CD99 polymorphisms significantly influence the probability to develop Ewing sarcoma in earlier age and patient disease progression

Ewing sarcoma (EWS), the second most common primary bone tumor in pediatric age, is known for its paucity of recurrent somatic abnormalities. Apart from the chimeric oncoprotein that derives from the fusion of EWS and FLI genes, recent genome-wide association studies have identified susceptibility variants near the EGR2 gene that regulate DNA binding of EWS-FLI. However, to induce transformation, EWS-FLI requires the presence of additional molecular events, including the expression of CD99, a cell surface molecule with critical relevance for the pathogenesis of EWS. High expression of CD99 is a common and distinctive feature of EWS cells, and it has largely been used for the differential diagnosis of the disease. The present study first links CD99 germline genetic variants to the susceptibility of EWS development and its progression. In particular, a panel of 25 single nucleotide polymorphisms has been genotyped in a case-control study. The CD99 rs311059 T variant was found to be significantly associated [P value = 0.0029; ORhet = 3.9 (95% CI 1.5-9.8) and ORhom = 5.3 (95% CI 1.2-23.7)] with EWS onset in patients less than 14 years old, while the CD99 rs312257-T was observed to be associated [P value = 0.0265; ORhet = 3.5 (95% CI 1.3-9.9)] with a reduced risk of relapse. Besides confirming the importance of CD99, our findings indicate that polymorphic variations in this gene may affect either development or progression of EWS, leading to further understanding of this cancer and development of better diagnostics/prognostics for children and adolescents with this devastating disease.


INTRODUCTION
Ewing's sarcoma (EWS) is a highly malignant musculoskeletal tumor that preferentially displays aggressive growth, with approximately 30% of patients harboring disseminated metastatic disease at the time of diagnosis, the most common sites being the lungs and/or bones [1]. EWS arises abruptly in pediatric age, with the peak of incidence in the second decade of life, and its natural history is still mostly unknown. The tumor is more common in Caucasians while it rarely appears in individuals of African or Asian heritage [2][3][4][5][6]. This observation together with reports indicating EWS in siblings or cousins [7,8], suggests that genetic susceptibility factors may exist for this tumor, particularly among European population. However, due to the rarity of the disease, only sporadic information is available. Genome-wide scan for EWS susceptibility loci identified two common variants associated with higher susceptibility to EWS in French population [9]: a variant mapping in 1p36. 22 and located proximal to the TARDBP gene (Tat activating regulatory DNA-binding protein, or TDPwww.impactjournals.com/oncotarget 43, transactive response DNA-binding protein) and a variant in 10q21, located in a 561kb LD block containing three described genes: ADO (encoding cysteamine dioxygenase), ZNF365 (encoding zinc-finger protein 365) and EGR2 (encoding early growth response protein 2). Mechanistic studies connecting expression of these gene variants to EWS susceptibility have clearly demonstrated a functional association between EGR2 and the oncogenic protein of EWS [10], the chimera EWS-FLI. EWS is indeed characterized by a chromosomal translocation between the EWS (locus 22q12) and ETS transcription factor genes, especially FLI1 (11q24) (occurring in 85% of cases), or ERG (21q22) (about 5-10% of cases) [11,12]. EWS-FLI gene fusion leads to the formation of a chimeric protein that drives the malignant phenotype by affecting both gene transcription [13][14][15] and RNA splicing [16,17] of specific downstream target genes. The crucial role of EWS-FLI in EWS pathogenesis has been confirmed also by the genomic landscape of EWS [18][19][20], which showed a very low somatic mutation rate in this neoplasia. Analysis of SNPs in the EWS gene revealed that the rs4820804 TT genotype could be associated with a higher propensity of EWS breakage, increasing the chances for the translocation to occur [21]. However, in recent years it has become clear that the EWS-FLI oncogenic activity is a necessary but insufficient condition. To induce transformation, EWS-FLI requires a permissive cellular background and the presence of additional molecular events, including disruption of p53 and CD99 expression [22,23]. CD99 is a cell surface molecule of 32 KDa [24] encoded by the pseudoautosomal MIC2 gene, which is involved in crucial biological processes like migration, cell death, transendothelial migration of leukocytes, differentiation of T cells and thymocytes and transport of surface molecules [25][26][27][28][29][30][31][32]. CD99 is constantly present at high levels in EWS cells [33,34] and its detection is routinely used for differential diagnosis. The EWS-FLI oncogenic activity [35] is facilitated by CD99 [23] and consistently, EWS-FLI maintains high levels of CD99 expression [23,36,37] either directly, through binding of CD99 promoter [23,37,38] or indirectly through miRNA regulation [39]. Abrogation of CD99 in EWS cells leads to terminal neural differentiation and severely reduces tumor growth and bone metastasis in mice [23], supporting a central role for CD99 in the pathogenesis of EWS.
In the present study, we examined the genetic influence of CD99 polymorphisms on EWS susceptibility in a representative sample of the Italian population. We were specifically interested to identify, in a case-control study, any potential genetic markers associated with age of diagnosis and to establish whether CD99 polymorphisms may influence the EWS disease progression. Analyses revealed for the first time evidence of two variant alleles in the CD99 gene: one (rs311059-T) strongly associated with earlier onset of EWS, and the other one (rs312257-T) related to patient event-free survival (EFS).

Impact of the CD99 SNPs on the risk to develop EWS
Genotyping was carried out on a cohort of 100 EWS patients (clinicopathological features are summarized in Table 1). The genotyping iPlex assay included 25 SNPs mapping on CD99 gene. Nine SNPs (rs311036, rs311092, rs311095, rs312199, rs313089, rs5982836, rs6567640, rs2267799, and rs311088) were excluded from statistical analyses because genotyping success rate was lower than 90%; the remaining 16 SNPs were anyhow adequate to detect association between CD99 and the risk of EWS. Deviation from Hardy-Weinberg equilibrium was observed for the SNPs rs5939307 and rs5939113 in both case-and control-group (P value < 0.001), although no apparent technical genotyping issues could be highlighted. In both circumstances the observed heterozygote frequency was lower than expected. Data from these SNPs were anyway tested for association, taking into account that caution is needed in case of deviation from the null hypothesis.
No evidence of association with alleles or genotypes was observed when EWS cases and controls were compared (Supplementary Table S1). Similarly, analysis carried out on haplotypes did not mark any association between EWS and CD99 SNP alleles (data not shown). However, when patients were sub-grouped by age (≤14 and >14 years old), a significantly higher frequency of the CD99 rs311059-T variant was observed in those showing an earlier onset of the disease (P value = 0.0029) [OR het = 3.9 (95% CI 1.5-9.8) and OR hom = 5.3 (95% CI 1.2-23.7)] ( Table 2). The association test was significant at the Bonferroni corrected threshold level for multiple testing (α < 0.0031). A similar trend was corroborated when the EWS patients with earlier onset of the disease were compared to the control group (Table 2).
In order to improve the coverage of exon 2 region, which contains the rs311059, four additional SNPs close to this polymorphism were selected and genotyped. No evidence of association was observed between these additional markers and the predisposition to EWS. However, haplotype analysis confirmed that the haplotype rs311059-T/rs311060-C showed the same level of association with the onset of EWS before 14 years of age [P = 0.0029; OR = 2.5 (95% CI 1.2-23.7)] that was previously found for the allele rs311059-T. Moreover, to verify that CD99 polymorphisms specifically influence the EWS onset in pediatric age, the genotyping was extended to patients with OS (clinicopathological features are summarized in Supplementary Table S2), which also frequently affects children and adolescents. No correlation was found between CD99 rs311059-T and the risk to develop OS (data not shown), further confirming that the association between this variant and higher probability of premature disease onset is a hallmark of EWS.

Impact of the CD99 SNPs on EWS disease progression and patient outcome
In the same cohort of EWS patients, we tested if CD99 polymorphisms could influence disease progression. The value of CD99 SNPs as prognostic biomarkers was calculated only in patients with localized EWS at diagnosis. Patients with metastatic disease at diagnosis were excluded from this analysis due to differences in treatment and the well-recognized negative impact of the presence of metastasis on patient outcome. The association analysis carried out on 78 patients with a minimum follow-up of 10 months (Table 3) showed that the variant rs312257-T is associated with a lower risk of EWS relapse [P = 0.0265; OR het = 3.5 (95% CI 1.3-9.9)]. Although this link was not sufficiently strong to be confirmed after Bonferroni correction for multiple testing, the prognostic value of the CD99 rs312257-T was proved to be associated with EFS of EWS patients when Kaplan-Meier survival curves were designed (Figure 1) (P = 0.03, log-rank test). Presence of rs312257 T variant allele as well as good response of tumors to neoadjuvant chemotherapy, which resulted significantly associated with clinical outcome in univariate analysis (Table 3), were confirmed as independent risk factors associated with good outcome by multivariate Cox's proportional hazards regression analysis (Table 4).

DISCUSSION
Although cancer is a complex disease resulting from the interplay of genetic, epigenetic, and environmental factors, the etiopathology of pediatric tumors supports the role of genetic and epigenetic alterations rather than environmental factors in the tumorigenesis. For EWS, race is one of the very few, well-recognized risk factor [40] and this supports the hypothesis of a genetic contribution to its etiology. However, familial cases of EWS are  [21,42]. Using a combination of high-throughput sequencing and integrative genomics, susceptibility variants near EGR2 gene have been identified [9]. More recently, Grünewald et al. [10] has demonstrated that EGR2 significantly influences EWS growth in mice and recognized an inherited variation on chromosome 10 that affects EWS susceptibility by facilitating the binding of the EWS-FLI oncoprotein to the EGR2 locus. Besides EWS-FLI, the other major common determinant of EWS is CD99 [23,43,44], which contributes substantially to EWS-FLI transformation.
In this paper, we have evaluated CD99 single nucleotide variations as potential risk factors for the development of EWS. We hypothesized that EWS occurring in pediatric age (i.e. ≤14 years) could be supported by a different genetic predisposition with respect to EWS occurring later in life. Indeed, we found   that the CD99 rs311059-T is significantly associated with an increased risk to develop the disease before 14 years of age. Rs311059 is a non-coding SNP and its functional relevance is unknown. As for the inherited variation on chromosome 10 near EGR2 [10], CD99 rs311059 variant allele may not affect the gene product sequence, but indirectly increasing susceptibility to EWS through epigenetic regulation of gene expression. Rs311059 maps inside the suppressor of zeste 12 (SUZ12) binding site, as shown on UCSC Genome Browser on Human Feb. 2009 (GRCh37/hg19) Assembly. SUZ12, a gene identified at the breakpoints of a recurrent chromosomal translocation reported in endometrial stromal sarcoma [45], encodes for a zinc finger protein belonging to the Polycomb Repressive Complex 2 (PRC2), which modulates chromatin structure by repressive mechanisms [46,47]. PRC2 specifically stimulates H3K27 trimethylation, which is associated with transcriptional repression, and can recruit DNMTs [48], thus acting as a platform for aberrant de novo promoter methylation. We speculate that the presence of the rs311059 T allele creates a mismatch in SUZ12 binding site, which might facilitate the expression of CD99, enabling the permissive condition for EWS-FLI transformation. This could explain why rs311059-T appears to be a susceptibility factor in the EWS onset in pediatric age. The significant level of rs311059-T presence in young patients was found to be a distinctive feature of this disease. In fact, in patients affected by OS, which shares with EWS both the skeletal localization and the peak onset in childhood and adolescence age, the genotyping test did not provide any evidence of association.
In EWS, CD99 is expressed at high levels in virtually all samples [34,49] but no correlation between CD99 expression and patient outcome has been highlighted so far. In this paper, we first identified the CD99 rs312257 T genetic variant as significantly associated with EFS. The presence of the T allele confers superior survival to patients with localized EWS. Although mechanistic studies are needed to explain this observation, our findings support the hypothesis that variations in CD99 gene may significantly affect the progression of EWS. Considering the rarity of the tumor, we offer this evidence to the scientific community for more extensive validation studies.

Clinical samples
A cohort of 100 unrelated Italian patients with confirmed diagnosis of localized (80 cases) or disseminated (20 cases) EWS treated at the Rizzoli Orthopaedic Institute (Bologna, Italy) was considered. All cases included showed a specific EWS-ETS fusion, EWS patients underwent local treatments (surgery; surgery plus radiotherapy; radiotherapy only, when the surgeon considered the lesion inoperable or due to patient refusal) and neo-adjuvant chemotherapy according to protocols that were previously reported in detail [50,51]. Clinicopathological data are shown in Table 1.
Patients with localized EWS were followed-up and clinical information updated (median follow-up: 76 months, range: 5-304 months). EFS was calculated from the date of initial diagnosis to clinical endpoint, which is considered as the time of occurrence of adverse events (defined as recurrence or metastases at any site for EFS). Histological response to chemotherapy was evaluated according to the method proposed by Picci et al. [52].
The control group consisted of 147 unrelated healthy volunteers with matching sex, ethnic origin and from the same geographical area (Italy). In addition, a cohort of 121 osteosarcoma (OS) samples was also enrolled for this study as a further comparison group. OS was chosen because it shares with EWS the same site of origin (bone) and a similar peak of incidence in juvenile age, despite being a very different type of tumor either from biological and genetic point of view [53]. Clinicopathological data of OS samples are shown in Supplementary Table S2. The ethical committee of the Rizzoli Institute approved the studies and informed consent was obtained.
Single nucleotide polymorphism genotyping DNA was extracted from peripheral blood leukocytes using standard DNAzol procedure (Thermo  Figure S1 shows the position of each of the 25 SNPs along the CD99 locus and haploblock structures [54]. Assay design and analysis of allele peaks were performed using specific Sequenom software (Sequenom, San Diego, California, USA). Primers were synthesized at Metabion (Martinsried, Germany). In order to improve the coverage of the gene, the genotyping of 4 additional SNPs selected by Haplotagger was carried out by using a high resolution melting curve (HRM) analysis approach.

Statistical analysis
The deviations from Hardy-Weinberg equilibrium for genotype distributions, in both patient and control groups, were examined using Pearson's χ 2 test.
Analysis for genotypic and haplotypic associations were performed using UNPHASED program (Version 3.1.7) which employs an allelic likelihood ratio test [55]. Odd ratios were calculated in order to estimate the level of association of the rare allele carriers, i.e. heterozygotes vs non-carriers, as well as homozygotes vs non carriers. The results were adjusted for multiple testing with the Bonferroni correction according to the number of tag SNPs analyzed. Bonferroni correction is considered an overly stringent adjustment when tests are not independent, such as in case of linkage disequilibrium between SNPs.
Patients were then sub-grouped by age at diagnosis into 2 groups (≤14 and >14 years old) and association analysis was performed. Comparisons as follows: 1) ≤14 vs >14 years old; 2) ≤14 years old vs control group; 3) >14 years old vs control group. The association was also tested for 4 additional SNPs (rs1136447, rs311060, rs6567640, and rs311074) selected by Haplotagger, covering the exon 2 and the boundaries. These four SNPs together with rs311057 and rs311059, mapping near to the 5' exon 2 boundary were considered for haplotype analysis (until six marker combinations).
To verify association between CD99 SNPs and patient prognosis, a subgroup of 78 patients with localized EWS having a minimum follow-up of 10 months was considered. For this subgroup, Kaplan-Meier and log-rank tests were used to draw and evaluate the significance of survival curves in EWS patients in relation to CD99 SNP alleles.