ALK in cancer: from function to therapeutic targeting

Abstract

Anaplastic lymphoma kinase (ALK) is a receptor tyrosine kinase (RTK) that acts as an oncogenic driver in solid and haematological malignancies in both children and adults. Although ALK-expressing (ALK⁺) tumours show strong initial responses to the series of ALK inhibitors currently available, many patients will develop resistance. In this Review, we discuss recent advances in ALK oncogenic signalling, together with existing and promising new modalities to treat ALK-driven tumours, including currently approved ALK-directed therapies, namely tyrosine kinase inhibitors, and novel approaches such as ALK-specific immune therapies. Although ALK inhibitors have changed the management and clinical history of ALK⁺ tumours, they are still insufficient to cure most of the patients. Therefore, more effort is needed to further improve outcomes and prevent the tumour resistance, recurrence and metastatic spread that many patients with ALK⁺ tumours experience. Here, we outline how a multipronged approach directed against ALK and other essential pathways that sustain the persistence of ALK⁺ tumours, together with potent or specific immunotherapies, could achieve this goal. We envision that the lessons learned from treating ALK⁺ tumours in the clinic could ultimately accelerate the implementation of innovative combination therapies to treat tumours driven by other tyrosine kinases or oncogenes with similar properties.

Introduction

Anaplastic lymphoma kinase (ALK) is a receptor tyrosine kinase (RTK) that belongs to the insulin receptor superfamily. ALK is expressed during embryogenesis and its signalling regulates the development of neurons in the central nervous system (CNS), in the spinal cord and in enteric neurons. After development, ALK remains only weakly expressed in these tissues^1-5 (Box 1). However, ALK expression is reactivated in tumours by various mechanisms. The first genetic alteration of the ALK gene was discovered in 1994 as a chromosomal translocation associated with anaplastic large-cell lymphoma (ALCL)⁶. Subsequently ALK gene alterations were identified in non-small-cell lung cancer (NSCLC)^7,8 and many other tumours¹. Overall, ALK gene alterations, including chromosomal rearrangements, single point mutations and amplifications, cause aberrant tyrosine kinase activation that leads to downstream oncogenic signalling mediated by key pathways, such as MAPK, PI3K and Janus kinase (JAK)–signal transducer and activator of transcription (STAT)^1,9. ALK genetic alterations or aberrant ALK expression define globally a subset of tumours that depend on ALK oncogenic activity, termed ALK⁺ tumours¹⁰. The dependency on ALK signalling in ALK⁺ tumours creates a window of opportunity for therapeutic intervention, and as such, clinical responses to ALK tyrosine kinase inhibitors (TKIs) are, at least initially, quite remarkable¹¹. Patients with ALK⁺ tumours treated with ALK TKIs typically achieve a remission state or durable disease control^12-15. Yet, tumour cells are not fully eradicated and drug-tolerant persister cells that survive ALK inhibition seed relapse or metastatic dissemination^16,17. Thus, ALK⁺ tumours represent a case study for precision oncology and a paradigm also for other cancers. Understanding the mechanisms of drug tolerance and resistance to ALK-directed targeted therapies could shed light on therapeutic innovations that could have broader implications beyond ALK⁺ cancers and be applied to other tumours with similar genetic alterations.

Box 1 ∣. The role of ALK in normal tissues.

The anaplastic lymphoma kinase (ALK) gene is located on the short arm of chromosome 2 (2p23) and encodes a receptor tyrosine kinase (RTK) within the insulin receptor superfamily. It is a single-chain receptor that is highly conserved across species^1,259. The physiological role of ALK is mainly restricted to the embryo where it plays a part in the development of the nervous system, spinal cord and gut. In adult life, it is minimally expressed in the same tissues^1,4,5. In Drosophila melanogaster, ALK is crucial for embryonal development and survival, whereas in Caenorhabditis elegans and Danio rerio it is not essential for viability but is crucial for nervous system development^260-264. ALK knockout mice have shown a mild defect in neurogenesis and testosterone development, demonstrating that ALK is not crucial for embryonal development^265,266.

In mice and humans, the ALK ligands ALKAL1 and ALKAL2 are potent activators of ALK physiological signalling, although both ligands also bind to the closely related receptor, leukocyte tyrosine kinase (LTK)^267-269. Recent works have described the receptor–ligand complex for ALKAL1 or ALKAL2 and ALK, but a consensus on the exact structure of this complex is still lacking^{2,3,236,259,268}. Despite the identification of ALK ligands, the physiological role of mammalian ALK still needs to be fully elucidated. Interestingly, more recent works have suggested a role in metabolism and body weight regulation and in pain signalling^270,271. In 2020, a large genome-wide association study (GWAS) conducted on metabolically healthy, thin individuals in an Estonian cohort revealed that genetic variants in the ALK gene are linked to thinness²⁷⁰. Notably, ALK knockout mice were resistant to diet-induced obesity through the hypothalamic control of lipolysis mediated by ALK, indicating that ALK acts as a regulator of energy expenditure and weight gain²⁷⁰. Other studies have shown that ALKAL gene expression was enhanced during inflammation or injury in sensory neurons for pain (peptidergic nociceptor) in mice and humans, and further discovered a role for the ALKAL2–ALK axis in the control of nociceptor-induced sensitization²⁷¹. Interestingly, treatment with one of the ALK tyrosine kinase inhibitors crizotinib or lorlatinib reverted the hyperalgesia induced by inflammation or nerve injury.

In this Review, we discuss recent advances in ALK tumour biology and drug development that have improved outcome in patients with ALK⁺ cancers. We discuss the oncogenic signalling of ALK in different tumours, the recent development of new ALK TKIs and other innovative ALK-targeted therapies. We also discuss shared and ALK-specific mechanisms of resistance to targeted therapies and finally we provide insights on the developing area of ALK-directed immunotherapies that could represent new approaches associated with ALK TKIs to further improve the outcome in patients with ALK⁺ tumours.

Genetic alterations of ALK in cancer

Genetic alterations of ALK in cancer include chromosomal rearrangements, activating point mutations or ALK protein overexpression that typically supports oncogenesis by inducing constitutive activation of the ALK tyrosine kinase domain (Fig. 1). Chromosomal rearrangements generate ALK fusion proteins in which the constitutive activation of ALK is typically mediated by a forced dimerization caused by the fusion partner. Single point mutations and overexpression induce conformational changes or spontaneous dimerization, respectively, that can activate ALK and be further enhanced by engagement through ALK ligands ALKAL1 and ALKAL2 (also known as FAM150A or AUGβ and FAM150B or AUGα, respectively)¹⁸. Overall, aberrant kinase activity leads to the hyperactivation and/or dysregulation of ALK downstream signalling pathways that ultimately become oncogenic by supporting cell transformation and cancer development. These pathways are largely shared with other RTKs (Fig. 2).

Fig. 1 ∣. Alterations of the ALK protein in tumours are classified as fusions, point mutations and overexpression.

In its native form, anaplastic lymphoma kinase (ALK) is a receptor tyrosine kinase expressed in the cell membrane. When ALK is overexpressed or mutated and full-length, it is also expressed here, whereas ALK fusions remain within the cytoplasm or in the nucleus. a, The structure of the full-length ALK tyrosine kinase receptor includes the extracellular domain, which is composed of meprin, A5 protein, receptor protein-tyrosine phosphatase-μ (MAM1), LDL, MAM2 and glycine-rich domain (GRD). The transmembrane (TM) domain then sits within the cell membrane, and the tyrosine kinase domain (TKD) constitutes the intracellular portion of ALK. Amino acid residue numbers are indicated at the boundary of each domain. A list of tumours in which the ALK receptor is overexpressed is provided. b, The most frequently annotated germline and somatic alterations for each type of tumour are indicated in the respective panels. Although the pathogenicity of several somatic mutations in thyroid cancer remains to be validated, the ALK^L1198F mutation is not constitutively active but does respond to ligand activation, whereas the ALK^G1201E mutation generates an unstable receptor with very low levels of kinase activity²⁵⁷. c, ALK fusions involve a partner gene (yellow) fused to the cytoplasmic portion of ALK (exons (ex) 20–29), which includes the ALK TKD (light blue). The most frequent fusions identified for each ALK⁺ tumour are shown in the figure. Other ALK fusions reported in ALK⁺ tumours at lower frequency include: KLC1–ALK, KIF5B–ALK, DCTN1–ALK, HIP1–ALK, STRN–ALK, TPR–ALK, GCC2–ALK, CTNNA2–ALK, LMO7–ALK in ALK⁺ non-small-cell lung cancer (NSCLC); TFG–ALK, MSN–ALK, TPM3–ALK, TPM4–ALK, ATIC–ALK, MYH9–ALK, CLTC–ALK, RNF213–ALK, TRAF1–ALK in ALK⁺ anaplastic large-cell lymphoma (ALCL); nucleophosmin (NPM1)–ALK, SQSTM1–ALK, SEC31A–ALK, PABPC1–ALK in ALK⁺ diffuse large B cell lymphoma (DLBCL); KIF5B–ALK, CLTC–ALK, ATIC–ALK, EML4–ALK, CARS1–ALK, FN1–ALK, RANBP2–ALK, TIMP3–ALK in ALK⁺ inflammatory myofibroblastic tumour (IMT); EML4–ALK, TFG–ALK in ALK⁺ thyroid cancer; EML4–ALK in ALK⁺ histiocytosis; TPM3–ALK, EML4–ALK, STRN–ALK in ALK⁺ renal cancer; LRRFIP1–ALK, DCTN1–ALK, PRKD3–ALK, ATIC–ALK in ALK⁺ glioblastoma (GBM); SPTBN1–ALK, DIAPH2–ALK, STRN–ALK, WDCP–ALK in ALK⁺ colorectal carcinoma (CRC); EML4–ALK in ALK⁺ breast cancer. Specific exon breakpoints are indicated for each gene, except for EML4 (ex*) and KIF5B (ex**), which can occur at multiple points. Ex*, exons 13, 20, 6a, 6b, 14, 2a, 2b, 13b, 14 of EML4 (ref. 258). Ex**, exons 15, 24 of KIF5B²⁵⁸.

Fig. 2 ∣. ALK oncogenic signalling depends on the type of genetic alteration and cellular context.

a, Activation of full-length anaplastic lymphoma kinase (ALK) can be triggered by ALKAL1 and ALKAL2, which effectively stimulate ALK downstream signalling pathways. ALK amplifications and ALK point mutations in the kinase domain can bypass the dependence on ligand-mediated activation. Full-length amplified or mutated oncogenic ALK upregulates downstream signalling pathways, including Janus kinase (JAK)–signal transducer and activator of transcription (STAT), PI3K–AKT and MAPK pathways leading to aberrant survival and proliferation. The SHP2 phosphatase is a key effector of ALK activity, through direct binding at the SH2 domain and modulation of downstream pathways. In neuroblastoma, ALK-activating mutations and ALK amplification lead to ERK5 activation to promote proliferation⁶⁸. They also lead to phosphorylation of ATR and CHK1, potentially through AKT (dashed arrow), to sustain an effective DNA damage response¹⁸. Therapeutic strategies for ALK⁺ tumours target key signalling pathways and are indicated in the figure. b, ALK fusions are generally localized to the cytoplasm and are activated independently of ligand. The constitutive kinase activity of ALK fusions triggers a context-dependent hyperactivation of ALK downstream pathways in tumour cells. Like full-length ALK, this includes JAK–STAT, PI3K–AKT and MAPK pathways in different tumour types. PTPN1 and PTPN2 phosphatases regulate SHP2 and ALK phosphorylation. Fused ALK represses mainly PTPN1 transcription¹¹⁰. SRC is activated by fused ALK and in turn triggers the MAPK pathway. The MAPK pathway has a fundamental role in all tumour types with ALK fusions. ALK fusions regulate the Rho GTPases and WASP only in anaplastic large-cell lymphoma (ALCL), leading to changes in cell morphology and migration. Moreover, ALK fusions in ALK⁺ ALCL induces expression of CD30 and represses CD45 through STAT3 and STAT5 signalling downstream of JAK. Therapeutic strategies are indicated. ADC, antibody–drug conjugate; C term, C terminus; N term, N terminus; NF-κB, nuclear factor-κB; PROTAC, proteolysis-targeting chimera; TK, tyrosine kinase; TKI, tyrosine kinase inhibitor.

ALK fusions

The most common ALK alteration found in human tumours is a chromosomal rearrangement that leads to gene fusion and the associated expression of new ALK chimeric proteins^1,9. ALK fusion proteins are present in both solid and haematological tumours, including T cell and B cell non-Hodgkin lymphomas, NSCLC, inflammatory myofibroblastic tumour (IMT), histiocytosis, glioblastoma (GBM), thyroid cancer, colorectal cancer, renal cancer and breast cancer, among others¹ (Fig. 1). Typically, the ALK gene breaks within intron 19, regardless of the partner gene and generates a new in-frame chimeric transcript in which the 3′ portion of the ALK gene, encoding the tyrosine kinase domain, fuses with the 5′ region of a partner gene⁹. The resulting ALK fusion protein contains the entire cytoplasmic portion of ALK, including the kinase domain, but lacks the extracellular and transmembrane domains. The partner gene typically provides a dimerization domain that induces constitutive activation of the ALK kinase domain^1,19.

Mechanistically, important factors that determine the selection of the partner gene and the preferential breakpoints in ALK intron 19 include the spatial genome reorganization near the ALK and partner gene breakpoints, the levels of expression of the ALK fusion determined by the transcriptional activity of the partner gene, as well as the stability of the resulting ALK fusion protein^19,20. Once the ALK fusion is generated, a selection process favours the survival and growth of cancer cells with an advantageous ALK fusion protein that is capable of triggering and maintaining a level of oncogenic signalling that is optimal for the cell type in which it occurs^19,20. More than 100 ALK partner genes have been reported so far, and their frequency varies across different ALK-rearranged tumours, with some fusions unique to a specific cancer type^1,9,21 (Fig. 1). Depending on the partner gene, the ALK fusion proteins can be localized to the membrane, nucleus, cytoplasm or subcellular granules, which may lead to alterations in protein–protein interactions, the downstream signalling cascade and protein stability⁹. ALK partner genes generate fusion proteins with different molecular weights and protein conformational structures. These differences, particularly in the putative oligomerization domains required for constitutive kinase activation, can alter the cellular phenotype and affect sensitivity to ALK TKIs^22,23.

In solid tumours, EML4–ALK fusion is the most frequent ALK rearrangement found and is detected in 3–7% of NSCLCs and in several other cancers ^7,24 (Fig. 1). In NSCLC, different EML4–ALK variants have been reported depending on the breakpoint in the EML4 gene^19,25-27. These EML4–ALK breakpoint variants have different structures²⁸ and are associated with different responses to ALK TKIs^19,25-27. For example, the EML4–ALK variant 3 that fuses exon 6 of EML4 with exon 20 of ALK (E6;A20) is characterized by a poorer response to ALK TKIs compared with other variants^29-31. Likewise, rare cases of ALK⁺ NSCLC have been described with breakpoints in intron 17 or intron 18 instead of the canonical intron 19 (refs. 32-34). These intron breakpoint variants are less sensitive to ALK TKIs than classical intron 19 variants. This is likely due to factors such as lower fusion protein stability, which results in low protein expression and phosphorylation levels, as well as weaker downstream oncogenic signalling^19,32-34. Haematological tumours driven by ALK rearrangements are overall less frequent than ALK-rearranged solid tumours. Nucleophosmin (NPM1)–ALK is the most common fusion, representing the most frequent driver genetic alteration in 80% of paediatric ALCL cases and 50% of adult ALCL cases⁶. Several other ALK fusions have been reported in ALCL and large B cell lymphoma (LBCL)^9,35. Recurrent KIF5B–ALK fusions have been documented in a subtype of ALK⁺ histiocytic neoplasms³⁶ (Fig. 1).

With the broader use of next-generation sequencing for various tumour types, the list of ALK fusions and the associated partner genes is constantly growing, and novel ALK fusions are discovered in an increasing number of tumour types. For example, in GBM, various ALK fusions such as protein phosphatase 1 catalytic subunit-β (PPP1CB–ALK), LLRFIP1–ALK, DCTN1–ALK and PRKD3–ALK, have been recently described in children and adults⁴. PPP1CB–ALK, LLRFIP1–ALK and DCTN1–ALK fusions promoted cell transformation in vitro in 2D cell cultures and in vivo in mice, and were sensitive to ALK inhibition, likely providing a new therapeutic option for ALK⁺ GBM^4,37. In contrast, the PRKD3–ALK gene fusion is an atypical fusion resulting from exons 1–10 of PRKD3 and exons 2–29 of ALK, therefore preserving part of the ALK extracellular and transmembrane regions in addition to the cytoplasmic kinase. However, its transforming properties and sensitivity to ALK TKIs still need to be proved. Although mutations and amplifications are the common mechanisms of activation of ALK in neuroblastoma, ALK fusions, such as TENM3–ALK, are rarely seen³⁸. These examples point out that the oncogenicity of newly discovered ALK fusions should be validated through functional assays of cell transformation and assessment of sensitivity to ALK TKIs to predict their therapeutic efficacy in patients¹⁹.

Mouse models of lymphoma and lung cancer have demonstrated that the expression of an ALK fusion is sufficient to drive tumorigenesis. Expression of the NPM1–ALK fusion selectively in T or B lymphocytes of mice was sufficient to support rapid lymphoma development and at high penetrance^39,40. Likewise, the expression of the EML4–ALK fusion in lung epithelial cells was sufficient to induce the formation of lung tumours at high penetrance within 8 weeks after birth^41-44.

ALK mutations, amplification and overexpression

ALK mutations and amplifications are overall quite rare and mostly limited to subsets of neuroblastoma, thyroid cancer and melanoma⁴⁵. Gain-of-function mutations typically occur within the kinase domain of the ALK full-length receptor, and most of these mutations lead to constitutive activation of the receptor in the absence of a ligand¹. In neuroblastoma, an extracranial tumour that is most commonly found in paediatric patients, ALK mutations can be germline or sporadic and are considered driver mutations in 8–16% of cases^46-49. Three hot-spot mutations within the ALK kinase domain represent almost 85% of all ALK mutations in neuroblastoma and can activate ALK kinase in a ligand-independent manner, that is, F1174 substitutions with V, L, S, I or C, F1245 substitutions with C, I, L or V and R1275 substitutions with L or Q. The mutant F1174L is found in tumours with a more aggressive phenotype compared with other tyrosine kinase mutants, likely due to increased ATP-binding affinity and therefore kinase activation^50,51. Because MYCN amplification is the most common genetic alteration in neuroblastoma, ALK mutations, more frequently the F1174L mutation, can also occur together with MYCN amplification^50,52-54. Neuroblastoma models in zebrafish and mice have demonstrated that activating ALK mutations cooperate with MYCN overexpression to promote neuroblastoma development^52-54. Remarkably, ALK and MYCN genes are located in proximity on chromosome 2p, and simultaneous amplification of both genes can occur in neuroblastoma as a result of the genetic amplification of the chromosome 2p region^52-54. Interestingly, ALKAL2 is located within the same chromosome 2p region, supporting the hypothesis that the combined enhancement of ALKAL2–ALK–MYCN signalling is a crucial component in neuroblastoma⁵⁵ (Box 1). Moreover, the frequency of ALK mutations in neuroblastoma that relapsed after treatment is higher than in treatment-naive neuroblastoma, suggesting that ALK mutations are associated with poorer response to therapy⁵⁶. Among other tumours, ALK mutations are found in about 10% of anaplastic thyroid cancer⁵⁷.

ALK gene amplifications, as in the presence of six or more copies of the ALK gene per cell⁵⁸, have been found in 4% of high-risk neuroblastomas and are frequently correlated with poor prognosis^50,59. Although ALK overexpression can lead to ligand-independent activation, it is thought that ALK ligands could further trigger the activation of overexpressed ALK¹⁸. ALK gene amplifications are also recurrent events in congenital and adult GBM (31%), breast cancer (13%) and colorectal cancer (3%), although their clinical significance is still under investigation⁴. Overexpression of ALK due to copy number gains (that is, the presence of three to five copies of the ALK gene per cell) has been reported in 25% of rhabdomyosarcomas and is usually correlated with poor prognosis^60,61. Finally, an alternative transcription initiation start site in ALK intron 19 (ALK^ATI) creates a novel ALK isoform that is expressed in 11% of melanomas⁶². The ALK^ATI isoform encodes the intracellular portion of ALK that includes the kinase domain but lacks a fusion partner. In the absence of a fusion partner, the ALK protein expressed by ALK^ATI is poorly phosphorylated and likely only contributes weak oncogenic signalling⁶³. Consequently, melanomas expressing ALK^ATI do not respond to ALK TKIs, which leave the use of ALK TKIs in these cases still unclear^63,64.

ALK oncogenic signalling

ALK oncogenic signalling has been elucidated primarily in the context of ALK fusions in various tumour types, such as NPM–ALK signalling in ALK⁺ ALCL and EML4–ALK signalling in NSCLC as these are the most studied cancers with ALK fusions. In contrast, the oncogenic signalling of ALK mutations or amplification has been studied more extensively in neuroblastoma. Oncogenic ALK signalling shares similarities between ALK fusions and ALK mutations or amplifications, and typically drives promotion of uncontrolled cellular proliferation through the MAPK pathway, and enhances cell survival, transcription and metabolism through PI3K–AKT and JAK–STAT signalling (Fig. 2). One or more of these three pathways are almost invariably activated in any tumour type driven by an ALK activation, with additional differences correlated with the specific signalling components present in the cell of origin of ALK⁺ tumours. For example, ALK⁺ ALCLs are T cell lymphomas, and ALK signalling exploits components of T cell receptor (TCR) signalling for its oncogenic activity¹. Thus, to promote oncogenesis, ALK signalling exploits common oncogenic pathways together with cell-specific signalling modules.

Downstream oncogenic pathways

The MAPK pathway.

The MAPK pathway is activated by ALK virtually in any tumour type^1,9. In ALK⁺ NSCLC, the MAPK pathway is the main downstream mediator of the EML4–ALK fusion⁶⁵ and its activation is achieved through the formation of membrane-less cytoplasmic protein granules. These granules recruit GAB1, GRB2 and SOS1 to EML4–ALK, thereby activating RAS and triggering MAPK pathway activation and downstream proliferation⁶⁶. In ALK⁺ ALCL, NPM–ALK activates the MAPK cascade not only through RAS by directly recruiting GRB2 and SHP2, but also indirectly through the activation of other RHO GTPase signalling complexes such as VAV1–RAC1 and CDC42–WASP, which are key signalling components in T cells⁶⁷. In neuroblastoma, oncogenic full-length ALK activates not only standard components of the MAPK pathway, such as ERK1/2, but also other MAPK modules such as MEK2/3–MEK5–ERK5 (refs. 68,69) (Fig. 2).

JAK–STAT pathway

The JAK–STAT pathway typically regulates cell survival and cytokine signalling and is required for a fully transformed phenotype in ALK⁺ tumours^9,70,71. Among the various STATs, STAT3 has a fundamental role in ALK⁺ ALCL because it regulates survival, cell cycle and the expression of several key molecules that contribute to the phenotype or biology of ALK⁺ ALCL^72-74. As a consequence, STAT3 is an essential gene in ALK⁺ ALCL that can be exploited as a therapeutic target because its knockdown through short hairpin RNA (shRNA) or degradation by STAT3 degraders directly leads to lymphoma cell cycle arrest followed by cell death in cell lines and patient-derived xenograft models (PDX models)^74-77. The control of cell cycle and apoptosis is mediated by STAT3 through transcriptional regulation of cyclin D1, BCL-2, BCL-x_L, survivin and MCL1 proteins^74,75. Notably, STAT3 also regulated the transcription of PDL1, which is typically highly expressed by ALK⁺ ALCL⁷⁸. Besides STAT3, STAT5 and STAT1 also contribute to the biology of ALK⁺ ALCL. STAT5 is activated by ALK to support lymphoma proliferation and cytokine signalling⁷⁹ whereas STAT1 is downregulated by ALK and acts as a tumour suppressor in ALK⁺ ALCL^80,81.

In other ALK⁺ tumours, the role of the JAK–STAT pathway is less studied and appears to be less essential for the tumour biology. In ALK⁺ NSCLC, STAT3 activation regulates survivin expression but its knockdown is associated with a limited effect on tumour cell survival and proliferation⁸².

PI3K–AKT pathway

In ALCL, NPM–ALK activates the phosphorylation of the p85 unit of PI3Kα and its downstream effector AKT, resulting in the induction of survival signals⁸³. In contrast, the expression of PI3Kδ and PI3Kγ is repressed by NPM–ALK and restored by ALK inhibition⁸⁴. In NSCLC, the membrane-less cytoplasmic protein granules formed by EML4–ALK also contain molecules of PLCγ and PI3K signalling pathways⁸⁵, and EML4–ALK induces the phosphorylation of AKT⁸⁶. The activation of the PI3K–AKT and MAPK pathways is responsible for the upregulation of PDL1 by EML4–ALK on lung tumour cells⁸⁷. In neuroblastoma, the PI3K–AKT pathway is activated by mutant ALK together with the MAPK pathway^68,88. The activation of PI3K–AKT by ALK creates a signalling crosstalk that further potentiates the activity of MYCN, thus possibly explaining why neuroblastomas with ALK mutations and MYCN amplifications have a high-risk phenotype associated with poor patient prognosis⁸⁹. AKT activation is also stimulated by ALKAL2, suggesting that this pathway could be relevant in the context of both physiological and oncogenic ALK signalling¹⁸ (Box 1).

T cell-specific signalling in ALK⁺ ALCL.

ALK⁺ ALCL originates from T cells and displays specific features related to its cell of origin including modifications or interactions with proteins involved in TCR signalling^9,72. Most ALK⁺ ALCLs are negative for the pan-T cell markers CD3, CD5 and CD7, and positive for CD2 and CD4. In addition, although ALK⁺ ALCLs do not typically express CD8, they show positivity typical of markers of cytotoxic T lymphocytes, such as for TIA1 and granzyme B^9,72. Although some of these features are common to any ALCL irrespective of whether an ALK fusion is present, experimental evidence has demonstrated a direct role of oncogenic ALK signalling to control this peculiar phenotype in ALK⁺ ALCL. NPM–ALK can modulate T cell identity by transcriptional and epigenetic control mediated by STAT3 to repress crucial TCR downstream molecules, such as CD3ε, ZAP70, LAT and SLP76 (refs. 58,90), and induce T cell reprogramming⁹¹. Accordingly, oncogenic ALK bypasses TCR activation and independently activates downstream mediators of TCR signalling such as VAV1–RAC1 and CDC42–WASP^6,73,92-95. NPM–ALK directly promotes CDC42 and RAC1 GTPase activity, which in turn induces activation of the MAPK pathways and increased polymerization of the actin filaments⁹³, enhancing cell proliferation and survival while controlling cell migration and morphology^93-95. Blockade of CDC42 and RAC1 GTPase reverts the anaplastic morphology of ALK⁺ ALCL and prevents systemic dissemination of tumour cells^67,93. In addition, NPM–ALK downregulates WASP expression through STAT3, resulting in a further increase in CDC42 and MAPK activation⁶⁷. These pathways represent therapeutic vulnerabilities that can by exploited in ALK⁺ ALCL (Fig. 2).

The transcriptional activity mediated by STAT3 further explains phenotypes seen in ALK⁺ ALCL. For example, the expression of CD30 is characteristic of ALCL and is sustained directly by NPM–ALK through parallel activation of STAT3 and the MAPK–JUNB signalling cascade⁹. Notably, CD30 is a target for ALCL therapy via drug-conjugated antibodies such as brentuximab vedotin⁹⁶.

Additional pathways identified in specific ALK⁺ tumours.

Additional crucial pathways have been described in selective ALK⁺ tumours but could likely have a broader role. In ALCL, NPM–ALK regulates metabolism of lipids such as squalene⁹⁷, NAD synthesis via NAMPT⁹⁸, MYC expression⁹⁹, AP-1 complex activation¹⁰⁰, mTOR¹⁰¹ and DNA methylation⁹⁰. In neuroblastoma, ALK induces phosphorylation of components of the DNA damage response¹⁰², sensitizing ALK-driven neuroblastoma to therapeutic ATR inhibition. ALK and ATR combined inhibition resulted in complete tumour regression in preclinical models, suggesting that ATR inhibition could offer therapeutic benefit for high-risk patients with neuroblastoma¹⁰². In NSCLC, EML4–ALK regulates the expression of HIFs¹⁰³, PDL1 on tumour cells⁸⁷ and the expression of vimentin and cadherins involved in epithelial–mesenchymal transition (EMT)^104,105 (Fig. 2).

Phosphatases and regulation of ALK phosphorylation

ALK is activated by phosphorylation of more than ten tyrosine residues within its cytoplasmic domain. These serve as docking sites for adaptor molecules that initiate downstream signalling^106,107. There is a fine balance in ALK phosphorylation and activation in tumours; decreased ALK phosphorylation or excessive ALK activation can both independently lead to cell cycle arrest and cell death^81,108,109. In ALK⁺ ALCL, excessive ALK activation leads to oncogenic stress and DNA damage, resulting in cell cycle arrest and cell death^81,108,109. Tyrosine phosphatases that dephosphorylated ALK help to maintain the right balance of ALK activation^110,111.

The SH3 domain-containing kinase-binding protein 1 (SHP1; encoded by the PTPN6 gene) is a negative regulator of NPM–ALK¹¹¹, and thus when SHP1 is lost, JAK–STAT signalling is enhanced¹¹². Accordingly, in almost all patients with ALK⁺ ALCL, SHP1 expression is silenced owing to STAT3-mediated methylation of the PTPN6 promoter thereby enhancing JAK–STAT signalling¹¹³. PTPN1 and PTPN2 are two tyrosine phosphatases that bind to and dephosphorylate NPM–ALK¹¹⁰ (Fig. 2). Genetic loss of PTPN1 and PTNP2 is not tolerated by ALK⁺ ALCL tumour cells likely because it results in excessive ALK signalling and cell death, suggesting that inhibition of PTPN1 or PTPN2 could be a novel therapeutic option for ALK⁺ ALCL^110,114-116. On the contrary, loss of PTPN1 and PTNP2 increases resistance to the ALK TKI crizotinib because it counteracts the blockade of ALK phosphorylation¹¹⁰. Notably, PTPN1 has also been reported to repress ALK phosphorylation in both ALK wild-type and ALK-mutated neuroblastoma¹¹⁷.

SHP2 (encoded by PTPN11) is another phosphatase that is crucial for oncogenic ALK signalling, although it does not control ALK phosphorylation directly¹¹⁸. SHP2 is hyperactivated by NPM–ALK in ALK⁺ALCL, contributing to proliferation and migration through MAPK and SRC signalling¹¹⁸. In ALK⁺ NSCLC, SHP2 promotes ALK TKI resistance through a compensatory, ALK-independent activation of the MAPK pathway⁸⁶. SHP2 inhibition with SHP099 can restore sensitivity to ALK TKI in ALK TKI-resistant tumours⁸⁶. The SHP2 inhibitor TNO155 has been tested in xenograft mouse models of ALK-mutant neuroblastoma with a limited effect on survival¹¹⁹. The transmembrane protein tyrosine phosphatase (PTPase) CD45 (encoded by the PTPRC gene) is a crucial molecule for TCR activation and T cell function. In ALK⁺ ALCL, CD45 expression is partially inhibited by NPM–ALK via STAT3 and contributes to the activation of ALK oncogenic signalling^120,121. Intriguingly, some partners in ALK fusions are themselves phosphatases, such as PPP1Cβ found in PPP1Cβ–ALK in GBM, but it is unknown whether PPP1Cβ in this fusion retains phosphatase activity⁴.

Therapies for ALK-positive tumours

The dependency on ALK signalling by human ALK⁺ tumours has been exploited to design tailored therapies, most notably, specific and potent ALK TKIs. ALK TKIs inhibit the tyrosine phosphorylation activity by competitively inhibiting ATP binding in the intracellular kinase domain¹²². Therefore, ALK TKIs inhibit activation of the wild-type receptor and, fundamentally, any ALK fusion that contains the intact kinase domain. Currently, there are four generations of ALK TKIs that demonstrate increased on-target potency and improved CNS penetration with each generation. Many of these are approved or in clinical trials for ALK⁺ tumours (Table 1). Although ALK TKIs have mainly been used to treat ALK⁺ NSCLC owing to the disease prevalence, they have subsequently been approved for the treatment of other ALK⁺ cancers.

Table 1 ∣. Targeted therapies for ALK⁺ tumours.

Drug	Target	Mechanism of action	Clinical status	Tumour type	Refs.
ALK TKIs
Crizotinib (1G TKI)	ALKa	Bind to the ATP-binding pocket to prevent kinase activation	FDA approved in 2011	NSCLC	13,14,127,128,242-244
			FDA approved in 2021	Therapy-refractory or relapsed paediatric ALCL
			FDA approved in 2022	IMT
Ceritinib (2G TKI)	ALK		FDA approved in 2014	NSCLC	245
Alectinib (2G TKI)	ALK		FDA approved in 2018	NSCLC	135,139,246
Brigatinib (2G TKI)	ALK		FDA approved in 2020	NSCLC	136,144
Ensartinib (2G TKI)	ALK		FDA approved in 2024	NSCL	141,247
Lorlatinib (3G TKI)	ALK		FDA approved 2021	NSCLC	147,248-250
TPX-0131 (4G TKI)	ALK		NCT04849273 terminated (adverse change in the risk to benefit ratio)	NSCLC	158,159
NVL-655 (neladalkib) (4G TKI)	ALK		NCT05384626 phase I/II trial	NSCLC	160,161,251
Multi-RTK inhibitors
Gilteritinib	ALK and FLT3	ATP-binding pocket	NCT06225427 Phase I trial	NSCLC	162-164
Entrectinib	ALK, NTRK and ROS1		NCT02568267 Phase II trial	ALK+ tumours	165,166,252
Iruplinalkib	ALK and ROS1		Approved in China in 2023	NSCLC	168
Envonalkib	ALK, MET and ROS1		Approved in China in 2024	NSCLC	169
Covalent drugs
ConB-1	ALK	Covalent bound to ALK Cys1259	Preclinical	NSCLC	173
BNP7787	ALK	Covalent bound to ALK Cys1156		NSCLC	253
PROTACs
TL13-112b	ALK	ALK degradation by recruitment of the ubiquitin–proteasome system	Preclinical	NSCLC ALCL Neuroblastoma	174
MS4077b				NSCLC ALCL	254
MS4078b
TD-004c				NSCLC ALCL	175,255
SIAIS117d				NSCLC	175
CPD-1224e			Preclinical		176
Combination therapies
Lorlatinib and crizotinib	ALK and MET	NA	NCT04292119 Phase I/II trial	NSCLC	187,200
Lorlatinib and binimetinib	ALK and MEK	NA	NCT04292119 Phase I/II trial	NSCLC	65,195,200-202
Alectinib and cobimetinib	ALK and MEK	NA	NCT03202940 Phase I/II trial	NSCLC
Brigatinib and binimetinib	ALK and MEK	NA	NCT04005144 Terminated (low accrual)	NSCLC
Lorlatinib and TNO155	ALK and SHP2	NA	NCT04292119 Phase I/II trial	NSCLC	116,200
Lorlatinib and PF-07284892	ALK and SHP2	NA	NCT04800822 Terminated (for strategic reasons)	NSCLC	203
Ceritinib and brentuximab vedotin	ALK and CD30	NA	NCT02729961 Phase I/II trial	ALCL	9,256

This table might not be exhaustive, but it includes the most widely used or relevant treatments for ALK⁺ tumours. ALCL, anaplastic large-cell lymphoma; ALK, anaplastic lymphoma kinase; G, generation; IMT, Inflammatory myofibroblastic tumour; NA, not applicable; NSCLC, non-small-cell lung cancer; PROTAC, proteolysis-targeting chimera; RTK, receptor tyrosine kinase.

^aAlso MET and ROS1 tyrosine kinase inhibitors (TKIs).

^bCeritinib–cereblon (CRBN) ligand.

^cCeritinib–von Hippel–Lindau (VHL) ligand.

^dBrigatinib–VHL ligand.

^eAmine-containing anaplastic lymphoma kinase (ALK) inhibitor (CRBN ligand).

ALK tyrosine kinase inhibitors

First-generation TKIs.

Although initially developed to inhibit MET, crizotinib is a multitarget TKI that inhibits MET, ALK and ROS1 (ref. 123). Crizotinib was tested and found to be superior to platinum-based chemotherapy as first-line therapy in ALK⁺ NSCLC¹³. Since then, several basket trials have proved the safety and superiority of crizotinib over chemotherapy in various paediatric refractory and recurrent ALK⁺ tumours, including ALCL, IMT and neuroblastoma, producing objective and durable responses in patients with ALCL and IMT^15,124-127. Notably, more than 80% of paediatric patients with ALK⁺ ALCL experienced complete response with crizotinib as single therapeutic agent^124,125. Adult patients with refractory or relapsed ALK⁺ ALCL following chemotherapy also responded well to crizotinib^14,128-131. In neuroblastoma, crizotinib had modest efficacy, likely because it develops ALK mutations that have since been found to be identical to those conferring resistance to crizotinib in ALK⁺ NSCLC^15,125,132. However, the efficacy of crizotinib in NSCLC and ALCL is reduced by the occurrence of resistance within the first or second year of treatment^14,133. Additionally, the poor CNS penetrance of crizotinib limited its efficacy in brain metastases and prompted the development of more potent ALK TKIs¹³⁴.

Second- and third-generation TKIs.

The occurrence of resistance while on treatment with crizotinib prompted the development of novel ALK TKIs. Like crizotinib, second-generation ALK TKIs, ceritinib, brigatinib, ensartinib and alectinib, are ATP-competitive inhibitors that bind to the ATP-binding pocket of the ALK kinase domain; however, they also bind even in the presence of mutations that confer crizotinib resistance^31,135-141. Clinically, they all showed superior efficacy compared with crizotinib and were subsequently approved as first-line therapy for patients with locally advanced or metastatic ALK⁺ NSCLC^31,135-140. Because brain metastases are a frequent complication in about 30–35% of patients with ALK⁺ NSCLC, second-generation ALK TKIs were also developed to have superior CNS penetrance compared with crizotinib^31,135-140. Currently, alectinib is the most widely used among all ALK TKIs for the treatment of NSCLC, and it displays good blood–brain barrier penetration owing to its increased lipophilic properties¹⁴². The second-generation TKIs ceritinib and brigatinib are also very effective in ALK⁺ ALCL^143,144.

Whereas second-generation TKIs target the primary driver and some resistance mutations, third-generation ALK TKIs are designed to have broader mutation coverage. Lorlatinib is a third-generation ALK TKI designed to specifically target ALK mutations that are refractory to first- and second-generation ALK TKIs and to enhance penetration of the blood–brain barrier^88,145. It has broader efficacy against most of the ALK mutations that confer resistance to second-generation ALK TKIs, including the most frequent, ALK^G1202R (refs. 145,146). In a randomized phase III study designed to compare first-line response to lorlatinib versus crizotinib in advanced ALK⁺ NSCLC, lorlatinib showed improved disease control and reduced risk of CNS progression¹⁴⁷. However, a broad range of CNS-related side effects have been reported in patients¹⁴⁸.

Second- and third-generation ALK TKIs have also proved efficacious in other tumours^15,125,144. In a phase I trial, ceritinib showed good results in terms of efficacy and safety in patients with chemorefractory and relapsed ALK⁺ ALCL, IMT and neuroblastoma¹⁴³. Likewise, in a phase II trial in patients with ALK⁺ ALCL, a complete response rate of 60% was achieved with alectinib¹⁴⁹ and 73% with brigatinib¹⁴⁴ as second-line treatments¹⁴⁴. In patients with ALK⁺ IMT, other ALK TKIs have proved effective although tested in small numbers owing to the rarity of the tumour¹⁵⁰. In patients with neuroblastoma, ongoing trials with lorlatinib showed tolerability as a monotherapy or in combination with chemotherapy and demonstrated efficacy to target neuroblastoma with a broad spectrum of ALK mutations including crizotinib-resistant ALK (ALK^F1174L) (refs. 151,152). Consistently, two lorlatinib analogues, LA7 and LA9, have shown selective activity for the most recurrent lorlatinib-resistant compound mutations ALK^G1202R, ALK^I1171N, ALK^I1171S and ALK^I1171T (ref. 153).

It is likely that second- and third-generation ALK TKIs might be effective against almost any additional ALK⁺ tumour in which ALK drives the oncogenic signalling (Fig. 2); however, their relative rarity together with the lack of systematic screening for ALK alterations in tumours complicates the design of large clinical studies. Alectinib and lorlatinib have both shown efficacy in PDX mouse models of ALK⁺ LBCL, and when tested in patients with ALK⁺ LBCL that is refractory to chemotherapy, they achieved complete remission¹⁵⁴. Patients with ALK⁺ histiocytosis also showed robust and durable responses to crizotinib and thus might also respond well to second- and third-generation ALK TKIs³⁶. Finally, ALK⁺ thyroid cancers also respond to ALK TKIs¹⁵⁵, and a case of ALK-fused glioma had a near complete response to lorlatinib, suggesting that ALK TKIs that penetrate the CNS warrant further investigation for the management of ALK⁺ glioma¹⁵⁶.

Fourth-generation TKIs.

Fourth-generation ALK TKIs are engineered to target compound ALK mutants that cause resistance to second- and third-generation ALK TKIs¹⁵⁷. Specifically, TPX-0131 and NVL-655 (also called neladalkib) have been developed to target ALK mutations in cis that confer resistance to lorlatinib^157,158. Both TPX-0131 and NVL-655 are designed to be ‘double mutant active’, meaning that they target two mutations occurring in the ALK kinase domain simultaneously. Indeed, through a different binding site, TPX-0131 and NVL-655 have demonstrated activity against compound ALK mutants in phase I and phase II clinical trials (NCT04849273 (ref. 159) and ALKOVE-1 (ref. 160), respectively) for patients with pre-treated advanced ALK⁺ NSCLC. The clinical trial testing TPX-0131 has been terminated because of a non-favourable risk–benefit profile. In contrast, data from patients in the ALKOVE-1 trial have shown that NVL-655 potently and selectively targets diverse ALK fusions and secondary ALK resistance mutations, including single and compound mutations involving ALK^G1202R and ALK^I1171N. Moreover, it demonstrates effective blood–brain barrier penetration and few off-target adverse effects in the CNS¹⁶¹.

Multitarget kinase inhibitors.

Multitarget kinase inhibitors, which inhibit ALK while also efficiently targeting other kinases, have also been evaluated in preclinical and clinical settings. Gilteritinib, a FLT3-kinase inhibitor approved for FLT3-mutated acute myelogenous leukaemia, also displayed potent cross-target inhibition of ALK(I1171N)–ALK(I1171S) compound mutants, but did not inhibit ALK(G1202R) and ALK(D1203N) mutants. Additionally, it showed antitumour efficacy superior to that of alectinib in ALK⁺ NSCLC^162,163. A phase I clinical trial (NCT06225427)¹⁶⁴ has recently started to assess its efficacy in patients with advanced ALK⁺ NSCLC who have progressed after ALK TKI treatment. Another multitarget kinase inhibitor is entrectinib, a newly developed pan-neurotrophic tyrosine receptor kinase (NTRK), ALK and ROS1 inhibitor that has shown efficacy in preclinical settings and phase I clinical trials in tumours with NTRK1, NTRK2, NTRK3, ALK or ROS1 alterations¹⁶⁵. Entrectinib is currently in a phase II basket trial (STARTRK-2; NCT02568267)¹⁶⁶ to evaluate efficacy in patients assigned to different basket groups according to tumour type and gene fusion (Table 1). Interestingly, entrectinib showed activity in a patient with a neuroblastoma carrying a germline variant in the ALKLAL2 gene and a chromosomal 2p gain with ALK and TRK activity¹⁶⁷. Iruplinalkib (WX-0593), a highly selective oral ALK and ROS1 TKI¹⁶⁸, and envonalkib (TQ-B3139), an inhibitor of ALK, ROS1 and MET¹⁶⁹, are both approved in China for the treatment of patients with locally advanced or metastatic ALK⁺ NSCLC who have progressed after prior treatment with crizotinib therapy or developed toxicities to crizotinib^168,169.

Novel ALK-centric therapies

Alternative approaches to target ALK with novel compounds such as covalent inhibitors and protein degraders are currently under development^170-172. These approaches are poised to overcome limitations of the current ALK TKIs, including the reversibility of ALK inhibition. Covalent drugs incorporate reactive functional groups that can form covalent bonds with the target, usually with a cysteine residue adjacent to the ATP-binding site, and inhibit the target activity¹⁷⁰. The ALK covalent drug ConB-1 has been specifically developed to bind to cysteine at 1259, adjacent to the ALK ATP-binding site. It is selective for ALK inhibition in vitro and showed improved antitumour activity compared with ceritinib in mouse models of ALK⁺ NSCLC¹⁷³ (Table 1).

Targeted protein degradation is an emerging technology aimed at oncogenic drivers in cancer. ALK degraders are based on the proteolysis-targeting chimera (PROTAC) technology, which brings a protein of interest into proximity to an E3 ubiquitin ligase, prompting the proteasome to degrade it. Various ALK degraders have been developed whereby an ALK TKI is conjugated to cereblon (CRBN), a substrate for E3 ubiquitin ligases, or von Hippel–Lindau (VHL), an E3 ubiquitin ligase^174,175. ALK degraders have typically been developed using second-generation ALK TKIs ceritinib and brigatinib as ALK binders (Table 1). They have demonstrated efficacy in vitro in various ALK⁺ tumour cell lines, including NSCLC, ALCL and neuroblastoma^174,176; however, the in vivo efficacy still needs to be evaluated. Optimization of pharmacological properties of ALK degraders can lead to the generation of catalytic degraders with improvement in their activity and different profile of action. For example, CPD-1224 was developed from TL13–122 and showed increased potency and capability of degrading ALK carrying compound mutations L1196M and G1202R, which is typically resistant to lorlatinib. CPD-1224 also showed good pharmacokinetics and reduced ALK⁺ tumour growth in mouse models¹⁷⁶. An ALK degrader should have a more potent effect than an ALK TKI alone, as it disrupts both the kinase-dependent and kinase-independent activities of ALK. However, although ALK degraders are an evolving field, it is unclear whether they might efficaciously replace ALK TKIs in treating patients with ALK⁺ tumours.

Another potential therapeutic approach for ALK⁺ tumours could leverage that hyperactivation of oncogenic driver pathways can induce oncogenic stress. Thus, hyperactivating ALK could induce cytotoxicity and cancer cell death¹⁷⁷. For example, in cell cultures with ALK overexpression or in mice grafted with ALK TKI-resistant ALK⁺ ALCL, ALK hyperphosphorylation after sudden withdrawal of ALK TKIs led to oncogenic stress, cell cycle arrest and apoptosis^{81,108,109,178}. In these conditions, increased ALK signalling induced hyperactivation of the MAPK pathway, which triggered a DNA damage response mediated by ATM–CHK2 and γH2AX, leading to apoptosis¹⁰⁸. Moreover, ALK hyperactivation after TKI withdrawal led to overactivation of STAT1, an oncosuppressor in ALK⁺ ALCL⁸¹. Similarly, the deletion of PTPN1 or PTPN2 caused lymphoma cell death by ALK hyperactivation¹¹⁰. Overall, these studies suggest that cancer cells require adequate levels of ALK activation for their fitness and that a swing between states of ALK inhibition and ALK hyperactivation could be exploited for therapeutic purposes.

Mechanisms of resistance to ALK-directed therapies

Despite the success of ALK inhibitors, resistance in tumours develops quite invariably. In some patients, resistance can be observed soon after the initiation of ALK TKI treatment, while in others resistance develops much later, even several years after response¹¹. By far the most studies that focus on mechanisms of resistance in ALK⁺ tumours have been carried out in the context of ALK⁺ NSCLC. The mechanisms of resistance in ALK⁺ tumours can be divided into four main mechanisms: first, ALK mutations affecting the kinase domain that impair binding of the ALK TKI; second, hyperactivation of compensatory pathways that compensate for ALK blockade; third, switch to a different histological subtype; and fourth, resistance by microenvironmental signals (Fig. 3). It is still unclear how and when these events, or a combination of them, occur temporally in tumours that develop rapid resistance compared with tumours that respond and remain in prolonged remission. In this latter situation, ALK⁺ tumour cells can survive in a dormant, non-dividing state for years⁸⁴, and thus, when assessing and targeting resistance to ALK TKIs, drug-tolerant persister cells must be addressed¹⁷⁹. Such persister cells can likely survive through intrinsic mechanisms of adaptation or through survival signals from the tumour microenvironment (TME)^84,179.

Fig. 3 ∣. ALK⁺ cancer cells can bypass signalling blockade achieved by ALK tyrosine kinase inhibitors through both cell-intrinsic and cell-extrinsic mechanisms.

a, Cell-intrinsic mechanisms include anaplastic lymphoma kinase (ALK) amplification and ALK point mutations that affect the ALK kinase domain in both fusions and full-length ALK (indicated by yellow stars). b, ALK-independent mechanisms include reactivation of upstream receptor tyrosine kinases (RTKs) through activation of SHP2 or downstream signalling through activating mutations in the MAPK pathway. YAP–TEAD signalling axis promotes the selection of ALK⁺ drug-tolerant persister cells. c, Cell-extrinsic mechanisms include tumour microenvironment (TME)-mediated survival. This can be triggered by C-C chemokine receptor type 7 (CCR7)–PI3Kγ signalling, whereby endothelial cells that produce the CCR7 ligands chemokine (C-C motif) ligand 19 (CCL19) and CCL21 protect tumour cells from apoptosis induced by ALK inhibitors in the perivascular niche. Human growth factor (HGF) and neuregulin 1 (NRG1) produced in the TME can contribute to resistance in ALK⁺ non-small-cell lung cancer (NSCLC) by engaging bypass pathways such as MET and epidermal growth factor receptor (EGFR), respectively. d, ALK⁺ tumours can further undergo morphological changes that confer resistance. Epithelial–mesenchymal transition (EMT) is associated with loss of E-cadherin and gain of vimentin and with expression of proteins that support the survival of persister cells. ALK⁺ NSCLC can also undergo transition to different histological subtypes including neuroendocrine differentiation or histological switch to small-cell lung cancer or to squamous cell carcinoma. CAF, cancer-associated fibroblast.

Mutations of the ALK kinase domain that impair binding of ALK TKIs

ALK mutations that affect the tyrosine kinase domain provide mechanisms for resistance to ALK TKIs, whereby the relative frequency of each mutation depends on the generation of the inhibitor and on whether ALK TKIs are used as first-line, second-line or third-line therapy. Notably, the sequential use of different ALK TKIs can favour the accumulation of subsequent ALK mutations over alternative mechanisms of resistance¹¹. For example, mutations are found in 42% of patients with alectinib used as first-line treatment but in 67% of patients who received alectinib as second-line treatment, suggesting that treatment-naive tumours are more likely to develop mutation-independent mechanisms of resistance to alectinib¹⁸⁰. Single and compound mutations that induce resistance to lorlatinib are more frequent when lorlatinib is used as second-line therapy, rather than as first-line treatment^153,181,182. Mutational hot spots within the kinase domain have been identified largely by sequencing of primary samples or circulating tumour DNA (ctDNA) from patients with ALK⁺ NSCLC^146,183. They show a broad pattern of mutational hot spots, even within resistance mechanisms for the same generation of ALK TKIs. These mutations almost invariably reduce the binding affinity of the ALK TKI by steric hindrance, or alter the kinase domain conformation and thus ATP binding¹⁴⁶ (Fig. 3). L1196M and G1269A compound changes in ALK restrict access to the ATP-binding pocket, reducing crizotinib binding¹⁸⁴. Mutations affecting the surface of ALK, such as G1202R, G1202 deletion, D1203N, S1206Y and S1206C, impair ALK TKI binding by disrupting the solvent-facing surface of ALK¹¹ (Fig. 4). Interestingly, although some mutations, such as G1202R, are found in tumours resistant to most ALK TKIs, other mutations are selective for specific ALK TKIs. For example, in ALK⁺ NSCLC, ALK^I1171N, ALK^I1171S and ALK^I1171T are mutations that are seen in alectinib-resistant tumours but not in ceritinib- or brigatinib-resistant tumours¹⁴⁶. Likewise, in ALK⁺ ALCL treated with crizotinib, resistant tumours are enriched for Q1064R and I1171N mutations that are not typically seen in other ALK⁺ tumours treated with crizotinib¹⁴.

Fig. 4 ∣. The crystal structures of the kinase domain of ALK and the mutations associated with resistance to various tyrosine kinase inhibitors.

a, Crystal structure of the human wild-type (WT) anaplastic lymphoma kinase (ALK) kinase domain. b-g, The structure of the ALK kinase domain is shown here in complex with various tyrosine kinase inhibitors (TKIs) (red). They are sorted according to their generation (b, first generation; c–e, second generation; f, third generation; g, fourth generation). In particular, panels b–f show the most frequently mutated residues responsible for the mechanisms of resistance to TKIs in ALK⁺ tumours, and panel g shows the two lorlatinib-resistant compound mutations L1196M and G1202R against which NVL-655 (neladalkib) has been predicted to be active. h, The common ALK mutations in neuroblastoma are depicted here (namely, F1174L, F1245V and R1275L). These can also occur as compound mutations F1174L and G1202R, F1174L and D1203N or F1174L and L1196M. The images were generated with Open-Source PyMOL (Schrödinger, LLC, New York, NY, USA) using the following Protein Data Bank (PDB) structures: 3L9P for WT human ALK and neuroblastoma, 2XP2 for crizotinib; 3AOX for alectinib; 6MX8 for brigatinib; 4MKC of ceritinib, 4CLJ for lorlatinib; 9GBE for NVL-655. Computational mutagenesis was applied to residues known to confer resistance to TKIs to highlight their potential impact on drug binding and efficacy. The mutated residues are represented concurrently for illustrative purposes.

In addition, compound mutations have been frequently observed in patients treated with third-generation ALK TKIs such as lorlatinib^153,181,182. It is likely that these mutations are compounded because the ALK gene initially acquires resistance mutations in response to first- or second-generation ALK TKIs, and subsequently acquires further mutations upon exposure to lorlatinib^153,181. Although fewer data are available for other types of ALK⁺ tumour, the pattern of newly acquired mutations affecting the kinase domain of ALK in patients with neuroblastoma who were treated with lorlatinib¹⁸⁵ indicates that ALK compound mutations can still be acquired with lorlatinib treatment in the absence of previous treatment with earlier-generation ALK TKIs. These include the following co-mutations in ALK: F1174L and G1202R, F1174L and D1203N, F1174C and G1202R, and F1174L and L1196M. Fourth-generation ALK TKIs might overcome some double or triple compound mutations that occur in cis in the ALK gene. TPX-0131 (ref. 158) and NVL-655 (refs. 157,186) have proved active in vitro against compound ALK mutations such as G1202R and L1196M, G1202R and G1269A, and G1202R and L1198F. Gilteritinib has also shown promising in vitro activity against ALK(I1171N) in combination with other mutations¹⁶².

Hyperactivation of compensatory pathways

With the growing potency of ALK inhibitors, an increasing number of tumours develop resistance through ALK-independent signalling mechanisms. These involve genetic activation, overexpression or engagement of autocrine feedback signalling pathways. The most common mechanism is through activation of an RTK with a similar signalling pattern to ALK. In ALK TKI-resistant NSCLC, this includes genetic amplification, overexpression or hyperactivation of MET, EGFR, IGF1R, HER2, HER3, KIT and FGFR^187-193. Activation can be also achieved by overexpressing the HER3 ligand, neuregulin 1 (ref. 194). Additionally, in patients with ALK⁺ NSCLC, activating mutations of MAP2K1 (refs. 65,195), SRC¹⁹⁵, BRAF¹⁹⁶, NF2 (ref. 197) or copy-gain mutations in wild-type KRAS or the loss of negative regulators of the MAPK pathway, such as DUSP6 (ref. 65), are further mechanisms that can activate the MAPK pathway to contribute to ALK TKI resistance. Activation of SHP2 downstream of several RTKs can also mediate resistance through compensatory pathways¹¹⁶. Inhibition of SHP2 overcomes resistance to ALK TKIs in cell lines⁸⁶, suggesting that SHP2 could be a target to bypass MAPK activation and circumvent ALK TKI resistance¹¹⁶. Finally, blockade of the focal adhesion kinase (FAK)–YAP signalling axis could also revert ALK TKI resistance and impair the survival of persister cells^198,199 (Fig. 3).

Together these data indicate that resistance is acquired through selection for mechanisms that reactivate the same essential pathways inhibited by ALK TKIs, supporting a combinatorial use of TKIs. Patients with ALK⁺ NSCLC who acquire MET amplification achieved a clinical response in a case study with two patients¹⁸⁷. Ongoing clinical trials include ALK TKIs in combination with MET inhibitors (NCT04292119)^187,200, MEK inhibitors (NCT04292119 (ref. 200), NCT03202940 (ref. 201) and NCT04005144 (ref. 202)) or SHP2 inhibitors (NCT04292119 (ref. 200) and NCT04800822 (refs. 203,204)). Preliminary results from a customized phase I clinical trial to test a novel allosteric SHP2 inhibitor (PF-07284892) in combination with lorlatinib showed a partial response in a patient with ALK⁺ NSCLC whose disease had progressed on lorlatinib monotherapy²⁰⁴. Data on ALK TKI bypass mechanisms for other ALK⁺ tumours are scant. However, neuroblastoma that developed resistance to lorlatinib shows enrichment of mutations in the MAPK and PI3K pathways, including oncogenic alterations of MET, BRAF, EGFR, FGFR1, PTPN11, PIK3CA and RAS GTPases^185,205. In lymphoma, activation of the JAK–STAT pathway through IL-10 signalling²⁰⁶ or PDGFR²⁰⁷ can promote resistance to ALK TKIs.

Histological switch and microenvironment-mediated resistance

Histological switch occurs in NSCLC when the tumours acquire histological features different from those of an adenocarcinoma and consequently resemble another subtype of lung cancer. This is a rare mechanism of resistance to ALK inhibitors in patients with ALK⁺ NSCLC; however, when it happens, ALK⁺ lung adenocarcinoma switches to small-cell lung cancer^208,209 or, more rarely, to squamous cell carcinoma²¹⁰. NSCLC with neuroendocrine differentiation²¹¹ has also been reported during treatment with almost all generations of ALK TKIs¹¹. During this transformation, tumours typically retain the same ALK genetic alterations, although ALK protein expression can be reduced^208,209. They likely become resistant to ALK TKIs by activating compensatory pathways upregulated by the histological switch. In other cancers, such as neuroblastoma, ALK signalling maintains tumour cells in an undifferentiated state by repressing DLG2, such that blocking ALK induces differentiation of neuroblasts into cells with lower migration and proliferation²¹². Finally, ALK⁺ NSCLC can acquire features associated with EMT in ALK TKI-resistant tumours^104,146,213. However, the mechanism by which EMT could drive resistance to ALK TKIs is still unclear and likely depends on expression of proteins that support the survival of persister cells.

There is growing evidence that the TME contributes to resistance to ALK inhibition. Environment-mediated drug resistance has been described in multiple tumour types²¹⁴ and in the case of ALK⁺ NSCLC is partially mediated by cancer-associated fibroblasts (CAFs) because in vitro co-culture of ALK⁺ tumour cells with CAFs or their conditioned medium substantially reduces their sensitivity to TKIs^215,216. In a preprint, stromal niches were shown to protect ALK⁺ lung cancers as tumour cells proximal to the stroma were less sensitive to alectinib owing to stroma-induced activation of MET in tumour cells²¹⁷. In ALK⁺ ALCL, compensatory activation of the MAPK pathway mediated by endothelial production of CCL19 and CCL21, which activates CCR7 and PI3Kγ signalling, suggests that the perivascular niche can amend resistance to ALK inhibition⁸⁴. CCL19 and CCL21 are secreted not only by endothelial cells but also by CAFs and macrophages⁸⁴, suggesting that multiple cell types in the TME could further foster a milieu that favours the persistence of ALK⁺ tumour cells during TKI therapy (Fig. 3). Some ALK fusions, such as EML4–ALK V1, EML4–ALK V3, KIF5B–ALK and TFG–ALK, might also trigger an inflammatory response in the TME by upregulating SERPINB4, which promotes cell survival by inhibiting natural killer cell-induced cell death²¹⁸.

ALK activation as secondary event that drives resistance to other targeted therapy in tumours

Retrospective sequencing analysis in naive untreated NSCLC demonstrates that ALK alterations are mutually exclusive with other known oncogenic driver genes, KRAS and EGFR²¹⁹. However, ALK alterations have been reported as secondary mechanisms of resistance in tumours treated with KRAS²²⁰ or EGFR²²¹ inhibitors. Blockade of such driver oncogenes through targeted therapy allows for the emergence of resistance clones that simultaneously harbour ALK alterations^220,221. These alterations are typically ALK rearrangements and can result from either the selection of rare pre-existing subclonal populations already containing an ALK rearrangement or de novo ALK alterations acquired during treatment¹⁹. Although extensive clinical confirmation is needed, we expect that tumours acquiring an ALK rearrangement as a mechanism of resistance to another targeted therapy will respond to ALK TKIs.

Emerging ALK-directed immunotherapies

Immune checkpoint inhibitors (ICIs) have become standard therapy for many human tumours, either as monotherapy or in combination with other drugs. In ALK⁺ tumours, some efficacy of ICIs has been reported in patients with ALK⁺ ALCL^222,223; however, patients with ALK⁺ NSCLCs do not typically respond to ICIs^224,225. This has prompted the development of new immunotherapeutic approaches to specifically target ALK⁺ cancers.

Several features make ALK an especially attractive target for immunotherapies, including its oncogenicity, its restricted expression to tumour tissue rather than healthy adult tissue and its immunogenicity^1,226. Initial evidence of ALK immunogenicity was found in the serum of patients with ALK⁺ ALCL who showed spontaneous immune responses to the ALK segment of the ALK fusion protein, and in titres of ALK antibodies from treatment-naive patients with ALK⁺ ALCL, where the titre inversely correlated with disease stage, number of circulating tumour cells and the cumulative incidence of relapse²²⁷. Furthermore, ALK-specific tumour-reactive T cells can be detected in peripheral blood mononuclear cells isolated from treatment-naive patients with ALK⁺ ALCL, but not from healthy volunteers²²⁸. Likely these spontaneous T cell responses are primed by the presentation of ALK peptides that originate from the fusion protein by major histocompatibility complex class I (MHC-I) molecules of tumour or antigen-presenting cells²²⁹. Remarkably, the presence of a pre-existing spontaneous immune response towards the ALK protein in patients with ALK⁺ ALCL inhibits dissemination and decreases the risk of relapse²²⁷. Fifteen to twenty per cent of patients with ALK⁺ NSCLC have high anti-ALK antibody levels, which correlates with improved survival²³⁰. Although antibodies specific for ALK likely do not have a direct biological impact on the disease owing to the cytoplasmic localization of the ALK fusions in ALCL and NSCLC, they are likely readouts of an existing immune response against ALK. Larger studies are needed to determine whether, like patients with ALK⁺ ALCL, ALK antibodies are prognostic indicators of risk of relapse in patients with ALK⁺ NSCLC. These data suggest that therapeutically increasing the immune response against ALK through immune-based therapies such as cancer vaccines or cellular immunotherapies could be beneficial for patients. Therefore, perhaps unsurprisingly, out of a priority-ranked list of 75 tumour antigens evaluated by the US National Cancer Institute for cancer vaccine development, ALK ranked fourth on the basis of its biological and immunogenicity properties²³¹. In the following sections, we discuss the emerging field of ALK-directed immunotherapies, some of which are being tested in clinical trials (Fig. 5).

Fig. 5 ∣. ALK is a promising target for immunotherapies, owing to its oncogenicity, its spontaneous immunogenicity and that its expression is restricted to tumour cells.

a, Anaplastic lymphoma kinase (ALK) immunotherapies against tumours expressing ALK fusions rely on the presentation of ALK peptides on tumour cells through major histocompatibility complex class I (MHC-I) or MHC-II. ALK vaccines, delivering the entire cytoplasmic portion of ALK or selected ALK peptides, boost priming of ALK-specific CD8⁺ T cells through presentation of ALK antigens by dendritic cells (DCs). T cells engineered to express T cell receptors (TCRs) that recognize ALK peptides displayed in MHC-I or MHC-II elicit anti-tumoural effects through the secretion of inflammatory cytokines such as IL-2, interferon-γ (IFNγ) and tumour necrosis factor (TNF) and through cytolytic effector function against tumour cells via perforin (PFN) and granzyme B (GZMB). b, Tumours expressing transmembrane full-length ALK can be targeted with anti-ALK blocking antibodies, anti-ALK antibody–drug conjugates or chimeric antigen receptor (CAR) T cells engineered to specifically bind to ALK. These approaches selectively target both wild-type and mutated ALK-expressing cells. ALCL, anaplastic largecell lymphoma; ER, endoplasmic reticulum; NSCLC, non-small-cell lung cancer; Ub, ubiquitin.

ALK vaccines

Evidence for immunogenicity of the ALK protein has prompted the development of ALK vaccines to boost the endogenous compensatory response. In principle, a shared-antigen ALK vaccine could be designed to target the wild-type protein or alternatively, a neoantigen ALK vaccine could be designed to target tumour-specific ALK mutants or ALK fusions. In the first option, the vaccine could be an mRNA encoding the entire ALK cytoplasmic portion that is invariably expressed in any ALK fusion. Conversely, shorter mRNA or peptides could be used to induce antigen presentation and immune responses against selected sequences of wild-type or mutant ALK (Fig. 5a).

In support of a shared-antigen approach, vaccination with DNA encoding the cytoplasmic domain of the wild-type ALK protein elicited CD8⁺ cytotoxic T cell responses and provided long-term protection and therapeutic benefit in mouse models of ALK⁺ lymphoma²³² and ALK⁺ lung cancer⁴². These activities required IFNγ but not the presence of B cells²³², suggesting that the antitumour efficacy of the ALK vaccine resides in its capacity to elicit specific T cell responses instead of antibodies produced by B cells. ALK immunogenic peptides that elicit CD4⁺ or CD8⁺ T cell responses have been identified in mice and humans²²⁹. A vaccine based on a single ALK peptide was sufficient to cure up to 70% of mouse models of ALK⁺ lung cancer and prevent metastatic spread to the brain in 100% of mice when combined with an ALK TKI and ICIs²²⁹. In these models, those 30% of mice that were not cured by the ALK vaccine exhibited MHC-I downregulation in tumour cells, enabling immune escape. However, that was reversible by IFNγ or STING activation²²⁹. By studying ALK peptides in human cells, peptides presented in HLA-A*0201 or HLA-B*0702 have been identified and demonstrated to be immunogenic when used as vaccines in mice transgenic for those human HLA alleles²²⁹. Notably, patients with ALK⁺ NSCLC may develop spontaneous T cell responses to these same peptides without vaccination, indicating that ALK⁺ tumours can elicit selective anti-ALK responses by the immune system²²⁹. It is likely that additional ALK peptides could be displayed by tumour cells in different tumour HLAs, therefore expanding the repertoire of ALK peptides and patient HLAs that are suitable for vaccination. An ALK vaccine based on the full cytoplasmic portion of ALK would be a broader approach that is not restricted to HLA-A*0201 or HLA-B*0702.

As for the neoantigen approach, ALK vaccines could be developed that encode recurring mutations in ALK, such as the recurrent mutations found in ALK TKI-resistant lung cancers¹¹ or the recurring activating mutations found in ALK⁺ neuroblastoma^46-49. A phase I trial to test an ALK vaccine against resistance-conferring ALK mutations in ALK⁺ NSCLC is ongoing (NCT05950139)²³³. Alternatively, vaccines against specific ALK fusions, such as the EML4–ALK or the NPM1–ALK fusion could be in principle used to target the neoantigen (or neoantigens) created by the fusion portion. As fusions are not expressed in normal tissues, this approach could further increase the safety of ALK vaccination.

ALK antibodies

For tumours that express the full-length receptor, such as neuroblastoma, Merkel cell carcinoma, GBM and rhabdomyosarcoma, ALK antibodies may be a therapeutic option either as blocking agents or in the form of antibody–drug conjugates (ADCs). Earlier studies in human cell lines showed that ALK antibodies could block ALK activity, inhibit downstream signalling and suppress neuroblastoma cell growth^106,234. An ADC directed to ALK reduced tumour growth by selectively binding and delivering the conjugated drug to ALK⁺ cells in mouse models of neuroblastoma²³⁵ (Fig. 5) showing the potential for ADCs that target ALK. Characterizing the ALK binding sites has provided mechanistic insight into the biological activity of ALK antibodies and will likely help in the rational design of new ALK antibodies with increased efficacy²³⁶. Additionally, the insights into the structural interactions between ALK and its ligands ALKAL1 and ALKAL2 can be leveraged to develop highly specific antibodies to block ALK signalling in tumours such as neuroblastoma^2,3. Despite this encouraging progress, drug-conjugated anti-ALK antibodies have not yet reached clinical trials.

ALK-specific CAR T cells and cellular therapies

For similar reasons to ALK vaccines and ALK antibodies, ALK is also amenable to targeting with cellular immunotherapies. Tumours that express the full-length ALK receptor on the surface, such as neuroblastoma, can be targeted by chimeric antigen receptor (CAR)-T cells specific to ALK. However, in tumours that express ALK fusions, and thus do not express ALK on the surface, ALK cannot be targeted using canonical CAR T cells (Fig. 5b). In earlier work in PDX mouse models of neuroblastoma, the efficacy of ALK.CAR T cells was limited for cells expressing low levels of surface ALK²³⁷. More recently however, ALK. CAR T cells generated with a different antibody with different binding properties have shown high efficacy as monotherapy against neuroblastoma cells expressing high levels of ALK on the cell surface, achieving a complete cure in preclinical mouse models of metastatic ALK⁺ neuroblastoma⁵. Not unexpectedly, ALK.CAR T cells were less effective against neuroblastoma cells with lower surface ALK expression. However, when ALK carries an activating mutation, treatment with an ALK TKI can increase the abundance of the ALK protein on the surface of tumour cells through receptor trafficking²³⁸ and increased ALK gene transcription^5,47. Consequently, treatment of neuroblastoma with an ALK TKI increases the density of ALK on the surface of tumour cells thereby facilitating targeting by ALK.CAR T cells⁵. A phase I/II clinical trial to treat patients with relapsed or refractory neuroblastoma with ALK.CAR T cells is ongoing (NCT06803875)²³⁹. In principle, ALK.CAR T cells could be used therapeutically for other tumours that express full-length ALK as a receptor on their surface, including but not limited to Merkel cell carcinoma, melanoma, Ewing sarcoma and some small-cell lung cancers.

In contrast, for tumours that express ALK only in the cytoplasm, T cell therapies could be developed by engineering T cells with TCRs that specifically recognize ALK peptides presented in selective HLAs, such as those restricted to HLA-A*0201 or HLA-B*0702 (ref. 229). Although this work is still far away from clinical application, initial results in vitro show the feasibility of cloning TCRs that recognize ALK peptides presented in select HLAs and the subsequent development of engineered CD8⁺ ALK. TCR T cells to target ALK⁺ NSCLC²⁴⁰ (Fig. 5a). As further support for this concept, recent work identified a TCR in patients with ALK⁺ ALCL that specifically recognizes a peptide generated by the NPM1–ALK fusion that is presented by the human MHC-II complex²⁴¹.

Conclusions and perspectives

Just over 30 years have passed since ALK was first cloned and characterized in a subtype of human lymphoma⁶. During this time, our understanding of its function in cancer has provided patients with ALK⁺ tumours with a better outcome than those with many other oncogene-driven tumours¹¹. So far, this progress has been made possible by the development of ALK inhibitors with increased potency, a deeper understanding of the biology of ALK⁺ tumours and the fact that patients with ALK⁺ tumours are typically of a young age and good fitness. Although ALK inhibitors are a spectacular example of the progress and potential of targeted therapies, the final step to achieve a definitive cure for ALK⁺ tumours has yet to be taken. The key challenge that remains is the development of additional therapies to eradicate drug-tolerant persister cells that survive ALK inhibition. To achieve this, we first need a better understanding of the tumour-intrinsic mechanisms that support the survival of these persister cells during ALK inhibition. Then, we would need to identify efficacious drugs or other modalities to cut off these mechanisms in combination with ALK TKIs; however, toxicity and tumour adaptation to combinatorial drug therapies will likely be a high bar to overcome. Alternatively, or simultaneously, we need to develop tumour-extrinsic orthogonal strategies to eradicate persister cells, such as leveraging the immune system through vaccines and cellular therapies or by targeting the tumour microenvironment. Successfully application of these strategies to ALK⁺ tumours would also establish a paradigm to improve the treatment of tumours driven by other tyrosine kinases or driver oncogenes.

Glossary

Antibody–drug conjugates (ADC).

Targeted therapies that link an antibody to a cytotoxic drug, delivering the drug specifically to cancer cells by binding to the antigen expressed on their surface.

Basket trials

Clinical trials that test the effectiveness of a single drug or treatment on multiple types of cancer that share a common genetic mutation, regardless of the cancer cell of origin or cancer type.

Breakpoint variants

Different, yet specific, locations within the same gene in which chromosomal breaks occur, leading to structural rearrangements such as translocations or fusions.

Circulating tumour DNA (ctDNA).

Fragments of DNA released into the bloodstream from cancer cells, which can be analysed to detect and monitor presence, progression and response to treatment.

Complete remission

The absence of all detectable signs of disease following cancer treatment.

Compound ALK mutants

The presence of two or more mutations within the same gene, occurring either on the same allele (cis) or different alleles (trans). They usually arise as a mechanism of resistance to targeted therapy.

Drug-tolerant persister cells

A small subpopulation of cancer cells that enter a dormant state, allowing them to survive harsh conditions or treatments, such as chemotherapy or tyrosine kinase inhibitor, and potentially cause recurrence or treatment failure.

First-line therapy

The initial treatment recommended for a particular disease or condition, typically based on its effectiveness and safety.

Oncogenic stress

A detrimental effect on tumour cell fitness that results from the overexpression or excessive activation of an oncogene, which can cause cellular stress, senescence or death, which can be exploited for cancer therapy.

Partner gene

One of the two genes involved in a fusion event, often due to chromosomal rearrangements, that can lead to abnormal gene function.

Patient-derived xenograft models (PDX models).

Models in which tumour tissue from a patient is implanted into immunodeficient mice, allowing researchers to study the tumour’s growth, drug response and biology in vivo while maintaining its original characteristics.

Steric hindrance

The prevention of chemical reactions or interactions owing to physical obstruction caused by the spatial arrangement of atoms or groups within a molecule.

Tyrosine kinase inhibitors (TKIs).

Small drugs capable of binding to the catalytic domain of a tyrosine kinase and disrupting the kinase activity and downstream signal transduction pathways.

Data availability

The crystal structures used for Fig. 4 were downloaded from PDB with the following accessions: 3L9P for WT human ALK and neuroblastoma, 2XP2 for crizotinib; 3AOX for alectinib; 6MX8 for brigatinib; 4MKC of ceritinib, 4CLJ for lorlatinib; 9GBE for NVL-655. PyMol session files of the above structures with the in silico mutated residues reported in Fig. 4 are available in the public figshare repository (https://figshare.com/projects/ALK/234473).