#AACR19: Genomic cfDNA analysis of aqueous humor in retinoblastoma (RB) predicts eye salvage


Liya Xu1, Jesse L. Berry2, Irsan Kooi3, A Linn Murphree2, Rishvanth K. Prabakar1, Mark W. Reid4, Kevin Stachelek4, Bao Han A. Le2, Lisa Welter1, Rima Jubran4, Thomas C. Lee2, Jonathan W. Kim2, Peter Kuhn1, David Cobrinik2, James B. Hicks1

1University of Southern California, Los Angeles, CA; 2Children’s Hospital Los Angeles and the USC Roski Eye Institute, Los Angeles, CA; 3Leiden, the Netherlands, Leiden, Netherlands; 4Children’s Hospital Los Angeles, Los Angeles, CA


Background: Retinoblastoma (RB) was one of the first tumors to demonstrate a genetic basis to the development of cancer. However, unlike other cancers, RB cannot be directly biopsied due to the high risk of extraocular cancer spread. Therefore, unless the eye is enucleated, tumor tissue is not evaluated for genetic and genomic changes and these alterations are not used to inform diagnosis or prognosis for this disease. However, in a 2017 publication in JAMA Ophthalmology, we demonstrated that tumor-derived cell-free DNA can be extracted from the aqueous humor (AH) of RB eyes, which is safe to remove even with active intraocular disease. The purpose of this current study (Berry JL, Xu L et al. Molec Canc Res 2018) was to identify somatic chromosomal copy number alterations (SCNA) in tumor-derived cell-free DNA in the AH of RB eyes and to correlate with clinical outcomes particularly tumor relapse requiring enucleation.

Methods: AH was extracted via paracentesis from RB eyes during intravitreal injection of chemotherapy or post enucleation. Shallow whole genome sequencing was performed to assess for cell-free tumor DNA fractions and highly-recurrent SCNAs in RB which include gain of 1q, 2p, 6p and loss of 13q and 16q. Globe salvage was recorded.

Results: 26 patients were included; 3 patients had both eyes included for 29 eyes. From these, 63 samples of AH were analyzed; 5 post-enucleations and 58 during intravitreal chemotherapy injection. Ultimately 13 eyes required enucleation and 16 eyes were salvaged. Follow-up ranged from 8-43 months (median 17 months).
The presence of any detectable SCNA was 92% in enucleated eyes versus 38% in salvaged eyes (p=0.006). 6p gain was the most common SCNA found in 77% of enucleated eyes versus 25% of salvaged eyes (p=0.0092). 6p gain was associated with a ten-fold increased odds of enucleation (OR=10, 95% CI:1.8-55.6). The mean amplitude of 6p gain was 1.47 in enucleated eyes versus 1.07 in salvaged eyes (p=0.001). The probability of ocular survival was higher in eyes without detectable SCNAs in the AH (p=0.0028).
Serum testing was positive for a germline mutation in 17 eyes of 14 patients with the following incidence of RB SCNAs 1q (35%); 2p (18%); 6p (47%); 13q (18%) and 16q (35%) versus 12 eyes of 12 patients without germline disease with 1q (33%); 2p (8%); 6p (50%); 13q (17%) and 16q (33%) (p=0.71).

Conclusions: This is the first study to show that clinical outcomes correlate with highly-recurrent SCNAs in the AH from RB eyes. This study suggests that the AH can reliably serve as a surrogate to tumor biopsy and improves upon current clinical staging to predict tumor response to therapy and the ability to salvage the eye. Unlike previous studies that suggest a greater incidence of RB SCNAs in patients with non-germline disease, this was not seen in this evaluation which may be because all previous work was done on tumor from enucleated eyes instead of salvaged eyes.

For a discussion with Dr. Xu, please visit her at AACR 2019: April 2, 2019, 8:00 AM – 12:00 PM – Section 6

Add to your AACR Calendar

Liya Xu, PhD

As a postdoctoral research associate in USC Michaelson CSI center, I started collaborating with Dr. Jesse Berry in 2017, to explore the using of aqueous humor from retinoblastoma (RB) eyes as liquid biopsy for the disease. RB is a disease in which malignant cells form in the tissues of the developing retina. Despite the fact that RB was the first tumor with a know genetic etiology, children with RB were never benefit from precision diagnosis and therapies because the intraocular space was inviolable and tumor biopsy was not available. The classification models for RB is not based on biopsy and does not consider any tumor-derived genetic markers for diagnosis or prognosis of treatment response. So far no knowledge of correlations between genetic changes and treatments. My collaboration with Dr. Berry bridge this gap by harnessing the power of next-generation sequencing to assay rare tumor-derived nucleic acid fragments in the aqueous fluid. Our work shows for the first time that this method is feasible in eyes with RB. Further more, we confirmed that genomic analysis of the AH samples reproducibly reflects the genomic state of the corresponding tumor and the highly recurrent “RB_somatic copy number alterations” have power to predict tumor response to therapy. This study is a remarkable application of the “liquid biopsy” concept in RB. More broadly, it provides a crucial proof of concept for the feasibility of this approach in ocular oncology.