A personalized medicine approach to drug repurposing for the treatment of breast cancer molecular subtypes

Introduction

Breast cancer is a major public health concern and a relevant mortality issue in young women worldwide [ ]. Mammary tumors are extremely heterogeneous at the histological, molecular, and systemic levels. To handle such variability, oncology researchers and clinicians have developed a number of prognostic and therapeutic approaches ranging from classifications based on clinical parameters or histopathologic markers (such as estrogen, progesterone, and epidermic human growth factor receptors) [ ] to classifications based on gene expression profiling [ ] as well as various attempts to patient-based phenotyping.

The main objective of these methods is to allow the design of improved therapeutic procedures. However, the lack of such options for certain subtypes—in particular for triple-negative tumors—represents an important source of frustration challenge for clinical oncologists who often have to resort to cytotoxic therapies with a large number of adverse side effects.

The development of new anticancer drugs, although extremely relevant, is a very slow and high-cost endeavor. In contrast, the repurposing of many approved drugs (both anticancer and non–anti-cancer) has become an effective strategy to broaden the options in the oncologic therapeutic spectrum with the undeniable advantages of being faster, cheaper, and faster to go through preclinical and clinical stages of validation protocols, tier studies, and clinical trials [ ]. Of particular relevance has been the strategy to develop tailor-made drug cocktails based on personalized medicine studies.

There is indeed a large body of evidence to support the claim that combination therapies are more effective against late-stage neoplastic tumors than single agents or sequential drugs combinations, given the large inter- and intratumoral heterogeneity among patients [ , ]. Tailor-made combination therapy is not without caveats, in particular given the fact that the development and testing procedures in the pharma industry are not, in general, designed with multitherapy in mind [ ].

Another important caveat for the rational design of multidrug repurposing approaches lies in the fact that it is a quite interdisciplinary task, even more reliant on computational biology, bioinformatics, and artificial intelligence (AI) than “usual” oncopharmacology. Clinical practitioners and pharmaceutical company officers need to become aware of this fact and adapt their current practices accordingly, take for instance, the wealth of information on chemical, pharmacological, and genomic databases.

Let us briefly analyze one (of many) instance in which data mining approaches become relevant. Most compound and drugs currently used in the clinic have a large number of off-target effects aside of its main therapeutic mechanism of action. Such off-target actions are indeed the basis for a lot of drug repurposing strategies. Computer-aided (CA) interrogations of the many large databases on drugs and mechanisms of action may allow the identification of (additional) specific targets [ , ].

Other computational strategies of drug repurposing include the combination of knowledge discovery in databases (KDD) with molecular profiling and modelization to identify novel drug–target interactions. Often machine learning (ML) algorithms are used to screen enormous catalogs of molecules to search for drug–target interactions. The combination of KDD/ML with high-throughput in vitro assay screening has shown to be an extremely effective strategy in multifactorial diseases such as cancer largely outperforming single-drug approaches [ ]. Aside from “single-shot” drug–target interactions, molecular and phenotypic heterogeneity in cancer tumors needs to be taken into account for the design of effective anticancer therapeutic interventions, for instance, when dealing with immunotherapy [ ]. Although highly publicized and extremely successful in some cases, the majority of cancer patients do not benefit from immunotherapy, quite likely by effects related to the immunosuppressive nature of their individual tumor microenvironment. Designs considering a personalized approach need to focus on deciphering the individually affected metabolic pathways. Li and coworkers have discussed quite extensively the way metabolic circuits regulate/deregulate intrinsic antitumor immunity pathways and how some (many, indeed) of these relationships have even reached the clinical trial stage (see, for instance, Table 1 in Ref. [ ]). Along the same lines, repurposing of immunomodulatory drugs such as thalidomide, lenalidomide, and pomalidomide has been highly enhanced by the extensive validation of computationally predicted biomarkers in patient-diverse subpopulations [ ]. There is of course the concern that the therapeutic efficacy of these treatments is highly heterogeneous [ ]. For this reason, these drugs are usually part of a polypharmacological treatment. For instance, Shen and coworkers have reported that the use of thalidomide increases delivery and efficacy of cisplatin [ ]. On the other hand, the main challenges for their wider use rely on the fact that there have been many reports of adverse drug effects, these include peripheral neurotoxicity [ ], as well as teratogenic and dermatological, among other effects [ ].

Of course, the challenges of anticancer drug repurposing do not end with the (many) molecular targets and off-target prediction technicalities. Barriers to repurposing often start with the availability of better diagnostic tools able to predict and stratify patients response to therapy. This is another field in which CA and AI approaches may result helpful [ ]. Aside from CA/AI tools, systems biology modeling may also allow for better phenotyping and prognostics, leading to better-suited drug repurposing designs [ , ].

The use of patient-derived genomic information to study how genetic alterations influence the routes to tumor progression and cell survival is key to uncover tumor-specific vulnerabilities that may lead to the development of narrowly targeted therapeutic interventions, including knowledge-driven drug repurposing [ ].

Once novel repurposing ideas have been found, there is the need to develop strategies to make these ideas reach the patients [ ]. Among the many caveats, three main questions have been considered particularly relevant [ ]:

• 1.

  How to establish the recommended dose to achieve anticancer activity, especially when repurpose drugs were not initially intended as antitumor drugs. In this regard, novel computational approaches based on a systems biology philosophy could be applied [ ].

• 2.

  How to deal with intellectual property, patent, and licensing issues, both in generic and proprietary treatments?

• 3.

  Since cancer-related clinical trials are usually more expensive, need longer follow-ups, and are very prone to failure than those of noncancer drugs, there may be sufficient financial incentives for the pharmaceutical stakeholders. This is so since most repurposed drugs, aside from being clinically significant, need to demonstrate higher cost-effectiveness ratios than newer treatments.

In this chapter, we will outline and discuss some of the major themes in this regard, particularly related to the development of translational bioinformatics strategies to guide clinical oncologist in the design of more effective and personalized therapies using repurposed drugs to treat breast cancer subtypes at the individual (personalized) level.

Mutation-specific therapies as an approach to personalized medicine in cancer: pros and cons

Ever since the discovery of the first cancer-associated mutations and oncogenes, one important tenet of anticancer therapy has been the discovery of such cancer “causal” mutations (in particular tumor drivers), to later try to figure out the structure of a “silver-bullet” drug, a molecule that may target tumors on an extremely specific fashion leaving nontumor cells unaffected. It actually seemed like a pretty good idea.

In Fig. 7.1 , a simplified view of a generalistic mutation profile–based approach to personalized drug repurposing to treat breast cancer tumors is presented. First, by means of high-precision DNA sequencing, a tumor-specific mutation is found in the genome of a patient. In the best scenario, it may be a mutation that is already known and annotated in a “cancer panel” so that its intrinsic molecular characteristics are available.

With this knowledge and after assessing that these mutation is absent in the germline, a targeted therapy may be developed, being this the finding of a monoclonal antibody able to recognize the effect of the mutation at the protein level [ ], the composition of an antibody–drug conjugate complex [ ], or the synthesis of a small molecule drug [ , ].

Once this therapy or combination of given therapies is known, one must interrogate the pharmacological databases for drug repurposing to look up for the existence of such pharmaceuticals and in due case its possible off-targets and side effects [ ].

Unfortunately, except on a few exceptional cases of highly penetrant mutations for which the response to such silver bullets has been equally exceptional, most cancer patients did not benefit by using these approaches [ , ].

Indeed, it has been argued that “… Precision oncology promises to pair individuals with cancer with drugs that target the specific mutations in their tumor, in the hope of producing long-lasting remission and extending their survival. The basic idea is to use genetic testing to link patients with the drugs that will work best for them, irrespective of the tissue of origin of their tumor. Enthusiasm has been fueled by reports of exceptional or super responders — individuals for whom experimental therapies seem to work spectacularly well … [yet] … Few patients benefit from precision oncology. Data from some 2600 people enrolled in a sequencing program at the MD Anderson Cancer Center in Houston, Texas, showed that just 6.4% were paired with a targeted drug for identified mutations ” [ , ].

The field is moving fast in recent times, and the general trends however are not apparently changing yet. A recent large-scale survey intended to evaluate the benefits of genome-driven oncology, the MOSCATO study [ , ], suggests that a genome-driven strategy for cancer therapy is able to improve clinical outcomes in a significant minority of patients who undergo molecular screening. The increase in progression-free survival rate as compared with the group with no screening is larger than 1.3 (30% increase) in about 33% of the patients with a targeted therapy matched to a genomic alteration. This seems to be good news and they are indeed, but only, for a fraction of patients. According to the MOSCATO study “…although these results are encouraging, only 7% of the successfully screened patients benefited from this approach…”. So, apparently figures go up from 6.4% to about 7% of success. This situation may nonetheless improve in the upcoming future with the advent of better, more precise tumor variation assessment and improved therapeutic designs.

These facts have eroded the optimism on mutation-centered personalized medicine that was prevailing years ago [ , ]. Mutational heterogeneity is indeed key to understand the challenges of this conceptually appealing approach. In recent times, mutational variability in breast cancer tumors has been unveiled at an unprecedented scale [ ].

It has been revealed that mutation induced by pharmacological treatment increases the severity and the drug resistance of metastatic tumors. This is so, since mutational burden enhances up to the double the relative contributions that secondary breast cancer mutational signatures present [ ].

In particular, the mutation frequency of well-known driver genes increases in metastatic breast cancer. Such changes in frequency are indeed significantly associated with previous pharmacological treatment [ , ]. It is actually known that the APOBEC family of APO enzymes plays a relevant role in these mutational heterogeneity [ , ].

To make things still more complex, the genomic region of APOBEC enzymes is one of abnormally high sequence heterogeneity [ ]. Breast cancer tumors in particular (and most epithelial tumors indeed) are also characterized by a complex subclonal structure [ ], a fact that may cause that even drugs perfectly targeting the mutational profile of a given clone may be unable to have a sustained therapeutic effect on these tumors.