Cell-Free RNA as a Novel Biomarker for Response to Therapy in Head & Neck Cancer

Abstract

Liquid biopsies are gaining more traction as non-invasive tools for the diagnosis and monitoring of cancer. In a new paradigm of cancer treatment, a synergistic botanical drug combination (APG-157) consisting of multiple molecules, is emerging as a new class of cancer therapeutics, targeting multiple pathways and providing a durable clinical response, wide therapeutic window and high level of safety. Monitoring the efficacy of such drugs involves assessing multiple molecules and cellular events simultaneously. We report, for the first time, a methodology that uses circulating plasma cell-free RNA (cfRNA) as a sensitive indicator of patient response upon drug treatment. Plasma was collected from six patients with head and neck cancer (HNC) and four healthy controls receiving three doses of 100 or 200 mg APG-157 or placebo through an oral mucosal route, before treatment and on multiple points post-dosing. Circulating cfRNA was extracted from plasma at 0-, 3- and 24-hours post-treatment, followed by RNA sequencing. We performed comparative analyses of the circulating transcriptome and were able to detect significant perturbation following APG-157 treatment. Transcripts associated with inflammatory response, leukocyte activation and cytokine were upregulated upon treatment with APG-157 in cancer patients, but not in healthy or placebo-treated patients. A platelet-related transcriptional signature could be detected in cancer patients but not in healthy individuals, indicating a platelet-centric pathway involved in the development of HNC. These results from a Phase 1 study are a proof of principle of the utility of cfRNAs as non-invasive circulating biomarkers for monitoring the efficacy of APG-157 in HNC.

Keywords: biomarkers; cfRNA; curcumin; head & neck; liquid biopsy.

Figure 1

Characteristics of the sequencing runs. Samples were run in four batches under different conditions. (A) Number of sequencing reads mapping to features of interest (exons) per sequencing batch, non-corrected (left), after mitochondrial transcript removal (right). (B) Intron to exon ratio per sequencing batch. (C) Principal component analysis after normalization and variance-stabilizing transformation of samples before (left) and after batch correction (right). (D) Distance matrices between samples before (left) and after (right) batch correction. Each row and column represent a single sample, and the diagonal represents samples that are identical. A higher distance score represents higher dissimilarity between a pair of samples. Sample annotation is shown by multiple variables, such as: batch, whether a patient has cancer or not, whether they received 100mg or 200mg of APG-157 or Placebo, treatment time point and patient ID. We have included the same batch and cross-batch technical replicates for a portion of the samples. (E) The distribution of exon occupancy across all genes with 2 or more exons; a value of 1 means full occupancy; a value between 0 and 1 means partial exon occupancy; genes with no exon occupancy have been removed from the dataset. (F) An IGV screenshot showing the coverage of ACTB as a representative gene with full exon occupancy.

Figure 2

Cell-type deconvolution from plasma cfRNA using two reference datasets and two deconvolution tools. Upper panel: a single cell RNAseq reference dataset from H&N cancer (28). Lower panel contains a reference from lymphocytes LM22 (27). Left panel shows results obtained by CIBERSORTx, right panel results obtained by GEDIT. Results are clustered by proximity. Each row represents a single sample, and an annotation panel is included to the left of each heatmap representing the: sequencing batch (batch), whether the patient has cancer (cancer), whether they received APG-157 (APG), which concentration (concentration), when the sample was collected (treatment) and the blinded patient ID. Some samples contain within batch/across batch technical replicates.

Figure 3

Differential expression analysis. (A) Volcano plot representing significantly differentially expressed genes in Cancer vs. Healthy patients at baseline. Genes are considered significant if they pass the FDR-corrected p-value of 0.1 and show an absolute fold difference of at least 2. (B) Volcano plot representing significantly differentially expressed genes in APG-treated Cancer patients after 24h vs. baseline. Genes are considered significant if they pass the FDR-corrected p-value of 0.1 and show an absolute fold difference of at least 2. (C) Overlap of differentially expressed genes (DEGs) between different comparisons and (D) heatmap showing the overlap in overrepresented terms between different comparisons. CvsNup = Cancer vs. Healthy patients before treatment, upregulated genes; YY3up/down APG-treated Cancer patients at 3h post-treatment vs. pre-treatment, up or downregulated genes; YY24up/down APG-treated Cancer patients at 24h post-treatment vs. pre-treatment, up or downregulated genes; NN24up/down Placebo-treated healthy patients at 24h post-treatment vs. pre-treatment, up or downregulated genes. (E) Bar plot showing the top most significantly enriched terms in cancer vs. healthy patients at baseline. The dataset is a combination of GO Biological Processes, GO cellular component, MsigDb and Jensen tissues databases. (F) Bar plot showing the top most significantly enriched terms in APG-treated cancer patients 24h post-treatment vs. pre-treatment. The dataset is a combination of GO Biological Processes, GO cellular component, MsigDb and Jensen tissues databases.