RADICAL ENZYMES Itaconyl-CoA forms a stable biradical in methylmalonyl-CoA … · RESEARCH ARTICLE RADICAL ENZYMES Itaconyl-CoA forms a stable biradical in methylmalonyl-CoA mutase

RESEARCH ARTICLE◥

RADICAL ENZYMES

Itaconyl-CoA forms a stable biradical inmethylmalonyl-CoA mutase and derails its activityand repairMarkus Ruetz1, Gregory C. Campanello1*, Meredith Purchal2, Hongying Shen3,4, Liam McDevitt1,Harsha Gouda1, Shoko Wakabayashi5, Junhao Zhu5, Eric J. Rubin5, Kurt Warncke6,Vamsi K. Mootha3,4, Markos Koutmos2,7, Ruma Banerjee1†

Itaconate is an immunometabolite with both anti-inflammatory and bactericidal effects. Its coenzymeA (CoA) derivative, itaconyl-CoA, inhibits B12-dependent methylmalonyl-CoA mutase (MCM) by anunknown mechanism. We demonstrate that itaconyl-CoA is a suicide inactivator of human andMycobacterium tuberculosis MCM, which forms a markedly air-stable biradical adduct with the5′-deoxyadenosyl moiety of the B12 coenzyme. Termination of the catalytic cycle in this way impairscommunication between MCM and its auxiliary repair proteins. Crystallography and spectroscopyof the inhibited enzyme are consistent with a metal-centered cobalt radical ~6 angstroms away from thetertiary carbon-centered radical and suggest a means of controlling radical trajectories during MCMcatalysis. Mycobacterial MCM thus joins enzymes in the glyoxylate shunt and the methylcitrate cycle astargets of itaconate in pathogen propionate metabolism.

The immunomodulatory and antimicro-bial effects of itaconate are evincingnewfound interest in a compound his-torically used as a precursor in polymersynthesis. Upon activation, immune cells

stimulate itaconate synthesis ~10-fold via acon-itate decarboxylase (Irg1)–catalyzed decarboxyl-ation of cis-aconitate, a tricarboxylic acid cycleintermediate (1). Itaconate activates Nrf2, in-hibits succinate dehydrogenase, and blocks thetranscription factor Ikbz, leading to a switchfrom a pro- to an anti-inflammatory state (2).The antimicrobial activity of itaconate is pur-portedly due to its inhibition of twomicrobe-specific targets: isocitrate lyase in the glyoxylateshunt and methylcitrate lyase in the methyl-citrate cycle, two enzymes that are needed forpathogen survival on acetate or propionate,respectively, as the sole carbon source (Fig. 1A)(3). Propionyl-coenzymeA (CoA) is derived fromcholesterol catabolism and is used by patho-gens like Mycobacterium tuberculosis (Mtb)for biomass production in the glucose-limitingconditions found in phagosomes (4).An unexpected intersection between itacon-

ate and B12-dependent propionate metabo-

lism was revealed recently by the demon-stration that itaconyl-CoA (I-CoA) is a potent in-hibitor of human 5′-deoxyadenosylcobalamin(AdoCbl)–dependentmethylmalonyl-CoAmutase(hMCM) (5). Itaconate can be cleared by a C5-

dicarboxylate pathway (6) via acylation to I-CoA,hydration to citramalyl-CoA, and cleavage toacetyl-CoA and pyruvate, a reaction catalyzedby citramalyl-CoA lyase, which is encoded bythe recently deorphaned citrate lyase beta-like (CLYBL) gene (5) (Fig. 1A). I-CoA (ormethylenesuccinyl-CoA) is an analog ofsuccinyl-CoA, which is interconverted tomethylmalonyl-CoA (M-CoA) by the isomeraseMCM. Ametabolic connection between CLYBLand B12 was initially revealed in a genome-wideassociation study (7) that showed a correlationbetween the biallelic loss of CLYBL, which hasan ~3 to 6% prevalence in certain populations(8), and B12 deficiency. Although I-CoA inhibi-tion ofMCMprovides amolecular link betweenmitochondrial B12 and C5-dicarboxylate metab-olism, it does not explain why inhibited MCMcannot be repaired by the auxiliary protein sys-tem that is dedicated for this function (9). Giventhe homology between bacterial and humanMCM, host-derived itaconate could also poten-tially target MCM in pathogenic bacteria, inwhich it is required for lipid breakdown.Itaconate induces electrophilic stress and

modifies small molecules and protein targetsby alkylating cysteine residues (10, 11). Here, wereport a radical suicide inactivation mechanismin which addition of the elusive 5′-deoxyadenosylradical (dAdo•) to I-CoA in MCM leads to anair-stable biradical comprising a tertiary car-bon radical coupled to the metal-centered cob

RESEARCH

Ruetz et al., Science 366, 589–593 (2019) 1 November 2019 1 of 5

1Department of Biological Chemistry, University of Michigan,Ann Arbor, MI 48109, USA. 2Department of Chemistry,University of Michigan, Ann Arbor, MI 48109, USA. 3HowardHughes Medical Institute and Department of MolecularBiology, Massachusetts General Hospital, Boston, MA 02114,USA. 4Broad Institute, Cambridge, MA 02142, USA.5Department of Immunology and Infectious Diseases,Harvard T. H. Chan School of Public Health, Cambridge, MA02115, USA. 6Department of Physics, Emory University,Atlanta, GA 30322, USA. 7Program in Biophysics, Universityof Michigan, Ann Arbor, MI 48109, USA.*Present address: Merck & Co., Inc., Kenilworth, NJ, USA.†Corresponding author. Email: [email protected]

Fig. 1. I-CoA inhibits human and Mtb MCM by forming an air-stable biradical. (A) Mtb pathways targetedby I-CoA. MCC, methylcitrate cycle; ICL, isocitrate lyase; S-CoA, succinyl-CoA;TCA, tricarboxylic acid cycle. (B andC) Titration of holo-MtbMCM (B)with 30mMbound AdoCbl or holo-hMCM (C) with 40 mMbound AdoCbl (black traces)with increasing concentrations of I-CoA. The intermediate spectra (gray) were recorded after 5 min of equilibration.(Insets) Representative plots of D528 or D530 nm versus I-CoA indicated stoichiometric binding [n = 2 (MtbMCM);n = 3 (hMCM)]. (D) EPR spectra of the 1 mM I-CoA-induced biradical on hMCM (375 mM) in the presence ofnatural abundance (top) or [13C]I-CoA (bottom). The experimental and simulated spectra are in black and gray,respectively. (E and F) Possible fates of dAdo• when I-CoA behaves as substrate (E), not observed, or inhibitor (F).

on June 3, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/

(II)alamin radical. We visualized Mtb MCMbound to a dAdo adduct of I-CoA at 2.0-Åresolution by x-ray crystallography. In addi-tion to identifying MCM as an antimicrobialtarget of itaconate and demonstrating its im-portance for Mtb growth on propionate, ourstudy provides molecular insights into howMCM controls radical trajectories during catal-ysis with its normal substrates to promote thedesired chemistry and suppress unwanted sidereactions.

I-CoA inactivates Mtb MCM by forming anair-stable biradical

MCM catalyzes the reversible isomerization ofM-CoA to succinyl-CoAviaanAdoCbl-dependentradical mechanism. To test our hypothesis thatbacterial MCM is also a target of the anti-microbial effect of itaconate, we cloned andexpressed Mtb MCM and the two auxiliaryproteins that load and repair AdoCbl (fig. S1A).The kinetic parameters of Mtb MCM are com-parable to those of the human homolog (fig.S1, B and C). Addition of I-CoA to Mtb MCM-AdoCbl led to a rapid shift in the absorptionmaximum (lmax), from 528 nm to 466 nm (Fig.1B), indicating stoichiometric binding (Fig. 1B,inset), as was also seen with hMCM (Fig. 1C)(5). Homolysis of the cobalt-carbon bond inAdoCbl, the first step in the MCM-catalyzedreaction, leads to formation of the radical pairdAdo• and cob(II)alamin, albeit with a differ-ent lmax (474 nm). We therefore used electronparamagnetic resonance (EPR) spectroscopy toidentify the 466-nm–absorbing products ofI-CoA–inactivated MCM.The EPR spectrum of MCM-bound cob(II)

alamin displays the typical eight-line hyperfinesplitting of the unpaired electron with the I =7/2 cobalt nucleus that is resolved in the high-field region of the spectrum (fig. S2, A and B).Additional superhyperfine splitting due to thecoordinating I = 1 lower axial nitrogen from ahistidine ligand is also observed. Notably, ad-dition of I-CoA to human or Mtb MCM led todistinct spectra (Fig. 1D, top, and fig. S2C) thathad the hallmarks of a hybrid triplet system.The spectra were reminiscent of the EPR spec-trum of a transient catalytic intermediatetrapped duringMCMturnover, exhibiting strongelectron-electron spin-coupling between theproduct succinyl-CoA radical and the low-spin cob(II)alamin (12). Similar biradicalintermediates have also been trapped in theAdoCbl-dependent enzymes glutamate mutase(13) and 2-methyleneglutarate mutase (14).The hyperfine multiplicity and the substantialg anisotropy identified the cobalamin compo-nent in the triplet spin system.The identity of the organic radical com-

ponent was further assessed with [13C]I-CoA(uniformly labeled in the itaconate carbons) or[13C]AdoCbl (uniformly labeled in the adenosylmoiety). Whereas 13C-labeled AdoCbl had no

effect on the EPR spectrum, [13C]I-CoA led toinhomogeneous broadening throughout theline shape, which is consistentwith appreciablemixing of cob(II)alamin and organic radicalquantum states in the strong electron-electroncoupling regime (Fig. 1D, bottom). These resultsindicate an absence of marked unpaired elec-tron spin density on the dAdo moiety, whichis in accord with the near-unit spin densityinferred for the electron-13C interaction in theradical pair formed from [13C]I-CoA. Spectralsimulations predict an interspin distance of 5to 6 Å, while the Euler angles position the or-ganic radical at an angle of 43° relative to theprincipal axis of the dz

2 orbital on cobalt. Inother AdoCbl-dependent isomerases, the dis-tance between the substrate and cobalamin

radicals range between 5 to 6 Å and 8 to 12 Å,respectively, necessitating small or large move-ments of the initially formed dAdo• to reachthe substrate hydrogen atom (H-atom) destinedfor abstraction (15).The I-CoA–induced biradical was air stable

for over 1 hour (figs. S3 and S4), unlike thetransient biradical formed during catalytic turn-over with M-CoA that was trapped by freeze-quenching (12). To understand the chemicalbasis of its unusual stability, we further inves-tigated the identity of the organic radical.

The adenosyl radical is stabilized by additionto itaconyl-CoA

Following generation of the dAdo•-cob(II)alaminradical pair onMCM, the isomerization reaction

Ruetz et al., Science 366, 589–593 (2019) 1 November 2019 2 of 5

Fig. 2. Crystallographic capture of a biradical in I-CoA-inactivated Mtb MCM. (A) Orientation of dAdo(green) in relation to the corrin ring (gray; pyrrole rings A to D and acetamides a and c are shown) in nativeMtb MCM. (B) 2Fo-Fc omit maps (blue) around B12 and I-CoA contoured at 1.5s. (C) Shift in B12 and rotationof the adenine ring from the coplanar (gray) to perpendicular (yellow) position relative to the corrin ring.(D) EPR spectra of Mtb MCM + I-CoA. (E) Geometry of B12 and the I-CoA–dAdo adduct in crystal.(F) Hydrogen bonding interactions in the MCM–I-CoA structure.

RESEARCH | RESEARCH ARTICLEon June 3, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/

is initiated byH-atom abstraction fromM-CoAby dAdo•, forming a substrate-centered radi-cal that undergoes rearrangement. The dAdo•is a primary and highly reactive alkyl radicalthat has eluded direct detection in all AdoCbl-dependent isomerases but recently was trappedin a radical S-adenosylmethionine enzyme (16).In principle, two potential reactions betweendAdo• and I-CoA can be considered (Fig. 1, Eand F). To distinguish between these mecha-nistic possibilities, the reaction products fromMtb and hMCM inactivated by I-CoA undersingle-turnover and aerobic conditions wereseparated by high-performance liquid chro-matography (fig. S5). Two major productpeaks with retention times of 22.9 and27.0 min (peaks 2 and 5) were identified;

Mtb MCM showed an additional peak at23.5 min (peak 2b).Matrix-assisted laser desorption/ionization–

mass spectrometry (MS) analysis confirmedthe second reaction pathway leading to anaddition product between dAdo and I-CoA(Fig. 1G), indicating that dAdo• adds to thedouble bond in I-CoA, yielding a tertiary car-bon radical additionally stabilized by de-localization onto the p-system of the adjacentcarboxylate. MS analysis revealed that peaks2a and 2b are isomers with the same mass/charge ratio (m/z) of 1145.4 (fig. S6), which is16 mass units higher than the expected massof the addition product (m/z 1129), indicatingincorporation of an oxygen atom. We assignpeak 2 as the hydroxyl derivative of the ad-

dition product and peak 5 (m/z 1099.3), whichis 30 Da lighter, indicating the formal loss offormaldehyde to the oxidative decarboxylationproduct of peak 2 (fig. S7A). MS analysis of thehMCM samples yielded similar results (fig. S6,D and E).Under anaerobic conditions, additional hydro-

phobic products were observed (fig. S8). Peak6 with an m/z of 1129.4 represents the ex-pected radical addition product, while peak5 and theminor peaks 7 and 8, withm/z valuesthat are two mass units lower (fig. S9), wereassigned to intramolecular cyclization and/orelimination products (fig. S7, B and C). Ado•cyclization products have been reported dur-ing anaerobic photolysis of AdoCbl (17) andduring the reaction of hydroxyl radicals withdAdo or deoxyguanosine (18), resulting in 5′,8-cyclopurine nucleosides.

Crystallographic capture of thebiradical on MCM

Given the protracted air stability of the biradical,we attempted to visualize the inhibited formby soaking AdoCbl-reconstituted Mtb MCMcrystals with I-CoA. We determined structuresof the unsoaked and I-CoA–soaked enzyme at1.9-Å and 2.0-Å resolution, respectively (tableS1). Mtb MCM, like the Propionibacteriumshermanii protein (19), is a heterodimer com-prising an a subunit that binds B12 and a b sub-unit that is inactive (fig. S10A). In the nativestructure, AdoCbl is boundwith its endogenousdimethylbenzimidazole tail inserted in a sidepocket while His-629 serves as the lower axialligand (Fig. 2A). On the opposite face, the 5′-carbon of the upper axial dAdo ligand is 2.5 Åaway from the cobalt atom, while the adeninegroup is coplanar with the corrin ring andoriented above pyrrole rings A and B.I-CoA binding induces a large conforma-

tional change in the AdoCbl-binding a subunit[Ca rootmean square deviation (RMSD), 1.47 Å],whereas the small subunit is almost unchanged(Ca RMSD, 0.26Å) (fig. S10A). Soaking in I-CoAdid not affect crystal stability, as there are nocrystal contacts in the region affected by itsbinding. The a subunit collapses around theI-CoA binding pocket, with the motion beinglargest at the periphery and smallest wherethe a and b subunits are proximal. The crys-tal structure of hMCM, which is a homodimerwith two B12 binding subunits, similarly closesin on its substrate, M-CoA (20).We assigned the electron density in the

active site to the adduct between I-CoA andthe 5′-carbon of dAdo (Fig. 2B and fig. S10B).The 5′-carbon of the dAdo moiety is rotatedalmost 180° away from the cobalt, and the dis-tance to the cobalt atom increases to 4.3 Å.The rotation places the 5′-carbon of dAdo 1.5 Åaway from the methylene group of I-CoA, in-dicating the presence of a covalent bond be-tween them.The geometry of the tertiary carbon

Ruetz et al., Science 366, 589–593 (2019) 1 November 2019 3 of 5

Fig. 3. I-CoA inactivation impairs MCM repair. (A) Scheme showing the role of the auxiliary proteins incofactor loading/off-loading to/from MCM. (B and C) Enzyme-monitored turnover by human (B) andMtb (C) MCM (black spectra) in the presence of M-CoA and human or Mtb CblA-GTP. Intermediatespectra (gray) were recorded every 2 min. Final spectra (red) were recorded at 1 hour. (D) Specificactivity (SA) of human and Mtb MCM after 1-hour preincubation without or with M-CoA (red versus blue)and subsequent addition of the repair system (orange). (E and F) Addition of I-CoA to hMCM-AdoCbl[black, (E)] or Mtb MCM-AdoCbl [black, (F)] results in inactive enzyme (gray). Further incubationover 1 hour causes only modest spectral changes (red). (G and H) At the end of the experiments in (E)and (F), the repair system was added for 20 min to human (G) and Mtb (H) MCM. The increase inabsorbance at 350 to 356 nm is indicative of OH2Cbl formation (red). (I) Same as in (D) but with I-CoA;in both panels, data represent means ± SD (n = 3).

RESEARCH | RESEARCH ARTICLEon June 3, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/

of I-CoA is planar, as expected for an sp2 carbon(Fig. 2B and fig. S10, C and D). dAdo and corrinare shifted by 2.1 Å relative to the corrin ring inthe structure without I-CoA. The acetamidegroup a on ring A is pushed up and in towardthe adenine ring (Fig. 2C), which causes theadenine to move from a parallel to an almostperpendicular position relative to the corrinplane (fig. S10E), as predicted computation-ally (21). A strong biradical EPR signal as-sociated with Mtb MCM crystals soakedwith I-CoA confirmed that the spin-coupledcarbon- and metal-centered radical pair canform in crystals (Fig. 2D). In the structurewith the adduct, the tertiary carbon is 6 Åfrom the cobalt and at an ~45° angle fromthe principal dz

2 orbital axis (Fig. 2E), in ex-cellent agreement with EPR simulations.To our knowledge, the only other enzyme-bound, carbon-centered radical that has beencrystallized is the acetyl-thiazolium cation rad-ical in pyruvate:ferredoxin oxidoreductase,which is formed via a one-electron transferto an iron-sulfur cluster (22).In the resting state, the C2′-OH in dAdo is

engaged in a hydrogen-bonding interactionbetween Gln346 and awater-mediated hydrogenbond network to His260 and Arg223 (fig. S10F).In the I-CoA–inactivated structure, the C2′-OHmaintains a hydrogen bond with only Gln346,and the adenine NH2 group forms hydrogenbonds to the backbone carbonyls of Gly107 andAla156 (Fig. 2F). These interactions likely orientdAdo• for H-atom abstraction in the catalyticcycle when M-CoA is present; however, whenI-CoA is present, it places dAdo• in close prox-imity to the double bond, setting up the radicaladdition reaction. Tyr105 forms a hydrogen bondwith the terminal carboxylate of I-CoA and alsoflanks the dAdomoiety. In the resting enzyme,Tyr105 points in the opposite direction, i.e., awayfrom the substrate. Arg223 also engages viaelectrostatic interactions with the terminalcarboxylate of I-CoA and the acetamide sidegroup c in the corrin ring. The structure hasimplications for howMCM controls the dAdo•radical trajectory to promote H-atom abstrac-tion fromM-CoA and concomitantly suppressesunwanted side reactions that could lead toradical extinction. As first predicted in com-putational studies (21), rotation and upwardmovement of the adenine ring (Fig. 2C) positionthe C5′-carbon radical for H-atom abstractionfrom substrate. In contrast toMCM, glutamatemutase (23) and diol dehydratase (24) useribose pseudorotation and N-glycosidic bondrotation, respectively, to bring the C5′-carbonof dAdo• to within van der Waal’s distance ofthe respective substrates.

I-CoA inhibits MCM repair

While cob(II)alamin is an intermediate in theMCM-catalyzed reaction, it also represents theinactive form of the enzyme when it becomes

decoupled from the dAdo moiety, and thus itfails to re-form AdoCbl at the end of thecatalytic cycle (9). Under these conditions, theauxiliary proteins CblA and ATR engage withMCM to off-load cob(II)alamin onto ATR forrepair (Fig. 3A). ATR catalyzes the adenosyl-ation of cob(I)alamin to form AdoCbl andthen transfers the cofactor to MCM to recon-stitute the holoenzyme (25, 26). Cofactor trans-fer in either direction between ATR andMCMrequires the heterotrimeric guanine nucleotide–binding protein chaperone CblA and is fueledby its guanosine triphosphatase (GTPase) activity(27). Because I-CoA leads to rapid inactivationofMCMand formation of cob(II)alamin, whichcannot re-form AdoCbl, we assessed whetherthe enzyme can be repaired by the ATR-CblAsystem.Addition ofM-CoA, to initiate catalytic turn-

over by MCM in the presence of CblA-GDP(guanosine diphosphate), led to small (hMCM)or no (Mtb MCM) changes in the absorptionspectra, indicating that these enzymes are re-sistant to oxidative inactivation (Fig. 3, B andC). Consistent with this finding, the specificactivities of both enzymes were reduced only~15% after 1 hour of preincubationwithM-CoA(Fig. 3D). By contrast, the P. shermanni andM. extorquensMCMaremuchmore prone toinactivation and accumulate aquocobalamin(OH2Cbl) during turnover (28). Addition of therepair system (ATR, ATP, and GTP) led to com-

plete recovery of MCM activity (Fig. 3D), con-sistent with successful off-loading of inactivecob(II)alamin from MCM followed by reload-ing of AdoCbl from the assay mixture. In con-trast to M-CoA, which supported catalyticturnover ofMCMwithminimal spectral changes,incubation of either human (Fig. 3E) or Mtb(Fig. 3F) MCM with I-CoA led to an imme-diate increase in absorption at 466 nm thatdid not change significantly over 1 hour andwas correlated with complete loss of activity(Fig. 3I). Addition of the respective humanandMtb repair proteins led to an increase inabsorbance at 350 to 356nmand530 to 534nm,signaling oxidation of cob(II)alamin to OH2Cbl(Fig. 3, G and H). By contrast, addition of therepair systems to the same enzymes incubatedwith M-CoA did not induce cofactor oxidation(fig. S11). Following repair, only 23% (human)and 38% (Mtb) of the initial MCM activity wasrecovered (Fig. 3I).To gain insights into why the repair process

is impeded by I-CoA but not M-CoA, we usedhMCM, which forms a stable complex withCblA when it is in need of repair but is free inthe active AdoCbl-bound state (fig. S12A) (27). Insize exclusion chromatography profiles, hMCMis a stand-alone dimer (173 kDa) in the presenceof M-CoA and the repair mixture (fig. S12B).However, in the presence of I-CoA, the MCMpeak broadens and shifts to 211 kDa (fig. S12C),indicating the presence of a 1:1 hMCM:CblA

Ruetz et al., Science 366, 589–593 (2019) 1 November 2019 4 of 5

Fig. 4. Itaconate inhibits B12-dependent Mtb and macrophage metabolism. (A) Vitamin B12 (10 mg/ml)stimulates growth of wild-type Mtb strain H37Rv on 0.2% propionate as the carbon source. OD, opticaldensity. (B and C) B12 concentration dependence of Mtb growth and its inhibition by itaconate. (D) Westernblot of Irg1 in Irg1 CRISPR knockdown (KD) RAW264.7 cells with or without LPS (10 ng/ml) stimulation for6 hours. Lrpprc, a mitochondrial protein, was used as the loading control. (E) Liquid chromatography–MS ofitaconate, I-CoA, and AdoCbl in control and Irg1 KD RAW264.7 cells with or without LPS stimulation for6 hours. Data represent means ± SD of three independent experiments. N.D., not detected.

RESEARCH | RESEARCH ARTICLEon June 3, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/

complex (mass of 237 kDa) and free MCM.The 173- and 211-kDa fractions have spectracorresponding to AdoCbl and the biradical,respectively (fig. S12, D and F). The 82-kDaATRfractions show that with M-CoA, very little co-factor is transferred toATR (fig. S12E), becausevery little inactive MCM forms. By contrast,whereas I-CoA completely inactivates MCM,very little cofactor is off-loaded to ATR (fig.S12G), indicating that the sustained presenceof the dAdo-I-CoA adduct on MCM impedescofactor repair.Hobbling of the repair system helps explain

why CLYBL deficiency is correlated with B12deficiency. In the absence of available MCMactive sites to off-load AdoCbl, ATR catalyzesan unusual sacrificial homolysis of the cobalt-carbon bond and sequesters cob(II)alamin, towhich it binds more tightly than AdoCbl (25).We propose that CLYBL deficiency increasesthe propensity of I-CoA–dependent MCM in-activation and thereby leads to AdoCbl deple-tion (5). How a change in the mitochondrialB12 pool (AdoCbl) is signaled to the cytoplasmand affects B12 levels systemically is, however,not known.

Itaconate inhibits vitamin B12-stimulated Mtbgrowth on propionate

To corroborate the in vitro evidence that itaco-nate inhibits Mtb MCM, we directly testedwhether exogenous itaconate can blunt B12-stimulated Mtb growth on propionate as thesole carbon source. As reported previously, vita-min B12 supplementation at concentrations aslow as 1 mg/ml stimulate growth of Mtb H37Rvon propionate (Fig. 4, A and B), which has beenattributed to the MCM-dependent pathwayfor propionate utilization (29). Growth stim-ulation was reduced in the presence of 1 mMitaconate and was completely inhibited at≥5 mM itaconate (Fig. 4C). Millimolar con-centrations of itaconate are endogenouslyproduced in activated macrophages (1), andIrg1-deficient mice, unlike controls, succumbearly to Mtb infection (30), suggesting thatsuch an inhibitory mechanism could be phys-iologically relevant.

MCM inhibition and AdoCbl depletionin macrophages require endogenousI-CoA synthesis

Mtb infection in mice elevates itaconate pro-duction in lungs (31), presumably throughIrg1 induction. To recapitulate the metabolicconsequence of endogenous itaconate pro-duction, we stimulated RAW264.7 cells withlipopolysaccharide (LPS), a potent activatorof Irg1 transcription and itaconate production(1). We previously found that LPS stimulationdepletes AdoCbl in macrophages (5). Using aCRISPR knockdown of Irg1 in RAW264.7 cells(Fig. 4D), we observed that AdoCbl depletionis dependent on Irg1, which is transcriptionally

upregulated in LPS-stimulated macrophages(1, 32). Treatment with LPS induced an increasein itaconate and I-CoA levels, which was sig-nificantly attenuated in Irg1 knockdown cells(Fig. 4E). Although LPS stimulation reducedAdoCbl to undetectable levels in control cells,it was not significantly changed in the Irg1knockdown cells (Fig. 4E). Together, these datasuggest that AdoCbl depletion and MCM in-hibition in macrophages is caused by endog-enous itaconate produced by Irg1 and inducedduring LPS stimulation.

Conclusions

AdoCbl is a radical initiator that generates the“working” dAdo• and “spectator” cob(II)alaminradical by homolytic cleavage of its cobalt-carbon bond. We found that I-CoA triggershomolytic cleavage of the cobalt-carbon bondin AdoCbl as in the normalMCM catalytic cycle,but proximity effects promote suicidal additionof dAdo• into its double bond. A chemically akin,albeit nonspecific,Michael additionmechanismhas been invoked to explain itaconate-inducedelectrophilic stress (10, 11). The combined ac-tion of itaconate and I-CoA onMtb propionatemetabolism would be predicted to result inincreased levels of toxic propionate/propionyl-CoAderived fromcholesterol-dependent growthof this pathogen in host phagosomes (29).Although the conditions are not known underwhich the Mtb pathway for de novo B12 bio-synthesis might be operative (33), Mtb canscavenge B12 from its host (34). Itaconate-induced B12 deficiency in host macrophagesthus might be a strategy for restricting path-ogen growth, in addition to targeting patho-gen enzymes involved in propionate metabolism(Fig. 1A); the relative importance of eachinhibitory arm is unknown, however.We spec-ulate that the CLYBL null background couldboost the efficacy of this pathogen containmentstrategy, explaining the prevalence of the nullgenotype in human populations (5). In thiscontext, it is noteworthy that the incidenceof active tuberculosis is reported to be mark-edly lower in patients with B12 deficiency dueto pernicious anemia (35).

REFERENCES AND NOTES

1. A. Michelucci et al., Proc. Natl. Acad. Sci. U.S.A. 110,7820–7825 (2013).

2. L. A. J. O’Neill, M. N. Artyomov, Nat. Rev. Immunol. 19, 273–281(2019).

3. R. Kumar, Int. J. Integr. Biol. 7, 69–72 (2009).

4. J. D. McKinney et al., Nature 406, 735–738 (2000).

5. H. Shen et al., Cell 171, 771–782.e11 (2017).

6. J. Sasikaran, M. Ziemski, P. K. Zadora, A. Fleig, I. A. Berg,Nat. Chem. Biol. 10, 371–377 (2014).

7. N. Grarup et al., PLOS Genet. 9, e1003530 (2013).

8. M. Lek et al., Nature 536, 285–291 (2016).

9. D. Padovani, R. Banerjee, Proc. Natl. Acad. Sci. U.S.A. 106,21567–21572 (2009).

10. E. L. Mills et al., Nature 556, 113–117 (2018).

11. M. Bambouskova et al., Nature 556, 501–504 (2018).

12. S. O. Mansoorabadi et al., Biochemistry 44, 3153–3158(2005).

13. H. Bothe et al., Biochemistry 37, 4105–4113 (1998).

14. W. Buckel et al., Eur. J. Inorg. Chem. 2006, 3622–3626(2006).

15. R. Banerjee, Biochemistry 40, 6191–6198 (2001).

16. H. Yang et al., J. Am. Chem. Soc. 141, 12139–12146(2019).

17. H. P. C. Hogenkamp, J. Biol. Chem. 238, 477–480(1963).

18. P. Jaruga, M. Dizdaroglu, DNA Repair 7, 1413–1425(2008).

19. F. Mancia et al., Structure 4, 339–350 (1996).

20. D. S. Froese et al., J. Biol. Chem. 285, 38204–38213(2010).

21. R. A. Kwiecien et al., J. Am. Chem. Soc. 128, 1287–1292(2006).

22. E. Chabrière et al., Science 294, 2559–2563 (2001).

23. K. Gruber, R. Reitzer, C. Kratky, Angew. Chem. Int. Ed. 40,3377–3380 (2001).

24. J. Masuda, N. Shibata, Y. Morimoto, T. Toraya, N. Yasuoka,Structure 8, 775–788 (2000).

25. G. C. Campanello et al., J. Am. Chem. Soc. 140, 13205–13208(2018).

26. D. Padovani, T. Labunska, B. A. Palfey, D. P. Ballou,R. Banerjee, Nat. Chem. Biol. 4, 194–196 (2008).

27. M. Ruetz et al., Cell Chem. Biol. 26, 960–969.e4 (2019).

28. D. Padovani, R. Banerjee, Biochemistry 45, 9300–9306(2006).

29. S. Savvi et al., J. Bacteriol. 190, 3886–3895 (2008).

30. S. Nair et al., J. Exp. Med. 215, 1035–1045 (2018).

31. J. H. Shin et al., J. Proteome Res. 10, 2238–2247(2011).

32. A. K. Jha et al., Immunity 42, 419–430 (2015).

33. K. Gopinath, A. Moosa, V. Mizrahi, D. F. Warner,Future Microbiol. 8, 1405–1418 (2013).

34. K. Gopinath et al., Open Biol. 3, 120175 (2013).

35. M. Barron, JAMA 100, 1590–1592 (1933).

ACKNOWLEDGMENTS

We acknowledge H. Sharma (University of Michigan) for hisassistance with crystallization and the GM/CA CAT at theAdvanced Light Source for beam time. We also acknowledgethe NIH Common Fund Metabolite Standards Synthesis Core(NHLBI contract no. HHSN268201300022C) for providing[13C]itaconate. Funding: This work was supported in part bygrants from the National Institutes of Health (RO1-DK45776 toR.B., 5 F32 GM113405 to G.C.C., K99-GM124296 to H.S.,R35GM122455 to V.K.M., R01-DK054514 to K.W., andU19AI107774 to E.J.R.) and start-up funds from the University ofMichigan (to M.K.). V.K.M. is an Investigator of the HowardHughes Medical Institute. Author contributions: M.R. andG.C.C. performed the majority of the biochemical studies,H.G. did the kinetic analysis of hMCM, and L.M. and M.R. clonedand expressed the Mtb genes used in this study. M.P. andM.K. were responsible for the crystallographic analyses,and K.W. was responsible for the EPR spectroscopic analysis.H.S. performed the macrophage experiments, and J.Z. andS.W. performed the Mtb growth experiments. M.R., G.C.C.,and R.B. wrote the manuscript. All authors were involved withdata analysis of experiments performed by them or in theirlaboratories and edited the manuscript. Competing interests:V.K.M. is a paid advisor to Janssen Pharmaceuticals and5AM Ventures and owns equity in Raze Therapeutics.Data and materials availability: All data are availablein the manuscript or supplementary materials. Thestructure factors and coordinates for Mtb MCM (6OXC)and Mtb MCM + I-CoA (6OXD) have been deposited in theProtein Data Bank.

SUPPLEMENTARY MATERIALS

science.sciencemag.org/content/366/6465/589/suppl/DC1Materials and MethodsFigs. S1 to S12Table S1References (36–59)

View/request a protocol for this paper from Bio-protocol.

19 May 2019; accepted 3 September 201910.1126/science.aay0934

Ruetz et al., Science 366, 589–593 (2019) 1 November 2019 5 of 5

RESEARCH | RESEARCH ARTICLEon June 3, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/content/366/6465/589/suppl/DC1https://en.bio-protocol.org/rap.aspx?eid=10.1126/science.aay0934http://science.sciencemag.org/

and repairItaconyl-CoA forms a stable biradical in methylmalonyl-CoA mutase and derails its activity

Wakabayashi, Junhao Zhu, Eric J. Rubin, Kurt Warncke, Vamsi K. Mootha, Markos Koutmos and Ruma BanerjeeMarkus Ruetz, Gregory C. Campanello, Meredith Purchal, Hongying Shen, Liam McDevitt, Harsha Gouda, Shoko

DOI: 10.1126/science.aay0934 (6465), 589-593.366Science

, this issue p. 589; see also p. 574Sciencehow macrophages resist Mtb infection.regeneration machinery. Itaconate blocks Mtb growth on propionate, and this inhibition mechanism may be relevant to

12species, which is incapable of completing the catalytic cycle and cannot be recycled by the endogenous coenzyme B. Itaconyl-CoA derails the normal radical reaction catalyzed by MCM, forming a long-lived, biradical12coenzyme B

enzyme methylmalonyl-CoA mutase (MCM), which uses the radical-generating cofactor adenosylcobalamin, or(see the Perspective by Boal). They found that the coenzyme A (CoA) derivative of itaconate can irreversibly inhibit the

(Mtb)Mycobacterium tuberculosissystems to inhibit propionate metabolism, a crucial metabolic pathway in pathogenic investigated how the immunometabolite itaconate might undermine these intricateet al.dedicated systems. Ruetz

Controlled radicals enable unusual enzymatic transformations, but radical generation and management requireItaconate brings metalloenzyme to a halt

ARTICLE TOOLS http://science.sciencemag.org/content/366/6465/589

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/10/30/366.6465.589.DC1

CONTENTRELATED http://science.sciencemag.org/content/sci/366/6465/574.full

REFERENCES

http://science.sciencemag.org/content/366/6465/589#BIBLThis article cites 58 articles, 8 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on June 3, 2021

http://science.sciencemag.org/

Dow

nloaded from