HLA-G+3027 polymorphism is associated with tumor relapse in pediatric Hodgkin's lymphoma

In this study, we tested whether polymorphisms in human leukocyte antigen G (HLA-G) were associated with event-free survival (EFS) in pediatric Hodgkin's lymphoma (HL). We evaluated the association of HLA-G 3′-UTR polymorphisms with EFS in 113 pediatric HL patients treated using the AIEOP LH-2004 protocol. Patients with the +3027-C/A genotype (rs17179101, UTR-7 haplotype) showed lower EFS than those with the +3027-C/C genotype (HR= 3.23, 95%CI: 0.99-10.54, P=0.012). Female patients and systemic B symptomatic patients with the HLA-G +3027 polymorphism showed lower EFS. Multivariate analysis showed that the +3027-A polymorphism (HR 3.17, 95%CI 1.16-8.66, P=0.025) was an independent prognostic factor. Immunohistochemical analysis showed that HL cells from patients with the +3027-C/A genotype did not express HLA-G. Moreover, HLA-G +3027 polymorphism improved EFS prediction when added to the algorithm for therapeutic group classification of pediatric HL patients. Our findings suggest HLA-G +3027 polymorphism is a prognostic marker in pediatric HL patients undergoing treatment according to LH-2004 protocol.


INTRODUCTION
Hodgkin lymphoma (HL) originates mostly from transformed B-lymphocytes and represents 6% of the childhood tumors [1]. Mononuclear Hodgkin (H) and multinucleated Reed-Sternberg (HRS) tumor cells represent 0.1%-2% of the total tumor infiltrating cells [2]. A substantial number of HL cases are linked to Epstein-Barr virus (EBV) [3]. A follow-up clinical study of the Italian Association of Pediatric Hematology-Oncology (AIEOP) trial LH-2004 for pediatric HL showed that 108 out of 769 (14%) enrolled children showed disease progression or relapse (37 progression and 71 relapses), in general 2 years after diagnosis [4]. Therefore, it is important to identify markers that predict relapse in HL patients in order to obtain better responses with less aggressive therapy, particularly considering the long term adverse effects of HL therapy such as secondary cancer, infertility and neurological toxicity [5]. Genetic instability is also associated with HL. Therefore, it is key to identify genetic factors that are involved in HL pathogenesis and response to therapy [6][7].
Class I histocompatibility leukocyte antigen (HLA) molecules play an important role in the ability of tumor cells to evade immune surveillance [8]. Moreover, natural killer cells (NK) that kill cancer cells expressing HLA class I, are locally inhibited in HL [9][10]. Therefore, antigenpresentation by the HLA complex and co-stimulatory molecule expression plays a key role in the etiology of HL, including familial aggregation and the epidemiology of HL [11][12][13][14][15]. However, it is unclear if HLA genes are associated with HL due to high degree of polymorphism, peculiar genomic distribution and high level of linkage disequilibrium (LD) among the HLA genes [16].
HLA-G is induced during transplantation [25,26], viral infection [27] and malignancies [28,29] and its expression correlates with the size and aggressiveness of solid tumor [30,31]. HLA-G is also expressed by immune cells in the tumor microenvironment by trogocitosis, thereby restricting their local proliferation and cytotoxicity [32]. Furthermore, its expression is regulated by several factors such as IL-10, IFN-γ, heat shock, hypoxia, oxidative stress and radiation, which further impacts immune response towards tumor and tumor-infiltrating cells [33]. HLA-G expression and its clinical significance has been extensively investigated in solid tumors, but its role in hematological malignancies has not been well established [34]. Many differences have been reported between solid and hematological cancers. These differences are probably because immune cells, unlike epithelial cells, express both HLA-G and HLA-G receptors (ILT2) that inhibit proliferation of B-cell lymphomas, myelomas, and B-cell leukemia [35]. HLA-G expression in classic HL patients has been investigated in 2 studies both regarding adult patients. These authors showed about 50% of positive-tumor cells for HLA-G expression and suggested that induction of HLA-G protein contributed to the modulation of immune responses observed in HL since HLA-G expression correlated with a better prognosis [36,37]. Thus, in hematological diseases, the role of HLA-G is more complex than in solid tumors and depends on the balance between inhibition of anti-tumor responses and the anti-proliferative effects on malignant B cells [28,35]. Since it is important to understand the significance of HLA-G expression in hematological malignancies, we investigated the association of HLA-G 3'-UTR polymorphisms with event-free survival (EFS) in pediatric HL patients.

Patient and disease characteristics
Patient and disease characteristics are listed in Table 1. The HL patients were sub-divided into classical HL subtypes such as nodular sclerosis (cHL-NS n=86), mixed cellularity (cHL-MC, n=7), nodular lymphocytepredominant (HL-LPn, n=9) and unspecified (HL, n=11) according to the WHO guidelines [38]. The median follow-up period was 3.5 y. Based on Ann Arbor staging, patients were subdivided into stages I (n=7), II (n= 54), III (n=22) and IV (n=30). Twenty-seven patients (24%) showed relapse or tumor progression during the follow-up and two patients (1.7%) developed secondary tumor.
The following co-variables were tested: +3010, +3027, +3142 polymorphisms, gender, age, histological WHO classification and AIEOP LH-2004 classification for treatment response. The +3027 A-variant or the UTR-7 haplotype encompasses the 14bp region that spans the Ins, +3003T, +3010C, +3027A, +3035T, +3142G, +3187A and +3196C polymorphic sites. Only significant results obtained in multivariate analysis were reported. www.impactjournals.com/oncotarget (BD; n=259) and HL patients (n=113). In the control subjects, there were no differences in the allele and gene frequencies based on age or gender for any of the eight polymorphisms (data not shown). However, subgroup analysis of HL patients by (i) gender showed that male gender was associated with the +3187-AA genotype, (ii) the +3010-CC+CG; +3027-AA+CA and +3142-GG+CG genotypes were associated with increased risk for EFS, (iii) the +3196-CC genotype correlated with higher Ann Arbor stages (Table 2). We observed that the +3010-C, +3027-A and +3142-G alleles were part of the unique 3'-UTR-7 haplotype of HLA-G encompassing the +3027-A variant [24]. Therefore, we chose the +3027 polymorphism rather than UTR-7 haplotype for further analysis.
We found that 27 (23.9%) of the 113 patients that received AIEOP LH-2004 treatment experienced a tumorevent in 1.2±1.048 y (range 0.4-6 y) after diagnosis. Kaplan Meier survival analysis demonstrated that the patients carrying the heterozygous +3027-C/A genotype showed lower EFS time than patients carrying the C/C genotype (HR=3.23, 95%CI: 0.99-10.54; P=0.012; Figure  1). Only 1 out of 113 patients carried the A/A genotype and were excluded from the analysis.

Inclusion of HLA-G +3027 polymorphism enhances EFS risk estimation by the clinical therapeutic group (GR)
According to the AIEOP LH-2004 criteria, patients were stratified into 3 therapeutic groups (GR) based on increased radiation dosage and aggressive chemotherapy (both length and intensity) to reduce long-term toxicities in younger patients [4]. The groups according to AIEOP LH-2004 risk score criteria were (1) GR1: stage IA or IIA  without mediastinal bulky disease or involvement of lung hilum lymph nodes, and less than four lymph nodal regions; (2) GR2: negative for GR1 or GR3 criteria; (3) GR3: stage IIIB or stage IV or mediastinal bulky disease. Therapy was performed based on interim or post-chemotherapy positron emission tomography (PET) analysis by quantitative Fluorodeoxyglucose (FDG) avidity of tumor mass.
In the present study, we included the HLA-G +3027 risk factor in the algorithm with the AIEOP LH-2004 GR score to predict the EFS risk ( Figure 3A). We performed Kaplan-Meier analysis of this new algorithm and compared it to AIEOP LH-2004 GR score alone ( Figure 3B). Since the GR1 and GR2 groups included less than 2 patients with +3027 C/A variant, we merged the GR1, GR2 and GR3 groups with the +3027 C/A variant as a unique GR123-C/A group. We stratified patients in the 4 categories with lower EFS by the presence of +3027-C/A genotype as shown in the Kaplan-Meier analysis ( Figure 3). Based on the +3027 genetic algorithm, 8 out of 15 patients (53%) with the +3027-C/A genotype, and 18 out of the 68 patients (26%) in the GR3 category showed lower EFS (≤2years). We compared the hazard ratio (HR) values obtained for both AIEOP LH-2004 GR and HLA-G +3027 algorithms and found better risk stratification with the HLA-G +3027 algorithm (Figure 4).

Overall survival estimates based on the HLA-G +3027 algorithm
Cox regression analysis showed that patients in the GR123-C/A and GR3-C/C groups were associated with decreased overall survival and lower EFS score according to the HLA-G +3027 algorithm ( Figure 5). Nine of the 113 patients that died belonged to the GR123-C/A and GR3-C/C groups.

Analysis of HLA-G Expression in malignant cells
Immunohistochemical analysis of HLA-G expression in tumor cells from 25 lymph node biopsies that were obtained at diagnosis showed that patients carrying a +3027 C/A genotype were all negative for HLA-G and only 5 cases having the wild type +3027 C/C genotype showed HLA-G expression (20%; Figure 6).

DISCUSSION
In this study, we showed that the +3027-A (UTR-7 haplotype) HLA-G polymorphism was associated with a low EFS in pediatric HL patients treated according to the AIEOP LH-2004 protocol, more evident in female gender and patients with systemic B symptoms. Multivariate analysis demonstrated that +3027-C/A genotype was an independent prognostic factor associated with poor EFS. Moreover, incorporating the HLA-G +3027 polymorphism into the AIEOP LH-2004 therapeutic stratification improved risk evaluation. Patients belonging to the GR123-C/A and the GR3 groups showed lower 5 yearoverall survival than patients belonging to GR1 and GR2 groups (9.7% vs. 0%). The effect of HLA-G expression on HL patient survival needs further investigation.
Many studies have reported association of polymorphisms in the HLA region with HL [39,40]. However, the association of variants in the 3'UTR of HLA-G with clinical outcome in HL has not been reported so far. Moreover, mechanisms that regulate HLA-G expression in B-cells are not well defined [37,41,42]. We demonstrated that HLA-G was not expressed in the tumor cells of HL patients with +3027-C/A genotype and present only in a minority of cases (20%) with the wild type +3027-C/C genotype. Previously, HLA-G expression was reported in adult HL patients belonging to Ann Arbor stages IIb-IV in the tumor microenvironment in patients showing positron emission tomography (PET)-positive (carried out after 2 cycles of standard chemotherapy) during follow-up [37]. Conversely, PET-negative patients showed higher HLA-G protein expression levels on tumor cells (50% of cases) suggesting that HLA-G protein expression inhibited tumor progression [37]. We demonstrated that HLA-G was not expressed in the lymph node of HL patients carrying the HLA-G +3027 C/A genotype. However, HLA-G was expressed in few cases (n=5) with +3027 C/C genotype and therefore, the role of the wild type HLA-G C/C genotype needs to be studied in greater detail. Caocci et al. showed that HLA-G expression was a positive prognostic factor in patients with advanced-stage classical HL [37]. In the present study, we found HLA-G-positive tumor cells in lower number of cases (n=5/25, 20%) than reported by Diesptra HLA-G expression can exert beneficial or detrimental effects of immune recognition, which depend on the pathological context or situation [34,43]. HLA-G expression suppresses immune response in case of allotransplantation and autoimmune disorders [44]. However, HLA-G expression would be detrimental in solid tumors and virally infected cells, where it suppresses immune response by inhibiting NK cells that kill tumor cells [45]. Correlation between HLA-G expression and clinical outcome in hematological malignancies is unclear [46]. One probable reason is that contrary to epithelial cells, immune cells express both HLA-G and its inhibitory receptors such as ILT-2 on their surface. In multiple myeloma and B-cell leukemia, interaction with HLA-G and ILT2 inhibitory receptor induces tumor cell apoptosis [35]. Moreover, since the HLA-G +3027-A variant was found associated with lower HLA-G expression [41], our data might suggest that patients with the +3027-C/A genotype are more prone to tumor relapse because more prone to decreased expression in HLA-G, which affects HL tumor cell apoptosis.
Of note, the impact of chemotherapy and radiotherapy on HLA-G expression in patients carrying the HLA-G +3027-C/C and the HLA-G +3027-C/A genotypes is not known. Previous reports have shown that gamma radiation and/or chemotherapy modulates surface HLA-G expression in B-cell lines (M8 and K562), melanoma cell lines (OCM-1A and JEG-3) and basal cell carcinoma of the skin [47,48]. Thus, it is possible that chemo/radiotherapy modulate HLA-G expression also in our patients having a good prognosis, but not in patients having the HLA-G +3027 C/A genotype for a reason still now unknown. However, it is also possible that +3027-C/A polymorphism could be in strong linkage disequilibrium with another critical gene, which causes the disease progression and poor EFS.
Overall these issues needs to be investigated in detail.
In conclusion, we demonstrated that +3027-A HLA-G polymorphism was associated with poor EFS in pediatric HL patients treated with the AIEOP LH-2004 protocol. In the future, comprehensive prospective studies are necessary to confirm that HLA-G polymorphisms are prognostic indicators in pediatric HL therapy.

Sample collection and Patient grouping criteria
Peripheral blood samples were collected from 113 pediatric HL patients (range 3-18 years) at the AIEOP Centers in the laboratory of Padua University from December 2004 to August 2014. Table 1 lists patient information regarding age, sex, the presence of a secondary tumor, relapse and/or progression tumor events, presence or absence of systemic symptoms and Hodgkin's lymphoma subtype as recommended by World Health Organization (WHO) [38]. The patients were also staged according to Ann Arbor classification [48] and allocated into 3 therapeutic groups (GR) based on risk [4]. These 3 groups were (1) GR1: stage IA or IIA without mediastinal bulky disease or involvement of lung hilum lymph nodes, and with less than four lymph nodal regions; (2) GR2: absence of criteria for GR1 or GR3; and (3) GR3: stage IIIB or stage IV or mediastinal bulky disease. The patients were treated according to the protocol approved for the AIEOP multicenter clinical trial LH-2004 [4].

Treatment protocol and follow-up
The GR1 patients (low risk) were administered 3 cycles of ABVD (Adriamycin, bleomycin, vinblastine, dacarbazine) without radiotherapy (RT) were administrated. In case of unsatisfactory response after 2 cycles of ABVD or residual disease after 3 cycles, GR1 patients received chemotherapy/RT or RT. The GR2 (intermediate risk) and GR3 (high risk) patients were administered 4 or 6 alternating cycles of COPP-ABV chemotherapy (cylcophosphamide, vincristine, procarbazine, prednisone and ABV) to ABVD treatment. Radiotherapy was administered with Involved Field Technique (IFRT) at cumulative dose of 14.4 or 25.2 Gy depending on the remission status (complete or partial) at the end of chemotherapy. Patients in complete remission at the end of chemotherapy were not administered radiotherapy. Complete response (CR) was defined as absence of clinical, radiological (ultrasound and/or CT scan evaluation) and FDG-PET evidence of disease. We enrolled 259 local blood donors (BD) as a control group. Study was approved by the ethics committee of all participating institutions. Written consent forms were obtained from parents or legal guardians of all patients.

Immunohistochemistry
Paraffin-embedded primary HL lymph node tissue was retrieved. HLA-G expression was assessed in 25 paraffin-embedded tissues by Immunohistochemical staining with primary 4H84 mouse monoclonal antibody (Abcam, HIER, 1:2000), which recognizes all the HLA-G isoforms. If intense membrane HLA-G staining was found in more than 10% of Hodgkin or Reed-Sternberg cells, then the sample was scored as positive.

Statistical analysis
The frequencies of HLA allele and genotype were compared between patients and controls chi-square test (VassarStats http://faculty.vassar.edu/lowry/VassarStats. html and the SKDM software (http://sourceforge.net/ projects/skdm). The HLA-G 3'UTR haplotypes were defined according to Castelli [50]. HLA haplotype frequencies were estimated by the ARLEQUIN ver 3.1 software [51]. Frequencies and distributions of clinical characteristics were calculated as the period of time after diagnosis. Kaplan-Meier survival analysis was performed to determine relapse rate and overall survival after diagnosis and treatment according to the LH-2004 protocol. We included age, gender, Ann Arbor stage, tumor response, presence of systemic symptom and WHO histopathological classification as potential co-variables in the univariate model. Multivariate Cox proportional hazards models were used to estimate hazard ratios (HRs) for EFS with 95% CIs adjusted for +3027 polymorphism and potential co-variables. The most frequent variable was automatically selected as the reference category. Clinical prognostic risk groups [4] currently used to stratify therapy were used to compare the +3027 HLA-G genetic stratification.