Nath et al. · Proceedings of the National Academy of Sciences · 2019

Discovering long noncoding RNA predictors of anticancer drug sensitivity beyond protein-coding genes

Nath Aritro, Lau Eunice Y. T., Lee Adam M., Geeleher Paul, Cho William C. S., Huang R. Stephanie

The study

What was asked, and what was found

Nath et al., Proceedings of the National Academy of Sciences, 2019, asked whether long non-coding RNAs carry information about anticancer drug response that protein-coding genes miss. Working across large cell line drug screens with tissue type and neighbouring coding genes held as covariates, they found that adding two long non-coding RNAs to the model raised the variance explained for erlotinib and gefitinib well above what the known EGFR markers give on their own. The two were EGFR antisense RNA 1 and the MIR205 host gene.

A correlation is not a mechanism, so the group knocked both transcripts down with AUMlnc sdASOs from AUM BioTech, in two human lung cancer lines chosen for opposite erlotinib behaviour, HCC 827 which responds strongly and NCI H2228 which responds weakly. The supplementary methods print both target sequences, name a scramble control of the same chemistry, and describe the dosing plainly: knockdown efficiency was worked out at 1, 5 and 10 μM, and the medium was replaced with fresh medium containing the oligonucleotide, with cells harvested at 24, 48 and 72 hours. No transfection reagent appears anywhere.

Knocking down either transcript slowed growth in both lines, significantly by 48 and 72 hours. Under erlotinib the picture inverted and the knockdown cells proliferated faster than control, which is what a marker of drug sensitivity should do when it is removed. The mechanism sat in EGFR isoform balance rather than in EGFR abundance: total EGFR message did not move, but the ratio of the full-length isoform to the truncated one, which lacks the kinase domain, fell significantly in both lines for both targets. Knocking down EGFR antisense RNA 1 also lowered the MIR205 host gene, so the two transcripts are linked.

Key findings

Two long non-coding RNAs came out of a genome-wide analysis as the best predictors of response to the EGFR inhibitors erlotinib and gefitinib, independent of the known protein-coding biomarkers, and AUMlnc sdASOs were used to test whether they do anything.(SI Appendix, Supplementary Methods, LncRNA knockdown section)
Knocking down either transcript slowed the growth of both lung cancer lines, significantly by 48 and 72 hours, against a scramble control of the same chemistry.(Results, in vitro validation section, page 22026, Fig. 6C)
Under erlotinib the effect reversed, with the knockdown cells proliferating faster than control, which is the behaviour a resistance biomarker predicts.(Results, in vitro validation section, page 22026, Fig. 6D)
Total EGFR message did not move after either knockdown, so the mechanism did not run through how much EGFR the cells made.(Results, in vitro validation section, page 22026, SI Appendix Fig. S6D)
What did move was the balance between the two EGFR isoforms: the ratio of the full-length isoform to the truncated one, which lacks the tyrosine kinase domain, fell significantly in both cell lines for both targets.(Results, in vitro validation section, page 22026, Fig. 6E)
Knocking down one target lowered the other, so the two transcripts are not independent.(Results, in vitro validation section, page 22026)

Product: AUMlnc sdASO
Targets: EGFR-AS1 · human · long non-coding RNA, ENSG00000224057MIR205HG · human · long non-coding RNA, ENSG00000230937
Models: HCC 827 and NCI H2228 human lung cancer cells, chosen for strong and weak erlotinib response · cell line
Delivery: the growth medium was replaced with fresh medium containing the oligonucleotide, with no transfection reagent, and cells were harvested at 24, 48 and 72 hours
Readouts: quantitative real-time PCR, cell growth over time, proliferation rate under erlotinib, EGFR isoform ratio by quantitative PCR
Measured: target sequences of the two oligonucleotides: TATACATTTCATCCCATTGAC for EGFR-AS1 and AAGATTGAGCCACTGCACTCC for MIR205HG (given 5 prime to 3 prime in the supplementary methods)concentrations tested for knockdown efficiency: 1, 5 and 10 μM (HCC827 and NCI H2228 cells)sampling times after adding the oligonucleotide: 24, 48 and 72 hours (to determine knockdown efficiency)cell growth after knocking down either long non-coding RNA: reduced, significantly slower at 48 and 72 hours (both cell lines, against the scramble control)proliferation under erlotinib after knockdown: elevated against control (both target transcripts, both cell lines)total EGFR message after knockdown: not significantly altered (consensus EGFR sequence, both knockdown lines)cross-talk between the two targets: MIR205HG fell significantly when EGFR-AS1 was knocked down (EGFR-AS1 knockdown cells)ratio of the full-length EGFR isoform A to the truncated isoform D after knockdown: significantly reduced in both cell lines for both targets (NCI H2228 and HCC 827, against control)proportion of variance in drug response explained once the two long non-coding RNAs are added to the model: 12 to 18% for erlotinib and 25 to 30% for gefitinib (combined with the EGFR functional events, across all cell lines)
Figures: Fig. 6C Growth of the two lung cancer lines over 72 hours after knocking down either long non-coding RNA with an AUMlnc sdASO.Fig. 6E The ratio of the full-length EGFR isoform to the truncated one in both cell lines after knocking down either target.
Citation: Nath et al. Discovering long noncoding RNA predictors of anticancer drug sensitivity beyond protein-coding genes. Proceedings of the National Academy of Sciences 2019;116:22020-22029. doi:10.1073/pnas.1909998116

For research use only. Not for use in diagnostic or therapeutic procedures.