Predicting tumor ER and HER2 status using a cell-free RNA liquid biopsy assay

Lee Schwartzberg^1,2, Jennifer Yen², Taylor Cavazos², Evan Boyle², Mehran Karimzadeh², Hamed Heydari², Magdalena Gebala² , Akshaya Krishnan², Anna Hartwig², Jeff Gregg², Lisa Fish², Fereydoun Hormozdiari², Alexander Lazar³, Babak Alipanahi²

¹Renown Health Oncology and Hematology Center, Reno, NV; ²Exai Bio Inc., Palo Alto, CA; ³The University of Texas MD Anderson Cancer Center, Houston, TX

Background

Treatment selection in advanced breast cancer is stronglydependent upon protein expression of Estrogen Receptor(ER ) and Human Epidermal Receptor 2 (HER2)
We present a novel method to infer biomarker expression from cell-free small RNAs (smRNAs), which include orphan non-coding RNAs (oncRNAs)^{1, 2}. These smRNAs are actively secreted and stable in blood as they are protected by extracellular vesicles and binding of lipo-protein complexes.
Our platform uses smRNAs as surrogates for tumor transcriptomes and can infer ER and HER expression in breast cancer patients from 1 ml of plasma

Methodology

We developed a cell-free smRNA assay with an automated,CLIA-certified lab workflow.
We screened the smRNA transcriptome to identify smRNAs associated with ESR1 and ERBB2 mRNA expression across breast cancer subtypes in tumor tissue from The Cancer Genome Atlas (TCGA) (n=540 breast cancer samples). ER and HER status in TCGA were determined using ASCO-CAP recommendations³ derived from available immunohistochemistry (IHC) and Fluorescent In Situ Hybridization (FISH) scores, where HER2-positive was defined as IHC 3+ or ISH amplified, and HER2-negative as IHC score of 0, 1+ or 2+ and ISH−.
We developed a model in tumor tissue and validated our expression scores with published RNA-seq in tumor tissue (TCGA), breast cancer cell lines with conditioned media (n=10 cell lines in triplicate).
We applied our scores to smRNA isolated from the plasma (1 ml) of breast cancer patients obtained from commercially purchased vendor samples and the provided IHC-derived ER and HER2 status.

Figure 1. Outline of wet lab workflow

Blood was drawn into EDTA plasma, and Streck plasma tubes. The resulting underwent bead-based nucleic acid extraction including on-bead DNase I treatment. Small RNA library prep was performed, followed by sequencing.

smRNA derived scores correlate with ER and HER2 expression in tumor and cell lines

Tumor tissue detection of categorical and quantitative ER and HER2 signal

Figure 2. Predicted ER and HER2 scores by tissue IHC and mRNA status.

ER and HER2 scores were calculated across breast cancer samples in TCGA. A) Predicted gene biomarker scores were stratified by ER or HER2 status, respectively, as determined by IHC. AUCs for ER and HER2 are 0.81 and 0.87, respectively. B) High correlation is observed between ESR1 and ERBB2 gene expression measured by RNA-Seq using Expectation Maximization (RSEM) on the x-axis, and smRNA-based gene biomarker scores on the y-axis. Scores are plotted on a log scale on X and Y axes for both scores.

Cell line and conditioned media detection of categorical and quantitative ER and HER2 signal

Figure 3. ER and HER2 signal is consistent in cell lines and paired conditioned media.

To confirm extracellular secretion of smRNAs for ER and HER2 scoring, gene biomarker scores were generated for smRNA isolated from breast cancer (n=8) and human mammary epithelial (n=2) cell lines and the paired conditioned media. A) ER and HER2 scores for both cell lines and conditioned media trended with known ER and HER2 status. B) To assess the quantitative nature of smRNA-derived scores in extracellular media, smRNAs isolated from the conditioned media of the protein biomarker positive cell line were titrated into those of a triple negative cell line (HCC1143), in triplicate. ER and HER2 scores trended with percentage of biomarker positive cancer cells. Scores were plotted on a log scale on both y-axes.

Plasma smRNA-derived scores are associated with tumor ER and HER2 status

Plasma detection of categorical ER and HER2 signal

Table 1. Characteristics of patient plasma samples.

Plasma samples were obtained from commercial vendors across stages and ER and HER2 status. Triple negative (ER, HER2, PR) patients were used as controls. The available sample cohort was enriched for earlier stage samples (Stage I and II, 84.3%).

Figure 4. ER and HER2 scores trend with protein expression status and correlate between replicate pairs in patient plasma samples.

A) ER and HER2 scores were generated from 1 ml of plasma of samples listed in Table 1. Scores were significantly higher in protein biomarker-positive compared to the protein-biomarker negative samples for both ER and HER2 (Mann-Whitney Test).

B) Plasma from two separate tubes were obtained from the same patient in the same blood draw. Correlation was observed between replicate pairs (Pearson R). Higher correlation was observed for ER in Stage II/IV than Stage I/II. Scores are plotted on a log scale on the y-axis for both scores.

Conclusions

We demonstrate the first use of cell-free small RNAs in an automated liquid biopsy assay to infer breast cancer protein expression biomarkers from transcriptional profiles.
This approach may provide an alternative option to infer protein expression and potential drug target activity from 1 ml of plasma.
Modeling improvements with additional samples is ongoing.

Disclosures:

LS is an advisor of Exai Bio and consultant of Foundation Medicine and Exact Sciences. JY, TC, AK, MG, LF, AH, FH are full-time employees of Exai Bio. BA is a co-founder, stockholder, and full-time employee of Exai Bio. EB is a consultant of Exai Bio. AL is an advisor of Exai Bio.

References:

Fish L, et al. Nature Med. 2018.
Wang J, et al. AACR. 2022.
Tan, R.S.Y.C. et al. BMC Med 2022.