Hypoxia drives malignant progression in part by promoting accumulation of the oncogenic transcription factor hypoxia inducible factor–1α (HIF-1α) in tumor cells. Tumor aggressiveness also relates to elevation of the cancer stem cell–associated membrane protein CD24, which has been causally implicated in tumor formation and metastasis in experimental models. Here, we link these two elements by showing that hypoxia induces CD24 expression through a functional hypoxia responsive element in the CD24 promoter. HIF-1α overexpression induced CD24 mRNA and protein under normoxic conditions, with this effect traced to a recruitment of endogenous HIF-1α to the CD24 promoter. Short hairpin RNA–mediated attenuation of HIF-1α or CD24 expression reduced cancer cell survival in vitro and in vivo at the levels of primary and metastatic tumor growth. CD24 overexpression in HIF-1α–depleted cancer cells rescued this decrease, whereas HIF-1α overexpression in CD24-depleted cells did not. Analysis of clinical tumor specimens revealed a correlation between HIF-1α and CD24 levels and an association of their coexpression to decreased patient survival. Our results establish a mechanistic linkage between 2 critically important molecules in cancer, identifying CD24 as a critical HIF-1α transcriptional target and biologic effector, strengthening the rationale to target CD24 for cancer therapy. Cancer Res; 72(21); 5600–12. ©2012 AACR.
Tumor hypoxia is a poor prognostic factor for patient outcome (1, 2). This is likely from hypoxic induction of angiogenic factors, chemokines, oncogenes, and other drivers of tumor progression (3) that confer metastatic competence (4). CD24 is a mucin-like cell surface protein consisting of a short, heavily glycosylated core (5) linked to plasma membrane raft domains via glycosyl–phosphatidylinositol. In immune lineages, CD24 has been implicated in signal transduction and lymphocyte development (6, 7). Multiple studies have also shown that CD24 is expressed in many human malignancies including breast, ovarian, colorectal, pancreatic, prostate, bladder cancer, and higher levels of expression are associated with increased aggressiveness and worse prognosis (5, 8). Although in breast cancer, ambiguity exists as cells expressing CD44+ CD24(−/low) display more stem cell likeness (9), studies have shown that CD24+ cells are associated with poor prognosis in breast (9) and most other epithelial malignancies (5). Moreover, recent studies in pancreatic, colon, and hepatocellular cancers have identified tumor-initiating CD24+ populations that are capable of self-renewal, differentiation, and metastasis (10). These CD24 phenotypes have been delineated mechanistically by use of xenograft studies showing how expression of CD24 in tumor cells enhances their metastatic ability (8, 11).
Despite its importance in cancer biology and clinical oncology (12, 13), little is known about the regulation of CD24. Having recently identified CD24 as an important transcriptional target (8) of the Ral GTPase pathway, here, we used bioinformatic analyses to initially associate a transcriptional signature of the Ral pathway with hypoxia inducible factor–1α (HIF-1α) signaling. On the basis of these suggestive data, we then tested the hypothesis that CD24 might also be regulated by HIF signaling. Using bioinformatic, in vitro, in vivo, as well as analysis of human specimens in this study, we show an important mechanistic link between 2 pivotal genes in tumor progression, HIF-1α and CD24, and identify the latter as an effector of HIF-1α–induced tumor growth and lung colonization.
Materials and Methods
Statistics and informatics
Group comparisons were made using 1-way ANOVA and Newman–Keuls test unless otherwise indicated. Survival estimates in animal experiments were compared by log rank test of equality of survival distributions. For CD24 and HIF-1α staining, Spearman rank-based correlation was used to test correlation, whereas survival distributions were plotted and tested by the log-rank test (Prism Version 5.0; GraphPad Software) or Cox proportional hazards regression analysis (Matlab Version 2010b; The Mathworks). Molecular concepts analysis (MCA; ref. 14) was conducted using genes overexpressed at least 2-fold in a core transcriptional signature of the Ral GTPase pathway (n = 32 unique genes; ref. 15; Supplementary Fig. S1) by uploading into Oncomine (16) as a custom concept. All Entrez gene IDs were used as the null set (n = 42490). Pairwise associations of the Ral Signature with all molecular concepts in Oncomine (including Oncomine gene expression signatures, Gene Ontology, KEGG Pathways, Human Protein Reference Database, etc.) were automatically computed. Overlap with molecular concepts in Oncomine was considered significant at P < 1E−4, and selected concepts were exported into Cytoscape v2.8.2 for network visualization.
Cell culture and hypoxia
All cancer cell lines used were purchased from American Type Culture Collection, cultured, and profiled as described (17). UMUC-3 CD24 cells were generated by stably transfecting CD24 construct in pcDNA 3.1 vector. Short hairpin RNA (shRNA)–expressing stable UMUC-3 cell lines were generated by transfecting CD24 shRNA or HIF-1α shRNA in PLKO1 vector (Sigma-Aldrich). For hypoxic treatment, cells were incubated with 1% O2 balanced with CO2 and nitrogen in hypoxia chambers (Precision Scientific) as described (18). We identified a putative hypoxia responsive element (HRE) on CD24 promoter using a promoter scanning software Genomatix (19). To determine the HRE promoter activity, cells were lysed in passive lysis buffer (Promega) and assayed for firefly luciferase according to the Promega Luciferase kit instructions. Normalization of nuclear material was achieved by a Cyquant cell proliferation assay conducted using the same lysate following manufacturer's instructions.
Cloning of the CD24 promoter, deletion mutation, and reporter assays
Genomic DNA from several human cell lines, described in Results section, was isolated using the Puregene DNA isolation kit (Gentra Systems) according to the manufacturer's instructions. The PCR reaction also included: 10× PfuUltra HF Reaction Buffer (Agilent), 100 mmol/L dNTP Mix (Agilent), 5% dimethyl sulfoxide (DMSO), 1 mol/L betaine solution (Sigma-Aldrich), and 100 ng of DNA. The primer used was as follows: F:5′-GAT-CGC-TAG-CCA-CGC-CCG-GCC-AAA-GTA-TTT-C-3′, R:5′-GAT-CAA-GCT-TCA-GGA-TGC-TGG-GTG-CTT- GGA-G-3′. PCR products were isolated in a 1% agarose gel and purified using a Qiaquick Gel Extraction Kit (Qiagen). Isolated products were digested at 37°C by the enzymes NheI and HindIII (New England Biolabs) and ligated into the pGL4.20 Luciferase Reporter Vector (Promega). CD24 promoters were sequenced at the University of Virginia DNA Sciences Core (Charlottesville, VA) using the BigDye Terminator v3.1 (Applied Biosystems) chemistry on a 3730 DNA Analyzer (Applied Biosystems). Primers for sequencing were: RV primer 3: 5′-CTA-GCA-AAA-TAG-GCT-GTC-CCC-3′, pGL4.20 R: 5′-CTT-AAT-GTT-TTT-GGC-ATC-TTC-C-3′, CD24promSF1: 5′-CTC-CTC-TTT-GTG-CCG-GTT-CAT-T-3′, CD24promSR1: 5′-CGG-TCC-TGG-AGC-AAG-TGC-A-3′. Sequences were submitted to GenBank with the following accession numbers:
To generate deletion fragments of the CD24 promoter, specific primers were designed to amplify increasingly larger segments of the promoter region, preserving the same NheI and HindIII cloning strategy for standardization. These were subcloned to yield reporter vectors encompassing 100 bp, 223 bp, 423 bp, 680 bp, 943 bp, and 1139 bp. The CD24 promoter deletion constructs isolated above were cloned into pGL4-Basic (Promega), and HRE binding site in the promoter was mutated according to site-directed mutagenesis kit (Stratagene) instructions.
RNA isolation and real-time quantitative PCR
RNA was isolated and quantitative reverse transcription-PCR (RT-PCR) conducted in the ABI Prism 7900HT Sequence Detection System from Applied Biosystems as described (20). Primers used for CD24 were forward, 5V-CAATATTAAATCTGCTGGAGTTTCATG-3V and reverse, 5V-TCCATATTTCTCAAGCCACATTCA-3. Primers used for HIF-1α were: forward, 5′-TGCTTGGTGCTGATTTGTGA-3′ and reverse, 5′-GGTCAGATGATCAGAGTCCA-3′.
Western blotting and immunohistochemistry
Western blot analyses were conducted as previously described (21). Anti-CD24 mouse monoclonal antibody (SWA-11) was obtained as a gift from Dr. Peter Altevogt (Tumor Immunology Programme, German Cancer Research Center, Heidelberg, Germany), mouse monoclonal anti-HIF-1α antibody (BD pharmingen; Catalog No. 610958) and HIF-2α (Novus Biologicals; Catalog No. NB100-132) was used at 1:2,000 dilution, mouse monoclonal α-tubulin (Abcam) was used at 1:2,000 dilution. Immunoblots were developed using Super Signal Femto (Pierce) and visualized using Alpha Innotech imaging. A tissue microarray containing well-annotated clinical specimens with follow-up (clinical and pathologic data on specimens described in ref. 22) containing quadruplicate, representative cores from 151 zinc formalin-fixed, paraffin-embedded primary bladder cancers was evaluated. Slides were incubated with anti-CD24 (SWA-11) hybridoma supernatant with immunoperoxidase detection and DAB chromogen (23). Scoring of CD24 based on intensity was conducted as described (8), ranging from negative to strong (0, 1+, 2+, and 3+). For staining with HIF-1α, the monoclonal anti–HIF-1α antibody was used at 1:800 dilution for 30 minutes after antigen retrieval in a DAKO Pascal Pressure Chamber. Scoring for nuclear positivity was conducted (0+: none, 1+: 1–10% nuclei, 2+: 10–50% nuclei, 3+: >50% nuclei) similar to that reported previously (24). Nuclear staining was counted as positive at any intensity.
Cell viability assays
To monitor the viability of the cells, 4 × 103 cells grown in 96-well plates were stained with Calcein AM (stains live cells green) and ethidium bromide homodimer (detects dead cells) using the “Live/Dead” assay system (Molecular probes) following manufacturer's instructions. Cells incubated with dyes for 15 minutes were exposed to fluorescence intensity measurements using excitation/emission filter sets 495/520 nm (Calcein AM) and 530/642 nm (EthD). Mean fluorescence ± SEM (n = 8 wells) was estimated using BioTek Synergy 2 microplate reader. The “Viability Index” was calculated as the ratio of live fluorescence divided by dead fluorescence.
In vivo human xenograft assays in immunocompromised mice
Six-week-old nude mice were injected subcutaneously with 106 cancer cells. Tumor dimensions were measured 2 times a week and tumor volumes calculated as described (25). Five- to 6-week-old male (for PC3 cells) or female (for UMUC3 cells) athymic nude mice (Ncr nu/nu) were obtained from the National Cancer Institute (Frederick, MD). Animals were maintained according to the University of Virginia Institutional Animal Care and Use Committee (IACUC) guidelines. To evaluate the ability of HIF-1 and CD24 expression–modified PC3 cells for metastasis, 2 × 105 cells in serum-free medium were inoculated in the left ventricle. Six-week-old nude mice were given tail vein injections with 2.5 × 106 UMUC3 cells as described. In addition, in separate experiments, 106 PC-3 cells were injected intraprostatically as described (26). Bioluminescent in vivo imaging using IVIS 100 scanner (Xenogen Corp.) was used to monitor metastasis weekly as explained previously (26). d-Luciferin Firefly, potassium salt (30 mg/mL luciferin; Biosynth International) was injected intraperitoneally at a dose of 150 mg/kg body weight approximately 10 to 15 minutes before imaging. Images were acquired in 5-minute intervals using IVIS 100 scanner as described (25). The “survival” endpoint was defined according to the IACUC criteria such as significant weight loss, pain unrelieved by analgesias, tumor erosion, etc.). At euthanasia, the lungs of animals inoculated with UMUC3 cells were removed by dissection away from adjacent organs and examined as described (27). Genomic DNA was extracted from mouse lungs and 12p real-time PCR was conducted as described (20).
Pimonidazole (hypoxyprobe) immunohistochemistry
Six-week-old nude mice were injected subcutaneously with 106 UMUC-3 cells. After the tumors grew to a diameter of 1 to 1.5 cm, the animals were injected intravenously with pimonidazole hydrochloride (hypoxyprobe; 60 mg/kg). Ninety minutes after the injection, animals were sacrificed, tumors dissected into 2 or 3 portions, and fixed with 10% neutral buffered formalin overnight, and embedded in paraffin. Four-micrometer thick sections were processed and stained according to the manufacturer's instructions as described (28). Consecutive sections from the same tumors were immunostained for CD24 as explained (8).
UMUC-3 cells exposed to hypoxia for 24 hours were treated with 2% formaldehyde for 10 minutes at 25°C. Chromatin immunoprecipitation was conducted using the ChIP-IT Express Kit Chromatin Immunoprecipitation Kit from Active Motif according to the manufacturer's instructions. Briefly, genomic DNA was enzymatically sheared for 4 minutes at 37°C. Rabbit IgG (Santa Cruz Biotechnology), RNA Pol II (Active Motif), and HIF-1α (BD pharmingen; Catalog No. 610958) and HIF-1β (Novus Biologicals; Catalog No. NB300-525) were incubated with the genomic DNA and magnetic beads for 4 hours at 4°C. After the recommended washings, the DNA–protein crosslinking was reversed and the protein was removed from the samples with proteinase K. Detection of CD24 promoter was done using quantitative real-time PCR with the following primers, F: 5′-ACGGCTATTGTGGCTTTCCTGGTAT-3′; R: 5′-GCTTGGAGAACCGCTGGCTC-3′. Conditions for qRT-PCR included 40 cycles with an annealing temperature of 58°C. DNA amplification was measured using SYBR Green reaction mix supplemented with 1 mol/L betaine and 5% DMSO on Applied Biosystems (GeneAmp PCR system 9700). DNA quantity was determined on the basis of a standard curve and normalized on the basis of the amount of input DNA from each sample.
All research involving human participants was approved by the University of Virginia Institutional Review Board. Informed consent was obtained and all clinical investigation conducted according to the principles expressed in the Declaration of Helsinki.
CD24 is induced by hypoxia in cancer cells and human tumor xenografts
We recently identified CD24 as an important transcriptional target of signaling downstream from the Ral GTPase pathway, although little is known regarding the regulation of CD24. To generate testable hypotheses for regulators of CD24, we used MCA (14), implemented in oncomine (16) to analyze a transcriptional signature of the Ral GTPase pathway (15). Genes overexpressed from this signature, composed of a core set genes regulated 2-fold by siRNA-mediated depletion of both RalA and RalB (29), were compared with other collections of biologically relevant gene sets (molecular concepts), for disproportionate overlap using MCA (14). Interestingly, several findings, shown in the molecular concepts map (Fig. 1A) implicate hypoxia as a key theme among concepts significantly associated with the Ral signature, including known hypoxia targets, genes involved in clear cell renal cell carcinoma (a neoplasm driven by HIF signaling), and genes regulated by the treatment of breast cancer cells with deferoxamine (a chemical hypoxia mimic). Supplementary Fig. S1 details the genes in the Ral transcriptional signature and their representation in individual molecular concepts.
Specifically, implicating hypoxia in regulation of CD24, a previous study identified CD24 among a multitude of genes induced by hypoxia in human umbilical cord vein endothelial cells (30). To determine whether a similar regulatory circuit is operative in cancer cells from 2 common malignancies showing CD24 overexpression (8, 31), we exposed human PC-3 (prostate cancer) and UMUC-3 (bladder cancer) cell lines to hypoxia, and evaluated CD24 mRNA and protein levels. Exposure of cells to hypoxia increased CD24 mRNA by 12 hours, reaching a maximum at 24 hours (Fig. 1B and Supplementary Fig. S2A). Protein levels were seen to parallel this increase in mRNA in UMUC-3, PC-3, and 2 additional cell lines, KU-7 and LNCaP, derived from human bladder and prostate cancers, respectively (Fig. 1B and Supplementary Fig. S2B). Next, we evaluated the association of tumor hypoxia to CD24 expression in vivo. Immunohistochemistry revealed the pattern of CD24 expression as a function of pimonidazole (hypoxyprobe) staining in human bladder cancer xenografts. Supporting the in vitro findings above, CD24 protein expression was found to be elevated in hypoxic areas of the tumor (Fig. 1C).
Isolation of the human CD24 gene promoter
Exposure of UMUC-3 and PC-3 cells to hypoxia also led to an increase in HIF-1α protein, but not mRNA, and this preceded the increase in CD24 expression (Fig. 1D and Supplementary Fig. S2A). Because elemental iron is a critical factor for proline hydroxylation of HIF-1α, iron chelators such as deferoxamine (DFO; ref. 32) are widely used in HIF-1 studies as surrogates of hypoxia (33). We hence treated UMUC-3 cells with 50 μmol/L DFO, which led to a 3-fold increase in CD24 protein expression after 12 hours (Supplementary Fig. S2C).
These findings suggested the presence of hypoxia responsive cis-acting elements (HRE) in the CD24 promoter. There have been 2 published CD24 promoter sequences, the first from a pooled sample of human DNA (34), whereas the second from a population of B lymphocytes (35). Unfortunately, these 2 promoters had considerable mismatched sequences between each other and the partial promoter from GenBank (Supplementary Fig. S3). Therefore, to determine the consensus CD24 promoter sequence in UMUC3 cells, we cloned a 1,139 bp region that encompassed the first 79 base pairs of the CD24 5′ untranslated region (UTR). Comparing the UMUC3 sequence to the previously reported sequences for the CD24 promoter, several mismatches and deletions were observed (Supplementary Fig. S3). To ensure a consensus CD24 promoter sequence, we cloned the same region in 4 additional cancer cell lines (LUL2, EJ, J82, and LNCaP) and 2 noncancerous cell lines: telomerase (TERT) immortalized urothelial cell line and human embryonic kidney cells (293T). Here, we report and then use a consensus promoter whose sequence is deposited in GenBank (accession numbers are included in the Materials and Methods section).
CD24 gene is a transcriptional target of HIF-1α in hypoxia
Using the cloned promoter, we generated and transiently transfected a series of CD24 5′-flanking sequence deletion mutants driving a luciferase reporter (Fig. 2A) into UMUC-3 cells and incubated these under normoxia and hypoxia for 24 hours. Figure 2B indicates that the sequence between −100 bp and −223 bp is critical for CD24 transcriptional induction in hypoxia. To determine whether these sequences harbored candidate HIF binding sites, we used Genomatix software (36) to determine if any canonical HIF-1 binding sites existed from −118 bp to −135 bp. We then used site-directed mutagenesis to obliterate the putative site that was found (CGTG to AAAA) in the luciferase reporter plasmid (Fig. 2C) and transfected it into UMUC-3 cells. The mutation of the HIF-1 binding site completely abolished the responsiveness to hypoxia (Fig. 2D), and also reduced basal expression, probably because of reduction of detectable levels of HIF-1 expression in normoxic cells. These results suggest that this putative HRE is necessary for the induction of CD24 gene in cancer cells.
We next wanted to investigate the dependency of CD24 expression on HIF-1α levels. Transient overexpression of HIF-1α transgene in UMUC-3 cells showed an approxiamtely 2-fold increase in CD24 mRNA expression and an approximately 3-fold increase in CD24 protein expression in normoxia (Fig. 3A and B). We evaluated 4 different shRNA oligos against HIF-1α, found similar effects, and selected HIF-1α sh-3 for further studies (Supplementary Fig. S5D). Absence of appreciable change in HIF-2α levels after forced induction or silencing of HIF-1α indicates that CD24 expression is independent of HIF-2α status (Fig. 3B). To determine the contribution of HIF-1α to hypoxic induction of CD24 expression, UMUC-3 cells transiently depleted of HIF-1α were exposed to hypoxia. HIF-1α–depleted UMUC-3 cells failed to induce CD24 mRNA or protein expression after exposure to hypoxia (Fig. 3A and B). To determine if HIF-1α regulates CD24 via a direct interaction with its promoter sequence, we conducted chromatin immunoprecipitation. UMUC-3 cells were exposed to normoxia and hypoxia, and chromatin complexes immunoprecipitated with human HIF-1α or HIF-1β antibody, and PCR amplification was conducted using specific primers to human CD24 promoter region encompassing the identified HRE site. Results showed the preferential binding of HIF-1α (Fig. 3C) and HIF-1β (Supplementary Fig. S4) to CD24 promoter in hypoxia compared with normoxia. This supports finding from previous studies that although HIF-1β (ARNT) is constitutively expressed, it heterodimerizes with HIF-1α and selectively binds to hypoxia-responsive promoter elements of selected genes (37). These results also indicate direct transcriptional regulation of endogenous CD24 expression by HIF-1α.
CD24 is a critical determinant in HIF-1α–driven tumor progression and metastasis
Both HIF-1α and CD24 overexpression are associated with poor prognosis in multiple human cancer types (38). We first evaluated the role of CD24 expression in experimental models of distant colonization of bladder and prostate cancer. The first using lung colonization of UMUC-3 cells after tail vein inoculation and the second using bone colonization after intraprostatic inoculation of PC3 cells. We silenced CD24 expression in UMUC-3 and PC-3 cells using lentiviral transduction of shRNA targeted to 3 different oligos against CD24 (CD24shRNA) or nontargeted scrambled control (NTshRNA). Expression of CD24 after transfecting these oligos were evaluated separately in UMUC-3 as well as PC-3 cells (Supplementary Fig. S5A). Effects of shRNAs on growth of UMUC-3 cells were assessed (Supplementary Fig. S5B). From these 3 different oligos, we chose to use CD24-sh-3 for all further studies. After injection of UMUC-3 cells, we monitored the animals weekly using bioluminescent imaging (BLI), which was shown previously to track well with histology (52), and after 8 weeks, sacrificed and evaluated their lungs by human-specific quantitative real-time PCR (27). At 8 weeks, mice injected with CD24-deficient UMUC-3 cells had lower levels of total photon radiance than control UMUC-3 NTshRNA cells and had significantly reduced numbers of residual tumor cells detectable by human-specific quantitative real-time PCR (P = 0.01; Fig. 4A). Similarly, CD24-depleted PC-3 cells showed significantly lower photon radiance 3 to 5 weeks after injection (Supplementry Fig. S5C) and higher cumulative survival compared with (P = 0.002) mice inoculated with corresponding control PC-3 cells (Fig. 4B). This result in prostate cells suggests a role of CD24 in lung colonization and supports our findings in bladder cancer cells (14) that CD24 expression affects lung retention of tumor cells.
HIF-1α overexpression promotes metastasis, whereas HIF-1α depletion reduces this phenotype via alteration of a cancer cell's ability to survive hypoxic conditions (39). To study the contribution of CD24 in HIF-1α–mediated viability during hypoxic stress in vitro and HIF-1α–dependent metastasis in vivo, we generated HIF-1α shRNA-expressing UMUC-3 cells. We silenced HIF-1α expression in UMUC-3 cells using lentiviral transduction of shRNA targeted to 3 different oligos against HIF-1α or NT scrambled control. Expression of HIF-1α after transfecting these oligos were evaluated in UMUC-3 (Supplementary Fig. S5D). Effects of these shRNAs on growth of UMUC-3 cells were assessed (Supplementary Fig. S5E). From these 3 different oligos, we chose to use HIF-1α-sh-3 for all studies. As expected, HIF-1α–depleted UMUC-3 cells expressed significantly less CD24 compared with cells expressing scrambled control sequence (Fig. 5A). CD24 depletion led to a reduction in cell viability in normoxia and an even more profound effect in hypoxic condition (P = 0.0320). Importantly, this was comparable to that observed with HIF-1α depletion in these cells (Fig. 5B). Furthermore, when CD24 expression was restored in HIF-1α–depleted UMUC-3 cells using a CD24 transgene, this rescued the reduced survival observed after HIF-1α depletion in normoxia and hypoxia (P = 0.04). When injected subcutaneously into mice, HIF-1α–depleted UMUC-3 cells produced significantly decreased average tumor burden at a given time compared with UMUC-3 NTshRNA cells. Notably, cells with depleted HIF-1α, but restored CD24 expression, displayed a comparable tumor burden at all tested times to UMUC-3 NTshRNA cells (P = 0.13; Fig. 5C).
The engineered UMUC-3 cells described earlier were then inoculated via tail vein in vivo and lung metastasis assessed. Visual quantitation of lung metastases revealed that more than 50% of mice inoculated with empty vector–transfected UMUC-3 cells had lesions at 8 weeks, whereas none of the mice receiving UMUC-3 HIF-1α shRNA cells had detectable deposits (Supplementary Table S1). Notably, 50% of the mice injected with UMUC-3 HIF-1α shRNA cells stably overexpressing CD24 had visible lung metastasis. Evaluation of the lungs using quantitative real-time PCR with 12p human-specific probe (27) showed a drastic reduction in human genomic DNA in HIF-1α–depleted UMUC-3 cells compared with their nontarget controls (SEM 47.8 ± 27 vs. 4,320.2 ± 1,256, respectively; P = 0.0081), whereas forced expression of CD24 in HIF-1α–depleted cells rescued metastatic proficiency (P = 0.031; Fig. 5D). Furthermore, we sought to evaluate the role of HIF-1α and CD24 in spontaneous metastasis in an orthotopic prostate cancer model. Previously characterized luciferase-expressing PC-3 cells with depleted HIF-1α and overexpressing CD24 were orthotopically implanted into prostate and assessed for their metastatic ability by BLI at various intervals. BLI signal and survival of animals (Fig. 5E–F) showed that HIF-1α and CD24 expression are important for local and metastatic growth. Moreover, forced expression of CD24 partially rescued metastatic ability of HIF-1α–depleted cells.
Next, to establish the requirement for CD24 in HIF-1α induced primary tumor growth and metastasis, we first generated stable HIF-1α–overexpressing, and also CD24-depleted HIF-1α–overexpressing UMUC-3 cells. HIF-1α overexpression in UMUC-3 cells led to a marginal increase in cell viability compared with NTshRNA cells, whereas CD24 depletion in HIF-1α–overexpressing cells reduced their survival (P = 0.021; Fig. 6B). When injected subcutaneously into mice, HIF-1α–overexpressing UMUC-3 cells marginally increased average tumor burden compared with UMUC-3 NTshRNA cells (Fig. 6C). However, CD24 depletion, even in the presence of forced HIF-1α expression, produced smaller tumors (Fig. 6C). These modified UMUC-3 cells were inoculated via tail vein in vivo and lung metastasis was assessed. As evaluated by visual assessment and quantitative real-time PCR with a 12p human-specific probe, UMUC-3 cells stably overexpressing HIF-1α displayed more colonies in lung compared with UMUC-pcDNA NTshRNA cells, whereas CD24 depletion even in the presence of forced HIF-1α expression, reduced their ability to colonize to lungs (P = 0.022; Fig. 6D). These results support the role of CD24 as an important downstream effector of HIF-1α–mediated survival and metastasis. Furthermore, PC-3 cells stably overexpressing HIF-1α, and also CD24-depleted HIF-1α–overexpressing UMUC-3 cells were orthotopically implanted in prostate and assessed for their metastatic ability by BLI at various intervals. BLI signal and survival of animals (Fig. 6E–F) showed that tumor bulk and animal survival after inoculation of PC-3 cells depends on elevated expression of HIF-1α, CD24, and that CD24 expression contributes to the ability of HIF-1α to promote metastasis. To our knowledge, this is the first demonstration of the effect of CD24 and HIF-1α in a prostate orthotopic model that leads to spontaneous metastasis.
Coexpression of CD24 and HIF-1α in human bladder cancers
Finally, we sought to determine whether the pattern of expression of CD24 from human tumors might support the model that HIF-1α drives the metastatic phenotype in part via induction of CD24 expression, as indicated by the experiments described earlier. Hence, we examined if there is any relationship between HIF-1α and CD24 protein expression by immunohistochemistry in human bladder cancer tissues, using a tissue microarray of archival tissues, using antibodies reported before (23, 40), and scored expression as detailed in the Materials and Methods. In total, 144 cases represented on the array showed interpretable staining for both antibodies; of these, 101 cases were urothelial carcinomas (the most common type of bladder cancer in Western countries and the histology from which bladder cell lines used experimentally herein were derived). As reported before (24), HIF-1α showed highly variable nuclear positivity with frequent background moderate to strong cytoplasmic positivity (Fig. 7A and B). Given its known function and prior reports in bladder cancer associating degree of nuclear positivity with key clinicopathologic variables, HIF-1α staining was scored as % nuclear positivity. CD24 showed variable staining among cases of urothelial carcinoma with cytoplasmic immunoreactivity (scored as 0+, 1+, 2+, 3+ on overall intensity as reported before; ref. 11),with examples shown in Fig. 7C and D).
Among urothelial cases, we observed a significant, positive correlation between CD24 staining and HIF-1α nuclear positivity, rs = 0.29, P = 0.003. In contrast, in nonurothelial cases, CD24 and nuclear HIF-1α were not correlated significantly (rs = 0.036, P = 0.84). Among the urothelial carcinomas, the degree of correlation between CD24 and HIF-1α was related to stage: among cases stage pTa, pT1, and pT2 (primary tumor that has not extended beyond the bladder), CD24 and HIF-1α were not significantly correlated (rs = 0.001, P = 1.0). In contrast, cases showing extravesical extension of the primary tumor (stages pT3 and pT4) showed the highest degree of correlation (rs = 0.49, P = 0 < 0.001). Importantly, the most prevalent pattern of expression of these proteins was of expression of both at more than 2+ level (N = 55), as opposed to many fewer cases showing expression of both proteins at low level (0+ or 1+, N = 14).
Given our observations regarding roles for HIF-1α, and downstream, CD24, in our experimental metastasis model, we were interested in whether staining patterns for these proteins were associated with overall survival in the first 5 years postcystectomy, where recurrence (metastasis)-free survival contributes most strongly to overall survival (41). For these analyses, we tested the association of HIF-1α and CD24 staining with overall survival at 60 months by the log-rank test. For HIF-1α, we observed a nonsignificant trend toward decreased survival in cases showing 2+ or 3+ nuclear staining (HIF-1α high) as opposed to cases showing 0+ and 1+ (HIF-1α low; Fig. 7Ei). For CD24, we observed a nonsignificant trend toward decreased survival in cases showing 2+ or 3+ staining (CD24 high) as opposed to cases showing 0+ and 1+ (CD24 low; Fig. 7Eii).
Given the observed prometastatic regulatory cascade of HIF-1α on CD24 expression observed above, we also tested the relationship between survival and a total, combined HIF/CD24 score (i.e., HIF-1α + CD24). Comparing cases showing staining consistent with induction of this prometastatic cascade (HIF/CD24 total ≥ 3, high, N = 85) to cases showing lack of induction of either or both of these proteins (HIF/CD24 total score <3, low, N = 16), we found a significant difference in survival (Fig. 7F, P = 0.01). Importantly, given the aforementioned association between increased correlation of HIF-1α and CD24 and high primary tumor stage, we tested the relationship between HIF/CD24 total score class (low vs. high) and survival at the 60-month endpoint, alone and with stage in univariate and multivariate Cox proportional hazards regressions models. Strikingly, not only was HIF/CD24 class significantly associated with survival (univariate Cox P = 0.02), but it maintained independent, significant association with survival, adjusting for primary tumor stage (multivariate Cox P = 0.01). The correlation of HIF-1α and CD24 expression in tumor tissues and association of their coexpression with poorer survival outcomes supports the clinical relevance of model system findings above that HIF-1α drives the metastatic phenotype, in part, via CD24 induction.
The results presented here are the first to show that CD24 is transcriptionally regulated by HIF-1α and that this former molecule is a critical downstream effector of HIF-1α effects on tumor growth and lung colonization. Furthermore, the clinical relevance of this pathway was supported by data from human bladder tumors, indicating that the clinical impact of HIF-1α expression is enhanced by CD24 coexpression. By linking CD24 with hypoxia, these findings can provide one explanation why CD24 expression is such a powerful prognostic factor in bladder and prostate cancer.
Friederichs and colleagues showed that CD24 may contribute to metastatic dissemination as a ligand for the adhesion molecules P- and E-selectin (42), which are expressed on activated platelets and endothelial cells. Interestingly, hypoxia and reoxygenation can induce P-selectin expression in endothelial cells (43), which may promote further cancer cell–endothelial interactions. Inhibition of CD24 by monoclonal antibodies or siRNAs has been reported to reduce primary tumor growth and metastasis of colon, pancreas, and bladder cancers (17, 44). In this study, CD24 depletion had a greater impact on cell viability in hypoxia-exposed than normoxic cells. This suggests that CD24 expression serves to protect cells from hypoxia and allow them to survive in such adverse conditions in addition to promoting their local growth and metastatic colonization. Interestingly, the increase in CD24 mRNA in normoxia after HIF-1α transgene transfection was higher than that observed in the control-transfected cells exposed to hypoxia. In contrast, CD24 protein levels were similar despite higher HIF-1α in HIF-1α transgene-transfected cells. This suggests that CD24 protein levels are controlled by other factors or alternate limiting mediators of protein synthesis exist.
Immunohistochemistry and gene array–based studies have shown CD24 expression is upregulated in bladder and prostate cancers and its expression correlates with poor prognosis (8). Furthermore, this study showed that CD24 expression is critical for tumor growth and migration. Altevogt and colleagues showed that CD24 expression reduces SDF-1–mediated cell migration and signaling via CXCR4 (45), a mechanism by which CD24 may elicit tumor promoter effect. A recent study also suggested that CD24-dependent extracellular signal-regulated kinases and p38 mitogen–activated protein kinase activations are required for colorectal cancer cell proliferation in vitro and in vivo (46). CD24-bound oligosaccharides act as a ligand for P-selectin, a cell adhesion molecule present on endothelial cell wall and platelets were shown to facilitate tumor passage through blood stream (47). Using a short-term lung colonization assay, we have shown that CD24-expressing cells preferentially bind to lungs compared with nonexpressors (11). However, the mechanisms that drive CD24 expression in human cancer are unclear. In this study, automated comparison of the core Ral transcriptional signature (manuscript in revision) with all molecular concepts in oncomine (including oncomine gene expression signatures, Gene Ontology, KEGG Pathways, Human Protein Reference Database, etc.) identified significant association with hypoxia-related molecular concepts. Similarly, a previous study based on broad transcriptomic analysis of human umbilical cord vein endothelial cells exposed in vitro to hypoxia, Scheurer and colleagues reported that CD24 is one of 65 genes in which mRNA increases with hypoxia (30). However, no further insights were provided as to whether this was a transcriptional effect or the mechanisms responsible. Here, we observed that hypoxia led to a significant upregulation of CD24 mRNA and protein expression in a panel of bladder and prostate cancer cells. In addition, CD24 immunoreactivity was found in hypoxic areas of human xenografts indicating this association is also operative in vivo. Furthermore, studies conducted in UMUC-3 bladder cancer cells uncovered a putative hypoxia responsive element in the 5′ flanking region of CD24 promoter. Using chromatin immunoprecipitation studies, we showed increased HIF-1 binding to the CD24 promoter after exposure to hypoxia. Together, these data suggest that at least for the length of promoter we had for our studies, the HRE site was a prime regulator of transcription.
As CD24, HIF-1α is elevated in advanced tumors and is associated with metastatic progression (2). HIF-1α–mediated gene expression allows cells to respond to hypoxic insult by increasing oxygen delivery via secretion of angiogenic factors or confer a survival advantage to cells faced with decreased oxygen availability (48). In prostate (49) and bladder cancer (50), HIF-1α regulates multiple cytokines and their receptors. By regulating key target genes, HIF-1α promotes the tumor growth at primary and distant sites. Our data shows recovery of primary tumor growth and lung colonization (Fig. 5C and D). Upon CD24 overexpression in HIF-1α–depleted cells indicates that CD24 plays an important role in HIF-1–mediated tumor growth regardless of tumor site.
Finally, data from our clinical cohort support the relevance of our experimental findings to human tumors. We observed a significant correlation of HIF-1α and CD24, correlating in vitro findings of HIF-1α regulation of CD24 as initially suggested by molecular concepts mapping. In addition, attempting to evaluate the output of this regulatory pathway by use of a total score adding scores for HIF-1α and CD24, we found that we could significantly stratify survival, independently of tumor stage. Importantly, the most prevalent pattern of expression of these proteins in bladder cancers was of coincident moderate to intense (2+ or 3+) staining of both, a finding of relevance to potential therapeutic strategies (2, 11), which seem rational and should be considered.
As many cancers are thought to develop and progress to metastasis from a small number of transformed, self-renewing “cancer stem cells” (51), these results implicate a role for the HIF-1α–CD24 axis in achieving that goal. In addition, the work presented here on human tumors shows the value of risk stratification based on HIF and CD24 protein expression and could serve to select patients for trials in the adjuvant or early metastatic setting with anti–HIF-1α or CD24-directed therapy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: S. Thomas, M. Harding, S. C. Smith, D. Theodorescu
Development of methodology: S. Thomas, M. Harding, S. C. Smith, J. B. Overdevest
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Thomas, S. C. Smith, J. B. Overdevest, M. D. Nitz, H. F. Frierson, G. Kristiansen
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Thomas, M. Harding, S. C. Smith, M. D. Nitz, S. A. Tomlins, G. Kristiansen
Writing, review, and/or revision of the manuscript: S. Thomas, S. C. Smith, J. B. Overdevest, M. D. Nitz, S. A. Tomlins, D. Theodorescu
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Thomas, D. Theodorescu
Study supervision: S. Thomas, D. Theodorescu
This work was supported in part by the NIH grants CA075115 and CA104106 to D. Theodorescu.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The authors thank Drs. Chris Moskaluk and Pat Pramoonjago of the University of Virginia Biorepository and Tissue Research Facility for their technical support.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received November 4, 2011.
- Revision received July 16, 2012.
- Accepted August 5, 2012.
- ©2012 American Association for Cancer Research.