Synthesis and Demonstration of the Biological Relevance of sp3‐rich Scaffolds Distantly Related to Natural Product Frameworks

Abstract The productive exploration of chemical space is an enduring challenge in chemical biology and medicinal chemistry. Natural products are biologically relevant, and their frameworks have facilitated chemical tool and drug discovery. A “top‐down” synthetic approach is described that enabled a range of complex bridged intermediates to be converted with high step efficiency into 26 diverse sp3‐rich scaffolds. The scaffolds have local natural product‐like features, but are only distantly related to specific natural product frameworks. To assess biological relevance, a set of 52 fragments was prepared, and screened by high‐throughput crystallography against three targets from two protein families (ATAD2, BRD1 and JMJD2D). In each case, 3D fragment hits were identified that would serve as distinctive starting points for ligand discovery. This demonstrates that frameworks that are distantly related to natural products can facilitate discovery of new biologically relevant regions within chemical space.


Introduction
Small molecules continuet od ominate Man's ability to treat disease, and can transform our understanding of fundamental biology.Y et historically,t he exploration of chemical space has been highly uneven, [1,2] in part because an arrow toolkito fr eliable reactions has underpinned molecular discovery. [3] Natural products can facilitate the identification of biologically relevant chemicals pace [4] since they have arisen through the functiondrivene volution of biosynthetic pathways. [5] Indeed,a round a third of recent small moleculed rugs have been inspired by natural products. [6] In biology-oriented synthesis, [7] natural product frameworks [4] inform the design of productives mall molecule screening collections and fragments ets that can be exploited in the discoveryo fl igands for unrelated protein targets. [8,9] In addition, synthetic approaches have been developed to convert specific natural products into alternative complex frameworks. [10] Natural product-inspiredc ompounds can provide highly distinctive startingp oints ford iscovery that contrasts tarkly with most synthetic screening compounds: [11] in particular,t he typicallyh ighf raction of sp 3 -hybridised carbons (Fsp 3 )i sa ttractive since it correlates with the successful translation of clinical candidates. [12] We envisaged a" top-down" synthetic approach in whicha lternative complex, yet readily accessible, intermediates would be convertedi nto many natural product-like scaffolds (Scheme 1). Initially, bridged scaffolds of general structure 2 wouldb ep reparedu sing intramolecular [5+ +2] cycloaddition reactions [13,14] (e.g. 1!2): ring cleavage( red;f or example, 2! 3), ring expansion (magenta;f or example, 2!4), annulation (blue;f or example, 2!5), or functional group modification (green;f or example, 2!6)w ould then yield diverse scaffolds. The approachc ontrastsw ith diversity-oriented strategies [15,16] in which building blocks are prepared ("built") and linked ("coupled") to yield intermediates which are then cyclised ("paired") to yield alternative scaffolds. Although diversity-oriented approaches to sp 3 -rich fragments have been devel-oped, [17] their biological relevance has only rarely [17b] been demonstrated. To undertake ap reliminary assessment of biological relevance, we have therefore prepared and screened a fragment set that is based on many of the scaffolds that are accessible using our synthetic approach.
An hierarchical tree was constructed to capture the relationship between the scaffolds ( Figure 1). Here, scaffolds were systematically simplified using an established protocol by removal of rings until ap arentm onocyclic ring system was ultimately obtained. [21] Twenty two different graph-node-bond frameworks (capturing atom and bond type) were represented, which were then simplified to give nine parent monocycles. The scaffolds are based on awide range of parent ring systems and there is thus significant diversity at each level of hierarchy of the scaffold tree. The exploitation of several different complex intermediates 2 was critical to the diversitythat was possible, fore xample by enabling variation of regiochemistry (e.g. 4/15, 8/16 and 10/14)a nd heteroatom position and identity (e.g. 11/17). Such an approachw ould unlikely be possible by modification of natural products, since several relateds tarting materials would be required in multi-gram quantities.
To compare with other screenings ets,w ed etermined natural product likenesss cores [11] for the deprotected scaffolds,a natural product screening library (4,460 compounds) and a commercial screening collection (278,365 largely synthetic compounds) (Figure 2, Panel A). The distribution of the scores for the scaffolds was broadly similart ot hat of the natural product screening libraryb ut highly distinctive from that of the large screening collection. The local structuralf eatures of our scaffoldsa re thus reminiscent of those found in natural products.D espite the high naturalp roduct likeness,h owever, only one of the 22 graph-node-bond frameworks is actually a Scheme1.Overview of our unified approach to diverse natural product-like scaffolds. Intramolecular [5+ +2] cycloadditionw ould yield alternative complex intermediates (e.g. 2)that would be converted into diverse scaffolds by ring cleavage (red;f or example, !3), ring expansion (magenta;for example, !4), annulation (blue;for example, !5)ora ddition/modification( green; for example, !6).
Scheme2.Synthesis of complex intermediatecycloadducts. Chem. Eur.J.2017, 23,15227 -15232 www.chemeurj.org 2017 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim substructure in the roughly 281,000c ompounds in the Dictionary of Natural Products [22] (Figure 1a nd Supporting Information). Indeed, significant simplification to mono-or bicyclic frameworks is required before sub-structures of naturalp roducts are found. Our scaffolds thus have high natural product likeness,b ut their frameworks lie in branches that augment the scaffold trees of natural product frameworks.

Synthesis of afragment set and high-throughput crystallographicscreens
To enablep reliminary assessment of biological relevance, we prepared 52 racemic fragments based on 23 of the scaffolds. Here, af ragment-baseda pproach was exploited to enable efficient exploration of chemical space accessible using our synthetic approach. The set was designed to have high shaped iversity,a nd to comprise fragments with controlled [23] size (13 to 19 heavy atoms) and lipophilicity (À1.5 < clogP < 3) (Supporting Information). The fragment set was significantly more three-dimensional (Supporting Information), [24] and more natural product-like ( Figure 2, Panel B), than commerciallya vailable fragments with the same heavy atom range.
The fragment set was screened against three protein targets from two different mechanisticc lasses involved in epigenetic biology:t he ATAD2 and BRD1 (also known as BRPF2) bromodomains, and the histone demethylase JMJD2D( also known as KDM4D). Here, the objective was to perform ap reliminarya ssessment of biological relevance rather than to provides pecific startingp oints ford iscovery.H igh expression levels of ATAD2 [25] and JMJD2 family members [26] correlate with poor outcomes in several cancers, whilst BRD1 is am ember of the BRPF familyo fs caffolding proteins whose role in acute myeloid leukemiai sn ow emerging. [27] The two bromodomains are contrasting targets:A TAD2 hasashallow N-acetyl lysine binding site and has been suggested to have particularly low druggability. [28] The three target proteins were all amenable to highthroughput protein crystallography.P rotein crystalsw ere soakedw ith the 52 individual racemic fragments, [29] picked and then subjected to automated X-ray diffraction. Fragment hits were identified through detection of additional electron density, [30] and inspection for polar interactions with the protein. Fragmenth its were successfully identified for each of the target proteins:t wo hits for JMJD2D, eight hits for the BRD1 bromodomain and seven hits for the ATAD2 bromodomain ( Figure 3a nd Supporting Information).
For the BRD1 bromodomain, the eight fragment hits targeted the N-acetyl lysine binding site (Panel B2, Figure 3a nd Supporting Information;s ee Ref. [27] and references therein for known ligands). The fragment hits contained severald ifferent N-acetyl mimetics that interacted with both N110 and, either directly or via ab ridging water,Y67.
For the ATAD2 bromodomain, the fragment screen yielded seven hits (based on six distinctf rameworks) that targeted the N-acetyl lysine binding site (Panel B3, Figure 3a nd Supporting Information). Six of the hits mirroredt he binding mode of the N-acetyl lysine side chain, [28a] interacting directly with N1064 and, via ab ridging water,w ith Y1021. Four hits were in common with BRD1 although, in three cases, the interaction networks were more extensive:f or example ent-23 made water-mediated interactions with V1008 and D1014 (Panel B3, top, Figure 3) and ent-25 interacted directly with E1017 ( Figure 4).
To compare directly with am ore conventional fragment set, we also screened 700 commercially availablefragments against ATAD2 by high-throughput crystallography and identified nine hits that targeted N-acetyl lysine binding site (Figure 4a nd Supporting Information). As ag roup, the interactions of these fragments parallel those of other sp 2 -rich fragments:they interact directlyw ith N1064 and/or,v ia ab ridging water,w ith Y1021 but make few additional polar contacts. [25,28,32] For screens against many targets, flatter fragments have been observed to have higher hit rates. [33] With ATAD2, however,asignificantly higher hit rate was observed with our shape-diverse natural product-like fragments (7/52) than with more conventional [34] flatter fragment sets (9/700). This outcomei sc onsistent with the low hit rate observed in ap revious fragment screen by NMR (65/13800, subsequently triaged to yield 12 hits with K d < 1mm). [28b] Figure 1. Hierarchical scaffold tree. The circles representframeworks at the graph-node-bond level (22 frameworks represented in the 26 scaffolds prepared, outer ring and boxed;simplified frameworks, other circles). The 22 frameworksare relatedt on ine parent (monocyclic) frameworks (identified using an established protocol, ref. 21). At each level of hierarchy,o ccurrenceass ubstructures of natural products is indicated (green, not found;orange, found in < 1% of natural products;red, found in > 1% of natural products). Figure 2. Natural product likeness of scaffolds and fragments. Panel A: Natural product-likeness scores for the 26 scaffolds (black), 4,460 natural products (green)and ac ommercial screening collection (278,365 largely synthetic compounds, grey). Panel B: Naturalp roduct likeness scores for the 52 fragments prepared (black), 1,236 commercially-available fragments( grey) and 128 naturalp roduct-inspiredfragments (green). Compounds are binned into 0.5 unit bins.

Conclusion
We have developed a" top-down" synthetic approach in which alternative complex, yet readily accessible, intermediates were convertedi nto many diverse scaffolds. These scaffolds have local natural product-like features, but are only distantly related to specificn atural product frameworks. As et of 52 fragments based on 23 of the scaffolds was screeneda gainst three epigenetic targets from two distinct protein families. In each case, hits were obtained that may provide distinctive opportunities for subsequent fragment growth. We have therefore demonstrated that frameworks that are distantly related to natural products can facilitate identification of novel regions of biologically relevant chemical space.S ynthetic approaches to such frameworks may thus help identify fertile chemical space for bioactive small molecule discoveryt hat is inaccessible to existing compound collections and biosynthetic pathways.