Biomineralogical signatures of breast microcalcifications

Abstract

Microcalcifications, primarily biogenic apatite, occur in cancerous and benign breast pathologies and are key mammographic indicators. Outside the clinic, numerous microcalcification compositional metrics (e.g., carbonate and metal content) are linked to malignancy, yet microcalcification formation is dependent on microenvironmental conditions, which are notoriously heterogeneous in breast cancer. We interrogate multiscale heterogeneity in 93 calcifications from 21 breast cancer patients using an omics-inspired approach: For each microcalcification, we define a "biomineralogical signature" combining metrics derived from Raman microscopy and energy-dispersive spectroscopy. We observe that (i) calcifications cluster into physiologically relevant groups reflecting tissue type and local malignancy; (ii) carbonate content exhibits substantial intratumor heterogeneity; (iii) trace metals including zinc, iron, and aluminum are enhanced in malignant-localized calcifications; and (iv) the lipid-to-protein ratio within calcifications is lower in patients with poor composite outcome, suggesting that there is potential clinical value in expanding research on calcification diagnostic metrics to include "mineral-entrapped" organic matrix.

Fig. 1.. Breast cancer patient sample flow (with anatomical context) and table of breast cancer patient data and pathology of calcification regions.

(A) Schematic depicting an example patient diagnosed with breast cancer, with calcifications (Calcs) present on the mammogram (MMG). Patient data collected at diagnosis are stored and later deidentified if the patient is selected for the study. (B) Schematic magnification of the calcification region from the mammogram in (A) showing a milk duct and lobule with a region of IDC (pink) and associated malignant mineral. Surgical excision of the tumor (dashed line) also extracts an example tumor-adjacent calcification situated in a benign lobule. (C) Schematic cross-section of the duct from (B) showing IDC. Cancer cells are both within and invading outside the duct into the surrounding connective tissue (stroma). Example malignant epithelial and nonepithelial calcifications are indicated. (D) Scheme showing that, upon surgical removal of the tumor (via lumpectomy or mastectomy), a portion was embedded, snap-frozen, and banked at −80°C. After selection for this study, tissue was serially cryosectioned for pathological and biomineralogical analysis. Note that the schematic is not to scale. (E) Table of patient and calcification data. Patient data collected at diagnosis and follow-up data (metastases) are shown (left; purple). For each patient, the number of analyzed calcification regions within sectioned tissue [as shown in (D)] is given and broken down by pathological classifications (right, pink) determined from the serial H&E section. Combined totals across patients for these classifications are shown in insets. Pathological classification is described in Supplementary Text (page 2, "Calcification classifications" section). Patient-specific symbols/markers and numbers are shown (left) and used throughout the study.

Fig. 2.. Schematic depicting data analysis flow.

The schematic shows how the pathological and biomineralogical signatures are generated for a sample benign (micro)calcification (Calc) region within sectioned tissue from patient 21. Using the H&E-stained section (right), the calcification region is classified by conventional histopathological evaluation as indicated. The corresponding unstained section undergoes light optical microscopy followed by rehydration and buffer immersion for nondestructive Raman microscopy (blue shaded area) of the calcification region (black rectangle). Electron microscopy and EDS mapping (green shaded area) are performed on the region (white rectangle) after sample dehydration. Both Raman and EDS produce hyperspectral datasets, spatial maps where each xy point (represented by overlaid grids) has a corresponding spectrum. Datasets are represented as cubes shown spatially binned and spectrally cropped for clarity. The spatial distribution of mineral and organic matrix components is visualized by integrating over isolated peaks associated with specific signatures. Example full-resolution false color peak area heatmaps for Raman and elemental maps for EDS are shown. For the Raman dataset, averaging the apatite-containing spectra (yellow shading on the gridded image) produces the “Raman colocalized matrix spectrum,” from which all organic matrix-containing compositional parameters are calculated (right). Because organic components spectrally overlap with mineral components, an additional demixed mineral spectrum is generated: the “Raman IBA spectrum,” from which all Raman mineral-specific compositional parameters are calculated. Last, for the EDS map, averaging the calcium-containing pixels (green shading on the gridded image) produces the “EDS spectrum,” from which all elemental ratios are calculated for the biomineralogical signature.

Fig. 3.. Example calcification pathologies and correlated Raman mapping of calcifications.

For each patient sample (columns), three serial sections were obtained: (A to E) H&E staining for pathology (cell nuclei: purple; connective and necrotic tissue: pink), (F to J) von Kossa staining for mineral identification (phosphate minerals: brown) with Nuclear Fast Red counterstaining (cell nuclei: pink), and (K to O) Raman mapping of calcifications for mineral and matrix compositional analysis, where false color heatmaps of indicated components were combined for visualization. Raman maps were taken from serially contiguous calcifications where possible [i.e., the same apparent calcification as that seen in von Kossa, as in (H) and (M)] or a nearby pathologically contiguous calcification. In addition, see Fig. 2 and figs. S12 and S13 for details of Raman processing and generation of color combination maps. Note that (M) is a maximum projection image of components from a 3D Raman scan (fig. S14), taken to capture the interface between cells and mineral. Refer to Fig. 1 for patient symbol/marker identification.

Fig. 4.. Hierarchical clustering of calcifications by biomineralogical signature.

Each row represents a biomineralogical signature of an individual calcification (75 total), while columns correspond to calcification compositional metrics derived from Raman microscopy and EDS mapping. Corresponding patient numbers are shown (left) and colored according to the malignancy of the tissue surrounding the given calcification. Local calcification tissue environment (Fig. 1) is indicated by an additional color column (which was not used for clustering). Parameters were standardized. Qualitative cluster assignments: cluster 1, invasive; cluster 2, reactive stroma; cluster 3, benign lobular; cluster 5, benign ductal; cluster 6, lower-grade cancer; cluster 7, large-area DCIS; cluster 8, intermediate- to high-grade DCIS. Clusters 4 and 9 are outlier clusters. A graphical legend (below the heatmap) shows, for each compositional parameter, the spectra from which the maximum (red) and minimum (blue) values (indicated) were calculated. Raman spectra are shown normalized to the apatite ν₁ phosphate peak, except for lipid/protein (normalized to the protein-rich 2940 cm⁻¹ region). EDS spectra are shown normalized to the calcium K-alpha peak. Na, Al, Zn, and Mg EDS parameters are ratios, calculated, for the purposes of this heatmap, as Ca/X (rather than X/Ca used elsewhere in this study). In addition to improved data normality, this was done to maintain the conventional direction of the calcium-to-phosphorus ratio while avoiding trivial inverse correlations. For interpretability, however, the color scale for the elemental data is reversed such that red corresponds to a high X/Ca.

Fig. 5.. Calcification carbonate content trends and heterogeneity.

(A) EDS Ca/P correlation with Raman carbonate-to-phosphate (Carb/Phos) over individual calcifications. (B) Carbonate-to-phosphate and (C) acid phosphate-to-phosphate versus malignancy (benign, “B”; malignant, “M”) by tissue type: epithelial (i.e., ductal and lobular) and nonepithelial (i.e., mixed stromal-epithelial, invasive duct-like structures, and stromal). For patients with multiple calcifications, calcification parameters carbonate-to-phosphate and Ca/P were aggregated by malignancy and tissue type (benign/epithelial, benign/nonepithelial, malignant/epithelial, malignant/nonepithelial) by taking the means. Box plots are overlaid, and P values were calculated using the Wilcoxon rank sum test. (D) H&E-stained section from patient 19 with high-grade IDC. Approximate locations of calcifications analyzed (in the serial section) are marked (triangles) and colored by the carbonate-to-phosphate ratio, with local calcification pathologies (including in situ components) indicated. (E) Carbonate-to-phosphate ratio versus calcification local pathology for patient 19. (F) From patient 19, a Raman-SEM composite image (original SE image and Raman carbonate-to-phosphate map; right), where the Raman carbonate-to-phosphate map was calculated on the basis of a stack of eight images spanning a depth of 16 μm, showing spatial distribution of carbonate within an individual calcification. (G) H&E-stained section from patient 21 with low-grade IDC (separated box is an invasive region outside the field of view of the primary image). Benign calcifications were also present, and the cancerous region is approximately outlined. (H) Carbonate-to-phosphate versus malignancy where benign calcification markers are colored by cross-sectional proximity to invasive cancer. (I) Raman-SEM composite image (original images; right) showing the spatial distribution of carbonate within an individual calcification in patient 21. Refer to Fig. 1 for patient marker identification.

Fig. 6.. Calcification trace element composition trends with local malignancy.

Plots showing log-transformed (A) Zn/Ca, (B) Fe/Ca, (C) Al/Ca, and (D) Na/Ca against malignancy of calcification locale: benign or malignant. Where patients had multiple calcifications, data were aggregated into benign or malignant groups within the patient by taking the mean. Box plots are overlaid, and P values were calculated using the Wilcoxon rank sum test. Refer to Fig. 1 for patient symbol/marker identification. (E to H) Elemental colocalization with example calcifications from four patients (indicated). The 5-kV SE images are shown above their corresponding elemental maps for calcium, phosphorus, and the elements of interest from (A) to (D). Refer to fig. S27 for corresponding histology and EDS matrix-mineral spectral comparisons and fig. S28 for additional elemental maps.

Fig. 7.. Lipid-to-protein trends with cancer severity.

Raman lipid-to-protein (lipid/protein) values for patients significantly differed based on clinical prognostic indicators: (A) lymph node status, (B) cancer histological grade, and (C) cancer stage. (D) Patients who later had breast cancer recurrence (all had metastases). (E) Combines poor outcomes from (A) to (D) to give a composite outcome. For (A) to (E), calcification data were averaged for each patient (where multiple calcifications were analyzed) regardless of tissue type or malignancy. Box plots are overlaid, and P values were calculated using the Wilcoxon rank sum test. (F) Example poor composite outcome patient calcification regions including local pathologies (H&E), mineral locations (von Kossa), and Raman maps of lipid/protein (shown, left to right, in order of increasing average lipid/protein values). For the Raman maps, pixels that were not colocalized with mineral were masked out (black). (G) Raman average spectra from the calcifications shown in the Raman lipid/protein maps from (F) and (H), normalized to the protein-rich peak at 2940 cm⁻¹, for patients with poor (blue, lower lipid-to-protein) and good (red, higher lipid-to-protein) composite outcomes. The pertinent lipid and protein peaks are indicated for clarity (although peak areas were used for the actual calculation). (H) Example good composite outcome patient calcification regions. Pathological classification of calcification regions for poor composite outcome patients (F): patient 8: sclerosing adenosis; patient 12: atretic lobule; patient 14: desmoplastic reactive stroma near papillary carcinoma. For good composite outcome patients (H): patient 13: low- to intermediate-grade DCIS (with freeze artifact evident); patient 18: low- to intermediate-grade DCIS; patient 16: duct ectasia. The fibroadenoma patient was excluded from lipid/protein measurements due to an artifactual skewing of the ratio by a lack of Raman-evident proteins other than collagen. Refer to Fig. 1 for patient symbol/marker identification.