Exploring novel genetic and hematological predictors of response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer

Abstract

Introduction: The standard treatment for locally advanced rectal cancer (LARC) is neoadjuvant chemoradiotherapy (nCRT). To select patients who would benefit the most from nCRT, there is a need for predictive biomarkers. The aim of this study was to evaluate the role of clinical, pathological, radiological, inflammation-related genetic, and hematological parameters in the prediction of post-nCRT response. Materials and methods: In silico analysis of published transcriptomics datasets was conducted to identify candidate genes, whose expression will be measured using quantitative Real Time PCR (qRT-PCR) in pretreatment formaline-fixed paraffin-embedded (FFPE) samples. In this study, 75 patients with LARC were prospectively included between June 2020-January 2022. Patients were assessed for tumor response in week 8 post-nCRT with pelvic MRI scan and rigid proctoscopy. For patients with a clinical complete response (cCR) and initially distant located tumor no immediate surgery was suggested ("watch and wait" approach). The response after surgery was assessed using histopathological tumor regression grading (TRG) categories from postoperative specimens by Mandard. Responders (R) were defined as patients with cCR without operative treatment, and those with TRG 1 and TRG 2 postoperative categories. Non-responders (NR) were patients classified as TRG 3-5. Results: Responders group comprised 35 patients (46.6%) and NR group 53.4% of patients. Analysis of published transcriptomics data identified genes that could predict response to treatment and their significance was assessed in our cohort by qRT-PCR. When comparison was made in the subgroup of patients who were operated (TRG1 vs. TRG4), the expression of IDO1 was significantly deregulated (p < 0.05). Among hematological parameters between R and NR a significant difference in the response was detected for neutrophil-to-monocyte ratio (NMR), initial basophil, eosinophil and monocyte counts (p < 0.01). According to MRI findings, non-responders more often presented with extramural vascular invasion (p < 0.05). Conclusion: Based on logistic regression model, factors associated with favorable response to nCRT were tumor morphology and hematological parameters which can be easily and routinely derived from initial laboratory results (NMR, eosinophil, basophil and monocyte counts) in a minimally invasive manner. Using various metrics, an aggregated score of the initial eosinophil, basophil, and monocyte counts demonstrated the best predictive performance.

Keywords: hematological parameters; inflammation; locally advanced rectal cancer; neoadjuvant chemoradiotherapy; predictive biomarkers.

FIGURE 1

Analysis of microarray datasets GSE45404_570 (A); GSE46862 (B); GSE139255 (C). The analysis was performed using the GEO2R functionality available in the GEO repository (https://www.ncbi.nlm.nih.gov/geo/geo2r). Corresponding volcano plots were automatically generated by the same application. No differentially expressed genes (adjusted p-value<0.05) were found in datasets GSE45404_570 and GSE46862. The top ten differentially expressed genes found in dataset GSE139255 according to logFC value were labeled in 1C. Upregulated genes were labeled red (IL6, SFRP2, IL11, IL1R1, SPP1, FGF7, IL2RB) and downregulated blue (CNTFR, PPP2R2C, DKK4).

FIGURE 2

GSEA enrichment plots for genes included in Hallmark inflammatory response pathway: GSE45404_570 (A); GSE46862 (B); GSE139255 (C). False Discovery Rate q-value FDR q-values were 0.661 (A), 0.056 (B), 0.054 (C). Pathways associated datasets A and B showed significant enrichment (meeting the threshold of FDR <0.25).

FIGURE 3

Venn diagram showing overlapping of inflammation-related genes between three analyzed datasets. Top 100 genes from selected datasets (ranked by the default Signal2Noise metric used in GSEA analysis), were extracted and overlapped using Venn diagram software. Cytoscape (version 3.10.0) was applied to evaluate the potential correlation between finally selected genes.

FIGURE 4

Gene expression levels of IL6, CXCL9, CYBB and IDO1 in responders (blue) and non-responders (green) normalized to GAPDH.

FIGURE 5

ROC curves for the absolute basophil count (A), absolute eosinophil count (B), absolute monocyte count (C) and NMR (D) in relation to response to treatment.

FIGURE 6

Performance of the composite scores with respect to various metrics: AUC - Area Under Curve; AUPRC - Area Under Precision-Recall Curve; RMSE - Root Mean Square Error; RFMDA - Random Forest Mean Decrease in Accuracy; Basoph. - Absolute basophil count; Eosin. - Absolute eosinophil count; Monoc. - Absolute monocyte count; N/M - Neutrophil-to-monocyte ratio; Score 1 - Absolute basophil + eosinophil count; Score 2 - Absolute basophil + monocyte count; Score 3 - Absolute eosinophil + monocyte count; Score 4 - Absolute basophil + eosinophil + monocyte count; Score 5 - Neutrophil-to-monocyte ratio + Absolute monocyte count; Score 6 - Neutrophil-to-monocyte ratio + Absolute eosinophil count; Score 7 - Neutrophil-to-monocyte ratio + Absolute basophil count; Score 8 - Neutrophil-to-monocyte ratio + Absolute monocyte + eosinophil count; Score 9 - Neutrophil-to-monocyte ratio + Absolute monocyte + Absolute basophil count; Score 10 - Neutrophil-to-monocyte ratio + Absolute eosinophil + basophil count; Score 11 - Neutrophil-to-monocyte ratio + Absolute monocyte + eosinophil count + basophil count; Supplementary Figure S1. Relationship between False Negative and True Positive Rates for Top Three Composite Scores at Different Cut-off Values: Score 2 - Absolute basophil + monocyte count; Score 3 - Absolute eosinophil + monocyte count; Score 4 - Absolute basophil + eosinophil + monocyte count.