Background

Breast cancer is not solely a disease of cancer cells. The extracellular matrix (ECM) surrounding cancer cells, particularly the density and arrangement of collagen, also plays a role in invasion and metastasis. The 'tumor-associated collagen signature' (TACS), characterized by the reorganization of collagen in the tumor-adjacent region into straight, dense bundles or alignment perpendicular to the tumor boundary, has been associated with poor prognosis.

However, conventional histopathological examinations disrupt the three-dimensional structure during section preparation, and it is difficult to quantitatively analyze the molecular and supramolecular arrangement of collagen based solely on stained images. Second-harmonic generation (SHG) microscopy, which allows observation of fibrous collagen without staining, is an alternative, but the analysis of polarization-resolved SHG (pSHG) data typically involves mathematical assumptions that collagen follows a cylindrical or orthorhombic symmetry. Given that the structure varies depending on the tissue, it is not always valid to assume that the results obtained by fitting to a specific symmetry model are accurate.

A research team in Italy presented an optical index to identify aligned collagen bundles without a theoretical structural model in a study published in Scientific Reports.

Key Findings

The researchers analyzed 30-micrometer-thick formalin-fixed paraffin-embedded sections made from surgical residual specimens of four breast cancer patients. They obtained pSHG images by changing the laser polarization angle in the tumor stroma and peritumoral tissue indicated by the pathologist, and defined the normalized SHG intensity difference between two consecutive minima in the polarization curve as 'polarization asymmetry' (PA). If the two signals with a 180-degree period are the same, PA is close to 0, and the larger the difference, the more disrupted the local symmetry of the collagen.

Homogeneous collagen with a wavy pattern showed a single peak centered around 0 in the PA distribution. In contrast, the straight, aligned bundles in the tumor showed a bimodal distribution with high absolute value regions on both sides. This asymmetry was not due to instrument error. The curves differed in the 180-degree interval, but the total SHG intensity was reproducible when the laser polarization was rotated 360 degrees.

The researchers compared PA with the angle of the minimum value in the curve, the phase shift φ0, and the second-order nonlinear susceptibility component |χyyy| obtained by fitting an orthorhombic symmetry model. Only pixels with a coefficient of determination R² of 0.99 or higher were included in the fitting analysis. In the aligned bundles, PA and |χyyy| were high, and φ0 formed a bimodal distribution with a difference of approximately 90 degrees. The standard deviation SD_PA and SD_φ0 per 20×20 micrometer tile were also higher in the tumor tissue than in the peritumoral tissue (p<0.001). This suggests that the macroscopic aligned bundles are formed by the local intersection and entanglement of microfibers in different directions.

Significance and Prospects

PA can be calculated using only the difference between the two signals without complex curve fitting or prior symmetry assumptions. When the researchers performed the same analysis on human cornea, which has well-known structural and mechanical properties, similar signals to those in breast cancer stroma were reproduced in areas with complex collagen arrangement, such as the lamellar layer. This suggests that it can be applied not only to specific tumors but also to fibrotic diseases and other collagen-rich tissues.

However, this study is an exploratory analysis of tissues from four patients. It remains to be verified whether PA can predict the grade, molecular subtype, recurrence, or survival of cancer. The pSHG itself has limitations in that it requires repeated acquisition of the same field of view at multiple polarization angles, which increases the acquisition time. The relationship between collagen signal and actual tissue stiffness has not been directly measured, so cross-validation with biomechanical tests and large-scale clinical cohorts is needed.