Cell-Free DNA Tumor Fraction in the Aqueous Humor Is Associated With Therapeutic Response in Retinoblastoma Patients

Abstract

Purpose: The aqueous humor (AH) liquid biopsy enables in vivo evaluation of tumor-derived cell-free DNA (cfDNA) from retinoblastoma (RB) eyes. Herein, we test our hypothesis that longitudinal dynamics of AH cfDNA-including tumor fraction (TFx) and somatic copy number alteration (SCNA) amplitude-correspond to therapeutic response.

Methods: Eyes with ≥3 AH extractions during intravitreal chemotherapy (IVM) or at secondary enucleation between 2015 to 2019 were included. AH cfDNA was sequenced to assess RB SCNA amplitude; ichorCNA software was used to estimate TFx. Eyes without SCNAs or with TFx < 0.10 across all samples were excluded. Therapeutic responses for each eye were determined from clinical records. Statistical analyses included Mann-Whitney U and Pearson correlation tests.

Results: Twenty eyes of 20 patients underwent ≥3 AH extractions; 6 eyes lacked SCNAs or had TFx < 0.10 throughout sampling and were excluded. Clinical progression was associated with significantly higher SCNA amplitudes and TFx values than regression (P ≤ 0.04). Relative increases in TFx (ΔTFx 1.86 ± 2.22) were associated with disease progression, whereas relative decreases in TFx (ΔTFx 0.53 ± 0.36) were associated with disease regression (P < 0.00001). A ≥15% increase in TFx relative to baseline during treatment was associated with an over 90-fold increased likelihood of clinical progression (odds ratio = 90.67, 95% confidence interval = 8.30-990.16, P = 0.0002). TFx and SCNA amplitude were significantly positively correlated throughout sampling (P ≤ 0.002).

Conclusions: Longitudinal changes in AH-derived cfDNA TFx and SCNA amplitude are concordant with clinical responses of intraocular RB during active therapy.

Translational relevance: Longitudinal evaluation of AH cfDNA may provide an objective, quantitative way to monitor therapeutic response and disease burden in RB patients.

Keywords: aqueous humor; liquid biopsy; retinoblastoma; somatic copy number alteration; tumor fraction.

Figure 1.

SCNA amplitude and TFx values correspond to clinical therapeutic response. (A) Genomic profiles of six sequential AH samples extracted from a single eye (case 42) during IVM therapy for tumor seeding (* = 6p gain or 16q loss). (B) Changes in SCNA amplitude correspond to clinical regression (R) and progression (P) at each sampling timepoint. (C) Changes in TFx correspond to clinical regression (R) and progression (P) at each sampling timepoint. (D) Fundus photographs of the affected eye over the course of 6 injections of IVM. Reduction of vitreous seeding and tumor size occurred between injections 2 through 5, whereas a large recurrence of the retinal tumor (arrow) was noted on the day of injection 6.

Figure 2.

Genomic profiles showing SCNAs from cell-free DNA in the aqueous humor of an eye with retinoblastoma (case 12). The genomic profiles generated by our established SCNA analysis protocol (A) were concordant with the genomic profiles generated by the ichorCNA pipeline (B). In ichorCNA profiles, the color of each data point corresponds to estimated integer copy number (green = 1 copy, blue = 2 copies, brown = 3 copies, red = 4+ copies).

Figure 3.

SCNA absolute amplitudes and corresponding TFx for cases 6 (A), 13 (B), 40 (C), and 12 (D). Both SCNA gains and losses are represented as absolute values, determined by |1 – ratio to the median|, for visualization purposes. Clinical progression (P) or regression (R) at each sampling timepoint is shown.

Figure 4.

TFx and SCNA absolute amplitudes from aqueous humor cell-free DNA are significantly positively correlated via Pearson correlation. Specifically, TFx was correlated with amplitudes of 6p gain (r(46) = 0.685, P < 0.00001), 1q gain (r(40) = 0.763, P < 0.00001), 16q loss (r(33) = 0.800, P < 0.00001), and 2p gain (r(23) = 0.588, P = 0.002). Values are shown as absolute amplitude irrespective of a gain or loss.