Phosphoprotein-based drug target activation mapping for precision oncology: a view to the future

KEYWORDS:

• Drug target
• activation mapping
• oncology
• phosphoproteomics

Historically, genomic-based tumor analyses have yielded critical information impacting treatment, such as HER2 amplification and transcription-based tumor classification in breast cancer, and identification of ALK/ROS/EGFR genomic alterations in non-small cell lung cancer (NSCLC). Most current precision oncology trials that incorporate molecular profiling rely solely on broad scale-genomic information such as copy number variation, exome panel-based profiling, mutational analysis, and transcription-based profiling to characterize individual lesions and guide treatment selection. However, this approach is limited in many respects, and key information about tumor biochemistry is often neglected.

The prevalence of most actionable genomic alterations in various cancers is low and sensitivity to targeted compounds may differ across tumors of different origin even in the presence of a common ‘driving’ event [1]. Broad-scale genomic profiling provides scientists and physicians with a complex snapshot of all genomic alterations acquired by a tumor beginning from the early phases of tumorigenesis. Consequently, discriminating between derangements that were necessary for the onset of the disease versus molecular events that hold potential therapeutic value at any given time is daunting. Likewise, aberrant signaling caused by alterations at the gene and transcript level can also be triggered by genome-independent mechanisms including the secretion of soluble factors or the establishment of adaptive feedback mechanisms within the context of the tumor microenvironment.

The concept of ‘precision’ strongly suggests that exactness is a pillar of any process developed under this umbrella. However, most precision oncology treatments rely on genomic measurements that are not tied directly to the mechanism of action of the drugs themselves. Precision oncology drugs do not target genes (‘the playbook’), they target proteins (‘the players’). It has been amply demonstrated that DNA alterations often are not transcribed, and mRNA expression changes are not strictly reflected at the protein level. Furthermore, DNA, RNA and even protein expression are unlinked to protein activation (e.g. phosphorylation) [2–4]. So when it comes to cancer, why should we limit our knowledge to a playbook, when we can actually follow the players in the field?

Regardless of the nature of the molecular events driving individual lesions, abnormal cellular functions in tumors are mainly attributable to protein kinase dysfunction resulting in aberrant signaling network activity. Under physiological conditions, kinase activation is tightly regulated through phosphorylation (or dephosphorylation) at specific sites leading to the sequential cascading activation of downstream kinases which ultimately affect gene transcription. Dysregulation of these kinase-driven signaling networks is a hallmark of tumor cells, and most FDA-approved and experimental targeted therapeutic agents, such as monoclonal antibodies and small molecule inhibitors, are designed to prevent the activation of key kinases or disrupt their interactions with ligands or downstream substrates. These aberrant phosphoprotein-driven signaling events are not captured by any current commercially available molecular profiling system.

Independent investigations have demonstrated that phosphoprotein mapping of the drug target activation landscape of tumor cells offers unique and highly valuable information for precision oncology. It can be used to identify predictive, prognostic, and therapeutic biomarkers, which might be missed even by traditional protein expression analyses [4–6]. For example, our group identified a cohort of breast cancer patients with HER2− lesions (measured by Immunohistochemistry [IHC]/Fluorescent in situ hybridization [FISH]) that had phospho-HER2 levels comparable to HER2+ tumors [4]. This apparent genome-independent activation of HER2 was accompanied by elevated activity of EGFR and HER3, suggesting heterodimerization across members of the HER receptor family, and observed activation of downstream effectors such as SHC and AKT. Identification of patients in whom genome-independent mechanisms may modulate activity of a drug target can have significant therapeutic implications and further refine treatment selection for cancer patients.

Given the complex and interconnected networks that kinase signaling generates, the activation of a receptor tyrosine kinase or a targetable cytosolic protein is relevant to the biology of a tumor only if it successfully activates its downstream effectors. Consequently, activation of a drug target should be evaluated along with its substrates. For example, phosphorylation of EGFR has shown predictive value as a single marker in NSCLC patients undergoing treatment with EGFR inhibitors even in patients that do not harbor an EGFR mutation [6,7]. In addition, phosphorylation of EGFR downstream substrates such as AKT, STAT3 and ERK 1/2 showed predictive value for response to anti-EGFR targeted treatment in NSCLC [7]. In hepatocellular carcinomas (HCC), elevated baseline levels of phosphorylated ERK1/2 have been demonstrated as a potential predictive marker for response to the RAF/VEGFR/PDGFR inhibitor sorafenib [8], and it has also been identified as a potential surrogate marker in peripheral blood mononuclear cells collected longitudinally from HCC patients undergoing sorafenib-based treatments [9].

The role of the PI3K/AKT/mTOR signaling axis in endocrine therapy resistance in ER+ breast cancer has received significant attention. Activation of the pathway is associated with resistance to endocrine modulators such as tamoxifen, but robust biomarkers for activity are lacking. The search for genomic predictors of resistance in this pathway has yielded mixed results. Analyses of PIK3CA mutations in a cohort of ER+ postmenopausal patients failed to show an association with tamoxifen response [10], while high mRNA levels of S6K2 and 4EBP1, downstream elements of the pathway, have been associated with poor outcome in ER+ patients [11]. At the protein level, several large studies have shown that response to endocrine therapy was associated with baseline phosphorylation levels of mTOR kinase and its substrate p70S6K, as well as total levels of 4EBP1 [2,11,12]. It was found that ER+ patients whose tumors had activated mTOR signaling received little or no benefit from tamoxifen, pointing to the need for prescreening these patients and perhaps combining mTOR therapies with ER-directed therapies from the outset.

With immune-based therapies paving a new era for precision medicine, the potential for kinase-driven analyses should not be underestimated. While expression of immune checkpoints and their ligands have been long considered, the main source of predictive markers to agents targeting the PD-L1/PD-1 axis, additional molecular testing appears necessary to achieve effective stratification to treatment. Amongst many, tumor mutation burden and interferon signaling are currently under intense scrutiny as molecular mechanisms that directly modulate tumor cell immune evasion and consequently affect response to agents modulating immune activation [13,14]. In recent work using quantitative IHC, Tumeh and colleagues have explored the role of phosphorylated STAT1 as a predictive marker of response to pembrolizumab in melanoma patients [15]. STAT1 phosphorylation in CD8⁺ T cells located at the invasive front of the tumor was greater in patients that benefited from treatment compared to patients whose tumors were resistant to the anti-PD-1 agent. This association was consistently detected in biopsies collected before and during treatment with pembrolizumab. Because T-cell activation and effective immune surveillance are orchestrated by the sequential activation of interconnected kinases and their downstream substrates, the development of companion/complementary tests able to capture these events may further increase our ability to identify patients that are ideal candidates for treatments with targeted immunotherapies.

Undoubtedly, the molecular mechanisms underlying tumor progression and response to therapy are extremely heterogeneous and dynamic as cancer cells adapt and prosper in the host microenvironment. Most compensatory and adaptive mechanisms as well as the cross-talk between cancer cells and surrounding stroma/immune cells are driven by unpredictable post-translational events. Given the multifaceted biology of the disease, improving cancer treatment cannot be achieved by genomic, proteomic or phosphoproteomic mapping in isolation. The concept of multi-omic molecular profiling, where these complementary data fields are evaluated concomitantly, should be espoused as a new paradigm for classifying tumors and identifying targetable alterations in individual patients.

A number of tools are currently available to the research community for mapping kinase-driven signaling activity [16], some of which are offered in CAP/CLIA-compliant or accredited laboratories [17]. The development of standardized panels to be used as companion/complementary diagnostic tools capable of directly measuring drug targets and downstream substrate activity should become a priority in precision oncology. For example, interferon gamma signaling and JAK/STAT phosphorylation-based activation should be routinely measured along with PD-L1 expression and tumor burden in patients receiving anti-PD-L1/PD-1 based treatments [13]. Similarly, activation of the mTOR pathway should be measured when considering endocrine therapy, and phosphorylation measurements of the HER2 receptor and its downstream substrates should be assessed as complementary tests to conventional IHC and FISH in breast cancer to guide therapeutic selection. New approaches to measure the state of the signaling architecture of tumor cells are evolving through the use of single cell analysis [18] and multi-parametric tissue imaging technologies that bring histomorphology and localization-based context to the analysis [19]. As these technologies mature to commercially available and CAP/CLIA accredited assays, their impact in precision medicine will be more fully realized.”

Ultimately, the use of phosphoprotein based CDx and therapeutic response predictors would optimally be placed in the context of mulit-omic analysis wherein genomic data and proteomic/phosphoprotein data are integrated together for more accurate response prediction. Indeed, in the context of genomic alterations, functional proteomic data concerning which drug target pathways are ‘in use’ and activated would be more powerful than the data in isolation. The use of multi-omic based molecular profiling, including phosphoprotein mapping, has recently been tested in clinical trials for metastatic breast cancer (i.e. NCT01074814, NCT01919749, NCT03195192) and has shown encouraging results and clinical benefits in patients that have progressed on standard of care [20]. These preliminary observations suggest that classification of tumors using a multi-omic and pathway-centered analysis may offer a new means for classifying tumors more accurately and for delivering more effective precision treatment to cancer patients.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.