Pseudonatural Products for Chemical Biology and Drug Discovery

Abstract

Natural product (NP) structures have provided invaluable inspiration for the discovery of bioactive compound discovery. In the pseudonatural product (PNP) concept unprecedented combinations of NP fragments combine the biological relevance of NPs with exploration of wider chemical space by fragment-based design. We describe the principles underlying the PNP design and discovery of selected PNPs with unexpected or novel bioactivity. Cheminformatic analyses of ChEMBL 32, the Enamine screening library, phase 1-3 clinical compounds, and approved drugs reveal that ca. 1/3 of historically developed biologically active compounds and of currently commercially available screening compounds are PNPs, and that PNPs are increasing over time. PNPs are 54% more likely to be found in clinical compounds versus nonclinical compounds, and 67% of recent clinical compounds are PNPs. 63% of the core scaffolds in recent clinical compounds are made up of just 176 NP fragments, which suggests that PNPs open up a multitude of unexploited opportunities for drug discovery.

1

Principles of biologically oriented synthesis (BIOS) and PNP design. ,

2

PNP scaffold design principles: a) NP fragments can be combined in various connectivity patterns, utilizing a fusion edge, spiro, or bridge with shared atoms. b) Combinations of NP fragments can involve different connection types with intervening atoms, including mono-, bi-, or tripodal connections. c) Fragment combinations can also include both connectivity patterns, such as bridged bipodal or bridge tripodal connections.

3

General PNP design principles. Pseudo NPs are generated via a combination of different connectivity patterns, varying connectivity points while maintaining the connectivity pattern, or through the combination of more than two fragments.

4

Combination of complementary design principles. a) Combination of the PNP and the CtD approaches yields diverse pseudosesquiterpenoid alkaloids. b) Combination of the PNP logic with the BIOS. c) Combination of DOS with the PNP logic giving diverse pseudonatural products (dPNPs). Structures 34, 35, 37, 38, and 39 as well as class 5 and class 8 contain an edge fusion of an aromatic and an aliphatic ring, and we consider them PNPs. (a–c) Adapted and modified from Liu et al., Aoyama et al., and Bag et al., − under CC-BY4.0 and CC-BY-NC-ND 4.0.

5

Design, synthesis, and biological evaluation of indomorphan PNPs. a) Design of indomorphan-PNPs combining the NP-fragments of morphan and indole. b) Synthesis of indomorphans. c) Concentration dependent 2DG uptake in the presence of different glupin stereoisomers. d) Cellular thermal shift assay (CETSA) to determine stabilization of glucose transporters by Glupin. e) Inhibition of glucose uptake and cancer cell growth through inhibition of GLUT-1/3 by Glupin. (c–e) Adapted from Ceballos et al. under CC-BY 4.0.

6

Design, synthesis, and biological evaluation of pyrroquinoline PNPs. a) Design of PQ PNPs 53 by combining NP fragments, pyrrolidine of kainic acid, and tetrahydroquinoline of virantmicin. PNPs with structure 50 contain an edge fusion of an aromatic ring and an aliphatic ring. b) Synthesis of PQ utilizing Lewis acid catalyzed Povarov reaction. c) Alpl gene expression declines in the presence of 1.5 μM purmorphamine (antagonist of Hh signaling pathway) and increasing Tafbromin (cis-diastereomer of 54) concentrations. d) BROMOscan panel profiling for isomeric mixture 54 at 10 μM. Data are percent inhibition of tracer binding to the respective bromodomain. e) Tafbromin exerts its inhibitory effect on osteoblast differentiation by selectively binding to BD2 of TAF1. This interaction reduces the expression of Hedgehog target genes Gli1 and Ptch1 and the osteogenic marker Alpl. Consequently, osteogenesis is suppressed, overcoming the inhibition of ribosome biogenesis. Figure 6e and caption 6d adapted from Patil et al. CC-BY-NC-ND 4.0.

7

Design, synthesis, and biological evaluation of pyrrolidino-myrtanol PNPs. a) Design of pyrrolidino-myrtanol PNPs. b) Synthetic route to iDegs 62 utilizing the [3 + 2] cycloaddition. c) Structure of screening hit iDeg-1 and IC₅₀ value in the Kyn assay in BxPC3 cells. d) iDeg-6 Kyn screening with corresponding result from the Kyn assay. e) Reduction of IDO1 protein levels through iDeg-6. f) Regulation of IDO1 by heme. Figure 7e–f adapted from Hennes et al. under CC-BY-ND 4.0.

8

Design, synthesis, and biological evaluation of piperidinathiazole PNPs. a) Design of piperidinothiazoles. PNPs with structure 67 contain an edge fusion of an aromatic and an aliphatic ring. b) Synthetic route toward amino piperidinothiazoles. c) Role of GRAMD1A and cholesterol in autophagosome formation: enrichment of PtdIns3P at the autophagosome initiation sites also increases levels of GRAMD1A near the endoplasmic reticulum (ER). GRAMD1A mediates cholesterol transport between the organelles and, therefore, promotes autophagosome biogenesis. Figure 8c adapted from Wu and Waldmann.

9

Design, synthesis, and biological evaluation of pyrrolidino- and pyrroline succinimide PNPs. a) Design of diverse pyrrolidino-succinimide PNPs. b) Synthetic route toward various polypyrrolidines. c) Target identification of Rhonin with samples 88a and 88b and affinity probes for the pull-down experiment 89a and 89b. d) Proposed mode of action for Rhonin 88b. e) Downregulation of Gli1 and Ptch1. f) Affinity-based enrichment of RHOGDI1 by affinity probe 89a is active compared to inactive 89b. Figure 9d–g modified and adapted from Akbarzadeh et al. under CC-BY 4.0.

10

Classes of pseudonatural products derived from ketone derivatives of fragment-sized NPs. PNPs QN-I, QD-I, and GF-I contain an edge fusion of an aromatic and an aliphatic ring. Figure 10 modified from Grigalunas et al. under CC-BY 4.0.

11

Tanimoto similarities of Morgan fingerprints of the PNP library. a) Intrasubclass comparisons. b) Intersubclass comparisons. c) PCA of indole containing PNPs. d) PCA of chromanone containing PNPs. Figure 11a–d adapted from Grigalunas et al. under CC-BY 4.0.

12

Most common NP fragments and fragment combinations identified in the ChEMBL v32 data set. a) Distribution of PNP status for the ChEMBL v32 data set, deduplicated by InChIKeys (y-axis scale in millions of compounds, 1e6). b-d) Data for PNP compounds, only. b) Distribution of most common NP fragment connection types in percent; cm: connection monopodal, fe: fusion edge, fb: fusion bridge, fs: fusion spiro, cbe: connection bipodal edge. c) Most common NP fragments, with percent occurrence in all PNP NP fragment combinations. d) Most common NP fragment combinations, with connection type and percent occurrence (see b for an explanation of connection types). Figure 12 and caption adapted from Pahl et al. under CC-BY 4.0.

13

Most common NP fragments and fragment combinations were identified in the Enamine screening library. a) Distribution of PNP status for the full Enamine data set (y-axis scale in millions of compounds, 1e6). b–d) Data for PNP compounds, only. b) Distribution of most common NP fragment connection types in percent; cm: connection monopodal, fe: fusion edge, fs: fusion spiro, fb: fusion bridge, cbe: connection bipodal edge. c) Most common NP fragments, with percent occurrence in all PNP NP fragment combinations. d) Most common NP fragment combinations, with connection type and percent occurrence (see b for explanation of connection types). Figure 13 and caption adapted from Pahl et al. under CC-BY 4.0.

14

a) Ring-type density (ring count normalized to number per 100 heavy atoms), b) aromatic/aliphatic compound type, and c) % PNP in clinical compounds (phase 1–3 and marketed drugs) from the ChEMBL v32 data set over time. Publication date is the first appearance in Scifinder. ^aAromatic/aliphatic classes: aliphatic = 0 aromatic rings; carboaromatic = ≥1 phenyl rings and 0 heteroaromatic rings; hetero/carboaromatic = ≥1 phenyl rings and ≥1 heteroaromatic rings; heteroaromatic = ≥1 heteroaromatic rings and 0 phenyl rings.

15

Distribution by publication period, pre- and post-1990, of PNP, NonPNP, NP, and NPL clinical compounds in property space defined by t-distributed stochastic neighbor embedding (t-SNE). 3-D coordinates derived from MW, ALogP, cx_LogD, HBA, HBD, PSA, RotB, Fsp3, stereocenters, normalized special score (nSPS), carboaliphatic rings, carboaromatic rings, heteroaliphatic rings, heteroaromatic rings, and aromatic_N atoms. t-SNE calculation was done using DataWarrior. Clinical compound publication dates are from Scifinder.

16

PNP_Status (defined in Section ) vs publication date. a) For clinical compounds phases 1–3 and marketed drugs and b) reference compounds acting at the same biological targets as the clinical compounds. c) % PNP in clinical compounds, reference compounds by target, and the [clinical-reference] difference; target class clinical compound numbers in parentheses. Figure adapted from Heinzke et al. under CC-BY 4.0.

17

58 most abundant post-2008 clinical NP fragments comprised 90.5% of total clinical NP fragment count. Shown for each fragment are count (% clinical fragments) (top); odds ratio vs reference compounds (lower left); p value (lower right; NS = not significant, p > 0.05). Clinical abundance increased (15, 26%): green, p < 0.01; blue, p = 0.01–0.05. Clinical abundance decreased (4, 7%): red (pink, p < 0.05). Canonical tautomers shown, as generated by RDKit. Figure 17 adapted from Heinzke et al. under CC-BY 4.0.

18

Overall changes in PNP_Status (a1, b1) and paired PNP_Status changes (a2, b2) seen in compilations of hit-to-candidate optimizations (a) and fragment-to-lead optimizations (b). In a), mean NP fragment counts are 2.05 for hits and 2.60 for candidates (p < 0.0001); mean NP likeness values are −0.90 for hits and −0.89 for candidates (not different). In b), mean NP fragment counts are 1.25 for fragments and 2.37 for leads (p < 0.0001); mean NP likeness values are −0.93 for hits and −0.92 for leads (not different).

19

Bicyclic NP fragments that were absent or rarely seen in the 2008 onward literature, comprising clinical compounds and those reference compounds acting at the biological targets accounting for clinical efficacy. The counts of fragment appearances in reference and clinical compounds are shown. Fragment numbering is taken from the full list (available in Supporting Information in ref. ).