Evidence for corrin biosynthesis in the last universal common ancestor

Abstract

Corrinoids are cobalt‐containing tetrapyrroles. They include adenosylcobalamin (vitamin B₁₂) and cobamides that function as cofactors and coenzymes for methyl transfer, radical‐dependent and redox reactions. Though cobamides are the most complex cofactors in nature, they are essential in the acetyl‐CoA pathway, thought to be the most ancient CO₂‐fixation pathway, where they perform a pterin‐to‐cobalt‐to‐nickel methyl transfer reaction catalyzed by the corrinoid iron–sulphur protein (CoFeS). CoFeS occurs in H₂‐dependent archaeal methanogens, the oldest microbial lineage by measure of physiology and carbon isotope data, dating corrinoids to ca. 3.5 billion years. However, CoFeS and cobamides are also essential in the acetyl‐CoA pathway of H₂‐dependent bacterial acetogens. To determine whether corrin biosynthesis was established before archaea and bacteria diverged, whether the pathways arose independently or whether cobamide biosynthesis was transferred from the archaeal to the bacterial lineage (or vice versa) during evolution, we investigated phylogenies and structural data for 26 enzymes of corrin ring and lower ligand biosynthesis. The data trace cobamide synthesis to the common ancestor of bacteria and archaea, placing it in the last universal common ancestor of all lifeforms (LUCA), while pterin‐dependent methyl synthesis pathways likely arose independently post‐LUCA in the lineages leading to bacteria and archaea. Enzymes of corrin biosynthesis were recruited from preexisting ancient pathways. Evolutionary forerunners of CoFeS function were likely Fe‐, Ni‐ and Co‐containing solid‐state surfaces, which, in the laboratory, catalyze the reactions of the acetyl‐CoA pathway from CO₂ to pyruvate under serpentinizing hydrothermal conditions. The data suggest that enzymatic corrin biosynthesis replaced insoluble solid‐state catalysts that tethered primordial CO₂ assimilation to the Earth's crust, suggesting a role for corrin synthesis in the origin of free‐living cells.

Cobalamin biosynthesis can be traced back to the last universal common ancestor. After a stage of chemical evolution of the acetyl‐CoA pathway, the enzymatic synthesis of cobalamin evolved, allowing life to be independent of the insoluble solid‐state catalysts that tethered primordial CO₂ assimilation to the Earth's crust, suggesting a role for corrin synthesis in the origin of free‐living cells, acetogens and methanogens.

Abbreviations

ACS

acetyl‐CoA synthase

AcsE

Methyl‐H₄F:CoFeS methyltransferase

BchE

anaerobic magnesium‐protoporphyrin IX monomethyl ester cyclase

BchQ

chlorophyll/bacteriochlorophyll a synthase

BchR

bacteriochlorophyllide d C‐12(1)‐methyltransferase

BioD

dethiobiotin synthetase

BtuR/CobA/PduO

corrinoid adenosyltransferase

BzaABCDEF

5‐hydroxybenzimidazole synthase

CbiA

cobyrinic acid a,c‐diamide synthetase

CbiB

adenosylcobinamide‐phosphate synthase

CbiC

precorrin‐8X/cobalt‐precorrin‐8 methylmutase

CbiD

cobalt‐precorrin‐5B (C1)‐methyltransferase

CbiE

cobalt‐precorrin‐7 (C5)‐methyltransferase

CbiF

precorrin‐4/cobalt‐precorrin‐4 C11‐methyltransferase

CbiG

cobalt‐precorrin 5A hydrolase

CbiH

precorrin‐3B C17‐methyltransferase/cobalt‐factor III methyltransferase

CbiJ

precorrin‐6A/cobalt‐precorrin‐6A reductase

CbiL

precorrin‐2/cobalt‐factor‐2 C20‐methyltransferase

CbiT

cobalt‐precorrin‐6B (C15)‐methyltransferase

CbiX_L/CbiK/CbiXs

sirohydrochlorin cobaltochelatase

CfbB

Ni‐sirohydrochlorin a,c‐diamide synthase

CobB

NAD‐dependent protein deacetylase/lipoamidase

CobC/CobZ

adenosylcobalamin/alpha‐ribazole phosphatase

CobQ

adenosylcobyric acid synthase

CobS

adenosylcobinamide‐GDP ribazoletransferase

CobU/CobY

adenosylcobinamide‐phosphate guanylyltransferase

CODH

carbon monoxide dehydrogenase

CoFeS

corrinoid iron–sulphur protein

DMB

5,6‐dimethylbenzimidazole

DsrAB

dissimilatory sulphite reductase

EutT

ethanolamine utilization cobalamin adenosyltransferase

FdhA

formate dehydrogenase

Fhs

N10‐formyl‐H₄F synthetase

FolD

5,10‐methenyl‐H₄F cyclohydrolase/dehydrogenase

Ftr

formylmethanofuran:H₄MPT formyltransferase

FwdBD

formylmethanofuran dehydrogenase

Ga

Giga annum

H₄F

tetrahydrofolate

H₄MPT

tetrahydromethanopterin

HBI

5‐hydroxybenzimidazole

HemH/PpfC/CpfC

ferrochelatase

HGT/LGT

horizonal/lateral gene transfer

Hmd

5,10‐methenyltetrahydromethanopterin hydrogenase

ICGs

informational core genes

LUCA

last universal common ancestor

MazG

nucleoside triphosphate diphosphatase

Mch

methenyl‐H₄MPT cyclohydrolase

Mer

5,10‐methylenetetrahydromethanopterin reductase

MetF

5,10‐methylene‐H₄F reductase

MetF

5,10‐methylenetetrahydrofolate reductase

MeTr

methyltransferase present in some acetogens

MinD

septum site‐determining protein

MoCo

molybdenum cofactor

Mtd

F₄₂₀‐dependent methylene‐H₄MPT dehydrogenase

ParA

chromosome partitioning protein

PFOR

pyruvate:ferredoxin oxidoreductase,

SAM

S‐adenosylmethionine

SirB/ShfC

sirohydrochlorin ferrochelatase

SirC

precorrin‐2 dehydrogenase

ThiC

phosphomethylpyrimidine synthase

UroIII

uroporphyrinogen III

UroM/SuMT/SirA

uroporphyrin‐III C‐methyltransferase

Introduction

Corrinoids are nature's most complex cofactors and coenzymes. They include cobalt‐containing cobamides, which differ with respect to the lower ligand: 5,6‐dimethylbenzimidazole in cobalamin, and 5‐hydroxybenzimidazole, 5‐methoxybenzimidazole, 5‐methylbenzimidazole, benzimidazole, 5‐methoxy‐6‐methylbenzimidazole, phenol, p‐cresol, 2‐methylsulfinyladenine, 2‐methylsulfonyladenine, 2‐methyladenine or adenosine in other cobamides [1, 2, 3, 4, 5, 6]. These cobalt‐containing corrins are mainly involved in methyl transfer [7, 8, 9, 10] and in radical‐based reactions [11], including the generation of deoxyribose nucleotides for DNA by adenosylcobalamin‐dependent ribonucleotide reductase [12, 13, 14]. Corrinoid‐dependent enzymes are also involved in a variety of other processes, such as the biosyntheses of amino acids (Met, [10, 15, 16]), queuosine [17], bacteriochlorophyll [18], hopanoids [19], polytheonamide [10, 20], antibacterials and antivirals [21, 22], as well as the detoxification of aromatic and aliphatic chlorinated organics [16, 23], and a range of fermentations: amino acids [15, 24, 25, 26, 27], one‐carbon compounds (e.g., methylamine, methanol and methylthiol [16]), ethanolamine and choline [16, 28, 29], nicotinic acid [16, 30], 1,2‐diols [16, 31, 32, 33] and fatty acids [16, 26, 27]. Corrinoids are required by many organisms but are synthesized by few [34, 35, 36, 37]. The biosynthetic pathway in anaerobes entails over 20 enzymatic steps if starting to count from uroporphyrinogen III, the first macrocyclic intermediate and the universal precursor of tetrapyrrole biosynthesis [38], the number of steps to the final product depending on the nature of the lower ligand [39]. Though the genes for corrin synthesis can be laterally transferred across lineages, and some prokaryotes have partial cobalamin synthesis, for example, from cobyric acid to adenosylcobalamin [40], by far the most common strategy to satisfy corrin requirements is to obtain the cofactor from corrin‐producing organisms via the environment [34]. The list of corrin auxotrophs includes prokaryotes [35, 36, 37, 41], algae [37, 42], protists [37, 43] and humans who strictly require cobalamin for only two enzymatic reactions: the isomerization of methylmalonyl‐CoA to succinyl‐CoA and the synthesis of methionine from homocysteine [44]. Corrins are thought to be ancient [45, 46], and earlier studies traced some corrin‐dependent enzymes to the last universal common ancestor, LUCA [47], but the issue of whether the enzymes of corrin biosynthesis themselves trace to LUCA is so far unresolved.

The evidence for corrin antiquity stems from its essential role in the acetyl‐CoA pathway, the most ancient among CO₂‐fixing pathways [48, 49, 50, 51]. In the acetyl‐CoA pathway of H₂‐dependent acetogens (bacteria) and H₂‐dependent methanogens (archaea), the corrinoid iron–sulphur protein (CoFeS) catalyzes a cobamide‐dependent cobalt‐to‐nickel methyl transfer reaction that is essential for acetyl‐CoA synthesis [52, 53] (Fig. 1A). The geological age of the corrins is given by methanogens, which are the most ancient lineage of microbes by the measure of physiology [46] and carbon isotope data. Methane isotopes [65] provide the earliest evidence for life on Earth, tracing methanogenesis, hence the acetyl‐CoA pathway, hence an ancestral CoFeS (judging by the homology between bacterial and archaeal CoFeS), hence corrins, into rocks 3.8 billion years of age. Isotopically ultralight carbon is also reported in 3.8 Ga sediment carbon [66, 67], which suggests the presence of the acetyl‐CoA pathway [68] without discriminating an acetogen or methanogen source. Phylogenetic studies implicate acetogens and methanogens as the most ancient lineages among modern bacteria and archaea, respectively [47]. Both acetogens and methanogens inhabit H₂‐producing hydrothermal systems today [69, 70] and might present windows into the physiology and habitat of the first cells [47, 69, 70, 71, 72].

Fig. 1.

The role of corrins in the acetyl‐CoA pathway. (A) The acetyl‐CoA pathway, redrawn from references [49, 54]. Methyl synthesis: pterin‐dependent methyl synthesis in the acetyl‐CoA pathway of bacteria and archaea from H₂ and CO₂ [49]. ACS, acetyl‐CoA synthase [53]; CODH, carbon monoxide dehydrogenase [55]; CoFeS, corrinoid iron–sulphur protein [52]; PFOR, pyruvate:ferredoxin oxidoreductase [54]. The methyl group is highlighted in red, the corrinoid cobalt atom in blue, the proximal nickel atom of the ACS active site [53] in green and enzyme names or steps are given in sepia letters. (B) Corrinoid‐dependent methyl transfer from a pterin cofactor to the active site metal cluster (A‐cluster) of acetyl‐CoA synthase (ACS) (PDB ID: 7ZKJ [56]) in the acetyl‐CoA pathway of acetogens and methanogens. The corrinoid iron–sulphur protein CoFeS catalyzes the methyl‐transfer reaction, in some cases aided by a separate methyltransferase (MeTr) [52, 57]. H₄F and H₄MPT: tetrahydrofolate and tetrahydromethanopterin, respectively. R1 = –H in H₄F and –CH₃ in H₄MPT.For the chemical structure of R2,in H4F and H4MPT, , see[58]. Methylcobalamin is shown in the base‐off configuration [59], with –R3 replacing the nucleotide loop. Methyl groups derived from S‐adenosylmethionine (SAM) are shown in blue. Fe atoms are shown in orange, S atoms in yellow and Ni atoms in green. Lower left: Examples of two cobamide lower ligands (–R3 in A); 5‐hydroxybenzimidazole (HBI) is found in the cobinamides of some methanogens [39], while 5,6‐dimethylbenzimidazole (DMB, the lower ligand base of cobalamin) is found in some bacteria [52]. (C) Native metals catalyze formate, acetate and pyruvate synthesis from H₂ and CO₂ over Ni⁰, Fe⁰, Co⁰, Ni₃Fe and other alloys [60, 61, 62, 63, 64], identifying a nonenzymatic, geochemical precursor of corrin‐dependent methyl transfer shown in panels A and B. More is known about the enzymatic reaction mechanisms [53] than about the metal‐catalyzed forerunners [60, 61, 62, 63, 64]. The catalyst‐bound methyl groups in panel C are probable intermediates in the reaction mechanism, as methanol and methane are sometimes observed as products [60, 61]. In some experiments, Fe₃S₄ and Fe₃O₄ are effective as catalysts, but it cannot be excluded that these catalysts are reduced to Fe⁰ at some sites during the reaction with H₂.

In addition to physiology [48, 49, 50, 51], isotopes [65, 66, 67] and phylogeny [47, 69, 70], metal catalysts provide an independent line of evidence in favour of the antiquity of the acetyl‐CoA pathway. Serpentinizing (H₂‐producing) hydrothermal systems synthesize formate [73], an intermediate of the acetyl‐CoA pathway [60], and methane, the end product of methanogenesis, from abiotic reactions of H₂ and CO₂ [74]. Though the catalysts that facilitate formate and methane synthesis in vents have not yet been identified in situ, native (zero‐valent) transition metals are thought to be the catalysts [60], for two reasons. First, Fe⁰, Ni⁰, Co⁰ and their alloys are naturally deposited in serpentinizing hydrothermal vents because of their highly reducing conditions [75]. Second, Fe⁰, Ni⁰, Ni₃Fe and silica‐supported Co‐Fe alloys can replace the proteins that synthesize formate, acetate, pyruvate from H₂ and CO₂ via the acetyl‐CoA pathway [60, 61, 62, 63, 64] (Fig. 1C). In water and in the absence of enzymes, the solid‐state metals generate methyl groups from H₂ and CO₂, producing formate, acetate, pyruvate and methane, replacing the function of over 100 enzymes [76]. Furthermore, Ni⁰ readily forms amino acids from 2‐oxoacids, NH₃ and H₂ [77], and Fe⁰ can reduce ferredoxin, functionally substituting for flavin‐based electron bifurcation, a complex process that requires multi‐enzyme systems capable of coupling exergonic and endergonic reactions [78] and reducing the low‐potential physiological donor of electrons in the acetyl‐CoA pathway, ferredoxin, with electrons from H₂ [79].

Countering the case for corrin antiquity, its biosynthetic pathway is the longest, and the product is structurally the most complex of any known tetrapyrrole [38, 80, 81], raising the question of its possible functional precursor during biochemical evolution (Fig. 1). Corrin synthesis itself holds no clear evidence for antiquity, other than its basal position in tetrapyrrole biogenesis [46, 82]. While some cofactors such as thiamine or pyridoxal phosphate participate in their own biosynthesis, forming small autocatalytic cycles [71], that is not the case for corrins. The pathway involves the introduction of eight methyl groups to the tetrapyrrole backbone, one of which aids methylene elimination and contraction of the porphyrin to the corrin ring [83, 84]. The seven methyl groups present in the corrinoid structure [9] (Fig. 1B) are donated by S‐adenosylmethionine (SAM), a cofactor of methyl transfers and radical reactions [85], yet SAM does not substitute for cobamide in either the CoFeS reaction [52] or in the pumping reaction at MtrA‐H in methanogens [86].

The age of methanogens and the identification of functional abiotic homologues of the acetyl‐CoA pathway point to simple environmental precursors of corrin‐dependent methyl transfer (Fig. 1C) but do not directly address corrin biosynthesis antiquity. To determine whether corrin synthesis required for the acetyl‐CoA pathway traces to the common ancestor of bacteria and archaea – the last universal common ancestor (LUCA) – or was the result of later lateral gene transfers [47], we have investigated the evolutionary history of the enzymes of corrin biosynthesis in acetogens and methanogens [38], including the synthesis of the lower ligand (Fig. 1B) [39] and CoFeS [87, 88].

Results and Discussion

There are two pathways of corrin synthesis, an O₂‐dependent (aerobic) pathway and an oxygen‐sensitive (anaerobic) pathway, where the aerobic pathway requires oxygen for ring contraction and cobalt insertion [38, 81]. Only the enzymes of the assumed‐to‐be‐more‐ancient anaerobic pathway are of interest in early biochemical evolution for three reasons. First, O₂ is a product of cyanobacterial photosynthesis, which was lacking [89] in the environment at the origin of the acetyl‐CoA pathway and methanogens > 3.5 billion years ago [65, 82]. Second, both abiotic [90] and enzymatic reactions to pyruvate synthesis along the acetyl‐CoA pathway are extremely O₂‐sensitive due to the involvement of FeS clusters in electron transfer and other radical‐generating reactions [91, 92], such that the corrins required for the pathway arose in the absence of O₂ (as with chlorophyll synthesis, the O₂‐dependent corrin synthesis reactions arose in response to O₂‐dependent inhibition of radical‐dependent reactions [93]). Third, acetogens and methanogens themselves, which depend upon CoFeS and the acetyl‐CoA pathway for growth [48, 49, 78], are extremely O₂‐sensitive and use the anaerobic corrin synthesis pathway. Because physiology [46], phylogeny [47, 69, 70], and carbon isotopes [65, 66, 68] point to acetogens and methanogens as the most ancient bacterial and archaeal lineages, respectively, and because both require cobamide [7, 8], the corresponding genes of anaerobic corrin biosynthesis were extracted from the genome of the model acetogen Moorella thermoacetica [94] and from the genome of the methanogen Methanococcus maripaludis [95].

Corrinoids are not generally viewed as likely products of prebiotic synthesis because of their complex structure and biosynthesis [38, 39, 96], which is outlined in Fig. 2 (and in Fig. S1; see Supporting Information for a description of the initial biosynthetic steps). Of note, the initial steps, from Uro‐III to the formation of sirohydrochlorin, are common to siroheme biosynthesis [99], and homologues of CbiA and CbiX are involved in F₄₃₀ biosynthesis [100, 101], highlighting the duplications that occurred during tetrapyrrole evolution. At a late stage of the pathway (Fig. 2), adenosylation (catalyzed by BtuR/CobA) introduces an adenosyl moiety as the upper ligand at this stage of the adenosylcobalamine synthesis pathway [38]. In the cobamides that are employed in the CoFeS reaction, the adenosyl upper ligand is missing, the corresponding Co coordination site being occupied by water or by the methyl group during the methyl transfer reaction [52]. Accordingly, some methanogens (and other archaea) lack BtuR/CobA homologues [36, 102]. The biosynthetic route depicted in Figs 2 and 3 follows that described for adenosylated intermediates [38, 96], the corresponding genes, except for BtuR/CobA, being present in methanogens and acetogens that are included in this study.

Fig. 2.

The anaerobic cobalamin biosynthesis pathway. (A) The schematic shows the abbreviations and EC numbers of enzymes involved in the biosynthesis of adenosylcobalamin as found in the KEGG database, starting with the conversion of uroporphyrinogen III to precorrin‐2 and the early insertion of cobalt. Lower ligand synthesis is shown in Fig. S1. (B) In previously published literature, other abbreviations like UroM [97] or SuMT [98] have been used to refer to uroporphyrinogen III methyltransferase (SirA). Abbreviations and full names of enzymes found in this article are displayed in the schematic. A short description of the steps and citations for the individual reactions is given in the Suppporting Information.

Fig. 3.

Structurally homologous and divergent enzymes in anaerobic cobalamin synthesis and the acetyl‐CoA pathway. Left panel: cobalamin synthesis pathway. Right panel: the acetyl‐CoA pathway. Structures from acetogens (blue) and methanogens (salmon) are shown superimposed [103], while non‐homologous structures are shown adjacent. Adjacent structures of the same colour point to non‐homologous enzymatic variants in the same prokaryotic domain, such as in the case of Hmd. Hmd is the Fe‐only hydrogenase of archaea that directly reduces methenyl‐H₄MPT to methylene‐H₄MPT, is expressed under nickel limitation, uses a unique iron‐guanylylpyridinol cofactor and is not homologous to any enzymes of the pathways shown here [104]. Grey shading highlights non‐homologous enzymes in methanogen‐acetogen comparisons. Structures were modelled with AlphaFold [105], except for FwdBD (PDB ID: 5T5I [106]) and CODH/ACS (PDB IDs: 3CF4 and 1MJG [107, 108]). Methanogen ACS and methyl synthesis were modelled based on sequences from Methanothermobacter marburgensis. The acetogen methyl branch is from Moorella thermoacetica. In cobalamin biosynthesis, methanogen sequences are from Methanococcus maripaludis (and Methanosarcina barkeri in the case of CobA, which is specific to adenosylcobalamin synthesis and missing in many methanogens; see text) and acetogen sequences from M. thermoacetica. In multimers, only the structurally homologous subunit was shown for clarity. TM‐scores of structural alignments can be found in Table S1. For the nomenclature of enzymes involved in anaerobic cobalamin synthesis, see the legend and pannel (B) of Fig. 2. See also Svetlitchnaia et al. [52] for a structural comparison of CoFeS with structurally similar proteins.

The anaerobic synthesis of the cobalamin lower ligand dimethylbenzimidazole (DMB) is performed by the bzaABCDE gene products [39], starting from 5‐aminoimidazole ribotide (AIR), an intermediate of thiamine and purine biosynthesis. In some organisms, only the product of bzaF, a radical SAM enzyme homologous to bzaA, bzaB and thiC products, is present [39, 109], generating 5‐hydroxybenzimidazole (HBI) as the cobamide lower ligand (Fig. 1C). Other organisms employ different intermediates of lower ligand synthesis [39], and the lower ligands of cobamides isolated from acetogens [7] and methanogens vary across species [8]. Here we analysed only the biosynthesis of the lower ligand DMB since it is found in both archaeal and bacterial organisms and has a large impact in marine settings [110]. Moreover, it was shown that at least in Salmonella enterica, the incorporation of DMB or of adenine as lower ligands is performed by the same enzymes, and, in DMB‐limiting conditions, adenine can replace DMB [111].

Homologies trace corrin synthesis to the acetogen‐methanogen common ancestor

Alignments of structures generated from amino acid sequences with AlphaFold [103, 105] show that the structures of enzymes for methanogen and acetogen cobamide synthesis superimpose well in all cases except one, indicating homology of the corrin synthesis pathways in methanogens and acetogens (Fig. 3, left panel). The same is true for the two enzymes of the carbonyl branch of the acetyl‐CoA pathway, acetyl‐CoA synthase (ACS) and carbon monoxide dehydrogenase (CODH) (Fig. 3, lower right). In addition, the subunits of the corrinoid iron–sulphur protein (CoFeS) that binds cobamide also appear homologous in the methanogen–acetogen comparison.

CoFeS catalyzes two cobamide‐dependent methyl‐transfer reactions, one from the pterin cofactor to the cobalt atom of the cobamide cofactor and a second from Co to the active‐site Ni atom of ACS, in some cases aided by an additional, separate methyltransferase in acetogens [52, 57, 112, 113]. The two CoFeS subunits of Methanococcus maripaludis [88] and M. thermoacetica [87] were scored as present in a target genome if the genome had best hits for both subunits, a condition fulfilled in many cases. CoFeS showed a similar phylogenetic distribution as ACS (EC 2.3.1.169) and CODH (EC 1.2.7.4) (Fig. 4), which were previously known to be homologous across the bacterial–archaeal divide based on sequence similarity [47, 114]. In methanogens, CoFeS is often part of a larger CODH/ACS complex [107]. The CoFeS subunits of Methanococcus maripaludis and M. thermoacetica show 31.4% and 37.8% amino acid sequence identity in global pairwise comparisons (methanogenic CODH/ACS complex subunits delta and gamma, respectively). The sequence homology of acetogenic and methanogenic CoFeS was previously reported [14, 42]. Their structural alignments using AlphaFold‐modelled structures [103, 105] (Fig. S2) showed TM‐scores of 0.931 (methanogen CODH/ACS complex subunit delta, acetogen CoFeS small subunit) and 0.836 (methanogen CODH/ACS complex subunit gamma, acetogen CoFeS large subunit), normalized by the length of the respective acetogen protein. These scores indicate that the structures share a common fold, confirming their homology [115]. The methanogen corrinoid‐iron sulphur protein CoFeS (AcsC) also shares a common fold with the methyl‐H₄F:CoFeS methyltransferase (AcsE), which aids acetogen CoFeS in the methyl transfer step from the pterin to the corrin cobalt (Fig. 3), as previously reported [52].

Fig. 4.

Presence‐absence pattern of enzymes of the anaerobic biosynthesis of cobalamin, acetyl‐CoA synthase, carbon monoxide dehydrogenase, CoFeS and lower ligand synthesis enzymes in methanogens and acetogens. Best blast hits for all archaeal and bacterial query sequences (see Materials and methods) were scored in data sets of 80 methanogens and 189 acetogens. Analysed enzymes are displayed on the X‐axis, and groups of methanogens (red) and acetogens (blue) are shown on the Y‐axis. Black indicates presence and white indicates absence of the enzyme. The column (*) labelled “CobA” covers homologs of CobA, PduO and EutT. Nonorthologous alternatives of CobU (CobY) and CobC (CobZ) [36] were also included.

The active corrins in the acetyl‐CoA pathway of acetogens and methanogens accept methyl groups at the open Co coordination site. In adenosylcobalamin, the central cobalt is coordinated by an upper ligand adenosyl moiety that is introduced by the corrinoid adenosyltransferase BtuR/CobA [116], which can also be replaced by two alternative enzymes, PduO [117] or EutT [118]. BtuR/CobA was generally missing in methanogen genomes from the Methanococcales (including our query species Methanococcus maripaludis, Fig. 4) and Methanobacteriales, which contain hydrogenotrophic species. This is consistent with a lack of adenosylcobalamin in several studied species [7]. CobU (adenosylcobinamide‐phosphate guanylyltransferase) and its non‐orthologous alternative CobY [36] lack structural homology across the methanogen‐acetogen divide (Fig. 3) and, like CobS (adenosylcobinamide‐GDP ribazoletransferase), are not widely distributed in reciprocal comparisons among acetogens and methanogens represented here (Fig. 4). Among the enzymes involved in the synthesis of the cobalamin lower ligand, only BzaF, which catalyzes the synthesis of 5‐hydroxybenzimidazole (HBI) [39], is well conserved across the bacterial–archaeal divide, reflecting natural variation in lower ligand synthesis observed in methanogens and acetogens [7, 8] relative to Eubacterium limosum, the species in which lower ligand synthesis was identified [39]. Archaeal and bacterial CobT proteins are not homologous, as previously reported [119]. The distribution of the genes for enzymes of corrin biosynthesis across the bacterial–archaeal divide at the 25% amino acid sequence identity threshold is shown in Fig. 4.

Independent origins of enzymatic methyl synthesis in acetogens and methanogens

The methyl synthesis branch of the acetyl‐CoA pathway provides a continuous supply of methyl groups for the CoFeS reaction in acetogen and methanogen carbon assimilation. The enzymes of methyl synthesis and those underpinning the cofactors synthesis, tetrahydrofolate (H₄F) in bacteria and tetrahydromethanopterin (H₄MPT) in archaea, were previously found to be non‐homologous across the bacterial–archaeal divide [50] by the measure of amino acid sequence conservation. Because the methanogen enzyme formylmethanofuran dehydrogenase reveals structural homology to formate dehydrogenase of the acetogen pathway [116], which was not detected at the amino acid sequence level, we reinvestigated the potential common ancestry of acetogen and methanogen methyl synthesis pathways on the basis of structural homology. Identifiers for genes of the methyl branch of the acetyl‐CoA pathway for the methanogen Methanothermobacter marburgensis and the acetogen M. thermoacetica were taken from previous studies [49, 94]. Archaeal formylmethanofuran dehydrogenase (FwdBD) and bacterial formate dehydrogenase (FdhA) were not reinvestigated as structural similarity has already been shown [104]. Alignments of structures modelled with AlphaFold [103, 105] showed that the reduction of the pterin‐bound methylene to a methyl group is also catalyzed by enzymes that share a common fold, with a TM‐score > 0.67 (Fig. 3). The electron donor for methylene‐H₄MPT reductase in methanogens (Mer) is the cofactor F₄₂₀; the donor for the acetogen 5,10‐methylene‐H₄F reductase (MetF) is unclear [120].

The remaining two steps of methyl synthesis in acetogens, N ¹⁰‐formyl‐H₄F synthetase (Fhs) and 5,10‐methenyl‐H₄F cyclohydrolase/dehydrogenase (FolD), correspond to three steps in methanogens, catalyzed by formylmethanofuran:H₄MPT formyltransferase (Ftr), methenyl‐H₄MPT cyclohydrolase (Mch) and F₄₂₀‐dependent methylene‐H₄MPT dehydrogenase (Mtd). These conversions are catalyzed by structurally non‐homologous enzymes in acetogens and methanogens (TM‐score < 0.35 in all cases) (Fig. 3). The non‐homology of these methyl synthesis enzymes in acetogens and methanogens, their use of different redox cofactors at individual steps, in addition to the divergent syntheses of their corresponding pterin cofactors, indicates that the pathways arose independently in the acetogen and methanogen lineages, respectively [50]. This in turn suggests that the common ancestor of acetogens and methanogens was dependent upon a supply of chemically reactive methyl groups provided by the environment [47, 50] and that the use of environmental methyl groups in acetyl and methane synthesis is more ancient than the pterin‐dependent biochemical pathways used to synthesize the methyl groups. Such geochemically supplied reactive methyl groups could include methanol, a soluble substrate for the acetyl‐CoA pathway in methanogens and acetogens [49] and a synthesis product from H₂ and CO₂ in geochemical analogues of the acetyl‐CoA pathway [60], hydrothermal methyl sulphide [71, 121], methyl thioacetate [122] or methyl groups bound to solid‐state metal catalysts that synthesize methyl groups from H₂ and CO₂ [60, 62, 63, 64].

The corrin synthesis pathways in acetogens and methanogens consist of homologous enzymes (Figs 3 and 4) that provide the cofactor for the corrin‐binding methyltransferase, CoFeS, which is also homologous (Fig. S2), as are the enzymes of the acetyl synthesis branch of the acetyl‐CoA pathway (CODH/ACS) (Fig. 3). In the methyl synthesis branch, the reduction of CO₂ to a formyl group (in methanogens) or formate (in acetogens) and the reduction of pterin‐bound methylene to a methyl group are both steps catalyzed by structurally homologous enzymes across the acetogen‐methanogen divide (Fig. 3, Fig. S2). Two steps of methyl synthesis are not homologous in the acetogen/methanogen comparison, indicating that these two steps “resulted from convergent developments” [48], after the divergence of the acetogen and methanogen lineages.

One might challenge the view that the acetyl‐CoA pathway is the oldest pathway of CO₂ fixation [48, 49, 50, 51]. However, the evidence in favour of its antiquity is unambiguous. There are seven known pathways of CO₂ fixation (the Calvin cycle, the reverse citric acid cycle, the acetyl‐CoA pathway, the dicarboxylate/4‐hydroxybutyrate cycle, the 3‐hydroxypropionate/4‐hydroxybutyrate cycle, the 3‐hydroxypropionate bicycle and the glycine reductase pathway [123]), but only one of them, the acetyl‐CoA pathway, occurs in both bacteria and archaea [49, 124]. Moreover, only one of those seven, again only the acetyl CoA pathway, operates in vitro to form its main products – formate, acetate and pyruvate [49, 90] – from H₂ and CO₂ without enzymes or cofactors, using only metals or minerals that occur in hydrothermal vents as catalysts [60, 61, 62, 63, 64, 125]. None of the CO₂‐fixing reactions of the other six pathways have been shown to operate in vitro without enzymes. In addition, the amount of pyruvate formed via metal‐catalyzed, non‐enzymatic versions of the acetyl‐CoA pathway can reach up to 200 μm [62], which is the physiological concentration of pyruvate in the cytosol of clostridial acetogens growing from H₂ and CO₂ via the acetyl‐CoA pathway [126]. Finally, on the point of the antiquity of the acetyl CoA pathway, the same metal catalysts that allow it to unfold without ATP, enzymes, or cofactors convert 2‐oxoacids to amino acids at room temperature [77], reduce NAD⁺ [127] and reduce ferredoxin with H₂ [79]. The other six pathways consume ATP during CO₂ fixation [49]; the acetyl‐CoA pathway is integral to ATP synthesis from H₂ and CO₂ in acetogens and methanogens [78, 128]. These findings offer clear experimental evidence that the acetyl‐CoA pathway is the most ancient among known CO₂‐fixing pathways, and no comparable findings for any of the other six pathways have been reported.

Phylogenies trace corrin synthesis to the acetogen‐methanogen common ancestor

To determine whether the corrin pathway of methanogens and acetogens traces to the last universal common ancestor, we constructed alignments and phylogenies based on genes for enzymes of anaerobic cobalamin biosynthesis and lower ligand biosynthesis (see Materials and methods) (Fig. 5). We recovered three classes of phylogenies: (a) bacterial–archaeal monophyly (nine cases: SirA, SirC, CbiX, CbiE, CbiB, CobS, BzaA, BzaB and BzaF), (b) bacterial–archaeal monophyly violated by a few recent transfers of bacterial genes into the same methanogen lineages or by transfers of archaeal genes into the acetogen lineage (seven cases: CbiJ, CbiT, CbiC, CbiA, CobA, CobQ and CobU/Y) and (c) highly interleaved phylogenies suggesting multiple LGT events across the bacterial–archaeal divide (nine cases: CbiL, CbiH, CbiF, CbiG, CbiD, BzaC, BzaD, BzaE and CobC).

Fig. 5.

Unrooted gene trees of enzymes involved in the anaerobic biosynthesis of cobalamin and lower ligand synthesis. The overview shows single‐gene trees containing sequences of methanogens (red) and acetogens (blue). Calculations of trees are based on the best blast hits of acetogen and methanogen query sequences against genome data sets of methanogens and acetogens. The phylogenies were computed with raxml version 8.2.12 and are grouped according to their category (monophyly, rarely violated monophyly, highly interleaved) and arranged corresponding to the order displayed in Fig. 2.

The nine cases of domain monophyly trace anaerobic corrin and lower ligand synthesis to the common ancestor of bacteria and archaea. Among those, SirA and SirC are involved in siroheme biosynthesis as well, a cofactor also identified in Weiss et al. [47] as being present in LUCA. The seven cases of rarely violated monophyly also trace anaerobic corrin and lower ligand synthesis to the common ancestor of bacteria and archaea, yet with phylogenies punctuated by late lateral gene transfer events, indicated with asterisks in Fig. 5. In some cases, these transfers involved the import of multiple genes from acetogens into the same methanogen species. For instance, the genes of CbiJ, CbiT and CbiC were transferred from acetogens into Methanomassiliicoccus intestinalis (GCF_000404225.1), Methanoplasma termitum (GCF_000800805.1), Methanomethylophilus alvus (GCF_000300255.2) and methanogenic archaeon ISO4‐H5 (GCF_001560915.1). A common ancestor of these four archaea acquired the bacterial corrin synthesis genes cbiJ, cbiT and cbiC, apparently in a single event, because the four archaea in question share a common ancestor in a species tree constructed from a concatenated alignment of 23 prokaryotic informational core genes (ICGs) (Fig. S3), constructed mainly from ribosomal proteins and aminoacyl‐tRNA synthetases (Table S2). In addition, a deep separation between archaeal and bacterial CbiA/CobB proteins was also recently observed in a phylogeny comprising homologous amidases of the different tetrapyrroles present in 2070 assemblies [129], where, besides CbiA/CobB, the CfbB protein, involved in the amidation of the F₄₃₀ cofactor of methyl‐coenzyme M reductase [100, 130], and the DsrN protein [131], responsible for the amidation of the siroheme cofactor present in DsrAB proteins [132], are present. This phylogeny clearly shows that the methanogenic CfbB proteins and the DsrN proteins derived from archaeal CbiA/CobB proteins and that archaeal and bacterial CbiA/CobB proteins form two large clades separated by the root. This indicates not only the presence of this protein in LUCA but also the ancestry of CbiA/CobA in relation to F₄₃₀. A closer inspection of the phylogeny also indicated events of interdomain CbiA/CobA transfers with 11 archaeal sequences found in four small clades inside a clade of 1879 bacterial CbiA/CobB proteins. If indeed the proteins trace back to LUCA, it is expected that, during the almost 4 billion years of evolution, several lateral gene transfers would have occurred, leading to possible synonymous replacements of the original gene by the newly acquired one, as observed in other cases [129, 133, 134]. Moreover, studies have indicated that an average of 42.5% of genes per species is affected by lateral gene transfer when considering recent transfers [135], and this number raises up to 81% when the oldest transfers are included [136]. Thus, cases in which only a few HGTs are identified can be considered as present in LUCA. Putative LGT events are highlighted in Fig. S3.

The remaining nine trees showed abundant interleaving of bacterial and archaeal corrin synthesis genes, indicating that the homologous genes are, in terms of structure and function, freely interchangeable across the bacterial–archaeal divide. This indicates in turn (a) conservation of function since their divergence from a common ancestral gene and (b) the presence and function of a member of that gene family in the bacterial–archaeal common ancestor. Despite the interleaving of methanogens and acetogens in individual trees, the concatenated tree of corrin biosynthesis enzymes recovers domain monophyly (Fig. S4). Given the possible non‐specificity of reactions of primordial enzymes, it can be envisaged that CbiE (tracing to LUCA) could have replaced (in LUCA) some of the homologous proteins that show intercalated archaeal‐bacterial phylogenies, such as CbiF, CbiH and CbiL. Phylogeny and functional equivalence indicate the presence of an almost complete corrin pathway in the bacterial–archaeal common ancestor with a partial lower ligand biosynthesis.

Cobamide synthesis, a hub in biochemical pathway evolution

None of the intermediates in the 26‐enzyme cobamide synthesis pathway are known to serve a biochemical function beyond their role as a pathway intermediate. This would present an evolutionary paradox, because all individual enzymes of the entire pathway would have to be functional before selection could set in – unless, of course, the domains of cobamide synthesis enzymes were recruited from simpler preexisting pathways. To see which pathways might have contributed genes to corrin evolution, we took three approaches.

First, a diamond blastp search of cobalamin biosynthesis proteins against a dataset of 404 autotrophic core reactions [137] uncovered homologues of CbiA in F₄₃₀ biosynthesis, homologues of CbiA and CobQ in biotin biosynthesis, and homologues of CbiG and CobQ in histidine metabolism (Table S3). Second, we mapped the sequences of the 355 protein clusters from Weiss et al. [47] to InterPro and compared the SUPERFAMILY, TIGR and PFAM assignments with those obtained for cobalamin biosynthesis proteins. We found that the ancient ‘P‐loop containing nucleoside triphosphate hydrolase’ domain, which is the only domain of CobU and BtuR/CobA proteins, is present in 37 different protein families from Weiss et al. [47] belonging to seven different functional categories (Fig. 6). CbiT belongs to the large superfamily of S‐adenosyl‐l‐methionine‐dependent (SAM‐dependent) methyltransferases (SSF53335) and is homologous with five protein families of Weiss et al. [47], including three involved in RNA modifications. CbiE shares homology with the remaining cobalamin methylases (SirA, CbiF, CbiH and CbiL), indicating ancient diversification of the family. CobQ is a member of the Class I glutamine amidotransferase‐like superfamily, which includes a domain of CbiA; the second, nucleotide‐binding domain of CbiA is also present in the data of Weiss et al. [47] in a cluster annotated as MinD. In total, 12 out of 18 cobalamin biosynthesis proteins belong to protein families that trace to LUCA [47], indicating that their ancestral module was already present in LUCA (Table S4).

Fig. 6.

Homologs of cobalamin biosynthesis enzymes in other pathways. The chemical structure of adenosylcobalamin is coloured according to the enzymes that are involved in the synthesis of the respective region of the cobalamin structure. Dashed arrows point to KEGG pathway categories linked to identified homologs of the corresponding enzymes.

Next, we looked for domains shared between cobalamin biosynthesis and domains present in 3939 sequenced genomes of cultured organisms using InterPro domain scans (Fig. 6). This revealed homologues of genes for anaerobic cobalamin biosynthesis in additional ancient pathways. Homologues of cobalamin biosynthesis are involved in amino acid synthesis (tyrosine and histidine) and in cofactor biosynthesis pathways for molybdopterin (MoCo), pyridoxal phosphate and biotin. The class III tetrapyrrole methylases CbiE, CbiF, CbiH, CbiL and SirA are homologous to diphthine synthase, 16S rRNA (cytidine(1402)‐2′‐O)‐methyltransferase and MazG proteins. The class I methyltransferase CbiT shares homology with a large number of SAM‐dependent RNA‐modifying methyltransferases, which is in accordance with the literature [138]. SirC contains a very common NAD(P)‐binding Rossmann‐fold domain (Table S5). CobQ is related to BioD (biotin pathway), an enzyme responsible for ring formation [139] and homologous to MinD, CobB and ParA proteins [140]. No homologs were identified for CbiG, CbiJ, CbiD and CbiB. In the pathway for synthesis of the lower ligand, BzaA, BzaB and BzaF are homologous to ThiC, which catalyzes the first committed step of thiamine biosynthesis [141]. Other homologues included enzymes involved in the synthesis of F₄₃₀, a Ni‐containing tetrapyrrole specific to the methane‐forming step of methanogenesis [142] whose synthesis shares common steps with cobamide biosynthesis [38].

The shared homologies of cob enzymes reflect links to cofactor synthesis, amino acid synthesis and RNA modification (Fig. 6), functions that were clearly present in LUCA [47, 137], indicating that the 26‐step cobamide synthesis pathway was assembled from protein domains involved in ancient preexisting pathways. Cobalamin synthesis genes were, however, also recruited later in evolution, for example, in chlorophyll biosynthesis (BchE, BchR and BchQ) of photosynthetic bacteria [38]. In addition, all tetrapyrrole class II chelatases derive from a common ancestor [143, 144], that evolved through duplication to give rise to CbiX_L, CbiK, HemH (PpfC/CpfC) and SirB (ShfC). Both Methanococcus maripaludis and M. thermoacetica use the short version of the enzyme (CbiX_s), having retained the primordial class II tetrapyrrole chelatase and further underscoring the common ancestry of the pathways in these organisms. Cobamide biosynthesis represents a hub of functional domains that was assembled early in the evolution of biochemical pathways and tapped for useful activities during lineage divergence.

The acetogen–methanogen common ancestor and LUCA

The data indicate that corrin biosynthesis was present in the acetogen‐methanogen common ancestor. This traces it to LUCA, because there are only three possibilities concerning the relationship between the common ancestor of bacteria and archaea and LUCA. (a) Bacteria arose from a specific lineage of archaea, making LUCA an archaeon. (b) Archaea arose from a specific lineage of bacteria, making LUCA a bacterium. (c) Neither domain arose from the other, such that LUCA was the common ancestor of bacteria and archaea sampled here. No current molecular phylogenetic studies support an origin of one prokaryotic domain from the other or one member of one prokaryotic domain from the other. Though debates exist about LUCA's organizational state [145, 146], very few studies directly address the physiology of LUCA [47] and numerous studies support possibility 3 [47, 147, 148], which, given the homologies (Fig. 3) and phylogenies (Fig. 5) presented here, would place corrin biosynthesis in the common ancestor of bacteria and archaea, hence in LUCA.

The presence of corrin synthesis in LUCA contrasts with the lack of homologous methyl synthesis branches across the bacterial and archaeal versions of the acetyl‐CoA pathway that delivers methyl groups for CoFeS (Fig. 3), suggesting that LUCA utilized geochemically supplied methyl groups [47, 50]. A recent study reported that LUCA possessed a nearly complete acetyl‐CoA pathway [149], although the genes they identified (in their supplemental information) were specifically for the archaeal acetyl‐CoA pathway and were often present in fewer than 10% of the bacterial plus bacterial‐metagenomic lineages sampled, indicating transfer of the archaeal genes to bacteria, rather than presence of the archaeal pathway in LUCA, which their method inferred. Many such archaea‐to‐bacteria transfers are known for the archaeal methyl synthesis pathway, starting with the O₂‐dependent bacterial methylotroph Methylobacterium extorquens, where genes for the entire methyl synthesis pathway and methanopterin cofactor synthesis were acquired by transfer from archaea [150]. Several other studies uncovered further examples in which individual genes of the archaeal methyl synthesis and methanopterin pathways had been transferred from archaea to bacterial recipients [50, 151, 152, 153, 154], which are often O₂‐dependent methylotrophs and methanotrophs. Lateral gene transfer is a general hurdle to identifying genes that were present in LUCA [47]. The reactions of methyl synthesis in Fig. 3 are reversible. The presence of homologous enzymes generating formate (FdhA, FwdBD) could reflect formate utilization (generation of CO₂ and reductant) as reductant in early serpentinizing environments [70]; the presence of homologous enzymes for synthesis of methylene‐pterins (MetF, Mer) (Fig. 3) could reflect methyl‐dependent generation of methylene‐H₄F and methylene‐H₄MPT as C1 units for biosyntheses.

Implications for early biochemical evolution

Earlier gene‐based investigations of corrin biosynthesis evolution [80, 155, 156, 157] have missed the role of corrinoids in CoFeS and the acetyl‐CoA pathway, even though their function in ancient pathways has long been evident [7, 8, 48, 49]. The present findings trace corrinoids to the earliest phases of biochemical evolution, to a time when some essential biochemical functions were performed by proteins, some were performed by cofactors alone and some were still performed by solid‐state metals [62, 64, 77, 90]. Isotope data trace the ultralight isotope signature of the acetyl‐CoA pathway [68] back to > 3.5 Ga [65, 66], consistent with the structural (Fig. 3) and phylogenetic (Fig. 5) evidence for the age of corrins.

The present data suggest a sequence of events as outlined in Fig. 7. Under the conditions of serpentinizing hydrothermal vents, transition metals alone convert H₂ and CO₂ into formate, acetate and pyruvate in water [60, 62, 63, 158]. Methyl synthesis and transfer can occur on solid‐state catalyst surfaces (Fig. 7A). The exergonic nature of carbon and energy metabolism from H₂‐dependent CO₂ fixation can, in principle, energetically support the origin of genes and proteins with the help of substrate‐level phosphorylation [71, 159]. Proteins, once present, can organize Fe and Ni into active sites through ligation to cysteine sulfhydryls and perform the functions of ferredoxin [79], CODH and ACS [60, 61], such that acetyl and C≡O synthesis become independent of the environmental surface but remain tied to solid‐state catalysts for methyl group supply from H₂ and CO₂ (Fig. 7B). In this stable intermediate configuration, enzymatic carbon metabolism remains tethered to catalysts on the Earth's crust. More ancient than enzymatic methyl synthesis, corrin synthesis could supply soluble and diffusible Co‐bound methyl groups, freeing ACS and acetyl synthesis from contact with the solid state and allowing acetyl synthesis to migrate to the aqueous phase (Fig. 7C). Pterin‐dependent enzymatic methyl synthesis from H₂ and CO₂ freed methyl synthesis from solid‐state catalysts, involving folate in the lineage leading to acetogens and methanopterin in the lineage leading to methanogens [50] (Fig. 7D). This required the origin of additional complexes essential for methanogens and acetogens, namely, hydrogenases and flavin‐based electron bifurcation [78], the precursor of which was Fe⁰‐dependent ferredoxin reduction with H₂ under hydrothermal conditions [79].

Fig. 7.

Early phases of biochemical evolution in serpentinizing hydrothermal vents. (A) Transition metals alone convert H₂ and CO₂ into acetate and pyruvate; methyl synthesis and transfer occur on the solid‐state catalyst surface. (B) Still tied to solid‐state catalysts for methyl supply, proteins incorporate Fe‐ and Ni‐atoms into their active sites, allowing the synthesis of carbon monoxide and acetyl to become soluble and independent of the environmental surface. (C) The origin of corrin synthesis generates a supply of mobile, cobalt‐bound methyl groups, releasing ACS from the solid state and enabling acetyl synthesis to migrate to the aqueous phase. (D) H₂‐dependent methyl formation from CO₂ also migrated to the aqueous phase with the origin of enzymatic methyl synthesis after the archaeal–bacterial divergence.

A critical observer might maintain that even enzymatic corrin synthesis is ‘just too complicated’ to be ancient [80], but the argument fails. Why? Corrins are far less complicated than ribosomes, with 1/2000th the molecular mass [160], and ribosome synthesis had to be in place before any of the homologous protein‐coding genes for corrin biosynthesis arose. Complexity is not an argument against antiquity per se, as the ribosome attests. At biochemical origins, complex enzymatic pathways require functional catalytic precursors. The present findings establish a catalytic continuity between natural geochemical environments [75] and the requirement for Fe, Ni and Co in enzymes and corrin cofactors of the acetyl‐CoA pathway, which might foster the identification of additional intermediate states in early biochemical evolution.

Materials and methods

Data

A dataset of 80 genomes of methanogens and 189 genomes of acetogens from the Reference Sequence database (September 2016) [161] was analyzed. The methanogens belong to five archaeal orders (including “Other Archaea”), and three classes of acetogens are represented. A taxonomy report can be found in Table S6. Sequences of enzymes involved in the anaerobic biosynthesis of cobalamin (Fig. 2; based on [38]), as well as acetyl‐CoA synthase (ACS), carbon monoxide dehydrogenase (CODH) and CoFeS of M. thermoacetica (acetogen representative) and Methanococcus maripaludis (methanogen representative), were acquired from the KEGG (Kyoto Encyclopedia of Genes and Genomes) [162] database in June 2023 (KEGG organisms mtho/mta and mmp, respectively). M. thermoacetica sirohydrochlorin cobaltochelatase CbiX was downloaded from GenBank (TYL17187.1). Acetogen sequences of enzymes involved in lower ligand synthesis were also acquired from the Reference Sequence database and GenBank. Methanogen query sequences were selected from blast results against the acetogen set of sequences. Sequences of enzymes involved in the methyl synthesis in M. thermoacetica [94] were acquired from the IMG/M database [163] and UniProt [164]. For Methanothermobacter marburgensis, sequences (as described in Ref. [49]) were downloaded from GenBank. A dataset of 3939 genomes downloaded from NCBI (November 2019) was analyzed for the presence of homologues of the biosynthesis of cobalamin but excluding the aerobic cobalamin biosynthesis pathway. All analyzed sequences, including the CobA alternatives PduO/EutT and the corresponding organisms, are listed in Table S7.

Sequence homology search in methanogens and acetogens

The representative archaeal and bacterial sequences were scored in methanogens and acetogens with diamond version 2.0.1.139 [165, 166]. The homology search was performed using an expectation value threshold of 1 × 10⁻¹⁰ and an identity threshold of 25%. For each enzyme, the best sequence hit was selected per genome for further analyses. interproscan (version 5.66‐9.8) [167] was run on the retrieved hits to obtain further functional annotations. Global sequence alignments of methanogen and acetogen CoFeS subunits were calculated based on the Needleman‐Wunsch algorithm [168].

Synteny analysis

Genomes containing best‐hit sequences were used for the synteny analysis, which was used in phylogenetic analysis. The best hits were mapped to their gene location files (feature tables), and a syntenic neighbourhood was defined using a window of two upstream and two downstream proteins in the neighbourhood of the proteins of interest (same chromosome). The identified syntenic neighbourhood for each best hit was extracted and mapped onto the phylogenies with the corresponding accessions, protein types, PFAM annotations and gene descriptions for a detailed phylogenetic analysis and better distinction of paralogous sequences.

Expanded sequence homology searches

diamond (version 2.18) was also run against the dataset from Wimmer et al. [137] with an identity cutoff of 25% and E‐value cutoffs of 10⁻¹⁰, 10⁻⁸ and 10⁻⁵, and the best hits were further analyzed. In addition, interproscan (version 5.66‐9.8) [167] was run on the dataset of 3939 genomes and on the sequences from the 355 protein families mapped to LUCA from Weiss et al. [47]. All hits that shared the same SUPERFAMILY annotations as cobalamin biosynthesis enzymes were further analysed.

Computation of multiple sequence alignments and phylogenetic trees

Gene trees containing sequences of methanogens and acetogens for 26 analysed enzymes involved in the anaerobic biosynthesis of cobalamin and lower ligand synthesis were computed based on the best hit selection of the homology search or from the synteny analysis. For that, multiple sequence alignments were made using mafft l‐ins‐i version 7.471 [169] and unrooted gene trees were reconstructed with raxml version 8.2.12 [170] using the PROTCATWAG model and a random starting seed (12345). The same pipeline was used to compute a concatenated tree of methanogens and acetogens with best hits for ≥ 17 enzymes. Single sequence alignments of the analysed enzymes were combined to obtain a large alignment for tree reconstruction. mafft and raxml were used to calculate a species tree of all 80 methanogens present in the dataset based on a concatenated alignment combining sequences of prokaryotic informational core gene families (ICGs; [93]), which were determined by using the verticality value [171], the number of genomes and the taxonomic distribution of a protein family. All 80 methanogens are represented by at least one sequence in 23 different prokaryotic informational core gene families (Table S2). The trees were visually modified with the use of the interactive tree of life [172].

Structural alignments of enzymes

In order to compare the structures of methanogenic and acetogenic enzymes involved in anaerobic cobalamin synthesis and methyl synthesis in the acetyl‐CoA pathway, structures were either obtained from the Protein Data Bank (PDB, [173]) or modelled using AlphaFold version 2.2.0 [105]. Structural alignments and TM‐scores were determined using us‐align version 20230609 [103]. Alignments were visually inspected, and figures were prepared using pymol version 2.5.7 (The PyMOL Molecular Graphics System, Version 2.5.7; Schrödinger, LLC (http://www.pymol.org/pymol)]). In those cases when the structures overlapped over one or more domains, but not the full‐length protein (usually reflected in significant differences in the TM scores normalized by the lengths of the two proteins), the overlapping domains were separately aligned to obtain a more accurate TM‐score. This allowed us to determine structural homology for the superimposed domains, in line with the modular nature of protein evolution.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

WFM designed research; LDM, VK, MK, NM, WFM and FLS performed research; LDM, VK, MK, NM, WFM and FLS analyzed data; and LDM, VK, MK, NM, WFM and FLS wrote the paper.

Supporting information

Fig. S1. The anaerobic cobalamin biosynthesis pathway.

Fig. S2. Structural alignments (25) of the two CoFeS subunits of M. maripaludis (mmp0980 and mmp0981, red) and M. thermoacetica (moth_1198 and moth_1201, blue).

Fig. S3. Species tree of 80 analyzed methanogens.

Fig. S4. Concatenated tree of the anaerobic route to cobalamin in methanogens and acetogens.

Table S1. TM‐scores for structural alignments of archaeal and bacterial enzymes of the methyl branch of the acetyl‐CoA pathway and the anaerobic cobalamin synthesis pathway.

Table S2. Prokaryotic informational core genes (ICG) and their functions. The genes were used to calculate a species tree comprising all analysed methanogens (Fig. S3).

Table S3. Diamond blast hits of cobalamin biosynthesis proteins vs the dataset of 404 autocatalytic core reactions.

Table S4. Homologues of cobalamin biosynthesis proteins found in 355 protein families traced to LUCA (Weiss et al. 2016). Sorted by functional category.

Table S5. Homologues of cobalamin synthesis proteins that share SUPERFAMILY annotation found in the dataset of 3939 genomes.

Table S6. Accession numbers and taxonomy of acetogen and methanogen genomes analysed in this work. The genomes are a subset of a Reference Sequence (RefSeq) data set of 2016.

Table S7. Query sequences used for homology searches and sources for structural comparisons.

Data availability statement

Alignments, phylogenies, commands, synteny script and the dataset of 3939 genomes are available at Figshare: https://doi.org/10.6084/m9.figshare.25765518.