Non-RB1 germline cancer predisposing variants found in retinoblastoma patients

Abstract

Purpose

It is well known that individuals with hereditary retinoblastoma are at lifelong high risk for developing subsequent malignant neoplasms (SMN). However, the role that non-RB1 germline variants play in tumorigenesis and SMN risk has not yet been studied. The purpose of this study is to report the frequency and spectrum of non-RB1 germline cancer predisposing variants in individuals with retinoblastoma (RB).

Methods

Retrospective data collection from institutional electronic medical records of 94 individuals seen at our institution with personal history of retinoblastoma, who had undergone next-generation sequencing germline analysis.

Results

The prevalence of individuals with cancer predisposition was 57% (54/94). Of these individuals, 76% (41/54) had a pathogenic/likely pathogenic (P/LP) variant only in the RB1 gene, 9% (5/54) harbored a P/LP variant only in a non-RB1 gene, and 11% (6/54) had both. No difference was found between patients with and without non-RB1 variants when comparing demographic and clinical characteristics, including time to SMN. Variants were found in 7 different genes, with only 1 variant repeating 3 times.

Conclusion

In this small cohort of patients with retinoblastoma, non-RB1 variants did not appear to augment tumorigenesis or disease progression. Larger studies are required to determine associations between specific variants and development of SMN.

Introduction

It is recognized that individuals with hereditary retinoblastoma (OMIM#180200) are at lifelong high risk for developing subsequent non-ocular malignancies.1, 2, 3, 4 Germline RB1 variants,⁴ radiation,^2,5 systemic chemotherapy,^1,5 and young age at treatment are important risk factors.⁶ Pan-cancer studies have shown that between 8% to 18% of individuals with retinoblastoma harbor pathogenic or likely pathogenic (P/LP) variants in cancer predisposition genes and that 34% to 59% of those variants may be unrelated to the primary cancer, depending on the study population.7, 8, 9 In fact, in an earlier report of individuals with retinoblastoma, we found 5 individuals who had pathogenic variants in cancer predisposing genes other than RB1.¹⁰ If and how these non-RB1 germline variants influence tumorigenesis, disease progression, and subsequent malignant neoplasms (SMN) has not yet been studied. The purpose of this study is to report the frequency and spectrum of non-RB1 germline cancer predisposing variants from individuals with hereditary (HR) and non-hereditary retinoblastoma (nHR) seen at our institution.

Materials and Methods

Retrospective data collection from institutional electronic medical records of 94 individuals seen at Memorial Sloan Kettering Cancer Center (MSKCC) with a personal history of retinoblastoma (RB), who had previously undergone next-generation sequencing (NGS) germline analysis using the MSK-IMPACT (Integrated Mutation Profiling of Actionable Cancer Targets) platform (MSKCC; NCT01775072).¹¹ Testing was offered to all individuals undergoing enucleation and those who for clinical reasons were considered to benefit from broad germline testing. The study was approved by MSKCC’s Institutional Review Board/Privacy Board.

MSK-IMPACT is an FDA approved, hybridization capture-based NGS panel designed to identify known P/LP germline variants in 90 cancer-associated genes5, 6, 7^,11 (the list of genes can be shared upon reasonable request). All the germline genes that are analyzed are clinically actionable. Germline variant classification follows the American College of Medical Genetics and Genomics guidelines.¹² Unless otherwise specified, only P/LP germline variants are reported.

Statistical Analysis

Patients' characteristics among HR, nHR, RB1, and non-RB1 groups were analyzed using χ² test, Fisher’s exact test, or Wilcoxon rank sum as appropriate.

Results

Description of the population

In this cohort of 94 individuals with retinoblastoma, 83 (88%) were ≤18 years and 11 (12%) were >18 years.

Participants’ demographics are presented in Table 1. There was an even distribution of individuals with HR (52%; n = 49) and nHR (48%; n = 45).

Table 1.

Demographics and Rate of non-RB1 variants in the HR and non-HR groups

Characteristic	Hereditary, n = 49a	Non-hereditary, n = 45a	Overall, N = 94a
Sex
Female	22 (45%)	21 (47%)	43 (46%)
Male	27 (55%)	24 (53%)	51 (54%)
Age at time of study (years)	7 (1-45)b	6 (0-24)b	7 (0-45)b
Age at RB diagnosis (years)	1 (0-4)b	1 (0-9)b	1 (0-9)b
Unknown	1	0	1
Race description
Asian-Far East, Indian Subcontinent	7 (16%)	9 (21%)	16 (18%)
Black or African American	6 (14%)	6 (14%)	12 (14%)
Other	4 (9%)	4 (9%)	8 (9%)
White	27 (61%)	24 (56%)	51 (59%)
Unknown	5	2	7
Ethnicity description
Non-Spanish; Non-Hispanic	40 (83%)	35 (80%)	75 (82%)
Spanish NOS; Hispanic NOS, Latino NOS	8 (17%)	9 (20%)	17 (18%)
Unknown	1	1	2
Laterality
Bilateral	41 (84%)	0 (0%)	41 (44%)
Trilateral	1 (2%)	0 (0%)	1 (1%)
Unilateral	7 (14%)	45 (100%)	52 (55%)
RB1 germline variant
Negative	2 (4%)	45 (100%)	47 (50%)
Positive	47 (96%)	0 (0%)	47 (50%)
Germline variant in gene other than RB1	6 12% (95%CI: 5%, 25%)	5 11% (95%CI: 4%, 25%)	P = .86c
CHEK2	1	0
ERCC3	0	1
FANCC	0	1
MITF	1	0
MSH3	1	1
MUTYH	3	0
NTHL1	0	2
Second Germline variant in gene other than RB1:TSC2 c.4570-1G>A	0 (0%)	1 (2%)

CI, confidence Interval; NOS, not otherwise specified.

^an (%).

^bMedian (Range).

^cPearson’s χ² test.

The HR group was defined by those with a germline RB1 variant and/or history of bilateral disease. This included 2 individuals with negative germline RB1 testing and bilateral disease, 7 with positive germline RB1 variants and unilateral disease, and 1 person with trilateral disease and a germline RB1 pathogenic variant. All others were defined within the nHR group.

Of note, hereditary groups refer to the presence or absence of P/LP variants in the RB1 gene. Two more groups are described based on the presence (non-RB1 variant group) or absence (control group) of P/LP variants in genes other than RB1.

Prevalence of non-RB1 P/LP variants

Eleven individuals (12%) had variants in a non-RB1 gene; 6 had HR and 5 had nHR. These represent 12% of all individuals with HR and 11% of individuals with nHR. The rate of non-RB1 variants was not significantly different between the 2 groups (Table 1).

Supplemental Table 1 compares clinical and demographic characteristics of patients with and without non-RB1 variants. No statistically significant difference was found between the 2 groups from the variables analyzed.

Development of secondary cancers

In an attempt to evaluate any impact on long-term outcomes, we looked for any association between having P/LP variants in RB1 and/or non-RB1 genes with the development of SMN (Figure 1). Only 13 individuals had developed a SMN at the time of our study, and 5 developed a third cancer. All individuals who developed SMN except 1 had HR; 3 individuals also had a non-RB1 variant.

Figure 1.

Cumulative incidence of second cancers. Because the age at second cancer is a censored observation, we used survival analysis to estimate the rate of SMN diagnosis in each of the groups and analyze any associations with presence of a non-RB1 variant. The cumulative incidence of SMN in the overall study population is presented in Figure 1. The cumulative incidence of SMN was 48% at 15 years in the hereditary group versus 0% in the non-hereditary group, although the difference was not statistically significant (P = .07). Likewise, there was no evidence of a significant difference in the time to second cancer between patients with and without non-RB1 variants in the overall study population nor within the subgroup of patients with hereditary disease. Tick marks represent censoring times. The subgroup of patients with non-hereditary disease was not analyzed because of the small number of second cancers in that group. Time to SMN was defined as the time from the age at RB diagnosis to the age at the first SMN diagnosis. Patients without SMN diagnoses were censored at the current age. Deaths before the occurrence of SMN were treated as competing risks. The cumulative incidence of SMN was estimated using an Aalen-Johansen estimator,¹³ and the curves were compared using a Gray’s test.¹⁴ Analyses were done using R v.3.6.3.

The mean age of SMN diagnosis was 13 years (12 and 24 years in the HR and non-HR groups, respectively).

Frequency of variants in RB1 and non-RB1 genes

The overall prevalence of individuals with cancer predisposition (individuals with HR and/or those with a non-RB1 variant) was 57% (n = 54), of which 11% (n = 6) had 2 different cancer predisposing variants, 42 (45%) had only an RB1 P/LP variant, and 5 (5%) had only a non-RB1 P/LP variant.

The frequencies of variants in RB1 and non-RB1 genes are presented in Table 1. Only 1 patient had reported variants in 2 non-RB1 genes: a LP variant in NTHL1 and a TSC2 variant, which was later reclassified as variant of uncertain significance.

Spectrum of variants in the non-RB1 group

A total of 9 non-RB1 heterozygous P/LP variants were found in 11 individuals; no one had biallelic variants. The variants were found in 7 different genes, MUTYH, MSH3, NTHL1, ERCC3, FANCC, CHEK2, and MITF, with 1 variant identified in each gene except for MSH3 and NTHL1, in which 2 different variants were found (Supplemental Table 3).

All non-RB1 variants in our study were unique, except for a MUTYH (NM_001128425) variant, c.1187G>A p.(Gly396Asp), which was found in 3 different individuals, all of whom had HR. Among them, 2 developed SMN (osteosarcoma and meningioma at 12 and 32 years, respectively). Only 1 other individual with a non-RB1 variant developed a SMN, a 23-year-old female with HR and a MSH3 variant who developed meningioma at age 20. None of the SMNs seemed to have any correlation with the respective non-RB1 gene identified on each individual, but they all have been previously reported as SMN in individuals with HR (Supplemental Tables 2 and 3).

Similarly, all RB1 variants were different when looking at the subgroup of individuals with HR in the non-RB1 variant group. The spectrum of RB1 variants within that group include 2 frameshift, 1 nonsense, 1 splice site, 1 exon-level deletion (all truncating variants), and 1 missense variant (believed to retain some residual activity and be associated with decreased penetrance15, 16, 17). All variants are spread along the RB1 gene from exon 4 to exon 20.¹⁸

Clinical, demographic, and molecular information of individuals with non-RB1 variants can be found in Supplemental Table 2.

Discussion

This represents the largest cohort of individuals with retinoblastoma in which the prevalence of germline non-RB1 P/LP variants has been reported to date. We found that 12% of individuals with retinoblastoma harbored a P/LP variant in a non-RB1 cancer predisposing gene, independent of their underlying germline predisposition to retinoblastoma. There was no significant difference of prevalence between individuals with HR and nHR (12% and 11%, respectively). Furthermore, our study has an ascertainment bias given by the nature of the IMPACT test, which was designed to be performed in individuals with available tumor sample; therefore, our study population is mainly composed by those who required enucleation (representing fewer than 5% of all individuals with retinoblastoma in our institution). No other treatment modalities or exposures were found to be overrepresented in our study population].

The overall prevalence of individuals with a cancer predisposition in our study is 57%, which, when compared with the prevalence of HR (52%), shows that a small but significant percentage (5%) of individuals with cancer predisposition can be missed if only RB1 directed testing is performed in persons with retinoblastoma. Considering that IMPACT only tests for actionable genes, some of these individuals and their families could be missing from surveillance interventions recommended based on each gene. These percentage is probably even higher given that (1) IMPACT only reports variants classified as P/LP and does not report variant of uncertain significance that may eventually be upgraded to pathogenic, and (2) testing was limited to the 90 gene version of IMPACT, but the current number of known cancer predisposition genes is significantly higher and continues to expand. Overall, this represents a significant number of persons being missed, considering that our cohort only represents 5% of the individuals with retinoblastoma seen at our institution. This supports the benefit of using broader gene panels for RB patients in the research setting, to facilitate for future studies to determine any gene variants that might be enriched in the RB population that could be considered for clinical testing in patients who would benefit from a diagnosis (ie, diagnosis would modify screening and surveillance recommendations).

We found that 11% of individuals with HR (6% of all individuals) harbored a second P/LP variant in a cancer predisposition gene. We have yet to determine the impact that non-RB1 variants can have in individuals with HR to understand if they confer an independently added risk of cancer or if they interact through common pathways and have disease modifying roles that could help predict disease outcomes and prognosis. We did not find any change in the mean age of retinoblastoma diagnosis when comparing individuals with and without non-RB1 variants that would suggest a modifying effect in the development of the primary tumor. Similarly, there was no significant difference between the same groups in the time to second cancer diagnosis among the overall study population nor within individuals with HR. However, these findings were not statistically significant, likely because of the low rate of second cancers in our study given a primarily pediatric population (mean age of 7 years), with second malignancies typically developing later in life (reported median range of age is 13-17 years).18, 19, 20, 21 Nonetheless, it should be noted that, among the 6 individuals with HR and a non-RB1 variant, 3 individuals are older than 10 years (all females) and all have developed a second cancer (Supplemental Table 3).22, 23, 24, 25, 26, 27, 28, 29

All of the genes in which a non-RB1 variant was identified have a different role in cell regulation. Only 1 is an oncogene (MITF), and the others are tumor suppressors, each involved in different mechanisms. Supplemental Table 2 provides information on each of the genes, including their role in carcinogenesis, and the known associated cancer risks. Although most of the genes in which variants were found are associated with autosomal recessive disease, these results provide valuable information for future family planning. Furthermore, screening recommendations also changed for some of the individuals (and their families) in whom the second variant was found in genes associated with autosomal dominant inheritance.

Given that RB1 not only initiates cancer in the retina but is also an important mechanism of cancer development in non-ocular cancers of later development,^14,30, 31, 32, 33, 34 functional studies can help determine if germline variants in genes with specific roles within the cell could somehow play a role in the pathogenesis leading to the second somatic hit in RB1, facilitating the development of second cancers. A possible role of germline MUTYH variants in the pathogenesis of retinoblastoma has been recently suggested by Newman et al³⁵ and Akdeniz et al,³⁴ from results of functional studies from tumor samples of individuals with hereditary retinoblastoma and germline MUTYH variants, that suggest an important role of the loss of MUTYH function in the tumorigenesis of retinoblastoma.^33,35 These findings along with the increased frequency of MUTYH variants in our cohort (found in half of the individuals with HR and a non-RB1 variant, with 2 of them who also developed a second cancer), could suggest a possible interaction between MUTYH and RB1 with an impact in tumorigenesis.

Further studies are required to identify those genes in which germline variants are more frequently found in persons with HR to determine good candidate genes for functional assessment, which might provide new insights into unknown biological mechanisms contributing to disease, which might help guide management and surveillance strategies in the future.

Conclusion

We found that 12% of individuals with a personal history of retinoblastoma harbored a P/LP variant in a non-RB1 cancer predisposing gene, and 6% of them had 2 different cancer predisposition syndromes. Although the size of our cohort limited the analysis in our study, there were no findings suggesting any association between having HR or non-HR and the presence of a non-RB1 variant and they likely represent 2 completely independent factors. However, further studies are required to determine possible associations with the presence of specific variants and the development of second cancers that could provide a better understanding of tumorigenesis in RB individuals and influence future management, screening, and surveillance recommendations.

Data Availability

All data relevant to the study are included in the article or uploaded as supplementary information. All data were deidentified for use in this study.

Conflict of Interest

All authors declare no conflicts of interest.

Additional Information

Supplementary Table 1

Characteristics of patients with and without a non-RB1 variant

Supplementary Table 2. Clinical information of non-RB1 genes

Supplementary Table 3. Individuals with non-RB1 variants: clinical characteristics and variant information