Optimizing fresh specimen staining for rapid identification of tumor biomarkers during surgery

Abstract

Rationale: Positive margin status due to incomplete removal of tumor tissue during breast conserving surgery (BCS) is a prevalent diagnosis usually requiring a second surgical procedure. These follow-up procedures increase the risk of morbidity and delay the use of adjuvant therapy; thus, significant efforts are underway to develop new intraoperative strategies for margin assessment to eliminate re-excision procedures. One strategy under development uses topical application of dual probe staining and a fluorescence imaging strategy termed dual probe difference specimen imaging (DDSI). DDSI uses a receptor-targeted fluorescent probe and an untargeted, spectrally-distinct fluorescent companion imaging agent topically applied to fresh resected specimens, where the fluorescence from each probe is imaged and a normalized difference image is computed to identify tumor-target distribution in the specimen margins. While previous reports suggested this approach is a promising new tool for surgical guidance, advancing the approach into the clinic requires methodical protocol optimization and further validation.

Methods: In the present study, we used breast cancer xenografts and receiver operator characteristic (ROC) curve analysis to evaluate a wide range of staining and imaging parameters, and completed a prospective validation study on multiple tumor phenotypes with different target expression. Imaging fluorophore-probe pair, concentration, and incubation times were systematically optimized using n=6 tissue specimen replicates per staining condition. Resulting tumor vs. normal adipose tissue diagnostic performance were reported and staining patterns were validated via receptor specific immunohistochemistry colocalization. Optimal staining conditions were tested in receptor positive and receptor negative cohorts to confirm specificity.

Results: The optimal staining conditions were found to be a one minute stain in a 200 nM probe solution (area under the curve (AUC) = 0.97), where the choice of fluorescent label combination did not significantly affect the diagnostic performance. Using an optimal threshold value determined from ROC curve analysis on a training data set, a prospective study on xenografts resulted in an AUC=0.95 for receptor positive tumors and an AUC = 0.50 for receptor negative (control) tumors, confirming the diagnostic performance of this novel imaging technique.

Conclusions: DDSI provides a robust, molecularly specific imaging methodology for identifying tumor tissue over benign mammary adipose tissue. Using a dual probe imaging strategy, nonspecific accumulation of targeted probe was corrected for and tumor vs. normal tissue diagnostic potential was improved, circumventing difficulties with ex vivo tissue specimen staining and allowing for rapid clinical translation of this promising technology for tumor margin detection during BCS procedures.

Keywords: breast cancer; breast conserving surgery; dual probe difference specimen imaging; dual probe imaging; fluorescence; image-guided surgery; tumor margin assessment.

Fig 1

Staining protocol and experimental conditions. (A) Schematic of DDSI staining protocol and imaging approach. The parameters for each step of the staining protocol are listed in their respective boxes. The dual-stain soak experimental parameters that were optimized are highlighted in red. DDSI image processing was performed by subtracting the untargeted image from the targeted image and then dividing by the untargeted image as shown. (B) Table of experimental conditions tested to optimize the staining protocol.

Fig 2

DDSI staining condition optimization. Representative color, fluorescence, and DDSI images of tumor and adipose tissue pairs following staining using a range of dual-stain soak concentrations and incubation times for (A) probe pair A (Herceptin-Cy3b, DkRb-AF647) and (B) probe pair B (Herceptin-AF647, DkRb-Cy3b). All images are representative of data collected for n=6 tumor and adipose tissue pairs per staining condition. All untargeted and targeted channel images are background corrected, normalized by their exposure time and calibration drop intensity, and displayed on equivalent color scales across staining conditions and probe pairs. DDSI images are displayed with equivalent color scales across staining conditions. Scale bars = 5 mm.

Fig 3

Immunohistochemical analysis and DDSI staining pattern validation. Representative color, fluorescence, DDSI, H&E, and HER2 targeted IHC images of tumor and adipose tissue pairs following staining using a range of dual-stain soak concentrations and incubation times for (A) probe pair A (Herceptin-Cy3b, DkRb-AF647) and (B) probe pair B (Herceptin-AF647, DkRb-Cy3b). All images are representative of data collected for n=6 tumor and adipose tissue pairs per staining condition. All untargeted and targeted channel images are background corrected, normalized by their exposure time and calibration drop intensity. DDSI images are displayed with equivalent color scales across staining conditions. H&E and IHC images were acquired from serial sections of the same tissue face imaged in the whole specimen DDSI images. H&E: hematoxylin and eosin; IHC: immunohistochemistry. Scale bars = 5 mm.

Fig 4

ROC curve analysis and optimal DDSI staining condition selection. ROC curves and AUC values for untargeted, targeted, and DDSI images of tumor vs. normal adipose tissue differentiation following staining using a range of dual-stain soak concentrations and incubation times for (A) probe pair A (Herceptin-Cy3b, DkRb-AF647) and (B) probe pair B (Herceptin-AF647, DkRb-Cy3b). (C) Tumor and normal tissue pixel intensity histograms, ROC curves, and AUC values for untargeted, targeted, and DDSI images following staining using probe pair B at 200 nM concentration and 1 min incubation time. Optimal points determined from ROC curve analysis are displayed on each ROC curve and as a vertical line on each pixel value histogram. ROC: receiver operator characteristic; AUC: area under curve; opt pt: optimal point.

Fig 5

HER2(+) and HER2(-) testing cohort for DDSI staining and IHC validation. (A) Representative color, fluorescence, DDSI, H&E, and HER2 targeted IHC images of MCF7-HER2 (HER2+) and MCF7 (HER2-) tumor and adipose tissue pairs following staining using the optimal staining condition (Probe pair B, 200 nM concentration, 1 min incubation time). All images are representative of data collected for n=6 tumor and adipose tissue pairs per tumor cell line. All untargeted and targeted channel images are background corrected, normalized by their exposure time and calibration drop intensity, and displayed on equivalent color scales. DDSI images are displayed with equivalent color scales. H&E and IHC images were acquired from serial sections of the same tissue face imaged in the whole DDSI specimen images. Scale bars = 5 mm. (B) Tumor and normal tissue pixel intensity histograms, (C) ROC curves, and AUC values for untargeted, targeted, and DDSI images corresponding to each cell line. Optimal points determined from ROC analysis are displayed on each ROC curve (blue marker) and as a vertical line on each pixel value histogram. The optimal point determined from the training cohort data is displayed on the MCF7-HER2 tumor specimen ROC curve to demonstrate the diagnostic reproducibility under optimal staining conditions (orange marker). H&E: Hematoxylin & Eosin; IHC: immunohistochemistry; ROC: receiver operator characteristic; AUC: area under curve; opt pt: optimal point.