Genomics-Driven Precision Medicine in Pediatric Solid Tumors

Simple Summary The detection of genomic aberrations in cancers has yielded a wealth of information to discover oncogenic drivers or pathogenic variants that are relevant for the development of precise treatment strategies. Recent studies have shown promising outcomes in adult cancer patients with well characterized cancer genetic biomarkers. However, the development of precise treatments for pediatric cancers is difficult due to the limited number of accessible samples and the fact that well-defined target genetic aberrations are limited. Here, we review the current landscape of pediatric precision oncology compared to adults and highlight the examples of single-arm and multiple-arm designs of pediatric precision treatments. Abstract Over the past decades, several study programs have conducted genetic testing in cancer patients to identify potential genetic targets for the development of precision therapeutic strategies. These biomarker-driven trials have demonstrated improved clinical outcomes and progression-free survival rates in various types of cancers, especially for adult malignancies. However, similar progress in pediatric cancers has been slow due to their distinguished mutation profiles compared to adults and the low frequency of recurrent genomic alterations. Recently, increased efforts to develop precision medicine for childhood malignancies have led to the identification of genomic alterations and transcriptomic profiles of pediatric patients which presents promising opportunities to study rare and difficult-to-access neoplasms. This review summarizes the current state of known and potential genetic markers for pediatric solid tumors and provides perspectives on precise therapeutic strategies that warrant further investigations.


Introduction
Cancer occurrence before the age of 20 years is rare, but it is one of the leading causes of disease-related mortality in children and adolescents globally [1,2]. Approximately 300,000 children aged 0-19 years old worldwide are diagnosed with cancer each year [1], 1.
Familial history of the same or related cancers; 2.
Occurrence of bilateral or multifocal cancers; 3.
Earlier age at disease onset; 4.
Physical suggestive of a predisposition syndrome; 5.
Appearance of specific tumor types corresponding to the genetic predisposition.
Somatic mutations are de novo genetic alterations that spontaneously develop in an individual cell over time and play a vital role in cancer development and progression [51]. Studies have shown that the number of genetic abnormalities identified in each cancer patient may increase over time, leading to tumor survival against the selective pressure of drug actions, thereby acquiring resistance and causing disease progression [13,52]. Commonly identified somatic mutations include those involved in RTK signaling (PDFGRA, ERBB2 and EGFR), MAPK signaling (NF1, KRAS, and MAP2K1), PI3K-mTOR signaling (PIK3CA, MTORC1/2 and PTEN), cell cycle (CDKN2A/B, RB1 and ATM), DNA maintenance (TP53), transcriptional regulators (MYC and MYCN), and epigenetic modifiers (SMARCB1 and ATRX) [12,53]. Cancers usually involve a different spectrum of mutation which are strongly associated with pathogenesis and disease prognosis. A pan-cancer analysis reported by Grobner et al. [33] showed that 93% of adult cancer patients harbor at least one significantly mutated gene, while only 47% presented such mutations in pediatric tumors. However, approximately 30% of recurrent hot-spot mutations in pediatrics overlapped with adult cancers, highlighting some potential druggable targets based on finding from adult cancers. Hence, advances in identifying and understanding oncogenic drivers and actionable mutations would further improve the current therapeutic strategies for the development of precision medicine in cancers.

Germline and Somatic Variants Classified as Druggable
In the context of defining mutational actionability, the relevant effects of genomic aberration participating in cancer phenotypes are considered. DNA aberrations include missense, nonsense, frameshift mutations, and chromosome rearrangements, with some changes affecting only a single DNA base that may or may not alter the protein's property and some point mutations completely abrogating protein expression. A wide variety of gene alterations have been detected such as activating point mutation in BRAF, ALK, EGFR and FGFR1 genes, high copy number gains in PDGFRA and ERBB2, loss-of-function mutation affecting PTEN, PTPN11, PIK3R1, and MTORC1, CDKN2A/2B deletions, or inframe expression of large indels (NOTCH1 and FOXA1) [12]. Other changes involving larger stretches of DNA may include rearrangements, deletions, or duplications of long stretches of DNA [54]. For example, exon skipping on MET exon 14 proto-oncogenes resulting from intronic mutation increases the protein lifespan and promotes MET activation in lung carcinogenesis [55].
The significance of genetic variants may vary depending upon their potential effects on cellular functions. An "actionable" mutation is defined as a genetic aberration that is potentially responsive to targeted therapy, while a "driver" mutation refers to variants that confer a growth advantage to cancer cells but may not be targetable with a specific treatment yet. Passenger mutation is used to designate cancer-neutral variations and is unlikely to be under selective pressure during the evolution of the cancerous cells [56,57]. The "passenger" mutation has the lowest tendency to impact protein function, most of which are synonymous substitutions; however, these mutations occur more frequently than driver or actionable mutations. Unraveling the passenger mutational paradigm has otherwise revealed the existence of pre-existing latent driver mutations in which certain combinations of the passenger mutations could indeed be functional drivers. One example is the non-hotspot, passenger mutation of the Akt1 gene at position L52R, C77F, and Q79K, which promotes its membrane localization similarly to the E17K driver. In contrast, the co-existence of D32Y, K39N, and P42T passenger mutations can lead to Akt conformational inactivation, suggesting that treatment decisions based only on genetics may overlook crucial actionable components [56,58]. In addition, silent mutations occurring near the donor splice junction could contrarily affect exon splicing. For example, T125T mutation in TP53 is a recurrent mutation that is generally considered a non-functional passenger event; however, its existence at the −1 donor site of exon 4 raises the possibility that this mutation affects splicing. Further integration with RNA-seq data demonstrated that T125T mutation resulted in the retention of intron 4 and introduced a premature stop codon such as nonsense-mediated decay [59]. Thus, aberrant splicing caused by silent mutations should be carefully evaluated during interpretation of the sequencing results.
The accumulated data of genetic composition data from the tumors of patients has become a growing compendium of molecular biomarkers for precise treatment with FDAapproved drugs. Figure 1 summarizes the actionable mutations currently approved by FDA consortium for targeted therapy in adult cancers and pediatric solid tumors. Common actionable genetic aberrations associated with the National Comprehensive Cancer Network (NCCN) guidelines or FDA-approved targeted therapies are extensively summarized in Table 2. The data were predominantly gathered from the OncoKB database and the representative cancer types, and levels of evidence were included [60].  Table  1). Note that targeted therapies against PTCH1 and ALK in medulloblastoma and neuroblastoma are currently undergoing clinical assessment and awaiting further approval.   Table 1). Note that targeted therapies against PTCH1 and ALK in medulloblastoma and neuroblastoma are currently undergoing clinical assessment and awaiting further approval.    a FDA-approved level 1 = FDA-recognized biomarker predictive of response to an FDA-approved drug in this indication; level 2 = Standard care biomarker recommended by the NCCN or other professional guidelines predictive of response to an FDA-approved drug in this indication; level 3 = Standard care or investigational biomarker predictive of response to an FDA-approved or investigational drug in another indication; level 4 = Compelling biological evidence supports the biomarkers as being predictive of response to a drug; level R1 = Standard care biomarker predictive of resistance to an FDA-approved drug in this indication; level R2 = Compelling clinical evidence supports the biomarker as being predictive of resistance to a drug. b FDA-approved for pediatrics used [61]. c Clinical trial in pediatrics.

Pediatric vs. Adult Cancer Development
Pediatric cancers reflect a heterogeneous group of disorders distinct from adult cancers in terms of cellular origins, genetic complexity, and specific driver alterations [62,63]. Pediatric malignancies typically occur in developing mesoderm rather than adult epithelia (ectoderm) and are often induced by inherited or sporadic errors during development [33]. Studies have quantified the mutation burden in many pediatric cancers, identifying approximately 5 to 10 protein-coding variants identified across multiple tumor types except in osteosarcoma, which showed an average of 25 protein-affecting mutations. In contrast, the average number of mutations in adult cancers ranges between 33 to 66 in pancreatic, colon, breast, and brain cancers while mutagen-caused adult tumors (such as melanoma and lung cancers) can include up to 200 protein-coding variants [64][65][66]. At diagnosis, patients with pediatric cancers tend to have less complexity on mutational spectra than those in adult cancers; however, with treatment-refractory tumors and recurrence-the mutation rates in pediatric tumors have increased to be comparable to adult tumors [67,68]. Moreover, the rare occurrence of pediatric cancers and the low frequency of recurrent genomic alterations have a great impact on the investigations and the availability of targeted agents. Thus, there is an urgent need to accelerate the pace of genomic data acquisition and clinical trials in children to design more effective strategies for pediatric precision oncology.

Somatic and Germline Mutations Identified in Pediatric Cancer Cohorts
Single nucleotide variations (SNVs) and small indels are the usual mutations identified in adult cancers. In contrast, childhood cancers show a relatively high prevalence of copy number aberrations (CNAs) and specific structural variations (SVs). Note that insertion and deletion lead to adding and removing at least one nucleotide to the gene, respectively, which can affect protein functions and contribute to carcinogenesis. Current data suggest that approximately 10% of pediatric cancers are caused by genetic predisposition [32]. Zhang et al. [31] revealed that 95 out of 1120 (8.5%) patients younger than 20 years of age harbor germline mutations in cancer-predisposing genes. Diets et al. [69] performed trio-based whole-exome sequencing on the germline DNA of 40 selected children with cancer and their parents. Of these, germline pathogenic mutations were identified in 20% (8/40) of children with cancer [69]. Similarly, Grobner et al. [33] reported that most germline variants were related to DNA repair genes from mismatch (MSH2, MSH6, PMS2) and double-stranded break (TP53, BRCA2, CHEK2) repair.
Using combined somatic and germline sequencing for children with solid tumors, Parsons et al. [32] identified actionable mutations in up to 40% (47/121) of pediatric solid tumor tissues. Likewise, Wong et al. [12] performed the combination of tumor and germline sequencing (WGS) and RNA sequencing (RNA-seq) to identify 968 reportable molecular aberrations (39.9% in both WGS and RNA-seq; 35.1% in WGS only and 25.0% in RNA-seq only) in 247 high-risk pediatric cancer patients with 252 tumor tissues. Interestingly, 93.7% of these patients had at least one germline or somatic aberration, 71.4% had therapeutic targets, and 5.2% had a change in diagnosis [12].
These cohort studies emphasized that comprehensive molecular profiling could resolve molecular aberration in high-risk pediatric cancer and provide clinical benefits in a significant number of patients. In the era of next-generation sequencing, publicly genomic data access is considered one of the keys to accelerate research. The St. Jude Cloud is one of the most promising data-sharing ecosystems, with genomic data from >10,000 pediatric patients with cancer and long-term survivors. When exploring the mutational profile of pediatric solid tumors, the resource has revealed common genetic alterations among the different cancer types, as shown in Table 3. This integrative view of genomic data could be further used to expedite studies of pediatric cancer-associated risk factors and initiate novel therapeutic investigations for improving treatment outcomes.

Predictive and Common Genetic Variant Abnormalities Identified in Pediatric Tumors
The reports of actionable mutations identified in various studies have ranged from 27% to 100%, depending on the study design [6]. Several methods have been adopted for comprehensive molecular analysis to discover the actionable mutations that result in the targeting of cancer-associated elements. Table 4 contains a comprehensive, up-to-date summary of genomic aberrations found in pediatric solid tumors, together with potential targeted treatments, based on several public databases [60,[70][71][72][73]. We systemically reviewed genomic alterations with high prevalence in pediatric cancers using comprehensive WES and RNA-seq data via the St. Jude Cloud (www.stjude.cloud; accessed on 26 September 2022) [70]. Importantly, the genomic point mutations and gene fusions reported by this public domain are unique and different from those variants identified in the OncoKB database (the mutational collection of adult cancers) [60]. In addition, the potential druggable targets of these significant genomic alterations required further testing in pediatric solid tumor patients. A significant number of studies [60,[69][70][71][72][73][74][75][76][77][78][79][80][81][82][83][84][85] were reported by the Clinical Interpretation of Variants in Cancer (CIViC) database (https://civicdb.org; accessed on 18 September 2022) [71] which matched genomic alteration and molecularly targeted therapies tested in pediatric patients. These treatment designs were translated from the clinical care of adults across different tumor types but harboring the same genetic dysregulation, which gave satisfactory clinical outcomes. For pediatric solid tumors with no clinical evident support or undruggable genomic alterations, we listed the potential targeted therapies based on the knowledge from adult cancers as suggested by cBioPortal (www.cbioportal.org; accessed on 30 April 2022) [72,73] and OncoKB (https://www.oncokb.org; accessed on 17 April 2022) [60] that should be considered for further investigation and optimization for pediatric treatments. As of now, fewer number of patients could hinder the availability of molecular characterization and statistically meaningful preclinical/clinical outcomes. However, this challenge can be overcome by the initiation of multi-institutional cooperation and international data sharing, which would enable clinicians to effectively explore optimized therapeutic interventions toward pediatric precision oncology.

Current Progress in Clinical Trials for Pediatric Precision Oncology
Genomic precision medicine has demonstrated preferential outcomes among ongoing genomic-driven clinical trials in adult cancers. Yet, clinical investigations based on pediatric tumor genetics are still lacking. Based on the patient genetic profile screening, scattered reports on molecularly defined pediatric patients are showing prominent responses to some targeted therapies. For example, targeting ALK has shown success in treatments of ALK(+) non-small cell lung cancers and also in childhood anaplastic large cell lymphoma (ALCL) and inflammatory myofibroblastic tumor using the ALK inhibitor crizotinib [92]. While ALK mutation is the most common somatic mutation in neuroblastoma, crizotinib was compromised due to the interference by common ALK mutation F1174 [93]. Since then, ceritinib, alectinib, brigatinib, and lorlatinib have been approved against advanced ALK+ NSCLC [94][95][96][97]. Intriguingly, the third-generation TKI that targets both ALK and ROS1, lorlatinib, has recently shown promise in patients with ALK mutated neuroblastoma, but most of the studies are still at phase I clinical trial. [98]. Nonetheless, repotrectinib, a next-generation ROS1/TRK inhibitor with >90-fold potency against ROS1 than crizotinib in NSCLC patients is also being tested for dose escalation in phase II clinical trial with patients aged ≥ 12 years [99]. Another promising example is the targeted therapy against Ras-Raf-MEK-ERK signaling cascade which include somatic BRAF alterations (BRAF V600E and BRAF fusions). The prototype for targeting BRAF V600E/K is cutaneous melanoma, where 40-60% of patients with these mutations are eligible for the FDA-approved BRAF-inhibitor, vemurafenib [100]. Low-grade-gliomas have been identified to contain multiple alterations in Ras-Raf-MEK-ERK pathway, and a single treatment of vemurafenib in malignant glioma resulted in tumor regression [85,101]. Recently, Jain et al. [102] reported that a combination of BRAF-inhibitor dabrafenib and MEK-inhibitor trametinib enhanced treatment efficacies in pediatric low-grade-glioma carrying KIAA1549-BRAF fusion. Additionally, several studies have utilized the combination of molecularly targeted agents and traditional chemotherapy or radiation to reduce the severe side effects caused by an intensive dose of chemo/radiotherapy while minimizing acquired drug resistance due to selective pressure (Table 5).    The following large-scale pediatric and young-adult precision oncology programs have been launched with multiple-arm trials for patients with matched molecular profiles: TAPUR (ClinicalTrials.gov identifier NCT02693535), NCI-COG Pediatric MATCH (NCT03155620), the Tumor-Agnostic Precision Immuno-Oncology and Somatic Targeting Rational for You (TAPISTRY) (NCT04589845). These global, multicenter, open-label, multicohort studies are now at phase II, and the treatment assignment has relied on the basis of relevant onco-genotypes as identified by a Clinical Laboratory Improvement Amendments (CLIA)-certified or a validated next-generation sequencing (NGS) assay. While the eligible criteria of TAPUR are open for patients aged 12 years old or older, most of the patients enrolled are reported to have adult cancer phenotypes [103][104][105]. In contrast, the NCI-COG Pediatric MATCH aims to evaluate the molecular-targeted therapies with selected biomarkers of childhood and young adult patients with a reported detection rate of actionable alterations of 31.5% from the first 1000 tumors screened. Assignments to treatment arms were made for 28% of patients screened and 13% of patients enrolled in the treatment trial [106]. In the TAPISTRY study, nine targeted treatments are being examined, and eleven non-randomized treatment arms are available for participants of all ages with locally advanced/metastatic solid tumors. The purpose of this study is to evaluate the safety and efficacy of different targeted therapies and immunotherapies in patients as single agents, but the results of the study are still to be released. Overall, the advancements in high-throughput sequencing technology have closed the gap between the current treatment paradigm and precision medicine, markedly improving rates of response, progression-free survival (PFS), and overall survival (OS) compared to traditional randomized trials. Moreover, the multicenter, open-label, multi-arm treatment designs can further benefit treatment strategies by yielding efficacy and toxicity data in a timely manner with cost-effectiveness. Therefore, in the future, international coordination will be crucial to generate a database to inform rational trial design and to evaluate the combination of treatments/interventions that ensure more favorable outcomes.
The current applications of precision study designs for pediatric cancers (summarized from clinicaltrials.gov; accessed on 17 August 2022) are shown as single-arm and multiplearm designs in Tables 5 and 6, respectively.

Challenges and Perspectives
Large-scale cancer sequencing studies such as the 1000 Genomes Project [107], The Cancer Genome Atlas (TCGA) [108], and the International Cancer Genome Consortium (ICGC) [109] provide an extensive landscape of tumor genomic profiles which substantially facilitate the predication of recurrent hot-spot mutations on the selected type of cancers. Other large databases aim to collect the profile of childhood cancers include St. Jude/Washington University Pediatric Cancer Genome Project (PCGP) [110] and NCI's Therapeutically Applicable Research to Generate Effective Treatments (TARGET) [53] which are accessible via the St. Jude Cloud (https://www.stjude.cloud, accessed on 26 September 2022) public data repository. These large-scale studies have confirmed that the spectra of genomic alterations and their relevant mechanisms differ in childhood tumors from those predominantly occurring in adult cancer-at least by half. Thus, the actionability of pediatric-driven mutations needs to be carefully interpreted before translating into a targeted treatment option.
Several challenges need to be addressed when researchers launch the study/trial for pediatric cancer treatment. Many pediatric cancers are rare, and finding the right patient population for the drugs is challenging. In fact, a small patient population and a prolonged trial duration are not uncommon issues in the settings of rare diseases and low-incidence pediatric cancers [111][112][113][114]. Optimal statistical designs for less stringent comparisons, for example, by relaxing type I error (higher than 5%) or power (lower than 80%) can still provide meaningful results from small but faster trials [111][112][113][114]. Implementing multi-arm multi-stage trial design would allow patients with poor prognosis to be stratified into multiple phase II arms; receiving the window-of-opportunity/experimental therapies and restaging by serial biopsies and molecular characterizations to inform ongoing treatment choices [113,114]. These approaches remain useful to increase the overall feasibility for rare disease trials, i.e., keeping the sample size as small as possible while maintaining the power and ability to address the trial objectives.
Only 45% of pediatric cancer driver genes are shared with adult cancers, suggesting that novel therapeutic agents are required for pediatric cancer. Additionally, pediatric cancers are often driven by structural variants that can be challenging to identify and target. Nonetheless, children with cancers have accumulated fewer genetic mutations, thus making genomic targeting simpler than adults [113]. In a broad view, cancer intrinsic targets (e.g., mutated oncogene, tumor suppressor, epigenetics, synthetic lethal, and DNA damage) play crucial roles in cancer pathogenesis and thus could serve as the key stones for drug development against childhood cancers [115]. Another approach in drug development strategy is a mechanisms-of-action (MoA)-driven approach which successfully exemplified the efficiency of nivolumab and larotrectinib as targeted anticancer drugs against programmed cell death protein-1 (PD1) and TRK receptors, respectively [116]. Nonetheless, lessons learned from adult cancers have warned us that many pediatric cancers would have failed to express mutated kinase targets, and resistance to targeted therapies would rapidly occur. Recently, newly emerging cancer targets have been discovered upon multidimensional complexity of the dynamic oncogenic states, for example, tumor archetypes, master regulators, cancer-associated protein-protein interactions, and metabolic vulnerabilities [115,[117][118][119][120]. The development of drugs against the emerging classes of cancer targets may deliver adjunct/complementary agents for combination with targeted therapeutic regimens [115]. The emergence of gene editing technologies such as transcription activatorlike effector nucleases (TALENS) and clustered regularly interspaced palindromic repeats (CRISPR) paired with the CRISPR-associated endonuclease 9 (CRISPR-CAS9) offer the powerful customizable therapeutic options to precisely edit the targeted genes [121][122][123], thus providing hope to all pediatric cancers to be benefited from genomic-driven precision medicine approach.
Comprehensive molecular profiling of the genetic variants/mutations, gene expression at both transcripts and protein levels, and perhaps information on post-translational modifications and metabolites are coordinately utilized to improve the accuracy of molecularly targeted agents. Challenges in this grand scheme, besides big data sharing and multi-omics integration, are interpreting complex high-dimensional data in the biological sense, prioritizing findings into actionable targets/pathways, and achieving the candidate compounds/drugs for precise treatment. Aberrant expression of messenger RNA associated with genomic changes could contribute to the biology of tumor progression. In most cases, RNA-seq analysis can increase the coverage number of variant curations, especially the comprehensive gene fusion discovery and tumor expression subgroup analysis, when compared to WGS alone [124]. A novel molecularly guided approach, so-called transcriptomic connectivity analysis, utilizes the power of RNA-seq to detect aberrant gene expression and employs transcriptomic reversal of cancer cells/tissues for repurposing FDA-approved drugs [125][126][127]. This molecularly guided therapeutic approach could be an asset for prioritizing the approved drugs for off-label use in childhood cancer trials.
Last but not least, it should be noted that new therapeutics often lack dosage guidelines for children [12]. Acknowledging children have different drug responses and tolerance profiles compared to adults, it is crucial to define the optimal dosages of new drugs/biologics (and the off-label use of FDA-approved medications) to achieve preferred therapeutic outcomes. Recent innovations in study designs (i.e., phase I dose-finding design for pediatric population, the potential inclusion of children in adult trials, cooperative group trials) [131][132][133][134], together with the regulatory initiatives in the United States (US) and the E.U. which encourage the development of novel anticancer therapies in children [134,135], provide guidance to address this challenge while accelerating the pace of genomic-driven precision medicine in pediatric oncology.

Conclusions
Essential questions that need to be addressed in applications of precision therapeutic program include the applicability of the genetic testing, the significance of the mutation variant, and the existence of an approved targeted therapy. Although targeted agents are approved for a set of tumors harboring specific mutations, future development of clinical guidelines may recommend these agents to be used off-label in different tumor types with the same mutations. Identifying the mutational signatures of pediatric solid tumors will open opportunities for new targeted therapeutic strategies since their malignant origin manifests differently from the adults. Similar genomic-driven precision medicine approaches have been launched by several institutes, while the long-term effects of many of those novel agents are just beginning to be evaluated. These treatments could improve survival and reduce toxicity in pediatric patients and maximize therapeutic advantages when incorporated into standard care.

Acknowledgments:
The graphical abstract was created with BioRender.com (accessed on 29 November 2022).

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.