Integrated Synthesis and Testing of Substituted Xanthine Based DPP4 Inhibitors: Application to Drug Discovery

Abstract

A novel integrated discovery platform has been used to synthesize and biologically assay a series of xanthine-derived dipeptidyl peptidase 4 (DPP4) antagonists. Design, synthesis, purification, quantitation, dilution, and bioassay have all been fully integrated to allow continuous automated operation. The system has been validated against a set of known DPP4 inhibitors and shown to give excellent correlation between traditional medicinal chemistry generated biological data and platform data. Each iterative loop of synthesis through biological assay took two hours in total, demonstrating rapid iterative structure–activity relationship generation.

Medicinal chemistry is an iterative process of synthesis and screening of novel molecules to optimize both biological and physical properties to ultimately yield a candidate molecule for the clinic. We have previously reported¹ a closed loop integrated approach to elements of medicinal chemistry whereby the synthesis process is fully integrated with the biological assay enabling the rapid and automated generation of structure–activity relationship (SAR) data. This approach enables a true closed loop approach to medicinal chemistry to be undertaken where molecules are designed for synthesis and screening based upon the emerging SAR in a true serial iterative manner.

Data published in this letter demonstrates both the reproducibility and consistency of automated SAR generation for a series of compounds with known activity against dipeptidyl peptidase 4 (DPP4).

Several groups have implemented online screening of drug-like compounds. For example, Kool et al. used online bioaffinity analysis to screen a library of fragments for activity against acetylcholine binding protein, and Guetzoyan et al. used frontal affinity chromatography to screen a range of γ-aminobutyric acid (GABA) agonists synthesized in flow.^2,2b In addition, the opportunity to improve the speed, efficiency, and quality of medicinal chemistry hypothesis testing by the direct integration of compound synthesis with biology and design algorithms has been recognized by several groups.^3,3b The integrated platform described in this letter is designed to select a compound from a drug-like chemical space and synthesize, purify, and biologically test the compound. The SAR generated is then used to select the next compound for synthesis and test. This offers the advantage of fully automated, fast serial generation of SAR for drug discovery with each design iteration taking no more than two hours from the start of synthesis to biological readout.

The integrated optimization platform (Figure 1) comprises a reagent autosampler and flow synthesis apparatus connected to a commercial high-performance liquid chromatography (HPLC) mass spectrometer. Liquid chromatography mass spectrometry (LC–MS) is used to purify and characterize synthesized material with an evaporative light-scattering detector (ELSD) to establish sample concentration.⁴ Purified material is subsequently reformatted to the correct concentration for biological assay using proprietary hardware. Biological IC₅₀ determination is achieved using a bespoke liquid handler coupled to a fluorescence plate reader. The IC₅₀ is determined by fitting a curve to the measured data using in house software written in Matlab.⁵ The integrated optimization platform is controlled from a workstation using an in-house designed software package.

Figure 1.

Cyclofluidic optimization platform.

In order to further validate the integrated synthesis and screening approach, a joint project was undertaken to replicate an existing data set in a blinded experiment. Sanofi-Aventis provided a series of compounds with known inhibitory activity against DPP4^6−6b for evaluation on the platform. The SAR generated would then be unblinded by Sanofi-Aventis to provide validation of the fully integrated approach.

The compounds to be synthesized are shown in Scheme 1. The first step of the synthesis used either boc-protected diamines or amino alcohols to displace 8-bromoxanthines. The in situ generated boc-protected intermediates were then deprotected to give the compound of interest for testing.

Scheme 1. General Synthesis of DPP4 Inhibitors.

The first evaluation was to establish consistency between cycles by repeatedly synthesizing and screening a single compound 3 (Scheme 2) ten times on the integrated platform. For each individual compound synthesis, the sample was assayed twice in alternate assay plate columns against human DPP4 to assess variance in the assay. By way of control, a discrete solid sample was assayed manually to provide comparator data.

Scheme 2. Synthesis of Diamino DPP4 Inhibitors.

The data from the experiment is displayed in Table 1, and Figure 2 clearly indicates that, with over 20 hours and ten cycles of both synthesis and screening on the platform, the data is consistent within normal error; the activity for a solid sample was 106 nM. This clearly illustrates that as well as giving very little variance, the IC₅₀ is in complete agreement with the discrete sample determined manually.

Table 1. Repeated Synthesis and Assay of Compound 3.

loop number	DPP4 (human) nM column 1	DPP4 (human) nM column 2
1	76	115
2	84	104
3	86	83
4	120	110
5	103	72
6	76	128
7	82	105
8	166	158
9	122	131
10	124	128

Figure 2.

Graph of IC₅₀ results for compound 3.

An example IC₅₀ curve for compound 3 generated during the run is shown in Figure 3 showing excellent fit.

Figure 3.

Example curve of human DPP4 inhibition.

With reproducibility now confirmed, a series of diamine modified xanthines were synthesized via a two-step synthetic protocol using a Vapourtec R4 flow chemistry system. An example reaction is shown in Scheme 2.

The first synthetic displacement was conducted on a 2 mL volume stainless steel coil heated at 150 °C with injections of 250 μL of the 8-bromosubstituted xanthine in N-methyl-2-pyrrolidone (NMP) and 250 μL solution of two equivalents of the desired boc-protected diamine in NMP. The flow rate for step 1 was 50 μL/min per pump to give a residence time of 20 min. Next, the boc-protected intermediate was mixed with 500 μL of 30% methanesulfonic acid in water added at 50 μL/min to deprotect the boc group. This reaction was conducted in a second 2 mL coil at 90 °C and the product subsequently run through a 20 μL injection loop into the LC–MS purification system switched at the point of maximum concentration. NMP was used as the system and wash solvent. A schematic of the setup is shown in Figure 4.

Figure 4.

Flow reactor setup for diamine reaction.

The total synthesis time per compound to purification was 50 min, and the total cycle time per sample was 120 min.

All 12 compounds were successfully synthesized and tested in 24 h total platform time. Yields varied between 3% and 38%. It should be noted that even low yields provided enough material for the compounds to be purified and assayed.

The results are displayed in Table 2.

Table 2. Results of Diamine Synthesis and Testing.

Next, the synthesis and testing of a series of amino alcohols was undertaken. An example synthetic route is shown in Scheme 3.

Scheme 3. Synthesis of Amino-Alcohol Modified Xanthines.

In this case, the initial reaction was very fast using KO^tBu as base and could be conducted in a 32 cm long 1 mm ID tube in ∼3 min. Subsequent deprotection of the boc group was again achieved using 30% methanesulfonic acid in water at 90 °C with yields for the two step process of between 0% and 23%. The overall synthesis success for this process was a respectable 82%.

A schematic of the reaction setup is shown in Figure 5.

Figure 5.

Reaction setup for amino-alcohol reaction.

The results from 21 iterations of the system are shown in Table 3.

Table 3. Results of Amino-Alcohol Synthesis and Testing.

^aSynthesis failed.

^bMass spec. failed to give expected molecular ion. Several of the compounds have repeated values to further demonstrate consistency.

Biological assays were carried out in 40 mM tris·HCl, pH 7.4 at 25 ± 2 °C, in a total assay volume of 50 μL using 384-well plates (Costar 3574). Routinely, compounds were titrated in assay buffer to which was added either enzyme (0.82 mU/mL porcine DPP4 (Sigma) or 34 U/mL human DPP4 (Sigma). Residual enzyme activity was monitored by adding substrate (H-Gly-Pro-AMC; Bachem) to give a final concentration equivalent to the K_M^app (K_M = 25 μM for human DPP4 and K_M^app = 35 μM for porcine DPP4). Data was analyzed by nonlinear regression using a four parameter logistic variable slope model for IC₅₀ determinations.

The integrated platform data was compared against Sanofi-Aventis data as shown in Table 4. The porcine data matched closely validating the technique (Figure 6). The historical human DPP4 data was limited, but where available, the platform data is also in agreement with previously reported values.

Table 4. Comparison of Sanofi-Aventis and Integrated Platform Data^a.

^a(P) = porcine data; (H) = human data.

Figure 6.

Correlation between Sanofi-Aventis and platform data.

Overall, the data obtained by the platform is fully consistent with reference data provided by Sanofi-Aventis using traditional means.

The synthesis was generally robust, and the integrated approach enables biological data to be measured from low yielding reactions. This additionally minimizes the optimization time needed to achieve a reliable, generic, flow synthesis method and ensures results are obtained from reagents with a range of reactivity.

The biological assay has also been shown to be both robust and stable over an extended period of time, essential for an automated environment.

A plot of the porcine (closed circle) and human (open square) Sanofi-Aventis (log IC₅₀^SA) and platform (log IC₅₀^CYC) inhibition data presented in Table 4 is shown in Figure 6. The lines represent the best fit linear regression analysis of the data.

In conclusion, a series of DPP4 inhibitors have been successfully synthesized, purified, and tested using the platform with an overall chemistry success rate of 93%. Each compound was synthesized and tested in two hours total. It took less than three days to synthesize and test the total of 29 compounds. Close correlation between integrated platform data and data generated within the corresponding traditional medicinal chemistry approach has validated the technique (R² = 0.9253 and 0.9893 for porcine and human data, respectively; Figure 6). Further work on a series of synthetic routes and various target classes is currently underway.

Supporting Information Available

General synthetic and analytical purity determination procedures for all synthesized compounds. This material is available free of charge via the Internet at http://pubs.acs.org.

Cyclofluidic is grateful for support from the UK Technology Strategy Board, UCB, and Pfizer.

The authors declare no competing financial interest.