How Compensatory Mechanisms and Adaptive Rewiring Have Shaped Our Understanding of Therapeutic Resistance in Cancer

The molecular mechanisms that orchestrate all cellular functions in multicellular organisms form a vast array of interconnected signaling cascades that interpret and transmit different stimuli in order to produce a response. Homeostatic regulation of these pathways is coordinated by measures to amplify, sustain, or counteract specific signals, which in turn control signal strength and duration to determine the biological outcome. In this regard, the landmark study by O’Reilly and colleagues published in Cancer Research in 2006 (1) has proven instrumental in understanding how cells regulate signaling output via negative feedback loops. In summary, O’Reilly and colleagues demonstrated that pharmacologic inhibition of mechanistic target of rapamycin (mTOR, formerly known as mammalian target of rapamycin) complex 1 (mTORC1) results in enhanced PI3K/AKT pathway activation by relieving mTORC1-induced inhibition of insulin and insulin-like growth factor (IGF) signaling. A prominent consequence of this line of research has been an increased understanding of how cancer cells employ nongenetic compensatory mechanisms to mediate adaptive resistance to targeted PI3K and mTOR inhibitor therapies, and has guided the development of novel dual inhibitors. Here, we highlight some of the major ramifications from this and other related studies.

The PI3K–AKT–mTOR signaling pathway is frequently altered in the majority of human cancers. The canonical PI3K–AKT–mTOR pathway is initiated by stimulation of receptor tyrosine kinases (RTK) or G protein-coupled receptors (GPCR), which recruit PI3K to the plasma membrane and induce its activation. Upon activation, class I PI3Ks phosphorylate the 3′ hydroxyl on phosphatidylinositol-4,5-bisphosphate (PIP₂) on the inner leaflet of the lipid bilayer to generate phosphatidyl inositol-3,4,5-triphosphate (PIP₃), which acts as a docking site for the activation of multiple downstream signaling cascades. The serine/threonine kinase AKT binds to PIP₃ on the plasma membrane, where it is activated by phosphorylation on threonine 308 (T308) by PIP₃-bound phosphoinositide-dependent protein kinase 1 (PDK1). In addition, AKT is phosphorylated on serine 473 (S473) by mTORC2. Activated AKT in turn phosphorylates numerous targets to promote cell survival, growth, and proliferation (Fig. 1).

The serine/threonine kinase mTOR forms two distinct complexes—mTORC1 and mTORC2—that differ in subunit composition and substrate specificity. Although the precise upstream mechanisms leading to mTORC2 activation are still unclear, mTORC2 phosphorylation of AKT on S473 is required for maximal AKT activity. AKT signaling activates mTORC1 primarily by phosphorylating and inactivating tuberous sclerosis complex 2 (TSC2) within a complex that also includes TSC1 and TBC1 domain family member D7 (TBC1D7), thereby blocking inactivation of the small GTPase Rheb, which is required for mTORC1 activation. Activated mTORC1 promotes cell growth and proliferation through phosphorylation of multiple targets that enhance anabolic metabolism. Two prominent mTORC1 targets include the ribosomal protein S6 kinase 1 (S6K) and eukaryotic translation initiation factor 4E-binding protein (4E-BP). In particular, mTORC1 increases protein biosynthesis through the combined activation of S6K, which stimulates unwinding of the 5′ untranslated region (UTR) on mRNA, and inhibition of 4E-BP via phosphorylation, which induces cap-dependent translation (Fig. 1).

Prior to the study by O’Reilly and colleagues, it had been observed that prolonged stimulation of the insulin or IGF1 pathway in normal cells elicits negative feedback downregulation of the pathway via degradation of the insulin receptor substrates 1 and 2 (IRS1/2) mediated by mTOR-S6K signaling (2). In addition, just a couple of months before publication of O’Reilly's study, Sun and colleagues reported in Cancer Research a similar finding, in which rapamycin treatment upregulated AKT S473 phosphorylation that could be reversed by PI3K inhibition (3). Likewise, Shi and colleagues (4) observed increased AKT T308 phosphorylation in response to rapamycin in vitro, which could be reversed by blocking the IGF1 receptor (IGF1R). The authors also found increased AKT T308 phosphorylation in mouse xenografts in response to the rapamycin derivative (rapalog) temsirolimus (CCI-779; ref. 4).

In their study, O’Reilly and colleagues described a mechanism by which rapamycin upregulates IRS-1 protein levels, thus amplifying insulin/IGF1 signaling and resulting in progressively enhanced AKT S473 phosphorylation, which could be reversed by PI3K or IGF1R inhibition. Importantly, the authors analyzed tumor biopsy samples from patients before and after treatment with the rapalog everolimus (RAD001), showing that mTOR inhibition with a rapalog is associated with markedly increased AKT phosphorylation in the clinic, which, due to the oncogenic potential of AKT hyperactivation, could lead to significantly detrimental results. Similar clinical results were later observed in a phase I trial of rapamycin in patients with PTEN-deficient glioblastoma, in which 7 out of 14 patients treated with rapamycin showed enhanced AKT activation, which correlated with a shorter time to progression (5).

In addition, further research has revealed yet additional layers of negative feedback and compensatory mechanisms regulating the PI3K–AKT–mTOR pathway. Phosphoproteomic studies identified the adapter protein Grb10 as a direct target of mTORC1 that is stabilized upon phosphorylation (6). Grb10 directly inhibits insulin/IGF1 signaling by blocking IRS1/2 binding to its receptors (6). Moreover, mTORC1 signaling can inhibit mTORC2 activity by inducing phosphorylation of Sin1, a component of mTORC2, by S6K, which results in mTORC2 dissociation and inhibition of AKT phosphorylation upon RTK signaling, including via insulin/IGF1, EGF, and platelet-derived growth factor (PDGF; ref. 7).

Of note, rapamycin and its derivatives bind to an allosteric site on mTORC1 by forming a complex with FK506 binding protein (FKBP12). Therefore, these compounds do not directly affect mTORC2, and their effects on mTORC1 activity differ between specific substrates. For example, S6K phosphorylation by mTORC1 is highly sensitive to rapamycin, while 4E-BP is only partially inhibited. In light of the findings described above, ATP-competitive mTOR kinase inhibitors that could target both mTORC1 and mTORC2 were developed. These compounds effectively block AKT S473 phosphorylation by mTORC2, but only transiently inhibit AKT T308 phosphorylation by PDK1 and AKT activity, as assessed by phosphorylation of AKT targets (8). These effects were due to a strong induction of RTK activity, including members of the EGFR family and insulin/IGF1 receptors (InsR/IGF1R), and a consequent upregulation of PI3K and activation of PDK1-mediated AKT signaling (8). These observations highlight even more complex regulatory mechanisms. Indeed, pharmacologic AKT inhibition was shown to upregulate RTK signaling via EGFR, HER3, and InsR/IGF1R (9). This upregulation was mediated primarily due to the reactivation of FOXO transcription factors upon loss of inhibition by AKT, thus resulting in upregulated FOXO-mediated RTK expression (9). In these cases, combined targeted RTK and AKT inhibition enhanced antitumor efficacy in vivo (9).

More recently, pharmacologic inhibition of the cyclin-dependent protein kinases 4 and 6 (CDK4/6) has been shown to upregulate signaling through EGFR family receptors and to enhance AKT phosphorylation by blocking phosphorylation and inhibition of TSC2, hence resulting in enhanced mTORC1 activity and reduced negative feedback inhibition of RTK signaling (10). Active Cyclin D-CDK4/6 complexes promote cell growth and proliferation by phosphorylating and inhibiting the RB tumor suppressor protein, thus enhancing E2F-mediated gene expression and cell cycle progression. Hence, the direct link between these two pathways provides yet another example of the importance of compensatory mechanisms in regulating cellular signaling.

Finally, PI3K-independent compensatory mechanisms have also been described. For example, treatment with the dual PI3K/mTOR inhibitor dactolisib (BEZ235) was shown to upregulate the ERK pathway as a consequence of enhanced HER2/HER3 activation (11). Indeed, inhibiting the PI3K–AKT–mTOR pathway often leads to upregulation of the ERK pathway.

Due to the major role of the PI3K–AKT–mTOR pathway in cancer, there is considerable interest in developing effective therapeutic inhibitors for use in the clinic. Rapamycin has been approved by the FDA as an immunosuppressant since 1999. In 2015, rapamycin was approved for the treatment of lymphangioleiomyomatosis, a rare disease characterized by progressive overgrowth of smooth muscle cells that primarily affects the lungs, kidneys, and lymphatic system. To date, the rapalogs temsirolimus and everolimus have been approved by the FDA for the treatment of renal cell carcinoma. Unfortunately, clinical development of pan-PI3K, dual PI3K/mTOR, and mTOR kinase inhibitors has been hampered by significant adverse toxicity. On the other hand, isoform-specific PI3K inhibitors have demonstrated improved tolerability and efficacy. The PI3Kδ inhibitor idelalisib, PI3Kα inhibitor alpelisib, and the PI3Kα/δ inhibitor copanlisib have been approved by the FDA for the treatment of leukemia/lymphoma, PIK3CA-mutant hormone receptor–positive breast cancer (in combination with hormone therapy) and relapsed follicular lymphoma, respectively. Lastly, AKT inhibitors have shown mixed responses, including two candidates—ipatasertib and capivasertib—with promising safety profiles and preliminary efficacy that have advanced to phase III clinical trials.

Numerous clinical trials are currently evaluating the combination of selective PI3K–AKT–mTOR pathway inhibitors with various kinds of targeted therapies and/or immunotherapy. In this respect, CDK4/6 inhibitors are particularly interesting. For example, pharmacologic CDK4/6 inhibition has been shown to synergize with HER2 blockade to overcome therapeutic resistance in preclinical models (10). This line of studies led to the phase II clinical trial monarcHER (NCT02675231) and the phase III trial PATINA (NCT02947685) to evaluate the rational combination of HER2 inhibition and a CDK4/6 inhibitor (and hormone therapy if appropriate) in HER2-positive breast cancers. Given the strength of compensatory mechanisms and adaptive rewiring in response to pharmacologic targeting of this pathway, combination therapies hold the most promise for achieving durable results.

J.S. Bergholz reports other support from Dale Family Foundation and personal fees from Geode Therapeutics Inc. outside the submitted work; in addition, J.S. Bergholz has a patent for DFCI 2180.001 (DFS-166.25) issued. J.J. Zhao is a founder and board director of Crimson Biotech and Geode Therapeutics. No other disclosures were reported.