OMIP-069 version 2: Update to the 40-color full Spectrum flow cytometry panel for deep immunophenotyping of major cell subsets in human peripheral blood

Abstract

Optimized Multicolor Immunofluorescence Panel (OMIP)-069 was the first optimized flow cytometry panel demonstrating that 40 different fluorochromes can be effectively used in combination, without compromising the resolution of each individual marker [1]. Since its publication, the panel has been adopted in some laboratories (personal communications and [2, 3]). Throughout its adoption, challenges that limited the ability to use the full panel have been reported.

The challenges can be grouped into two main categories: reagent availability and reagent performance. Concerning reagent availability, CD20 Pacific Orange™ (Thermo Fisher Cat. MHCD2030), CD25 Phycoerythrin (PE) Alexa Fluor™ 700 (Thermo Fisher Cat. MHCD2524), CD24 PE-Alexa Fluor™ 610 (Thermo Fisher Cat. MHCD2422), and CD127 Allophycocyanin (APC) R700 (BD Biosciences Cat. 565185) have been reported to be often on back order. Regarding reagent performance, HLA-DR PE/Fire™ 810 (Biolegend Cat. 307683) and TCRγδ Peridinin-Chlorophyll-Protein (PerCP) eFluor® 710 (Thermo Fisher Cat. 46–9959-42), have shown technical issues, due to tandem degradation or unexpected spectrum signatures, respectively. Finally, it was also documented that spread between fluorescein isothiocyanate (FITC) and Brilliant™ Blue (BB) BB515, the most challenging fluorochrome combination in the panel (similarity index 0.98; spillover spread value of BB515 into FITC 29), was not always consistent with the one reported in the publication, making it difficult to use that fluorochrome pair.

These issues were investigated, and solutions were identified to enable the use of the full 40-color panel without compromises in performance. First, regarding the PE/Fire 810 tandem stability, the manufacturer stated the fluorochrome meets specifications and the authors did not observe tandem degradation when the reagent was used before its expiration date. To better understand the issues reported by the readership, a stability test was performed with exposure to fixation and light. No stability issues were observed with fixation with paraformaldehyde (1% or 4% in Phosphate buffered saline (PBS)). However, changes in the PE/Fire 810 spectrum and higher spillover values into PE were observed starting at 2 h of light exposure. Tandem breakdown can have serious impacts on the data, in this case specifically with any PE conjugates. For comparison, the PE/Cyanine 7 (Cy7) reagent included in the panel was tested in parallel and exhibited a more stable behavior (Figure 1A). The data suggests that PE/Fire 810 might need to be handled with even more care than other tandems. Next, multiple spectra were observed when using TCRγδ PerCP-eFluor710, without consistency across lots (Figure 1B). Data unmixing and TCRγδ population resolution are highly impacted by the presence of multiple spectra (Figure 1C). Finally, we observed that the use of BB515 and FITC in combination leads to increase in spread not only for those two fluorochromes, but also from Spark Blue 550 and PE. The spread is reduced when either BB515 or FITC are removed (Figure 1D).

To overcome the described performance issues and the reagent availability limitations, a rigorous investigation was done to identify fluorochromes and reagents that could be used as alternatives. The goal was to preserve all the markers included in the original publication and to ensure that the panel had equal or better performance. The steps used to identify alternative fluorochromes were the same as described in the original publication: (a) the spectrum needed to be unique, (b) new fluorochromes could not introduce significant spread into the other fluorochromes already in the panel, and (c) the brightness needed to be adequate to address the antigen densities of the new fluorochrome pairs. In addition, stability, tendency to aggregate, and interaction with other fluorochromes were evaluated. Five reagents had a direct fluorochrome replacement with new fluorochromes having similar emission spectrum and similar or higher CD4 resolution compared with the original fluor: Spark Violet 538 for Pacific Orange; cFluor blue yellow green (BYG) BYG710 for PE-Alexa Fluor 700; cFluor R720 for APC-R700; and PerCP-Vio 700 for PerCP-eFluor 710. cFluor Yellow Green (YG) YG610 was identified as an option to replace PE-Alexa Fluor 610, with low excitation by the blue laser, reducing the spread into blue laser excited fluorochromes with similar emission. To overcome the high spread between FITC and BB515, FITC was replaced by cFluor B532. We observed a significantly lower similarity index and lower spillover spread value between BB515 and cFluor B532 (similarity index of 0.97 for BB515-FITC vs. 0.89 for BB515-cFluor 532; [Figure S2A,B]). With careful PE/Fire 810 handling in sample staining, we did not observe stability issues during the development of OMIP-069. However, because of the performance reports we received regarding this reagent, we evaluated alternative fluorochromes to replace it. As mentioned before, we focused on fluorochromes with similar 561 nm excitation and far-red emission, stable to exposure to light and fixation, and with minimal spread impact into other fluorochromes. cFluor BYG750 was selected as a replacement. Resolution was first assessed by measuring stain index of peripheral blood mononuclear cells (PBMCs) labeled with anti-human CD4 conjugated to the new fluorochromes, as shown in Figure S1. Overall, spread was reduced in critical areas, such as the impact on BB515, when using original panel (Spillover spreading value (SSV) from cFluor B532 into BB515 10.99 [Figure S2C]) versus the revised panel (SSV from FITC into BB515 2.95, [Figure S2D]). Figure S2E shows the reduced spread into BB515 and Spark Blue 550 when FITC is replaced using CD4 single stained cells. Similar observations were seen for the markers in the panel when using cells stained with the full 40-color fluorochrome combination (Figure S2F).

Reagents of the same specificities and equivalent clones conjugated to the new fluorochromes were then tested. As shown in Figure 2, most of the new conjugates showed a similar resolution and pattern at optimal titer compared with the predicate reagent. However, the staining patterns for CD57 cFluor B532, CD24 cFluor YG610, and TCRγδ PerCP-Vio 700 were slightly different compared to the original reagents (highlighted with asterisks in Figure 5 and Figure S5). For CD57 cFluor B532, we evaluated its performance compared to PE (considered as a gold standard), FITC and Spark Blue 515 conjugates, all from the same clone. Cells were also stained with CD3 to evaluate expression in T and non-T cells. (Figure 3A,B). The staining pattern obtained with CD57 cFluor B532 was equivalent to the ones with the PE and Spark Blue 550 reagents, whereas the original CD57 FITC exhibited a lower percentage of positive cells in both compartments. We next investigated the differences in performance in a multicolor panel that allowed us to identify different T cell subsets. We found that CD57 cFluor B532 performed very similarly to the PE conjugate across T-cell subsets. Based on all these findings, CD57 cFluor B532 was identified as a suitable and superior reagent for the panel compared to the original CD57 FITC. Concerning CD24 cFluor YG610, co-staining with CD19, to identify B cells, and comparison with the original reagent revealed that the staining was specific and provided adequate resolution, and that the new reagent had less non-specific binding to non-B cells and hence included in the final panel (Figure 3C). Finally regarding TCRγδ, although changes in the staining pattern were observed between the original and replacement reagents, the frequencies of positive events remained unchanged (Figure S5). Of note, it was not possible to keep the clone consistent for this marker, but both clones (B1.1-original panel vs. REA591-revised panel) are pan identifiers of all circulating TCRγδ. We hypothesize that the two clones are binding to different conformational epitopes causing the differences in staining profiles, with three diagonals varying in median fluorescence intensity (MFI) observed using the B1.1 clone, in contrast to one single diagonal using REA591 clone. Confirmation that delta variants Vd1 and Vd2 are equally identified with both clones is presented in OMIP-xxx by the expression of TIGIT, CD38, CD27, and CD45RA within the CD3+ TCRγδ+ population.

As a result of this careful evaluation, the following reagents were identified as suitable alternatives: CD20 Pacific Orange™ was replaced by CD20 Spark Violet™ 538 (Biolegend Cat. 302,373), CD25 PE Alexa Fluor™ 700 by CD25 cFluor® BYG710 (Cytek Biosciences Cat. R7-20660), CD24 PE Alexa Fluor™ 610 by CD24 cFluor® YG610 (Cytek Biosciences Cat. R7-20658), TCRγδ PerCP-eFluor™ 710 by TCRγδ PerCP-Vio® 700 (Miltenyi Cat. 130–113-514), and CD127 APC-R700 by CD127 cFluor® R720 (Cytek Biosciences Cat. R7-20664) [3]. Two specificities were reassigned: HLA-DR PE/Fire™ 810 was replaced by HLA-DR BV480 (BD Biosciences Cat. 752499) and IgD BV480 by IgD cFluor® BYG750 (Cytek Biosciences Cat. R7-20662). Final panel designs (Figure S3A,B) are shown. A list of reagents suggested for this revised panel is included in Table 1. Titration results for the new eight reagents are shown in Figure S4.

After the replacement fluorochromes and reagents were thoroughly tested in the single-color scenario, their performance was evaluated in fully stained samples comparing the two 40-color combinations (the one originally published in OMIP-069, and the revised one with the replacement fluorochromes). Panel optimization was conducted as described in the original paper, including evaluation of spillover spread, marker resolution assessment, and evaluation of the ability to quantify all described populations by manual gating. A side-by-side comparison of the gating strategy used in OMIP-069 was included for both panels, displaying all populations of interest, all makers in the panel and, highlighting in blue frames, all reagent modifications (Figure S5). Definition of gate boundaries and identification of subsets of interest using manual gating was equally straightforward. Careful evaluation of performance across different donors allowed us to conclude that resolution obtained with this new combination of fluorochromes is comparable with the original panel. Of note, during assay optimization, modifications were introduced to the staining protocol as well as to the gating strategy to avoid double counting events. For detailed information please refer to OMIP-XXX [4]. In addition, two computational algorithms were used to evaluate if both panels can resolve similarly the heterogeneity of immune subsets identifiable with this combination of markers by high-dimensional analysis. FlowSOM was used for clustering [5] and Uniform Manifold Approximation Projection (UMAP) for dimensionality reduction [6]. Following the analysis strategy presented in OMIP-069, Figure 4 shows representative UMAP plots, overlaid with 34 cell clusters for the entire CD45+ population comparing both panels (original and revised). With both panels, all the populations gated manually can be identified as unique clusters, including subsets identified at different frequencies, from high (natural killer [NK] cell populations, B cell memory subsets, monocytes, CD4+ or CD8+naïve, and memory T cell subsets) to low (γδ T cells, dendritic cells [DC] subsets, basophils, or innate lymphoid cells [ILCs]). UMAP visualization simplified the performance comparison between the original and revised panel, showing similar phenotype across corresponding islands, as well as correlation in island location, size, and proximity in the lymphoid and myeloid sub compartments. Figure 5 shows in detail the performance of highly overlapping fluorochrome pairs in both panels, revealing high resolution for all the newly tested combinations.

We report that with the identified fluorochrome substitutions, the resolution of all cellular subsets identified with OMIP-069 was preserved, and in some cases improved. We carefully determined that the updated version of the panel has comparable performance, optimal resolution, and robust identification of all subsets of interest. Moreover, based on the experience with OMIP-069, revisions of OMIPs over time seem suitable and critical to keep panels relevant, address issues encountered by the readership or the authors, and benefit from advancements, such as the development of new reagents after publication.

Lily M. Park: Methodology; validation; software; data curation; formal analysis. Joanne Lannigan: Conceptualization; writing – review and editing; data curation. Quentin Low: Methodology; data curation; visualization; formal analysis. Maria C. Jaimes: Conceptualization; methodology; funding acquisition; formal analysis; project administration; resources; supervision; writing – original draft; writing – review and editing. Diana L. Bonilla: Writing – original draft; writing – review and editing; conceptualization; methodology; formal analysis; visualization; supervision; data curation; software; validation.