Longitudinal aqueous humor sampling reflects treatment response in retinoblastoma patients

Abstract

Purpose : The aqueous humor (AH) liquid biopsy for retinoblastoma (RB) facilitates in vivo sampling of tumor-derived cell-free DNA (cfDNA). We previously demonstrated that tumor-derived somatic copy number alterations (SCNAs) in AH predict poor ocular salvage. However, whether AH sampling can be used to monitor clinical response throughout therapy remains unknown. In this retrospective study, we hypothesize that SCNA amplitude (amp) and tumor fraction (TFx) of RB cfDNA in the AH correspond to treatment response over time.

Methods : Eyes that required intravitreal chemotherapy (IVM) for tumor seeding and had ≥3 sequential AH samples extracted during active IVM therapy and/or at secondary enucleation between 2015-2019 were included. Shallow whole-genome sequencing of AH cfDNA was performed to assess RB SCNA amp, and ichorCNA software estimated cfDNA TFx based on SCNAs present. Eyes without SCNAs in any sample were removed from final analysis. For each eye, clinical treatment response (regression vs progression) at each sample timepoint relative to the initial sample timepoint was determined from clinical records. Mann-Whitney U test was used for statistical analyses.

Results : Twenty eyes of 20 patients had ≥3 AH extractions during the study period; 6 eyes had no SCNAs and were excluded from final analysis. Disease progression at any timepoint was associated with significantly higher SCNA amp (1.65±0.34) and TFx (0.57±0.36) than disease regression (SCNA amp 1.22±0.34, P=0.0002; TFx 0.29±0.34, P=0.01) (Fig 1). Relative increases in SCNA amp (Δamp 1.19±0.35) and TFx (ΔTFx 3.68±5.27) from start to end of longitudinal AH sampling were associated with overall disease progression, whereas relative decreases in SCNA amp (Δamp 0.78±0.08; P=0.005) and TFx (ΔTFx 0.55±0.52; P=0.006) were associated with overall disease regression.

Conclusions : We demonstrate that increases and decreases in AH-derived SCNA amp and TFx during therapy tend to correspond to clinical progression and regression of disease, respectively. Our findings suggest that longitudinal AH evaluation throughout active RB therapy may provide a novel quantitative method of monitoring RB treatment response.

This is a 2020 ARVO Annual Meeting abstract.

 

Fig 1. (A) Genomic profiles of 6 sequential AH samples from an eye receiving IVM for RB seeding (* = 6p gain). (B) In the eye from (A), TFx and SCNA (6p) amp correspond to disease regression (R) and progression (P) at each sampling timepoint.