Substrate-dependent incorporation of carbon and hydrogen for lipid biosynthesis by Methanosarcina barkeri.

Dual stable isotope probing has been used to infer rates of microbial biomass production and modes of carbon fixation. In order to validate this approach for assessing archaeal production, the methanogenic archaeon Methanosarcina barkeri was grown either with H2 , acetate or methanol with D2 O and 13 C-dissolved inorganic carbon (DIC). Our results revealed unexpectedly low D incorporation into lipids, with the net fraction of water-derived hydrogen amounting to 0.357±0.042, 0.226±0.003 and 0.393±0.029 for growth on H2 /CO2 , acetate and methanol, respectively. The variability in net water H assimilation into lipids during growth of M. barkeri on different substrates is possibly attributed to different Gibbs free energy yields, such that higher energy yield promoted the exchange of hydrogen between medium water and lipids. Because NADPH likely serves as the portal for H transfer, increased NADPH production and/or turnover associated with high energy yield may explain the apparent differences in net water H assimilation into lipids. The variable DIC and water H incorporation into M. barkeri lipids imply systematic, metabolic patterns of isotope incorporation and suggest that the ratio of 13 C-DIC vs D2 O assimilation in environmental samples may serve as a proxy for microbial energetics in addition to microbial production and carbon assimilation pathways. This article is protected by copyright. All rights reserved.


Introduction
Stable isotopic compositions of microbial lipids provide valuable metabolic and taxonomic information that helps to decipher the role of microorganisms in biogeochemical cycles (Hayes, 1993;Hinrichs et al., 1999;Dawson et al., 2015). The 13 C/ 12 C ratio, expressed as δ 13 C, of microbial biomass is primarily determined by the carbon sources, fixation pathways and physiological conditions (e.g., Hinrichs et al., 1999;Hayes, 2001;Boschker and Middelburg, 2002;Schouten et al., 2004;Londry et al., 2008;Blaser et al., 2015), whereas the ratio of stable hydrogen isotopes (deuterium/protium ratio; D/H; expressed as δD) is determined by the water and substrate-based sources of hydrogen and the interactions of central metabolic pathways (Valentine, 2009;Zhang et al., 2009a;Wijker et al., 2019). The fractionation factors that determine the δD value of lipids versus source water vary systematically with specific metabolisms, showing a decrease in the order: heterotrophic growth on TCA-cycle precursors and intermediates > heterotrophic growth on sugars > photoautotrophy > chemoautotrophy (Valentine et al., 2004a;Sessions and Hayes, 2005;Zhang et al., 2009a).
Stable isotope probing (SIP) of diagnostic molecules (lipids or nucleic acids) with labelled substrates has been widely used because it can trace the activity of microbial communities and provide an estimate of key metabolic processes (Boschker et al., 1998;Kreuzer-Martin, 2007;Taubert et al., 2018). Widely used 13 C-labelled substrates include dissolved inorganic carbon (DIC) and organic compounds such as glucose, amino acids and bulk biomass . However, the utilization of organic substrates in SIP experiments distorts the in situ nutrition conditions, particularly for organic matter-lean microbial habitats, and it selects for a subpopulation of the microbial community, which probably characterizes a specific metabolic process rather than the entire community (Kopf et al., 2015).
Labeling experiments with heavy water (D 2 O) have been an alternative to earlier established methods (e.g., 13 C and 15 N) to estimate total microbial biosynthesis activity Kopf et al., 2015;Wegener et al., 2016) and can even be used for sorting and identifying single active cells (Berry et al., 2015;Kopf et al., 2016;Taubert et al., 2018). Lipid-based SIP with D 2 O provides a bulk, quantitative estimate of microbial community production, without altering natural conditions or selecting for certain autotrophic or heterotrophic C metabolisms, via the measurement of D incorporation into lipids via established gas chromatography-isotope ratio mass spectrometry. Water-derived hydrogen incorporation into lipids is essentially mediated by NADPH in the central metabolism (Zhang et al., 2009a;Berry et al., 2015;Wijker et al., 2019), where NADPH gains hydrogen from organic substrates or water-derived hydrogen via diverse metabolic processes such as glycolysis, the oxidative pentose phosphate pathway and the TCA cycle (Zhang et al., 2009a;Spaans et al., 2015;Wijker et al., 2019). Isotope labeling studies of fatty acid biosynthesis in heterotrophic bacteria indicate that NADPH is the most important H source (~50%) and controls the isotope effect during lipid biosynthesis, while the medium water (~25%) is of less importance relative to NADPH (Saito et al., 1980;Valentine, 2009;Zhang et al., 2009a;Osburn et al., 2016). In contrast, this is not true for archaea that derive their lipid hydrogen mainly from acetyl-CoA (~60%; Fig. S1, Jain et al., 2014;Rodriguez and Leyh, 2014;Vinokur et al., 2016). Despite this, the resulting δD values of lipids are generally well correlated with that of the medium water source (Valentine et al., 2004b;Zhang et al., 2009a;Zhang et al., 2009b;Pagani, 2013a, 2013b), highlighting that water plays an additional indirect role (Zhang et al., 2009a) in determining lipid hydrogen isotopic composition by potential hydrogen exchange between NADPH and water. Due to the difficulty in determining the hydrogen isotopic composition of NADPH, two ultimate H sources (i.e., external water and organic substrates) are assumed to be incorporated into lipids, with the final lipid H isotopic composition determined by isotope effects of metabolic fluxes and the enzymes involved (Sessions and Hayes, 2005;Zhang et al., 2009a;Wijker et al., 2019). Therefore, the fractionation factor between water and lipids (α l/w ) and the net water-derived hydrogen contribution (X w ) have to be considered for hydrogen assimilation during lipid biosynthesis (Sessions and Hayes, 2005;Zhang et al., 2009a;Kopf et al., 2015Kopf et al., , 2016. In order to better understand this relationship, a physiological parameter 'water hydrogen assimilation constant (a w )' was defined by Kopf et al. (2015), which was obtained from the regression of the hydrogen isotopic composition of lipids and water. Thus, a w represents a net combination of the net hydrogen isotope fractionation factor and water hydrogen assimilation, and can indirectly provide the information of water-derived hydrogen for lipid biosynthesis (Sessions and Hayes, 2005;Zhang et al., 2009a;Kopf et al., 2015Kopf et al., , 2016, which is relevant for the application of D 2 O labeling in environmental studies. Reported a w values of bacterial and eukaryotic fatty acids (Zhang and Sachs, 2007;Zhang et al., 2009a;Heinzelmann et al., 2015), alkenones (Englebrecht and Sachs, 2005;Schouten et al., 2006;van der Meer et al., 2015) and sterols (Sessions et al., 2002;Sachs and Schwab, 2011;Sachs and Kawka, 2015) in pure strain incubations and under different growth conditions range from 0.4 to 0.9 (Zhang and Sachs, 2007;Zhang et al., 2009a;Zhang et al., 2009b;Kopf et al., 2015Kopf et al., , 2016. The parameter a w is defined specifically for the application in isotope labeling work, and the diverse a w values necessitate the use of a correction factor in order to accurately estimate microbial lipid production via the D 2 O labeling approach (Kopf et al., 2015(Kopf et al., , 2016Wegener et al., 2016).
Combined amendments with both 13 C-DIC and D 2 O, also described as dual SIP, were developed to garner metabolic information without changing the in situ nutrient conditions Wegener et al., 2012;Huguet et al., 2017). In these dual-SIP applications, the assumptions were: (i) independent of the carbon substrate utilized, most H incorporated into microbial lipids derives from water (>80%); (ii) assimilation of inorganic carbon (IC) into lipids of microbial heterotrophs is minor (<30%); and (iii) the ratio of 13 C (from IC) versus D (from water) incorporated into lipid biomarkers ( 13 C-IC/ D 2 O assim ) can distinguish the central microbial metabolism (~1 for autotrophs vs. <0.3 for heterotrophs). However, the diverse values of a w imply the cut-off value for the distinction of the microbial metabolism could be variable. Similarly, an unconstrained contribution of IC during heterotrophic growth can complicate the use of 13 C-IC/ D 2 O assim for isotope labeling experiments. These assumptions, based on experiments with the bacterium Desulfosarcina variabilis , have not yet been validated for other microbes, especially for archaea.
Archaea play an important role in the biogeochemical carbon cycle in the ocean and in sediments (Könneke et al., 2005;Thauer et al., 2008;Offre et al., 2013;Yu et al., 2018). In particular, methanogenic archaea, a widespread group of strictly anaerobic Euryarchaeota, can utilize diverse carbon substrates for energy metabolism like CO/CO 2 and H 2 , formate, acetate and methanol or methylamines (Thauer et al., 2008), while methanotrophic archaea, on the other hand, turn over methane, which is represented by characteristic carbon isotopic compositions of archaeal lipids (Hinrichs et al., 1999;Elvert et al., 2000;Méhay et al., 2013;Zhuang et al., 2016). To our best knowledge, there is only one study on the a w , investigating archaeol (AR) from a halophilic archaeon, Haloarcula marismortui (Dirghangi and Pagani, 2013b), but lipid a w values of methanogens or other autotrophic archaea have rarely been constrained (Kaneko et al., 2011;Könneke et al., 2012). Thus, the correction factor of water hydrogen assimilation based on bacterial metabolisms might be not feasible when archaeal activity is studied using D 2 O labelling. Moreover, if archaea also assimilate DIC when grown heterotrophically, as observed for the bacterium Desulfosarcina variabilis (cf. Wegener et al., 2012), it is necessary to investigate the contribution of DIC for archaeal lipid biosynthesis during heterotrophic growth. For methanogenic archaea, lipid biosynthesis is associated with an IC demand not only for autotrophic but also for heterotrophic growth, i.e., IC is fixed into the intermediate acetyl-CoA catalysed by the carbon monoxide dehydrogenase (CODH) enzyme, and further utilized for lipid biosynthesis (Thauer, 1998;Yin et al., 2019). However, it is not clear how much IC is flowing into lipid biosynthesis of methanogenic archaea during heterotrophic growth (Londry et al., 2008), and it thus hinders the dual SIP application and the elucidation of the carbon metabolic routes of methanogenic and, more generally, archaeal lipid biosynthesis in environmental samples.
In this study, we examined the methanogenic archaeon Methanosarcina barkeri cultured with diverse carbon substrates to better understand the mechanisms of carbon and hydrogen fixation into archaeal lipids using the dual SIP approach. Specifically, we (i) quantified the contribution of IC during lipid biosynthesis of M. barkeri grown under different substrate conditions (H 2 /CO 2 , acetate or methanol), (ii) followed the assimilation of water derived hydrogen into archaeal lipids produced under these conditions, and (iii) evaluated the ratios of 13 C-DIC and D 2 O derived from carbon assimilation and lipid production rates respectively. Based on the acquired information, we were able to extend the utilization of the dual SIP approach to study the microbial activity of archaeal community members in natural settings.

Metabolic activity and development of methane carbon isotopic composition
Methanol was completely consumed by M. barkeri within 4 days and methane was produced relatively rapidly, increasing to~20% (v/v) of the headspace (Fig. 1A). Methane was produced more slowly when acetate or H 2 / CO 2 were supplied as substrates (Fig. 1A). This indicates that M. barkeri grew most rapidly on methanol, followed by H 2 /CO 2 and acetate, which is consistent with observations by Londry et al. (2008). More details about the incubation of M. barkeri are provided in the Supporting Information Experimental Procedures. The δ 13 C values of CH 4 at the time of harvest varied with the different 13 Clabelled carbon substrates. When growing on acetate and methanol with 13 C-bicarbonate the methane isotopic compositions were only moderately affected (Table S1). The δ 13 C-CH 4 values of +38 to +215‰ suggest a minor contribution of IC to methane production for growth on the two organic substrates and are a result of back fluxes in the enzymatic chain of methanogenesis (Zehnder and Brock, 1979).
Substrate-dependent production of lipids AR and hydroxyarchaeol (OH-AR) were the two dominant lipids of the polar lipid fraction in all experiments; the relative abundance of OH-AR (60%-70%) was higher than that for AR (30%-40%). Isoprenoid alkanes and alkenes were detected in the apolar fraction. Their distribution varied according to the carbon substrate. When grown on methanol, unsaturated pentamethylicosane derivatives (unsPMIs) with 1-3 double bonds accounted for 73%-96% (avg. 85%) of the total pentamethylicosanes (PMIs), whereas growth on H 2 /CO 2 yielded only minor amounts of unsPMIs (<10%), and no unsPMIs were detected in incubations with acetate ( Fig. 1B). This suggests the potential of unsPMIs as specific biomarker during the growth of M. barkeri on methanol.

Carbon isotopic composition of archaeal lipids during growth on specific substrates
When grown on unlabeled substrates, the net isotope fractionation between lipids and substrates varied substantially. For growth on acetate, lipids were enriched in 13 C relative to acetate with fractionation factor ε lipid/acetate of +0.85‰ to +3.62‰ (avg. +2.5‰; Supporting information Table S2). In contrast, lipids were depleted in 13 C with ε lipid/substrate of −24.6‰ and −21.1‰ for H 2 /CO 2 and methanol respectively. Londry et al. (2008) also observed that lipids were enriched in 13 C on acetate but depleted on H 2 /CO 2 , however, the values of ε lipid/substrate were larger than in our study, e.g., ε lipid/CO2 was up to −48.4‰ (Londry et al., 2008) versus −24.6‰ for growth on H 2 / CO 2 in our study (Supporting information Table S2).
The incorporation of 13 C-labelled substrates with similar labeling strength during the growth on different energy substrates caused varying labeling intensity of the produced lipids. When grown on H 2 /CO 2 with 13 C-bicarbonate, the δ 13 C of lipids (+1590‰ to +2290‰) was slightly lower than the δ 13 C values of total inorganic carbon (TIC) (δ 13 C-TIC from +2410‰ to +2740‰). Based on the mass balance equation (Supporting Information Experimental Procedures), the contribution of TIC (X TIC ) in PMI, AR and OH-AR was 85.6 AE 8.5%, 87.8 AE 7.3% and 81.9 AE 11.9% respectively, with an average of X TIC = 86.8 AE 7.4% (Fig. 2). If a kinetic isotope effect was considered (e.g., α = 1.052 for acetyl-CoA production from CO 2 by CODH; cf. Hayes, 2001), X TIC would be more than 90%, suggesting that IC was the major carbon source for growth on H 2 /CO 2 , although the growth medium was supplemented with complex carbon substrates (e.g., yeast extracts and casitone).
During growth of M. barkeri on acetate (δ 13 C = −34.2‰) and labelled 13 C-bicarbonate (δ 13 C-3230‰ to 4370‰) lipids with δ 13 C values of +670‰ to +730‰ were produced (Supporting information Table S1). This suggests that acetate was the major carbon source under these conditions. It was combined with minor utilization of smaller amounts of IC, and indeed X TIC for the biosynthesis of PMI, AR and OH-AR was 19.1 AE 2.1%, 20.9 AE 2.1% and 18.9 AE 2.3% respectively (Fig. 2), with an average of 19.6 AE 2.2%. When M. barkeri was grown with 2-13 C-acetate (+1330‰ < δ 13 C-acetate < +1830‰), the δ 13 C values of PMI, AR and OH-AR were between +1540‰ and +2180‰ (for all data see Supporting information Table S1). The 13 C content of these lipids was higher than that of bulk acetate, suggesting that the methyl group of acetate is predominantly used for lipid synthesis. Based on mass balance equations (Equations 2 and 3 in Supporting Information Experimental Procedures), the contribution of methyl C (acetate-C2; X C2 ) and carboxyl C (acetate-C1; X C1 ) for lipid biosynthesis was on average 56.7 AE 1.7% and 23.7 AE 1.7% respectively (Fig. 2).
When grown on methanol with either 13 C-bicarbonate (+4250‰ < δ 13 C-TIC< +8740‰) or 13 C-methanol (+4370‰ < δ 13 C-methanol < +5560‰), the δ 13 C values of unsPMIs, AR and OH-AR ranged from +1440‰ to +2960‰ (for all data see Supporting information Table S1) and were thus much lower than the label strength of IC or methanol. It The molar fraction of M. barkeri lipid carbon derived from inorganic carbon (X TIC ; %) during growth on H 2 /CO 2 , acetate or methanol. Values determined by 13 C-substrate labeling according to Equations 1-3 in Supporting Information Experimental Procedures. n.d., not detected. X TIC for lipids estimated from the 2-13 C-acetate experiment is not listed because X TIC was not derived independently. Error bars are standard deviation from replicate samples (n = 3-5).
suggests that during growth on methanol M. barkeri uses both IC and methanol for the biosynthesis of its lipids. Based on the 13 C-bicarbonate experiment, X TIC for the biosynthesis of unsPMIs, AR and OH-AR was 36.4 AE 2.6%, 36.3 AE 3.2%, 35.5 AE 3.3% respectively (Fig. 2), with an average of 36.2 AE 3.3%. By contrast, in the 13 C-methanol experiment, X TIC for the biosynthesis of unsPMIs, AR and OH-AR was on average 46.2 AE 5.7% (Fig. 2). The difference could be due to the different isotopic fractionation effects between IC and lipids and methanol and lipids respectively, which were not considered in our calculation (cf. Equation 1 in Supporting Information Experimental Procedures). Notably, when 13 C-bicarbonate or 13 C-methanol was used, δ 13 C-PMI was positive but always much lower (+7‰ to +1110‰; Supporting information Table S1) than that of other lipids, and even not related to the label strength of the carbon substrates. This led to an inconsistent estimate of X TIC for PMI biosynthesis: the experiments with 13 C-bicarbonate yielded low values of X TIC (7.6 AE 8.7%; Fig. 2) and those with 13 C-methanol yielded high values (92.0 AE 10.2%; Fig. 2). It indicates that unsPMIs were probably synthesized by M. barkeri during growth on methanol but PMI was not. unsPMIs, which are frequently observed in cold seep environments (Elvert et al., 1999(Elvert et al., , 2000Pancost et al., 2001), might bear some potential as biomarker for the utilization of methanol or potentially other methylated substrates. Because neither bicarbonate nor methanol was apparent substrates for PMI synthesis during growth on methanol, the weighted average X TIC for lipid biosynthesis (41.2 AE 6.7%) considered only the values determined for unsPMIs, AR and OH-AR via the two labeling strategies (i.e., 13 C-bicarbonate and 13 Cmethanol).

Hydrogen isotopic composition of archaeal lipids during growth on various substrates
The δD of lipids varied with substrates when grown in unlabeled water medium (δD = −50.6‰) such that the hydrogen isotope fractionation factor between lipids and water was larger during growth on H 2 /CO 2 (ε lipid/water = −279‰) compared with growth on acetate (ε lipid/water = −209‰) and methanol (ε lipid/water = −205‰; for all data see Supporting Information Table S3). Given that net fractionation factors of acetate and methanol are not as large as for H 2 /CO 2 , the fractionation factor between substrate and lipid could be larger than −200‰. This net isotope fractionation factor is an organism level fractionation factor that differs from the water hydrogen assimilation factor α w (Equation 5 in Supporting Information Experimental Procedures). The varying fractionation factors for different substrates are consistent with the pattern that chemoautotrophy is generally associated with higher lipid-water hydrogen isotope fractionation than heterotrophy (Zhang et al., 2009a). This supports the notion that central metabolisms affect the archaeal lipid hydrogen isotope composition (Zhang et al., 2009a;Wijker et al., 2019). During growth on deuterated medium (1% D 2 O; equal to a δD of~+ 64,000‰), the δD values of all three archaeal lipids ranged from +600 to +25 600‰ (Supporting information Table S1). Additional experiments with reduced label content (0.2% and 0.5% D 2 O) yielded similar differences, indicating that this observation is not caused by an excessive labeling strength. This shows that lipid hydrogen of M. barkeri derives from multiple sources and not from water alone. Notably, for growth on methanol with a 1% D 2 O treatment, δD-PMI was substantially lower and more variable (δD = +600 to +14,500‰) than that of unsPMIs, AR and OH-AR, suggesting that PMI was not primarily produced during growth on methanol, as also indicated by the 13 C labeling experiment.
In order to understand water-derived hydrogen contribution to lipid biosynthesis, a linear regression of the fractional D abundance (F D ) between water and lipids was applied (see Supporting information Experimental Procedures; Zhang et al., 2009a;Kopf et al., 2015;Kopf et al., 2016). During growth on H 2 /CO 2 , acetate or methanol, F D of lipids correlated well with that of water in the medium (Fig. 3). The slope of regression was defined as the water hydrogen assimilation factor α w , which represents the contribution of water hydrogen to lipid biosynthesis (Kopf et al., 2015(Kopf et al., , 2016. The average α w for PMI, AR and OH-AR during the growth on H 2 /CO 2 and acetate was 0.357 AE 0.042 and 0.226 AE 0.003 respectively (Fig.  3). However, when grown on methanol, the F PMI D was not strongly correlated with water (R 2 = 0.35; Fig. 3) and α w for PMI was lower (avg. 0.065 AE 0.065) than that for unsPMIs, AR and OH-AR (avg. 0.393 AE 0.029; Fig. 3). The large deviations of α w for PMI are consistent with the questionable estimates of X TIC for PMI during growth on methanol (see above).

Substantial contribution of IC in lipid biosynthesis
IC fixation by methanogenic archaea is powered by H 2 consumption (electrons or protons) via the Wood-Ljungdahl pathway with CODH/acetyl-CoA synthetase (ACS) (Thauer, 1998;Ragsdale and Pierce, 2008 figure refers to the contribution of total inorganic carbon (TIC). The carbon flow in the figure is explained as below: (i) X TIC ≈ 100% and 50% were the IC contributions for acetate biosynthesis for growth on H 2 /CO 2 and methanol respectively, determined from the growth on H 2 /CO 2 plus 13 C-DIC, and methanol plus 13 C-DIC or 13 C-methanol (cf. Equation 1 in Supporting Information Experimental Procedures). (ii) X TIC ≈ 26% is IC contribution to acetate-C1 due to isotopic exchange between 13 CO 2 and acetate-C1, determined from the growth on acetate plus 13 C-DIC (cf. Equation 1 in Supporting Information Experimental Procedures). and methanol to acetate synthesis are consistent with the notion that methanol and CO 2 form the methyl and carboxyl groups of acetate respectively (Kenealy and Zeikus, 1982;Fig. 4). Furthermore, 13 C-bicarbonate was incorporated into acetate during growth on methanol, indicating that M. barkeri utilized external IC for acetate production, even though CO 2 can be derived from methanol catabolically. In contrast, the contribution of IC to lipid biosynthesis (X TIC = 86.8 AE 7.4%) was less than that for acetate production during growth on H 2 /CO 2 , suggesting that organic substrates (e.g., yeast extract, casitone and cysteine) were essential components of the growth medium that contribute to lipid biosynthesis, e.g., the glycerol backbone of lipids, which accounts for 3 C atoms in AR and thus roughly 7% of lipid C. The contribution of IC to lipids during growth on methanol (X TIC = 36.2 AE 3.3%) was also less than that for acetate production, implying 13 C-label dilution and/or loss of IC during the condensation of acetyl-CoA to isoprenoid moieties (Fig. S1).
Because acetate can be directly converted to acetyl-CoA by acetate kinase and phosphotransacetylase during acetoclastic methanogenesis (Fig. 4), it seems that IC is not required to form acetyl-CoA. However, the incorporation of 13 C-bicarbonate into lipids during growth on acetate suggests that IC may be assimilated via 'isotopic exchange' between CO 2 and the carbonyl moiety of acetyl-CoA (Eikmanns and Thauer, 1984), which is supported by the observation of 13 C enriched acetate at harvest for the treatment with 13 C-bicarbonate and natural acetate (Supporting information Table S1). No carbon exchange between other one-carbon substrates (e.g., formate and CO) and acetate was observed by methanogens (Eikmanns and Thauer, 1984), suggesting that this exchange requires electrochemical proton potential (Laufer et al., 1987). Another possibility is that the biosynthesis of acetyl-CoA actually included the disassembly and reassembly of acetyl-CoA for the further transfer of methyl group catalysed by nickel-containing CODH enzymes (Ragsdale and Wood, 1985;Ragsdale and Pierce, 2008). Therefore, this isotopic exchange mechanism may serve as a portal of IC assimilation in acetoclastic methanogenesis. A simple mass balance (e.g., Equation 1 in Supporting information Experimental Procedures) based on the 13 C enrichment of acetate (Supporting information Table S1) suggests that 26.1 AE 2.8% of acetate C was derived from IC by the end of the growth phase. The lower X TIC value for lipid biosynthesis (19.6 AE 2.2%) relative to that for acetate production may result from the loss of CO 2 via cleavage of carboxylic group of acetate during lipid biosynthesis via the mevalonate (MVA) pathway (Supporting information Fig. S1). This is also implied from the varying contribution of acetate-C1 (carboxylic carbon; 23.7 AE 1.7%) and acetate-C2 (methyl carbon; 56.7 AE 1.7%) to lipids. The higher degree of incorporation of acetate-C2 suggests that these lipids were predominantly synthesized via the classic MVA pathway, where acetate-C2 and acetate-C1 are incorporated via acetyl-CoA into the five-carbon isoprenoid unit at a ratio of 1.5 (Koga and Morii, 2007). The much higher ratio 2.4 of acetate-C2 to C1 suggested by our findings implies relatively higher decarboxylation of acetate-C1 during the biosynthesis of GGPP from acetyl-CoA ( Fig. 4; Supporting information Fig. S1). Preferential incorporation of acetate-C2 into membrane lipids is consistent with previous investigations of Methanospirillum hungatei (Ekiel et al., 1983) and Sulfolobus solfataricus (Rosa and Gambacorta, 1986).

Constraints of water-derived hydrogen to lipid biosynthesis
The relatively low water-derived hydrogen contribution observed in M. barkeri (<40%, Fig. 3) compared with acetogens (59%; Valentine et al., 2004a) or bacteria (>40%; Zhang et al., 2009a) suggests that the majority of lipid hydrogen was derived from the substrate (i.e., H 2 , acetate and/or methanol). For hydrogenotrophic methanogenesis, we expect the molar hydrogen contribution of water to lipids (X w ; Equation 4 in Supporting Information Experimental Procedures) to be 100% if there was a very rapid hydrogen exchange between H 2 and HDO (Valentine et al., 2004a). For hydrogenotrophic M. barkeri, the overall hydrogen isotope fractionation (ε l/w = (1 -a w ) × 1000‰) between lipid and water determined in the current study for PMI, AR and OH-AR ranged from −669 to −627‰. It suggests that a large fractionation effect could play an important role for water hydrogen assimilation during hydrogenotrophic growth. The isotope effect for D incorporation into archaeal lipids is substantially larger than observed for growth on H 2 /CO 2 by the acetogenic bacterium Sporomusa sp. (ε l/w = −410‰; Valentine et al., 2004a) as well as other bacteria (Zhang et al., 2009a). First, H 2 could be directly incorporated into intermediate methylene-H 4 MPT via the reduction from H 2 to F420-H 2 by the coenzyme F420 (Thauer et al., 2008). Furthermore, methylene-H 4 MPT is converted to acetyl-CoA. Thus, H 2 gas could partially contribute to lipid H. Although acetogenic bacteria and methanogenic archaea use the same pathway to fix H 2 /CO 2 (i.e., Wood-Ljungdahl pathway; Ragsdale and Pierce, 2008), the higher hydrogenase efficiency (or molecular H 2 utilization efficiency) by methanogenic archaea (~50%; Valentine et al., 2004b) relative to acetogenic bacteria (<20%; Valentine et al., 2004a) may indicate less time for isotope exchange between H 2 and HDO and more H 2 molecular contribution into lipid by M. barkeri. The relatively low a w of M. barkeri growing on H 2 /CO 2 may therefore be achieved by the rapid assimilation of H 2 and thus less time and larger isotopic fractionation during H 2 /HDO equilibration.
During acetoclastic and methylotrophic methanogenesis, M. barkeri appears to rather lower the water-derived hydrogen contribution by incorporating hydrogen from the methyl groups of acetate and methanol into lipids ( Fig. 5;  Supporting Information Fig. S1). We estimate that high portions of hydrogen are contributed through the substrate via the C skeleton of acetyl-CoA (50 of 88 hydrogen atoms in AR,~57%) and NADPH (30 of 88 hydrogen atoms in AR, 34%) may be assimilated into archaeal lipids (Supporting Information Fig. S1; Jain et al., 2014;Rodriguez and Leyh, 2014;Vinokur et al., 2016), with the remainder deriving from the water via the reduction of double bonds catalysed by geranylgeranyl diphosphate reductase (about 9%; 8 of 88 hydrogen atoms, Supporting Information Fig. S1). This contrasts the relative contribution of hydrogen sources to bacterial fatty acid biosynthesis (50%, 25%, and 25% for NADPH, substrate, and water respectively; Valentine, 2009;Zhang et al., 2009a;Wijker et al., 2019). The relatively high contribution of hydrogen atoms derived from sources other than water to archaeal methanogen lipids is consistent with >50% of lipid C being derived from the 13 Clabelled methyl groups of methanol or acetate (see Results). Therefore, the incorporation of water-derived hydrogen into the archaeal methanogen cell membrane is likely buffered by the contribution of substrate methyl group-derived hydrogen to the acetyl-CoA precursor of the MVA pathway, yielding relatively low a w values compared with other microbes (e.g., Valentine et al., 2004a;Zhang et al., 2009a). However, this mechanism alone cannot explain the difference in a w between acetoclastic versus methylotrophic methanogenesis as well as net hydrogen isotope fractionation between lipid and water (Fig. 3), unless the isotope fractionation of metabolic water and its contribution to transhydrogenation reactions in the MVA pathway are completely opposite for the two metabolisms ( Fig. 4 and Fig. S1). A more likely driver of the observed isotope effects is the process of NADPH-mediated exchange of hydrogen ions between water and lipids (Valentine, 2009;Zhang et al., 2009a;Wijker et al., 2019;Supporting Information Fig. S1).
Archaeal methanogens can metabolize substrates to acetyl-CoA via pathways that do not produce NADPH (Weimer and Zeikus, 1978;Fig. 4). Genome analysis suggests that M. barkeri and other methanogens are unlikely to generate NADPH similarly to bacteria, who gain energy via glycolysis, the oxidative pentose phosphate pathway, and/or the TCA cycle (Allen et al., 2009;Zhang et al., 2009a;Sakai et al., 2011;Spaans et al., 2015). Instead, the exergonic reduction of NADP + to NADPH coupled with ferredoxin by transhydrogenase NADH-dependent ferredoxin:NADP oxidoreductase (NfnAB) is likely the main NADPH production pathway for archaea (Huang et al., 2012;Spaans et al., 2015). Ferredoxin is widespread among methanogens containing cytochromes (e.g., Methanosarcina species) and additionally plays an important role in methane production (Thauer et al., 2008). NfnAB is thus coupled to microbial energy conservation (Buckel and Thauer, 2013;Spaans et al., 2015) and is hypothesized to have a significant effect on the hydrogen isotopic compositions of lipids from anaerobic bacteria (e.g., sulfate-reducing bacteria; Leavitt et al., 2016;Osburn et al., 2016). Accordingly, NADPH generated via NfnAB can exchange hydrogen ions with intercellular water (Fig. 4), thereby introducing both kinetic and equilibrium isotope effects. A fast turnover of NADPH presumably results in a kinetic isotope effect, while low turnover in an equilibrium isotope effect. As such, modulations of NADPH pool size, turnover time, and D/H isotopic composition may therefore explain the observed relationship between a l/w exhibited by M. barkeri and energy supply (or power requirement expressed in J/day; Hoehler and Jørgensen, 2013;LaRowe and Amend, 2015;Lever et al., 2015;Fig. 5).
A mass balance of hydrogen sources of lipids in M. barkeri (adopted from Wijker et al., 2019) indicates that the combined hydrogen isotope effects of NADPH production (ε NADPH/w ) during acetoclastic versus methylotrophic methanogenesis are offset by approximately 300‰ and may be as large as −635 AE 60 and −305 AE 155‰ respectively (Supporting Information Text). The relationship between a w and energy demand could result from either a decreasing residual pool of NADPH for lipid biosynthesis during rapid growth on methanol, which would yield smaller ε NADPH/w via Rayleigh fractionation, or an increased amount of time for NADPH to equilibrate with HDO. These findings are consistent with those by Penning et al. (2005), who demonstrated that Relationship of power derived from methanogenesis (J day -1 ) and water hydrogen assimilation factor a w . The power supply is calculated by multiplying the methane production rate (mmol CH 4 day −1 ) by Gibbs free energy potential (kJ mol −1 CH 4 ) under initial conditions (details see Supporting Information Table S5 and Text S1). the carbon isotopic fractionation of methane produced by methanogenic cultures was controlled by the Gibbs free energy change of methane formation and potentially also by the intracellular availability of CO 2 (Valentine et al.,2004b). In summary, our parallel C and H isotope data for hydrogenotrophic, acetoclastic and methylotrophic methanogenesis suggest that ferredoxinregulated hydrogen exchange between extracellular water and intracellular NADPH may ultimately determine the assimilation of hydrogen into membrane lipids.

Implication for the dual SIP approach
Using dual SIP with D 2 O and 13 C-DIC, the ratio of 13 C versus D incorporation into lipids ( 13 C-IC/D 2 O assim ) can distinguish autotrophic versus heterotrophic modes of carbon assimilation (cf. Wegener et al., 2012). Based on experiments utilizing the sulfate-reducing bacterium D. variabilis, 13 C-IC/D 2 O assim values were identified to be 1 and 0.3 during autotrophic and heterotrophic growth respectively  and thus an 13 C-IC/ D 2 O assim value of 0.3 was set as a 'cut-off' value to estimate the contribution of heterotrophic IC fixation. In contrast to experiments by Wegener et al. (2012), M. barkeri exhibited a very different pattern of 13 C and D incorporation with values for 13 C-IC/D 2 O assim of 2.4, 0.9 and 1.1 for growth on H 2 /CO 2 , acetate and methanol respectively (Fig. 6), owing to the relatively lower contribution of water-derived hydrogen to lipid biosynthesis. In order to account for the incorporation of hydrogen from sources other than water into membrane lipids, such as acetyl-CoA, we applied an a w correction factor of 0.33, based on the average of all three growth conditions. Accordingly, 13 C-IC/D 2 O assim values were reduced to 0.78, 0.28 and 0.30 for growth on H 2 /CO 2 , acetate and methanol respectively. This indicates that future estimates of archaeal lipid production based on D incorporation should consider the water hydrogen assimilation factor (a w ) (Kopf et al., 2016;Wegener et al., 2016) to distinguish between archaeal auto-and hetero-trophy. We suggest a correction factor a w of 0.67  and 0.33 (this study) to be used while estimating bacterial and methanogenic archaeal lipid production in marine environmental samples respectively. In conclusion, the application of dual SIP employing D 2 O and 13 C-DIC should consider variations of water-derived hydrogen discrimination caused by different microorganisms as wells as carbon substrates, which complicates the use of 13 C-IC/D 2 O assim as a proxy estimating the contribution of heterotrophic versus autotrophic lipid sources. thank X. P. Mollar, J. Wendt, H. Taubner, E. Schefuß and M. Kölling at MARUM for their support for δ 13 C and δD measurements. We also thank M. Lever in the Department of Environmental System Sciences at Swiss Federal Institute of Technology Zurich, Switzerland, for Gibbs energy calculation. We are very grateful to Sebastian Kopf and an anonymous reviewer for their constructive comments and stimulating suggestions that substantially improved the manuscript. Appendix 2. Supplementary figures and tables. Fig. S1. Diether lipid biosynthetic pathway emphasizing carbon and hydrogen sources (Jain et al., 2014;Rodriguez and Leyh, 2014;Vinokur et al., 2016). Acetoacetyl-CoA tautomerizes and exchanges hydrogen with intercellular water, resulting in the hydrogen isotopic modification (Rodriguez and Leyh, 2014;Wang et al., 2013). Green, brown, and black H atoms are from water, NADPH and acetyl-CoA respectively. Bold hydrogen atoms refer to the possibility of isotopic exchange with water via tautomerization (Rodriguez and Leyh, 2014). The hydrogen atoms in the glycerol moiety labelled with question marks (?) may be from NADPH (Koga and Morii, 2007). Fig. S2. Calibration curve of acetate and methanol standards on LC-IRMS. The x axis refers to carbon amount of acetate or methanol, the molar concentration (μM) of acetate and methanol was converted according to the carbon ratio of acetate and methanol of 0.40 and 0.374 respectively. Fig. S3. Calibration curve of expected δD of water amended with different D 2 O and measured δD. Fig. S4. Hydrogen isotope ratio (R(D/H)) between NADPH and water for growth on acetate and methanol. Table S1. δ 13 C values of methane and acetate at time of harvest, and δ 13 C and δD values of lipids produced by M. barkeri during the growth on labelled substrates (ca. 0.5%) and deuterated water (D 2 O, 1%, 0.5% and 0.2%). n.d. = not detected. n.a. = not analysed. Table S2. The Carbon isotope fractionation factor of lipid produced by M. barkeri grown on non-labelled H 2 /CO 2 , acetate and methanol. n.d., not detected. n.r., not reported. Table S3. The hydrogen isotope fractionation factor between lipids and water produced by M. barkeri when grown on different substrates in natural water medium (δD = −50.6‰; red dash line). n.d., not detected. Table S4. The mix ratio (%) of methane in headspace during incubation. The acceleration (r, % d −1 ) of methane accumulation in headspace was calculated during periods of linear increases (black values) in individual incubations via the linear regression. This rate was converted to methane production rate (r 2 , μmol d −1 ; Table S5) based on the ideal gas equation with headspace of 150 ml, constant pressure of 1 bar and temperature (308.15 K). The grey text indicates values that were not within the range of linear increase of methane and were excluded from rate calculations (cf. Penger et al., 2014). Table S5. Summary of methane production rate (r 2 ), Gibbs energy, power supply conditions and water hydrogen assimilation factor (a w ).