Small molecule sequestration of the intrinsically disordered protein, p27^Kip1, within soluble oligomers

Abstract

Proteins that exhibit intrinsically disordered regions (IDRs) are prevalent in the human proteome and perform diverse biological functions, including signaling and regulation. Due to these important roles, misregulation of intrinsically disordered proteins (IDPs) is associated with myriad human diseases, including neurodegeneration and cancer. The inherent flexibility of IDPs limits the applicability of the traditional structure-based drug design paradigm; therefore, IDPs have long been considered “undruggable”. Using NMR spectroscopy and other methods, we previously discovered small, drug-like molecules that bind specifically, albeit weakly, to dynamic clusters of aromatic residues within p27^Kip1 (p27), an archetypal disordered protein involved in cell cycle regulation. Here, using synthetic chemistry, NMR spectroscopy and other biophysical methods, we discovered elaborated analogs of our previously reported molecules with 30-fold increased affinity for p27 (apparent K_d = 57 ± 19 μM). Strikingly, using analytical ultracentrifugation methods, we showed that the highest affinity compounds caused p27 to form soluble, disordered oligomers. Based on these observations, we propose that sequestration within soluble oligomers may represent a general strategy for therapeutically targeting disease-associated IDPs in the future.

Graphical Abstract

Introduction

Proteins that exhibit intrinsically disordered regions (IDRs) are prevalent in the human proteome and perform wide ranging functions that control cellular behavior, including signaling and regulation [1]. The levels of proteins with IDRs were previously shown to be tightly controlled at many points of regulation [2], highlighting the need for proper control of their levels for normal cellular behavior. Consistent with this need, their overexpression was generally shown to be deleterious to the growth of yeast (Saccharomyces cerevisiae), flies (Drosophila melanogaster) and worms (Caenorhabditis elegans), and to lead to phenotypic changes in mice and to be associated with cancer in humans [3]. Further, early results from Dunker and co-workers associated proteins with IDRs with cancer [4], and it is now well established that certain fully intrinsically disordered proteins (IDPs) self-associate to form fibrillar structures in association with neurodegenerative diseases (e.g., Tau in Alzheimer’s disease [5]; α-synuclein in Parkinson’s disease [6]; and FUS and TDP-43 in amyotrophic lateral sclerosis and frontotemporal dementia [7]). Beyond self-association and fibril formation as a disease mechanism, the activity of a variety of transcription factors with IDRs is altered in cancer, including c-Myc and p53 [8], and fusion oncogenes, which often arise through fusion of disordered transactivation domains to folded DNA binding or other chromatin binding domains, are drivers of cellular transformation in a wide variety of cancers [9]. Despite knowledge of these disease associations, limited attention has been given to the development of strategies for therapeutically targeting IDPs and proteins with IDRs [10] and they are often considered “undruggable” [9]. Even with the challenges of targeting IDPs, several reports have identified small molecules that can bind IDPs through a combination of computational and high-throughput screening methods [11-13].

We previously reported the discovery of a series of small molecules that bind weakly but specifically to a region of the IDP, p27^Kip1, which is a regulator of the cyclin-dependent kinases (Cdks) that control cell division in humans [14]. p27 is mislocalized from the nucleus to the cytoplasm in certain cancers, where it interacts with RhoA and alters cell motility [15]. Further, expression of p27 in cells of the inner ear prevents reentry into the cell division cycle that could otherwise enable hearing regeneration in hearing damaged individuals [16]. Small molecules that bind to p27 and inhibit interactions with its Cdk partners in hearing cells and RhoA in cancer cells thus could have therapeutic applications. Toward this goal, we previously used nuclear magnetic resonance (NMR) spectroscopy-based fragment screening to identify ~40 aromatic heterocycles that bind to the Cdk2 binding region of p27, termed p27-D2 (Fig. 1A, residues 60-93) [17], which contains eight aromatic residues. One of the identified compounds (SJ403) was shown to inhibit the binding of p27 to Cdk2, which potently inhibits catalytic activity, leading to kinase activation. Detailed analysis using NMR relaxation methods showed that the aromatic residues within p27-D2 form dynamic clusters that exchange between multiple configurations on the low microsecond timescale. Importantly, the small molecule SJ403 preferentially bound to one aromatic residue cluster configuration, shifting the conformational equilibrium to populate this state more highly [18]. These results suggest that aromatic heterocyclic compounds integrate within the clustered aromatic residues of p27-D2, altering its dynamic conformational landscape. The compounds included in our published studies [14, 18] are of relatively low molecular weight with limited functionality for interactions with the sidechains of aromatic amino acids. We hypothesized that derivatization of these original compounds to increase their chemical functionality based upon cheminformatics analyses would yield compounds with increased affinity for p27. Here, we report novel chemical compounds that engage aromatic and other residues within p27, bind with higher affinity than the original compounds, and promote formation of soluble p27 oligomers. We used NMR spectroscopy and analytical ultracentrifugation (AUC) to characterize the structure, dynamics and stoichiometry of the small molecule-induced soluble oligomers of p27. Our results demonstrate for the first time that an IDP can be sequestered within soluble oligomers by a small molecule and may represent a general mechanism for therapeutically targeting IDPs in the future.

Figure 1.

A. Scheme of the kinase inhibitory domain of p27 (p27-KID) showing regions D1 (binds to cyclin A), D2 (binds to Cdk2), and linker helix (LH). Also depicted are p27 residues within the subdomain D2 that interact with small molecule hits. B. Strategy for Group 1 scaffold optimization showing SAR-by-catalog and SAR-by-synthesis approaches. Structural features identified upon testing commercially available compounds (SAR-by-catalog) are depicted in blue while those identified upon testing newly synthesized compounds are given in magenta. Green boxes show isosteric moieties.

Results

Identification of improved p27-binding small molecule through cheminformatics analyses and chemical synthesis.

Screening of commercially-available compounds identified using cheminformatics methods (termed SAR-by-catalog)

We analyzed one Group 2 (G2) and three Group 1 (G1) p27-binding compounds from our earlier report [14] using sub-structure filters to identify additional compounds with similar chemical features (termed SAR-by-catalog, Suppl. Fig. 1). This analysis led to the purchase of ~160 additional compounds (Supp. Table 1), which were screened for binding to the kinase inhibitory domain (KID, Fig. 1A) of p27 (¹⁵N-labeled p27-KID) using 2D ¹H-¹⁵N HSQC NMR. Most of the newly identified p27-binding compounds (termed “hits”) exhibited patterns of NMR chemical shift perturbations (CSPs) similar to those of prior G1 & G2 compounds; however, analogs of the G1.1 scaffold (e.g., SJ710, Fig. 1B, Suppl. Fig. 1A) exhibited improved solubility and p27-KID binding (based upon larger and/or additional CSPs, data not shown). Purchased compounds (termed “Analog-by-catalog”, or “ABC” compounds) related to the G1.1 scaffold displayed varied Ring 1 substituents and a few included an aromatic Ring 3 (Supp. Table 1, Suppl. Fig. 1D). Notably, Ring 3 aromaticity slightly enhanced interactions with the “FY₈₈Y” region of p27-KID (Suppl. Fig. 2A, ABC-1). Further screening identified ABC-2 (Fig. 1B, Suppl. Table 1), which contains a 2,5-substituted furan heterocycle between the tricyclic core and the carboxylic functional group found in ABC-1. For the G2 scaffold, purchased analogs displayed varied substitutions on both Rings 1 and 3 as well as varied size and heteroatom composition of the 3^rd ring of the tricyclic core (Suppl. Table 1, Suppl. Fig. 1E). Most new G2 analogs exhibited high aqueous solubility but none caused greater NMR CSPs than the parent compound, SJ403.

Chemical synthesis of novel p27-binding compounds (termed SAR-by-synthesis)

We leveraged our new screening results (summarized in Suppl. Fig. 1), together with the concept of bioisostere replacement ([19] and references therein), to derivatize the G1.1 scaffold through chemical synthesis. Inclusion of an aromatic Ring 3 offers access to diverse analogs using commercially available 4,5-substituted-2-bromo-benzoic acids. Condensation of resorcinols with bromobenzoic acids [20] yielded tricyclic intermediates, which enabled further chemical elaboration (Scheme 1A). Isostere replacement of the -COOH functional group of ABC-1 with a tetrazole moiety (Fig. 1, SJ747; see Scheme 1B for synthetic scheme) enhanced interactions with p27 (Suppl. Fig. 3). SJ747 caused NMR CSPs exclusively for amide groups of residues within the D2 sub-domain of p27-KID (p27-D2) that were more extensive than for either of the original G1 and G2 scaffolds (compare Suppl. Fig. 3 with Figs. 1, 2 from ref. [14]). Comparison of CSP data for compounds with and lacking the tetrazole moiety indicated that this carboxylic acid isostere interacted with N66 and W76 of p27 (Suppl. Fig. 3). p27-KID contains two tryptophan residues; therefore, we reasoned that incorporation of a second tetrazole moiety into G1.1 analogs might improve p27 binding. Compound SJ749 (Scheme 2) displays two tetrazole moieties and exhibited enhanced interactions with the W₆₀N₆₁ region and increased affinity for p27-KID (Fig. 2A, vide infra). Further elaboration to include the furan ring of compound ABC-2 afforded the bis(carboxyl) compound, SJ755 (Scheme 2), and the corresponding bis(tetrazole) compound, SJ757 (Scheme 3).

Scheme 1.

Synthetic route for preparation of optimized G1.1 tricyclic core (A) and compound SJ747 (B)

Reagents and conditions: i) 1. NaOH 2.5 M, 100 °C, 15 min; 2. 10% CuSO₄. 9 h, 100 °C; ii) K₂CO₃. DMF, 2-chloropropanenitrile, 6 h, 60 °C; iii) NaN₃, NH₄Cl, DMF, 9 h, 100 °C

Scheme 2.

Synthetic route for preparation of compounds SJ749 and SJ755

Reagents and conditions: i) BBr₃, CH₂Cl₂ overnight; ii) K₂CO₃. DMF, 2-chloropropanenitnle, 6 h, 60 °C; iii) NaN₃, NH₄Cl, DMF, 9 h, 100 °C; iv) K₂CO₃, DMF, methyl 5-(chloromethyl) furan-2-carboxylate, 6 h, 60 °C; v) NaOH, MeOH

Figure 2.

Synthetic compounds generated by growing G1.1 scaffold interact specifically with p27-KID. Chemical shift perturbation and peak intensity loss histograms obtained by analysis of ¹H-¹⁵N HSQC spectra of ¹⁵N-p27-KID upon interaction with SJ749 (A, B), SJ755 (C, D), and SJ757 (E, F), respectively.

Scheme 3.

Synthetic route for preparation of compounds SJ757

Reagents and conditions: i) BBr₃, CH₂Cl2; ii) K₂CO₃, DMF, 5-(chloromethyl)furan-2-carbonitrile, 6 h, 60 °C; iii) NaN₃, NH₄Cl, DMF, 9 h, 100 °C.

Synthetic compounds sequester p27-KID within soluble oligomers

In addition to causing CSPs, binding of SJ749 also caused selective resonance broadening in 2D ¹H-¹⁵N HSQC NMR spectra of p27-KID (Fig. 2A, B; Suppl. Fig. 5A). Analysis of CSP and resonance intensities upon titration of SJ749 (Suppl. Fig 5B, C) yielded apparent K_d values of 392 ± 128 μM and 291 ± 76 μM, respectively, for binding to p27-KID (Table 1). These values reflect approximately 10-fold higher affinity for p27-KID than the original, parent compound, SJ710 [14]. Analysis using sedimentation velocity analytical ultracentrifugation (SV-AUC) showed that isolated p27-KID is monomeric (Fig. 3A, black trace; S = 1.04, Suppl. Table 2) and that, upon addition of excess SJ749, the sedimentation coefficient (S) shifted to a slightly higher value, corresponding to a monomeric p27 species bound to SJ749 (Fig. 3A, magenta trace; S = 1.16, Suppl. Table 2). Two-dimensional (2D) analysis of SV-AUC data [21] revealed that SJ749 caused compaction of monomeric p27-KID, with two conformers with different shape factor values (f/f₀) observed (f/f₀ = 1.66 & 1.55; compare Figs. 3D & E). In addition, a peak corresponding to small population of dimeric p27-KID appeared in the c(s) versus S data with SJ749 (18%, S=1.82; Fig. 3A, magenta trace; Suppl. Table 2) but low signal intensity prevented 2D data analysis. Thus, SJ749 caused compaction of p27-KID and promoted formation of a minor dimeric species.

Table 1.

Apparent Equilibrium dissociation constant (K_d) values obtained from analysis of 2D ¹H-¹⁵N HSQC spectra recorded as compounds were titrated into ¹⁵N-p27-KID. K_d values were obtained by fitting a single-site binding model to chemical shift perturbation (CSP) and relative peak intensity (I/I₀) values.

Compound	Kd (from CSP)	Kd (from I/I0)
SJ710	4.8 ± 1.3 mM	n. d.
SJ749	392 ± 128 μM	291 ± 76 μM
SJ755	338 ± 50 μM	229 ± 56 μM
SJ757	140 ± 80 μM	57 ± 19 μM

Figure 3.

Sedimentation velocity analytical ultracentrifugation (SV-AUC) revealed that synthetic compounds induced formation of soluble oligomers of p27-KID. (A, B, C). SV-AUC profiles (fringe displacement) for p27-KID alone (black) and with compounds SJ749 (A, magenta), SJ755 (B, blue), and SJ757(C, green). Two-dimensional size-and-shape distribution analyses for the color-coded, boxed regions in A, B & C are shown in D-G.

The two compounds with inserted furan rings, SJ755 and SJ757, bound to the D2 region of p27-KID (Fig. 2C-F; Suppl. Figs. 7A, 8A) with apparent K_d values of 338 ± 50 μM and 140 ± 80 μM, respectively (based upon analysis of CSP data; Table 1). Further, SV-AUC analysis showed that these compounds caused formation of soluble oligomers of p27-KID with a range of molecular sizes (Fig. 3B, C, F & G). With SJ755, a highly populated, compact dimer formed (52%; Fig. 3B, F; Suppl. Table 2), while multiple, larger species (with masses ranging from 36,811 Da to 118,300 Da, corresponding to oligomers containing from approximately three to ten p27-KID molecules) were observed with SJ757 (90% of p27-KID molecules; Fig. 3C, G; Suppl. Table 2). Notably, 1D ¹H NMR analyses showed that the furan ring-containing compounds did not self-aggregate (Suppl. Fig. 6). NMR CSP and peak intensity data indicated that SJ749, SJ755, and SJ757 interacted with similar regions of p27-KID, including residues near W60, N66, W76 and Y88; however, due to formation of high molecular weight soluble oligomers, peak broadening was more pronounced for p27-KID in the presence of SJ757, which was also associated with smaller CSP values for the observed resonances (Fig. 2E, F). Also, SJ757 broadened resonances of residues in other regions of p27-KID, including region D1 (residues 27-34), which binds to cyclin A, and region LH, which forms a kinked α-helix linking D1 and D2 (residues 38-59) when bound to Cdk2/cyclin A.

Interestingly, while SJ749, SJ755, and SJ757 significantly perturbed backbone HN resonances for W60 and W76 of p27-KID (CSP values and/or peak intensity values; Fig. 2; Suppl. Figs. 6-8), only SJ749, which lacks a furan ring, also perturbed chemical shift values of sidechain indole resonances (Fig. 2A, C, E; Suppl. Figs. 6-8). These results suggest that the furan rings in SJ755 and SJ757 influence their interactions with the indole sidechains of the two tryptophan residues. Strikingly, mutation of these two residues to alanine (p27-KID-W60A-W76A) abrogated binding to SJ749 and SJ755 (Suppl. Figs. 9, 10) but not to SJ757 (Suppl. Fig. 11). SJ757 bound to p27-KID-W60A-W76A with slightly reduced affinity relative to the wild-type protein but still engaged native residues within the D2 region as well as the D1 and LH regions. These results indicate that SJ757, by virtue of its more elaborate chemical structure, engages residues within multiple regions of p27-KID, enabling sequestration of multiple p27-KID molecules within soluble oligomers.

Tryptophan residues are critical for p27 Cdk2 inhibitory function

Our previously published p27 binding compounds [14], as well as the new analogs reported herein, all interact with the two tryptophan residues within p27-KID. When p27 is bound to Cdk2/cyclin A [22], the sidechains of these residues are protected from solvent by packing against the surface of Cdk2, suggesting that they are important contributors to the Gibbs free energy of binding (ΔG). We tested this hypothesis by using isothermal titration calorimetry (ITC) to monitor the binding of p27-KID and p27-D2 in which W60, W76, or both tryptophan residues were mutated to alanine (A) to either Cdk2 or the Cdk2/cyclin A complex. The W to A mutations within the p27-D2 construct, which lacks the D1 region that binds tightly to cyclin A [17], abrogated binding to Cdk2 (Suppl. Fig. 12B) and Cdk2/cyclin A (Fig. 4A; Table 2), demonstrating that the two tryptophan residues are important contributors to the ΔG of binding. In the context of p27-KID, the mutations abrogated binding to Cdk2 (Suppl. Fig. 12A) and caused a reduction of the values of the enthalpy of binding (ΔH) to Cdk2/cyclin A (Fig. 4B, Table 2), consistent with reduced binding of the mutated regions of p27-KID to Cdk2 within the Cdk2/cyclin A complex. With Cdk2/cyclin A, the ΔG of binding values for the p27-KID constructs were very similar due to binding of the native D1 region to cyclin A within the Cdk2/cyclin A complex (Table 2).

Figure 4.

Mutation of W60 and/or W76 to alanine affects binding of p27 to Cdk2/cyclin A. (A, B) Isothermal titration calorimetry data and binding isotherms for interaction of Cdk2/cyclin A with p27-D2 (A) and p27-KID (B). The p27-D2 mutants do not bind Cdk2/cyclin A (A), while the p27-KID mutants still bind Cdk2/cyclin A due to interactions between the p27-D1 region with cyclin A (B).

Table 2.

Thermodynamic parameters obtained using isothermal calorimetry at 25 °C.

Interaction	Kd (nM)	ΔG (kcal/mol)	ΔH (kcal/mol)	−TΔS (kcal/mol/deg)
Cdk2/cyclin A + p27-KID	12.9 ± 0.2	−10.7 ± 0.1	−43.6 ± 0.1	32.9 ± 0.1
Cdk2/cyclin A + p27-KID-W60A	16.9 ± 0.6	−10.6 ± 0.1	−35.3 ± 0.3	24.7 ± 0.3
Cdk2/cyclin A + p27-KID-W76A	20.3 ± 2.5	−10.5 ± 0.1	−30.3 ± 0.8	19.8 ± 0.9
Cdk2/cyclin A + p27-KID-W60A-W76A	17.0 ± 1.6	−10.6 ± 0.1	−24.2 ± 0.1	13.6 ± 0.1
Cdk2/cyclin A + p27-D2	133.6 ± 23.5	−9.4 ± 0.1	−24.7 ± 1.8	15.3 ± 1.9
Cdk2/cyclin A + P27-D2-W60A	Binding not detected	-	-	-
Cdk2/cyclin A + P27-D2-W76A	Binding not detected	-	-	-
Cdk2/cyclin A + P27-D2-W60A-W76A	Binding not detected	-	-	-
Cdk2 + p27-KID	180.8 ± 21.7	−9.2 ± 0.1	−23.2 ± 0.1	14 ± 0.1
Cdk2 + p27-KID-W60A	Binding not detected	-	-	-
Cdk2 + p27-KID-W76A	Binding not detected	-	-	-
Cdk2 + p27-KID-W60A-W76A	Binding not detected	-	-	-
Cdk2 + p27-D2	230.5 ± 67.4	−9.1 ± 0.2	−24.5 ± 0.7	15.4 ± 0.9
Cdk2 + p27-D2-W60A	Binding not detected	-	-	-
Cdk2 + p27-D2-W76A	Binding not detected	-	-	-
Cdk2 + p27-D2-W60A-W76A	Binding not detected	-	-	-

We next determined the effect of the W to A mutations in p27-KID on inhibition of the kinase activity of Cdk2/cyclin A toward the substrate Histone H1 (HH1). Wild-type p27-KID is a potent inhibitor of Cdk2/cyclin A [23], with an IC₅₀ value of 1.9 ± 0.3 nM in the current experiments (Fig. 5, Suppl. Fig. 13, Table 3), while p27-D2, which binds only to Cdk2 within the Cdk2/cyclin A complex, exhibited an IC₅₀ value of 67 ± 22 nM (Suppl. Fig. 13). As expected based upon our ITC results, mutation of individual or both tryptophan residues in p27-D2 abrogated inhibitory activity (Suppl. Fig. 13F-J). However, in the context of p27-KID, the individual W to A mutations only slightly affected Cdk2 inhibitory activity (IC₅₀ values of 1.7 ± 0.3 nM and 4.3 ± 1.0 nM for the W60A and W76A mutants, respectively; Table 3), while the dual W to A mutant exhibited an IC₅₀ value of 36 ± 8 nM but, even at saturating concentrations, was unable to fully inhibit Cdk2/cyclin A (Fig. 5, Suppl. Fig. 13). The results with the p27-KID mutants suggest that, despite disruption of interactions with Cdk2 due to the individual W to A mutations, Y88 at the C-terminal end of the KID is able to bind within that ATP binding pocket of Cdk2, while the D1 region is bound to cyclin A, and inhibit its catalytic activity [17, 22]. Mutation of both tryptophan residues disrupts binding to Cdk2 further, apparently limiting access of Y88 to Cdk2’s active site, increasing the IC₅₀ value and preventing full Cdk2 inhibition (Fig. 5, Suppl. Fig. 13, Table 3). Together, the ITC and Cdk2 inhibition assay results demonstrate that the two tryptophan residues within the D2 region of p27-KID are major contributors to the thermodynamics of binding to Cdk2 within the Cdk2/cyclin A complex and are critical for full inhibition of Cdk2 activity.

Figure 5.

Mutation of W60 and W76 in p27-KID to alanine reduced inhibitory potency and prevented full inhibition of Cdk2 catalytic activity toward the substrate, Histone H1. The catalytic activity of Cdk2 within the Cdk2/cyclin A complex at the lowest concentration of p27-KID (black data points) and p27-KID-W60A-W76A (green data points) was normalized to 100%.

Table 3.

Half maximal Cdk2/cyclin A inhibitory concentration (IC₅₀) values of p27-KID variants using Histone H1 as substrate.

P27-KID-	IC50(nM)
WT	1.9 ± 0.3
W60A	1.7 ± 0.3
W76A	4.3 ± 1
W60A-W76A	36 ± 8

Discussion

Our previous NMR-based fragment screening efforts identified aromatic heterocycles that bound weakly but specifically to dynamic clusters of aromatic residues within the D2 region of p27 [14, 18]. Here, by screening compounds with similar aromatic heterocyclic core structures but with different ring substituents, we identified compounds (ABC-1 & ABC-2; Fig. 1B) that bound with higher affinty and engaged a larger number residues within p27-KID. Further, we reasoned that duplication of the substituents associated with enhanced binding on the aromatic heterocyclic core would improve binding. In addition, we hypothesized that replacement of the carboxcylic acid moiety of ABC-1 and ABC-2 with an isosteric tetrazole moiety would also enhance binding to p27-KID. To test these ideas, we synthesized the compounds, SJ749, SJ755 and SJ757 (Fig. 1B), each with a common aromatic heterocyclic core substituted with two phenyl ether moieties rich in H-bond donors and acceptors. The aromatic heterocyclic core of these compounds preserved and enhanced interactions with aromatic residues in p27-KID (near residues W60, W76 and Y88) and extended them to also include residues near N66 through introduction of the two tetrazole moieties in SJ749 (Fig. 2A, B). Introduction of two carboxy-furan moieties in SJ755 (Fig. 1B) enhanced interactions further (Fig. 2C, D). Interestingly, the p27-KID binding enhancements caused by these two types of aromatic heterocyclic core substituents were also associated with compound-dependent formation of soluble dimers of p27-KID (Fig. 3; Suppl. Table 2). This property, compound binding-dependent formation of soluble protein oligomers, was further enhanced in the compound, SJ757, in which the terminal carboxy moiety of SJ755 was replaced with an isosteric tetrazole moiety. Remarkably, this compound sequestered 90% of p27-KID molecules within an array of soluble oligomers comprised of between three and approximately ten p27-KID molecules and an indeterminate number of compound molecules. We propose that these chemically multivalent compounds interact with dynamic clusters of aromatic amino acids in different p27-KID molecules, causing formation of soluble dimers and, with SJ757, higher order soluble oligomers. Because p27-KID dynamically fluctuates between multiple conformations involving clusters of different aromatic amino acids [18], the two tetrazole-furan moieties of SJ757 may weakly and non-covalently cross-link multiple protein molecules. The large size of SJ757 requires that its multivalency for binding to aromatic amino acids be fulfilled by multiple p27-KID molecules. We term this behavior “disordered protein sequestration by soluble oligomerizaton” and propose that this may be a general approach for chemical modulation of IDPs.

p27-KID folds upon binding to Cdk2/cyclin A, with the “RxL” motif within the D1 region binding to a conserved pocket on the surface of cyclin A and the D2 region adopting extensive secondary structure in the course of forming an extensive interface with Cdk2, ultimately positioning Y88 in the ATP binding pocket for kinase inhibition [17, 22]. Eight aromatic residues contribute to the interface between the D2 region of p27-KID and Cdk2, and most of these are the residues engaged by the synthetic compounds identified herein. While, in the absence of Cdk2/cyclin A, these aromatic residues form disordered, constantly fluctuating clusters [18], they are poised to adopt specific, ordered conformations on the surface of Cdk2. Some of the aromatic residues within the D2 region of p27-KID are essential for interactions with Cdk2 (e.g., W60 and W76; Fig. 4, 5; Suppl. Fig. 12, 13; Tables 2, 3) and these are key mediators of interactions with the compounds reported herein (Suppl. Figs. 9-11). We propose that small molecule-dependent sequestration through soluble oligomerization provides a general approach for targeting IDPs that experience folding upon binding to their functional partners. The residues within disordered protein regions that participate in specific partner recognition and folding upon binding can, in principle, be leveraged for recognition by chemical moieties within chemical compounds. The high enrichment of aromatic residues within the D2 region of p27-KID is exceptional relative to the usual amino acid compositional bias of disordered protein regions [1]; however, many IDPs display short linear motifs (SLiMs) with conserved sequences that mediate specific folding upon partner binding and these SLiMs often contain amino acids not typically associated with disorder. For example, the N-terminal transactivation domain of p53 contains a SLiM with conserved aromatic and hydrophobic residues that mediates binding to Mdm2 [24]. Further, many viral proteins contain multiple, conserved SLiMs and other, longer interaction regions [25, 26], that could possibly be targeted for sequestration by chemical compounds with binding features like those reported herein for binding p27-KID. Our experimental strategy, which involves NMR-based fragment screening, cheminformatics analysis, molecular elaboration through chemical synthesis and detailed biophysical characterization of protein:compound interactions, is readily adaptable to other disease-associated IDPs. We acknowledge that our top compound, SJ757, binds p27-KID relatively weakly (apparent K_d values of 140 ± 80 μM and 57 ± 19 μM from NMR CSP and I/I₀ data, respectively; Table 1); however, the modest chemical synthesis campaign employed herein resulted in binding affinity improvements of more than an order magnitude. Investment of much more extensive chemical synthesis resources could more thoroughly explore chemical space and potentially increase affinity for p27-KID further, providing opportunities for modulating p27 function in cells. The mechanism of sequestration through soluble oligomerization presented herein differs from the entropy-driven small molecule:disordered protein interaction mechanism discussed by Vendruscolo and co-workers [27] and Liu and co-workers [28], and thus provides an additional strategy for consideration when seeking to therapeutically intervene in human diseases involving disordered proteins.

Materials and Methods

Preparation of proteins

The p27 constructs were expressed in E. Coli (BL21/DE3) with an N-terminal 6xHis affinity tag after sub-cloning into pET28a (Novagen) using established procedures [17]. This included p27-KID (residues 22-105 of human p27) and p27-D2 (residues 58-105 of human p27) and the following mutants: W60A, W76A, W60A-W76A. Isotope-labeled proteins (¹⁵N and ¹³C/¹⁵N) were expressed in a MOPS-based minimal media using established procedures [23]. All p27 constructs were purified by nickel affinity chromatography, digested with thrombin to remove the 6xHis tag, and further purified using reverse-phase high performance liquid chromatography (HPLC) using a C4 column (Vydac) and 0.1% trifluoroacetic acid-containing water/acetonitrile solvent system. Protein concentrations were determined by UV absorbance at 280 nm under denaturing conditions using a molar extinction coefficient of 15,470 M⁻¹ cm⁻¹ for p27-KID and p27-D2; 9,970 M⁻¹ cm⁻¹ for p27 variants with a single tryptophan residue; and 4,470 M⁻¹ cm⁻¹ for p27 variants without a tryptophan residue. Full length human Cdk2, active Cdk2 (phosphorylated at threonine 160), and truncated human cyclin A (residues 173–432) were expressed and purified using established protocols [17, 29].

SAR-by-analogs: Cheminformatics analysis

Based on the central cores of Group 1 (G1.1, G1.2, and G1.3) and Group 2 (G2) hits, respectively, substructure searches were performed in SciFinder and selected compounds (Suppl. Table 1) were purchased from commercial vendors: Vitas-M, Vitascreen, LLC, University of Illinois Urbana-Champaign Research Park, 2001 South First Street, Suite 201, Champaign, IL, 61820; Ambinter, Ambinter c/o Greenpharma, 3, allée du titane, 45100 Orléans, FRANCE; ChemBridge, ChemBridge Corporation, 11199 Sorrento Valley Rd., Suite 206, San Diego, CA 92121, USA; PrincetonBio, Princeton BioMolecular Research, Inc., 475 Wall Street, Princeton, NJ 08540, USA; Alinda, SRC Alinda, Kuskovskaya Street, 20A, entrance 2B, office 409 111141, Moscow, Rusia; Specs, Specs US Compound Management Facility, 14900 Burbridge Road SE, Cumberland, MD 21502, USA; and Maybridge, Fisher Scientific UK Ltd, T/A Maybridge, Bishop Meadow Road, Loughborough, Leicestershire, LE11 5RG, UK. The Group 1 and Group 2 analog types are defined in [14].

Chemical synthesis of novel G1.1 compounds

All materials were purchased from commercial suppliers and used without further purification. Pre-purification and QC analyses were done on a Waters Acquity UPLC/PDA/ELSD/MS system carried out with a BEH C18 2.1 x 50 mm column using gradient elution. Purification of compounds was carried out by normal phase column chromatography using pre-pack SNAP silica cartridges on a Biotage Isolera system. Reported yields were not optimized. Structures were determined by NMR spectroscopy and purity was determined by LC-MS/ELSD. NMR spectra (1D ¹H and ¹³C, respectively, and 2D ¹H-¹³C HSQC and ¹H-¹³C HMBC, respectively) were recorded on a Bruker 600 MHz spectrometer equipped with TCI cryogenic gradient probe and processed and analyzed using Bruker Topspin software.

General procedure for generation of 6H-benzo[c]chromen-6-one tricyclic core (1-3).

In a 5 mL glass vial, resorcinol derivatives (2 mmol) and o-bromobenzoic derivatives (1 mmol) were mixed with water (0.7 mL) and NaOH (5N, 0.7 mL) and heated at 100 °C for 15 minutes. CuSO₄ (10 % aqueous solution, 0.2 mL) was added to the reaction mixture, the vial was sealed and further heated at 100 °C for 6 hours. The reaction mixture was cooled on ice and the precipitate was filtered, washed with cold water, and dried in vacuo.

3-hydroxy-4,9-dimethyl-6H-benzo[c]chromen-6-one (1) (yield 51 %)

¹H-NMR δ 2.214 (3 H, s), 2.514 (3 H, s), 6.898 (1 H, d, ³J = 8.63 Hz), 7.381 (1 H, d, ³J = 7.94 Hz), 7.998 (1 H, d, ³J = 8.63 Hz), 8.083 (1 H, s), 8.092 (1 H, d, ³J = 7.94), 10.27 (1 H, s).

¹³C-NMR δ 8.903, 22.326, 109.758, 111.916, 112.633, 116.8, 121.808, 122.291, 129.325, 130.233, 136.163, 146.568, 151.035, 158.492, 161.318.

MS (ES⁺): m/z = 241.21.

3-hydroxy-8-methoxy-4,9-dimethyl-6H-benzo[c]chromen-6-one (2) (yield 70%)

¹H-NMR δ 2.216 (3 H, s), 2.358(3 H, s), 3.933 (3H, s), 6.885 (1 H, d, ³J = 8.6 Hz), 7.56 (1 H, s), 7.928 (1 H, d, ³J = 8.6 Hz), 8.095 (1 H, s), 10.08 (1 H, s).

¹³C-NMR δ 8.904, 17.401, 56.273, 108.69, 110.89, 111.833, 112.525, 118.152, 121.159, 124.421, 129.41, 136.3, 150.179, 157.276, 157.584, 161.306.

MS (ES⁺): m/z = 271.21.

3-hydroxy-8-methoxy-1,4,9-trimethyl-6H-benzo[c]chromen-6-one (3) (yield 59%)

¹H-NMR δ 2.193 (3 H, s), 2.373 (3 H, s), 2.744 (3 H, s), 3.945 (3 H, s), 6.732 (1 H, s), 7.67 (1 H, s), 8.139 (1 H, s), 10.05 (1 H, s).

¹³C-NMR δ 9.094, 17.719, 25.681, 56.2, 109.246, 109.39, 109.933, 116.486, 119.255, 128.158, 130.414, 133.485, 135.104, 150.986, 156.196, 156.69, 161.16.

MS (ES⁺): m/z = 285.11.

Synthesis of 2-((4,9-dimethyl-6-oxo-6H-benzo[c]chromen-3-yl)oxy)propanenitrile (4)

In a 5 mL glass vial, under nitrogen, derivative 1 (0.5 mmol) was mixed with potassium carbonate (0.6 mmol) and 1 mL anhydrous DMF and stirred for 15 minutes at 70 °C. 2-chloropropanenitrile (1 mmol) was added and the reaction mixture was further stirred for 6 hours at 70 °C. The reaction was subsequently cooled, diluted with EtOAc (10 mL) and washed three times with water (3 x 10 mL). The organic phase was dried over anhydrous MgSO4, filtered, and concentrated under reduced pressure. Pure compound was obtained by flash chromatography using an EtOAc/hexane/MeOH gradient (yield 68 %).

¹H-NMR δ 1.786 (3 H, d, ³J = 6.57 Hz), 2.284 (3 H, s), 2.545 (3 H, s), 5.622 (1 H, q, ³J = 6.57 Hz), 7.281 (1 H, d, ³J = 8.84 Hz), 7.475 (1 H, d, ³J = 8.01 Hz), 8.134 (1 H, d, ³J = 8.01 Hz), 8.238 (1 H, s), 8.279 (1 H, d, ³J = 8.84 Hz).

¹³C-NMR δ 9.013, 19.909, 22.317, 63.567, 110.802, 113.365, 115.393, 117.619, 119.619, 122.367, 123.047, 130.303, 130.408, 135.2, 146.852, 150.509, 155.942, 160.862.

MS (ES⁺): m/z = 294.26.

Synthesis of 3-(1-(1H-tetrazol-5-yl)ethoxy)-4,9-dimethyl-6H-benzo[c]chromen-6-one (SJ982747; SJ747)

In a 5 mL glass vial, under nitrogen, nitrile derivative 4 (0.3 mmol) was mixed with NaN₃ (0.6 mmol), NH₄Cl (0.6 mmol), and anhydrous DMF (2mL) and stirred overnight at 100 °C. The reaction was cooled down and iced water (3 mL) added. The reaction mixture was adjusted to pH 11 with NaOH and filtered through celite. The product separated upon acidification to pH 2 as an off-white solid (yield 65 %).

¹H-NMR δ 1.762 (3 H, d, ³J = 6.47 Hz), 2.234 (3 H, s), 2.48 (3 H, s), 6.08 (1 H, q, ³J = 6.47 Hz), 7.099 (1 H, d), 7.406 (1 H, d), 8.069 (3 H, m).

¹³C-NMR δ 9.043, 20.647, 22.237, 68.052, 110.939, 112.409, 115.474, 117.222, 121.984, 122.69, 130.211, 130.235, 135.225, 146.956, 150.342, 156.611, 158.041, 161.112.

MS (ES⁺): m/z = 337.10.

Synthesis of 3,8-dihydroxy-4,9-dimethyl-6H-benzo[c]chromen-6-one (5)

In a two-necked round bottom flask, under nitrogen, methoxy derivative 2 (0.5 mmol) was mixed with anhydrous CH₂Cl₂ (2mL), and BBr₃ (1 mmol). The reaction mixture was stirred overnight at room temperature. The reaction was diluted with CH₂Cl₂ (3 mL) treated with water (3 mL), and then extracted with EtOAc. The organic phase was dried over anhydrous MgSO4, filtered, and concentrated under reduced pressure. Pure compound was obtained by flash chromatography using an EtOAc/hexane/MeOH gradient (yield 58 %).

¹H-NMR δ 2.2 (3 H, s), 2.327 (3 H, s), 6.86 (1 H, d, ³J = 8.56 Hz), 7.537 (1 H, s), 7.88 (1 H, d, ³J = 8.56 Hz), 8.013 (1 H, s), 9.979 (1 H, s), 10.186 (1 H, s).

¹³C-NMR δ 8.906, 17.351, 110.427, 111.777, 112.425, 112.961, 118.173, 120.793, 124.492, 127.959, 135.099, 149.893, 155.983, 156.828, 161.29.

MS (ES⁺): m/z = 257.15.

Synthesis of 2,2'-((4,9-dimethyl-6-oxo-6H-benzo[c]chromene-3,8-diyl)bis(oxy))-dipropanenitrile (6)

In a 5 mL glass vial, under nitrogen, derivative 5 (0.25 mmol) was mixed with potassium carbonate (0.6 mmol) and 1 mL anhydrous DMF and stirred for 15 minutes at 70 °C. 2-chloropropanenitrile (1 mmol) was added and the reaction mixture was further stirred for 12 hours at 70 °C. The reaction was subsequently cooled, diluted with EtOAc (10 mL) and washed three times with water (3x10 mL). The organic phase was dried over anhydrous MgSO₄, filtered, and concentrated under reduced pressure. Pure compound was obtained by flash chromatography using an EtOAc/hexane/MeOH gradient (yield 61 %).

¹H-NMR δ 1.78 (6 H, b), 2.268 (3 H, s), 2.397 (3 H, s), 5.570 (1 H, q, ³J = 6.59 Hz), 5.677 (1 H, q, ³J = 6.59 Hz), 7.25 (1 H, d, ³J = 8.86 Hz), 7.816 (1 H, s), 8.177 (1 H, d, ³J = 8.86 Hz), 8.272 (1 H, s).

¹³C-NMR δ 8.964, 17.188, 19.803, 19.91, 63.152, 63.64, 110.96, 111.875, 113.131, 115.44, 119.003, 119.44, 119.602, 121.924, 125.747, 130.048, 137.159, 149.884, 154.926, 155.5, 160.612.

MS (ES⁺): m/z = 363.02.

Synthesis of 3,8-bis(1-(1H-tetrazol-5-yl)ethoxy)-4,9-dimethyl-6H-benzo[c]chromen-6-one (SJ982749; SJ749)

In a 5 mL glass vial, under nitrogen, bis-nitrile derivative 6 (0.3 mmol) was mixed with NaN₃ (1.2 mmol), NH₄Cl (1.2 mmol), and anhydrous DMF (2mL) and stirred overnight at 100 °C. The reaction was cooled and iced water (3 mL) added. The reaction mixture was adjusted to pH 11 with NaOH and filtered through celite. The product separated upon acidification to pH 2 as a light-yellow precipitate (yield 60 %).

¹H-NMR δ 1.758 (3 H, d, ³J = 6.68 Hz), 1.776 (3 H, d, ³J = 6.68 Hz), 2.238 (3 H, s), 2.379 (3 H, s), 6.072 (1 H, q, ³J = 6.68 Hz), 6.137 (1 H, q, ³J = 6.68 Hz), 7.098 (1 H, d, ³J = 8.86 Hz), 7.646 (1 H, s), 8.018 (1 H, d, ³J = 8.86 Hz), 8.149 (1 H, s).

¹³C-NMR δ 9.116, 17.465, 20.610, 20.651, 67.814, 68.139, 111.082, 111.796, 112.407, 115.399, 118.711, 121.559, 125.307, 129.416, 137.528, 149.758, 155.545, 156.094, 158.002, 158.171, 160.746.

MS (ES⁺): m/z = 449.10.

Synthesis of dimethyl 5,5'-(((4,9-dimethyl-6-oxo-6H-benzo[c]chromene-3,8diyl)-bis-(oxy))bis-(methylene))bis(furan-2-carboxylate) (7)

In a 5 mL glass vial, under nitrogen, derivative 5 (0.25 mmol) was mixed with potassium carbonate (0.6 mmol) and 1 mL anhydrous DMF and stirred for 15 minutes at 70 °C. Methyl 5-(chloromethyl) furan-2-carboxylate (1 mmol) was added and the reaction mixture was further stirred for 12 hours at 70 °C. The reaction was subsequently cooled, diluted with EtOAc (10 mL) and washed three times with water (3 x 10 mL). The organic phase was dried over anhydrous MgSO₄, filtered, and concentrated under reduced pressure. Pure compound was obtained by flash chromatography using an EtOAc/hexane/MeOH gradient (yield 54 %).

¹H-NMR δ 2.24 (3 H, s), 2.367 (3 H, s), 3.833 (6 H, s), 5.327 (2 H, s), 5.371 (2 H, s), 6.834 (2 H, d ³J = 3.06 Hz), 7.252 (1 H, d, ³J = 8.85 Hz), 7.335 (1 H, d, ³J = 3.06 Hz), 7.348 (1 H, d, ³J = 3.06 Hz), 7.776 (1 H, s), 8.14 (1 H, d, ³J = 8.85 Hz), 8.249 (1 H, s). ¹³C-NMR δ 8.893, 17.387, 52.524, 62.891, 63.207, 110.012, 110.66, 112.215, 113.04, 113.173, 114.353, 118.727, 119.711, 119.733, 121.611, 125.18, 129.328, 136.762, 144.525, 144.586, 149.705, 154.906, 155.151, 156.524, 157.13, 158.854, 160.922.

MS (ES⁺): m/z = 533.07.

Synthesis of 5,5'-(((4,9-dimethyl-6-oxo-6H-benzo[c]chromene-3,8-diyl)bis(oxy))bis-(methylene))-bis(furan-2-carboxylic acid) (SJ982755; SJ755)

In a 10 mL glass vial, derivative 7 (0.15 mmol) was disolved in MeOH (3 mL) and NaOH (1 mmol) was added. The reaction mixture was stirred at room temperature for 12 hours. Solvent was removed under reduced pressure; the residue was resuspended in water (2 mL), pH adjusted to 11, and solution filtered through celite. The product separated upon acidification to pH 2 as an off-white precipitate (yield 72 %).

¹H-NMR δ 2.239 (3 H, s), 2.366 (3 H, s), 5.3 (2 H, s), 5.342 (2 H, s), 6.787 (2 H, d, ³J = 3.18 Hz), 7.223 (1 H, d, ³J = 3.18 Hz), 7.236 (1 H, d, ³J = 3.18 Hz), 7.251 (1 H, d, ³J = 8.84 Hz), 7.771 (1 H, s), 8.134 (1 H, d, ³J = 8.84 Hz), 8.239 (1 H, s).

¹³C-NMR δ 8.974, 17.381, 62.959, 63.26, 110.046, 110.653, 112.167, 112.926, 113.039, 114.365, 118.687, 119.043, 119.021, 121.588, 125.137, 129.301, 136.827, 145.788, 145.851, 149.682, 154.265, 154.518, 156.549, 157.161, 159.839, 160.975.

MS (ES⁺): m/z = 505.19.

Synthesis of 3,8-dihydroxy-1,4,9-trimethyl-6H-benzo[c]chromen-6-one (8)

In a two-necked round bottom flask, under nitrogen, methoxy derivative 3 (0.5 mmol) was mixed with anhydrous CH₂Cl₂ (2 mL), and BBr₃ (1 mmol). The reaction mixture was stirred overnight at room temperature. The reaction was diluted with CH₂Cl₂ (3 mL), treated with water (3 mL), and then extracted with EtOAc. The organic phase was dried over anhydrous MgSO4, filtered, and concentrated under reduced pressure. Pure compound was obtained by flash chromatography using an EtOAc/hexane/MeOH gradient (yield 48 %).

¹H-NMR δ 2.169 (3 H, s), 2.405 (3 H, s), 2.705 (3 H, s), 6.699 (1 H, s), 7.639 (1 H, s), 8.04 (1 H, s), 9.862 (1 H, s), 10.18 (1 H, s).

¹³C-NMR δ 9.08, 17.675, 25.683, 109.697, 109.877, 113.767, 116.251, 119.256, 128.374, 128.909, 131.115, 133.826, 150.662, 155.161, 155.456, 161.11.

MS (ES⁺): m/z = 271.21.

Synthesis of 5,5'-(((1,4,9-trimethyl-6-oxo-6H-benzo[c]chromene-3,8-diyl)bis(oxy))bis-(methylene))-bis(furan-2-carbonitrile) (9)

In a 5 mL glass vial, under nitrogen, derivative 8 (0.25 mmol) was mixed with potassium carbonate (0.6 mmol) and 1 mL anhydrous DMF and stirred for 15 minutes at 70 °C. 5-(chloromethyl)furan-2-carbonitrile (2 mmol) was added and the reaction mixture was further stirred for 12 hours at 70 °C. The reaction was subsequently cooled, diluted with EtOAc (10 mL) and washed three times with water (3 x 10 mL). The organic phase was dried over anhydrous MgSO₄, filtered, and concentrated under reduced pressure. Pure compound was obtained by flash chromatography using an EtOAc/ hexane/MeOH gradient (yield 42 %).

¹H-NMR δ 2.223 (3 H, s), 2.394 (3 H, s), 2.865 (3 H, s), 5.327 (2 H, s), 5.417 (2 H, s), 6.931 (2 H, m), 7.131 (1 H, s), 7.654 (2 H, m), 7.878 (1 H, s), 8.238 (1 H, s).

¹³C-NMR δ 9.071, 17.684, 25.82, 62.479, 62.67, 111.345, 111.525, 112.242, 112.277, 112.593, 112.741, 113.802, 119.899, 125.13, 125.152, 125.614, 125.651, 128.95, 130.367, 134.336, 135.614, 150.469, 155.541, 155.57, 156.625, 156.812, 160.749, 162.921.

MS (ES⁺): m/z = 481.15.

Synthesis of 3,8-bis((5-(1H-tetrazol-5-yl)furan-2-yl)methoxy)-1,4,9-trimethyl-6H-benzo[c]chromen-6-one (SJ982757, SJ757)

In a 5 mL glass vial, under nitrogen, bis-nitrile derivative 6 (0.15 mmol) was mixed with NaN₃ (0.6 mmol), NH₄Cl (0.6 mmol), and anhydrous DMF (1 mL) and stirred overnight at 100 °C. The reaction was cooled and iced water (3 mL) added. The reaction mixture was adjusted to pH 11 with NaOH and filtered through celite. The product separated upon acidification to pH 2 as a light-yellow precipitate (yield 41 %).

¹H-NMR δ 2.223 (3 H, s), 2.39 (3 H, s), 2.877 (3 H, s), 5.296 (2 H, s), 5.372 (2 H, s), 6.807 (1 H, d, ³J = 2.91 Hz), 6.83 (1 H, d, ³J = 2.91 Hz), 7.015 (2 H, b), 7.185 (1 H, s), 7.918 (1 H, s), 8.232 (1 H, s).

¹³C-NMR δ 9.180, 17.853, 25.894, 63.139, 63.307, 107.081, 107.091, 111.207, 111.259, 112.055, 112.860, 113.067, 113.870, 119.793, 128.813, 130.135, 134.175, 135.624, 148.781, 149.066, 150.005, 150.092, 150.455, 154.911, 155.917, 156.043, 160.915.

MS (ES+): m/z = 567.16.

NMR experiments

All NMR experiments were performed at 298 K (25 °C) using a Bruker Avance 600 MHz spectrometer equipped with TCI cryogenic gradient probe. NMR spectra were processed using Bruker Topspin software and analyzed using computer-aided resonance assignment (CARA) software [30]. Two-dimensional (2D) ¹H-¹⁵N HSQC NMR experiments were performed using a SampleJet sample changer. Compounds were dissolved at 50 mM each in DMSO-D₆ and were mixed using a Gilson 215 liquid handler with buffer (20 mM Na phosphate, pH 6.5, 200 mM NaCl, 10 % ²H₂O, 5 mM DTT-D₁₀) containing ¹⁵N-p27-KID protein (100 μM) to give final compound concentration of 1 mM. Three-dimensional (3D) backbone triple-resonance experiments were performed to establish resonance assignments for the p27 constructs. 2D ¹H-¹⁵N HSQC NMR titrations of optimized compounds SJ749, SJ755, and SJ757, respectively, into ¹⁵N-p27-KID were recorded to determine p27-KID:compound affinity. The following molar ratios of ¹⁵N-p27-KID to compound were used: 1:0, 1:0.5, 1:1, 1:2, 1:4, 1:6, and 1:8. Chemical shift perturbation values were quantified using the equation $\Delta \delta =\sqrt{\Delta {\delta }_H^2+(0.2\Delta {\delta }_N{)}^2}$. The statistical significance was based on a threshold defined as the average CSP value plus two times the standard deviation of the mean (Δδ_ave + 2σ). All resonances that exhibited chemical shift perturbations greater than Δδ_ave + 2σ were subsequently plotted against compound concentration and non-linear fitting was performed using GraphPad Prizm 7 software utilizing a simple 1:1 binding model to quantify p27-KID:small molecule interactions and obtain apparent equilibrium dissociation constant (K_d) values. The WaterLOGSY experiments to assess compound aggregation in the absence of protein (Suppl. Fig. 6) were performed at a compound concentration of 200 μM with a 1.25 s NOE mixing time.

Analytical Ultracentrifugation (AUC)

Sedimentation velocity experiments were conducted in a ProteomeLab XL-I analytical ultracentrifuge (Beckman Coulter, Indianapolis, IN) following standard protocols unless mentioned otherwise [31]. Protein samples (100 μM) in buffer containing 20 mM sodium phosphate pH 6.5, 200 mM NaCl, 5 mM DTT, 2% DMSO without and with compounds (ratio 1:8) were loaded into cell assemblies comprised of double sector charcoal-filled centerpieces with a 12 mm path length and sapphire windows. The density and viscosity of the ultracentrifugation buffer at 20 °C were measured with a DMA 5000M density meter and an AMVn viscometer (both Anton Paar, Graz, Austria), respectively. The cell assemblies, containing identical sample and reference buffer volumes of 390 μL, were placed in a rotor and temperature equilibrated at rest at 20 °C for 2 hours before it was accelerated from 0 to 50,000 rpm. Rayleigh interference optical data were collected at 1-minute intervals for 12 hours. The velocity data were modeled with diffusion-deconvoluted sedimentation coefficient (S) distributions c(s) in SEDFIT (https://sedfitsedphat.nibib.nih.gov/software/default.aspx), using algebraic noise decomposition and with signal-average frictional ratio and meniscus position refined with non-linear regression [32]. The s-value was corrected for time, temperature and radial position and finite acceleration of the rotor was accounted for in the evaluation of Lamm equation solutions [33]. Maximum entropy regularization was applied at a confidence level of P-0.68. The partial specific volume of the protein, based on its amino acid composition, was calculated in SEDFIT. All plots were created in GUSSI [34] (http://biophysics.swmed.edu/MBR/software.html, kindly provided by Dr. Chad Brautigam). A two-dimensional size-shape distribution, c(s, f/f₀) (with one dimension the s-distribution and the other the frictional ratio (f/f₀)) was calculated with an equidistant f/f₀-grid of 0.1 steps that varied from 1 to 3, a linear S-grid from 0.2 to 6 S with 100 S-values, and Tikhonov-Phillips regularization at one standard deviation. The velocity data were transformed to c(M,f/f₀) distributions with M the molecular weight, f/f₀ the frictional ratio, S the sedimentation coefficient and plotted as contour plots. The dotted lines of c(M, f/f₀) indicate constant S values.

Isothermal calorimetry (ITC)

ITC experiments were performed using a MicroCal ITC200 calorimeter with p27 variants in the syringe and Cdk2/cyclin A and Cdk2, respectively, in the cell. Prior to each set of experiments, protein samples were extensively dialyzed together against ITC buffer containing 20 mM HEPES, pH 7.5, 300 mM NaCl, 5 mM TCEP. A standard titration consisted of 19 injections of 2-μl of p27 variant (100 μM) into a solution of Cdk2/cyclin A (10 μM) or Cdk2 (10 μM), respectively. The time interval between injections was 180 s. Experiments were performed at 25 °C. Thermodynamic parameters were obtained by fitting the raw data using Origin software (OriginLab) according to the manufacturer's instructions using a 1:1 binding model to obtain values of the enthalpy of binding (ΔH), Gibbs free energy of binding (ΔG), entropy of binding (ΔS), and stoichiometry factor (N). Experiments were performed in triplicate and mean and standard deviations of the mean values of these parameters are reported.

Cdk2/cyclin A kinase activity assays

Cdk2/Cyclin A (100 pM) was mixed with Histone H1 (50 μM; EMD Millipore) and varied amounts of p27 constructs and incubated for 3 hours at 4 °C. Subsequently, ATP (50 μM total concentration, of which 10 μCi γ ³²P-ATP (PerkinElmer, Inc.)) was added to each reaction and further incubated for 30 minutes at 35 °C. Each reaction had a total volume of 20 μL. The sample buffer contained 20 mM HEPES pH 7.3, 25 mM sodium β-glycerolphosphate, 15 mM MgCl₂, 16 mM EGTA, 0.5 mM Na₃VO₄ and 10 mM DTT. The reactions were quenched by addition of SDS-gel loading buffer (7 μL) and then analyzed by SDS-PAGE (10 μL). The gels were dried at 70 °C under vacuum and a phosphoimager (GE Healthcare, Piscataway, NJ) was used to quantify the ³²P-Histone H1 bands. IC₅₀ values were determined by curve fitting using GraphPad Prizm 7 software. Experiments were performed in triplicate and mean IC₅₀ and standard deviations of the mean values are reported.

Research Highlights.

• Due to their lack of stable structure, IDPs are often considered ‘undruggable’

• A multivalent small molecule binds clusters of aromatic residues within p27, an IDP

• These small molecules sequester p27 within soluble oligomers

• Both hydrophobic and polar moieties mediate these small molecule/IDP interactions

• We propose sequestration within soluble oligomers as a strategy of drugging IDPs