Purpose:EML4–ALK fusions define a subset of lung cancers that can be effectively treated with anaplastic lymphoma kinase (ALK) inhibitors. Unfortunately, the duration of response is heterogeneous and acquired resistance limits their ultimate efficacy. Thus, a better understanding of resistance mechanisms will help to enhance tumor control in EML4–ALK-positive tumors.
Experimental Design: By applying orthogonal functional mutagenesis screening approaches, we screened for mutations inducing resistance to the aminopyridine PF02341066 (crizotinib) and/or the diaminopyrimidine TAE684.
Results: Here, we show that the resistance mutation, L1196M, as well as other crizotinib resistance mutations (F1174L and G1269S), are highly sensitive to the structurally unrelated ALK inhibitor TAE684. In addition, we identified two novel EML4–ALK resistance mutations (L1198P and D1203N), which unlike previously reported mutations, induced resistance to both ALK inhibitors. An independent resistance screen in ALK-mutant neuroblastoma cells yielded the same L1198P resistance mutation but defined two additional mutations conferring resistance to TAE684 but not to PF02341066.
Conclusions: Our results show that different ALK resistance mutations as well as different ALK inhibitors impact the therapeutic efficacy in the setting of EML4–ALK fusions and ALK mutations. Clin Cancer Res; 17(23); 7394–401. ©2011 AACR.
Our results show that previously described ALK resistance mutations as well as newly characterized mutations show a differential pattern of sensitivity to the ALK inhibitors PF02341066 (currently in the clinic) and TAE684. Some mutations induce resistance to both kinase inhibitors, others only to one of the two inhibitors used in this study. Thus, the individual resistance mutation should influence the use of the appropriate ALK inhibitor to enhance tumor control and patient benefit in lung cancer, neuroblastoma, and potentially other cancers with ALK aberrations.
Lung cancers bearing epidermal growth factor receptor (EGFR) mutations or EML4–ALK fusions can be effectively treated with EGFR and anaplastic lymphoma kinase (ALK) inhibitors, respectively (1, 2). Different oncogenic EML4–ALK fusion variants have been described, which induce ALK dependency (3–5). All of these variants contain the complete kinase domain of ALK, which is predominantly fused to the N-terminus of echinoderm microtubule–associated protein-like 4 (EML4), leading to constitutive activation of the kinase (6, 7). In addition, activating ALK mutations have been described in neuroblastoma. These mutations are assumed to be driver mutations and may be amenable to therapeutic ALK inhibition (8–12). Unfortunately, all patients treated with targeted therapeutics will eventually relapse, in most cases due to the emergence of acquired genetic alterations conferring resistance (13, 14). The knowledge about the actual resistance mechanism is, however, a prerequisite for development of secondary treatment strategies that can overcome resistance (15, 16). The aminopyridine ALK inhibitor, PF02341066 (crizotinib), is currently undergoing evaluation in phase III clinical trials (1). Until now, only 2 resistance mutations within the ALK domain have been identified in a single EML4–ALK-positive lung cancer patient with acquired crizotinib resistance as well as in a cell culture model, and another mutation was found in a patient with an inflammatory myofibroblastic tumor harboring a RANBP2–ALK translocation (13, 17, 18). Structural modeling, however, suggests that diaminopyrimidine scaffolds, such as TAE684 (19), should still be able to bind to the mutated kinase. We therefore tested the known PF02341066 resistance mutations for sensitivity to TAE684 and conducted orthogonal mutagenesis screens to identify novel PF02341066 and/or TAE684 resistance mutations that show differential sensitivity patterns to these ALK inhibitors. This information may provide a mechanistic rationale for the development of second-generation ALK inhibitors.
Materials and Methods
cDNA and cell lines
pMA-3FLAG-EML4-ALK v1 plasmid was cloned into the retroviral pBabe puro backbone. pDONR-EML4-ALK v3a was cloned into the pBabe Gateway puro backbone. Full-length human wild-type ALK and ALKF1174L cDNAs were kindly provided by Rogier Versteeg (Academic Medical Centre-AMC, Amsterdam, the Netherlands). Site-directed mutagenesis was carried out as described previously (20). Ba/F3 cell lines were established as described previously (20). SH-SY5Y neuroblastoma cells were cultured in Dulbecco's Modified Eagle's Medium supplemented with 8% serum, cells were validated by sequencing of ALK (data not shown). H3122 were cultured as described previously (21). Human cell lines have been tested by single-nucleotide polymorphism-based genotyping.
PF02341066 (racemic mixture) was purchased from Selleck Chemicals, and TAE684 was purchased from Axon Medchem.
Immunoblotting was carried out by standard procedures (22). The following antibodies were used: phospho (p)-ERK, ERK1, and ERK2 from Santa Cruz Biotechnology. Antibodies against phospho (p)-ALK Tyr1604, p-ALK Tyr1278/1282/1283 (23), phospho (p)-AKT Ser473, and total AKT were from Cell Signaling; actin from MP Biomedical, and total ALK from Cell Signaling and Bethyl Laboratories.
Saturation mutagenesis (24, 25) was carried out by propagation of EML4–ALK cDNA containing plasmids in the mismatch repair–deficient E. coli strain XL1-Red (Stratagene). Bacteria were grown for 48 or 72 hours. The resulting plasmids were expanded in XL1-Blue bacteria and packaged in retroviruses followed by infection of Ba/F3 cells and subsequent selection of stable cell lines in the absence of interleukin 3 and presence of compound (750/1,000/1,500 nmol/L of PF02341066) to only allow proliferation of resistant clones. Mutant inserts were recovered from drug-resistant polyclonal lines, pooled, and sequenced on a GS Flex instrument. The raw data were aligned and visualized by IGV.
In an orthogonal chemical mutagenesis screen, EML4–ALK-expressing Ba/F3 cells were treated with N-ethyl-N-nitrosourea (ENU; 100 μg/mL) overnight and subsequently cultured in the presence of ALK inhibitor (750/1,000/1,500 nmol/L of PF02341066) to select for resistant clones. Inserts were PCR-amplified and sequenced as described earlier. For differentiation of the clonal origin of the 2 mutations found, dideoxy sequencing was carried out for each polyclonal-resistant clone separately.
For the PCR-based random mutagenesis screen, a NotI fragment containing the ALKF1174L open reading frame was cloned into pMX-IRES-blasticidin. To generate an ALKF1174L cDNA library with random mutations restricted to the kinase domain, the ALKF1174L kinase domain was amplified by an error-prone PCR using the following primer pair: ALKKin-fwd GGCATCATGATTGTGTACCG and ALKKin-rev TCTTTTTGGTGGGTTTCTCTG. The PCR products were digested with the restriction enzymes BlpI and BsiWI and cloned into the pMX–ALK backbone. To achieve a sufficient representation of random mutations of ALKF1174L, we pooled approximately 4 × 105 bacterial clones and isolated the pooled plasmid DNA. SH-SY5Y cells stably infected with the ALKF1174L mutant library were seeded at low density and incubated with TAE684 (100 nmol/L). Resistant colonies were pooled, and genomic DNA was isolated with DNAzol (Invitrogen). The ALK domain was recovered from the genomic DNA of the pooled TAE684-resistant colonies by PCR using the primers indicated earlier. The PCR products were cloned into the pMX–ALK backbone and sequenced. To identify recurring mutations, we sequenced 82 bacterial clones covering the entire region that was targeted for random mutagenesis.
Colony formation assay
For colony formation assays, SH-SY5Y cells were seeded at low density and treated with the various concentrations of the ALK inhibitors or left untreated. At reaching confluence, the cells were fixed with formaldehyde, stained with crystal violet, and photographed.
Ba/F3 viability assays were conducted as described previously (26) measuring cellular ATP content (CellTiter-Glo; Promega) after 96 hours of treatment.
SH-SY5Y cells were seeded at high density (4 × 103 per 384-well plate), and ALK inhibitors were added 24 hours after seeding. After 5 days, cell viability was determined with CellTiter-Blue (Promega) according to the manufacturer's recommendations. CellTiter-Blue signals were detected with the EnVision Multilabel Reader (PerkinElmer).
EML4–ALK-expressing cells were treated for ALK inhibitors for the indicated duration before staining with trypan blue and counting of negative (viable) and positive (dead) cells.
The modeling of resistance mutations into ALK [PDB-codes: 2XP2 (crizotinib) and 2XB7 (TAE684)] was carried out with PyMol (Schroedinger).
The L1196M mutation has recently been reported to confer acquired resistance in an EML4–ALK-positive patient treated with the ALK inhibitor crizotinib (13). This mutation was also found in an experimental model of acquired resistance involving the EML4–ALK mutant lung cancer cell line H3122 (18). We first expressed the predominant crizotinib resistance mutations EML4–ALKL1196M and EML4–ALKF1174L (13, 17, 18) in the EML4–ALK-positive non–small cell lung carcinoma (NSCLC) cell line H3122 (Supplementary Fig. S1A) and determined the sensitivity of the resulting mutants to the structurally unrelated ALK inhibitor TAE684. As expected, both mutations induced resistance to PF02341066, whereas EML4–ALKF1174L was sensitive to higher PF02341066 concentrations. However, confirming previous reports (18), the L1196M mutant retained high sensitivity to TAE684. Furthermore, H3122EML4–ALKF1174L cells were also found to be exceptionally sensitive to this structurally different compound (Supplementary Fig. S1B). To validate these findings in an independent model, we expressed these mutations in Ba/F3 cells and treated them with both ALK inhibitors. Furthermore, we introduced a mutation at position G1269 (G1269S) adjacent to the DFG (asp-phe-gly) motif in EML4–ALK, which we predicted to induce resistance to PF02341066 but not to TAE684 due to steric hindrance (Supplementary Fig. S2). Of the resulting Ba/F3 mutants, EML4–ALKF1174L (17) and EML4–ALKG1269S showed an increase in ALK phosphorylation, whereas the phosphorylation levels of EML4–ALKL1196M (13) were comparable with those of the wild-type kinase (Fig. 1A). Confirming previous reports (13, 17, 18), all mutations induced resistance to PF02341066, with F1174L leading to a slight increase in resistance and L1196M and G1269S leading to a high level of resistance (Fig. 1B). Confirming and extending our results in the H3122 cells, the 3 mutants were highly sensitive to TAE684 with the 2 activating mutants EML4–ALKF1174L and EML4–ALKG1269S again being particularly sensitive to treatment with TAE684 compared with the original EML4–ALK variant (Fig. 1C). Accordingly, phosphorylation of mutated EML4–ALK remained unchanged after treatment with up to 2.5 μmol/L of PF02341066 but disappeared under treatment with TAE684 at concentrations as low as 30 nmol/L in all EML4–ALK variants (Fig. 1D).
Structural modeling of G1269S suggests a steric clash with PF02341066 but to a much lesser degree with TAE684 as the underlying mechanism of resistance (Supplementary Fig. S2). Furthermore, the activating nature of this mutation should–by itself–reduce binding of PF02341066 because this compound binds the inactive conformation of the kinase. We speculate that resistance to TAE684 might be induced by a larger amino acid than Ser at this position (Supplementary Fig. S2). A similar mechanism of steric hindrance might be induced by L1196M, preventing the binding of PF02341066 (13) but allowing the binding of TAE684 (Fig. 2A and B). The differential resistance for EML4–ALKF1174L has been discussed previously (17). Thus, resistance mediated by L1196M, F1174L, and G1269S may be overcome by ALK inhibitors that circumvent interference with the side chains of certain amino acids of the ALK and are also able to bind the active kinase confirmation (Fig. 2A and B; ref. 27).
To discover additional mutations conferring ALK inhibitor resistance, we conducted a saturation mutagenesis screen (24, 25) to express randomly mutated versions of EML4–ALK in Ba/F3 cells and selected for cells that were resistant to ALK inhibition (Supplementary Fig. S3A). This approach yielded 2 novel resistance mutations (L1198P, 49% and D1203N, 12%) at high prevalence (Fig.2A and B, Supplementary Fig. S4). The L1198P mutation induced a similar degree of resistance to the ALK inhibitors PF02341066 [crizotinib; concentration needed to reduce the growth of treated cells to half that of untreated cells (GI50), 3,396 nmol/L] and TAE684 (GI50, 624 nmol/L; Fig. 3A and B). Counting of viable cells following treatment confirmed this observation (Fig. 3C). Resistance was reflected by an increase in basal kinase activity (Supplementary Fig. S5) as well as sustained phosphorylation of ALK at concentrations up to 2.5 μmol/L of PF02341066 and 300 nmol/L of TAE684 in immunoblotting assays (Fig. 3D). The second mutation, D1203N, induced resistance to PF02341066 (GI50, 3,357 nmol/L) and to TAE684 (GI50, 604 nmol/L; Fig. 3A and B). This mutation led to a shift toward higher compound concentrations needed for inhibitor-mediated dephosphorylation as well, although this effect was less pronounced than in the L1198P mutation and no basal kinase activation was observed (Supplementary Fig. S6A and S6B).
As an orthogonal approach, we mutagenized Ba/F3 cells expressing EML4–ALK v3a with the chemical mutagen ENU (28) and selected for resistance mutations by subsequent culture in PF02341066 (Supplementary Fig. S3B). Most of the resulting resistant polyclonal cell lines expressed the mutation L1198P and were highly resistant to ALK inhibition (Supplementary Fig. S7).
We also carried out a PCR-based random mutagenesis screen on the ALKF1174L kinase domain, followed by ectopic expression of mutated ALK in the ALKF1174L-mutant neuroblastoma cell line SH-SY5Y (Supplementary Fig. S8A). The F1174L mutation increases the basal activation of ALK (Fig. 1A), thereby reducing the binding of PF02341066, which tends to bind to the inactive conformation of ALK (29). Nevertheless, SH-SY5Y cells are still responsive to higher concentrations of PF02341066, consistent with our previous findings in EML4–ALKF1174L (Fig. 1B). In contrast, SH-SY5Y cells are highly sensitive to TAE684 (30). To screen for resistance, SH-SY5Y cells stably expressing mutagenized ALK were treated with TAE684 and surviving clones were analyzed by PCR-based cloning and sequencing (Supplementary Fig. S8B). We identified approximately 2 mutations per clone with G1123S, G1123D, and L1198P as single mutations and Y1278H mutations only in the context of G1123S or G1123D (data not shown). We next expressed ALKF1174L, ALKF1174L/L1198P, ALKF1174L/G1123S, and ALKF1174L/G1123D in the original SH-SY5Y cell line. All 3 mutations induced a high level of resistance to TAE684 in colony formation assays (Fig. 4A, left), and short-term viability assays revealed an intermediate level of resistance to TAE684 in those cells expressing G1123S or G1123D, whereas the expression of L1198P induced strong resistance (Fig. 4B, left). Immunoblotting showed that all 3 mutations prevented inhibition of extracellular signal–regulated kinase (ERK) and AKT by TAE684 (Fig. 5A). Interestingly, only the L1198P mutation, but not G1123S/D, induced crossresistance to PF02341066, as shown by colony formation assays (Fig. 4A, right), viability assays (Fig. 4B, right), and immunoblotting (Fig. 5B).
We next applied structural modeling to identify the mechanism of resistance induced by the L1198P mutation that was the most prominent resistance mutation across our saturation mutagenesis screens. Most in-cis resistance mechanisms show direct interactions with the inhibitor or shift the kinase equilibrium toward a more active conformation (14). The L1198P mutation is localized in the hinge region of the kinase domain at a position where both inhibitors form key hydrogen bonds to the backbone of the protein to mimic ATP binding (Fig. 2A and B). The L1198P mutation is located next to E1197, whose side chain forms hydrogen bonds to K1267 and R1181 (Supplementary Fig. S9). Such a polar interaction was recently described as a “molecular brake” to keep the kinase domain in an inactive state and proposed to be a common regulatory mechanism in receptor tyrosine kinases (31). Although P1198 is unlikely to directly participate in this inhibitory network, we speculate that it perturbs its function by restricting backbone conformations of neighboring amino acids (32), thereby shifting the kinase toward an active conformation. This notion is compatible with our observation of increased basal EML4–ALK phosphorylation in this mutant (Supplementary Fig. S5) and may explain resistance to PF02341066, an inhibitor that is known to bind to and stabilize inactive kinase conformations (29). ALK crystal structures suggest the methoxy group of TAE684 to bind in a small cavity of the hinge region (19, 23). Our modeling studies show that the L1198P mutation decreases the space occupied by this amino acid and results in suboptimal TAE684 binding. In general, L1198P seems to prevent ALK inhibition by all ATP-competitive analogues with different scaffolds.
In contrast, the D1203N mutation is in close proximity to both compounds at the lip of the ATP pocket (Fig. 2A and B), but the charged side chain points away from the inhibitor cores toward the solvent. Because of the fact that the mutation did not lead to increased basal kinase phosphorylation, the underlying resistance mechanism remains unclear at this point.
The 2 mutations that were only found in the neuroblastoma resistance screen (G1123S/D) are located in the glycine-rich loop, which is known to be crucial for ATP and ligand binding (33) and are the first mutations described that induce resistance to TAE684, but not to PF02341066. Although PF02341066 does not directly interact with the glycine-rich loop, the sulfonated aniline moiety of TAE684 makes hydrophobic interactions to the Gly1123-His1124 segment (23). Our modeling studies indicate that mutations in this part of the protein are (i) likely to sterically impede ATP binding and/or (ii) alter the dynamics of the glycine-rich loop and thus perturb interactions with inhibitors that require a particular conformation of the loop for binding (Fig. 2A and B).
Here, we show that the recently described PF02341066 resistance mutants, L1196M and F1174L, retain exquisite sensitivity to a structurally different ALK inhibitor, TAE684. Thus, compounds developed on the basis of these structural considerations may have the potential to overcome resistance to PF02341066 when caused by these mutations.
By using complementary mutagenesis approaches, we have furthermore identified novel resistance mutations in EML4–ALK and ALK that induce a high level of resistance to both structurally unrelated ALK inhibitors and one of the 2 inhibitors only. We have also provided mechanistic evidence for these observations, which are based on structural models of compound binding to the kinase and on biochemical analyses of kinase activity. Therefore, we predict that some patients with acquired in-cis crizotinib resistance mutations will respond to diaminopyrimidine-based ALK inhibitors. Others, depending on the respective resistance mutation, will not. Thus, further development is required to develop compounds that are capable of overcoming resistance mediated by these novel mutations. Furthermore, the individual resistance mutation may dictate the use of the appropriate ALK inhibitor.
In summary, we have shown that structurally diverse ALK inhibitors can elicit strikingly different cytotoxic potency in genotypically defined EML4–ALK and ALK mutants. This observation highlights that the development and application of ALK inhibitors should take into account individual resistance mutations to enhance tumor control and patient benefit in lung cancer, neuroblastoma, and potentially other cancers with ALK aberrations.
Disclosure of Potential conflicts of Interest
R.K. Thomas received consulting and lecture fees from Sanofi-Aventis, Merck, Roche, Boehringer Ingelheim, AstraZeneca, Atlas-Biolabs and received research support from AstraZeneca, EOS, Merck. W. Pao has consulted for MolecularMD and received research funding from XCovery. D. Rauh received research grants from Merck Serono, Merk Sharp & Dohme, and Beyer Schering Pharma. No potential conflicts of interest were disclosed by the other authors.
The authors thank Caroline Emery for technical advice.
This work was supported by the German Ministry of Science and Education (BMBF) as part of the NGFNplus program (grant 01GS08100), by the Max Planck Society (M.I.F.A.NEUR8061), by the Fritz-Thyssen-Stiftung Grant 10.08.2.175, the Deutsche Forschungsgemeinschaft (DFG) through SFB, and by the Ministry for Innovation, Science, Research and Technology of the State of Nordrhein-Westfalen (MIWT, 4000-12/09), all to R.K. Thomas. W. Pao received support from Vanderbilt's Specialized Program of Research Excellence in Lung Cancer grant (CA90949), by the NCI Lung SPORE grant (CA70907) and the Vanderbilt-Ingram Cancer Center Core grant (P30-CA68485). C.M. Lovly was supported by the ASCO Young Investigator Award (ASCO YIA) and Uniting Against Lung Cancer (UALC). M. Hölzel was supported by the Dutch Cancer Society (KWF).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.