Activity-Directed Synthesis with Intermolecular Reactions: Development of a Fragment into a Range of Androgen Receptor Agonists

Activity-directed synthesis (ADS), a novel discovery approach in which bioactive molecules emerge in parallel with associated syntheses, was exploited to develop a weakly binding fragment into novel androgen receptor agonists. Harnessing promiscuous intermolecular reactions of carbenoid compounds enabled highly efficient exploration of chemical space. Four substrates were prepared, yet exploited in 326 reactions to explore diverse chemical space; guided by bioactivity alone, the products of just nine of the reactions were purified to reveal diverse novel agonists with up to 125-fold improved activity. Remarkably, one agonist stemmed from a novel enantioselective transformation; this is the first time that an asymmetric reaction has been discovered solely on the basis of the biological activity of the product. It was shown that ADS is a significant addition to the lead generation toolkit, enabling the efficient and rapid discovery of novel, yet synthetically accessible, bioactive chemotypes.

The discovery of biologically active molecules typically involves the synthesis and testing of many compounds,e ach individually crafted to optimize the arrangement of functionality.S uch workflows tend to induce scientists to exploit al imited palette [1][2][3] of reliable chemical transformations.A s ad irect consequence,d esigned arrays comprise compounds that are readily prepared, tending to limit diversity and, potentially,tofocus on unproductive areas of chemical space.
We recently introduced activity-directed synthesis (ADS), [4] an ovel discovery approach in which bioactive small molecules emerge together with associated syntheses. ADS is iterative,b orrowing concepts from biosynthetic pathway evolution. [5] In each round, the components of diverse reaction arrays are widely varied;b ye xploiting reactions with many possible outcomes,d iverse chemical space is explored. After catalyst removal, the crude product mixtures are screened, and reactions that yield active products inform subsequent reaction array design. Only reactions that yield bioactive product mixtures are ever scaled up to enable the characterization and identification of the responsible products.A DS is function-driven, [6] focusing resources on reactions that yield bioactive products.
We recently harnessed intramolecular metal-catalyzed carbenoid reactions in the ADS of androgen receptor (AR) agonists.A DS drove the discovery of both novel ligandsbased on scaffolds with no annotated activity against the receptor-and associated high-yielding syntheses (Fig-ure 1A). However,i nt his validation work, reliance on intramolecular reactions meant that the structural diversity of possible products was largely encoded by the substrates used. We envisaged that ADS would be enhanced by exploiting intermolecular reactions as the range of possible reaction outcomes-and thus the chemical space exploredwould be dramatically increased. It was proposed to exploit ADS to drive the productive elaboration of the 4-cyano-3trifluoromethylphenylacetamide fragment found in many modulators of AR. [7] Although the core motif displayed only modest agonism (1;E C 50 = 92 AE 13 mm ;E C 50 = concentration of ligand needed to induce the half-maximal observed effect), we reasoned that intermolecular reactions of related diazo acetamides could help identify productive strategies for fragment growth ( Figure 1C). [8][9][10] The a-diazo amides 2-5 incorporate the 4-cyano-3-trifluoromethylphenyl fragment and bear agroup (N-methyl or N-cyclopropyl) expected to suppress intramolecular reactions. [11] In round one,weperformed an array of 192 reactions randomly chosen from 480 possible combinations of four substrates (2-5), ten co-substrates (6a-6i or no co-substrate), six catalysts,a nd two solvents (CH 2 Cl 2 or toluene;F igure 2A). Thec o-substrates were selected on the basis of diversity of possible intermolecular reactions with metal carbenoids [12][13][14] and the catalysts on the basis of their diverse reactivity.W ei nitially showed that the diazo substrates and Abstract: Activity-directed synthesis (ADS), anovel discovery approach in whichbioactive molecules emerge in parallel with associated syntheses,w as exploited to develop aw eakly binding fragment into novel androgen receptor agonists.H arnessing promiscuous intermolecular reactions of carbenoid compounds enabled highly efficient exploration of chemical space. Four substrates were prepared, yet exploited in 326 reactions to explore diverse chemical space;guided by bioactivity alone,the products of just nine of the reactions were purified to reveal diverse novel agonists with up to 125-fold improved activity. Remarkably,o ne agonist stemmed from an ovel enantioselective transformation;t his is the first time that an asymmetric reaction has been discovered solely on the basis of the biological activity of the product. It was shown that ADS is as ignificant addition to the lead generation toolkit, enabling the efficient and rapid discovery of novel, yet synthetically accessible,b ioactive chemotypes. Angewandte co-substrates were all inactive in our assay (see the Supporting Information). Thea rray was performed in 96-well plates with reactions involving ad iazo substrate (100 mm), ac osubstrate (1.0 m), and ac atalyst (1 mm). To demonstrate significant exploration of chemical space,w es howed that ar andom selection of reactions yielded several products, including,ing eneral, those of both inter-a nd intramolecular reactions (see the Supporting Information). After 48 h, the crude reaction mixtures were scavenged to remove metal contaminants,e vaporated, and assayed for agonism of AR (total concentration of products:1 0mm in 1% DMSO in pH 7.5 aqueous buffer) using an established assay. [15] Remarkably,o nly two of the 192 reactions yielded products that were significantly active ( Figure 2B;s ee also the Supporting Information): both reactions involved diazo substrate 3,[ Rh 2 {(S)-DOSP} 4 ], and CH 2 Cl 2 ,t ogether with either cyclohexene (6a)o ri ndole (6f)a st he co-substrate. These data suggest that the fate of 3-both with 6a and 6fdepends critically on the specific catalyst and solvent used, which determine whether the substrate is steered towards active products.T he lack of activity without ac o-substrate strongly suggests that the active products are derived from intermolecular reactions.A tt his stage,n op roducts were isolated or identified;i nstead, the most promising reactions informed subsequent reaction array design.
Round two focused on N-methyl diazo substrate 3 and its closely-related N-cyclopropyl variant 5,t he co-substrates cyclohexene (6a)a nd indole (6f)a sw ell as structurally related compounds,a nd rhodium carboxylate catalysts.T he 86 reactions were randomly chosen from 360 possible combinations of the two substrates,1 8c o-substrates,f ive catalysts,a nd two solvents (CH 2 Cl 2 or toluene;F igure 3A). To drive the development of efficient activity-directed syntheses,t he crude product mixtures were assayed at twofold lower total product concentration (5 mm). Five promising combinations of substrate and co-substrate were identified: N-methyl diazo substrate 3 with dihydronaphthalene (6m), dihydropyran (6o), or indene (6r), and N-cyclopropyl diazo substrate 5 with indole (6f)or7-azaindole (6n;see Figure 3B and the Supporting Information). All of these combinations were superior to the most promising combinations from round one.
In round three,substrate 3 was combined with all possible combinations of 12 co-substrates (6a,6f , 6m, 6n, 6o, 6r,and six structurally related compounds) and four catalysts in CH 2 Cl 2 ( Figure 4A). After scavenging and evaporating,t he crude product mixtures were assayed at five-fold lower total product concentration (1 mm)t oi ncrease the selection pressure ( Figure 4B). With co-substrate 6e' ',t he substrate yielded active product mixtures with all four catalysts,w ith the highest activity for [Rh 2 (OAc) 4 ]. However,with dihydropyran 6f' ',s ignificant activity was only observed when [Rh 2 {(R)-DOSP} 4 ]had been used;remarkably,nos ignificant activity was observed with the enantiomeric catalyst, suggesting that the most active product is chiral and that [Rh 2 {(R)-DOSP} 4 ]c atalyzes the formation of the more active enantiomer.
To understand the basis for the emergence of bioactive structures and associated synthetic routes,p rioritized reactions were scaled up from all three rounds of ADS.T he reactions were performed on fifty-fold larger scale;t he products were purified by column chromatography and their dose-dependent activities determined (Table 1). In all but one . .

Angewandte
Communications case,aproduct whose activity accounted for that observed in the original array was obtained in good (71-82 %) yield.
However,t he combination of 5, 6n,a nd [Rh 2 (OAc) 4 ]( from round two) gave the a,b-unsaturated g-lactam 9 in low (18 %) yield together with recovered starting material 5 (67 %): here, the observed activity must have stemmed from g-lactam 9 (EC 50 = 790 AE 60 nm)r ather than the recovered starting material 5 (EC 50 > 500 mm). Crucially,i ne ach case,w ew ere confident that the product that was responsible for the observed activity had been identified.
In round one,two reactions had yielded product mixtures with significant activity ( Figure 5). In each case,the bioactive product was formed in an intermolecular reaction. Amide 7 was the product of C À Hinsertion into the 3-position of indole (6f), whereas amide 8 was formed by cyclopropanation of cyclohexene (6a). In the other 190 reactions,t he product mixtures did not have biological activity that was detectable in the assay.Analysis of aselection of reactions in round one

Angewandte
Chemie had revealed that aw ide range of products had been produced:t hus,s ignificant chemical space had been explored-yet discarded-in the search for bioactive products.
In round two,the reaction array had been informed by the reactions that yielded amides 7 and 8.T hus,s uccess with cyclohexene (6a)l ed to the inclusion of dihydronaphthalene (6m), dihydropyran (6o), and indene (6r)a sa lternative cosubstrates.I ne ach case,s creening at lower total product concentrations (5 mm)h ad enabled the identification of alternative cyclopropanation products (10-12)w ith significantly higher biological activities than 8.I nterestingly,w ith substrate 5,t he introduction of indole (6f)o r7 -azaindole (6n)changed the outcome of the reaction. Rather than acting as co-substrates,these additives steered the reaction towards the rearranged product (9) [16,17] of an intramolecular C À H insertion:These reactions were discovered because product 9 is apotent partial agonist.
In round three,dihydropyran 6f' ' and benzopyran 6e' ' had been included as co-substrates on the basis of their similarity to dihydropyran (6o)a nd dihydronaphthalene (6m). The additional functional groups in both 6e' ' and 6f' ' offer opportunities for different types of intermolecular reactions. In both cases,t hese new possibilities were exploited in the formation of more bioactive products:amide 13,which is the product of an OÀHi nsertion, [18,19] and oxazole 14,w hich is formed by reaction with the nitrile moiety of 6e' '. [20] Our retrospective analysis revealed two mechanisms by which activity-directed syntheses can develop.F irst, by varying the specific catalyst and solvent used, the yield of the bioactive products may be optimized. As an example,i n round three,the activity of the product mixtures derived from substrate 3 and co-substrate 6e' ' varied widely.W ith [Rh 2 (OAc) 4 ]a st he catalyst, oxazole 14 (EC 50 = 730 AE 30 nm) was formed in 75 %y ield;h owever,w ith other catalysts,t he observed activity was lower, presumably as ar esult of 14 being produced in al ower yield. Thec hoice of total product concentration in the assay is crucial as,tooptimize the yield of abioactive product, it is important to screen at concentrations that do not saturate the target protein.
Second, ADS can be used to optimize the structure of bioactive products.F or example,t he cyclopropanes 10, 11, and 12,which were formed in round two,all had significantly higher activities than cyclopropane 8,w hich was formed in round one.F urthermore,i ntroducing reactants with new functional groups can offer new opportunities for forming more active products.F or example,amide 13 and oxazole 14 were both formed in types of reaction that were not possible in round two.Such possibilities can expand the chemical space that is explored during the optimization process,enabling the emergence of new synthetically accessible bioactive chemotypes.T he emergence of oxazole 14 was particularly interesting:S uch heterocycles are widely exploited as peptidomimetics in medicinal chemistry. [21][22][23][24] In future ADS campaigns, the inclusion of more functionalized substrates in round one may increase the diversity of the explored chemical space from the outset. It remains to be seen whether other promiscuous reaction classes can be configured to support the discovery of novel bioactive chemotypes using ADS. [ c] Yield of purified product (see Figure 5f or the structures).
[d] Dose-dependent activity of the purified product.
[e] Additional products were also isolatedw hose activity was not significant (see the SupportingInformation).
[f ]See the SupportingInformation for the products obtained with 6fin place of 6n.
[i]Activity of the purified reaction product that had 56 % ee. . .

Angewandte Communications
In round three,the activity of the product mixture derived from substrate 3 and racemic dihydropyran 6f' ' was dependent upon the enantiomer of the [Rh 2 (DOSP) 4 ]c atalyst used. HPLC analysis on ac hiral stationary phase showed that kinetic resolution of the dihydropyran had occurred to give product 13 in 56 % ee.T od etermine the absolute configuration and activity of both enantiomers,w ep repared [25,26] samples of both enantiomers of 6f' ' and hence 13:Itwas found that (S)-13 was approximately five-fold more active than (R)-13 [(S)-13:E C 50 = 860 AE 40 nm ;( R)-13:E C 50 = 4.5 AE 0.2 mm] and had indeed been selectively formed in the kinetic resolution of dihydropyran 6f' ' catalyzed by [Rh 2 {(R)-DOSP} 4 ]. To the best of our knowledge,e nantioselective OÀHi nsertion reactions of rhodium carbenoids have not previously been described: [27] Hence,A DS has enabled an ovel asymmetric transformation to be identified for the first time solely on the basis of the biological activity of aproduct.
To demonstrate the value of ADS in lead generation, we performed al imited SAR study inspired by amide 13 and oxazole 14.Arange of 17 analogues was prepared by Rhcatalyzed reactions of a-diazo amide 3 with either alcohols or nitriles.T he activity of the analogues spanned well over two orders of magnitude in both series,a llowing key structural features to be identified ( Figure 6; see also the Supporting Information, Table S2). Therefore,A DS can enable the discovery of bioactive chemotypes that provide options for subsequent lead optimization.
In conclusion, intermolecular reactions add significant value to ADS by dramatically expanding the range of reaction outcomes.B ye xploiting intermolecular reactions of diazo acetamides,ADS enabled the rapid elaboration of afragment (1)toproduce sub-micromolar agonists,all based on chemotypes with no previously reported activity against AR. In total, just four substrates were prepared, yet exploited in at otal of 326 reactions to explore diverse chemical space; however, the products of just nine reactions were purified to reveal arange of novel bioactive ligands.Injust three rounds of ADS,t he activity of the fragment was improved by up to afactor of 125. Retrospective analysis showed that ADS had enabled the efficient exploration of chemical space and facilitated the identification of increasingly active AR agonists,t ogether with associated syntheses.R emarkably,t he approach also led to the discovery of an ovel asymmetric transformation on the basis of the biological activity of the product alone.O verall, ADS is as ignificant addition to the lead generation toolkit, enabling the rapid discovery of novel, yet synthetically accessible,bioactive small molecules. Figure 6. Structures of the most active analoguesa nd summaryo fthe limited SAR study.S ee the Supporting Information for information on analogues 21-37.A r= 4-cyano-3-trifluoromethylphenyl.