Quantifying CDK inhibitor selectivity in live cells

Concerted multidisciplinary efforts have led to the development of Cyclin-Dependent Kinase inhibitors (CDKi’s) as small molecule drugs and chemical probes of intracellular CDK function. However, conflicting data has been reported on the inhibitory potency of CDKi’s and a systematic characterization of affinity and selectivity against intracellular CDKs is lacking. We have developed a panel of cell-permeable energy transfer probes to quantify target occupancy for all 21 human CDKs in live cells, and present a comprehensive evaluation of intracellular isozyme potency and selectivity for a collection of 46 clinically-advanced CDKi’s and tool molecules. We observed unexpected intracellular activity profiles for a number of CDKi’s, offering avenues for repurposing of highly potent molecules as probes for previously unreported targets. Overall, we provide a broadly applicable method for evaluating the selectivity of CDK inhibitors in living cells, and present a refined set of tool molecules to study CDK function.


Target Engagement Analysis in MCF7 cells.
On the day prior to transfection, 0.1 mL / well of MCF7 cells (ATCC) were seeded into 96-well plates (Corning 3917) in DMEM + 10% FBS at densities of 25,000 cells per well. After 20 hours of incubation, cells were transfected with NLuc-CDK6 or NLuc-CDK4 plasmids in combination with Cyclin D1 expression plasmid using conditions similar to those described for HEK293 cells, but using Viafect transfection reagent and a 6:1 Viafect:DNA ratio (volume:mass). 20 hours post transfection, medium was replaced with Opti-MEM medium. The target engagement assay conditions and probe 2 concentrations were identical to those used for HEK293 cells (0.063µM).

Endogenous phospho-Rb analysis in MCF7 cells.
The level of phospho-Rb (Ser 807/811) was measured with Lumit Immunoassay Cellular System as follows; 50,000 MCF-7 cells were seeded in 160 µl complete growth medium (DMEM + 10% FBS) per well into a 96 well plate and incubated overnight. The next day, the medium was replaced with phenolred free DMEM medium without serum and the cells were incubated overnight. Then, the cells were treated with various concentration of the compounds in 200 µl volume in the presence of 25 ng/ml hEGF (Promega, G5021) for 24 hours. After treatment, the samples in the plates were analyzed with Lumit Immunoassay Cellular System (Promega, W1331) to detect phospho-Rb following the manufacturer's recommendations and as described 5 . The following antibodies were purchased from Cell Signaling Technology Inc. and were used at 150 ng/ml: Rabbit anti-phospho-Rb (#8516) and Mouse anti-Rb (#9309). Briefly, the medium was replaced with 40 µl of 1X immunoassay buffer and 10 µl of lysis buffer was added. The plates were mixed vigorously for 20 minutes. Then, 50 µl of an antibody mix in 1X immunoassay buffer containing two primary antibodies against the phospho-Rb protein and the two Lumit secondary antibodies (Lumit Anti-Mouse Ab-SmBiT and Anti-rabbit Ab-LgBiT) was added to the lysates. The plates were incubated at 23°C for 90 minutes, followed by the addition of 25 µl of Lumit detection reagent. Luminescence was measured after 2 minutes using a plate-reading luminometer.
Data represents % Rb phosphorylation normalized to the control wells that contained vehicle treated cells (100%) and the 0% phosphorylation wells where all the detection reagents were added except for primary antibodies.

General Chemical Synthesis Information
All solvents were purchased from Sigma or Fisher Scientific and used without purification. AT7519 was purchased from MedKoo Biosciences, Morrisville NC. NanoBRET® 590 SE was obtained from Promega Corp. Madison, WI. 1 H-NMR spectra were recorded on a Bruker Avance 400 MHz spectrometer or a Bruker Ascend 400 MHz spectrometer. Chemical shifts (δ) are quoted in parts per million (ppm) and referenced to the residual solvent peak. Multiplicities are denoted as s-singlet, d-doublet, t-triplet, qquartet and quin-quintet and derivatives thereof (br denotes a broad resonance peak). Coupling constants are given in Hz and round to the nearest 0.1 Hz. Mass spectra were recorded on a Waters SQ Detector 2 (LC-MS) and purity (≥95 %) determined by reverse-phase high pressure liquid chromatography (RP-HPLC) using a Kinetex 5 µm EVO C18 100 Å LC Column 30 × 2.1 mm column or a Phenomenex Synergi 2.5 µm Max-RP 100 Å LC column. High resolution mass spectrum (HRMS) were recorded on a SCIEX Triple TOF 5600 spectrometer. Compounds were purified on a Waters LC Prep 150 using a Waters XBridge Prep C18 OBD 30x250mm column. Standard Method 1: Initial -90% aqueous (0.1% TFA in H2O), 10% acetonitrile to 0% aqueous, 100% acetonitrile, 30 min linear gradient.
The organic layer was dried over Na2SO4, absorbed on celite, concentrated to dryness and was subjected to silica gel flash chromatography using a 10% MeOH/DCM gradient. The purified product was concentrated to give the product (34.0 mg, 98.1%) as a purple solid. 1
NanoBRET 590 SE (12.3 mg, 0.0288 mmol) was added and the mixture was stirred in the dark for 2 hrs.
The reaction mixture was diluted to 8 mL with 1:1:0.01 water, ACN, TFA and was subjected to reversephase preparative HPLC purification using Standard Method 1. Product containing fractions were pooled and concentrated under reduced pressure to give a purple film that was treated with 10 mL ACN and concentrated to dryness three times. The resulting film was dried overnight under high vacuum to give the product (14.8 mg, 81.6%) as a purple solid.

Supplementary Figures
Supplementary Figure 1 Supplementary Figure 1. Linker evaluation for energy transfer probes. For each base compound, linker length, properties, or positioning was evaluated by screening for specific BRET signals across the panel of CDKs. Probes were added to cells expressing CDK/NanoLuc fusions at a concentration of 0.5µM in the presence of 20µM of the unlabeled parent compound to demonstrate specificity. For each probe/CDK pair, specific BRET is reported as a relative BRET signal by normalizing the raw BRET value for the tracer only samples (n = 2) to the raw BRET value in the presence of 20µM of the unlabeled parent compound (n = 2). Probe structures are depicted in panels A, B, C, and D, with corresponding relative BRET signals tabulated in panel E. Relative BRET signals in panel E represent values measured in a single (n = 1) biological experiment. Source data are provided as a Source Data File. Energy probes were characterized as described in the methods section. Briefly, energy probes were titrated onto HEK-293 cells expressing CDK/NanoLuc fusions up to a maximum concentration of 1µM (the functional performance limit of most of these energy transfer probes, see Supplementary Figure 5). The BRET ratio was measured as a function of energy probe concentration, and the EC50 value of the tracer was interpolated using Equation 1 (see methods section). For some CDK assays, the BRET was not saturable before reaching the energy probe solubility threshold (see Supplementary Figure 5), and the EC50 value is reported as undefined. In these cases, the absence of impact of tracer concentration on compound IC50 (Supplementary Figure 3)

Supplementary Figure 3. Impact of Energy Transfer Probe Concentration on the Apparent Compound IC50 for Preferred CDK/Probe Pairs.
Energy probes were characterized as described in the methods section. Briefly, a dilution series of test compound was added to HEK-293 cells containing an increasing concentration of energy transfer probe. The BRET ratio was measured as a function of test compound dose and the compound IC50 at each probe concentration was interpolated using equation 1 (see methods section). The data in each plot was collected in a single biological experiment (n = 1), with each individual data point as a single technical replicate (n = 1). In all cases, optimized probe concentrations chosen for CDK compound profiling in this study (Supplementary Table 1) provided a balance of sufficient assay window with minimal right-shifting of apparent compound IC50. Source data are provided as a Source Data File. In most cases, linear fits to the IC50 replots did not yield a meaningful Cheng-Prusoff relationship, as evidenced by the lack of either a slope meaningfully higher than "0" or a concentration-dependent increase in apparent compound IC50, suggesting that the probe concentrations used for IC50 profiling (Supplementary Table 1 Energy probes were characterized as described in the methods section. Briefly, HEK-293 cells were transfected with mixtures containing 1 part NanoLuc fusion vector and 9 parts of either cyclin/regulator vector (CDK + Cyclin samples) or a promoterless carrier DNA (CDK only samples). A dilution series of probe was added to the cells, after which BRET was measured and plotted as a function of probe concentration. In some cases where data was collected on different luminometers (e.g. CDKs 14−18), data is reported as signal fold change by normalizing each individual BRET value to the background BRET value on each instrument. In most cases, co-expression of a cyclin or regulator protein was found to influence the BRET value or probe potency. In the case of CDK10, CDK11A, and CDK11B, influence of co-expression on assay behavior was negligible. In virtually all cases (except for CDK7, which is explored further in Figure S9), cyclin or regulator co-expression was included for compound IC50 profiling. Dose response curves in each panel were measured in a single biological experiment (n = 1). Individual data points in Panel G (both curves), and Panels Q−U (Carrier DNA) represent technical singlicates (n = 1). The individual points in the Cyclin E1 curve for Panel C represent the mean of technical duplicates (n = 2). All other data points represent the mean ± S.D. of 4 technical replicates (n = 4). Source data are provided as a Source Data File.

Supplementary Figure 9
Supplementary Figure 9. Influence of Cyclin H and MAT1 Co-Expression on the CDK7 Assay. The influence of Cyclin H and/or MAT1 co-expression on the CDK7 assay was evaluated in HEK-293 cells using probe 5. NanoLuc-CDK7 fusion vector was co-transfected into HEK-293 cells with cyclin H, MAT1, and/or a promoterless carrier DNA at the ratios indicated. Probe 5 was added at a concentration of 1µM in the presence or absence of 20µM of Dinaciclib to demonstrate specificity. Specific BRET is reported as a relative BRET signal by normalizing the raw BRET value for the tracer only sample to the raw BRET value in the presence of 20µM of the unlabeled parent compound. Individual data points were measured in a single biological experiment (n = 1) and represent technical singlicates (n = 1). Cyclin H coexpression was found to reduce the BRET ratio regardless of MAT1 co-expression. MAT1 co-expression did not impact the BRET ratio regardless of Cyclin H co-expression. The highest BRET ratio was found when CDK7 was transfected in the absence of regulators. Source data are provided as a Source Data File.  a Relative BRET signals for each CDK/NanoLuc fusion after incubation with optimized energy transfer probes. Each probe was screened at a concentration of 0.5µM in the presence or absence of 20µM unlabeled parent compound, and the relative BRET signal calculated as described in the legend of Supplementary Figure 1. Values represent a single biological experiment (n = 1) with technical replicates (n = 2) for both the tracer only samples and the samples with 20µM of the unlabeled parent compound. Source data are provided as a Source Data File. Z' values were measured in a single biological experiment (n = 1) with technical quadruplicates (n = 4) at the probe concentration specified above in the presence or absence of a saturating (≥10 M) dose of unlabeled derivative.