Proteome-wide Identification of Off-Targets of a Potent EGFR^L858R/T790M Mutant Inhibitor

Abstract

Target identification is an essential step in drug discovery. It facilitates an understanding of drug action and potential toxicities and offers opportunities to repurpose drug candidates. HP-1, a potent EGFR^L858R/T790M (epidermal growth factor receptor) mutant inhibitor, was developed by the group in an effort to treat acquired resistance in nonsmall cell lung cancer (NSCLC), but its cellular off-targets were not identified. An activity-based probe, HJ-1, was created followed by chemical proteomics and bioimaging studies. A total of 13 protein hits, including EGFR and NT5DC1, were identified by pull-down/LC-MS. Subsequent validation experiments indicated the involvement of a major off-target, NT5DC1, in the biological function of HP-1.

Kinases, a major class of oncogenic signaling proteins, serve as drug targets for various cancers.¹ Many drugs that have been developed inhibit kinases by reacting with active cysteine residues.² Patients with nonsmall cell lung cancers (NSCLCs) harboring epidermal growth factor receptor (EGFR) mutations are highly responsive to first- and second-generation EGFR kinase inhibitors. Thus a third-generation inhibitor of the EGFR, HP-1, was designed by the group to inhibit the drug-resistant EGFR T790M mutant with improved pharmacokinetic properties (Figure 1A).^3,4 It was developed in an effort to treat the acquired resistance of NSCLC. However, the cellular targets were still unknown, resulting in a poor understanding of the drug action and potential toxicities. Activity-based protein profiling (ABPP) is a powerful method for target identification on a proteome-wide scale.⁵ During the past decade, ABPP has been successfully used in target identification with many natural products and synthetic compounds and has provided information that helps us to understand their mechanism of action.⁶ To facilitate the application of ABPP in covalent inhibitors, we recently developed a suite of minimalist linkers and novel electrophiles that are suitable for irreversible inhibitors and have been used to develop high-quality activity-based probes that are useful in target identification.⁷⁻⁹

Figure 1.

(A) Structures of the parent HP-1 and the probe HJ-1. (B) In vitro kinase inhibitory activities (IC₅₀) of compounds against EGFR^T790M and cell antiproliferation activities (IC₅₀) of compounds against H1975 harboring EGFR^T790M. (C) Concentration-dependent and (D) time-dependent labeling of HJ-1 in live H1975 cells. FL, in-gel fluorescence scanning; CBB, Coomassie gel. (E) Competitive labeling of HJ-1 in the presence of HP-1. (F) Live-cell imaging of H1975 cells treated with HJ-1 in the presence or absence of HP-1. H1975 cells were treated with 5 μM HJ-1 for 1 h. The excitation wavelength of the probe channel was 550 nm. Scale bar = 10 μm. (G) Pull-down labeling with HJ-1 in H1975 cells. PD, pull-down experiments. (H) MS-spectrometry-based profiling of HJ-1 binding proteins (5 μM probe concentration). (I) MS spectrometry-based profiling of HJ-1 binding proteins in the presence of 5 × HP-1 (5 μM probe concentration).

To characterize the protein targets of HP-1 in live cells, we created an activity-based probe HJ-1 based on the parent compound HP-1 (Figure 1A). Previous studies have shown that modification of a water-soluble structural motif of kinase inhibitors does not compromise their biological activity.^4,10 Accordingly, an alkyne handle was incorporated into the piperazine of HP-1 and used for subsequent protein enrichment and target identification (Figure 1A). The synthetic method that was used is similar to that of previously reported procedures (Scheme S1).^3,4

We first evaluated the biological activity of the HJ-1 probe using the parent compound HP-1 as a control. This was carried out by standard in vitro kinase inhibition assays and cell-based CCK8 antiproliferation assays using H1975 cells. As shown in Figure 1B, HJ-1 with IC₅₀ = 8.04 nM displays inhibition comparable to that of HP-1 (IC₅₀ = 5.27 nM) against EGFR T790M mutant kinase. The antiproliferative activity against H1975 NSCLC cells harboring the EGFR T790M mutant was also measured. Two EGFR inhibitors, afatinib and AZD9291, were tested as positive controls. HJ-1 and HP-1 also exhibited similar antiproliferative activities with IC₅₀ values of 0.82 and 0.13 μM, respectively (Figure 1B). These results proved that the probe largely retains the bioactivity of the parent inhibitor.

Proteome profiling and bioimaging experiments were carried out to evaluate the performance of HJ-1 in live cells. After incubation of the probe with H1975 cells at various concentrations for different time periods, the cells were lysed and click-reacted with TAMRA-azide. The labeled proteomes were then separated by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE) and analyzed by fluorescence imaging. As shown in Figure 1C,D, two major fluorescent bands (*) at ∼180 and ∼50 kDa were detected, highlighting the excellent selectivity of the probe. The ∼180 kDa band presumably corresponds to the known target, EGFR T790M. Time- and concentration-dependent experiments demonstrated that a probe concentration as low as 0.1 μM or 10 min of incubation time is sufficient to produce detectable labeling bands, signifying excellent labeling efficiency (Figure 1C,D). The fluorescent bands were abolished upon treatment with the parent compound HP-1 (Figure 1E, Figure S6), which proves that the probe shares the same set of protein targets as the parent compound under in situ conditions. Imaging experiments were carried out to track the subcellular locations of HJ-1 in live cells. H1975 cells were incubated with HJ-1 in the presence or absence of parent compound HP-1 for 1 h (5 μM probe concentration), and the cells were then fixed and permeabilized. After a click reaction with TAMRA-azide, the cells were then imaged by microscopy. Strong fluorescence signals were observed in the cell cytosol, and relatively weak fluorescence was detected in the nucleolus from the probe-treated samples (Figure 1F) but not from the control samples treated with dimethylsulfoxide (DMSO) and HP-1, implying that the main cellular targets are in the cell cytosol. Taken together, these competitive labeling and cellular imaging experiments confirmed that the probe HJ-1 is suitable for target identification of the parent inhibitor.

Subsequently, large-scale pull-down/LC-MS experiments were then carried out to identify the protein targets. On the basis of the results of protein labeling and imaging, a 5 μM probe can produce a notable fluorescence signal, and thus this probe concentration was used in the proteomics studies. Live H1975 cells were incubated with HJ-1 in the presence or absence of the parent compound HP-1 for 1 h followed by cell lysis and conjugation to a biotin-TAMRA-azide tag under CuAAC conditions. The biotin-labeled proteins were then enriched by streptavidin beads. The samples were first separated by SDS-PAGE and analyzed by in-gel fluorescence scanning. As shown in Figure 1G, the major labeling proteins were successfully enriched by this pull-down procedure, indicating that the samples can be analyzed by LC-MS. After trypsin digestion, the resulting peptides were analyzed by LC-MS/MS. The protein hits were refined by corresponding volcano plots as (log2−) of the enrichment ratio (probe/DMSO) and competitive ratio (probe/probe + inhibitor) against statistical significance (−log10 p value) and the LFQ (label-free quantification) intensity.¹¹ Only protein hits with (log2−) > 3, (−log10 p value) > 2, and LFQ intensity (probe) > 1 × 10⁷ in the enrichment experiment were further considered. On the basis of these criteria, 13 protein hits for HJ-1 were obtained including NT5DC1, MAVS, RPS19, MAP4, MCM2, HADHA, and the known target EGFR (Figure 1H,I, Tables S1 and S2). After analysis of the protein hits, we found that the 51 kDa protein, a 5′-nucleotidase domain-containing protein 1 (NT5DC1), corresponds to the major band observed in the labeling experiments (Figure 1C–E). Both EGFR and NT5DC1 were also detected in the competitive profiling experiment, indicating the high reliability of these data (Figure 1I, Table S2). The NT5DC1 gene has been reported to be associated with chronic obstructive pulmonary disease (COPD)¹² and neuropsychiatric disorders,¹³⁻¹⁵ such as attention-deficit/hyperactivity and bipolar disorders. Most of the protein hits, such as NT5DC1 and RPS19, are located in the cell cytosol. This agrees well with the imaging results (Figure 1F), indicating the high reliability of these protein hits. It is worth noting that these protein hits possess important functions, such as RNA synthesis/degradation and protein synthesis (Table 1). It is possible that the anticancer effects produced by this compound involve a combination of these protein hits.

Table 1. Protein Hits Identified by Pull-Down/LC-MS and Their Functions^a.

gene name	mol. weight (kDa)	subcellular locations	protein function
NT5DC1	51.844	cytosol
EGFR	129.17	plasma membrane, nucleus	cell proliferation
MAVS	56.527	mitochondrion	necessary in the virus-triggered β-interferon signaling pathways
RPS19	7.8309	nucleus, cytosol	protein synthesis
RRP12	143.7	nucleus
MAP4	121	plasma membrane, cytosol	microtubule assembly and stabilization
FAM120A	121.89	cytosol
MCM2	93.998	nucleus	eukaryotic genome replication
LARS1	134.46	nucleus, cytosol	tRNA synthesis
CNOT2	19.981	nucleus, cytosol	mRNA synthesis and degradation
HADHA	82.999	mitochondrion	mitochondrial β-oxidation of long chain fatty acids
AHNAK	629.09	plasma membrane, nucleus, cytosol	blood–brain barrier formation, cardiac calcium channel regulation
TXN	11.737	nucleus, cytosol	redox reactions

^aBold protein hits correspond to the major labeling bands in proteome labeling profiles (* marked bands).

NT5DC1 is the main cellular off-target and has been reported to interact with different covalent inhibitors.^16,17 Subsequent validation experiments were therefore carried out with this hit protein. Pull-down/Western blotting (WB) experiments with the corresponding antibodies proved that the major labeling bands at ∼180 and ∼50 kDa were EGFR and NT5DC1, respectively (Figure 2A). A cellular thermal shift-binding assay (CETSA) proved that HP-1 and HJ-1 both interact with NT5DC1, inducing a positive shift in the melting temperature (Figure 2B, Figure S7), which indicated protein stabilization upon binding with the compounds. Labeling of recombinant NT5DC1 proved that at a concentration as low as 50 pmol, the protein can be successfully detected by the 0.1 μM probe, which again demonstrated excellent labeling efficiency (Figure 2C). Consistent with the results of labeling of live cells (Figure 1E), the labeling bands can be completely competed away by treatment with the parent inhibitor, indicating that the bands had been produced by real probe binding, not by nonspecific labeling (Figure 2D). The labeled recombinant protein was digested with trypsin, and the resulting mixture was analyzed by LC-MS/MS to determine the binding pattern of the probe with NT5DC1. C119 was identified with a molecular weight increase corresponding to the addition of HJ-1-labeled adducts (Figure 2E). To further confirm this binding model, we carried out labeling of the mutant protein using the procedures described above. As shown in Figure 2F, mutation of C119 lead to a failure of labeling NT5DC1 (∼50 kDa band), which demonstrated that the binding occurred at C119. Docking experiments (Figure 2G) proved that both HJ-1 and HP-1 could accommodate C119 in NT5DC1.

Figure 2.

Target validation in H1975 cells. (A) Target validation of NT5DC1 and EGFR by pull-down/WB. (B) Thermal shift binding assay of HP-1 and HJ-1 with H1975 cells. (C) Concentration-dependent and (D) competition labeling of recombinant NT5DC1 with HJ-1 in the presence of HP-1. (E) MS spectra of peptide labeled by HJ-1 and modification of the Cys119 residue of NT5DC1. (F) Labeling of the NT5DC1 wild type (WT) and C119A mutant with HJ-1. (G) Docking experiments to predict the binding mode of HJ-1 and HP-1 with NT5DC1.

Finally, functional validation experiments were carried out to evaluate the relationship between this target and HP-1. As shown in Figure 3A, knockdown of NT5DC1 lead to a weak labeling profile toward the ∼50 kDa band. Knockdown of NT5DC1 resulted in inhibition of the proliferation of H1975 cells and could sensitize the cancer cells to different concentrations of HP-1 or HJ-1 (Figure 3B,C). Knockdown of NT5DC1 in H1975 cells had no effect on the cell cycle but induced cell apoptosis (Figure 3D,E). These results indicated that NT5DC1 is essential for cell proliferation and is potentially involved in the drug activity. To evaluate if the probe could selectively label NT5DC1 in different cancer cells, we carried out labeling profiles of other cancer cells such as H1299, HeLa, HuH-7, MCF-7, MDA-MB-453, and MV4-11. As shown in Figure 3E and Figure S4, the probes could predominantly label NT5DC1 in different cancer cells. Interestingly, the labeling fluorescence intensity is correlated with the expression level of the NT5DC1 protein in different cancer cells (Figure 3F,G), indicating that the probe can be an efficient reagent to detect the expression level of NT5DC1 in various cancer cells.

Figure 3.

(A) Labeling performance of the probe in the presence or absence of knockdown reagent. NC, scrambled siRNA as a negative control. (B) Knock-down of NT5DC1 by using siRNA NT5DC1. NT5DC1 expression was evaluated by WB (upper gel). Cell viability was tested in siRNA-NC- and siRNA-treated NT5DC1 H1975 cells, with cells treated with scrambled siRNA as a negative control (NC). (C) Cell viability of compounds in H1975 cells with or without knock-down reagents. Cells treated with DMSO were used as negative controls. (D) Cell cycle of H1975 cells treated with siRNA NT5DC1 for 48 h. (E) Apoptosis of H1975 cells treated with siRNA NT5DC1 for 72 h. Cells treated with siRNA NC were used as negative controls. (F) In situ labeling of different cancer cells with HJ-1 (1 μM). (G) NT5DC1 expression level in different cell lines evaluated by WB.

In conclusion, using a chemical proteomics approach, we have identified NT5DC1 as a major off-target of a potent EGFR^L858R/T790M mutant inhibitor in H1975 cells. Further functional validation experiments indicated that this protein could be involved in the drug action and cell apoptosis. The protein hits identified in this study would be valuable clues for understanding the activity and safety of the drug candidate. Furthermore, HJ-1 appears to be the first probe that is reported to selectively label NT5DC1 in various cancer cells and could be a useful reagent for the highly efficient detection of NT5DC1 expression.

Glossary

Abbreviations

ABPP

activity-based protein profiling

EGFR

epidermal growth factor receptor

IC₅₀

the half maximal (50%) inhibitory concentration of a substance

TAMRA

carboxytetramethylrhodamine

FL

in-gel fluorescence scanning

CBB

Coomassie gel

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00651.

• Synthetic procedures and compound characterization, biological methods, and ¹H NMR and ¹³C NMR spectra of probes (PDF)

The authors declare no competing financial interest.