Thio-Linked UDP-Peptide Conjugates as O-GlcNAc Transferase Inhibitors.

O-GlcNAc transferase (OGT) is an essential glycosyltransferase that installs the O-GlcNAc post-translational modification on the nucleocytoplasmic proteome. We report the development of S-linked UDP-peptide conjugates as potent bisubstrate OGT inhibitors. These compounds were assembled in a modular fashion by photoinitiated thiol-ene conjugation of allyl-UDP and optimal acceptor peptides in which the acceptor serine was replaced with cysteine. The conjugate VTPVC(S-propyl-UDP)TA ( Ki = 1.3 μM) inhibits the OGT activity in HeLa cell lysates. Linear fusions of this conjugate with cell penetrating peptides were explored as prototypes of cell-penetrant OGT inhibitors. A crystal structure of human OGT with the inhibitor revealed mimicry of the interactions seen in the pseudo-Michaelis complex. Furthermore, a fluorophore-tagged derivative of the inhibitor works as a high affinity probe in a fluorescence polarimetry hOGT assay.

The reversible attachment of β-N-acetylglucosamine (O-GlcNAc) to serine and threonine residues of nucleocytoplasmic proteins is a highly conserved and dynamic post-translational modification (PTM) found in higher eukaryotes.[1,2] The transfer of O-GlcNAc from the donor substrate UDP-GlcNAc onto substrate proteins is catalyzed by the O-GlcNAc transferase (OGT), while the removal is catalyzed by the O-GlcNAc hydrolase (OGA).[3] Both enzymes possess substantial substrate promiscuity, which enables this single pair of enzymes to control the O-GlcNAc proteome linked to the regulation of a range of cellular processes.[4,5] Protein O-GlcNAcylation is essential for proper development of the animal embryo,[6−8] and mutations in the O-GlcNAc transferase have recently been associated with intellectual disability.[9−11] Multiple metabolic pathologies, such as diabetes[12,13] and cancer,[14−16] as well as neurological diseases[17,18] such as Alzheimer’s and Parkinson’s have been linked to dysregulation of protein O-GlcNAcylation.

Elucidation of the precise biological role(s) of protein O-GlcNAcylation is markedly hampered by the lack of potent and selective inhibitors of OGT.[19] The unusual catalytic machinery of the enzyme, which only seems to be fully assembled on the binding of the donor substrate,[20] has proven to be difficult to target. Previously, high-throughput compound screening studies have yielded cell penetrant small molecules like BZX[21,22] and OSMI1.[23] The former renders OGT inactive by cross-linking Lys842 and Cys917 in the active site. Given its reactivity, BZX may also target enzymes that contain similar Lys/Cys pairs. OSMI1 inhibits OGT through an as of yet unknown mechanism and was also shown to affect cell surface glycosylation.[23] The most widely used OGT inhibitor, the per-acetylated derivative of 5-thio-N-acetylglucosamine, hijacks the hexosamine biosynthetic pathway to yield the donor substrate analogue UDP-5S-GlcNAc that is inherently nonselective in the context of several other families of essential UDP-GlcNAc processing glycosyl transferases present in the cell.[24]

Recently, we have reported the synthesis and kinetic evaluation of bisubstrate OGT inhibitors that have been conceived by tethering the anchoring UDP moiety of the donor substrate to the serine side chain of a series of heptapeptides derived from several human OGT (hOGT) acceptor proteins.[25] The bisubstrate UDP–peptide conjugate VTPVS(O-propyl-UDP)TA 1 acted as a competitive hOGT inhibitor in vitro (IC50 = 18 μM[25]).

To develop these UDP–peptide conjugates into useful chemical biology tools two problems need to be addressed. First, the potency needs to be increased to effectively compete with the donor substrate UDP-GlcNAc (Kd = 16 μM[20]). Second, the intrinsic cell penetrance of these conjugates is limited by their size and the negative charge of the UDP moiety. To address the first problem, an expedient access to a diverse set of bisubstrate inhibitors was required to enable efficient structure–activity studies. To improve cell penetrance, we aimed to fuse the bisubstrate “warhead” to a selection of cell penetrating peptides (CPP) to achieve cellular uptake.

In our original approach, the phosphorylated “stretched serine” building block was incorporated into the backbone of a peptide with subsequent formation of the pyrophosphate bond in a reaction with an activated UMP derivative (Figure ). To improve synthetic efficiency we sought an alternative modular strategy that would enable installation of tethered UDP as the single moiety on a variety of peptide backbones. Advantageously, the suggested convergent approach would also facilitate the assembly of the proposed CPP-bisubstrate inhibitors by solid phase peptide synthesis (SPPS) followed by attachment of the tethered UDP as the last synthetic step.

Figure 1

Synthesis and structure of the O-linked UDP–peptide conjugate 1.[25]

As the chemical structure of the O-linked UDP–peptide conjugates showed limited opportunity for modular assembly, we instead opted to explore novel S-linked UDP–peptide conjugates that could be accessed by chemoselective cysteine modification with a thiol-reactive tethered UDP derivative (Figure ). Evaluation of three mainstay techniques for carbon–sulfur bond formation such as Michael addition (Path A), phosphine promoted disulfide contraction[26] (Path B), and “click” thiol–ene conjugation (TEC)[27−30] (Path C) in the context of construction of such thiol reactive UDP derivatives revealed inherent limitations of the first two approaches (Figure ). Indeed, introduction of a mandatory electron withdrawing group would disrupt the topological integrity of the linker in the proposed Michael type reactive derivative. In turn, disulfide contraction, widely used for preparation of S-linked glycoproteins, seemed to be less attractive for peptide substrates, as it would unavoidably result in the loss of the stereochemical integrity of the linking cysteine. In fact, only photoinitiated TEC reaction of cysteine containing peptides with simplistic allyl-UDP 4 held promise for an efficient assembly of the envisaged conjugates.

Figure 2

Selected possible approaches to the synthesis of S-linked UDP–peptide conjugates. In path A an electron withdrawing group in the Michael acceptor would disrupt the topological integrity of the linker. Path B leads to isomerization at Cys.[26] Path C allows for the efficient construction of the target molecule.

To synthesize the requisite allyl-UDP 4 the Bogachev–Kiessling[31,32] protocol was used by coupling UMP N-methyl imidazolide 3 with the crude bis-triethylammonium salt of allyl phosphoric acid, prepared in turn from allyl bis-fluorenemethyl phosphate 2 (Scheme , Scheme S8).

Scheme 1

Chemical Synthesis of Allyl-UDP

(i) Et3N-DCM, 4:1, 16 h; (ii) a) Bogachev–Kiessling coupling, 3 h; b) MeOH, Et3N, H2O, 16 h, 44%.

The structure of 4 was unambiguously assigned by a set of spectroscopic methods (Figures S1−S3); the presence of a pyrophosphate bond was confirmed by a typical pair of doublets (σ −11, −12, JP,P = 21 Hz) in the 31P NMR spectrum. The photoinitiated TEC reaction between 4 and 20% molar excess of the substrate peptide VTPVCTA 5 was initially attempted at 10 mM in 0.1 M acetate buffer (pH 4) in the presence of 20% of different water-soluble photoinitiators VA-044, DMPA (50% MeOH additive), or LAP.[33] The reactions were irradiated using an 8 W Camag TLC reader lamp at 366 nm and monitored by LC-MS. We found that LAP efficiently mediated the conjugation, resulting in clean formation of the expected product 6 (m/z 1175) after only 10 min (Scheme A).

Scheme 2

Photoinitiated Thiol–Ene Conjugation of Cysteine Containing Peptides with Allyl-UDP

(i) 8 W 366 nm, 10 min, 50–55%; (ii) 5(6)-fluorescein NHS ester, DMF, 0.15 M NaHCO3, 3 h, 85%.

On the other hand, <5% conversion was detected with VA-044, while no reaction took place with DMPA. The TEC reaction was equally efficient when no excess of the substrate peptide was present and with only 10% of the photoinitiator. No products were observed in the absence of LAP or without UV light. The reaction was scaled up to 0.05 mmol (20 mM in water), and the S-linked UDP–peptide conjugate 6 was isolated by preparative HPLC in 56% yield. The structure of the product was verified by 1H and 13C NMR spectra, proving the presence of the peptide backbone and the UDP moiety (Figures S4, S5). The 31P spectrum showed two broadened singlets at σ −10.94 and −11.48, confirming the integrity of the pyrophosphate bond.

The inhibitory potency of 6 was assessed by the previously described fluorometric OGT activity assay.[25] Unexpectedly, the S-linked UDP peptide conjugate 6 appeared to be an almost 10-fold more potent hOGT inhibitor (IC50 = 2 μM) than its O-linked progenitor 1(25) (Figure A). To provide a negative control and on the assumption that hOGT evolved to bind D-configured UDP, we synthesized the L-uridine congener 7 starting from allyl-L-UDP 4-L (Scheme A, Scheme S10), which showed no inhibition of hOGT (Figure A). Next, we demonstrated that 6 is also a hOGT inhibitor in a cell-free system, reducing protein O-GlcNAcylation in a dose-dependent manner by about 50% at 1 mM (Figure B, Figure S6).

Figure 3

(A) Dose–response curves of inhibition of hOGT activity in the presence of increasing concentrations of 6 and 7. Errors shown represent the s.e.m. of three replicates. (B) Western blot analysis of cell lysate in the presence and absence of 6, 7, and UDP-5S-GlcNAc. HeLa cell lysates were O-GlcNAc stripped by the potent bacterial O-GlcNAcase CpOGA, whose activity was then blocked by the potent OGA inhibitor GlcNAcstatin G. These stripped lysates were then re-O-GlcNAcylated by recombinant OGT and blotted for protein O-GlcNAcylation using the RL2 antibody. HSP60 was used as a loading control.

Next, we attempted the assembly of the potentially cell penetrating bisubstrate hOGT inhibitors. To this end, starting from peptides featuring a VTPVCTA C-terminus and either Penetratin (RQIKIWFQNRRMKWKK - 8/9)[34] or TAT (YGRKKRRQRRR - 10)[35] peptides at the N-terminus, we synthesized S-linked UDP peptide conjugates 11–13 by TEC reaction with allyl-UDP 4 (Scheme C).

Encouragingly, the conjugates 11 and 13 (IC50[13] = 5 μM, Figure S7) were shown to be almost as potent hOGT inhibitors in vitro as the parent compound 6. However, neither 11 nor 13 had a noticeable effect on total O-GlcNAcylation in cell cultures at concentrations up to 1 mM (Figure S8). Moreover, microscopy of the HeLa cells treated with the 5-fluorescein thioureide (Flut) labeled conjugate 12 for 24 h revealed fluorescent puncta (Figure S8). Taken together, these data suggest that while 12 and, implicitly 11 and 13, could cross the cell membrane, they remain trapped in the early endosomes[36,37] and therefore cannot target cytosolic hOGT.

Discovery of the potent hOGT binder 6 also offers an opportunity for development of a sensitive hOGT fluorescence polarimetry assay (FP). We explored the fluorescently tagged derivative 17 of the inhibitor as a high affinity FP probe. To this end, the peptide 14 was efficiently transformed into corresponding S-linked UDP–peptide conjugate by TEC reaction with allyl-UDP 4. Next, the N-terminal 6-aminohexanoyl (ahx) residue was tagged with 5(6)-fluorescein NHS ester to give the requisite 5(6)-fluorescein carboxamide (Floc) labeled derivative 17 (Scheme B). The same reaction sequence was performed with 14 and allyl-L-UDP 4-L to give the negative control compound 18.

Direct binding affinity of compound 17 to hOGT was measured, yielding a KD of 1.6 μM, whereas no quantifiable binding could be detected for the corresponding negative control 18 (Figure A). The observed value is in good agreement with the IC50 of 2 μM established in our activity assay.

Figure 4

(A) FP assay showing the binding of 17 and 18 to hOGT. (B) Dose–response curves from the FP assay showing the displacement of 17 by UDP-5S-GlcNAc, 1 and 6. Errors shown represent the s.e.m. of at least three replicates.

To verify whether the probe 17 could be used in a workable FP hOGT assay we demonstrated that the available competitive inhibitors UDP-5S-GlcNAc[24] and 1 displace 17 from the complex with the enzyme in a dose-dependent manner (Figure B). The derived values of Ki = 7 ± 2 μM (1) and Ki = 5 ± 2 μM (UDP-5S-GlcNAc) were in agreement with the previously reported affinities,[20,25] validating this assay.

Next, we used this assay platform to examine the effect of the peptide sequence on the inhibitory potency of the S-linked UDP–peptide conjugates. We compiled a set of peptides derived from the known hOGT substrate proteins RB2, TAB1, CK2, and KER, as well as a previously published hOGT sequon[38] (Table ). The topology of the heptapeptides in the 7-S5C subseries (19–23) was chosen to replicate that of the peptide 5. The nonapeptides 9-S5C (24–26) were designed with a centrally located Cys. The pair of the tridecapeptides 27 and 28 was included to explore distant regions of the binding groove. The TEC reaction of allyl-UDP 4 with all substrates worked routinely well to give an array of the novel S-linked UDP–peptide conjugates 29–38 (Table , Table S1) that were evaluated by the modified FP assay (Table , Figure ).

Table 1

Inhibition of hOGT with UDP–Peptide Bisubstrate Conjugates 6 and 29–38

	sequence	origin/Cys position	Cys peptide	product	Ki (μM)
1	VTPVCTA	RB2_7-S5C	5	6	1.3 ± 0.8
2	VPTVCTA	pseudo RB2_7-S5C	19	29	12.8 ± 4.5
3	SVPYCSA	TAB1_7-S5C	20	30	2.3 ± 1.6
4	VTPVCSA	CK2_7-S5C	21	31	1.2 ± 1.3
5	VTPVCRA	SEQ_7-S5C	22	32	1.3 ± 0.5
6	PVFTCRS	KER_7-S5C	23	33	12.5 ± 7
7	VTPVCTATH	RB2_9-S5C	24	34	12.7 ± 2.3
8	SVPYCSAQS	TAB1_9-S5C	25	35	15 ± 2.3
9	PVFTCRSAA	KER_9-S5C	26	36	56 ± 32
10	PVCTATHSLSRLH	RB2_13-S3C	27	37	80 ± 46
11	KENSPAVTPVCTA	RB2_13-S11C	28	38	1.8 ± 1.3

Figure 5

Schematic of the synthesis of UDP–peptide bisubstrate conjugates shown in Table .

The collated data (Table ) suggest that conjugates derived from the heptapeptides 7-S5C cover the minimal structure of the bisubstrate inhibitor until there is proline in the −2 position, as the binding potency remains in the range of 1 ± 0.5 μM (note the potency drop in the conjugate 29 (entry 2)). These data, taken together, are in good agreement with previous findings that emphasize the importance of a proline in the −2 position.[38] Notably, C-terminal elongation of conjugates in 9-S5C and 13-SC3 series results in a steady potency drop, while N-terminal extension did not affect the binding of the 13-S11C conjugate 38.

To reveal likely reasons for the enhanced potency of the S-linked UDP–peptide conjugates we collected high-resolution synchrotron diffraction data of crystals of hOGT in complex with 6 (1.85 Å, Rwork/Rfree = 0.22/0.25) or its O-linked progenitor 1 (1.68 Å, Rwork/Rfree = 0.19/0.22) (Figure , Table S2). Structure solution by molecular replacement and subsequent refinement revealed continuous |Fo|–|Fc| electron density for both ligands allowing the unambiguous placement of each (Figure S9). The fully refined models revealed both conjugates bind to the OGT active site in a conformation closely resembling the previously reported pseudo-Michaelis complex of hOGT with UDP-5S-GlcNAc and an acceptor peptide[20] (Figure ). The largest atomic shift between the UDP moieties of conjugates 1 and 6 and the corresponding substrates/substrate analogues is 0.7–0.8 Å. Shifts were also observed between the positions of the linking oxygen and sulfur (0.8 Å) and the positions of the linker C1 atoms (1.1 Å). Overall, in the conjugate 1 the linker adopts a synclinal conformation (dihedral angle O–C1–C2–C3 of 72°), while in 6 the linker adopts an antiperiplanar conformation (dihedral angle S–C1–C2–C3 171.4°). This difference may contribute to the increased potency of 6, as the antiperiplanar conformation of the thio-propyl linker seen in 6 is energetically more favorable than the synclinal conformation of the oxy-propyl linker in 1.

Figure 6

Crystal structures of the pseudo-Michaelis complex (PDB 5C1D(38)) and bisubstrate inhibitors 1 and 6 bound to hOGT. The protein is shown as a grey cartoon overlaid with a grey surface, the peptide part is colored in blue, the linker/sugar is colored in green, and the UDP moiety is colored in magenta.

In conclusion, we developed a modular synthetic approach to S-linked UDP–peptide conjugates using a photoinitiated TEC reaction between allyl-UDP and cysteine containing peptides. A series of conjugates sharing the XXPXC(S-propyl-UDP)XX scaffold showed binding potency toward hOGT in the 1 ± 0.5 μM range in vitro, which places them among the most potent hOGT inhibitors reported to date. The increased potency of the conjugate VTPVC(S-propyl-UDP)TA in comparison with its O-linked progenitor correlates with the more relaxed conformation of the thiopropyl linker. We have shown that the chosen S-linked conjugate inhibits recombinant hOGT activity in HeLa cell lysates and successfully assembled two prototypes of the cell-penetrating bisubstrate hOGT inhibitors that feature Penetratin and TAT peptides C-terminally extended with the VPTVC(S-propyl-UDP)TA motif. Finally, we developed a practical variant of the fluorescence polarimetry hOGT assay using a fluorescently labeled derivative of the bisubstrate conjugate as a high affinity probe suitable for high-throughput screening. Further optimization of the structure of the novel hOGT bisubstrate inhibitors as well as evaluating of different types of cell penetrating vehicles could lead to further improvements of this scaffold.