Variability in retinoblastoma genome stability is driven by age and not heritability

Abstract

Retinoblastoma (RB) is a childhood intraocular cancer initiated by biallelic inactivation of the RB tumor suppressor gene (RB1^−/−). RB can be hereditary (germline RB1 pathogenic allele is present) or non-hereditary. Somatic copy number alterations (SCNAs) contribute to subsequent tumorigenesis. Previous studies of only enucleated RB eyes have reported associations between heritability status and the prevalence of SCNAs. Herein, we use an aqueous humor (AH) liquid biopsy to investigate RB genomic profiles in the context of germline RB1 status, age, and International Intraocular Retinoblastoma Classification (IIRC) clinical grouping for both enucleated and salvaged eyes. Between 2014 and 2019, AH was sampled from a total of 54 eyes of 50 patients. Germline RB1 status was determined from clinical blood testing, and cell-free DNA from AH was analyzed for SCNAs. Of the 50 patients, 23 (46.0%; 27 eyes) had hereditary RB, and 27 (54.0%, 27 eyes) had non-hereditary RB. Median age at diagnosis was comparable between hereditary (13 ± 10 months) and non-hereditary (13 ± 8 months) eyes (P = 0.818). There was no significant difference in the prevalence or number of SCNAs based on (1) hereditary status (P > 0.56) or (2) IIRC grouping (P > 0.47). There was, however, a significant correlation between patient age at diagnosis, and (1) number of total SCNAs (r[52] = 0.672, P < 0.00001) and (2) number of highly-recurrent RB SCNAs (r[52] = 0.616, P < 0.00001). This evidence does not support the theory that specific molecular or genomic subtypes exist between hereditary and non-hereditary RB; rather, the prevalence of genomic alterations in RB eyes is strongly related to patient age at diagnosis.

1 |. INTRODUCTION

Retinoblastoma (RB) is a cancer of the developing pediatric retina and is the most common childhood intraocular malignancy. Biallelic inactivation of the RB tumor suppressor gene (RB1^−/−) initiates tumor formation in the vast majority of RB cases,¹ whereas additional genomic events (termed somatic copy number alterations, or SCNAs) contribute to further tumor growth and progression.^2,3 In approximately 40% of patients, RB is considered hereditary due to the presence of a germline RB1 pathogenic allele and consists generally of bilateral, multifocal disease.⁴ The remaining non-hereditary patients require somatic inactivation of both RB1 alleles for cancer initiation and present with unilateral disease. It is well accepted that non-hereditary patients often present later than hereditary patients, even without a positive family history. In general, unilateral patients present at 24 months of age, whereas bilateral hereditary patients present at approximately 12 months.⁵ This disparity was part of the basis of Knudson’s hypothesis and the early understanding of tumor suppressor genes.¹

From molecular analyses of enucleated eyes, multiple studies have proposed the existence of molecular subtypes within RB1^−/− RB. For example, van der Wal et al. divided RB tumors into high and low chromosomal instability groups, with hereditary cases possessing low-level instability and non-hereditary cases showing more frequent chromosomal aberrations.⁶ Subsequent studies also showed that bilateral cases had fewer chromosomal changes compared to unilateral, non-hereditary cases.^7,8 More recently, molecular profiling of tumor tissue suggested variability in genomic stability between hereditary and non-hereditary RB tumors—with older non-hereditary RBs containing significantly more SCNAs than younger hereditary RBs.⁹ Based on these findings, Mol et al. proposed that the differences in SCNA prevalence could be due to underlying germline RB1 status of patients; although the true factor responsible for this disparity in genomic stability was difficult to determine, as both heritability status and age at diagnosis were highly related in their dataset.⁹

Until recently, molecular and genomic profiling of RB has been limited solely to enucleated eyes. This is because direct tissue biopsy of active RB tumors is strictly contraindicated due to the risk of causing extraocular tumor seeding and orbital relapse.¹⁰ However, with our development of the aqueous humor (AH) liquid biopsy for RB, we are now able to obtain in vivo tumor-derived cell-free DNA (cfDNA) from the AH and generate genomic profiles for both enucleated and salvaged (non-enucleated) RB eyes.^11,12 From our liquid biopsy platform, we have shown that the presence of highly recurrent RB SCNAs (gain of 1q, 2p and 6p; loss of 13q and 16q) is significantly predictive of poor ocular survival—with 6p gain being the most predictive of enucleation.¹² Given this association between genomic alterations and prognosis, a clear understanding of how SCNA patterns relate to other clinical features of RB is paramount. Therefore, the aim of this study was to analyze genomic profiles from the AH of both enucleated and salvaged eyes in the context of germline RB1 status, age, and International Intraocular Retinoblastoma Classification (IIRC) grouping.¹³ From our patient population at a large RB treatment center, we provide compelling evidence that the prevalence of SCNAs is related to age at diagnosis and not heritability status or IIRC grouping of disease.

2 |. MATERIALS AND METHODS

2.1 |. Patient and sample characteristics

This research was performed under Institutional Review Board approval and adhered to the tenets of the declaration of Helsinki. Written informed consent was obtained from the parents of all RB patients before inclusion in the study.

All patients diagnosed with RB between December 2014 and December 2019 at Children’s Hospital Los Angeles (CHLA) were included. As described previously,^11,12 samples consisted of 0.1 mL AH extracted via clear corneal paracentesis during routine intravitreal melphalan (IVM) treatment,¹⁴ at primary or secondary enucleation, or at diagnosis. For eyes that had multiple AH samples extracted throughout therapy, the first available sample was used according to our established protocol.¹² Patients with RB who were treated at CHLA but did not have AH extracted were excluded. Clinical records were reviewed retrospectively for patient information, including age at diagnosis and IIRC group. Germline RB1 status was determined from routine clinical blood testing as part of the standard RB diagnostic work-up. Patients were defined as hereditary when a RB1 pathogenic allele was present in peripheral blood, whereas patients without a RB1 pathogenic allele in peripheral blood were considered non-hereditary. It should be noted that a minority (5%−10%) of non-hereditary patients could possess a low-level mosaic pathogenic allele that was not detected on peripheral blood analysis.¹⁵

2.2 |. Isolation and sequencing of AH cfDNA

The process of AH cfDNA collection and processing for genomic profiling has been described in depth previously.^11,12 Based on established methods of SCNA analysis,^16,17 isolated cfDNA underwent shallow whole-genome sequencing for copy number profiling. As described, SCNAs were considered to be present at 20% deflection from a baseline human genome.^11,12

2.3 |. Copy number alteration heatmap

Clustering was performed in R using the heatmap.2 function in the ggplots package. Samples were clustered by Ward’s method with Manhattan distance by their median centered data. Cutoff for gains and losses were 1.20 and 0.80 over the median, respectively.¹⁷

2.4 |. Statistical analyses

Differences in categorical variables were analyzed with Fisher’s exact test, and Mann-Whitney U test was used to compare continuous variables. Pearson correlation was performed to assess the relationship between age at diagnosis and number of SCNAs.

3 |. RESULTS

3.1 |. Patient characteristics

A total of 54 eyes of 50 patients were included in the study (Figure 1a; see Supplementary Table 1). Twenty-three of the 50 patients (46.0%; 27 eyes) had hereditary RB, and 27 of the 50 patients (54.0%; 27 eyes) had non-hereditary RB. All non-hereditary patients had unilateral disease; 5 of the 27 (18.5%) hereditary cases were unilateral, and the remaining cases were bilateral. The median age at diagnosis overall was 13 ± 9 months (range 0–59 months), with 24 of 54 eyes (44.4%) diagnosed at <12 months of age and 30 of 54 eyes (55.6%) diagnosed at ≥12 months. There was no difference in median age at diagnosis between hereditary (13 ± 10 months; range 0–38 months) and non-hereditary (13 ± 8 months; range 2–59 months) patients (P = 0.818). Bilateral RB tended to be diagnosed at younger ages (11.5 ± 8 months; range 0–38 months) than unilateral RB (18 ± 9.5 months; range 2–59 months) regardless of germline RB1 status, although this difference was not statistically significant (P = 0.14). According to the IIRC scheme, there were 2 of the 54 (3.7%) group B eyes, 4 of the 54 (7.4%) group C eyes, 38 of the 54 (70.4%) group D eyes, and 10 of the 54 (18.5%) group E eyes (with groups D and E representing the most advanced disease).¹³

FIGURE 1.

(A) Summary of clinical characteristics of the 54 eyes, including germline RB1 status (hereditary or non-hereditary), age at diagnosis (<12 months or ≥12 months), and International Intraocular Retinoblastoma Classification (IIRC) grouping (groups B, C, D, or E). We have drawn a line to illustrate how this figure should be interpreted: the line represents a sample patient who has hereditary disease (red), was diagnosed at <12 months of age (light blue), and has Group E RB (light green). (B–E) Somatic copy number alteration (SCNA) heatmaps representing chromosomal gains in red and chromosomal losses in blue. There was no significant difference in number of SCNAs between hereditary (B) and non-hereditary (C) eyes. There was, however, a significantly lower number of SCNAs in eyes diagnosed at <12 months of age (D) compared to ≥12 months (E)

3.2 |. Heritability status and SCNAs

Similar to previous publications,^2,12 6p gain was the most common highly-recurrent RB SCNA (28/54 eyes, 51.9%), followed by 1q gain (23/54 eyes, 42.6%), 16q loss (17/54 eyes, 31.5%), 2p gain (11/54 eyes, 20.4%), and 13q loss (6/54 eyes, 11.1%). Gain of MYCN on chromosome 2p was the only recurrent focal alteration in our study population, present in two hereditary eyes and two non-hereditary eyes; all four of these eyes had proven RB1^−/− RB on post-enucleation tissue analysis, suggestive of a secondary amplification of MYCN.¹⁸ There were no significant differences in the prevalence of RB SCNAs between hereditary and non-hereditary eyes (P > 0.56; see Table 1). In addition, the median numbers of total SCNAs and RB SCNAs per eye were not significantly different between hereditary (total SCNAs: 2 ± 2; RB SCNAs: 1 ± 1) and non-hereditary (total SCNAs: 3 ± 2; RB SCNAs: 2 ± 1) eyes (P = 0.857 and 0.430, respectively; Figure 1b and c). Even in a sub-analysis of only enucleated eyes (similar to the Mol et al. study⁹), there were still no significant differences in median numbers of total SCNAs or RB SCNAs between hereditary (total SCNAs: 4 ± 3; RB SCNAs; 2 ± 1) and non-hereditary (total SCNAs: 5 ± 2; RB SCNAs; 3 ± 1) eyes (P = 0.857 and 0.313, respectively).

TABLE 1.

Prevalence of highly recurrent RB SCNAs, including gain of 1q, 2p, and 6p, and loss of 13q and 16q, in different clinical subgroups. Significant (<0.05) P-values are bolded

	1q	2p	6p	13q	16q	Any RB SCNA	Totals
Hereditary	37.0%	18.5%	48.1%	11.1%	25.9%	66.7%	27
Non-hereditary	48.1%	22.2%	55.6%	11.1%	37.0%	70.4%	27
P-value	0.583	1	0.786	1	0.559	1
<12 months	4.2%	16.7%	20.8%	0.0%	0.0%	33.3%	24
>12 months	73.3%	23.3%	76.7%	20.0%	56.7%	96.7%	30
P-value	<0.00001	0.736	<0.00001	0.028	<0.00001	<0.00001
Group D	44.7%	23.7%	55.3%	10.5%	34.2%	71.1%	38
Group E	30.0%	10.0%	50.0%	10.0%	20.0%	70.0%	10
P-value	0.488	0.664	1	1	0.472	1

Abbreviations: RB, retinoblastoma; SCNA, somatic copy number alteration.

3.3 |. Age at diagnosis and SCNAs

All highly recurrent RB SCNAs except for 2p gain were significantly more prevalent in eyes diagnosed at ≥12 months of age rather than <12 months (P < 0.03; see Table 1). Eyes with higher ages at diagnosis (≥12 months) had significantly greater median numbers of total SCNAs and RB SCNAs (total SCNAs: 5 ± 2.5; RB SCNAs: 2.5 ± 0.5) than eyes diagnosed at <12 months (total SCNAs: 1 ± 1; RB SCNAs: 0 ± 0; P < 0.00001 for both; Figure 1d and e). Overall, bivariate analysis with Pearson correlation—which evaluated age as a continuous variable—demonstrated strong positive correlations between age at diagnosis and (1) number of total SCNAs (r[52] = 0.672, P < 0.00001) and (2) number of RB SCNAs (r[52] = 0.616, P < 0.00001).

3.4 |. IIRC grouping and SCNAs

As above, the majority of eyes were classified as either group D (70.4%) or E (18.5%), consistent with advanced intraocular disease. There were no significant differences in the prevalence of RB SCNAs between group D and E eyes (P > 0.47; see Table 1). In addition, the median numbers of total SCNAs and RB SCNAs per eye were not significantly different between group D (total SCNAs: 3.5 ± 3.5; RB SCNAs: 2 ± 1) and group E (total SCNAs: 1 ± 0; RB SCNAs: 1 ± 0) eyes (P = 0.401 and 0.373, respectively).

4 |. DISCUSSION

Previous studies have shown a significant correlation between heritability status and genomic stability, although these analyses were limited to enucleated RB tumor tissue.^6–9 Our AH-based in vivo assay, however, allows us to investigate this relationship on a more comprehensive scale. In a larger cohort of both enucleated and salvaged eyes, we demonstrate that the presence of SCNAs is clearly associated with age at diagnosis and not germline RB1 status or IIRC group.

In a previous study, Mol et al. noted that older non-hereditary patients tended to have more SCNAs, although the close association between heritability and age in their study masked the true reason for SCNA differences.⁹ Our study, however, included hereditary and non-hereditary cases that showed no difference in age at diagnosis, which effectively eliminates the potential confounding factor. It is possible that this disparity in age distribution between our patient populations is due to differences in inclusion criteria between studies. For example, Mol et al. specifically selected their hereditary tumor samples from patients with bilateral RB and very low ages at diagnosis and non-hereditary tumor samples from patients with unilateral RB and very high ages at diagnosis,⁹ whereas we included all patients with RB from whom AH was extracted, regardless of age or laterality. In addition, although the age distribution in our study population differs from that of Mol et al.,⁹ we are not the first to demonstrate a lack of correlation between age at diagnosis and heritability status. A large study of unilateral RB at our institution¹⁹ as well as studies in Germany²⁰ and Belgium²¹ have all shown that age at diagnosis alone is a poor predictor of RB1 status. Based on our current findings, age is likely a strong confounding factor in the association between heritability status and genomic stability, and we therefore recommend that age at diagnosis be explicitly accounted for in future RB genomic profiling studies.

Importantly, we demonstrate a significant correlation between age at diagnosis and genomic instability, with tumors in older patients having significantly more SCNAs than tumors in younger patients. This corroborates the findings by Kooi et al., who similarly associated age (over heritability) with genomic instability in a large meta-analysis of SCNAs from enucleated eyes.²² The significant association between age and SCNAs in our study has potential biological implications related to RB development and growth. In particular, our findings suggest that the prevalence of SCNAs in an individual tumor may be related to the time since the tumor cell of origin has been without functional RB1 and its corresponding protein, pRB. Given that the cell of origin for RB is sensitive to RB1 loss during gestation,²³ the loss of pRB (coinciding with the second hit to RB1) is considered a rate-limiting step in RB development. However, there is evidence that loss of the RB1 tumor suppressor gene alone initiates a premalignant RB1^−/− retinoma phase,^3,7,23 and there is a second rate-limiting step prior to expansive tumor growth that is likely related to the development of one or more chromosomal alterations.^2,3,7 Although the transition from retinoma to active RB is generally attributed to the accumulation of SCNAs—which are believed to drive RB growth and progression^3,7—previous studies have also demonstrated that SCNAs are present (although typically fewer in number) in the retinoma stage before RB formation.^3,7 It is possible that individuals whose RB tumors are diagnosed at later ages have had undetected retinomas for longer periods of time before escaping the retinoma phase and transitioning to malignant RB, perhaps allowing for more SCNAs to develop and accumulate. Further investigations regarding time to RB formation and SCNA development would be useful to clarify this potential biological relationship.

Our findings indicate that the classification of RB (RB1^−/−) tumors into specific molecular or genomic categories based on heritability status or IIRC group alone is not appropriate. From RNA and DNA profiling of enucleated RB1^−/− tumors, Kooi et al. similarly concluded that RB should not be categorized into distinct molecular subtypes but, rather, should be described according to their stage of progression over time.²⁴ A recent study using the AH liquid biopsy demonstrated that a gain of chromosome 6p in RB eyes is strongly predictive of poor ocular survival.^12,25 Given this finding, an argument could be made for categorizing tumors as either 6p-gain positive or negative. However, we emphasize that this distinction is likely not related to germline status of disease, as neither hereditary nor non-hereditary eyes were preferentially associated with the presence of 6p gain.

For multiple reasons, our results add a unique perspective to the previous analyses of RB heritability status and genomics. First, our patient cohort is over twice the size of the previous analysis by Mol et al., which examined a total of only 21 RB tumors.⁹ In addition, all previous studies of RB profiling in relation to heritability status were based on tumor biopsies taken post-enucleation.^6–9 In contrast, our findings are derived from AH liquid biopsy analyses of both enucleated and salvaged RB eyes—not only those that were surgically removed. Because we included patients regardless of age at diagnosis or enucleation status, we believe that our findings are more generalizable to RB tumors compared to prior analyses. However, it is important to note that in both this study and the analysis by Mol et al.,⁹ the majority of studied eyes had advanced (IIRC group D or E) disease. Therefore, it is not yet clear whether our findings can be directly applied to less advanced (IIRC groups A, B, or C) eyes. A large multicenter global RB study showed that the vast majority of patients with RB present with advanced disease, with only 6.7% of eyes having group A or B disease at presentation worldwide.²⁶ Patient presentation at earlier stages is more typically seen in familial RB when the child is being screened; however, enucleation is rarely required with these less advanced eyes; therefore, it is rare to have tumor tissue available for analysis. As we gather more data in vivo from the AH, we hope to determine whether less advanced eyes show a similar age-related progression of SCNAs in hereditary and non-hereditary RB patients.

With the expansion of sequencing technologies and an increased emphasis on molecular diagnosis and prognostication of RB tumors in recent years,^11,12,27 the use of genomic profiling to help guide clinical decision-making in the future seems ever more conceivable. Therefore, it is important that we have an appropriate understanding of RB genomics in the context of other genetic and clinical features of RB. Based on our findings, we conclude that age is strongly associated with the occurrence of genomic alterations, suggesting that SCNAs may accumulate in RB tumors over time. Larger prospective, multicenter studies would be useful to evaluate the role that AH sampling should play in genomic profiling and clinical management of RB patients in the future.