Introduction

Breast cancer is the most common cancer in women being rated as the second most frequent type of cancer in the world. Besides the inter-tumor heterogeneity, mammary tumors also show high heterogeneity at intratumoral levels¹, especially triple-negative breast cancer. Intratumoral heterogeneity refers to sub-populations of tumor cells with different malignancy potential and responsiveness to treatment. Environmental selection forces, such as hypoxia in the tumor microenvironment provide pressures that drive natural selection, creating these sub-populations of tumor cells². Thus, solid tumors frequently exhibit variable perfusion deficits and consequently, heterogeneous regions of hypoxia, which may have a profound impact on the evolution of the tumor microenvironment and is a barrier to therapies. In this study, tumor subregions (habitats) were identified using distribution clustering of multiple MRI pulse sequences and a 3D printed approach for co-registration, in order to correlate these habitats to histological features.

Methods

Human triple-negative breast cancer cells (MDA-MB-231) were injected into the right mammary gland of the athymic nude mice. The mice were imaged on 7T horizontal magnet (Agilent ASR 310) and Bruker electronics (BioSpec AV3HD), using a 35 mm birdcage coil (Doty Scientific). Anatomical T2 axial images of the mice were obtained with a TurboRARE sequence with TR/TE 2000/48ms, 11 slices, 1 mm slice thickness, FOV=32x32 mm. Also, diffusion-weighted MRI (DWI) (TR/TE=700/15.7 ms and b values of 100, 500 and 750 s/mm2) and T₁ weighted dynamic contrast enhanced (DCE) upon injection of 0.2 mmol/kg Gd-DTPA-BMA (Omniscan®). All parameter maps were obtained using nonlinear least squares pixel-by-pixel fit (Levenburg-Marquardt) to the corresponding functions for T₂, T₂* and ADC. DCE maps were obtained through analysis of the pre- and post-contrast agent bolus time-series on a pixel-by-pixel basis. Parameter maps (Fig 1b-g) values within the tumor volume of interest were used to simultaneous classify 4 clusters. The algorithm was based on Gaussian mixture model fit with an expectation maximization algorithm. All values within the tumor were used for clustering, whereas a single slice is shown (Fig 2). After MRI scans, the mice received an intraperitoneal injection of pimonidazole (60 mg/kg) and were euthanized after 1 hour. The axial T2 images were used to manually segment the tumor contours and create a tumor mold printed in a 3D printer Object24 (Stratasys Ltd., USA), designed to cut the tumor into slices of 2 mm thickness, which are corresponding to two MRI slices (1 mm each). Tumor slices were stained with H&E and with pimonidazole and an endothelial cells marker (CD-31) by immunohistochemistry (IHC). MATLAB code was used to estimate the optimal 3D alignment of MRI images and histological slice.

Results & Discussion

Fig. 1 shows the different MRI sequences used to create the habitats. Pixels in the ADC map that have low values likely have high cellularity, while pixels with high ADCs have low cellularity (Fig. 1d). Low enhancement identifies regions with low vascularization and/or low permeability, while regions with high enhancement can be identified with higher vascularization and permeability. Distinct hypoxic habitats in histology and IHC (Fig 2) were identified, and are corroborated by clusters 2 and 4. Cluster 2 shows low values in ADC (high cellularity) and low enhancement in DCE integral, while cluster 4 also shows low DCE integral and higher values in ADC than cluster 2 (Fig. 2a-b). The MRI and histological images show some differences in the shape, probably because the shrinkage during the tissue fixation process, but qualitatively, the positive staining of pimonidazole showed a similar pattern observed to the hypoxic areas in habitats (Fig. 2e). Areas with high enhancement (clusters green and yellow) (Fig. 2d) showed negative staining to pimonidazole in histology (Fig. 2f). Interestingly, a specific region with high enhancement (green) can be observed in agreement with negative staining for pimonidazole and positive staining for CD-31 (Fig 2g,i).

Conclusion

Non-invasive methods to measure spatially dependent differences in vascular perfusion, permeability, cell density, necrotic state that define intratumoral spatial heterogeneity, which can provide relevant information for making decisions during a course of treatment. The identification of habitats using multiple imaging sequences permits the quantification of the intratumoral spatial heterogeneity, which can be qualitatively comparable to histological assessment of hypoxia. Thus, this approach could potentially be used as a non-invasive imaging method for the monitoring of the intratumoral heterogeneity following the therapy in breast cancer.