Selective Inhibitors of a Human Prolyl Hydroxylase (OGFOD1) Involved in Ribosomal Decoding

Abstract Human prolyl hydroxylases are involved in the modification of transcription factors, procollagen, and ribosomal proteins, and are current medicinal chemistry targets. To date, there are few reports on inhibitors selective for the different types of prolyl hydroxylases. We report a structurally informed template‐based strategy for the development of inhibitors selective for the human ribosomal prolyl hydroxylase OGFOD1. These inhibitors did not target the other human oxygenases tested, including the structurally similar hypoxia‐inducible transcription factor prolyl hydroxylase, PHD2.


Introduction
In humans and other animals, prolyl hydroxylases (PHs) play critical roles in collagen biosynthesis and hypoxias ensing. [1][2][3] The PHs are Fe II and 2-oxoglutarate (2OG) dependent oxygenases, which normally produce succinate and CO 2 as coproducts. [4] The procollagen PHs (CP3H and CP4 H) hydroxylate either C3 or C4 of prolyl residues;t he latter is essentialf or maintenance of the collagen tripleh elix secondary structure ( Figure 1A). [5,6] In humans, the prolyl hydroxylased omain enzymes (PHD1, 2, and 3) act as oxygen sensors in the chronic response to hypoxia by catalyzing oxygen-limited hydroxylation of prolyl residues in the hypoxia-inducible factor-a (HIFa)s ub-units of HIF transcription factors, leading to HIFa degradation in aerobic conditions ( Figure 1A). [1] The human 2OG oxygenase OGFOD1 hasb een recently shownt oh ydroxylate Pro-62 of the ribosomal protein RPS23. [7][8][9][10] Pro-62 RPS23 is situatedi nt he ribosomal decoding site, which is responsible for ensuring the fidelity of mRNA codon recognition by tRNA and release factor proteins during protein synthesis. [11,12] While the role of this hydroxylation in Figure 1. Prolyl hydroxylase reactionsa nd structures:( A) Regio-and stereoselectivity of hydroxylations catalyzedbydifferent types of prolylh ydroxylases. Each hydroxylationisc oupled to the conversion of 2-oxoglutarate( 2OG) and O 2 into succinate and CO 2 .OGFOD1 acts on ar ibosomal protein, the CPHs act on procollagen, and the PHDs act on the hypoxia induciblef actor (HIF) transcription factors. [5] (B) Structures of 2OG and the 2OG analogue Noxalylglycine (NOG). (C, D) Views from the crystallographically observed active sitesofO GFOD1 (PDB 4NHX) [10] and PHD2( PDB 5L9R) [15] showing the interactionsb etween active site residues, the bound metal [Mn II substituting for Fe II ], and NOG. human and animal cells is not yet understood, the Saccharomyces cerevisiae OGFOD1 homologueT pa1p is proposed to catalyzed ihydroxylation of the corresponding prolyl residue, and to regulate translational accuracy in an mRNA sequence context-dependentmanner. [8,13,14] Of the~60-70 human 2OG oxygenases,s ome are current medicinal chemistryt argets,i ncluding enzymes involved in chromatin modification and lipid metabolism. [5,16] Inhibition of the procollagen hydroxylasesi su nder consideration as a target to limit the overproduction of collagen associatedw ith certain cancers andf ibrotic diseases. [17] The PHDs are presently being targeted for the treatment of hypoxia-related diseases, with inhibitors in late-stage clinicalt rials for anaemia. [18] If OGFOD1i si ndeed involved in mRNA codon recognition, as suggested based on studies of yeast homologues, [8,13] smallmolecule-mediated inhibition of ribosomal hydroxylation could prove useful for the treatment of diseases such as muscular dystrophy that are caused by premature stop-codons through nonsense suppression. [19] However, due to the uncertainty regardingt he specific roles of OGFOD1a nd OGFOD1-catalysed hydroxylation in animals, it is unclear how exactly its inhibition might manifest. Thus, such OGFOD1 inhibitors are also of interest as chemical probes to decipher the biological role of RPS23 hydroxylation, as wella st hose of other recently reported ribosome-associated hydroxylations. [20][21][22][23] As the 2OG oxygenases are involved in diverseb iological processes,d eveloping inhibitors selective for particular oxygenases is an important therapeutic consideration. The available biophysical evidence, principally from crystallography, implies that key features in the active sites of the different types of human PHsa re substantially,b ut not completely,c onserved. [4] Therefore, there is the potentialt hat inhibitors targeting OGFOD1 could also interfere with hypoxia sensing and collagen biosynthesis throughi nhibition of other PHs (or vice versa).T od ate, no detailede vidence for inhibitors selective for the different typeso fh uman PHs have been reported. Here, we establish the viability of as tructure-guided template-based approachf or the development of selective OGFOD1 inhibitors which do not target the otherh uman oxygenases tested, including the human hypoxia-sensing enzymePHD2.

Results
To assess the viability of developingi nhibitors selective forparticular PHs, with af ocus on OGFOD1, we first compared crystal structures of OGFOD1a nd PHD2. [10,24,25] Although there are differences in the OGFOD1 and PHD2 active sites, the binding modes of 2OG [and the 2OG analogue N-oxalylglycine (NOG)] appear largely conserved ( Figure 1B-D). The 2-oxoacid group of 2OG (or NOG) binds to the metal in ab identate manner, while the 2OG C-5 carboxylate is positioned to interactw ith conserved tyrosine and arginine residues (Tyr169 and Arg230 in OGFOD1, Tyr329 and Arg383 in PHD2;F igure 1C,D). These comparisons, along with those shown foro ther human oxygenases, [4] suggestt hat inadvertent inhibition of the PHDs may represent ac hallengei nd eveloping selectiveO GFOD1 inhibitors, and vice versa.
While many reported PHD2 inhibitors possess ag lycinamide side chain, the triketone plant growth/HPPD inhibitors do not ( Figure 2A). Modeling results suggest that the sulcotrione 2 methyl sulfonyl group and the nitro group of mesotrione 3 and nitisinone 1 may mimic the 2OG/glycinamide side-chain binding at their active sites ( Figure 2B). It is possible that the enzymea ctive site may accommodate the side chains of nitisinone 1,s ulcotrione 2,a nd mesotrione 3,w hich are bulkier than the glycinamide side chain of FG2216 7;h owever,i nt he absence of structurali nformation, we cannotp reclude the possibility that these inhibitors with bulky side chains interact with the enzymeinana lternate orientation.
The importance of the N-alkyl substituents for inhibition was examined with barbiturates bearing C-5 cyclopentyl substituents ( Figure 4C). Analogues bearing N,N'-dimethyl ( 29), N,N'-diethyl (30), and N,N'-dicyclohexyl (31)g roups were prepared ( Figure 4C); analogues 30 and 31 were less potent than 29, suggesting that substituents larger than methylg roups may not be favorable forO GFOD1i nhibition ( Figure S9). Note, however,t hat GSK1278863 6 potently inhibits OGFOD1d espite the presence of N,N'-dicyclohexyl groups (Figure 2A). [33] Modeling suggestst hat the aryl side chainso f25 and CCT3 likely do not fit in the OGFOD12 OG binding site due to steric constraints;i nstead, the aryl ring may bind in the substrate binding site, potentially contributing to the selectivity of these inhibitors. Varying the C-5 side chain with different mono-and bicyclic aromatic and saturated substituents (32-45)d id not have as ubstantial impact on potency ( Figure 4D,F igure S10). Similarly,c hanging the nature of the C-5 link from ak etone to an amide (i.e., 46), or extending the carbon chain length beyondt wo carbons between the carbonyl and the substituent had little effect ( Figure 4D).
To validate the in vitro inhibition results, direct bindingi nteractions between inhibitors CCT3 and CCT4 and OGFOD1a nd PHD2 were assessed by NMR analysis. While both inhibitors were observed to strongly bind to OGFOD1,a so bserved using 1 HC arr-Purcell-Meiboom-Gill (CPMG) analyses ( Figure S13), only weak binding to PHD2 was observed by water-Ligand Observe Gradient Spectroscopy (wLOGSY) experiments (Figure S14). Competitione xperiments between the inhibitors and enzyme-bound 2OG were conducted by monitoring the recovery of the enzyme-free2 OG methylenepeak at 2.35 ppm using CPMG-edited 1 H-NMR upon addition of the inhibitor. [35] The results indicate that CCT3 and CCT4 are capable of displacing bound2 OG from the active site of OGFOD1, but not from that of PHD2 ( Figure 5C,S 15).
We examined the potential inhibition of PHD enzymes by CCT3 and CCT4 using the HeLa human cell line. [36] Compared to the known PHD inhibitor IOX4, [34] CCT3 and CCT4 did not  Figure 5B). PHD-catalyzed hydroxylation targets HIFa for proteasomal degradation, indicating that these compounds do not inhibit the PHDs in cells. Based on aM DR1-MDCK assay (performed by Cyprotex, UK;s ee the Supporting Information), CCT3 and CCT4 demonstrate good cell permeability properties, and are predicted to be permeable to the blood brain barrier.L iver microsome stabilitys tudies indicate only low levels of clearance of CCT3 and CCT4 (Cyprotex, UK).

Discussion
Our resultsd emonstrate the viability of at emplate-based approach for the development of selective 2OG oxygenase/prolyl hydroxylase inhibitors capable of differentiating between closely related active sites, such as those of the human prolyl hydroxylases OGFOD1a nd PHD2.F urthermore, the optimized inhibitors also did not inhibit the other 2OG oxygenases tested, including other ribosomal oxygenases, as well as histone demethylases. The resultss uggestt hat specific inhibitor 'templates' maybepreferred for certain oxygenases or oxygenase subfamilies, as supported by work implying differential selectivity betweenP Ha nd JmjC histoned emethylase inhibitors. [37] This preference for particulart emplates may even extend to enzymesw ith closely related actives ites. Appropriately modified barbiturate-based inhibitors may selectively inhibit OGFOD1b ecause of their ability to support substituents, which extend towards active-site residues present in OGFOD1 but not in PHD2. By contrast, the glycinamide side chain present in many PHD2 inhibitors (including severalc ompounds in clinicalt rials, e.g., 6 and 7)i sc learly not required for potent OGFOD1i nhibition. It should also be noted that potent PHD2 inhibitors without ag lycinamide side chain are known. [33,34] In future work, it will be of interesttofurther explore the selectivity of the compounds reported here. In this regard, studies with the procollagen C-4 (and C-3) prolyl hydroxylases are of particulari nterest, especiallya st he procollagen C-4 PHs are potentialtherapeutic targets. [38] There is considerable academic and pharmaceutical interest in developing chemical probe compounds to investigate the biological functions of 2OG oxygenases. [39] Our resultss uggest that the developmento fl eads based on known pharmaceutical and agrochemical 'templates' (someofw hich can penetrate the blood-brain barrier) with well-studied physicochemical properties, such as barbiturates,w ill be ap roductive strategy. The combination of the tricarbonyl barbiturate template of the PHD2 inhibitor GSK1278863 6 with the side chains of agrochemicalo xygenase inhibitors (e.g.,p rohexadione-calcium), followed by subsequent optimization,y ielded potent and selective OGFOD1i nhibitors. Futurew ork can now be focused on applying these OGFOD1i nhibitors to investigate the biological roles of OGFOD1, and applying as imilar inhibitord evelopment strategy to other ribosomal oxygenases. Based on what is observedi nt hesef unctional studies, further optimization of these inhibitors may be warranted (e.g.,i fp enetration of the blood-brain barrierisd esirable).
It is important to note that many PHD2 inhibitors reported in the literature, including those screened in this work, and those currently in clinicalt rials, also inhibitO GFOD1. [33] Indeed, they may also inhibit other human prolyl hydroxylases and 2OG oxygenases, including those for which assays are currently not available. [10,40] From ac linicalp erspective,i ti sa lso important to note that the barbiturate-related 'triketone' HPPD inhibitor nitisinone, which is used in the treatment of type Ityrosinaemia, [27] is an OGFOD1i nhibitor,s omething that mightb e taken into consideration if nitisinone successors with improved properties are pursued. In the presentw ork, we have demonstratedt hat it is possible to attain selectivity between different 2OG oxygenases,w ith lead compounds that inhibit OGFOD1, but not PHD2.S uch "biochemicals electivity" is not necessarily an issue with clinicala pplicationsa st he desired pharmacological effect/safety profilem ay be achieved by controlling metabolism and tissue distribution.H owever,w ep ropose that, at least for chronic applications, biochemical selectivity could and shouldb eo ptimized during the development of 2OG oxygenase inhibitors. We also hope that inhibitors selective forp articular 2OG oxygenases may help enable their individual biological roles to be deciphered.