A bifunctional salvage pathway for two distinct S-adenosylmethionine byproducts that is widespread in bacteria, including pathogenic Escherichia coli

Summary

S-adenosyl-L-methionine (SAM) is a necessary co-substrate for numerous essential enzymatic reactions including protein and nucleotide methylations, secondary metabolite synthesis, and radical-mediated processes. Radical SAM enzymes produce 5’-deoxyadenosine, and SAM-dependent enzymes for polyamine, neurotransmitter, and quorum sensing compound synthesis produce 5’-methylthioadenosine as byproducts. Both are inhibitory and must be addressed by all cells. This work establishes a bifunctional oxygen-independent salvage pathway for 5’-deoxyadenosine and 5’-methylthioadenosine in both Rhodospirillum rubrum and Extraintestinal Pathogenic Escherichia coli. Homologous genes for this pathway are widespread in bacteria, notably pathogenic strains within several families. A phosphorylase (Rhodospirillum rubrum) or separate nucleoside and kinase (Escherichia coli) followed by an isomerase and aldolase sequentially function to salvage these two wasteful and inhibitory compounds into adenine, dihydroxyacetone phosphate and acetaldehyde or (2-methylthio)acetaldehyde during both aerobic and anaerobic growth. Both SAM byproducts are metabolized with equal affinity during aerobic and anaerobic growth conditions, suggesting that the dual-purpose salvage pathway plays a central role in numerous environments, notably the human body during infection. Our newly discovered bifunctional oxygen-independent pathway, widespread in bacteria, salvages at least two byproducts of SAM-dependent enzymes for carbon and sulfur salvage, contributing to cell growth.

Abbreviated Summary:

The cofactor S-adenosyl-L-methionine (SAM) plays an essential role in nucleotide methylation, polyamine formation, and radical SAM enzymes, during which distinct inhibitory byproducts are formed. This work defines a dual-purpose pathway for two SAM byproducts, 5’-methylthioadenosine and 5’-deoxyadenosine, that is widespread in bacteria, including Extraintestinal Pathogenic Escherichia coli. This pathway results in precursor metabolites of nucleotides, central carbon metabolism, and methionine to support cell growth. The pathway appears functional in multiple environments, notably during human infection.

Introduction

Numerous enzymes utilize S-adenosyl-L-methionine (SAM) as a cofactor. SAM serves as a methyl-donor for DNA, RNA, lipid, and protein methylation. It also serves as a homoserine lactone donor for quorum sensing compounds, an aminopropyl-donor for polyamine synthesis, and a source of 5’-deoxyadenosyl radicals for Radical SAM enzymes (Fontecave, et al., 2004). Methyl-transfer from SAM results in S-adenosylhomocysteine (SAH); polyamine and quorum sensing compound biosynthesis result in 5’-methylthioadenosine (MTA) production; and utilization of 5’-deoxyadenosyl radicals result in 5’-deoxyadenosine (5’dAdo) formation. SAH, MTA, and 5’dAdo are each inhibitory byproducts (Fig. 1) (Parveen and Cornell 2010). Accumulation of these byproducts leads to defects in growth and metabolism (Choi-Rhee and Cronan, 2005; Challand et al., 2009), and elimination via export constitutes a loss of valuable organic carbon, nitrogen, and sulfur.

Fig. 1.

Salvage of SAM byproducts. A) In most organisms, the SAH is recycled to methionine by a variation of the Methionine Cycle (Miller et al., 2015; Miller et al., 2013; Sun et al., 2004; Vendeville et al., 2005; Beeston et al., 2002). B) For 5’dAdo (this study) and MTA (North et al., 2017), a dual-purpose DHAP shunt recycles each compound into the central carbon metabolite, dihydroxyacetone phosphate (DHAP), for carbon salvage, and acetaldehyde or (2-methylthio)acetaldehyde, respectively, for additional carbon or sulfur salvage. C) Further metabolism of (2-methylthio)acetaldehyde varies by organism (North et al., 2017; Miller et. al. 2018; North J.A. unpublished results). Rp, Rhodopseudomonas palustris; Ru, Rhodospirillum rubrum; Ec, Escherichia coli. Protein designations and EC numbers are provided where available.

Detoxification and sulfur salvage from SAH appear to occur in all organisms, except for certain obligate endosymbionts, by the Methionine Cycle (Fig. 1A) (Sun et al., 2004; Vendeville et al., 2005; Beeston et al., 2002). However, the means by which organisms, particularly bacteria and archaea, manage MTA and 5’dAdo is less understood. In commensal E. coli strains (e.g. K12), the pfs gene product, a variable SAH/MTA/5’dAdo nucleosidase, catalyzes cleavage of SAH, MTA and 5’dAdo into S-ribosylhomocysteine, 5-methylthioribose, and 5-deoxyribose, respectively (Fig. 1; nucleosidase, brown)) (Choi-Rhee and Cronan, 2005; Challand et al., 2009). This detoxifies the nucleoside and salvages the adenine; 5-methylthioribose and 5-deoxyribose appear to have no further fate beyond export (Choi-Rhee and Cronan, 2005; Hughes, 2006). The homologous nucleosidase from Mycobacterium tuberculosis is also catalytically efficient with both MTA and 5’dAdo (Namanja-Magliano et. al., 2016). Eukaryotes and some bacteria employ one of several variations of the Universal Methionine Salvage Pathway (MSP) to recycle MTA back into methionine (Fig. 2B) (Sekowska et al., 2004; Guan et al., 2012; Guan at al., 2011). However, the oxygen dependence of the Universal MSP dioxygenase (Fig. 2B; MtnD) (Myers et al., 1993) renders it ineffective for MTA and putative 5’dAdo salvage under anaerobic growth conditions where Radical SAM enzymes predominantly function (Broderick et al., 2014).

Fig. 2.

Other known methionine salvage pathways. MTA metabolism is initiated by phosphorylases (MtnP), nucleosidases (MtnN), kinases (MtnK), and isomerases (MtnA), homologous to DHAP shunt enzymes with 5-methylthioribulose-1-phosphate as a common intermediate. Some organisms such as P. aeruginosa and M. jannaschii employ an MTA deaminase (MtaD) and 5’-methylthioinsoine phosphorylase (MtiP) (Guan et. al., 2011; Guan et. al., 2012; Miller et. al., 2013; Miller et. al., 2018). A) MTA-isoprenoid shunt from R. rubrum in which a methylthio-glutathione intermediate is reduced to release methanethiol (CH₃SH) by the proposed gamma-glutamyl cycle as the immediate precursor to methionine (Erb et al., 2012; Warlick et al., 2012; North et al., 2016; Cho et. al. 2015). B) Universal methionine salvage pathway and variations thereof (Albers, 2009; Sekowska et al., 2004). C) RubisCO-dependent MTA metabolism pathway from R. rubrum (Singh and Tabita, 2010; Dey et al., 2015). Protein designations and EC numbers are provided where available.

Recently our group identified three novel MSPs for MTA, the “MTA-isoprenoid shunt” (functions aerobically or anaerobically), the “dihydroxyacetone phosphate (DHAP)-ethylene shunt” (functions anaerobically), and the “DHAP-methanethiol shunt” (functions aerobically), in the photosynthetic alphaproteobacteria, Rhodospirillum rubrum and Rhodopseudomonas palustris (Figs. 1C and 2A) (North et al., 2017; Miller A et al., 2018; Erb et al., 2012). The two DHAP shunt-based MSPs salvage MTA via a phosphorylase, an isomerase, and a novel class II aldolase to form DHAP and (2-methylthio)acetaldehyde as key intermediates (Fig. 1B). In this study we demonstrate that MTA and 5’dAdo, two of the three major byproducts of SAM-dependent enzymatic activity, are both salvaged as adenine, DHAP, and an aldehyde intermediate in R. rubrum by the sequential activity of the phosphorylase, isomerase, and novel aldolase, forming a DHAP shunt (Fig. 1B). Furthermore, this work reveals that nearly half of all Extraintestinal Pathogenic E. coli (ExPEC) isolates, unlike commensal E. coli strains, employ a variation of the DHAP shunt pathway to metabolize the ubiquitous E. coli SAH/MTA/5’dAdo nucleosidase products, 5-methylthioribose and 5-deoxyribose, into DHAP and corresponding aldehyde. This enables growth of ExPEC strains and identifies 5’dAdo and MTA, common urine and blood metabolites, as a previously unidentified carbon and energy sources for E. coli, suggesting a possible role for the DHAP shunt during infection and in human health.

Results

MTA and 5’dAdo salvage requires the DHAP shunt –

Previous studies demonstrated that R. rubrum and R. palustris encode genes (mtnP, mtnA, and ald2) for enzymes that catalyze the initial steps of two distinct anaerobic MSPs for MTA metabolism (Fig. 1B-C), with DHAP and (2-methylthio)acetaldehyde as key products (North et al., 2017; Miller A et al., 2018). Given previous in vitro indications that these same enzymes may also function in 5’dAdo metabolism (Erb et al., 2012), we simultaneously quantified the intracellular levels of the SAM metabolism byproducts, MTA and 5’dAdo, in R. rubrum. The wild type strain maintained ~0.2 and ~0.1 micromole per gram cell dry weight of 5’dAdo and MTA, respectively, under both aerobic and anaerobic growth conditions as observed by HPLC (Figs. 3A and S1) and verified by mass spectroscopy (Fig. S2). Upon individual deletion of the mtnP, mtnA, or ald2 genes, both MTA and 5’dAdo accumulated extracellularly to levels above that of the wild type strain (Figs. 3A and S1). In all strains, 5’dAdo was consistently 2-fold higher in abundance than MTA, indicating that R. rubrum must routinely address the consequences of more 5’dAdo from Radical SAM reactions than MTA derived from polyamine and quorum sensing compound synthesis. In addition, as indicated by the mtnP deletion strain, cells produced 75-fold more 5’dAdo and MTA during anaerobic growth versus aerobic growth (Figs. 3A and S1). Plasmid-based gene complementation of each respective gene deletion strain with the relevant gene restored 5’dAdo and MTA abundance to wild-type levels, indicating functional restoration of a common salvage pathway to metabolize both SAM byproducts (Figs. 3A and S1). This underpins the utility of the phosphorylase, isomerase, and aldolase as an oxygen-independent dual-purpose metabolic pathway in contrast to the oxygen-dependent Universal MSP and the MTA-isoprenoid shunt MSP (Fig. 2A-B).

Fig. 3.

MTA and 5’dAdo metabolism in R. rubrum requires DHAP shunt gene products MtnP, MtnA, and Ald2. A) Quantification of MTA and 5’dAdo abundance in R. rubrum grown both aerobically and anaerobically. Error bars are standard deviation for n = 3 independent growth experiments. B) Identification of the specific production of [5-H³]-5-deoxyribose-1-phosphate (5dR-1P) and [5-H³]-5-deoxyribulose-1-phosphate (5dRu-1P) by the DHAP shunt phosphorylase (MtnP) and isomerase (MtnA), respectively, upon feeding with [5’-H³]-5’-deoxyadenosine and resolving products via HILIC chromatography. See Fig. S3 for full feeding time series. C) Identification by HILIC chromatography of [2-H³]-acetaldehyde produced from 5dRu-1P by the DHAP shunt aldolase (Ald2). [2-H³]-acetate and [2-H³]-ethanol are subsequently formed as further confirmed by ion exclusion chromatography (Fig. S4A). D) Gas chromatography verification of acetaldehyde produced by the sequential activity of purified R. rubrum phosphorylase (RrMtnP), isomerase (RrMtnA), and aldolase (RrAld2) with 5’dAdo. The identity of acetaldehyde was further validated by specific conversion to ethanol by Saccharomyces cerevisiae alcohol dehydrogenase (ScADH) coupled to nicotinamide adenine dinucleotide (NADH) oxidation. E) Specific activity of R. rubrum and E. coli 5-methylthioribulose-1-phosphate/5-deoxyribulose-1-phosphate aldolases (RrAld2 and EcAld2) as a function of metal cofactor. E: purified recombinant enzyme produced by E. coli grown in lysogeny broth without any supplemented trace metals; E*: recombinant enzyme produced in the presence of supplemented trace metals in the growth media; E_D: dialyzed apoenzyme reconstituted with 0.1 mM of the indicated metal. MTRu-1P: 5-methylthioribulose-1-phosphate; 5dRu-1P: 5-deoxyribulose-1-phosphate. Error bars are standard deviation for n = 3 independent enzyme preparations. F) R. rubrum DHAP Shunt pathway for 5’dAdo with location of the tritium label indicated (*), utilized for metabolite detection via HPLC.

The DHAP shunt links to carbon metabolism –

The metabolism of 5’dAdo was followed by feeding R. rubrum anaerobically with [5’-H³]-5’-deoxyadensoine and using HILIC chromatography to resolve intracellular products. 5’dAdo was metabolized to 5-deoxyribose-1-phosphate (Fig. 3B; 5dR-1P), 5-deoxyribulose-1-phosphate (Fig. 3B; 5dRu-1P), and a mixture of acetaldehyde, ethanol, and acetate (Fig. 3C) after comparing retention times to synthesized standards. Initiation of 5’dAdo metabolism and specific production of 5-deoxyribose-1-phosphate was abolished upon deletion of the phosphorylase (ΔmtnP; Fig. 2B-C), establishing the phosphorylase (MtnP) as the first committed step of a dual-purpose MTA and 5’dAdo salvage pathway. Deletion of the isomerase (MtnA) arrested 5’dAdo metabolism beyond 5-deoxyribose-1-phosphate (ΔmtnA; Fig. 2B-C), and deletion of the aldolase (Ald2) prevented metabolism beyond 5-deoxyribulose-1-phosphate (Δald2; Fig. 2B-C). The synthesis of these products from 5’dAdo by MtnP, MtnA, and Ald2 is analogous to each enzyme’s respective function in MTA metabolism previously reported under both aerobic and anaerobic conditions (Fig. 1B; DHAP Shunt) (North et al., 2017; Miller A et al., 2018). Thus, these three enzymes constitute a dual-purpose “DHAP shunt”, where MTA is metabolized to adenine, DHAP, and (2-methylthio)acetaldehyde for purine, carbon, and sulfur salvage, and 5’dAdo is metabolized to adenine, DHAP, and acetaldehyde for purine and carbon salvage.

DHAP shunt enzyme characterization–

In order to verify the genetic and whole-cell feeding evidence for the proposed pathway, each R. rubrum gene (mtnP, mtnA, ald2) was recombinantly expressed and proteins purified from E. coli. The products derived from DHAP shunt enzymatic activities with MTA and reconstitution of the full pathway converting MTA to DHAP and (2-methylthio)acetaldehyde were previously characterized (North et al., 2017; Erb et al., 2012). The enzyme reaction products derived from 5’dAdo metabolism were verified by GC (Ald2; Fig. 3D), HPLC-mass spectrometry (MtnP; Fig. S5) and specific enzyme assays (Fig. S8). The full pathway for conversion of 5’dAdo to DHAP and acetaldehyde was reconstituted in vivo with the phosphorylase, isomerase, and aldolase (Fig. 3D).

For the phosphorylase, MTA and 5’dAdo both served as catalytically efficient substrates (Table 1; RrMtnP), while the remaining major SAM byproduct, SAH, did not serve as a substrate. Therefore, RrMtnP is in fact a promiscuous MTA and 5’dAdo phosphorylase that produces 5-methylthioribose-1-phosphate and 5-deoxyribose-1-phosphate, respectively. This is analogous to the promiscuity of the human orthologue: HsMtnP K_m-MTA = 4 μM, K_m-5’dAdo = 23 μM, V_max-MTA/V_max-5’dAdo = 0.56 (Savarese et. al., 1981), Similarly, 5-deoxyribose-1-phosphate and 5-methylthioribose-1-phosphate both served as effective substrates for the isomerase (Table 1; RrMtnA), and 5-deoxyribulose-1-phosphate and 5-methylthioribulose-1-phosphate both were effective aldolase substrates (Fig. 3E and Table 1; RrAld2). Recombinant R. rubrum aldolase holoenzyme produced by E. coli had a 5-fold higher specific activity for 5-deoxyribulose-1-phosphate versus 5-methylthioribulose-1-phosphate (Fig. 3E; E and E*). Further analysis of metal cofactor dependence on substrate selectivity and catalytic activity using aldolase apoenzyme reconstituted with various metal cofactors revealed that Co(II) and Ni(II) conferred the highest activity for both substrates (Fig. 3E; E_D). Mn(II) and Co(II) conferred the least difference in selectivity between the two substrates (3- and 5-fold difference). Ni(II), Zn(II), Ca(II), Fe(III) had intermediate selectivity (10-fold difference). And Fe(II) conferred the highest selectivity (30-fold difference) (Fig. 3E). The ability of the R. rubrum phosphorylase, isomerase, and aldolase to selectively and efficiently catalyze the metabolism of both 5’dAdo and MTA to adenine, DHAP, and the corresponding aldehyde species further confirms our in vivo observations that together these three enzymes form a dual-purpose DHAP shunt for carbon and sulfur salvage (Fig. 1B).

Table 1:

Kinetic parameters of DHAP shunt Enzymes from Rhodospirillum rubrum (Rr) and Escherichia coli (Ec).

Pathway	Enzyme	Substrate (Co-Substrate)	Kcat(s−1)	KM(μM)	kcat/KM(M−1s−1)
R. rubrum DHAPShunt	Phosphorylase RrMtnP	S-adenosyl-L-homocysteine	N.A.*	N.A.*	N.A.*
		5’-deoxyadenosine#	6.7	40	1.7 × 105
		5’-methylthioadenosine#	4.5	14	3.2 × 105
		adenine	0.06	41	0.014 × 105
	Isomerase RrMtnA	5-deoxyribose-l-phosphate#	0.18	171	1.1 × 103
		5-methylthioribose-l-phosphate#	0.1	134	0.75 × 103
		ribose-l-phosphate	0.03	>2000	< 0.02 × 103
	Aldolase RrAld2	5-deoxyribulose-1-phosphate	1.94	229	85 × 103
		5-meththioribulose-l-phosphate	0.94	288	1.7 ×103
		ribulose-1-phosphate	0.44	>7000	< 0.06 × 103
E. coli DHAPShunt	Nucleosidase EcMtnN	S-adenosyi-L-homocysteine	84	5.4	0.15× 108
		5’-deoxyadenosine	57	4.7	0.12 × 108
		5’ -methylthioadenosine	150	0.83	1.8 × 108
		adenine	0.24	93	0.00003 × 108
	Kinase EcMtnK	S-ribosyl-L-homocysteine (ATP)	5.4	2265	0.024 × l05
		5-deoxyribose (ATP)	16.1	25	63 × 103
		5-methylthioribose (ATP)	10.6	186	0.57 × 105
		ribose (ATP)	11.6	18181	0.0064 × 105
		ATP (s-deoxyribose)	14.1	83	1.7 × 105
	Isomerase EcMtnA	5-deoxyribose-l-phosphate	2.02	222	9.1 × 103
		5-methylthioribose-1-phosphate	0.49	50	9.8 × 103
		ribose-l-phosphate	0.67	>4900	< 0.14 × 103
	Aldolase EcAld2	5-deoxyribulose-1-phosphate	22	598	3.7 × 104
		5-meththioribulose-1-phosphate	22	275	9.8 × 104
		ribu1ose-1-phosphate	8.7	827	1.1×104
Fucose Metabolism	Aldolase Ec FucA	5 -deoxyribulose-1-phosphate	3.87	980	3.9 × 103
		5-methylthioribulose-1-phosphate	3.9	1160	34 × 103
		ribulose-1-phosphate	0.17	> 1800	< 0.09 × 103
B. subtilis Universal Methionine Selvage Pathway	Kinase Bs MtnK	S-ribosy4-L-homocysteine (ATP)	3.0	946	0.0031 × 106
		5-deoxyribose (ATP)	3.4	74	0.45 × 106
		5-methylthioribose (ATP)	3.8	24	1.6 × 106
		ribose (ATP)	4.1	6474	0.00064 × 106
		ATP (5-methylthioribose)	1.9	90	0.021 × 106
	Isomerase BsMtnA	5-deoxyribose-l-phosphate	0.45	> 847	<0.053 × 104
		5-meththioribose-1-phosphate	3.9	114	3.5 × 104
		ribose-l-phosphate	0.021	>8000	< 0.00026 × 104
	Dehydratase Bs MtnB	5 -deoxyribulose-1-phosphate	0.053	315	0.0020 × 105
		5-meththioribulose-1-phosphate	16	24	6.7 × 105
		ribulose-1-phosphate	0.0067	1171	0.000057 × 105

^*N.A. No activity detected up to 20 µM enzyme, 10 mM substrate.

^#Kinetic parameter values originally reported elsewhere (Erb et al., 2012).

Kinetic parameters were determined by nonlinear weighted regression fit to a Michaelis-Menten model. Weights were the standard deviation for n = 3 independent experiments (Figs. S8-S10). Standard error for all kinetic parameters were less than 15 %. Literature values for B. subtilis known enzyme parameters are given it Supplementary Table S1.

A DHAP shunt in Pathogenic E. coli –

A surprising observation was that in several genera, pathogenic bacterial strains possessed a putative DHAP shunt gene cluster (mtnK, mtnA, and ald2), whereas non-pathogenic strains of these genera did not. For example, amongst firmicutes, DHAP shunt genes were found in Clostridium tetani, C. botulinum, Bacillus thuringiensis, B. cereus, and B. anthracis (Fig. S6A and Table S2). Strikingly, putative DHAP shunt genes (mtnK, mtnA, and ald2) were observed in ExPEC strains such as those of the predominant ST131 lineage (Fig. 4A) (Petty et al., 2014). Blast and phylogenetic analysis of available E. coli genomes in the NCBI database (Fig. S6B) revealed that while only <0.1% of analyzed pathogenic intestinal E. coli isolates (1 of 1376) contained the putative gene cluster, 42% of analyzed ExPEC isolates possessed it (471 of 1133). In all cases, the putative DHAP shunt gene cluster was located at the distal end of the ExPEC tRNA-leuX genomic island (Fig. 4A and Fig. S6C) (Totsika et al., 2011), suggesting a common ancestry.

Fig. 4.

A DHAP shunt in Extraintestinal Pathogenic E. coli (ExPEC) strains. A) Sequence alignment of highly virulent ExPEC ST131 and ExPEC ATCC 25922 compared to commensal strain K12 MG1655. Light gray, 100% identity; red, SNPs. Regions of insertion into previously identified tRNA_Leu Genomic Island indicated in dark gray; deletions indicated by black arrows. Alignment and visualization performed using NCBI blastn. Inset: gene cluster containing DHAP shunt genes. See Fig. S6 and Table S4 for other ExPEC isolates with DHAP shunt genes. B) Identification by reverse-phase chromatography of characteristic DHAP shunt metabolite (2-methylthio)ethanol (MT-EtOH) in the ExPEC strain (ATCC 25922) upon feeding with [methyl-C¹⁴]-5’-methylthioadenosine compared to commensal strain K12 and the ExPEC kinase deletion strain (ATCC 25922 ΔmtnK) where the DHAP shunt is inactivated. C) Identification by HILIC chromatography of specific production of 5-deoxyribose-1-phosphate (5dR-1P) and 5-deoxyribulose-1-phosphate (5dRu-1P) by the DHAP shunt kinase (MtnK) and isomerase (MtnA), respectively, upon feeding the ExPEC strain (ATCC 25922) with [5’-H³]-5’-deoxyadenosine. See Fig. S7 for full feeding timeseries. D) Identification by HILIC chromatography of specific production of acetaldehyde (observed as ethanol) from 5dRu-1P by DHAP shunt aldolase (Ald2) upon feeding the ExPEC strain (ATCC 25922) with [5’-H³]-5’-deoxyadenosine. Identity of ethanol was further confirmed by ion exclusion chromatography (Fig. S4B). E) Identification by HILIC chromatography of metabolites produced upon feeding the EXPEC nucleosidase deletion strain (ATCC 25922 Δpfs) with [5’-H³]-5’-deoxyadenosine. F) Comparative growth studies of the ExPEC strain (ATCC 25922) versus the E. coli commensal strain (K12 BW25113). - 5dR: cells grown without addition of 5-deoxyribose. G) E. coli DHAP Shunt pathway for 5’dAdo with location of the tritium label indicated (*), utilized for metabolite detection via HPLC.

To determine if the putative DHAP shunt gene cluster was expressed and constituted a functional salvage pathway for both MTA and 5’dAdo, E. coli commensal strain (K12 JM109) and attenuated ExPEC strain (ATCC 25922), which is non-pathogenic in humans but of low to moderate virulence in mice (Ashby et al., 1994; Hof et al., 1986), was fed aerobically with labeled MTA as previously done for R. rubrum (Fig. 4B) (North et al., 2017). In the commensal strain, MTA was rapidly and completely metabolized to 5-methylthioribose via the MTA/SAH/5’dAdo nucleosidase, as previously reported for other commensal strains (Hughes 2006). In stark contrast, the ExPEC strain metabolized MTA to 5-methylthioribose and subsequently to (2-methylthio)ethanol as the terminal product. Inactivation of the putative kinase gene (ATCC 25922 ΔmtnK), the first in the proposed salvage pathway, arrested MTA metabolism at 5-methylthioribose, identical to commensal strains (Fig. 4B).

5’dAdo was processively metabolized by the ExPEC strain to 5-deoxyribose, 5-deoxyribose-1-phosphate, 5-deoxyribulose-1-phosphate, and DHAP plus acetaldehyde (observed as ethanol) (Fig. 4C-D). Inactivation of the putative DHAP shunt kinase (ATCC 25922 ΔmtnK) eliminated the specific production of 5-deoxyribose-1-phosphate from 5-deoxyribose. Inactivation of the isomerase (ATCC 25922 ΔmtnA) eliminated the specific production of 5-deoxyribulose-1-phosphate. And, inactivation of the aldolase (ATCC 25922 Δald2) lowered acetaldehyde (and DHAP) formation 10-fold (Fig. 3C-D). The small amount of pathway activity in the aldolase deletion strain can be attributed to E. coli L-fuculose-1-phosphate aldolase, which has modest activity with 5-deoxyribulose-1-phosphate as a substrate (Table 1; EcFucA) (Erb et al., 2012; Beaudoin et al., 2018). Inactivation of the MTA/SAH/5’dAdo nucleosidase gene (pfs) common to all E. coli resulted in > 10-fold slower metabolism of 5’dAdo to 5-deoxyribose (Fig. 4E). The residual nucleosidase activity is likely due to a non-specific nucleosidase as previously observed for commensal E. coli pfs deletion strains (Choi-Rhee and Cronan, 2005).

Clearly the E. coli MTA/SAH/5’dAdo nucleosidase products, 5-methylthioribose and 5-deoxyribose, are further metabolized by ExPEC strains via the dual-purpose DHAP shunt analogous to R. rubrum for potential carbon and sulfur salvage. This enables the ExPEC strain (ATCC 25922) to grow with 5-deoxyribose as sole carbon source (Fig. 4F). In contrast, the E. coli commensal strain (K12 BW25113) cannot utilize 5-deoxyribose as sole carbon source. Inactivation of each individual DHAP shunt specific gene (mtnK, mtnA, or ald2) precluded growth of the ExPEC strain with 5-deoxyribose (Fig. 4F), establishing 5-deoxyribose as a heretofore unidentified growth substrate and the DHAP shunt as a functional carbon salvage pathway in ExPEC strains. Evidently, while purified FucA is active in vitro with 5-deoxy- and 5-metylthioribulose-1-phosphate (Table 1; EcFucA), it’s in vivo expression and/or activity is insufficient to compensate for the DHAP shunt aldolase, underpinning the two enzymes’ distinct role in their respective pathways. For sulfur salvage from MTA, metabolism beyond (2-methylthio)ethanol has not been observed (Fig. 4B), and the ExPEC strain could not utilize MTA, 5-methylthioribose, or (2-methylthio)ethanol as a sole sulfur source for methionine biosynthesis in growth experiments. At present, it is unknown if (2-methylthio)ethanol in other ExPEC strains is simply an exported terminal product, serves as a precursor to methionine, or is further metabolized into another product for other purposes.

DHAP shunt versus Universal MSP homologues –

While the in vivo function of the ubiquitous E. coli MTA/SAH/5’dAdo nucleosidase (MtnN) in 5’dAdo cleavage has been previously established (Choi-Rhee and Cronan, 2005), its substrate preference has only been partially characterized (Challand et al., 2009). The ExPEC (ATCC 25922) nucleosidase has 100% amino acid identity to the commensal E. coli (K12) nucleosidase, allowing for direct comparison. Purification and enzyme analysis of the ExPEC MtnN (Table 1; EcMtnN) resulted in catalytic properties for MTA and SAH in close agreement with those previously reported (K_m-MTA = 0.4 μM, k_cat-MTA = 151 s⁻¹; K_m-SAH = 4.3 μM, k_cat-SAH ~ 61 s⁻¹) (Della Ragione et al., 1985). The catalytic properties with 5’dAdo as the substrate were nearly the same as for SAH (Table 1; EcMtnN), compared to a previous report of a much higher K_m (33 μM) (Challand et al., 2009) using a different method. This established that 5’dAdo is an efficient substrate of the MTA/SAH/5’dAdo nucleosidase likely in all E. coli and other bacteria, confirming previous reports (Choi-Rhee and Cronan, 2005; Challand et al., 2009).

Both 5-deoxyribose and 5-methylthioribose were found to catalytically efficient substrates for EcMtnK (Table 1), consistent with this enzyme acting as the start of the dual-purpose DHAP shunt in the ExPEC (ATCC25922) strain. S-ribosyl-homocysteine was a very poor substrate, indicating its normal disposal and salvage as methionine through the Methionine Cycle (Fig. 1A). Glycerol was also a weak substrate for the kinase as previously reported for the B. subtilis enzyme (Sekowska et al., 2001). Glycerol was removed from the enzyme and buffers prior to performing all assays. This also enabled accurate measurement of the intrinsic ATPase activity in the absence of any co-substrate (0.04 s⁻¹ for both the E. coli and B. subtilis kinase). This was significantly lower than previous reports with the B. thuringiensis enzyme, which included glycerol in the reaction mixture, likely overestimating the intrinsic ATPase activity (Beaudoin et al., 2018). The ExPEC (ATCC 25922) isomerase equally preferred 5-deoxyribose-1-phosphate and 5-methylthioribose-1-phosphate (Table 1; EcMtnA).

The aldolase exhibited similar preference for 5-deoxyribulose-1-phosphate and 5-methylthioribulose-1-phosphate (Table 1; EcAld2), irrespective of the metal cofactor (Fig. 3E; EcAld2), and with 10-fold greater catalytic efficiency compared to FucA (Table S1; EcFucA). Cobalt(II) not only resulted in the highest catalytic activity, but it was also the preferred metal cofactor. Natively purified Ald2 had the same maximal activity seen with the cobalt(II)-reconstituted apoenzyme and both were violet in color, indicative of a complexed cobalt metal cofactor (Fig. 3E). Currently, the protein residues coordinating the cobalt an its oxidation state are unknown. Recent characterization of an orthologous B. thuringiensis DHAP shunt aldolase revealed manganese(II) as the native cofactor, which conferred maximal activity (Beaudoin et al., 2018). However, the alternate metal cofactor preference by the R. rubrum and E. coli aldolases (Fig. 3E) indicates metal cofactor preference varies among the orthologous DHAP shunt aldolases, likely resulting from the specific MTA and 5’dAdo abundance levels encountered by each organism.

The catalytic efficiencies and substrate preferences of the DHAP shunt enzymes contrast with the B. subtilis Universal MSP homologues. The B. subtilis kinase, which was catalytically active with both 5-methylthioribose and 5-deoxyribose, preferred the former (Table 1; BsMtnK). Strikingly the B. subtilis isomerase had a > 60-fold higher substrate specificity for 5-methylthioribose-1-phosphate versus the 5-deoxy- species (Table 1; BsMtnA), and the B. subtilis 5-methylthioribulose-1-phosphate dehydratase, the first unique enzyme of the Universal MSP (Fig. 2B), had a 3000-fold higher specificity for 5-methylthioribulose-1-phosphate (Table 1; BsMtnB). This indicates that the kinase (MtnK), isomerase (MtnA), and dehydratase (MtnB) of the Universal MSP are primarily poised for recycling MTA to methionine during aerobic growth, whereas the DHAP shunt is employable for MTA and/or 5’dAdo during both aerobic and anaerobic growth.

Discussion

All organisms are faced with SAH, MTA, and 5’dAdo as inhibitory byproducts of routine enzymatic processes involving SAM as a co-substrate (Fontecave et al., 2004; Parveen and Cornell, 2010). The DHAP shunt in both R. rubrum (Fig. 3) and E. coli ExPEC strains (Fig. 4) is specific for the salvage of MTA and 5’dAdo, with disposal of SAH relegated to the Methionine Cycle (Table 1 and Fig. 1A). Thus, organisms with the DHAP shunt and the Methionine Cycle can utilize all major byproducts of SAM-dependent enzymes to regenerate nucleotides from the salvaged adenine or hypoxanthine, regenerate SAM from the salvaged methionine, and support further cell growth and metabolism through a novel shunt into central carbon metabolism via DHAP and acetaldehyde.

The DHAP shunt is widespread, occurring in at least 6 different bacterial phyla and in > 10 % of sequenced species (Fig. S6A) (North et al., 2017; Beaudoin et al., 2018). Like the Methionine Cycle, the DHAP shunt is oxygen independent for functionality in both oxic and anoxic environments (Fig. 3A). Previous R. rubrum transcriptomic and proteomic analyses demonstrate that the DHAP shunt genes appear to be constitutively expressed and gene products produced during both aerobic and anaerobic growth, irrespective of extracellular MTA and 5’dAdo concentration, for continual salvage of MTA and 5’dAdo (Erb et al., 2012; Cho et al., 2015; North et al., 2016). Given the propensity of 5Ado formation by Radical SAM enzymes during anaerobic metabolism (Figs. 3A and S1) (Broderick et al, 2014), this underpins the utility of the DHAP shunt in anoxic environments, and likely explains the observation that the DHAP shunt genes are enriched in facultative and obligate anaerobic bacteria (North et al., 2017; Beaudoin et al., 2018). This also contrasts to the oxygen-dependent Universal MSP, which is employed by nearly all eukaryotes and some bacteria (e.g. Bacillus, Klebsiella, Pseudomonas) for strict aerobic recycling of MTA to methionine (Fig. 2B) (Guan et al., 2012; Guan et al., 2011; Albers, 2008). Most eukaryotes appear to possess only a few Radical SAM enzymes (8 in humans versus 25 in R. rubrum), which are highly regulated, suggesting that SAH and MTA are the primary SAM byproducts (Broderick et al., 2014). This is consistent with the selectivity of the Universal MSP for MTA and the observation that in humans 5’dAdo is not metabolized in vivo by the Universal MSP beyond the phosphorylase step. The resulting 5-deoxyribose-1-phosphate is the primary terminal product, but may be further converted to 5’-deoxyinsoine via reaction with hypoxanthine by a purine nucleoside phosphorylase (Plagemann and Wohlhueter, 1983; Savarese et al., 1981). A recent report interpreted the presence of both the Universal MSP genes and DHAP shunt genes in B. thuringiensis as distinct pathways for disposal of MTA and 5’dAdo, respectively (Beaudoin et al., 2018). While this is likely a specific case for aerobic growth in this organism, the DHAP shunt is the probable route for both compounds during anaerobic growth when the Universal MSP is inactive. Indeed, the current study provides a unified role of the DHAP shunt genes in both MTA and 5’dAdo salvage (Fig. 1B), reconciling seemingly disparate reports of their function in MTA metabolism and 5’dAdo metabolism (North et al., 2017; Miller A et al., 2018; Beaudoin et al., 2018).

For MTA metabolism, the DHAP shunt constitutes one of several recently discovered anaerobic salvage pathways. The MTA-isoprenoid shunt (Fig. 2A) converts MTA to 1-deoxyxylulose −5-P (isoprenoid precursor) and methanethiol (methionine precursor) under both aerobic and anaerobic growth conditions (North et al., 2017; Erb et al., 2012; North et al., 2016; Cho et al., 2015; Warlick et al., 2012). In R. rubrum, RuBisCO itself appears to be involved in a distinct, albeit unresolved, anaerobic MTA metabolism pathway (Fig. 2C) (Singh and Tabita, 2010; Dey et al., 2015). In the archaeon, Methancaldococcus jannaschii, MTA deaminase (MtaD), 5’-methylthioinosine/5’-deoxyinosine phosphorylase (MtiP), and an isomerase (MtnA), analogous to the beginning of the Pseudomonas aeruginosa Universal MSP variation, appear to function in both MTA and 5’dAdo metabolism (Fig. 2B) (Guan et al., 2012; Guan et al., 2011; Miller et al., 2013; Miller D et. al., 2018). The remainder of the pathway remains elusive. Salvage of 5’dAdo is likely critical in archaea given the large number of Radical SAM enzymes encoded in the genome (e.g. 30 out of 1726 genes in M. jannaschii). Even though archaea do not appear to have a homolog of the aldolase based upon sequence similarity, it is altogether likely that they employ a DHAP shunt analog by virtue of a functionally analogous aldolase.

The common feature of these pathways is the initiation of MTA and 5’dAdo metabolism by the phosphorylase (or deaminase and phosphorylase) and isomerase, as verified for the DHAP shunt (this work) and M. jannaschii enzymes (Miller et al., 2013; Miller D et. al., 2018). Therefore, it is plausible that 5’dAdo can proceed through the MTA-isoprenoid shunt (Fig. 2A) in analogous fashion to produce 1-deoxyxylulose-5-phosphate for isoprenoids, and through the RuBisCO-dependent pathway (Fig. 2C) to produce lactate or pyruvate for central carbon metabolism. Subsequent studies will likely identify these and additional unknown variations in MTA and 5’dAdo metabolism, further illuminating the bacterial diversity in SAM byproduct recycling for needed sulfur and carbon salvage and production of novel metabolites.

Lastly, this study reveals a surprising enrichment of DHAP shunt genes in pathogenic strains of several organisms, including Bacillus, Clostridium, and E. coli (Fig. S6 and Table S2). Putative MTA salvage genes were first identified in a 1999 France neonatal meningitis ExPEC isolate (S88) as putative mtnK (ECS88_4901), mtnA (ECS88_4901), and mtnB (ECS88_4903) orthologues of the universal MSP (Touchon et al., 2009). However, the appearance of an incomplete MSP was perplexing, and the potential functionality speculative (Sekowska et al., 2019). The E. coli studies presented here (Fig. 4) demonstrate that this gene cluster, present in 42% of all ExPEC isolates but nearly absent in intestinal pathogenic E. coli isolates (Fig. S6B-C), is a functional DHAP shunt for carbon salvage as DHAP from MTA and 5’dAdo and acetaldehyde from 5’dAdo. Further studies are required to determine the utility, if any, for (2-methylthio)acetaldehyde produced from MTA.

In the extraintestinal environment of urine, blood, and cerebrospinal fluid, available carbon and sulfur sources are often limiting or present in compounds such as urea, organic acids, purines, and amino acids typically not encountered by commensal E. coli strains (Mann et al., 2017; Alteri et al., 2009). Thus, ExPEC cells are faced with chemical challenges that they must tolerate and/or develop specific metabolic strategies compared to commensals. While ExPEC strains without DHAP shunt genes are clearly able to cause infection in such challenging environments, the DHAP shunt could provide and additional advantage via carbon (and potentially sulfur) salvage of native MTA and 5’dAdo produced by essential polyamine and Radical SAM enzyme reactions. Many pathogenic strains exhibit elevated or altered polyamine synthesis profiles compared to commensals (Bower and Mulvey, 2006; Compilongo et al., 2014), and that Radical SAM enzymatic activities are necessary for synthesis of key metabolites such as biotin (Choi-Rhee and Cronan, 2005; Challand et al., 2009). The DHAP shunt also serves as an effective scavenger of extracellular 5-deoxyribose, enabling cell maintenance and growth of ExPEC strains (Fig. 4F). DHAP shunt genes in ExPEC strains are expressed during urinary tract infection and may be upregulated compared to growth in rich media (Subashchandrabose et al., 2014; Bielecki et al., 2014). Host-produced MTA and 5’dAdo (also as 5’-deoxyinosine) are common components of blood and urine (0.02 – 0.24 μmol per mmol creatinine) (Liebich et al., 1997; Lee et al., 2004), and 5-methylthioribose and 5-deoxyribose are exported metabolites from commensal E. coli (Choi-Rhee and Cronan, 2005; Hughes, 2006). Therefore, these modified ribose compounds, which appear otherwise inaccessible to nonpathogenic strains, are available to ExPEC strains and likely other pathogenic species with DHAP shunt genes to aid cell growth and maintenance in challenging environments (Fig. 4 and Table S2). The baseline information provided by this study will further future efforts to determine if the MTA and 5’dAdo salvage reactions play any specific role in pathogenesis. Certainly, their presence in only pathogenic members of several genera (Fig. S6 and Table S2 other tables) is tantalizing and might suggest potential specific therapeutic targets. Moreover, knowledge of the regulation and role of the salvage reactions in metabolism in conjunction with specific DHAP shunt gene deletions in highly virulent ExPEC strains derived from the current studies will make future efforts with ExPEC/UPEC infections using standard animal models feasible (Hung et al., 2009; Knappe et al., 2012).

Experimental Procedures

Fine Chemicals –

All chemicals and reference standards were from Sigma-Aldrich, [5’-H³]-5’-deoxyadensoine was from Moravek, and [methyl-C¹⁴]-5’-methylthioadenosine was from Perkin-Elmer. [methyl-C¹⁴]-labeled 5-methylthioribose, 5-methylthioribose-1-phosphate, 5-methylthioribulose-1-phosphate, (2-methylthio)acetaldehyde, and (2-methylthio)ethanol were synthesized enzymatically from [methyl-C¹⁴]-5’-methylthioadenosine and purified exactly as previously described (North et al., 2017; Erb et al., 2012). [5-H³]-labeled 5-deoxy-analogs were synthesized enzymatically from [5’-H³]-5’-deoxyadensoine and purified in the same manner. [2-H³]-acetate and [2-H³]-ethanol were enzymatically synthesized from [2-H³]-acetaldehyde using S. cerevisiae aldehyde dehydrogenase (Sigma) or alcohol dehydrogenase (Sigma), respectively, following manufacturer’s directions.

Fine Chemical Analysis –

Specific production of 5-deoxyribose-1-phosphate from 5’dAdo by purified R. rubrum phosphorylase was verified by LC-ESI-FTMS (Orbitrap, Thermo-Fisher) using a zic-pHILIC column as previously described and outline below (Erb et al., 2012). Specific production of 5-deoxyribulose-1-phosphate by purified R. rubrum isomerase was verified by coupled assay with E. coli FucA, triose phosphate isomerase, and glycerol-3-phosphate dehydrogenase as previously described (Erb et al., 2012). Formation of acetaldehyde, ethanol, and acetate by purified R. rubrum aldolase and the respective alcohol or aldehyde dehydrogenase was verified by gas chromatography on a DB-wax capillary column (Agilent) attached to a Shimadzu GC-14A system with helium carrier gas. Compounds were separated by a linear thermal gradient from 40 ° to 180 °C at Δ10 °C min⁻¹, detected using a flame ionization detector, and identified based on the unique retention time of reference standards. High resolution/accurate mass measurements were made using a Q-Exactive MS (Thermo-Fisher Scientific) and custom built microLC (Eksignet) and HTS-PAL autosampler (LEAP Technologies) using a custom packed 0.5 × 150 mm Zic-pHILIC (Merck SeQuant and Higgins Analytical) with 10 mM ammonium bicarbonate in water (A) and 10 mM ammonium bicarbonate in 95% acetonitrile (B). One microliter of sample was loaded on the column then eluted by holding at 100% B for 5 minutes followed by a ramp down to 90% B over 20 minutes before washing and re-equilibrating the column by ramping to 100 over three minutes then back to 100% B over one minute with a final hold at 100%B for 16 minutes. High-resolution profile data were collected in full-scan polarity switching mode at a resolution setting of 140,000 (at m/z 200). Data were analyzed using the QualBrowser application of Xcalibur (Thermo-Fisher Scientific).

Purified enzymes –

Expression plasmids (pET28) for R. rubrum phosphorylase (RrMtnP), isomerase (RrMtnA), and aldolase (RrAld2), and E. coli fuculose phosphate aldolase (Ec FucA) were as previously reported (Erb et al., 2012), as were expression plasmids for B. subtilis kinase (BsMtnK), isomerase (BsMtnA), dehydratase (BsMtnB), and enolase (BsMtnW) (Imker et al., 2007). Genes for E. coli (ATCC 25922) MTA/SAH/5’dAdo nucleosidase (EcMtnN), kinase (EcMtnK), isomerase (EcMtnA), and aldolase (EcAld2) were cloned into pET28 using primers listed in Table S3. All plasmids were transformed into E. coli Rosetta(DE3) pLysS expression host (Novagen) and strains were grown at 37 °C with shaking at 300 r.p.m. in lysogeny broth supplemented with appropriate antibiotics and 10 ml L⁻¹ Ormerod’s trace elements solution where indicated (Ormerod et al., 1961). When cultures reached an optical density at 600 nm of ~0.6, cells were chilled to 18 °C, induced with 250 μM IPTG, and induction was allowed to proceed at 18 °C with shaking at 200 r.p.m. for 12–16 h. Cells were chilled on ice, harvested by centrifugation, lysed with a French pressure cell, and purified by nickel chromatography as previously reported (Erb et al., 2012). Purified enzymes were concentrated into 20 mM Tris pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, and 300 mM NaCl (TED300) by centrifugal concentration device at 4 °C (Amicon, Millipore), quantified by UV absorbance based on calculated extinction coefficient at 280 nm, assayed for purity by SDS-PAGE (Fig. S11), and stored at −80 °C in 20 % glycerol.

Strains and growth conditions –

Rhodospirillum rubrum wild type (ATCC 11170) and phosphorylase (ΔmtnP), isomerase (ΔmtnA), and aldolase (Δald2) deletion strains were as previously described (North et al., 2017). R. rubrum was grown at 30 °C under aerobic chemoheterotrophic conditions in the dark or anaerobic, photoheterotrophic conditions with 2000 lux incandescent illumination in sulfur-free Ormerod’s malate minimal media supplemented with 1 mM ammonium sulfate (Dey et al., 2015). R. rubrum gene complementation studies were performed using plasmid pMTAP expressing the respective gene as previously described (North et al., 2017). E. coli commensal strains (K12 JM109, Agilent) and (K12 BW25113) (Baba et al., 2006) and E. coli ExPEC strain (ATCC 25922) were routinely cultured in Lysogeny Broth at 37 °C. Marker-less deletion of the individual nucleosidase (Δpfs), kinase (ΔmtnK), isomerase (ΔmtnA), and aldolase (Δald2) was performed using the λ-red system with primers listed in Table S3 (Datsenko and Wanner, 2000).

For R. rubrum growth experiments, cells were initially grown to mid-log phase at 30 °C in aerobic 50 ml culture tubes with shaking at 280 r.p.m. or in static 30 ml anaerobic culture tubes flushed with nitrogen and containing 10 ml of sulfur-free Ormerod’s malate minimal media supplemented with 1 mM ammonium sulfate. Cells were washed 3 times either aerobically or anaerobically with sulfur-free Ormerod’s malate minimal media and used to inoculate 10 ml media to an initial culture optical density at 660 nm (OD_660nm) of ~0.02 under the same conditions.

For E. coli growth experiments, cells were initially grown aerobically in 5 ml Lysogeny Broth at 37 °C in the dark with shaking at 300 r.p.m. Cells were then grown twice in the dark at 37 °C with shaking at 300 r.p.m. in 25 ml aerobic culture tubes or 30 ml anaerobic culture tubes flushed with nitrogen and containing 5 ml of sulfur-free, carbon-free modified M9 media (per liter water; 6.2 g Na₂HPO₄, 3.0 g KH₂PO₄, 0.5 g NaCl, 1.0 g NH₄Cl, 0.4 g MgCl₂ ◦ 6H₂O, 14.7 mg CaCl₂ ◦ 2H₂O, 6.8 mg FeCl₃ ◦ 6H₂O, 2.2 mg ZnCl₂, 22.3 mg EDTA ◦ 2Na2H₂O, 1.37 mg NiCl₂ ◦ 4H₂O, 1.24 mg MnCl₂ ◦ 4H₂O, 1.4 mg H₃BO₄, 0.137 mg CuCl₂ ◦ 2H₂O, 0.37 mg NaMoO₄ ◦ 2H₂O, 0.124 mg CoCl₂ ◦ 6H₂O, 1 mg thiamine) supplemented with 1 mM ammonium sulfate and 25 mM glucose or 40 mM glycerol. Cultures were then washed 3 times either aerobically or anaerobically with sulfur-free, carbon-free modified M9 media and used to inoculate 2 ml of sulfur-free, carbon-free modified M9 media in 10 ml aerobic or anaerobic culture tubes flushed with nitrogen to an initial OD_400nm ~ 0.02. Sulfur source (ammonium sulfate, MTA, 2-methylthioethanol) was added to 1 mM and carbon source (glucose or 5-deoxyribose) was added to 25 mM final concentration. In the case of 5-deoxyribose growth experiments, 5 mM additions were made every 6 hours for 24 hours to 25 mM total, as a single initial addition of 25 mM resulted in long lag times, presumably due to an inhibitory effect of non-physiological levels of 5-deoxyribose as previously reported (Beaudoin, et. al., 2018). All anaerobic manipulations were performed in an anaerobic chamber (Coy Laboratories).

For feeding experiments, E. coli strains were grown in the dark at 37 °C with shaking at 300 r.p.m. to late log phase (OD_600nm ~ 1.0) in 50 ml aerobic cultures tubes containing 10 ml sulfur-free, carbon-free modified M9 media supplemented with 1 mM ammonium sulfate and 25 mM glucose. R. rubrum strains were grown photoheterotrophically to late log phase (OD_660nm ~ 1.2) in 15 ml anaerobic culture tubes containing 15 ml sulfur-free Ormerod’s malate minimal media supplemented with 0.1 mM ammonium sulfate.

Quantification of native MTA and 5’dAdo –

During R. rubrum growth experiments (Fig. S1), 0.5 ml of culture was collected, cells were rapidly removed by centrifugation at 10,000 g, and the supernatant was applied to a C18 reverse phase column (Altima) connected to a Shimadzu Prominence HPLC with UV detection at 260 nm. Nucleotides were eluted on a linear gradient of 1%−25% acetonitrile in 20 mM ammonium acetate, pH 6.8. Nucleotides were identified by retention time based on known standards, collected, and verified by ESI-QTOF mass spectrometry (Bruker maXis, Fig. S2). Concentration of MTA and 5’dAdo was calculated from peak absorbances calibrated against concentration curves of reference standards. One liter of culture at an optical density of 1 measured at 660 nm (OD_660nm) corresponded to 0.529 ± 0.087 g of cell dry weight (CDW). This value was used to calculate the concentration of MTA and 5’dAdo in μmol per g CDW.

Feeding experiments –

R. rubrum feeding experiments with [5’-H³]-5’-deoxyadenosine were performed as previously described for MTA under anaerobic conditions (North et al., 2017). Briefly, cells were grown anaerobically, photoheterotrophically as described above. Cells were harvested anaerobically by centrifugation, washed 3 times anaerobically with sulfur-free Ormerod’s malate minimal media, and resuspended to a final OD_660nm ~ 10. Aliquots of 200 μl of cells were sealed in 1 ml serum vials in an anaerobic chamber (Coy Laboratories). 5’-deoxyadenosine was added to 100 μM and [5’-H³]-5’-deoxyadenosine to 10 μM final concentration, and cells were incubated at 30 °C, 2000 lux incandescent illumination for the indicated amount of time. Cells and accompanying media were flash frozen, and metabolites were extracted by addition of 95 % acetonitrile, 0.2 N ammonium hydroxide to 80% acetonitrile final concentration. Samples were vortexed at room temperature for 5 minutes followed by incubation at −20 °C for 20 minutes. Debris was removed by centrifugation at 21000 g before extracted metabolites were analyzed by liquid chromatography. Similarly, E. coli cells were grown aerobically in modified M9 media as described above, washed three times with sulfur-free, carbon-free modified M9 media, and resuspended to a final OD_600nm ~ 10 in the same media supplemented with 1 mM ammonium sulfate and 25 mM glucose. Unlabeled and labeled MTA or 5’dAdo were added to the same concentrations indicated above and cells were placed in a conical 5 ml glass vial, shaken at 100 r.p.m. at 37 °C, and continuously bubbled with a stream of purified air. Fractions of 200 μl cells were collected at the indicated time, flash frozen, and process as described for R. rubrum.

Metabolite detection and analysis –

Metabolites emanating from [5’-methyl-C¹⁴]-5’-methythioadenosine and [5’-H³]-5’-deoxyadenosine were resolved by HPLC on a Shimadzu Prominence HPLC with inline UV-absorbance detector and β-ram scintillation counter (IN/US Systems). Phosphorylated species, MTA, 5-methylthioribose, 5’dAdo, and 5-deoxyribose were resolved by zic-pHILIC (Millipore) as previously described (Miller A et al., 2018). MTA, 5-methylthioribose, and (2-methylthio)ethanol were resolved by C18 Reverse Phase HPLC (Alltima) as previously described after removal of sample acetonitrile by centrifugal vacuum concentration (North et al., 2017). Ethanol, acetaldehyde, and acetate were separated by Resex ion exclusion chromatography (Phenomenex) isocratically with 0.05 N H₂SO₄ after removal of sample acetonitrile by centrifugal vacuum concentration. Metabolites were identified by retention time against pure synthesized H³- and C¹⁴- labeled standards (see Fine Chemicals).

Enzyme Kinetics –

Enzyme kinetics were performed essentially as previously described by either following the nucleosidic cleavage of MTA, 5’dAdo, SAH, or adenine (nucleosidase or phosphorylase assay) at 275 nm (ε = 1600 M⁻¹ cm⁻¹) (Erb et al., 2012), following the formation of ADP by NADH reduction at 340 nm (ε = 6220 M⁻¹ cm⁻¹) in an assay coupled with pyruvate kinase and lactate dehydrogenase (MtnK kinase assay) (Ku et al., 2004), following the formation of DHAP by NADH reduction via coupled assay with triose phosphate isomerase and glycerol-3-phosphate dehydrogenase (MtnA isomerase, Ald2 aldolase, and FucA aldolase assay) (Erb et al., 2012), or following the formation of 2,3-diketopentyl-1-phosphate compounds (MtnB dehydratase assay) in a assay coupled with B. subtilis MtnW enolase followed at 280 nm (ε = 9500 M⁻¹ cm⁻¹) (Ashida et al., 2008) (Figs. S8-S10). Alternatively, the complementary formation of the aldehyde species was followed by NADH reduction via coupled assay with alcohol dehydrogenase (Sigma) with the same results (MtnA isomerase and Ald2 aldolase assays). Assays were performed on a spectrophotometer (Carey 100 UV-VIS, Agilent) at 30 °C in a 400 μl total reaction volume, 1 cm quartz cuvette, with 0.1 s integration and 1 nm bandwidth. MtnB dehydratase assays were also performed at 35 °C due to lack of sufficient measurable activity with 5-deoxyribulose-1-phosphate and ribulose-1-phosphate. E. coli nucleosidase assays were performed in a 3 ml reaction volume with 1 s integration and 2 nm bandwith for sufficient signal at the low substrate concentrations (< 1 μM) required for K_m measures. All kinetics experiments were performed in triplicate and data were fit by nonlinear weighted regression (MatLab) to a Michaelis-Menten model. For the MtnK kinase assay, glycerol was removed from all enzymes by centrifugal concentration device (Amicon, Millipore) and omitted from all buffers. For apoenzyme formation of the aldolases, E. coli and R. rubrum Ald2 was dialyzed in a microdialysis chamber for 24 hours in TED300 + 100 mM EDTA at 4 °C followed by dialysis for 24 hours into TED300 buffer pre-treated with Chelex Resin (Biorad). Holoenzyme was reconstituted using 0.1 mM of the indicated metal cofactor for 1 hour before use in assays.

Assay conditions:

The R. rubrum phosphorylase (RrMtnP) and B. subtilis dehydratase (BsMtnB) were performed as previously reported (Erb et al., 2012; Ashida et al., 2008). Assay of E. coli MtnN; 40 mM K-phosphate pH 7.8, 0.1 M MgCl₂, 0.1 – 6.0 μg enzyme, and 0 – 500 μM MTA, SAH, 5’dAdo or adenosine. Assay of E. coli and B. subtilis MtnK: 100 mM MOPS-KOH pH 7.8, 5 mM MgCl₂, 0.4 mM NADH, 1 mM phospho-enol pyruvate, 1 mM ATP, 10 μM E. coli MtnN, 2.5–5 U pyruvate kinase and lactate dehydrogenase (Sigma), 0.2–0.4 μg enzyme and 0 – 90 mM MTA, SAH, 5’dAdo or adenosine. The kinase substrate was produced in situ by incubating the assay mixture for 2 minutes before the addition of MtnK. Assay of E. coli, R. rubrum, and B. subtilis MtnA: 50 mM K-phosphate pH 7.8, 5mM MgCl₂, 0.3 mM NADH, 0.3 mM CoCl₂, 5 μM FucA, and 10 μM S. cerevisase alcohol dehydrogenase (Sigma) or 2.5 U triose phosphate isomerase and glycerol-3-phosphate dehydrogenase (Sigma), and 2.0–4.0 μg enzyme. The substrates for MtnA were formed from 10 – 50 mM of MTA, 5’dAdo, or adenosine by incubation in 100 mM K-phosphate pH 7.8, 10 – 50 mM MgCl₂, 10 – 50 mM ATP, 30 μM E. coli MtnN, and 30 μM E. coli MtnK at 35 °C for 1 hr. Reaction completeness and substrate purity was assayed by C18 Reverse Phase HPLC before use in MtnA isomerase assay. Assay of E. coli Ald2 and FucA and R. rubrum Ald2: 40 mM K-phosphate pH 7.8, 5 mM MgCl₂, 10 μM R. rubrum MtnP and MtnA, 0.3 mM NADH, 0.3 mM CoCl₂, 10 μM S. cerevisase alcohol dehydrogenase (Sigma) or 2.5 U triose phosphate isomerase and glycerol-3-phosphate dehydrogenase (Sigma), and 0.03–10.0 μg enzyme. The aldolase substrate was produced in situ by incubating the assay mixture for 2 minutes before the addition of FucA or Ald2.

Bioinformatics –

The NCBI and Kegg reference protein databases were queried by protein BLAST using the R. rubrum MtnP, MtnA, or Ald2, or E. coli MtnK protein sequences. Organisms with putative DHAP shunt were identified in which MtnP or MtnK, and MtnA, and Ald2 each had an e-value < e-20 and the corresponding genes were all co-localized in the same gene cluster on the organism’s genome (Table S2). For E. coli, the NCBI E. coli sequenced genome database was queried by nucleotide BLAST using the ATCC 25922 DNA sequence of the DHAP Shunt genes to identify other isolates containing the DHAP Shunt (Table S4) with coverage > 400 bp and e-value < e-177; sequences with larger e-values and smaller coverage did not correspond to the DHAP shunt genes. The E. coli isolates with DHAP shunt genes (Table S4) were then correlated with all E. coli isolates listed in the NCBI database (Table S5), listed ExPEC isolates in the NCBI database from curated collections (Table S6), and all listed intestinal pathogenic E. coli isolates in the NCBI database (Table S7) to determine the frequency of DHAP Shunt genes in the intestinal versus extraintestinal pathogenic E. coli groups (Table S6B).