Malignant melanoma is a cancer that arises from melanocyte cells in a complex but well-studied process, and which can only be successfully treated prior to metastasis as it is highly resistant to conventional therapies. A number of recent reports have indicated that members of the HOX family of homeodomain-containing transcription factors are deregulated in melanoma, and may actually be required to maintain proliferation. In this report, we describe the use of a novel, cell-permeable antagonist of the interaction between HOX proteins and PBX, a second homeodomain-containing transcription factor that modifies HOX activity. This antagonist can block the growth of murine B16 cells and trigger apoptosis both in vitro and in vivo when administered to mice with flank tumors. [Cancer Res 2007;67(12):5806–13]
Melanoma is a cancer of the melanocytes, a population of melanin-producing cells that reside in the epidermal basement membrane of the skin (reviewed in ref 1). Its superficial location often makes excision the first choice of treatment for malignant melanoma, and is highly successful if done before malignant cells have breached the epidermal basement membrane. After this event though, malignant cells frequently and quickly metastasize, spreading via the sentient lymph nodes to a variety of targets including the lungs and brain. Despite considerable advances in our understanding of melanoma genetics, thus far, there are no effective treatments for metastatic melanoma, and the prognosis after the disease has progressed to this stage is extremely poor ( 1).
In addition to genes that regulate the cell cycle and cell signaling pathways, there is a growing interest in the deregulation of transcription factors in melanoma. These include transcription factors that have previously been identified as oncogenes, as well as other transcription factors that are expressed primarily in embryonic tissues but maybe re-expressed in malignant cells. Of particular note are the HOX genes, a family of homeodomain-containing transcription factors that determine the identity of cells along different embryonic axes, including the developing nervous system and limb ( 2). HOX gene deregulation has been observed in a number of cancers (reviewed in ref 3), including melanomas ( 4– 7).
Mammals have 39 HOX genes, many of which have partly or fully redundant functions ( 2). For this reason, it has generally proven difficult to fully assess the function of HOX genes in some contexts using conventional genetic approaches. To overcome this difficulty, we have designed a small, cell-permeable peptide (HXR9) that antagonizes the interaction between HOX and a second transcription factor (PBX), which binds to HOX proteins in paralogue groups 1 to 8. HOX/PBX dimers have significantly greater binding affinity and specificity for target DNA sequences than the HOX monomer alone ( 8– 10). Here, we show that HXR9 triggers apoptosis in melanoma cells both in vitro and in vivo.
Materials and Methods
Culturing and maintenance of cell lines. B16F10 was cultured as described previously ( 11). Briefly, cells were maintained in RPMI medium (Life Technologies) supplemented with 10% fetal calf serum, 2 mmol/L of l-glutamine (Life Technologies), and 1% penicillin/streptomycin (Life Technologies). Cells were maintained at 37°C in a 10% CO2 incubator. “Primary cells” were derived from a visceral malignant melanoma. This was resected and half of the tissue was used to confirm the disease phenotype through standard histopathologic techniques, whereas the second was treated to separate individual cells. These cells were cultured using the same conditions described above for B16F10 except that DMEM replaced RPMI.
Cord blood–derived CD133+ cells were enriched through positive selection using a MiniMACS separation system (Miltenyi Biotec). Enriched hematopoietic stem cells (HSC; 2 × 103) were suspended in 300 μL of Iscove's modified Dulbecco's medium supplemented with 2% fetal bovine serum (Life Technologies/Invitrogen) and were mixed vigorously with 3 mL of MethoCult GF H4434 containing recombinant cytokines and erythropoietin (StemCell Technologies). Cells were incubated with 60 μmol/L of HXR9 or CXR9 at the start of the experiment and the medium was changed after 24 h (start of day 2).
In vitro treatments. Cells were plated at a density of 2.6 × 104 cells/mL in six-well plates in 1.5 mL of media, and were allowed to adhere overnight. Cells received 60 μmol/L of HXR9, or the control peptide CXR9 for 2 h, and were harvested for subsequent analyses. For annexin V and nuclear staining, B16 cells were plated at a density of 9 × 104 cells/mL in 25 cm2 culture flasks. After 24 h of recovery, flasks for nuclear staining were treated as mentioned above for 2 h. Then, cells were detached using trypsin/EDTA solution (Sigma) and 200 μL aliquots were used to prepare cytospins at 800 rpm for 5 min using a Cytospin 4 (ThermoShandon). Cells used for annexin V staining were treated with 60 μmol/L of HXR9 alone for 2 h or after the pretreatment with Z-VAD-fmk (50 μmol/L; Bachem) for 45 min. RNA extraction was done using a Qiagen kit.
cDNA synthesis. RNA was reverse-transcribed as described previously. Briefly, RNA was first denatured by heating to 65°C for 5 min. One to five micrograms of RNA was incubated in a volume of 50 μL at 37°C for 1 h with final concentrations of 10 mmol/L of DTT, 1 mmol/L of deoxynucleotide triphosphate mix, as well as 100 μg/μL of poly-T primers, 200 units of reverse transcriptase (Invitrogen), and 40 units of RNaseOUT (Invitrogen). The cDNA synthesis reaction was terminated by placing tubes at 65°C for 5 min.
Real-time PCR. Quantitative reverse transcription-PCR was done using the Stratagene MX4000 real-time PCR machine. The Stratagene MX4000 measures PCR product accumulation during the exponential phase of the reaction, prior to the amplification becoming vulnerable to limited reagents and cycling variability. Fluorescence increases in accordance with increasing levels of PCR product.
Half-life determination of HXR9. Each peptide was added to freshly prepared human plasma to a final concentration of 1 mmol/L in a total volume of 40 μL. Samples were incubated at 37°C and reactions were stopped by adding 60 μL of water and 0.9 mL of protein gel loading buffer (Bio-Rad) containing 1% 2-mercaptoethanol, and heating to 95°C for 4 min. Two microliters of each sample were run on a 22% acrylamide gel.
Analysis of cell death and apoptosis. The assessment of cell viability was done using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Promega), MTS (Promega) assay or the lactose dehydrogenase (LDH) cytotoxicity detection kit (Roche) according to the manufacturer's instructions. Apoptotic morphology was determined using scanning electron microscopy, fluorescence microscopy, and flow cytometry. Changes in nuclear morphology were assessed by fixing B16 cytospins with ice-cold methanol at room temperature for 5 min. The fixed cells were stained with Hoechst 33258 (2 μg/mL; Sigma) and mounted using Vectashield mounting medium (Vector) and analyzed by fluorescence microscopy (Nikon Eclipse TE 2000-S). B16 cells were also processed using the annexin V-FITC apoptosis detection kit as described by the manufacturer (Oncogene Research Products, Calbiochem). Samples were analyzed using a Beckman Coulter Epics XL flow cytometer (argon laser, excitation wavelength 488 nm, and used FL-1 and FL-3 detectors).
Silencing of c-Fos RNA. B16 cells were seeded at 3 × 104 cells/mL in a 96-well plate and allowed to recover for 24 h. Cells were pretreated with a mixture of oligonucleotides complementary to Fos (asFos) or with a random, noncoding oligonucleotides as a control (ascon) for 2 h at a final concentration of 1 μmol/L. B16 cells were then treated with 60 μmol/L of CXR9 or HXR9 for 2 h and the assessment of cell viability was done using the LDH cytotoxicity detection kit. From these cells, the RNA was isolated and samples analyzed by quantitative PCR.
Mice and in vivo trial. B16 cells in exponential growth phase were injected s.c. at a dose of 1 × 106 cells into C57black/6 mice. Tumor growth was monitored daily. Once tumors were established, mice received 10 mg/kg of HXR9 or CXR9 i.v. via the tail vein, twice weekly (10 mice per group). Mice were euthanized when tumor volume exceeded 1,200 mm3 and tumors were excised and fixed for histopathologic analysis. Control groups received CXR9 and PBS accordingly.
Band shift assays. Whole protein extracts were made from excised tumors and used with double-stranded oligonucleotide probes labeled with streptavidin, based on previously described methods ( 4). The forward sequences of these oligos were: HP (containing a HOX/PBX consensus binding site)—GGACA AACTG AAGGC AGAGC TGATT TATGG CACAC ACACA AGAAT GGACA AACCC GTGAG, HPC (containing an altered HOX/PBX binding site as a control)—GGACA AACTG AAGGC AGAGC GCTCC GTTAA CACAC ACACA AGAAT GGACA AACCC GTGAG. Complexes were supershifted using the anti-PBX1,2,3 antibody (c-20; Santa Cruz Biotechnology) at a dilution of 1:2,000.
Statistical analysis of the data. Data are given as means ± SE of three independent experiments. Comparison of treatments against controls was made using one-way ANOVA followed by Fisher's LSD post hoc test with the SPSS statistical package. The significance level chosen was P < 0.05.
HXR9: an antagonist of HOX/PBX interaction. In order to block PBX-dependent HOX function, we designed a cell-permeable peptide that mimics the “hexapeptide” sequence found in HOX proteins of paralogue groups 1 to 9 ( 8– 10, 12– 16). The rationale for this design was that HXR9 should act as a specific, competitive inhibitor of the HOX/PBX interaction, and thus prevent HOX/PBX dimers from binding to DNA. For cell penetration, the hexapeptide was linked to a polyarginine (R9) sequence, which has previously been shown to act as an efficient peptide delivery system ( 17). As a control, we designed a second peptide, CXR9, which lacks a functional hexapeptide portion. In order to increase the stability of these peptides, the NH2-terminal- and COOH-terminal–most amino acids were synthesized as D-isomers.
The stability of the peptides was assessed by incubating each with human plasma at 37°C. The amounts of peptide remaining after different lengths of time were assessed by SDS-PAGE, and indicated that the half-life of both HXR9 and CXR9 was ∼12 h (Supplementary data 1).
In order to test the specificity of HXR9, we treated murine B16 melanoma cells with 60 μmol/L of HXR9 or CXR9 for 1 h and extracted protein from the cells. HOXD9, a known PBX-binding partner that is expressed in melanoma ( 7), was then immunoprecipitated using an anti-PBX antibody. This technique permits the recovery of HOXD9 in CXR9-treated cells but not in HXR9-treated cells, indicating that HXR9 does indeed block the interaction between PBX and HOX ( Fig. 1A ). As a further control on specificity, we also immunoprecipitated the glucocorticoid receptor (GR). This binds to PBX through an interaction that does not depend on the presence of the hexapeptide motif in the GR protein ( 18), and thus, this particular interaction should not be blocked by HXR9, and indeed, it is not ( Fig. 1A).
As an additional test of the peptide's selectivity, we studied its binding to a double-stranded DNA probe containing a HOX/PBX consensus binding site. If the hexapeptide does prevent the interaction of HOX and PBX, it should also disrupt the binding of these proteins to HOX/PBX DNA consensus sites. In order to test this, a band-shift type assay was used. B16 cells were treated with 60 μmol/L of HXR9 or CXR9 for 1 h, and then a cell-free extract was prepared. A streptavidin-labeled DNA probe was added that contains either a HOX/PBX consensus site (HP) or a modified sequence that does not match the consensus site (HPC; Fig. 1B). The addition of the HP probe to the CXR9 or control B16 lysates, followed by resolution on a nondenaturing acrylamide gel resulted in a significant reduction in mobility. This indicates that one or more proteins had bound to it. In order to characterize these proteins, a “supershift” assay was done, whereby an anti-PBX antibody was added to the lysate. This resulted in a further shift in mobility equivalent to a single antibody molecule binding, and indicating that each probe-binding protein complex contains a single PBX protein.
Importantly, whereas proteins in the cellular extracts from untreated and CXR9-treated cells bound to the probe, those from the HXR9 extracts did not. This suggests that HXR9 does indeed prevent the formation of HOX/PBX dimers and their subsequent binding to DNA.
In order to confirm that HXR9 can penetrate B16F10 cells, we used an anti-HXR9 antibody combined with the fluorescent FITC tag to examine its intracellular distribution ( Fig. 1D). This shows that HXR9 is present both in the cytoplasm and the nucleus.
HXR9 triggers apoptosis in B16 and primary melanoma cells. We investigated the antiproliferative activity of HXR9 on B16 cells. The IC50 for HXR9 and CXR9 were found to be 20 and 200 μmol/L, respectively. Cells treated with 60 μmol/L of HXR9 for 2 h exhibited a notable change in morphology, including a reduction in dendritic processes (data not shown). Staining with Hoechst dye also revealed a large number of HXR9-treated cells displaying nuclear apoptotic morphology (chromatin condensation; Fig. 2A ). An additional sign of apoptosis is the appearance of phosphatidylserine in the cytoplasmic surface of the cell membrane ( 19). This can be detected by fluorescently labeled Annexin V protein and cell sorting using fluorescence-activated cell sorting. Using this technique with B16F10 cells treated for 2 h with 60 μmol/L of HXR9 revealed that a significant proportion of cells were in late phases of apoptosis ( Fig. 2B). A significant protection from HXR9-induced cell death by apoptosis was observed by the pretreatment of B16F10 cells with the pan-caspase inhibitor Z-VAD-fmk ( Fig. 2B). CXR9 had no effect on proliferation and did not cause apoptosis.
In addition to the derived B16 murine melanoma cells, we also tested the ability of HXR9 to trigger apoptosis in primary melanoma cells derived from a human malignant melanoma removed for histologic analysis. Cells were separated from this tissue and cultured until they reached 80% confluence and then treated with 60 μmol/L of HXR9 for 2 h, after which they exhibited morphology consistent with cell death ( Fig. 2C). Staining of nuclei with Hoechst dye also revealed the presence of apoptotic morphology ( Fig. 2C). Additionally, an IC50 determination for HXR9 was done using the LDH cytotoxicity assay. This revealed that the IC50 for the cells was 50 μmol/L.
HXR9 causes specific transcriptional changes. In order to examine changes in gene transcription on HXR9 treatment, B16F10 cells were treated for 2 h with 60 μmol/L of HXR9 or CXR9. RNA was extracted for analysis by Microarray, using the Mouse Genome 430A 2.0 Array (Affymetrix), which contains 14,000 characterized mouse genes. Although the majority of genes failed to show any significant changes in transcription, 22 showed an increase in transcription (P < 0.05). Most notably, these included oncogenes Fos and Jun. In order to confirm that these genes were up-regulated in HXR9-treated B16 cells, quantitative PCR was used to measure the relative number of transcripts in RNA extracted from cells ( Fig. 3A ); this confirmed that Fos, Jun, Dusp1, and Atf1 were all significantly up-regulated in response to HXR9.
HXR9 blocks tumor growth and triggers apoptosis in vivo. The B16 F10 murine tumor is a well-established experimental in vivo model of melanoma, and is also one of the most aggressive murine tumors ( 20, 21). Subcutaneous inoculation was used to introduce 1 × 105 cells, and mice were treated with twice weekly i.v. doses of HXR9 or CXR9 at 10 mg/kg once tumors were palpable (10 mice per group). At the end point of these experiments, HXR9-treated tumors showed a significant degree of growth retardation ( Fig. 4A ), and histologic analyses of excised tumors revealed a significantly higher level of cell death in HXR9-treated mice ( Fig. 4B).
In order to assess whether the hexapeptide could block HOX/PBX binding in vivo, tumors were injected directly with HXR9 at 10 mg/kg, and were then used to produce nuclear lysates 2 h after treatment. Band shift assays (as detailed above) revealed that protein extracted from these tumors could not bind a DNA probe containing a HOX/PBX consensus binding site, whereas an equivalent extract from CXR9-treated tumors could ( Fig. 1C), indicating that HXR9 specifically disrupts the formation of HOX/PBX/DNA complexes in vivo. Quantitative PCR analysis of RNA extracted from these tumors revealed that Fos, Jun, Dusp1, and Atf3 were up-regulated, mirroring those results obtained in vitro ( Fig. 3B).
Up-regulation of Fos is a key event in HXR9-mediated apoptosis. Previous studies have suggested that an increase in Fos transcription could trigger apoptosis ( 22– 29). We therefore tested whether Fos up-regulation might be responsible for HXR9-mediated apoptosis by ablating Fos RNA in B16 cells prior to HXR9 treatment by incubating cells with a mixture of antisense oligonucleotides that were complementary to Fos (asFos), and with a second set of random, noncoding oligonucleotides as a control (ascon). Treatment with the former caused a reduction in Fos mRNA in HXR9-treated cells to a level that was equivalent to that found in CXR9-treated cells, whereas the ascon set did not ( Fig. 5A ). This reduction in the number of Fos transcripts resulted in a partial but significant protection from HXR9-induced cell death ( Fig. 5B).
HXR9 does not trigger apoptosis in CD133+ cells derived from human umbilical cord blood. A number of previous studies have indicated that HOX genes are essential in the correct proliferation and differentiation of blood progenitor cells. In particular, HOXB4 has been shown to promote the proliferation of highly purified populations of HSC ( 30, 31). In light of these observations, we decided to test the toxicity of HXR9 on a population of cells affinity-purified from human umbilical cord blood using CD133, a marker of primitive blood progenitor cells ( 32). CD133+ cells continued to proliferate when treated with HXR9, although at a lower rate than CXR9-treated cells ( Fig. 6 ), and no significant differences in cell death were observed for the two treatments (data not shown).
No overt toxicity was apparent in HXR9-treated mice. HXR9 toxicity was assessed by i.v. administration to MF-1 mice at 15 mg/kg for 10 days. As HOX gene function is important in hemopoiesis, close attention was paid to potential adverse affects in the bone marrow and peripheral blood. Peripheral blood samples were taken prior to administering HXR9. After 10 days of treatment consisting of daily i.v. doses of 15 mg/kg, no adverse side effects were seen in HXR9- or CXR9-treated mice. Full blood counts were within expected reference ranges seen in the strain of mice used, and histologic analysis of the liver also revealed no abnormalities (data not shown).
In this study, we have shown that antagonizing the interaction between HOX and PBX proteins triggers apoptosis in malignant melanoma cells, both in vitro and in vivo. HOX genes are also known to promote the proliferation of normal HSCs however, raising the possibility that antagonizing the same interaction in these cells could likewise be cytotoxic ( 30, 31). Our findings indicate that this is not the case though, as purified HSC populations continue to proliferate even in the presence of HXR9, and that there is no overt toxicity observed in mice. These findings concur with previous work indicating that a mutant version of HOXB4 lacking a functional hexapeptide was actually more effective in promoting HSC proliferation than the wild-type sequence ( 33). Taken together, these results suggest that targeting the interaction between HOX and PBX could be selective for the malignant phenotype, in a way that direct knockout of HOX genes would not.
How does antagonizing the HOX/PBX interaction trigger apoptosis in B16 melanoma cells? Microarray analysis of cells treated with HXR9 has identified a number of target genes that become dysregulated, including the oncogene c-Fos, the expression of which increases 2.6-fold 2 h after HXR9 treatment. This increase in c-Fos transcription must be at least partly responsible for the induction of apoptosis, as knocking down its expression in B16 melanoma cells results in a considerable reduction in their sensitivity to HXR9. Fos and its binding partner Jun are both members of the bZIP superfamily of transcription factors, which are characterized by a basic DNA-binding domain combined with a leucine zipper region ( 34). JUN can homodimerize to form the activator protein (AP-1) transcription factor. FOS cannot homodimerize, but it could also form part of the AP-1 transcription factor by binding JUN. AP-1 activates the expression of a number of genes involved in cell cycle regulation, including cyclin D1 ( 35). Conversely, AP-1 also mediates apoptosis in a wide range of cell types including lymphocytes, fibroblasts, neurons, and retinal cells ( 22– 29). The mechanistic basis of AP-1–mediated apoptosis is unclear, and depends in part on the actual composition of the AP-1 dimer. Jun-containing AP-1 can promote the transcription of the Fas ligand (FasL), which promotes cell death through the FasL/Fas receptor pathway in lymphocytes, fibroblasts, and neurons ( 22– 26). Fos overexpression induces apoptosis in hepatocytes ( 36), and is also required for Myc-induced cell death in hepatoma cells ( 37). Furthermore, the re-expression of Fos in a number of established tumor cell lines can be anti-oncogenic through the activation of proapoptotic genes, although this seems to involve a different mechanism from that used by Jun ( 38).
In addition to c-Fos, a number of other target genes are also up-regulated by HXR9, of which Dusp1 and Atf3 show by far the greatest increase. The former encodes a dual specificity phosphatase, a protein that can remove phosphate residues from serine, threonine, and tyrosine residues of a wide range of substrates and can block signaling through the mitogen-activated kinase pathway, thereby blocking cellular proliferation (reviewed in ref. 39). Atf3 is a member of the activating transcription factor/cyclic AMP–responsive element binding protein family of transcription factors that are known for their role in stress response, including DNA damage ( 40). It has recently been shown that Atf3 can stabilize p53 by preventing its ubiquitination ( 41), and promote its tumor suppressor functions, including the induction of apoptosis ( 42). Atf3 also promotes cell cycle arrest and apoptosis in UV-treated fibroblast cells, and blocks Ras-mediated transformation ( 43).
Many of the other target genes identified in the screen also have antitumor properties. Kruppel-like factor 4 is a transcription factor that is known to be down-regulated in many colorectal cancers, inhibits the activity of the oncogenic transcription factor β-catenin, and prevents xenograft tumor growth in athymic nude mice ( 44). Mad is an inhibitor of another oncogenic transcription factor, myc, and also has a role in blocking proliferation (reviewed in ref. 45). Finally, Drak2 is a serine/threonine kinase with a high degree of sequence and functional homology to the death-associated kinases involved in apoptosis, and can induce apoptosis when overexpressed in NIH 3T3 cells ( 46).
The predominance of tumor suppressor and proapoptotic genes among HXR9 targets, together with the ability of HXR9 to induce apoptosis in B16 melanoma cells, suggests that the function of HOX gene expression may itself be to inhibit the intrinsic tumor–suppressing pathways. The relative importance of these targets is likely to vary from one cell type to another depending on the genetic background, but a generalized inhibition of many tumor suppressing pathways could account for the wide range of tumors that overexpress HOX genes. Together with the data presented here showing that HXR9 could selectively kill malignant cells, these observations suggest that the HOX/PBX interaction may present an important target in cancer therapy.
Grant support: Association for International Cancer Research (04-382), Cancer Research UK (C7822/A3832), and the Prostate Project.
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.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received November 17, 2006.
- Revision received April 10, 2007.
- Accepted April 19, 2007.
- ©2007 American Association for Cancer Research.