Promise and Problems of Bcl-2 Antisense Therapy

Ever since its implication in multidrug chemoresistance in tumors, the bcl-2 gene (also known as BCL2) has stood out among molecular targets in oncology ( 1 ). The protein encoded by bcl-2 is a potent suppressor of apoptosis and is found at inappropriately high levels in probably more than half of all cancers in humans [reviewed in ( 2 )]. Gene transfer-mediated overexpression of Bcl-2 in a wide variety of tumor cell lines has been shown to confer marked resistance to the cytotoxic actions of essentially every type of anticancer drug presently available, as well as γ irradiation [reviewed in ( 3 )]. High levels of Bcl-2 have also been associated with progression to hormone independence in prostate cancers ( 4 , 5 ), presumably reflecting the consequences of blocking the apoptotic cell death that would normally ensue upon androgen deprivation in prostatic epithelial cells. Moreover, the relationship between Bcl-2 and chemoresistance has been borne out by clinical correlative studies showing that elevated expression of Bcl-2 can be associated with shorter survival and other indicators of worse clinical outcome in patients with at least some types of cancer, including aggressive non-Hodgkin's lymphomas, acute myelogenous leukemias, and adenocarcinomas of the prostate ( 6 – 13 ). Despite more than 10 years of effort, the biochemical mechanism by which the Bcl-2 protein prevents apoptosis remains enigmatic. Recent structural and functional studies ( 14 – 16 ) suggest that Bcl-2 may be a channel protein that regulates the transport of ions or proteins across the intracellular membranes where Bcl-2 resides, i.e., the outer mitochondrial membrane, nuclear envelope, and endoplasmic reticulum. However, Bcl-2 clearly has other functions related to its ability to interact physically with other proteins, including other members of the Bcl-2 family and a variety of unrelated proteins, which include a protein kinase (Raf-1), a phosphatase (calcineurin), a proteaseinteracting protein (CED-4), and a p53-binding protein (p53- BP2) [reviewed in ( 17 )]. In part because of confusion over the biochemical mechanism of Bcl-2, approaches for interfering with Bcl-2 that do not depend on such knowledge have been sought by academicians and pharmaceutical companies alike. Enter antisense oligonucleotides—short, synthetic, singlestranded DNA molecules that theoretically can interfere with gene expression via heteroduplex formation with complementary sequences within target messenger RNAs (mRNAs). Antisense oligonucleotides were first applied against Bcl-2 in 1990, showing in a human leukemia cell line that antisense-mediated reductions in Bcl-2 protein levels could promote cell death and reduce cell growth in vitro ( 18 ). Since then, similar types of experiments have been performed, and similar results were obtained with the use of either natural phosphodiester oligonucleotides or nuclease-resistant phosphorothioate-based oligonucleotides in freshly isolated acute myelogenous leukemia cells and in established non-Hodgkin lymphoma and prostate cancer cell lines ( 19 – 21 ). In those studies, antisense-mediated reductions in Bcl-2 protein levels were shown to sensitize malignant cells to traditional chemotherapeutic agents. Moreover, antisense Bcl-2 oligonucleotides have been successfully used in a SCID mouse model (i.e., mouse strain with severe combined immunodeficiency) for eradicating t(14;18)-containing human B-cell lymphoma cells in vivo under conditions comparable to minimalresidual- disease situations ( 22 ). Although demonstrating what appear to be sequence-specific antisense actions, applications of antisense technology to Bcl-2 have suffered from poor potency. This problem presumably can be attributed to two issues: 1) uptake and intracellular compartmentalization and 2) accessibility of the antisense oligonucleotide to its intended hybridization site on target Bcl-2 mRNA molecules. In this issue of the Journal, Ziegler et al. ( 23 ) report on the significant strides that they have made toward reducing the latter problem. Their approach entailed randomly choosing 13 different sites along the Bcl-2 mRNA for antisense attack and then synthesizing and testing the corresponding 20-mer antisense phosphorothioate oligonucleotides for inhibition of Bcl-2 production and tumor cell growth and survival. Among these 13 antisense compounds, one exhibited clearly superior activity, with an IC 50 (i.e., concentration necessary to inhibit cell growth by 50%) of probably 0.1 µ M or lower for inhibition of tumor cell line growth in vitro. The somewhat surprising observation was that this most potent of the 13 oligomers tested recognized a hybridization site within the coding region of the Bcl-2 mRNA. Although we tend to think of mRNAs as linear molecules, they are anything but that. All RNAs are highly structured molecules in cells, usually as a result of extensive intramolecular hybridization of complementary regions. Previous applications of Bcl-2 antisense approaches have relied on empirical targeting of oligonucleotides against the translation-initiation region of mRNAs, where the ATG start codon lies. The rationale for choosing this site is that mRNAs are touted to be single stranded in this region, and thus this area should be assessable for antisense oligonucleotide hybridization. In fact, however, this may not be the case for many mRNAs, including Bcl-2—particularly since the Bcl-2 gene can be regulated at a translational level under some circumstances ( 24 ). At present, we know very little about the structures of all but a few RNAs (such as the transfer RNAs), and computer-based prediction models have proved virtually worthless to date. Moreover, the longer the mRNA, the less accurate the predictions—a real problem when one is dealing with Bcl-2 mRNAs, which range from approximately 5 to 8 kilobases in length. What the results of Ziegler et al. indirectly suggest, therefore, is that a single-stranded area resides within Bcl-2 mRNA molecules near the region corresponding to codons 141-147. Head-to-head comparisons of this coding-region antisense oligonucleotide with the traditional translation start-site oligonucleotides revealed that its potency is perhaps as much as a log greater. This approach of empirically testing multiple oligonucleotides is financially unthinkable for most academic researchers, but it has become routine for pharmaceutical companies ( 25 ). In fact, with an mRNA target the length of Bcl-2, the examination of a mere 13 oligomers represents only the start of what ought to be done. Nevertheless, Ziegler et al. have identified a site that deserves further exploration through testing of additional antisense oligonucleotides targeted against this general area in the coding region of Bcl-2 mRNAs. Although making advances on how to deal with solving the problem of optimal site selection, the study described by Ziegler et al. in this issue does nothing about the issue of oligonucleotide uptake and intracellular compartmentalization. Natural phosphodiester oligonucleotides and nuclease-resistant phosphorothioate oligonucleotides are polyanionic compounds that, by all rights, should not cross biologic membranes. Indeed, the studies by Stein and others have provided abundant evidence that most of the antisense material that enters cells probably never escapes the endosomal compartment [see, for example, ( 26 )]. Ziegler et al. ( 23 ) used the now-popular trick of delivering their antisense compounds into tumor cells with cationic lipids—an approach that is inappropriate for systemic delivery of antisense oligonucleotides, at least with currently available liposome technology, but that may nevertheless have merit for local delivery (intracranial, intrabronchial, or intraperitoneal). One caveat should be raised about the data of Ziegler et al. which imply that an oligomer targeting the 141-147 coding region was more potent because of the site it targeted on Bcl-2 mRNAs. Given that measurements were not performed of the relative amounts of antisense phosphorothioates actually delivered into effective compartments within tumor cells, we cannot exclude the unlikely possibility that this particular antisense oligonucleotide was more efficacious because of alterations in its intracellular trafficking behavior. Although substantial efforts have been made to improve the membrane penetrability of synthetic antisense oligonucleotides through novel chemical modifications, most changes that enhance uptake simultaneously reduce hybridization efficiency or abrogate the capacity of the compound to initiate an RNase H-like degradation of the target mRNA and thus have proved unsuccessful. Nevertheless, rumors abound that the issue of uptake and compartmentalization is less of a problem when antisense phosphorothioates are used in vivo, and thus the potency of this new Bcl-2 antisense phosphorothioate molecule deserves evaluation in appropriate animal models. An additional accomplishment of the study by Ziegler et al. is that it provides the first proof of concept data that Bcl-2 expression may be critical for maintaining the survival of small-cell lung cancer cells. It has been estimated that approximately 90% of these incurable tumors express Bcl-2 at high levels, but expression does not necessarily equate with necessity. These new data thus lend additional weight to the choice of Bcl-2 as a drug-discovery target for this particular type of cancer and add it to the growing list of such cancers that include to date at least subgroups of Bcl-2-expressing non-Hodgkin's lymphomas, acute myelogenous leukemias, prostate cancers, and breast cancers ( 19 – 21 , 27 , 28 ). It is important to remember, however, that a necessity for Bcl-2 for tumor cell survival in vitro does not guarantee a requirement for Bcl-2 in vivo; thus, the in vitro studies need to be confirmed in vivo by use of animal models. In this regard, Walker et al. ( 29 ) have provided evidence that the local tumor microenvironment can provide cell survival signals, via growth factors and probably through cell adhesion mediated by integrins, which collaborate with Bcl-2 overexpression to protect cells from apoptosis induced by chemotherapeutic drugs. In some cases, these environmental signals may result in increased expression of other anti-apoptotic members of the Bcl-2 family, such as Bcl-X L , that are impervious to Bcl-2 antisense oligonucleotides because of nucleotide sequence differences. Clearly, therefore, caution is warranted where interpretation of the potential clinical significance of these observations is concerned. Moreover, it deserves emphasis that, in some instances, in which an antisense-mediated reduction in Bcl-2 protein levels was by itself inadequate for inducing tumor cell apoptosis, it did nevertheless increase sensitivity to chemotherapeutic drugs ( 19 , 21 , 27 , 28 ). Bcl-2 gene knockout experiments in mice ( 30 , 31 ) have revealed the same phenomenon where sensitivity to radiation is concerned. The potential need, therefore, to combine Bcl-2 antisense therapy with traditional chemotherapy or radiotherapy creates additional complications for clinical trial design but should not be an insurmountable problem. Within the past year, the first phase I trial of a Bcl-2 antisense phosphorothioate oligonucleotide was conducted by Webb et al. ( 32 ) on patients with relapsed and refractory lymphomas. The oligomer employed for that trial targeted the translationinitiation region of Bcl-2 mRNAs, which probably was not the optimal site based on the new findings reported by Ziegler et al. in this issue of the Journal. Moreover, the phase I format dictated that the antisense compound was delivered alone, without traditional chemotherapy. Nevertheless, evidence of in vivo decreases in Bcl-2 protein levels within the tumors of some patients was obtained, and antitumor responses were observed in two of the nine patients. Toxic side effects were minimal. A major issue in any such study is whether the effects seen were truly obtained through an antisense mechanism or were obtained through other effects—particularly given the tendency of phosphorothioates to interact with and modify the functions of proteins and their potential for modulating immune cell function through sequence-specific but non-antisense mechanisms ( 33 – 36 ). If we have learned anything in the past decade of working with phosphorothioates, it is that the two cardinal rules of pharmacology clearly apply when interpreting any results obtained with these compounds: 1) No drug has a single mechanism of action, and 2) enough of anything will block anything. Despite these caveats, the very preliminary phase I results provide encouragement that Bcl-2 antisense therapeutics may eventually find a role in the practice of clinical oncology. Further optimization and clinical testing of Bcl-2 antisense compounds therefore are warranted.