Discovery of Mcl-1 Inhibitors Offer Hope for New Cancer Therapeutics
By: Carol A. Rouzer, VICB Communications
Published: January 8, 2013
Fragment-based methods and structure-based design lead to inhibitors of myeloid cell leukemia 1 (Mcl-1), a protein that allows cancer cells to avoid programmed cell death.
One of the most common abnormalities found in human cancer is the overexpression of a protein called Mcl-1 (myeloid cell leukemia 1). Mcl-1 overexpression prevents cancer cells from undergoing programmed cell death (apoptosis), allowing them to survive despite widespread genetic damage and contributing to their malignant behavior. Clearly, an inhibitor of Mcl-1 that suppresses the aberrant survival and growth of cancer cells overexpressing this protein holds promise as a novel chemotherapeutic agent. To this end, Vanderbilt Institute of Chemical Biology member Steve Fesik and his lab have applied their expertise in fragment-based discovery and structure-based design to the task of finding a small molecule inhibitor of Mcl-1 activity [A. Friberg et al. (2012) J. Med. Chem., published online December 17, DOI:10.1021/jm301448p].
Mcl-1 is a member of the Bcl-2 family of proteins, which was first recognized through the discovery of a frequent form of DNA damage found in B cell leukemias. This damage involved a reciprocal translocation (exchange) of genetic material between chromosomes 14 and 18 that placed the gene for the Bcl-2 (B cell leukemia-2) protein next to the antibody heavy chain gene enhancer. Since B cells are specialized to make antibodies, this enhancer is very active, and the translocation leads to the production of large amounts of Bcl-2. Researchers rapidly discovered that overexpression of Bcl-2 inhibits apoptosis, thereby contributing to the survival of leukemia cells bearing the translocation.
The discovery of Bcl-2 overexpression in B cell leukemias ultimately revealed an entire family of related proteins that regulate the apoptotic pathway. The Bcl-2 family includes proapoptotic members (such as Bax and Bak) which, upon activation, form a homo-oligomer in the outer mitochondrial membrane that leads to pore formation and the escape of mitochondrial contents, including cytochrome C. This is an early step in triggering apoptosis. Antiapoptotic members of the Bcl-2 family (such as Bcl-2, Bcl-xL, and Mcl-1) block the activity of Bax and Bak, while a third set of proteins (such as Bid, Bim, Bik, and Bad) exert additional regulatory functions (Figure 1).
Figure 1. Examples of how Bcl-2 proteins interact to control programmed cell death (apoptosis). A. An apoptotic signal activates the proapoptotic protein Bax, causing it to bind to the outer mitochondrial membrane. If it oligomerizes with other Bax proteins, it will form a pore in the membrane allowing release of contents, including the protein cytochrome C, which is an early step in the apoptotic pathway. B. The antiapoptotic protein Bcl-2 binds to Bax and prevents its self-oligomerization. C. Other proteins, such as Bid can bind to Bcl-2, causing it to release Bax. D. No longer bound to Bcl-2, Bax self-oligomerizes, causing the release of cytchrome C (CC).
The overexpression of antiapoptotic Bcl-2 family members in numerous cancers has led to the development of an inhibitor of Bcl-2 and Bcl-xL, ABT-263, and a selective Bcl-2 inhibitor (ABT-199) that are currently showing promise in early clinical trials. However, these inhibitors do not block the activity of Mcl-1, which is actually the most frequently overexpressed Bcl-2 family member in human cancer. In fact, Mcl-1 overexpression conveys resistance to therapy with ABT-263. Some Mcl-1 inhibitors have been reported, but none is currently in clinical trial, so the Fesik laboratory’s work addresses an important unmet clinical need. The fact that Mcl-1 exerts its effects through protein-protein interactions over large surfaces makes the discovery of high affinity small molecule inhibitors particularly challenging. The fragment-based approach used by the Fesik lab is well-suited to meet this challenge.
Fragment-based discovery focuses on identifying fragments (molecules of molecular weight <300) that bind to a target molecule with at least millimolar affinity. The fragments provide the starting point for establishing some level of interaction between the target protein and small molecules and serve as an initial test of a target’s “druggability”. The Fesik lab expressed recombinant Mcl-1 (amino acids 172-327) that was uniformly labeled with 15N. The label allowed them to use a technique called SOFAST 1H-15N-HMQC NMR to screen a fragment library of >13,800 compounds. This NMR technique provides two-dimensional data that correlate the 1H and 15N signal from every N-H bond in the protein. Plots of the data (Figure 2) reveal changes in the protein that occur upon binding of a fragment through a shift in the position of the signals. This rapid and sensitive screening approach yielded 132 hits from 11 chemical classes and provided assurance that Mcl-1 is druggable.
Figure 2. 1H-15N HMQC NMR spectra of 15N-labeled Mcl-1 in the presence (red) and absence (black) of a fragment exhibiting a 22 μM binding affinity. Note that in most cases, the red and black signals are superimposed. Signals that are observed in distinct locations indicate changes in position or environment of the relevant region of the protein due to fragment binding. Reprinted with permission from Friberg et al. (2012) J. Med. Chem., published online December 17, DOI:10.1021/jm301448p. Copyright 2012, American Chemical Society.
Antiapoptotic Bcl-2 family proteins interact with multiple other proteins that possess a BH3 domain. Ironically, Mcl-1 binds particularly well to a helical peptide comprising its own BH3 domain. The Fesik lab exploited this interaction by monitoring the binding of fluorescently labeled Mcl-1-BH3 peptide with Mcl-1 through the technique of fluorescence polarization anisotropy (FPA). The FPA assay allowed them to evaluate the ability of each of their fragment hits to interfere with Mcl-1’s interactions with the BH3 domains of other proteins. As the assay was being developed, a careful analysis of the NMR data revealed two classes of fragment hits that appeared to bind to Mcl-1 in different regions of the same binding pocket. Class I comprised 6,5-fused heterocyclic carboxylic acids, while Class II comprised hydrophobic aromatics tethered by a linker to a polar functional group, most frequently a carboxylic acid (Figure 3). The FPA assay provided the means to rapidly assess the binding affinities of analogues of the hits in these two fragment classes, allowing structure-activity relationships (SAR) to be elucidated.
Figure 3. Examples of a Class I and Class II compound and the merging of these two to form compound 60, which led to a marked increase in affinity for Mcl-1. Also shown is compound 53, the most potent Mcl-1 inhibitor obtained in this study.
To better understand the interaction of the Class I and II compounds with Mcl-1, the Fesik group expressed 15N,13C-labeled Mcl-1 for use in NMR-based structural studies. The results confirmed that the hydrophobic portions of the two classes of compounds bind at distinct but nearby sites in the BH3 binding pocket of Mcl-1, while the carboxylate groups of both classes shared an interaction with arginine-263 (Figure 4). This shared interaction explained why fragments from Class I and Class II could not bind to Mcl-1 at the same time. The results also suggested that synthesis of a merged compound by connecting the aromatic group of a Class II compound through a linker to the 3-position of a Class I heterocycle should produce a molecule that binds at both sites simultaneously and with greater affinity (Figure 3). After first using the SAR data to obtain optimized Class I and Class II compounds, the investigators began synthesizing merged molecules. The results led to increases in affinity of over two orders of magnitude. In addition, the merged molecules were selective for Mcl-1, exhibiting poor binding to either Bcl-2 or Bcl-xL. This finding is particularly important, because the Bcl-xL inhibitory activity of ABT-263 has resulted in toxicity. Further structure optimization of the merged compounds yielded inhibitors with sub-micromolar potency and 15- to 270-fold selectivity for Mcl-1 inhibition (Figure 3).
Figure 4. NMR-based models of the binding of a representative Class I (A) and Class II (B) molecule to Mcl-1. Interactions with important amino acid residues in the protein are indicated. Reprinted with permission from Friberg et al. (2012) J. Med. Chem., published online December 17, DOI:10.1021/jm301448p. Copyright 2012, American Chemical Society.
X-ray crystal structures of high affinity inhibitors confirmed the expected binding mode, with the two hydrophobic ring systems occupying distinct sites in the BH3 binding pocket and the carboxylate interacting with arginine-263 (Figure 5). The investigators were interested to find that binding of the inhibitors induced changes in Mcl-1 structure that facilitate the interaction between arginine-263 and the carboxylate of the inhibitor. These changes were not readily predicted from the structure of the BH3 peptide bound to Mcl-1. A similar binding pocket is present in both Bcl-2 and Bcl-xL; however the pocket is not as deep or as large in these proteins as it is in Mcl-1. This may explain the basis of the Mcl-1 selectivity of the newly discovered inhibitors. Also of interest was the finding that even the most potent of the inhibitors interact with only a portion of the Mcl-1 binding pocket. This leaves a large region of the pocket that could be exploited for further improvements in binding affinity. Indeed, the Bcl-2/Bcl-xL inhibitor ABT-737, similar in structure to ABT-263, is a much larger molecule than any of those studied here and interacts with its target over a substantially greater area. An obvious next step will be to screen the fragment library to search for compounds that bind to Mcl-1 in the presence of one of the high affinity inhibitors such as compound 53 or compound 60. This approach may lead to the discovery of a fragment that interacts with a region of the binding pocket that has not yet been exploited. Successful linking of such a fragment with compound 53, 60, or an appropriate analogue should lead to another major improvement in potency.
These promising results provide the framework for the discovery of potent Mcl-1 inhibitors suitable for clinical evaluation.
Figure 5. X-Ray crystal structure of compound 60 (A and B) and compound 60 (A and C) bound to Mcl-1. (A) shows a merged structure for both compounds. (B) and (C) show the protein surface interacting with each individual molecule. Reprinted with permission from Friberg et al. (2012) J. Med. Chem., published online December 17, DOI:10.1021/jm301448p. Copyright 2012, American Chemical Society.