Introduction

Until now, the proportion of cancers in children and adolescents attributable to an underlying genetic syndrome or inherited susceptibility is unclear. In the early 1990s, the inherited fraction of childhood cancer was estimated at 1–10 % [29]. A recent report from the Pediatric Cancer Genome Project/St. Jude Children’s Research Hospital determined an incidence of 16.0 % in patients with solid tumors, 8.6 % with brain tumors, and 3.9 % with leukemias. The report initially focused on 23 well-known cancer predisposition genes and 8 genes that predispose to pediatric cancer with a high penetrance [47]. The most frequently affected genes included TP53, APC, and BRCA2. Additional analyses were expanded to 565 genes that are known to play a role in various steps and pathways of cellular transformation. Identified variants were classified as pathologic, likely pathologic, uncertain significance, likely benign, and benign. Taking the larger gene-set into account, the overall prevalence of an inherited mutation increased only slightly, with a pathologic or likely pathologic variant being detected in 8.6 % of all patients and 4.6 % of patients with leukemia. However, the spectrum of tumors sequenced was not numerically representative of the spectrum of childhood tumors, and the mutation frequencies may be skewed accordingly. In a hereditary cancer risk assessment study in survivors of childhood cancer, a genetic counselor considered 29 % of the survivors as eligible for further genetics evaluation [19].

However, in the era of high-throughput sequencing in which new cancer susceptibility syndromes (CSS) and mechanisms are increasingly discovered—did we so far maybe just see the tip of the iceberg?

Current clinical approach to CSS

Pediatric oncologists consider children diagnosed with cancer and their families for inherited cancer susceptibility according to well-established criteria [20]. These comprise patient-specific constellations including (i) rare tumors commonly associated with cancer predisposition (e.g., adrenocortical carcinoma), (ii) bilateral or multifocal tumors (e.g., Wilms’ tumor), (iii) cancer diagnosis at a younger than expected age (e.g., thyroid carcinoma), (iv) multiple synchronous or metachronous tumors, (v) additional conditions (e.g., axillary freckling) indicative of an underlying syndrome, and (vi) suspicious family features. These might include (a) familial clustering of the same or closely related cancers, (b) cancer diagnoses in two or more first-degree relatives, (c) tumor patterns associated with a specific cancer predisposition syndrome, (d) exceptional young age at diagnosis, (e) sibling with childhood cancer, and (f) consanguineous parents.

Li-Fraumeni syndrome (LFS) is one of the most striking familial cancer predisposition syndromes. It is clinically and genetically heterogeneous and characterized by autosomal dominant inheritance and early onset of tumors, multiple tumors within one individual, and multiple affected family members. LFS presents with a variety of tumor types with soft tissue sarcomas, osteosarcomas, breast cancer, brain tumors, leukemia, and adrenocortical carcinoma being the most common tumor types. Comprehensive surveillance protocols have been implemented and proven efficiency in terms of superior survival [46]. Table 1 lists common hereditary cancer susceptibility syndromes sorted by the underlying mechanism. The American College of Medical Genetics and Genomics and the National Society of Genetic Counselors just published the latest referral indications for cancer predisposition assessment [13]. However, due to de novo mutations, incomplete penetrance of inherited mutations, and variable phenotype/genotype correlations, the family history may not in all cases be helpful. For example, up to 25 % de novo events of TP53 mutations are reported in Li-Fraumeni syndrome [6]. In most other cases of CSS, however, the proportion of inherited susceptibility versus de novo mutations remains unknown.

Table 1 Examples of common hereditary cancer predisposition syndromes
Full size table

Personalized medicine

With the ongoing efforts to personalize cancer medicine, comprehensive genome analyses will increasingly find its way into daily clinical routine in pediatric oncology. In the recently established German INdividualized therapy FOr Relapsed Malignancies in childhood (INFORM) project, this idea has been introduced for pediatric patients with relapsed or refractory high-risk disease without further standard of care therapy options. Individual tumor samples are characterized on the molecular level by next-generation sequencing (NGS) to establish a “fingerprint” of the tumor to identify promising targets for a successful relapse therapy [10].

Other such examples in which the detection of specific mutations has already led to a change of therapy of course also exist. Recently, a new leukemia subtype of high-risk B-precursor acute lymphoblastic leukemia (ALL), called Ph-like ALL, was characterized. Besides its Ph- or BCR-ABL-like transcriptional profile, no translocation t(9;22) or BCR/ABL rearrangement, respectively, is present. Instead, multiple other genetic alterations can be detected, which are potentially druggable by tyrosine kinase inhibitors or other targeted therapies [18, 24, 36, 37]. In pediatric low-grade astrocytoma, the BRAF V600E-mutation was identified as a frequent genomic aberration activating the MAPK pathway. Tumors carrying this mutation show significantly increased BRAF and CCND1 levels [33]. Since its discovery, the BRAF V600E-mutation has been described in an increasing number of pediatric central nervous system (CNS) tumors [8, 11, 40, 41]. Targeted therapies such as the BRAF inhibitor vemurafenib and MEK1/2 inhibitors are available and some encouraging examples of effective therapies even in very aggressive tumor types have already been reported, such as the successful treatment of a 12-year-old child with relapsed glioblastoma multiforme with vemurafenib [38]. With the identification of a highly recurrent genetic alteration and its resulting fusion protein in ependymoma, the C11orf95-RELA protein, a further potentially druggable target was identified and specific therapy will hopefully be available in the near future [31]. We might also hypothesize that children with hereditary cancer syndromes like the so-called rasopathies might soon benefit from targeted therapy, as the underlying genetic alterations are highly recurrent [1, 9].

Next-generation sequencing

Due to rapid technical advances in the field of NGS, tumor (including leukemia) genomes can nowadays comprehensively be analyzed within few days. Today’s state of the art in high-throughput sequencing already allows the usage of whole genome sequencing for research projects and of whole exome sequencing for daily clinical routine. However, the likelihood of identifying contemplable mutations is highly dependent on the relative ability of the sequencing approach to find these mutations. Computational processing, analyzing, and interpreting the massive amounts of data and genetic variants produced by NGS still remains challenging and requires comparisons with databases such as dbSNP and 1000 genomes project [3, 16]. Another valuable resource in interpreting own experimental data is the ExAC browser provided by the Broad Institute at www.exac.broadinstitute.org. It meanwhile provides exome data from more than 60,000 unrelated individuals. Before definitive conclusions can be drawn, the functional consequences of identified mutations on protein structure and function often have to be demonstrated experimentally [43]. In addition, a frequent conceptual misunderstanding relates to the fact that even a mutation with profound impact on protein function does not automatically proves its pathogenicity and disease-causing effect.

Each of us carries an average of approximately 3000 single nucleotide polymorphisms (SNPs) in terms of individual SNPs. To generate a personal cancer genome signature for molecular targeted therapy, it is important to discriminate between these individual SNPs and somatic (tumor) mutations. Thus, comparing the NGS data of tumor versus germline DNA is a condition sine qua non to identify the somatically acquired genetic variants of the tumor.

However, NGS not only generates focused genetic results with precise clinical implications for treatment but also so-called incidental findings with possible, limited, or unknown clinical impact or might even uncover an underlying susceptibility to cancer and other hereditary diseases. Such incidental findings are divided into “anticipatable” and “unanticipatable” ones. The former is a finding that is known to be associated with the test and is possible to be found. The latter could not have been anticipated given the current state of scientific knowledge [17]. Hence, treating physicians will increasingly be faced with such incidental genetic findings and the difficulties of interpreting and reporting these results.

Moreover, the pediatric oncologist is confronted with one new situation in particular: the fact that genetic events within the tumor genome itself can point toward underlying germline cancer susceptibility. Thus, even if not initially aimed to detect a CSS, the somatic “molecular make-up” of the tumor genome may suggest a germline mutation in a CSS gene.

Up to now, there are two well-known findings in tumor genetics which can be indicative that the tumor arose on the basis of the child’s germline alteration, (a) chromothripsis and (b) a high total number of mutational events which exceeds that of other samples of the same tumor type.

  1. (a)

    The phenomenon of chromothripsis was first reported by Stephens in 2011 [44]. The term “chromothripsis” (“chromo” from chromosome; “thripsis” for shattering into pieces) describes the shattering of a chromosome or a chromosomal region into tens to hundreds of pieces and locally clustered reassembling of some of the genomic fragments while others are lost to the cell.

According to Stephens [44], chromothripsis is defined by six features: (1) rearrangements localized within the genome, (2) oscillating changes of the copy number profile between one and two copies, whereby (3) loss of heterozygosity (LOH) causes a copy number of one, and retaining heterozygosity a copy number of two, (4) clustering of breakpoints across the chromosome, (5) conjunction of two remote chromosome fragments, and (6) joining rearrangements between two chromosome arms with clustering at the breakpoints. Rapid oscillations between copy number states one and two within the whole or parts of the chromosome characterizes the copy number profile in case of chromothripsis.

In contrast to common theories of cancer evolution through progressive accumulation of genomic alterations such as oncogene activation and tumor suppressor loss through environmental and lifestyle factors in adults, chromothripsis as a single catastrophic event might be involved in the development of a variety of cancers in childhood. It can cause the formation of new gene fusions, disruption of tumor suppressors, and amplification of oncogenes [35, 44]. In adults, 2–3 % of all cancers show evidence of chromothripsis; in bone cancers, this incidence is especially high with 25 % [44]. The impact of chromothripsis on cancer gene function and cancer development in childhood has already been demonstrated for many different tumor entities, e.g., ALL, AML, ependymoma, medulloblastoma, neuroblastoma, and retinoblastoma [4, 23, 26, 28, 30, 31, 35]. In addition, chromothripsis has been associated with poor prognosis in neuroblastoma [28]. A list of pediatric tumors, in which chromothripsis has been described, is given in Table 2. Conversely, alterations in TP53 have been shown for low-hypodiploid ALL but without chromothriptic pattern [15].

Table 2 Examples of (pediatric) tumors associated with chromothripsis
Full size table
  1. (b)

    To provide a comprehensive landscape of somatic genomic alterations (termed mutational signatures) in cancer genomes, numerous cancers have been profiled by DNA sequencing [2, 34, 45]. The occurring genomic alterations are presumably caused by defective DNA replication or repair and exogenous or endogenous mutagen exposure and include substitutions, insertions or deletions, rearrangements, copy number alterations, completely new sequences from exogenous sources, and combinations of all these possibilities. The prevalence of such mutations is highly variable between cancer (sub)types [2, 22]. Due to extensive exposure to carcinogens, small cell lung cancer (tobacco) and malignant melanoma (ultraviolet light) show the highest somatic mutation prevalence (over 100/Megabase (Mb)). In contrast, the mutation rate in pediatric cancers is lowest (0.1/Mb; approximately one change across the entire exome) as chronic mutagenic exposure plays a minor part in carcinogenesis in childhood [22]. An outline of mutation frequencies in various (pediatric) cancer types is given in Table 3.

    Table 3 Examples of mutation frequencies in (pediatric) tumors
    Full size table

Alexandrov et al. [2] described a mutational signature with very large numbers of substitutions and small indels, the latter at short nucleotide repeats and with overlapping microhomology at breakpoint junctions, termed “microsatellite instability,” which is characteristic of cancers with defective DNA mismatch repair and may suggest constitutional mismatch repair-deficiency syndrome (CMMR-D) in childhood.

As was shown by Rausch et al. [35], the single nucleotide variant (SNV) rate of children with Sonic-Hedgehog medulloblastoma (SHH-MB) is clearly higher (24 tumor-specific SNVs) in the case of inherited TP53 mutations compared to sporadic pediatric medulloblastoma samples (average 5.7 non-synonymous SNVs per sample; [32]). Thus, comparing the patient’s SNV with the average SNV rate of a given tumor entity, an increased mutation frequency (SNV rate) detected by NGS of the tumor again may point to an underlying CSS (Li-Fraumeni syndrome).

Ethical and legal issues

“Are our other children at an increased risk of developing cancer?” Parents of a child diagnosed with cancer frequently raise this question. Up to now, pediatric oncologists mostly reassure them that cancer in children usually is not hereditary but an exceptionally bad stroke of fate. However, will this statement still hold true in the future with ever-increasing knowledge about underlying cancer predisposition syndromes and inherited cancer susceptibilities in childhood?

The incidental finding of chromothripsis and its association with Li-Fraumeni syndrome in SHH-MB patients very well demonstrates the far-reaching consequences of translational research and genetic testing in pediatric oncology with its challenges for scientists, treating physicians, and the affected child and his entire family.

By detecting chromothripsis in a tumor, further genetic testing for germline p53 mutations is highly advisable as this phenomenon might be attributable to an underlying Li-Fraumeni syndrome. The latter obviously represents an important piece of clinical information as it will guide treatment, surveillance, and further early cancer screening programs [21, 46].

According to the recommendations of national and international human genetic societies and the legislation of most European countries, prior to genetic testing, the child (wherever possible) and the parents must be informed in detail, preferences as to which findings should be reported must be assessed, and written informed consent must be obtained. This is a well-established standard of care for targeted molecular testing an affected individual or suspected carrier for a specific hereditary condition. However, NGS is likely, apart from the initial indication to perform it, to uncover incidental findings, such as an underlying CSS as well as non-cancer-related germline mutations (e.g., CFTR, Huntington’s disease) with varying clinical importance for the patient. In order to comply with the aforementioned recommendations, this would require extensive genetic counseling of the child/parents of a child diagnosed with cancer undergoing NGS of the tumor prior to testing, which would have to encompass both incidental findings with possible, limited, or unknown clinical impact and numerous results unrelated to the indication for NGS [42]. We believe that this is highly impractical in the daily life of a pediatric hemato-oncologist as disclosing the diagnosis of cancer itself is overwhelming and dramatically limits the child’s/parents’ receptivity, and NGS of the tumor often has to be initiated at the time of diagnosis. However, whenever NGS is initiated, the treating physician has an obligation to discuss the full range of generated data and the possibility of incidental findings and its disclosure with the child/parents. Furthermore, the ordering physician is responsible for obtaining informed consent and providing pre- and post-test counseling. Thus, regarding the child’s/parents’ autonomy and both their right to access all NGS data and their “right not to know,” they should be informed of the benefits, risks, and alternatives of genetic testing in detail [5, 7, 12]. When the patient/parent refuses to be informed about incidental findings, even if disclosure leads to beneficial interventions, the physician must ensure that adequate information has preceded this refusal. However, most clinicians do not have sufficient training in NGS and need to be extensively trained for clinical translation and reporting of NGS data.

In contrast to the standards for genetic testing in adults, predictive testing in pediatric patients is only recommended when the disease is associated with childhood onset and only with available effective screening and/or intervention options [7, 39]. Refraining from predictive testing of children allows them to autonomously make this decision once they reach adulthood.

Last but not least, identifying children with hereditary cancer predispositions has immediate consequences for the entire family (siblings, parents, and extended family) [20, 25, 42]. Due to the young age of the index patient, potentially affected relatives might as well be young and yet asymptomatic. Having been tested themselves might—depending on the outcome—influence their family planning but will of course also provide an excellent opportunity to initiate early cancer surveillance programs which they will benefit from. However, genetic testing and tumor surveillance can have deeply affecting psychological consequences for the child and the family, emotional support should thus be in place for the families.

Clear legislation on returning genetic results in oncology are still missing. Lolkema et al. have thoroughly addressed the accompanying ethical, legal, and counseling challenges [25]. Comprehensive ethical recommendations on how to report research results to patients and parents are, for example, given by the American College of Medical Genetics and Genomics, the Boston Children’s Hospital, the American Academy of Pediatrics, the “EURAT” (Ethical and legal aspects of whole human genome sequencing) project of the Marsilius Kolleg of Heidelberg University, and the Leopoldina National Academy of Sciences Germany [5, 12, 14, 27, 39]. However, their practical implementation in day-to-day clinical life remains challenging.

Conclusions

Genetic testing and translational research in pediatric oncology provides new and exciting insights into the evolution and pathogenesis of childhood cancer. On the other hand, it can incidentally uncover an underlying cancer susceptibility syndrome with implications not only for the child but also for the entire family. Pediatric oncologists should therefore increase their awareness of chances and risks that accompany the increasingly wide clinical implementation of NGS platforms [42, 43].