Molecular heterogeneity of pediatric choroid plexus carcinomas determines the distinctions in clinical course and prognosis

Abstract

Background

Choroid plexus carcinomas (CPCs) are rare aggressive pediatric tumors of the brain with no treatment standards. Genetic profiling of CPCs is often confined to possible association with Li–Fraumeni syndrome, though only about a half of CPCs develop from syndromic predispositions. Whole-chromosome gains and losses typical of CPCs reflect genomic instability of these tumors, but only partially explain the aggressive clinical course.

Methods

This retrospective study enrolled 25 pediatric patients with CPC, receiving treatment between January 2009 and June 2022. Molecular-genetic testing was performed for 20 cases with available tumor tissue and encompassed mutational status, chromosomal aberrations, and gene expression profiles. We analyzed several factors presumably influencing the outcomes, including molecular profiles and clinical parameters. The median follow-up constituted 5.2 years (absolute range 2.8–12.6 years).

Results

All studied CPCs had smooth mutational profiles with the only recurrent event being TP53 variants, either germline or somatic, encountered in 13 cases. Unbalanced whole-chromosome aberrations, 
notably multiple monosomies, were highly typical. In 7 tumors, chromosome losses were combined with complex genomic rearrangements: segmental gains and losses or signs of chromothripsis. This phenomenon was associated with extremely low 5-year survival: 20.0 ± 17.9% vs 85.7 ± 13.2%; P = .009. Transcriptomically, the cohort split into 2 polar clusters Ped_CPC1 and Ped_CPC2 differing by survival: 31.3 ± 17.8% vs 100%; P = .012.

Conclusion

CPCs split into at least 2 molecular subtypes distinguished both genomically and transcriptomically. Clusterization of the tumors into Ped_CPC1 and Ped_CPC2 significantly correlates with survival. The distinction may prove relevant in clinical trials for dedicated and patient-oriented optimization of clinical protocols for these rare tumors.

Choroid plexus carcinomas (CPCs), the rare epithelial tumors derived from the choroid plexus of the brain, account for 0.3%–1% of central nervous system (CNS) tumors in children and adolescents. CPCs mostly arise in children under 3 years and seldom in adults.¹ The fifth edition of the WHO Classification of CNS Tumors ranks CPCs as Grade 3 malignancies which should meet at least 4 of the following 5 essential criteria: high cellularity, loss of papillary growth patterns, necrosis, nuclear pleomorphism, and high mitosis rate.² A major proportion of CPCs harbor pathogenic variants in TP53, either germline (Li–Fraumeni tumor predisposition syndrome, LFS; OMIM: 151623) or somatic.^3–10TP53 mutations have been recognized as adverse prognostic factor in CPC, moreover, zygosity-dependent, with homozygous TP53-mutated tumors showing worse outcomes.⁷TP53-mutated tumors often reveal chromosome instability, up to chromothripsis. TP53-mutated tumors of the choroid plexus are characterized by whole-chromosome gains and losses, with whole-chromosome losses and copy-neutral losses of heterozygosity (LOHs) specifically characteristic of CPC.^7,11,12

DNA methylation profiling is currently considered the most powerful tool for CNS tumor characterization. Choroid plexus tumors split into 3 methylation classes: (1) supratentorial pediatric low-risk choroid plexus tumors—“pediatric A”: choroid plexus papilloma (CPP) and atypical choroid plexus papilloma (aCPP); (2) supratentorial pediatric high-risk choroid plexus tumors—“pediatric B”: CPP, aCPP, and CPC; (3) infratentorial low-risk choroid plexus tumors in adults—CPP and aCPP, “adult.” However, pediatric tumors classified as CPC are almost indistinguishable by their DNA methylation patterns and identify with “pediatric B” methylation class independently of TP53 status, which so far makes the characteristic epigenetic signature clinically irrelevant.¹³ Gene expression profiling supports the principal distinction between CPP and CPC while revealing no distinctions among CPCs.⁷

The choice of management strategies for CPC is challenging—the rarity of these tumors rules out the use of randomized clinical trials for candidate drugs and clinical protocols. This crucial point is reflected by the lack of standard chemotherapy (CT) schemes for CPC, as development of such protocols would require multicenter prospective studies on the efficacy and toxicity of various cytostatic drug combinations. The “realistic” management strategies for CPC involve maximal safe resection followed by adjuvant therapy, with the options including CT and radiation therapy (RT),^1,6,14–17 although the advantages of RT, especially in LFS, are questionable due to very high risks of secondary malignancies.^6,18,19 Several studies identify the extent of surgical resection as a critical prognostic factor, affecting both event-free, and overall survival (respectively, EFS and OS).^1,14,15 Nevertheless, CPC diagnosed in under-3-year-olds reveals no correlation of survival with the extent of surgical resection, as well as metastasis or RT, with TP53 status being the only significant variable to be used as the basis for risk stratification in clinical trials.²⁰

Due to the rarity of choroid plexus tumors, the majority of studies enroll mixed cohorts of patients of variable age and histological type of the tumor. In this study, we focus exclusively on pediatric patients with morphologically verified diagnosis of CPC. The study aimed at comparative assessment of molecular and clinical parameters possibly affecting the outcomes, including mutational landscapes, cytogenetic signs, and gene expression signatures, as well as extent of tumor resection, CT/RT regimens, and heritability status (LFS).

Methods

Patient Enrollment and Sample Collection Procedures

The retrospective study (patient data analysis, molecular tests), carried out on the premises of the Dmitry Rogachev National Medical Research Center of Pediatric Hematology, Oncology and Immunology (a national reference clinical research center), was approved by Local Ethical Review Board and complied with the Declaration of Helsinki. The Center performs nation-wide visualization and histopathological and molecular reference diagnostics for patients (pts) with malignant and non-malignant neoplasms of various localization, including CNS, as well as blood disorders. During the time period from January 2009 to June 2022, 25, 43, and 86 pts were diagnosed with, respectively, CPC, aCPP, and CPP based on neuro-visualization, morphological, and genetic criteria, among approximately 12,000 pts with CNS tumors admitted to the Center. All patients received treatment in regional hospitals specialized in pediatric. For 20 pts, formalin-fixed paraffin-embedded (FFPE) tumor tissue samples were available; for the rest 5 pts, the only studied biological material was peripheral blood as a source of constitutional DNA. In 1 pt, CPC resulted from malignant transformation of aCPP, with FFPE samples of both tumors available for analysis.

Mutational Profiling

The primary analysis involved FFPE tumor tissues (available for 20 pts). Total DNA and RNA were extracted with FFPE RNA/DNA Purification Plus Kits (Norgen Biotek, Canada). Mutation profiles were assessed by high-throughput DNA sequencing using a QIAseq custom gene panel (Qiagen, Germany) with PCR-based targeted enrichment for genes of interest selected by etiological and pathogenetic relevance to pediatric solid tumors; the list of genes is given in Supplementary Table 1. The paired-end sequencing was carried out in a NextSeq 500 instrument with NextSeq 500/550 Mid Output Kit v2.5 (300 Cycles) (Illumina, USA). The reads were aligned to human genome, version hg19, and the variants were identified using smcounter and mutec algorithms.

Genetic testing for TP53 mutation status of non-tumor DNA isolated from white blood cells was applied given (1) a clinically relevant TP53 variant identified in tumor-derived DNA or (2) non-availability of tumor tissue for molecular-genetic testing. Genomic DNA, extracted from peripheral blood samples using AmpliPrime DNA-Sorb-B kits (NextBio, Russia), was used for amplicon Sanger sequencing in ABI 3500xl Genetic Analyzer (Applied Biosystems, USA). In cases of confirmed germline status of the identified TP53 mutation, genetic blood testing of parents of the proband was performed to clarify the mutation history (inherited or de novo) and to screen for LFS carriers in families.

Copy Number Variation Analysis

Comparative genomic hybridization (CGH) on CytoSure Cancer +SNP Array 4 × 180k (OGT, UK) used 1 µg of genomic DNA from FFPE tumor tissue. Data visualization and interpretation were performed in CytoSure software. The tumors were classified based on the profiles of unbalanced chromosome rearrangements as near-diploid (44–46 chromosomes), hyperdiploid (47–52 chromosomes), hypodiploid (34–42 chromosomes), or hypodiploid with complex rearrangements (including chromothripsis and segmental copy number alterations).

Targeted mRNA Expression Analysis

Gene expression profiling with nCounter PanCancer Pathways Panel (NanoString, USA) started with 300 ng of total RNA for each FFPE tumor. RNA hybridization to barcoded probes was followed by automated purification and immobilization of the RNA-probe complexes in a sample cartridge using nCounter Prep Station, scanning of the fluorescent signals at maximal resolution using nCounter Digital Analyzer, and data analysis by nSolver 4.0 software using nCounter Advanced Analysis module (version 2.0.115). The clustering parameters were as follows: distance metric—Euclidean distance, linkage method—average. The list of differentially expressed genes was loaded in the Reactome online tool for gene ontology analysis and pathway mapping.²¹ The transcriptomic profiles for CPC were assessed against a reference cohort of well-characterized ependymomas (ST-EPN-ZFTA, n = 7; PF-EPN-B, n = 4, SP-MPE, n = 3), pilocytic astrocytomas (KIAA1549-BRAF-positive, n = 17; BRAF V600E-positive, n = 5, FGFR1-TACC1-positive, n = 1; MEF2D-NTRK1-positive, n = 1), infant-type hemispheric high-grade gliomas (TRK-rearranged, n = 3), HGNET-BCOR (n = 4), and medulloblastomas (WNT-activated, n = 16; SHH-activated, n = 2; non-WNT, non-SHH-activated, n = 3).

Clinical Management

The management scheme for CPC includes maximally safe tumor resection, CT and RT. Due to the rarity of these tumors, there is no standardized chemoradiation treatment for CPC.

The extent of surgical resection was assessed on the basis of early postoperative MRI and categorized as gross total resection (GTR, no residual tumor), subtotal resection (STR, 50%–90% of the tumor removed) or partial resection (PR, <50% of the tumor removed).

CT with CarbEV backbone of etoposide, vincristine, and carboplatin (6 cycles) demonstrated superior efficacy compared to CycEV of etoposide, vincristine, and cyclophosphamide in CPT-SIOP-2000 trial.²² Accordingly, CarbEV regimen was considered as a reference scheme of CT for subsequent comparison analysis. Miscellaneous CT regimens were prescribed by regional therapists in compliance with CPT-SIOP-2009 (NCT01014767) terminated early (2019) due to insufficient enrollment.²³

RT was performed in 14 pts (including 10 newly diagnosed cases and 4 pts in relapse); for 7 pts, the decision was to refrain from RT. The RT option (local, to a total dose of 54 Gy in 30 fractions, 1.8 Gy/fraction) was applicable to patients older than 3 years with M0 stage of the disease. Cranio-spinal irradiation to a total dose of 36 Gy in 20 fractions, 1.8 Gy/fraction, with extra 54.4 Gy on tumor bed and 49.6 Gy on metastatic foci, 1.6 Gy/fraction, was applied in metastatic CPC.

Statistics

Statistical processing of the data was accomplished with XLstat software (Addinsoft, USA). Comparison of categorical variable distributions between the groups was carried out by χ ² test with Yates correction. Survival rates were estimated by Kaplan-Meier method with time-to-event comparisons carried out using log-rank tests. Cumulative incidences of relapse/progression (CIRPs) for the groups were compared by Gray’s test. Individual EFS spans were recorded as time from initial MRI till adverse event (relapse or progression of the main disease, second(ary) malignancy, death or date last seen); the time-to-relapse/progression spans were determined as time from initial MRI to disease-related adverse event; OS spans were determined as time from initial MRI till death.

Results

Patient Characterization

The median age at diagnosis was 2.5 years (absolute range 0.6–13 years). The primary foci were localized in lateral ventricles (23 pts) or third ventricle of the brain (2 pts). Metastatic spread of the tumor was diagnosed in 9 pts; localizations of secondary lesions are listed in Supplementary Document 1. CPC was the primary neoplasm in all pts except one, aged 5 years 9 months at diagnosis, with CPC resulting from malignant transformation of a repeatedly resected congenital aCPP. CONSORT diagram showing diagnostic procedures and therapeutic interventions is given in Figure 1.

Landscape of Genetic Variants in CPC

FFPE tumor tissues were available for 20/25 pts. The corresponding DNA samples were tested for genetic variants known as probable molecular drivers of pediatric CNS malignancies (Supplementary Table 1). The only marker recurrently mutated in CPC was TP53, with TP53^mut status detected in 11/20 pts. The identified mutations included 8 single-nucleotide missense variants affecting the DNA-binding domain of p53, 2 splice site variants (affecting both donor and acceptor positions) and a TP53 (NM_001276760.2) c.876 + 312C>G substitution. This variant of uncertain significance corresponds to c.1037C>G p.S346W missense variant in non-canonical TP53 NM_001126113.2 transcript, but spares splice sites (maximum SpliceAI score 0.38 for donor gain). Sanger sequencing of constitutional DNA revealed germline TP53^mut status in 5/11 pts and no mutations in 4/11 pts (implicating somatic TP53^mut status). For 2/11 pts, constitutional DNA was not available for the germline vs somatic specification.

For 5/25 pts, the only available biological material was peripheral blood; of those, the sequencing of coding exons and splice sites of TP53 revealed germline pathogenic variants in 2/5 pts.

Thus, the tests identified germline TP53 mutations in 7 pts (g_TP53^mut-positive, 28%), somatic mutations in 4 pts (s_TP53^mut-positive, 16%), TP53 mutations of unknown status in 2 pts (x_TP53^mut-positive, 8%), and unaltered TP53 sequences in 12 pts (TP53^wt, 48%). Of 11 pts with TP53 variant identified in tumor DNA, 10 pts had variant allele frequency (VAF) above 50% (63%–95%) and 1 pt had VAF as low as 13%. The identified TP53 variants (designated by standard nomenclature, varnomen.hgvs.org) are listed in Table 1, with status and germline histories specified (if applicable). All other identified genetic variants were non-recurrent and of uncertain clinical significance (Supplementary Table 2).

Landscape of Unbalanced Chromosome Rearrangements and Heterozygosity Losses

The analysis of unbalanced chromosome rearrangements divided CPCs into 4 non-intersecting groups: near-diploid (n = 2), hyperdiploid (n = 3), hypodiploid (n = 6), and hypodiploid with complex chromosome rearrangements (n = 7). In 2/20 samples, DNA quality was too low for CGH.

Of 2 pts with near-diploid chromosome profiles, Pt #4 had balanced tumor genome and Pt #22 had monosomies 3 and 10, and neither of them displayed chromosome 17 rearrangements, though Pt #4 harbored a somatic splice site variant in TP53 (Chr17). Hyperdiploid CPCs (3 pts) were distinguished by multiple trisomies (Pt #17: +7,+11,+14,+18; Pt #23: +1,+5,+8,+9,+12,+14,+20; Pt #25: +1,+2,+4,+10,+13,+14,+18,+20,+21), accompanied by multiple monosomies in 2 cases (Pt #23: −3 −7,−11,−13,−15,−18; Pt #25: −15,−16,−17,−19,−22), whereas Pt #17 presented with tetrasomy 12. Generic features of hyperdiploid CPCs included the increase in Chr14 copy number and the lack of nucleotide substitutions in TP53.

Hypodiploid tumors with or without extra segmental chromosome rearrangements or signs of chromothripsis were the predominant cytogenetic type of CPC for the studied cohort (13 cases out of 20). The group of hypodiploid tumors without extra cytogenetic signs corresponding to complex chromosome rearrangements (6 pts) harbored multiple monosomies including recurrent losses of chromosomes 3, 6, 9, 11, 15, 16, 17, and 22. None of the patients harbored a trisomy or shared a profile of unbalanced aberrations with other tumors within the group. TP53 was affected in 5/6 cases; the alterations included single-copy deletions due to monosomy 17 (n = 5) and missense mutations (n = 3). The group of hypodiploid tumors with complex chromosome rearrangements (7 pts) harbored recurrent monosomies 3, 6, 11, 13, 16, 17, and 22. The encountered complex rearrangements of various nature (chromothripsis, local segmental chromosome rearrangements, copy-neutral LOH) most often mapped to 2, 4, 5p, 10, 15q, and 18q chromosomes. By contrast with hypodiploid CPCs without extra rearrangements, these tumors presented with whole-chromosome duplications (+1, +7, +20, +21) or increased copy number of individual chromosome arms (20q) or isochromosome formation (iso5p, iso8p). Patients of this group, all except the patient with CPC developed by malignant transformation of aCPP, had hemizygous mutations in TP53 (unbacked by normal allele due to monosomy 17). Of note, TP53 variants (somatic in 3 pts) could be associated with chromothripsis on chromosomes 2 and 15 (2 pts) or copy-neutral LOH on chromosomes 1, 2, 4, 10, 12, 14, 16, 20, 21, 22 (1 pt). Cytogenetic profiles for the entire cohort are summarized in Figure 2.

Targeted Gene Expression Profiling

The targeted transcriptomic profiling of pediatric CPCs identified 2 unique clusters, Ped_CPC1 and Ped_CPC2, characterized by fundamentally different expression of genes involved in major oncogenic signaling pathways (Figure 3A). Ped_CPC1 cluster encompassed 11 pts, mostly (8/11) with mutated TP53 and/or hypodiploid genome with complex chromosome rearrangements (P = .021), whereas Ped_CPC2 (9 pts) encompassed lower proportion of TP53 mutated tumors (3/9, P = .190) and none of the cases with complex chromosome rearrangements (P = .021). CPCs arising in the context of LFS were tightly clustered with sporadic CPCs harboring somatic variants in TP53, which indicates the lack of unique molecular distinctions for such tumors arising in patients with inherited or de novo germline pathogenic mutation in TP53. Of note, Pt #10 with TP53 variant initially considered to be of uncertain clinical significance (TP53 (NM_001276760.2) c.876 + 312C>G) had hypodiploid tumor with complex chromosome rearrangements, affiliated to Ped_CPC1 expression cluster. Considerable similarity of this tumor to LFS-associated CPCs revealed by targeted transcriptomic profiling increases the probability of oncogenic relevance for this variant. The split of CPCs into 2 expression clusters was robust (ie, reliable and reproducible) enough to stay preserved upon combined analysis of the studied CPC samples with other well-characterized CNS tumors (Figure 3C).

Ped_CPC2 cluster was distinguished by hyperexpression of many genes selected for the target panel, whereas none of the panel marker genes was expressed in Ped_CPC1 stronger than in Ped_CPC2 (Figure 3B, Supplementary Figure 1, Supplementary Table 3). A total of 423 genes of those included in the panel revealed increased expression in Ped_CPC2 tumors (at least 2-fold, as assessed by Log2 of normalized levels). Gene-set analysis affiliated these markers to oncogenic signaling pathways PI3K-AKT, MAPK, and Cell_cycle/Apoptosis. Pathway analysis performed using the Reactome online tool revealed enrichment with transcripts involved in Interleukin signaling and PI3K, MAPK, TP53, and NOTCH2 cascades (Supplementary Document 2).

Surgery and Adjuvant Treatment

Surgery outcomes at diagnosis included GTR in 12 pts, STR in 7 pts, and PR in 6 pts. Additionally, 3 cases were subject to repeated intervention affording GTR in 2 pts (one post-STR and one post-PR; Table 1). Thus, GTR and STR were achieved in, respectively, 14 (56%) and 6 (24%) of the cases. In 5 pts (20%), the surgery was limited to PR in connection with large tumor volume, CNS metastasis, tumor hypervascularization, and/or severe clinical complications. Fourteen pts (56%) started CarbEV after surgery as reference CT, whereas the rest received other CT schemes. Four pts still continue with the treatment and are not included in survival analysis. In 3 pts, the CarbEV backbone was interrupted at <6 cycles for reasons including a switch to a different CT scheme due to emergence of second tumor (pilocytic astrocytoma of the cerebellopontine angle in Pt #3) or CT refusal by parents/guardians due to high toxicity of the regimen (hematological toxicity grade 3–4 combined to encephalopathy in Pt #7 and Pt #8). These truncated schemes were classified as miscellaneous CT.

Of 10 pts treated with >6 cycles of reference CarbEV, only 3 pts received RT first-line, though 2 pts (one of them with LFS) received RT as salvage therapy at relapse. Of 11 pts treated with miscellaneous CT regimens, 7 pts received RT first-line (including 2 pts g_TP53^mut-positive and 1 pt x_TP53^mut-positive) and other 2 pts (one of them with LFS) received RT at relapse.

Patient Outcomes and Prognostic Factors

The median follow-up was 5.2 years (absolute range 2.8–12.6 years) from the enrollment. At the time of collecting the follow-up data, 10 pts (48%) had progression within the absolute range of 0.5–4.6 years; of those, 7 pts died (6 pts of main disease and 1 pt of secondary HGG) and 3 pts live: 1 pt (#20) 7.4 years post-diagnosis currently without disease and 2 pts (#8 and #21), respectively, 2.8 and 5.1 years post-diagnosis with stabilized disease. Of 11 pts without main disease progression, 1 pt (#11) died of treatment-related complications, 1 pt with LFS presented with radiation-induced HGG 6.3 years from the primary diagnosis of CPC (#1, alive in critical condition) and the other 9 pts live (8 pts without disease and 1 pt with residual disease). Four patients continue to receive therapy: just completing 6 cycles of CarbEV (Pt #12), receiving the third and the fourth cycles of CarbEV (Pt #13, Pt #23), and the fifth cycle of the CycEV/CarbEV alternating regimen (Pt #25).

The extent of tumor resection (achieving GTR vs STR and PR) was the only significant clinical risk factor for CPC progression as assessed through EFS: 61.2 ± 14.0% vs 14.3 ± 13.2%, P = .024. The lack of beneficial effect of GTR on OS (76.2 ± 12.2% vs 38.1 ± 19.9%, P = .094; Supplementary Figure 2A and B) probably reflects the non-uniformity of subsequent CT/RT schemes. Other conventional risk factors (metastatic spread, CT regimen, RT) had no significant influence on EFS and OS (Supplementary Figure 2C–F).

Molecular-genetic features of the tumors, by contrast, revealed high relevance to the risks of adversity (including tumor progression) and death in patients with CPC. The well-characterized molecular-genetic marker of unfavorable prognosis, mutated TP53, significantly affected 5-year EFS and OS independently of the origin (germline or somatic): 24.2 ± 13.8% vs 70.0 ± 14.5%, P = .027 and 38.2 ± 16.4% vs 90.0 ± 9.5%, P = .016 for TP53^mut and TP53^wt tumors, respectively (Figure 4A and B). Patients with hypodiploid tumors showed decreased 5-year EFS and OS, both 52.0 ± 15.6% vs 100%, although these trends were below the level of significance: P = .112 and P = .217, respectively; Supplementary Figure 2G and H. At the same time, the extremely pronounced genetic instability of tumor cells manifested in the form of hypodiploid genome combined to complex chromosome rearrangements was associated with dramatically reduced survival: 5-year EFS 20.0 ± 17.9% vs 87.5 ± 10.5%, P = .007 and OS 20.0 ± 17.9% vs 87.5 ± 13.2%, P = .009 (Figure 4C and D).

Our study identifies the transcriptomic cluster affiliation of a tumor, determined by targeted expression profiling of genes involved in major oncogenic signaling pathways, as integral genetic indicator and a unifying diagnostic link between TP53^mut status and karyotype. The Ped_CPC1 expression signature was associated with highly aggressive biological behaviors and unfavorable prognosis (5-year EFS 25.0 ± 15.3%, OS 31.3 ± 17.8%), whereas the alternative Ped_CPC2 signature was associated with favorable course (5-year EFS and OS of 100%; P = .003 and P = .012, respectively; Figure 4E and F) and the lack of recurrence (CIRP 0% vs 50.0 ± 16.4%; P = .034; Supplementary Figure 2I). The only adverse event encountered in Ped_CPC2 was the radiation-induced secondary HGG in a patient with LFS.

Molecular Evolution of aCPP Progressing into CPC

The cohort included 1 patient (#18) with CPC resulting from malignant transformation of a pre-existing congenital aCPP. The original tumor was identified by neurosonography at the age of 2 months as a mass in left lateral ventricle, eventually verified as aCPP by morphological examination. At 47 months post-diagnosis, the patient relapsed with a similar mass, once again morphologically verified as aCPP. After another 19 months, the patient presented with new progression of the disease, this time morphologically identified as CPC. The constitutional DNA sequencing revealed no LFS.

Tumor tissue samples of aCPP (collected at relapse) and its derivative malignant CPC were available for comparative molecular profiling. High-throughput targeted sequencing of tumor DNA revealed no genetic variants of any clinical significance (neither confirmed, nor uncertain). The SNP array-based CGH qualified both samples as hypodiploid tumors with complex chromosome rearrangements. The common cytogenetic aberrations included 1p, 3, 4q, 5q, 6q, 8q, and 10q chromosome losses and segmental deletions mapping to 4p15.33p11, 5p15.33p13.3, 5p13.1p11, 6p25.3p22.2, 6p21.31p11.1, 10p12.1p11.1, 12q21.1q21.31, and 18q21.33q23. The malignant transformation was accompanied by major (+1q, +8p, −9, −22q) and subclonal (−11, −13q, −15q, −16, −21q) rearrangements (Figure 5). The change in the cytogenetic status of chromosome 1 is noteworthy, with the short arm (1p) deleted in both samples and a 1q43q44 region lost in aCPP. By contrast, in its derivative CPC, an increase in the long arm (1q) copy number is accompanied by restoration of the normal 1q43q44 copy number albeit apparently through a duplication preserving the hemizygous status (qualified as LOH). The targeted gene expression profiling affiliated #18 aCPP with Ped_CPC1 cluster qualified as prognostically unfavorable; the affiliation was preserved upon malignant transformation of the tumor into CPC (Figure 3A, scarlet subcluster).

After being diagnosed with CPC, the patient underwent radical surgery and received 6 cycles of reference CarbEV CT followed by RT. The patient is currently in full sustained remission over a 4-year follow-up.

Discussion

Tumors of the choroid plexus are rare neoplasms of varying malignant potential, with CPCs being the worst. The majority of CPCs (estimated 72%) arise at the age under 20.¹ Due to the rarity of choroid plexus tumors, studies on their clinical and biological aspects usually enroll mixed cohorts of patients differing by age and histological type of the tumor. Our study enrolled exclusively pediatric patients with CPC.

Several known clinical and genetic risk factors for CPC, determining the unfavorable course, include post-surgical residual tumor tissue, metastatic disease, and LFS.^1,3,20 The benefit of GTR (as compared to STR and PR) on the survival has been demonstrated,^1,14,24 and our data support this point, at least with regard to EFS. The value of “aggressive” surgical intervention was challenged by the prospective multicenter Choroid Plexus Tumor Study CPT-SIOP-2000, where the presence of residual tumor tissue after surgery did not interfere with survival or risks of relapse provided a well-standardized CT. Undoubtedly, the highly effective systemic CT is prerequisite for neutralization of the prognostic significance of extent of surgical resection in CPC; thus, advanced standardization of CT schemes for CPC is endorsed in terms of neurosurgical risk alleviation.²²

Until recently, the choice of CT backbone for scheduled treatment of CPC was largely empirical. The CPT-SIOP 2009 trial (NCT01014767) supposed a 4-armed randomization for CT protocols using cytotoxic drugs of profoundly different action mechanisms.²³ A SJYC07 prospective study used a risk-stratified, multimodal treatment scheme with non-myeloablative HD MTX-based induction therapy and consolidation with RT or further CT. The authors conclude that non-myeloablative MTX-containing CT is an effective and well-tolerated treatment option.²⁰ CPT-SIOP-2000, a prominent CPT registry and interventional prospective clinical trial of CT for CPC, identified CarbEV regimen as the most effective, feasible, and tolerable scheme. In this study, GTR significance was outperformed by the impact of CT.²² In our study, the CarbEV scheme was applied in about one-half of the cohort, but showed no advantage compared with other regimens, probably due to significant variability of the latter.

RT, generally an important option in CNS tumor management, has no established value for CPC. RT is not in the list of standard options for CPC and is applied selectively, chiefly in patients with metastatic disease and/or residual tumor volume after surgery and CT.^17,22 Liu et al. show that adjuvant RT has no unambiguous advantages in the treatment of younger children with CPC.²⁰ Bahar et al. retrospectively evaluate the effectiveness and sequelae of RT in patients with LFS-associated CPC, applied first-line or at relapse, to reveal major disadvantages of RT in such patients, presenting with either CPC progression or radiation-induced secondary malignancies in most cases.¹⁹ Among our patients with germline mutations in TP53, RT was applied in 2 pts first-line (#1 and #5) and in other 2 pts at relapse (#2 and #9). The outcomes were adverse: 2 cases of the main disease progression and 2 cases of secondary radiation-induced HGG (including a previously reported clinical case of disease-free survival for 6.3 years post-diagnosis).²⁵ Nevertheless, despite the lack of survival benefits at the cohort level, for a subcohort of patients with somatic TP53 mutations or TP53^mut-negative, RT afforded full sustained remission in 6/8 pts and stabilization in 1/8 pt (median follow-up 5.1 years). Thus, identification of TP53 variants and determination of their status (germline or somatic) provide relevant predictors for RT effectiveness.

TP53 mutation status is a well-established prognostic factor in CPC. Although malignant tumors of the choroid plexus are typical for LFS, the majority of CPC cases arise sporadically without links to tumor predisposition syndromes. At that, sporadic CPCs often (in up to 50% of the cases) present with somatic TP53 variants, which supports the pathogenetic significance of TP53 status in CPC.³ The prevalence patterns of TP53 mutations in our cohort show full consistency with the published evidence: 48% TP53^wt, 28% with germline mutations, 16% with somatic mutations, and 8% with TP53 mutations of unknown origin. All germline TP53 variants identified by us in this study were missense mutations in the p53 DNA-binding domain-encoding sequence, whereas somatic mutations included both missense and splice site variants. Finding a TP53 mutation in tumor tissue was associated with poor outcome independently of the mutation origin (germline or somatic). Remarkably, in all cases except Pt #4, the TP53 variants had VAF 63%–95% corresponding to LOH through the loss of Chr17. As demonstrated by Merino et al., survival rates in patients with TP53-mutated CPC correlate with the non-redundant TP53 copy number: the loss of normal allele leads to a catastrophic drop in survival.⁷ Investigating the global DNA methylation profile, Pienkowska et al. identified 2 groups of CPC: containing homozygous TP53-mutated cases either harboring both heterozygous TP53^mut and TP53^wt tumors. The patient’s outcomes were unequal between these groups, demonstrating the worst one in the first group.²⁶ The only patient with low TP53^mut VAF in our cohort is alive without symptoms of the disease for 6.1 years post-diagnosis.

By CPC karyotypes, the studied cohort roughly splits into 2 major groups: (1) hyperdiploid, with multiple whole-chromosome gains, and (2) hypodiploid, with multiple monosomies. The group of hyperdiploid CPCs is cytogenetically similar to benign tumors of the choroid plexus. Merino et al. studied prognostic value of the ploidy in CPC to observe no correlation with survival.⁷ Ruland et al. report adverse effects of 12q losses on OS in CPC and also identify multiple short genomic regions subject to copy number alterations associated with poor prognosis.¹¹ Our study reveals segmental chromosome rearrangements in 7 patients. Along with signs of chromothripsis, such events can be regarded as indicators of the ultimate genomic instability. Of note, 6/7 pts harbored either germline or somatic mutations in TP53, and the only TP53^wt patient was unique in the developing of CPC through malignant progression of aCPP. Overall, hypodiploid CPCs had worse outcomes, especially so when harboring complex rearrangements associated with extremely poor survival.

The transcriptomic and epigenetic diversity of CPCs was previously described by Merino et al.⁷ The tumors were distributed into 2 clusters differing in representation of TP53-mutated tumors and partially reproducing the ploidy-based subdivision. One cluster, enriched in hypodiploid tumors, was characterized by elevated expression of genes involved in cellular metabolism, cellular signaling, and cell migration pathways. The other cluster comprised predominantly hyperdiploid and diploid tumors, characterized as over-proliferative and immune-infiltrated according to the expression profiling data. At that, such gene expression-based classification was of no apparent prognostic value.⁷ Cornelius et al. revealed elevated expression of PDGFRB, FGF2, KRAS, MAP2K2, HDAC3/8, and MLST8, indicative of MAPK, mTOR, and HDAC signaling pathway activation, in an infant with CPC. These data provided the grounds for non-specific therapy with a cocktail of inhibitors (sirolimus, thalidomide, sunitinib, and vorinostat) affording sustained response and long-term survival.²⁷ The clinical effect supports the pathogenetic significance of the activated signaling pathways and plausible identification of the tumor with the “hyper-proliferative” expression subgroup.

In the current study, we obtained a convincing clusterization of CPCs into Ped_CPC1 and Ped_CPC2 by using a selective transcriptomic approach. The choice in favor of the targeted approach was made because of the reliability of FFPE-tissue derived mRNA quantification by the NanoString technology. Relatively short sequences (up to 100 bases) required for hybridization of the probes allow successful investigation of both intact and fragmented mRNA molecules. Previously published data²⁸ and our own experiments carried out on various tumor types (including HGG and neuroblastoma) revealed no differences in the results of transcripts quantification in mRNA samples extracted from FFPE and snap frozen tumor tissue (unpublished data).

Ped_CPC1 cluster was enriched with TP53-mutated and hypodiploid tumors, although both categories were non-confined to this cluster. Meanwhile, all hypodiploid CPCs with complex rearrangements encountered in this study fell into Ped_CPC1 cluster. The other cluster, Ped_CPC2, was heterogeneous in terms of ploidy and characterized by elevated expression of multiple genes involved in chief proliferative signaling pathways and immune response regulation. Despite the “hyper-proliferative” molecular profiles of Ped_CPC2 tumors, the cluster showed enhanced survival rates indicating lower aggressiveness of the tumors in both biological and clinical aspects. Furthermore, none of the cases assigned to Ped_CPC2 showed disease progression or recurrence, although one of the patients, with LFS, developed secondary HGG in the aftermath of RT.

Conclusion

The study demonstrates considerable molecular heterogeneity of pediatric CPCs, involving large chromosome rearrangements, mutations in TP53 (the only recurrently mutated marker), and gene expression profiles of the tumor. The analysis of transcriptomic signatures in CPC appears to provide an integral diagnostic tool, with each cluster showing enrichment with specific patterns of genetic rearrangement. Overall, the impact of molecular features on the outcomes exceeded that of the conventional risk factors: patients with Ped_CPC2 tumors demonstrated superior survival on the current therapeutic regimens, although we should be careful considering de-escalation of the anti-tumor treatment. By contrast, patients with Ped_CPC1 tumors had dismal outcome despite of the modern multimodal treatment and may benefit from innovative and experimental approaches.

Funding

The study was supported by Foundation for support and development in the field of pediatric hematology, oncology and immunology “Science for Children.”

Conflict of Interest statement: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Authorship statement: M.Z., A.V., and A.D. developed the concept and wrote the manuscript; A.Sh. performed reference morphological examinations; M.Z. and A.D. performed molecular tests; L.Y. directed the diagnosis and analyzed patient data for Li–Fraumeni syndrome; A.V., A.S., L.P., and A.K. directed patient care and clinical decision-making; K.V., N.U., and A.D. performed statistical analysis; G.N. and A.D. supervised data interpretation and manuscript writing. All authors critically read the manuscript and approved the final version.

References

Author notes