Genomic landscape of HRD/hypermutated pediatric cancer

Literature Review

Canonical signs of homologous recombinant deficiency (HRD) from broad germline sequencing studies of pediatric cancer

A defect in a single gene such as BRCA1 or BRCA2 does not result in a single signature – it gives rise to mutational signatures of all classes, including SNV, Indel, and SV types (Davis, 2017; Cuppen, 2020).

• SBS3, a uniform pattern of mutations across all 96 possible substitution types, was first discovered to be correlated with BRCA1/2 mutations in breast cancer and later extended to pancreatic, ovarian, and gastric cancer in adult cohorts. However, a later study demonstrated that most samples in the top quartile of SBS3 activity did not exhibit deleterious BRCA1/2 mutations, suggesting that other events might also contribute to SBS3 including germline nonsense and frameshift variants in PALB2 but not ATM or CHEK2 and epigenetic silencing of RAD51C and BRCA1 by promoter methylation (Polak, 2017). SBS3 is said to not be as specific for HRD as previously believed, and using SBS3 alone to classify tumors as HRD has been shown to overestimate its prevalence in esophagus, lung, head, and neck, and uterine cancers (Degasperi, 2020).
• SBS8, characterized by C>A substitutions, has also been associated with HRD, although to a lesser extent than SBS3.
• ID6, an excess of large deletions (>3 bp) with flanking micro-homology at the junction of the deletion (Davis, 2017) or del.mh.bimh.2.5 was found to be the most important predictor of HRD (Cuppen, 2020).
• SV3, comprised of very short (<10 kb) simple tandem duplications between is associated with tumors that have a BRCA1 (but not BRCA2) mutation (Nik-Zainal, 2016).

BRCA carriers in the literature so far

1. DKFZ: Waszak (2018)

• Patient characteristics (Cohort/Individual patient descriptions, and available demographic and clinical characteristics): 1,022 patients with medulloblastoma (MB) were analyzed for cancer predisposition genes, but only 375 WGS with >100 SNVs and low quantification error were retained for signature analysis. Mutations in BRCA2 were discovered in 11 (1%) of 1,022 patients – 10 pediatric patients and 1 adult, median age at diagnosis was 5-7 years. Germline mutations in BRCA2 showed no strict associations with age.
• Clinical info available for 4 patients with compound heterozygous mutations of BRCA2: ICGC_MB309 was diagnosed at the age of 4.4 years with MBSHH and had microcephaly, short stature, and café-au-lait spots. The patient further developed severe skin and mucosal toxicity, repeated sepsis, and severe hematologic toxicity during treatment. The patient had a positive family history of cancer (ALL, AML, nephroblastoma, breast cancer, ALK-positive lung cancer). SF_D1212 was diagnosed at the age of 2.4 years with MBSHH and had mild “elfin-like” facies, and skin pigmentation (multiple brown punctate lesions) on the leg, groin, labia, abdomen, and back. The patient had a positive family history of cancer with one sibling being diagnosed with neuroblastoma and another sibling with MB and Wilm’s tumor. SF_D2245 was diagnosed at the age of 2.8 years with MBSHH and had short stature, “elfin-like” facies, and polydactyly. The patient had no family history of cancer. SJMB012 had no cardinal physical features of Fanconi anemia. He was diagnosed with MBSHH when he was 8.5 years old and received radiation therapy, chemotherapy, and stem cell rescue. He had persistent unexplained macrocytosis of his red blood cells, yet without anemia or any other pancytopenias. The latter would be, in principle, consistent with signs of Fanconi anemia, yet this patient was not officially diagnosed as such and is a long-term MB survivor (12 years). Of note, the father was diagnosed with Lynch syndrome and he is carrier of a germline mutation in MLH1, yet the mutation was not transmitted to the affected child.
• Tumor type (Histology and molecular classification into consensus subgroups): Diverse subgroups of medulloblastoma observed among BRCA2 carriers: MB (SHH), MB (Group 3), MB (Group 4). However, 4/11 patients with compound heterozygous mutations of BRCA2 developed exclusively MB (SHH) and exhibited worse progression free/overall survival compared with 7/11 patients who had heterozygous germline mutations (100% survival). Patients with germline mutations in BRCA2 (n=9) and PALB2 (n=3) were diagnosed with classic medulloblastoma (5/12), desmoplastic medulloblastoma (4/12), large cell/anaplastic medulloblastoma (2/12) and medulloblastoma with extensive nodularity (1/12).
• Signature biomarkers of HRD (SBS3, SBS8, ID6, SV3): MBs were classified as HRD if most (>50%) of somatic SNVs were assigned to SBS3 or 8. 9 of the 12 BRCA2- and PALB2- deficient tumors showed evidence for an HRD-like mutation spectrum. Analysis of tumor-derived somatic mutation patterns in all four compound heterozygous BRCA2 cases revealed signs of HR-deficiency (59.0%, 66.7%, 73.5%, and 81.8% of somatic mutations assigned to mutational signatures 3 and 8). SVs were called using Delly, but exposures not described in the report. =
• Other biomarkers of HRD (e.g. R-loops, DNA-RNA hybrid structures): Chromothripsis was investigated in this paper for the cohort and reported in Li-Fraumeni syndrome patients (TP53 carriers) but were not reported specifically for BRCA2 carriers.

2. MSKCC: Miala (2021)

• Patient characteristics (Cohort/Individual patient descriptions, and available demographic and clinical characteristics): 751 pan-cancer solid tumor (139 CNS, 612 non-CNS) pediatric patients 0-19 y/o (median age 8.2) were analyzed for germline pathogenic variants. Cancer distribution in cohort included all major groups of pediatric solid tumors including sarcoma (229), neuroblastoma (182), CNS tumors (143), retinoblastoma (70), and other rare solid tumors (143). 138 (18% of total) patients reported germline P/LP variants in cancer predisposition genes. 34 (49%) retinoblastoma, 39 (21%) CNS tumors, 28 (15%) neuroblastoma, 28 (12%) sarcoma, 27 (19%) of patients with other rare solid tumors. Moderate- or high-penetrance P/LP variants were detected in DNA damage repair genes (n=24, 3.2% of patients; BRCA1, n=2; BRCA2, n=2; MLH1, n=2; MSH2, n=1; MSH6, n=1; PMS2, n=4). BRCA1 patients: Patient 317, alive, M, neuroblastoma, age 1 at dx, heterozygous tumor status; Patient 467, alive, M, osteoma, age 16 at dx, indeterminate tumor status, APC tested negative, PGF rectal cancer (60s) and PGM breast cancer (40s), AJ-European. BRCA2 patients: Patient 20, alive, F, CNS-astrocytoma, age 0 at dx, heterozygous tumor status, AJ-European; Patient 156, alive, F, neuroblastoma, age 3 at dx, ganglioneuroblastoma, gain of mutation tumor status, mother has BRCA2 mutation.
• Germline testing patients with cancer is critical for distinguishing conditions e.g., NF1 and CMMRD, which can be phenocopies of each other. For example, a child presenting with numerous cafe au lait spots and leukemia may have either condition, but treatment and screening recommendations for proband and family members will differ depending on the germline diagnosis. Besides the known associations of causal germline mutations (e.g. NF1, CMMRD), broad tumor-normal sequencing has revealed novel associations. Some findings likely represent population detection and do not play a role in the pathogenesis of cancer. Population detection is expected, especially in large studies given the frequency of common germline predisposition; BRCA1/BRCA2 mutations are found in 2-3% (1 in 40) individuals of Ashkenazi Jewish ancestry. In pediatric oncology, the main successes of germline-targeted therapies so far have been CMMRD-related tumors responding to immunotherapy, NF1-associated inoperable plexiform neurofibromas responding to selumetinib/NF-1 associated optic pathway gliomas responding to MEK inhibitors. The efficacy of other targeted therapies for example against BRCA1- and BRCA2-associated pediatric tumors (medulloblastomas, typically) remains preclinical and theoretical at this time (March 2021).
• Signature biomarkers of HRD (SBS3, SBS8, ID6, SV3): No signature analysis undertaken in this paper.
• Other biomarkers of HRD (e.g. R-loops, DNA-RNA hybrid structures): None mentioned.

3. St Jude: Zhang et al (2015)

• Patient characteristics (Cohort/Individual patient descriptions, and available demographic and clinical characteristics): 1120 pan-cancer pediatric patients 0-19 y/o (median age 6.9 years) were analyzed for germline pathogenic variants by WGS (595 patients) and WXS (456) or both (69). Cancer distribution in cohort included leukemia (588), CNS tumors (245), and non-CNS solid tumors (287). 95 (8.5% of total) patients reported germline P/LP variants in cancer predisposition genes (565 analyzed). A total of eight children had germline mutations in the adult-onset cancer-predisposition genes BRCA1, BRCA2, and PALB2. The spectrum of cancers observed in these children included leukemia, CNS tumors, neuroblastoma, osteosarcoma, and rhabdomyosarcoma. Although biallelic mutations of BRCA1/2 and PALB2 are known to cause Fanconi’s anemia, there were no germline mutations or deletions involving the second alleles of these genes in any of the affected patients. BRCA1, BRCA2, and PALB2 are not normally examined in children because they are thought to be predisposition genes for adult cancer. Magnusson et al described a high prevalence of childhood cancer in families with germline BRCA2 mutations, and Brooks et al reported 20 cases of pediatric cancer among 379 families, members of which had a mutation in either BRCA1 or BRCA2. These reports suggest that pathogenic mutations in BRCA1 and BRCA2 are more common in pediatric cancer than has been recognized previously and that they potentially underpin a broader spectrum of cancer phenotypes.
• Signature biomarkers of HRD (SBS3, SBS8, ID6, SV3): No signature analysis undertaken in this paper.
• Other biomarkers of HRD (e.g. R-loops, DNA-RNA hybrid structures): None mentioned.

4. Michigan: Mody et al (2015)

• Patient characteristics (Cohort/Individual patient descriptions, and available demographic and clinical characteristics): 91 relapsed, refractory or rare pan-cancer pediatric patients 0-22 y/o (median age 11.5 years) were analyzed for germline pathogenic variants by WXS and RNAseq. Cancer distribution in cohort included hematological (28) and solid tumor (63) malignancies. 9 (9.8% of total) patients reported incidental germline P/LP variants in cancer predisposition genes. 1 patient (patient #21) with melanoma and established BRAF mutation (p.V600E) had a BRAF-associated protein 1 BAP1 mutation (p.D567X). BAP1 mutations are implicated in cancer predisposition for malignant mesothelioma, atypical melanocytic tumors, uveal melanoma, and cutaneous melanoma. The patient had a family history of cancer, including her mother who was diagnosed with ovarian cancer when she was 44 years of age.
• Signature biomarkers of HRD (SBS3, SBS8, ID6, SV3): No signature analysis undertaken in this paper.
• Other biomarkers of HRD (e.g. R-loops, DNA-RNA hybrid structures): None mentioned.

5. Baylor: Parsons et al (2016)

• Patient characteristics (Cohort/Individual patient descriptions, and available demographic and clinical characteristics): 150 patients (mean age 7.4 years) had tumor and matched normal samples undergo WES through BASIC3 (Baylor College of Medicine Advancing Sequencing in Childhood Cancer Care) with CNS (56) and non-CNS (94) representation. 15 cases (10%) revealed diagnostic germline findings – 13 P/LP mutations in cancer susceptibility genes were discovered. 2 patients reported germline P/LP dominant mutations in BRCA1, and 1 with BRCA2 variant, all of which were inherited from parents. There was no evidence of loss of heterozygosity (LOH) in any of the 3 tumors. A number of children carried pathogenic mutations in adult-onset cancer genes, including the above 3 (2%) of 150 participants with BRCA1 or BRCA2 mutations. Polymorphisms in BARD1 which encodes a BRCA1-interacting protein has been associated with neuroblastoma risk, but 1 study reported no increase in childhood cancer diagnosis in breast cancer kindreds with BRCA1/2 mutations (Brooks, 2006). The 3 different tumor types, absence of LOH, and no statistically significant increase in BRCA1/2 mutations compared with a nonpediatric cancer cohort would suggest that this may be an incidental finding.
• Tumor type (Histology and molecular classification into consensus subgroups): Patient #176888: BRCA1, heterozygous mutation in autosomal dominant disorder, c.68_69delAG, p.E23Vfs, diagnosed with neuroblastoma, family history of breast cancer, VAF (N/T) of (0.43/0.53); Patient #1185828: BRCA1, heterozygous mutation in autosomal dominant disorder, c.697_698del, p.V233fs, diagnosed with anaplastic medulloblastoma, family history of breast cancer, VAF (N/T) of (0.5/0.35); Patient #364059: BRCA2, heterozygous mutation in autosomal dominant disorder, c.1278delA, p.D427fs, diagnosed with Ewing sarcoma, family history of breast cancer and ovarian cancer, VAF (N/T) of (0.6/0.34)
• Signature biomarkers of HRD (SBS3, SBS8, ID6, SV3): No signature analysis undertaken in this paper.
• Other biomarkers of HRD (e.g. R-loops, DNA-RNA hybrid structures): None mentioned.

6. Columbia: Oberg et al (2016)

• Patient characteristics (Cohort/Individual patient descriptions, and available demographic and clinical characteristics): 101 patients (median age 8.0 years) with 120 cases of cancer (85 primary disease, 35 relapse/refractory disease) underwent WES of matched tumor normal samples and RNAseq of tumor for identification of sequence variants, fusion transcripts, relative gene expression, and copy number variation. Cases were predominantly pediatric patients with solid tumors (64%); sarcoma (n=17) was the most common diagnostic sub-category followed by brain tumors (n=16). 90 patients had germline tissue sequenced and 18 (20%) were found to have clinically impactful germline alterations in cancer predisposition genes: 11/57 patients with solid tumors (19%) and 7/33 patients with hematologic conditions (21%). BRCA1 germline mutations were detected in brain as well as hepatic cancers (see below).
• The outcomes for children with cancer have steadily improved to the present time when more than 80% of all pediatric oncology patients are cured. Nonetheless, cancer remains the leading cause of disease-related death in children. Moreover this success comes at a price; two-thirds of all survivors have some long-term sequelae attributable to their treatment. The requirement to further improve existing outcomes and to decrease toxicity underscores the need for the current national initiative in precision medicine to include pediatric oncology patients.
• Tumor type (Histology and molecular classification into consensus subgroups): Patient #14-59462: BRCA1, frameshift mutation, c.68_69delAG, p.E23Vfs, diagnosed with nested stromal epithelial tumor of the liver (rare), at age 18 y/o (M); Patient #1185828: BRCA1, frameshift mutation, c.5587_5594delGTCAGCACT, p.V1863Lfs*35, diagnosed with ependymoma, at age 17 y/o (F)
• Signature biomarkers of HRD (SBS3, SBS8, ID6, SV3): No signature analysis undertaken in this paper.
• Other biomarkers of HRD (e.g. R-loops, DNA-RNA hybrid structures): None mentioned.

7. Australia: Wong et al (2020)

• Patient characteristics (Cohort/Individual patient descriptions, and available demographic and clinical characteristics): 247 Australian patients' 252 tumors representing high-risk pediatric cancer cases (<30% chance of 5-year survival after diagnosis) underwent tumor and germline WGS and RNAseq. Tumors were grouped into five main categories: CNS (n=92), sarcoma (n=62), non-sarcomatous extracranial solid (n=35), neuroblastoma (n=20) and hematological malignancy (n=43). Patients were enrolled at various diseases stages: initial diagnosis (47.2%), relapse (41.3%) or refractory (9.9%). A few (1.6%) were treatment-induced secondary tumors. 16.2% of patients (40/247) were identified to have pathogenic cancer-predisposing variants, a higher rate than previously described via WXS. These 40 patients harbored 52 pathogenic germline cancer predisposition alleles. Germline findings were made in all major pediatric cancer types, from a rate of 12.9% of patients with sarcoma to 21.9% in solid tumors. Strikingly, the risk variant was already known tot he family in only 14/40 (35%) patients and the variant was known in that individual in 11/40 (27.5%) patients. The potential availability of new targeted therapies and immunotherapies has ushered in the era of precision medicine, where the molecular profile of a tumor helps guide patient management. The hypothesis is that matching treatments to molecular changes in the tumor results in more effective cancer control and less long-term treatment-related side effects. There are unique challenges to personalizing pediatric cancer treatment: firstly, only 45% of pediatric cancer driver genes are shared with adult cancers, suggesting that new therapeutic agents are required for pediatric cancer; second, pediatric cancers are often driven by structural variants that can be challenging to identify and target; and third, many new targeted drugs lack dosage guidelines and efficacy data in children.
• Tumor type (Histology and molecular classification into consensus subgroups): 4 out of the 247 patients were identified to have BRCA2 mutations. Represented cancer types included 2 CNS, 1 NBL, and. 1 solid tumor.
• Signature biomarkers of HRD (SBS3, SBS8, ID6, SV3): All tumors with BRCA2 SNV/small indel mutations exhibited the associated SBS3 exposures.
• Other biomarkers of HRD (e.g. R-loops, DNA-RNA hybrid structures): None reported.

The origins of hypermutation in pediatric cancers

There is no agreed on definition of “hypermutation”. Larger unbiased cohorts are needed to define cut-offs for hypermutation, not only to understand its frequency across cancer but also to uncover some common mutagenic processes associated with hypermutation, “whose temporal order is important but typically unknown”. In general, understanding early drivers of hypermutation may be useful for predicting the cancer’s evolutionary trajectory and accumulation of additional mutations (Shlien) but this is difficult given that the sheer number of passenger variants can obscure true drivers – this may be easier in pediatric cancer.

A goal: To better understand the characteristics, biomarkers, and corresponding treatment options for hyper-mutated cancers (SNV/Indel/SV-wise), specifically in the pediatric context. There was a “Comprehensive Analysis of Hypermutation” conducted for both adult and pediatric cancer in Cell 2017 by Adam Shlien, which follows up an analysis of bMMRD in pediatric patients by Shlien in 2015.

Shlien’s 2017 study analyzed 2,885 childhood cancers for hypermutation (panel sequencing) and found 160 cases with >10 Mut/MB (which was taken as hypermutation) in types not typically associated with elevated numbers of mutations including sarcomas, germ cell tumors, nephroblastomas, neuroblastomas. >100 Mut/Mb (which was taken as ultrahypermutation) and only included three tumor types malignant gliomas, colorectal cancers, and leukemias/lymphomas.

Biallelic DNA mismatch-repair deficiencies (bMMRD): Adam Shlien and Uri Tabori (2015) analyzed genomes of 17 inherited cancers (12 patients), and 17 non-neoplastic tissues from patients for which the matched tumor was not available (18 patients). bMMRD is also known as CMMRD and is a rare disorder that greatly increases the risk of developing one or more types of cancer in children and young adults. The cancers most common in CMMRD syndrome is cancers of the colon and rectum, brain, and blood. People with CMMRD syndrome may develop multiple noncancerous growths in the colon that are likely to become malignant over time. Many people with CMMRD syndrome develop features similar to those occurring in neurofibromatosis type 1 (NF1) i.e., cafe au lait pigmentation. It is relatively rare with >200 individuals reported in the literature. Shlien and Tabori found that of the 17 bMMRD cancers, all 10 ultra-hypermutated malignant brain tumors from children with inherited biallelic mismatch repair deficiency (the first component which prevents point mutations during replication) reported exhibited massive numbers of substitution (>250/Mb), which was greater than all childhood (typically ~1/Mb) and most adult cancers (typically ~10/Mb) with >7000 analyzed. The few adult cancer types with >100/Mb include LGG (low grade glioma), COREAD (colorectal adenocarcinoma), STAD (stomach adenocarcinoma), LUAD (lung adenocarcinoma), and LUSC (lung squamous cell carcinoma). Mutation frequencies were calculated in 1-Mb bins and there was no evidence of localized hypermutation (kategies). All ultra-hypermutated bMMRD cancers acquired early somatic driver mutations in DNA polymerase e and d (the second component which prevents point mutations during replication). All samples harbored hypermutation at low allelic fraction (<20%) indicating recent and explosive accumulation of mutations after POLE or POLD1 mutation (identified as the driver mutations). Sequential tumor biopsy analysis revealed that bMMRD/polymerase-mutant cancers rapidly amass an excess of simultaneous mutations (~600 mutations/cell division), reaching then plateauing at 20,000 exonic mutations in less than 6 months. bMMRD tumors were devoid of the copy number alterations typically observed in childhood brain cancers and were microsatellite stable. Authors estimate that the probability of observing ultra-hypermutation in a child with sporadic, non-bMMRD was <10e-13. “To our knowledge, ours (2015) is the first report of a tumor genome profile that can be used to infer germline mutational status. This ultra-hypermutated phenotype occurs rapidly and is limited to substitutions, making it distinct from other tumors which carry a variety of mutation types that typically accumulate in a slow and stepwise manner to provide sufficient clonal advantage.”

Other inherited germline predisposition syndromes besides bMMRD (mutations in one of the four mismatch repair genes – MLH1, MSH2, MSH6, or PMS2) which was studied above. Other such syndromes might include Lynch syndrome, and polymerase proofreading-associated polyposis (PPAP). Although data gathered by the bMMRD consortium indicates all malignant bMMRD cancers are hypermutant, it is not known if the same is true for Lynch syndrome and PPAP. Li et al in 2019 produced the first genomic landscape of a bMMRD-like Lynch syndrome case in a single 16 year old here, in which it resulted in ultrahypermutated GBM and similarly a microsatellite stable genome. Other reports of bMMRD mimicry have surfaced, including a 13 year old who developed an ultrahypermutated malignant glioneuronal tumor (two PMS2 variants) and a nonhypermutated glioma, with distinct somatic second hits in the remaining allele) by Wu (2014). However, Wu did not report the mutational signature and potential somatic mutations in replication DNA polymerase genes. Lynch syndrome is a single heterozygous pathogenic germline variant in one of the MMR genes and was previously affiliated with only colorectal cancer. This case illustrates the possibility that Lynch syndrome patients can develop early onset brain cancer, and reiterates the importance of germline testing to differentiate bMMRD from Lynch syndrome as the two predispositions have different prognoses, surveillance recommendations, and inheritance risks. It remains unclear which ICI marker (TMB, PD-1/PD-L1 or MMRD) serves as a better indicator of response for GBM patients.

Therapy-induced – better map out what kind of therapies lead to what kind of patterns of hypermutation (SNV/SV/Indel).

Carcinogens (more common in adult cancers than pediatric cancers) such as UV light causing melanoma, tobacco smoke causing lung cancer.

Data Analysis

SV Analysis

Undertake de novo signature extraction (SigProfiler) and visualization (chromoscope) for each batch of patients with available SV calls.

By tumor type:

• Neuroblastoma from KIDSFIRST (20 samples)
• B-Acute Lymphoblastic Leukemia from KIDSFIRST (81 samples)
• Pan-cancer from Thatikonda (785 samples)

By DNA repair deficiency of interest:

• HRD samples [germline BRCA1/2] from the following cancer center’s papers of germline testing in pediatric tumors – 1. DKFZ: Waszak (2018), 2. MSKCC: Miala (2021), 3. St Jude: Zhang et al (2015), 4. Michigan: Mody et al (2015), 5. Baylor: Parsons et al (2016), 6. Columbia: Oberg et al (2016), 7. Australia: Wong et al (2020)
• MMRD [hypermutated samples >10 Mut/Mb and/or with germline MLH1, MSH2, MSH6, or PMS2] from Shlien paper