c.1810C>T Polymorphism of NTRK1 Gene is associated with reduced Survival in Neuroblastoma Patients

Background TrkA (encoded by NTRK1 gene), the high-affinity tyrosine kinase receptor for neurotrophins, is involved in neural crest cell differentiation. Its expression has been reported to be associated with a favourable prognosis in neuroblastoma. Therefore, the entire coding sequence of NTRK1 gene has been analysed in order to identify mutations and/or polymorphisms which may alter TrkA receptor expression. Methods DNA was extracted from neuroblastomas of 55 Polish and 114 Italian patients and from peripheral blood leukocytes of 158 healthy controls. Denaturing High-Performance Liquid Chromatography (DHPLC) and Single-Strand Conformation Polymorphism (SSCP) analysis were used to screen for sequence variants. Genetic changes were confirmed by direct sequencing and correlated with biological and clinical data. Results Three previously reported and nine new single nucleotide polymorphisms were detected. c.1810C>T polymorphism present in 8.7% of cases was found to be an independent marker of disease recurrence (OR = 13.3; p = 0.009) associated with lower survival rates (HR = 4.45 p = 0.041). c.1810C>T polymorphism's unfavourable prognostic value was most significant in patients under 18 months of age with no MYCN amplification (HR = 26; p = 0.008). In-silico analysis of the c.1810C>T polymorphism suggests that the substitution of the corresponding amino acid residue within the conservative region of the tyrosine kinase domain might theoretically interfere with the functioning of the TrkA protein. Conclusions NTRK1 c.1810C>T polymorphism appears to be a new independent prognostic factor of poor outcome in neuroblastoma, especially in children under 18 months of age with no MYCN amplification.


Background
Neuroblastoma (NB), a tumour of the peripheral nervous system arising from embryonal neural crest cells (NCCs), is the foremost malignant neoplasm of neonate and infant [1]. It demonstrates considerable diversity in clinical behaviour, ranging from spontaneous regression or maturation to rapid progression despite aggressive therapy [2]. Among molecular factors influencing the clinical outcome of neuroblastoma, the expression of TrkA has been associated with favourable prognosis [3].
TrkA is a member of a Trk family of tyrosine kinase receptor for nerve growth factor (NGF) and other neurotrophins. TrkA primarily regulates growth, differentiation and programmed cell death of neurons in both peripheral and central nervous systems [4]. Formation of the neuronal network during maturation of the peripheral nervous system is a complex process that is rigorously governed by the interactions of neurotrophins with their respective Trk receptors [5]. Interestingly, TrkA expression may be involved both in regulation of cell differentiation and in induction of programmed cell death of NCCs sympathoadrenal lineage. Consequently, alterations in the genes of the Trk family may result in oncogenic transformation of the neural crest derivatives [6]. In NBs, the amount of NGF in the tumour microenvironment and expression of TrkA receptor has a profound effect on cellular behaviour [7]. Tumour cells expressing TrkA undergo cell differentiation in the presence of NGF [8], while withdrawal of NGF induces apoptosis. Finally, as previously reported, TrkA expression is a crucial factor for spontaneous regression of neuroblastoma [5].
Human TrkA coding gene (named NTRK1) maps to chromosome 1q21-q22 and is composed of 17 exons [9,10]. Neither germ-line nor somatic gain-of-function mutations of the NTRK1 gene have been observed in any human neoplasia [11][12][13][14][15], except for one case of a de novo AML. That patient was found to carry NTRK1 S677N mutation that might interfere with the tyrosine kinase (TK) activity [11]. A small number of polymorphisms of unknown physiological significance has been reported in a few cases of prostate cancer, papillary thyroid carcinoma and NB [13][14][15]. Inborn inactivating mutations of NTRK1 gene are responsible for neurodevelopmental abnormalities known as CIPA (Congenital Insensitivity to Pain with Anhidrosis; MIM#256800), a rare recessive genetic disease [16]. Interestingly, the clinical profile of CIPA does not include predisposition to cancer.
Although many publications provide information on the role of TrkA expression in NB [17], only one pilot study analysed partially NTRK1 gene sequence in NBs [13]. Therefore, we performed mutational analysis of the entire coding sequence of the NTRK1 gene to identify somatic mutations and/or polymorphisms which might alter TrkA expression and to evaluate clinical and biological consequences of their incidence for the patients with NB.

Methods
Fifty-five NB patients treated in Polish paediatric oncologic centres and 114 NB patients retrieved from the Italian Neuroblastoma Group were enrolled into the study. Main clinical and biological characteristics of the patients, including age, stage of the disease, histology subtype in accordance with International Neuroblastoma Pathology Classification (INPC), MYCN amplification status and 1p36 chromosome deletion, are shown in Table 1. A control group of 158 healthy anonymous adult volunteers without history of hereditary cancer or other chronic disorders was included into the study on the basis of a detailed questionnaire and clinical examination. Written informed consent was obtained from the relevant guardians of the children and from patients and controls themselves, whenever eligible. The study was approved by the Ethics Committee of the Medical University of Gdańsk, Poland.
DNA was extracted from fresh-frozen and/or paraffinembedded tumour samples obtained during initial surgery procedures and from leukocytes of healthy controls. Tumour DNA was analysed because peripheral blood samples were not readily attainable from NB patients representing a number of oncologic centres from two countries. After digestion with proteinase K, DNA was purified by standard phenol-chloroform procedure [18]. All coding exons and flanking intronic (ca. 50 bp) sequences of NTRK1 gene were screened for mutations using two complementary techniques: SSCP (single-strand conformation polymorphism) analysis and DHPLC (denaturing high performance liquid chromatography). SSCP was performed in 12% polyacrylamide gels run at 200V for 2-4 hrs in 4°C followed by silver staining, as previously described [15]. DHPLC was performed on an automated WAVE System 3500 (Transgenomic Inc., San Jose, CA). Samples showing variation in SSCP banding pattern and/ or abnormal profile of elution by DHPLC as well as a few randomly selected negative samples were subject to sequencing. DNA sequencing reactions were performed using capillary-based ABI 3100 Genetic Analyser (Applied Biosystems Inc., Foster City, CA).
Selected bioinformatic tools were used to assess the effect of sequence variants on the structure and function of the receptor. Alignment analysis of 4606 sequences similar to TrkA retrieved from Pfam database [19] [22].
In all statistical analyses, allele frequencies were compared between case and control subjects within each population by chi-squared tests with adequate corrections, Fisher exact probability tests with Freeman-Halton extension for 2 × 3 tables and in the two populations combined by the Cochran-Mantel-Haenszel statistics. MAF (minor allele frequency) <5% was the criterion for exclusion from further statistical analysis. Unconditional logistic regression was used to examine the association between NTRK1 genotypes and clinical and biological characteristics of the disease. The dichotomous NB prognostic risk factors used as covariates are summarised in Table 1. The effect of the genotypes on long-time survival rates were analysed using Kaplan-Meier survival probability estimates, log-rank tests and Cox hazard proportional models. In the analysis of the two populations combined, an indicator variable for the study population (Poland versus Italy) was added to the models.
For all non-synonymous variants (i.e. c.482G>A, c.505G>A, c.1785G>A, c.1810C>T, c.1858_1859delinsTC) in silico structure-function analyses have been performed. c.482G>A codes the last residue of the third leucine-rich motifs region. It is the last residue of the α-helix and the replaced amino acid has similar physical characteristics, thus the probability of its influence on protein structure is rather low. Similarly, c.505G>A most likely does not affect the protein structure, since it encodes a substitution of a residue in a non-structuralised fragment of the protein. The remaining three non-synonymous SNPs encode residues of the TK kinase, among which residue 604 (coded by c.1810C>T) is the only one evolutionarily conserved. We found an experimentally determined structure of MuSK TK 1luf [23] to have the highest similarity with the TrkA TK domain. Sequence alignment revealed that all of the mutated TK residues lie in the kinase insert (TrkA residues 602-622) at the C-terminal lobe. This is a region of variable size and unknown function in many receptor TKs [23]. Homology modelling of TrkA TK domain, based on 1luf structure file, allowed us to visualise the localisation of His604 relative to the domain's functional regions ( Figure 2).  Figure 1 Sequence variants of the NTRK1 gene identified in the present study. The structure of TrkA receptor. Extracellular part of the protein consists of: two cysteine clusters (Cys); three leucine-rich motifs (LRM) and two immunoglobulin-like C2 domains (Ig-C2). It is followed by a short transmembrane portion (TM), and intracellular juxtamembrane domain (JM) and tyrosine kinase domain (TK). * -newly identified sequence variants, circled -the SNPs having MAF (minor allele frequency) value >5%, arrows -exonic localization of the variants, dishes -intronic localization of the variants.
Homology model of the TrkA TK domain Figure 2 Homology model of the TrkA TK domain. The model was obtained using rat MuSK kinase structure 1luf [23]. Activation loop (orange) and catalytic loop (blue) are indicated in accordance to de Pablo et al. [31]. The kinase insert fragment of kinase domain constitutes a loop joining two alpha-helices.
Most of its residues are missing in the crystal structure presumably due to the loop's high conformational flexibility and therefore not present in homology model (depicted as green beads). The His604 residue (red) is located at the end of structurally determined helical region, distant from ATPase active site (arrow). It may, however, be involved in interactions with residues of the unstructured loop region. Although the function of this loop is not yet known, it may be speculated that it is involved in specific for TrkA interactions with secondary messengers or regulatory proteins. Homology model was visualized with VMD program [32].
was associated with shorter survival rates ( Figure 3).

Discussion
The basic alteration in NB is developmental arrest of neuronal differentiation and unrestricted proliferation that most probably result from altered expression of genetic factors. Studies of familial cases of NB allowed for identification of two NB susceptibility genes (ALK, PHOX2B) yet their mutations are present in about 10% of sporadic tumours [24,25]. Inability to identify a single major specific tumour suppressor gene as a triggering agent of NB initiation [2] points toward the role played by genes involved in normal noradrenergic development. NTRK1 gene encoding TrkA neurotrophin receptor for NGF, is referred to as a candidate predisposition gene [5]. In this study, we investigated the genetic variability at the NTRK1 locus in human NBs.
Previously, only one study analysed sequence of NTRK1 gene in NBs [13] and that analysis was limited to the region coding TK domain (exons [13][14][15][16]. This and two other studies of the NTRK1 sequence: in prostate cancer [15] and medullary thyroid carcinoma [14], utilised SSCP as the sole screening method. However, current standards of molecular diagnostics do not recommend SSCP as a single screening procedure for detection of sequence variants because of its low sensitivity of 50-86%, while DHPLC is considered to be the method of choice because of its sensitivity exceeding 99% [26]. In our study, we used both methods. DHPLC vastly outperformed SSCP, allowing for detection of a notably larger number of sequence variants (11/12 versus 5/12 SNPs identified).
Although TrkA expression in NBs was reported significantly variable even at the cellular level within a single case [27], the genetic variability at the NTRK1 gene locus was found to be very limited. Non-synonymous polymorphisms were detected in less than 10% of the analysed cases. Furthermore, tag SNPs identified by HapMap project as well as SNPs present in NCBI dbSNP in vast majority are located in the intronic sections of the gene or do not result in amino acid substitution. Besides, according to the results of bioinformatic analyses, the identified non-synonymous polymorphisms, with the exception of c.1810C>T, have minimal theoretical chance of affecting the protein's structure and/or functions. These findings confirm the conservative character of NTRK1 gene sequence; however, it is possible that significant sequence variants are located elsewhere, for instance in currently unknown 3' flanking regulatory elements.
Even though c.1810C>T has MAF of 6%, from comprehensive statistical analysis of the associations with NB clinical and biological prognostic factors it appears to be an independent unfavourable risk factor in NB patients. c.1810T allele has been detected in 8.7% of sporadic tumours, which is a similar frequency to the mutations observed for the two known familial NB susceptibility genes (ALK = 12.4%, PHOX2b = 4.3%) [24,25]. Presence of c.1810T allele correlated with inferior 5-year OS and EFS and with more frequent relapse of the disease. From the analyses of the Cox hazard regression models, it appears that c.1810T provides additive information to MYCN amplification status and age of the patient, allowing for finest prediction of the eventual outcome of the disease. Interestingly, the stage of the disease was less informative and could be omitted in the model based on minimal set of predictive parameters. The lack of prognostic significance of stage is most probably due to strong associations with other prognostic variables -MYCN amplification and age. These associations were not observed for the NTRK1 c.1810C>T polymorphism.
The additive value of c.1810T allele to the current clinical prognostic markers should be, however, assessed on a prospective, more coherent, group of patients. In the current study, patients were enrolled in two countries over 20 years, thus they received a range of treatment protocols. Also, histopathological assessment was modified over the time and not all the samples could be assessed in accordance with the revised INPC. Interestingly, IT patients were diagnosed with lower stages of the disease and at a younger age, which is most probably related to better access to medical imaging techniques in Italy and thus higher incidence of incidentalomas diagnosed before overt clinical symptoms of advanced disease became apparent. The above-presented discrepancy in the patient population and the fact that the study was retrospective might, however, result in some limitation of the force of the study.
The mechanisms through which c.1810T variant could trigger and/or modify oncogenic events in embryonal NCC are only hypothetical at this time; however, our preliminary data provide an interesting starting point for further studies, possibly using a genome-wide association methodology.
Several points strongly support the significance of our finding; a) first, the evidence of the observed correlations was highly significant, both in univariate and multivariate analyses. Similarly, in the previous study of 63 NBs [13], there was a trend for the c.1810T to be present in cases having advanced stages of the disease; b) there were no major differences in allelic frequencies among two European populations that would explain these findings on the basis of population stratification. Linkage disequilibrium between c.1674G>A and c.1810C>T was present in both groups of patients, and also in four populations from HapMap database, thus the results obtained in both patient groups most probably refer to the same genetic effect; c) the bioinformatic analysis of the c.1810C>T showed that substitution of the corresponding amino acid (His→Tyr) affects a conservative residue of the TK domain. This substitution changes the chemical properties of a residue in conserved position, which may possibly alter the function of TK domain. From analysis of our homology model as well as from the results of in vitro studies by Mardy et al. [28], it appears that c.1810C>T does not inhibit the ability for autophosphorylation of TK domain. It may, however, be involved in interactions with the residues of the kinase insert loop. The exact role of this fragment of TK, specific for TrkA, remains unknown though it may be involved in signal transduction through interfering with the adaptor molecules.
Finally, a few methodological considerations regarding peculiarities of gene association studies performed on a paediatric oncologic experimental group should be pointed out. The relatively low incidence of neuroblastoma of 10.9 cases per million children yearly [29] affects the available size of patient sample group. In order to overcome this limitation, patients representing two European populations were enrolled in the study after initial validation of their correspondence using the Cochran-Mantel-Haenszel analysis. Another important methodological issue is related to the design of an adequate control group. If age-matched child controls had been used in the study, a screening for clinically inert tumour would have been required, however this might have created an ethical dilemma. Therefore, similarly to other studies in the field [30], healthy adult volunteers were chosen as controls, even if such a decision was related to some decrease in the force of the study. Moreover, even though the study was aimed at detection of constitutional mutations, since the available material collected by both national tissue banks consisted of primary NB tumours, tumour DNA was analysed. A similar approach is applied in studies of other researchers [24,25] because tumour biology is the area of primary interest from the clinical point of view. Besides, tumour samples are the routine material available for molecular assessment in research setting. Nevertheless, supplementary nested study of the corresponding germline DNA, derived from peripheral blood leukocytes of the cases where such material was available, was performed to provide confirmation of the constitutional character of the analysed SNPs.