Turning liabilities into opportunities: Off-target based drug repurposing in cancer

Abstract

Targeted drugs and precision medicine have transformed the landscape of cancer therapy and significantly improved patient outcomes in many cases. However, as therapies are becoming more and more tailored to smaller patient populations and acquired resistance is limiting the duration of clinical responses, there is an ever increasing demand for new drugs, which is not easily met considering steadily rising drug attrition rates and development costs. Considering these challenges drug repurposing is an attractive complementary approach to traditional drug discovery that can satisfy some of these needs. This is facilitated by the fact that most targeted drugs, despite their implicit connotation, are not singularly specific, but rather display a wide spectrum of target selectivity. Importantly, some of the unintended drug “off-targets” are known anticancer targets in their own right. Others are becoming recognized as such in the process of elucidating off-target mechanisms that in fact are responsible for a drug’s anticancer activity, thereby revealing potentially new cancer vulnerabilities. Harnessing such beneficial off-target effects can therefore lead to novel and promising precision medicine approaches. Here, we will discuss experimental and computational methods that are employed to specifically develop single target and network-based off-target repurposing strategies, for instance with drug combinations or polypharmacology drugs. By illustrating concrete examples that have led to clinical translation we will furthermore examine the various scientific and non-scientific factors that cumulatively determine the success of these efforts and thus can inform the future development of new and potentially lifesaving off-target based drug repurposing strategies for cancers that constitute important unmet medical needs.

1. Introduction

In the past two decades groundbreaking advancements in molecular biology and genomics technologies, powerfully manifested in the sequencing of the human genome, have enabled the birth of the concept of personalized or precision medicine^1–4, which means to give the right drug to the right patient at the right time, and which revolutionized particularly the field of cancer research and clinical oncology. The development of potent new targeted drugs, such as the therapeutic antibody trastuzumab for HER2-positive breast cancer and the tyrosine kinase inhibitor (TKI) imatinib for BCR-ABL-driven chronic myeloid leukemia (CML)^5–8, has not only dramatically prolonged and improved the lives of many cancer patients, but also created profound excitement among scientists in cancer biology and drug discovery across academia, government and pharmaceutical and biotechnology industry. These developments were to large extent fueled by next generation high-throughput DNA and RNA sequencing technologies^9–12, which enabled initiatives like The Cancer Genome Atlas (TCGA) and provided a new, genomically founded and mechanistic understanding of the root causes of many cancers and resulted in the identification of an increasing number of novel cancer targets^13–16. Arguably, these exciting discoveries could not have had such paradigm-changing impact if not accompanied by similarly important advances in medicinal chemistry and structural biology. These enabled the synthesis of small molecule inhibitors of cancer-related proteins, predominantly tyrosine kinases that in contrast to early inhibitors like staurosporine exhibited not only high potency, but also improved selectivity for the desired targets.

Although these successes created a wave of discoveries of new actionable targets and targeted drugs for various cancers, it quickly became apparent that few of these novel targeted therapies had curative potential. Rather, drug resistance developed in many cases rapidly through a plethora of mechanisms^17–19, as has been powerfully illustrated for imatinib in CML as well as the BRAF inhibitor vemurafenib in melanoma^20–22. This further increased the demand for new drugs that could overcome resistance to first generation targeted agents. At the same time, these aspects unveiled the limitations and challenges of traditional drug discovery approaches within the pharmaceutical and biotechnology sector^23–25, most notably steadily increasing research and development costs, high attrition rates along the course of clinical development, including failure in expensive late-stage clinical trials, thinning drug pipelines and long development times ranging in average between 10 to 15 years from lead discovery to final market approval²⁶. Considering that rapid development of drug resistance, besides of its obvious clinical ramifications, also means a reduced ability of pharmaceutical companies to recover their initial investments into individual drugs and that precision/personalized medicine with its smaller market shares for each drug per definitionem embodies a profoundly different concept from previous and highly lucrative “one drug fits all” blockbuster drug paradigms of the past, this cumulatively put increasing financial pressure on drug discovery companies.

In light of these challenges, the concept of drug repurposing (also known as ‘drug repositioning’) has gained much attention and momentum in recent years^{25; 27–29}. ‘Drug repurposing’ refers to the identification of new therapeutic applications of existing or investigational drugs beyond their initially intended indication. This may even allow ‘drug rescuing’ of compounds that have failed in initial clinical trials. Identifying new therapeutic uses for existing molecules has obvious benefits for patients that may otherwise not have any efficacious therapy options, but it also provides a powerful avenue towards new research tools for interrogating underlying disease biology and not least it provides an opportunity for pharmaceutical industry to increase their earnings from a particular drug. Thus, drug repurposing can complement the undisputedly valuable traditional drug discovery approach while at the same time relieving some of the aforementioned financial pressure on drug discovery companies. This does not mean that drug repurposing approaches do not require additional investments in themselves or are without risks. In fact, drug repurposing approaches are similarly likely to fail in phase II or III trials for lack of efficacy, which is the major reason for drug failure in recent years accounting for ~50–60%³⁰, as traditional drug development approaches for a drug’s original indication^{23; 31}. However, as drug repurposing candidates have gone through significant preclinical testing as well as phase I safety studies, the risk for toxicity-associated failure is usually low. Furthermore, pharmacokinetics and pharmacodynamics profiles, formulations and dosing schedules are at this point well established, which cuts down significantly on further costs and development times^{24; 25}. As a consequence, there is a plethora of successful drug repurposing cases in cancer and many other diseases, which have been comprehensively reviewed elsewhere^{29; 32–35}. Early successes in this field have in addition inspired dedicated efforts to harness this concept in the form of public-private partnerships by the Medical Research Council (MRC) in the United Kingdom and the National Institutes of Health (NIH) in the United States, which founded the National Center for Advancing Translational Sciences (NCATS), with various pharmaceutical companies^{36; 37}.

The majority of repurposing projects can be traced back to a new mechanistic understanding on the biology of a drug’s intended (“canonical”, “cognate”) target or pathway in a different disease. Examples include repurposing of the immunosuppressant mTOR inhibitor rapamycin and its derivative everolimus for renal cancer^{38; 39}, which led to FDA approval, and the still ongoing investigations into the antidiabetic drug metformin as a stimulant of AMPK signaling in various cancers⁴⁰, aspirin as a COX-2 inhibitor for the prevention of colorectal cancer⁴¹, and α-difluoromethylornithine (DFMO) as an inhibitor of ornithine decarboxylase in neuroblastoma^{42; 43}. In all of these cases the pharmacological target protein or pathway is the same in the original and the repurposed indication (Figure 1A). Notably though, small molecule drugs, even clinically approved ones, can have widely varying target selectivity profiles displaying a range of unintended non-canonical targets⁴⁴. This is the case even for drugs that may fall into the category “targeted therapies”^45–49, a term which suggests high precision. Non-canonical or “off-” targets are often regarded as a liability and are usually considered to be responsible for toxic side effects. There are many documented cases where this is indeed true and some undesired off-targets, so-called “anti-targets” like the ion channel protein hERG that is associated with cardiac toxicity, are routinely screened against in early drug development. However, inhibition of many off-targets is biologically neutral and does not cause meaningful effects as even multi-targeted drugs can be relatively safe depending primarily on what particular off-targets are inhibited and in what biological context. Moreover, in some curious cases inhibition of unintended targets may elicit unexpected beneficial effects in a different disease context⁴⁴, and thus provide a rationale for drug repurposing (Figure 1B). What makes this aspect particularly interesting is the identification of an unanticipated opportunity to elucidate previously hidden biological disease mechanisms. It is these cases of off-target, rather than on-target, based drug repurposing that we would like to discuss in this review, as they, in many cases, spark or at least contribute to these discoveries. While the definitions of “on-target” and “off-target” are straightforward enough (Box 1), a closer look in this context is necessary. For instance, can the receptor tyrosine kinase (RTK) REarranged during Transfection (RET) be considered an off-target of the multi-targeted drug cabozantinib? Similarly, can KIT be considered an off-target of sunitinib? Strictly speaking, both cases clearly fulfill the offered definition as cabozantinib was developed with the intention to inhibit MET and VEGFR-2, not RET, and sunitinib was developed as an inhibitor of VEGFR and PDGFR RTKs, not KIT. However, as most new drug candidates are usually screened against a broader panel of proteins to determine target selectivity, RET and KIT were early on identified to be off-targets of these two drugs^50–52. While this is true for many other proteins, too, the fact that both RET and KIT are inhibited potently enough by these drugs and that at the time of the first publication of cabozantinib and sunitinib, both RTKs were already associated with specific diseases, such as thyroid cancer and gastrointestinal stromal tumors (GIST)^{53; 54}, respectively, the therapeutic opportunities of inhibiting these nominal off-targets was immediately recognized and ultimately led to successful repurposing for these tumor types^55–58. Thus, there is a temporal element to assigning this designation and RET and KIT are, with some justification, not widely perceived as off-targets of cabozantinib and sunitinib. For the purpose of this review we will therefore exclude cases like these where the off-target repurposing potential is recognized immediately within the first report of a compound, primarily because for this to be the case the underlying biology has to be sufficiently understood a priori. We will instead focus our attention on those drug-off-target relationships, for which repurposing opportunities (and thus the underlying biology) were initially non-obvious and which were discovered only in the course of later studies. Specifically, this means that while we consider ROS1 as a crizotinib off-target that led to successful repurposing in non-small cell lung cancer (NSCLC) with oncogenic ROS1-fusions (see below), the same does not apply any more to the next-generation TKI lorlatinib as, due to the success of crizotinib, this compound was developed with ROS1 as a secondary target in mind^{59; 60}.

Figure 1. On-target and off-target based drug repurposing.

A. Cartoon illustrating the repurposing of drugs (A; B) for different cancers based on their intended targets A and B playing a functional role in cancer. The initial indication may be another cancer or a cancer-unrelated disease. B. Cartoon illustrating the repurposing of a multi-targeted drug (C) for a specific type of cancer based on inhibition of an off-target (G) that plays a functional role in that cancer. The initial indication may be another cancer or a cancer-unrelated disease.

Box 1. Glossary.

Drug repurposing/drug repositioning - the process of finding new therapeutic applications of existing or investigational drugs beyond their original medical indication

Drug promiscuity - binding of a given drug to multiple targets

Cognate target(s) - intended primary or canonical pharmacological target(s). Inhibition causes on-target effects.

Off-target - unintended drug targets; also referred to as “non-canonical” target

Anti-target - an off-target that is not just unintended, but also undesired as its engagement will produce detrimental effects (e.g. toxicity; reduction of efficacy)

Polypharmacology - binding of a given drug to, and functional modulation of, two or more targets that mechanistically contribute to the overall therapeutic effects within a specific indication

In some cases, the biological activity of a drug is not the result of inhibiting only one, but two or more targets, which may be intended targets, off-targets or a combination of both. This phenomenon is referred to as “polypharmacology”, which was initially defined as the binding of a compound to multiple targets⁶¹. However, there have been significant advancements in the field since then, suggesting that, while some drugs are more selective than others, essentially every drug binds more than one target and the term is accordingly used currently in different ways. We would therefore like to define it here more narrowly as the binding of a drug to and functional modulation of two or more targets that are mechanistically relevant for the overall biological or therapeutic drug effect in a defined disease (Box 1). Importantly, in this definition “polypharmacology” is not equivalent to “multi-targeted” as it also requires a mechanistic contribution of multiple targets. Traditionally, drug discovery pursues highly selective and potent drugs for one specific target as such high-affinity drugs are considered to maximize intended target inhibition and minimize unwanted off-target binding that may lead to side effects. This “one disease - one target - one drug” model has been successful in the control of cancers that are driven by a strong oncogenic target (Figure 2A). However, inhibition of a single target is often not sufficient to generate optimal therapeutic benefit where the disease displays polygenicity (e.g. cancer, psychiatric diseases) or involves complex biological signaling networks and feedback loops^62–67. For instance, RTKs usually share the same canonical downstream signaling pathways and inhibition of one RTK driver can be compensated for by activation of a different RTK. Accordingly, EGFR-mediated bypass signaling confers resistance to ALK or ROS1 inhibition by crizotinib in fusion-kinase driven NSCLC^68–70, and MET signaling can counteract EGFR inhibition by gefitinib in EGFR-mutant NSCLC⁷¹. Similarly, feedback activation of EGFR in BRAF-mutant colorectal cancer causes primary resistance to the BRAF inhibitor vemurafenib⁷². In the case of the PI3K–AKT–mTOR pathway, it is noteworthy that inhibition of mTORC1, for instance by everolimus, or knockdown of its downstream target S6K releases a negative feedback loop and paradoxically leads to upregulation of AKT signaling via e.g. mTORC2 activation^73–75. Therefore, treating such complex or “smart cancers” requires similarly complex, “smart” therapeutic approaches that target the extended oncogenic signaling network^{61; 76; 77}. This cannot be accomplished by inhibiting a single protein, but cumulative and even partial effects of drugs on multiple targets, including also less potent ones, can exceed that of single target engagement and thus overcome primary and acquired drug resistance⁶⁴. Multi-targeted approaches, for instance by inhibition of IGF1R, HDACs or AURKA^{78; 79}, can also suppress drug-tolerant persister cells, which survive single target inhibition and from which outright drug-resistant cell populations can arise over time^{80; 81}. Moreover, multi-targeted drugs can be, maybe unintuitively, less toxic than single-targeted drugs as has been observed with non-selective COX-1/COX-2 inhibitors (e.g. naproxen) that compared to selective COX-2 inhibitors (e.g. rofecoxib) are associated with lower cardiovascular risk^{82; 83}. This kind of multi-targeted or network pharmacology may be achieved by developing either drug combinations for different relevant targets or single drugs with polypharmacology mechanisms of action (MoAs) (Figure 2B). Drug combinations consisting of two or more targeted agents can represent an effective strategy in improving the therapeutic response of patients, reducing the likelihood of drug resistance and prolonging remission^84–86. Despite the encouraging success of many drug combinations, clinical translation can be limited by various aspects. For instance, it can be difficult to achieve sufficient target tissue concentrations simultaneously for two or more drugs. In addition, establishment of formulations and dosing schedules is more challenging and can negatively impact patient compliance. There may also be additional regulatory, patent or business strategy hurdles that would have to be overcome, particularly if drugs are developed by different pharmaceutical companies and if they are at different stages of clinical development (different patent life, sharing of profits and investments into trials). Finally, one of the most common issues is that of added toxicity, which may be dose-limiting for one or more drugs within a combination and thus limit the efficacy of the entire drug combination. This is especially a concern for drugs with overlapping toxicity profiles. At the same time, considering that the attempt to engage two or more different targets with one drug requires some structural compromises, selectivity may be harder to achieve for a polypharmacology drug compared to a drug designed towards inhibiting a single target. This in turn may increase the probability to inhibit anti-targets and elicit toxicity. However, it is also conceivable that two single-targeted drugs together can be less selective than a single polypharmacology drug. In summary, although rational design of polypharmacology drugs and avoiding increased toxicity is technically challenging and in some cases may actually not be achievable, single polypharmacology drugs may be more advantageous than drug combinations in many cases, as reviewed comprehensively also elsewhere^{61; 87; 88}. Interestingly in the context of repurposing, there are not only some powerful examples where drug polypharmacology has been intentionally designed, but also some recent cases where drugs may be repurposed for specific cancers due to previously unknown and unanticipated polypharmacology mechanisms, which - sometimes exclusively - involved several off-targets and unveiled novel network vulnerabilities of these cancer cells (Box 2).

Figure 2. Single target and network pharmacology strategies for cancer therapy.

A. Monotherapy for targeting a single, strong oncogene driver leading to cancer cell death. B. Network-based targeting strategies for cancers and inhibition of oncogenic pathways driven by multiple oncoproteins of varying strength (arrow width) can be implemented either by combination therapy with more than one drug (upper panel) or by a single drug with polypharmacology mechanism of action against more than one cancer-relevant target (lower panel).

Box 2. Polypharmacology Drugs.

Successful design of polypharmacology drugs [on-target]

1. Dasatinib was developed as a dual ABL/SRC inhibitor²⁴¹ as, due to structural similarities between these kinases, SRC inhibitors could potentially inhibit imatinib-resistant BCR-ABL mutants in CML³⁴¹. In addition, BCR-ABL can activate LYN and other SFKs³⁴², which have been implicated in imatinib resistance in CML and BCR-ABL-positive ALL^{343; 344}. Similar rationales led to development of bosutinib (approved by FDA and EMA) and bafetinib^345–347.

2. Sunitinib is a multikinase inhibitor approved by the FDA and/or EMA for GIST, renal cell carcinoma (RCC) and pancreatic cancer. Independent of targeting specific oncoproteins (e.g. mutant KIT in GIST), sunitinib was designed for dual inhibition of VEGFR and PDGFR RTKs⁵⁰, which are critical for tumor neoangiogenesis and combined inhibition of which in the tumor microenvironment produces potent antitumor effects in various cancers³⁴⁸. Similarly, sorafenib inhibits multiple kinases, including VEGFRs, PDGFRβ, and exhibits potent antiangiogenic and antitumor effects, although it was primarily designed to target RAF1^{349; 350}. Sorafenib is approved for RCC, hepatocellular carcinoma, and thyroid cancer.

3. Using iterative chemical synthesis, X-ray crystallography and kinase target profiling, Apsel et al. designed several polypharmacology compounds. PP121 inhibits multiple tyrosine and lipid kinases, such as ABL, mTOR and PI3K, and can override BCR-ABL^T315I-mediated imatinib resistance in CML cells³⁵¹.

4. Fusing two functional pharmacophores generated bispecific hybrid molecules, some of which are in preclinical or clinical tests, such as the HDAC/EGFR/HER2 inhibitor CUDC-101^{352; 353} and the HDAC/PI3K/AKT inhibitor CUDC-907³⁵⁴, and the dual HDAC/JAK inhibitor EY3238³⁵⁵. Whereas these compounds combine HDAC and kinase inhibitor scaffolds, the former can also be integrated with inducers of reactive oxygen species³⁵⁶, topoisomerase inhibitors³⁵⁷, or BET inhibitors³⁵⁸. Serendipitous off-target inhibition of BET proteins has also been reported for several kinase inhibitors, such as the clinically advanced JAK2 inhibitor fedratinib^{359; 360}, suggesting a possibility to design polypharmacology drugs with selectivity across the kinase and BET protein families³⁶¹.

Repurposing opportunities due to novel polypharmacology-based mechanisms [off-target]

1. The ALK inhibitor ceritinib is approved for crizotinib-resistant, EML4-ALK-positive NSCLC^{303; 305}. Independently, ceritinib was found to display moderate cellular efficacy also in EML4-ALK-negative NSCLC cells. Integrated functional proteomics and network analysis elucidated its polypharmacology in these cells through the off-targets RSK1/2, IGF1R and FAK1. Identification of the downstream effector YB1, which mediates taxol resistance, led to the rational design of synergistic combination with taxanes, detection of a potential mechanistic biomarker and clinical translation¹³¹.

2. Midostaurin, originally characterized as a PKC inhibitor and subsequently found to inhibit many other kinases, such as FLT3 and KIT^{179; 180}, was approved for FLT3-mutated AML and SM. Using a similar approach as for ceritinib (see above), Ctortecka et al. found that midostaurin’s cellular potency in NSCLC cells resulted from polypharmacology involving its off-targets TBK1, PDPK1, and AURKA. This led to identification of midostaurin synergy in these cells with an inhibitor of PLK1, a converging downstream signaling protein of AURKA and TBK1¹³⁰.

3. Foretinib and its structural analogue cabozantinib were designed as multi-kinase inhibitors, specifically for MET and VEGFRs, which cooperate to promote tumor angiogenesis^{51; 293; 362}. While cabozantinib received regulatory approval for thyroid cancer and RCC, clinical development of foretinib was subsequently discontinued. However, a recent study in NSCLC observed distinct cellular activities of these almost identical drugs. A systems chemical biology approach identified foretinib’s MoA to involve multiple off-targets, namely AURKB, MEK1/2 and FER, which were also targeted by cabozantinib, but less potently¹³³. Because AURKB is a relevant target in MYC-amplified SCLC³⁶³, this led to the identification of foretinib synergy with a more potent AURKB inhibitor in MYC-amplified SCLC.

Having introduced the rationale for drug repurposing and its potential specifically with regard to clinical translation and discovery of new cancer biology arising from repurposing approaches that are based on engaging one or more drug off-targets, we will in the following discuss strategies and methods, both computational and experimental, that are employed to identify new repurposing opportunities. We will furthermore illustrate several cases of off-target based repurposing that have been translated into clinical studies for various cancers or even received regulatory approval and finally review those factors, which have enabled or prevented the clinical success of these endeavors and which therefore will be informative for future off-target based repurposing efforts.

2. Current strategies towards drug repurposing

Before we turn our attention to specific cases of off-target based repurposing that have been translated into clinical studies, it is worthwhile discussing the different strategies developed to identify drug repurposing opportunities, which can be roughly divided into computational and experimental approaches.

Computational Approaches

There is a variety of computational or in silico approaches that have been widely used in the exploration of drug repurposing opportunities, starting from drug target identification, prediction of new indications, to mechanistic investigation in a given indication, which have been excellently reviewed in several recent articles and will be therefore covered here only in brief^{25; 32; 89}. In silico methods exhibit many advantages over experimental approaches with reduced requirements regarding cost, time and labor. These approaches are heavily driven by large quantities of data, such as various omics data sets (e.g. gene expression), the chemical structure space, protein interaction databases, electronic health records, the scientific literature, etc. Powerful examples include the mining of side effect profiles of marketed drugs for potentially shared targets and subsequent target-informed repurposing opportunities⁹⁰ or harnessing gene expression signature-based Connectivity Map approaches to identify cellular targets and activities of existing compounds in new disease contexts^91–93. Other systems biology approaches towards interrogating complex datasets and identifying new drug indications use for instance ‘network pharmacology’ that harnesses drug-gene or drug-protein correlations^{61; 76; 77; 94–96}. The DEMAND hybrid method combines gene expression analysis with network perturbation upon drug treatment to elucidate drug-target interactions⁹⁷. The similarity ensemble approach (SEA) on the other hand focusses on the chemical structures of small molecule drugs or ligands based on the hypothesis that ligands with similar structures are likely to have similar targets^{98; 99}. Application of this concept has been shown to successfully identify multiple new drug-target interactions, which can lead to rapid repurposing if a new target is associated with a specific disease. In contrast, the Repurposing Drugs in Oncology (ReDO) project is pursuing a literature-based approach to find licensed non-cancer drugs, which may also have utility in cancer based on prior reports of anticancer activity or indirect evidence based on activity against cancer-related pathways¹⁰⁰. Finally, over recent years much effort by academia and the pharmaceutical industry has gone into artificial intelligence (AI) techniques with the goal to improve the performance and extend the utility of computational methods and mine the inventory of either existing/known drugs or discontinued drug candidates to discover repurposing opportunities including new indications in cancer (Box 3)^{101; 102}.

Box 3. Artificial intelligence (AI) approaches in drug repurposing.

Al-Ali et al. used support vector machines (SVM), a pattern recognition algorithm, and found some kinase inhibitors to exhibit multi-targeted mechanisms and promote neurite outgrowth suggesting the potential for the treatment of neurodegenerative diseases³⁶⁴. Combining the SVM approach with cell-based compound screening allowed several cancer-selective kinase targets to be identified and subsequently functionally validated in triple-negative breast cancer cell lines³⁶⁵. As the selective target dependencies observed with this method are directly correlated with functional compound screening, the identified targets are inherently actionable thereby potentially providing a short path towards repurposing of hit compounds for the queried disease.

The clinical outcome search space (COSS) platform for drug repositioning is composed of literature search, analysis, and data visualization modules that can generate a multi-level, mechanism-based rationale to support drug-disease associations^{366; 367}. Applying this method, BVA201, an anti-depressant approved in Russia, is being repurposed to treat multiple sclerosis and the antibiotic BVA501 was found to display favorable anticancer activities in glioblastoma, melanoma, and thyroid cancer.

Lim et al. employed an approach termed REMAP, which can perform large-scale off-target predictions using chemical and protein space and their interactome. They proposed by comparison with known cancer drugs that seven cancer-unrelated drugs have repurposing potential for cancer, most of which were supported by experimental evidence³⁶⁸.

Another strategy applies proteome-wide docking of small molecule ligands to available protein X-ray crystal structures in order to identify new drug-target relationships and repurposing opportunities. For instance, application of this approach suggested that the antiviral drug nelfinavir exhibits potential as an antifibrosis compound through inhibition of TGFβRI³⁶⁹. AI in the context of drug repurposing is a rapidly developing, albeit still new, area and it will be interesting to follow its contributions to this field over the next years.

Experimental Approaches

Although targeted approaches have been successful in specific cases, such as imatinib repurposing for gastrointestinal stromal tumors (GIST) (see below), phenotypic screening has been widely recognized as a powerful discovery tool^{103; 104}, which can also drive drug repurposing approaches. Phenotypic screening has been successfully performed with a variety of model systems. In the context of cancer, readouts usually include cancer cell viability, proliferation or death, tumor growth, or more specific phenotypes, which are sometimes preferred due to facilitated mechanistic follow up¹⁰⁵, such as using specific cancer-associated isogenic cell line pairs¹⁰⁶ or processes like cytokinesis¹⁰⁷. Among the most popular cancer models are cancer cell lines^{108; 109}, and several landmark, large-scale cell line screening studies have revealed new drug-gene relationships and repurposing opportunities^110–117. Notably, the results of several of these high-throughput screens can be mined through public access portals, such as the Cancer Cell line Encyclopedia (CCLE)^{110; 113}, the Cancer Therapeutics Response Portal (CTRP)¹¹¹, Genomics of Drug Sensitivity in Cancer Web (GDSC)¹¹⁸, and the NCI Transcriptional Pharmacodynamics Workbench (NCI TPW)¹¹⁹. Although these studies have been highly informative, they screened relatively limited 2D cell culture models. More recent efforts have also utilized more complex 3D models and conditionally reprogrammed cell line cultures derived from patient tumor tissue for small molecule screening^120–122, the latter of which also have potential to inform individual therapy decisions¹²³. Along those lines, but economically restrictive for most researchers, Gao et al. recently reported on an HTS screen using a large panel of ~1000 patient-derived tumor xenograft mouse models, which was found to be predictive of clinical trial responses with the screened drugs¹²⁴. However, organism-based cancer screens have also been described using more accessible, specialized fly or zebrafish models^125–129. These different phenotypic screening approaches have been or can be readily used for drug repurposing, particularly when they are further supported by the identification of a potential biomarker derived from drug-gene correlations. However, this has been mostly implemented for on-target-based repurposing, as for instance for PARP1 inhibitors in Ewing’s sarcoma¹¹⁰. To query off-target-based drug effects, it is useful to build some target redundancy into the compound screening library, which will enable distinguishing on- from off-target effects, and subsequently analyze the data for unique, rather than common, drug activities, which has been described in some recent studies^130–133. Interestingly, this kind of strategy does not only successfully detect anticancer off-target effects in general, but also identifies those non-canonical anticancer activities, which are the result of complex polypharmacology mechanisms that involve multiple off-targets^{130; 131; 133}. This, in turn, emphasizes the need for subsequent unbiased target identification and deconvolution approaches that are critical for any off-target-based repurposing approach and in particular for polypharmacology-driven mechanisms.

In addition to the computational approaches discussed above, there are many experimental methods that are being used to identify drug targets in an unbiased way, also as primary screening assays for target identification and drug repurposing, which are reviewed in detail elsewhere^134–136. While gene expression-based techniques like the Connectivity Map have been shown to be highly useful for target identification, they have limited utility for dissecting off-target effects that are specific for certain cell types or polypharmacology mechanisms. Recombinant protein binding or catalytic enzyme screening assay panels can detect new drug-protein interactions and at the same time provide functional validation of these interactions^{45; 47–49}. However, they are by definition limited to defined subsets of proteins, such as kinases or bromodomain proteins, and therefore may or may not capture the target underlying any observed cellular activity. In contrast, proteomics approaches are essentially unbiased, provide broad coverage of the target space and accordingly have demonstrated high utility to identify drug off-targets. Protein microarrays were first reported in 2000¹³⁷ and have been used in drug discovery for drug target identification and validation¹³⁸, but have in recent years been largely replaced in popularity by mass spectrometry (MS)-based chemical proteomics, which also reflect the cellular context of a target, such as post-translational modifications, mutation status and protein expression level¹³⁹. A chemical proteomics approach, which is widely used for the identification of targets within enzyme families with catalytic activity, is activity-based protein profiling (ABPP)^{140; 141}. Although powerful, this assay, like screening recombinant proteins, is inherently biased towards the detection of defined target classes as ABPP probes selectively label active enzymes, such as phospholipases and kinases^{142; 143}. This also applies in a similar fashion to mixed drug affinity matrix approaches, such as kinobeads^{144; 145}, which combine several broad-spectrum kinase inhibitors to enrich much of the kinome and other ATP-binding proteins. In contrast, the recently developed cellular thermal shift assay (CETSA) can cover several thousand proteins and has been demonstrated to be able to identify drug targets and their protein binding partners in an unbiased way¹⁴⁶. Like ABPP and the kinobeads approach, CETSA does not require chemical modification of a compound of interest. However, the biochemical rules governing susceptibility of individual targets to this approach are still not completely understood and not all known drug targets are being captured. Although they require chemical modification, compound-centric approaches, which can be also modified by integrating quantitative MS for determination of dissociation constants¹⁴⁷, have been shown to have broad utility across multiple drug and target classes^148–151, which makes them well suited for unbiased off-target protein identification of phenotypic screening hits.

3. Clinical translation of off-target-based repurposing in cancer

As mentioned above, there is a large number of studies that report new drug repurposing opportunities, even if limiting this to off-target based repurposing only, and it would exceed the capacity of this article to attempt to cover them all. However, the purpose of this review is not to provide comprehensive coverage of the field, but to illustrate the concept, potential strategies, clinical potential and challenges of off-target based drug repurposing approaches in cancer. To this end, it is most informative to discuss those cases that have been translated into subsequent clinical studies as they more readily allow us to glean some insight into the factors that led to ultimate success, i.e. regulatory approval for the new indication, or in some cases failure to do so, which we expect will both be equally helpful for designing future repurposing projects. Below we are describing some of the most iconic clinical attempts of off-target based drug repurposing (Table 1). The majority of these involve kinase inhibitors, for two specific reasons: 1. Many kinases have been identified to drive various cancers and thus are considered highly relevant cancer targets. 2. Kinase inhibitors are notoriously unselective (although there are exceptions) and therefore make particularly good candidates for off-target based drug repurposing efforts. The path for non-kinase inhibitors, however, frequently stems from reports of unexpected anti-cancer activity in patients being treated for non-cancer related maladies. These cases cover a yet wider range of potential targets and are equally, if not more, interesting. We will therefore discuss examples of both, kinase inhibitors and non-kinase inhibitors, that are being evaluated for off-target repurposing in cancer.

Table 1.

Summary of clinical off-target based repurposing studies

Drug	Original Target	Original Disease	Off-Target(s)	Repurposed Disease	Clinical Status	Revenue
Crizotinib	MET	Solid Tumors	EML4-ALK	NSCLC	Approved	>$3 bio (2012–18)
			ROS1 Fusions	NSCLC	Approved
Imatinib	BCR-ABL	CML	KIT	GIST	Approved	N/A for GIST alone
Midostaurin	PKC	Solid Tumors	FLT3	FLT3+ AML	Approved	N/A
			KIT	SM & MCL	Approved
Thalidomide	Unknown	Morning Sickness	Immune Modulation (CRBN)	ENL, MM	Approved	~$4 bio (2005–18)
Lenalidomide				MM, MDS (5q), MCL	Approved	>$53 bio (2006–18)
Pomalidomide				MM	Approved	~$7 bio (2013–18)
Erlotinib	EGFR	NSCLC	JAK2	MDS, AML	Discontinued (Phase II)	-
Tozasertib	AURKA-C	Multiple Cancers	BCR-ABLT315I	CML & Ph+ ALL	Discontinued (Phase II)	-
Desipramine	SERT, NET	Depression	Various Receptors	SCLC	Discontinued (Phase II)	-
Dasatinib	SRC, ABL	CML & Ph+ ALL	DDR2	NSCLC	Ongoing	-
			BTK	CLL	Ongoing	-
Ibrutinib	BTK	MCL, CLL, WMG	EGFR	NSCLC	Ongoing	-
			ERBB2	ERBB2+BC & GEC	Ongoing	-
Axitinib	VEGFRs	RCC	BCR-ABLT315I	CML	Ongoing	-
Ponatinib	BCR-ABLT315I	CML	RET	RET+ MTC & NSCLC	Ongoing	-
Cabozantinib	MET, VEGFR2	MTC, RCC, HCC	ROS1 Fusions, NTRK Fusions	NSCLC	Ongoing	-
Ceritinib	EML4-ALK	NSCLC	ROS1,	NSCLC	Ongoing	-
			IGF1R, FAK, RSK1/2	NSCLC	Ongoing	-
			IGF1R, ACK1	Melanoma	Ongoing	-
Itraconazole	Lanosterol-14a-de-methylase	Fungal Infections	SMO	Multiple Cancers, BCC	Ongoing	-
Disulfiram	ALDH	Alcoholism	NPL4	Multiple Cancers	Ongoing	-
Chloroquine	Heme	Malaria	PPT1	Multiple Cancers	Ongoing	-
Artemisinin	Reactive Oxygen Species	Malaria	BAD	Multiple Cancers	Ongoing	-
Sulfasalazine	Unknown	IBD, RA	SPR	Cancer Pain	Ongoing	-

BC: breast cancer; BCC: basal cell carcinoma; GEC: gastroesophageal cancer; HCC: hepatocellular carcinoma; IBD: inflammatory bowel disease; MTC: medullary thyroid cancer; RA: rheumatoid arthritis; RCC: renal cell carcinoma; WMG: Waldenstrom’s macroglobulinemia; N/A: not available.

Crizotinib

Crizotinib arguably represents one of the most powerful examples of the transformative potential of off-target based drug repurposing as this TKI was originally developed as a potent inhibitor of the RTK MET^{152; 153}, which is involved in many cancer-related processes including proliferation and metastasis, but ultimately received regulatory approval by the US Food and Drug Administration (FDA) in its capacity as an inhibitor of the fusion kinase EML4-ALK. Although crizotinib’s ability to inhibit NPM-ALK fusions in anaplastic large-cell lymphoma (ALCL) cell lines had been noted, the chromosomal translocation leading to the EML4-ALK fusion oncogene in 3–7% of NSCLC and recognition of its transforming activity were only identified shortly thereafter¹⁵⁴. Further studies demonstrated that crizotinib could successfully shut down oncogenic signaling and inhibit the growth of EML4-ALK fusion-positive NSCLC cells¹⁵⁵. Subsequent phase I clinical trials focusing on EML4-ALK-rearranged lung cancer patients showed, in addition to a favorable safety profile, tumor shrinkage or stable disease for the majority of patients^{156; 157}. The impressive responses observed in these studies, which were later confirmed in a phase III study¹⁵⁸, led to the accelerated approval of crizotinib for EML4-ALK-positive NSCLC in the US in 2011 and Europe in 2012, eventually replacing conventional chemotherapy as first-line therapy.

Inspired by this success, Bergethon et al. showed that in preclinical models, as well as in a NSCLC patient, crizotinib also inhibits the ALK-related RTK ROS1¹⁵⁹, which constituted a potent off-target of crizotinib that had not been picked up in the initial reports^{152; 153}. Similar to ALK, ROS1 tyrosine kinase activity is constitutively activated as the result of chromosomal translocation events, which had been described previously to serve as oncogenic drivers in NSCLC^{155; 160}, but in contrast to EML4-ALK involve a variety of different fusion partners and affect a different, and smaller, subpopulation of NSCLC patients^{161; 162}. The finding that crizotinib can potently inhibit the growth of ROS1-rearranged NSCLC cells yielded a different phase I clinical trial specifically for patients with advanced ROS1-fusion positive NSCLC where crizotinib again showed pronounced antitumor activity¹⁶³, which led to crizotinib obtaining FDA approval as first-line therapy in 2016 also for ROS1-rearranged NSCLC. In fact, inhibition of ROS1-fusions by crizotinib has proven to be highly effective and generally produces more durable responses than most other targeted therapies even when compared to crizotinib treatment of ALK-rearranged NSCLC, which has been in the meantime replaced as first-line therapy by yet more potent second-generation ALK inhibitors, such as alectinib, initially developed to overcome ALK point mutation-mediated crizotinib resistance^{164; 165}. However, the twofold success of crizotinib repurposing in NSCLC, which was based on the ALK- and ROS1-fusion kinases as off-targets, has made a major clinical impact and has sparked a vibrant research field on targeting oncogenic fusion kinases not just in lung cancer, but in a broad range of tumor types¹⁶².

Imatinib

The development of the ABL TKI imatinib and its groundbreaking success in BCR-ABL-driven chronic myeloid leukemia (CML) has revolutionized the field of cancer research and is widely regarded as the paradigm for targeted therapy^{7; 166; 167}. Not surprisingly, therefore, the enthusiasm generated by imatinib also rapidly kindled many repurposing studies for this drug. While imatinib was first rationally developed to inhibit BCR-ABL and PDGFR^{6; 168}, it was subsequently found to be also a similarly potent inhibitor of KIT, which shares high homology with PDGFR¹⁶⁹ and for which gain-of-function mutations causing malignant transformation had shortly before been described in gastrointestinal stromal tumors (GIST)⁵⁴. After observing sustained objective responses in more than half of patients with advanced unresectable or metastatic GIST¹⁷⁰, imatinib was further tested in GIST patients with activating mutations of KIT or PDGFRA, which can also show sensitivity to imatinib¹⁷¹, where it also produced significant clinical benefit¹⁷². These studies led to regulatory approval of imatinib for the treatment of GIST in 2002. Interestingly, although the more potent second-generation BCR-ABL inhibitor nilotinib was found to be superior to imatinib in patients with CML¹⁷³, a randomized phase III trial in unresectable or metastatic GIST did not recommend use of nilotinib over imatinib as first-line treatment for GIST¹⁷⁴, illustrating that just because a drug is superior over another for inhibiting their cognate target, the same is not always true for therapeutically relevant off-targets. In addition, imatinib in combination with the anti-estrogen therapy letrozole for hormone receptor-positive metastatic breast cancer expressing KIT or PDGFRβ was well tolerated, but appeared to have limited efficacy¹⁷⁵. In contrast, clinical studies of imatinib in metastatic melanoma with KIT aberrations showed significant benefit^{176; 177}, although this has been suggested to be limited to KIT mutations rather than cases with gene amplification only¹⁷⁸. Thus, while repurposing of imatinib for cancer indications driven by its off-target KIT have been highly successful, particularly for GIST, it is also apparent that understanding the biological context of the respective off-target and its relative importance for driving specific cancers is critical and cannot be implied based on what is known in other tumor types.

Midostaurin

Midostaurin (PKC412) was originally identified as an inhibitor of protein kinase C (PKC)¹⁷⁹ and subsequently shown to inhibit other kinases including VEGFR2, PDGFR and KIT¹⁸⁰. Interestingly, development of midostaurin as a PKC or VEGFR2 inhibitor in various cancer types ultimately failed due to insufficient efficacy¹⁸¹. However, having screened a panel of small molecules, Weisberg et al. demonstrated the efficacy of midostaurin in inhibiting mutant FLT3 receptor as a novel off-target in leukemia cells¹⁸². This was based on the identification of activating mutations in the RTK FLT3, particularly internal tandem duplications (ITD) in the juxtamembrane domain, in approximately 30% of acute myeloid leukemia (AML) patients¹⁸³. This finding was followed by clinical observations showing efficacy of midostaurin in AML patients with activating FLT3 mutations, although it was recognized that combination with conventional chemotherapy may be required¹⁸⁴. Subsequently, midostaurin was confirmed to be safe and efficacious in clinical trials and in 2017 received regulatory approval for use in FLT3-mutant AML^{181; 185; 186}.

As described above, another off-target of midostaurin is KIT. D816V mutations of KIT, which are resistant to imatinib¹⁸⁷, have been found to play a crucial role in the majority of systemic mastocytosis (SM) cases, a rare hematologic neoplasm that can develop into mast cell leukemia (MCL). Midostaurin was shown to inhibit KIT D816V and block growth of KIT^D816V-transformed cells at nanomolar concentrations¹⁸⁸. This led to a multicenter phase II study of midostaurin in aggressive SM (ASM)/MCL patients in which midostaurin was safe and elicited clinically relevant and durable responses^{189; 190}. Following these studies, midostaurin was approved also for advanced SM and MCL^{181; 191}. It is interesting to note that, similar to crizotinib, midostaurin was ultimately approved for treating malignancies that were driven by off-targets, but not actually for diseases driven by its cognate target PKC, which powerfully illustrates the potential of investigating drug off-target effects. In contrast to crizotinib and imatinib, however, midostaurin’s journey took almost 20 years, which in part is the consequence of a complicated international study in AML. Unfortunately, this also limited the ability and interest of the midostaurin developing pharma company, in this case Novartis, to pursue further repurposing opportunities with midostaurin, such as following up the unexpected observation that midostaurin can potently inhibit the EGFR T790M gatekeeper mutant, but not wild-type (wt) EGFR^{48; 192; 193}, which constituted a high unmet medical need in NSCLC at that time.

Thalidomide

Thalidomide is the infamous founding member of a (re)emerging class of drugs, the history of which is dating back to 1956, when it was introduced as a sedative and hypnotic that also alleviates morning sickness in pregnant women^{194; 195}. However, just as imatinib is the paradigm for the success of targeted therapies, thalidomide can be considered the paradigm for severe drug toxicity as it exhibits dramatic teratogenic properties and led to thousands of infants that were born with limb deformations^{196; 197}, such as phocomelia, which later resulted in thalidomide withdrawal from the market¹⁹⁸. Strictly speaking, the thalidomide case does not qualify as off-target-based repurposing considering that the pharmacological target was unknown for the longest time. However, its initial indication was based on thalidomide structural and medicinal similarity with barbiturates, which implied a similar target. After thalidomide had made a serendipitous comeback in 1965 as a treatment for erythema nodosum leprosum, an immunological complication of leprosy¹⁹⁹, for which it gained FDA approval in 1998, it was also recognized that thalidomide displays potent immunomodulatory properties (hence the term ImmunoModulatory Drug or “IMiD”) by e.g. inhibiting TNFα synthesis^{200; 201}. Subsequently, a possible association of thalidomide with malformations due to inhibition of angiogenesis was identified, which was likely independent of TNFα²⁰². As multiple myeloma (MM) features enhanced neovascularization in the bone marrow, thalidomide was repurposed for the treatment of MM²⁰³, for which it showed remarkable efficacy as single agent and in combination with dexamethasone and eventually obtained regulatory approval. These successes subsequently led to the development of more potent and stable IMiDs, namely lenalidomide and pomalidomide, which can enhance the cytokine release and anti-myeloma activity of thalidomide^204–208 and are also effective in mantle cell lymphoma (MCL) and myelodysplastic syndrome (MDS)^{209; 210}. While these analogues were being developed in the clinics other studies were progressing on identifying the MoA of thalidomide and the IMiDs. Ito et al. proposed that thalidomide causes its teratogenic effects by binding to the E3 ubiquitin ligase cereblon (CRBN) in complex with damaged DNA binding protein 1 (DDB1) and Cul4A thereby inhibiting its ubiquitin ligase activity²¹¹. Further refinement of this mechanism showed that IMiD binding to CRBN leads to recruitment, ubiquitination and selective degradation of the lymphoid transcription factors IKZF1 and IKFZ3^212–214. Based on this discovery, Winter et al. designed novel thalidomide-based proteolysis-targeting chimeric molecules (PROTACs) that efficiently and selectively led to degradation of the transcriptional coactivator BRD4 or the chaperone FKBP12²¹⁵. This provided a powerful demonstration of the broad applicability and potential of the PROTAC approach²¹⁶ for the design of novel drug candidates and unique chemical probes and invigorated academia and pharmaceutical industry alike as potentially any protein of interest, which could be considered engineered indirect thalidomide off-targets, could be targeted for degradation. Even though this is a developing field and many design rules have to be still established, there is a lot of momentum and the first PROTACs are entering clinical trials already, a truly remarkable development considering thalidomide’s tragic past.

Erlotinib

Erlotinib has been developed as a specific TKI for the epidermal growth factor receptor (EGFR)²¹⁷. EGFR shows TKI-sensitive oncogenic gain-of-function mutations within the tyrosine kinase domain in about 10–15% of NSCLC patients^{218; 219}, which represents erlotinib’s primary clinical indication and which transformed the field of lung cancer therapy²²⁰. Interestingly, it has been observed that erlotinib is also an inhibitor of JAK2 V617F, the dominant mutation in polycythemia vera (PV), and inhibited PV cell growth²²¹. In addition, Boehrer et al. suggested that erlotinib exhibits antineoplastic off-target effects in AML and MDS via JAK2²²², although other mechanisms via inhibition of other targets have been proposed as well²²³. Following these findings, there have been multiple efforts to repurpose erlotinib for targeting of JAK2 in the clinics. Although erlotinib appeared to induce a significant number of responses in higher-risk MDS/AML having failed azacitidine treatment in a phase I/II study²²⁴, Komrokji et al. found despite general tolerability only modest single-agent activity of erlotinib in a phase II study in MDS²²⁵. Similarly, erlotinib as a single agent was found to have only limited clinical efficacy in patients with relapsed/refractory AML in a phase II clinical trial²²⁶. Collectively, these studies suggest that while JAK2 may constitute in general an interesting off-target of erlotinib, it may not have the required potency for producing clinically meaningful effects and/or that JAK2 may not be a strong enough driver for AML, which is in line with reports that also the potent JAK2 inhibitor ruxolitinib may only exhibit clinical efficacy in patients who developed AML from myeloproliferative neoplasms (MPN)²²⁷. Accordingly, there are no currently ongoing erlotinib off-target based clinical trials in AML or MDS.

Tozasertib

Tozasertib (VX680) is a potent and selective small-molecule inhibitor of Aurora kinases and considered to possess potential to be used in the treatment of multiple human malignancies, including colorectal cancer, lymphoma and acute leukemias²²⁸. By screening clinically relevant mutants against a panel of small molecules and X-ray crystallography, tozasertib was subsequently shown to also bind and have inhibitory activity against BCR-ABL bearing the T315I “gatekeeper” mutation^{229; 230}. This mutant confers broad cross-resistance against the first and second-generation BCR-ABL inhibitors imatinib, dasatinib and nilotinib and represents the most pressing remaining challenge in CML therapy²³¹. Early case reports showed that tozasertib is active in patients with CML and ALL with the BCR-ABL T315I mutation^{232; 233}. However, although tozasertib also showed significant activity in patients with CML or ALL expressing BCR–ABL T315I in a phase I/II trial²³⁴, Seymour et al. afterwards observed minimal efficacy of tozasertib and only at higher, intolerable doses²³⁵. In addition, there have been reports of cardiac toxicity, such as QTc interval prolongation, which together led to discontinuation of further clinical development of tozasertib for BCR-ABL T315I-mutant leukemias²³⁶.

Desipramine

Desipramine is a member of the class of tricyclic antidepressants (TCAs). The TCAs act mainly as serotonin and norepinephrine reuptake inhibitors (SNRIs) by blocking the serotonin transporter (SERT) and the norepinephrine transporter (NET) resulting in enhanced neurotransmission²³⁷. TCAs like imipramine and desipramine have been observed in several studies to show anticancer effects for instance in glioma and colorectal cancer cells^{238; 239}, although the underlying mechanisms were not well understood in these studies. Later, Jahchan et al. utilized an innovative bioinformatic drug repurposing approach to compare disease and drug-induced gene expression signatures in small cell lung cancer (SCLC)²⁴⁰. This method identified TCAs to elicit gene expression changes that were anticorrelated to SCLC suggesting TCAs might be efficient drugs to treat this cancer. Indeed, they found TCAs to exhibit in vitro and in vivo anticancer activity in SCLC, which they attributed to a polypharmacology mechanism through inhibition of the cognate targets SERT and NET, but also off-targets like certain histamine, adrenergic, acetylcholine and serotonin receptors²⁴⁰. These observations led to a subsequent phase IIa trial of desipramine in SCLC and other high-grade neuroendocrine tumors. Unfortunately, this trial had to be terminated as SCLC patients were not able to tolerate clinically relevant doses of desipramine and lower drug doses yielded no efficacy, which illustrates that drug doses that are generally safe for the original indication may be too toxic in the context of specific other diseases.

Whereas for the examples discussed above clinical studies have been completed, which allow us to evaluate the success and, in some cases, the encountered challenges, multiple off-target based repurposing attempts for cancer have been translated into clinical trials, which are still ongoing and for which the final outcome is therefore not yet known. However, as it is already a major achievement to launch a clinical trial based on one’s preclinical research, reviewing the trajectory of these projects can allow a better understanding of what aspects facilitated clinical translation. Below, we will therefore discuss several of these cases.

Dasatinib

Dasatinib, a multi-kinase inhibitor that was originally designed as a dual SRC/ABL inhibitor²⁴¹, received accelerated FDA approval in 2006 as second-line therapy for adults in all phases of CML or with Philadelphia-chromosome-positive (Ph+)-acute lymphoblastic leukemia (ALL). It has since been approved for first-line treatment in both adults and children, thus exemplifying its therapeutic safety when taken at lower doses that are sufficient to inhibit BCR-ABL and SRC-family kinases (SFKs). Having shortly thereafter found activity of dasatinib in NSCLC cells due to a combined inhibition of SFKs and its weak off-target EGFR, Haura et al. began a clinical trial to test the combination of dasatinib and the more potent EGFR inhibitor erlotinib (Figure 3A)^242–244. Two patients from this trial had partial responses; one with an EGFR mutation, which can explain the response to erlotinib-based therapy, the other without. This latter patient was instead found to carry a novel oncogenic discoidin domain receptor 2 (DDR2) kinase domain mutation²⁴⁵. Importantly, DDR2 and its closely related kinase DDR1 had previously been identified as potent dasatinib off-targets^{144; 246; 247}. However, this was the first study to identify and quantify the prevalence of DDR2 mutations in lung squamous cell cancer (LUSQ), where they represent approximately 4% of all cases²⁴⁵. In light of these discoveries, dasatinib monotherapy clinical trials in LUSQ and other DDR2-mutant cancers were initiated. Unfortunately, at the dose used (140 mg vs 100 mg for CML), limited efficacy was achieved before toxicities, such as pleural effusions and immunosuppression leading to life-threatening infections led to trial termination. These findings were well aligned with inhibition of SFKs and dasatinib’s potent off-target Bruton’s tyrosine kinase (BTK)^{248; 249}. It has now been shown that several compensatory signals are up-regulated upon dasatinib treatment in LUSQ cells, which limit the efficacy of targeting DDR2²⁵⁰. In particular, the RTKs EGFR and MET are activated suggesting that inhibition of DDR2 alone is not going to be an effective strategy. Rather, a combination of dasatinib, or a yet more potent DDR2 inhibitor^{251; 252}, and a potent MET or EGFR inhibitor (as given to the first patient with DDR2-mutant LUSQ) may be needed for successful DDR2-based repurposing of dasatinib²⁵⁰. This would be similar to what has been described in BRAF-mutant colorectal cancer where EGFR signaling was found to mediate primary vemurafenib resistance⁷². However, given that dasatinib is included in the active MATCH Screening Trial, where patients are matched to various targeted inhibitors based on their genetic profiles, e.g. DDR2 mutations, efforts to determine dasatinib efficacy are still somewhat ongoing.

Figure 3. Off-target based drug repurposing initiatives for dasatinib and ibrutinib in various cancers.

A. Repurposing of the dual SRC/ABL inhibitor, developed to treat CML, for various B cell malignancies and NSCLC due to its ability to also inhibit BTK and EGFR (weak), respectively. Combination therapy with the potent EGFR TKI erlotinib led to the identification of mutant DDR2 as a novel cancer driver in LUSQ. B. Repurposing of the irreversible BTK inhibitor ibrutinib, developed to treat different B cell malignancies, for breast cancer (with trastuzumab) and EGFR-mutant NSCLC due to its ability to also inhibit HER2 and EGFR, respectively. * indicates oncogenic mutation; ** indicates gene amplification; light bulb indicates discovery of new drug-protein interaction.

In some cases, repurposing is based on an off-target that was originally reported to cause an unwanted side effect. Interestingly, BTK, which was found to be involved in the immunosuppressive side effects of dasatinib^247–249, has been implicated as a cancer driver in, for example, diffuse large B-cell lymphoma (DLBCL) and chronic lymphocytic leukemia (CLL). In fact, the first-in-class covalent BTK inhibitor, ibrutinib (PCI-32765), was also in its infancy at this time^253–255. Interestingly, dasatinib has been shown to work in both ibrutinib-sensitive and - resistant DLBCL and CLL cells^256–259. However, some reports suggest that activity in these cells is not only due to BTK inhibition, but that also SFKs play a role. For instance, in DLBCL cells carrying the BTK-C481S ibrutinib-resistant mutation, the autophosphorylation site (Y233) of BTK-C481S remains phosphorylated upon dasatinib treatment, while downstream FYN inhibition mediates dasatinib activity²⁵⁷. This suggests a polypharmacology effect in wild-type BTK cells, where inhibition of both SFKs and BTK contributes to the therapeutic activity of dasatinib. A clinical trial using dasatinib in CLL where SFKs, BTK and the related TEC kinase are being evaluated as biomarkers for response has been completed, but results have not yet been reported. These observations with dasatinib are powerful examples of how off-targets can inform new research directions, identify responder patients, lead to the elucidation of new cancer biology mechanisms and even turn initially toxic off-target effects into therapeutic approaches for other malignancies in a similar way as seen with thalidomide. However, they also illustrate that such repurposing strategies, just as much as conventional targeting approaches, may be limited due to adaptive signaling that causes resistance to single-agent therapies.

Ibrutinib

Ibrutinib represents an exciting “rags-to-riches” development story. Originally designed as a tool compound to study BTK biology and to develop potent, non-covalent, BTK inhibitors, ibrutinib was purchased from Celera Genomics in 2006 by Pharmacyclics for $3 million (cash and stock). Just 5 years later, in 2011, Johnson and Johnson paid Pharmacyclics almost $1 billion (upfront and milestones) to co-develop ibrutinib for CLL and non-Hodgkin’s lymphoma²⁶⁰. Having received FDA approval for mantle cell lymphoma (2013), CLL (2014) and Waldenstrom’s macroglobulinemia (2015), AbbVie outbid Johnson and Johnson to buy Pharmacyclics for $21 billion and its success as a BTK inhibitor continues. In addition, inhibition of ibrutinib’s off-targets EGFR and ERBB2/HER2 was indicated from kinase assays in the original report, but it wasn’t until 2014 that ibrutinib was recognized as a potential therapeutic option for EGFR-mutant NSCLC or ERBB2-positive breast cancer (Figure 3B)^{254; 261; 262}. In terms of EGFR-driven cancer, it is uncertain, if ibrutinib can compete with the potent and mutant-selective EGFR inhibitor osimertinib (AZD9291), which in 2018 received FDA approval for first-line treatment of EGFR-mutant NSCLC; however, there remains at least one clinical trial for ibrutinib in previously treated EGFR-mutant NSCLC, which is still active^263–266. It could be that combining ibrutinib with other compounds, such as the gold complex auranofin, a thioredoxin reductase 1 inhibitor that inhibits AKT/mTOR signaling in lung cancer cells, will be necessary for effective repurposing in EGFR-mutant NSCLC²⁶⁷. Interestingly, ibrutinib’s ability to inhibit myeloid-derived suppressor cell and mast cell function via BTK and the related interleukin-2-inducible T-cell kinase (ITK) could support the anti-tumorigenic potential of ibrutinib as an ERBB2/HER2 inhibitor^268–270, which would represent an intriguing polypharmacology mechanism involving targets in tumor microenvironment cells as well as cancer cells, and clinical trials are currently recruiting for ERBB2-driven gastroesophageal cancer and metastatic breast cancer. Finally, using ibrutinib as an ERBB4 inhibitor in ERBB4-expressing cancers may be possible and ERBB4 is being monitored as one of several possible biomarkers in a phase II study for ibrutinib in refractory metastatic cutaneous melanoma^{271; 272}. While the future looks bright for ibrutinib, paying attention to protocols for handling side-effects coming from years of use in hematologic malignancies will certainly benefit its further off-target repurposing potential²⁷³.

Axitinib

Axitinib, an angiogenesis inhibitor whose initial targets were reported to be the VEGFR kinases, PDGFR-β, KIT and FGFR-1, was approved in 2012 for advanced renal cell carcinoma^{274; 275}. While screening 252 clinically relevant drugs for BCR-ABL T315I-gatekeeper inhibitors, Pemovska et al. noticed that axitinib, as opposed to other VEGFR inhibitors, showed strong potency against the assayed cells, suggesting an off-target effect²⁷⁶. Indeed, they found that while axitinib binds the T315I mutant in a range similar to VEGFR2, and at a much lower concentration than parental BCR-ABL, it binds in a very different conformation to ABL than VEGFR2 and differently than other ABL inhibitors, including ponatinib, the only approved BCR-ABL^T315I inhibitor. This suggested a different wt/mutant-binding profile between the two drugs²⁷⁶. It has since been shown that axitinib has a very narrow BCR-ABL mutant binding profile and often at concentrations that are clinically unachievable²⁷⁷. However, it could be an option for patients who cannot tolerate ponatinib treatment, which is important given the propensity for patients to develop resistance via this important mechanism. Furthermore, a clinical phase I/II trial in combination with bosutinib for CML patients is currently ongoing, which is particularly interesting as a previous study had found that bosutinib in combination with the Aurora kinase and BCR-ABL^T315I inhibitor danusertib exhibited strong synergy in BCR-ABL^T315I-positive CML cells due to additional effects of bosutinib on its off-target MEK1/2²⁷⁸.

Ponatinib

Imatinib, as has been mentioned, has revolutionized personalized medicine and specifically CML therapy. Still, disease progression via BCR-ABL mutation rendering the fusion kinase resistant to inhibition remains a major problem even with the development of second-line BCR-ABL inhibitors such as nilotinib and dasatinib. Ponatinib was therefore developed to inhibit a panel of BCR-ABL mutations, but in particular the highly resistant T315I-gatekeeper mutation^{279; 280}. From the beginning, it was understood that ponatinib was a multi-kinase inhibitor with activity also against the therapeutically relevant VEGFR, FGFR, PDGFR and FLT3 kinases^279–283. Given that ponatinib was designed specifically to target gatekeeper mutations, which are commonly observed upon initial drug therapy in many kinases, such as EGFR, NPM-ALK and EML4-ALK^{164; 284–286}, de Falco et al. and Mologni et al. both tested and found that ponatinib was a potent inhibitor of the RET gatekeeper V804M and V804L mutants as well as wt RET^{287; 288}. While sporadically found in lung and colon cancer, RET-fusions and germline mutations are the main oncogenic kinases in different thyroid cancer subtypes²⁸⁹ and ponatinib is in phase II clinical trials for multiple tumors, including RET mutation-positive thyroid cancer. Results from a recently completed phase II trial of ponatinib in RET-fusion-positive NSCLC, which was also found to be sensitive to ponatinib²⁹⁰, are not available yet. A recent review summarizes nicely the broad application of ponatinib based on both on- and off-target driven diseases²⁹¹.

Cabozantinib

Cabozantinib is approved for treating thyroid cancer, renal cell carcinoma, and hepatocellular carcinoma. It has been designed as an inhibitor of mainly MET and VEGFR2, but also RET, AXL, KIT, FLT3, and TIE2 among others were originally recognized as important therapeutic targets of cabozantinib^{51; 52; 58}. Interestingly, VEGFR2 and MET inhibition have been suggested to work together to elicit cabozantinib’s therapeutic effect, as tumors have been shown to compensate for VEGFR inhibition through MET signaling, making cabozantinib an excellent example of effective drug polypharmacology^{51; 52; 292; 293}. More recently, cabozantinib has been recognized also for its ability to inhibit ROS1- and NTRK-fusion kinases that have become resistant to the first-line inhibitor crizotinib or larotrectinib, respectively^{51; 52; 292–297}. While ROS1-fusions are present in 1–2% of NSCLC¹⁵⁹, NTRK-fusions have been identified at low rates in several cancers^{298; 299}. However, some cancers, such as mammary analogue secretory carcinomas are 100% positive for NTRK-fusions, thus illustrating the importance of inhibiting these kinases^{162; 300}. Cabozantinib crosses the blood-brain barrier and case studies have shown it to be effective in patients with ROS1-positive, crizotinib- (and ceritinib-) resistant NSCLC primary tumors and brain metastases^{301; 302}. At least two clinical trials are currently ongoing for cabozantinib in ROS1- or NTRK-fusion positive NSCLC and based on these off-targets cabozantinib is one of few promising contenders for second-line therapy in these tumors.

Ceritinib

Ceritinib belongs to the growing family of FDA-approved ALK inhibitors that includes crizotinib, alectinib, brigatinib and lorlatinib and was the first ALK inhibitor approved for crizotinib-resistant EML4-ALK-positive NSCLC^303–305. As with cabozantinib, ceritinib has also been found to inhibit ROS1 as a major off-target^{60; 306}. Accordingly, ceritinib is in several clinical trials for ROS1-fusion-positive NSCLC and first results suggest potent clinical efficacy of ceritinib also for this indication³⁰⁷.

In addition to ROS1, other off-targets of ceritinib have been identified. For instance, IGF1R, reported early on³⁰⁵, has been shown to contribute to the overall therapeutic effect of ceritinib in EML4-ALK-positive NSCLC cells³⁰⁸. Different studies also found that IGF1R inhibition is responsible for ceritinib activity in rhabdomyosarcoma and hepatocellular carcinoma^{309; 310}. Through phenotypic screening followed by an integrated functional proteomics approach, another study found that inhibition of IGF1R together with the additional off-targets FAK and RSK1/2 contributes to a more complex polypharmacology mechanism of ceritinib leading to its activity in EML4-ALK-negative NSCLC cells¹³¹. It was also observed that these targets modulate a downstream pathway involved in paclitaxel resistance and accordingly that combination with taxanes can lead to pronounced synergy. This was particularly the case in cells with high FAK autophosphorylation, which led to a phase I/II clinical trial for EML4-ALK-negative, EGFR-wild-type NSCLC. Finally, IGF1R and/or ACK1 appear to contribute to ceritinib activity in BRAF/NRAS-wild-type melanoma, in which ceritinib in combination with the MEK inhibitor trametinib showed pronounced in vitro and in vivo anti-tumor activity³¹¹. A phase II clinical trial of single-agent ceritinib in unresectable, refractory melanoma has been initiated based on this study although ceritinib may have greater potential in this disease when combined with MEK inhibitors.

Itraconazole

Itraconazole is an anti-fungal drug used in the treatment of systemic and superficial fungal infections that inhibits lanosterol-14α-demethylase, an enzyme involved in fungal membrane synthesis^{312; 313}. While itraconazole’s inhibitory potential is much higher for fungal 14α-demethylase, it also inhibits human 14α-demethylase (involved in the biosynthesis of cholesterol) and this is believed to be involved in its reported anti-angiogenesis activity³¹⁴. As an anticancer off-target effect, on the other hand, itraconazole has been shown through a phenotypic screen to block hedgehog pathway signaling by inhibiting smoothened (SMO)³¹⁵. This activity seems to be different than that shown by other SMO inhibitors such as cyclopamine, suggesting a possibility for designing hedgehog inhibitor combination strategies although synergy with cisplatin has also been described^{316; 317}. Multiple clinical studies of itraconazole are currently ongoing in various malignancies and there are several reviews that nicely summarize the current status of itraconazole repurposing in cancer^{312; 318; 319}. Notably, an open-label, exploratory phase II trial of oral itraconazole has reported activity for the treatment of basal cell carcinoma³²⁰, but there have been also reports with significant, particularly cardiovascular, toxicities, which may limit itraconazole’s utility in malignant diseases^320–322.

Disulfiram

Disulfiram was originally approved in the 1950’s to treat alcoholism. Inhibition of aldehyde dehydrogenase by disulfiram leads to accumulation of acetaldehyde, causing severe “hangover” symptoms upon alcohol ingestion³²³. Many theories have been suggested for the anticancer effects seen in patients treated with disulfiram, but in particular it has been proposed to be a proteasome inhibitor through a cell-based high-content screen^{323; 324}. Along these lines, Skrott et al. showed that the active metabolite of disulfiram, the ditiocarb–copper complex, is binding and inhibiting NPL4, a p97 segregase adaptor, which functions upstream of the proteasome, thus leading to several overlapping, but also unique, properties compared to “true” proteasome inhibitors³²⁵. Disulfiram is currently being pursued for its anticancer properties in several active clinical trials and knowing its true target is expected to assist in developing mechanistic biomarkers for patient selection.

Chloroquine

In use since around 1940 to treat malaria, chloroquine was first tested as an anticancer agent for glioblastoma multiforme (GBM) in the late 1990’s. It has been in several clinical trials since then for various cancers as a single agent and in combination with chemotherapeutic drugs^{326; 327}. Different from its use as an antimalarial, which is based on its ability to bind heme inside the parasitic food processing system and thereby blocking hemoglobin digestion, chloroquine has been used as an autophagy inhibitor in the context of cancer^{326; 327}. Until recently, however, the actual molecular target of chloroquine and its related drugs, such as hydroxychloroquine and dimeric chloroquine analogues, was unknown. In 2017, using a chemical proteomics approach Rebecca et al. reported that palmitoyl-protein thioesterase 1 (PPT1), a protein involved in lysosomal degradation, may be that (off-)target^{328; 329}. Importantly, they found that dimeric chloroquine analogues have better lysosomal inhibitory properties, blocking autophagy, but also inhibiting lysosomal-driven mTORC1 signaling^328–330. It will be interesting to see how knowledge of the direct cancer target of chloroquines will be utilized in further development of these compounds for potentially biomarker-guided clinical trials.

Artemisinin

Artemisinin is a natural product isolated from the sweet wormwood plant, which has been used in Chinese traditional medicine to treat malaria for thousands of years. Artemisinin as the active compound was discovered in the early 1970’s by Tu Youyou, who was co-awarded the 2015 Nobel Prize in Medicine for this discovery^{331; 332}. Most people believe that the antimalarial effects of artemisinins are due to the production of reactive oxygen species (ROS) derived from the degradation of the endoperoxide bridge, and this theory has also been put forward, at least partly, for artemisinin’s antitumor properties^{331; 333}. Challenging this hypothesis, Gotsbacher et al. recently used a reverse chemical proteomics screen and identified a direct molecular target for artemisinin, namely the BCL-2 antagonist of cell death promoter BAD³³⁴. Since BAD is sequestered and inhibited upon phosphorylation, the ability of artemisinin to reduce phosphorylation of BAD, thereby increasing its pro-apoptotic activity, may allow for combination with other drugs that target the anti-apoptotic program prevalent in cancer cells³³⁴. Several clinical trials are currently active pursuing artemisinin (artesunate) activity in various cancers, such as colorectal cancer.

Sulfasalazine

Sulfasalazine is an anti-inflammatory drug used in inflammatory bowel disease and rheumatoid arthritis. In 2011, Chidley et al. discovered through a yeast three-hybrid target screen that the molecular target inhibited by sulfasalazine is sepiapterin reductase (SPR)^{335; 336}. This protein is involved in the synthesis of the cofactor tetrahydrobiopterin (BH4), which is increased, for instance, in inflammation and pain^{337; 338}. This pathway is also upregulated in neuroblastoma³³⁹. Although SPR is most likely also driving activity in the original indication and therefore strictly speaking can probably not be considered a sulfasalazine “off-target”, knowledge of the specific target may lead to more appropriate, and therefore more successful, development of sulfasalazine in cancer clinical trials, which are currently ongoing for instance in the pain management associated with breast cancer.

4. Conclusions

Having discussed various off-target-based repurposing examples in detail, it is worth extracting some specific aspects that can inform future repurposing efforts by highlighting opportunities and helping to avoid pitfalls. So, what can we learn from these cases? First, it should be noted that despite the stigma associated with off-targets and the initial cringe with which many scientists, specifically in the field of drug discovery, react, the end results from such efforts can be far from incremental, marginal or detrimental. Rather, they can have a profound positive impact, obviously on the lives of cancer patients who may not have otherwise effective therapies available, but also conceptually. Successful off-target-based repurposing can lead to elucidation of new mechanisms in cancer biology, inspire entirely new research directions and drug development programs or rekindle interest in existing fields by giving them a new spin²⁵. This is best illustrated by crizotinib’s success in ALK- and ROS1-fusion positive NSCLC, which given the significant survival benefits achieved by crizotinib allowed the field to tackle acquired drug resistance in these tumors as a means to further prolong patient survival; a challenge that was in turn successfully met by the scientific and pharmaceutical community and in record time led to the development of several next-generation ALK and ROS1 inhibitors that have received regulatory approval, namely ceritinib, alectinib, brigatinib and lorlatinib for ALK-fusion positive, and most recently entrectinib for ROS1-fusion positive NSCLC. It is fairly safe to say that these accomplishments furthermore inspired new drug development efforts for other oncogenic fusion kinases, such as those involving RET and NTRK family kinases, thereby leading to the recent approval of larotrectinib and entrectinib for NTRK-fusion driven cancers. Likewise, the elucidation of thalidomide’s MoA as a consequence of major repurposing efforts and its employment as a targeted protein degrader have fueled the PROTAC field and led to nothing short of a surge in drug development programs that exploit this concept. It is also noteworthy that some drugs, for instance crizotinib and midostaurin, were repurposed and received approval due to their abilities to inhibit particular off-targets, not their cognate targets, thus highlighting the potential relevance of specifically elucidating off-target effects^{62; 87}. Moreover, as demonstrated by dasatinib (BTK) and thalidomide (CRBN), elucidating the mechanisms underlying (off-target based) toxic drug side effects can lead to new repurposing initiatives^{25; 87}.

These examples demonstrate that it is important to develop a detailed understanding of the targets and cellular MoAs of drugs⁶². While this is essentially always true, for off-target based repurposing approaches it is particularly critical to evaluate and functionally validate a drug-target interaction also in the correct biological context to ascertain that the off-target is inhibited potently enough and at the same time is a strong enough disease driver. This aspect may underlie the challenges encountered with erlotinib in JAK2-mutant AML or with imatinib in the context of breast cancer expressing KIT or PDGFRβ, the latter of which are validated imatinib targets with demonstrated clinical relevance when mutated for instance in GIST. This emphasizes that, as would be expected, the mechanism of oncogenic activation of a target is highly relevant and that gene fusions, mutations, gene amplification and (over)expression are not necessarily of equal importance across different cancers. In addition, some drugs display good potency for overexpressed or amplified wild-type targets, but not certain mutant forms of these, or vice versa. This is nicely exemplified by dasatinib and axitinib, which have opposite preferences for wild-type BCR-ABL and its T315I gatekeeper mutant, and by imatinib and midostaurin with regard to wild-type and D816V KIT.

Like with targeted therapies in general, it is furthermore apparent that drug repurposing is most successful when a predictive biomarker is available to assist in directing the drug to the right patient population^{62; 340}. Of particularly high utility are genetic biomarkers, such as the presence of EML4-ALK translocations in NSCLC, activating KIT mutations in GIST or FLT3-ITD mutations in AML for indicating the use of crizotinib, imatinib and midostaurin, respectively. Although such markers may still not be sufficient to guarantee success in every case, as seen with erlotinib in JAK2-mutant AML and dasatinib in DDR2-mutant NSCLC for instance due to compensatory signaling, development of a biomarker in the context of drug repurposing is likely going to increase the chance for success of such effort. As the best predictive biomarkers have a strong mechanistic link, this again underscores the importance of obtaining detailed knowledge of the drug MoA and relevant target(s) and it will be interesting to follow future advances for instance with regard to repurposing of chloroquine or artemisinin now that specific drug targets of these have been identified in cancer cells.

The example of dasatinib repurposing for DDR2-mutant lung cancer, and similarly of tozasertib in BCR-ABL gatekeeper mutant CML, also shows that the potency of a drug for an off-target of interest has to be carefully considered in the context of clinically tolerable drug concentrations and drug toxicity, particularly as drugs have not been optimized for inhibition of off-targets and thus will often show weaker potency for these⁶². However, despite the ultimate failure of tozasertib, these efforts were not futile as the preclinical and clinical studies with tozasertib provided important proof of concept for efficacy and feasibility of targeting the BCR-ABL gatekeeper mutant and thus significantly supported the successful development of ponatinib, which currently constitutes the drug of last resort for CML patients. Cross-fertilization is also apparent for the development of ibrutinib and dasatinib as BTK inhibitors in hematological malignancies. Furthermore, if a drug is already approved for a different kind of cancer, like in the case of ibrutinib repurposing for lung or breast cancer, there will be already significant clinical expertise for handling its toxic side effects. However, as was seen with the curious case of desipramine in SCLC, there may be unique and unforeseeably low thresholds or potentially even different manifestations of toxicity in the context of a new disease^{25; 29}.

The obvious solution for overcoming suboptimal potency against cancer-relevant off-targets is to replace the current drug with one that is known to have higher potency for this target. Unfortunately, such a compound is often not available. An alternative strategy that has been applied with some success to address this issue has been the design of drug combination therapies that provide a second hit either downstream of the relevant off-target or in a parallel pathway that contributes to oncogenic signaling or is involved in adaptive resistance through compensatory signaling. Notably, midostaurin has received regulatory approval for FLT3-ITD positive AML only in combination with conventional chemotherapy. Similarly, the combination of axitinib as a BCR-ABL^T315I inhibitor with bosutinib is currently under clinical investigation for CML and a clinical trial of ceritinib with docetaxel, which enhances ceritinib’s polypharmacology-based efficacy, is ongoing in EML4-ALK-negative NSCLC.

The latter example also illustrates another, non-scientific, but equally critical factor for being able to successfully repurpose drugs, which is timing. As ceritinib had only relatively shortly before received regulatory approval for EML4-ALK-positive NSCLC, the intended indication, and thus secured a revenue flow, the pharmaceutical partner, in this case Novartis, was at that time interested in exploring and sponsoring additional therapeutic opportunities with ceritinib. This would have been unlikely to be of similar interest, if ceritinib had not received its initial approval yet or, conversely, if its patent protection period had been approaching its end. This latter aspect undoubtedly also facilitated the success of crizotinib as an ALK inhibitor considering that crizotinib drug development as a MET inhibitor was already in full swing when the EML4-ALK translocation was discovered as an oncogenic driver in NSCLC. It is remarkable, though, that Pfizer as the pharmaceutical company developing crizotinib, committed to exploring ALK-positive NSCLC already before crizotinib was approved, to some extent certainly also a testament to the strong validation of EML4-ALK as an oncogene. On the other hand, repurposing of midostaurin for EGFR-mutant NSCLC, which has developed resistance to EGFR TKI through acquisition of the T790M gatekeeper mutation, was not carried forward despite an at that time important unmet clinical need. This was probably due to the fact that midostaurin had not yet received its initial regulatory approval while at the same time its clinical development had taken unusually long thereby shortening its remaining patent life, factors that together would shift the risk versus potential benefit ratio and make an additional investment less favorable. In addition, there may have been a conflict with the pharmaceutical’s internal drug development pipeline, which contained nazartinib, a different (and newer) EGFR gatekeeper inhibitor. These cases demonstrate some of many, sometimes conflicting and often not transparent factors^{23; 25}, which determine how much interest a pharmaceutical company can afford to pay to a new repurposing opportunity.

What these considerations furthermore highlight is that successful drug repurposing, be it off- or on-target, is a major team effort that involves multiple players with a spectrum of expertise^{23; 25}. This is true already for the preclinical studies, which often start with a screen and continue with extensive phenotypic and mechanistic in vitro and in vivo validation before clinical translation can be considered. At this stage collaboration with pharmaceutical partners and their willingness/ability to sponsor clinical trials becomes critical as the costs for such trials usually exceed the budget of academic investigators, but also as they possess detailed knowledge of their drugs, for instance with regard to DMPK properties, side effects, administration and dosing schedules, which may not be available to the public. In addition, the repurposing approaches that are most successful involve clinician or physician-scientist collaborators or principal investigators, preferably at an institution with the staff and infrastructure to perform clinical studies. They have the necessary disease expertise, including detailed knowledge of the clinical needs and challenges, and access to the appropriate patient population and thus can evaluate the translational potential of any drug repurposing opportunity much better than a basic scientist and can point a project in a more impactful direction. Thus, the earlier a clinician is being involved in such an effort, the higher the potential for successful translation. Beyond this, clinicians often have established relationships with pharmaceutical companies through medical science liaisons whose job it is specifically to ease these interactions and promote clinical collaborations. Thus, their involvement can also provide access to communication channels, which can greatly facilitate collaboration between academic and pharmaceutical partners.

In summary, we here provide an overview of the rationale, methods applied and specific examples for off-target based repurposing in cancer. In doing so we furthermore discussed multiple scientific and non-scientific factors that either positively or negatively influence this process, some of which requiring careful consideration particularly for off-target based approaches. However, we hope we were also able to illustrate that off-targets should not just be viewed as potential causes of toxic side effects, but that off-target based drug repurposing efforts can be highly transformative and reveal novel or deepen our understanding of existing cancer biology mechanisms, alleviate the financial strain on pharmaceutical industry and, above all, provide urgently required new therapeutic approaches for cancer patients with previously unmet medical needs.