Quantifying Tumor Heterogeneity via MRI Habitats to Characterize Microenvironmental Alterations in HER2+ Breast Cancer

Abstract

This study identifies physiological habitats using quantitative magnetic resonance imaging (MRI) to elucidate intertumoral differences and characterize microenvironmental response to targeted and cytotoxic therapy. BT-474 human epidermal growth factor receptor 2 (HER2+) breast tumors were imaged before and during treatment (trastuzumab, paclitaxel) with diffusion-weighted MRI and dynamic contrast-enhanced MRI to measure tumor cellularity and vascularity, respectively. Tumors were stained for anti-CD31, anti-ɑSMA, anti-CD45, anti-F4/80, anti-pimonidazole, and H&E. MRI data was clustered to identify and label each habitat in terms of vascularity and cellularity. Pre-treatment habitat composition was used stratify tumors into two "tumor imaging phenotypes" (Type 1, Type 2). Type 1 tumors showed significantly higher percent tumor volume of the high-vascularity high-cellularity (HV-HC) habitat compared to Type 2 tumors, and significantly lower volume of low-vascularity high-cellularity (LV-HC) and low-vascularity low-cellularity (LV-LC) habitats. Tumor phenotypes showed significant differences in treatment response, in both changes in tumor volume and physiological composition. Significant positive correlations were found between histological stains and tumor habitats. These findings suggest that the differential baseline imaging phenotypes can predict response to therapy. Specifically, the Type 1 phenotype indicates increased sensitivity to targeted or cytotoxic therapy compared to Type 2 tumors.

Keywords: BT-474; diffusion-weighted MRI; dynamic contrast-enhanced MRI; habitat imaging; immunofluorescence; immunohistochemistry; paclitaxel; trastuzumab.

Figure 1

Pipeline for habitat imaging analysis and identification of tumor imaging phenotypes. Longitudinal DW- and DCE-MRI data was acquired on days 0 (pre-treatment), 1, and 4 and processed for each mouse to generate quantitative parameter maps (A, top row). Tumors were excised for immunohistochemistry (IHC) and immunofluorescence (IF) analysis at day 5. Multiparametric image data were clustered to identify tumor habitats (B, middle row). Baseline imaging data (day 0) of tumors were used to cluster tumors into two phenotypes based on tumor habitat composition of cellularity and vascularity (C, bottom row left). Tumors were separated by treatment group alone (pooled) or in addition to tumor imaging phenotype (Type 1, Type 2), and longitudinal analysis of tumor composition was analyzed in terms of the percent tumor volume each habitat comprised (D, bottom row right).

Figure 2

Characterization of tumor habitats. The mean value of each quantitative imaging parameter was used to determine the physiology represented by each habitat (A). Identified tumor habitats were labeled in terms of high or low “vascularity” (K^trans, k_ep) and “cellularity” (ADC, v_e). Three tumor habitats were identified: high vascularity–high cellularity (HV-HC), low vascularity–high cellularity (LV-HC), and low vascularity–low cellularity (LV-LC). Error bars show standard deviation and ** indicates a p < 0.01. Representative parameter maps and corresponding habitat maps are shown in (B), with the HV-HC habitat shown in red, LV-HC in green, and LV-LC in blue. The units for K^trans are mL (blood)/mL (tissue)/min, and the units for the ADC are mm²/s.

Figure 3

Discovery of tumor imaging phenotypes. (A) shows the linear relationship between tumor habitats at baseline. Each point represents a tumor at day 0 (pre-treatment), with tumor phenotypes represented by shape (Type 1: black circles, Type 2: white squares). A negative linear correlation was observed between the percent tumor volume of LV-HC and HV-HC habitats (r = −0.75, p < 0.01) and LV-LC and HV-HC habitats (r = −0.80, p < 0.01). A positive linear correlation was observed between the percent tumor volume of LV-LC and LV-HC habitats (r = 0.33, p < 0.01). (B) shows the resulting dendrogram and heatmap from agglomerative clustering of tumors (columns within the heatmap) using baseline tumor habitat information (rows), from which two phenotypes were identified. These tumor imaging phenotypes were designated as Type 1 (black) and Type 2 (white). (C) shows representative habitat maps of a Type 1 and Type 2 tumor at day 0 alongside a corresponding pie chart of whole tumor composition. Type 1 tumors showed significantly higher proportions of the HV-HC habitat compared to Type 2 tumors as well as decreased LV-HC and LV-LC habitats (D). Error bars show 95% confidence interval and ** indicates p < 0.01.

Figure 4

Longitudinal tumor response to targeted and cytotoxic therapies. (A) shows the median percent change in tumor volume over 30 days post-treatment for control, trastuzumab-treated, and paclitaxel-treated tumors. At 30 days after the initiation of therapy, tumors treated with paclitaxel showed a significant decrease in tumor growth compared to control tumors (p < 0.01). Treatment with trastuzumab yielded a longitudinal decrease in tumor volume, significantly lower than both control and paclitaxel-treated tumors (p < 0.01) at day 30. (B) shows the median percent change in tumor volume for control and trastuzumab- and paclitaxel-treated tumors over the course of the imaging study. Trastuzumab-treated tumors showed a longitudinal decrease in tumor volume at day 4 (p < 0.05) compared to baseline and significant decreases in tumor volume compared to control tumors at day 4 (p < 0.05). (C) shows the median percent change in tumor volume for each treatment group, as in (B), separated by tumor imaging phenotype. Type 1 control tumors showed a significant increase in tumor volume at day 1 (p < 0.05) compared to baseline. Type 1 tumors treated with trastuzumab showed a significant decrease in tumor volume compared to control tumors at day 4 (p < 0.05). Type 2 tumors showed no significant changes in tumor volume over the course of the MRI study. Error bars show standard error, ** indicates p < 0.01, * indicates p < 0.05.

Figure 5

Longitudinal alterations in tumor composition of Type 1 tumors in response to cytotoxic and targeted therapies. The left column of each row (A–C) shows the median percent tumor volume of each habitat (HV-HC, LV-HC, LV-LC) for Type 1 tumors at days 0, 1, and 4. The right column of each row (A–C) shows representative habitat maps on days 0, 1, and 4 for Type 1 tumors. Longitudinal alterations in tumor composition are shown for control (row A), paclitaxel-treated (row B), and trastuzumab-treated (row C) tumors. Type 1 tumors treated with paclitaxel showed a longitudinal decrease in the percent tumor volume of the HV-HC habitat (p < 0.01) and a longitudinal increase in the percent tumor volume of the LV-LC habitat (p < 0.01). Type 1 tumors treated with trastuzumab showed a longitudinal increase in the percent tumor volume of the LV-LC habitat by day 4 (p < 0.05). No significant longitudinal alterations in tumor habitat composition were observed Type 1 control tumors. Error bars show standard deviation, ** indicates p < 0.01.

Figure 6

Longitudinal alterations in tumor composition of Type 2 tumors in response to cytotoxic and targeted therapies. The left column of each row (A–C) shows the median percent tumor volume of each habitat (HV-HC, LV-HC, LV-LC) for Type 2 tumors, at days 0, 1, and 4. The right column of each row (A–C) shows representative habitat maps at days 0, 1, and 4 for Type 2 tumors. Longitudinal alterations in tumor composition are shown for control (row A), paclitaxel-treated (row B), and trastuzumab-treated (row C) tumors. Type 2 tumors treated with trastuzumab showed a longitudinal decrease in the percent tumor volume of the LV-HC habitat at day 1 (p < 0.01) and day 4 (p < 0.05) compared to baseline. No significant longitudinal alterations in tumor habitat composition were observed Type 2 control tumors. Error bars show standard deviation, ** indicates p < 0.01, * indicates p < 0.05.

Figure 7

Correlation between MRI tumor habitats and immunofluorescence (IF) staining. For each tumor, the percent tumor volume of each habitat was calculated (A). For each tumor section and IF stain, stain-positive regions were quantified as the percent viable tissue area (A). Correlations were calculated between the percent tumor volume of each habitat and the percent viable tissue area of each IF stain. The percent tumor volume of the HV-HC habitat showed a significant positive correlation with the percent viable tissue area of the CD31+ (r² = 0.49, p = 0.03) and CD45+, F4/80+ (r² = 0.47, p = 0.04) regions (B, from left to right) as well as the vascular maturation index (r² = 0.57, p = 0.01). The percent tumor volume of the LV-HC habitat showed a significant positive correlation with the percent viable tissue area of pimonidazole+ regions (r² = 0.60, p < 0.01) (panel B, far right).