Low-Dose IFN-γ Induces Tumor MHC Expression in Metastatic Malignant Melanoma

The response rate to IFN-γ in metastatic melanoma is low (0–12%) across a range of doses (1, 2, 3, 4). In contrast, IFN-α produces responses rates of 12–22% in metastatic melanoma (5), and unlike IFN-γ is of benefit in the adjuvant setting. This lack of effect is perhaps surprising, because IFN-γ has substantially more immunomodulatory and antiproliferative effects than IFN-α. One possibility is that effective antitumor doses of IFN-γ do not equate with doses approaching the maximal tolerated dose. There is evidence for this. In animal models, the antitumor effects of IFN-γ exhibit a bell-shaped dose-response curve (6). Likewise, induction of various markers of immune activation by IFN-γ exhibit bell-shaped dose-response curves (6, 7, 8). In humans receiving IFN-γ, bell-shaped dose-response curves are described for lymphocyte and monocyte activation (9), antibody-dependent monocyte cytotoxicity, Fc receptor expression (10), and induction of serum β-2-microglobulin and neopterin levels (11). These studies have not uniformly defined an optimum dose for immune activation, but almost all suggest that IFN-γ 100 μg/m² is associated with near maximal activity, at least in terms of these parameters.

Animal data indicate that daily IFN-γ therapy is less effective in suppressing melanoma growth than triweekly therapy (6). Animal and clinical data also indicate that repeated once daily injections eventually down-regulate immune parameters that are initially up-regulated by IFN-γ (8, 11, 12, 13). Similarly, triweekly IFN-γ, at doses between 10–500 μg initially up-regulated serum neopterin and β-2-microglobulin levels, but these were suppressed with continued therapy (11), whereas once weekly dosing induced prolonged immune activation, which was not down-regulated with repeated dosing (14). Hence, the antitumor effects of IFN-γ may be schedule dependent.

HLA expression is down-regulated in a significant proportion of melanomas (15, 16, 17, 18). It is postulated that this enables tumor progression. The immune modulatory properties of IFN-γ include up-regulation of MHC class I and II expression, induction of MHC class II expression on tissue devoid of it, and enhancement of macrophage function. These factors are of presumed importance in the antitumor effects.

This study was performed: (a) to test the hypothesis that low melanoma response rates to IFN-γ might be related to nonoptimal dosing and schedules; and (b) to analyze the ability of IFN-γ to overcome one mechanism of immune escape, HLA down-regulation. Because 100 μg/m² once weekly appears to have optimal immune effects, we performed a Phase II study of this dose schedule in patients with metastatic melanoma. Evidence for HLA induction was sought by measuring changes in tumor HLA expression and serum β-2-microglobulin, a marker of MHC class I up-regulation. In addition, we assessed changes in serum neopterin, a marker of macrophage-monocyte activation.

Recombinant human IFN-γ (Immukin; Boehringer Ingelheim) was supplied in vials containing an isotonic solution of 200 μg/ml. One-hundred μg/m² (maximum dose 200 μg) was administered s.c. every week for a maximum of 6 months.

Patients had histologically proven melanoma with biopsy accessible cutaneous or lymph node metastases. Other eligibility criteria included: age >18 years, WHO performance status ≤2, life expectancy of >3 months, and no chemotherapy or investigational agents in the previous 30 days (40 days for mitomycin C or nitrosoureas). Eligible patients had to have a white cell count >3 × 10⁹/liter, neutrophil count >1.5 × 10⁹/liter, hemoglobin >10 g/dl, platelet count >100 × 10⁹/liter, a normal serum creatinine and bilirubin, and serum transaminases <2 × upper limit of normal. Pregnant females, patients positive for hepatitis B, hepatitis C, or HIV, and those with autoimmune disease, or receiving corticosteroids or other hormonal treatment were ineligible.

A treatment cycle comprised 28 days of weekly dosing. During the first 24 h of treatment, patients were observed in hospital. Patients were reviewed at 7 days, and thereafter at monthly intervals. At each clinic visit, symptoms were assessed and clinical examination performed. Full blood counts and routine biochemistry were checked at the same time points. Providing there was no evidence of disease progression, treatment was continued for a maximum of 6 months. Patients underwent clinical response evaluation monthly and radiological restaging every 3 months.

Standard WHO criteria for objective response assessment were used (19). NCI-CTC³ were used to grade toxicity.

Plasma was obtained into heparinized plastic tubes, containing preservative-free heparin, and assayed for β-2-microglobulin and neopterin, at baseline, and 4 and 24 h, 1 week and monthly after commencing therapy.

Neopterin levels were measured by an enzyme immunoassay kit (ICN Pharmaceuticals Inc., Diagnostics Division). The assay uses a specific antineopterin antibody to bind its corresponding antigen. The assay range was from 0.5 to 100 ng/ml. Mean neopterin level in a random sampling of 80 healthy volunteers was 1.4 ± 0.6 (SD) ng/ml.

β-Microglobulin was measured by rate nephelometry using a Beckman Specific Protein Analyser. Reference normal range is 0–3 mg/liter with a lower limit of detection of 1 mg/ml.

In each patient, an accessible lesion underwent FNA or core biopsy before treatment, 24 h after the first dose, 1 week later, and 1, 3, and 6 months after commencing therapy. FNA material was suspended in 2 ml of RPMI 1640 (Life Technologies, Inc., Biocult Ltd.) kept on ice and cytospins prepared within 1 h, fixed in acetone, and stored at −70°C. Core biopsies were performed using a 14 or 16 FG Temno automated biopsy needle (Bauer Medical International, Santo Domingo, Dominican Republic). Samples were placed in Nunc freezing vials and snap frozen in liquid nitrogen. Cryosections (8 μm) from the frozen core biopsies were fixed in acetone and stored at −70°C.

Primary antibodies used were: anti-HLA-class I (W6/32), anti-HLA-class II (CR3/43), anti-ICAM-1 (6.5.B5), and mouse IgG1, IgG2a, and IgG2b isotype controls (Dako, Cambridge, UK). Anti-β-2-microglobulin (BBM.1) and anti-HLA-A2.1 (BB7.2) hybridomas were obtained from the American Tissue Culture Collection (Manassas, VA). Anti-B and C heavy chains (HC10), and anti-A2 heavy chain (HCA2) were the gift of Dr. Hidde Ploegh, Amsterdam, the Netherlands. The ABComplex/alkaline phosphatase anti-alkaline phosphatase system (Dako) was used for visualization as per the manufacturer’s instructions. The samples from each patient were run together simultaneously with cytospins of human monocyte-derived dendritic cells as positive control slides for each antibody. All of the slides were reviewed with a senior pathologist (K. G.) and scored positive if >50% melanoma cells stained for that antibody.

The study was approved by the Central Oxford Research Ethics Committee and conducted according to the declaration of Helsinki. Each patient provided written informed consent.

To ensure a low probability of erroneously rejecting a treatment that is active in 20% of patients, at least 14 patients were treated, according to principles described previously (20). Data were compared by Mann Whitney t tests.

The characteristics of the patients are listed in Table 1. Tumor response data were available for all of the patients (Table 2). Three patients attained a CR. The remaining patients had progressive disease. All three of the responding patients had low volume disease. One had disease confined to skin (4 s.c. deposits), which resolved within 1 month of commencing treatment. He continued on treatment for 6 months, and the CR has been maintained for 36 months. The second patient had a single s.c. metastases and a distant lymph node metastases. The lymph node was impalpable 1 month after commencing treatment, and the skin deposit, which was a biopsy-proven recurrence, had reduced to fulfill criteria for a PR. This did not change in size for the remaining 5 months of treatment. It was then excised, and there was no evidence of melanoma in the specimen. There has been no relapse during 53 months of follow-up. The third patient had previously developed rapid local recurrence after axillary clearance of malignant nodes, and at the start of IFN-γ treatment had multiple small (<1 cm) lung lesions, which were not biopsied, but assumed to be metastases. After 1 month of treatment, the axillary mass had reduced by >50% and at 3 months was no longer palpable. The pulmonary metastases had regressed at 3 months, and by 9 months had resolved, leaving residual scarring. The response has been maintained for 25 months.

The treatment was well tolerated with no grade 3–4 toxicities (Table 3). The majority of patients developed mild flu-like symptoms for 24 h after treatment, which responded well to paracetamol. There were no treatment-associated biochemical changes. One patient developed transient grade 2 lymphopenia, and 2 patients developed grade 1 lymphopenia during treatment.

Plasma β-2-microglobulin and neopterin levels were measured 4 h, 24 h, 7 days, and 28 days after commencing treatment. Data were available for β-2-microglobulin levels in 19 patients. Mean β-2-microglobulin levels (Fig. 1) were significantly elevated compared with baseline (mean = 0.5 ± 0.73 mg/ml) only at 24 h after commencing treatment (mean = 1.29 ± 0.83 mg/ml; P < 0.0015). Neopterin levels were available for 12 patients (Fig. 2) and were significantly elevated compared with baseline (mean = 1.64 ± 0.68 ng/ml) only at 24 h (mean = 2.98 ± 1.1 ng/ml; P < 0.0067). It is important to note that levels checked at 7 and 28 days were from samples obtained immediately before receiving IFN-γ.

Tumors from 18 patients were evaluable. Four patients had two lesions biopsied and in all 4, the two lesions gave consistent results.

Only 2 patients had tumor biopsy samples, which were W6/32 negative (α chain monomorphic determinate) before treatment (Table 4). Tumor from both became strongly positive, one 7 days and the other 3 months after commencing IFN-γ. The mechanism of class I up-regulation was assessed using antibodies to other components of the class I complex. For 1 patient, β-2-microglobulin expression was not detected before treatment and up-regulated 3 months after commencing IFN-γ (Table 4; Fig. 3). For the other patient, B and C heavy chain expression was undetectable before, and up-regulated after IFN-γ treatment.

To determine whether there was selective allelic loss, tumor HLA-A2.1 expression was assessed in the 6 HLA-A2.1-positive patients (Table 4). Before treatment, tumors from 2 patients were W6/32-positive and HLA-A2.1-negative. One of these stained HLA-A2.1-positive 7 days after commencing IFN-γ, and the other remained negative throughout treatment. In both patients, W6/32 staining was strong.

Tumors from 14 of 18 patients stained negative for class II expression before treatment (Table 5). Tumors from 6 patients (43%) became positive with IFN-γ treatment (Fig. 4). The rapidity of class II up-regulation after commencing IFN-γ ranged from 1 day to 3 months. In 4 patients, tumor biopsies were available from times after conversion from class II negative to positive, and in all 4 tumors the up-regulation persisted.

Tumors from 6 patients were evaluated for ICAM-1 expression, because this is up-regulated in vitro by IFN-γ. Before treatment, tumors from 4 patients stained ICAM-1-positive and during treatment 2 of these became negative. Tumors from 2 patients were negative for ICAM-1 before treatment, stained positive after 1 month, but stained negative after 3 months of treatment.

For technical reasons and/or because of complete tumor responses, it was not possible to correlate tumor responses with tumor antigen induction. There appeared to be no significant relationship between neopterin, and β-2-microglobulin up-regulation and tumor antigen induction (Table 5).

In this study, IFN-γ induced tumor HLA expression in approximately half the patients. Likewise, induction of the surrogate markers of monocyte activation (neopterin) and HLA class I up-regulation (β-2-microglobulin) was observed in most patients tested. These effects are consistent with the physiological properties of IFN-γ. The response rate in this study (13%) was low, but the 3 responses were durable and complete. Such durable and CR rates are reported for various immune therapies in a small proportion of patients with metastatic melanoma (21). It is salient that all 3 of the responding patients had low volume disease, and responses were not observed in patients with visceral disease. Previous trials of higher-dose, more frequent IFN-γ schedules in metastatic melanoma have reported response rates of 0–12% (1, 2, 3). Amalgamating responses from these previous trials, there were 7 PRs and 1 CR among 140 treated patients. Two of these responses were observed in patients (total 14 evaluable) receiving 0.01 mg/m². Hence, the current low-dose weekly regime appears as effective as higher-dose, more frequent schedules, and although this was a small trial, may be more effective.

This is the first study to show that IFN-γ up-regulates HLA expression in tumors in vivo. This induction was generally detectable 7 days after the first IFN-γ treatment and persisted throughout the study. It is important to note that apart from the subset of patients who had biopsies 24 h after starting IFN-γ, the remaining biopsies were performed 7 days after IFN-γ dosing, indicated that tumor HLA induction is a long-lived effect of IFN-γ. However, there was no correlation between tumor HLA induction and responses.

In this study, 90% of the tumors expressed HLA class I, as determined by staining with Mab W6/32, which recognizes a conformational epitope on the intact MHC molecule containing both β-2-microglobulin and the heavy chain. This accords with several previous studies (15, 16, 17). Others have reported a lower rate of class I expression in metastases compared with primary tumors and postulated class I down-regulation as a mechanism for immune escape (22). In the 2 patients with W6/32-negative tumors at baseline, there was loss of tumor β-2-microglobulin expression in 1 and loss of B and C heavy chain expression in the other. IFN-γ appeared to induce class I expression in these 2 patients by reversing this selective loss. Studies in the subset of HLA-A2.1-positive patients showed selective allelic loss in tumors from 2 patients. The observation of complex changes in class I expression are in accord with other studies in melanomas (22), and it is of interest that IFN-γ restored allelic expression in 1 of these patients.

Tumors from 14 of 18 patients did not express HLA class II before IFN-γ therapy. In half, class II was up-regulated by IFN-γ. Other studies report class II expression in 0–80% of human melanomas (18, 23). This variation must, in part, reflect different staining techniques and criteria for positivity. In some series, class II appears to be expressed more in locoregional metastases than in primary melanomas and associated with a worse prognosis (23). In contrast, others found class II expression correlated with a better prognosis (24) or had no correlation with prognosis (15). Therefore, the prognostic significance of class II expression in melanoma is unclear. In other tumor types, loss of class II expression correlates with a worse prognosis (25, 26).

Two previous clinical studies assessed the effects of IFN-γ on melanoma HLA expression. Kim et al. (18) studied 12 patients receiving IFN-γ 0.1–0.3 mg/m²/day. No tumors expressed class II antigens before treatment, and there was no up-regulation during IFN-γ treatment. Furthermore, a physiological effect of IFN-γ is induction of class II on normal keratinocytes, and this was only observed in 2 patients. IFN-γ 3–5 mg/m² thrice weekly also failed to induce tumor class II expression (27). Such data suggest that the IFN-γ dose and schedule used in the current study is optimal for tumor MHC induction.

In animal models, MHC class II expression induced by IFN-γ restored melanoma immunogenicity. In our patients, class II up-regulation was not associated with clinical responses. One possibility is that the induced antigens were unable to functionally present antigen, because melanocytes are not professional antigen-presenting cells. Antigen presentation by nonprofessional antigen-presenting cells can fail to induce efferent immune reactivity and can induce tolerance. Hence, neoexpression of class II may be a tumor promotion mechanism. However, in patients with melanoma receiving other immune therapies, such as IL-2 or granulocyte macrophage colony stimulating factor, induction of class II expression correlated with responses (28, 29).

Neopterin is a surrogate marker of monocyte activation (30) and β-2-microglobulin a marker of MHC class I induction. Induction of both neopterin and β-2-microglobulin levels was detected in the majority of patients, but only 24 h after commencing IFN-γ. This implies that IFN-γ significantly activates monocytes and up-regulates tissue HLA class I expression, but the effect is short lived. These observations are in accord with previous studies (11). The short-lived induction of β-2-microglobulin contrasts with the results in tumors, where IFN-γ-induced HLA up-regulation persisted throughout the study. Because there was no correlation between tumor antigen induction, and neopterin and β-2-microglobulin induction, these surrogate markers are not useful correlates of IFN-γ-induced events in tumors.

High response rates in metastatic melanoma have been reported recently with combination cytokine and combination chemoimmunotherapy regimens (31, 32, 33). Given that durable CRs were observed in the current study, IFN-γ at the dose and schedule used here would be a useful addition to combination chemoimmunotherapy regimens, because it should not significantly exacerbate toxicity. In addition, multiple mechanisms are defective in antigen presentation by melanoma cells, and IFN-γ reverses a significant proportion of these defects (34). Hence, the drug could also have a role in combination with other immune approaches, such as vaccine treatment, with the aim of enhancing immune reactions to tumor antigens. Some indication of the potential for this dose of IFN-γ was indicated by its use in combination with cytotoxic chemotherapy in the treatment of advanced ovarian cancer, where it significantly enhanced response rates and survival (35).

To conclude, the study shows that significant immune modulation by IFN-γ does not correlate with tumor responses and that serum surrogate marker changes do not reflect tumor events. The durable and long-lived responses, clear demonstration of MHC up-regulation, and low toxicity indicate that weekly IFN-γ 100 μg/m² would be a potentially useful addition to chemoimmunotherapy regimens.

We thank Boehringer Ingleheim for the supply of interferon gamma. We acknowledge support from the research nursing staff and the data team for conducting the trial, and finally to patients who participated in the trial supported by Cancer Research UK.