Fig 2: Complex mammary-specific protein expression patterns from the tissue of origin are maintained in culture.Four normal mammary tissues were digested to single cells and fixed with paraformaldehyde. In parallel, organoid cultures were generated from each of the tissues, passaged four times, digested to single cells, and fixed. Protein expression levels of 38 markers associated with mammary tumorigenesis and development were analyzed at the single-cell level in the organoid cultures and the matched tissues of origin using cytometry by time of flight (CyTOF). a Force-directed layout of organoid cultures and matched tissues of origin, where each dot represents one cell, and the closeness in space is related to their similarity in terms of protein expression patterns. Cells are colored by X-shift-defined clusters. b Delineation of the major mammary populations on the force-directed layout based on expression of standard protein markers for these cell types. c Examples of lineage-specific markers used to determine the mammary populations shown in b, including luminal (CK8 and EpCAM), mature luminal (GATA3), and basal/stem (CK14, SMA, and CD49f), as well as stromal markers (CD31, CD45, and CD140b) (warmer colors = higher expression levels). d For comparison, a normal mammary tissue was digested to single cells and fixed, and in parallel a two-dimensional culture of human mammary epithelial cells (HMECs, cultured in Lonza MEGM) was generated and passaged two times. HMECs and the matching tissue of origin were also analyzed using CyTOF as in a. The experimental schematic, delineation of cell types on force-directed layout, and cells from the tissue or cells from the culture are highlighted in the different panels for both the organoids (left column) and the HMECs (right column). e Correlation between the protein expression profiles of each HMEC or organoid cell and expression signatures derived from the major epithelial clusters in matched primary tissue. Box plots (center line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range) show the maximum r value of each cell to the major epithelial clusters, stratified by sample. Statistical significance was assessed by two-sided Mann–Whitney test (***p < 2.2e-16).

Fig 3: Computational search for pathogenic mutations that potentiate amyloid aggregation.a, Method for screening for mutations in LARKS that enable the conversion to a steric zipper. (i) The human proteome was screened for sequences capable of adopting LARKS-like structures, a structural motif underlying reversible aggregation. (ii) Mutational databases were queried to identify disease-linked variants that occur in the identified LARKS regions. (iii) The wild-type and mutant LARKS sequences were scored for their capacity to form steric zippers, the core structural motif underlying irreversible amyloid aggregation. (iv) Mutations in LARKS sequences that increase their propensity to form steric zippers were identified and further characterized. b, Plots of the steric-zipper scores for pathogenic mutations in LARKS (yellow) compared with benign SNPs (gray). The insert highlights a population of mutations with an increased likelihood of zipper formation. Many known mutations linked to amyloid aggregation were identified, including those in FUS, TDP-43, hnRNPA1, and hnRNPA2. c. Heat map of pathogenic mutations occurring in LARKS (n = 1,691). Transition of wild-type glycine to polar or hydrophobic residues, and proline to leucine mutations, are the most commonly observed variants. d, Selected proteins with LARKS containing pathogenic mutations predicted to increase zipper formation, highlighted in red. Like proteins already characterized to reversibly and irreversibly aggregate (FUS, TDP-43, and hnRNPA), the computational screen also identified zipper-promoting mutations in LARKS in the intermediate filament protein KRT8, suggesting that it is in the same class as the other proteins.Source data

Fig 4: Breast epithelial cell types form distinct structures in organoid culture.a Bright field microscopy image of a representative organoid culture demonstrating multiple structure types, highlighted by red arrowheads. b Representative examples of organoid morphology for each of the indicated mammary cell lineages. c Cells from an organoid culture (representative of n = 9 cultures) were stained for EpCAM and CD49f, and sorted using FACS into mature luminal (ML, red), luminal progenitor (LP, orange), and basal/stem cell (BS, green) populations. Cells were recultured as organoids, and then assessed by bright field microscopy (n = 9), or confocal microscopy after staining for the indicated markers (column 1, columns 2–3, and columns 4–6 are from separate images, n = 3). Scale bars = 100 µm. d, e Cells of the indicated types were recultured after sorting for >6 weeks, then imaged by bright field microscopy in d, (representative of n = 9, scale bar = 100 µm), or analyzed for EpCAM and CD49f expression by flow cytometry in e. Cultures with luminal progenitor cells that differentiated to generate other mammary cell types were ORG7, ORG43, ORG46, ORG48, ORG49, and ORG51. ORG41, ORG42, and ORG50 had luminal progenitor cells that did not differentiate to generate other mammary cell types. f Luminal progenitor cells (EpCAM+ CD49f+) sorted from two different organoid cultures were recultured as organoids, stained for CK8 (green) or EpCAM (green), CK14 (red), and with DAPI (blue), and imaged by confocal microscopy (representative of n = 6, scale bar = 100 µm). g Cells were double-sorted for EpCAM and CD49f, then re-grown and assayed as in e.

Fig 5: Radial profiles of x-ray powder diffraction of KRT8 aggregates.A slight difference between the wild-type and mutants forms of KRT8 aggregates can be observed between the rings ranging 4.3-4.6 Å. The G55A mutant also produces a much weaker ring near 10 Å compared to the others. Data for graphs is available as source data. Source data