Mutational spectrum of ATRX aberrations in neuroblastoma and associated patient and tumor characteristics.

Neuroblastoma is the most common extracranial solid tumor in children. The chromatin remodeler ATRX is frequently mutated in high-risk patients with a poor prognosis. Although many studies have reported ATRX aberrations and the associated clinical characteristics in neuroblastoma, a comprehensive overview is currently lacking. In this study, we extensively characterize the mutational spectrum of ATRX aberrations in neuroblastoma tumors reported in previous studies and present an overview of patient and tumor characteristics. We collected the data of a total of 127 neuroblastoma patients and three cell lines with ATRX aberrations originating from 20 papers. We subdivide the ATRX aberrations into nonsense, missense, and multiexon deletions (MEDs) and show that 68% of them are MEDs. Of these MEDs, 75% are predicted to be in-frame. Furthermore, we identify a missense mutational hotspot region in the helicase domain. We also confirm that all three ATRX mutation types are more often identified in patients diagnosed at an older age, but still approximately 40% of the patients are aged 5 years or younger at diagnosis. Surprisingly, we found that 11q deletions are enriched in neuroblastomas with ATRX deletions compared to a reference cohort, but not in neuroblastomas with ATRX point mutations. Taken together, our data emphasizes a distinct ATRX mutation spectrum in neuroblastoma, which should be considered when studying molecular phenotypes and therapeutic strategies.

alternative lengthening of telomeres

B‐cell acute lymphoblastic leukemia

in‐frame deletion

in‐frame fusion

multiexon deletion

MYC‐N amplification

segmental chromosomal aberration

X chromosomal inactivation

INTRODUCTION

Neuroblastoma is the most common extracranial solid tumor in children and has an incidence of 10.2 cases per million under the age of 15 years. , It arises from progenitor cells during fetal adrenal neuroblast development. Clinical, genetic, and pathologic characteristics are used for risk stratification and treatment allocation. Although the 5‐year overall survival rate for patients has improved over recent years to 81%, for patients with high‐risk disease survival remains 50% despite multimodal treatment regimens.

High‐risk tumors can be subclassified in four categories based on genomic aberrations: MYCN amplifications (37%), TERT rearrangements (23%), ATRX deletions (11%), and having none of the previously mentioned aberrations. , All these genetic aberrations are independently associated with a poor clinical outcome for patients. Both MNA and TERT rearrangements induce elevated expression of TERT and thereby facilitate telomere maintenance, one of the hallmarks of cancer. In contrast, ATRX aberrations can induce ALT, which is a homologous recombination‐based telomere maintenance mechanism that occurs in 5–15% of cancer. The MNA/TERT rearrangements and ATRX aberrations are mutually exclusive, whereas MNA and TERT rearrangements sporadically co‐occur.

ATRX, located on the X‐chromosome, is a gene that encodes for an ATP‐dependent helicase of the SWI/SNF family of chromatin remodelers. The coding sequence, excluding the introns, comprises 7479 base pairs and consists of 35 exons. ATRX forms a chromatin remodeling complex with DAXX, which performs the replication‐independent deposition of the histone variant H3.3 at pericentromeric and telomeric regions. , In addition, ATRX is both directly and indirectly involved in a plethora of nuclear processes, such as removing G‐quadruplexes, inhibiting macroH2A1 deposition, removing R‐loops, modifying histones, methylating DNA, and resolving replication fork stalling.

This wide spectrum of ATRX functions might be reflected by the large shift in the expression of hundreds of genes that has been observed in cell lines derived from neuroblastoma patients with large MEDs resulting in IFDs of the ATRX gene. These IFDs do not lead to an absence of protein, but rather result in a shorter protein product. In neuroblastoma, both ATRX MEDs and point mutations have been reported, and these aberrations are strongly associated with ALT development , , , , , , , , , , , , , , , , , , , , . ATRX aberrations are also present in various other pediatric , , , , , , , , , , , , , , , , , , , , , , , and adult cancers.

Although several studies investigating genomic alterations in neuroblastoma have reported ATRX aberrations, the number of patients was often limited and a comprehensive analysis of the spectrum of ATRX mutations and patient characteristics is currently lacking. Therefore, in this meta‐analysis we extensively characterized the mutational spectrum of ATRX aberrations in neuroblastoma tumors described in previously published works and compared these with other pediatric cancers. Furthermore, we present an overview of the patient characteristics and co‐occurring aberrations in this neuroblastoma subgroup.

MATERIALS AND METHODS

Data acquisition and analysis

To obtain all reported pediatric cancers with known genetic ATRX aberrations, we carried out a thorough search in several databases. The last search was undertaken on April 30, 2020 and all identified papers were manually assessed for patients with known genetic ATRX aberrations. Twenty‐one tumors with ATRX deletions described by Brunner et al. and Abbasi et al. were reanalyzed for this study for breakpoints and deleted exons. From all identified papers with neuroblastoma patients with ATRX aberrations, we collected the specific genetic mutation and a selection of the tumor/patient data (Table S1). The ATRX aberrations were reported in our paper based on the human reference genome Hg19 and transcript variant 1 (NM_000489.6). As multiple published papers can use the same dataset, we carefully assessed the origin of each patient and in that manner removed duplicates. To create the overviews of the ATRX point mutations and MEDs we used the genomic positions for plotting. For the analysis of the predicted effect of the missense mutation we performed PROVEAN protein batch analysis and for every analysis on the patient and tumor data all available input data were used (Table S1) and missing data were removed per analysis. All the data were analyzed and plotted using R 4.0.2 and ggplot2_3.3.5. For further details see Document S1.

RESULTS

ATRX aberrations in neuroblastoma and other pediatric cancers

After collecting all the data on ATRX aberrations from neuroblastoma reports, we identified 127 unique patients and three cell lines from a total of 20 papers and belonging to 16 cohorts. Apart from two tumors, that were stage 2 and 3, all tumors were classified as stage 4. For comparison we also included the ATRX aberrations identified in other pediatric cancers of 122 tumors and one cell line (Table S1). These data originated from 23 studies and comprised nine different tumor types, mainly osteosarcoma, high‐grade glioma, or B‐ALL (Figure 1A). Tumors with ATRX aberrations can be divided in three subgroups based on their mutation type: nonsense/frameshift mutations, missense mutations, or MEDs. We defined MEDs as deletions involving at least two complete exons; all reported smaller deletions induce frameshifts and were therefore included within the nonsense/frameshift category. We observed that 68% of the neuroblastoma samples contained MEDs, whereas the MEDs only comprised 8% in other pediatric cancers, with the majority belonging to osteosarcomas (Figure 1B and Table S1). This indicated that the MEDs are almost exclusively present in neuroblastoma. Furthermore, these data showed that point mutations were less frequent in neuroblastoma, only constituting 32% of the cases, while making up to 92% of the cases in other pediatric cancers (Figure 1B). Overall, this indicates that ATRX aberrations in neuroblastoma have a different distribution of mutation types, with a high frequency of MEDs.

FIGURE 1

Patient and mutation type distribution in our compiled dataset. (A) Distribution in percentages of all patients within our dataset across the different pediatric cancer types. ACT, adrenocortical tumor; AML, acute myeloid leukemia; B‐ALL, B‐cell acute lymphoblastic leukemia; ETMR, embryonic tumor with multilayered rosettes; HGG, high‐grade glioma; LGG, low‐grade glioma; MB, medulloblastoma; NB, neuroblastoma; OS, osteosarcoma; T‐ALL, T‐cell acute lymphoblastic leukemia. *≤1%. (B) Distribution in percentages of the three ATRX mutation types in neuroblastoma (left) and all other pediatric cancers (right)

Identification of a mutational hotspot region in the helicase domain for ATRX missense aberrations in neuroblastoma

To further analyze the ATRX mutational spectrum in neuroblastoma and other pediatric cancers, we determined the frequency and distribution of nonsense/frameshift mutations as well as missense mutations across the gene.

For neuroblastoma, 23 nonsense/frameshift ATRX mutations were reported, whereas 59 cases were described for all other pediatric cancers (Table S1). These aberrations were randomly distributed across the ATRX gene, which was also the case for the other pediatric cancers (Figure 2A). Recurrent nonsense/frameshift ATRX mutations were very rare and none of the recurrent mutations occurred in more than two patients. The only exception was R2386* (Table S1), which was present in three osteosarcoma patients, including two siblings with a germline mutation.

FIGURE 2

Pediatric point mutation landscape within the ATRX gene. (A) Distribution of nonsense and frameshift mutations of all pediatric cancers within our dataset across the ATRX gene. (B) Distribution of missense mutations of all pediatric cancers within our dataset across the ATRX gene. In both panels, neuroblastoma mutations are displayed above the bar and below the bar mutations that were identified in other pediatric cancers are displayed. All mutations were plotted based on DNA positions. ACT, adrenocortical tumor; AML, acute myeloid leukemia; B‐ALL, B‐cell acute lymphoblastic leukemia; ETMR, embryonic tumor with multilayered rosettes; HGG, high‐grade glioma; LGG, low‐grade glioma; MB, medulloblastoma; NB, neuroblastoma; OS, osteosarcoma; T‐ALL, T‐cell acute lymphoblastic leukemia

In addition, ATRX missense mutations were identified in 17 neuroblastoma tumors and one cell line and 54 other pediatric tumors (Table S1). When these mutations were mapped against the ATRX gene, we observed a mutational hotspot region for neuroblastoma in the helicase domain (56% of missense mutations), especially the C‐terminal helicase domain where nearly 45% of all reported missense mutations were located (Figure 2B). Missense mutations were significantly enriched in the helicase domain for neuroblastoma (binomial test, p = 8.633e‐06). Furthermore, 38% and 36% of all ATRX missense mutations for high‐graded glioma and osteosarcoma, respectively, were also restricted to the helicase domains. In summary, this strongly supports the presence of a mutational hotspot region within the ATRX helicase domain across pediatric cancers.

To date no complete 3D structure of the ATRX protein exists, therefore we were unable to assess structural consequences of the missense mutations and were limited to mutation prediction software to indicate potential biological effects. Of the 18 missense mutations identified in neuroblastoma tumors, 13 were predicted to have a deleterious effect by both the SIFT and PROVEAN analyses (Table S2) and two were reported as neutral by both algorithms. Overall, we can conclude that ATRX missense mutations are predominantly present in the helicase domain and that they are predicted to disturb protein function, whereas nonsense mutations are randomly distributed across the gene and could result in absence of ATRX protein production.

Majority of recurrent ATRX MEDs in neuroblastoma are predicted in‐frame

ATRX MEDs are observed in both neuroblastoma and osteosarcoma. Therefore, we undertook a detailed analysis of the ATRX MEDs identified in 77 neuroblastoma tumors and two cell lines and six osteosarcoma tumors and one cell line. For neuroblastoma, 29 unique MEDs were identified (Figures 3A and S1 and Table S1). Most commonly a deletion of exon 2–10 (28 of 79, 35.4%, MED#1) was observed, which was predicted to be in‐frame (called exon 1–11 IFF in Qadeer et al.) and to produce a mutant protein product. The deletion of exon 2–9 also occurred frequently (11 tumors and two cell lines, MED#2). Although this deletion is predicted to be out‐of‐frame, on the mRNA level alternative splicing with skipping of exon 10 was identified, thus resulting in the same mutant protein product as for a deletion of exon 2–106. The third most common deletion consists of exons 3–9 (2–10 IFF, five tumors and one cell line, MED#3), followed by several deletions that only occurred in two patients. The remaining 19 distinct deletions occurred only in a single case (Figure S1A). We also identified one MED that precisely covered the entire ATRX gene (Table S1, MED#29). We never observed a tumor with the same MED occurring in homozygosity. Of all deletions, 75% (59 of 79) was shown or predicted to be in‐frame and therefore could result in mutant protein production in most of the tumors. The minimal deleted region, the part that is lost in most cases, was exon 8 and 9 (Figure 3B). Upon close examination of all deletions, two distinct subgroups can be identified, namely the ones missing exon 8–9 and cases with smaller deletions that included loss of exon 11–12 (Figures 3A,B, S1, and S2). Interestingly, this is in sharp contrast to the missense mutations that are enriched in the helicase domain, while the deletions mostly involve the N‐terminal domains. This suggests a whole other type of functional consequence on the molecular level, as different protein‐interacting domains are affected and therefore could lead to different impairments of ATRX’s functions.

FIGURE 3

Mutational landscape of recurrent multiexon deletions (MEDs) in neuroblastoma. (A) Overview of all recurrent MEDs in our dataset that occur in more than one patient. Numbers (#1–#9) indicate the unique MED identifier, and the number before every deletion indicates the number of patients with that specific aberration. Symbols after every deletion indicate whether the deletion is predicted to be in‐frame (Y) or not (N). Y* is predicted to be in‐frame because of exon skipping of exon 10. (B) Exon landscape plot showing the presence of each exon in all neuroblastoma MEDs in our dataset, including the deletions displayed in Figure S1

Subsequently, we also wanted to examine deletions in other pediatric cancers and found that large deletions only occurred in osteosarcoma and B‐ALL (Table S1). However, in B‐ALL there were three out‐of‐frame MEDs, all different from the MEDs in neuroblastoma. In osteosarcoma there were seven reported MEDs, of which two were also reported in neuroblastoma, namely the predicted out‐of‐frame deletion of exon 2–15 (MED#6) and the predicted IFD of 2–13 (MED#5), of which the latter was the only recurrent aberration for osteosarcoma. The remaining four deletions were all predicted to be out‐of‐frame. In summary, we determined that MEDs were mostly observed in neuroblastoma and rarely occur in other pediatric cancers. Additionally, we found that 75% of ATRX deletions in neuroblastoma are predicted to be in‐frame and that there are two patterns of deletions, namely those lacking exons 8–9 or exons 11–12.

Correlations of ATRX mutation subtypes with clinical characteristics

There were several reports that showed that patients with neuroblastoma carrying ATRX aberrations are diagnosed at an older age. , , We wanted to investigate this for all three mutation subtypes separately. As a reference cohort for this analysis we used the cohort of Pugh et al. This cohort consisted of 240 stage 4 tumors, of which 24 had an ATRX aberration. After removing these samples, which were included in our ATRX subtypes, we had a reference cohort of 216 tumors. We took this cohort as it consisted of only stage 4 tumors, which was also almost the case for the ATRX tumors collected from published reports. Additionally, it was one of the very few studies reporting many of the SCAs observed in neuroblastoma. As none of the collected patients with ATRX aberrations belonged to a single cohort, the use of this reference cohort might introduce a systematic bias. We observed that there was a difference in the distribution of the age at diagnosis between the reference cohort and all three ATRX mutation subtypes (Mann–Whitney U‐test, deletion p = 2.6e‐13, nonsense p = 7.7e‐05 and missense p = 0.00053) (Figure 4A). Interestingly, this difference was found for all three subtypes, even for the patients with missense aberrations for which we had relatively few data points.

FIGURE 4

Correlations between ATRX mutation subtypes with clinical characteristics in neuroblastoma. (A) Violin plot showing the distribution of age at diagnosis (years) for each of the mutation types. Red diamonds show medians: ATRX WT, 3.18 years; Deletion, 5 years; Nonsense/frameshift, 5.28 years; Missense, 11.7 years; p values (Mann–Whitney U‐test) are displayed. (B) Percentage of patients belonging to each age category for every mutation type. Fisher’s exact tests were carried out for each age category (18–60 months, 5–12 years, and >12 years) to compare the mutation types. Both p values and adjusted p values can be found in Table S3. (C) Distribution of gender in percentages within the ATRX aberrant subtypes. (D) Kaplan–Meier curves of event‐free survival (left) and overall survival (right) of the ATRX aberrant subtypes

To confirm the different age distribution, we separated the patients with and without ATRX aberrations into three categories (18–60 months, 5–12 years, and >12 years) according to their age at diagnosis (Figure 4B; no patients were younger than 18 months of age in this cohort). We chose a cut‐off of 5 years as 90% of neuroblastoma patients are younger than 5 years old at diagnosis, and a cut‐off of 12 years because of the start of puberty. Eighty‐three percent of patients in the reference cohort (ATRX WT) were 5 years or younger, whereas less than 42% of patients were in this age group for the three ATRX mutation types. Compared to the reference cohort, the ATRX subtypes had a significantly different contribution of each age category (Table S3), except for the 5–12 years age category, where there was no significant difference between the reference cohort and the missense and nonsense subtype (Fisher’s exact test, p adj = 0.88 and p adj = 0.12, respectively). This could, however, be explained by the relatively low number of patients for these subtypes. Overall, this indicates that all three ATRX subtypes occurred more frequently in patients diagnosed at an older age, but still two‐fifth of the patients were diagnosed below 5 years of age.

The ATRX gene is located on the X chromosome and therefore ATRX aberrations in neuroblastoma might affect male individuals more often than female individuals. To investigate this, we compared the male : female ratio in all three mutation types to the ratio reported in a large neuroblastoma cohort, which is 1.1:1. Interestingly, a higher male : female ratio than expected, with a ratio of 1.77:1 (Figure 4C), was detected in patients with ATRX MEDs (binominal test, p = 0.040). In contrast, the male : female ratio was not found to be skewed when ATRX was affected by missense or nonsense mutations (Figure 4C; binominal tests; p = 0.79 and p = 0.64, respectively). Our reference cohort had a relatively high male : female ratio of 1.6:1. However, male : female ratios of 1.2:1, 1.3:1, and 1.65:1  have also been reported in neuroblastoma cohorts and when we used these ratios the deletion subgroup was no longer significantly different. However, the ratio of 1.1:1 that we used for our analysis was derived from a large cohort across different continents and therefore seems more reliable. Thus, our analysis suggests that ATRX deletions, but not other ATRX aberrations, might occur more often in male than female patients. Nevertheless, more thorough investigation is necessary in the future.

Patients with ATRX aberrant neuroblastoma are known to have poor event‐free and overall survival. , Here we assessed the event‐free and overall survival for each mutation type separately (Figure 4D). For event‐free survival, there was no difference between the three groups (Gehan–Breslow–Wilcoxon test, p = 0.098). However, a significantly better overall survival (Gehan–Breslow–Wilcoxon test, p = 0.022) of patients with missense mutations was observed as compared to the other two mutation types. Thus, patients with ATRX missense mutations might have a slightly better prognosis.

Neuroblastomas with ATRX deletions more often have 11q deletion and/or 2p gain

Segmental chromosomal aberrations are common in neuroblastoma and several of these recurrent aberrations are associated with a poor prognosis, such as 17q gain, 11q deletion, 2p gain, and 1p deletion. To investigate the co‐occurrence of these SCAs in the three subgroups of ATRX aberrant tumors and one cell line (CHLA‐90), we plotted their frequencies compared to the reference cohort. This comparison showed no significant differences in the proportion of tumors with 17q gain, or 1p or 3p deletion (Figure 5 and Table S4). In contrast, 11q deletion (p adj = 1.2694e‐10) and 2p gain (p adj = 0.021) more often occurred in tumors with ATRX deletions compared to the other two ATRX mutation types or the reference cohort (Figure 5). This was most striking for the 11q deletion, as almost 90% of the patients with an ATRX deletion carried this SCA. 11q deletion and MNA are almost mutually exclusive, and the reference cohort contains approximately 36% MNA tumors; in contrast, MNA is rarely reported for ATRX aberrant tumors. Therefore, the observed enrichment of 11q deletions in tumors with ATRX deletions might result from a bias caused by MNA. To exclude this possibility, we carried out the analysis again after excluding all MNA tumors from the reference cohort. Even though the percentage of patients with 11q deletion increased in the reference cohort, there was still a significant difference for both 11q deletion (p adj = 0.0004) and 2p gain (p adj = 0.02) between the ATRX deleted tumors and the reference cohort (Figure S3 and Table S5). Thus, this was a strong indication that the co‐occurrence of 11q deletion occurred more frequently in the deletion subtype than in the other two ATRX mutation types.

FIGURE 5

Co‐occurrence of ATRX aberrations with segmental chromosomal aberrations (SCAs). Percentage of patients with neuroblastoma‐specific SCAs for the ATRX mutation types and the ATRX WT reference cohort (including MYC‐N amplified tumors). Only significant p values after multiple testing correction are displayed (Fisher’s exact tests). del, deletion

Additionally, it is important to consider the status of the X chromosome, as ATRX is localized on it. In male individuals, a single ATRX aberration will abolish WT ATRX expression, however, female individuals normally have two alleles. Even though the high prevalence of in‐frame MEDs might suggest a gain‐of‐function/dominant effect, this still must be experimentally confirmed. The X chromosomal status was reported for 24 of 47 female patients in our study. For one female patient (Table S1, AMC802T), both alleles contained an ATRX aberration; for another female patient (PASTKW), two ATRX aberrations were also reported and potentially inactivated both alleles (could also be on the same allele). From the remaining 22 female patients, seven tumors lost one X chromosome and therefore the detected ATRX aberrations will abolish WT ATRX expression. However, we also observed two girls and one boy in which the tumor acquired one X chromosome. Thus, not always both ATRX alleles are no affected by aberrations in all tumors; nevertheless, ALT has been reported in at least some of these tumors, suggesting a potential dominant effect.

Finally, we also investigated the co‐occurrence of MNA, ALK, and TP53 mutations with ATRX aberrations. MYC‐N amplification was reported in only 4 of 121 ATRX aberrant tumors (two nonsense, one missense, and one MED tumor) (Table S1), which could be attributed to heterogeneous MNA as previously described. ALK status and TP53 status were only reported in a subset of the ATRX aberrant tumors (71 and 46 tumors, respectively; Table S1). ALK mutations occurred in 14 of these tumors, including all ATRX mutation types. Interestingly, TP53 mutations were only identified in two samples, which are the cell lines SK‐N‐MM and CHLA‐90 carrying an ATRX IFD. Thus, neuroblastoma‐typical SCAs and mutations co‐occur with ATRX aberrations and specifically 11q deletions are more frequently observed within tumors with ATRX deletions.

All types of ATRX aberrations are associated with ALT

ATRXaberrations are often associated with ALT and neuroblastomas with ATRX aberrations are more likely to have ALT than ATRX WT tumors. To investigate the frequency of ALT in the ATRX aberrant neuroblastomas from this study, we checked whether their ALT status was reported. The ALT status was assessed in a total of 44 tumors with ATRX aberrations, and a total of 43 samples were confirmed to have ALT (Table S1). The only ALT‐negative tumor contained a missense mutation at L407F, which was predicted as neutral and damaging by the PROVEAN and SIFT algorithms, respectively. The negative ALT status of this tumor might support a neutral role of this mutation in the tumor biology, although it could also be a false negative sample. Overall, these results confirm the strong association between the co‐occurrence of ALT with ATRX aberrations in neuroblastoma.

DISCUSSION

The aim of this study was to characterize the mutational spectrum of ATRX aberrations in neuroblastoma tumors and to present an overview of the accompanying tumor and patient characteristics. In this manner we identified a mutational hotspot region for ATRX missense mutations within the helicase domain. Interestingly, we determined that nearly 70% of the ATRX aberrations in neuroblastoma are MEDs, compared to only approximately 10% in other pediatric cancers. Additionally, we found that 75% of these MEDs are predicted or known to be in‐frame and could result in protein production. Our separate comparison of the three mutational subtypes with ATRX WT tumors revealed that 11q deletion only co‐occurs more frequently with ATRX deletions. Finally, we identified that, even though there is a higher frequency of ATRX aberrations in patients diagnosed at an older age, still approximately 40% of patients are younger than 5 years of age at diagnosis.

Although ATRX aberrations are frequently point mutations in most pediatric cancers, they represent a minority of the ATRX aberrations in neuroblastoma. We showed that in neuroblastoma the ATRX nonsense mutations were equally distributed across the ATRX gene. The effect of ATRX nonsense mutations in neuroblastoma has not been extensively studied. It is likely that these mutations lead to nonsense‐mediated mRNA decay or to an unstable protein prone to degradation. Unlike the random distribution of ATRX nonsense mutations, we identified the helicase domain as a missense mutational hotspot region in neuroblastoma. It is currently unknown whether missense mutations lead to reduced or absence of ATRX protein production in neuroblastoma. Interestingly, in patients with ATR‐X syndrome, a disease characterized by moderate intellectual disability and mild symptoms of α‐thalassemia, most of the missense mutations result in an unstable protein, more prone to degradation and resulting in reduced protein abundance. As the helicase domain is required for the DNA translocation activity of ATRX, defects in this domain could lead to disruption of histone variant H3.3 incorporation or other remodeling processes, such as removal of G‐quadruplexes or R‐loops. Therefore, ATRX point mutation models could illuminate the precise effects of the patient‐specific aberrations.

Multiexon deletions of the ATRX gene are almost exclusively present in neuroblastoma and rare in other pediatric malignancies investigated in this study. We determined that 75% of these deletions are known or predicted to be in‐frame in neuroblastoma. Several studies have reported that these in‐frame ATRX deletions still result in protein expression. , This high percentage of IFDs indicates that there could be a gain‐of‐function effect of these aberrations. This effect might have been identified by Qadeer et al., who postulated that ATRX IFFs are still able to localize to chromatin with an alternative binding pattern. We also showed that the deletions can be divided into two subgroups, namely the more frequent larger deletions including exons 8–9, and the less common smaller deletions lacking exons 11–12. These smaller deletions, which so far have not been well studied, could have different molecular effects in neuroblastoma cells as compared to bigger deletions and therefore might need other therapies. The low frequency of these smaller deletions could be explained by these being much harder to detect in genome sequencing data, which is the most often used technique to identify such aberrations. Additionally, we determined that 25% of the deletions are predicted to be out‐of‐frame; nonetheless some could still produce in‐frame transcripts by using exon skipping, which has been reported for the second most common deletion of exon 2–9 (MED#2).

In our study we determined that there was a trend towards a gender bias in the group of patients with ATRX deletions, however, this was heavily dependent on the used reference ratio and therefore should be investigated on a larger cohort. In addition, it is often not reported whether female patients have one or two affected ATRX alleles. This is also complicated by whole X chromosome losses or gains and random XCI that occurs during each cell division in female cells. Interestingly, ATRX is involved in the maintenance of XCI and both ATRX alleles might be transcribed in female patients. Support for this comes from a study that generated a conditional heterozygous KO of ATRX in female cells of the 8‐ to 16‐cell state of developing mice that resulted in abrogation of XCI in extraembryonic tissues. Perturbed XCI has already been observed in several cancers including breast and colorectal cancer. Thus, also for neuroblastoma, it will be important to determine whether the ATRX aberrant allele is transcribed in all or a proportion of the tumor cells, which might be relevant for treatment strategies.

It has previously been described that patients with neuroblastoma diagnosed at an older age more often had ATRX aberrations than younger patients. , , Our study confirmed this observation for each of the three mutation types independently in a cohort of 119 patients with ATRX aberrations. However, we also identified that almost 40% of the patients were younger than 5 years of age, which is much more frequent than described in earlier publications that contained fewer patients. , Furthermore, patients with ATRX aberrations have been reported to have a very poor overall and event‐free survival. , We observed this as well for the three mutation types regarding event‐free survival, but patients with a missense mutation had a better overall survival. An explanation could be that some of the observed missense mutations might be passenger variants that do not affect function, therefore patients might have been misclassified as ATRX aberrant.

In neuroblastoma there are few recurring mutations, but almost all high‐risk patients carry SCAs, of which many have been associated with poor overall survival. , Here we showed that almost 90% of the tumors with an ATRX deletion also have a 11q deletion, which was more frequent than in the reference cohort. The association of ATRX aberrations and 11q deletions has very recently been reported, , but in our study this is only the case for tumors with ATRX deletions and not for ATRX point mutations. The minimal region of overlap lost in 11q deletions contains a lot of genes involved in apoptosis and in cell cycle progression. ATRX aberrations are known to result in defective replication restart and concomitantly lead to increased DNA damage. Therefore, it could be evolutionary advantageous to acquire an 11q deletion when ATRX is aberrant to reduce the levels of DNA damage‐induced apoptosis, although it is also possible that the 11q deletion is already present before the ATRX deletion occurs and therefore provides the appropriate molecular environment. Why this co‐occurrence is more frequent for ATRX deletions than for point mutations and the order of the mutational events needs further investigation. We also show that ATRX deletions and 2p gain often co‐occur. Interestingly, in our dataset there were four tumors in which both an ATRX aberration and MNA were present. However, MNA and ATRX aberrations are reported to be clinically mutually exclusive , and elevated MYCN expression and ATRX loss were shown to be synthetically lethal in mouse models and cell lines. Nevertheless, for some tumors, heterogeneity of MNA and ATRX aberrations has been observed, , which could be a potential explanation for the patients with both aberrations in our dataset. Another possibility is that, in some instances, MNA might not result in high MYCN protein expression. Finally, TP53 mutations are very rare in our dataset, and are only present in two cell lines, where they might have been introduced in vitro. In contrast to glioma, where there is a strong co‐occurrence of TP53 and ATRX aberrations, we did not find such an association for neuroblastoma.

The major limitation of this study is the lack of data of both patient and tumor characteristics for many of the categories that we investigated, such as overall survival and SCA status. We are especially lacking a lot of information for the less frequently observed SCAs (2p gain and 3p del) and for the mutational status of ALK and TP53. This lack of information might have influenced our outcomes and therefore our results should be re‐evaluated in the future. Most importantly, the X chromosomal or ATRX copy number status are frequently not reported, even though this might be important information for female patients. In the future, it would be of great value if more complete and more standardized clinical and tumor data would be reported and if sequencing data would be publicly available.

In conclusion, this study shows that ATRX MEDs are the most frequent type of ATRX aberrations in neuroblastoma and that 75% of these are expected to produce in‐frame deleted protein products with a potential gain of function. We also identified a more frequent co‐occurrence of 11q deletions and ATRX deletions in neuroblastoma tumors. Overall, our study highlights that the different types of ATRX aberrations have distinct characteristics. It emphasizes the need for in vitro and in vivo models of these different aberrations, which might cause unique molecular phenotypes and require distinct treatments.

DISCLOSURE

The authors declare that there are no competing interests.

Table S1

Click here for additional data file.

Supplementary Material

Click here for additional data file.