OleD Loki as a Catalyst for Hydroxamate Glycosylation

Abstract

Herein we describe the ability of the permissive glycosyltransferase (GT) OleD Loki to convert a diverse set of > 15 histone deacetylase (HDAC) inhibitors (HDACis) into their corresponding hydroxamate glycosyl esters. Representative glycosyl esters were subsequently evaluated in assays for cancer cell line cytotoxicity, chemical and enzymatic stability, and axolotl embryo tail regeneration. Computational substrate docking models were predictive of enzyme-catalyzed turnover and suggest certain HDACis may form unproductive, potentially inhibitory, complexes with GTs.

Graphical Abstract

HDAC inhibitor sweetener: Hydroxamate-based HDAC inhibitors (HDACis) were identified as glycosyltransferase (OleD Loki) substrates, and corresponding hydroxamate glycosyl ester products were evaluated in assays for cancer cell line cytotoxicity, chemical/enzymatic stability, and axolotl (salamander) embryo tail regeneration. Computational docking models suggest certain HDACis to form unproductive, potentially inhibitory, complexes with glycosyltransferases.

Sugar nucleotide-dependent glycosyltransferases (Leloir GTs)^[1] catalyze the regio/stereospecific transfer of sugars from activated sugar-nucleotide donors to hydroxy groups,^[2] amines,^[3] thiols,^[4] and activated carbon nucleophiles.^[5] GTs are prevalent in nature and contribute to diverse cellular functions including, but not limited to, cellular signaling, molecular recognition, energy/metabolite storage, and drug resistance/detoxification. These proficient biocatalysts have also been used for glycoside synthesis as exemplified by applications of the permissive microbial detoxifying GTs YjiC^[6] and OleD.^[7] Within this context, OleD acceptor and sugar nucleotide donor permissivity has been further enhanced via directed evolution and structure-based approaches.^[7g,j,k] An OleD-based transglycosylation strategy has also been developed^[7f,i] that exploits GT-catalyzed reaction reversibility, and the ability of enhanced OleD mutants (OleD ASP, TDP16 and Loki)^{[7a–f,i–k]} to efficiently use simple colorimetric or fluorescent substrates. This transglycosylation platform also enabled a plate-based screen to identify new OleD substrates and improved OleD-catalyzed syntheses via shifting the reaction equilibria toward desired glycoside product.^[7a–f]

One such recent transglycosylation screen identified the histone deacetylase inhibitor (HDACi) trichostatin A as a new OleD Loki substrate and subsequently demonstrated the efficient OleD-catalyzed synthesis of the corresponding glucopyranosyl hydroxamate trichostatin C (the only naturally occurring hydroxamate glycoside isolated to date, produced by Streptomyces platensis No. 145 and Streptomyces strain Y-50).^[7a,8] Although HDACis have broad clinical utility, their corresponding glycosides have not been extensively studied. To further probe the synthetic utility of OleD, herein we describe the evaluation of 21 structurally diverse hydroxamate-based HDACis as potential OleD substrates. This study revealed OleD Loki to turnover 17 of the hydroxamates tested to their corresponding hydroxamate glycosyl esters. Subsequent bioactivity studies revealed hydroxamate glucosylation to decrease cytotoxicity by 20- to > 3000-fold. The corresponding hydroxamates were also found to be resistant to glycosidase-catalyzed hydrolysis and afford moderate acid and base stability. Cumulatively, this study advances a new OleD catalytic function and the first fundamental stability and bioactivity studies for hydroxamate glycosides.

Using 2-chloro-4-nitrophenyl β-d-glucoside (ClNP β-d-Glc) and the OleD variant Loki,^[7a–e] 21 representative hydroxamate-based HDACis were evaluated as putative substrates via a simple colorimetric transglycosylation screen under the control of a catalytic amount of UDP (Figure 1 A). Assays (1 mm hydroxamate, 2 mm ClNP β-d-Glc, 0.1 mm UDP, 25 mm Tris pH 8.0, 5 mm MgCl₂, 0.25 μm OleD Loki, 30 μL total volume, 30°C, 1 h, A₄₁₀) were conducted in triplicate in 384 well plates. Each plate also contained a positive (4-methylumbeliferone; 4-MeUmb)^{[7a–g,i–k]} and negative (DMSO) control. Seventeen preliminary primary hits were identified (ΔA₄₁₀ > 3 standard deviations above the negative control) in this first-pass screen (Figure 1 B). Crude reactions identified as preliminary primary hits were subsequently analyzed by LC–MS. For all 17 primary hits, the observation of a single major glucoside product by LC–MS served as key validation of the colorimetric screen (Figure S1 and Table S1 in the Supporting Information). Representatives 1–5 (Figure 2) were subsequently selected as models for scale-up and structure elucidation based on turnover and commercial availability.

Figure 1.

A) Schematic of the colorimetric screen. B) Reaction progress as monitored via ΔA₄₁₀ using the standard assay format (see panel A; 1 mm hydroxamate, 2 mm CINP β-d-Glc, 0.1 mm UDP, 25 mm Tris pH 8.0, 5 mm MgCl₂, 0.25 μm OleD Loki, 30 μL total volume, 30 °C, 1 h). Vehicle alone (no acceptor) served as the negative control and a well-characterized OleD substrate (4-MeUmb) as the positive control. Assays were conducted in triplicate. GT, OleD Loki; ClNP β-d-Glc, 2-chloro-4-nitrophenol β-d-glucose; UDP, uridine diphosphate. The bars in red represent the compounds selected for scale-up and characterization. Hydroxamate structures are illustrated in Figure 2.

Figure 2.

HDACis tested as putative OleD Loki substrates. Compounds in black were identified as hits in the colorimetric screen and subsequently confirmed by LC–MS. Percent conversion (in parentheses) was determined via HPLC peak integration. Compounds in grey lacked turnover.

Scale-up reactions were each accomplished in a total volume of 30 mL (1 mm 1–5, 20 mm Tris·HCl, pH 8.0, 5 mm MgCl₂, 2 mm CINP-β-d-glucose, 0.1 mm UDP, 1 μm OleD Loki, 30 °C, 12–24 h) with reaction progress monitored via ΔA₄₁₀. Upon completion of the reaction, products and residual reactants were captured by solid phase extraction and the resulting mixture purified by semipreparative reversed-phase HPLC (Supporting Information Method B). Glycoside products were collected and subjected to HR-ESI-MS to establish the molecular formula for compounds 1a–5a (Figure 3) as 1a, C₂₇H₃₃N₃O₁₀ [m/z 560.2240 (M + H)⁺]; 2 a, C₂₈H₃₅N₃O₈ [m/z 542.2489 (M + H)⁺]; 3a, C₃₀H₃₇N₃O₉ [m/z 584.2601 (M + H)⁺]; 4a, C₂₇H₃₃N₃O₇ [m/z 512.2392 (M + H)⁺]; and 5a, C₂₆H₃₁N₃O₇ [m/z 498.2237 (M + H)⁺], respectively (Table S1). Comparison of the 1a–5a 1D and 2D NMR to that of the corresponding parental hydroxamates (1–5; (Figures S6–S50) revealed 1a–5a signatures consistent atypical glucosides and a lack of the 1–5 hydroxamate hydroxy ¹H NMR resonance (δ_H = 8.5). Evidence of a 1 a–5 a β-O-glucoside derived from the key anomeric ¹H (near δ_H = 4.5, J=6.6–8 Hz) and ¹³C (δ_C = 102–106) resonances (Tables S2–S6, Figure S4) typical of β-O-glucosides.^[9] As comparators, typical anomeric carbon ¹³C resonances for C- and N-glycosides range from δ_C = 70–85^[10] and 85–95,^[7a,11] respectively. It is also important to note that 1 and 3–5 contain only a single accessible nucleophilic OH (that of the hydroxamate). Consistent with a glycosidic bond comprised of two heteroatoms,^[12] HMBC correlations between the sugar and aglycone structure were not observed for 1 a–5 a. Cumulatively, these data provide strong support for the proposed structures and are also consistent with the previously characterized naturally occurring hydroxamate glycoside trichostatin C.^[7a,8]

Figure 3.

Representative hydroxamate glycosyl ethers selected for scale-up and further study.

The original trichostatin C structure elucidation studies revealed susceptibility of the naturally occurring hydroxamate glucoside to strong acid (3 n HCl or 40% methanolic HCl).^[8b] However, to the best of our knowledge, the lability of hydroxamate glycosides has not been extensively studied. Thus, model hydroxamate glucosides 3 a and 4 a were further evaluated for pH stability and susceptibility to glucosidase-catalyzed hydrolysis. In this analysis, glucosides 3 a, 4 a and commercially available control 4-methylumbelliferyl-glucopyranosides were stable at pH 5, 7.5 and 10 (≤ 2% degradation over 30 days, room temperature; Figure S2). Consistent with the prior trichostatin C precedent, all glucosides were unstable to concentrated H₂SO₄ (≈50% glucoside hydrolysis within 5 min based on thin layer chromatography, data not shown). This limited study suggested similar pH liabilities among hydroxamate and umbelliferone glucosides. Hydroxamate glucosides 3 a and 4 a were also resistant to α- and β-glucosidase-catalyzed hydrolysis while corresponding controls 4-methylumbelliferyl α-d-glucopyranoside and 4-methylumbelliferyl β-d-glucopyranoside were rapidly hydrolyzed by α- or β-glucosidase, respectively (Figure S3). As glucosidases are typically permissive to substantial aglycon structural diversity,^[13] this result may suggest unique electronic and/or steric features contribute to the glycosidase resistance of hydroxamate glycosides.

HDACs catalyze histone lysine deacetylation and function as critical cellular epigenetic modifiers.^[14] The agents illustrated in Figure 2 are prototypical HDACis and include three clinically approved anticancer drugs (belinostat, panobinostat and vorinostat). Hydroxamates function as reversible inhibitors by chelating the key HDAC active-site Zn²⁺ and display potent cancer cell line antiproliferative activities. Yet, while trichostatin A and its hydroxamate glycoside trichostatin C were both reported to increase histone H4 acetylation in B cells and induce erythroleukemia differentiation,^[15] the impact of hydroxamate glycosylation on biological activity has not been extensively studied. To address this, compounds 1a–5a and their parental hydroxamates 1–5 were evaluated in standard cancer cell line cytotoxicity assays against both human colorectal (HCT116) and non-small-cell lung (A549) cancer cell lines (Figure 4). Similar to the previously observed decrease in potency invoked via trichostatin A glycosylation, glucosylation of 1–5 led to a 20- to > 3000-fold decrease in potency. Glucoside potency trends mirrored that of the parental HDACis with the glucoside of the pan-HDACi panobinostat (4 A549 IC₅₀ 0.8 nm; HCT116 IC₅₀ 1.8 nm) identified as the most cytotoxic hydroxamate glycosyl ester (4a A549 IC₅₀ > 1.0 μm; HCT116 IC₅₀ > 1.9 μm).

Figure 4.

Comparative cytotoxicity of representative hydroxamates and hydroxamate glycosyl ethers toward A) human colorectal (HTC116) and B) non-small cell lung (A549 cells) cancer cell lines.

Using a Mexican axolotl (Ambystoma mexicanum) embryo tail regeneration (ETR) assay,^[16] HDACis have also been identified as potent inhibitors of tail regeneration. Additional subsequent HDACi-based chemical genetic and microarray studies highlighted the importance of HDAC activity at the time of tail amputation to regulate the initial transcriptional response to injury and regeneration in the axolotl model.^[16e] To investigate the impact of hydroxamate glycosyl esters within this context, compounds 1a–5a and their parental hydroxamates 1–5 were evaluated in the ETR assay.^[15] Tail-amputated embryos were incubated in microtiter plates in the absence (vehicle control, DMSO) or presence of 10 μm test agent (1–5, 1a–5a) and imaged on day 1 (pre-treatment) and day 7. Consistent with the prior study, parental hydroxamates inhibited tail regeneration (Figure S5). In contrast, hydroxamate glycosides were inactive at the concentration tested, potentially due to reduced potency (consistent with Figure 4) and/or in vivo exposure.

Computational substrate docking models (Figure 5) are also consistent with empirically determined turnover (or lack thereof). Specifically, these models highlight that the representative models for high (abexinostat, Figure 5 A), moderate (tepoxalin, Figure 5 B), and low (MC1568, Figure 5 C) OleD turnover all orient the hydroxamate within close proximity to the key OleD catalytic H19-D110 acid-base pair. Conversely, the predicted high affinity binding mode of a ligand that lacked turnover with OleD (tubacin) revealed a catalytically unproductive conformation (Figure 5D). This computational model may implicate tubacin and/or related pharmacophores as a potential new starting point for GT inhibitor discovery.

Figure 5.

The predicted binding OleD binding models for A) abexinostat (high turnover, Figures 1 and 2, 1), B) tepoxalin (moderate turnover, Figures 1 and 2), C) MC1568 (low turnover, Figures 1 and 2) and D) tubacin (no turnover, Figures 1 and 2) based on the wild-type OleD-erythromycin ligand-bound crystal structure (PDB ID: 2IYF). Yellow dashed lines reflect putative key binding/catalytic contacts (distances in Å), including that of the ligand hydroxamate with the putative OleD active site base (His19 side chain of the active-site H19–D110 acid–base pair).

In summary, this study expands on the prior discovery of trichostatin A as an OleD Loki substrate.^[7a] Consistent with the prior comparison of the catalytic competencies for trichostatin A (k_cat/K_M = 1.4×10⁴ μm⁻¹s⁻¹) to the parental OleD Loki acceptor 4-methylumbelliferone (k_cat/K_M = 2.2×10⁴ μm⁻¹ s⁻¹), the current study highlights OleD Loki as an efficient and permissive biocatalyst for HDACi glucosylation as a basis for exploring glycoconjugate tumor targeting and/or prodrug strategies.^[17] Hydroxamate glycosylation is also expected to circumvent the Lossen rearrangement to a highly reactive isocyanate, a reaction known to contribute to nonspecific alkylation and corresponding off-target toxicity observed by HDACis.^[14f] Most hydroxamates also suffer from rapid metabolism and clearance, where UDP-glucuronosyltransferase (UGT)-catalyzed hydroxamate glucuronidation plays a major role.^[18] Within this context, the OleD Loki platform may also serve as a potential screen to facilitate the discovery of hydroxamate analogues less susceptible to enzymatic glycosylation. These cumulative factors suggest OleD Loki may add to the repertoire of medicinal chemistry tools to advance HDACis for cancer,^[19] immunotherapy,^[20] neurological disorders,^[21] and/or infectious disease.^[22]