced with FeCl 2 (1.3 mM).
Acetate, formate, hydrogen or their different combinations were used as electron donor(s) and Fe(III) citrate was used as the electron acceptor (Table S1). Sodium acetate and sodium formate were added from 1 M sterile anoxic stock solutions. When hydrogen was used as the electron donor, 17.9% of the headspace was filled with pure hydrogen (18.1 KPa) to reach 10 mmol hydrogen per litre of liquid medium. G. sulfurreducens was adapted to each culture conditions by at least five subsequent transfers (10% v/v) with the corresponding electron donor(s). All incubations were done in triplicate. Growth was determined by quantifying depletion of electron donors and production of Fe(II).

Analytical methods
Acetate and formate were analysed using highperformance liquid chromatograph (HPLC) using a Dionex UHPLC system (Molenaar et al., 2019). Headspace hydrogen was measured by gas chromatography (GC). Headspace samples (0.2 ml) taken with a 1 ml syringe were analysed using a Compact GC 4.0 (Global Analyser Solutions, The Netherlands). A molsieve 5A column operated at 100 C coupled to a Carboxen 1010 pre-column was used for hydrogen determination, and detection was done via a thermal conductivity detector. Fe(II) and Fe(III) were quantified with the ferrozine colorimetric method (Stookey, 1970) with absorbance at 562 nm using a U-1500 spectrophotometer (Hitachi, Chiyoda, Tokyo, Japan).

Proteomic analyses
Protein extraction and separation. G. sulfurreducens grown on a single electron donor (acetate, formate or hydrogen) and a mixture of all three electron donors (Table S1) was used for whole cell proteomic analyses. Cells from triplicate cultures were harvested at the end of the exponential growth phase by centrifugation at 16 000g for 20 min at 4 C. Cell pellets were washed twice with 20 mM Tris-HCl (pH 7.5) and stored at −80 C until further use. For protein extraction, cell pellets were resuspended in 0.5 ml SDT-lysis buffer composed of 100 mM Tris/HCl, pH 7.5, 4% w/v sodium dodecylsulfate (SDS), SIGMAFAST™, Protease Inhibitor Cocktail Tablet (Sigma-Aldrich, Missouri), and 0.1 M dithiothreitol (DTT). The suspension was subsequently sonicated as described previously (Rupakula et al., 2013). Proteins were denatured by heating at 95 C for 5 min, followed by centrifugation at 16000× g for 10 min. The samples were loaded on a 10% precast polyacrylamide gel (10% Mini-PROTEAN® TGX™ Precast Protein Gels, Bio-Rad) with Mini-PROTEAN Tetra system (Bio-Rad Laboratories B.V., Veenendaal, The Netherlands). The electrophoresis procedure was done following the precast gels manufacturer's instructions by loading 40 μl of each sample in gel slots (an empty lane left next to each loaded lane). Gels were stained with Coomassie brilliant blue R250 (Merck, Darmstadt, Germany) and scanned using a G:BOX Chemi XT4 (Syngene, Cambridge, UK). To quantify the intensity of each entire lane, the GeneSys software version 1.5.5.0 (GeneTools version 4.03.01) was used. The ratio of intensity of the lanes to the lane with the highest intensity was calculated and used to prepare series of gels that had less than 10% difference in the intensity between all lanes (Sedano-Núñez et al., 2018).
In-gel trypsin digestion and peptide purification. SDS-PAGE-separated proteins were subjected to in-gel trypsin digestion and peptide purification with slight modifications to a protocol described previously (Rupakula et al., 2013). All solutions were prepared in 50 mM NH 4 HCO 3 unless otherwise stated. Disulfide bridges of proteins in each gel were reduced with DTT solution (10 mM, pH 7.6) at 60 C for 1 h, alkylated with iodoacetamide solution (10 mM in 100 mM Tris-HCl, pH 8.0) in the dark at room temperature for 1 h with shaking at 100 rpm. After each step, gels were rinsed three times with ultra-pure water. Gels were cut into individual lanes and each gel lane was cut into three equal slices. Each slice was further processed to pieces of about 1 mm 3 cubes and transferred to a 0.5 ml protein LoBind tube (Eppendorf, Hamburg, Germany).
Enzymatic digestion was done by adding 50 μl of 5 ng/μl trypsin solution (Sequencing grade modified trypsin; Promega, Madison, WI, USA) to each tube, and incubating overnight at room temperature with gentle shaking. In order to stop trypsin digestion, 10% (v/v) trifluoroacetic acid solution in H 2 O was added to the supernatant to lower the pH (<5). Peptides were concentrated using a C18 Empore membrane (Empore™ Extraction Disc, 3 M, VWR Amsterdam, the Netherlands) in StageTip columns with additional C18 column material as described previously (Rappsilber et al., 2007;Lu et al., 2011). Peptides were eluted with 50% (v/v) acetonitrile (in 0.1% formic acid in H 2 O) from the column. Finally, a SpeedVac vacuum centrifuge was used to remove the acetonitrile from the samples and to concentrate the peptides. To obtain enough volume for LC-MS/ MS analysis, the samples volume were adjusted to 50 μl with 0.1% (V/V) formic acid. Samples were analysed using nano-liquid chromatography (Proxeon EASY nLC-MS) connected to a LTQ-Orbitrap XL (Thermo Fisher Scientific, Bremen, Germany) mass spectrometry as described previously (Lu et al., 2011).

LC-MS data analysis
The protein database of G. sulfurreducens PCA was downloaded from Uniprot (http://www.uniprot.org). An additional database with protein sequences of common contaminants (trypsin, human keratins and bovine serum albumin) was also included in the database search. MaxQuant v. 1.5.2.8 was used to analyse MS and MS/MS spectra. False Discovery Rate (FDR) of less than 1% were set at both peptide and protein levels. Modifications for acetylation (Protein N-term), deamidation (N, Q) and oxidation (M) were allowed to be used for protein identification and quantification. All other quantification settings were kept default. Filtering and further bioinformatics and statistical analysis were performed with Perseus v.1.5.0.8. Proteins included in the analysis contain at least two identified peptides of which at least one is unique and at least one unmodified. Reversed hits and contaminants were filtered out. Protein groups were filtered to require three valid values in at least one experimental group. Label-free quantification (LFQ) intensities (values normalized with respect to the total amount of protein and all of its identified peptides) were used to analyse the abundance of proteins in the fractions and further statistical comparisons among conditions. LFQ intensities were transformed to logarithmic values base 10. The non-existing LFQ intensity values have been replaced with values obtained by applying a normal distribution down shift of 1.8 and a width of 0.3 (Perseus default values). Relative protein quantification of sample to control was conducted with PERSEUS by applying twosample t tests using the 'log LFQ intensity' columns obtained with an FDR threshold set to 0.05 and S0 = 1. S0 = 1 indicates that both the average protein abundance ratio and the t test P value have equal weight to decide whether the protein is significantly different between the two different growth conditions. Proteins significantly (P < 0.05) present under given condition were mentioned throughout the text. To determine the proteins fold change between two conditions, log protein abundance ratio of those conditions were considered. In order to make sensible ratio calculations possible, protein groups with a logarithmic LFQ intensity of zero of all treatments were deleted from the MaxQuant result table. To analyse proteins detected in each experimental group (Venn diagrams) proteins were considered when they had been detected in two samples in each experimental group. Zscore normalization in which the mean of each row (where each row is a protein in triplicate and in different conditions) is subtracted from each value and the result divided by the standard deviation of the row was applied before clustering of the samples; Hierarchical clustering was performed using relative intensity LFQ data, Pearson's correlation as a distance metric and agglomerative hierarchical clustering. For the enrichment analysis, only proteins detected in at least two samples in each experimental group were considered. The enrichment analysis was performed using a hypergeometric function to model background probabilities as implemented in the R 'phyper' function. PCA, heatmaps, hierarchical trees and enrichment analysis were done using R (Team, 2016). The R package Venn Diagram was used for Venn diagrams (Chen and Boutros, 2011); 'prcomp' command was used for PCA, and 'hclust' was used for hierarchical clustering. The protein groups of G. sulfurreducens are listed in Supplementary Table S2. Proteins detected in our proteomic analysis were used to prepare the heatmaps. IV pilus biogenesis protein PilO (PilO), type IV pilus secretin lipoprotein PilQ (PilQ), type IV pilus assembly lipoprotein PilP (PilP), sensor histidine kinase PilS (PAS, HisKA, HATPase_c) (PilS), sigma-54-dependent transcriptional response regulator PilR (REC, sigma54 interaction, HTH8) (PilR), twitching motility pilus retraction protein (pilT-1), twitching motility pilus retraction protein (PilT-4). Table S1. Cultivation conditions for G. sulfurreducens used in this study. The growth conditions shown in bold were used for proteomic analysis.