Abstract Background Inhibitors of apoptosis proteins (IAPs) are a family of antiapoptotic proteins modulating cell cycle, signal transduction and apoptosis. Dysregulated IAPs have been reported to contribute to tumor progression and chemoresistance in various cancers. However, existing studies were sporadic and only focus on one specific cancer with one particular gene in the IAPs family. A systematic investigation on the co-expression pattern, regulation frameworks on various pathways, prognostic utility on patient outcomes, and predictive value on drug sensitivity among all the IAPs across multiple tumor types was lacking. Methods Leveraging The Cancer Genome Atlas data with comprehensive genomic characterizations on 9714 patients across 32 tumor types and the Genomics of Drug Sensitivity in Cancer data with both genomic characterizations and drug sensitivity data on > 1000 cell lines, we investigated the co-expression pattern of IAPs, their regulations of apoptosis as well as other pathways and clinical relevance of IAPs for therapeutics development. Results We discovered diverse expression pattern among IAPs, varied spectrum of apoptosis regulations through IAPs and extensive regulations beyond apoptosis involving immune response, cell cycle, gene expression and DNA damage repair. Importantly, IAPs were strong prognostic factors for patient survival and tumor stage in several tumor types including brain, liver, kidney, breast and lung cancer. Further, several IAPs were found to be predictive of sensitivity to BCL-2 inhibitors (BIRC3, BIRC5, BIRC6, and BIRC7) as well as RIPK1 inhibitors (BIRC3 and BIRC6). Conclusion Together, our work revealed the landscape of regulations, prognostic utilities and therapeutic relevance of IAPs across multiple tumor types. Keywords: Drug sensitivity, Gene expression regulation, Inhibitors of apoptosis proteins, Personalized cancer therapy, Prognostic factors Background Inhibitors of apoptosis proteins (IAPs) are a family of regulators controlling multiple biological pathways involving various physiologic and pathologic conditions. To date, eight mammalian IAPs have been identified: BIRC1 (NAIP/NLRB), BIRC2 (cellular IAP1/cIAP1/human IAP2), BIRC3 (cellular IAP2/cIAP2/human IAP1), BIRC4 (X-linked IAP/XIAP), BIRC5 (survivin), BIRC6 (apollon/BRUCE), BIRC7 (livin/melanoma-IAP, also called ML-IAP/KIAP), and BIRC8 (testis-specific IAP/Ts-IAP/hILP-2) [[37]1]. The defining characteristic of IAPs is the existence of the BIR (baculovirus IAP repeat) domain. In addition to the BIR domain, IAPs also contain other important domains including RING (C-terminal Ring zinc-finger domain), CARD (caspase recruitment domain) and UBC (C-terminal ubiquitin-conjugating domain) as depicted in Additional file [38]1: Figure S1. IAPs impose apoptosis regulation through three well-characterized apoptosis pathways: (1) the extrinsic pathway, also known as the death receptor pathway, which activates caspase-8 and further propagates the apoptosis signal to effector caspases including caspase-3 upon the stimulation of death receptors; (2) the intrinsic pathway of apoptosis, also called the mitochondrial pathway, which starts with the release of apoptogenic factors such as cytochrome c into the cytosol triggering activation of caspase-3; and (3) the execution phase of apoptosis, which controls the breakdown of cells leading to cell death. The functions of IAPs in apoptosis have been wildly investigated in human cancers as therapeutic targets for anti-cancer drug development. Besides the roles in apoptosis regulation, growing evidence has demonstrated that IAPs are more than just inhibitors of apoptosis proteins and they can regulate various biological processes beyond apoptosis including immune response, cellular stress, translation, transcription, cell proliferation, differentiation, motility and signal transduction [[39]2]. Futher, recent studies also found that IAPs have opposing roles in cancer serving as both oncogenes and tumor suppressors [[40]3]. However, these studies were sporadic and sometimes contradictory. A systematic investigation of the complexity of apoptosis regulations by IAPs and their therapeutic relevance across various human cancers would provide a consolidated view and thus be invaluable for deepening our understandings. The Cancer Genome Atlas (TCGA) is a large collection of human tumor data containing nearly 10,000 tumor patients from 32 different tumor types [[41]4]. TCGA serves as an open repository of large-scale genome sequencing and multi-modal molecular profilings of human cancers. In addition to the TCGA patient data, the Genomics of Drug Sensitivity in Cancer (GDSC) database further provides large scale molecular characterization and drug profiling data for more than a thousand cell lines [[42]5]. The emergence of such large scale genomic data for both patient and cancer cell lines greatly facilites multiple integrated genomic investigations across multiple cancers [[43]6]. In this work, we leveraged the TCGA data to comprehensively characterize the global expression pattern of 7 IAPs across 32 types of human cancers. We investigated the co-expression pattern among IAPs demonstrating both similarity and differences in terms of expression dynamics, which served as the foundation of differential IAPs regulation in various pathways. We further identified several candidate miRNAs regulating IAPs, some of which had been experimentally validated in previous studies. For clinical translations, we found IAPs were strong prognostic markers for patient survival and tumor stage, with BIRC5 being the most potent biomarker for patient survival in multiple tumor types. Additionally, we found that IAPs predicted chemosensitivity of several anti-cancer drugs, further demonstrating the great potential of IAPs for therapeutics development. Results Global expression patterns among IAPs As a gene family, members of IAPs are characterized by the presence of one to three BIR domains and optionally a RING and CARD domain [[44]7]. Despite of the similarity of gene structure, the gene expression pattern of IAPs as a whole is rarely studied. To systematically investigate the co-expression pattern of IAPs which forms the basis of orchestrated apoptosis regulation, we employed hierarchical clustering of these genes across 32 human cancers consisting of 9714 tumor samples (data summary provided in Additional file [45]6: Table S1) and catalogued the frequency of any gene pair being clustered together. We observed that except BIRC5 and BIRC7, there were broad expression similarity among IAPs (Fig. [46]1). For example, BIRC2 was clustered with BIRC3, XIAP and BIRC6 in more than 50% of the cancers. The strongest coherence was observed between XIAP and BIRC6 which were clustered together in 97% of the cancers. This was further confirmed by Spearman’s rank correlation analysis which quantified the actual strength of co-expression. XIAP and BIRC6 was strongly co-expressed with an average correlation over 0.5 (Fig. [47]1d and e). There was moderate co-expression among the NAIP-BIRC2-BIRC3-XIAP tetrad. Fig. 1. [48]Fig. 1 [49]Open in a new tab Global Expression Pattern of IAPs. The 7 IAPs have distinct expression pattern across 32 cancer types. Using GBM as an example, hierarchical clustering (b) using Ward’s linkage and correlation-based distance identified 2 clusters based on the criteria of maximum average Silhouette width (a). c Divergent correlation among the 7 IAPs in GBM. d Example association between XIAP and BIRC6 among 32 cancer types. All these correlations are statistically significant with p < 0.001. e A global view of the similarity of IAP expression patterns. The upper right triangle shows the frequency of any pair of IAPs being clustered together among 32 cancers. The lower left triangle shows the average correlation between any 2 pair of IAPs. Circle size indicate magnitude (frequency or correlation). Color indicates correlation direction, with red for positive correlation, blue for negative correlation and values indicated in the color legend. All correlations are computed using Spearman rank correlation Unlike other IAPs, BIRC5 and BIRC7 had weak or even negative correlation with others (Fig. [50]1e). For example, BIRC5 showed moderate anti-correlation with XIAP, NAIP, and BIRC6 while BIRC7 showed weak anti-correlation with BIRC6 (Additional file [51]2: Figure S2A, Fig. [52]1e). Between BIRC5 and BIRC7, both positive and negative correlations were observed across the 32 cancers, demonstrating a histology dependent co-expression pattern (Additional file [53]2: Figure S2B). Our global co-expression analysis of IAPs revealed distinct expression pattern of BIRC5 and BIRC7 as well as the synchronized expression for NAIP, BIRC2, BIRC3, XIAP and BIRC6. The demonstrated complexities of IAPs at the expression level provided immense possibilities for apoptosis regulation and potential mechanisms of oncogenesis from aberrant expression of IAPs. Spectrum of apoptosis regulations by IAPs IAPs are best known for their ability to regulate apoptosis. In an effort to systematically investigate the spectrum of apoptosis regulations by IAPs across various human cancers, we performed gene set enrichment analysis [[54]8] using high quality pathways manually curated by experts provided by the Reactome database [[55]9]. A total of 25 genes in the extrinsic pathway of apoptosis, 36 genes in the intrinsic pathway of apoptosis, and 47 genes in the execution phase of apoptosis were extracted (Fig. [56]2). Genes in these apopotosis pathways are provided in Additional file [57]8: Table S3. Fig. 2. [58]Fig. 2 [59]Open in a new tab Spectrum of Apoptosis Regulations By IAPs. Gene set enrichment analysis identified different mechanisms for apoptosis regulations by different IAPs in different cancers (a) through three major apoptosis pathways including the extrinsic pathway of apoptosis (25 genes), the intrinsic pathway of apoptosis (36 genes) and the execution phase of apoptosis (47 genes) as depicted in (b). Percentage value on top of (a) summarized the proportion of cancers with at least one apoptosis pathways being regulated by a given IAP. The right hand bars on (a) summarized number of apoptosis pathways being regulated in each cancer. The intrinsic pathway of apoptosis (35.7%) as well as the extrinsic pathway of apoptosis (29.0%) were more frequently regulated by IAPs than the execution phase of apoptosis (18.3%) across all cancers. Here percentages were computed across all cancer-IAP regulation pairs. The frequency of apoptosis pathway regulations varied among different IAPs, with BIRC3 and BIRC6 both regulating 36.4% of cancers while BIRC7 only regulating 12.5% of cancers through at least one of the apoptosis pathways. Enriched pathways were identified with adjusted p value less than 0.05 Overall, the intrinsic pathway of apoptosis (35.7%) as well as the extrinsic pathway of apoptosis (29.0%) were more frequently regulated by IAPs than the execution phase of apoptosis (18.3%) across all cancers (proportion test P = 0.00018, Fig. [60]2a). The frequency of apoptosis pathway regulations varied among different IAPs, with BIRC3 and BIRC6 both regulating 36.4% of cancers while BIRC7 only regulating 12.5% of cancers in at least one of the apoptosis pathways (overall proportion test P = 0.0022). Mechanisms of apoptosis regulations through IAPs also differed. In particular, BIRC3 and NAIP mostly focused on the extrinsic pathway of apoptosis (average frequency of 59.4%) while BIRC6 and XIAP were heavily involved in the intrinsic pathway of apoptosis (average frequency of 48.4%). In contrast, there seemed to be a balanced distribution of the three different pathways for BIRC2 and BIRC5. The varied preferences of IAPs on regulating