Deploying Fluorescent Nucleoside Analogues for High Throughput Inhibitor Screening

Abstract

High-throughput small molecule screening in drug discovery processes commonly rely on fluorescence-based methods including fluorescent polarization and fluorescence/Förster resonance energy transfer. These techniques use highly accessible instrumentation, however may suffer from high false negative rates and background signals or, may involve complex schemes for the introduction of fluorophore pairs. Herein, we present the synthesis and application of fluorescent nucleoside analogues as the foundation for directed approaches for competitive binding analyses. The general approach describes selective fluorescent environment-sensitive (ES) nucleoside analogues that are adaptable to diverse enzymes that act on nucleoside-based substrates. We demonstrate screening a set of uridine analogues and development of an assay for fragment-based lead discovery with the TcdB glycosyltransferase (GT), an enzyme associated with virulence in Clostridium difficile. The uridine-based probe used for this HTS has a K_D of 7.2 μM with the TcdB GT and shows a >30-fold increase in fluorescence intensity upon binding. The ES-based probe assay is benchmarked against two other screening approaches.

ToC Graphic

Introduction:

Fluorescence-based analytical approaches are ubiquitous in biochemical and biological studies and form the mainstay of small-molecule ligand discovery efforts in chemical biology and medicinal chemistry research. Fluorescence is particularly useful in high throughput screening applications where alternative approaches, such as those involving radioactivity or complex instrumentation, are less accessible. Methods including fluorescence polarization (FP) ^[1] and differential scanning fluorimetry (DSF) ^[2] analyses are widely applied for initial steps in small molecule screening. FP assays are based on changes in the orientation and mobility of ligands upon binding to macromolecules, thereby providing information on solution state interactions.^[3] FP plays an important role in the drug discovery process; when adapted to a multi-well plate reader format, FP assays provide relatively low cost analyses across a wide range of screening conditions.^[4] In DSF, changes in protein stability due to small molecule binding are quantified by assessing thermal shifts in the presence of a dye, such as SYPRO® Orange, which becomes more fluorescent when bound to unfolded proteins. Applications of both FP and DSF in high-throughput screening (HTS) and drug development are well known and have been thoroughly reviewed.^[5]

Despite the advantages of the aforementioned methods, there are notable drawbacks. For example, with FP there may be fluorophore aggregation,^[6] unacceptable false-negative rates from high background due to excess fluorophore,^[7] small variations in buffer and pH influencing the polarization of free fluorophore,^[2] and in particular for fragment-based screening, competition between the large excess of screening compounds and fluorophores for hydrophobic patches on protein surfaces.^[8] In DSF, thermal shifts may be very small (<1 °C) and information on the site of small-molecule binding is not defined by the assay. In light of these issues, there remains a need for equally accessible and complementary methods for HTS screening for small-molecule protein ligands.

As an alternative to FP and DSF, enzyme-targeted ligands that have been elaborated with environment-sensitive fluorophores (ESFs) can provide advantages for HTS. ESFs have been extensively used as solvent polarity indicators^[9] and as probes for investigating protein interactions and dynamics.^[10] ESFs are highly sensitive to the polarity of the local microenvironment due to modulation of the fluorophore quantum yield, which impacts signal brightness. Some ESFs may also undergo changes in emission wavelengths and thus are designated as solvatochromic.^[11] Importantly, for ESFs that show an increased quantum yields with decreased solvent polarity, an excess of unbound ESF-containing reagents in HTS assays does not significantly contribute to background fluorescence,^[12] thereby mitigating the problem of false negatives arising in some FP-based screens.^[13] Additionally, ESF probes that are selective for a specific hydrophobic binding pocket (i.e. ligand targeted) on a protein can compete for binding against candidate ligands used in HTS, providing an alternative readout to nonspecific fluorophore-dependent measurements to determine thermal stabilization by ligands in DSF-based assays.^{[2, 14]} Finally, ligand-targeted ESFs enable validation of competitive binding at a protein binding site by observed decreases in fluorescence in the presence of the ligand.

Selectivity of the ESF probe for an enzyme active site can be achieved by incorporating key features of the cognate substrate scaffold. In this report, we describe ESF-derivatized nucleosides as probes for HTS of enzymes that use sugar nucleotides as substrates including the plethora of glycosyltransferases (GTs) that are implicated in complex glycan biosynthesis (Figure 1). First, we present the synthesis and screening of four nucleoside analogs as probes for HTS. Then, we exemplify the application of ESF-conjugated nucleoside probes in an HTS assay of the N-terminal GT domain of TcdB. TcdB is a polyprotein produced by Clostridium difficile, a Gram-positive gastrointestinal pathogen that is currently a leading cause of hospital-acquired infections in developed countries.^[15] TcdB is a primary virulence factor for C. difficile, and is a member of a family of glucosylating toxins that inactivate small GTPases responsible for eukaryotic signaling pathways.^[16] TcdB is a potential pharmacological target for inhibition to prevent the host target glucosylation that is associated with bacterial virulence,^[17] and because of the current medical relevance of C. difficile.^[18]

Figure 1.

A) Probes synthesized for this study. Probes are based on a synthetically-accessible C5’ amino-uridine analogue coupled to DMAP and DMN ESFs via varied linkers. B) Application of tailored uridine-based probes for HTS of enzymes selective for uridine-based enzyme substrates.

The ESFs employed in the current studies are based on N,N-4-dimethylamino-phthalimide (DMAP) and N,N-4-dimethylamino-naphthalimide (DMN) derivatives.^[19] These imides exhibit minimal fluorescence in polar aqueous environments – thus fluorescence is “turned off” in an unbound state – and enhanced fluorescence in hydrophobic environments (e.g. protein binding sites) – thus fluorescence is “turned on” in a bound state.^[20] These fluorophores also show a small (40–80 nm) hypsochromic shifts.^[19] DMAP and DMN are ideal fluorophores for substrate analogue preparation due to their relatively small size, which is less likely to perturb a natural substrate scaffold.^{[10a, 21]} Additionally, synthesis can be accomplished while simultaneously minimizing the overall effort by utilizing standard coupling methods, such as amide coupling or Cu(I)-catalyzed azide-alkyne cycloaddition (Cu(I)AAC),^[22] and a common nucleoside building block - in this case a uridine analog, with a primary amine handle at the ribose C-5’.The general approach described herein should be readily adaptable to diverse enzymes that act on nucleoside-based substrates and should be compatible with HTS discovery efforts.

Results

Uridine diphosphate glucose (UDP-Glc) is the glucosyl donor substrate for the TcdB GT (K_m 21 μM),^[23] therefore we chose a uridine core as the key structural feature for active-site recognition and conjugated both 4-DMAP^[21a] and 4-DMN^[21b] ESFs via varied linkers to screen for binding and fluorescence. The ESFs were appended to the ribose C-5’ of uridine, using the modified amino-uridine derivative (4), by standard amide coupling methods or Cu(I)AAC (Scheme 1) to afford probes 5 through 8. When conjugated, each of the probes exhibits comparable fluorescence properties to those of the unmodified dyes in water or buffer (data not shown).

Scheme 1.

Syntheses of ESF probes. Reagents and conditions: (a) Acetone, sulfuric acid; (b) p-Toluenesulfonyl chloride, 4-(dimethylamino)pyridine, pyridine, dichloromethane; (c) Sodium azide, 40 °C, DMF; (d) Pd(OH)-C, isopropyl alcohol, H₂O, formic acid; (e) N,N-Diisopropylethylamine, hydroxybenzotriazole, HBTU, DMF, 4-(N,N-dimethylamino)phthalic anhydride; (f) Trifluoroacetic acid, MeOH; (g) CuSO₄, tris(3-hydroxypropyltriazolylmethyl)amine, sodium ascorbate, 7: compound 10, 8: compound 11; (h) i. Fmoc-Gly-OH, N,N-diisopropylethylamine, HBTU, DMF, ii. 20% piperidine in DMF; (i) 4-(N,N-Dimethylamino)phthalic anhydride, dioxane, reflux, 10: Propargyl amine, 11: 1-Amino-but-3-yne.

For the presented assay, the best probe for the TcdB GT was 5; it showed a low background signal in buffer and the fluorescence signal upon binding to the TcdB GT was linear across the μM concentration range (Figure S2). As determined by fluorescence titration, probe 5 bound to the TcdB GT with a K_D of 7.2 ± 1.2 μM and, at saturation showed a >30-fold increase in fluorescence at 512 nm relative to the unbound state (Figure 2). In general, we anticipate that signal changes with the ESF-based probes above 5-fold would be sufficient for general adaptation to HTS assays as this would provide statistically robust data for analysis. Addition of competing UDP-Glc reduced fluorescence to unbound levels showing that binding of probe 5 is both competitive and reversible (Figure S3). Probe displacement with UDP-Glc also serves as an ideal positive control for further assay development towards HTS applications.

Figure 2.

A) Fluorescence spectra showing titration of probe 5 with 20 μM TcdB GT. B) Binding of 5 with 20 μM TcdB GT measured at 518 nm.

Probes 6 - 8 did not signal the TcdB GT, however, binding to other select GTs and phosphoglycosyl transferases (PGTs) was observed (Figure S1). All fluorescence studies included TWEEN-20 (0.09%) to minimize non-specific binding.^[6] Full experimental details for fluorescence studies can be found in the Supplementary Information.

Towards the development of an HTS method, the ESF probe assay was transferred to a multiwell-plate reader format; volume was scaled to 20 μL and optimal working concentrations for the TcdB GT and 5 were found to be 10 μM each. The Maybridge Ro3 1000 compound diversity library, which includes stock solutions of fragments dissolved in DMSO to a final concentration of 100 mM, was used for the screen. The screen was carried out at 5 mM final fragment concentration, which is typical in fragment screening.^[24] The working assay solution, with the ESF probe and the fragments (at 5 mM), included DMSO at 5.5% of the final volume.

Each microwell plate in the HTS included analyses of the following samples: background fragment fluorescence, positive controls, negative controls, and maximum expected signal control. At high concentrations many fragments are fluorescent, so measurement of the fragment fluorescence in assay buffer with no added enzyme was used as the background. This background was subtracted from the fluorescence of the working assay, which included the ESF probe, enzyme, fragment, and assay buffer. Additionally, UDP-Glc (at 1 mM and 100 μM) was used as a positive control for probe displacement on each microwell plate of the HTS. Blank wells containing assay buffer and 5.5% DMSO were added to each plate as a negative control. Lastly, wells with probe 5, the TcdB GT and 5.5% DMSO in the absence of fragment were included to provide the maximum signal for ratio-based calculations.

Hits were measured as the relative fluorescence ratio of background-corrected displacement signal to maximum signal (Eq. S1). The distribution of percent probe displaced (proportional to fragment binding) from the 1000-member library showed 34 fragments with a signal equal to, or more than 60% of the maximum observable signal (Figure S4). In HTS compound screening assays, validation of apparent hits is critical because false positives can arise from experimental artefacts, such as those due to fragment insolubility and fluorescence quenching.^[8] In this case, the selection of 34 apparent hits was reduced to 24 hits after 10 fragments were determined to be false positives due to fluorescence quenching. In addition to the 24 apparent hits, five additional fragments were randomly selected as negative controls for further validation studies. The 24 apparent hits are numbered 1–24 and the additional five negative controls are numbered 25-29. The table also includes DMSO at 5.5% as a blank and additional negative control, numbered 30 (Table 1).

Table 1.

Results of ESF probe (5) binding, DSF analysis and biochemical assay determination for 29 fragments

#	% probe displaced	probe displaced	Δ Tm	Tm hits	Inhibition at 800 μM	#	% probe displaced	probe displaced	Δ Tm	Tm hits	Inhibition at 800 μM
1	61	yes	NC	NC	yes	16	78	yes	−3.5	yes	no
2	97	yes	NC	NC	NC	17	98	yes	NC	NC	NC
3	90	yes	−0.2	no	yes	18	80	yes	NC	NC	NC
4	60	yes	NC	NC	NC	19	88	yes	NC	NC	yes
5	81	yes	1.0	yes	no	20	79	yes	−3.8	yes	no
6	82	yes	NC	NC	NC	21	76	yes	0.3	no	no
7	90	yes	0.3	no	NC	22	95	yes	1.2	yes	no
8	66	yes	−2.5	yes	no	23	94	yes	0.5	no	NC
9	61	yes	NC	NC	yes	24	68	yes	1.2	yes	NC
10	93	yes	1.8	yes	NC	25	10	no	0.2	no	no
11	87	yes	NC	NC	yes	26	85[b]	no	−2.4	yes	no
12	61	yes	NC	NC	no	27	10	no	0.1	no	no
13	83	yes	NC	NC	no	28	9	no	0.3	no	no
14	74	yes	0.8	no	no	29	14	no	0.7	no	no
15	95	yes	−0.7	no	NC	30[c]	0	no	40.5[c]	no	no

^[a]NC: not compatible with DSF or biochemical assay at concentration of probe employed.

^[b]Concentration dependence validation showed this fragment to be a false positive as a fluorescence quencher in the ESF assay.

^[c]Control with equivalent volume of DMSO.

^[d]T_m for TcdB.

Hits and negative controls (1-30) from the ESF probe assay were tested with a DSF assay using SYPRO® Orange. In DSF, the protein T_m is the half maximum fluorescence signal found at complete protein denaturation. Samples exhibit a shifted T_m if stabilized or destabilized in the presence of a screening compound, calculated as ΔT_m.^[8] Table 1 includes the ΔT_m (determined in triplicate) of TcdB in the presence of fragments 1-29. The T_m of TcdB with an equivalent amount of DMSO, but without a fragment, was used as a control. Typically, fragments that stabilize the native conformation of a protein increase the T_m. However, some fragment binders may destabilize the native conformation upon binding to a non-native state of the enzyme.^[25] Such fragments are still considered apparent hits, and are subject to further evaluation. DSF analyses may be complicated by interference from highly fluorescent fragments.^[4] In this case the ΔT_m values cannot be determined and are thus recorded as not compatible (ΔT_m: NC) with the DSF assay. A total of 11 fragments, of the 29 studied (Table 1), were deemed not compatible with DSF analysis, underscoring the need for alternative approaches for HTS of fragments and other weak binders. Thus, in each of these cases, the ESF probe assay was able to provide a read-out for fragments that would have been eliminated from the HTS based solely on DSF incompatibility.

The 29 fragments were also subjected a biochemical activity assay. For this, we applied the Promega UDP-Glo^® assay, which detects UDP release over the course of a GT reaction. Because this is a multienzyme-based detection assay, an equivalent high concentration of fragment to that used in ESF and DSF analyses was not compatible due to excessive off-target inhibition of assay enzymes. This is a well-known complication of fragment screening by activity analysis due to the high concentrations (typically 5 mM) that are necessary for hit identification in initial steps.^[6] The activity of TcdB GT was tested in the presence of fragments at 800 μM. Fragments are typically weak binders, so high GT inhibition is not anticipated (Figure S5). Off-target inhibition of the Glo® reagent enzymes by the fragments was also tested and corrected for, however, in several cases we noted off-target inhibition of the Glo® assay at over 50% and these cases were deemed not compatible (NC) with the fragment assay (for complete error analysis see SI). Of the 29 fragments, 10 were found to be NC with the biochemical activity assay. This incompatibility rate is comparable that observed with the DSF assay and again underscores the value of the ESF probe-based approach. Structures of the fragment hits are available in the “Supplemental Information” and can now be further elaborated for structure-activity relationship (SAR) studies towards inhibitor development for TcdB (Figure S6).

In summary, although some of the hits from the ESF probe screen of the TcdB GT can be further validated using DSF or biochemical assay-based approaches, the significant rates of assay incompatibility with the high concentrations required for initial fragment-based screening highlights the challenges. From the 24 ESF-based leads, 30% were validated as hits by DSF and, an additional 46% could not be analyzed by DSF due to fluorescence interference from the fragments. The enzyme-based biochemical assay also showed high incompatibility; 42% of the fragments showed excessive background inhibition of the assay itself (>50% even at 800 μM) and in this case only 20% of the ESF hits could be supported by activity analysis in the presence of 800 μM fragment. These percentages serve to highlight the potential of the ESF probes to address the challenges of low affinity fragments, a challenge that has been documented between various HTS assays.^{[6, 26]}

Conclusions

We have described a robust ESF-based assay for enzymes that act on nucleoside-based substrates such as the ubiquitous NDP-sugars that are extensively exploited as glycosyl donor substrates in the assembly of complex glycan structures and for post-translational protein modification. The exemplar assay uses a readily-accessible nucleoside analogue, which is modified with the environment-sensitive DMAP and shows a K_D of 7.2 ± 1.2 μM with a >30-fold increase in fluorescence signal upon binding to the active site of the TcdB GT domain from the C. difficile TcdB toxin. These properties make this probe highly amenable to HTS. The use of the probe has been demonstrated in a fragment-based lead discovery screen using the 1000-member Ro3 Maybridge library; the ESF probe assay revealed 24 leads for further development.

ESF nucleoside analogues, such as those presented in this report, can be applied to a diverse range of enzymes that act on nucleoside-based substrates. A subset of these enzymes includes glycosyltransferases and phosphoglycosyl transferases involved in the glycosylation pathways of bacterial pathogens that use NDP-sugars as substrates.^[27] Many of these glycosylation-related enzymes are potential antibiotic drug targets due to the key roles played in bacterial glycoconjugate biosynthesis. The strategy for nucleoside-based ESF probe development provides valuable reagents for HTS applied to inhibitor development towards these nucleoside-accepting enzymes.