Secretase

All data sets were recorded using 300 keV electrons. The γ-secretase, β-galactosidase, and mitoribosome data sets were recorded manually; the complex-I data set was recorded automatically using the EPU software from FEI. For all data sets, fields of views that showed signs of significant drift, charging, or astigmatism were discarded. For the γ-secretase data, this assessment was made after alignment using the algorithm by Li et al. (2013) (link). Movies on the Falcon-II detectors on the Polara and Titan Krios microscopes were intercepted using a system that was developed in-house (Bai et al., 2013 (link)). CTF parameters were estimated using CTFFIND3 (Mindell and Grigorieff, 2003 (link)), and the particles were picked in a semi-automated manner, using EMAN2 (Tang et al., 2007 (link)) for the mitoribosome, and RELION for the three other data sets. Selection of particles for the final 3D reconstruction was performed using reference-free 2D class averaging and 3D classification in RELION (Scheres, 2012 (link)), and the final maps before and after movie processing were calculated using RELION’s 3D auto-refine, followed by automated B-factor sharpening (Rosenthal and Henderson, 2003 (link)) and correction for the MTF of the detector. All resolutions were based on the gold-standard FSC = 0.143 criterion (Scheres and Chen, 2012 (link)), and FSC curves were corrected for the effects of soft masking by high-resolution noise substitution (Chen et al., 2013 (link)). Density figures were made using UCSF Chimera (Pettersen et al., 2004 (link)).

Unfiltered reconstructions for TRPV1 (EMD-5778), γ-secretase (EMD-3061) and β-galactosidase (EMD-2984) were obtained from the EMDB model challenge website (http://challenges.emdatabank.org). Initial map targets for coordinate refinement were generated by applying uniform filtering to the original EMDB entries. The deposited map for TRPV1 (EMD-5778) was globally sharpened using an additional B-factor of −100 Ų. Coordinate refinement was performed using real-space refinement against the initial map as implemented in cctbx/PHENIX (Adams et al., 2010 (link)) with additional restraints on secondary structure. Residue-grouped atomic B-factors were refined in reciprocal space. For TRPV1 and β-galactosidase, non-crystallography symmetry (NCS) restraints were employed to account for rotational and dihedral symmetry. Reference restraints for PETG were obtained with phenix.elbow from the crystal structure of 2-phenylethyl 1-thio-β-D-galactopyranoside (Brito et al., 2011 (link)). Building of glycans used glycan modeling tools available in Coot. Local resolution for all maps was estimated by local FSC computations using an in-house Python program locres.py. Local resolution mapped to the coordinate models using the measure_mapValues function in UCSF Chimera (Pettersen et al., 2004 (link)), and the residue-averaged resolution was used to assign local resolution-scaled reference restraints. Model rebuilding was performed in Coot (Emsley and Cowtan, 2004 (link)). Half-set reconstructions for cross validation were obtained from the EMDB model challenge. Model maps were generated by inverse Fourier transform using B-factor-weighted electron form factors (Colliex et al., 2006 ). The FSC_ref between model map and half-set 3D reconstructions was used to assess over-fitting. To this end, coordinates were first refined against the full map. Coordinates were then randomly displaced by a maximum of 0.5 Å and subjected to three cycles of real-space refinement against one of the half maps (work map) using the same protocol outlined above. The other half map (test map) was used as the test map for cross-validation. Following refinement, we computed the FSC of the refined model against the work map (FSC_work) and the cross-validated FSC between refined model and the test map (FSC_test). All FSCs were computed using a structure mask obtained from the respective EMDB entry that was low-passed filtered to 60 Å.
To compare model refinement against deposited EMDB entries and LocScale maps, PDB-deposited coordinate models associated with the respective EMDB entry were randomly perturbed by applying atom shifts of up to 0.4 Å to serve as starting models. Five iterations of local and global real-space coordinate refinement against the respective EMDB or LocScale map were each followed by refinement of atomic B-factors in reciprocal space. As above, secondary structure restraints, resolution-dependent weights and NCS restraints were employed. EMRinger scores were computed using phenix.em_ringer (Barad et al., 2015 (link)); all other validation scores were obtained using MOLPROBITY (Chen et al., 2010 (link)). All cross-validation FSC and real-space correlations were computed against the original reconstruction and the deposited EMDB map, respectively.

Human mammary immortalized cells (MCF10A), and the TNBC cell line MDA-MB-231 were purchased from American Type Culture Collection. MCF10A cells were grown in a 1:1 mixture of Dulbecco’s modified Eagle’s medium (DMEM) and Ham’s F12 medium with 20 ng mL⁻¹ human epidermal growth factor, 100 ng mL⁻¹ cholera toxin, 0.01 mg mL⁻¹ bovine insulin, 500 ng mL⁻¹ hydrocortisone, and 5% horse serum. MDA-MB-231 cells were grown in DMEM containing L-glutamine and sodium pyruvate, supplemented with 10% fetal bovine serum and 1% antibiotic and antimycotic solution in a humidified atmosphere of 5% CO2 at 37°C in an incubator. Human BCSC cells: ALDH⁺ and CD44⁺/CD24⁻, and BC cells: ALDH- were purchased from Celprogen (San Pedro, CA, United States) and maintained in human BCSC expansion and undifferentiation media. DAPT (γ-secretase), cycloheximide (CHX), chloroquine (CQ), and MG132 were purchased from Sigma (St. Louis, MO).