Melanoma exerts immune-suppressive effects to facilitate tumor progression and metastatic spread. We studied these effects on dendritic cell (DC) and T-cell subsets in 36 melanoma sentinel lymph node (SLN) from 28 stage I–III melanoma patients and determined their clinical significance. Four conventional DC subsets, plasmacytoid DCs, and CD4+, CD8+, and regulatory T cells (Tregs), were analyzed by flow cytometry. We correlated these data to clinical parameters and determined their effect on local and distant melanoma recurrence, with a median follow-up of 75 months. In stage I and II melanoma, increased Breslow thickness (i.e., invasion depth of the primary melanoma) was associated with progressive suppression of skin-derived migratory CD1a+ DC subsets. In contrast, LN-resident DC subsets and T cells were only affected once metastasis to the SLN had occurred. In stage III patients, increased CD4:CD8 ratios in concert with the accumulation of Tregs resulted in decreased CD8:Treg ratios. On follow-up, lower frequencies of migratory DC subsets proved related to local melanoma recurrence, whereas reduced maturation of LN-resident DC subsets was associated with distant recurrence and melanoma-specific survival. In conclusion, melanoma-mediated suppression of migratory DC subsets in the SLN precedes local spread, whereas suppression of LN-resident DC subsets follows regional spread and precedes further melanoma dissemination to distant sites. This study offers a rationale to target migratory as well as LN-resident DC subsets for early immunotherapeutic interventions to prevent melanoma recurrence and spread. Cancer Immunol Res; 5(11); 969–77. ©2017 AACR.
The sentinel lymph node (SLN) is a pivotal site not just in terms of early metastasis but also as the birthplace of antitumor immunity (1). Its importance is particularly clear for melanoma, being one of the most immunogenic tumors (2–4). The SLN is not only the most probable site for early metastasis, but also the location where melanoma (neo-)antigens first drain and melanoma-derived antigen-presenting cells (APC) first migrate. It is thus the site where melanoma antigens are first presented to the naïve immune system and the critical initial decision between activation and tolerance is made. The spread of melanoma through the lymphatic system may be facilitated by a bias toward tolerance that is apparent in tumor-draining lymph nodes (TDLNs) and most notably in the SLN. The SLN is converted into an immune privileged site by the primary melanoma (5). Melanoma-derived factors suppress the activation and maturation of dendritic cells (DC), inducing cross-tolerance and precluding an effective antitumor immune response (1, 6, 7). In addition, immune suppressive cells like regulatory T cells (Treg) induce regional and systemic tolerance (1, 6, 7).
Although much has yet to be learned about the different DC subsets present in human lymph nodes, we and others have characterized different conventional DC (cDC) subsets in SLN-derived single-cell suspensions in relation to their ability to stimulate T cells (8, 9). Based on their phenotype, we can identify both CD1a+ Langerhans cells (LC) and dermal DC (DDC) in SLN suspensions. Both subsets migrate from the skin to draining LN and prime antigen-specific Th cells or CTLs (10, 11). Besides these two migratory subsets, draining LNs contain two CD1a− LN-resident cDC subsets (CD14− and CD14+; ref. 8), likely recruited from blood precursors (12). In a comparative analysis, the LN-resident cDC subsets were more powerful T-cell stimulators in terms of allogeneic T-cell priming and IFNγ induction, despite their lower phenotypic maturation in the steady state than the skin-derived subsets (8). In addition, these CD1a− LN-resident cDC subsets express BDCA3 and the C-type lectin receptor CLEC9A, and correspond to the CD8α+ DC subset in the murine spleen, which has powerful CTL cross-priming ability, an important feature for the generation of antitumor immunity. In vitro studies have demonstrated the ability of human BDCA3+CLEC9A+ DCs to cross-prime CTLs and thus confirmed their homology with the murine CD8α+ DC subset (13–15).
In contrast to the growing knowledge on the distinct functional abilities of human DC subsets in lymph nodes, little is known of their interaction with tumors and the phenotypic and functional consequences. In this study, we analyzed DC subsets by multi-parameter flow cytometry in 36 SLNs from 28 patients with pathologically confirmed stage I–III melanoma. In addition, the SLN T-cell content, including Tregs, was analyzed. SLN cells were obtained using a previously described scraping technique, without interfering with routine pathologic diagnostics (16, 17).
We found evidence for sequential changes in DC and T-cell subsets in conjunction with melanoma progression. Given the central role of DCs in balancing an effective antitumor immune response with cross-tolerance (1, 5), we hypothesized that the melanoma related effects on migratory and LN-resident DC subsets could affect melanoma development and spread in these early-stage patients. Frequencies of migratory subsets proved to be related to local melanoma recurrence, whereas the maturation status of LN-resident DC subsets was related to distant recurrence and melanoma-specific survival. Together, our findings offer a rationale and possible targets for immune-based strategies in the adjuvant treatment of early-stage melanoma.
Materials and Methods
Human SLN material was obtained after written informed consent from clinically stage I/II melanoma patients undergoing an SLN triple-technique procedure (16) at the VU University Medical Center. Invasion depth of the primary tumor (Breslow thickness, expressed in mm) was determined by the local pathologist prior to referral of the patients to the VU University Medical Center. SLNs were sampled in the context of 2 clinical phase II trials (ISRCTN63321797), conducted between May 2004 and June 2007, which were approved by the institutional review board of the VU University Medical Center (12, 18). These studies were performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendment. In both studies, patients who had undergone previous immunotherapy or chemotherapy were excluded as well as patients receiving immunosuppressive medication or suffering from any autoimmune disorder. For this study, we used immune profiling data from the placebo control patients who received injections with plain saline at the excision site of the primary tumor prior to re-excision and SLN biopsy. The data from these patients were supplemented with SLN profiling data from an additional 14 off-study and untreated melanoma patients who underwent an SLN procedure around the time of the above listed phase II trials between March 2004 and January 2011. Of note, immune and clinical parameters were not significantly different between these groups (Supplementary Table S1). The materials were obtained under supervision of a pathologist, without interfering with diagnostic procedures, according to a protocol approved by the scientific committee of the Department of Pathology and following National Academic Hospital regulations for the proper use of clinical materials for scientific research.
Harvesting SLN cells
Immediately after removal, SLNs were collected in sterile ice-cold complete medium, composed of Iscove's modified Dulbecco's Medium supplemented with 25 mmol/L HEPES buffer (BioWhittaker/CAMBREX) with 10% FCS, 50 IU/mL penicillin-streptomycin, 1.6 mmol/L L-glutamine, and 0.05 mmol/L ß-mercaptoethanol. Before routine histopathologic examination of the SLN, viable cells were isolated by a cytological scraping method as previously described (16). In short, SLNs were bisected lengthwise with a surgical scalpel, and the cutting surface of the SLN was scraped 10 times with a surgical blade (size no. 22; Swann Morton Ltd.). SLN cells were rinsed from the blade with medium containing 0.1% DNAse I, 0.14% Collagenase A (Boehringer Mannheim), and 5% FCS, incubated for 45 minutes at 37°C, and subsequently placed in PBS with 5 mmol/L EDTA for 10 minutes on ice. Finally, the SLN cells were washed twice in complete medium, counted, and further processed.
After cell harvesting all SLNs were worked up according to the EORTC MG pathology protocol designed by Cook and colleagues (19). Immediately after harvesting, the SLNs were fixed for 24 hours in buffered formalin and subsequently embedded in paraffin. From each face of the lymph node, five serial step sections of 4 μm each were cut with 50 μm intervals between different numbers of sections. Finally, all sections were stained with hematoxylin and eosin and S100 and/or MelanA. All SLNs with tumor burden were reviewed by an experienced pathologist. SLN tumor load was classified according to the Rotterdam criteria (i.e., the maximum diameter of the largest SLN tumor lesion was determined; ref. 20).
Flow-cytometric DC and T-cell analyses
Freshly isolated SLN cells were directly stained with antibodies labeled with either FITC, PE, PE-CY5.5, PerCP-CY5.5 or APC and analyzed by flow cytometry at 100,000 or 200,000 events per measurement, as previously described (12, 18). The following monoclonal antibodies with matching isotype control antibodies were used: CD3, CD4, CD8, CD11c, CD14, CD25, CD123 (Becton Dickinson), CD1a, CD40, CD80, CD86 (PharMingen), CD83 (Immunotech), BDCA-2/CD303 (Miltenyi Biotec). For Treg detection in patient samples predating 2005, CTLA4 was stained intracellularly using the BD Cytofix/Cytoperm Kit following the manufacturer's instructions. From 2005 onwards, intracellular FoxP3 staining was performed using the eBioscience PE-antihuman FoxP3 staining set following the manufacturer's instructions; see gating procedures in Supplementary Fig. S1A. In 22 SLN, both CD3+CD4+CD25hi CTLA4+ and FoxP3+ Tregs were analyzed. In a comparative analysis, percentages correlated (P = 0.0002) and did not differ significantly between these Treg detection methodologies (Supplementary Fig. S1B).
For comparisons between melanoma disease stages, we applied one-way ANOVA tests with post-hoc Tukey for comparisons between two groups. Correlations were determined by use of the Pearson r test. For comparisons between two groups, Student t test was applied. Differences in Kaplan–Meier survival curves were analyzed with the log-rank test. Differences were considered significant when P < 0.05.
Patient and SLN characteristics
A total of 28 patients with clinical stage I and II melanoma were included in this study (clinical characteristics summarized in Table 1). From these patients, 36 SLNs were examined. In 8 patients, two SLNs were identified, excised, and sampled. In 10 patients, the SLN proved to be tumor positive (36%). Thus, after pathologic examination, 10 patients were histopathologically diagnosed with stage I, 8 with stage II, and 10 with stage III disease according to the AJCC staging system for melanoma. No significant differences in gender, age, or interval between primary tumor excision and SLN procedure were found between these three groups. Breslow thickness, as a measure of the primary tumor size/invasion depth, was significantly higher in stage II and III compared with stage I patients (P < 0.05), but comparably high in stage II and III patients (see Table 1). Clinical follow-up data were available for all patients with a median follow-up of 75 months.
|.||All .||Stage I .||Stage II .||Stage III .||Pa .|
|SLN tumor burden (mm)||1.21||N/A||N/A||1.21||N/A|
|.||All .||Stage I .||Stage II .||Stage III .||Pa .|
|SLN tumor burden (mm)||1.21||N/A||N/A||1.21||N/A|
aBy one-way ANOVA
SLN DC subset status by melanoma disease stage
Frequencies and maturation and activation state of five SLN DC subsets were assessed: two skin-derived migratory CD1a+ conventional DC subsets (i.e., CD1ahiCD11c+CD14− LC and CD1a+CD11chiCD14− DDC), two LN-resident CD1a− conventional subsets (i.e., CD1a−CD11c+BDCA3+CD14− and CD1a−CD11c+CD14+), and CD123hiBDCA2+ plasmacytoid DC (pDC; ref. 8). Gating procedures applied for these subsets are in Supplementary Fig. S2A. To explore possible effects of disease progression on these DC subsets, we grouped assessed DC parameters according to disease stage following the AJCC staging system. Figure 1A shows the frequencies and maturation status of the different cDC subsets [exemplified by CD83 expression, both by percentage and mean fluorescence intensity (MFI)]. Histograms showing examples of CD83 or CD40 expression levels on the DC subsets are shown in Supplementary Fig. S2B. No significant differences in cDC subset frequencies between the disease stages were found (Fig. 1A). Skin-derived migratory CD1a+ DC subsets showed a significant decrease in maturation and activation states in stage II as compared with stage I patients, but no differences were found between stage II and III patients; in contrast, LN-resident cDC subsets only showed significantly decreased maturation and activation markers in stage III melanoma (Fig. 1B). The pDC maturation state (exemplified by CD40) was significantly decreased in stage III, compared with stage I, melanoma (P = 0.05; Fig. 1C). No differences in frequencies of any of the other common SLN leukocyte subsets were found between disease stages, indicating that no shifts in the lymph node composition could account for the observed changes in DC subset frequencies (Supplementary Fig. S3).
Primary melanoma growth progressively suppresses skin-derived SLN DC subsets
Breslow thickness is one of the prognostic factors in melanoma patients and an important parameter in the AJCC staging system. We used this parameter as a measure of primary tumor burden to investigate its effect on the different DC subsets.
In line with the observed frequencies, maturation and activation of the cDC subsets in stage I and II melanoma patients (Fig. 1A and B), we only observed significant inverse correlations between Breslow thickness and the maturation state of LCs and DDCs as measured by CD83 expression levels (Fig. 2A and B).
Lymph node tumor burden affects LN-resident DC subsets
Lymph node tumor burden can be expressed as the maximum diameter of SLN tumor deposits. This parameter has been shown to be an independent prognostic factor in stage III melanoma patients (21). We used this parameter to investigate the effects of SLN tumor burden on different DC subsets and T cells in the SLN. Whereas no effect of Breslow thickness (i.e., primary melanoma tumor burden) was found on CD1a− LN-resident DC subsets, SLN tumor burden did affect these subsets. DC frequencies and activation marker levels from tumor positive SLNs were plotted against the maximum SLN tumor size. A significant positive correlation was observed with frequencies of CD1a− cDC subsets (CD1a−CD11c+BDCA3+CD14− and CD1a−CD11c+CD14+; Fig. 2C), whereas the maturation status of these DC subsets decreased with increasing metastatic burden (Fig. 2D), suggesting recruitment and accumulation of suppressed, tolerogenic LN-resident DC in lymph nodes with growing metastases.
Lower rates of mature migratory cDC in proximal than in distal SLN
In 7 patients, two SLNs were identified and we could, using a combination of SLN location, uptake of Patent Blue dye, and γ probe counts, identify a more proximal and distal draining SLN in relation to the tumor (i.e., SLN1 and SLN2). Three of these patients had stage I or II melanoma, 4 had stage III disease (see Supplementary Table S2 for details). In all stage I and II patients, migratory DDCs proved to be more numerous in the distal as compared with the proximal SLN, as well as significantly more activated (as measured by CD40 and CD80, see Fig. 3A). In contrast, no differences were seen for stage III patients (Fig. 3B). These observations indicate suppression of the migratory cDC subsets by the primary tumor, preceding local metastasis.
Increased SLN Treg rates in stage III melanoma patients
In patients with tumor positive SLNs (i.e., stage III melanoma patients), we observed an increase in numbers of CD4+ T cells coinciding with decreased frequencies of CD8+ T cells resulting in significantly increased CD4:CD8 ratios (Fig. 4A). CD4+ T-cell frequencies correlated significantly with SLN tumor burden (Fig. 4B). In conjunction with increased CD4:CD8 ratios, Tregs (defined as CD4+CD25hiCTLA4+ or CD4+CD25hiFoxP3+ T cells, see Materials and Methods and Supplementary Fig. S1A for applied gating) were significantly increased in the SLNs of stage III melanoma patients (Fig. 4C), resulting in significantly decreased CD8:Treg ratios as compared with stage I and II patients (Fig. 4D). The differences between tumor stages when assessing Tregs based on a CD4+CD25hiFoxP3+ phenotype in a limited number of patients were similar to those shown in Fig. 4C (Supplementary Fig. S4). Of note, Treg rates in the SLN were significantly inversely correlated with percentages of mature CD14−BDCA3+ LN-resident DCs (by CD83 expression, Fig. 4E).
Migratory and resident cDC subsets are differentially related to local or distant recurrence
To determine the clinical relevance of the observed melanoma-induced immune suppressive effects, we investigated local and distant melanoma recurrence intervals. Frequencies of skin-derived migratory DC subsets proved to be associated with local melanoma recurrence. Patients harboring above median frequencies of these subsets in their SLN experienced an increase in local recurrence-free survival (LRFS, Fig. 5A and B). In keeping with these findings, when patients were grouped according to experiencing local recurrence or not, LC and DDC frequencies were higher in the recurrence-free patients (Supplementary Fig. S5A) as were the expression levels (albeit modestly) of the maturation/activation markers CD83, CD40, CD86, and CD80 (Supplementary Fig. S5B).
The maturation state of both LN-resident cDC subsets was found to be related to distant recurrence-free survival (DRFS, Fig. 5C and D). Four patients with below median expression of CD83 on these LN-resident cDCs experienced distant disease recurrence. All these patients eventually died from melanoma. In contrast, none of the patients with above medium CD83 expression levels experienced distant recurrence. When grouping patients by their experiencing distant recurrence or not, no differences in frequencies of LN-resident cDC subsets were apparent (Supplementary Fig. S6A), whereas maturation/activation markers were lower on these subsets in patients who ultimately experienced distant recurrence (Supplementary Fig. S6B). No differences in local or distant recurrence were noted for frequencies or maturation status of any of the other DC or T-cell subsets, including Tregs and CD8:Treg ratios.
Early work by Cochran and coworkers indicated that lymph nodes close to melanoma tumors were more immunologically compromised than more distal ones (1, 22, 23). In their studies of melanoma SLN, they observed a reduction in the frequencies of paracortical DCs as well as in the complexity of their dendritic processes. These studies were largely based on morphology and immunohistochemistry, providing a rough measure of DC frequency and activation status but not accurate differentiation between different DC subsets.
Given the complex composition of the DC compartment in human lymph nodes (8, 9), we used multi-color flow-cytometric analyses to investigate melanoma-induced effects on different DC and T-cell subsets. Combining this analysis with the AJCC melanoma staging system, we identified a stepwise pattern in which the primary melanoma first suppresses skin-derived DC subsets. The inverse correlations found between the activation state of these DC subsets and Breslow thickness shows that this suppression progresses as the primary melanoma grows in situ. This is in agreement with the studies by Cochran, Essner and colleagues and with subsequent reports, which showed immune suppressive changes in the SLN preceding lymphatic spread (1, 24, 25).
The clinical relevance of these findings is demonstrated by the observation that frequencies of CD1a+ DC subsets were related to local melanoma recurrence (Fig. 5A and B). Although, in contrast to the activation status of these DC subsets, frequencies did not significantly change in relation to tumor stage, we did find a trend toward lower skin-derived DC frequencies in higher stage patients (Figs. 1A and 2A). Also, in 3 patients in whom two SLNs were harvested, skin-derived DC frequencies were lower in the first, proximal draining SLN, compared with the second, distal SLN downstream (Fig. 3A). These findings are in agreement with reports showing lower DC frequencies in the SLN of melanoma patients (26), and might either be explained by hampered DC migration, for example, through inhibition of lymphatic vessel formation (27), or by tumor-derived TGFβ-1 mediating local DC apoptosis (28). Together, our observations provide evidence for a role of CD1a+ skin-derived DCs in local melanoma control. They are also in agreement with reports on the protective effect of CD1a+DC-LAMP+ DCs in the melanoma SLN, although these studies didn't discriminate between local and distant recurrence (29, 30).
Once melanoma has metastasized to the SLN, additional immune effector subsets are affected. In such patients, unlike in stage I and II patients, we observed suppression of LN-resident DC subsets, higher CD4:CD8 ratios, and increased frequencies of Tregs, leading to decreased CD8:Treg ratios. Suppression and frequencies of CD1a− LN-resident DC subsets correlated with SLN tumor burden, suggesting a direct metastasis-related effect. These results contrast with some studies that did not observe differences in DC activation between tumor positive and negative melanoma SLN (24, 31). However, those studies made use of immunohistochemistry and could not discriminate between different DC subsets. Other studies have also found increased frequencies of Tregs in tumor positive SLN (32, 33), but conflicting reports exist (31). Again, this might be due to the use of immunohistochemistry as Tregs are best detected by use of a complete marker set by flow cytometry (34).
The tolerogenic milieu of the SLN thus imposed by the melanoma may, as proposed by Munn and colleagues (5), exert a tolerizing influence on the systemic antitumor immune response, facilitating further tumor dissemination. This stepwise pattern of consecutive local and systemic immune suppression following melanoma progression through stages I to III, however, does not explain the occurrence of distant metastasis in stage I and II patients (35). In our study, 2 stage II patients presented with visceral metastasis after 22 and 67 months of follow-up. Both patients had low expression levels of maturation markers on LN-resident DC subsets, in keeping with the apparent relationship of immune suppression of these subsets with distant metastasis (Fig. 5C and D). Indeed, for the whole study group, distant melanoma recurrence (and melanoma-associated death), irrespective of tumor stage, was related to the maturation state of LN-resident DC subsets and not to frequencies or the activation state of skin-derived migratory DC subsets.
Previous reports have described various prognostic markers in the melanoma SLN. Indoleamine 2,3-dioxygenase (IDO) expression has been proposed (36) and appears correlated to Treg activity, which may be induced by IDO-expressing pDCs (37, 38). In agreement with Gerlini and colleagues (37), we did observe higher frequencies of immature pDC in higher disease stages (Fig. 1C) and although we did not determine IDO expression levels, we also observed higher Treg rates in tumor-involved SLN (Fig. 4C). However, out of all DC subsets in the melanoma SLN, a relationship with Treg rates was only found for the CD14−BDCA3+ LN-resident subset: low maturation state of this subset was associated with high Treg rates (Fig. 4E). Whereas others have found high Treg rates in tumor positive melanoma SLN to be associated with poor prognosis (31), we did not find a relationship between Treg frequencies or CD8:Treg ratios and disease recurrence. Thus, in early stage melanoma, LN-resident DC subsets appear to direct effective systemic antitumor immunity, with more prognostic consequences than Treg levels.
Human BDCA3+CLEC9A+ cDCs have been identified as cross-presenting DCs that most closely resemble the splenic murine CD8α+ cDC subset, established as responsible for in vivo cross-priming of potent CD8+ cytotoxic effector T cells (13–15). DC-mediated cross-presentation of tumor-derived (neo-)antigens to CD8+ T cells is vital for the elicitation of an antitumor immune response and the activation status of these cross-presenting DC is paramount to the decision between the induction of immunity against tumor epitopes or the induction of cross-tolerance (5). Therefore, the apparent association between distant disease recurrence and the maturation state of the CD14−BDCA3+ LN-resident cDC subset in melanoma SLN (which also expresses CLEC9A, ref. 12) may be explained by its directive role in the immune cascade.
A disconnect between local and distant control of melanoma spread and growth has been reported in some cases. In a case study, Judge and colleagues reported on a patient presenting with local melanoma growth who nevertheless survived for more than a decade after palliative amputation of the local tumor masses, only to subsequently die of unrelated causes (39). Whereas the local tumors were poorly infiltrated, systemic NY-ESO-1–specific T-cell responses were detectable by IFNγ Elispot read-out between 2 and 6 years after amputation. This study suggests that local and systemic antitumor immunity may be differentially regulated, which is in line with our observations in relation to migratory versus LN-resident cDC subsets. Whereas LN-resident subsets appear to be vital for the generation of systemic immunity, providing protection against distant metastases, migratory subsets appear to direct local antitumor immunity, possibly through their ability to imprint and license primed effector T cells for selective homing to, and functional activity in, cutaneous compartments (40, 41).
Based on our observations, we propose that therapies aimed at local control should preferentially target skin-derived migratory DC subsets, whereas therapies aimed at distant control should target the LN-resident cDC subsets. In previous interventional phase II studies in early-stage melanoma patients, we showed that intradermal delivery of Granulocyte/Macrophage-colony stimulating factor (GM-CSF) resulted in the activation and increased migration of skin-derived DC subsets (42), whereas intradermal delivery of the TLR9 agonist CpG-B (CPG7909) selectively induced increased frequencies and activation of the LN-resident cDC subsets (12, 18). In line with their reported functional abilities (14), frequencies of in vivo recruited and matured LN-resident cDC correlated with increased ex vivo cross-presenting capacity of SLN suspensions (12). We have followed patients participating in these randomized and placebo controlled trials, and have found prolonged melanoma-specific survival and systemic tumor control in groups treated with CpG-B (43).
In conclusion, our insights into the immunologic events accompanying local and regional melanoma progression offer a rationale and suggest cellular targets for early immunotherapeutic interventions designed to prevent local or distant melanoma dissemination and recurrence.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: M.F.C.M. van den Hout, B.D. Koster, B.G. Molenkamp, A.J.M. van den Eertwegh, R.J. Scheper, P.A. van Leeuwen, M.P. van den Tol, T.D. de Gruijl
Development of methodology: M.F.C.M. van den Hout, B.G. Molenkamp, P.A. van Leeuwen, M.P. van den Tol, T.D. de Gruijl
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.F.C.M. van den Hout, B. Sluijter, B.G. Molenkamp, A.J.M. van den Eertwegh, P.A. van Leeuwen, M.P. van den Tol, T.D. de Gruijl
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.F.C.M. van den Hout, B.D. Koster, B.G. Molenkamp, R. van de Ven, A.J.M. van den Eertwegh, M.P. van den Tol, T.D. de Gruijl
Writing, review, and/or revision of the manuscript: M.F.C.M. van den Hout, B.D. Koster, B. Sluijter, B.G. Molenkamp, R. van de Ven, A.J.M. van den Eertwegh, M.P. van den Tol, T.D. de Gruijl
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.F.C.M. van den Hout, B.D. Koster, M.P. van den Tol, T.D. de Gruijl
Study supervision: A.J.M. van den Eertwegh, P.A. van Leeuwen, M.P. van den Tol, T.D. de Gruijl
This study was supported by Stichting Vivax (Vivax 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.