Abstract
Three-dimensional protein localization intricately determines the functional coordination of cellular processes. The complex spatial context of protein landscape has been assessed by multiplexed immunofluorescent staining1–3 or mass spectrometry4, applied to 2D cell culture with limited physiological relevance5 or tissue sections. Here, we present 3D SPECS, an automated technology for 3D Spatial characterization of Protein Expression Changes by microscopic Screening. This workflow encompasses iterative antibody staining of proteins, high-content imaging, and machine learning based classification of mitotic states. This is followed by mapping of spatial protein localization into a spherical, cellular coordinate system, the basis used for model-based prediction of spatially resolved affinities of various mitotic proteins. As a proof-of-concept, we mapped twelve epitopes in 3D cultured epithelial breast spheroids and investigated the network effects of mitotic cancer drugs with known limited success in clinical trials6–8. Our approach reveals novel insights into spindle fragility and global chromatin stress, and predicts unknown interactions between proteins in specific mitotic pathways. 3D SPECS’s ability to map potential drug targets by multiplexed immunofluorescence in 3D cell cultured models combined with our automized high content assay will inspire future functional protein expression and drug assays.
Iterative antibody labeling overcomes the spectral limit of total number of fluorescent antibodies that can be applied simultaneously to individual cells1,3. We successfully extend this technique of chemically bleached fluorescently labeled antibodies, to 3D cell cultured spheroids in Matrigel9 combined with drug treatment (see Supplementary Table 1). Our setup (Fig. 1a) uses confocal laser scanning microscopy together with automated pre-screen classification by machine learning and motorized in-built micro pipetting robot to identify and comprehensively stain mitotic phases. Naturally, mitotic cells self-organize, structure and typically segregate in different orientations, rendering a direct comparison of different data sets impossible. To overcome this limitation, we propose a novel representation named SpheriCell for spatial alignment of subcellular images. It infers a spherical coordinate system of the cellular space, using the spindle axis perpendicular to the metaphase plate to define the orientation of the vertical axis, while the metaphase plate defines the equatorial plane. Within this spherical coordinate system, protein concentrations are measured as 3D partitions in three symmetrical sets of spherical sectors and six shells. For enhanced visualization, mean values of 3D partitions are projected onto a two-dimensional longitudinal plane (Fig. 1b). Hence, we screened 6,272 image stacks, identified 1,217 mitotic events resulting in 284,778 mean intensity values of 3D partitions. We illustrate differences between tumorigenic and non-tumorigenic 3D cell lines in more physiological conditions still accessible by high-throughput screening5. The applied human epithelial MCF10 breast cancer progression model compares near-diploid cell line MCF10A10 forming polarized spheroids11 with the tumorigenic line MCF10CA, which bear activating mutations of HRAS and PIK3CA, amplified MYC, but wild-type TP5310.
Using 3D SPECS, we first studied sub-cellular differences in protein localizations in the tumorigenic and non-tumorigenic 3D cell lines distinguishing between the cell cycle states metaphase and segregation. Here we assessed co-localization affinities of proteins and their preferred localization to subcellular compartments by our mathematical modeling approach. We thus unraveled network-wide treatment effects on mitotic spindle organization, spindle assembly checkpoint (SAC), and complementary cell fate indicators. The spindle assembly checkpoint (SAC) control includes e.g. the chromosome passenger complex (CPC) comprising BIRC5, Borealin, INCENP, and Aurora B12, which inhibit anaphase promoting complex (APC/C) most efficiently through mitotic checkpoint complex (MCC) containing BUB1 β (BUBR1)13. Failures in these checkpoints can lead to disruption of mitosis and subsequent autophagic or necrotic events.
We confirmed known localizations of cellular proteins and known mitotic checkpoints for untreated cells as described before12–21, supporting the utility of 3D SPECS (Fig. 2a). MCF10CA staining patterns resembled those of MCF10A, showing a slightly reduced average DAPI signal due to increased cell size. Highest protein concentration increases were observed for γ-H2AX and Aurora A, contrasted by a reduction strongest for γ-Tubulin (Fig. 2b). Increased levels of γ-H2AX16, a marker for double strand breaks, may reflect higher chromosomal stress. Higher intensity levels of Aurora A, which is upregulated during mitosis and localizes mostly towards centrosomes14, in MCF10CA is consistent with previously described effects upon activation of Raf-122, downstream of the oncogenic RAS pathway.
We then analyzed the effects of twelve targeted inhibitors on mitotic proteins of dividing MCF10A and MCF10CA cells (Fig. 3a,b), specifically on protein concentrations and preferred localizations. To compare spatial distribution patterns of protein intensities, in each cell, 18 subcellular compartments were defined by a combination of six eccentricity shells with three orientations relative to the division plane. In analogy to calculating a center of mass, we specified measures of spatial intensity distributions (Supplementary Note 1). Fig. 3a visualizes significant concentration fold changes and spatial changes in eccentricity and orientation compared between cell lines, mitotic phases and for inhibitors relative to controls. Notably, comparisons between cell lines showed more pronounced effects on concentration fold changes than on spatial distributions. MCF10CA cells showed significantly higher γ-H2AX and Aurora A concentrations, but lower γ-Tubulin concentrations. Changes in eccentricity and orientation of the localization pattern could be observed in MCF10CA relative to MCF10A cells mostly during metaphase. Obviously, as indicated for the comparison between segregation and metaphase, the mitotic phase strongly influenced the spatial distributions of most observed proteins. Changes in the distribution pattern for DAPI and γ-H2AX reflect the movement of the nucleus towards the cell division axis and to higher eccentricity while the other proteins move closer to the cell division plane.
During segregation, relative to metaphase, both cell lines showed a known decrease in CDC20 concentration23 and elevated BIRC5 concentrations, whereas the CENP-A was only increased in MCF10A cells. Inhibitors mostly increased the concentrations of several proteins but hardly affected their spatial distributions (Fig. 3b). Here, the inhibitions of master regulator AURKB12 and also Haspin known to be implicated in Aurora B positioning12 showed a prominent effect across nearly all proteins in MCF10A as well as MCF10CA cells. Moreover, we detected broad effects of increased DNA damage by Topoisomerase II poisoning24. Interestingly, MCF10CA cells appeared to be more sensitive to mitotic spindle interference, reflected by effects on Aurora A and CENP-E. Concordantly, inhibition of Aurora A affected more proteins in MCF10CA spheroids compared to MCF10A. MCF10CA showed more pronounced effects of PLK1 inhibition. Contrarily, although KIF11 (Eg5) and KIFC1 (HSET) facilitate separation and clustering of centrosomes25, the effects due to inhibition of KIF11 were restricted to MCF10A cells. High natural levels of γ-H2AX intensity in MCF10CA were not increased by our treatments as observed in MCF10A. Also, BIRC5 concentration was affected in MCF10A but not in MCF10CA cells. In contrast to effects on concentrations, the eccentricity and the orientation of distribution patterns were less affected by inhibitions except for Haspin and PLK1 being two notable exceptions.
To study intracellular distributions and localized interactions between proteins, we developed a non-linear model that was calibrated with our partitioned mitotic protein intensity data (Supplementary Note 1). To capture interactions between proteins in a simplified manner, the model describes concentrations at steady state for monomers, homo- and heterodimers of all proteins bound to subcellular compartments with first and second order interactions. Affinities of every protein to subcellular compartments, and affinities between proteins were estimated by model fitting. To predict new affinities between proteins, we started by fitting a model of interactions from literature in Ingenuity Pathway Analysis (IPA) regarded as ground truth. This initial model was fitted to our untreated control cells. Then, by sequential forward selection, new mutual affinities between proteins were additionally included in the model and pertained if model fits were significantly improved, based on likelihood-ratio testing. Fig. 3c visualizes the 19 known interactions overlaid with all 16 additionally predicted mutual affinities. Interestingly, colocalization affinities between species do not necessarily imply functional relations and pathway interactions do not require high affinities. For example, we identified known interactions of γ-Tubulin with CDC2026 as well as known DNA-binding of BIRC512 (Fig. 3c). While we triggered DNA damage pathways with Topoisomerase II poisoning and inhibition of CHK127, indications of cell fate could be inferred from double strand break marker γ-H2AX16, necrosis associated HMGB117, and autophagic vesicle marker MAP1LC3A (LC3)18. BIRC5 was predicted by the model to interact with LC3, which links mitotic surveillance and autophagy pathways18. The model predicted interactions of γ-H2AX, γ-Tubulin and β-Tubulin with several other proteins, which might indicate indirect interactions with the mitotic spindle or the cytoskeleton. Coefficients describing mutual affinities between proteins are depicted in Fig. 3d. Importantly, the fraction of a protein that is localized to a mitotic bin due to mutual interactions with other proteins does not only depend on affinity coefficients but may require high affinities of interacting species to the respective mitotic bin. Estimated fractions of proteins recruited to subcellular compartments due to mutual interactions with other proteins, as well as estimated affinities to subcellular compartments are shown in Supplementary Fig. 1. The highest values of mutual affinities with other proteins were found between CENP-E molecules as described earlier28 and for the newly predicted binding of β-Tubulin to HMGB1 (Fig. 3d).
We next inspected changes in mutual affinities between mitotic proteins and their affinities to subcellular compartments upon drug treatment. To this end, the model with known and additionally predicted interactions was fitted to datasets from cells treated with inhibitors to estimate mutual affinities between proteins and to subcellular compartments. In Fig. 3e, estimated affinity coefficients for PLK1 as an exemplary inhibitor are visualized. Inhibition of PLK1 affected the localization patterns of several proteins and caused a strong specific shift in mutual affinities among several studied proteins in comparison to untreated cells (Fig. 3d,e). In those cases, many estimates of mutual affinities decreased, indicated by changes to blue color code. Contrarily, affinities to subcellular compartments during metaphase and segregation showed almost no differences to untreated cells (Supplementary Fig. 1e-h). It is tempting to speculate that effects of PLK1 inhibition are mediated through its involvement in spindle network formation29. Specifically, predicted affinity of β-Tubulin to γ-H2AX, HMGB1 and INCENP decreased, and chromosome affinity of BIRC5 appears to be reduced by inhibition of PLK1, whereas the predicted affinity of INCENP to HMGB1 is increased (Fig. 3e). Reduced chromosome affinity of BIRC5 after PLK1 inhibition is in accordance with the finding that phosphorylation of BIRC5 by PLK1 is required for a proper chromosome alignment during mitosis12.
To summarize, we present 3D SPECS, a high-content screening assay employing automated iterative antibody labeling in 3D cell cultures. It allowed us to compare system-wide interactions between twelve proteins of two cell lines in two mitotic phases, upon twelve individual treatments. High automation comprises detection of mitoses, iterative staining and imaging, 3D partitioning, modeling and visualization using SpheriCell, a novel approach that does not require antibody image segmentation, nor alignment of cell division in 3D. This explorative approach recapitulated prior knowledge on proteins involved in mitosis and allowed the generation of novel hypotheses in mitotic pathway signaling. Most prominently, we discovered up-regulation of γ-H2AX in tumorigenic MCF10CA cells compared to MCF10A. Further, γ-H2AX has a higher sensitivity to interference in MCF10A, which in turn appears to have a more robust spindle apparatus. Our novel combined imaging and mathematical modeling approach allowed us to disentangle inhibitor-mediated protein localization and binding affinity changes and showed that changes in affinities between proteins (Fig. 3d,e) were more pronounced than changes in individual protein localizations (Supplementary Fig. 2d,e). As an example, we focused on the measured inhibitions of PLK1 activity, which is responsible for establishing the mitotic spindle and which is frequently hyper-activated in cancer30. Subsequent reduction in chromatin affinity of BIRC5 could be explained by its dependency on PLK1 phosphorylation12, most likely intertwined with CPC function.
Our method can be readily extended to directly determine the activity of proteins by phospho-specific antibodies. For a more fine-grained assessment of protein localization additional nuclear or membrane staining can be easily integrated into 3D SPECS. The SpheriCell approach that renders intuitively simple and comprehensive visualization of protein localization in cell division can also be amended by including polar landmarks of non-dividing cells. Taken together we have demonstrated 3D SPECS as a novel workflow unraveling thus unprecedented levels of details in changes of protein localization and interaction upon drug treatment of three-dimensional cell cultures.
Author contributions
L.M. and C.C conceived the experiments and subcellular visualization strategy. L.M., K.J. and M.W. established antibody staining and drug treatment protocols. L.M. developed automated iterative staining workflow, conducted experiments, and image analysis. S.K. developed the interaction model and performed statistical tests. C.C. and R.E. supervised this project. L.M., S.K., C.C, and R.E. wrote the manuscript. All authors commented on the manuscript.
Competing financial interests
The authors declare no competing financial interests.
Online Methods
Mitotic proteins were assessed after 48h drug treatment by iterative immunofluorescence labeling. The antibodies were either labeled with one of DyLight 550 / Cy3 or DyLight 650 / Cy5. Both types of dyes could be used interchangeably in terms of excitation and emission spectra.
While Matrigel is essential for acinar growth of spheroids11, it also dissolves quickly when the bleaching solution is applied. Therefore, we have used DyLight instead of Cy or Alexa31 labeled antibodies, as they bleach much faster and have a very strong fluorescence signal nevertheless. Applying the bleaching solution significantly longer than 5 minutes at a time typically dissolved the Matrigel carrying the spheroids.
All treatments were imaged at 30 spheroids that showed at least one mitosis each. 196 stacks per well, eight wells per round and four rounds resulted in 6272 image stacks with 21 slices each that were automatically pre-screened for mitotic events. Iterative high resolution images of 30 positions per well in eight wells in each of four rounds totaled in 960 identified spheroids that were imaged each with 31 slices, after staining and after bleaching in six iterations.
For analysis and visualization, every mitosis was aligned along its division plane for a spherical neighborhood that contains the cell division in equatorial axis (see Fig. 1b).
3D cell culture and drug treatment
Human mammary epithelial MCF10A pBabePuro cells were kindly obtained from Zev Gartner Lab; MCF10CA1d.cl1 (MCF10CA) cells from Karmanos Cancer Institute. Eight well Lab-Tek Chambered Coverglass slides (Sigma 155411) were treated with 2 M NaOH for 20 min and rinsed twice for 10 min with MilliQ water. Ten μl Matrigel (growth factor reduced, phenol red-free, Corning 356231) per well was added on ice with pipette tips pre-cooled to -20 °C. MCF10A and CA cells were seeded with 2% Matrigel in Growth Medium overnight. Growth medium was adapted from Debnath et al.32 and is based on DMEM/F12 (no phenol red, Gibco 21041-33), with 5% Horse Serum (Gibco 16050-122), 20 ng/ml EGF (Sigma E9644-.2MG), 0.5 mg/ml Hydrocortisone (Sigma H0888-1g), 100 ng/ml Cholera Toxin (Sigma C8052-1MG), 10 μg/ml Insulin (Life Technologies 12585014). For the inhibition experiments, the cells were treated for 48h at one day after seeding.
Inhibitors
Drugs, suppliers, and concentrations used were Barasertib (Aurora B inhibitor; alternative name AZD1152-HQPS; SelleckChem S1147; 1.11 nM); CHR-6494 (Haspin inhibitor; MedChem Express HY-15217; 500 nM); CW069 (HSET inhibitor; SelleckChem S7336; 25.0 μM); Etoposide (Topoisomerase II inhibitor; SelleckChem S1225; 333 nM); GSK461364 (PLK1 inhibitor; SelleckChem S2193; 2.20 nM); GSK923295 (CENP-E inhibitor; SelleckChem S7090; 3.20 nM); Ispinesib (KIF11 inhibitor; alternative name SB-715992; SelleckChem S1452; 1.70 nM); MK-5108 (Aurora A inhibitor; alternative name VX-689; SelleckChem S2770; 0.576 nM); MK-8776 (CHK1 inhibitor; alternative name SCH 900776; SelleckChem S2735; 9.00 nM); Paclitaxel (microtubule inhibitor; SelleckChem S1150; 2.67 nM); Vinblastine (microtubule inhibitor; Sigma V1377; 2.40 nM); and YM155 (BIRC5 inhibitor; SelleckChem S1130; 0.540 nM).
Antibodies and labeling kits
Antibodies were conjugated with DyLight 550 and 650 Microscale labeling kits per supplier reference manual (Sigma, 84531 and 84536, respectively) unless otherwise stated. Antibody targets, dilutions, supplier, and conjugation method in iterative staining order were CENP-E (1:400; Abnova MAB1924; conjugated DyLight 550); BubR1 (1:600; Thermo Fisher MA5-16036; pre-conjugated with DyLight 650); beta-Tubulin (1:5000; Abcam ab11309; pre-conjugated with Cy3); CDC20 (1:400; Bethyl A301-179A; conjugated DyLight 550); gamma-Tubulin (1:12000; Abcam ab176404; pre-conjugated with Cy3); LC3A, microtubule-associated proteins 1A/1B light chain 3A (1:400; Novus NB100-2331; conjugated DyLight 650); BIRC5 (1:1000; Abcam ab176402; pre-conjugated with Cy3) INCENP (1:1000; Thermo Fisher MA5-17100; conjugated DyLight 650); Aurora A (1:6000; Abcam ab176375; pre-conjugated with Cy3); CENP-A (1:500; Abnova PAB18324; conjugated DyLight 650); HMGB1 (1:3000; Abcam ab176398; pre-conjugated with Cy3); H2AX (1:2500; Cell Signaling 9718BF; conjugated DyLight 650).
Iterative antibody labeling
Cell fixation was based on a protocol from Debnath et al.32, with 1.85% formaldehyde solution (Sigma 252549) added to the medium for 10 minutes. Cells were rinsed twice with PBS and permeabilized for 10 min at RT with 0.5% TX-100 pre-chilled to 8 °C, washed three times with PBS-glycine (130 mM NaCl, 7 mM Na2HPO4, 3.5 mM NaH2PO4, 100 mM glycine) for 10 minutes, and blocked overnight at RT in a blocking solution consisting of IF-wash solution33 (130 mM NaCl, 7 mM Na2HPO4, 3.5 mM NaH2PO4, 7.7 mM NaN3, 0.1% BSA, 0.2% Triton X-100, 0.05% Tween 20) with 10% goat serum (Sigma G9023-10ML) and 1:1000 DAPI (Sigma D8417-1MG), inside an opaque EMBL microscope incubation chamber. For each iteration, two antibodies were diluted in freshly prepared blocking solution and stored in a slide within a 4x LabTek holder (EMBLEM LTT-01 and LTT-02). They were automatically pipetted into the wells by a peristaltic pump of the ProCellcare 5030 system (ProDesign) and incubated for 3h, washed twice with IF-wash for 5 min and three times with PBS for 5 min. After imaging, freshly prepared H2O2 bleaching solution34 containing 3% H2O2 (AppliChem, Cat. No. 121076) and 0.1M Na2CO3/NaHCO3 buffer at pH ≈ 10 was stored in another LabTek. It was automatically applied for 5 minutes and washed twice with PBS for 5 minutes. Standard incubator light source was switched on during bleaching with Energenie EG-PM2. Pipetting positions were planned with Zeiss Zen blue (www.zeiss.com/zen) and pipetting workflow was implemented in LabView (www.ni.com/labview).
Pre-screen
During blocking, slides were imaged with a Yokogawa CSU-X1 spinning disc unit attached to a Zeiss Observer Z1 inside an EMBL incubation chamber. 196 image stacks of 401.6 × 400 × 60 μm were taken per well with a plan-apochromat 20x/0.8 NA objective. Stack slices had 1004×1002 pixels, step size was 3 μm and exposure 40 ms. Candidate mitotic positions were detected via their DAPI signal by a custom KNIME workflow and selected or expanded manually if necessary. The automatic selection excludes monolayer slices and uses a supervised tree ensemble classifier (comparable to a random forest). For each treatment and cell line, 30 positions of spheroids with at least one mitosis each were selected for imaging during the iterative staining workflow.
Acquisition of iterative staining images
After each round of bleaching or staining, spheroids were automatically imaged with a laser scanning confocal Zeiss LSM 780 connected to the same Axio Observer as the spinning disc unit. Stack dimensions were 106.07 × 106.07 × 60 μm, with 512 × 512 pixels per slice and 2 μm Z steps. Objective was plan-apochromat 20x/0.8 NA, pixel dwell time 3.15 μs, and pinhole 32 μm. Emission spectra were taken at 410 – 489 nm (DAPI), 560 – 586, 586 – 612, and 612 – 630 (three parts of Cy3 / DyLight 550), and 638 – 758 nm (Cy5 / DyLight 650).
Image processing
Splitting the emission spectrum from 560 to 630 nm in three parts allowed for post-acquisition exposure correction. Only for BIRC5 it was necessary to exclude the strongest emission channel from the labeling image. All remaining split channels were averaged. Mitotic DAPI signal was segmented in 3D by a customized region growing algorithm35 with manual seeds and borders to closely neighboring nuclei, especially in Z-direction. Segmented areas were manually annotated with their mitotic phase and ana- / telophases joined. Assignment of segregating split chromatin regions to a single dividing cell was verified with β-Tubulin staining. Registration of consecutive stacks per imaging position used subpixel alignment of Fiji36 plugin Correct 3D Drift37 followed by MultiStackReg38 with scaled rotation. Registration of the stacks was verified manually. For the spherical neighborhood, images were interpolated linearly in Z to match the X/Y pixel dimensions. Spherical segment angles were generated by Recursive Zonal Equal Area Sphere Partitioning Toolbox39 in MATLAB (www.mathworks.com) for 180 areas. A custom R script was created to join the areas to segments and to fit them in size and orientation to the individual mitoses, which were identified with 3D ellipsoid fit of the 3D ImageJ suite40. Metaphases could use those values as-is, but the size of segregating cells was overestimated by the ellipsoid fit and replaced by the centroid distances of their individual chromatin regions. Their 3D orientation used the average of the first two eigenvectors and the normalized centroid to centroid vector as third. The six spherical neighborhood shells grow linearly in their radius from the mitotic center, and the inner four span the identified nucleus area. Missing cells due to loss of Matrigel or failed registration were removed from further analysis. All image processing steps were embedded in a KNIME35 workflow.
Visualization
Antibody intensities are depicted as color-coded mean values for three distinct segment classes (phi levels, Fig. 1b). Decreased intensity dependent on imaging depth was corrected by the mean DAPI intensity within the individual spherical neighborhood and mean DAPI intensity per well in pre-screen images after z-projection. To avoid an artificial increase in background signal of antibodies, DAPI intensities below a minimum threshold were excluded. Highlighting of partitions was determined by the control intensity over all rounds. Data was visualized with the packages ggplot241, EBImage42, and shiny (shiny.rstudio.com) for R (www.r-project.org).
Statistics and mathematical modeling
To test for significance of comparisons between controls and inhibitor treatments, we applied two-sample t-tests. For comparisons between two groups, intensity values for all subcellular compartments were collected within each group. In total, 54 comparisons were made between measurements of each protein. Therefore, t-tests with p<0.05/54 were regarded as significant (Bonferroni correction for multiple testing). The interaction model was implemented in MATLAB. For model calibration, we applied the solver lsqnonlin using the trust-region-reflective algorithm. Details on the model formulation and fitting can be found in Supplementary Note 1.
Interpretation
Known interactions form literature were generated through the use of QIAGEN’s Ingenuity® Pathway Analysis (IPA®, QIAGEN Redwood City, www.qiagen.com/ingenuity). They are supported by at least one reference from the literature, from a textbook, or from canonical information stored in the Ingenuity Pathways Knowledge Base.
Data and code availability
Spherical neighborhoods, their image sources, and custom software code is available upon request.
Acknowledgements
We thank Sabine Aschenbrenner for support with lab techniques, Siegfried Winkler, Leo Burger, and Helmuth Schaar for microscopy hardware, Antonio Politi for imaging advice and ZEN black macro interface, Maria Maier for assistance with 3D renderings, Christian Dietz for continued development of KNIME image processing, and Clarissa Liesche and Joël Beaudoin for critical comments. The authors acknowledge support by the state of Baden-Württemberg through bwHPC.