Expression and Prognostic Significance of IAP-Family Genes in Human Cancers and Myeloid Leukemias¹

Suppression of apoptosis contributes to carcinogenesis by several mechanisms, including prolonging cell lifespan (thus facilitating the accumulation of gene mutations), permitting growth factor-independent cell survival, promoting resistance to immune-based cytotoxicity, and allowing disobeyance of cell cycle checkpoints that would normally induce apoptosis (1, 2, 3, 4). Defects in apoptotic mechanisms also play an important role in resistance to chemotherapy and radiation (1).

The IAPs⁴ are a family of antiapoptotic proteins that are conserved across evolution, with homologues found in both vertebrate and invertebrate animal species (5). The baculovirus IAPs, Cp-IAP and Op-IAP, were the first members of this family to be identified based on their ability to functionally complement defects in the cell death inhibitor, p35, a baculovirus protein that binds to and inhibits caspase-family cell death proteases (6, 7). Subsequently, five human (XIAP,cIAP1, cIAP2, NAIP, and Survivin) and two Drosophila IAP homologues have been identified, which have been demonstrated to inhibit cell death (4, 8, 9, 10, 11, 12, 13, 14). The human IAPs, XIAP,cIAP1, and cIAP2, have been reported to bind and potently inhibit caspase-3 and caspase-7 with Kis in the range of 0.2–10 nm (15, 16). These IAP-inhibitable caspases operate in the distal portions of apoptotic protease cascades,functioning typically as effectors rather than initiators of apoptosis (17, 18, 19). Although quantitative studies are lacking, the IAP-family member Survivin also binds and inhibits some effector caspases (14, 20). Moreover, at least some IAPs, such as XIAP, are capable of binding and suppressing specific initiator caspases such as caspase-9, the pinnacle caspase in the cytochrome c/mitochondrial pathway for apoptosis (21).

The common structural feature of all IAP family members is an ∼70 amino acid zinc-binding fold termed the BIR domain, which is present in one to three copies (22). Using a mutagenesis approach, we showed previously that the second of the three BIR domains (BIR2) of XIAP is necessary and sufficient for inhibiting the effector caspase-3 and caspase-7 (23), implying that a single BIR domain can possess antiapoptotic activity. However, for caspase-9 suppression by XIAP, the third BIR domain is required (24).

To learn more about the importance of IAPs in cancer and leukemia, we examined the expression of several members of the IAP family in the well-characterized NCI panel of 60 human tumor cell lines, correlating their expression at either the mRNA or protein levels with other tumor-related genes and with in vitro chemosensitivity data for 30,000 compounds. In addition, the prognostic significance of one of the IAPs, XIAP, was examined in AML, revealing correlations with clinical outcome, thus suggesting that analysis of IAP-family proteins may provide predictive information about responses to chemotherapy and survival for at least some subgroups of patients with AML.

Patient-derived AML samples (n = 97) were randomly selected from a large assortment of patient samples collected,Ficoll-purified, and cryopreserved between 1983 and 1995. Only samples with >15% blasts were selected for further analysis(n = 78) to avoid contributions from contaminating normal cells. The median follow-up of survivors is nearly 5 years. Clinical characteristics of these patients are presented below.

Immunoblotting of AML patient samples derived from 78 newly diagnosed patients with AML and 10 normal individuals was carried out using cell lysates from mononuclear fractions of peripheral blood generated by Ficoll separation. Whole-cell lysates from 5 × 10⁵cells were electrophoresed through 8–14% SDS-PAGE gradient gels and electroblotted to Immobilon polyvinylidene difluoride membranes(Millipore, Bedford, MA) using a semi-dry transfer apparatus at 0.8 mA/cm² for 1.5 h (25). Each gel included a XIAP-expressing positive control cell line (HeLa), two to three peripheral blood mononuclear cell samples from normal individuals, and molecular weight markers. The membranes were blocked in Tris-buffered saline with 0.05% Tween 20 (TBST) and 3% nonfat dry milk (Blotto) at 4°C for 4 h and then exposed overnight to a mouse anti-XIAP monoclonal antibody (Pharmingen, San Diego, CA) at a 1:350 dilution at 4°C overnight. Subsequently, the membranes were washed twice in Blotto, exposed to sheep antimouse IgG conjugated to horseradish peroxidase (1:4000) for 1 h, washed in Blotto and TBST, and then exposed to SuperSignal West Pico substrate chemiluminescence mixture for 1 min according to the directions of the manufacturer (Pierce Corp., Rockford, IL). Images were collected on a ChemiImager 4400 system (Alpha Innotech, San Leandro, CA), and densitometry was performed using the best image. To normalize for variation in antibody concentration or time of exposure, the XIAP signal from the patient was normalized against the XIAP signal of the control cell line HeLa. Results are expressed in terms of this ratio.

Detergent lysates were prepared in the presence of protease inhibitors from established tumor cell lines essentially as described (26). After normalization for total protein content (50μg/lane), samples were subjected to SDS-PAGE/immunoblot analysis using monoclonal anti-IAP antibodies specific for XIAP (Transduction Laboratories, Lexington, KY) or for cIAP1 and cIAP2 (R&D Systems,Minneapolis, MN). Data on X-ray films were quantified by scanning densitometry using the IS-1000 image analysis system (Alpha Innotech Co.), and the results from a standard curve generated using purified recombinant GST-XIAP, GST-cIAP1, or GST-cIAP2 protein were used to estimate the amounts (ng) of XIAP, cIAP1, and cIAP2 protein per 50 μg of total protein. Data from two independent protein standard-containing blots were within 10% agreement.

Total RNA was isolated from the NCI 60 tumor cell line panel or freshly isolated AML cells using the RNeasy Mini kit (Qiagen, Germany),according to the manufacturer’s manual. After normalization for total mRNA content (12 μg/lane), samples were subjected to RPA, as described in the manufacturer’s manual (Riboquant; Pharmingen, La Jolla, CA). Briefly, a multiprobe was used for the T7 polymerase-directed synthesis of a ³²P-labeled antisense RNA probe set. The probe set was hybridized in excess to target RNA in solution, after which free-probe and other single-stranded RNA were digested with RNases. The remaining “RNase-protected” probes were purified, resolved on denaturing polyacrylamide gels (4.75%), and quantified by autoradiography and phosphorimaging (Bio-Rad Laboratories, Inc.; Molecular Analyst, Version 2.1). The quantity of each mRNA species in the original RNA sample was then determined based on the intensity of the appropriately sized, protected probe fragment relative to the loading control (GAPDH).

Comparison of IAP mRNA and protein levels with chemosensitivity data for the 60 cell line panel (NCI) resulted in a rank order of drugs from the standard agent database of the NCI. Pearson correlation coefficients and two-tail Ps were used to assign possible significance to these data. P ≤ 0.005 was considered significant in all analyses.

For analysis of patient data, results were rank-ordered according to XIAP protein levels and divided into thirds, thus assessing possible threshold effects of XIAP on survival while avoiding searching for optimal cutpoints for levels of XIAP (27). The top two-thirds demonstrated similar outcomes, and therefore those arms were collapsed. Unadjusted survival analyses were performed using Kaplan-Meier plots (28), and comparisons of survival between patient subgroups were made using the log-rank test (29). For comparisons of clinical variables between patients within the lowest third of XIAP expression versusthe top two-thirds, either Student’s t test orχ ² analysis was performed. All computations were performed using Statistica version 5.1 M (StatSoft Inc., Tulsa, OK).

We analyzed the levels of XIAP, NAIP, cIAP1, and cIAP2 mRNAs in the NCI 60 tumor cell line panel by RNase Protection Assay (RPA). Examples are shown in Fig. 1. GAPDH was included as a control to ensure equal loading. Data on X-ray films were quantified by scanning densitometry (Fig. 2). The mRNAs for XIAP and cIAP1 were detectable at variable levels in all cell lines analyzed (Fig. 2, A and B). In contrast, cIAP2 was more restricted in its expression. cIAP2 mRNA was abundant, for example, in most CNS and many renal tumor cell lines but almost undetectable in ovarian and breast cancer lines (Fig. 2 C). Thus, in contrast to XIAP and cIAP1 mRNAs that were present in all tumor lines, the presence of cIAP2 mRNA was detectable in only 34(56%) of the 60 tumor cell lines within the NCI screening panel. NAIP mRNA was undetectable in the entire panel of 60 tumor cell lines but was plentiful in some normal tissues, such as peripheral blood lymphocytes (not shown).

To compare the relative levels of IAP mRNAs and proteins in the 60 cell line panel, we analyzed the relative levels of XIAP, cIAP1, and cIAP2 proteins by immunoblot assay using monoclonal antibodies. Data on X-ray films were quantified by scanning densitometry and compared with results obtained from defined amounts of recombinant protein (standard curve) to estimate the amount (ng) of IAP protein per 50 μg of total protein. Examples of immunoblot data are shown in Fig. 3,A. Higher levels of XIAP protein were generally found in renal cancer and melanoma cell lines,whereas lower levels of XIAP were typically present in CNS tumor cell lines (Fig. 3,B). cIAP1 protein was expressed at higher levels in colon cancers, whereas lower levels were found in melanoma cell lines Fig. 3.. The expression of cIAP2 was far more restricted; cIAP2 protein was detectable in only 5 (8%) of the 60 tumor cell lines analyzed (MCF7, HT29, A549, NCI H322M, and K562). Thus, protein data for cIAP2 did not correlate with mRNA data.

Similarly, XIAP and cIAP1 protein levels did not correlate with mRNA levels. For example, XIAP mRNA levels were above the mean for the NCI 60 cell screening panel in the leukemia lines HL60 and RPMI-8226, the lung cancer cell lines NCI-H322M and NCI-H460, the colon lines COLO 205 and HCC-2998, the brain tumor lines SF-539 and SNB-75, and the breast cancer lines MCF7 and HS 578T, whereas XIAP protein levels were below the mean in these same tumor lines (compare Figs. 2,A and 3 B). Conversely, XIAP mRNA levels were below the mean of tumor cell lines in the leukemia lines K562 and MOLT4, the lung cancer line HOP-62, the colon cancer line KM12, the melanoma cell lines LOX-IMVI, MALME-3M, and M14, the ovarian line OVCAR-3, the renal cancer lines 786-0, ACHN, CAKI-1, and TK10, the prostate cancer PC3, and the breast cancer line MDA-MB 231, whereas XIAP protein levels were above the mean within these same tumor cell lines.

XIAP and cIAP1 protein levels were correlated with cytotoxicity data for compounds tested against the NCI 60 cell screening panel. Because cIAP-2 protein was found in only five cell lines, it was excluded from statistical analysis. Several nucleoside DNA chain-terminating drugs were positively correlated with XIAP protein levels, suggesting that drug sensitivity may be associated with expression of this protein(Table 1A). The most significant correlation was between XIAP protein levels and in vitrochemosensitivity to cytarabine, indicating a possible role for XIAP in conferring sensitivity to this agent in vitro (Pearson correlation coefficient, 0.44; P (two-tail) < 0.0006). For cIAP1, the most significant correlation was seen between cIAP1 protein levels and in vitro chemoresistance to carboplatin(P < 0.002) and cisplatin (P < 0.002;Table 1B). Significant associations between higher levels of cIAP1 protein and resistance to several DNA alkylating agents and the topoisomerase inhibitor VP-16 were also observed (Table 1B).⁵ No significant correlations with drug cytotoxicity data and IAP-family mRNA levels were found.

A comparison of the relative levels of IAP-family mRNAs and proteins with the expression of more than 30 genes evaluated previously in the NCI 60 tumor cell line screening panel failed to reveal any significant correlations (Pearson, −0.10). This search for correlations included p53, Bcl-2, Jun, Fos, Ras, and Rb.⁶

The observed correlation of higher levels of XIAP protein with greater sensitivity to nucleoside analogue drugs was paradoxical in terms of a priori expectations based on gene transfection evidence that overexpression of XIAP protects tumor cell lines from apoptosis induced by various anticancer drugs (reviewed in Refs. 30, and 31). Because XIAP protein levels showed the greatest correlation with cytarabine (AraC), we explored the relation of XIAP expression to in vivo drug responses in a clinical context where AraC-based therapy is standardly used, i.e., in patients with AML (32). All patients were treated with high-dose AraC-containing regimen. The relative levels of XIAP protein were compared with clinical responses to chemotherapy, consisting of high-dose AraC plus idarubicin, using leukemia cell samples derived from 78 AML patients. Expression of XIAP protein was detected by immunoblotting in 76 of 78 samples tested; however, the level of XIAP was very low (<5% of the control cell line) in 20 cases. The range of expression of XIAP was heterogeneous across all French-American-British and cytogenetic categories, except for promyelocytic leukemia where there was a narrow range of expression (not shown). Bcl-2, Bax,caspase-2, and caspase-3 expression have been measured previously for most of these patients (25, 33, 34). No significant correlation between expression of these other apoptosis-related proteins and XIAP protein levels was found (data not shown).

To assess possible threshold effects of XIAP protein on survival while avoiding searching for optimal cutpoints for expression, results were rank-ordered and then divided into thirds. Because data derived from patients with XIAP levels in the top two-thirds were similar (not shown), they were collapsed into one arm for further analysis and comparison. Patients with lower XIAP levels had significantly longer survival (median survival, 133 versus 52.5 weeks; P = 0.05) and a tendency toward longer median remission duration (87 versus 52.5 weeks; P = 0.13)than those with higher levels of XIAP. (Table 2 and Figure 4). For example, 42% of the patients with low XIAP expression are alive today, as compared with 23% with high XIAP expression (P = 0.004; Table 2). Patients with low levels of XIAP had only a slightly higher complete remission rate with induction chemotherapy (73% versus 65%; P = 0.42) but a lower relapse rate (42%versus 65%; P = 0.17) compared with patients with XIAP levels in the top two-thirds (Table 2). However,neither the complete remission rate nor relapse rate data reached statistical significance. No statistically significant differences were noted between the XIAP high and low groups with respect to age,cytogenetics, frequency of an antecedent hematological disorder, or gender; however, patients with higher XIAP were more likely to have poor Zubrod performance status (Table 2).

From this evaluation, we conclude that higher levels of XIAP protein tend to be associated with shorter remission durations and shorter overall survival in AML patients treated with AraC-containing regimens. Thus, these findings are at odds with in vitro correlations derived from analysis of the NCI 60 cell screening panel and are fitting with expectations based on knowledge of the antiapoptotic activity of XIAP in cells.

In this study, we show that expression of several members of the human IAP family of antiapoptotic genes is differentially regulated within the NCI panel of 60 human tumor cell lines and in freshly isolated human leukemia samples. XIAP and cIAP1 were widely expressed in tumor cell lines. However, the relative levels of these IAP-family members were variable among the tumor lines evaluated, implying differential regulation of the expression of these antiapoptotic genes within cancers. In contrast to XIAP and cIAP1, the expression of cIAP2 appears to be more restricted because cIAP2 mRNA was detectable in only approximately half of the analyzed tumor cell lines, and cIAP2 protein was found at detectable levels in only 5 (8%) of tumor cell lines. In comparison, NAIP mRNA was undetectable in all tumor cell lines tested. Thus, this particular member of the IAP family appears not be to a major participant in solid tumors. The IAP-family member Survivin has been shown previously to be widely expressed in the tumor cell lines of the NCI 60 cell line screening panel (14).

Interestingly, IAP mRNA and protein levels were commonly noncongruent,suggesting the likelihood of posttranscriptional regulation of the expression of these antiapoptotic genes, perhaps through translational mechanisms or by differential rates of protein turnover. In this regard, evidence of translational control of XIAP has been reported recently, revealing the presence of an internal ribosome entry site within the XIAP mRNA (35). This lack of correlation between XIAP, cIAP1, and cIAP2 has important implications with respect to any future attempts to use cDNA arrays or related technologies for assessing IAP-family gene expression in cancers. The finding that cIAP1 protein but not cIAP1 mRNA levels were associated with resistance to several anticancer drugs among the tumor cell lines of the NCI 60 cell screening panel lends additional support to the argument that measurements of protein and not mRNA levels are critical to understanding the role of this family of antiapoptotic genes in cancer.

The variability in IAP protein levels seen here among the NCI’s 60 human tumor cell lines suggests that further correlative studies of IAP expression with clinical outcome and with other biomarkers should be informative with regards to assessing the prognostic significance of these antiapoptotic proteins. Moreover, the lack of correlation of IAPs with other known tumor-related genes assessed previously in the NCI 60 cell panel (e.g., c-Jun, Ras, Fos, p53, and Rb) raises the possibility that IAPs could serve as independent risk factors for some types of malignancies, as suggested recently for Survivin (11, 36, 37).

The paradoxical association of XIAP protein levels with sensitivity rather than resistance to AraC and some other nucleoside analogues among the NCI panel of 60 tumor cell lines remains unexplained. We might speculate that a fortuitous correlation exists between XIAP and levels of other proteins that might sensitize cells to such agents,such as nucleotide kinases and phosphatases. Moreover, the correlation between XIAP levels and sensitivity of cell lines to various drugs in vitro does not take into account the clinical usefulness of this approach, e.g., AraC is not the cytotoxic agent of choice for solid tumors. Given the overwhelming evidence that XIAP is a caspase-inhibiting, antiapoptotic protein (reviewed in Refs. 30, and 31), this observation illustrates the difficulty in drawing conclusions about protein function from correlative approaches. In this regard, the NCI 60 tumor cell line panel was assembled originally for the purpose of identifying compounds with potential antitumor activity (38). In recent years, however, attempts have been made to use bioinformatics and genomics technologies to reveal associations between cytotoxic responses of cancer cell lines to compounds and various biomarkers. Future efforts of this type may ultimately provide a molecular explanation for the paradoxical correlation of XIAP protein with greater sensitivity to nucleoside agents.

Although little attempt has been made thus far to compare the levels of IAPs in tumors with other biomarkers or with clinical outcome, it has been reported that Survivin expression in neuroblastomas correlates with clinically more aggressive, histologically unfavorable disease (37). Moreover, higher levels of Survivin protein as determined by immunostaining and p53 accumulation (indicative of mutant p53) were positively correlated in a survey of gastric cancers (36), implying an association of Survivin with more aggressive disease.

Prompted by an unexpected inverse correlation between XIAP protein levels and resistance to AraC in the NCI 60 cell screening panel, we compared the expression of this IAP-family member in AML blasts derived from newly diagnosed, untreated patients, making correlations with clinical outcome. Patients with low levels of XIAP (lowest third)enjoyed a significantly longer survival and tended to have longer median remission durations. These data suggest that XIAP may have prognostic potential in AML, a finding that should be evaluated in additional studies involving larger numbers of AML patients and extended to other types of cancer. Moreover, these in vivodata are fitting with expectations based on knowledge of the antiapoptotic activity of XIAP in cells.

These results thus warrant further research on both the functions and prognostic relevance of XIAP and other IAP-family genes in cancer and leukemia.

We thank members of the laboratory and Zeev Estrov for helpful discussions, Sunil Patel for aid with immunoblot analysis, and Tim Myers and Ed Sausville for aid with statistical analysis of the cell line data.