p16^INK4a Expression and Absence of Activated B-RAF Are Independent Predictors of Chemosensitivity in Melanoma Tumors

Abstract

Metastatic cutaneous melanoma is highly resistant to cytotoxic drugs, and this contributes to poor prognosis. In vivo studies on the chemosensitivity of metastatic melanoma are rare and hampered by poor response rates to systemic chemotherapeutics. Patients who undergo isolated limb infusion (ILI) with cytotoxic drugs show high response rates and are, therefore, a good cohort for studying chemosensitivity in vivo. We used tumors from patients who underwent ILI to study the role of melanoma tumor-suppressor genes and oncogenes on melanoma chemosensitivity. Prospectively acquired tumors from 30 patients who subsequently underwent ILI with melphalan and actinomycin-D for metastatic melanoma were investigated for mRNA expression levels of p14^ARF, p16^INK4a, and MITFm. The mutation status of B-RAF, N-RAS, and PTEN were also determined. A high percentage of tumors had activating mutations in either B-RAF (15/30) or N-RAS (10/30) and only two tumors carried altered PTEN. High expression of p16^INK4a and absence of an activating B-RAF mutation independently predicted response to treatment. Further, inducible expression of p16^INK4a sensitized a melanoma cell line to death induced by melphalan or actinomycin-D. This study shows that high expression of p16^INK4a or the absence of activated B-RAF correlates with in vivo response of melanoma to cytotoxic drugs.

Introduction

Metastatic cutaneous melanoma has a poor prognosis with 2-year survival of less than 30% in patients with visceral involvement [1]. This reflects the marked resistance of the disease to cytotoxic drugs. Single-agent treatment with the alkylating agent dacarbazine (5-(3,3-dimethyl-l-triazeno)-imidazole-4-carboxamide) remains the standard best systemic therapy for metastatic melanoma [2]. Although overall response rates to dacarbazine are less than 15% [3,4], certain patients have highly sensitive tumors, and rare complete remissions are often sustained [5]. To date, there are a few studies of in vivo molecular determinants of chemosensitive melanomas, and the mechanisms involved in melanoma chemoresistance have not yet been elucidated.

We sought molecular correlates of chemosensitivity in a unique cohort of prospectively acquired tumor samples from patients with metastatic melanoma receiving isolated limb infusion (ILI) chemotherapy [6]. Isolated limb infusion is a complex and technically demanding technique performed as a palliative procedure for extensive inoperable locoregional recurrence in a limb, usually from multiple in-transit seeding in cutaneous lymphatic vessels. A tourniquet is applied to isolate the circulation of the affected limb, allowing very high doses of cytotoxic drugs to be administered locally, while minimizing toxic systemic effects [6]. Isolated limb infusion is only performed in a few centers around the world on highly selected patients, and although cohort size will always be limited, it provides a unique platform for the study of chemosensitivity in melanoma. Firstly, unlike most clinical situations, pretreatment fresh-frozen tumor samples are readily available. Secondly, response rates to ILI are >50% [6], allowing greater statistical power to assess molecular correlates of response than in systemic chemotherapy, where response rates are <15% [7].

We selected for this analysis key candidate genes known to be important in melanomagenesis and also linked to the regulation of chemosensitivity [7]. We analyzed the expression of MITFm and the p14^ARF and p16^INK4a melanoma tumor-suppressor genes and the status of B-RAF, N-RAS, and PTEN.

The INK4a/ARF locus on chromosome 9p is the most frequently deleted region in melanoma [8] and is inherited in mutated form in 39% of melanoma-prone families [9]. INK4a/ARF encodes two tumor-suppressor proteins, p14^ARF and p16^INK4a, both of which have been shown to enhance the chemosensitivity of human cancer cells [10–13]. p14^ARF accumulates in response to oncogenic stimuli and stabilizes p53, leading to cell cycle arrest or apoptosis (reviewed in Sherr [14]). p16^INK4a activates the retinoblastoma pathway by inhibiting cyclin-dependent kinases 4 and 6 [15] leading to cell cycle arrest and, in some instances, cell death [16].

The microphthalmia-associated transcription factor (MITF) regulates development and differentiation of melanocytes [17] and is deregulated in melanoma. Increased levels of MITF may contribute to melanoma progression as MITF induces expression of the antiapoptotic molecule, Bcl-2 [18], and MITF ablation sensitizes melanoma cells to cytotoxic drugs [19]. Microphthalmia-associated transcription factor amplification is more frequent in metastatic melanoma and correlates with decreased patient survival [19]. In normal melanocytes, MITF transcriptionally activates p16^INK4a to promote cell cycle arrest [20] and, as expected, MITF amplification is accompanied by p16^INK4a inactivation in melanoma cell lines [19].

Activating N-RAS and B-RAF mutations are the most common oncogenic mutations in melanoma. Up to 80% of benign nevi [21] and 25% to 66% of melanomas contain activating B-RAF mutations [22,23]. Greater than 89% of these mutations alter a single amino acid (V600E and V600K) and another 5% to 6% of melanoma-associated B-RAF alterations affect exon 11 [24]. Activating N-RAS mutations occur in 5% to 30% of melanomas [23,25,26], with the highly recurrent mutations affecting Gly-12, Ala-18, and Gln-61, accounting for approximately 12%, 5%, and 70% of melanoma-associated mutations, respectively [24]. Activated B-RAF may also co-operate with loss of the PTEN tumor-suppressor in promoting melanoma development. PTEN attenuates phosphoinositide 3-kinase (PI(3)K/AKT) signaling, and the simultaneous activation of B-RAF and loss of PTEN simulate N-RAS activation to promote melanoma development [27,28]. The high frequency of N-RAS and B-RAF mutations in melanoma and the fact that these mutations have been correlated with poor prognosis [23] indicate that they are potential therapeutic targets, and yet there have been few studies investigating N-RAS/B-RAF status and melanoma response to common chemotherapeutic agents.

Materials and Methods

Patients

Fresh-frozen tumor excision biopsies were obtained from 30 consecutive consenting patients undergoing elective therapeutic ILI at the Sydney Melanoma Unit. These patients had not undergone prior systemic therapies. Catheters were inserted percutaneously into the axial artery and vein of the affected limb, and a pneumatic tourniquet was inflated proximally. Melphalan (7.5-mg/L tissue) and actinomycin-D (75-µg/L tissue) were rapidly infused into the arterial catheter while the limb was isolated. The infusate was then circulated for 20 to 30 minutes by catheter using a syringe attached to a three-way tap. Finally, the limb was flushed with saline before the tourniquet was released and the catheters were removed [6].

Patient details are shown in Table 1. All had lower limb skin metastases and were eligible for the study if at least one lesion greater than 10 mm³ in volume could be excised, leaving at least one other lesion in the same field that could be evaluated for response. Response was determined by standard response evaluation criteria in solid tumors at 4-week intervals until disease progression, recorded by digital photography, and confirmed by independent review of these records. The study was approved by the Sydney South West Area Health Service, Protocol No. X00-0274 and was initiated in March 2001.

Table 1.

Details of Patients Treated by ILI with Melphalan and Actinomycin-D.

	Responders	Nonresponders
Male/Female	5:16	2:7
Age, median (range), years	75 (48–93)	78 (46–86)
Best response, n	CR, 5	SD,* 7
	PR, 16	PD, 2
Mean time to best response	57 days	N/A
Mean time to PD in ILI field	272 days in responders	94 days in nonresponders
	(from 3 CR and 10 PR;	(from 2 PD and 5 SD;
	674 days in CR,	118 days in SD,
	152 days in PR)	49 days in PD)

CR indicates complete response; NA, not applicable; PD, progressive disease; PR, partial response; SD, stable disease.

^*Mean duration of stable disease was 74 days (derived from 5 SD).

RNA Extraction and cDNA Synthesis

The excised lesion was frozen immediately in liquid nitrogen. Samples were trimmed of surrounding tissue, and RNA was extracted using Trizol (Sigma, St. Louis, MO) and purified using RNeasy columns (Qiagen, Valencia, CA). Samples with evidence of melanin were cleaned by heating to 65°C for 5 minutes during Trizol extraction or by using Bio-Spin P-30 Tris columns (Bio-Rad, Hercules, CA). cDNA was synthesized from 100 ng to 1 µg of RNA using Superscript III (Invitrogen, Carlsbad, CA) with oligo-dT primers. No-SuperScript III enzyme controls were performed to ensure there was no contaminating genomic DNA.

Real-time Reverse Transcription-Polymerase Chain Reaction

Real-time reverse transcription-polymerase chain reaction (RT-PCR) was performed using a Corbett Rotorgene 3000 with Platinum SYBR Green qPCR Master Mix (Invitrogen) for MITF. For the remaining genes, TaqMan probes were used; the master mix used for p16^INK4a, 18S, and hP0 was Platinum qPCR SuperMix-UDG (Invitrogen); and Brilliant qPCR Master Mix (Stratagene, La Jolla, CA) was used for p14^ARF. DMSO 3% was added to p16^INK4a and p14^ARF reactions, and Mg²⁺ increased to 5 mM in p14^ARF reactions. Primers used were as follows:

Equal loading was ensured by adjusting samples to the same average expression levels for the ribosomal proteins human P0/36B4 and 18S. hP0 and 18S levels were measured by real-time RT-PCR using VIC-labeled predeveloped assay reagent probes (Applied Biosystems, Foster City, CA).

Mutation Status

Mutation status of N-RAS at codons 12, 16, 18, and 61 was determined by RT-PCR cleaned up with ExoSapIT (USB Corporation, Cleveland, OH) and sequenced on an ABI PRISM 3100 (Applied Biosystems). Primers used were as follows: N-RAS⁶¹_fwd: CTG ACA ATC CAG CTA ATC, N-RAS⁶¹_rev: GTC TTT TAC TCG CTT AAT CTG, N-RAS^ex1_fwd: AGC TTG AGG TTC TTG CTG GT and N-RAS^ex1_rev: CTC TCA TGG CAC TGT ACT CT.

Mutation status of B-RAF at codon 600 was determined by single-strand conformation polymorphism analysis. A 228-bp fragment of B-RAF spanning codon 600 was amplified by RT-PCR (forward primer: ATA GAT ATT GCA CGA CAG AC; reverse primer: ATC TTG CAT TCT GAT GAC TTC), placed in formamide/EDTA loading buffer, and denatured at 95°C for 10 minutes then placed on ice for 2 minutes. Samples were run on an SSCP gel (10% polyacrylamide gel containing 1x Tris-glycine buffer and 5% glycerol) for 3.5 hours at 22°C and then the gel was silver-stained. Samples showing aberrant bands on the gel were sequenced to confirm the mutation.

Mutation status of BRAF exon 11 and part of exon 12 was determined by sequencing RT-PCR products as detailed above. BRAF exon 11/12 primers were as follows: BRAF^exon11/12_fwd: GAC GGG ACT CGA GTG ATG AT, BRAF^exon11/12_rev: CTG CTG AGG TGT AGG TGC TG.

Mutation status of PTEN cDNA was determined by sequencing RT-PCR products as described in the study of Liu and Kagan [29]. Partial PTEN sequences were obtained for tumor samples 006 and 026 owing to limited RNA.

All sequences were analyzed by visual inspection and using the Australian National Genomic Information Service BioManager.1

Statistical Analysis

The data were analyzed using SPSS for Windows 15.0. Two-tailed tests with a significance level of 5% were used throughout. p16^INK4a, p14^ARF, and MITFm expression levels were log-transformed to approximate normality before analysis. Response was considered as a dichotomous outcome, with patients showing complete or partial response classified as responding and patients with stable disease or progressive disease classified as not responding. Logistic regression analysis was used to test for the association between all variables and the presence of response. The odds ratios (ORs) and their 95% confidence intervals (CIs) were used to quantify the degree of association between the independent predictors of response identified using multiple logistic regression with backward stepwise variable selection. Mann-Whitney correlation was used to quantify the degree of association between B-RAF and the expression of p16^INK4a, p14^ARF, and MITFm.

Cell Culture

The WMM1175_p16^INK4a cell clone carries the stably integrated-p16^INK4a gene under isopropyl-β-d-1 thiogalacto pyranoside-inducible expression control and has been described previously [30]. Melanoma cells were grown in DMEM/10% FBS in a 37°C incubator with 5% CO₂.

In cytotoxicity assays, cells were exposed to 1-mM IPTG for 4 days, followed by the addition of medium containing IPTG and drugs to a final concentration of 25-nM actinomycin-D or 200-µM Melphalan for a further 24 hours.

Cell Cycle and Apoptosis Analysis

For cell cycle analysis, cells were fixed in 70% ethanol at 4°C for at least 1 hour, washed in PBS, and stained with propidium iodide (50 µg/mL) containing ribonuclease A (50 µg/mL). DNA content from at least 2000 cells was analyzed using ModFIT software (Verify Software House, Topsham, ME). Numbers of cells with sub-G₁ content were determined using CellQuest software (Becton Dickinson, Franklin Lakes, NJ). Annexin V staining was performed as detailed by the manufacturer (Sigma). Western blots were probed for p16^INK4a (sc-467; Santa Cruz Biotechnology, Santa Cruz, CA) and β-actin (AC-74; Sigma).

Results

Patient Response

Fresh-frozen tumor biopsy samples from 30 melanoma patients who subsequently underwent ILI were analyzed. Tumor response to ILI was classified as responsive, which included tumors showing complete response (n = 5) and partial response (n = 16), and non-responsive, which incorporated stable disease (n = 7) and disease that progressed (n = 2; Table 1).

N-RAS and B-RAF Are Frequently Mutated in Melanoma

Most N-RAS and B-RAF mutations in melanoma occur at codons 61 (Q61K and Q61R) and 600 (V600E and V600K), respectively [23,31]. In addition, a small proportion of melanomas carries alterations affecting exon 11 of B-RAF and N-RAS at Gly-12 and Ala-18 [24]. We analyzed tumor samples for mutations affecting these codons using SSCP (Figure 1) and sequencing (Figure 2). Of the 30 tumors, 10 (33%) showed mutation in N-RAS and only codon 61 was altered. Another 15 (50%) melanomas contained B-RAF alterations; 9 with V600E, 4 with V600K, 1 with G466V, and 1 with L485F (Table 2). As expected, activation of N-RAS and B-RAF in our panel of melanoma tumors was mutually exclusive, and all mutations were heterozygous (Figure 2).

Figure 1.

SSCP analysis of a B-RAF RT-PCR fragment encompassing codon 600. Wild type (wt) control B-RAF PCR product was generated from the WMM1215 melanoma cell line and the B-RAF^V600E mutant control was amplified from a B-RAF^V600E expression plasmid. Amplified B-RAF cDNA from tumor samples are also shown with V600E and V600K mutation-positive tumors marked E and K, respectively. The remaining unmarked tumor samples were wild type at codon 600.

Figure 2.

Sequence analyses of amplified B-RAF and N-RAS transcripts. The nucleotide and codon changes encoded are shown on the left of each sequence. Numbering is based on GenBank accession NM_004333 (B-RAF) and NM_002524 (N-RAS). wt indicates wild type sequence.

Table 2.

B-RAF and N-RAS Mutation Status of 30 Melanomas.

Mutation	Nucleotide	n	Tumors Samples
N-RASQ61R	A435G	8	004, 010, 011, 018, 019, 042, 044, 046
N-RASQ61K	C434A	1	030
N-RASQ61K	AC433-434TA	1	001
B-RAFV600E	T1860A	9	002, 003, 007, 008, 015, 022, 023, 038, 041
B-RAFV600K	GT1859-60AA	4	005, 025, 027, 043
B-RAFG466V	G1458T	1	012
B-RAFL485F	G1516T	1	028
No mutation		5	006, 021, 026, 032, 045
Total		30

Expression Analysis of p16^INK4a, p14^ARF, and MITF

There was considerable variation in p14^ARF and p16^INK4a transcript expression (Figure 3). Most tumors (approximately 60%) displayed very low expression of both p14^ARF and p16^INK4a. In most cases, the expression of these transcripts was comparable, which is expected considering that these genes share genomic sequence. Tumors 004, 005, 007, 012, 030, and 041 showed discordant levels of the p16^INK4a and p14^ARF transcripts. The expression of MITF was also variable with only four tumors (006, 008, 025, and 045), displaying very low levels of this transcript (Figure 3).

Figure 3.

Relative expression levels of p14^ARF, p16^INK4a, and MITFm in melanoma tumors was determined using quantitative real-time RT-PCR. The mutation status of B-RAF and N-RAS for each tumor is also shown (N = mutant N-RAS, B = mutant B-RAF, - = wild type N-RAS and B-RAF). Tumor response classification is also indicated. CR indicates complete response; PD, progressive disease; PR, partial response; SD, stable disease.

Absence of B-RAF Mutation and p16^INK4a Expression Correlate with Better Response

We investigated whether the presence of N-RAS or B-RAF mutation or the expression level of p16^INK4a, p14^ARF, or MITF predicted response to chemotherapeutic drugs. The independent predictors of response in this tumor set were the absence of a B-RAF mutation and high log-transformed p16^INK4a expression (Table 3). N-RAS mutation was also weakly associated with response; although on multivariate analysis, B-RAF was the more powerful predictor, with seven of nine nonresponsive tumors harboring an activated B-RAF mutant. The expression levels of p14^ARF and MITF did not correlate with response. Further, a Mann-Whitney test showed that there was no correlation between the presence of B-RAF mutation and p16^INK4a, p14^ARF, or MITF expression (P = .861, P = .423, and P = .922, respectively).

Table 3.

Best-Fitting Multiple Logistic Regression Model of Response Together with Adjusted ORs and Their 95% CIs for Independent Predictors of Response.

	B	SE	P	OR	95% CI for OR
					Lower	Upper
BRAF mutation	-2.44	1.18	.039	0.087	0.009	0.88
log p16INK4a	0.60	0.26	.022	1.83	1.1	3.06
Constant	-0.882	1.27	.49	0.41

Mutations Affecting the PTEN Tumor-Suppressor Are Not Common in Melanoma

Considering that oncogenic mutations affecting N-RAS and B-RAF were not equivalent in predicting melanoma response, it seemed likely that the MAP kinase-signaling cascade, which is activated by both N-RAS and B-RAF, may not significantly influence treatment response. The main signaling cascade differentially activated by N-RAS and B-RAF is the PI(3)K/AKT pathway, and the integrity of this pathway was analyzed by screening the PTEN tumor-suppressor, which attenuates PI3K signaling. As expected, PTEN mutations were not common in our panel of melanomas and were identified in only three tumors (003, 005, and 045). The 045 tumor, which was wild type for both N-RAS and B-RAF, expressed wild type PTEN and PTEN with the G44G silent amino acid change (data not shown). Tumor 003 expressed BRAF^V600E and carried the homozygous PTEN mutant I253N, and tumor 005 expressed oncogenic B-RAF^V600K and was heterozygous for the PTEN P213S alteration (Figure 4).

Figure 4.

Sequence analyses of amplified PTEN transcripts derived from tumor samples 003 and 005. The nucleotide and codon changes encoded are shown on the left of each sequence. Numbering is based on accession NM_000314. wt indicates wild type sequence.

p16^INK4a Expression Induces Melanoma Cell Death

To investigate the impact of p16^INK4a expression on melanoma cell survival, we used a stable p16^INK4a-inducible melanoma line. Expression of p16^INK4a in this WMM1175_p16^INK4a cell clone was induced with 1-mM IPTG for a 5-day period (Figure 5A). As shown in Figure 5B, induction of p16^INK4a led to potent G₁ cell cycle arrest and cell death, as determined by the increased sub-G₁ population. Accumulation of p16^INK4a also sensitized WMM1175_p16^INK4a cells to death in response to the cytotoxic drugs melphalan and actinomycin-D. In particular, the sub-G₁ population increased from 18% in the presence of p16^INK4a alone to 26% with the addition of actinomycin-D and 30% in the presence of p16^INK4a and melphalan (Figure 5B). Consistent with these data, a significantly higher percentage of Annexin V-positive apoptotic cells was also observed in response to actinomycin-D or melphalan when p16^INK4a was induced (Figure 5C).

Figure 5.

Induced expression of p16^INK4a promotes arrest and sensitizes cells to cell death. (A) WMM1175_p16^INK4a melanoma cells were induced to express p16^INK4a with the addition of 1-mM IPTG to the medium for 4 days, then treated with IPTG and actinomycin-D, melphalan or carrier control for 24 hours. Accumulation of p16^INK4a was determined by Western blot analysis, and detection of β-actin was used to demonstrate equal protein loading. (B) The cell cycle distribution of the WMM1175_p16^INK4a cells was examined using propidium iodide (PI) staining. Cells were left untreated (left panels) or incubated with medium containing 1-mM IPTG for 4 days (right panels). Induced and uninduced cells were then treated with ethanol (Control), 25-nM actinomycin-D (ActD), or 200-µM melphalan (Mel) with or without IPTG for a further 24 hours. (C) Apoptosis assays of the WMM1175_p16^INK4a cells, monitored by Annexin V-FITC staining. Cells were left untreated (-IPTG) or incubated with medium containing 1-mM IPTG for 4 days (+IPTG). Induced and uninduced cells were then treated with ethanol (Control), 25-nM actinomycin-D (ActD), or 200-µM melphalan (Mel) with or without IPTG for a further 24 hours. Error bars represent SD from at least two independent experiments.

Discussion

In this study, we investigated the in vivo response of melanoma metastases to cytotoxic drugs administered by ILI to determine whether response correlated with the expression of p16^INK4a, p14^ARF, and MITFm and the mutation status of N-RAS, B-RAF, and PTEN. We have shown that high p16^INK4a expression and the absence of mutant active B-RAF correlate with chemotherapeutic response of melanoma tumors. The fact that melanomas commonly lack p16^INK4a and carry active B-RAF may therefore contribute to chemoresistance.

The high prevalence of B-RAF and N-RAS mutations (25/30) found in our melanoma samples is similar to other studies [22,23,26]. Likewise, the Gln-61 and Val-600 codons of N-RAS and B-RAF, respectively, were the common targets of alteration in our melanoma tumors (Table 2). In addition, one tumor contained the B-RAF G466V mutation and another, the L485F B-RAF mutant. The G466V mutation has been identified in melanoma and lung cancer [22,32,33], whereas the L485F mutation has not been reported in melanomas previously [22,24] but has been identified in individuals with the sporadic developmental disorder cardiofaciocutaneous syndrome [34,35]. More importantly, like the V600 B-RAF mutations, both G466V and L485F are within the B-RAF kinase domain and have elevated kinase activity in vivo [34–37]. N-RAS and B-RAF mutations were mutually exclusive, as previously reported [22], presumably because both these gene products signal through the MAP kinase pathway.

The differential role of B-RAF and N-RAS in predicting melanoma chemosensitivity (this study) and survival [38] may involve the PI(3)K/AKT pathway. Unlike B-RAF, N-RAS activates the PI(3)K/AKT pathway and several studies have implicated activation of the PI(3)K pathway as another crucial event in the progression of melanoma [39]. In particular, RAS activation of the PI(3)K/AKT pathway controls the activation of endoplasmic reticulum stress (ER-stress) response [39], which can activate cytoprotective or cytotoxic effects depending on the cellular environment [40]. When ER-stress is persistent or excessive, it can trigger cell death. For instance, the ER-stress-inducer thapsigargin selectively enhanced tumor necrosis-related apoptosis-inducing ligand (TRAIL)-induced apoptosis in melanoma cells [41]. B-RAF and RAS also differ in their capacity to promote growth arrest in normal human cells [42] and in their ability to produce invasive melanocytic neoplasia [27]. Finally, microarray studies show melanomas with an activating N-RAS mutation have different gene expression patterns to those with activating B-RAF mutations [43], and tumors containing N-RAS^Q61R are less dependent on MEK than tumors with BRAF^V600E [44].

The combined activation of B-RAF and loss of PTEN may simulate oncogenic N-RAS activity to simultaneously activate the MAP kinase and PI(3)K pathways. Accordingly, melanomas rarely express oncogenic N-RAS and altered PTEN [45], whereas they frequently carry oncogenic B-RAF and mutated PTEN [27]. In our sample set, only two melanomas carried missense PTEN mutations. These mutations (P213S and I253N) presumably inactivate PTEN because they are located in the PTEN lipid-binding domain and have been identified in a human glioma and endometrial cancer [46,47]. The two tumors with PTEN loss also expressed oncogenic forms of B-RAF, but patient response did not relate to PTEN status; 005 showed complete response, whereas 003 had stable disease after therapy. Considering the low frequency of PTEN alterations detected in melanomas, a significantly larger set of melanoma tumor samples needs to be analyzed to accurately define the contribution of PTEN and the PI(3) K pathway to melanoma chemosensitivity. This is particularly important because the combination of B-RAF inhibitors with PI(3)K inhibitors have been shown to co-operate in preventing melanoma cell proliferation and consistently enhanced melanoma chemosensitivity and suppressed invasive tumor growth [48].

Our set of melanoma tumor samples displayed heterogeneity in MITF expression with no apparent correlation with chemosensitivity. Inhibition of MITF function in melanoma may trigger CDK2-mediated growth arrest [49] or apoptosis through Bcl-2 down-regulation [18]. Accordingly, MITF loss has been shown to sensitize melanoma cells to the cytotoxic agents, cisplatin and docetaxel [19]. Alternatively, increased MITF expression has been shown to induce cell cycle arrest by activating the expression of the CDK inhibitor p16^INK4a [20]. Interestingly, MITF-M expression may be down-regulated during the spontaneous regression of melanoma [50]. We observed no association between MITF and p16^INK4a expression in our melanoma tumors (data not shown).

MITF activity is also regulated by the MAP kinase pathway; activation of MAP kinase signaling phosphorylates MITF, which simultaneously increases its transactivation potential and targets it for degradation through the 26S proteasome [51,52]. Thus, MITF amplification, which is found in 20% to 30% of melanomas, is often accompanied by B-RAF mutation and p16^INK4a loss [19]. However, we did not see any correlation between MITF and B-RAF expression levels in this study (data not shown).

Ectopic expression of p14^ARF enhances chemosensitivity in human tumor cell lines [10] and p14^ARF expression has been associated with improved prognosis in several cancers, including acute myeloid leukemias [53,54]. However, no association between p14^ARF expression and response was seen in our panel of melanoma tumors. Further, in a recent microarray study of melanoma tumors, p14^ARF expression was not associated with survival [55]. Although p14^ARF is frequently lost in human melanomas, this usually occurs in combination with p16^INK4a loss. Moreover, most mutations affecting the INK4a/ARF locus target p16^INK4a for inactivation [56], indicating that, in humans, p16^INK4a is the critical INK4a/ARF tumor-suppressor.

p16^INK4a loss correlates with poor prognosis in human cancers such as melanoma, leukemia, and renal cancers [57,58]. Although there is no necessary biological connection between poor prognosis and chemoresistance, it is possible that the high frequency of loss of p16^INK4a in melanoma is a direct contributor to treatment failure. Consistent with this hypothesis, we found that retention of p16^INK4a was associated with melanoma chemosensitivity and that accumulation of p16^INK4a promoted the death of melanoma cells. More importantly, p16^9INK4a co-operated with the cytotoxic drugs melphalan and actinomycin-D to enhance melanoma cell death. The mechanism by which p16^INK4a expression promotes cell death in response to drugs requires investigation. This is especially relevant because melanoma cells have usually lost p16^INK4a and display an intrinsic resistance to drug-induced cell death.

Although most tumor samples displayed comparable levels of p14^ARF and p16^INK4a message, only p16^INK4a expression predicted disease response. This is because two nonresponders (012 and 041) accumulated high levels of p14^ARF mRNA but no p16^INK4a transcript (Figure 3). The intimate genomic organization of the p16^INK4a and p14^ARF genes allows for co-ordinated expression [59], although independent promoters primarily drive their expression [60]. We have previously shown that disruption of this co-ordinate expression is associated with melanoma predisposition [59], but in this report, we found no evidence that discordant expression of p16^INK4a and p14^ARF correlated with treatment response.

In summary, this study shows that p16^INK4a expression and absence of activating B-RAF mutation correlate with in vivo response of metastatic melanoma exposed to high doses of locally administered cytotoxic drugs. This study reinforces the importance of the INK4a/ARF locus and highlights the PI(3)k/AKT pathway, in the regulation and execution of apoptosis in melanoma, and suggests directions to improve its responsiveness to conventional chemotherapeutic agents.

MITFm_fwd:	CCG TCT CTC ACT GGA TTG GTG
MITFm_rev:	GCT TGC TGT ATG TGG TAC TTG G
p16INK4a_fwd:	GCC CAA CGC ACC GAA TAG

p16INK4a_rev:	ACG GGT CGG GTG AGA GTG
p16INK4a_probe:	FAM6-TCA TGA TGA TGG GCA GCG
	CC-TAMRA (A/B 450025)
p14ARF_fwd:	CTA CTG AGG AGC CAG CGT CT
p14ARF_rev:	ACG GGT CGG GTG AGA GTG
p14ARF_probe:	FAM6-TCA TGA TGA TGG GCA GCG
	CC-TAMRA (A/B 450025)