FCGR Polymorphisms Influence Response to IL2 in Metastatic Renal Cell Carcinoma

This article is featured in Highlights of This Issue, p. 2129

Patients with metastatic renal cell carcinoma (mRCC) show a 14% to 22% response to the standard high-dose regimen of aldesleukin (IL2). On the basis of retrospective analyses of patients with mRCC that had been treated with IL2, potential biomarkers that may be useful for response predictions were assessed in the "SELECT" clinical trial conducted by the Cytokine Working Group. Patients with mRCC were entered into the “SELECT” clinical trial of high-dose IL2 (HD-IL2) to prospectively determine whether certain clinical and pathologic criteria [the type of mRCC tumor (clear cell vs. non–clear cell), carbonic anhydrase 9 (CA-9) tumor staining and CA-9 polymorphism status, programmed death ligand 1 (PD-L1) tumor staining, B7-H3 tumor staining, as well as other criteria] are associated with response to IL2. As noted in the clinical report, correlations were observed between the immune cell modulatory ligand, PD-L1, and immune response, and the study produced a response rate of 25% (1). In an effort to further identify genetic markers that might associate with efficacy of the HD-IL2 treatment for patients with mRCC, and potentially identify immunologic mechanisms involved in the response, we sought to identify genotypic factors that may influence the immune activity of HD-IL2 therapy. In this study, we genotyped SNPs found in certain activating Fc-gamma receptor (FCGR) genes (FCGR2A, FCGR3A, and FCGR2C).

Variably expressed on immune cells, FCGRs bind the Fc fragment of IgG antibodies (2–4). Upon engagement and cross-linking, activating FCGRs transmit signaling within the immune cell and initiate immune activation (5–7). FCGR2A (expressed on dendritic cells, macrophages, monocytes, neutrophils, and eosinophils), FCGR3A [expressed on natural killer (NK) cells and macrophages], and FCGR2C (also expressed on NK cells) are all activating FCGRs (4, 8). The SNPs found in both FCGR2A and FCGR3A genes convey differential binding affinities for the Fc portion of antibody. The FCGR2A SNP encodes amino acids of either histidine (H) or arginine (R) at position 131 of the FCGR2A protein (FCGR2A-H131R, rs1801274), and the FCGR3A SNP encodes either valine (V) or phenylaline (F) at amino acid 158 of FCGR3A (FCGR3A-V158R, rs396991; refs. 9–12). The FCGR2A-H and FCGR3A-V receptors each have higher binding affinities to human IgG than do the FCGR2A-R and FCGR3A-F receptors, respectively (2, 4, 11). This stronger binding affinity results in more potent in vitro antibody-dependent cell-mediated cytotoxicity (ADCC) and tumor cell death (13, 14). In some clinical trials involving various chimeric or humanized mAbs specific for head and neck, colorectal, or B-cell malignancies, both FCGR2A-H and FCGR3A-V SNPs are associated with improved clinical response (13–16). Similarly, in a trial of an idiotypic vaccine for B-cell lymphoma, designed to induce endogenous anti-idiotypic antibody, better outcome was seen for patients with the higher-affinity FCGR2A-H and FCGR3A-V SNPs (16). Alternatively, other studies have found no association of FCGR2A-H/R or FCGR3A-V/F SNP genotype with patient response to immunotherapy (17, 18).

The FCGR2C gene has an SNP in exon 3 (c.169 C<T, rs759550223) that influences the expression of FCGR2C on NK cell surfaces (19–21). The presence of a “C” nucleotide in this SNP leads to an open reading frame, enabling the expression of the FCGR2C receptor. In contrast, a “T” nucleotide creates a stop codon, resulting in the lack of expression, for that allele (19, 20, 22). A minority of individuals (20%–40%) have the “C” allele (either FCGR2C-C/C or C/T genotype) and, thus, have FCGR2C expressed on their NK cells (22–25). When expressed, FCGR2C is capable of inducing ADCC after receptor cross-linking (22, 23, 25). Although the SNP of FCGR2C genotype has been correlated with patient response to immunomodulatory therapy for autoimmune-based diseases (23, 26–30), little has been published regarding the role of FCGR2C expression in cancer immunotherapy.

In this study of patients with mRCC who received HD-IL2, we looked for associations of patient FCGR2A, FCGR3A, and FCGR2C genotypes with clinical outcome. We found that higher-affinity FCGR genotypes resulted in improved tumor shrinkage and overall survival (OS). These findings suggest a potential role for the cells expressing these FCGRs in the clinical response of patients with mRCC to HD-IL2 therapy.

A total of 106 patients from the SELECT trial had DNA available for genotyping, along with clinical data for correlative analyses. DNA was isolated from peripheral blood mononuclear cells following the manufacturer's protocol of the DNeasy Blood and Tissue Kit (Qiagen). DNA was kept at 4°C during the time of analyses, and later was transferred to −80°C for long-term storage after completion of the analyses.

All SNP genotyping was performed on a StepOnePlus quantitative PCR machine (ABI/Life Technologies). The FCGR2A SNP was determined using TaqMan primer/probes available from ABI/Life Technologies and used as per the manufacturer's protocol. For both FCGR3A and FCGR2C, Rnase H primers and probes for each gene were developed in our laboratory to allow for specific amplification of each gene. For genotyping the FCGR3A SNP, Rnase H primers were developed to specifically amplify FCGR3A while not coamplifying FCGR3B. These primers were paired with specific probes to determine the SNP (31). For genotyping the FCGR2C-C/T SNP, Rnase H primers were developed to specifically amplify this gene while not coamplifying FCGR2B. Primers and probes for both FCGR3A and FCGR2C were designed through Integrated DNA Technologies (IDTDNA). Specific method details can be found in Erbe AK and colleagues, 2016 (31). Genotyping was conducted in a blinded manner, where those individuals that determined the genotype of the patients did not have access to the clinical outcome data. Genotypes were verified for Hardy–Weinberg equilibrium agreement (32); specific genotype results can be found in Supplementary Table S1.

The clinical results of the SELECT trial have been published (1). As noted in the clinical report, patients received 600,000 IU/kg/dose of IL2 (Prometheus Laboratories Inc., San Diego, CA) IV every 8 hours for 5 days for a maximum of 14 doses (1). Clinical data for percent tumor shrinkage, progression-free survival (PFS), and response rates [complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD)] and overall survival (OS) were obtained from the clinical dataset (33, 34). Data were updated through October 31, 2013. The clinical characteristics of the patient subset we analyzed (106 patients of the 120 patients in the original trial) are similar to the clinical characteristics of those observed in the original study (Table 1). For our analysis of percent tumor shrinkage and disease control rate [(DCR) = CR + PR + SD vs. PD], two patients did not have percent tumor shrinkage clinical data available, and two patients did not have DCR data, and thus were excluded from the percent tumor shrinkage and DCR analyses (n = 104 for percent tumor shrinkage and n = 104 for DCR). As noted in the original clinical report, all patients who achieved CR, PR, or SD for more than 6 months had their CT scans audited by independent radiologists to confirm their response and response duration (1). The protocol was approved by the human Investigational Review Board at each participating site, and voluntary written informed consent was obtained from each patient.

The clinical outcomes assessed included percent tumor shrinkage and OS. The percent tumor shrinkage was defined as the percent change in tumor size from baseline to maximum shrinkage. OS was defined as the time in months from the date of treatment initiation to the date of death or was censored at the date of last contact with the patient. PFS was defined as the time in months from the date of treatment initiation until the date that disease progression criteria were met or the date of death without progression, or was censored at the date of the last disease assessment without evidence of progression. For evaluation of response, patients with a CR or PR were classified as responders and those with PD or SD were classified as nonresponders; correlations of genotype with overall response included genotype parameters with CR/PR versus SD/PD, and correlations of genotype with DCR included genotype parameters with CR/PR/SD versus PD. The association between percent tumor shrinkage and genotyping predictors was evaluated using two-sample t tests. The Kaplan–Meier method was used for estimation of the survival distribution for OS. For the survival plots, the tick marks along each line indicate patients censored; each drop of the line indicates a clinical event (i.e., patient death). Log-rank tests were used to assess the association between genotyping predictors and OS and PFS. The association between binary response and the genotyping predictors was evaluated using logistic regression. The association between FCGR genotypes was assessed using Fisher exact test. Changes in tumor size were represented using box plots, which show the 25th percentile (Q1; bottom of box), the 50th percentile (Q2; bolded black line), the 75th percentile (Q3; top of box), and the mean (red cross inside the box). The lower and upper short horizontal red lines represent the minimum and maximum values, excluding the outlying high and low values. Outlying values [i.e., those that are a distance of more than 1.5 × (Q3−Q1) from the box] are shown as circles outside the horizontal lines. For these analyses in this study, no adjustments in reported P values were made for multiplicity of testing.

In these patients with mRCC treated with HD-IL2, we found that individuals homozygous for the high-affinity FCGR3A-V/V allele had significantly prolonged OS compared with those having only one or no copy of this high-affinity allele (FCGR3A-V/V: 73.4 months vs. FCGR3A-V/F or F/F: 40.6 months; P = 0.03, Table 2). In addition, although not significant, the percent tumor shrinkage for FCGR3A-V/V patients was greater than that for FCGR3A-V/F or F/F patients, (FCGR3A-V/V: 32.8% vs. FCGR3A-V/F or F/F: 9.4%; P = 0.21, Table 2). As SD is considered an indicator of clinical benefit for mRCC patients treated with HD-IL2 (33, 35), we looked for associations with DCR by comparing the different FCGR genotypes with clinical outcome grouped as CR/PR/SD versus PD. The probability of being classified as CR/PR/SD was more likely if FCGR3A-V/V as compared with FCGR3A-V/F or F/F, although this difference was not significant (P = 0.20, Table 2). For FCGR2A, those with the higher-affinity genotype compared with low-affinity receptors (H/H vs. H/R or R/R, respectively) had increased tumor shrinkage, although these differences were not significant (FCGR2A-H/H: 25.8% vs. FCGR2A-H/R or R/R: 7.1%; P = 0.18, Table 2). There was no difference in OS based on FCGR2A genotype. Finally, there was a trend for increased likelihood of being classified as CR/PR/SD if FCGR2A-H/H as compared with FCGR2A-H/R or R/R (P = 0.07, Table 2). Patients that express the FCGR2C receptor [those with at least one copy of the C allele (C/C or C/T)] had prolonged OS as compared with those that did not express FCGR2C on their NK cell surface (those with an FCGR2C genotype of T/T); however, these differences were not significant (FCGR2C-C/C or C/T: 73.3 months vs. FCGR2C-T/T: 40.6 months; P = 0.12, Table 2). There were no differences in either percent tumor shrinkage or DCR based on FCGR2C genotype (Table 2). In addition, there were no significant associations of any of the individual FCGR polymorphisms with PFS (FCGR2A H/H vs. H/R or R/R: P = 0.98, FCGR3A VV vs. V/V or V/F: P = 0.62 and FCGR2C C/C or C/T vs. T/T: P = 0.92, Supplementary Table S2).

Upon antibody recognition and binding, there may be cross-talk between cells that express FCGR2A and cells that express FCGR3A that can influence NK-cell response (36). Prior studies of patients treated with mAb have reported associations between FCGR SNPs and clinical outcome when both genotypes of FCGR2A and FCGR3A were combined for the analyses (37, 38). We compared individuals that were homozygous for either the H allele of higher-affinity FCGR2A or for the V allele of higher-affinity FCGR3A (group 1 in Fig. 1A) with individuals that were not homozygous for either the higher-affinity allele of FCGR2A or FCGR3A (group 2 in Fig. 1A). We found significantly improved tumor shrinkage in group 1 versus group 2 (P < 0.05, Fig. 1B,). In addition, group 1 also showed prolonged OS versus group 2, but this was not significant (P = 0.17, Fig. 1C). Finally, there was a significant association of having an improved DCR for patients with a more favorable FCGR3A/2A genotype (group 1 vs. group 2: P = 0.03, Fig. 1D). For PFS, there were no significant differences for FCGR3A/2A polymorphisms when comparing group 1 versus group 2 (P = 0.40, Supplementary Table S2).

NK-cell ADCC capabilities can be enhanced if FCGR2C is expressed on the cell surface (24). As NK cells can express both FCGR3A and FCGR2C, we considered whether patient outcome was influenced by the combined genotypes for FCGR3A and FCGR2C. Patients that have two copies of the high-affinity FCGR3A allele (V/V) or one copy of the high-affinity FCGR3A allele (V/F) and at least one copy of FCGR2C (C/C or C/T), or two copies of FCGR2C (C/C) are identified as group 3 (boxes I, II, III, IV, V, and VII) in Fig. 2A. All other patients are identified as group 4 and include those with genotypes that have only one copy of the high-affinity FCGR3A allele (V/F) and have no copy of FCGR2C (T/T), and those that have no copy of the high-affinity FCGR3A allele (F/F) and have only one or no copy of FCGR2C (C/T or T/T; boxes VI, VIII, and VIIII in Fig. 2A). Although there was no difference between group 3 and group 4 for percent tumor shrinkage (Fig. 2B), group 3 showed significantly prolonged OS compared with group 4 (P = 0.01, Fig. 2C). However, there was no association of DCR with FCGR3A/2C group 3 versus group 4 (P = 0.95, Fig. 2D), nor was there an association with PFS for FCGR3A/2C polymorphisms when comparing group 3 versus group 4 (P = 0.78; Supplementary Table S2).

On the basis of our findings that patients with FCGR2A and FCGR3A genotypes in homozygous form resulted in prolonged OS (although not statistically significant) and significantly improved percent tumor shrinkage (Fig. 1), as well as our finding that high-affinity FCGR3A-V in combination with the expression of FCGR2C resulted in significant improvement in the length of OS (Fig. 2), we further assessed the combined influence of all three of these FCGR genotypes on patient response. To simultaneously consider the genotype combinations for all three FCGRs studied here, we categorized patients into “favorable” and “unfavorable” groups (Fig. 3A) based on the genotypic patterns presented in Figs. 1A and 2A. The favorable group (shaded in Fig. 3A) included all patients homozygous for FCGR3A V/V or FCGR2A H/H, as well as patients with at least two higher-affinity alleles of FCGR3A or FCGR2A (at least one copy of FCGR3A-V and at least one copy of FCGR2A-H) with FCGR2C expression (C/C or C/T), namely V/F-H/R patients, if they also expressed FCGR2C (C/C or C/T). This corresponded to 42 favorable-genotype patients. The remaining 64 patients (unshaded in Fig. 3A) are designated as unfavorable genotype.

Patients in the favorable FCGR genotype group had a significantly improved percent tumor shrinkage as compared with those with “unfavorable” FCGR genotype (28.5% vs. 1.7%; P = 0.03, Fig. 3B). Patients in the favorable group also showed a trend toward improved OS (Fig. 3C; 56.0 vs. 37.4 months for favorable vs. unfavorable groups; P = 0.07). There was a near trend for association with either CR/PR/SD response if patients had an FCGR2A/3A/2C favorable genotype versus an unfavorable genotype (P = 0.14, Fig. 3D). Those with a favorable FCGR2A/3A/2C genotype had PFS of 5.5 months as compared with the PFS of 2.6 months for those with an unfavorable FCGR2A/3A/2C genotype; however, this difference was not significant (P = 0.30, Supplementary Table S2).

The waterfall plot of percent tumor shrinkage (Fig. 4) demonstrates that those patients in the FCGR2A/3A/2C favorable group (red bars) are more prominent on the right (tumor shrinkage) side of the graph. Conversely, those in the FCGR2A/3A/2C unfavorable genotype group (blue bars) are more prominent on the left (tumor growth) side of the graph. Gillespie (2012) suggests that waterfall plot visualization can be useful for attempts to personalize treatments for individual patients (34, 39).

Although both the genotypes of the SNPs on FCGR2A and FCGR3A have been implicated in some analyses of the clinical antitumor response to tumor-reactive mAb immunotherapy, we believe this is the first study to show a potential association of favorable FCGR genotype with clinical outcome in the antitumor use of single-agent HD-IL2, without mAb administration. Moreover, FCGR2C expression based on SNP status has not yet been shown to influence clinical response to immunotherapeutics in cancer patients, in particular in patients not treated with mAb.

The data presented in Fig. 3B–D show a significant association with percent tumor shrinkage and a trend with OS when simultaneously considering genotypes for all three of these loci (FCGR2A, 3A, and 2C). The finding that there are associations of FCGR genotype with the clinical outcome parameters of both tumor shrinkage and OS appears to involve all three of these FCGR genes. This is consistent with data in Table 2, showing a significant role for FCGR3A in OS, as well as trends in improved percent tumor shrinkage for FCGR2A, and improved OS for FCGR2C (although not statistically significant).

Responses to immunotherapy in patients with mRCC are most often observed in patients with mRCC who have clear cell histology (40). In this study, of the 106 patients with mRCC reported and analyzed for FCGR genotypes (n = 104 for % tumor shrinakge, n = 106 for OS), 100 had clear cell mRCC (n = 98 for % tumor shrinkage, n = 100 for OS), and six had non–clear cell mRCC. None of the patients with mRCC with non–clear cell histology responded to the HD-IL2 treatment (1). Therefore, we also assessed the role of FCGR genotypes within those patients that have clear cell mRCC (excluding those with non–clear cell mRCC; Supplementary Table S3). Within this clear cell subset, there was a significant association in the DCR for FCGR2A H/H versus H/R or R/R (P = 0.05; Supplementary Table S3). Of note, when comparing FCGR2A/3A/2C “favorable” versus “unfavorable” genotypes with the duration of OS within these patients with clear cell mRCC, there was a statistically significant difference where those with favorable FCGR2A/3A/2C polymorphisms had both improved percent tumor shrinkage (P = 0.03) and prolonged OS (P = 0.04) as compared with those with unfavorable FCGR2A/3A/2C polymorphisms as well as a trend for improved DCR (P = 0.10, Supplementary Table S3).

Although we found significant associations of FCGR genotype with all of these clinical parameters (% tumor shrinkage, OS, and DCR), we did not see significant associations of patient FCGR2A, 3A, and 2C SNP genotype with PFS, nor did we see differences when comparing FCGR genotype with patient overall response rate when considering CR/PR versus SD/PD (data not shown). The comparison of CR/PR versus SD/PD categorically divides patients based on percent tumor shrinkage data; this type of evaluation is a binary comparison of “responders” (complete or partial responders) versus “nonresponders” (stable or progressive disease), and it does not include those with SD as “responders.” Hughes and colleagues (2015) found that patients with melanoma or stage IV RCC that had SD had significantly longer OS than patients with PD, and they suggest that those patients with SD should be considered as having clinical benefit. By including patients with SD with those with CR/PR, we did find associations of DCR with outcome, some of them significant, for FCGR genotypes. Moreover, the waterfall analysis (Fig. 4), which scores each patient based on their maximum amount of percent tumor shrinkage, is based on quantitative measures, and it is considered by some to be more sensitive compared with the overall response rate (34, 39). This may account for the genotypic associations with percent tumor shrinkage, but not with overall response status, found in this study.

Such associations of favorable FCGR genotypes and clinical outcome with HD-IL2 treatment do not prove a causal link. McDermott and colleagues, 2015, reported that in the original cohort of patients treated with HD-IL2, in addition to the HD-IL2, 80 patients also received VEGF-targeted therapy. This additional VEGF-targeted therapy may have contributed to the OS length found in those individuals that were treated with it. Beyond the additional treatment measures (i.e., VEGF-targeted treatment) that may have influenced clinical response differences, genotypes that show an association with a clinical condition may do so because of their linkage disequilibrium to nearby loci that were not directly genotyped, yet influence the clinical associations seen. The FCGR genes are located in close proximity to each other on chromosome 1q23, with FCGR2A located upstream [with genomic coordinates (GRCh38): 1:161,505,41-161,524,048] of FCGR3A [with genomic coordinates (GRCh38): 1:161,541,759-161,550,623] followed by FCGR2C [with genomic coordinates (GRCh38): 1:161,581,339-161,601,220] (41). Using the genotype data for this population (Supplementary Table S1), we found a trend toward linkage disequilibrium between FCGR3A and FCGR2A (P = 0.08), a significant disequilibrium between FCGR3A and FCGR2C (P < 0.01), and no significant disequilibrium between FCGR2A and FCGR2C (P = 0.70; data not shown). This linkage disequilibrium involving these three genes could contribute to the favorable FCGR genotype grouping found in this study, as shown in Fig. 3. Furthermore, although unlikely, these favorable FCGR gene alleles that are associated with better outcome in this study could potentially be in linkage disequilibrium with a favorable allele for some separate (non-FCGR) gene that might actually be responsible for the improved outcome we observe in association with the favorable FCGR genotype. The fact that some of the associations that we have observed are significant while others are trends suggests that the effect of the favorable FCGR genes is one of the several factors involved in the antitumor activity of HD-IL2 in some patients (but not others) with mRCC.

This association of outcome with favorable FCGRs suggests that greater functionality of FCGRs, due to higher affinity (for FCGR2A and FCGR3A) and expression of FCGR2C, may be playing a role in at least part of the antitumor activity of HD-IL2. Our current understanding of these FCGRs is that they function primarily through engaging antigen-bound IgG, transmitting an activating signal, and inducing cellular responses, such as the induction of ADCC (by NK cells, neutrophils, and monocytes/macrophages) or antibody-dependent cellular phagocytosis (ADCP), and the uptake of antigens by antigen-presenting cells through immobilized, bound IgG molecules, resulting in antigen processing and presentation (3, 4, 6, 24, 42). In each of these settings, an antigen-reactive antibody (either endogenous or passively administered) is needed for antibody/FCGR-facilitated ADCC, ADCP, or antigen processing. The data presented here, showing that HD-IL2–treated mRCC patients with more functional FCGR genotypes showed increased tumor shrinkage and prolonged OS compared with those with less functional FCGR genotypes, support the hypothesis that some of these patients may have formed endogenous antibodies, reactive with their autochthonous mRCC that were capable of mediating ADCC, ADCP, and/or antigen presentation. The in vivo antitumor activity of such endogenous antitumor antibodies would potentially be enhanced by the presence of more favorable FCGRs.

In 1955, Graham JB and Graham RM suggested that some gynecologic oncology patients formed endogenous antibodies recognizing autologous tumor antigens, but these endogenous antibodies did not recognize the tumor antigens derived from tumors of similar histology from other patients (43). Since that time, several endogenous antibodies that are reactive against well-described and conserved shared tumor antigens have been identified, including antigens on RCC (44–46). For example, Knutson and colleagues (2016) recently showed that for patients with HER2⁺ breast cancer, a combination therapy that included chemotherapy together with trastuzumab (mAb against HER2) induced, in 69% of patients, endogenous IgG-antibodies directed against HER2 and a subset of endogenous shared tumor-associated antigens; this endogenous antibody response was associated with improved disease outcome (47). However, for most tumor types, methods to readily demonstrate and quantify the presence and functional activity of endogenous antibodies against the unique neoantigens present on patients' autochthonous tumors, for the full cohort of patients enrolled in a trial such as this one, remain elusive. Thus, in this retrospective study, with no access to patient sera or to patient tumor tissue, we have not attempted to evaluate the presence or functionality of endogenous antibody to autochthonous tumor.

The interplay of several immune cell types, through engagement of their FCGRs via antibody-bound antigen recognition, creates the potential for a successful immunotherapeutic response following treatment with mAb (48). On the basis of the associations reported here, of more functional FCGRs being associated with improved outcome with HD-IL2 therapy, we hypothesize the following immunologic pathways may be involved. First, the presence of preexisting endogenous tumor-reactive IgG antibodies might enable IL2 to induce augmented ADCC and ADCP, which would be enhanced by the presence of more functional FCGR alleles through cross-talk of NK cells (expressing FCGR3A and potentially FCGR2C) and monocytes (expressing FCGR2A; ref. 36). Alternatively, the preexisting antitumor antibodies might facilitate tumor antigen presentation and induction of an adaptive (dendritic cell, T cell, and potentially B cell) response, which could be augmented by IL2 treatment and more functional FCGR. Finally, in some patients, more than one of these mechanisms could be at work simultaneously. The FCGR genotype combinations identified here have the potential to serve as biomarkers to personalize immunotherapeutics for cancer treatment (49). Future studies validating this association of favorable FCGR genotype with outcome, as well as prospective efforts to evaluate sera from all treated patients for functional antibody reactive to tumor (particularly autochthonous tumor), will be needed to test these hypotheses and determine whether they lead to actionable clinical modifications in this approach toward immunotherapy.

No potential conflicts of interest were disclosed.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NCI.

Conception and design: A.K. Erbe, W. Wang, J.A. Hank, M. Atkins, P.M. Sondel

Development of methodology: A.K. Erbe, M. Atkins, P.M. Sondel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.K. Erbe, J. Goldberg, M. Gallenberger, Y. Song, M. Atkins, A. Carlson, J.W. Mier, D.J. Panka, D.F. McDermott, P.M. Sondel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.K. Erbe, J. Goldberg, K. Kim, L. Carmichael, D. Hess, E.A. Mendonca, S.-C. Cheng, D.F. McDermott, P.M. Sondel

Writing, review, and/or revision of the manuscript: A.K. Erbe, W. Wang, J. Goldberg, M. Gallenberger, K. Kim, L. Carmichael, E.A. Mendonca, J.A. Hank, S.-C. Cheng, S. Signoretti, M. Atkins, D.J. Panka, D.F. McDermott, P.M. Sondel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.K. Erbe, K. Kim, J.A. Hank, S. Signoretti, A. Carlson, D.J. Panka, P.M. Sondel

Study supervision: A.K. Erbe, P.M. Sondel

This research was supported by The Institute for Clinical and Translational Research, Hyundai Hope on Wheels Grant, Midwest Athletes Against Childhood Cancer, Stand Up 2 Cancer, The St. Baldrick's Foundation, American Association of Cancer Research, University of Wisconsin-Madison Carbone Cancer Center; by NCI-Cytokine Working Group, and supported in part by Public Health Service grants CA014520, CA021115, CA180820, CA180794, CA180799, CA14958, CA166105, and CA197078, from the NCI, the NIH, and the Department of Health and Human Services.

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.