key: cord-0321055-7991388o
authors: Radtke, Andrea J.; Kandov, Evelyn; Lowekamp, Bradley; Speranza, Emily; Chu, Colin J.; Gola, Anita; Thakur, Nishant; Shih, Rochelle; Yao, Li; Yaniv, Ziv Rafael; Beuschel, Rebecca T.; Kabat, Juraj; Croteau, Joshua; Davis, Jeremy; Hernandez, Jonathan M.; Germain, Ronald N.
title: IBEX – A versatile multi-plex optical imaging approach for deep phenotyping and spatial analysis of cells in complex tissues
date: 2020-11-20
journal: bioRxiv
DOI: 10.1101/2020.11.20.390690
sha: a0b91c968152df47236bdb40009e9d4e47b07d4d
doc_id: 321055
cord_uid: 7991388o

The diverse composition of mammalian tissues poses challenges for understanding the cell-cell interactions required for organ homeostasis and how spatial relationships are perturbed during disease. Existing methods such as single-cell genomics, lacking a spatial context, and traditional immunofluorescence, capturing only 2-6 molecular features, cannot resolve these issues. Imaging technologies have been developed to address these problems, but each possesses limitations that constrain widespread use. Here we report a new method that overcomes major impediments to highly multi-plex tissue imaging. Iterative Bleaching Extends multi-pleXity (IBEX) uses an iterative staining and chemical bleaching method to enable high resolution imaging of >65 parameters in the same tissue section without physical degradation. IBEX can be employed with various types of conventional microscopes and permits use of both commercially available and user-generated antibodies in an ‘open’ system to allow easy adjustment of staining panels based on ongoing marker discovery efforts. We show how IBEX can also be used with amplified staining methods for imaging strongly fixed tissues with limited epitope retention and with oligonucleotide-based staining, allowing potential cross-referencing between flow cytometry, Cellular Indexing of Transcriptomes and Epitopes by Sequencing (CITE-Seq), and IBEX analysis of the same tissue. To facilitate data processing, we provide an open source platform for automated registration of iterative images. IBEX thus represents a technology that can be rapidly integrated into most current laboratory workflows to achieve high content imaging to reveal the complex cellular landscape of diverse organs and tissues. Significance Statement Single cell flow cytometry and genomic methods are rapidly increasing our knowledge of the diversity of cell types in metazoan tissues. However, suitably robust methods for placing these cells in a spatial context that reveal how their localization and putative interactions contribute to tissue physiology and pathology are still lacking. Here we provide a readily accessible pipeline (IBEX) for highly multi-plex immunofluorescent imaging that enables a fine-grained analysis of cells in their tissue context. Additionally, we describe extensions of the IBEX workflow to handle hard to image tissue preparations and a method to facilitate direct integration of the imaging data with flow cytometry and sequencing technologies.

of >65 parameters in the same tissue section without physical degradation. IBEX can be employed 48 with various types of conventional microscopes and permits use of both commercially available and 49 user-generated antibodies in an 'open' system to allow easy adjustment of staining panels based 50 on ongoing marker discovery efforts. We show how IBEX can also be used with amplified staining 51 methods for imaging strongly fixed tissues with limited epitope retention and with oligonucleotide-52 based staining, allowing potential cross-referencing between flow cytometry, Cellular Indexing of 53 Transcriptomes and Epitopes by Sequencing (CITE-Seq), and IBEX analysis of the same tissue.

To facilitate data processing, we provide an open source platform for automated registration of 55 iterative images. IBEX thus represents a technology that can be rapidly integrated into most current 56 laboratory workflows to achieve high content imaging to reveal the complex cellular landscape of 57 diverse organs and tissues.

Significance Statement 59 Single cell flow cytometry and genomic methods are rapidly increasing our knowledge of 60 the diversity of cell types in metazoan tissues. However, suitably robust methods for placing these 61 cells in a spatial context that reveal how their localization and putative interactions contribute to 62 tissue physiology and pathology are still lacking. Here we provide a readily accessible pipeline 63 (IBEX) for highly multi-plex immunofluorescent imaging that enables a fine-grained analysis of cells 64 in their tissue context. Additionally, we describe extensions of the IBEX workflow to handle hard to Introduction 69 Mammalian tissues are composed of a wide variety of cell types, presenting a major 70 challenge to understanding the cell-cell interactions required for homeostasis as well as the 71 compositional changes associated with disease. To address this complexity, several multi-plexed 3 imaging methods utilizing conventional microscopes and commercially available antibodies have 73 been described to overcome the target detection limitations of conventional immunohistochemistry 74 (IHC) or immunofluorescence (IF) imaging (1) (2) (3) (4) (5) (6) (7) (8) . The majority of these methods generate high 75 dimensional datasets through an iterative, multi-step process (a cycle) that includes: 1) 76 immunolabeling with antibodies, 2) image acquisition, and 3) fluorophore inactivation or 77 antibody/chromogen removal. While these methods are capable of generating high dimensional 78 datasets, they are greatly limited by the number of markers visualized per cycle, length of time 79 required for each cycle, or involve special fluid-handling platforms not generally available to most 80 laboratories (1) . Commercial systems based on the co-detection by indexing (CODEX) method (9) 81 have facilitated the acquisition of multi-plex imaging data by providing a fully automated instrument 82 for cyclic imaging. Despite this advancement, the proprietary nature of this method imposes 83 constraints on the reagents available for use as well as the number of markers to be imaged for 84 each round. Furthermore, cyclic imaging methods that employ a small number of markers per cycle 85 (<3) may result in tissue loss due to the stress of repeated fluid exchanges. To this end, novel 86 imaging techniques such as multi-plexed ion beam imaging (MIBI) (10) and imaging mass 87 cytometry (IMC) (11) enable the capture of multi-parameter data without cyclic imaging. However, 88 both of these methods require specialized instrumentation and consumables, with the latter often 89 again limited in breadth to choices made by the supplier, not the investigator. This constrains their 90 capacity for broadly analyzing human or experimental animal tissues with respect to lists of 91 validated antibodies, the ability to work across various established protocols for tissue processing, 92 and the capacity for real time changes to the epitope target list based on data emerging from high 93 content methods such as single cell RNA sequencing (scRNA-Seq).

To facilitate the increasing need for high content analysis of tissues for projects such as 95 the Human Cell Atlas and others, the field needs a fully open and extensible method for multi-plex 96 imaging. Our laboratory has extensively characterized murine and human immune responses using 97 quantitative multi-parameter imaging of fixed frozen samples (12) (13) (14) (15) (16) (17) (18) 

Iterative imaging methods typically use either fluorophore bleaching or 120 antibody/chromogen removal to achieve multi-parameter datasets (1) (2) (3) (4) (5) (6) (7) (8) . Due to the harsh and 121 variable conditions required to remove chromogens and antibodies with diverse target affinities, we 122 pursued a strategy based on fluorophore bleaching. To achieve an efficient means to increase the 123 number of markers visualized on a single section, we sought a fluorophore inactivation method that 124 could bleach a wide range of fluorophores in minutes without epitope loss or tissue destruction.

While H2O2 in alkaline solution has been reported to inactivate Cy3-and Cy5-conjugated antibodies 126 in human FFPE samples (3), we observed significant tissue loss using this formulation over multiple 

Hoechst required more than 120 minutes for significant loss of fluorescence signal (Table S1 ),

permitting these probes to be used as fiducials for alignment of images emerging from iterative 139 cycles.

To prevent tissue destruction over multiple cycles, we evaluated several different tissue 141 adhesives and found that chrome gelatin alum securely adhered tissues to glass coverslips and 142 slides, permitting more than 15 cycles to be performed with no appreciable loss to the tissue (Fig. 143 S1B-C). We next reduced the antibody labeling time from 6-12 hours to 30-45 minutes by designing 144 programs for a non-heating microwave that facilitates rapid antibody penetration into the section.

Finally, although IBEX was designed to simply bleach the fluorophores, it was important to assess 

The resulting method, IBEX, reduces the fluorophore inactivation and antibody labeling 155 steps to less than 1 hour (Fig. 1A) . We first tested IBEX in practical use by examining the image 156 quality that could be obtained from a 3 cycle analysis of mouse LNs (Fig. 1B) , a tissue with which 157 we had extensive experience using multi-parameter, single cycle staining and image collection.

This initial test used 6-8 markers per cycle and showed that all fluorophores, except for AF594 and 

In the case of animal studies, it is also very useful to be able to integrate antibody staining 198 with imaging of fluorescent marker proteins expressed by engineered cells transferred into animals 199 or expressed in situ. We therefore next investigated whether the IBEX method could be used to 

To determine how IBEX performs using sections from a variety of tissues, we performed 209 3-5 cycle IBEX experiments on murine spleen, thymus, lung, small intestine, and liver tissue 210 sections ( Fig. 3A -B, Movies S1-S5, Table S2 ). It is important to note that the cycle and marker 211 numbers described here are provided as a proof-of-concept and do not reflect technical limitations 212 of the method. The antibody panels were designed to capture the major cellular populations and 213 structures present in each organ and fluorophores were chosen to avoid native tissue 214 autofluorescence. Organ-specific fiducials were selected based on expression throughout the 215 tissue, e.g., EpCAM to mark the epithelium of the small intestine and laminin for the liver sinusoids.

Collectively, these data confirm the ability to use IBEX to obtain high quality, multi-plexed imaging 217 datasets from a wide range of tissues. 

To assess the quality of data generated by the IBEX method, we employed the open 233 source, computational histology topography cytometry analysis toolbox (histoCAT) to quantify 234 differences in LN organization resulting from immunization (27). Individual cells were segmented 235 based on membrane and nuclear labels with Ilastik (28) and CellProfiler (29) Table S3 ). Additionally, we observed 264 subcellular resolution for PC-specific markers (membrane: CD138, nuclear: IRF4, cytoplasmic: 265 IgA1, IgA2) present in distinct cycles with no epitope loss, as evidenced by our ability to label the 266 immune marker CD45 with two different antibody clones present in cycles 9 and 19 (Movie S8).

The utility of this method is further exemplified by our ability to characterize the complex stroma of incubation with an unconjugated primary, followed by a horseradish peroxidase (HRP)-conjugated 280 secondary, and deposition of Opal fluorophore in the tissue, is an attractive method for detection 281 of very low levels of specific proteins (36). We first tested this method by staining for endogenous 282 levels of the chemokine CXCL9 in the liver sinusoids of mice (Fig. S7A) , which showed a signal not 283 readily detected with direct or indirect staining methods. Further, because Opal IHC is well 284 described for the imaging of fixed (FFPE) human tissues (37, 38), we next evaluated whether this 285 method could be expanded upon to achieve multi-parameter imaging of tissues from high 9 containment facilities. Due to the extreme fixation conditions required to inactivate select agents 287 such as the Ebola virus (10% formalin for 8 days), the majority of stainable epitopes are lost in 288 these tissues (39, 40). To overcome this significant technical limitation, we developed the Opal-289 plex method that is based on the IBEX pipeline. Opal-plex extends the usual fluorophore limitations 290 of Opal by combining multi-plex Opal IHC with cycles of IBEX-based bleaching to eliminate signal 291 from the following LiBH4-sensitive dyes (Opal 570, 650, and 690) while utilizing the LiBH4-resistant 292 dye (Opal 540) as a fiducial (Fig. 6A, Fig. S7B ). Using this approach, we achieved single cell 293 resolution of 10 unique markers in heavily fixed mouse LNs (Fig. 6B, Movie S9 ).

We next evaluated whether oligonucleotide-conjugated antibodies, including those used 295 for CITE-Seq, are compatible with our IBEX workflow. While immunolabeling with oligonucleotide-296 conjugated antibodies is well established (9), the use of a large number of commercially available 297 TotalSeqA TM antibodies with publicly available oligo-tag sequences, the employment of non-

proprietary buffers for hybridization and dehybridization, and the use of a wide spectrum of 299 fluorophore-labeled complementary oligonucleotides provides a truly 'open source' system with 300 many advantages. In particular, the imaging method described here applies the same antibodies 301 used for scRNA-Seq, permitting direct comparison between imaging and CITE-Seq datasets while 302 providing a much-needed spatial context for the cell populations identified. Using this approach, 303 we were able to achieve high quality tissue staining with 5 unique fluorophores (Fig. S7C ). This 304 method can be directly integrated into our IBEX protocol, alongside fluorophore-conjugated 305 antibodies when CITE-Seq antibodies to desired targets do not exist, as LiBH4 bleaching leaves 306 oligonucleotide binding intact (Fig. 6C-D) . Importantly, the quality of staining achieved with 307 oligonucleotide-conjugated antibodies, even after multiple cycles of LiBH4 bleaching, is comparable 308 to conventional IF as quantitative differences, e.g., higher expression of MHCII on DCs versus B 309 cells, can still be observed (Fig. 6D, Fig. S7D , Movie S9). In summary, this protocol improves upon 

Importantly, LiBH4 treatment does not cause tissue or epitope loss as evidenced by our ability to 327 obtain highly multi-plexed data over several cycles in a wide range of tissues with a very large 328 number of antibodies. Third, and integral to the preservation of tissue integrity through multiple fluid 329 handling cycles, was the use of the tissue adhesive chrome gelatin alum. Importantly, this adhesive 330 adheres delicate tissues to the slide or coverslip surface while maintaining key anatomical features.

Finally, the SimpleITK workflow described here represents a significant advancement for the 332 registration of images obtained via cyclic IF methods. In addition to offering flexibility in terms of 333 the repeated markers (membrane, nuclear, structural) used, it provides alignment of markers 334 present on the same cell but not utilized as the fiducial. This is a critical standard for all high 335 dimensional imaging methods because multiple markers are often required to phenotype a 336 particular cell type and staining for the relevant epitopes may occur in different imaging cycles.

In addition to developing an efficient method for highly multi-plexed imaging, the IBEX 338 workflow, unlike commercial all-in-one systems (9-11), offers flexibility in terms of cellular markers, 

Given that the barcode sequences for TotalSeqA TM antibodies are disclosed, and a wide range of 349 fluorophore-conjugated oligos is readily available, fluorophore and antibody pairing can be fully 350 customized to match microscope configuration, epitope abundance, and unique tissue 351 characteristics. Taken together, the oligonucleotide-staining method described here provides a

completely 'open' method to achieve highly multi-plexed IF imaging using the same antibodies 353 employed for flow cytometry and/or CITE-Seq, enabling effective cross-referencing of datasets 354 derived from these complementary technologies.

As a proof-of-concept, we have used the IBEX workflow to examine such issues as the 356 visualization of difficult to extract myeloid populations in various tissues, changes in immune cell 357 composition following immune perturbation, and detection of low abundance epitopes. For the first 358 application, we were able to visualize tissue-resident macrophages that are difficult to characterize 359 using other methods such as flow cytometry because of their limited recovery upon enzymatic 360 11 tissue digestion (12). Using the panels of antibodies outlined here, we were able to deeply 361 phenotype medullary (CD169 + F4/80 + CD11b + Lyve-1 +/-) and subcapsular sinus (CD169 + F4/80 -362 CD11b + ) macrophages in the LNs (41) as well as alveolar (SiglecF + CD11b -CD11c + ) and interstitial 363 (CD11b + CD11c + MHCII + ) macrophages of the lung (42). Additionally, we show that the IBEX 364 method can be scaled to capture ultra-high content imaging in human tissues. The ability to survey 365 large areas of human tissue is critically important as all possible information needs to be extracted 366 to provide maximally useful clinical and research data. 

In summary, IBEX constitutes a versatile technique for obtaining high content imaging data 382 using conventional microscopes and commercially available antibodies. In addition to providing a 383 valuable resource for studying tissue-based immunity in animal models of disease, ongoing studies 384 have shown the value of the IBEX method to provide a spatially-defined assessment of complex 385 cell phenotypes from diverse organs including lung, kidney, heart, and lymphoid tissues from The alignment of all IBEX panels to a common coordinate system was performed 437 using SimpleITK (23, 24) . To facilitate registration, we utilized a common channel present in all 438 panels. As the images may differ by a significant translational motion, we use a Fourier domain- 

(RCA# 2020-0333). CJC is supported as a UK-US Fulbright Scholar. We would like to thank Dr. 

Topological proteomics, toponomics, MELK-technology

Analyzing proteome topology and function by automated 473 multidimensional fluorescence microscopy

Highly multiplexed single-cell analysis of formalin-fixed, paraffin-475 embedded cancer tissue

Highly multiplexed imaging of single cells using 477 a high-throughput cyclic immunofluorescence method

Highly multiplexed immunofluorescence imaging of human tissues and 479 tumors using t-CyCIF and conventional optical microscopes

Multi-phenotypic 481 subtyping of circulating tumor cells using sequential fluorescent quenching and 482 restaining

Quantitative multiplex immunohistochemistry reveals myeloid-484 inflamed tumor-immune complexity associated with poor prognosis

Multiplexed protein maps link subcellular 487 organization to cellular states

Deep profiling of mouse splenic architecture with CODEX multiplexed 489 imaging

Multiplexed ion beam imaging of human breast tumors

Highly multiplexed imaging of tumor tissues with subcellular resolution 493 by mass cytometry

Histo-cytometry: a method 495 for highly multiplex quantitative tissue imaging analysis applied to dendritic cell subset 496 microanatomy in lymph nodes

A 498 spatially-organized multicellular innate immune response in lymph nodes limits systemic 499 pathogen spread

Innate and adaptive lymphocytes sequentially shape the gut microbiota and 501 lipid metabolism

The chemoattractant receptor Ebi2 drives intranodal naive CD4(+) T 503 cell peripheralization to promote effective adaptive immunity

Resident 506 macrophages cloak tissue microlesions to prevent neutrophil-driven inflammatory damage

Follicular CD8 T cells accumulate in HIV infection and can kill infected 509 cells in vitro via bispecific antibodies

Spatial distribution and function of T follicular regulatory cells in human lymph 511 nodes

Multiplex, quantitative cellular analysis in large tissue 513 volumes with clearing-enhanced 3D microscopy (Ce3D)

Simultaneous epitope and transcriptome measurement in single cells

A pyramid approach to subpixel registration based 518 on intensity

Efficient subpixel image registration 520 algorithms

fluorophore signal remaining after 15 minutes of LiBH4 treatment. Data are pooled from 2 similar

Figure 2. Image alignment with SimpleITK image registration pipeline

Workflow for SimpleITK image registration pipeline. (B) Confocal images showing JOJO-1 and 618 CD4 from 3 consecutive IBEX cycles before and after alignment using the nuclear marker JOJO-1 619 as a fidicual across all 3 cycles. CD4 was also repeated and shows cell-cell registration

Cycle (C), scale bar is 50 µm. Cross correlation similarity matrices before and after 621 alignment with JOJO-1 for JOJO-1 and CD4 channels

Confocal images from IBEX experiments in various mouse 629 organs

Visualization and quantification of LN populations using IBEX and histoCAT 684 following immune perturbation

Confocal images of pLNs from naïve and SRBC-immunized mice from 10 cycle (C)

Scale bars from left to right: 100 µm, 25 µm, 100 µm, and 50 µm. (B) t-689 SNE plots from naïve and immunized LNs identified by Phenograph clustering using segmented 690 cells in histoCAT

Single plots show separation of representative markers into discrete clusters with color map 692 showing relative expression levels based on Z-score normalized marker intensity values

Phenograph clusters identified by histoCAT were phenotyped based on marker expression and 694 expressed as a proportion of lineage. Tfh (T follicular helper), MΦ (macrophage), SCS (subcapsular 695 sinus), MSM (Medullary sinus), DC (dendritic cell), dDC (dermal DC). Data are from one experiment 696 and representative of 2 similar experiments

Figure 5. IBEX scales to capture ultra-high content imaging in human tissues

Scale bar is 500 µm (left), 100 µm (Box 1), or 50 707 µm (Box 2). (B) Representative confocal images from human mesenteric LN obtained by IBEX 708 method (20 cycle 66 parameters). Scale bars (500 or 50 µm)

See Movie S8 for additional details

Figure 6. Incorporation of Opal fluorophores and oligo-conjugated antibodies into IBEX 721 workflow

Opal-plex imaging method consisting of several rounds of labeling with marker-specific primary 723 antibodies, an HRP-conjugated secondary antibody, Opal dyes, and antibody stripping for each 724 marker-Opal fluorophore pair followed by cycles of IBEX (imaging, removal of coverslip, and 725 bleaching). (B) Representative images from a 10 parameter 4 cycle Opal-plex experiment throughout cycles 1-4 and served as a fiducial (*)

Scale bars (400 µm, top-left panel or 50 µm). Data are representative of 3 similar 733 experiments