Synthesis of 2-oxoglutarate derivatives and their evaluation as cosubstrates and inhibitors of human aspartate/asparagine-β-hydroxylase

2-Oxoglutarate (2OG) is involved in biological processes including oxidations catalyzed by 2OG oxygenases for which it is a cosubstrate. Eukaryotic 2OG oxygenases have roles in collagen biosynthesis, lipid metabolism, DNA/RNA modification, transcriptional regulation, and the hypoxic response. Aspartate/asparagine-β-hydroxylase (AspH) is a human 2OG oxygenase catalyzing post-translational hydroxylation of Asp/Asn-residues in epidermal growth factor-like domains (EGFDs) in the endoplasmic reticulum. AspH is of chemical interest, because its Fe(ii) cofactor is complexed by two rather than the typical three residues. AspH is upregulated in hypoxia and is a prognostic marker on the surface of cancer cells. We describe studies on how derivatives of its natural 2OG cosubstrate modulate AspH activity. An efficient synthesis of C3- and/or C4-substituted 2OG derivatives, proceeding via cyanosulfur ylid intermediates, is reported. Mass spectrometry-based AspH assays with >30 2OG derivatives reveal that some efficiently inhibit AspH via competing with 2OG as evidenced by crystallographic and solution analyses. Other 2OG derivatives can substitute for 2OG enabling substrate hydroxylation. The results show that subtle changes, e.g. methyl- to ethyl-substitution, can significantly alter the balance between catalysis and inhibition. 3-Methyl-2OG, a natural product present in human nutrition, was the most efficient alternative cosubstrate identified; crystallographic analyses reveal the binding mode of (R)-3-methyl-2OG and other 2OG derivatives to AspH and inform on the balance between turnover and inhibition. The results will enable the use of 2OG derivatives as mechanistic probes for other 2OG utilizing enzymes and suggest 2-oxoacids other than 2OG may be employed by some 2OG oxygenases in vivo.

Introduction 2-Oxoglutarate (2OG, a-ketoglutarate; 1, Fig. 1a) is an integral metabolite in most of biology including prokaryotes, archaea, and eukaryotes; 1 2OG is crucially involved in cellular energy homeostasis and small-molecule metabolism 1 and can act as a signaling molecule linking nitrogen and carbon metabolism. 2 2OG is an intermediate of the tricarboxylic acid (TCA) cycle where it is produced by isocitrate dehydrogenase (IDH)catalyzed decarboxylation of D-isocitrate; 2OG is converted to succinyl-CoA and CO 2 by the 2OG dehydrogenase complex. Reductions in cellular 2OG coupled with the formation of (R)-2hydroxyglutarate, which occur as a result of IDH mutations, 3 are associated with changes in epigenetic regulation (e.g. DNA and histone methylation status) and certain types of cancer. 3b,4 These effects are proposed to be mediated, at least in part, by modulation of the activities of enzymes that rely on 2OG as a (co-)substrate, other than the 2OG dehydrogenase complex. Such enzymes include aminotransferases (e.g. branched-chain aminotransferases, BCATs), 5 which convert 2OG to glutamate, and 2OG dependent oxygenases. 6 The latter couple the conversion of 2OG and O 2 to succinate and CO 2 with substrate hydroxylation or demethylation via hydroxylation. 7 There are approximately 60-70 assigned human 2OG oxygenases, all studied members of which likely employ Fe(II) as a cofactor and which have diverse roles, ranging from DNA/RNA modication and damage repair, 8 histone/chromatin modication, 9 lipid metabolism, 10 post-translational modication of proteins with important functions in the extracellular matrix, 7,11 to hypoxia sensing. 12 The human 2OG oxygenase aspartate/asparagine-bhydroxylase (AspH, BAH, HAAH) 13 is highly unusual amongst human 2OG dependent hydroxylases, because its Fe(II) cofactor is complexed by only two residues (His679 and His725) rather than by the typical triad of ligands (HXD/E/H) found in other human 2OG oxygenases (Fig. 1b). 7, 14 AspH catalyzes the post-translational hydroxylation of specic Aspand Asn-residues in epidermal growth factor-like domains (EGFDs) of its substrate proteins in the endoplasmic reticulum ( Fig. 1a). 15 AspH is of signicant interest from a cancer research perspective, because its levels are upregulated in invasive cancers (e.g. hepatocellular carcinoma 16 and pancreatic cancer 17 ) and it is translocated to the cell surface where it can be used as a diagnostic and prognostic marker. 18 Mouse models 19 and heritable genetic diseases associated with mutations likely effecting AspH catalysis (e.g. Traboulsi syndrome) 20 suggest that the Notch signaling pathway may be involved in transmitting the effect of AspH on cancer invasiveness. AspH levels are regulated by hypoxia which is a characteristic of many tumor cells. 21 Thus, AspH appears to be an attractive medicinal chemistry and diagnostic target for certain types of cancer.
Recent studies have provided crystallographic and solutionbased evidence that AspH accepts EGFD substrates with an unusual non-canonical disulde connectivity (i.e. Cys 1-2, 3-4, 5-6) rather than the well-characterized canonical disulde connectivity (i.e. Cys 1-3, 2-4, 5-6; ESI Fig. S1 †). 22 Highthroughput MS assays were established to monitor the catalytic activity of AspH using stable thioether-linked cyclic peptide substrate analogues mimicking the central non-canonical macrocyclic Cys 3-4 EGFD disulde (ESI Fig. S1 †). 23 Kinetic studies have revealed that AspH is sensitive towards subtle changes in oxygen availability and thus it is a candidate oxygenase for involvement in hypoxia sensing. 23 Previous studies have revealed that differences in the cosubstrate binding sites of 2OG oxygenases can be exploited for stereoselective selective inhibition employing 2OG analogues, e.g. N-oxalylamino acids. 24 Analysis of reported AspH crystal structures wherein 2OG is replaced by a close 2OG analogue, i.e. N-oxalylglycine (NOG, 3; Fig. 1c), 22 suggest that the AspH active site is sufficiently spacious to Fig. 1 The AspH active site and stoichiometry of its reaction. (a) AspH catalyzes the post-translational hydroxylation of Asn-and Asp-residues in epidermal growth factor (EGF)-like domains; (b) analysis of an AspH:substrate (human Factor X, hFX) crystal structure (PDB ID: 5JQY) 22 reveals that two AspH residues (His679 and His725) coordinate the active site metal rather than the typical three residues found in other human 2OG dependent hydroxylases. In the crystallographic analyses, Mn substitutes for Fe and NOG (3) for 2OG (1); (c) N-oxalylglycine (NOG, 3). accommodate substituents at the 2OG C3-and/or C4position, in part due to its unusual Fe(II)-binding geometry (Fig. 1b). We were therefore interested to test this by exploring how a diverse set of 2OG derivatives interact with AspH.
To our knowledge, only a limited number of studies describing how 2OG derivatives bearing substituents at the C3and/or C4-position modulate the activities of 2OG oxygenases (i.e. human JmjC histone N 3 -methyl lysine demethylase 4A, KDM4A; 25 human factor inhibiting the hypoxia-inducible transcription factor HIF-a, FIH; 24 and rat g-butyrobetaine dioxygenase, BBOX 26 ) are reported. In part, this likely reects a lack of a simple synthetic method to access these types of 2OG derivatives. Some prior syntheses of C3/C4-substituted 2OG derivatives have relied inter alia on alkylation (Fig. 2a) 25 and Michael reactions (Fig. 2b) 27 to access key synthetic intermediates; the corresponding 2OG derivatives (6) were obtained aer saponi-cation. Other approaches rely on oxidation reactions using ozone 28 or sodium periodate 29 as oxidants to convert Michael acceptors (10) into 2OG derivatives (Fig. 2c). The described syntheses are frequently associated with limited scalability, low overall chemical yields, and/or narrow substrate scopes due to harsh reaction conditions requiring, for example, the use of strong bases and acids, 25,30 high pressure, 27 or strong oxidants. 28a-c,29 Cyanophosphorous ylids are reported as valuable intermediates for the synthesis of a-keto acids, 31 however, only one example of a cyanophosphorous ylid (i.e. 11) being converted into a 2OG derivative is reported (Fig. 2c), 32 possibly reecting limitations associated with the conversion of cyanophosphorous ylids into a-keto acids which requires strong oxidants. [31][32][33] The use of cyanophosphorous ylids has been largely superseded by the corresponding cyanosulfur ylids, which can be oxidized under milder conditions. 33a, 34 We thus envisaged that cyanosulfur ylids could be used for the synthesis of C3/C4-substituted 2OG derivatives.
Here we report the use of cyanosulfur ylids as intermediates that enable the facile synthesis of multiple 2OG derivatives bearing a diverse set of substituents at the C3-and/or C4positions. The synthetic 2OG derivatives were used to modulate the activity of recombinant human AspH. Kinetic and crystallographic studies were employed to elucidate the mechanisms by which the 2OG derivatives modulate AspH activity and to garner information of the active site requirements of AspH. The results reveal an unexpectedly diverse set of 2OG derivatives can bind at the AspH active site and that subtle differences in the 2OG substitution pattern can cause signicant disturbances in the balance between productive catalysis and inhibition.

Synthesis of 2OG derivatives
A versatile and scalable synthetic route to access 2OG derivatives was developed employing cyanosulfur ylids as key intermediates (Scheme 1). The route employs mono-methyl dicarboxylic acid half-esters 13 as starting materials which were either commercially available or synthesized by established reactions, i.e. nucleophilic openings of the requisite symmetric cyclic anhydrides, formylation reactions of aryl iodides, 35 Heck couplings 36 of aryl iodides with orthogonally protected itaconates, 37 or Horner-Wadsworth-Emmons (HWE) 38 reactions (ESI Fig. S2 †). In our work, racemic mixtures of mono-methyl dicarboxylic acid half-esters 13 bearing stereogenic carbon atoms at the 2OG C3-or C4-equivalent position were employed.
Cyanosulfur ylids 15, which are the key intermediates in our strategy, were obtained by reaction of mono-methyl dicarboxylic acid half-esters 13 with the reported tetrahydrothiophene bromide salt 14 33a in yields ranging from 11 to 95%. T3P 39 was chosen as the coupling reagent because it is suited for use with sterically hindered carboxylic acids, including those bearing substituents at the carboxylate aposition. For some substrates, T3P-derived byproducts interfered with the purication process; however, these were completely removed aer the subsequent reaction by chromatography.
The cyanosulfur ylids 15 were converted into the corresponding dimethyl dicarboxylic acid esters 12 using oxone 33a as a mild oxidation reagent in methanol, in part to avoid ester exchange (Scheme 1). The dimethyl esters 12 were obtained in high purity aer column chromatography in yields ranging from 39 to 98%. Lithium hydroxide-mediated saponication of dimethyl dicarboxylic acid esters 12 afforded the desired 2OG derivatives 6 (Scheme 1). The 2OG derivatives were obtained in sufficient purity aer removal of excess base by acidic ion exchange chromatography yielding salt-free dicarboxylic acids 6 (ESI †), which are suitable for performing in vitro biochemical experiments with AspH. The 2OG derivatives and their synthetic precursors were stable when stored at À20 C for more than six months.

Scope of the synthesis
Following its development, the synthetic route was used to synthesize a diverse set of 2OG derivatives bearing aliphatic substituents at the 2OG C3-and/or C4-positions (Table 1, entries 1-21). The aliphatic substituents varied in both length and steric bulk of the carbon chain. Furthermore, 2OG derivatives were synthesized in which the C3/C4 ethylene unit of 2OG The synthesis of 2OG derivatives bearing acid-labile or some oxidation-prone moieties was challenging. For example, during the cyanosulfur ylid oxidation reaction, both ketal and silyl ether alcohol protecting groups were cleaved and nitrogencontaining heteroaromatic rings (e.g. pyridines) formed Noxides. Nonetheless, the oxidation conditions were sufficiently mild to tolerate alkenes (36; Table 1, entry 21) and substituted thiophenes (37 and 38; Table 1, entries 22 and 23), which constitutes an advantage compared to many prior syntheses of 2OG derivatives (Fig. 2).

AspH inhibition studies
We then evaluated the potential of the synthesized racemic 2OG derivatives to inhibit AspH by measuring AspH substrate depletion and product formation (i.e. by monitoring a +16 Da mass shi) using an established solid phase extraction coupled to mass spectrometry (SPE-MS) AspH inhibition assay. 41 Half maximum inhibitory concentrations (IC 50 -values) for all the synthetic 2OG derivatives prepared were determined (Table 1).
Interestingly, 4-benzyl-2OG (32) inhibits AspH signicantly more efficiently (IC 50 $ 0.4 mM; Table 1, entry 17) than its Noxalyl analogue N-oxalyl-D-phenylalanine (NOFD, IC 50 $ 15.5 mM), 41a which is a reported inhibitor of human FIH. 24 An opposite trend was observed for FIH, for which NOFD was a substantially more efficient inhibitor than 32, 24 revealing the context dependent effect of the same 2OG substitutions. The structures of these two inhibitors are very similar: the C3 methylene-unit of 4-benzyl-2OG (32) is substituted for an NHgroup in NOFD; 24 however, 32 was prepared as a racemic mixture whereas NOFD was used in enantiopure D-form. To investigate the effect which the NH-group present in NOG and NOFD imposes on AspH inhibition, while excluding possible interference from the stereochemistry of the inhibitors (including with respect of C3-racemisation of the chiral 2oxoacids), a derivative of 4,4-dimethyl-2OG (34) was thus synthesized in which the C3 methylene-unit was replaced with an NH-group (N-oxalyl-a-methylalanine, 49; ESI Fig To dene whether the mechanism by which the 2OG derivatives inhibit AspH involves competition with 2OG for binding to the AspH active site, the effect of altered 2OG concentrations on the IC 50 -values of AspH was investigated. The IC 50 -values of four potent AspH inhibitors (i.e. 17, 29, 33, and 34) were determined at 2OG assay concentrations of 3, 200, 400, and 600 mM ( Fig. 3a; ESI Table S1 †). The results reveal an ascending linear dependence of the IC 50 -values on the 2OG assay concentration suggesting that the 2OG derivatives inhibit AspH by competing with 2OG for binding to the active site. This is in agreement with a Hill coefficient 42 analysis of the AspH inhibition curves which indicates that the 2OG derivatives did not inhibit AspH by forming colloidal aggregates; the Hill coefficients are in the range of the expected 'ideal' value À1 (Fig. 3b). 43 Neither the position nor the size of the C3/C4-substituent of the 2OG derivatives had a detrimental effect on the linear dependence of their AspH IC 50 -values on the 2OG concentration (Fig. 3a). Efficient inhibition of AspH at higher 2OG assay concentrations was observed for 2OG derivatives 33 and 34 (IC 50 $ 5.6 and 18.7 mM at 0.6 mM 2OG assay concentration, respectively; ESI Table S1, † entries 3 and 4).
Increasing the steric bulk of the 2OG derivative, while maintaining the relative arrangement of the two carboxylate groups, decreases the catalytic efficiency of the cosubstrate analogue, as revealed by the comparison of the phenyl-ring 2OG derivative 41 ($80%) with its bridged bicyclo[2.2.2]octane analogue 47 ($8%). Derivatives of 41 bearing substituents ortho to the ketone (e.g. as in 42, $45%) seemed to be more efficient AspH cosubstrates than those bearing substituents meta to the ketone (e.g. as in 43, $8%). Increasing the size of the substituents ortho to the ketone of 2OG derivative 41 (i.e. ortho-F, 42; ortho-Br, 44; ortho methyl 46) results in a noticeable decrease in the efficiency to replace 2OG as an AspH cosubstrate (i.e. 42: $45%; 44: $20%; 46: $5%); all derivatives of 41 were signicantly less efficient with respect to the parent compound ($80%).
Kinetic analyses of the two most efficient 2OG substitute AspH cosubstrates identied (i.e. 16 and 41) were performed. Maximum velocities (v app max ) and Michaelis constants (K app m ) were determined employing SPE-MS turnover assays, albeit under modied conditions than previously reported for 2OG as a cosubstrate (Fig. 4d-f), 23 as L-ascorbic acid (LAA), which is commonly added to 2OG oxygenase assays, was included in the assay buffer. The presence of LAA affected the kinetic parameters for 2OG when compared to the previous parameters ( Table  2, entry 1). The v app max (2OG) did not change signicantly, being $16.8 Â 10 À3 mM s À1 in the absence of LAA 23 and $15.0 Â 10 À3 mM s À1 in the presence of LAA. The K app m (2OG)-value in the presence of LAA was approximately double ($1.3 mM) compared to that in the absence of LAA ($0.6 mM). 23 However, this K app m (2OG)-value is still in the range of those values reported for most other human 2OG oxygenases including for the HIFa prolyl hydroxylases and FIH (1-25 mM) 44 and bovine AspH ($5 mM). 45 Fig. 3 Inhibition of AspH by 2OG derivatives. (a) The AspH IC 50 -values for the 2OG derivatives 17 (black diamonds), 29 (red squares), 33 (orange circles), and 34 (green triangles) depend on the 2OG concentration. AspH inhibition assays were performed in triplicate as described in the ESI † using 3, 200, 400, and 600 mM 2OG. IC 50 -values are summarized in ESI Table S1; † (b) representative dose-response curves used to determine IC 50 -values for the 2OG derivatives 17 (black diamonds), 29 (red squares), 33 (orange circles), and 34 (green triangles) at a 2OG assay concentration of 3 mM 2OG. Three dose-response curves each composed of technical duplicates were independently determined using SPE-MS AspH inhibition assays, performed as described in the ESI † and manifest high Z 0 -factors 40 and signal-to-noise ratios (ESI Fig. S3 †).
Compared to the K app m -value of AspH for 2OG ($1.3 mM; Table  2, entry 1), the K app m -value for 3-methyl-2OG (16) was about ve times lower ($0.3 mM; Table 2, entry 2), indicating a higher affinity of AspH for 16 compared to 2OG. By contrast, the K app mvalue of AspH for the 2OG derivative 41 was $47 times higher ($62 mM; Table 2, entry 3) than that for 2OG, indicating much a Mean of three independent runs (n ¼ 3; mean AE SD). b AspH assays were performed as described in the ESI using 0.1 mM His 6 -AspH 315-758 and 2.0 mM hFX-CP 101-119 (ESI Fig. S1d) as a substrate. AspH assays were performed as described in the ESI, † data are shown as the mean of three independent runs (n ¼ 3; mean AE standard deviation, SD). The results are summarized in Table 2 and the peptide hydroxylation rates are shown in ESI Fig. S6. † less efficient binding. All three K app m -values range signicantly below reported 2OG concentrations in healthy cells (up to >1 mM), 46 which however vary substantially, but are in the approximate range of reported physiological 2OG levels in human plasma (9-12 mM 2OG). 47 Based on the determined concentration of active AspH (90.8 AE 13.7 nM for an original estimated AspH assay concentration of 100 nM AspH), 23 turnover numbers (catalytic constants, k cat ) and specicity constants (k cat /K m ) were calculated for the 2OG derivatives ( Table 2). Comparison of the k cat -values reveals that the impact of the 2OG derivatives on k cat -values is notably smaller than on K app m -values which indicates that efficient AspHcatalyzed substrate hydroxylation is still feasible with the 2OG derivatives (Table 2): for AspH, the k cat -value for 16 ($0.08 s À1 ; Table 2, entry 2) was about half the k cat for 2OG ($0.17 s À1 ; Table 2, entry 1), whereas the k cat for 41 ($0.04 s À1 ; Table 2, entry 3) was about a quarter of that for 2OG. Comparison of the k cat /K m -values clearly reveals the potential of 2OG derivatives to substitute for 2OG itself, with racemic 16 being of similar efficiency to 2OG. Although 41 is a much less efficient substrate, its conversion reveals the potential for unexpected cosubstrate utilization by 2OG oxygenases.

Crystallography
The AspH turnover assays indicated that the synthetic 2OG derivatives compete with 2OG for binding to the AspH active site. To investigate the divergent effects of C3/C4-substituted 2OG derivatives on AspH catalysis, i.e. AspH inhibition or promoting AspH activity, crystallographic studies were initiated. In the reported AspH crystal structures, the natural AspH cosubstrate 2OG was substituted for a 2OG competing inhibitor (e.g. NOG, 2,4-PDCA or L-malate), 22,41a but an AspH crystal structure complexed with 2OG has not previously been reported.
To enable comparisons of how 2OG and the 2OG derivatives bind AspH, AspH was rst crystallized in the presence of 2OG with the natural AspH cofactor Fe(II) being replaced by Mn(II). AspH crystallized in the absence of substrate in the P2 1 2 1 2 1 space group (AspH:2OG; 2.1Å resolution), the structure, as were the subsequently described structures, was solved by molecular replacement using a reported AspH structure (PDB ID: 5JZA) 22 as a search model (ESI Fig. S7 †). Clear electron density corresponding to 2OG was observed (Fig. 5a); the C5-carboxylate of 2OG being positioned to form a salt bridge with the side chain of Arg735 (2.4 and 3.0Å) and to interact with Ser668 (2.6Å), which is part of an 'RXS motif' present in some other 2OG oxygenases. 48 The C1-carboxylate of 2OG is positioned to interact with Arg688 (2.7 and 2.9Å) and His690 (3.3Å). The C1carboxylate of 2OG and the 2OG C2-carbonyl group complex the Mn ion in a bidentate manner (1.6 and 2.5Å; Fig. 5a). Two water molecules also coordinate the Mn ion (2.1 and 2.4Å) along with the two anticipated residues His679 and His725 (2.2 and 2.1Å; Fig. 5a). Thus, the AspH:2OG structure supports the proposal (based on the AspH structures in complex with the 2OG analogue NOG 22 ) that the active site metal, when bound to the natural AspH cosubstrate 2OG, is complexed by only two AspH residues rather than by the typical triad of ligands (HXD/E/H) found in other human 2OG hydroxylases. 7,14 Superimposition of the AspH:2OG structure with the reported AspH:NOG 22 and AspH:L-malate 22 structures reveals that AspH adopts similar conformations in all structures and that 2OG binds the AspH active site in a similar manner to NOG (Ca RMSD ¼ 0.21 and 0.21Å, ESI Fig. S7 †).
To investigate the effects of substrate binding in the presence of 2OG, AspH was crystallized in the presence of 2OG, Mn(II), and the synthetic hFX-EGFD1 86-124 -4Ser substrate 22 (ESI Fig. S1c †), which mimics the EGFD1 of the reported AspH substrate human coagulation factor X (hFX). 49 The structure was solved by molecular replacement using a reported structure (PDB ID: 5JTC) 22 as a search model (P2 1 2 1 2 1 space group; 2.3Å resolution). The active site region manifested electron density for both 2OG (Fig. 5b) and for the hFX-EGFD1 86-124 -4Ser peptide (ESI Fig. S8 †). Substrate binding to AspH affects the relative alignment of the oxygenase and TPR domains, i.e. the distance between the Ca atoms of Leu433 on TPR repeat a6 and Pro756 in the AspH C terminal region decreases from $20Å to $14Å upon substrate binding (ESI Fig. S9 †). Evidence for an induced t substrate binding mechanism involving this conformational change has been described when the cosubstrate 2OG was substituted for NOG. 22 The signicant conformational changes in the AspH oxygenase domain triggered by substrate binding (ESI Fig. S10 †) do not affect the observed mode of 2OG binding in the active site (Fig. 5c); the binding modes of both 2OG and NOG are very similar, both in the presence or absence of substrate (ESI Fig. S9 †). Interestingly, whilst in the 2OG complex structure the substrate residue Asp103 hFX was observed in a single conformation whereas in the analogous NOG structure it was observed in two conformations (ESI Fig. S9 †).
High resolution crystal structures of AspH complexed with 3methyl-2OG (16)  and S13 †). A similar change in the relative alignment of the AspH oxygenase and TPR domains upon substrate binding was observed when 2OG was substituted for 16 (ESI Fig. S13 †).
Thus, the combined crystallographic analyses reveal that the C3/C4-substituted 2OG derivatives bind AspH in the same general manner as 2OG, with near identical binding modes for the oxalyl-groups and the C5-carboxylates. The different conformations for the C3-and C4-methylenes observed, however, do not correlate with catalysis versus inhibition (assuming the crystallographic binding modes reect those in solution). Thus, whilst the C3-and C4-methylenes of (R)-16 and (R)-17 adopt a very similar conformation, which is distinct from that of 2OG and 34, 16 is a cosubstrate, whereas 17 is an inhibitor.
Hydrophobic interactions made by the C3-methyl and C3ethyl groups of (R)-16 and (R)-17, respectively, and of one of the methyl groups of 34 with AspH, are consistent with the tighter binding of 16 as compared with 2OG and judged by K app m comparison (Table 2). However, the k cat for 16 is approximately half that of 2OG whereas 17 and 34 are inhibitors and/or poor substrates, respectively (Table 2); the reason for these differences is uncertain, but it may reect slower binding of O 2 , or a slower subsequent step during catalysis, e.g. reaction of the ferryl intermediate proposed to be present in the 2OG oxygenase catalytic cycle or release of the succinate coproduct (ESI Fig. S21 †). Given the reduced k cat for 16, it is reasonable to propose that the 'k cat ' for the C3-ethyl substituted inhibitor 17 will be reduced even further, potentially approaching zero, potentially as a consequence of particular strong interaction with the hydrophobic pocket.

Discussion
The use of C3/C4-substituted 2OG derivatives has been a productive strategy to investigate inter alia the function, mechanism, and inhibition of aminotransferases 30,51 and dehydrogenases 51a,51b,52 that employ 2OG as a substrate or cosubstrate, but has been employed to a much lesser extent with 2OG oxygenases. [24][25][26] By contrast with 2OG derivatives bearing major structural modications of the 2-oxo-1,5-dicarboxylic acid scaffold, 53 the synthetic accessibility of C3/C4-substituted 2OG derivatives has been hitherto limited, a factor which might have hampered detailed biochemical studies on their effects on 2OG oxygenases. To address this need, we developed an efficient synthesis of C3/C4-substituted 2OG derivatives relying on the use of cyanosulfur ylids 33a,34 as key intermediates (Scheme 1). Our synthesis compares favorably to reported syntheses of 2OG derivatives (Fig. 2); 25,[27][28][29][30] it is scalable, affords the 2OG derivatives and their synthetic precursors in high purity suitable for biochemical applications, and avoids the use of strong bases, acids, and oxidants. The broad substrate scope of the synthesis reects the mild reaction conditions; for example, 2OG derivatives bearing oxidation-prone olens and thiophenes were readily synthesized (36-38; Table 1). No special laboratory equipment, such as an ozone generator, is required for the synthesis rendering it particularly user-friendly. We prepared racemic mixtures of C3/C4-substituted 2OG derivatives, however, the same synthetic strategy could be applied for the synthesis of enantiopure C4-substituted 2OG derivatives using enantiopure mono-methyl dicarboxylic acid half-esters 13 as starting materials, which, for example, can be obtained by asymmetric hydrogenation reactions from itaconates. 54 The C3/ C4-substituted 2OG derivatives were obtained as dicarboxylic acids (6, Scheme 1) and dimethyl dicarboxylates (12, Scheme 1). The former are useful for experiments with isolated enzymes as performed in this study, whereas the latter can be used for cellbased and in vivo experiments, due to their improved cell-wall penetrating abilities. 55 Many of the C3/C4-substituted 2OG derivatives synthesized inhibited human AspH by a mechanism involving competition with 2OG for binding the active site as revealed by inhibition assays performed at variable 2OG concentrations and crystallographic studies ( Fig. 3 and 6). In general, the C4-substituted 2OG derivatives were more efficient AspH inhibitors than the C3-substituted 2OG derivatives (Table 1). The C4-substituted 2OG derivatives were also more potent AspH inhibitors than the corresponding C4-substituted NOG derivatives (ESI Fig. S4 †). NOG and other N-oxalyl amino acids are plant natural products and it is proposed that they may act as enzyme inhibitors in vivo. 56 Our observations thus raise the possibility that naturally occurring 2-oxoacids may be biologically relevant modulators of the activities of 2OG oxygenases and related enzymes.
Our results indicate that some 2OG derivatives efficiently inhibit AspH at physiologically relevant 2OG levels which range from 9-12 mM 2OG in human plasma 47 to >1 mM 2OG in cells. 46 For example, 2OG derivative 33 inhibits AspH with 0.6 mM 2OG in the assay (IC 50 $ 5.6 mM; ESI Table S1 †). 2OG competitive AspH inhibitors, such as 33, 34, or optimized variants of them, might be useful from a therapeutic perspective, because their inhibitory effect is unlikely to be compromised by mutations. Indeed an AspH mutation associated with Traboulsi syndrome occurs in the 2OG binding site (i.e. R735W) and is likely inactivating. 20a At high enzyme concentrations, AspH converts some of the inhibitors (e.g. 34) slowly into the corresponding succinate derivatives (ESI Fig. S20 †), indicating that the hydrophobic interactions of the 2OG C3/C4-substituents with AspH stabilize the AspH:2OG derivative complexes and/or reduce the rotational exibility of the cosubstrate necessary to enable its oxidative decarboxylation. This observation is interesting because in some contexts, e.g. the inhibition of HIF-a prolyl hydroxylases, compounds that do not completely block activity even when present in excess, including poor cosubstrates, may actually be desirable, as they may help avoid overdose. 57 C3/C4-substituted 2OG derivatives bear the potential to be used as small-molecule probes 55a in cells or in vivo to modulate the catalytic activity of 2OG oxygenases, provided selective interaction can be achieved. In this regard, the 2OG derivatives might, at least partially, be selective AspH inhibitors considering that C4-substituted 2OG derivatives were reported to not inhibit wild-type KDM4A, 25 which is a 2OG dependent histone demethylase. 58 Note that, NOG derivatives have been used in vitro and in cellular experiments to modulate the catalytic activities of FIH 24 as well as of the KDM4 59 and ten-eleventranslocation (TET) enzymes 60 with some selectivity. Our results indicate that C4-substituted 2OG derivatives are more efficient in inhibiting AspH than the corresponding C4substituted NOG derivatives (ESI Fig. S4 †), as opposed to human FIH which also catalyzes the hydroxylation of Asp-and Asn-residues and for which NOG derivatives were more potent inhibitors. 24 With regard to enzymes other than 2OG oxygenases, the AspH inhibitor 4,4-dimethyl-2OG (34), for example, does not efficiently substitute for 2OG in glutamic oxaloacetic aminotransferase catalyzed transamination reactions suggesting that it might not interfere with the catalytic activities of human aminotransferases. 61 The lysine metabolite 2-oxoadipate, which is based on a C6 rather than a C5 carbon skeleton as in 2OG, is capable of acting as a relatively poor cosubstrate for procollagen prolyl hydroxylases, 62 phytanoyl-CoA 2-hydroxylase (PAHX), 63 and for a bacterial ethylene-forming 2OG dependent enzyme. 64 However, to our knowledge, C3/C4-substituted 2OG derivatives have so far not been reported to substitute for 2OG as a cosubstrate for wild-type 2OG oxygenases. The observation that several of our synthetic 2OG derivatives efficiently substituted for 2OG as an AspH cosubstrate, with 2OG derivatives 16 and 41 being the most efficient (ESI Table S2 †), is therefore of more general interest.
The observation that 4-carboxyphenylglyoxylic acid (41) can promote turnover of a 2OG oxygenase is remarkable because of its cyclic aromatic scaffold. 41 is thus a promising candidate to modulate AspH activity in vivo, as it might display selectivity for AspH over other 2OG oxygenases because of its distinctive structure. The oxidative decarboxylation of 41 by AspH is reminiscent of the reaction catalyzed by 4-hydroxyphenyl pyruvate dioxygenase (HPPD), 65 which is from a different class of Fe(II)-dependent oxygenases. A regioisomer of 41, 3-carboxyphenylglyoxylic acid (40), is a plant metabolite; 66 thus, it is possible that 41 or related aromatic compounds may modulate the catalytic activity of 2OG oxygenases in vivo.
The k cat /K m -value of AspH for 3-methyl-2OG (16) is approximately three times higher than for 2OG, suggesting the feasibility of 2OG derivatives to selectively enhance AspH (or indeed other 2OG oxygenase) catalysis in cells or on the cell surface of cancer cells in the presence of 2OG (Table 2). 16 is of particular interest because it is a reported ingredient of human nutrition (i.e. it is present in honey); 67 thus, 16 might modulate the activity of AspH and potentially other 2OG oxygenases in humans. 2-Methylsuccinate, which is formed by the AspHcatalyzed oxidative decarboxylation from 16 (ESI Fig. S5 †), has been detected in human urine 68 and is inter alia used as a biomarker for metabolic diseases as it is a product of other metabolic pathways (i.e. isoleucine catabolism); 69 this, however, does not rule out the possibility that some 2-methylsuccinate might originate from the oxidative decarboxylation of 16 catalyzed by 2OG oxygenases. 16 is also a proposed precursor of 3methyl glutamate, which is incorporated in natural products such as polytheonamide A and B 70 and daptomycin. 71 16 is likely biosynthesized in microorganisms through the direct reaction of 2OG and SAM, 72 suggesting that it might also be biosynthesized by animals in a similar manner. The results thus raise the possibility that 16 or other C3/C4-substituted 2OG derivatives, including their corresponding glutamate derivatives, are human/animal metabolites and/or are bioavailable through nutrition or the gut microbiome.
In comparison with 16, its isomer 4-methyl-2OG (26), which is a reported metabolite in plants 73 and ingredient of wine, 74 is substantially less efficient in substituting for 2OG as a cosubstrate (ESI Table S2 †), highlighting that the position of the 2OG substituent determines the ability of the 2OG derivatives to serve as an alternative AspH cosubstrate.
AspH was co-crystallized in the presence of 3-methyl-2OG (16) affording structures with the highest resolution reported for AspH so far (1.5 and 1.8Å; ESI Fig. S11 and S12 †). The crystallographic studies revealed that 16 substitutes for, and binds similarly to, 2OG in the AspH active site (Fig. 5). However, the C3-and C4-methylenes of 16 occupy different conformations orienting the 2OG C3-methyl-substituent towards Val676 and Val727, so enabling hydrophobic interactions. Superimposition of 16 and 3-ethyl-2OG (17), which is a potent AspH inhibitor, reveals that both 2OG derivatives adopt a similar conformation when bound to AspH (Fig. 6c). Despite the use of racemic 16 and 17 for crystallizations, in both cases the (3R)enantiomers were observed by crystallography, with the C3-alkyl substituents interacting with one face of a hydrophobic pocket (ESI Fig. S16c †). Thus, subtle changes in the structure of the 2OG derivative can have a pronounced effect on AspH catalysis, resulting in either efficient substrate hydroxylation or efficient AspH inhibition. The exact factors that determine whether a particular 2OG derivative inhibits or enables AspH catalysis are unknown, but might relate to differences in oxygen binding, modulation of steps aer oxygen binding or to the stability of the AspH:succinate derivative complexes (ESI Fig. S21 †).
The potential of compensating catalytically inactivating (with 2OG as a cosubstrate) mutants of 2OG oxygenases with 2-oxoacids has been exemplied in the case of PAHX. 63 Our results thus raise to possibility that naturally occurring 2OG derivatives/analogues may modulate the activity of 2OG oxygenases and related enzymes (e.g. HPPD 65 ), either by acting as inhibitors or by replacing 2OG in catalysis. Indeed, it is possible that some of the 2OG oxygenases without assigned biochemical functions (e.g. Jumonji C domain-containing protein 1c, JMJD1C) 75 or with unusual Fe(II) binding site geometries (e.g. PHD nger protein 2 (PHF2) 76 and hairless 77 ) use 2-oxoacid cosubstrates other than 2OG. Future work will focus on exploring these possibilities using the 2OG derivatives described here and others prepared by the cyanosulfur ylid methodology.

Conclusions
A user-friendly efficient synthesis of C3/C4-substituted 2OG derivatives based on the use of cyanosulfur ylids was developed and employed to afford a diverse set of 2OG derivatives for detailed biochemical and structural investigations on the cosubstrate selectivity of the human 2OG oxygenase AspH. The overall results reveal that C3/C4-substituted 2OG derivatives can have profound effects on AspH catalysis and, by implication, likely other 2OG oxygenases. Simple alkyl substituents, e.g. methyl at the 2OG C3-or C4-postion, enables retention of productive catalysis, likely in a stereoselective manner. By contrast, e.g. 2OG C4-dimethylation leads to inhibition/low levels of cosubstrate activity. The use of 2OG derivatives may thus inform on both the catalytic mechanisms and biological roles of AspH and other 2OG oxygenases, and aid in the development of new types of small-molecules that modulate 2OG oxygenase activity.

Author contributions
L. B. synthesized the 2OG derivatives and the AspH substrate, produced recombinant human AspH, and performed AspH assays and crystallizations. Y. N. solved and rened the AspH crystal structures. All authors analyzed data. L. B. and C. J. S. wrote the manuscript with help from Y. N.

Conflicts of interest
There are no conicts to declare.