Local Adjuvant Treatment with Low-Dose CpG-B Offers Durable Protection against Disease Recurrence in Clinical Stage I–II Melanoma: Data from Two Randomized Phase II Trials

This article is featured in Highlights of This Issue, p. 5659

Malignant melanoma is a considerable healthcare problem in Caucasian populations, and incidence rates are expected to continue to rise (1). In the Netherlands alone, mortality has more than doubled and incidence has almost quadrupled since the early 1990s (2). The prognosis of newly diagnosed patients depends on the pathologic stage of the disease (3). Of all newly diagnosed cases, 85% to 90% present with localized disease. The 10-year survival rates of early-stage localized melanoma (pathologic stage IA–IIC) range from 95% to 40% (3), depending on the Breslow thickness, the presence of ulceration, and the number of mitoses per mm² in the primary tumor. When regional lymphatic metastases are present (pathologic stage IIIA–C), the 10-year survival rates drop to 68% to 24% (3), confirming the insidious nature of the disease. Unfortunately, the treatment options for pathologic stage I–III melanoma are still very limited because there is no widely accepted and implemented adjuvant treatment option that can safely lower the risk of distant recurrence.

We and others have reported on the advantages of locally administered immunotherapy. Topical, intradermal (i.d.), intranodal, or intratumoral administration of Toll-like receptor agonists (like CpG; refs. 4–7) and also of cytokines (like GM-CSF and IFNα; refs. 4, 8–11) has been clinically explored as monotherapy, in combination, or as an adjuvant to vaccines against melanoma. Because of the lower dose and serum levels, local treatment reduces the chances of autoimmune adverse events without compromising antitumor immune protection (12). After surgical removal of the primary tumor, early-stage melanoma lends itself to local immune potentiation of the tumor-draining lymph nodes, and in particular of the sentinel lymph node (SLN), which bears the brunt of tumor-induced immune suppression but is also a rich source of antitumor immune effector cells (13).

In an effort to harness antitumor immunity in the SLN and in more downstream tumor-draining lymph nodes, two consecutive single-blinded randomized phase II clinical trials were conducted to investigate the effects of i.d. administration of the dendritic cell (DC)–stimulatory compound unmethylated CpG type B oligodeoxynucleotide CPG 7909 (CpG-B), alone or combined with GM-CSF, at the primary melanoma excision site prior to routine re-excision and sentinel node biopsy (SNB) in clinical stage I–II melanoma patients (see Fig. 1 for an overview). Because the inclusion of patients took place prior to SNB, clinical stage I–II patients were included that could potentially be diagnosed with pathologic stage III disease after SNB. We compared posttreatment immune parameters of the treated patients with patients who received plain saline (0.9% NaCl) and observed activation in the SLN of in particular lymph node (LN)–resident DC subsets, as well as increased antimelanoma CD8⁺ T-cell rates in the SLN and peripheral blood (PB; refs. 5, 6, 14). With only minor and transient toxicities, these trials showed that i.d. administration of CpG-B, with or without GM-CSF, was safe and a biologically active treatment option with possibly protective effects against systemic tumor spread. Here, we report on the outcome of a meta-analysis of local and systemic DC and T-cell responses and the recurrence-free survival (RFS) of the patients who participated in these two trials.

Fifty-two patients with clinical stage I–II melanoma (according to the criteria of the American Joint Committee on Cancer) who were scheduled to undergo SNB were enrolled in two consecutive single-center, single-blinded, randomized, and placebo (saline)-controlled phase II clinical trials between June 2004 and June 2007 at the VU University Medical Center in Amsterdam, The Netherlands (registered as ISRCTN63321797; for an overview of these trials we refer to Fig. 1). At the time of these trials, every melanoma patient who entered our hospital was scheduled for an SNB, and there was no clinical cutoff (e.g., Breslow) to qualify for the procedure, and as a consequence, to be enrolled in the clinical studies. The anticipated biological efficacy and primary endpoint, i.e., DC and melanoma antigen-specific CD8⁺ T-cell activation in the SLN, were observed in the CpG-B treatment arms of both trials, regardless of the dose level, combination with GM-CSF, or dosing schedule (4–6, 14), and we therefore combined all participating patients for the current clinical follow-up analysis (Fig. 1).

The SLN tumor status of all patients was yet to be determined at the time of randomization. Some of the patients were diagnosed with stage III melanoma only after the SNB (Table 2). Although preoperative ultrasound examinations were not part of the protocol, patients who were suspected of having LN metastases at physical and/or ultrasound examination were excluded from participation in the trials as well as patients who had undergone previous immunotherapy or chemotherapy and patients who had received immunosuppressive medication or were suffering from any autoimmune disorder. For an overview of patient and tumor characteristics, we refer to Table 1. The studies were approved by the Institutional Review Board of the VU University Medical Center, and written-informed consent was obtained from each patient before treatment in accordance with the Declaration of Helsinki.

In the first trial (hereinafter referred to as “trial I”), patients were randomly assigned (1:1) to receive CpG-B or plain saline. In trial II, patients were randomly assigned (1:1:1) to receive CpG-B, CpG-B and GM-CSF, or plain saline (Fig. 1). Because patients were recruited during a prespecified period, treatment allocation in both trials did not have the exact 1:1 and 1:1:1 ratio. Sealed opaque envelopes were used for allocation concealment. Within both studies, the injections had the same volume and were visually identical. Only the patients were masked to treatment allocation.

In trial I, 24 patients were assigned randomly to receive one preoperative intradermal injection at the tumor excision site of either 8 mg CpG-B (PF-3512676, formerly named CPG 7909, Coley Pharmaceutical Group; n = 11) or a saline placebo (n = 13), 7 days prior to SNB (5, 6). Trial II was a three-arm clinical study in which 28 patients were randomized to receive intradermal injection of either 1 mg CpG-B (PF-3512676, Coley Pharmaceutical Group) alone (n = 10) or in combination with 100 μg GM-CSF (Leukine, Berlex Laboratories Inc.; n = 9), or a saline placebo (n = 9; ref. 4); the injections were given directly adjacent to the scar of the primary melanoma excision, 7 and 2 days before surgery.

Frequencies and activation state (by CD83 expression levels) of migratory DC subsets [i.e., skin-derived Langerhans cells (LC) and dermal DC (DDC)] and LN-resident CD14⁻ and CD14⁺ conventional DC (cDC) subsets in SLN-derived single-cell suspensions were determined by four-color flow cytometry as previously described (4, 15). In short, LC were gated as CD11c^intCD1a^hi, DDC as CD11c^hiCD1a^int, CD14⁻ LN-resident DC as CD1a⁻CD11c^hiCD14⁻, and CD14⁺LN-resident DC as CD1a⁻CD11c^hiCD14⁺. Based on these new cDC subset definitions, available FACS data from trial 1 (5) were reanalyzed and combined with the previously published cDC data from trial 2 (4) in a meta-analysis of placebo versus treated patient groups. Cryopreserved SLN cells from patients participating in trial I were used for flow-cytometric reanalysis of the newly identified CD11c^hiCD1a⁻ LN-cDC subsets; of note, only for 3 patients from the placebo group of this trial, cryopreserved SLN cells were available for this analysis. Induced T-cell responses to a panel of HLA-A2–restricted epitopes from shared melanoma antigens were determined as previously described in SLN suspensions and in pre- and posttreatment peripheral blood mononuclear cells (6, 14). In brief, after polyclonal anti–CD3- and –CD28-mediated expansion of SLN-derived T cells (in order to obtain sufficient numbers of cells from all patients), CD8⁺ T-cell reactivity was measured against the melanoma-associated epitopes gp100_154–162, gp100_209–217, gp100_280–288, MAGE-A3_271–279, MART-1_26–35, NY-ESO-1_157–165, and Tyrosinase_369–377, either by PE-labeled HLA-A2 tetramer binding and flow cytometric read-out (Trial II, tetramers kindly provided by Dr. Ton Schumacher, Netherlands Cancer Institute, Amsterdam, the Netherlands) or by overnight IFNγ Elispot read-out (Trial I). Elispot assay and Tm-binding methodologies, as well as definitions of positive responses, were all as previously described (6, 14). T-cell data from trials I and II were combined in a meta-analysis of placebo versus treated patient groups.

All patient and tumor characteristics (Table 1) were collected from the patients' medical records and pathology reports. Patient follow-up data were collected from medical records or retrieved via the patients' general practitioner. RFS and distant RFS (DRFS) were recorded as time (in months) since the combined re-excision and SNB until diagnosis of any melanoma recurrence.

The primary outcome measure of this follow-up analysis was RFS of the treated patients compared with RFS of the saline control groups from the two trials (see Fig. 1 and Table 1). We also analyzed the pathologically confirmed stage I–II melanoma patients (no LN metastases found at pathologic examination) and patients who only received CpG-B without GM-CSF separately.

Differences in patient characteristics and immune reactivity parameters between the study groups were analyzed with the two-tailed Student t test, χ² test, or the two-tailed Fisher exact probability test (when a subgroup had only 5 or less patients). HRs and differences in survival curves were analyzed with the log-rank test. Differences were considered significant when P < 0.05. Microsoft Excel (version 2010) and GraphPad Prism (Version 6.02) were used for all graphs, tables, and analyses.

In total 52 patients were included in the current analysis, who were enrolled in two randomized phase II trials and received either injections with CpG-B alone (n = 21) or in combination with GM-CSF (n = 9), or a saline placebo (n = 22, see Fig. 1). At baseline, there were no significant differences in clinical characteristics between the saline-administered patients (placebo group) and the CpG-B–treated patients (treated group), see Table 1. However, when the SNB procedure was performed (1 week after the experimental injections), a remarkable and significant difference in the number of tumor-positive SLN (pathologic stage III) was observed, i.e., 3 of 30 (10%) in the treated group versus 8 of 22 (36%) in the placebo group (P = 0.04), see Table 2.

We hypothesized that the observed difference in tumor involvement between the placebo and treated groups might well have been due to the successful boosting of locoregional antitumor immunity. We previously showed for trial II that local CpG-B administration at the tumor excision site led to selective recruitment and activation of both CD14⁺ and CD14⁻ LN-resident cDC subsets with cross-presentation ability (4). With this knowledge, we reanalyzed the flow cytometric DC measurements in SLN from patients participating in the first trial. We now report that we found similar results for trial I (data not shown). A pooled analysis of the treated versus placebo groups from both trials clearly confirmed the selective recruitment and activation of the CD14⁻ and CD14⁺ LN-resident cDC subsets in the SLN of patients receiving local CpG-B injections (Fig. 2A). T-cell responses against HLA-A2–restricted epitopes from shared melanoma antigens (MART-1, Tyrosinase, gp100, MAGE-A3, and NY-ESO-1) were determined in available HLA-A2⁺ SLN and peripheral blood (PB) samples, by IFNγ Elispot read-out (in trial I) or by Tm binding (in trial II). As we previously observed a 74% to 88% concordance between positive responses based on parallel Elispot and Tm-binding read-outs (8, 14), we performed a pooled analysis of the T-cell response rates from trials I and II (Fig. 2B). Of note, the PB T-cell reactivity data from trial II were not previously reported. Consistent with the recruitment and activation of LN-resident cDC subsets, reported to have particular T-cell (cross-)priming abilities (4, 16), and the lower number of patients with tumor-involved SLN, higher melanoma-specific T-cell response rates were found in SLN of CpG-B–treated patients (consistently in both trials, data not shown). Importantly, PB melanoma-specific T-cell response rates were also significantly increased posttreatment in the CpG-B–treated group (7–21 days after the experimental injections), indicative of systemic immune protection following local CpG-B administration, which is consistent with observations previously reported for trial I (6).

Identical enrollment criteria and consistent immune modulating effects of CpG-B in both trials further supported a pooled clinical follow-up analysis. After a median clinical follow-up of 88.8 months (range, 5–133 months), a clearly improved RFS was found for the treated patients as compared with the placebo controls (HR, 0.16; 0.06–0.65; P = 0.008), see Fig. 3A. In the treated group, we recorded two melanoma recurrences: the first patient was treated in trial I (8 mg CpG-B) and had a first recurrence at 5 months after the SNB. The patient was one of the three treated patients who were actually shown to have a tumor-positive SLN. The patient died 14 months after SNB from diffuse melanoma metastases. The second patient was treated in trial II (1 mg CpG-B only arm) and had a first recurrence at 26 months after the SNB. The patient had a tumor-negative SLN but died at 40 months after SNB from metastatic disease. In the placebo group, 9 of the 22 patients suffered a disease recurrence. Three of these patients had a locoregional metastasis only (i.e., excision site, regional LN, or in-transit) and were alive without evidence of disease at the time of follow-up. Six had distant metastases and had died from metastasized melanoma at the time of follow-up. When we next analyzed the pathologically confirmed stage I–II patients separately, we still found a significant difference in RFS in favor of the treated group versus the placebo control group (HR, 0.12; 0.03–0.75; P = 0.02; see Fig. 3B). Among these patients, we found one recurrence in the treated group and five with recurrences in the placebo group of whom three had died from metastasized melanoma at the time of the last follow-up. Also for patients treated with CpG-B only (excluding the CpG/GM-CSF combination arm from trial II), clinical benefit was established: an improved RFS compared with the placebo control group (HR, 0.22; 0.09–0.91; P = 0.03; see Fig. 3C). Of note, although it failed to reach significance, also DRFS in treated patients was higher than in placebo controls (HR, 0.26; 0.07–1.06; P = 0.076), as shown in Fig. 3D. The strong prognostic value of this parameter was confirmed by the fact that all patients who developed distant recurrences eventually succumbed to the disease.

The clinical follow-up analysis of two randomized and placebo-controlled phase II trials presented in this article demonstrates that adjuvant treatment of clinically stage I–II melanoma patients with intradermal injection of CpG-B may provide protection against (distant) disease recurrence. These findings are highly relevant because there is no widely accepted and applied adjuvant treatment option for this patient group. Of all newly diagnosed melanoma cases, 85% to 90% present clinically with localized disease, which, despite surgical removal with curative intent, still leads to disease recurrence at a later date in one third of the cases (3). Even patients with a confirmed negative SLN after surgery still run a reported 16% risk of recurrences (17, 18). Local recurrences of melanoma are already associated with a 48.5% mortality rate at 5 years of follow-up (3). For distant metastases, this rate even rises to 90% (19). Although these figures are likely to drop due to recent developments in the treatment of advanced-stage melanoma patients, in particular with immune checkpoint inhibitors, these costly treatment regimens are often associated with serious autoimmune adverse events and, as yet, in a substantial proportion of cases do not prevent eventual progression of the disease (20, 21). Even though thin melanoma (<1 mm) carries a low risk of recurrence (on average 3%–5%), it is not negligible (in particular when one considers the sheer volume of this newly diagnosed group; refs. 1–3) and certainly justifies the single low-dose administration of CpG-B as adjuvant treatment.

We found an improved and highly durable RFS in the treated group: the latest (and only second) recurrence was recorded at 26 months after SNB. Our data even suggest that the treatment afforded protection against distant recurrence (which is associated with a dismal prognosis; ref. 19), as evidenced by a considerable, though with these patient numbers non-significant, DRFS advantage. The RFS curves of the treated group show a plateau in their tail which is typically seen in survival curves of immunotherapeutic approaches in metastasized melanoma (21). This strengthens our hypothesis that the patients in the treated group are in fact benefiting from durable systemic immune protection.

The patients in the treated group also had significantly less tumor-positive SLN (P = 0.04), whereas other known prognostic factors like older age, gender, melanoma of the head/neck and trunk, Breslow thickness, and tumor ulceration did not differ significantly between the groups (Table 1; refs. 17, 18, 22). Previously, we reported that the immune status of the SLN from the patients who were treated with CpG-B was significantly enhanced and that their SLN were much bulkier (4, 5). This leads us to the hypothesis that the lower number of tumor-positive SLN in the adjuvant treatment group is due to the rapid (i.e., within the 7 days between local adjuvant treatment and SNB) elimination of the clinically occult nodal metastases by SLN-resident memory T cells that were boosted by local CpG-B administration as demonstrated for the combined trial I and II results in Fig. 2B. This is actually in line with recent reports that T cells, once released from immune-suppressive constrictions, can rapidly clear even bulky tumors (23), and is further supported by experimental in vivo evidence demonstrating rapid effector differentiation from resident memory T cells in a noncognate and type-I IFN-dependent manner (24). Locally administered CpG-B may act on various levels of innate and adaptive immunity to curb metastatic spread, by boosting DC activation (5, 6) and recruitment (5), by natural killer (NK)–cell activation (14), and the rapid induction of systemic T-cell responses against melanoma-associated antigens (6, 14), as previously shown upon intratumoral delivery of CpG-B (with an abscopal effect on distant tumors; ref. 7) or when administered in conjunction with a vaccine (25, 26). Indeed, the combined immune monitoring data from trial I and II clearly point to a robust effect on the recruitment and activation of in particular the CD14⁻ LN-resident cDC subset, which we and others have identified as the possible human equivalent of the murine CD8α subset with cross-priming abilities (4, 27). This subset has recently been shown by Gajewski and colleagues to be vital for effective antimelanoma immunity and to be recruited and activated by type-1 IFN responses, which are also induced by local CpG-B administration (4, 28). Thus, our data point to a type-I IFN-mediated activation and boosting of both locoregional and systemic antimelanoma T-cell responses following local application of CpG-B (Fig. 2), which may have secured the apparent protection against later disease recurrence (Fig. 3). We have not endeavored to relate the measured T-cell immunity directly to RFS as the immune monitoring in this study entailed only a subset of HLA-A2⁺ patients and did not include neoantigens, now known to be vital T-cell targets for tumor protection (29). Moreover, observed increases in effector NK cells may also be involved in the prolonged RFS (6, 14). Careful immune monitoring in the context of a planned larger randomized phase III trial may shed further light on underlying immune mechanisms and may lead to the identification of useful predictive biomarkers.

The absence of tumor cells in the SLN is a good predictor for improved survival (19). To rule out that the adjuvant treatment with locally administered CpG-B had led to downstaging without RFS benefit or that the distribution of stage III patients was uneven between the treatment and placebo arms by mere chance, we also analyzed the histologically confirmed stage I–II patients separately (Fig. 3B). The observed RFS increase in the treated group of this subanalysis (P = 0.02) strengthens the hypothesis that there was no downstaging without RFS benefit, in which case the opposite trend might have been expected. It also provides further evidence for the induction of durable systemic antimelanoma immunity, preventing late recurrences that are observed even at these lowest stages of the disease. Although the number of tumor-involved SLN in the placebo group of our study was somewhat higher than generally reported (2, 3), we were able to preclude that this might have resulted from the saline injections as we found a similar frequency of stage III disease among untreated patients undergoing re-excision and SNB at the same time as the conduct of these trials (data not shown). The strict randomization of the enrolled patients, the complete equivalence in patient and tumor characteristics between the placebo and treatment groups, and the clear immune activating effects of the administered CpG-B, both at a locoregional and systemic levels, bolster our confidence that the observed differences in clinical outcome are due to the protective effects of local CpG-B treatment. Nevertheless, caution is called for in view of the small numbers of patients enrolled in these trials, and our interpretation of the clinical outcome data should therefore be regarded as hypothesis generating, requiring confirmation in a larger randomized phase III trial.

The observation of significantly improved RFS by CpG-B monotherapy (Fig. 3C) makes CpG-B the clear candidate of choice to take forward for further clinical testing. However, we have found evidence to suggest that the combination of GM-CSF and CpG-B may induce even greater antitumor activity than CpG-B alone. GM-CSF, unlike CpG-B, enhanced activation of migratory skin-derived cDC in the SLN (4, 5), and combined treatment with CpG-B and GM-CSF induced superior activation of LNR-cDC, which in turn was related to a superior capacity for cross-presentation to CD8⁺ effector T cells (4). Moreover, although the CpG-B/GM-CSF combination group in trial II was too small (n = 9) to make any firm conclusions, no recurrences were observed in patients receiving the combination treatment (see Supplementary Fig. S1 for the RFS of the separate arms in trials I and II).

In conclusion, local administration of CpG-B, aimed at immune potentiation of the SLN, may provide a safe and widely applicable adjuvant treatment for early-stage melanoma (i.e., clinical stages I and II), affording long-term protection against locoregional and distant recurrence. To deliver on this remarkable promise of intradermally administered CpG-B and translate it into the new standard treatment for patients with clinical stage I–II melanoma, the conduct and long-term follow-up of a randomized phase III trial are required.

No potential conflicts of interest were disclosed.

Conception and design: B.D. Koster, B.G. Molenkamp, R.J.C.L.M. Vuylsteke, A. Baars, P.A.M. van Leeuwen, R.J. Scheper, M.P. van den Tol, A.J.M. van den Eertwegh, T.D. de Gruijl

Development of methodology: B.D. Koster, M.F.C.M. van den Hout, B.G. Molenkamp, P.A.M. van Leeuwen, M.P. van den Tol, A.J.M. van den Eertwegh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.D. Koster, M.F.C.M. van den Hout, B.J.R. Sluijter, B.G. Molenkamp, R.J.C.L.M. Vuylsteke, P.A.M. van Leeuwen, M.P. van den Tol, A.J.M. van den Eertwegh, T.D. de Gruijl

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.D. Koster, M.F.C.M. van den Hout, B.G. Molenkamp, A. Baars, P.A.M. van Leeuwen, M.P. van den Tol, A.J.M. van den Eertwegh, T.D. de Gruijl

Writing, review, and/or revision of the manuscript: B.D. Koster, B.J.R. Sluijter, B.G. Molenkamp, A. Baars, M.P. van den Tol, A.J.M. van den Eertwegh, T.D. de Gruijl

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.D. Koster, M.F.C.M. van den Hout, B.J.R. Sluijter, B.G. Molenkamp, R.J.C.L.M. Vuylsteke, P.A.M. van Leeuwen, M.P. van den Tol

Study supervision: B.G. Molenkamp, R.J.C.L.M. Vuylsteke, P.A.M. van Leeuwen, M.P. van den Tol, A.J.M. van den Eertwegh, T.D. de Gruijl

The authors thank Pepijn Wijnands and Sinéad Lougheed for excellent technical assistance, Dr. Hans van der Vliet for critical reading of the article, and Dr. Art Krieg for provision of CPG 7909, for invaluable advice and discussions, and for his contribution to the writing of this article.

This study was supported by the Fritz Ahlqvist Foundation.

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.