Nanomolar Inhibitors of the Transcription Factor STAT5b with High Selectivity over STAT5a

Src homology 2 (SH2) domains play a central role in signal transduction. Although many SH2 domains have been validated as drug targets, their structural similarity makes development of specific inhibitors difficult. The cancer-relevant transcription factors STAT5a and STAT5b are particularly challenging small-molecule targets because their SH2 domains are 93 % identical on the amino acid level. Here we present the natural product-inspired development of the low-nanomolar inhibitor Stafib-1, as the first small molecule which inhibits the STAT5b SH2 domain (Ki=44 nm) with more than 50-fold selectivity over STAT5a. The binding site of the core moiety of Stafib-1 was validated by functional analysis of point mutants. A prodrug of Stafib-1 was shown to inhibit STAT5b with high selectivity over STAT5a in tumor cells. Stafib-1 provides the first demonstration that naturally occurring SH2 domains with more than 90 % sequence identity can be selectively targeted with small organic molecules.

Abstract: Src homology 2(SH2) domains play acentral role in signal transduction. Although many SH2 domains have been validated as drug targets,t heir structural similarity makes development of specific inhibitors difficult. The cancer-relevant transcription factors STAT5a and STAT5b are particularly challenging small-molecule targets because their SH2 domains are 93 %i dentical on the amino acid level. Here we present the natural product-inspired development of the lownanomolar inhibitor Stafib-1, as the first small molecule which inhibits the STAT5b SH2 domain (K i = 44 nm)with more than 50-fold selectivity over STAT5a. The binding site of the core moiety of Stafib-1 was validated by functional analysis of point mutants.Aprodrug of Stafib-1 was shown to inhibit STAT5b with high selectivity over STAT5a in tumor cells.S tafib-1 provides the first demonstration that naturally occurring SH2 domains with more than 90 %s equence identity can be selectively targeted with small organic molecules.
Src homology 2( SH2) domains are highly homologous regions,approximately 100 amino acids long, which are found in asignificant subset of signaling proteins.SH2 domains play afundamental role in intracellular signaling via recognition of specific phosphotyrosine-containing peptide motifs. [1] The conserved nature and similar binding preferences of these domains pose an enormous challenge for specific inhibitor development. [2] Ap articularly challenging example is given by the SH2 domains of the transcription factors STAT5a and STAT5b, [3] which are essential for STAT5a/b signaling.T hese are 93 %i dentical on the amino acid level (see Figure S1 in the Supporting Information), and bind to the same peptide motifs. [4] Consequently,STAT5a and STAT5b are often jointly referred to as "STAT5", despite tissue-specific expression patterns and an umber of non-redundant biological functions. [3,5] Antisense oligonucleotides directed against STAT5b, but not STAT5a, inhibited tumor growth in mice. [6] Most recently,s elective inhibition of Stat5b has been proposed as an ovel therapeutic approach to combat Bcr-Abl positive leukemias. [7] Small-molecule inhibitors targeting the SH2 domain of STAT5b with specificity over the STAT5a SH2 domain would be useful for investigating the differential roles of STAT5a and STAT5b under diverse conditions,a nd could serve as lead compounds for drug development.
While as mall number of STAT5S H2 domain inhibitors have been reported, [8] none of these studies has disclosed selectivity for one STAT5protein over the other. We recently reported Fosfosal as aS TAT5b SH2 domain inhibitor (K i = 17.4 mm) [9] (Figure 1a). Structure-activity relationships of Fosfosal revealed that deleting the carboxylic acid retained partial activity,b ut deleting the phosphate led to ac omplete loss of activity against STAT5b.I na na ttempt to increase activity,w es ubstituted the carboxylic acid group of Fosfosal with asecond phosphate.This structural change is contrary to standard medicinal chemistry,w hich aims to replace phosphate groups with isosteric groups of lesser charge.However, the resulting compound, catechol bisphosphate (1), inhibited STAT5b with an inhibitory constant K i of only 0.9 AE 0.1 mm (Figures 1a,b,T able 1, and see Tables S1 and S2 in the Supporting Information). Surprisingly,a nalysis of 1 against STAT5a in an ewly developed binding assay (see Figure S2 and Supporting Methods in the Supporting Information) revealed 35-fold lower activity against the SH2 domain of STAT5a (K i = 34 AE 3 mm). TheS H2 domains of other STAT family members and the tyrosine kinase Lck [10] were only inhibited at higher concentrations.P ositioning the two phosphate groups of 1 in the meta position, as represented by resorcinol bisphosphate (2), resulted in as trong decrease of activity against STAT5b (K i = 27 AE 4 mm,T able 1), with concomitant loss of selectivity.R eplacement of the two phosphate groups by methylene diphosphonates,a si nt he compound 3,r esulted in virtually complete loss of activity against STAT5b (Table 1a nd Table S1).
We carried out molecular docking studies using Auto-Dock Vina [11] and the previously described homology model of the STAT5b SH2 domain [9] based on the crystal structure of STAT5a. [12] Thet op three docking results placed catechol bisphosphate into the phosphotyrosine binding pocket of the STAT5b SH2 domain, consistent with electrostatic interactions between STAT5b Arg618 and both phosphate groups ( Figure 1c). Thearginine residue in position 618 of STAT5b is highly conserved in all SH2 domains and forms essential electrostatic interactions with the phosphotyrosine moiety of its natural interaction partners. [13] Alignment of the STAT5b homology model [9] with the X-ray structure of STAT5a [12] suggested that small differences in the primary structure of STAT5a and STAT5b might cause ap artial change in secondary structure in the vicinity of the binding pocket for  The selectivity factor was calculated as Conversion of IC 50 values into K i values was carried out as described. [15] n.a. = not applicable.

Angewandte
Chemie 1 (see Figure S3 in the Supporting Information). Detailed comparative structural analysis by crystallographic or NMRbased methods would be required to identify the subtle structural differences contributing to specificity. In attempts to further increase the submicromolar affinity of 1 for STAT5b,wesought inspiration from natural products with an explicit or masked 1,2-dihydroxyphenyl motif such as dopamine,h ydrocaffeic acid, piperine,a nd l-DOPA( see Scheme S1 in the Supporting Information). Since our synthetic methodology for bisphosphate generation involved hydrogenolysis of dibenzylphosphate esters, [14] any double bonds in the natural products are saturated in our derivatives. TheN -acylatedd opamine derivative 4 (K i = 0.69 AE 0.04 mm), the hydrocaffeic acid derivative 5 (K i = 0.82 AE 0.05 mm), and the tetrahydropiperine derivative 6 (K i = 0.73 AE 0.07 mm)d isplayed similar or marginally higher activities than 1 (Table 1 and Table S1). However, the Fmoc-protected dopamine bisphosphate 7 (K i = 0.45 AE 0.04 mm)a nd Fmoc-protected l-Dopa methyl ester bisphosphate 8 (K i = 0.46 AE 0.08 mm)a re approximately twice as active as 1.Docking studies suggested p-stacking interactions with Tr p641 as ap otential cause,a nd suggested that shifting the amide bond of 7 closer to the catechol moiety might facilitate hydrogen bonds between the inhibitor and the protein backbone at amino acid positions 642 and 644 (see Figure S4 in the Supporting Information). Indeed, 9 (K i = 0.21 AE 0.04 mm)istwofold more active than 7. All catechol bisphosphate derivatives,except 5,display strong selectivity for STAT5b over STAT5a (Table 1a nd Table S1).
Before further compound optimization, we verified the binding site of 1 (Figure 1c). TheB ODIPY-FL-labeled derivative 10 (see Scheme S2 in the Supporting Information) was designed as at racer molecule for direct binding assays based on fluorescence polarization, and was found to bind to wild-type STAT5b with high affinity (K d = 0.86 AE 0.08 mm ; Figure 2). Consistent with the binding mode proposed by docking (Figure 1c), binding of 10 to the STAT5b point mutant Arg618Ala was markedly reduced as compared to Stat5b wild-type.B inding to STAT5b Arg618Lys,w ith ap ositive charge one carbon-carbon bond closer to the protein backbone than in wild-type STAT5b,w as partially reduced. Binding to STAT5b Trp641Ala was significantly reduced, suggesting that Tr p641 is important for binding of the catechol bisphosphate core.The weak affinity of wild-type STAT5a for 10 is in line with the results of the competitionbased fluorescence polarization assay (Figure 1b and Table 1). These results indicate that the catechol bisphosphate binds to the phosphotyrosine binding pocket of the STAT5b SH2 domain, confirming its selectivity for STAT5b over STAT5a.
Replacement of the Fmoc group of 9 with an aphthyl group resulted in the compounds 11 (K i = 0.24 AE 0.01 mm)and 12 (K i = 0.28 AE 0.02 mm ;T able 1). In order to further improve the activity of the compounds,w ee nvisioned targeting the hydrophobic pocket created by Phe633 and Ty r665 by an aromatic group.W edesigned compound 13 by extending the core of 12 via an N-phenyl carboxamide group (Figure 3a,b) for synthesis in an eight-step procedure (Figure 3c). Compound 12 was chosen over 11 as at emplate,s ince it displays more favorable geometry for extending to the adjacent hydrophobic pocket. Compound 13 displayed fourfold higher activity against STAT5b (K i = 0.044 AE 0.001 mm)t han 12 (Figure 3d,T able 1), and displays 55-fold selectivity for STAT5b over STAT5a (K i (STAT5a) = 2.42 AE 0.05 mm ;s ee Table S3 in the Supporting Information), and even higher selectivity against other SH2 domains.R emoval of both phosphate groups as in 14 led to ac omplete loss of activity. Monophosphorylated 15,w hich is based on 12,w as also significantly less active,and lost specificity for STAT5b.These data, as well as data from additional control compounds, demonstrate the fundamental importance of the bisphoshorylated core structure for selective STAT5b inhibition (see Table S4 in the Supporting Information).
Thehigh activity and selectivity of 13 prompted us to carry out acomparison with the natural ligand QDTpYLVLDKWL (16), derived from the EPO receptor. [16] Thef luorescence polarization assays we used for STAT5a and STAT5b are both based on STAT5a/b binding to the core sequence of this peptide motif,p YLVLDKWL. [17] Both STAT5a (K d = 133 AE 26 nm)a nd STAT5b (K d = 103 AE 13 nm)b ind to 5-carboxylfluorescein-pYLVLDKWL with similar affinities.T he inhibitory constants of the peptide QDTpYLVLDKWL (16) against the STAT5p roteins were also similar (K i (STAT5a): 0.41 AE 0.09 mm, K i (STAT5b) = 0.54 AE 0.08 mm), demonstrating that the natural peptide sequence does not differentiate  Ak ey event in STAT signaling is STAT binding to activated cell surface receptors and/or non-receptor tyrosine kinases (Figure 4a). [18] Binding is required for STAT phosphorylation at the conserved tyrosine residue C-terminal of their SH2 domain (STAT5a: Tyr694;STAT5b:T yr699). Theefficiency of STAT SH2 domain inhibitors can thus be investigated by analyzing the phosphorylation state of STAT proteins,u sing phosphospecific antibodies for the conserved tyrosine.H owever,a nalysis of endogenous Stat5a and Stat5b phosphorylation is hampered by the lack of antibodies able to selectively recognize tyrosine-phosphorylated STAT5b but not tyrosine-phosphorylated STAT5a (or vice versa). All current commercial antibodies recognizing phosphorylated STAT5b (pTyr699) and/or STAT5a (pTyr694) have been raised against ap hosphotyrosinecontaining peptide which is identical in both proteins.I no rder to analyze the relative activity of small molecules against the individual STAT5p roteins,w et ransfected K562 cells with expression vectors for the fusion proteins STAT5a-GFP or STAT5b-GFP.L ike endogenous STAT5a/b,t he fusion proteins are phosphorylated at the conserved tyrosine residue.H owever, the presence of the GFP tag increases the molecular weight, allowing the fusion proteins to be distinguished from endogenous Stat5 on aW estern blot.
To test its activity in cell-based assays, 13 was converted into the corresponding cell-permeable pivaloyloxymethyl (POM) ester 17, which is designed to liberate 13 inside the cell after cleavage by intracellular esterases (Figure 4b). [19] To rule out the possibility that inhibitory effects of 17 on STAT5b phosphorylation might be caused by the release of pivalic acid or formaldehyde generated during the cleavage,t he pivaloyloxymethylester 18,b ased on the inactive bisphosphonate 3,w as prepared as ac ontrol compound. Neither 17 nor 18 displayed activity against STAT5b in vitro (see Table S1). Tr eatment of Stat5b-GFP-transfected K562 cells with 17 showed adose-dependent decrease of Ty r699 phosphorylation of STAT5b-GFP (Figure 4c,d). Timecourse experiments indicated as teady increase in inhibition during the first hour of exposure to 17,consistent with cleavage of the POM groups to liberate the active agent, and inhibition was still observed after 8hours (see Figure S5 in the Supporting Information). Then egative These data demonstrate that the high specificity of 13 observed in vitro is also maintained under cellular conditions. Since endogenous STAT5c onsists of both STAT5a and STAT5b,the effect of 17 on endogenous STAT5 (Figure 4g,h) is lower than the effect on STAT5b-GFP (Figure 4c,d). Based on the relative degrees of inhibition of endogenous STAT5 and STAT5b-GFP,w ec onclude that the majority of phosphorylated endogenous STAT5i nK 562 cells is STAT5b. Studies quantifying the relative amounts of STAT5a and STAT5b and their phosphorylation levels in K562 cells have not been published.
In summary,w eh ave developed 13 as the first small molecule that can differentiate between the two highly homologous STAT5p roteins. 13,d ubbed Stafib-1 (for "Stat five b inhibitor-1"), displays low-nanomolar activity against STAT5b (K i = 44 AE 1nm), with more than 50-fold selectivity over STAT5a. This is the first report describing strongly divergent affinities of the SH2 domains of STAT5a/b for ac hemical entity,b oth in vitro and in cultured human cells.
Theb inding site of the core of the inhibitors,c atechol bisphosphate (1), on STAT5b was characterized by point mutants,which represents the first point mutant validation of as mall-molecule inhibitor of aS TAT SH2 domain. The pivaloyloxymethylester of Stafib-1, dubbed Pomstafib-1 (17), inhibited tyrosine phosphorylation of aS TAT5b-GFP fusion protein with high selectivity over the corresponding STAT5a-GFP fusion protein in human leukemia cells. 17 will be av aluable tool for dissecting the functions of STAT5b and STAT5a in mammalian cells.T ot he best of our knowledge, our work represents the first demonstration that small molecules can differentiate between naturally occurring SH2 domains which have more than 90 %a mino acid sequence identity.W eb elieve that our data will significantly influence the scientific communitys perception of the potential of small molecules as potent and selective inhibitors of highly similar protein-protein interaction domains.  . .