Quantitative, Wide-Spectrum Kinase Profiling in Live Cells for Assessing the Effect of Cellular ATP on Target Engagement

1Kinases represent the largest group of enzymes in the human proteome and are functionally integral to signal transduction. Dysregulated kinase activity is commonly associated with a number of human cancers and autoimmune diseases. Consequently, the human kinome represents one of the broadest sources of therapeutic targets for small-molecule modulation.

2The majority of clinically relevant kinase inhibitors engage their target proteins at the kinase active site, and compete with high intracellular concentrations (1–10 mM) of the cosubstrate ATP. Selective modulation of individual kinases represents a challenge due to the highly conserved structure of the ATP binding pocket, most notably within the protein kinase family. Engagement of off-target kinases is therefore commonly observed, and collateral engagement is often implicated in drug side effects. Kinase inhibitors may also demonstrate polypharmacology over a number of intracellular targets, a feature that often underlies the therapeutic potential of lead drug candidates. As our understanding of the therapeutic relevance of the human kinome deepens, a growing need has emerged for methods capable of quantifying kinase inhibitor occupancy, selectivity, and affinity within the cellular environment where engagement would naturally occur.

3Traditional biochemical techniques may fail to predict kinase engagement in cells. Although methods with isolated proteins are generally robust and quantitative, they do not measure compound permeability, nor encompass the physiological complexity of the full-length kinase protein in the presence of the cellular milieu and regulatory circuits that modulate kinase function. For example, reliance on isolated catalytic kinase domains may pose certain challenges for assessing the binding characteristics of allosteric inhibitors that occupy domains distal to the ATP pocket (i.e., type III/IV inhibitors) or for inhibitors that show activation-state-dependent engagement. Moreover, acellular approaches operate under defined assay conditions, typically at subsaturating concentrations of ATP. Intracellular ATP concentrations are likely non-uniform (exceeding 1 mM in some cellular compartments), and the affinity (i.e., KM) of most kinases for ATP resides in the 10–100 μM range. The interdependence of these factors leads to a challenge for use of biochemical assays as accurate proxies for cellular kinase engagement studies. Consequently, engagement potencies for ATP-competitive kinase inhibitors are often markedly shifted for their cellular targets, and biochemical kinase methodologies may fail to accurately predict cellular potency. A number of newer approaches have therefore been developed in an attempt to more accurately assess compound engagement at protein kinases expressed in human cells.

4Mass spectrometry (MS)-based chemoproteomics approaches that utilize kinase inhibitor matrices may offer a more physiologically relevant assessment of kinase engagement, as they query endogenous kinases from cell extracts. However, such methods require disruption of plasma membrane integrity, and therefore dilution of the key cellular cofactors (such as ATP) that are known to influence cellular potency. The cellular thermal shift assay (CETSA) represents a significant advancement for intact cell engagement studies, but does not assess engagement under equilibrium conditions nor provide quantitation of target occupancy or affinity. CETSA is also reportedly prone to a number of noteworthy false-negative results for multikinase inhibitors, limiting its utility in the context of broad kinome profiling. More recent MS-based methods using cell-permeable kinase affinity probes that function in the presence of endogenous ATP offer a more compelling advancement for live-cell profiling. However, the formation of a covalent adduct between the reporter probe and the target protein is required prior to lytic analysis via MS. Consequently, such approaches do not allow a quantitative assessment of kinase engagement under a thermodynamic equilibrium with the test molecule. Ideally, methods for quantifying kinase target engagement should be performed within intact cells and offer a quantitative measure of fractional occupancy and affinity for the compound of interest.

5Here we report the outcome of the first broad-spectrum, equilibrium-based analysis of inhibitor occupancy in living cells, without dissolution of the cell membrane. Our approach utilizes an energy transfer technique (NanoBRET) that reports on kinase engagement in real time, and enables quantitative inhibitor profiling with a simplified work flow for 178 kinases including over 40 integral membrane receptors. Broad kinome profiling of clinically relevant kinase inhibitors in live cells provides a mechanistic framework for the impact of microenvironmental ATP concentrations on cellular target engagement for human kinases.

6Traditional biochemical assays provide a convenient and robust platform for quantitation of binding or enzymatic inhibition, but do not allow for assessments of compound permeability or affinity in the presence of cellular factors that may influence potency. To develop a broad platform to measure kinase engagement in live cells, we exploited a recently described technique using bioluminescence resonance energy transfer (NanoBRET) as a proximity-based measure of compound binding. In brief, a quantitative analysis can be performed in live cells via energy transfer between a 19-kDa luciferase (NanoLuc, Nluc)-tagged target protein and a cell-permeable fluorescent energy transfer probe introduced to the culture medium (Figure 1). Compound binding results in competitive displacement of the probe and a loss of energy transfer in live cells. The measurement is exclusive to only intact cells, based on the addition of a cell-impermeable inhibitor of Nluc (which gates out signals originating from targets in dead cells or cell culture debris). Previously, this technique has successfully been used to quantify compound engagement over a number of intracellular target proteins, and is compatible with scalable work flows and microplate-compatible instrumentation.

7To enable a kinome-wide capability, a number of cell-permeable, fluorescent energy transfer probes were developed based on a diverse set of ATP-competitive kinase inhibitors (Figure 2A). Type I kinase inhibitors generally bind with minimal conformational bias, and were thus selected as inhibitor scaffolds for the development of broad-coverage energy transfer probes. With the exception of probe 4, all probes were all derived from type I kinase inhibitors (Figure 2A).

8Target specificity is achieved based on the dependence of proximity between the Nluc and the energy transfer probe. Each probe was characterized in cells via energy transfer, using a library of plasmid DNAs encoding N- or C-terminal genetic fusions of full-length kinases with Nluc. To determine engagement over this kinase library, we determined the apparent affinity of each energy transfer probe by titration onto live HEK293 cells expressing full-length kinase/Nluc fusion protein. As shown in Figure 2B (and Table S1), currently 178 full-length kinases can be queried using this energy technique, with good representation across the kinome subfamilies (Figures 2, 3, 4, and S1). This approach is even well suited for large, multidomain proteins such as LRRK2 (286 kDa) and IRE1α/ERN1 (Table S1). Forty full-length receptor tyrosine kinase (RTK) assays were achieved, representing the broadest cellular target engagement approach for these integral membrane receptors. Bioluminescence imaging analysis (Figure S2) confirmed appropriate subcellular localization for kinases used in this study, including DDR1 (plasma membrane), IRE1α (ER), and p38α/MAPK14 (cytoplasm). The ability to interrogate kinase engagement in such complex cellular compartments may offer a general advantage over MS approaches, which are often more limited to highly soluble or abundant kinases. In total, 178 cellular kinases were successfully queried with only six energy transfer probes. However, not all kinases tested against the probe set yielded robust assay signals. With increasing characterization of publicly available kinase inhibitor scaffolds, it should be possible in the future to develop a more diverse set of probes that span a larger fraction of the kinome.

9Use of reversible, ATP-competitive energy transfer probes should enable a quantitative measure of intracellular target occupancy. As demonstrated in Figure 3, the Cheng-Prusoff relationship between the reporter and the unlabeled molecule can be established in live cells, which allows determination of the apparent dissociation constant (KD) for a test compound. For DDR1 as an example, the linearized Cheng-Prusoff analysis yielded an apparent KD of 0.5 nM for dasatinib in live HEK293 cells. At energy transfer probe concentrations at or below KD, this technique should serve as measure of apparent occupancy for the selected kinase in live cells, provided that the expression level of the target kinase does not impact the measured occupancy values. In cell culture, intracellular target engagement potency (i.e., apparent KD) should be independent of intracellular kinase concentration, provided that the kinase levels are sufficiently low. We established that our target engagement results measured with transfected cells remained constant over a range of expression levels, and were similar to those results measured with a target kinase expressed from an endogenous genetic locus. Using the nuclear kinase CDK2 as an example, we were able to titrate expression to levels flanking that of a homozygous cell line engineered to express Nluc as a fusion with CDK2 protein from an endogenous genetic locus (knocked in via CRISPR/Cas9). In transfected cells, target engagement results were nearly identical over this broad expression range, and agreed closely with results generated with endogenously expressed CDK2-Nluc protein (Figure S2).

10Biochemical analysis of kinase selectivity and affinity may fail to predict engagement in live cells. For the development of kinase chemical probes for cell biology research, kinase inhibitors must display sufficient intracellular selectivity at the intended target. We applied the energy transfer technique to query affinity and selectivity of the novel chemical probe 6j (Figure S3), a putative selective inhibitor for DDR1. In biochemical measurements using the truncated domain of DDR1, 6j demonstrates 20-fold selectivity over DDR2. However, as shown in Figure S3, this selectivity index was considerably lower in cells, suggesting that phenotypic responses may arise from inhibition of a combination of both DDR1 and DDR2. These results further indicate that intracellular kinase selectivity for a given chemical probe may deviate from biochemical predictions.

11For large, multidomain kinases, intracellular energy transfer method should address many of the shortcomings of biochemical assay techniques traditionally used to characterize inhibitor selectivity. For example, the reliance on segregated kinase domains in biochemical formats has been shown to limit the ability to detect broad classes of ATP-competitive inhibitors that may bind outside the ATP pocket or exhibit conformation-state-dependent binding. For many kinases (e.g., RTKs), isolation of full-length kinase for biochemical analysis may be problematic.

12To demonstrate the broad applicability of the energy transfer assay for quantitation of engagement for kinase inhibitors utilizing diverse binding mechanisms, we performed equilibrium-based analysis of engagement for a set of full-length kinase targets in live cells. This target panel was composed of ABL, PIK3CA (PI3Kα), BTK, RIPK1, LIMK1, and LCK. As shown in Figure 4 (and Figure S1), affinity profiles were successfully established at each kinase. Use of full-length kinases allowed us to query the engagement of various types of ATP-competitive inhibitors, which bind sites within, adjacent to, or distal to the nucleotide binding pocket (type I/II, III, and IV, respectively). Analysis of RIPK1 engagement revealed binding of various type I inhibitors, as well as the type II inhibitor ponatinib and the type III inhibitor necrostatin-1 (nec-1). Similarly, we successfully measured engagement of inhibitors at LIMK1 that utilize diverse mechanisms. Binding of both the type I inhibitor (LIMKi3) and the type III inhibitor TH251 was detected in live cells. For ABL kinase, engagement of type I inhibitors (dasatinib), type II inhibitors (imatinib, ponatinib) and the type IV inhibitor GNF-2 were successfully measured. GNF-2 engages ABL kinase in the myristoyl binding pocket, located distally to the ATP binding site. Only partial displacement of GNF-2 was observed against the energy transfer probe, possibly due to the unique binding mode of GNF-2 at ABL. It has been previously established that both nec-1 and GNF-2 are functionally competitive with ATP, consistent with our ability to detect their kinase interactions using type I, ATP-competitive energy transfer probes. These results support use of the energy transfer technique for querying engagement of compounds that bind via mechanisms dissimilar from that of the energy transfer probe, as long as the binding of the compound and the probe are mutually exclusive.

13For energy transfer probes derived from type I kinase inhibitor scaffolds, compound potencies should be generally unaffected by the probe chemotype. RIPK1 binds multiple energy transfer probes developed in this study, and therefore served to demonstrate this independence. For the type I inhibitors measured against RIPK1, engagement potency was similar (within 2-fold) when using probe 1 or probe 2 (Figure S3).

14As discussed earlier, measurements of binding or inhibition of kinases in isolation may fail to accurately predict engagement in the complex milieu of live cells. To demonstrate that the energy transfer method could establish a functional link between physical binding and modulation of intracellular kinase activity, we performed a correlation analysis between target engagement via energy transfer and kinase function using a more traditional cellular pathway analysis. In the absence of suitable biophysical methods, cellular pathway analysis tools may serve as proxies to infer intracellular target engagement for kinases. For example, substrate phosphorylation analysis has previously served as a measure of intracellular inhibition for the targets of the multikinase inhibitor (R)-crizotinib. Comparison of cellular phospho-ELISA results versus enzymatic half-maximal inhibitory concentration (IC50) values revealed significant potency offsets for crizotinib for nearly all of the queried kinases (Table S2). The cellular offset was most notable for kinase LCK, which demonstrated a >3-log shift in cellular potency (3 μM) compared with the reported biochemical value (<1 nM).

15We used energy transfer to measure intracellular target engagement potencies against the same group of tyrosine kinases analyzed via phospho-ELISA. Consistent with intracellular phosphorylation analysis, highest engagement potencies were observed for the primary targets MET and ALK, with potencies ranging over approximately three orders of magnitude for the collateral targets (Figures 5A and S3). In contrast to the acellular enzymatic analysis, live-cell engagement analysis in HEK293 cells revealed engagement potencies in close agreement (R2 = 0.95) with published cellular potency values determined by phospho-ELISA (Figures 5B and S3). These results indicate that kinase engagement measured via energy transfer can be predictive of inhibition of intracellular kinase activity. With the emergence of new genome editing and expression techniques, the BRET method should be expandable to a variety of cell types when a specific cellular context or native expression level is required to accurately reflect cell physiology.

16The potency offset observed between cellular and acellular analysis of kinase inhibition implies that selectivity may be generally improved within cells for promiscuous, ATP-competitive kinase inhibitors. To test this hypothesis, we performed a quantitative, kinome-wide analysis of intracellular occupancy for (R)-crizotinib, and compared our results with published crizotinib profiles using isolated kinases or kinase domains. Such comprehensive biochemical analyses provide a strong foundation for complementary cellular mechanistic studies. Our intracellular target engagement analysis included a diversity set comprising 178 full-length kinases spanning each kinase subfamily, including the well-characterized crizotinib targets MET and ALK. Since engagement is dose dependent, we selected a clinically relevant (Cmax) dose of 1 μM crizotinib for the cellular profiling. For each kinase tested, a concentration of energy transfer probe was introduced to the culture medium at an optimized concentration as described in Table S1. Once the appropriate energy transfer probe concentration is determined, the energy transfer technique is easily scaled in a microplate format for single-dose profiling of target occupancy, and can be performed for 178 kinases in a single day (see STAR Methods). Use of plasmid DNA as a source of each cellular kinase fusion circumvents the need for protein purification. Consequently, these features support a relatively simple platform for broad-spectrum profiling in live cells.

17Fractional occupancies were determined for each kinase in live cells, and results were compared with published values from previous biochemical binding studies. Of the 178 kinases in our analysis, only 13 targets were engaged above 50% occupancy at a dose of 1 μM crizotinib (Figure 6A and Table S1). In contrast, previous biochemical analysis of crizotinib suggested that 69 of our queried kinases would be engaged over the same 50% occupancy threshold at 1 μM. As expected, the kinases MET and ALK were highly occupied by 1 μM crizotinib in cells. However, a number of putative collateral crizotinib targets such as LCK and the pseudokinase EPHB6 were unoccupied in cells despite previous reports of strong biochemical affinity (KD = 30 nM and 6 nM, respectively). To establish that the improved spectrum of activity observed in cells was not unique to crizotinib, we also performed a systematic, broad-spectrum profiling of the cellular targets of dasatinib at a clinically relevant dose of 100 nM. Consistent with the observations for crizotinib, our analysis of dasatinib revealed improved selectivity compared with previous biochemical profiling results (Figure 6B and Table S1), yielding 22 targets engaged in cells compared with 44 targets engaged biochemically at the 50% occupancy threshold. Our dasatinib engagement results provided a dataset for comparison with previous work using MS-based profiling approaches, such as those that utilize thermal denaturation or covalent capture. Thermal profiling in live cells failed to identify the BCR-ABL oncoprotein as a direct target of dasatinib, whereas energy transfer confirmed high occupancy of ABL at a 100-nM dose. Our analysis also revealed a broader engagement profile compared with a covalent capture probe approach in live Jurkat cells, which identified that only six kinases occupied >50% at 100 nM dasatinib. Numerous RTKs were engaged with high occupancy with 100 nM dasatinib in our analysis. It is possible that many of these integral membrane proteins could be expressed at low levels or could be incompatible for target capture, and therefore are undetectable by MS. These factors may lead to the narrower spectrum of engagement compared with our results.

18Our cellular profiling results were highly reproducible over the kinase diversity set (Figure S4), supporting the energy transfer technique as a robust approach to live-cell kinase profiling. Moreover, these results support a cellular selectivity mechanism that may be a general feature of ATP-competitive kinase inhibitors, and underscore the value of broad kinase profiling in live cells for determination of inhibitor selectivity.

19The occupancy profile and pronounced potency offset observed for crizotinib suggests that endogenous ATP could be contributing to target engagement interference. However, this interpretation is confounded by a number of interdependent factors resident to cells that must be considered. First, ATP interference is presumably influenced by the intrinsic affinity of ATP for each kinase, representing a key variable for cellular target engagement. Moreover, the KM of ATP is known to be heavily influenced by kinase activation state, which may exist in a dynamic equilibrium in cells. Finally, local ATP concentrations may vary depending on the cellular localization of the target, suggesting that potency shifts may be challenging to functionally deconvolute in a cellular environment and difficult to simulate biochemically.

20Using the energy transfer technique, we performed a mechanistic analysis of ATP interference on cellular kinase engagement. As shown in Figure 6A, the cellular engagement profile for crizotinib revealed a number of putative, high-affinity targets that were surprisingly unoccupied in cells. For example, the affinity of the tyrosine kinase LCK and pseudokinase EPHB6 for crizotinib are reported to be 30 nM and 6 nM, respectively, in biochemical analyses, but both unexpectedly show low cellular occupancy (<10%) at a dose of 1 μM. Conversely, a number of relatively poor-affinity crizotinib targets such as MuSK (230 nM), CASK (140 nM), and TYRO3 (800 nM) displayed high cellular occupancy under these conditions. This discordance suggests that a cellular factor such as ATP may represent a key variable influencing cellular potency for these kinases. We therefore attempted to determine the relative influences of the cosubstrate ATP on engagement profiles for these individual kinases and pseudokinases.

21To investigate this discordance between cellular and acellular engagement, we assessed the influence of ATP on the engagement profile for crizotinib in live cells. This was accomplished by performing live-cell target engagement analysis at physiological ATP levels, along with a similar analysis performed in cells depleted of ATP. To modulate intracellular ATP concentrations, we performed our analysis with live cells depleted of intracellular ATP through inhibition of glycolysis and mitochondrial ATP synthesis. This approach preserves plasma membrane integrity (assessed by exclusion of a cell-impermeable DNA dye) and cellular architecture, but lowers intracellular ATP levels up to 20-fold (Figure S5) over the duration of the assay (assessed by quantitation of total [ATP]). As shown in Figure 7 (and Figure S6), ATP-depleted cells with full membrane integrity showed an enhanced (i.e., left-shifted) engagement profile for crizotinib for the kinases LCK and EPHB6, suggesting interference from cellular ATP. Moreover, analysis using permeabilized cells (more severely depleted of ATP) resulted in a more pronounced left-shifted crizotinib affinity for these kinases (Figures 7 and S6). The KD for crizotinib obtained using permeabilized cells (14 nM and 70 nM for EPHB6 and LCK, respectively) were in close agreement with those reported from biochemical analysis, suggesting that the Nluc tag has minimal influence on the binding behavior of these kinase targets and that the energy transfer approach in a permeabilized cell format provides a suitable biochemical proxy for determination of compound affinity in full-length proteins (Figures S6 and S7). Moreover, we were also able to confirm ATP affinity for both LCK and EPHB6 by titration of exogenous nucleotide in the permeabilized cell format (Figures S6 and S7).

22Compared with the strong affinity targets LCK and EPHB6, the reported biochemical affinity of MuSK for crizotinib is relatively weak. Thus, the high intracellular occupancy of MuSK with 1 μM crizotinib was unexpected. To address this discordance, we investigated whether MuSK was similarly competitive with cellular ATP. In contrast to cellular target engagement results with LCK and EPHB6, MuSK demonstrated negligible interference with ATP, as measured in all ATP-depleted conditions (Figures 7 and S6). Corroborating this result, MuSK showed no observed binding to exogenous ATP titrated into permeabilized cells (Figures S6 and S7). It is noteworthy that the reported KM of ATP for MuSK is highly dependent on kinase activation state (14 μM for active kinase compared with >3,000 μM for inactive kinase), raising the possibility of activity dependence on intracellular target engagement for MuSK. Our results support that MuSK (along with a number of other kinases) may exist in an inactive state that is relatively insensitive to competition with endogenous ATP, at least under unstimulated conditions. Furthermore, these results indicate that reported ATP KM values using isolated kinase domains may not always reflect cellular physiology and the dynamics of intracellular kinase activation state. However, as this analysis was performed in HEK293 cells, it may be challenging to assess activation-state-dependent binding for all queried kinases. Assessments of MuSK engagement in muscle cells could make it possible to investigate engagement potency under active versus inactive kinase states. As new genome editing techniques progress, it should be feasible to introduce this energy transfer technique into such physiologically relevant cell models when required. Our successful use of CRISPR/Cas9 for the engineering of endogenous CDK2 for the energy transfer technique (Figure S2) supports this as a feasible direction.

23In this study, we utilized a live-cell energy transfer technique to perform a broad-coverage, quantitative analysis of intracellular kinase target engagement. Unlike existing methods, this technique is compatible with analysis under equilibrium conditions, thus allowing for determination of fractional target occupancy and affinity for various classes of ATP-competitive inhibitors in live cells. Broad-spectrum analysis of target occupancy uncovered an improved selectivity profile for clinically relevant multikinase inhibitors. A systematic analysis of engagement affinity in ATP-depleted cells allowed us to examine the influence of local cellular ATP concentrations on the putative targets of the multikinase inhibitor crizotinib. Moreover, the target engagement results obtained for crizotinib translated well with published cellular potency values measured via phospho-ELISA. These findings support cellular target engagement profiling as a more accurate predictor of intracellular drug selectivity compared with standard biochemical approaches, offering a potentially valuable tool to guide medicinal chemistry during the development of chemical probes or clinical drug candidates. With a quantitative and accurate kinase target engagement technique, cellular target engagement profiling should be performed as standard practice for human kinases.

24Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact Matthew B. Robers. Matt.Robers@promega.com

25HEK-293 cells (ATCC) or HeLa cells (ATCC) were cultured in DMEM (Gibco) + 10% FBS (GE Healthcare), with incubation in a humidified, 37°C/5% CO2 incubator.

26N- or C-terminal NanoLuc/Kinase fusions were encoded in pFN31K or pFC32K expression vectors (Promega), including flexible Gly-Ser-Ser-Gly linkers between Nluc and each full-length kinase. Optimal orientations for each construct are described in Table S1. For cellular BRET target engagement experiments, HEK-293 or HeLa cells were transfected with NLuc/target fusion constructs using FuGENE HD (Promega) according to the manufacturer’s protocol. Briefly, Nluc/target fusion constructs were diluted into Transfection Carrier DNA (Promega) at a mass ratio of 1:10 (mass/mass), after which FuGENE HD was added at a ratio of 1:3 (μg DNA: μL FuGENE HD). 1 part (vol) of FuGENE HD complexes thus formed were combined with 20 parts (vol) of HEK-293 cells suspended at a density of 2 x 105 per mL, followed by incubation in a humidified, 37°C/5% CO2 incubator for 20 hr. For broad kinase profiling experiments, kinase transfections were performed in 96-well plates using plasmid DNAs arrayed based on energy transfer probe affinity. To simplify the work flow, energy transfer probes were binned based on their optimal concentrations for each kinase target. Based on these groupings, the work flow could be simplified and 178 individual kinases could be queried by a single person, in a single day. The analysis can be performed without any automated liquid handling instruments or robotics. Following transfection, cells were washed and resuspended in Opti-MEM. BRET assays were performed in white, 96-well plates (Corning) at a density of 2 x 104 cells/well. All chemical inhibitors were prepared as concentrated stock solutions in DMSO (Sigma-Aldrich) and diluted in Opti-MEM (unless otherwise noted) to prepare working stocks. Cells were equilibrated for 2 hr with energy transfer probes and test compound prior to BRET measurements. Energy transfer probes were prepared at a working concentration of 20X in tracer dilution buffer (12.5 mM HEPES, 31.25% PEG-400, pH 7.5). For target engagement analysis, the energy transfer probes were added to the cells at concentrations optimized for each target, as described in Table S1. For analysis of DDR1 and DDR2 with compound 6j, energy transfer probe 6 was used at a concentration of 300 nM and 330 nM, respectively. To measure BRET, NanoBRET NanoGlo Substrate and Extracellular NanoLuc Inhibitor (Promega) were added according to the manufacturer’s recommended protocol, and filtered luminescence was measured on a GloMax Discover luminometer equipped with 450 nm BP filter (donor) and 600 nm LP filter (acceptor), using 0.5 s integration time. Milli-BRET units (mBU) are calculated by multiplying the raw BRET values by 1000. Apparent tracer affinity values (EC50) were determined using the sigmoidal dose-response (variable slope) equation available in GraphPad Prism (Equation 1);

27Competitive displacement data were then plotted with GraphPad Prism software and data were fit to Equation 1 to determine the IC50 value.

28For fractional occupancy determination in kinase profiling experiments, the following equation (Equation 2) was used;where X = BRET in the presence of the test compound and energy transfer probe, Y = BRET in the presence of only energy transfer probe, and Z = BRET in the absence of the energy transfer probe and test compound. Predicted biochemical occupancy at the chosen drug dose for kinase profiling experiments was determined from the Kd value reported previously using a variation of the Langmuir Isotherm (Equation 3);

29For all BRET data shown, no individual data points were omitted.

30In order to evaluate the effect of endogenous ATP on target engagement parameters, cells were depleted of ATP by treatment with rotenone, a known inhibitor of mitochondrial ATP synthesis that blocks complex I. This method was adapted from previous reports that use a glucose-free galactose medium to sensitize cultured mammalian cells to mitochondrial toxicants, and adjusted to fit the BRET based target engagement workflow described above. Briefly, HEK-293 cells expressing the kinase/NLuc fusion of interest were prepared by transient transfection as described in the Cell transfection, treatments, and BRET measurements section. The cells were harvested by trypsinization, pelleted by centrifugation, and resuspended in galactose medium, which consisted of 1 mM HEPES (Gibco), 2 mM L-glutamine (Gibco), 1 mM sodium pyruvate (Gibco), and 10 mM D-(+)-galactose (Sigma) in glucose-free DMEM without phenol red (Gibco A14430-01). The cell density was adjusted to 2 x 105 cells/mL, after which 1000X rotenone (1 mM in DMSO) was added to achieve a final concentration of 1 μM in the cell suspension. Subsequently, the cells were used in target engagement workflows as described above in the cell transfection, treatments, and BRET measurements section, except that working stocks of all chemical inhibitors, NanoGlo Substrate, and Extracellular NanoLuc Inhibitor were prepared in galactose medium rather than OptiMEM. Relative ATP levels and membrane integrity of the cell suspension were monitored using a multiplex of the Cell Titer-Glo 2.0 Assay (Promega) and the CellTox Green assay (Promega) performed as described in the manufacturer’s protocols. After a 2 hour treatment, this protocol routinely reduces the total ATP in the cell suspension by 90−95% relative to those maintained in glucose rich media, and without loss of membrane integrity (Figure S5). For assays in permeabilized format, cells were permeabilized by treatment with 50 μg/mL digitonin, and Extracellular NanoLuc Inhibitor was omitted from the NanoGlo Substrate solution.

31To assess the target engagement behavior of CDK2-NanoLuc fusion expressed at endogenous levels, NanoLuc was introduced as a fusion by homologous directed repair of a CRISPR-Cas9 double-strand break at the CDK2 locus in HEK293 cells. The genome edit was designed to insert the NanoLuc coding sequence and thereby create a fusion of NanoLuc to the C-terminus of the CDK2 protein. The Cas9 guide sequence crRNA was directed to the sequence GGCTATCAGAGTCGAAGATG (the PAM site is GGG) and generates a double strand break 9 bp upstream of the stop codon for CDK2. The donor DNA sequence for homologous directed repair was contained in a circular double-stranded plasmid DNA, and included 500bp of genomic CDK2 DNA sequence upstream of the stop codon, the NanoLuc coding sequence including a four amino acid flexible peptide linker (GSSG), and 450bp of CDK2 3’UTR genomic sequence. Cas9 protein (Aldevron His-sNLS-Myc-spCas9) was incubated with tracrRNA:crRNA (IDT) at equimolar ratios to create a ribonucleoprotein (RNP) complex. The RNP and donor DNA plasmid was introduced to 2x105 HEK293 cells via electroporation using a Lonza 4D NucleoFector. After 6 days, the electroporated cells were diluted and plated in 96 well plates to isolate single colonies. Single colony isolates showing light production three-fold over background were collected and tested for stable growth and light generation, and a small number of clonal cell lines were screened by PCR for NanoLuc insertion using CDK2 genomic primers that were outside the homology regions used in the donor DNA plasmid. One clonal cell line, B10, was isolated that was homozygous for insertion, Figure 1A. The PCR amplicon from the homozygous insertion CDK2-NanoLuc cell line was remade with the high fidelity DNA polymerase Q5 (New England Biolabs) and ligated into a bacterial plasmid for DNA sequencing. 24 plasmid clones were sequenced, and all 24 showed perfect homology throughout the CDK2 regions upstream and downstream of the NanoLuc insertion point. Protein was isolated from the homozygous insertion CDK2-NanoLuc cell line to perform a Western using anti-CDK2 (Figure S2). The antibody stained blot showed that CDK2-staining band now ran at the expected higher molecular weight of the CDK2-NanoLuc fusion. PCR Characterization; Primers to the CDK2 locus were directed to 540 bp upstream of the CDK2 stop codon (GACCAGCTCTTCCGGATCTTTCGGA) and 567 bp downstream of the CDK2 stop codon (GCTGAGTCTTCAGTCTCCCAGCCT). In HEK-293, the resulting amplicon is 1,110 bp; a successful insertion of the NanoLuc coding region, 534bp, would result in an amplicon size of 1,644 bp. Clonal line B10 shows only the 1,644 bp amplicon, and represents homozygous insertion of NanoLuc with no “wild-type” amplicon. Clonal line C5 is included to illustrate a heterozygous insertion of NanoLuc with both amplicons present. The PCR conditions used 100ng of genomic DNA combined with 20 pmol of each primer and GoTaq® Hot Start Green Master Mix (Promega) in a total volume of 50μl. 30 cycles of PCR were performed with an annealing temperature of 61°C for 20 seconds and an extension temperature of 72°C for 1 minute. 5 μl of the reactions were run on a 1.2% agarose gel stained with Ethidium Bromide; image capture was done with a LAS4010 (GE). Western Blots were performed to assess the molecular weight of the CDK2-NanoLuc fusion protein in the clonal cell lines. The expected molecular weight of CDK2 is 34KDa and the CDK-NanoLuc fusion is 53KDa. Cells from the clonal lines B10 and C5 were harvested by trypsinization, the cell density determined, and 1−2 x 106 cells were pelleted, washed with PBS, pelleted again and the cell pellets were frozen at −80°C for storage. Samples were thawed and re-suspended in PBS containing 0.1% BSA, then added to an equal volume of Loading Buffer (Bio-RAD). Samples were separated by SDS/PAGE (BIO-RAD) according to the manufacturer’s recommendations and included the SeeBlue Plus2 molecular weight marker (Invitrogen). Western blots were performed using the SuperSignal West Pico Detection Kit (Thermo) following the manufacturers recommendations. Briefly, after electrophoresis, proteins were transferred to a PVDF membrane and blocked overnight at 4°C. Blots were incubated in primary antibody (Cell Signaling Technologies #25465) for one hour at room temperature on a rocking platform, then washed six times. Blots were similarly incubated in secondary antibody (HRP-conjugate, 1:1,000 dilution), washed six times and briefly exposed to ECL substrate. Image capture was done with a LAS4010 (GE). The loading control for the CDK2 Western is shown immediately below and followed the same protocol as above, but the primary antibody was against p38 (Abcam ab170099).

32All imaging experiments were performed using the LV200 bioluminescence imaging system (Olympus) equipped with an ImagEM X2 EM-CCD camera (Hamamatsu) and a 60x, 1.4 NA objective. HEK293 cells transiently transfected with the indicated expression construct (1:10 dilution in carrier DNA) were plated onto Nunc Lab-Tek II 8-well chambered coverslips (Thermo Fisher) in 400 μl growth medium (DMEM + 10% FBS) at a density of 8Χ105 cells per well. After 24 h of incubation at 37 °C, 100 μl of Nano-Glo Live Cell Reagent (Promega) was added. All images were acquired with cellSens software (Olympus) using electron multiplying (EM) gain of 600 and an exposure time of 3 seconds. Each image was generated using a average projection of 10 images. Generation of average projections, linear adjustments of dynamic range, and pseudocolor rendering were performed using Image J image processing software (Fiji package). Scale bar = 20 μm.

33All solvents were purchased from Sigma or Fisher Scientific and used without purification. NanoBRET 590 SE was obtained from Promega Corp. Madison, WI. 1H-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, q-quartet 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.

342-(4-((4-((4-aminobutyl)amino)-6-((5-cyclopropyl-1H-pyrazol-3-yl)amino)pyrimidin-2-yl)amino)phenyl)acetonitrile

35To a 5 mL pressure tube was added 1,4-diaminobutane (0.9 g, 10.3 mmol) followed by 2-(4-((4-chloro-6-((5-cyclopropyl-1H-pyrazol-3-yl)amino)pyrimidin-2-yl)amino)phenyl)acetonitrile (50.0 mg, 0.14 mmol). The tube was capped and the mixture was heated to 120 °C overnight. The tube was cooled to ambient temperature, uncapped and was diluted with 50 mL of ethyl acetate, giving a yellow precipitate. The crude mixture was transferred to a 250 mL separatory funnel and was washed with a saturated aqueous solution of ammonium chloride (25 mL). The organic phase was retained, washed with 25 mL of water then 25 mL of brine solution. The organic phase was dried over Na2SO4, filtered and concentrated to give a yellow-white solid. The solid was dissolved in water (8.0 mL) and was subjected to RP-HPLC preparative purification using Standard Method 1. Product containing fractions were analyzed by HPLC, pooled and were concentrated to afford the product (22.7 mg, 39.8%) as a yellow solid. 1H NMR (400 MHz, methanol-d) δ 7.90 (s, 1H), 7.63 (d, J = 8.3 Hz, 2H), 7.41 (d, J = 8.3 Hz, 2H), 5.60 (s, 1H), 3.91 (s, 2H), 3.45 (s, 2H), 2.98 (m, 2H), 1.93 (m, 1H), 1.73 (m, 4H), 1.02 (m, 2H), 0.72 (m, 2H). HRMS (ESI) calcd for C22H28N9 [M+H]+: 418.2468. Found: 418.2452.

362-(4-((4-((4-aminobutyl)amino)-6-((5-cyclopropyl-1H-pyrazol-3-yl)amino)pyrimidin-2-yl)amino)phenyl)acetonitrile-NanoBRET 590(Energy Transfer Probe 1)

37To a stirred solution of 2-(4-((4-((4-aminobutyl)amino)-6-((5-cyclopropyl-1H-pyrazol-3-yl)amino)pyrimidin-2-yl)amino)phenyl)acetonitrile (3.9 mg, 0.009 mmol) in anhydrous DMF (1.0 mL) was added N-N-diisopropylethylamine (8.4 μL, 0.05 mmol). The mixture was stirred for 10 min and NanoBRET 590 SE (4.4 mg, 0.01 mmol) was added. The reaction mixture was capped and stirred in the dark for 3 hrs. The reaction was quenched with 6 mL of a solution of acetonitrile, water, TFA (1:1:0.01) and was subjected to RP-HPLC preparative purification using Standard Method 1. Product containing fractions were analyzed by HPLC for purity, pooled and were concentrated under reduced pressure affording compound 1 (2.8 mg, 40.9%) as a blue film. 1H NMR (400 MHz, methanol-d) δ 7.58 (m, 2H), 7.37 (d, J = 8.3 Hz, 2H), 7.16 (m, 4H), 6.97 (d, J = 4.6 Hz, 1H), 6.87 (m, 1H), 6.34 (m, 2H), 6.30 (d, J = 4.0 Hz, 1H) 3.41 (s, 2H), 2.62 (t, J = 7.5 Hz, 1H) 1.91 (m, 1H), 1.59 (s, 4H), 1.29 (s, 1H), 1.02 (m, 2H), 0.90 (m, 2H), 0.72 (m, 2H). HRMS (ESI) calcd for C38H40BF2N12O [M+H]+: 729.3509. Found: 729.3512.

38tert-butyl (18-(4-((6-(2,6-dichlorophenyl)-8-methyl-7-oxo-7,8-dihydropyrido[2,3-d]pyrimidin-2-yl)amino)phenoxy)-15-oxo-3,6,9,12-tetraoxa-16-azaoctadecyl)carbamate

39To a 5 mL amber vial was added 2,5-dioxopyrrolidin-1-yl 2,2-dimethyl-4-oxo-3,8,11,14,17-pentaoxa-5-azaicosan-20-oate (10.0 mg, 0.022 mmol), 2-((4-(2-aminoethoxy)phenyl)amino)-6-(2,6-dichlorophenyl)-8-methylpyrido[2,3-d]pyrimidin-7(8H)-one (4.7 mg, 0.01 mmol) and DMF (0.5 mL). To the stirred solution was added N-N-diisopropylethylamine (9.0 μL, 0.05 mmol) and the mixture was allowed to stir for 2 hrs. The reaction mixture was diluted with a solution of acetonitrile, water, TFA (1:1:0.01) to 6 mL and was subjected to RP-HPLC preparative purification using Standard Method 1. Product containing fractions were analyzed by HPLC for purity, pooled and concentrated to afford the product (5.2 mg, 62.8%) as a yellow solid. HRMS (ESI) calcd for C38H49Cl2N6O9 [M+H]+: 803.2938. Found: 803.2929.

401-amino-N-(2-(4-((6-(2,6-dichlorophenyl)-8-methyl-7-oxo-7,8-dihydropyrido[2,3-d]pyrimidin-2-yl)amino)phenoxy)ethyl)-3,6,9,12-tetraoxapentadecan-15-amide, TFA salt

41tert-butyl (18-(4-((6-(2,6-dichlorophenyl)-8-methyl-7-oxo-7,8-dihydropyrido[2,3-d]pyrimidin-2-yl)amino)phenoxy)-15-oxo-3,6,9,12-tetraoxa-16-azaoctadecyl)carbamate was taken up in DCM (1.0 mL) and diisopropylsilane (100 μL) was added. To the stirred solution was added trifluoroacetic acid (1.0 mL) and the reaction was allowed to stir for 90 min. Volatiles were removed under reduced pressure giving a colorless oil that was treated with diethyl ether (10 mL) affording a white solid. The ether was decanted and the resulting solid was dried under high vacuum for 10 min. An additional 10 mL portion of diethyl ether was added to the reaction residue, decanted and the resulting solid was dried under high vacuum for 1 hr.

421-amino-N-(2-(4-((6-(2,6-dichlorophenyl)-8-methyl-7-oxo-7,8-dihydropyrido[2,3-d]pyrimidin-2-yl)amino)phenoxy)ethyl)-3,6,9,12-tetraoxapentadecan-15-amide-NanoBRET 590 (Energy Transfer Probe 2)

43Crude 1-amino-N-(2-(4-((6-(2,6-dichlorophenyl)-8-methyl-7-oxo-7,8-dihydropyrido[2,3-d]pyrimidin-2-yl)amino)phenoxy)ethyl)-3,6,9,12-tetraoxapentadecan-15-amide-TFA salt was taken up in DMF (1.0 mL), was treated with N-N-diisopropylethylamine (11.3 μL, 0.07 mmol) and the mixture was stirred for 10 min. NanoBRET 590 SE (3.3 mg, 0.008 mmol) was added and the mixture was allowed to stir in the dark for 2 hr. The reaction mixture was diluted with a solution of acetonitrile, water, TFA (1:1:0.01) to 6 mL and was subjected to RP-HPLC prep purification using Standard Method 1. Product containing fractions were analyzed by HPLC for purity, pooled and were concentrated under reduced pressure affording compound 2 (1.2 mg, 18% -two steps) as a blue film. 1H NMR (400 MHz, methanol-d) δ 8.68 (s, 1H), 7.73 (s, 1H), 7.65 (s, 1H), 7.50 (d, J = 0.9 Hz, 1H), 7.48 (s, 1H), 7.40 (m, 1H), 7.19 (m, 1H) 7.18 (s, 1H) 7.15 (s, 1H), 6.95 (m, 3H), 6.88 (d, J = 3.9 Hz, 1H), 6.33 (m, 1H), 6.30 (d, J = 4.0 Hz, 1H), 4.04 (t, J = 10.8 Hz, 2H), 3.69 (m, 2H), 3.55 (m, 17H), 3.61 (m, 3H), 3.26 (m, 2H), 2.63 (t, J = 7.6 Hz, 2H), 2.45 (t, J = 6.0 Hz, 2H). HRMS (ESI) calcd for C49H53BCl2F2N9O8 [M+H]+: 1014.3455. Found: 1014.3461.

443-(4-morpholino-6-(piperazin-1-ylmethyl)thieno[3,2-d]pyrimidin-2-yl)phenol-NanoBRET 590 (Energy Transfer Probe 3)

45To a stirred solution of 3-(4-morpholino-6-(piperazin-1-ylmethyl)thieno[3,2-d]pyrimidin-2-yl)phenol (8.0 mg, 0.019 mmol) in anhydrous DMF (1.0 mL) was added N-N-diisopropylethylamine (17.0 μL, 0.097 mmol). The mixture was stirred for 10 min, NanoBRET 590 SE (9.1 mg, 0.021 mmol) was added and the reaction mixture was capped and stirred in the dark for 2 hr. The reaction was quenched with 6 mL of a solution of acetonitrile, water, TFA (1:1:0.01) and was subjected to RP-HPLC preparative purification using Standard Method 1. Product containing fractions were analyzed by HPLC for purity, pooled and were concentrated under reduced pressure affording compound 3 (9.9 mg, 70.0%) as a blue film. 1H NMR (400 MHz, methanol-d) δ 7.70 (s, 1H) 7.67 (s, 1H), 7.47 (s, 1H), 7.43 (t, J = 7.3 Hz, 1H), 7.25 (s, 1H), 7.21 (m, 2H), 7.18 (m, 1H), 7.10 (m, 1H), 7.02 (d, J = 4.6 Hz, 1H), 6.93 (D, J = 4.0 Hz, 1H), 6.33 (m, 2H), 4.27 (t, J = 4.8 Hz, 4H), 4.08 (s, 2H), 3.92 (t, J = 5.0 Hz, 4H), 3.71 (s, 2H), 3.64 (s, 1H), 2.86 (t, J = 7.8 Hz, 2H), 2.68 (t, J = 5.9 Hz, 4H). HRMS (ESI) calcd for C37H38BF2N8O3S [M+H]+: 723.2849. Found: 723.2830.

46N-(3-((4-(3-aminopropyl)piperazin-1-yl)methyl)-2-(trifluoromethyl)phenyl)-3-(imidazo[1,2-b]pyridazin-3-ylethynyl)-4-methylbenzamide-NanoBRET 590 (Energy Transfer Probe 4)

47N-N-diisopropylethylamine (5.6 μL, 0.032 mmol) was added to a solution of N-(4-((4-(3-aminopropyl)piperazin-1-yl)methyl)-3-(trifluoromethyl)phenyl)-3-(imidazo[1,2-b]pyridazin-3-ylethynyl)-4-methylbenzamide (7.5 mg, 0.013 mmol) in anhydrous DMF (1 mL) and the mixture stirred for 10 min. NanoBRET 590 SE (5.2 mg, 0.012 mmol) was added and the reaction stirred in the dark for 4 hrs followed by removal of the solvent under vacuum. The crude product was purified via preparative TLC (5 % methanol in dichloromethane) affording compound 4 (10.7 mg, 93 %) as a purple solid. 1H NMR (400 MHz, CDCl3) δ 10.40 (s, 1H), 8.50 (dd, J = 4.4, 1.6 Hz, 1H), 8.07 (m, 2H), 8.01 (dd, J = 9.2, 1.6 Hz, 1H), 7.93 (br s, 1H), 7.82 (m, 3H), 7.63 (d, J = 9.0 Hz, 1H), 7.42 (d, J = 8.0 Hz, 1H), 7.19 – 7.12 (m, 3H), 7.03 (d, J = 4.6 Hz, 1H), 7.00 – 6.95 (m, 2H), 6.85 (m, 2H), 6.40 – 6.28 (m, 2H), 3.64 (s, 2H), 3.50 (s, 2H), 3.35 (t, J = 7.3 Hz, 4H), 2.43 – 2.36 (m, 9H), 1.62 (p, J = 6.2 Hz, 2H), 1.26 (s, 4H). MS (ESI) calcd for C47H45BF5N10O2 [M + H]+: 887.37. Found: 887.73.

48Tert-butyl (15-(4-(4-((5-chloro-4-((2-(methylcarbamoyl)phenyl)amino)pyrimidin-2-yl)amino)phenyl)piperazin-1-yl)-15-oxo-3,6,9,12-tetraoxapentadecyl)carbamate.

49To a round bottom flask was charged 2,2-dimethyl-4-oxo-3,8,11,14,17-pentaoxa-5-azaicosan-20-oic acid (20.0 mg, 0.046 mmol), 2-((5-chloro-2-((4-(piperazin-1-yl)phenyl)amino)pyrimidin-4-yl)amino)-N-methylbenzamide (25.0 mg, 0.069 mmol) and HATU (20.8 mg,.058 mmol). The solids were taken up in anhydrous DMF (1.0 mL), N,N-diisopropylethylamine (23.9 μL, 0.137 mmol) was added and the mixture was allowed to stir for 3 hr. Volatiles were removed under reduced pressure and the resulting residue was dried under high vacuum for 2 hr. The crude was subjected to flash chromatography (0% to 10% methanol/DCM gradient). Product containing fractions were pooled and concentrated affording the product (25.0 mg, 55.8%) as a white solid. 1H NMR (400 MHz, CDCl3) δ 11.19 (s, 1H), 8.63 (d, J = 8.4 Hz, 1H), 8.00 (s, 1H), 7.68 (m, 1H), 7.50 (d, J = 1.6 Hz, 1H), 7.44 (d, J = 8.6 Hz, 2H), 7.37 (t, J = 7.9 Hz, 1H), 7.06 (t, J = 7.5 Hz, 1H), 6.88 (d, J = 8.7 Hz, 2H), 6.46 (m, 1H), 5.15 (s, 1H), 3.79 (m, 6H), 3.59 (m, 2H), 3.50 (m, 4H) 3.28 (m, 4H), 3.09 (m, 4H), 3.00 (d, J = 4.8 Hz, 2H), 2.64 (m, 3H), 1.56 (d, J = 6.7 Hz, 2H), 1.43 (s, 15H) 1.05 (m, 2H). MS (ESI) calcd for C38H54ClN8O8 [M + H]+: 785.38. Found: 785.60.

502-((2-((4-(4-(1-amino-3,6,9,12-tetraoxapentadecan-15-oyl)piperazin-1-yl)phenyl)amino)-5-chloropyrimidin-4-yl)amino)-N-methylbenzamide-TFA salt

51Tert-butyl (15-(4-(4-((5-chloro-4-((2-(methylcarbamoyl)phenyl)amino)pyrimidin-2-yl)amino)phenyl)piperazin-1-yl)-15-oxo-3,6,9,12-tetraoxapentadecyl)carbamate (25.0 mg, 0.032 mmol) was dissolved in 2.5 mL of DCM and triisopropylsilane (250 μL, 0.031 mmol) and TFA (2.5 mL) was added. The mixture was capped and stirred for 3 hr. Volatiles were removed under reduced pressure giving a yellow oil and 5 mL of diethyl ether was added giving a white precipitate. The ether was decanted and an additional 5 mL portion was added. The suspension was sonicated and the ether was decanted. The resulting residue was dried under high vacuum overnight affording the product (18.9 mg, 77.9%) as a white solid. MS (ESI) calcd for C33H46ClN8O6 [M + H]+: 785.23. Found: 785.32.

522-((2-((4-(4-(1-amino-3,6,9,12-tetraoxapentadecan-15-oyl)piperazin-1-yl)phenyl)amino)-5-chloropyrimidin-4-yl)amino)-N-methylbenzamide-NanoBRET 590 (Energy Transfer Probe 5)

53Primary amine 2-((2-((4-(4-(1-amino-3,6,9,12-tetraoxapentadecan-15-oyl)piperazin-1-yl)phenyl)amino)-5-chloropyrimidin-4-yl)amino)-N-methylbenzamide (55.0 mg, 0.08 mmol) was charged into a round bottom flask and was taken up in anhydrous DMF (5.0 mL). To the stirred solution was added N,N-diisopropylethylamine (69.9 uL, 0.40 mmol) then NaoBRET 590 SE (41.1 mg, 0.096 mmol). The mixture was capped and stirred for 2 hr, judged complete by the disappearance of the starting amine by HPLC. The reaction was quenched with 6 mL of a solution of acetonitrile, water, TFA (1:1:0.01) and was subjected to RP-HPLC prep purification using Standard Method 1. Product containing fractions were analyzed by HPLC for purity, pooled and were concentrated under reduced pressure affording the product 5 (64.8 mg, 81.0%) as a blue film. 1H NMR (400 MHz, DMSO-d) δ 11.60 (s, 1H), 11.40 (s, 1H), 9.24 (s, 1H), 8.73 (m, 2H), 8.16 (s, 1H), 8.00 (t, J = 5.6 Hz, 1H), 7.74 (dd, J = 1.6 Hz, J = 7.9 Hz, 1H), 7.48 (m, 4H), 7.43 (s, 1H), 7.36 (m, 1H), 7.32 (d, J = 4.6 Hz, 1H), 7.16 (d, J = 4.6 Hz, 1H), 7.12 (m, 1H), 7.01 (d, J = 4.0 Hz, 1H), 6.90 (d, J = 9.0 Hz, 2H), 6.33 (m, 2H), 3.61 (m, 6H), 3.51 (m, 2H), 3.41 (t, J = 5.8 Hz, 2H), 3.21 (m, 2H), 3.13 (m, 2H), 3.06 (m, 2H), 3.01 (m, 2H), 2.81 (d, J = 4.4 Hz, 2H), 2.61 (t, J = 6.6 Hz, 2H). MS (ESI): calcd for C49H58BClF2N11O7 [M]+ 996.43. Found: 996.55.

54Tert-butyl (2-(2-hydroxyethoxy)ethyl)carbamate

55To a stirred solution of 2-(2-Aminoethoxy)ethanol (27.5 g, 261.6 mmol) in EtOAc (100 mL), cooled to 0°C, was added a mixture of Di-tert-butyl dicarbonate (60.0 g, 274.6 mmol) in EtOAc (100 mL) dropwise over 1 hr. The mix warmed to rt and was left stirred for 3.5 hr. The crude mixture was washed with 3 × 200 mL portions of water then portion of brine (100 mL). The organic phase was dried over Na2SO4 and volatiles were removed under reduced pressure to provide 52.0 g of the Boc-amine (52.01 g, 96.9%) as a colorless oil. 1H NMR (300 MHz, CDCl3) δ 3.74 (t, J = 3.0 Hz, 2H), 3.56 (m, 4H), 3.31 (m, 2H), 1.45 (s, 9H).

56Tert-butyl (2-(2-bromoethoxy)ethyl)carbamate

57To a stirred solution of tert-butyl (2-(2-hydroxyethoxy)ethyl)carbamate (10.0 g, 48.7 mmol) in THF (100 mL) was added tertabromomethane (24.24 g, 73.1 mmol). The resulting solution was cooled to 0°C and triphenylphosphine (19.17 g, 73.1 mmol) was added in 3 x 5 g portions at 0 min, 15 min and 30 min. At 45 min, a 4.17 g was added. The reaction was allowed to warm to rt and left stirred for 6 hr. The crude reaction mixture was filtered and the filtrate was concentrated to a yellow solid. The solid was dissolved in DCM, absorbed on celite and was dried. The compound was subjected to flash chromatography (0 to 5% MeOH/DCM) affording the product (8.76 g, 67.1%) as a yellow oil. 1H NMR (400 MHz, CDCl3) δ 3.78 (t, J = 6.0 Hz, 2H), 3.56 (t, J = 5.12 Hz, 2H), 3.47 (t, J = 6.0 Hz, 2H) 3.33 (m, 2H), 1.45 (s, 9H).

58Tert-butyl (2-(2-(4-(2,2,2-trifluoroacetyl)piperazin-1-yl)ethoxy)ethyl)carbamate

59To a solution of tert-butyl (2-(2-bromoethoxy)ethyl)carbamate (8.83 g, 32.9 mmol) and trifluoroacetylpiperazine (5.0 g, 27.5 mmol) in ACN (50 mL) was added cesium carbonate (10.7 g, 32.9 mmol) and KI (227.8 g, 1.4 mmol). The mixture was heated to 60°C with stirring for 4 hr. Volatiles were removed under reduced pressure and the resulting residue was dried under hi-vacuum. To the white residue was added EtOAc (100 mL) giving a white suspension that was filtered. The filtrate was washed with two portions of water (50 mL), dried over Na2SO4 and was concentrated to a yellow oil that was diluted with DCM, absorbed on celite and was subjected to flash chromatography (0 to 5% MeOH/DCM) affording the product (8.15 g, 80.4%) as a colorless oil. 1H NMR (400 MHz, CDCl3) δ 3.67 (t, J= 5.2, 2H), 3.59 (t, J = 5.0 Hz, 2H), 3.54 (t, J = 5.4 Hz, 2H), 3.47 (t, J = 5.2 Hz, 2H), 3.25 (m, 4H), 2.55 (t, J = 5.4 Hz, 2H), 2.51 (m, 4H), 1.39 (s, 9H).

60Tert-butyl (2-(2-(piperazin-1-yl)ethoxy)ethyl)carbamate

61To a stirred solution of tert-butyl (2-(2-(4-(2,2,2-trifluoroacetyl)piperazin-1-yl)ethoxy)ethyl)carbamate (5.0 g, 13.5 mmol) in methanol (30 mL) and water (2 mL) was added potassium carbonate (9.35 g, 67.7 mmol). The mixture was stirred for 2 hr then concentrated to dryness. To the residue was added EtOAC (100 mL) giving a white suspension that was filtered then concentrated to ∼10 mL, absorbed on celite and dried. The crude was subjected to flash chromatography (0 to 5% MeOH/DCM) affording the product (2.6 g, 70.0%) as a colorless oil. 1H NMR (400 MHz, CDCl3) δ 3.59 (t, J = 5.6 Hz, 2H), 3.51 (t, J = 5.2 Hz, 2H) 3.30 (m, 2H), 2.92 (t, J = 4.9 Hz, 4H), 2.56 (t, J = 5.7Hz, 2H), 2.48 (m, 4H), 1.44 (s, 9H).

62Tert-butyl (2-(2-(4-(6-((5-((2-chloro-6-methylphenyl)carbamoyl)thiazol-2-yl)amino)-2-methylpyrimidin-4-yl)piperazin-1-yl)ethoxy)ethyl)carbamate

63Tert-butyl (2-(2-(piperazin-1-yl)ethoxy)ethyl)carbamate (2.9 g, 10.7 mmol) and N-(2-Chloro-6-methylphenyl)-2-[(6-chloro-2-methyl-4-pyrimidinyl)amino]-5-thiazolecarboxamide (1.1 g, 2.7 mmol) were charged into a round bottom flask. To the solids was added 2-propanol (5 mL), ethylene glycol (5 mL) and diisopropylethylamine (1.9 mL, 10.7 mmol). The mixture was heated to 110°C overnight giving a yellow solid after cooling. To the solid was added 50 mL of water giving a yellow solid that was collected by vacuum filtration and dried under high vacuum overnight. The crude residue was taken up in DCM/methanol, absorbed on celite and was dried to a free-flowing solid. The mixture was subjected to flash chromatography (0 to 5% MeOH/DCM) affording the product (1.08 g, 64.3%) as a white solid. 1H NMR (400 MHz, CDCl3) δ 9.91(s, 1H), 9.85 (s, 1H), 8.24 (s, 1H), 7.41 (d, J = 1.9 Hz, 1H), 7.29 (m, 2H), 6.88 (t, J = 7.6 Hz, 1H), 6.15 (s, 1H), 4.35 (d, J = 9.4 Hz, 9.4 Hz, 2H), 3.74 (d, J = 4.7 Hz, 2H), 3.61 (m, 2H), 3.47 (t, J = 5.8 Hz, 2H) 3.37 (m, 2H), 3.28 (m, 2H), 3.14 (m, 4H), 2.45 (s, 3H), 2.24 (s, 3H) 1.39 (s, 9H). MS (ESI): calcd for C29H40ClN8O4S [M+H]+: 631.26. Found: 631.25.

642-((6-(4-(2-(2-aminoethoxy)ethyl)piperazin-1-yl)-2-methylpyrimidin-4-yl)amino)-N-(2-chloro-6-methylphenyl)thiazole-5-carboxamide-TFA salt

65Tert-butyl (2-(2-(4-(6-((5-((2-chloro-6-methylphenyl)carbamoyl)thiazol-2-yl)amino)-2-methylpyrimidin-4-yl)piperazin-1-yl)ethoxy)ethyl)carbamate (75.0 mg, 0.119 mmol) was dissolved in 2.5 mL of DCM and triisopropylsilane (250 μL, 0.031 mmol) and TFA (2.5 mL) was added. The mixture was capped and stirred for 3 hr. Volatiles were removed under reduced pressure giving a yellow oil and 5 mL of diethyl ether was added giving a white precipitate. The ether was decanted and an additional 5 mL portion was added. The suspension was sonicated and the ether was decanted. The resulting residue was dried under high vacuum overnight affording the product (63.1 mg, 100%) as a yellow solid. MS (ESI): calcd for C24H32ClN8O2S [M+H]+: 531.21. Found: 531.18.

662-((6-(4-(2-(2-aminoethoxy)ethyl)piperazin-1-yl)-2-methylpyrimidin-4-yl)amino)-N-(2-chloro-6-methylphenyl)thiazole-5-carboxamide-NanoBRET 590 (Energy Transfer Probe 6)

67To a solution of 2-((6-(4-(2-(2-aminoethoxy)ethyl)piperazin-1-yl)-2-methylpyrimidin-4-yl)amino)-N-(2-chloro-6-methylphenyl)thiazole-5-carboxamide (10.0 mg, 0.024 mmol) in anhydrous DMF (2 mL) was added N,N-diisopropylethylamine (8.2 μL, 0.047 mmol). The solution was stirred for ten minutes, NanoBRET 590 SE (10.0 mg, 0.024 mmol) was added, the mixture was capped and stirred for 2 hrs, judged complete by the disappearance of the starting amine by HPLC. The reaction was quenched with 6 mL of a solution of acetonitrile, water, TFA (1:1:0.01) and was subjected to RP-HPLC preparative purification using Standard Method 1. Product containing fractions were analyzed by HPLC for purity, pooled and were concentrated under reduced pressure affording the product (11.6 mg, 73.2%) as a blue film. 1H NMR (400 MHz, d6 DMSO-d) δ 11.46 (s, 1H), 11.41 (s, 1H), 9.87 (s, 1H), 8.21 (s, 1H), 8.01 (t, J = 5.7 Hz, 1H), 7.44 (dd, J = 2.0 Hz, J = 7.5 Hz, 1H), 7.37 (s, 1H), 7.33 (d, J = 4.5 Hz, 1H), 7.27 (m, 3H), 7.16 (d, J = 4.6 Hz, 1H), 7.01 (d, J = 3.9 Hz, 1H), 6.34 (d, J = 4.0 Hz, 2H), 6.04 (s, 1H), 3.54 (t, J = 6.0 Hz, 2H), 3.50 (s, 4H), 3.43 (t, J = 6.0 Hz, 2H), 3.29 (s, 1H), 3.24 (m, 2H), 3.15 (m, 3H), 2.54 (s, 4H), 2.40 (s, 3H), 2.24 (s, 3H). MS (ESI): calcd for C40H44BClF2N11O3S [M+H]+: 842.31. Found: 842.24

68Data from multiple independent experiments (N) are presented as mean values +/- standard error of the mean (SE) and data involving technical replicates are presented as mean +/- standard deviation (SD) as indicated in the figure legends. The number of experimental or technical replicates for each experiment is also described in each individual figure legend. Apparent affinity values were determined using the sigmoidal dose-response (variable slope) equation available in GraphPad Prism (Version 7). Linear regression analyses were determined using Graphpad Prism (Version 7).

69J.D.V. and M.B.R. designed experiments and wrote the paper. J.D.V., C.R.C., J.W., C.A.Z., J.R.H., M.R.I., K.Z., T.M., T.A.K., K.G.H., R.F.O., M.S., P.O., M.C., C.I.W., B.-T.B., T.H., C.G., K.D., D.H.D., K.V.M.H., T.M.W., S.K., S.M., P.L.M., F.F., K.V.W., M.B.R. contributed to the design and/or execution of experiments.