Survivin study: What is the next wave?


  • Fengzhi Li

    Corresponding author
    1. Department of Pharmacology and Therapeutics, Grace Cancer Drug Center, Roswell Park Cancer Institute, Elm & Carlton Street, Buffalo, New York
    • Department of Pharmacology and Therapeutics, Roswell Park Cancer Institute, Elm & Carlton Street, Buffalo, New York.
    Search for more papers by this author


Survivin, a novel member of inhibitor of apoptosis (IAP) protein family, is aberrantly expressed in cancer but undetectable in normal, differentiated adult tissues. Current studies suggest that survivin is implicated in both control of apoptosis and regulation of cell division. However, due to some inconsistent observations on survivin subcellular localization, there is debate about survivin's function in regulating apoptosis, cell division, or both. This review will discuss concepts, experimental methods, and interesting results that unify the different notions about survivin localization and function or point out gaps of knowledge about controversial issues. The author also intends to review various aspects of survivin studies, which were not emphasized or sufficiently discussed in previous reviews on survivin, and update recent developments that may reveal new applications of disease-oriented therapeutics in the coming years. J. Cell. Physiol. 197: 8–29, 2003 © 2003 Wiley-Liss, Inc.

A new protein family, designated inhibitor of apoptosis (IAP), was recognized in recent years. The prototype IAP member was originally identified in baculoviral genomes in 1993 (Crook et al., 1993; Birnbaum et al., 1994). Subsequently, new IAP family members were found in yeast, nematodes, fruit flies, and humans [for more information, see recent reviews (Deveraux and Reed, 1999; Salvesen and Duckett, 2002)]. All IAP family members share one or more signature motifs termed baculovirus IAP repeat (BIR) which consists of a conserved sequence of about 70 amino acids. Eight human IAP family members have been identified so far. They are c-IAP1, c-IAP2 (Rothe et al., 1995), XIAP (Duckett et al., 1996), NAIP (Liston et al., 1996), survivin (Ambrosini et al., 1997), apollon (Chen et al., 1999), ML-IAP/livin (Vucic et al., 2000; Kasof and Gomes, 2001), and ILP-2 (Richter et al., 2001).

Among the eight human IAP members, survivin and ML-IAP/livin appear to be differentially expressed in cancer/transformed cells versus normal differentiated tissues. However, survivin has a number of distinct features that it does not share with ML-IAP/livin. These features are summarized below:

  • 1The overall protein identity of ML-IAP/livin is 34.7% similarity to XIAP but only 26.3% to survivin (Kasof and Gomes, 2001). Consistent with this fact, Smac/DIABLO exhibits strong binding to and inhibition of XIAP and ML-IAP/livin but shows much weaker binding (if any) to survivin (Du et al., 2000; Verhagen et al., 2000; Vucic et al., 2002). In this regard, the previous notion that survivin binds to Smac for replacing other IAP proteins in apoptosis inhibition is unlikely.
  • 2Similarly to c-IAP1, c-IAP2, XIAP, and ILP-2, ML-IAP has a RING finger motif at the carboxyl terminal of the protein. On the other hand, survivin has a long alpha-helix motif without a RING finger and is the only IAP member which homodimerizes in solution (Chantalat et al., 2000; Muchmore et al., 2000; Verdecia et al., 2000). The function of the RING finger motif in ML-IAP is not known, but the general role of this motif is for protein–protein interaction and has ubiquitin ligase (E3) activity to ubiqutinate other proteins as well as itself for degradation (Joazeiro and Weissman, 2000). The long alpha-helix motif at the carboxyl terminal of survivin is involved in interaction with polymerized microtubulin (Li et al., 1998, 1999). Consistent with these observations, ML-IAP does not appear to have similar localization to survivin (Vucic et al., 2000). Intriguingly, while survivin has no known E3 activity, survivin can be ubiquitinated and degraded by the proteasome [(Zhao et al., 2000) and Chen et al. in press]. Clearly, it will be valuable to identify the mechanism and the ubiquitination enzymes responsible for survivin degradation.
  • 3The expression of ML-IAP/livin appears to be predominantly restricted in melanoma and absent or low in other cancer tissue/cell types (Vucic et al., 2000; Kasof and Gomes, 2001) while survivin is ubiquitously expressed in all common human cancers (Table 1).
  • 4Survivin expression is cell cycle-regulated (a unique feature among the IAP family) with robust expression in G2/M phase while there is no apparent ML-IAP/livin to cell cycle regulation (Vucic et al., 2000; Kasof and Gomes, 2001).
  • 5Survivin appears to be involved in regulation of apoptosis as well as cell division (Li et al., 1998, 1999) (see below) while the known function of ML-IAP/livin is restricted to the regulation of apoptosis (Vucic et al., 2000; Kasof and Gomes, 2001).

In this article, efforts will focus on reviewing various aspects of survivin studies which were not emphasized or sufficiently discussed in previous reviews (Reed and Bischoff, 2000; Altieri, 2001), and update the recent new progress and debate about survivin studies without a detailed review of the basic information of survivin on gene and protein structure, expression, regulation, and function. However, this detailed basic information about survivin can be found from a previous review (Altieri et al., 1999). In order to comprehensively understand the function of survivin, the survivin-type IAP members from yeast, nematodes, and flies will be briefly reviewed and discussed as appropriate.


The accumulated data from the characterization of survivin expression in human cancer tissues in the past few years reveal an overwhelming consistent observation that the expression of survivin is enhanced in various human cancers in comparison with the adjacent normal tissues. Since the methods used in these studies are largely immunohistochemical and RT-PCR, it was found in some studies that the adjacent normal human tissues express survivin as well but at much lower amounts. Nevertheless, most data obtained from these studies suggest that survivin expression in cancer appears to be associated with unfavorable clinicopathological parameters, such as poor prognosis with progressive diseases and shorter patient survival rates. These data are summarized in Table 1. Here, it must be pointed out that with the increased numbers of publications about survivin in the last 3 years, inconsistent observations have been reported. Below is a summary and discussion when possible, of these data.

Table 1. Expression of survivin in cancer, which is associated with cancer progression and malignancy
Cancer typesSurvivin expressionCorrelation withReferences
  1. ALL, acute lymphocytic leukemia; AML, acute myeloid leukemia; ATL, adult T-cell leukaemia; BCC, basal cell carcinomas; B-CLL, B-cell chronic lymphocytic leukemia; HAK, hypertrophic actinic keratosis; HCC, hepatocellular carcinoma; IPMT, intraductal papillary-mucinous tumor; LSCC, laryngeal squamos cell carcinoa; NSCLC, non-small-cell lung cancer; PDC, pancreatic duct cell adenocarcinoma; SACL, small adenocarcinoma of lung; SCC, squamous cell carcinomas; SLN, sentinel lymph nodes; STS, soft tissue sarcomas.

Neuroblastoma34/72 (47%)Unfavorable disease and high grade; recurrence; survivin:Fas ratio increase (may be a better prognostic parameter)Adida et al. (1998), Islam et al. (2000a,b), Azuhata et al. (2001), Sandler et al. (2002)
Brain31/39 (80%); More than 50%; 34/43 (79%, RT-PCR)p53 expression; proliferation and poor prognosis; poor patient survival rate; malignant gradeChakravarti et al. (2002), Das et al. (2002), Sasaki et al. (2002), Kajiwara et al. (2003)
Esophageal and laryngealCancer versus normal (expression level) = 4 versus 1 (RT-PCR, P < 0.0001, n = 51); 67/102 (65.7%, LSCC); 67/84 (nuclei, 80%);Poor patient survival rate and recurrence; drug resistance; unfavorable disease; proliferative indexKato et al. (2001), Dong et al. (2002), Ikeguchi and Kaibara (2002), Poetker et al. (2002), Beardsmore et al. (2003), Grabowski et al. (2003)
Gastric60/174 (34.5%); 34/50 (68%, mRNA)Mut-p53 and Bcl-2 positive; unfavorable prognosis; reversed apoptotic index; diffuse type; cyclo-oxygenase 2 expressionLu et al. (1998), Wakana et al. (2002), Yu et al. (2002)
HCC14/20 (70%); 21/51 (41%); 19/61 (31%)Poor prognosis and recurrence; poor patient survival rate and cancer progressionIto et al. (2000), Ikeguchi et al. (2002a,c,d)
Colon, colorectal, and rectal91/171 (53%); 92/144 (63.5%, RT-PCR); 30/49 (61.2%)Bcl-2 positive; tumor development; poor patient survival rate (survivin-negative, 94%; positive, 48%); reversed apoptotic indexKawasaki et al. (1998), Sarela et al. (2000, 2001), Kawasaki et al. (2001), Rodel et al. (2002)
Bladder28/36 (78%)High grade and recurrenceSwana et al. (1999)
RenalTumor recurrence with high survivin:Fas ratioTakamizawa et al. (2001)
Uterine, ovarian, and endometrial42/42 (100%, RT-PCR); 41(66.7–100%); 7/33 (benign tumors); 11/23 (borderline tumors); 24/47, 90/124 (ovarian carcinomas)Carcinogenesis? Cancer progression; poor patient survival rate; unfavorable clinicopathological parameter; taxol resistanceSaitoh et al. (1999), Kim et al. (2002), Sui et al. (2002), Takai et al. (2002a,b), Zaffaroni et al. (2002)
NSCLC and SACL71/83 (85.5%, RT-PCR)Poor patient survival rate; venous invasionMonzo et al. (1999), Ikehara et al. (2002)
Breast118/167 (70.7%); 37/41 (90.2%); 27/39 (69%, RT-PCR)Bcl-2 expression but not p53 mutations; decreased apoptotic index; reversed patient survival rate; lymphatic invasionTanaka et al. (2000), Izawa et al. (2002), Nasu et al. (2002)
Pancreatic20/26 (77%, PDC); 9/16 (56%, IPMT); 46/52 (88%)Unfavorable disease; decreased apoptotic index; tumor malignance; Mut-p53 but not Bcl-2; both proliferation and apoptotic indexSatoh et al. (2001), Sarela et al. (2002)
Prostate-related100% (benign and malign., 57)Cancer malignance?Xing et al. (2001)
Non-melanoma17/21 (81%, BCC); 24/26 (92%, SCC); 19/19 (100%, premalign., SCC, HAK); 83/135 (61%, SCC)BCC and SCC development; high grade and undifferentiated; lymph node metastasisGrossman et al. (1999b), Lo Muzio et al. (2001)
Melanoma28/30 (93%, malignant); 11/11 (100%, invasive); 22/36 (61%, SLN)Cancer development and malignancy; short disease-free period and poor patient reversed survival rateChiodino et al. (1999), Grossman et al. (1999a), Gradilone et al. (2003)
STSPoor patient survival rate; telomerase expression, high grade and deathKappler et al. (2001), Wurl et al. (2002)
B-cell lymphoma134/223 (60%); 22/27 (80%)Poor prognostic marker; poor patient survival rate (?); cell cycle regulatorsAdida et al. (2000a), Kuttler et al. (2002)
Leukemia75/125 (60%, AML); 16/18 (89%, AML); 17/31 (55%, AML); 11/16 (69%, ALL)Negative prognostic factor; poor differentiation; short disease-free survival rateAdida et al. (2000b), Carter et al. (2001), Mori et al. (2002)

It was demonstrated that survivin expression in bladder cancer (transitional-cell carcinoma, TCC) is an attractive poor prognostic marker for bladder cancer recurrence, and urinary detection of survivin appears to be a valid test for prognosis (Swana et al., 1999; Smith et al., 2001). A similar study showed that nuclear staining for survivin was detected in 26 of 45 TCCs and in 2 of 14 cases of TCC in situ but was not detected in healthy bladder mucosa. However, TCC patients with a nuclear pattern of survivin localization had a greater period of disease-free survival (27.2 months) than was observed in TCC patients without nuclear staining of survivin (9.9 months) although not statistically significant (Lehner et al., 2002). In addition, a recent study of bladder cancer relapse with 30 patients by RT-PCR revealed that livin (also called ML-IAP) was positive in 7/30 patients (23%) while survivin was positive in 9/30 (30%). When the relationship between gene expression and time of relapse was examined, it was found that livin but not survivin had statistical significance for predicting relapse (Gazzaniga et al., 2003). Therefore, further study involved in more bladder cancer patients by multiple evaluating approaches would be required for a confirmed conclusion. However, it must be pointed out that the finding that survivin positivity in the nucleus is associated with a long period of disease-free survival is not only contrary to initial findings (Swana et al., 1999), but also to the current evidence that survivin is required for cancer cell division (see below). In addition, some authors found that the disease-free interval was significantly shorter when survivin expression was observed in the nucleus in studies of clear cell adenocarcinomas and serous adenocarcinomas of the ovary (Yoshida et al., 2001), and oesophageal squamous cell carcinoma (Grabowski et al., 2003).

In colon/colorectal cancer, the aberrant expression of survivin appears to be associated with Bcl-2 expression, tumor development, and poor patient survival rates (Table 1). Consistent with these findings, induction of the natural survivin antisense gene, epr-1 expression in a human colon cancer cell line inhibits survivin expression (Yamamoto et al., 2002). The reduction of survivin induced a concomitant decrease in cell proliferation, an increase in apoptosis, and sensitivity to anticancer agents (Yamamoto et al., 2002). However, a study using immunohistochemical detection of normal colonic mucosa as well as benign, hyperplastic, premalignant, and malignant lesions of the colon showed that survivin was detected in all cases of normal colonic mucosa (20/20), hyperplastic polyps (20/20), adenomatous polyps (20/20), and in both well differentiated and moderately differentiated colonic adenocarcinomas (20/20) although the spatial expression pattern was distinct (Gianani et al., 2001). While survivin expression was mostly restricted to crypts in the normal colonic mucosa, all epithelial cells showed uniformly intense staining for survivin in hyperplastic polyps. Adenomas and adenocarcinomas showed a heterogeneous staining pattern from cell-to-cell and gland-to-gland (Gianani et al., 2001). Based on these observations, the authors concluded that the expression of survivin is not a specific marker of adenocarcinoma of the colon (Gianani et al., 2001). Does the distinct expression pattern of survivin mean anything? A recent study demonstrated that wild type (wt) adenomatous polyposis coli (APC) down-regulates survivin expression via APC/beta-catenin/TCF-4 signaling (Zhang et al., 2001). Using normal colonic epithelium, they found that survivin is preferentially expressed in the lower crypt, which inversely correlates with wt-APC's expression pattern. These findings clearly argue that wt-APC, by progressively decreasing survivin and increasing apoptosis from the bottom of the crypt to the top, may limit the population size of stem cells in the lower crypt; mutant APC, which is common in colon cancer, may allow aberrant survivin expression leading to over-expansion of cells and initiating tumorigenesis (see below) (Zhang et al., 2001).

A number of reports indicated that survivin expression is associated with unfavorable prognosis and reduced apoptotic index in gastric cancer (Table 1), but some results inconsistent with these observations were also reported. One report showed that the nuclear staining of survivin (109/133 cases, 82.0%) is associated with a favorable prognosis but the cytoplasmic location of survivin (117/133, 88.0%) has no correlation with cancer progression or diagnosis (Okada et al., 2001). However, this contradicts the concept of the current findings on survivin function in mitosis, anti-apoptosis, and proliferation (see below). Another report (although not in gastric cancer) indicated that 14 of 20 (70%) hepatocellular carcinomas (HCC) showed positive nuclear staining for survivin, which was strongly correlated with the proliferation index (Ito et al., 2000). Promoting cell proliferation by survivin was further supported by the result that overexpression of survivin resulted in a decrease in the G0/G1 phase and an increase in the S phase in all four HCC cell lines tested (Ito et al., 2000).

In summary, while it is currently difficult to unify these inconsistent data with data in Table 1, the following observations may help to guide the direction of the future studies and interpretation of the data. There are three splicing variants for survivin (Fig. 1) (Mahotka et al., 1999), which may have differential localization, expression, and function in cancer cells (Kappler et al., 2001; Krieg et al., 2002; Mahotka et al., 2002a,b). Consistent with this possibility, using antibodies directed to the distinct survivin sequences, two pools of survivin protein (cytoplasmic and nuclear) could be identified (Fortugno et al., 2002) (see below).

Figure 1.

Comparison of the protein structure of survivin, survivin-2B, and survivin-ΔEx3. As shown, survivin protein encoded by four exons (Fig. 3) consists of a N-terminal domain, a BIR domain (aa38-87), and a C-terminal α-helix tail. Survivin-2B is a survivin-splicing variant and generated through an insertion of exon 2B (69 bp, 23 aa, Fig. 3) between exon 2 and 3 in frame. The insertion of exon 2B disrupts the BIR domain. Survivin-ΔEx3 is another survivin-splicing variant and encoded by the survivin exon 1, 2, and 4 without exon 3. Exon 3 escape results in a truncated BIR domain from amino acid position 73 and a frameshift from exon 4 that generates a novel C-terminal protein sequence with 64 amino acids in survivin-ΔEx3.


The initial mechanistic studies found that survivin is cell cycle-regulated with a robust expression in G2/M phase (Li et al., 1998). Consistent with this finding, immunofluorescence and confocal microscopy techniques demonstrated that survivin is associated with the microtubule organization center in interphase, centrosomes (mitotic poles), and mitotic spindles at metaphase, and relocates to midbodies in later telephase (Li et al., 1998, 1999). Consistently, survivin interacts with polymerized microtubules in vitro in a specific and saturable manner that is regulated by microtubule dynamics (Li et al., 1998). The localization of survivin appears to be required for its function in regulation of apoptosis because displacement of the endogenous wild type survivin by overexpression of the survivin C84A dominant negative mutant results in loss of survivin's anti-apoptosis function, increased caspase-3 activity and apoptosis during mitosis (Li et al., 1998). Further studies indicated that disruption of survivin localization by forced expression of the survivin C84A dominant negative mutant or inhibition of survivin expression by survivin antisense approaches results in a cell-division defect as well, which is characterized by centrosome deregulation, multipolar mitotic spindles, and multinucleated, polyploid cells (Li et al., 1999; Chen et al., 2000). These initial observations suggest survivin may be involved in both control of apoptosis and regulation of cell division during mitosis. However, the detailed mechanism is unknown.

Survivin and apoptosis pathway

Control of apoptosis and cell cycle by survivin has been demonstrated by many studies in multiple systems (Ambrosini et al., 1997, 1998; Tamm et al., 1998; Grossman et al., 1999a,b; Kobayashi et al., 1999; Li et al., 1998, 1999; Mahotka et al., 1999; Chen et al., 2000; Islam et al., 2000b; Olie et al., 2000; Suzuki et al., 2000a,b; Grossman et al., 2001a,b; Jiang et al., 2001; Mesri et al., 2001b; O'Connor et al., 2000a, 2002; Shankar et al., 2001; Shin et al., 2001; Aoki et al., 2002; Xia et al., 2002; Zhou et al., 2002). The initial evidence showing survivin inhibits apoptosis comes from the fact that forced expression of survivin in a mouse IL-3-dependent pre-B cell line, BaF3 counteracts apoptosis induced by deprivation of interleukin-3 (Ambrosini et al., 1997), suggesting survivin may be involved in apoptosis inhibition during G1/S transition as well. Consistent with this possibility, two recent studies demonstrated that survivin could interact with the cell cycle regulator Cdk4, leading to Cdk2/Cyclin E activation and Rb phosphorylation. Forced overexpression of survivin resulted in an accelerated S phase and resistance to G1 arrest (Suzuki et al., 2000a,b). In addition, the finding that upregulation of survivin by hematopoietic growth factors in quiescent CD34+ hematopoietic stem and progenitor cells occurs before cell cycle entry (Fukuda and Pelus, 2002), also supports a role for survivin in early cell cycle entry. Whether this is an unusual example for normal undifferentiated cells is unknown. Nevertheless, it is clear that further exploration of the role and mechanism of survivin in G1/S transition will enrich the current understanding of survivin function in the regulation of apoptosis and cell cycle control.

It must be emphasized that the majority of the current studies indicated that the inhibition of apoptosis by survivin is during mitosis and appears to be involved in a novel mitotic spindle assembly checkpoint. Initial studies using thymidine block/release experiments indicated that interference of endogenous survivin by forced expression of the survivin C84A dominant negative mutant or survivin antisense induces a peak of caspase-3 activation at the G2/M phase (Li et al., 1998). This suggests that apoptosis induced by inhibiting survivin occurs during mitosis. Consistent with this view, recent work has revealed that survivin has a CDC2/cyclinB1-specific phosphorylation site (T34PARMA39), which appears to be phosphorylated at the Thr34 position in vitro as well as in vivo during mitosis (O'Connor et al., 2000a). This phosphorylation is required for survivin to associate with polymerized microtubules/mitotic spindle and inhibit caspase-9 activity during mitosis (Fig. 2) (O'Connor et al., 2000a). Overexpression of a survivin Thr34Ala (T34A) mutant, which can still bind to the mitotic apparatus but is unable to be phosphorylated at Thr34 position by CDC2/cyclinB1, resulted in dissociation of an endogenous survivin–caspase-9 complex on the mitotic apparatus, and caspase-9-dependent apoptosis during mitosis (O'Connor et al., 2000a). A more recent study has generated even deeper insight into this survival pathway and found that increased survivin expression and CDC2 activity by microtubule-stabilizing anticancer drug taxol during mitosis preserved a survival pathway in cancer cells (O'Connor et al., 2002). Inhibition of CDC2 activity by pharmacologic inhibitors (purvalanol A), genetic modification, or molecular gene targeting resulted in the inability to phosphorylate survivin on Thr34, and a massive apoptosis without disruption of mitotic spindle (O'Connor et al., 2002). Interestingly, a follow-up paper showed a relative broad cell cycle Cdk inhibitor, flavopiridol, inhibits the same survivin T34 phosphorylation pathway in the induction of apoptotic cancer cell death (Wall et al., 2003). More interestingly, adriamycin (doxorubicin), a DNA damage-inducing anticancer drug, which was demonstrated to transcriptionally downregulate survivin in a number of cell lines with wild type p53 (wt-53) (Hoffman et al., 2002; Mirza et al., 2002; Zhou et al., 2002) can synergistically induce apoptosis and inhibit xenograft tumor growth in sequential combination with flavopiridol in association with the survivin T34 phosphorylation survival pathway (Wall et al., 2003). A recent study on the mechanism of apoptosis induced by flavopiridol suggests that multiple events may be attributed to flavopiridol-induced apoptosis (Wittmann et al., 2003). Nevertheless, the discovery of this novel mitotic spindle checkpoint survival pathway maintained by CDC2/cyclinB1–survivin interactions during mitosis in cancer cells (Fig. 2) may provide a new basis for a sequential regimen treatment of human cancer with anticancer drugs and CDC2 inhibitors.

Figure 2.

A hypothetic model to unify the function and subcellular localization of survivin and survivin-splicing variants. (1) In metaphase, survivin binds to the mitotic spindle fibers on which survivin is associated with and phosphorylated by CDC2/cyclin B1 at the survivin Thr34 position. The phosphorylated survivin offers survivin the ability to physically interact and functionally inhibit the caspase-9 activity. However, for any reason, if survivin fails to be phosphorylated during this stage, a caspase-9-involved, mitochondria-mediated intrinsic apoptotic pathway (apoptosome) will be initiated. In this context, survivin counteracts a default apoptotic checkpoint during mitosis. (2) During prometaphase and metaphase, survivin and survivin-ΔEx3 together with Aurora B and INCENP bind to kinetochore at the midpoint of the two opposite-facing kinetochore discs, which is likely involved in the latter chromosome segregation at the transition of metaphase to anaphase. Alternatively, the mitotic spindle-binding survivin may participate in chromosome arrangement and segregation by interaction with tubulin and other kinases such as Aurora-2 (Aur-2). (3) During anaphase, the chromosomes are pulled toward the spindle poles, the kinetochore-associated survivin, and survivin-ΔEx3 together with INCENP and Aurora B will be transferred to spindle midzone on which these proteins will be condensed and associated with a bound structure (midbody) at telephase. These and other related processes may be involved in cytokinesis.

Survivin appears to be able to control the activation/activity of caspase-9 during mitosis (O'Connor et al., 2000a, 2002). Would survivin be able to inhibit downstream effector protease caspase activation/activity as well? The answer is perhaps yes but not very clear because contradictory observations exist. Forced expression of survivin can protect 293 cells from apoptosis induced by overexpression of procaspase-3 and -7 and inhibit the processing of these zymogens into active caspases. In vitro experiments showed that survivin binds to caspase-3 and -7, but not to the upstream initiator protease caspase-8, suggesting a specific interaction (Tamm et al., 1998). It was also reported that survivin neither binds to nor inhibits caspase-3 activity/activation in the in vitro experiments (Banks et al., 2000; Verdecia et al., 2000). The reason these studies failed to demonstrate the physical interaction and functional inhibition of caspase-3 by survivin is not clear. However, Shin et al., 2001 found that a preincubation of survivin for at least 20 min at 25°C before adding substrate is critical for observation of the inhibitory effects of survivin on caspase-3 and -7 in vitro. In their experiments, the survivin protein purified from E. coli exhibits the inhibitory effects on caspase-3 and -7 activity, suggesting that post-modification of survivin may not be necessary for inhibition of these effector protease caspases by survivin. Regardless, although in vivo data from native conditions instead of transfection-mediated overexpression to show the physical interaction of survivin with caspase-3 and -7 in mammalian cell system is lacking, a heterologous yeast expression assay showed that survivin could suppress caspase-3-mediated cytotoxicity but was weaker than c-IAP1, c-IAP2, and XIAP (Wright et al., 2000). The same result was also obtained in mammalian cells expressing an auto-activating caspase-3 (Wright et al., 2000). In addition, a recent study reported that the sensitivity of TRAIL-induced apoptosis in renal cell carcinoma associates with the expression level of survivin (Griffith et al., 2002). Downregulation of survivin expression in TRAIL-resistant cells sensitizes TRAIL-induced apoptotic cell death and overexpression of survivin in TRAIL-sensitive cells increases TRAIL resistance (Griffith et al., 2002). In consideration of the fact that TRAIL-induced cell death involves a number of effector protease caspase activation including caspase-3 and -7, these data suggest that survivin may regulate the sensitivity of TRAIL-induced cell death though control of caspase-3 and -7 activation/activity. Taken together, the evidence suggests survivin is probably able to inhibit caspase-3 and -7 activation/activity in vitro and in vivo, but the inhibitory effects of survivin on caspase-3 and -7 might be weaker than that from c-IAP1, c-IAP2, and XIAP.

An interesting question is whether the inhibition of caspase activation/activity by survivin is the only way for survivin to exert its inhibitory function of apoptosis. A report on the role of inhibition of survivin and apoptosis in neural tumor cells communicated some interesting observations (Shankar et al., 2001). Downregulation of survivin by survivin antisense oligonucleotides (SAO) in human neuroblastoma cells revealed that a 74% of the SAO-treated cells were trypan blue positive but there was no PARP cleavage and caspase-3 activation (confirmed by the fact that co-administration of zVAD-fmk, a pan-caspase inhibitor, with SAO did not inhibit cell death). This data suggests that a caspase-independent mechanism of cell death may be involved. In contrast, targeting of survivin by SAO in an oligodendroglioma cell line, TC620, induced PARP cleavage and apoptosis, which could be blocked by administration of zVAD-fmk (Shankar et al., 2001). These findings raised the possibility that survivin may regulate both caspase-dependent and -independent cell death pathways, which may be dependent on the cell type. Further investigation of the possibility that survivin is involved in the regulation of caspase-independent cell death pathway should enrich our understanding of survivin function in apoptosis control.

Survivin localization, mitosis, and cell division

Evidence indicating a role for survivin in cancer cell division was developed from studies with survivin antisense or dominant negative mutant approaches that resulted in cells with centrosome deregulation, multipolar mitotic spindles, and multinucleated or giant nuclei (Li et al., 1999). While the mechanism of survivin's role in cell division during mitosis requires a further elucidation, the initial findings evoked interest in the scientific community. Subsequently, a debate about survivin function in mitosis was initiated as a result of observations of survivin localization during the cell cycle.

Using cultural HeLa cells and antibody TO65, raised against peptide Ala3–Ile19 of survivin, by immunofluorescence microscopy analysis, Uren et al., 2000 found that beginning from the late prophase until metaphase, survivin colocalizes with CENP-B, a constitutive component of the centromere. In anaphase, CENP-B moved along with the separating sister chromosome toward the poles, but survivin was retained between the separated sister-chromosome at the mitotic spindle midzone. In the later anaphase and telophase, survivin was located at the midbody (a midzone microtubule bundling structure) and then degraded after telophase (Uren et al., 2000). Using microtubule-stabilizing and -destabilizing agents, these investigators also demonstrated that the localization of survivin to the centromere from late prophase until metaphase did not depend on the integrity of the microtubules and that the location of survivin on the centromere is at the inner centromere region between the two kinetochore discs along Z-axis (Uren et al., 2000). This location is consistent with the survivin behavior that while CENP-B moved toward the poles, survivin retained at mitotic spindle midzone although the significance of such localization is currently unknown. In contrast from previous findings that INCENP (inner centromere protein) was located along the Y-axis (Cooke et al., 1987) or beneath the kinetochore discs along the X-axis (Earnshaw and Cooke, 1991; Eckley et al., 1997), Uren et al., 2000 found that INCENP was colocalized with survivin along the Z-axis at the inner centromere region. This observation suggests a possible interaction between these two proteins. It is surprising that using the anti-survivin TO65 antibody, these investigators did not observe survivin in association with the mitotic spindle during metaphase (Uren et al., 2000) (see below).

Consistent with the above findings, two recent papers revealed a similar survivin localization (Skoufias et al., 2000; Wheatley et al., 2001). The first paper used ectopic human hemagglutinin (HA)-tagged survivin expression in either HeLa or NIH3T3 cells and survivin localization was monitored by immunofluorescence analysis using anti-HA antibody (Skoufias et al., 2000). The second report employed a survivin polyclonal antibody (pAb, NOVUS Biologicals, Littleton, CO) for the immunofluorescence analysis of survivin localization in HeLa cells (Wheatley et al., 2001). Interestingly, both reports failed to locate survivin on the mitotic spindle during metaphase. They concluded that survivin is a chromosome passenger protein (Skoufias et al., 2000; Wheatley et al., 2001). However, using the survivin monoclonal antibody, 8E2, which was used in the initial publications (Li et al., 1998, 1999), Wheatley et al. observed survivin binds to the mitotic spindle at metaphase but not to the centromere (Wheatley et al., 2001), in agreement with the previous findings. Regardless, in support of their conclusion that survivin is a chromosome passenger protein, Wheatley et al. further demonstrated that survivin binds directly to both Aurora-B and INCENP (two well known chromosome passenger proteins) in yeast two-hybrid and in vitro pull-down assays, and that disruption of INCENP localization by forced expression of a INCENP1-405 mutant mislocates survivin to the chromosomes from the centromeres without transferring it to the spindle midzone in anaphase (Wheatley et al., 2001).

Recently, two more reports provided new evidence by immunoprecipitation and immunoblotting experiments that survivin, INCENP, and Aurora B interact each other in vivo (Bolton et al., 2002), and that in the presence of survivin, Aurora B kinase activity appears to be enhanced over ten times (Bolton et al., 2002; Chen et al., 2002). Survivin could directly interact with Aurora B in vitro (Chen et al., 2002), which confirmed the previous findings (Wheatley et al., 2001). The increase of Aurora B kinase activity by survivin required the phosphorylation of Aurora B itself (Bolton et al., 2002). Interestingly, while these authors could not find that Aurora B can phosphorylate survivin in vitro or in vivo (Chen et al., 2002; Wheatley et al., 2001, a recent report showed that Ark1p, an Aurora B homolog in S. pombe, mediated robust phosphorylation of a N-terminal fragment containing the BIR domain (1-208) of Bir1 protein (a survivin-type IAP in S. pombe) in vitro (Leverson et al., 2002).

Interestingly, while investigators consistently demonstrated that survivin, Aurora B, and INCENP interact each other in a complex (Wheatley et al., 2001; Bolton et al., 2002), the strength of the interaction of survivin with Aurora B or INCENP varied among different reports. By in vitro GST-survivin pull down experiments, it was shown that survivin binds to Aurora B tightly and the interaction is resistant to 3 M NaCl, while survivin exhibits a weak interaction with INCENP and the complex is dissociated above 0.6 M NaCl (Wheatley et al., 2001). Ectopic expression of a INCENP1-405 mutant showed displacement of the Aurora B from centromere, which results in Aurora B dispersal throughout the cell in the prometaphase, but survivin is still associated with chromosome (although not concentrated on the centromere) (Wheatley et al., 2001). If the interaction of survivin with Aurora B were much stronger than that with INCENP in vivo, one would expect that survivin would disperse with Aurora together in the cell as well. Therefore, presumably, the interaction of survivin with Aurora B is weaker than that with INCENP in vivo. Consistent with this view, using cell extract immunoprecipitation (IP) and immunoblotting approaches, Bolton et al., 2002 found that survivin binds to Aurora B weakly (dissociation above 0.5 M NaCl) than that with which INCENP binds to Aurora B. Moreover, these investigators found that the N-terminal 1-98 amino acids of Aurora B are required for interaction with either INCENP or survivin. In contrast, Chen et al. showed that the Aurora B66-344 mutant without the N-terminal part directly interacts with survivin by in vitro pull down experiments, but results from in vivo approaches such as IP were not provided (Chen et al., 2002).

An interesting question is which one of the three molecules (survivin, INCENP, and Aurora B) is the key protein for anchoring the other two proteins. Incomplete data have suggested that INCENP might be a scaffold protein for survivin and Aurora B to anchor to the centromere during mitosis (Wheatley et al., 2001; Bolton et al., 2002). However, inhibition of survivin expression in vivo by survivin antisense mislocates Aurora B from chromosome (presumably from centromere kinetochore) to a diffused pattern in the cell (Chen et al., 2002). This observation suggests that survivin is required for anchoring Aurora B to a protein (INCENP?) on the centromere. Regardless, survivin binding to Aurora B appears to be required for Aurora B kinase activation to phosphorylate its substrates such as histone H3 (Bolton et al., 2002; Chen et al., 2002).

Here, it is necessary to separately mention a recent report, which used an unusual survivinDsRed fusion protein to study the localization and function of survivin during cell cycle (Temme et al., 2003). Overexpression of survivinDsRed in HeLa cells showed no colocalization of survivin with mitotic spindle but association with chromosomes during mitosis (Temme et al., 2003). This result is not exactly consistent with the finding that survivin associates kinetochores (Skoufias et al., 2000; Uren et al., 2000; Wheatley et al., 2001), or with the initial finding that survivin associated mitotic spindle fibers (Li et al., 1998). Moreover, studies of survivin function using the DsRed-tagged survivin CDC2-phosphorylation site mutant (survivin-T34A) showed that overexpression of this mutant did not induce apoptosis but rather protected cells from apoptosis induced by a number of anticancer drugs including cisplatin and doxorubicin (Temme et al., 2003). These unusual observations imply their experimental system may be inappropriate. This is likely because the large DsRed tag brings survivin into a large tetramer complex from its natural small dimmers. In addition, their function results have been disapproved by a number of recent studies using this mutant in vitro and in vivo (O'Connor et al., 2000a, 2002; Grossman et al., 2001b; Mesri et al., 2001b).

The notion that survivin is solely chromosome passenger protein also comes from the observation that the survivin-type IAP, Bir1/bir1 proteins (BIRP) from nematode (C. elegans) (Fraser et al., 1999; Speliotes et al., 2000) and yeast (S. cerevisiae and S. pombe) (Uren et al., 1999; Yoon and Carbon, 1999; Li et al., 2000; Leverson et al., 2002) are exclusively involved in chromosome segregation and cell division/cytokinesis. Using RNA interference technology to knockout BIR-1/bir-1 in C. elegans, resultant embryos and fertilized oocytes could not complete cytokinesis and became multinucleate/highly polyploid giant cells (Fraser et al., 1999; Speliotes et al., 2000), similar to cells lacking the Aurora-like kinase, AIR-2 (Speliotes et al., 2000). Without BIR-1, AIR-2 also mislocates from the chromosome (Speliotes et al., 2000), suggesting anchor properties for BIR-1. Interestingly, the cytokinesis defect induced by BIR-1 defect could be partially rescued by expression of human survivin (Fraser et al., 1999; Speliotes et al., 2000). In S. cerevisiae, BIR1 was found to interact with the kinetochore protein, Ndc10p (a substrate of Aurora-like kinase Ipl1p) in a yeast two-hybrid screen (Yoon and Carbon, 1999). Interestingly, the C-terminal domain (550-954), but not the conserved BIR domain, of BIR1 is responsible for the interaction with Ndc10p and for the rescue of chromosome segregation defects induced by BIR1 knockout (Yoon and Carbon, 1999). Consistent with this observation, the cell division defect and growth retardation induced by BIR1 knockout can be rescued by ectopic expression of BIR1 but not by its human counterpart, survivin (Li et al., 2000). Likewise, in S. pombe, bir1 haploidy is lethal, and the bir1 deletion mutant spore is unable to finish metaphase to anaphase transition because of a failure of spindle microtubule elongation (cut phenotype) (Uren et al., 1999). Interestingly, in contrast to a role of the survivin-type IAP proteins from nematode and yeast in cytokinesis, a survivin-type IAP member, deterin from flies (D. melanogaster) was demonstrated to be solely involved in the regulation of apoptosis (Jones et al., 2000). This observation strengthened the notion that survivin is involved in both regulation of apoptosis and cell division.

Now, an important issue is how to explain or unify the inconsistent results about survivin localization during mitosis that have been reported. This question appears to be addressed by a recent report. In this report, Fortugno et al., 2002 found that survivin exists in immunochemically distinct subcellular pools. In subcellular fractionation and Western blot experiments with HeLa cells, mAb 32.1, a monoclonal antibody specific to the survivin sequence Ala3–Ile19, exclusively recognized a 16.5 kDa survivin band from the nuclear fraction but not from the cytosolic and the centrosome-enriched fractions. In contrast, mAbs 59 or 60, monoclonal antibodies specific to survivin sequence Cys57–Trp67, exclusively recognized cytosol and centrosome-associated survivin but not survivin from the nuclear extracts (NEs) similarly to the 8E2 mAb used previously (Li et al., 1998, 1999). Interestingly, 9B1D9, a survivin mAb generated by immunization of BALB/c mice with full-length survivin protein, could recognize the chromosome-associated survivin but not the midbody-assocated survivin (Temme et al., 2003). On the other hand, rabbit polyclonal antibodies (pAb) NOVUS (NOVUS Biologicals) and BTD, raised against full-length recombinant human survivin, reacted with both post-nuclear (cytosolic and cytoskeletal extracts) and nuclear pools of survivin (Fortugno et al., 2002). Consistent with these observations, immunofluorescence and (confocal) microscopy experiments monitoring the in situ localization of endogenous survivin during the cell cycle by mAb 8E2 recognized a filamentous pattern in interphase, spindle poles, and spindle microtutubles at metaphase and anaphase, and stained midbodies at telophase (Fortugno et al., 2002), in agreement with the previous observations (Li et al., 1998; Wheatley et al., 2001). In contrast to the mAb 8E2 pattern, mAb 32.1 exhibited a punctate labeling of the centromere kinetochore at metaphase, which was then transferred to the spindle midzone at anaphase, and midbodies at telephase (Fortugno et al., 2002). This pattern is identical to that of Uren et al. using a survivin antibody raised against survivin Ala3–Ile19 (Uren et al., 2000).

In agreement with the Western blot experiments above, polyclonal antibodies, pAb NOVUS (NOVUS Biologicals) and BTD simultaneously label spindle poles and spindle fibers at metaphase, spindle midzone at anaphase, and midbodies at telephase (Fortugno et al., 2002). However, it will be interesting to know whether these pAb could associate with centromores/kinetochores. Interestingly, more detailed subcellular fractionation and Western blot experiments showed that the nuclear pool of survivin recognized by pAb NOVUS is situated in the nucleoplasm, but not in DNase released proteins, nuclear matrix proteins, or outer nuclear matrix proteins. The ratio of nucleoplasmic survivin versus cytosolic survivin is about 1/6 (Fortugno et al., 2002). This raised an intriguing question concerning the mechanism for the survivin translocation to the nuclei and subsequent localization there.

Using thymidine block/release, subcellular fractionation and Western blot experiments, Fortugno et al., 2002 demonstrated that survivin immunoprecipitated by mAb 32.1 is not associated with or phosphorylated by CDC2/CyclinB1 from either asynchronized or taxol-synchronized HeLa cells but survivin immunoprecipitated by the pAb NOVUS does. These results suggest that although both cytoplasmic and nucleoplasmic survivin exists, only the cytoplasmic survivin is phosphorylated at the Thr34 site.

How could survivin exist in two immuno-distinct pools? One possibility is that the post-translational modifications of survivin may affect survivin conformation and epitope accessibility in cytoplasmic versus nucleoplasmic survivin. In support of this possibility, there are a number of potential kinase phosphorylation and acetylation sites in the survivin protein sequence. Another possibility is that the cytoplasmic pool of survivin is likely the “ortho-” survivin or both survivin and survivin-2B, a survivin splicing variant with an extra 23 amino acid insertion in frame between exon 2 and 3 (Fig. 1) (Mahotka et al., 1999), and the nucleoplasmic pool of survivin is possibly involved in the survivin-ΔEx3, another survivin splicing variant without exon 3 but with a novel carboxy-terminal peptide tail derived by a frameshift from exon 4 (Fig. 1) (Mahotka et al., 1999). This may be a more likely explanation since a number of current observations are consistent with this possibility. It appears that survivin or both survivin and survivin-2B employ a CRM1-mediated nuclear export of survivin to counteract its nuclear translocation and as an overall result, survivin and possibly survivin-2B as well, stay in the cytoplasm (Rodriguez et al., 2002). In contrast, survivin-ΔEx3 possesses a bipartite nuclear localization signal (NLS, underlined) sequence in an R/K-rich region (81RRKNLRKLRRK91). It was demonstrated that this structure is required and sufficient for retaining survivin-ΔEx3 in the nucleus (Rodriguez et al., 2002), and expression of a GFP–survivin-ΔEx3 revealed a cell cycle-dependent nuclear accumulation of survivin-ΔEx3 (Mahotka et al., 2002b). However, it should be pointed out that a recent report has a different view of survivin-ΔEx3 (Wang et al., 2002). The authors indicated that a K7 protein encoded by the K7 open reading frame of Kaposi's sarcoma-associated herpesvirus is structurally related to survivin-ΔEx3. Based on the alignment data, it was noted that both K7 and survivin-ΔEx3 contain a potential mitochondrial-targeting sequence, an N-terminal region of the BIR domain and a putative Bcl-2 homology-like domain. Therefore, survivin-ΔEx3 was speculated to be the cellular homolog of K7 protein (Wang et al., 2002). While these investigators demonstrated K7 could localize to mitochondria, endoplasmic reticulum, and the nuclear membrane, there was no data provided for survivin-ΔEx3. In addition, the putative Bcl-2 homology-like domain of the K7 protein does not have the R/K-rich region, but survivin-ΔEx3 does. Therefore, all these claims remain to be verified by experiments. In addition, quantitative RT-PCR experiments indicated that the ratio of survivin mRNA versus survivin-ΔEx3 mRNA is approximately 5–7:1 in HeLa cells (unpublished data), which is consistent with the observation reported by Fortugno et al. that the ratio of cytoplasmic survivin versus nucleoplasmic survivin is ∼6:1 (Fortugno et al., 2002). Since three survivin variants have the same N-terminal sequence (Fig. 1), which could be potentially recognized by survivin antibodies used currently, this observation raises a possibility that the nuclear survivin is survivin-ΔEx3 or both survivin and survivin-ΔEx3.

In summary, the notion that survivin is solely a chromosome passenger protein (Skoufias et al., 2000; Uren et al., 2000; Adams et al., 2001; Wheatley et al., 2001) is neither complete nor consistent with the localization pattern of survivin during mitosis. Survivin appears to be involved in both regulation of apoptosis and cell division, which are associated with multi-facet subcellular localizations (Fig. 2). However, the involvement of survivin in cell division is mechanistically much less clear than that of survivin in anti-apoptosis. Further investigation of the differential function of survivin and survivin variants in the regulation of apoptosis and cell division will be required to obtain a uniform view of survivin.

Cell phenotype and targeting survivin with antibody

One report found that microinjection of a survivin polyclonal antibody (pAb, Alpha Diagnostics International Inc., San Antonio, TX) shortens the duration of metaphase and prolongs the duration of both anaphase and cytokenesis without significant effects on the duration of prometaphase in the onset of mitotic cells (Kallio et al., 2001). In contrast, another report reached the opposite conclusion as microinjection of survivin pAb NOVUS (NOVUS Biologicals) into the early mitotic HeLa cells prolonged the duration of both prometaphase and metaphase without significant effects on the duration of anaphase (Giodini et al., 2002). The reason for this contradictory observation is not clear. Regardless, microinjection of synchronized HeLa cells in the nucleus with mAb 32.1 (a monoclonal antibody that specifically recognizes the kinetochore-associated survivin protein (Fortugno et al., 2002)) has no effect on mitotic spindle assembly or cytokinesis (Giodini et al., 2002). The observation that specific inhibition of the kinetochore-associated survivin does not result in a defect in mitotic spindle assembly and cytokinesis, suggests that the cytoplasmic pool of survivin is responsible for the phenotype induced by microinjection of anti-survivin pAb NOVUS. However, this notion may not be able to reconcile with the concept that the kinetochore-associated survivin is a key regulator for Aurora B kinase activity and involved in chromosome segregation and cytokinesis (Uren et al., 2000; Wheatley et al., 2001; Bolton et al., 2002; Chen et al., 2002), since Aurora B was demonstrated to be a key regulator in these processes (Kallio et al., 2002; Murata-Hori and Wang, 2002a,b). In addition, it is interesting that injection of pAb NOVUS resulted in shortened mitotic spindles, and overexpression of survivin resulted in a similar observation that reduces pole-to-pole distance at metaphase of HeLa cells (Giodini et al., 2002). Thus, it will require more investigation to reveal how, at the molecular level, survivin is involved in microtubule organization, mitotic progression, metaphase-anaphase transition, and perhaps cytokinesis. In short, molecular clarification of the function of survivin and survivin variants will be required to lay a strong basis for a more rational application of therapeutics using survivin as a target. Hopefully, this will be a focus in the coming years.


Transcriptional regulation of survivin

The complexity of survivin function in anti-apoptosis and cell division, which is distinctly associated with its subcellular localization in the cancer cells, implies that it may be a challenge to find an effective way to inhibit the function of survivin for cancer therapeutics. To this end, investigation of transcriptional and post-transcriptional regulation of survivin, and targeting its aberrant expression may be able to reveal effective approaches for cancer therapeutics. In this context, an important question to be answered is what is the mechanism by which survivin is expressed in cancer but not in normal, differentiated adult tissues. Using genomic DNA as templates, which were isolated from normal, fetal, and cancer tissues, and digested with methylation-sensitive restriction enzymes, Sac II (two sites in the proximal promoter region of survivin) and Eco0109 I (one site in the proximal promoter region and two sites in the intron I of survivin), PCR amplification with a pair of primers flanking these methylation-sensitive sites indicated that no PCR products could be amplified (Li and Altieri, 1999b). This result suggests that survivin gene is not methylated at these sites. Interestingly, in a later report, using genomic DNA as templates, which were isolated from normal and cancerous ovarian tissues, and digested with methylation-sensitive restriction enzyme, Hpa II (one site in the proximal promoter region, one in exon 1 and one in intron 1 of survivin respectively), PCR amplification with a pair of primers overlapping with the Hpa II sites in the promoter region and intron 1 respectively, resulted in a differential methylation pattern on Hpa II sites (Hattori et al., 2001). DNA from the cancerous tissues is methylation-free on Hpa II sites in most cases (37/43, 86%), but DNA from the normal tissues is methylated on Hpa II site (41/43, 95%). The different observations from these two reports raise a possibility that the survivin gene may have a limited and selective CpG methylation pattern in the normal tissues, which is removed during the transition of normal cells to the cancer cells (Fig. 3). The consistent finding that survivin is methylation-free in cancer cells provides an example does not favor the notion that CpG island-associated genes, in particular tumor suppressor or related genes, are often hypermethylated in association with silencing of these genes in cancer (Clark and Melki, 2002), but does favor the new view that cancer-associated DNA hypomethylation is as prevalent as cancer-linked hypermethylation (Ehrlich, 2002).

Figure 3.

Transcriptional regulation of survivin gene expression in cancer cells. The accumulated data suggest that (1) survivin gene may be partially methylated in the CpG island in normal adult cells but demethylated in cancer cells; (2) p53 mutation may contribute to the enhanced expression of survivin in cancer cells since wild type p53 (wt-53) transcriptionally suppresses survivin expression; (3) the enhanced β-catenin induced by adenomatous polyposis coli (APC) mutation may transcriptionally upregulate survivin by cooperation with TCF-4; (4) a general model for the constitutive and regulated expression of survivin in cancer cells can be resulted from any of the following events—virus infection, growth factor/cytokine paracrine/autocrine loop, tyrosine kinase receptor mutation, activation or gene amplification, and constitutive activation of the relevant signaling transducer proteins. The activated signaling proteins will upregulate survivin gene expression by either directly activating the existed TFs (transcription factors) or indirectly inducing the relevant TFs' transcription, which can in turn contribute survivin expression.

Survivin is constitutively expressed in various cancers (Table 1) and its expression is cell cycle-regulated with a robust expression in the G2/M phase of the cell cycle (Li et al., 1998; Li and Altieri, 1999a; Otaki et al., 2000). An issue is what controls the constitutive and cell cycle-regulated expression of survivin in cancer cells and how does it work? Previous reports revealed that Sp1 is a major player for the constitutive expression of survivin when evaluated in a luciferase reporter assay using an isolated human survivin core promoter (230 bp) (Li and Altieri, 1999b). Recent study suggests, a much more complex scenario appears to govern the constitutive expression of survivin involving multiple transcription factor interactions in and outside of the core promoter of survivin (unpublished data) (Fig. 3). It was demonstrated that the cell cycle-regulated expression of survivin is attributed to the existence of multiple CDE (cell cycle dependent element) and CHR (cell cycle gene homology region) motifs in the core promoter region of survivin (Li et al., 1998), but what links the cis-acting protein factors binding on these motifs for the regulation of survivin expression is currently unknown. It must be emphasized that CDE/CHR motifs in the core promoter of survivin only contribute a 2–5-fold increase in G2/M versus in G1 in the survivin promoter-luciferase assay experiments but Northern blot indicated that survivin mRNA is over a 40-fold increased in G2/M versus in G1 phase (Li et al., 1998). This observation suggests that cis-acting DNA elements rather then CDE/CHR motifs in the core promoter, contribute to the cell cycle-dependent expression of survivin as well. Consistent with this view, it appears that certain areas outside of the core promoter of survivin are involved in both its constitutive and cell cycle regulation (Fig. 3). A similar scenario for the constitutive and cell cycle-regulated expression of survivin appears to exist in the mouse as well (Li and Altieri, 1999a; Otaki et al., 2000).

Currently, the detailed mechanism for the control of survivin transcription in cancer is largely unknown. A great deal of efforts will be required in this research area in order to apply the inhibition of survivin transcription in cancer as a novel approach for cancer therapeutics. Based on the present information, a possible model for survivin transcription control is that a diversity of events may result in constitutive activation of upstream signaling in cancer cells which may trigger the activation and transcription of appropriate transcription factors (TFs) to turn on survivin expression in cancer (Fig. 3). In support of this hypothesis, a recent report showed that inhibition of the constitutive activation of STAT3 by forced expression of a dominant negative Stat3 or by pharmacological Stat3 inhibitor transcriptionally represses survivin expression without effect on Bcl-2, Bcl-XL, and Mcl-1 protein expression, and triggers caspase-dependent apoptotic cell death in AIDS-defining primary effusion lymphoma (PEL), which can be rescued by forced expression of survivin (Aoki et al., 2002). Consistent with this report, Zhu et al., 2003 found that Vpr, a human immunodeficiency virus type 1 (HIV-1) accessory protein can transcriptionally upregulate survivin assessed by both survivin promoter-luciferase assay and endogenous survivin expression. The upregulation of survivin by Vpr is also associated with an increase of G2/M cell population and caffeine, a Vpr inhibitor, can ablate survivin promoter activity and decrease the G2/M cell population (Zhu et al., 2003). The finding in this particular case suggest that the constitutive expression of survivin in virus-associated cancer could be due to the virus infection of normal cells (Fig. 3). Here, it should be emphasized that although most normal, differentiated adult tissues do not express survivin, a few adult tissues do express survivin including spleen (unpublished data), testis, thymus, placenta (Ambrosini et al., 1997; Kobayashi et al., 1999) and normal colonic crypt (Zhang et al., 2001). However, limited studies suggested that the regulation of survivin expression between cancer cells and survivin-positive normal adult issues is possibly distinct (Fig. 4). To this end, if a differential regulation of survivin transcription between cancer and survivin-positive normal adult tissues could be confirmed, transcriptional disruption of survivin expression in cancer may have advantages over other approaches in terms of cancer therapeutics.

Figure 4.

Differential DNA–protein interactions in Southwestern blot using the 6.2-kb human survivin promoter as a probe. Nuclear extracts (NEs), from survivin-expressing cancer cells (HeLa and ABE8.1/2, A and B) or survivin-expressing normal tissues (thymus and spleen, C) and from survivin-negative untransformed cells (18Lu and NIH3T3, A and B) or survivin-negative normal tissues (brain, liver and kidney, C), were separated on SDS–PAGE gel and transferred to Immobilon-P filter (Millipore Co.; Bedford, MA). After renaturation of the immobilized proteins on the filter, the renatured proteins on the Immobilon-P filter were hybridized with α-32P-dCTP-fill-in-labeled DNA probes from the previously characterized 6.2-kb human survivin promoter region digested with Msp I to generate a size range from 68 to 1,105 bp (18 DNA fragments in a mix). The sources and amounts of NEs and the molecular weights of protein markers are indicated. As shown, while NEs from survivin-negative cells or tissues have little interaction with survivin promoter probes, NEs from survivin-positive cells or tissues exhibit strong interactions with survivin promoter probes. However, the DNA–protein interaction pattern is distinct between survivin-positive cancer cells and normal adult tissues while NEs from spleen and thymus show a similar pattern. This result provides a possibility that the regulation of survivin expression in cancer cells is distinct from that in normal adult tissues. For the Southwestern blot experiments, the final concentration of non-specific cold DNA competitors to block the non-specific binding to the filter or probes during prehybridization and hybridization is poly I-C (30 μg/ml), herring testes DNA (10 μg/ml), and calf thymus DNA (9.5 μg/ml). HeLa: human epithelial carcinoma; 18Lu: human lung fibroblast; ABE8.1/2: mouse pre-B cell lymphoma; NIH3T3: mouse fibroblast.

A number of reports showed that the expression of survivin can be regulated by growth factors/cytokines (Carter et al., 2001; Sharief and Semra, 2002), anticancer agents (Ikeguchi et al., 2002b; Notarbartolo et al., 2002), hormones (Formby and Wiley, 1999), and kinase inhibitors (Decker et al., 2003), but the underlying mechanism whereby survivin is regulated is unclear. Two interesting examples with mechanistic characterizations are discussed here.

The fact that colorectal cancers (CRCs) frequently display APC mutation, inhibition of apoptosis and increased expression of survivin, provides a possibility that inhibition of apoptosis by mutant APC may employ the deregulated survivin expression and that wild type APC (wt-APC) may repress survivin expression in normal colonic epithelium. Using the HT-29 CRC cell line with a mutant endogenous APC but containing a zinc-inducible wt-APC (HT29-APC), Zhang et al., 2001 found that induction of wt-APC expression in HT29-APC cells but not in a control cell line with an inducible lacZ gene, suppresses both survivin mRNA and protein expression within 8 h after zinc treatment, and that transfection-mediated expression of a dominant-negative TCF-4 gene construct in HT29 cells to block endogenous TCF-4-induced gene transcription can significantly repress survivin expression. These results suggest that wt-APC downregulates survivin via the APC/β-catenin/TCF-4 signaling pathway. In support of this possibility, there is a TCF-4 binding motif with a perfect match in the proximal promoter region of survivin. Using immunohistochemistry and RT-PCR, these investigators further demonstrated that in normal colonic epithelium, survivin expression is high in the crypt areas, low in the middle areas, and very weak at the surface of the crypt. This is inversely correlated with the expression pattern of wt-APC (Zhang et al., 2001). Thus, wt-APC, by progressively decreasing survivin and increasing apoptosis from crypt bottom to top, may limit the population size of the proliferative cells in the lower crypt. However, mutant APC would release the suppression of survivin expression and lead to aberrant survivin expression allowing overproduction of the proliferative stem cells and initiation of tumorigenesis. In support of this notion, investigation of the premalignant crypt phenotype in familial adenomatous polyposis patients indicated that an increase in crypt stem cell number, not changes in the rate of cell cycle proliferation, differentiation, or apoptosis of the non-stem cell population, simulates this abnormality (Boman et al., 2001).

It was reported that survivin expression is associated with mutant p53 accumulation in gastric cancer (Lu et al., 1998) and during colorectal tumorigenesis (Kawasaki et al., 2001), and that survivin inhibition by progesterone is associated with wt-p53 increase (Formby and Wiley, 1999). Consistent with these observations, it was found that wt-p53 can transcriptionally repress survivin expression and that doxorubicin-induced DNA damage employs a p53-survivin signaling pathway to regulate cell cycle and apoptosis in a number of cancer cell lines (Hoffman et al., 2002; Mirza et al., 2002; Zhou et al., 2002).

A number of different cell model systems were used to demonstrate the wt-p53-dependent suppression of survivin expression. Hoffman et al. used a human lung adenocarcinoma cell line H1299ts-p53 (Val138), which is derived from p53-null H1299 cells by stable transfection of a temperature-sensitive p53 allele encoding valine at codon 138. The p53 protein in these cells exists in a mutant conformation at 39°C, but in a wild type conformation at 32°C. They demonstrated with this cell line that survivin mRNA (Northern blot) is significantly reduced when cells are cultured at 32°C in comparison with levels from the cells cultured at 39°C. However, the XIAP mRNA level was unchanged under the same condition (Hoffman et al., 2002). Mirza et al. used a human ovarian cancer cell line 2774qw1 with a mutant p53. After infection of either replication-deficient adenovirus expressing wt-p53 (rAd-p53) or adenovirus containing empty vector (rAd-control), real time RT-PCR revealed that survivin mRNA decreased 3.6–6.7-fold over a period of 16–24-h post-infection (Mirza et al., 2002). Hoffman et al. also demonstrated that downregulation of survivin expression by p53 but not by the p53-homolog p73β, does not require induction of G1 arrest but rather associates a G2/M arrest (Hoffman et al., 2002). Survivin promoter-luciferase reporter assays revealed that a survivin core promoter (−105 to −15) harboring a potential p53-binding site is required and sufficient for conveying the inhibitory effects by p53, and deletion of this site abrogates transcriptional repression of survivin promoter by p53 (Hoffman et al., 2002). Immunobinding (McKay assay) and chromosome immunoprecipitation (CHIP) experiments demonstrated that p53 protein physically interacts with the potential p53-binding site, which is sequence-specific (Hoffman et al., 2002). Interestingly, deletion of Sin3 (a corepressor factor) binding domain of p53 impairs its ability to suppress survivin, suggesting that Sin3 recruitment of HDACs (histone deacetylases) is involved in the inhibitory effect of p53 on survivin (Hoffman et al., 2002). It should be pointed out that with a different CHIP protocol, Mirza et al. did not find p53 could physically associate the p53 binding site in survivin core promoter region (Mirza et al., 2002). However, they reached a similar conclusion that HDACs (HDAC-2 in their cells) may be involved in the repressing effects of p53 on survivin expression (Mirza et al., 2002).

A critical issue is what are the consequences of the repression of survivin expression by p53? Zhou et al. explored the cell cycle regulation and apoptosis induced by doxorubicin treatment in a set of ALL (acute lymphoblastic leukemia) cell lines (Zhou et al., 2002). They found that doxorubicin treatment of ALL cells induces a p53 status-dependent cell cycle arrest and apoptosis. EU-6 (mutant p53) and EU-4 (p53 null) cells exhibited a G2/M arrest without significant sub-G1 apoptotic cell increase within 24-h treatment by doxorubicin. In contrast, Eu-3 (wt-p53) exhibited a continuous sub-G1 increase with an apparent G2/M-associated cell death under the same treatment over a 24-h treatment (Zhou et al., 2002). These investigators demonstrated that doxorubicin treatment was associated with wt-p53 accumulation and suppression of survivin expression, and that downregulation of survivin by doxorubicin-induced wt-p53 accumulation triggers an apoptosis pathway including caspase-3 activity, PARP cleavage, and enhanced Annexin V staining. Survivin antisense-mediated downregulation of survivin in mutant p53 cells or forced expression of wt-p53 in p53 null cells could recapitulate the survivin pathway in the sensitization of these cells to doxorubicin treatment-induced apoptosis (Zhou et al., 2002). Thus, p53-survivin pathway appears to be an important pathway for p53 manipulation of apoptosis. Loss of wt-p53 function in tumor cells may contribute to up-regulation of survivin and resistance to anticancer agents (Zhou et al., 2002).

Post-transcriptional regulation of survivin

It is interesting to consider the post-transcriptional control of survivin at the pre-mRNA splicing level in light of the diverse aspects of survivin function and localization covered above. The survivin gene has four exons and alternative splicing of its pre-mRNA can produce three different survivin mRNAs, which encode three distinct survivin proteins (survivin, survivin-2B, and survivin-ΔEx3 in human; survivin140, survivin121, and survivin40 in mouse) (Mahotka et al., 1999; Conway et al., 2000). Human survivin has 142 amino acids (aa) derived from exon 1–4, survivin-2B (167 aa) has an extra 23 aa derived from a 69 bp cryptic exon (2B) within intron 2 (spliced into survivin mRNA between exon 1 and 2), and survivin-ΔEx3 (137 aa) is a frameshift readthrough variant due to exon 3 escape (Figs. 1 and 3) (Mahotka et al., 1999). In contrast, while mouse survivin (survivin140) is derived from exon 1–4, survivin121 is from exon 1–3 and part of intron 3, and survivin40 is a frameshift-truncating variant due to exon 2 escape (Conway et al., 2000). These variants are broadly expressed in cancer cell lines together with the “ortho-” survivin. The general ratio of survivin: survivin-2B: survivin-ΔEx3 in human cancer cell line is about 5–7:1:1 (unpublished data). Currently, it is not known how these survivin variants are regulated and how they function in comparison with survivin. Limited data suggest that survivin-2B might act as a dominant negative and survivin-ΔEx3 might be a more regulated variant of the parental molecule. In soft tissue sarcoma (STS), RT-PCR revealed that 36 out of 56 STS (64%) were survivin positive but no survivin-2B could be detected. Only 15 out of 36 survivin positive STS samples could be detected with survivin-ΔEx3 expression (Kappler et al., 2001), suggesting survivin and survivin variants are differentially regulated in STS. In gastric cancer, RT-PCR from 30 different gastric carcinoma samples indicated that all gastric carcinomas, irrespective of their histological types, grades, or stages, express survivin-ΔEx3, survivin-2B and survivin, the latter being the dominant transcript (Krieg et al., 2002). However, in comparing the disease stages I + II with III + IV, expression of survivin and survivin-ΔEx3 remained unchanged while there was a significant stage-dependent decrease in the expression of survivin-2B (Krieg et al., 2002). These limited observations imply that exploring the regulation and function of survivin variants may become an important research area in the future.


There are multiple strategies to utilize survivin gene as a cancer therapeutic target as survivin is overexpressed in cancer but undetectable in normal, differentiated adult tissues (Table 1) (Velculescu et al., 1999). To this end, the survivin promoter may be an ideal tool to specifically drive cytotoxic gene expression in cancer cells by gene therapy. Consistent with this notion, Bao et al. used a survivin promoter to drive a human secreted alkaline phosphatase (SEAP) reporter gene and demonstrated that the survivin promoter was more active in all cancer cell lines tested than in normal ovarian surface epithelial cells or mouse 3T3 cells (Bao et al., 2002). Using a human A2780 ovarian cancer cell line stably transfected with the survivin promoter-SEAP reporter construct, they further revealed that the injected cells in the mouse ovaries show increased SEAP activity with time and tumor growth (Bao et al., 2002). Using a survivin promoter–luciferase reporter construct, Zhu et al. addressed a key point about the potential in vivo toxicity to host tissues for gene therapy by survivin promoter. They showed that survivin promoter activity is not only “liver off” but also very low in other organs (Zeng Bian Zhu, University of Alabama at Birmingham, 2003, personal communication). Furthermore, Yang et al. has made a survivin promoter construct driving the expression of an autocatalytic Rev-caspase 3 gene. Expression of this construct specifically kills tumor cells but not normal cells (Lily Yang, Emory University, 2003, personal communication).

It has been demonstrated that the expression of survivin in cancer associates with poor prognosis, cancer progression/recurrence, drug resistance, and a short patient survival rate (Table 1). These observations triggered enthusiasm in the scientific community for using survivin itself as a target for cancer therapeutics. The current approaches are summarized below:

  • 1Survivin antisense approach. This approach was either a survivin antisense cDNA expression vector (expressing a partial or full-length survivin cDNA in the reverse direction) (Ambrosini et al., 1998; Kanwar et al., 2001; Yamamoto et al., 2002) or synthesized survivin antisense oligonucleotides (SAO)(Li et al., 1999; Chen et al., 2000; Olie et al., 2000; Xia et al., 2002). The first approach is suitable for making cancer cell models for validation of in vivo survivin function. The latter is a more suitable approach for pre-clinical, in vitro and pharmacological use as well as cancer treatment although the formulation of the antisense oligonucleotides remains to be improved.
  • 2Survivin dominant negative mutant (DNM) approaches. To date, the most effective survivin dominant negative mutant is the survivin CDC2 phosphorylation site mutant (survivin-T34A), which was used in a number of cell and mouse models (O'Connor et al., 2000a, 2002; Grossman et al., 2001b; Mesri et al., 2001b). The initial work also identified a survivin BIR domain-disrupting mutant (survivin-C84A), which has also been used in a number of models (Li et al., 1998, 1999; Grossman et al., 1999a; Kanwar et al., 2001). It must be emphasized that this approach is a good tool for research to demonstrate the principles of the survivin pathway, which may be used for therapeutic application such as combination of anticancer drugs with CDC2 inhibitor (O'Connor et al., 2002), but the survivin dominant negative mutant tools are not likely to be suitable for direct cancer therapy.
  • 3Survivin ribozyme approach. This approach involves survivin mRNA-specific ribozyme cleavage of survivin mRNA to decrease survivin expression (Pennati et al., 2002; Choi et al., 2003). This may be an effective alternative tool for research purposes but also appears unsuitable for clinic treatment of cancer.
  • 4Survivin RNA interference (RNAi) approach. Recently studies suggest that RNAi technology is a powerful approach to silence mammalian gene expression for gene function studies (Zamore, 2001; McManus and Sharp, 2002). Two approaches can effectively inhibit expression of the targeted genes in mammalian cells without activation of the non-specific interferon response. These approaches are the in vitro synthesized 21–25 nucleotide (nt) double-stranded RNAs with 2-nt 3′ overhangs (small inhibitory RNA, siRNA) or the vector-based expression of the 21–29 nt short hairpin double-stranded RNA (shRNA) driven by a RNA polymerase III-mediated promoter from the human H1 or U6 RNA genes. Limited studies suggest that survivin expression can be inhibited by both approaches (X. Ling and F. Li, 94th AACR Annual Meeting, 2003). It appears that both shRNA and siRNA approaches will be developed into a regular, useful tool for studying gene function in research. However, since synthesis of RNA oligonucleotides is much more expensive than that of DNA oligonucleotides, the application of the siRNA approach will be restricted even though there is potential application for siRNA to human diseases. Here, the author would like to point out that although there is a possibility for survivin gene therapy by forced expression of survivin antagonists (antisense, DNM, ribozyme, and shRNA), this may not be developed into a clinic routine approach for cancer treatment.
  • 5Approaches using small organic compounds or other small antagonists such as small peptides. This will be a very exciting area in the coming years to pursue and would also appear to be the most practical way to suitable approaches for clinical application. The potential small chemical molecules include those that either transcriptionally or post-transcriptionally inhibit survivin expression or abrogate survivin function such as disruption of survivin–caspase interactions. The first example for a chemical molecule to inhibit survivin expression is doxorubicin (adriamycin) (Hoffman et al., 2002; Mirza et al., 2002; Zhou et al., 2002). There is no example so far for any small molecule that directly disrupts the survivin–caspase-9 interaction. Indirect disruption of survivin–caspase-9 apoptosome interaction can be obtained by abrogation of CDC2 kinase activity with purvalanol A (O'Connor et al., 2002). However, CDC2 phosphorylates many substrates, thus inhibition of its kinase activity may not be applicable to cancer therapeutics targeting survivin.
  • 6Survivin-derived peptide-specific cancer immunotherapy. This is an emerging area of research that has generated high interest. A detailed review with discussion of this topic is provided below.

The expression of survivin in cancer (Table 1) but not in normal, differentiated adult tissues (Ambrosini et al., 1997; Velculescu et al., 1999) also implies that survivin is a potential tumor-associated antigen (TAA) for tumor vaccination. Since it is known that the activated CD8+ cytotoxic T lymphocytes (CTL) can recognize and lyse tumor cells that present peptide epitopes derived from TAA, it is critical to demonstrate that the survivin peptide presenting cells such as monocyte-derived dendritic cells (DCs) can activate specific naïve T cells to generate cytotoxic T cells. To this end, Schmitz et al. used recombinant His-tag-purified survivin protein to pulse (5–10 μg/ml for 3 days) monocyte-derived DCs and stimulated the CD8+ T cells with the survivin-pulsed DCs in vitro. They demonstrated that such stimulated CD8+ CTLs could efficiently lyse a survivin-negative EBV-transformed B lymphoblastoid cell line (EBV-BLCL) transfected with a survivin expression vector but not with the empty expression vector control (Schmitz et al., 2000). They further identified a survivin peptide epitope (95ELTLGEFLKL104, Sur9), which mediates survivin peptide-loaded (peptide-transporter-deficient) T2 cell lysis by CD8+ CTLs that are activated either by survivin peptide-pulsed DCs or by recombinant survivin protein-pulsed DCs. In addition, the CD8+ CTLs exhibited a selectivity to lyse EBV-BLCL transfected by survivin expression vector but not by empty vector as well (Schmitz et al., 2000), suggesting Sur9 may be a bonafide endogenous epitope for HLA-A*0201 (HLA-A2). Together, these observations demonstrated the concept that survivin-derived peptide epitopes presented on DCs could mature CD8+ T cells into CTLs for efficiently killing survivin-positive tumor cells.

These in vitro studies (Schmitz et al., 2000) were supported and extended by two recent reports that test the spontaneous CTL responses against survivin-derived peptide epitopes in cancer patients (Andersen et al., 2001a,b). Using an assay based on loading an effective peptide into T2 cells will stabilize the class I MHC molecule and in turn the cells will be lysed by CTL (Fig. 5), the authors screened for potential survivin-derived complementary peptides to the HLA-A2, and identified a number of such peptides exhibiting an affinity to HLA-A2. These peptides include Sur1 (96LTLGEFLKL104) and Sur3 (130KVRRAIEQL138) with a C50 (concentration of a peptide required for half-maximal binding to HLA-A2) of over 100 μm, Sur6 (101FLKLDRERA109), and Sur8 (5TLPPAWQPFL14) with a C50 of 30 μm and Sur9 (95ELTLGEFLKL104) with a C50 of 10 μm, as well as two Sur1 analogs, Sur1L2 and Sur1M2 (C50, 1 μm), in which the natural T of Sur1 at position 2 was replaced with L or M (Andersen et al., 2001a). Using survivin peptides Sur1, Sur9, and Sur1M2 in an ELISPOT assay which measures the numbers of survivin peptide-specific IFN-γ-releasing effector cells (CTL), they demonstrated that peripheral blood lymphocytes (PBLs) isolated from 6 of 7 HLA-A2 positive chronic lymphatic leukemia (CLL) patients cross-reacted with Sur1 (2/6 samples), Sur9 (6/6), and Sur1M2 (5/6), and that PBL isolated from 6 of 10 HLA-A2 positive breast cancer patients cross-reacted with Sur1 (2/6), Sur9 (3/6), and Sur1M2 (3/6) (Andersen et al., 2001a,b). Interestingly, T cells isolated from tumor-infiltrated lymph nodes from 3 of 6 melanoma patients (50%) cross-reacted with sur1 (2/3 samples) and Sur9 (3/3) (Andersen et al., 2001a), and similarly, PBLs isolated from 7 of 14 melanoma patients (50%) cross-reacted against Sur1 (2/7), Sur9 (3/7), and Sur1M2 (5/7) as well (Andersen et al., 2001b). These results suggest that reactive T cells (CTL) could effectively release from the original site into the circulation system of the melanoma patients.

Figure 5.

Cancer immunotherapy with survivin-derived peptide epitopes. The data from recent publications showed that (1) CD8+ naïve T cells can be primed by survivin-derived antigen presenting cells (monocyte-derived dentritic cells and cancer cells) in the site of primary tumor, and be possibly able to proliferate since small CTL clusters can be found in the primary tumor tissues; (2) survivin antigen-activated CD8+ T cells (CTL) can also be detected in the sentinel lymph notes containing metastatic solid tumor cells (tumor-infiltrated lymph notes); (3) tumor cells with different origin (such as breast cancer versus melanoma) appear to be able to present a similar set of survivin-derived peptide epitopes since CTLs isolated from one origin can lyse HLA-matched tumor cells from different origin. These observations suggest that survivin antigen-activated T cells (CTL) can be released from lymph nodes into circulation system, and home to or stay in the site of primary tumor with a possible proliferation on the site. These findings lay an interesting new research area for tumor immunotherapy with survivin-derived peptide epitopes.

Several interesting points may be worthy to mention. First, the fact that in most cases, Sur1 and 9 do not cross-react against reactive T cells (CTL) of the same patient suggests a proportion of T cells can distinguish the two peptides. On the other hand, the CTL from a few patients can respond to both peptides. This suggests that both Sur1 and Sur9 are processed and presented but that the immunodominance of these peptides may be distinct from patient to patient. Second, the C50 value of a peptide binding to HLA-A2 may not always be inversely proportional to the response intensity in the ELISPOT assay. For example, the PBL from one melanoma patient strongly reacted with Sur1 but showed weak reaction with Sur9 (Andersen et al., 2001a), and PBL from two breast cancer patients strongly reacted with Sur1 but had no reaction with Sur9 (Andersen et al., 2001b). Finally, the PBLs' response rate from CLL patients (6/7, 86%) was quite different from those of the melanoma and breast cancer patients in which only 50% of the patients showed a response. This could be due to several possibilities. The patients may have had a very low spontaneous CTL response. If this is the case, investigation of the underlying mechanism is intriguing. Alternatively, distinct class I MHC-restricted survivin peptide epitopes may mediate the responses and be presented by HLA-A2 or other HLA in these patients. Consistent with this possibility, Hirohashi et al. recently identified a survivin-2B-derived peptide epitope (survivin-2B80-88, AYACNTSTL) that shows a distinct affinity to HLA-A24 (Hirohashi et al., 2002). They demonstrated that survivin-2B80-88-loaded monocyte-derived autologous DCs could prime CD8+ T cells isolated from the blood of both 5 HLA-A24-positive patients with distinct cancers and 2 of 4 HLA-A24-positive healthy donors, and that the primed T cells (CTL) can efficiently lyse survivin-2B80-88-loaded HLA-A24-positive cells but not the survivin-2B80-88-loaded HLA-A24-negative cells (Hirohashi et al., 2002). This observation suggests that survivin or survivin variant-derived peptides for distinct HLA types may also be important for epitope-specific cancer immunotherapy. Therefore, alternative experimental approaches such as proteasome-mediated digestion analysis (Kessler et al., 2001) of survivin or survivin variant-derived peptides may help to determine new and potentially effective peptide epitopes.

The feasibility for cancer vaccination using survivin-derived peptide epitopes were further strengthened by the fact that HLA-A2/survivin-reactive T cells can be located in the primary cancer tissues and that the isolated reactive CTLs from one origin can lyse HLA-matched tumor cells from a distinct origin (Fig. 5) (Andersen et al., 2001b). These investigations utilized probes generated with biotinylated HLA-A2 bound with Sur1M2 or with melanoma-associated antigen peptide gp100154-163 (control) which were then multimerized with streptavidin–FITC-conjugated dextran molecules to generate multivalent HLA–dextran compounds (MHC/Sur1M2-complexes). Immunochemical and confocal analysis of the primary melanoma tumor lesion, the sentinel lymph node of a stage III melanoma patient, and a primary breast cancer lesion detected the presence of survivin-reactive CTLs with small clusters from the CD8+ T cell pool in tumor tissue or lymph note. In contrast, the MHC/gp100154-163-complexes could not locate any positive CTL in the same breast tumor lesion nor could MHC/Sur1M2-complexes locate positive CTL from the HLA-A2-negative tumor lesion (Andersen et al., 2001b). This finding is important as it suggests that survivin peptide-presenting cells in vivo (such as DCs and cancer cells) can prime the CD8+ T cells, which can home to (and stay in) the site of action, and that a local expansion of survivin antigen-activated CTLs is possible. Furthermore, CTLs purified from melanoma-infiltrated lymph nodes by magnetic beads coated with HLA-A2/sur1M2 complexes can specifically lyse both HLA-A2-positive melanoma (FM3) and breast cancer (MCF-7) cells but can not lyse HLA-A2-negative melanoma (FM45) or breast cancer (BT-20) cells (Andersen et al., 2001b). This finding strongly suggests that cancer cells from distinct origins are able to process and present the same endogenous survivin peptide epitopes. Therefore, identification and validation of the most effective endogenous proteosome-processed survivin peptides or modified ones will be critical for survivin antigen-specific cancer immunotherapy.

An important concern is the potential side effects that might occur following survivin-derived peptide immunization. However, there is little evidence for toxicity thus far. This may not be surprising since the expression of survivin is largely restricted to tumor tissues (Table 1). Second, immune reaction against survivin peptides can only be detected in HLA-A2-positive tumor patients but not in HLA-A2-positive healthy donors and neither of the patients in the study showed any signs of autoimmunity although they hosted a T-cell response against survivin (Andersen et al., 2001a,b). Finally, consistent with these observations, 11 of 51 sera from lung cancer patients (21.6%), 4 of 49 sera from colorectal cancer patients (8.2%) (Rohayem et al., 2000) and 25 of 63 sera from gastrointestinal cancer patients (39.7%) (Yagihashi et al., 2001) reacted with purified recombinant survivin detected by ELISA but the reaction was negative with sera from all 93 healthy blood donors (Rohayem et al., 2000; Yagihashi et al., 2001). These results suggest survivin is a major cancer antigen and is able to initiate both T and B cell responses in cancer patients. Interestingly, it seems that the production of anti-survivin antibody in patients is not proportional to the survivin expression in a particular cancer. For example, survivin expression is observed in 53–63.5% in colorectal cancer, 85.5% in lung cancer, and 70.7–90.2% in breast cancer (Table 1). However, reactivity of patient sera with survivin protein is 4% (4/49), 21.6% (11/51) (Rohayem et al., 2000), and 6.3% (1/16) (Yagihashi et al., 2001) #140 in these tumor types, respectively. This might suggest that the type and site of tumor origin can affect B cell responses for the production of anti-survivin antibody. However, it is also possible that this may not be relevant since different patients were involved in these studies, and that some of these differences may merely reflect the MHC alleles that are present in patients with different types of tumor and those may determine the response to survivin peptides.

In addition, based on the fact that molecular targeting of survivin in cancer induces cell death, combination of molecular targeting of survivin with immunotherapy may significantly improve the previously used immunotherapy approaches. To this end, Kanwar et al., 2001 #54found that large tumors (ϕ 1 cm) established from a mouse EL-4 thymic lymphoma cell xenograft show the presence of NK cells and high levels of survivin in comparison with small tumors (ϕ 0.2 cm). Intratumoral expression of mouse survivin antisense vector in large tumor xenografts stimulated the production of infiltrating CD8+ leukocyte and moderately enhanced CD4+ T cells by checking the splenocytes, whereas intratumoral expression of mouse survivin C84A dominant negative mutant vector resulted only in a similar level of CD4+ T cells. However, intratumoral expression of B7-1 gene vector in large tumor xenografts stimulated the production of CD8+, CD4+, and NK cells). Combination of survivin targeting with B7-1 immunotherapy speeds up the complete regression of small tumors and inhibits large tumor growth. Interestingly, molecular targeting by survivin antisense seems much more effective than the C84A mutant for large tumors in these studies (Kanwar et al., 2001). The reason for this result may be that survivin antisense targeting can prime CD8+ T cells but survivin C84A mutant does not. However, why survivin antisense could do this but not the survivin C84A mutant, is not clear. Although the authors stated that determination of the particular leukocyte subset produced may provide insights into the mechanisms responsible for antitumor immunity, the reason that survivin antisense but not survivin C84A mutant stimulated CD8+ T cells in xenograft mice was not provided in the study.

In summary (Fig. 5), the current data suggest that (1) tumors of different origin appear to be able to process and present the same set of survivin-derived peptide epitopes; (2) CD8+ T cells and monocyte-derived DCs can home to the primary solid tumor site, which provides the feasibility for cancer cells to act as survivin-antigen presenting cells to prime CD8+ T cells; (3) metastatic tumor cells can act as survivin antigen presenting cells to mature CD8+ T cells in the immune system (lymph node) into CTL, and these can be released into circulation system and home to the primary tumor site.


Survivin, endothelial cell, and angiogenesis

While the baseline level of survivin expression in normal endothelial cells (EC) is barely detectable, vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) appeared to induce survivin mRNA and protein expression in the range of 10–20-fold (Tran et al., 1999; O'Connor et al., 2000b). In contrast, these growth factors had no significant effect on XIAP (Tran et al., 1999), and tumor necrosis factor α (TNF-α) and interleukin-1 could not induce survivin expression in EC (O'Connor et al., 2000b). Three-dimensional EC culture experiments indicated that formation of three-dimensional vascular tubes in a rat-tail type I collagen/fibronectin matrix gel is associated with strong induction of survivin (O'Connor et al., 2000b). Immunohistochemical analysis of skin biopsies containing granulation tissue or normal non-inflamed skin by anti-survivin antibody revealed that survivin is strongly expressed in the cytoplasm of EC of newly formed capillaries of skin tissue as well as in large vessels. In contrast, analysis of non-proliferating capillaries of non-inflamed normal skin revealed a minimal expression of survivin in EC (O'Connor et al., 2000b). In this context, antisense oligonucleotide targeting of survivin expression induced by VEGF suppresses VEGF-mediated EC protection and vascular tube-like structure formation (Mesri et al., 2001a).

Interestingly, unlike most angiogenic regulators such as bFGF and VEGF, angiopoietin-1 (Ang-1) does not stimulate EC growth but rather promotes stabilization of vascular networks and branching morphogenesis in vivo and in vitro. However, the functional mechanism is not very clear (Yancopoulos et al., 2000; Ward and Dumont, 2002). Papapetropoulos et al., 2000 found that treatment of EC enhances Akt phosphorylation, which could be inhibited by the PI3 kinase inhibitor, Wortmannin or by a dominant negative mutant of the PI3 kinase p85 subunit. They also reported that Ang1-stimulated phosphorylation of Akt is antagonized by preincubation of Ang-1 with soluble Tie2 receptor but not with soluble Tie-1, suggesting the event is through a Ang1/Tie2/PI3 kinase/Akt pathway. They observed that Ang-1 upregulates survivin mRNA and protein expression but not Bcl-2 through the PI3/Akt pathway, and that the survivin increase is required for EC viability and cytoprotection against apoptosis induced by a combination of tumor TNF-α and cycloheximide (CHX) (Papapetropoulos et al., 2000). Consistent with these findings, a recent report showed that Ang-1 through Tie-2 receptor attenuated serum deprivation-induced apoptosis by inhibition of the activation of caspase-9, -7, and -3, but not caspase-8. This can be reversed by the PI-3 kinase inhibitor Wortmannin (Harfouche et al., 2002), suggesting survivin may be involved in this process. Ang-1 exposure upregulated the expression of survivin, but not XIAP and Bcl-2 family proteins, reduced the cytosolic levels of Smac, but not that of cytochrome c (Harfouche et al., 2002).

It was also shown that induction of survivin expression by VEGF and bFGF through a PI3/Akt-dependent pathway in EC is cytoprotective against apoptosis induced by chemotherapeutic drugs (Tran et al., 2002). This work also indicated that survivin expression maintains microtubule structure integrity in EC, which may be required for EC viability and cytoprotection (Tran et al., 2002). Another report showed that hypoxia followed by reoxygenation accelerates EC tubular morphogenesis and induces survivin expression. Survivin induction can be inhibited by NF-kB and PI3 kinase inhibitors (Zhu et al., 2001). While the PI3/Akt pathway was shown to be involved in survivin induction by VEGF, bFGF, Agn-1, and hypoxia/reoxygenation, interleukin-11 (IL-11) could upregulate survivin expression through a Stat3-dependent pathway in human umbilical vein EC and in IL-11 transgenic mouse lung tissues as well as in a human abdominal skin xenograft mouse model (Mahboubi et al., 2001). Overexpression of a phospho-site mutant of Stat3 but not a phospho-site mutant of Stat1 could ablate the effect of IL-11 on survivin induction (Mahboubi et al., 2001). Taken together, these observations suggest a role of survivin in maintaining cell viability and cytoprotection in both proliferative and non-proliferative states of EC.

In summary, given that many types of tumors excrete excessive VEGF and that ablation of apoptosis initiates tumorigenesis, these results indicate the feasibility that targeting survivin for solid tumor treatment may have advantages over other therapeutic targets in that both induction of tumor cell death and regression of tumor vascular network may occur simultaneously. On the other hand, excessive apoptosis in EC was shown to be involved in many EC-associated diseases including atherosclerosis (Choy et al., 2001; Ruberg et al., 2002). In this context, pharmacological protection of EC viability may improve the EC-associated cardiovascular diseases (Guevara et al., 2001).

Survivin, smooth muscle cell (SMC), and neointima formation

Interestingly, while survivin expression in EC appears to have a role in protection of EC viability, survivin may also participate in SMC proliferation and neointima formation in acute vascular injury (Blanc-Brude et al., 2002). Using both rabbit balloon iliofemoral artery injury and mouse wire femoral artery injury models, Blanc-Brude et al. (2002) showed that injury-induced survivin expression is as early as 2 days after injury and neointima becomes visible at day 7. Maximum survivin expression and neointima formation were obtained at 14 days after injury and thereafter decreased but was sustained beyond 6 weeks. Immunohistochemical analysis indicated that survivin induction appears to parallel PDGF-B production. Consistent with these observations, these investigators showed that PDGF-AB induces survivin expression in SMC, which can be modulated by Actinomycin D, MEK, and PI3 inhibitors (PD098059 & LY294002) but not by p38 MAPK inhibitor (SB203580) (Blanc-Brude et al., 2002), suggesting that PDGF-AB transcriptionally upregulates survivin at least in part through both the Erk1/2 and Akt signaling pathways. What is the significance of survivin induction by PDGF-AB? Further studies demonstrated that SMC treated with PDGF-AB or infected with replication-deficient adenovirus-survivin expression vector counteracts apoptotic SMC death induced by C2-ceramide or TNF-α/CHX treatments (Blanc-Brude et al., 2002). Alternatively, they showed that expression of a survivin phospho-site mutant (T34A) reverses the protective effect of either PDGF-AB or serum on SMC and induces SMC apoptosis (Blanc-Brude et al., 2002). Similarly, adenovirus-mediated delivery of survivin T34A in vivo at the injury sites could inhibit neointima formation and induce apoptosis as well (Blanc-Brude et al., 2002). These in vitro and in vivo observations imply that survivin reexpression during acute artery injury may also have a role in the initiation of chronic vascular diseases including atherosclerosis.

Growing evidence has revealed that the formation of atherosclerosis and cancer shares many similarities (Ross et al., 2001). (1) Oxidative stress-induced cellular damage has been implicated in atherogenesis and neoplasia, and cigarette smoking and excessive fat intake are associated with the pathogenesis of both diseases; (2) cell proliferation pathways including genes involved in the G1/S transition checkpoint have been linked to plaque progression, stenosis, and restenosis after angioplasty as well as in cancer progression; (3) alterations in cell adhesion molecules (integrins, cadherin–catenins) have been also linked to plaque formation and thrombosis as well as to tumor invasion, progression, and metastasis; (4) altered expression of proteases such as t-PA and u-PA has been implicated in atherosclerotic plaque expansion and hemorrhage and in the invasion and metastasis of malignancy; (5) ligand–receptor kinase interactions and downstream signaling such as MAPK and Akt pathways have been associated with early atherosclerotic lesions as well as cancer development and spread; (6) angiogenesis modulators have recently been connected to plaque expansion and restenosis of atherosclerotic lesions as well as local and metastatic tumor expansion. In addition, established anticancer treatments, such as the use of growth factor inhibitors and radiation, were recently introduced into atherosclerosis therapeutic strategies to prevent restenosis after angioplasty and endarterectomy. These similarities support a notion that atherosclerosis and cancer are far more closely aligned than previously believed and that many therapeutic approaches used for treatment of cancer may be efficacious for cardiovascular diseases including atherosclerosis. In this regard, it is not surprising that survivin has been demonstrated to be involved in both cancer and vascular disease process. Further elucidation of the role of survivin in cardiovascular diseases will enrich our understanding of survivin function.


The rapidly growing numbers of publications on survivin study in the last 6 years (Fig. 6) suggest that survivin has indeed attracted a substantial attention in the cancer research community. This is largely attributed to its unique structure (special BIR and homodimer), expression (cancer-associated), regulation (cell cycle-associated), and function (regulating apoptosis as well as cell division). Overall, these studies can be categorized into the following five research areas:

Figure 6.

The growing numbers of survivin publications each year from 1997 to 2002.

  • 1The expression pattern of survivin in various cancers. Substantial data has been collected (Table 1). Although some inconsistent reports exist, the overwhelming reports support the concept that survivin expression in cancer is associated poor prognosis, cancer progression/recurrence, drug resistance, and a short patient survival rate. The information collected in these studies will provide a basis for guiding clinical applications. New research directions in this category involving cancer will probably focus on the association of particular cancers with the differential expression of survivin variants.
  • 2Survivin subcellular localization and function. This will become an even more important area for active investigation in the coming years. The finding that survivin exists in immunochemically distinct subcellular pools can provide explanations for current inconsistent observations on survivin subcellular localization during mitosis. However, this finding has raised more new questions than have been solved. For example, what is the distinct role of two pools of survivin and how are they controlled? Why do some antibodies only recognize one of the two survivin pools while others recognize both? Is the localization of survivin on the mitotic spindle only for anti-apoptosis through inhibition of caspase-9 or is it involved in microtubule stability and/or chromosome segregation as well? Is survivin localization on the centromere/kinetochore also involved in chromosome segregation and cytokinesis or in microtubule polymerization and mitotic spindle formation/stability or in both? What is the localization and function of the two survivin variants (survivin-2B and survivin-ΔEx3) in these processes? If any of these questions generates a “yes” answer, the next question will be how does it work? Some of these questions will probably be answered or partially answered in the coming years.
  • 3Regulation of survivin expression. In view of the complex localization and function of survivin, control of its expression may be an important alternative approach to modulate survivin function in cancer treatment. Although data accumulated thus far suggest that the expression of survivin in cancer is largely controlled at the transcriptional level, the molecular mechanism by which the survivin gene is regulated in cancer is largely unknown. Extensive efforts will be required in the coming years to catch up with the need for cancer therapeutics through control of survivin expression.
  • 4Survivin and cancer therapeutics. This is an immature area so far but likely to get substantial attention and progress in the coming years.
  • 5Survivin and cardiovascular diseases. Although a number of interesting reports have been published in this category, the potential role of survivin in the development of cardiovascular diseases will require substantial investigation and validation. It is likely getting more attentions in the coming years.


The author thanks Dr. Michael G. Brattain for helpful comments and critical reading of the review, and the colleagues and collaborators who have shared their data prior to publication for inclusion in this article.