Total Synthesis, Structure, and Biological Activity of Adenosylrhodibalamin, the Non‐Natural Rhodium Homologue of Coenzyme B12

Abstract B12 is unique among the vitamins as it is biosynthesized only by certain prokaryotes. The complexity of its synthesis relates to its distinctive cobalt corrin structure, which is essential for B12 biochemistry and renders coenzyme B12 (AdoCbl) so intriguingly suitable for enzymatic radical reactions. However, why is cobalt so fit for its role in B12‐dependent enzymes? To address this question, we considered the substitution of cobalt in AdoCbl with rhodium to generate the rhodium analogue 5′‐deoxy‐5′‐adenosylrhodibalamin (AdoRbl). AdoRbl was prepared by de novo total synthesis involving both biological and chemical steps. AdoRbl was found to be inactive in vivo in microbial bioassays for methionine synthase and acted as an in vitro inhibitor of an AdoCbl‐dependent diol dehydratase. Solution NMR studies of AdoRbl revealed a structure similar to that of AdoCbl. However, the crystal structure of AdoRbl revealed a conspicuously better fit of the corrin ligand for RhIII than for CoIII, challenging the current views concerning the evolution of corrins.

The biochemical activity of the biological forms of B 12 is based on the pivotal role played by the cobalt center bound by the corrin ring. [1] However,w hy is cobalt, rather than any other metal, so suited to its role in B 12 ? [1b] This old question has posed aformidable challenge. [1,2] Interestingly,cobalt was given its name because German miners found it in ores contaminated with arsenic, and believed it was added malevolently by "Kobolds", or goblins.T oa ddress the "cobalt question", we considered the replacement of cobalt by its heavier Group IX homologue rhodium. Thes pecific suitability of coenzyme B 12 (5'-deoxy-5'-adenosylcobalamin, AdoCbl;F igure 1) as ac atalytic radical source by enzymecontrolled homolytic cleavage of its Co À Cb ond [3] suggested that its rhodium homologue,5 '-deoxy-5'-adenosyl-rhodibalamin (AdoRbl), would be ap articularly interesting target. AdoRbl was first prepared in the 1970s via metal-free hydrogenobalamin, which was isolated in low yields from cultures of Chromatium vinosum grown in cobalt-free media, but incompletely characterized. [2a] Unfortunately,v arious alternative strategies to generate metal analogues of the natural corrinoids by removal of the Co center of vitamin B 12 derivatives have not been successful (see e.g. Ref. [4]); therefore,anovel approach for its preparation was required. Herein, we describe ac oncise total synthesis of AdoRbl through as trategical combination of biological and chemical means,a nd report its structural and basic biological properties.I ndeed, as described below,b ya sking "Why not rhodium?", we have addressed ar elated fundamental question concerning the evolutionary selection and adaptation of corrins.
Complementary chemical and biological methods were developed for the synthesis of 5'-deoxy-5'adenosylrhodibalamin (AdoRbl;F igure 2). Initially,h ydrogenobyrinic acid a,cdiamide (Hbad) was synthesized de novo and in vivo using an engineered E. coli strain containing the ten genes (cobA-I-G-J-M-F-K-L-H-B)that encode the enzymes for the biosynthesis of cobalamin from the endogenous biosynthetic intermediate uroporphyrinogen III. [5] From around 30 Lo fc ulture,8 8.2 mg of Hbad were obtained. Hbad was converted into rhodibyrinic acid a,c-diamide by chemical insertion of Rh I . [6] Orange-red dicyanorhodi(III)byrinic acid a,c-diamide (CN 2 -Rhbad) [6b] was obtained in 75 %yield, but unfortunately proved to be resistant to refunctionalization into adenosylrhodi(III)byrinic acid a,c-diamide (AdoRhbad). However,A doRhbad was synthetically accessible when rhodibyrinic acid a,c-diamide was isolated as the dichlorosubstituted Rh III corrinoid (DCRhbad), which was characterized by UV/Vis spectroscopy and mass spectrometry.Reduction of DCRhbad with sodium borohydride in deoxygenated solution led to the light yellow Rh I corrinoid. Tr eatment of the latter with 5'-iodo-5'-deoxyadenosine (0 8 8C!RT)g ave an orange reaction mixture from which AdoRhbad could be isolated in 75 %y ield. Them olecular formula of AdoRhbad was confirmed by ESI mass spectrometry (m/z 1230.3 [M] + ). Its UV/Vis spectrum was similar to those of the dichloro and dicyano Rh III analogues (see the Supporting Information, Figure S1). Them etal-bound methylene group of the 5'-deoxyadenosyl moiety of AdoRhbad gave rise to two characteristic multiplets at high field in the 1 HN MR spectrum, which were assigned to the diastereotopic protons of this CH 2 group ( Figure S2). Both of these resonances of AdoRhbad showed ad iagnostic 1.7 Hz coupling with 103 Rh (I = 1/2). With the Rh analogue of adenosylcobyrinic acid a,c-diamide in hand, the cobalamin (Cbl) biosynthetic pathway was again employed, namely in the form of CobQ, [7] to specifically amidate four of the remaining five side-chain carboxyl groups, thus converting AdoRhbad into adenosylrhodi(III)byric acid (AdoRhby;s ee Figure 2). Indeed, AdoRhbad served remarkably well as ap seudosubstrate for CobQ and furnished AdoRhby in 92 %y ield. Ther egiospecific fourfold amidation of the peripheral side chains was confirmed by ESI mass spectrometry and 1 H NMR spectroscopy.
5'-Deoxy-5'-adenosylrhodibalamin (AdoRbl), the Rh analogue of coenzyme B 12 (AdoCbl), was prepared by chemical conjugation of AdoRhby with the B 12 nucleotide moiety. [1a, 8] This was achieved by activation of AdoRhby with the carbodiimide reagent 1ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) in the presence of the B 12 nucleotide,f urnishing the orange-red AdoRbl in 79 %y ield. TheU V/Vis spectrum of isolated AdoRbl showed absorption maxima at l = 512, 491, and 350 nm ( Figure S1), as reported previously. [2] Thespectrum of AdoRbl is surprisingly similar to that of cyanocobalamin (vitamin B 12 ), but differs significantly from that of coenzyme B 12 (AdoCbl). As also noted earlier, [3] AdoRbl does not decompose when exposed to daylight in aerated solutions,in contrast to the photosensitive AdoCbl. [9] The5 00 MHz 1 HN MR spectrum of AdoRbl in D 2 O contained resonances for all carbon-bound Hatoms,i ncluding the characteristic doublet-and triplet-like signals at high field for the Rh-bound CH 2 group of the Ado ligand ( Figure 3). This latter finding is in striking contrast to the spectrum reported in the earlier work on AdoRbl, where such signals were not recorded. [2] Thehigh-field resonances of the  Tables S1-S3), indicating similar conformations of the corrin ligand and both axially bound groups for AdoRbl and AdoCbl. Thes tructures of AdoRbl and coenzyme B 12 (AdoCbl) [10] in aqueous solution are therefore very similar.
AdoRbl crystallized from as olution of water/acetonitrile as dark red monoclinic prisms (space group C2, No.5). Awell resolved X-ray crystal structure was obtained for AdoRbl, which is the first of any metal analogue of the Cbl series.I t confirmed the NMR-derived chemical constitution of AdoRbl and showed basic structural features similar to those of AdoCbl [11] (Figure 4a nd Figure S6). As in AdoCbl, the Ado group of AdoRbl is positioned roughly above ring C of the corrin ring. However,i nA doRbl, it has been rotated counterclockwise by about 248 8 ( Figure S7), and is held in position by an unprecedented hydrogen-bonded dimerization interface involving the Ado moiety and both of the side chains of ring B( Figure S9). Thus the Ado group is closer to ring B in AdoRbl than in AdoCbl, [11] and ring Badopts aconformation unparalleled in Cbls (see below).
Thebiological activity of AdoRbl as an AdoCbl analogue was initially investigated in the form of ab ioassay [12] by monitoring the activity of methionine synthase (MetH), and subsequently by adirect enzymatic assay with diol (propanediol) dehydratase (PduCDE). ForMetH, which utilizes aB 12 cofactor, we employed ap late-based microbiological bioassay that uses a Salmonella enterica cbiB metE reporter strain that is reliant upon exogenous cobalamin (Cbl) for its MetH when grown on minimal media. The size of the growth circles observed on these plates is related logarithmically to the quantity of applied Cbl ( Figure 5).
Addition of AdoRbl alone to the bioassay plates did not promote any growth. However,w hen AdoRbl was applied in close proximity to an equivalent amount of vitamin B 12 (CNCbl), agrowth inhibition zone around the AdoRbl application point was observed. Increasing the concentration of AdoRbl resulted in greater inhibition ( Figure 5). Unexpectedly,amixture of CNCbl and AdoRbl resulted in al arger but more diffuse growth circle.T hese observed growth patterns indicate that 1) AdoRbl is not converted into an active cofactor form for methionine synthase,a nd that 2) AdoRbl acts as an inhibitor for Cbl either by preventing the uptake of Cbl from the medium or by competing for the active site of methionine synthase.I ndeed, the larger growth circles that were observed when CNCbl was mixed with an excess of AdoRbl can be explained best by the ability of this analogue to actively interact with the regulation of Cbl uptake through aB 12 riboswitch. [13] In E. coli and S. enterica,the btuB riboswitch acts as af eedback control mechanism, with AdoCbl as the preferred ligand, [13,14] to switch off the production of the outer-membrane B 12 transporter.T he increased growth circles on the bioassay plates are consistent with AdoRbl reducing the level of Cbl uptake.
Thee ffect of AdoRbl on the activity of AdoCbl-dependent enzymes was investigated by studying the Citrobacter freundii 1,2-propanediol dehydratase (Figures S10-S12). The kinetic constants for the reaction catalyzed by purified 1,2propanediol dehydratase were determined by non-linear regression. Thee nzyme was found to be inactive with AdoRbl as apseudo-coenzyme.H owever,int he presence of AdoCbl, the enzyme was active,with a K m value of 3.0 mm for AdoCbl and k cat = 358 s À1 (based on an a 2 b 2 g 2 quaternary structure). Both AdoRbl and vitamin B 12 were found to be  competitive inhibitors of the enzyme,with K i values of 6.9 mm for AdoRbl and 2.5 mm for vitamin B 12 .These results confirm that AdoRbl acts as an inhibitory analogue of AdoCbl and is unable to catalyze the propanediol dehydratase reaction.
Theb iological roles of cobalt [15] and the functional forms of B 12 [16] appear to be largely interdependent and remarkably exclusive.T hus the question of why cobalt, rather than any other transition metal, is found in B 12 has spurred interest in developing metal analogues of B 12 . [6] In this respect, the Group IX metal rhodium represents ap rime substitute.W e thus developed as trategy for the total synthesis of 5'-deoxy-5'-adenosylrhodibalamin (AdoRbl), which is based on aconcise sequence of biological and chemical steps.S tructural studies with AdoRbl in solution (by NMR) and in the crystal confirmed the expected structural similarity to AdoCbl. However,t he crystal structure of AdoRbl also revealed some remarkable consequences when Co III is replaced with alarger Rh III ion in AdoRbl. As expected, all six bonds to the metal center of AdoRbl were longer than those in AdoCbl, which is consistent with the 0.06 l arger covalent radius of Rh. [17] In fact, the four equatorial bonds were found to be elongated by 0.082(5) a nd the axial bonds by only 0.035-(5) c ompared to those in AdoCbl ( Figure 6). Furthermore,the flatter corrin ligand of AdoRbl displays ar ecord small fold angle of 5.9(2)8 8 (13.38 8 in AdoCbl;s ee Figures 7, S8, and S9). [18] Ring BofAdoRbl exhibits astriking reversed conformational twist when compared to the structures of AdoCbl and other natural corrinoids.R ing Bo f AdoRbl is also flattened, and its acetamide and propionamide substituents are both in ap seudo-equatorial position. In contrast, the NMR solution structure indicated ar ing B conformation in AdoRbl matching that in AdoCbl, [10] and revealed no sign of the unprecedented "reversed" twist. Thus ring BofAdoRbl is,infact, flexible and undergoes conformational inversion in the crystal.
Hence,acomparison of the AdoRbl and AdoCbl structures reveals that, counterintuitively,t he larger Rh III ion fits better into the corrin ligand than the biologically relevant Co III ion. Eschenmoser and Kratky have analyzed the fundamental structural consequences of am utual misfit between small coordinated metal ions (e.g.,low-spin Ni II )and the coordination hole of tetrapyrrolic macrocycles. [1a, 19] They suggested that contraction of the macrocycle leads to nonplanar ("saddle-shaped") porphyrins and ac orrelated conformational change of the four pyrrolic rings.S imilar conformational effects of the mutual misfit of the size of the metal ion and the coordination hole have been observed for ar ange of porphyrinoid metal complexes [20] and have been   recognized to be an important factor in modifying the biological activity of metal porphyrinoids. [21] Thec rystal structure of AdoRbl indicates that the coordination hole of the corrin ligand is slightly too large for the coordination of Co III ions.I nf act, in natural Co III corrinoids,asignificant corrin fold (13.38 8 in AdoCbl) is also apparent, as are notable twists of all four pyrrolic rings [22] ( Figure S8). These twists are most prominent in ring B, where ac onsistent conformational twist "in phase" with the nonplanar corrin macrocycle is observed. [1a, 22a] In contrast, the crystallographic studies with AdoRbl suggest as ignificant conformational relaxation of the corrin macrocycle when adapting its structure to the coordination requirements of Rh III . Nature has evolved the unique "constitutional ring contraction" of the corrin ligand [1a, 23] to reduce its hole size and to accommodate cobalt. [24] However,a sdiscussed here, the structural data are consistent with an additional conformational adaptation of the corrin ligand to meet the effective size of the coordinated Co III ions.H ence,t he observed better fit of Rh III over Co III suggests that the corrin ligand of cobalamin may not primarily be targeted by Nature to Co III .A sr hodium is not considered to be an element that is essential for life on Earth, [15] the interaction of the corrin ligand with Co,rather than Rh, ions deserves close attention, including reduced Co II and Co I forms.AdoCbl and cob(II)alamin feature very similar cobalt corrin structures, [25] but crystallographic data of aCo I corrin are not yet available (see,f or example,R ef.[16a]). As light expansion of the coordination hole of the corrin ligand has been calculated to assist the reduction of Co III and Co II corrins. [26] It is thus tempting to suggest that corrins may display ap articular fit and stabilization for the polarizable Co I ions,the action center of the enigmatic "supernucleophilic" Co I corrins. [1b, 16a,b,27] By analogy with the idea of enzymes evolving to stabilize at ransition state to lower the activation energy of the reaction, [28] the proposed ability of stabilizing the Co I state over Co III and Co II forms would be ac rucial aspect of the corrin ligand in enzyme reactions with Co I corrin intermediates, [26,27,29] which are difficult to generate in ab iological environment. [14,29] This property would have allowed the selection of B 12 to be fine-tuned for its role as an essential organometallic catalyst for the prebiotic chemistry of life, [30] in line with the proposed antiquity of cobalt corrins as ancient cofactors. [1a] Themolecular recipe for the biosynthesis of coenzyme B 12 (AdoCbl) is confined to the genomes of only certain prokaryotes. [31] By combining it with an engineered E. coli strain, ac oncise biological/chemical synthesis pathway to AdoRbl became available.A doRbl was characterized as as tructural, but not functional, mimic of the B 12 coenzyme AdoCbl. Thec oenzyme inactivity of the largely isostructural Rh analogue of coenzyme B 12 ,i nc ombination with the inhibitory action of AdoRbl, suggests inefficient Rh À Cbond homolysis of the enzyme-bound AdoRbl. Thed etermination [32] of the strength of the Rh À Cb ond in AdoRbl will provide an experimental test for this conclusion.
Having re-addressed the fundamental question of "Why cobalt?", [1] perhaps we should now ask:"Whynot rhodium or another metal?" Metal analogues of the cobalamins (metbalamins) are believed to be inactive as cofactors,w hich is consistent with our studies on AdoRbl. Indeed, some metbalamins have been shown to inhibit bacterial growth. [6] Suitably structured metbalamins may thus represent effective B 12 antimetabolites or "antivitamins B 12 ", [33] which are of growing interest in view of recent detailed structural studies concerning remarkable "novel" biological functions of Cbls. [34] Our combined biological and chemical synthesis approach to the "rhodium problem" has opened anew entry to metbalamins and other metallocorrinoids,a ne xciting though poorly explored territory in the multifaceted B 12 field.

Experimental Section
See the Supporting Informationf or materials, instruments,s trains used, construction of plasmids,d etails of synthetic and enzymatic procedures,spectroscopy,a nd X-ray crystallography.
X-ray crystallography:C CDC 1450631 (AdoRbl) containst he supplementary crystallographic data for this paper.T hese data are provided free of charge by TheC ambridgeC rystallographic Data Centre.