Molecular genomic landscape of pediatric solid tumors in Chinese patients: implications for clinical significance

Pediatric solid tumors are significantly different from adult tumors. Studies have revealed genomic aberrations in pediatric solid tumors, but these analyses were based on Western populations. Currently, it is not known to what extent the existing genomic findings represent differences in ethnic backgrounds. We retrospectively analyzed the basic clinical characteristics of the patients, including age, cancer type, and sex distribution, and further analyzed the somatic and germline mutations of cancer-related genes in a Chinese pediatric cohort. In addition, we investigated the clinical significance of genomic mutations on therapeutic, prognostic, diagnostic, and preventive actions. Our study enrolled 318 pediatric patients, including 234 patients with CNS tumors and 84 patients with non-CNS tumors. Somatic mutation analysis showed that there were significant differences in mutation types between CNS tumors and non-CNS tumors. P/LP germline variants were identified in 8.49% of patients. In total, 42.8% patients prompted diagnostic, 37.7% patients prompted prognostic, 58.2% patients prompted therapeutic, and 8.5% patients prompted tumor-predisposing and preventive, and we found that genomic findings might improve clinical management. Our study is the first large-scale study to analyze the landscape of genetic mutations in pediatric patients with solid tumors in China. Genomic findings in CNS and non-CNS solid pediatric tumors provide evidence for the clinical classification and individualized treatment of pediatric tumors, and they will facilitate improvement of clinical management. Data presented in this study should serve as a reference to guide the future design of clinical trials.


Introduction
Pediatric tumors are simply divided into two categories: hematological malignancies (leukemia and lymphoma) and solid tumors (central nervous system (CNS) tumors and non-CNS tumors) (Surrey et al. 2019;Gutiérrez-Jimeno et al. 2021). Lancet Oncology published the first report assessing the global burden of cancer in children and adolescents, which stated that pediatric tumors were the sixth largest cancer burden globally, after adult lung, liver, stomach, colon, and breast cancers (GBD 2017Childhood Cancer Collaborators 2019. Meanwhile, the same report also showed that pediatric tumors were the ninth largest childhood disease burden in the world (GBD 2017Childhood Cancer Collaborators 2019. Several studies have confirmed that pediatric tumors are significantly different from adult tumors in terms of epidemiology, pathogenesis, prognosis, and molecular characteristics (Pfister et al. 2022;Vellichirammal et al. 2021;Jones et al. 2019). Therefore, classification of pediatric tumors was first proposed in the fifth edition of World Health Organization (WHO) based on morphological, immunohistochemical, and molecular features, indicating that pediatric tumors focus on multimodal diagnosis and individualized treatment (Pfister et al. 2022).
Compared with adult tumors, pediatric tumors have a better prognosis, and the > 5-year survival rate of patients with pediatric tumors is as high as 80% (Suh et al. 2020;Berbegall et al. 2014). For example, the 5-year survival rate of rhabdomyosarcoma in children is > 70%, while that of rhabdomyosarcoma in adults is only approximately 40% (Walterhouse and Watson 2007;Ferrari et al. 2003). Studies have shown that pediatric tumors are distinct from adult tumors in genetic and molecular features (Surrey et al. 2019;George et al. 2019). Unlike adult tumors, which are predominantly epithelial in origin, pediatric tumors are often mesodermal or neuroectodermal in origin, and approximately 8-12% of patients are known to have a hereditary predisposition to cancer caused by a pathogenic germline variant (Pfister et al. 2022;Gröbner et al. 2018;Fiala et al. 2021;Zhang et al. 2015;Parsons et al. 2016). Adult tumors often arise as a result of the accumulation of genetic mutations over time. However, pediatric tumors often result from maturation arrest of immature cell types. The tumor mutational burden (TMB) of pediatric tumors is significantly lower than that of adult tumors (Surrey et al. 2019;Chang et al. 2016; Thomas et al. 2007). Pediatric tumors are often driven by a single gene mutation that is heterogeneous, with only 30% of the genetic mutations being identical (Surrey et al. 2019;Gröbner et al. 2018). Universal adult cancer gene panels are now routinely used in pediatric oncology (Harris et al. 2016;Ortiz et al. 2016). However, the molecular differences between adult tumors and pediatric tumors make the usefulness of this application extremely limited in the clinic. For example, the hallmark mutation in childhood high-grade gliomas is histone H3 mutation, whereas adult high-grade gliomas are often driven by IDH mutations (Roux et al. 2020). An increasing number of researchers are paying attention to this phenomenon, focusing on developing cancer gene panels suitable for pediatric tumors (George et al. 2019). There are more than 150 targeted drugs for adult cancers in the United States; however, treatment options for childhood tumors are limited. Understanding the genetic landscape and epigenetic signatures unique to childhood tumors may help expand the therapeutic options of childhood cancers (Filbin and Monje 2019).
Currently, the landscape of genetic mutations in pediatric tumors is still under investigation, and none of these cohorts are Chinese (George et al. 2019;Gröbner et al. 2018;Chang et al. 2016;Thomas et al. 2007;Harris et al. 2016;Ortiz et al. 2016;Roux et al. 2020). Because of the heterogeneity and dissimilarity of pediatric tumors at the molecular genetic level, these studies are limited in clinical significance for the Chinese population. In our study, next-generation sequencing (NGS) was used to explore the molecular characteristics of 318 patients with pediatric solid tumors, which is the first large-scale study in Chinese children. We analyzed the clinical factors associated with the distribution of cancer types, such as sex and age. In addition, the study focused on mapping the molecular landscape of pediatric solid tumors and analyzing the molecular characteristics of different tumor types. Conclusions drawn from this study provided evidence for the clinical classification and rationale for individualized treatment of pediatric tumors.

Patient information and sample collection
We retrospectively recruited 318 pediatric patients with CNS tumors or non-CNS solid tumors from June 2019 to August 2021. The baseline characteristics of all patients were collected, including age, sex, and cancer type. Comprehensive genomic profiling of tumor tissue and paired blood was performed by WES, 539 tumor-related gene panel or 131-gene panel. Informed consent was obtained from all guardians of all individual participants recruited in our study.

DNA extraction and targeted next-generation sequencing
DNA was prepared from tumor [either formalin-fixed paraffin-embedded (FFPE) tumor blocks/slides or frozen fresh tissues] and paired blood samples and sequenced using WES, hybridization capture-based targeted NGS of the 539-gene panel or 131-gene panel by the CAP (College of American Pathologists)-accredited central laboratory at Jiangsu Simcere Diagnostics Co., Ltd (Nanjing, China). Genomic DNA (gDNA) of tumor tissues was extracted using the Tissue Sample DNA Extraction Kit (Kai Shuo), and gDNA of leucocytes was extracted using the MagMAXTM DNA Multi-Sample Ultra Kit (Thermo). The extracted gDNA was quantified with a Qubit dsDNA HS Assay Kit on a Qubit Fluorometer (Thermo Fisher Scientific), and the quality of gDNA was evaluated by Agilent 4200 TapeStation (Agilent).
Then, 15-200 ng gDNA was digested into 200-350 bp fragments by enzyme treatment, and fragmented gDNA was used to prepare sequencing libraries by the probe hybridization capture method with commercial reagents and customized probes. To ligate indexed paired-end adaptors for Illumina platform (SimcereDx), end repair was performed by KAPA HyperPlus DNA Library Prep Kit (Roche Diagnostics), and A-tailing was performed by VAHTSTM Universal DNA Library Prep Kit for Illumina ® (Vazyme). Unligated adaptors were removed by size selection with Agencourt AMPure XP beads (Beckman Coulter). The ligation products were amplified by PCR to form a pre-library for hybridization. The final qualified DNA libraries were then sequenced on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA) by 150 bp paired-end sequencing according to the manufacturer's instructions.
The minimum confidence threshold of variants and Indels calling was set to 2% in tissue. Tumor mutation burden (TMB) calculated in WES and the 539 panel was counted by summing all base substitutions and Indels in the coding region of targeted genes, excluding synonymous alterations, alterations of AF < 0.02 and alterations listed as known somatic alterations in COSMIC. To determine microsatellite instability (MSI) status, homopolymer repeat loci with adequate coverage on the panel were selected, and reads that were successfully mapped to each of the loci were extracted from the deduplicated BAM file. Msisensor was employed to evaluate the distribution of read counts among various repeat lengths and determine the stability of each locus. The MSI score was defined as the percentage of unstable loci (Niu et al. 2014). Any sample with an MSI score ≥ 0.2 in WES, ≥ 0.15 in the 539-gene panel, or ≥ 0.1 in the 131 panel was classified as MSI-H (MicroSatellite Instability-High).

Variant classification
According to the joint consensus recommendation of AMP (Association for Molecular Pathology), CAP (College of American Pathologists), and ASCO (American Society of Clinical Oncology), the identified somatic variants were categorized into four tiers based on their clinical impact on cancer diagnosis, prognosis, and treatment options ). Tier I: variants with strong clinical significance; tier II: variants with potential clinical significance; tier III: variants with unknown clinical significance; and tier IV: benign or likely benign variants.
Germline variants were classified into five levels based on American College of Medical Genetics and Genomics (ACMG) guidelines, including pathogenic (P), likely pathogenic (LP), variants of unknown significance (VUS), likely benign, and benign (Richards et al. 2015).

Statistical analysis
All statistical analyses were performed using R (4.0.5) and Excel. Differences in TMB and P/LP among subgroups were analyzed by the Wilcoxon test, and p values of < 0.05 were considered statistically significant. The genome-wide circos of fusion were displayed using circlize (R package). The oncoprint of the germline (P/LP) and somatic events was generated using ComplexHeatmap (R package). Other figures were generated with the R package ggplot2 or Excel.

Patient enrollment
A total of 318 patients with pediatric tumors were enrolled in our study from June 2019 to August 2021, of which 183 were males and 135 were females. Samples from 234 enrolled patients with CNS tumors and 84 with solid tumors (non-CNS) were examined using a 131-gene panel, 539-gene panel or WES. Blood was collected from all 318 patients as controls and analyzed for germline and somatic mutations. Patient categorization and analysis processes are shown in Fig. 1A, and the classification of somatic and germline variations and variation assessment are shown in Fig. 1B.
The median age of all enrolled patients was 9 years, the median age of patients with CNS tumors was 8 years, and the age range of all patients was from 0 to 18 years. The age distribution of patients with CNS tumors had two peaks, one at 3 years and the other at 10 years. The median age of patients with solid tumors was 13 years, and the patients' age ranged from 0 to 18 years. The incidence of solid tumors after 12 years was higher than that before 12 years. Both the total number of enrolled patients and the number of patients with CNS tumors were the highest at 3 years of age, while the number of patients with solid tumors was highest at 13 and 17 years of age. The age distribution of patients with CNS tumors and solid tumors is shown in Fig. 1C.
Out of 234 patients with CNS tumors enrolled in this study, 134 were males and 100 females. Out of 84 patients with solid tumor enrolled in this study, 49 were males and 35 females. Within the entire population of the enrolled patients, there were more males (183) than females (135). There were also more male patients than female patients in each subgroup. Glioma (132/234) was the most common CNS tumor, followed by embryonal tumors (47/234) and ependymal tumors (27/234). Tumors of soft tissue and bone were the most common solid tumors (60/84), followed by kidney cancer (13/84). The distribution of cancer types and sex is shown in Fig. 1D.
Subtypes of each carcinoma are shown in Supplementary  Table S1.
In CNS tumors, the proportions of mutation types were SNV (40.62%), CNV (38.73%), fusion (11.36%), and Indel (9.29%). In solid tumors, the proportions of mutation types were CNV (70.16%), SNV (22.38%), Indel (3.92%), and fusion (3.54%). There were significant differences in mutation types between CNS tumors and solid tumors (p value = 3.53E-31). The proportion of CNV in solid tumors was higher than that in CNS tumors (p value = 5.66E-32). The fusion type in CNS tumors and solid tumors also varied, with the highest incidence of BRAF fusion in CNS tumors and EWSR1 fusion in solid tumors.

Identification of gene fusions and CNV
The chromosomal distribution of fusion in CNS tumors is shown in Fig. 2C, and the chromosomal distribution of fusion in solid tumors is shown in Fig. 2D. The distribution of fusions in each chromosome is shown in Supplementary Figure S1. Fusion in CNS tumors occurred frequently on chromosomes 7 and 11, and fusion in solid tumors most commonly occurred on chromosomes 11, 22 and 5. The protein domain structures of the most common BRAF fusion in CNS tumors and the most common EWSR1 fusion in solid tumors are shown in Supplementary Figure S2. The novel fusions are shown in Supplementary  Table S2. The significance of these novel fusions remains to be explored.
The distribution of CNVs on chromosomes is shown in Fig. 2E. CNV occurred on chromosomes (chr) 1-13 and 17-19, and CNV appeared most frequently on chr2, followed by chr5 and chr9. Copy-number gains occurred of germline and somatic mutations, which occurred in 2% or more than 2%, is shown. Genes are classified into pathway categories and ordered by the proportions of mutations in the specified gene. The bar plot above shows the number of mutations in each patient and aberration types are highlighted. The bar plot on the right shows the number of mutations in each indicated gene and aberration types are is highlighted. The bottom heatmaps show the distribution of cancer type, age, and sex of patients (legend shown on the right) more frequently than copy-number losses, and only the APC gene had both copy-number loss and gain.

TMB analysis and mutational landscape
We analyzed the TMB in 91 patients with CNS tumors and 82 patients with solid tumors whose tumors were analyzed by a 539-gene panel or WES. The median TMB was 0.71/ Mb (0-12.13/Mb) in CNS tumors and 0.74/Mb (0-19.12/ Mb) in solid tumors, and there were no significant differences between the two groups (p value = 0.29) (Supplementary Figure S3).
The mutational landscape of all enrolled patients with pediatric tumors is shown in Fig. 2F. Patients with CNS tumors harboring BRAF fusions rarely had other co-mutations.

Characteristics of germline variants of tumor predisposition genes
We analyzed the germline SNV/Indel variants of 90 tumor predisposition genes in 318 patients in our study. The pathogenic (P) and likely pathogenic (LP) variants were classified according to the ACMG guidelines. P/LP germline variants were identified in 8.49% of patients (27/318), which is not lower rate than previously described in the Western population (Table 1). Among patients with CNS tumors, the frequency of P/LP germline variants was 9.40% (22/234); for patients with solid tumors, the frequency was 5.95% (5/84) (Fig. 3A). All variants were heterozygous and details are described in Supplementary Table S3.
As predictive biomarkers for immunotherapy, associations between TMB/MSI and germline variants have been established in some adult tumors. Three cases (P128, P182, P307) with germline heterozygous P/LP variants in mismatch repair (MMR) gene resulted in tumors with MMR gene deficiency, but their microsatellites were stable, and only P128 had high TMB (TMB = 12.13). The median TMB of pediatric patients with DDR was 0.71. In addition, there was no significant difference in TMB between the P/LP variant group (median TMB = 0.71) and the non-P/LP variant group (median TMB = 0.71) in our cohort (p = 0.7941273) (Fig. 3C).
With increasing use of multigene panels analyzed by NGS, a growing number of P/LP variants were found to be related to adult-onset diseases, such as BRCA1/2-tumors,  Table S3).

Genomic findings support clinical management improvement
We investigated the clinical significance of genomic findings in tumor tissue and paired blood samples obtained from 318 pediatric patients on therapeutic, prognostic, diagnostic, and preventive actions. Variants were classified   Fig. 4A). Genomic findings also prompted a change of diagnosis. 42.8% were the biomarkers of molecular diagnosis. In CNS tumors, four gliomas with specific BRAF fusions supported a diagnosis of low-grade glioma; a GBM with the C11orf95-RELA fusion suggested a diagnosis of ependymoma. We identified histone H3 G34R mutations resulting in a diagnosis of diffuse hemispheric glioma, H3 G34-mutant. A high-grade glioma harboring an SMARCB1 loss-of-function mutation and chr22q11.1-q13.33 deletion supported a change of diagnosis to atypical teratoid rhabdoid tumors (ATRT).
Most patients with P/LP germline variants in our study did not continue further genetic counseling and cascade testing. To date, 3 germline variants have been confirmed as inherited, 1 germline variant was de novo, and the remaining 23 are unknown (Supplementary Table S3). We reviewed and found that 67% (18/27) of families might accept cancer surveillance according to the results of cascade testing (Supplementary Table S3).

Discussion
Our study is the first large-scale study to analyze the landscape of genetic mutations in pediatric patients with solid tumors in China. In our study, 318 pediatric patients with CNS or non-CNS solid tumors were enrolled, and the basic clinical characteristics of the patients were retrospectively analyzed, including age, cancer type, and sex distribution of patients in the CNS and non-CNS solid tumor groups. Somatic and germline mutations of cancer-related genes in CNS and non-CNS solid tumors were further analyzed. The characteristics of somatic mutations in the two groups were analyzed, including tumor driver gene mutations and different mutation types (SNV, Indel, fusion, and CNV), as well as TMB and MSI. The germline mutations were classified into P/LP and non-P groups according to ACMG guidelines. Then, the frequency of P/LP in the CNS and non-CNS solid tumor groups and the overall frequency were analyzed and compared with those in previously published literature. TMB and MSI in the P/LP and non-P groups were analyzed. In addition, we investigated the clinical significance of genomic mutations on therapeutic, prognostic, diagnostic, and preventive actions. The proportion of mutations with respective clinical significances was analyzed, and we found that genomic findings might improve clinical management.
Cancer is the major cause of death in children worldwide. The incidence of pediatric cancer tends to increase with time and varies considerably in different regions, cancer types, sexes, ages, and racial and ethnic groups (Steliarova-Foucher et al. 2017). Pediatric tumors differ significantly in many aspects from those occurring in adults. For the first time, in the newest fifth edition of the World Health Organization (WHO) classification of tumors, childhood tumors are covered in a separate volume (Pfister et al. 2022;Alaggio et al. 2021). Genomic sequencing studies have highlighted remarkable heterogeneity in genetic alterations between pediatric and adult cancers (Sweet-Cordero and Biegel 2019). Pediatric cancers typically have a higher proportion of germline alterations in cancer predisposition-related genes and fewer somatic mutations than adult cancers (Gröbner et al. 2018). Most genomic studies of pediatric tumors have focused on high-risk diseases or specific tumor types. In our study, we analyzed the genomic landscape of solid tumors from pediatric patients in China for the first time, unselected for tumor type. Somatic mutation analysis showed that there were significant differences in mutation types between CNS tumors and non-CNS solid tumors, and the proportion of CNVs in the non-CNS solid tumor group was higher than that in the CNS tumor group. The fusion types also varied, with the highest incidence of BRAF fusion in CNS tumors and EWSR1 fusion in non-CNS solid tumors. Differences in fusion may be related to the cancer types enrolled.
The genetic events driving tumors are different between pediatric and adult cancers (Gröbner et al. 2018;Ma et al. 2018;Johnson et al. 2017). It has been reported that diverse genomic lesions drive pediatric cancers using comprehensive DNA and RNA genomic profiling (Newman et al. 2021). However, the comprehensive genomic profiling is currently beyond the capability of most cancer clinical services. As molecular diagnosis has far-reaching impacts on grading, classification and therapy, tumor diagnostics can no longer be made based on pathologic diagnosis alone, and integrated diagnosis has been devised (Pfister et al. 2022;Louis et al. 2021Louis et al. , 2018. In our study, we investigated the clinical significance of genomic findings using a tumorassociated multiple driver gene detection panel on treatment options, prognosis, diagnosis, and prevention. Genomic findings supported diagnosis, and 42.8% patients prompted diagnosis in total pediatric patients, especially in CNS tumors and tumors of soft tissue and bone. Although pediatric solid tumors have a lower frequency of somatic mutations than most adult cancers, pediatric solid tumors also harbor potentially useful biomarkers for targetable therapies (Parsons et al. 2016;Wong et al. 2020).

3
In all pediatric patients, 58.2% (185/318) were associated with therapeutic options, and the biomarkers with potential targetable therapies were prominently concentrated in the PI3K/MAPK/mTOR pathway and the cell cycle pathway. Although stratified targeted therapy is currently rarely used as the first-line therapy in patients with pediatric tumors, there is also an urgent need for personalized genomic profiling analysis for each pediatric tumor patient to increase diagnostic accuracy and to apply potential precision therapies more effectively and less harmfully.
Checkpoint inhibitors, including PD-L1 (programmed death-ligand 1) and PD-1 (programmed cell death protein 1) inhibitors, have emerged and have shown promise as therapeutic options for treating multiple cancer types (Bagchi et al. 2021). PD-L1 expression in tumor cells, MSI, and TMB have been reported as biomarkers of the response to immunotherapy (Doroshow et al. 2021;Chan et al. 2019). However, the majority of pediatric solid tumors show a low TMB and very limited immune cell infiltration and are often considered immunologically "cold" tumors (Johnson et al. 2017;Grabovska et al. 2020;Wienke et al. 2021). Therefore, pediatric solid tumors might have difficulty benefiting from immunotherapy unless when mismatch repair (MMR) or POLE, or POLD1 mutations occur resulting in high TMB. In our study, the median TMB in the CNS tumor group and the non-CNS solid tumor group were both low. One case harbored germline heterozygous MMR gene P/LP variants, and one case with mediastinal peritoneal germ cell tumor showed high TMB. In addition, there was no MSI-H found in our study in all patients enrolled, even if they harbored MMR gene deficiency, which suggested possible differences from adults.
Previous studies have found that heritable germline predisposition occurs in 8-12% of patients with pediatric solid tumors (Fiala et al. 2021;Zhang et al. 2015;Parsons et al. 2016). In our study, P/LP germline variants in tumor predisposition genes were identified in 8.49% of total pediatric patients; among patients with CNS tumors, the frequency was 9.40% and among patients with non-CNS solid tumors, the frequency was 5.95%. The overall rate of 8.49% is not lower than the described western population rate. Genes with the most prevalent P/LP variants were NF1 and TP53, and the signaling pathway most frequently affected was the DNA damage repair pathway. The analysis of germline predisposition is dependent on the cancer types' enrolled patients and genes included for analysis, as well as the variant interpretation.
There are several limitations in our study. Our study was a retrospective study, and the panel used was designed for generic application in adults. The cancer type of enrolled patients was not random, which may be related to the study center and whether the patients were tested. The number of enrolled patients with non-CNS solid tumors is limited. WES detection was not performed in all patients enrolled, and more patients received panel NGS detection instead.
In conclusion, our study is the first large-scale study in Chinese patients with pediatric tumors to analyze the landscape of genomic mutations, including somatic and germline mutations, as well as their clinical significance. There was no significant difference between our results and foreign children's genomes reported in the previous studies. Genomic findings in CNS and non-CNS solid pediatric tumors are expected to improve clinical management, which provides a reference direction for the future design of clinical trials and is an important application of gene panels in the evaluation and management of children.