1. Cancer Biology
  2. Computational and Systems Biology
Download icon

Highly multiplexed immunofluorescence imaging of human tissues and tumors using t-CyCIF and conventional optical microscopes

  1. Jia-Ren Lin
  2. Benjamin Izar
  3. Shu Wang
  4. Clarence Yapp
  5. Shaolin Mei
  6. Parin M Shah
  7. Sandro Santagata
  8. Peter K Sorger  Is a corresponding author
  1. Harvard Medical School, United States
  2. Dana-Farber Cancer Institute, United States
  3. Broad Institute of MIT and Harvard, United States
  4. Harvard University, United States
  5. Brigham and Women’s Hospital, Harvard Medical School, United States
Tools and Resources
Cite this article as: eLife 2018;7:e31657 doi: 10.7554/eLife.31657
12 figures, 4 tables and 7 additional files


Figure 1 with 2 supplements
Steps in the t-CyCIF process.

(A) Schematic of the cyclic process whereby t-CyCIF images are assembled via multiple rounds of four-color imaging. (B) Image of human tonsil prior to pre-staining and then over the course of three rounds of t-CyCIF. The dashed circle highlights a region with auto-fluorescence in both green and red channels (used for Alexa-488 and Alexa-647, respectively) and corresponds to a strong background signal. With subsequent inactivation and staining cycles (three cycles shown here), this background signal becomes progressively less intense; the phenomenon of decreasing background signal and increasing signal-to-noise ratio as cycle number increases was observed in several staining settings (see also Figure 1—figure supplement 1).

Figure 1—figure supplement 1
Reduction in background signal intensity with repeated cycles of bleaching.

(A–C) Intensity distributions for three fluorescence channels (FITC/Alexa-488, Cy3/Alexa-555 and Cy5/Alexa-647) prior to pre-bleaching (blue), and after 1, 2 or 3 cycles of bleaching (black, red and green lines, respectively). With increasing number of bleaching cycles, the background signal is reduced 10- to 100-fold.

Figure 1—figure supplement 2
t-CyCIF using antibodies labelled with Zenon Alexa-555 Fab fragments.

A human tonsil specimen was stained with unconjugated anti-CD11b antibody and then with Alexa-488 conjugated anti-Rabbit secondary antibody (left) or with the same anti-CD11b antibody following incubation with Zenon Alexa-555 (ThermoFischer; right). The Zenon Fab fragments generate non-covalent immune complexes that ‘label’ the primary antibody in a manner that is stable to subsequent processing steps.

Figure 2 with 2 supplements
Multi-scale imaging of t-CyCIF specimens.

(A) Bright-field H&E image of a metastasectomy specimen that includes a large metastatic melanoma lesion and adjacent benign tissue. The H&E staining was performed after the same specimen had undergone t-CyCIF. (B) Representative t-CyCIF staining of the specimen shown in (A) stitched together using the Ashlar software from 165 successive CyteFinder fields using a 20X/0.8NA objective. (C) One field from (B) at the tumor-normal junction demonstrating staining for S100-postive malignant cells, α-SMA positive stroma, T lymphocytes (positive for CD3, CD4 and CD8), and the proliferation marker phospho-RB (pRB). (D) A melanoma tumor imaged on a GE INCell Analyzer 6000 confocal microscope to demonstrate sub-cellular and sub-organelle structures. This specimen was stained with phospho-Tyrosine (pTyr), Lamin A/C and p-Aurora A/B/C and imaged with a 60X/0.95NA objective. pTyr is localized in membrane in patches associated with receptor-tyrosine kinase, visible here as red punctate structures. Lamin A/C is a nuclear membrane protein that outlines the vicinity of the cell nucleus in this image. Aurora kinases A/B/C coordinate centromere and centrosome function and are visible in this image bound to chromosomes within a nucleus of a mitotic cell in prophase (yellow arrow). (E) Staining of a melanoma sample using the GE OMX Blaze structured illumination microscope with a 60X/1.42NA objective shows heterogeneity of structural proteins of the nucleus, including as Lamin B and Lamin A/C (indicated by yellow arrows) and part of the nuclear pore complex (NUP98) that measures ~120 nm in total size and indirectly allows the visualization of nuclear pores (indicated by non-continuous staining of NUP98). (F) Staining of a patient-derived mouse xenograft breast tumor using the OMX Blaze with a 60x/1.42NA objective shows a spindle in a mitotic cell (beta-tubulin in red) as well as vesicles staining positive for VEGFR2 (in cyan) and punctuate expression of the EGFR in the plasma membrane (in green).

Figure 2—figure supplement 1
Flat-field and shading correction for stitched images.

A large melanoma specimen (same as shown specimen as in Figure 2A–B) was imaged with a 20X/0.8NA objective on a CyteFinder, and a total 165 frame images were assembled to one image using the ASHLAR algorithm. Representative channels (Hoechst 33342 and S100-Alexa-488) of the stitched images before (left) and after (right) correction for uneven illumination using the BaSiC algorithm (Peng et al., 2017) (see Materials and methods).

Figure 2—figure supplement 2
OMX super-resolution t-CyCIF images.

(A) Single-channel images for the composite image shown in Figure 2E. This melanoma sample was imaged using the GE OMX Blaze structured illumination microscope with 60X/1.42NA objectives. (B). Single-channel images for the composite image shown in Figure 2F. This breast cancer xenograft sample was also imaged using the OMX Blaze.

t-CyCIF imaging of normal tissues.

(A) Selected images of a tonsil specimen subjected to 10-cycle t-CyCIF to demonstrate tissue, cellular, and subcellular localization of tissue and immune markers (see Supplementary file 1 for a list of antibodies). (B) Selected cycles from (A) demonstrating sub-nuclear features (Ki67 staining, cycle 1), immune cell distribution (cycle 2), structural proteins (E-Cadherin and Vimentin, cycle 5) and nuclear vs. cytosolic localization of transcription factors (NF-kB, cycle 6). (C) Five-cycle t-CyCIF of human skin to show the tight localization of some auto-fluorescence signals (Cycle 0), the elimination of these signals after pre-staining (Cycle 1), and the dispersal of rare cell types within a complex layered tissue (see Supplementary file 1 for a list of the antibodies).

Figure 4 with 1 supplement
Efficacy of fluorophore inactivation and preservation of tissue integrity.

(A) Exemplary image of a human tonsil stained with PCNA-Alexa 488 that underwent 0, 15, 30 or 60 min of fluorophore inactivation. (B) Effect of bleaching duration on the distribution of anti-PCNA-Alexa 488 staining intensities for samples used in (A). The distribution is computed from mean values for the fluorescence intensities across all cells in the image that were successfully segmented. The gray band denotes the range of background florescence intensities (below 6.2 in log scale). (C) Effect of bleaching duration on mean intensity for nine antibodies conjugated to Alexa fluor 488, efluor 570 or Alexa fluor 647. Intensities were determined as in (B). The gray band denotes the range of background florescence intensities. (D) Impact of t-CyCIF cycle number on tissue integrity for four exemplary tissue cores. Nuclei present in the first cycle are labeled in red and those present after the 10th cycle are in green. The numbers at the bottom of the images represent nuclear counts in cycle 1 (red) and cycle 10 (green), respectively. (E) Impact of t-CyCIF cycle number on the integrity of a TMA containing 48 biopsies obtained from 16 different healthy and tumor tissues (see Materials and methods for TMA details) stained with 10 rounds of t-CyCIF. The number of nuclei remaining in each core was computed relative to the starting value; small fluctuations in cell count explain values > 1.0 and arise from errors in image segmentation. Data for six different breast cores is shown to the right. (F) Nuclear staining of a melanoma specimen subjected to 20 cycles of t-CyCIF emphasizes the preservation of tissue integrity (22 ± 4%). (G) Selected images of the specimen in (F) from cycles 0, 5, 15 and 20.

Figure 4—source data 1

Mean intensity versus bleach time for multiple antibodies (Figure 4C).

Figure 4—source data 2

Intensity distribution for single cells versus bleach time for one antibody (Figure 4B).

Figure 4—source data 3

Cell counts dependent on number of staining cycles (Figure 4E).

Figure 4—figure supplement 1
Impact of bleaching time on fluorophore inactivation.

Intensity distributions for specimens stained with an antibody against Ki67 coupled to the following fluorophores: (A) Alexa-488, (B) Alexa-570 and (C) Alexa-647 prior to bleaching (blue curves) and after 15, 30 or 45 min of bleaching (black, red and green curves respectively). The distributions were calculated from the average fluorescence intensities of single cells as determined after image segmentation.

Design of a 16-cyle experiment used to assess the reliability of t-CyCIF data.

(A) t-CyCIF experiment involving two immediately adjacent tissue slices cut from the same block of tonsil tissue (Slide A and Slide B). The antibodies used in each cycle are shown (antibodies are described in Supplementary file 2). Highlighted in blue are cycles in which the same antibodies were used on slides A and B at the same time to assess reproducibility. Highlighted in yellow are cycles in which antibodies targeting PCNA, Vimentin and Tubulin were used repeatedly on both slides A and B to assess repeatability. Blue arrows connecting Slides A and B show how antibodies were swapped among cycles. (B) Representative images of Slide A (top panels) and Slide B specimens (bottom panels) after each t-CyCIF cycle. The color coding highlighting specific cycles is the same as in A.

Figure 6 with 1 supplement
Impact of cycle number on repeatability, reproducibility and strength of t-CyCIF immuno-staining.

(A) Plots on left: comparison of staining intensity for anti-PCNA Alexa 488 (top), anti-vimentin Alexa 555 (middle) and anti-tubulin Alexa 647 (bottom) in cycle 3 vs. 16 and cycle 7 vs. 12 of the 16-cycle t-CyCIF experiment show in Figure 5. Intensity values were integrated across whole cells and the comparison is made on a cell-by-cell basis. Spearman’s correlation coefficients are shown. Plots in middle: intensity distributions at cycles 3 (blue), 7 (yellow), 12 (red) and 16 (green); intensity values were integrated across whole cells to construct the distribution. Box plots to right: estimated dynamic range at four cycle numbers 3, 7, 12, 16. Red lines denote median intensity values (across 56 frames), boxes denote the upper and lower quartiles, whiskers indicate values outside the upper/lower quartile within 1.5 standard deviations, and red dots represent outliers. (B) Representative images showing anti-tubulin Alexa 647 staining at four t-CyCIF cycles; original images are shown on the left (representing the same exposure time and approximately the same illumination) and images scaled by histogram equalization to similar intensity ranges are shown on the right. (C) Image for anti-CD45RO-Alexa 555 at cycles 5 and 15 scaled to similar intensity ranges as described in (B); the dynamic range (DR) of the cycle 15 image is ~3.3 fold lower than that of the Cycle 5 image, but shows similar morphology. (D) Intensity distributions for selected antibodies that were used in different cycles on Slides A and B. Colors denote the degree of concordance between the slides ranging from high (overlap >0.8 in yellow; PCNA), slightly increased or decreased with increasing cycle (overlap 0.6 to 0.8 in light blue or light red; S100 and SMA) or substantially increased or decreased (overlap <0.6 in red or blue; VEGFR2 and CD45RO). (E) Summary of effects of cycle number on antibody staining based on the degree of overlap in intensity distributions (the overlap integral); color coding is the same as in (D). (F) Effect of cycle number and specimen identity on overlap integrals for all antibodies and all cycles assayed. The red line denotes the median intensity value, boxes denote the upper/lower quartiles, and whiskers indicate values outside the upper/lower quartile and within 1.5 standard deviations, and red dots represent outliers. All the numeric data in Figures 5 and 6 are available in a Jupyter notebook; see Code Availability section of Materials and methods for details.

Figure 6—source data 1

Single-cell intensity data used in Figure 6.

Figure 6—figure supplement 1
Comparison of staining intensities across different cycles at a single-cell level.

Images come from the antibody-swap and repeat experiment showing in Figures 5 and 6. Comparison of single-cell level intensities for (A) PCNA Alexa-488, (B) Vimentin Alexa-555 and (C) Tubulin Alexa-647 stained in four different cycles (cycle 3, 7, 12 and16). For any given cycle pair, single-cell intensities density plots are shown. Spearman’s correlation coefficients (rho) and the mean intensity ratios between two cycles are shown.

Figure 7 with 1 supplement
t-CyCIF of a large resection specimen from a patient with pancreatic cancer.

(A) H&E staining of pancreatic ductal adenocarcinoma (PDAC) resection specimen that includes portions of cancer and non-malignant pancreatic tissue and small intestine. (B) The entire sample comprising 143 stitched 10X fields of view is shown. Fields that were used for downstream analysis are highlighted by yellow boxes. (C) A representative field of normal intestine across 8 t-CyCIF rounds; see Supplementary file 3 for a list of antibodies. (D) Segmentation data for four antibodies; the color indicates fluorescence intensity (blue = low, red = high). (E) Quantitative single-cell signal intensities of 24 proteins (rows) measured in ~4×103 cells (columns) from panel (C). The Pearson correlation coefficient for each measured protein with E-cadherin (at a single-cell level) is shown numerically. Known dichotomies are evident such as anti-correlated expression of epithelial (E-Cadherin) and mesenchymal (Vimentin) proteins. Proteins highlighted in red are further analyzed in Figure 8.

Figure 7—source data 1

Single-cell intensity data used in Figure 7E.

Figure 7—source data 2

Single-cell intensity data used in Figures 7 and 8.

Figure 7—figure supplement 1
t-CyCIF for examining large resection specimens of a human pancreatic cancer.

(A) Representative frame of small intestine from the PDAC resection shown in Figure 7 with images for PCNA, beta-catenin, Ki67 and pERk shown. These frames correspond to the segmented panels shown in 7D. (B) Creation of nuclear masks following identification of nuclei. Left panel: Hoechst image from t-CyCIF cycle one on the PDAC resection sample; middle panel: a binarized nuclear mask from cycle 8 (the final cycle in this t-CyCIF experiment); right panel: the overlay image of Hoechst stain (blue) and the cycle eight nuclear mask (yellow). The final cycle is used to create the mask used in this analysis so that the same cells can be tracked through all t-CyCIF cycles despite ~15% overall cell loss by cycle 8. Thus, regions in the overlay that show up in blue correspond in most cases to cells that are lost in the course of t-CyCIF and not to failure to identity and segment cells correctly.

High-dimensional single-cell analysis of human pancreatic cancer sample with t-CyCIF.

(A) t-SNE plots of cells derived from small intestine (left) or the PDAC region (right) of the specimen shown in Figure 7 with the fluorescence intensities for markers of proliferation (PCNA and Ki67) and signaling (pERK and β-catenin) overlaid on the plots as heat maps. In both tissue types, there exists substantial heterogeneity: circled areas indicate the relationship between pERK and β-catenin levels in cells and represent positive (‘a’), negative (‘b’) or no association (‘c’) between these markers. (B) Representative frames of normal pancreas and pancreatic ductal adenocarcinoma from the 8-cycle t-CyCIF staining of the same resection specimen from Figure 7. (C) t-SNE representation and clustering of single cells from normal pancreatic tissue (red), small intestine (blue) and pancreatic cancer (green). Projected onto the origin of each cell in t-SNE space are intensity measures for selected markers demonstrating distinct staining patterns. (D) Fluorescence intensity distributions for selected markers in small intestine, pancreas and PDAC.

Figure 8—source data 1

Single-cell data in FCS format (Figure 8C–E).

Figure 9 with 1 supplement
Spatial distribution of immune infiltrates and checkpoint proteins.

(A) Low-magnification image of a clear cell renal cancer subjected to 12-cycle t-CyCIF (see Supplementary file 4 for a list of antibodies). Regions high in α-smooth muscle actin (α-SMA) correspond to stromal components of the tumor, those low in α-SMA represent regions enriched for malignant cells. (B) Representative images from selected t-CyCIF channels are shown. (C) Quantitative assessment of total lymphocytic cell infiltrates (CD3+ cells), CD8+ T lymphocytes, cells expressing PD-1 or its ligand PD-L1 or the VEGFR2 for the entire tumor or for α-SMAhigh and α-SMAlow regions. VEGFR2 is a protein primarily expressed in endothelial cells and is targeted in the treatment of renal cell cancer. The error bars represent the S.E.M. derived from 100 rounds of bootstrapping. (D) Density plot for CD3 and CD8 expression on single cells in the tumor (left) or stromal domains (right). (E) Centroids of CD3+ or CD3+CD8+ cells in blue or dark blue as well as cells staining as SMAhigh or SMAlow (gray and light-gray, respectively) used to define the stromal and tumor regions. (F) Centroids of PD-1+ and PD-L1+ cells are shown in red and green, respectively. (G) Results of a K-nearest neighbor algorithm used to compute areas in which PD-1+ and PD-L1+ cells lie within ~10 µm of each other and with high spatial density (in yellow) and thus, are potentially positioned to interact at a molecular level.

Figure 9—source data 1

Immune cell counts from bootstrapping in tumor and stroma regions (Figure 9C).

Figure 9—source data 2

Single-cell intensity data used in Figure 9.

Figure 9—figure supplement 1
Spatial analysis of PD-1 and PD-L1 expressing cells.

This figure is relevant to the t-CyCIF study on a clear cell cancer shown in Figure 9. To compare co-localization of PD-1 and PD-L1 in tumor vs stroma, we computed the probability that PD1+ and PDL1+ cells would co-occur within a radius of ~10 um, normalizing for the difference in the total number of PD1+ and PDL1+ cells in the two tissue regions. We interpret the spatial density of PD1+ or PDL1+ cells at each point in space as proportional to the probability of their occurring there (see Figure 9E–F). The co-occurrence density at a point (Figure 9G) is therefore the product of the spatial densities for PD1+ or PDL1 +cells at that point. For simplicity, regions corresponding to tumor and stroma regions were defined by a diagonal line that separated the upper-right and lower-left regions of the tissue; this corresponded closely to α-SMA-low (tumor) and α-SMA-high (tumor) domains (left panel). We found an We found an e^(0.98)=2.7 fold difference in the distributions of co-occurrence densities between the two regions, representing an effect size of 0.94 as measured by Hedge's g; this represents a significant and potentially meaningful fold-change (right panel).

Figure 10 with 1 supplement
Eight-cycle t-CyCIF of a tissue microarray (TMA) including 13 normal tissues and corresponding tumor types.

The TMA includes normal tissue types, and corresponding high- and low-grade tumors, for a total of 39 specimens (see Supplementary file 3 for antibodies and Supplementary file 5 for specifications of the TMA). (A) Selected images of different tissues illustrating the quality of t-CyCIF images (additional examples shown in Figure 9—figure supplement 1; full data available online at www.cycif.org). (B) t-SNE plot of single-cell intensities of all 39 cores; data were analyzed using the CYT package (see Materials and methods). Tissues of origin and corresponding malignant lesions were labeled as follows: BL, bladder cancer; BR, breast cancer CO, Colorectal adenocarcinoma, KI, clear cell renal cancer, LI, hepatocellular carcinoma, LU, lung adenocarcinoma, LY, lymphoma, OV, high-grade serous adenocarcinoma of the ovary, PA, pancreatic ductal adenocarcinoma, PR, prostate adenocarcinoma, UT, uterine cancer, SK, skin cancer (melanoma), ST, stomach (gastric) cancer. Numbers refer to sample type; ‘1’ to normal tissue, ‘2’ to -grade tumors and ‘3’ to high-grade tumors. (C) Detail from panel B of normal kidney tissue (KI1) a low-grade tumor (KI2) and a high-grade tumor (KI3) (D) Detail from panel B of normal ovary (OV1) low-grade tumor (OV2) and high-grade tumor (OV3). (E) t-SNE plot from Panel B coded to show the distributions of all normal, low-grade and high-grade tumors. (F) tSNE clustering of normal pancreas (PA1) and pancreatic cancers (low-grade, PA2, and high-grade, PA3) and normal stomach (ST1) and gastric cancers (ST2 and ST3, respectively) showing intermingling of high-grade cells.

Figure 10—source data 1

Single-cell intensity data used in Figure 10.

Figure 10—figure supplement 1
Gallery of exemplary tissues imaged on the TMA described in Figure 10.

Gallery of exemplary tissues imaged on the TMA described in Figure 10.

Molecular heterogeneity in a single GBM tumor.

(A) Representative low-magnification image of a GBM specimen generated from 221 stitched 10X frames; the sample was subjected to 10 rounds of t-CyCIF using antibodies listed in Supplementary file 6. (B) Magnification of frame 152 (whose position is marked with a white box in panel A) showing staining of pERK, pRB and EGFR; lower panel shows a further magnification to allow single cells to be identified. (C) Normalized Shannon entropy of each of 221 fields of view to determine the extent of variability in signal intensity for 1000 cells randomly selected from that field for each of the antibodies shown. The size of the circles denotes the number of cells in the field and the color represents the value of the normalized Shannon entropy (data are shown only for those fields with more than 1000 cells; see Materials and methods for details).

Figure 11—source data 1

Normalized entropy data shown in Figure 11C.

Figure 11—source data 2

Single-cell intensity data used in Figure 11 and 12.

Figure 12 with 2 supplements
Spatial distribution of molecular phenotypes in a single GBM.

(A) Clustering of intensity values for 30 antibodies in a 10-cycle t-CyCIF analysis integrated over each whole cell based on images shown in Figure 11. Intensity values were clustered using expected-maximization with Gaussian mixtures (EMGM), yielding eight clusters, of which four clusters accounted for the majority of cells. The intensity scale shows the average level for each intensity feature in that cluster. The number of cells in the cluster is shown as a percentage of all cells in the tumor (bottom of panel). An analogous analysis is shown for 12 clusters in Figure 12—figure supplement 2. (B) EMGM clusters (in color code) mapped back to the positions of individual cells in the tumor. The coordinate system is the same as in Figure 11A. The positions of seven macroscopic regions (R1-R7) representing distinct lobes of the tumor are also shown. (C) Magnified view of Frame 147 from region R5 with EMGM cluster assignment for each cell in the frame; dots represent the centroids of single cells. (D) The proportional representation of EMGM clusters in each tumor region as defined in panel (B).

Figure 12—source data 1

Ratios of EMGM clusters in different regions of a GBM (Figure 12D).

Figure 12—figure supplement 1
Determination of cluster number for semi-supervised clustering using expectation–maximization Gaussian mixture (EMGM) modeling.

To determine an appropriate number of clusters (k) for analysis of the GBM tumor shown in Figure 12 we determined negative log-likelihood-ratio for various values of k. Due to the large sample size, likelihood-ratio tests were not helpful in choosing k. Thus, for each choice of cluster number n, the likelihood-ratio was calculated for a Gaussian mixture model with n = k-1 and with n = k and the ratio then plotted relative to k. The EMGM algorithm was initialized 30 times for each value of k and it converged in all instances. The inflection at k = 8 (red arrow) suggested that inclusion of additional clusters (k > 8) explains a smaller, distinct source of variation in the data. The plot is shown on a logarithmic scale to better visualize the range of the log-likelihood ratios, and should not be confused with the logarithm already applied to the likelihood ratios themselves.

Figure 12—figure supplement 2
Spatial distribution of molecular phenotypes in a single GBM.

This analysis is directly analogous to the analysis shown in Figure 11, but uses 12 clusters for EMGM analysis rather than 8. (A) Intensity values from the tumor in Figure 10 were clustered using expected-maximization with Gaussian mixtures (EMGM) with k = 12. The number of cells in each cluster is shown as a percentage of all cells in the tumor. (B) EMGM clusters (in color code) mapped back to singles cells and their positions in the tumor. The coordinate system is the same as in Figure 10. The positions of seven macroscopic regions (R1-R7) representing distinct lobes of the tumour are shown. (C) Magnified view of Frame 147 from region R5 with EMGM cluster assignment for each cell in the frame shown as a dot. (D) The proportional representation of EMGM clusters in each tumor region as defined in Panel B.



Table 1
Microscopes used in this study and their properties.
InstrumentTypeObjectiveField of viewNominal
RareCyte CytefinderSlide Scanner10X/0.3 NA1.6 × 1.4 mm1.06 µm
20X/0.8NA0.8 × 0.7 mm0.40 µm
40X/0.6 NA0.42 × 0.35 mm0.53 µm
GE INCell Analyzer 6000Confocal60X/0.95 NA0.22 × 0.22 mm0.21 µm
GE OMX BlazeStructured
Illumination Microscope
60 × 1.42 NA0.08 × 0.08 mm0.11 µm
  1. *Except in the case of the OMX Blaze, nominal resolution was calculated using the formula (r) = 0.61λ/NA for widefield and (r) = 0.4λ/NA for confocal microscopy with λ = 520 nm. Actual resolution depends on optical properties and thickness of sample, alignment and quality of the optical components in the light path. For structured illumination microscopy, actual resolution depends on accurate matching of immersion oil refractive index with sample in the Cy3 channel and use of an optimal point spread function during reconstruction process. The resolution in other channels will be sub-nominal.

Table 2
List of antibodies tested and validated for t-CyCIF.
Antibody nameTarget proteinPerformanceVendorCatalog no.CloneFluorophoreResearch resource
Bax-488Bax*BioLegend6336032D2Alexa Fluor 488AB_2562171
CD11b-488CD11b*AbcamAB204271EPR1344Alexa Fluor 488
CD4-488CD4*R and D SystemsFAB8165GPolyclonalAlexa Fluor 488
CD8a-488CD8*eBioscience53-0008-80AMC908Alexa Fluor 488AB_2574412
cJUN-488cJUN*AbcamAB193780E254Alexa Fluor 488
CK18-488Cytokeratin 18*eBioscience53-9815-80LDK18Alexa Fluor 488AB_2574480
CK8-FITCCytokeratin 8*eBioscience11-9938-80LP3KFITCAB_10548518
CycD1-488CycD1*AbcamAB190194EPR2241Alexa Fluor 488
Ecad-488E-Cadherin*CST319924E10Alexa Fluor 488AB_10691457
EGFR-488EGFR*CST5616D38B1Alexa Fluor 488AB_10691853
EpCAM-488EpCAM*CST5198VU1D9Alexa Fluor 488AB_10692105
HES1-488HES1*AbcamAB196328EPR4226Alexa Fluor 488
Ki67-488Ki67*CST11882D3B5Alexa Fluor 488AB_2687824
LaminA/C-488Lamin A/C*CST86174C11Alexa Fluor 488AB_10997529
LaminB1-488Lamin B1*AbcamAB194106EPR8985(B)Alexa Fluor 488
mCD3E-FITCms_CD3E*BioLegend100306145–2 C11FITCAB_312671
mCD4-488ms_CD4*BioLegend100532RM4-5Alexa Fluor 488AB_493373
MET-488c-MET*CST8494D1C2Alexa Fluor 488AB_10999405
mF4/80-488ms_F4/80*BioLegend123120BM8Alexa Fluor 488AB_893479
MITF-488MITF*AbcamAB201675D5Alexa Fluor 488
Ncad-488N-Cadherin*BioLegend3508098C11Alexa Fluor 488AB_11218797
p53-488p53*CST54297F5Alexa Fluor 488AB_10695458
PCNA-488PCNA*CST8580PC10Alexa Fluor 488AB_11178664
PD1-488PD1*CST15131D3W4UAlexa Fluor 488
PDI-488PDI*CST5051C81H6Alexa Fluor 488AB_10950503
pERK-488pERK(T202/Y204)*CST4344D13.14.4EAlexa Fluor 488AB_10695876
pNDG1-488pNDG1(T346)*CST6992D98G11Alexa Fluor 488AB_10827648
POL2A-488POL2A*Novus BiologicalsNB200-598AF4884H8Alexa Fluor 488AB_2167465
pS6(S240/244)−488pS6(240/244)*CST5018D68F8Alexa Fluor 488AB_10695861
S100a-488S100alpha*AbcamAB207367EPR5251Alexa Fluor 488
SQSTM1-488SQSTM1/p62*CST8833D1D9E3Alexa Fluor 488
STAT3-488STAT3*CST14047B3Z2GAlexa Fluor 488
Survivin-488Survivin*CST281071G4B7Alexa Fluor 488AB_10691462
Catenin-488β-Catenin*CST2849L54E2Alexa Fluor 488AB_10693296
Actin-555Actin*CST804613E5Alexa Fluor 555AB_11179208
CD11c-570CD11c*eBioscience41-9761-80118/A5eFluor 570AB_2573632
CD3D-555CD3D*AbcamAB208514EP4426Alexa Fluor 555
CD4-570CD4*eBioscience41-2444-80N1UG0eFluor 570AB_2573601
CD45-PECD45*R and D SystemsFAB1430P-1002D1PEAB_2237898
CK7-555Cytokeratin 7*AbcamAB209601EPR17078Alexa Fluor 555
cMYC-555cMYC*AbcamAB201780Y69Alexa Fluor 555
E2F1-555E2F1*AbcamAB208078EPR3818(3)Alexa Fluor 555
Ecad-555E-Cadherin*CST429524E10Alexa Fluor 555
FOXO1a-555FOXO1a*AbcamAB207244EP927YAlexa Fluor 555
FOXP3-570FOXP3*eBioscience41-4777-80236A/E7eFluor 570AB_2573608
GFAP-570GFAP*eBioscience41-9892-80GA5eFluor 570AB_2573655
KAP1-594KAP1*BioLegend61930420A1Alexa Fluor 594AB_2563298
Keratin-555pan-Keratin*CST3478C11Alexa Fluor 555AB_10829040
Keratin-570pan-Keratin*eBioscience41-9003-80AE1/AE3eFluor 570AB_11217482
Ki67-570Ki67*eBioscience41-5699-8020Raj1eFluor 570AB_11220088
LC3-555LC3*CST13173D3U4CAlexa Fluor 555
MAP2-570MAP2*eBioscience41-9763-80AP20eFluor 570AB_2573634
pAUR-555pAUR1/2/3(T288/T2*CST13464D13A11Alexa Fluor 555
PDL1-555PD-L1/CD274*AbcamAB21335828–8Alexa Fluor 555
pH3-555pH3(S10)*CST3475D2C8Alexa Fluor 555AB_10694639
pRB-555pRB(S807/811)*CST8957D20B12Alexa Fluor 555
pS6(235/236)–555pS6(235/236)*CST3985D57.2.2EAlexa Fluor 555AB_10693792
S6-555S6*CST698954D2Alexa Fluor 555AB_10828226
SQSTM1-555SQSTM1/p62*AbcamAB203430EPR4844Alexa Fluor 555
VEGFR2-555VEGFR2*CST12872D5B1Alexa Fluor 555
Vimentin-555Vimentin*CST9855D21H3Alexa Fluor 555AB_10859896
Vinculin-570Vinculin*eBioscience41-9777-807F9eFluor 570AB_2573646
AKT-647AKT*CST5186C67E7Alexa Fluor 647AB_10695877
aSMA-660aSMA*eBioscience50-9760-801A4eFluor 660AB_2574361
B220-647CD45R/B220*BioLegend103226RA3-6B2Alexa Fluor 647AB_389330
Bcl2-647Bcl2*BioLegend658705100Alexa Fluor 647AB_2563279
Catenin-647Beta-Catenin*CST4627L54E2Alexa Fluor 647AB_10691326
CD20-660CD20*eBioscience50-0202-80L26eFluor 660AB_11151691
CD45-647CD45*BioLegend304020HI30Alexa Fluor 647AB_493034
CD8a-660CD8*eBioscience50-0008-80AMC908eFluor 660AB_2574148
CK5-647Cytokeratin 5*AbcamAB193895EP1601YAlexa Fluor 647
CoIIV-647Collagen IV*eBioscience51-9871-801042Alexa Fluor 647AB_10854267
COXIV-647COXIV*CST75613E11Alexa Fluor 647AB_10994876
cPARP-647cPARP*CST6987D64E10Alexa Fluor 647AB_10858215
FOXA2-660FOXA2*eBioscience50-4778-823C10eFluor 660AB_2574221
FOXP3-647FOXP3*BioLegend320113206DAlexa Fluor 647AB_439753
gH2ax-647H2ax(S139)*CST972020E3Alexa Fluor 647AB_10692910
gH2ax-647H2ax(S139)*BioLegend6134072F3Alexa Fluor 647AB_2114994
HES1-647HES1*AbcamAB196577EPR4226Alexa Fluor 647
Ki67-647Ki67*CST12075D3B5Alexa Fluor 647
Ki67-647Ki67*BioLegend350509Ki-67Alexa Fluor 647AB_10900810
mCD45-647ms_CD45*BioLegend10312430-F11Alexa Fluor 647AB_493533
mCD4-647ms_CD4*BioLegend100426GK1.5Alexa Fluor 647AB_493519
mEPCAM-647ms_EPCAM*BioLegend118211G8.8Alexa Fluor 647AB_1134104
MHCI-647MHCI/HLAA*AbcamAB199837EP1395YAlexa Fluor 647
MHCII-647MHCII*AbcamAB201347EPR11226Alexa Fluor 647
mLy6C-647ms_Ly6C*BioLegend128009HK1.4Alexa Fluor 647AB_1236551
mTOR-647mTOR*CST50487C10Alexa Fluor 647AB_10828101
NFkB-647NFkB (p65)*AbcamAB190589E379Alexa Fluor 647
NGFR-647NGFR/CD271*AbcamAB195180EP1039YAlexa Fluor 647
NUP98-647NUP98*CST13393C39A3Alexa Fluor 647
p21-647p21*CST858712D1Alexa Fluor 647AB_10892861
p27-647p27*AbcamAB194234Y236Alexa Fluor 647
pATM-660pATM(S1981)*eBioscience50-9046-4110H11.E12eFluor 660AB_2574312
PAX8-647PAX8*AbcamAB215953EPR18715Alexa Fluor 647
PDL1-647PD-L1/CD274*CST15005E1L3NAlexa Fluor 647
pMK2-647pMK2(T334)*CST432027B7Alexa Fluor 647AB_10695401
pmTOR-660pmTOR(S2448)*eBioscience50-9718-41MRRBYeFluor 660AB_2574351
pS6_235–647pS6(S235/S236)*CST4851D57.2.2EAlexa Fluor 647AB_10695457
pSTAT3-647pSTAT3(Y705)*CST4324D3A7Alexa Fluor 647AB_10694637
pTyr-647p-Tyrosine*CST9415p-Tyr-100Alexa Fluor 647AB_10693160
S100A4-647S100A4*AbcamAB196168EPR2761(2)Alexa Fluor 647
Survivin-647Survivin*CST286671G4B7Alexa Fluor 647AB_10698609
TUBB3-647TUBB3*BioLegend657405AA10Alexa Fluor 647AB_2563609
Tubulin-647beta-Tubulin*CST36249F3Alexa Fluor 647AB_10694204
Vimentin-647Vimentin*BioLegend677807O91D3Alexa Fluor 647AB_2616801
anti-14-3-314-3-3*Santa CruzSC-629-GPolyclonalN/DAB_630820
anti-5HMC5HMC*Active Motif39769PolyclonalN/DAB_10013602
anti-MCM6MCM6*Santa CruzSC-9843PolyclonalN/DAB_2142543
anti-pRBpRB(S807/811)*Santa CruzSC-16670PolyclonalN/DAB_655250
anti-pRPA32pRPA32 (S4/S8)*BethylIHC-00422PolyclonalN/DAB_1659840
anti-VEGFR1VEGFR1/FLT1*Santa CruzSC-31173PolyclonalN/DAB_2106885
Abeta-488Beta-Amyloid (1-16)BioLegend8030136E10Alexa Fluor 488AB_2564765
BrdU-488BrdUBioLegend3641053D4Alexa Fluor 488AB_2564499
cCasp3-488cCasp3R and D SystemsIC835G-025269518Alexa Fluor 488
CD11b-488CD11bBioLegend101219M1/70Alexa Fluor 488AB_493545
CD123-488CD123BioLegend3060356H6Alexa Fluor 488AB_2629569
CD80-FITCCD80R and D SystemsFAB140F37711FITCAB_357027
CD8a-488CD8aeBioscience53-0086-41OKT8Alexa Fluor 488AB_10547060
CDC2-FITCCDC2/p34Santa CruzSC-54 FITC17FITCAB_627224
CycB1-FITCCycB1Santa CruzSC-752 FITCPolyclonalFITCAB_2072134
FN-488FibronectionAbcamAB198933F1Alexa Fluor 488
IFNG-488Interferron-GammaBioLegend5025174S.B3Alexa Fluor 488AB_493030
mCD8a-488ms_CD8aBioLegend10072653–6.7Alexa Fluor 488AB_493423
Nestin-488NestineBioscience53-9843-8010C2Alexa Fluor 488AB_1834347
NeuN-488NeuNMilliporeMAB377XA60Alexa Fluor 488AB_2149209
PR-488PR/PGRAbcamAB199224YR85Alexa Fluor 488
Snail1-488Snail1eBioscience53-9859-8020C8Alexa Fluor 488AB_2574482
TNFa-488TNFaBioLegend502917MAb11Alexa Fluor 488AB_493122
AR-555ARCST8956D6F11Alexa Fluor 555AB_11129223
CD11b-555CD11bAbcamAB206616EPR1344Alexa Fluor 555
CD31-PECD31R and D SystemsFAB3567P-0259G11PEAB_2279388
CD45R-e570CD45R/B220eBioscience41-0452-80RA3-6B2eFluor 570AB_2573598
CK19-570Cytokeratin 19eBioscience41-9898-80BA17eFluor 570AB_11218678
HER2-570HER2eBioscience41-9757-80MJD2eFluor 570AB_2573628
pMAPK (T202/Y204)pERK1/2(T202/Y20CST14095197G2PE
pMAPK (Y204/Y187)pERK1/2(Y204/Y18CST75165D1H6GPE
ABCC1-647ABCC1BioLegend370203QCRL-2Alexa Fluor 647AB_2566664
AnnexinV-674N/DBioLegend640911NAAlexa Fluor 647AB_2561293
CD103-647CD103BioLegend350209Ber-ACT8Alexa Fluor 647AB_10640870
CD25-647CD25BioLegend302617BC96Alexa Fluor 647AB_493046
CD8a-647CD8aBioLegend344725SK1Alexa Fluor 647AB_2563451
CD8a-647CD8aR and D SystemsFAB1509R-02537006Alexa Fluor 647
CycE-660CycEeBioscience50-9714-80HE12eFluor 660AB_2574350
HIF1-647HIF1BioLegend359705546–16Alexa Fluor 647AB_2563331
HP1-647HP1AbcamAB198391EPR5777Alexa Fluor 647
NGFR-647NGFR/CD271BD560326C40-1457Alexa Fluor 647AB_1645403
pBTK-660pBTK(Y551/Y511)eBioscience50-9015-80M4G3LNeFluor 660AB_2574306
PD1-647PD1AbcamAB201825EPR4877 (2)Alexa Fluor 647
PR-660PR/PGReBioscience50-9764-80KMC912eFluor 660AB_2574363
RUNX3-660RUNX3eBioscience50-9817-80R3-5G4eFluor 660AB_2574383
SOX2-647SOX2AbcamAB192075PolyclonalAlexa Fluor 647
anti-AxlAxlR and DAF154PolyclonalN/DAB_354852
anti-eEF2KeEF2KSanta CruzSC-21642K-19N/DAB_640043
anti-FKHRL1FKHRL1Santa CruzSC-9812PolyclonalN/DAB_640608
anti-HMB45HMB45AbcamAB732HMB45 + M2- 7C10 + M2-
anti-HSP90bHSP90bSanta CruzSC-1057D-19N/DAB_2121392
anti-LAMP2LAMP2R and DAF6228PolyclonalN/DAB_10971818
anti-pCTDPol II CTD(S2)Active Motif610833E10N/DAB_2687450
anti-TIA1TIA1Santa CruzSC-1751PolyclonalN/DAB_2201433
anti-TLR3TLR3Santa CruzSC-8691PolyclonalN/DAB_2240700
CD45RO-488CD45ROBioLegend304212UCHL1Alexa Fluor 488AB_528823
CD8a-488CD8BioLegend301024RPA-T8Alexa Fluor 488AB_2561282
cJUN-FITCcJUNSanta CruzSC-1694 FITCPolyclonalFITCAB_631263
FOXP3-488FOXP3BioLegend320011150DAlexa Fluor 488AB_439747
MITF-488MITFNovus BiologicalsNB100-56561AF48821D1418Alexa Fluor 488AB_838580
NCAM-488NCAM/CD56AbcamAB200333EPR2566Alexa Fluor 488
PD1-488PD-1BioLegend367407NAT105Alexa Fluor 488AB_2566677
PD1-488PD-1BioLegend329935EH12.2H7Alexa Fluor 488AB_2563593
pERK-488pERK(T202/Y204)CST4374E10Alexa Fluor 488AB_10705598
pERK-488pERK(T202/Y204)CST4780137F5Alexa Fluor 488AB_10705598
SOX2-488SOX2BioLegend65610914A6A34Alexa Fluor 488AB_2563956
cMyc-TRITCcMYCSanta CruzSC-40 TRITC9E10TRITCAB_627268
cPARP-555cPARPCST6894D64E10Alexa Fluor 555AB_10830735
GATA3-594GATA3BioLegend65381616E10A23Alexa Fluor 594AB_2563353
Oct4-555OCT_4CST4439C30A3Alexa Fluor 555AB_10922586
p21-555p21CST849312D1Alexa Fluor 555AB_10860074
PDGFRb-555PDGFRbAbcamAB206874Y92Alexa Fluor 555
pSTAT1-555pSTAT1CST818358D6Alexa Fluor 555AB_10860600
cCasp3-647cCasp3CST9602D3E9Alexa Fluor 647AB_2687881
CD3-647CD3BioLegend300422UCHT1Alexa Fluor 647AB_493092
CD3-660CD3eBioscience50-0037-41OKT3eFluor 660AB_2574150
ER-647ERAbcamAB205851EPR4097Alexa Fluor 647
FOXO3a-647FOXO3aAbcamAB196539EP1949YAlexa Fluor 647
GZMA-e660Granzyme AThermoFisher50-9177-41CB9eFluor 660AB_2574330
GZMB-647Granzyme_BBioLegend515405GB11Alexa Fluor 647AB_2294995
GZMB-APCGranzyme_BR and D SystemsIC29051A356412APCAB_894691
HER2-647HER2BioLegend32441224D2Alexa Fluor 647AB_2262300
mCD49b-647ms_CD49bBioLegend103511HMα2Alexa Fluor 647AB_528830
NCAM-647NCAM/CD56BioLegend3625135.1H11Alexa Fluor 647AB_2564086
NCAM-e660NCAM/CD56ThermoFisher50-0565-805tukon56eFluor 660AB_2574160
pAKT-647pAKTCST4075D9EAlexa Fluor 647AB_10691856
pERK-647pERK (T202/Y204)CST4375E10Alexa Fluor 647AB_10706777
pERK-647pERK (T202/Y204)BioLegend3695036B8B69Alexa Fluor 647AB_2571895
pIKBa-660pIKBaeBioscience50-9035-41RILYB3ReFluor 660AB_2574310
YAP-647YAPCST38707SD8H1XAlexa Fluor 647
anit-pcJUNp-cJUNSanta CruzSC-822KM-1N/DAB_627262
anti-CXCR5CXCR5R and DMAB-190-SP51505N/DAB_2292654
anti-GZMBGranzyme BAbcamAB4059PolyclonalN/DAB_304251
anti-PD1PD-1R and DAF1021PolyclonalN/DAB_354541
  1. *Show positive/correct signals in multiple samples/tissues.

    †Show positive/correct signals in some but not all samples tested.

  2. ‡Show no signal or incorrect signals in most samples tested.

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional information
Biological sample
(human tissue specimen)
TMA:TMA-1207Protein BiotechnologiesCat: TMA-1207http://www.proteinbiotechnologies.com/pdf/TMA-1207.pdf
Biological sample
(human tissue specimen)
TMA:MTU481BiomaxCat: MTU-481https://www.biomax.us/tissue-arrays/Multiple_Organ/MTU481
AntibodyAlexa-488 anti-Rabbit
antibodies (Fab)
ThermoFisher ScientificCat: A-11034
Dilution 1:2000
AntibodyAlexa-555 anti-Rat
ThermoFisher ScientificCat: A-21434
Dilution 1:2000
AntibodyAlexa-647 anti-Mouse
antibodies (Fab)
ThermoFisher ScientificCat: A-21236
Dilution 1:2000
Chemical compound,
Hoechst 33342ThermoFisher ScientificCat: H3570https://www.thermofisher.com/order/catalog/product/H3570
Software, algorithmImageJPMID:22930834RRID: SCR_003070https://imagej.nih.gov/ij/
Software, algorithmMatlabMathWorks, Inc.RRID:SCR_001622
Software, algorithmAshlarLaboratory of Systems
Pharmacology, Harvard
Medical School
RRID:SCR_016266https://github.com/sorgerlab/ashlar (copy archived at
Software, algorithmBaSiCHelmholtz Zentrum
RRID: SCR_016371https://www.nature.com/articles/ncomms14836
Otherwww.cycif.orgLaboratory of Systems
Pharmacology, Harvard
Medical School
RRID:SCR_016267Online resource for
cyclic immunofluorescence
Otherlincs.hms.harvard.eduHMS LINCS CenterRRID:SCR_016370Additional data/image
resource for t-CyCIF
Table 3
Breakdown of individual steps performed for dewaxing and antigen retrieval on a Leica BOND.
StepReagentSupplierIncubation (min)Temp. (°C)
1*No ReagentN/D3060
2BOND Dewax SolutionLeica060
3BOND Dewax SolutionLeica0R.T.
4BOND Dewax SolutionLeica0R.T.
5200 proof ethanolUser*0R.T.
6200 proof ethanolUser*0R.T.
7200 proof ethanolUser*0R.T.
8Bond Wash SolutionLeica0R.T.
9Bond Wash SolutionLeica0R.T.
10Bond Wash SolutionLeica0R.T.
11Bond ER1 solutionLeica099
12Bond ER1 solutionLeica099
13Bond ER1 solutionLeica2099
14Bond ER1 solutionLeica0R.T.
15Bond Wash SolutionLeica0R.T.
16Bond Wash SolutionLeica0R.T.
17Bond Wash SolutionLeica0R.T.
18Bond Wash SolutionLeica0R.T.
19Bond Wash SolutionLeica0R.T.
20IF BlockUser*30R.T.
21Antibody MixUser*60R.T.
22Bond Wash SolutionLeica0R.T.
23Bond Wash SolutionLeica0R.T.
24Bond Wash SolutionLeica0R.T.
25Hoechst SolutionUser*30R.T.
26Bond Wash SolutionLeica0R.T.
27Bond Wash SolutionLeica0R.T.
28Bond Wash SolutionLeica0R.T.

Data availability

All data generated or analyzed during this study are included in the manuscript and supporting files. Intensity data used to generate figures is available in supplementary materials and can be downloaded from the HMS LINCS Center Publication Page (http://lincs.hms.harvard.edu/lin-elife-2018/) (RRID:SCR_016370). The images described are available at http://www.cycif.org/ (RRID:SCR_016267) and via and OMERO server as described at the LINCS Publication Page.

Additional files

Supplementary file 1

List of antibodies used for staining in Figure 3.

Supplementary file 2

List of antibodies used for staining in Figures 5 and 6.

Supplementary file 3

List of antibodies used for staining in Figures 7, 8 and 10.

Supplementary file 4

List of antibodies used for staining in Figure 9.

Supplementary file 5

Descriptions of TMA shown in Figure 10.

Supplementary file 6

List of antibodies used for staining in Figures 11 and 12.

Transparent reporting form

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)