Protocol for generating high-quality, 45-color spectral flow cytometry data for unsupervised clustering to investigate aging in human PBMCs

Summary

Spectral flow cytometry enables the use of high-parameter panels. Using a Sony ID7000-optimized 45-color panel, we present a protocol for acquiring and preprocessing high-quality, high-parameter spectral flow cytometry data. We describe steps for batch-learning-assisted comparative cluster identification to study age-induced differences in human peripheral mononuclear cells. We detail spectral flow cytometry procedures, including sample staining, spectral reference acquisition, and unmixing. We also provide data preprocessing steps, including biexponential axis optimization, which is important for accurate dimensionality reduction and cluster identification.

Graphical abstract

Highlights

• •Instructions on how to acquire high-quality, reusable spectral references
• •Steps for staining human PBMCs with a 45-color panel to study immune aging
• •Guidance on how to unmix and preprocess high-quality spectral datasets
• •Procedures for comparative, unsupervised clustering and statistical analysis

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

Spectral flow cytometry enables the use of high-parameter panels. Using a Sony ID7000-optimized 45-color panel, we present a protocol for acquiring and preprocessing high-quality, high-parameter spectral flow cytometry data. We describe steps for batch-learning-assisted comparative cluster identification to study age-induced differences in human peripheral mononuclear cells. We detail spectral flow cytometry procedures, including sample staining, spectral reference acquisition, and unmixing. We also provide data preprocessing steps, including biexponential axis optimization, which is important for accurate dimensionality reduction and cluster identification.

Before you begin

This protocol is aimed at investigators using cryopreserved PBMCs to study immune aging. Panel design rules were followed as previously described.¹ The 45-color aging panel (Table 1; Figure 1) was designed using the IntelliPanel^TM panel design tool provided by FluoroFinder and optimized on a 6-laser-configurated Sony ID7000^TM Spectral Cell Analyzer. Unmixing performance was assessed using beads, and the similarity index matrix as well as spillover and total spreading matrices² were calculated in FlowJo (SSM & TSM, Figure 1; Figure S1 and S2). All conjugates and the viability dye were titrated on a single donor, and the three-step staining protocol was empirically optimized guided by the following principles: First, all antibodies were added to the cells and, based on the staining profiles, the conjugates were divided into three master mixes, placing low-abundance antigens into the first master mix and easily stained antigens into the last master mix. Difficult-to-stain mostly comprised weakly expressed markers that were placed in the first staining step and included the chemokine receptors CXCR5 and CCR7, as well as CD62L, CD28, and γδTCR. Antigens with robust staining performance, including well-expressed lineage markers such as CD8, CD14, or CD56, were placed into the last (third) master mix. Fluorochrome size was considered to reduce steric hindrance. Larger fluorochromes, including the protein-based fluorochromes, such as PE or APC, and their tandem conjugates, were preferentially placed into the last master mix, but those with poor staining performance were placed in the first master mix. Similarly, smaller fluorochromes, such as brilliant violet or ultraviolet dyes, were preferentially placed in the second staining step (see Table 2 for the optimized and titrated staining groups).

Table 1.

Target human lymphocyte cell markers, assigned fluorochromes, ID7000 detection channels and biexponential axis parameters

No.	Excitation laser	Detection channel	Marker	Clone	Color/Format	Description (cell population)	Negative area	Extra negative area	Axis maximum
1-A	320	320–06	Streptavidin	N/A	QdotTM 525	N/A	2.58	0.09	70465
1-B	N/A	N/A	CD73	AD2	Biotin	Various, e.g., T and B cells	N/A	N/A	N/A
2	355	355–01	CD14	S18004B	Spark UV TM 387	Monocytes	2.29	0.28	422508
3	355	355–05	CD28	CD28.2	Brilliant UltravioletTM 496	T cell differentiation	2.34	0	36504
4	355	355–11	ICAM-1	HA58	Brilliant UltravioletTM 563	Monocytes and lymphocytes	2.2	0	152198
5	355	355–13	CD4	RPA-T4	Brilliant UltravioletTM 615	T helper cells and monocytes	2.26	0	104262
6	355	355–18	CD3	UCHT1	Brilliant UltravioletTM 661	Pan T cell marker	2.42	0.21	272462
7	355	355–25	TCRγδ	11F2	Brilliant UltravioletTM 737	γδ T cells	2.32	0	23312
8	355	355–31	CD11b	ICRF44	Brilliant UltravioletTM 805	Monocytes	2.3	0	16811
9	405	405–01	CD57	QA17A04	Brilliant VioletTM 421	Senescent T and NK cells	1.87	0.19	1.00E+06
10	405	405–02	CD45RO	UCHL1	Pacific BlueTM	Memory T cells	2.02	0	25999
11	405	405–04	CD20	2H7	Brilliant VioletTM 480	B cells	2.3	0.08	108901
12	405	405–05	CD62L	FMC46	StarBrightTM Violet 515	Various, e.g., naïve T cells, NK cells, and basophils	2.44	0.17	104889
13	405	405–07	CD8a	QA18A37	SparkVioletTM 538	Cytotoxic T and NK cells	2.2	0	23200
14	405	405–11	CD45RA	HI100	Brilliant VioletTM 570	T cell differentiation and B cells	2.39	0.09	192466
15	405	405–13	CD19	HIB19	Super Bright 600	B cells	2.51	0.01	33786
16	405	405–14	CD11c	BU15	mFluorTM Violet 610	Various, e.g., monocytes and dendritic cells	2.62	0.13	47751
17	405	405–18	CD26	L272	Brilliant VioletTM 650	T cells (e.g., MAIT, γδ T) and activated NK cells	2.23	0.03	24531
18	405	405–23	CD1c	L161	Brilliant VioletTM 711	Conventional dendritic cells (cDC2)	2.47	0.14	38485
19	405	405–26	CD24	ML5	Brilliant VioletTM 750	B cell differentiation	1.96	0.05	26316
20	405	405–29	CXCR5	J252D4	Brilliant VioletTM 785	B and T cells	1.98	0	7175
21	488	488–05	IgD	IA6-2	FITC	B cell differentiation	2.47	0	239686
22	488	488–07	CD123	7G3	RealBlueTM 545	Plasmacytoid dendritic cells and basophils	2.3	0.2	129293
23	405	488–12	CD33	WM53	StarBrightTM Blue 580	Monocytes and myeloid-derived suppressor cells	2.4	0.21	106130
24	488	488–14	CD38	HB7	RealBlueTM 613	Various, e.g., activated T cells and B cell differentiation, monocytes, NK cells	2.59	0.17	1000000
25	488	488–19	CD5	UCHT2	NovaFluorTM Blue 660/120S	T and B1 B cells	2.3	0.23	151941
26	488	488–22	CD159c	134591	RealBlueTM 705	T and NK cells	2.6	0	37270
27	488	488–26	CD16	3G8	RealBlueTM 744	Various, e.g., monocytes, T cells, and NK cell subsets	2.11	0.2	118976
28	488	488–29	CD148	A3	RealBlueTM 780	Various, e.g., monocytes, dendritic cells	2.38	0	48172
29	488	488–30	CD10	SN5c	StarBrightTM Blue 810	Immature B cells	2.61	0.25	82336
30	488	488–32	CD44	IM7	PerCP/FireTM 806	Broadly expressed on lymphocytes, monocytes, and non-leukocyte subsets	2.51	0	114780
31	561	561–11	CD95	DX2	PE	Various; death receptor, e.g., T memory stem cells and monocytes	2.45	0.24	231030
32	561	561–12	KLRG1	Z7-205.rMAb	RealYellowTM 586	Various, e.g., Senescent and memory T, NK cells	2.76	0.35	163255
33	561	561–14	CD161	HP-3G10	PE-eFluorTM 610	MAIT T and NK cells	2.93	0.24	73975
34	561	561–18	HLA DR	L243	PE/FireTM 640	Various antigen-presenting cell types, e.g., dendritic cells, monocytes, activated T cells & B cells	2.48	0.13	227707
35	561	561–19	CD56	5.1H11	PE/Cyanine5	T and NK cell subsets	2.44	0.25	496575
36	561	561–22	CD244	eBioC1.7 (C1.7)	PE-Cyanine5.5	Various, e.g., monocytes, T and NK cell subsets, and basophils	2.62	0.01	29285
37	561	561–23	CD31	WM59	StarBrightTM Yellow 720	Various, e.g., monocytes, T cell subsets (recent thymic emigrants), and platelets	2.5	0.22	113990
38	561	561–24	CD103	B-Ly7	NovaFluorTM Yellow 730	Tissue-resident memory T cells and B cell subsets	2.5	0.22	113990
39	561	561–29	CCR7	G043H7	PE/Cyanine7	T cell differentiation marker, B cells	2.35	0	28499
40	561	561–32	CD27	O323	PE/FireTM 810	T and B cell differentiation	2.32	0	22278
41	637	637–18	CD34	581	APC	Hematopoietic stem cells	2.9	0.55	208689
42	637	637–19	CD141	M80	Alexa FluorTM 647	Conventional dendritic cells (cDC1)	2.77	0.25	493834
43	637	637–23	CD94	HP-3D9	Red 718	NK and CD8 T cell subsets	2.5	0.01	64275
44	637	637–27	Viability	N/A	Zombie NIRTM	Viability marker	2.4	0.32	178069
45	637	637–32	CD45	HI30	APC/FireTM 810	Pan leukocyte marker	2.02	0	132185

Figure 1.

Optimization of a 45-color spectral panel

(A) “Spectral Viewer” overview of all 45 fluorescence spectra acquired using a six-laser–configured ID7000 (320, 355, 405, 488, 561, and 637 nm) with 182 detector channels.

(B) Spillover spreading matrix calculated in FlowJo V10.10.0 from single stains.

(C) Titration of antibodies in human PBMCs. The viability dye Zombie NIR and CD73-biotin/Qdot 525 streptavidin were independently optimized (see Table 2 for final volumes).

Table 2.

Master mix compositions for staining

Marker	Fluorochrome/dye	Volume (μL) per 3 M cells (1 donor)	Staining concentration (μg/mL)∗	Notes on staining performance
Step 1
CD4	BUV615	2	1	High density antigen but did not stain well in step 3, best in step 1, titrated concentration halved
CD159c	RB705	1	2	Low density antigen
CXCR5	BV785	4	1.6	Low density antigen
CD1c	BV711	4	2	Low density antigen, best performance in step 1, titrated concentration quadrupled
CD161	PE-eFluor610	1	1	Low density antigen, best performance in step 1, needed bright fluorochrome
CD38	RB613	2	4	Best performance in step 1, titrated concentration doubled. Needed bright fluorochrome
CCR7	PE/Cy7	2	4	Low density antigen
γδTCR	BUV737	4	8	Low density antigen, best performance in step 1
CD28	BUV496	2	4	Low density antigen, sensitive to steric hindrance, difficult to stain with several tested conjugates
CD62L	SBV515	4	None provided	Low density antigen, best performance in step 1
CD10	SBB810	2	None provided	Low density antigen
Staining buffer		Add to 100 μL
Step 2
ICAM1	BUV563	4	8	Small fluorochrome; titrated concentration doubled
CD103	NFY730	2	1.6	Best performance in step 2, titrated concentration doubled
CD20	BV480	1	2	Small fluorochrome, best performance in step 2
CD3	BUV661	2	2	Small fluorochrome, best performance in step 2
CD45RO	Pacific Blue	2	10	Small fluorochrome, titrated concentration halved
KLRG1	RY586	1	0.5	Small fluorochrome, best performance in step 2, needed bright fluorochrome
CD33	SBB580	2	None provided	Small fluorochrome
HLADR	PE/Fire640	2	1	Needed bright fluorochrome, best performance in step 2
CD5	NFB660/120S	2	1.6	Best performance in step 2, minor non-specific binding to monocytes
CD123	RB545	2	4	Small fluorochrome
CD141	AF647	2	2	Small fluorochrome
CD31	SBY720	1	None provided	Small fluorochrome
CD24	BV750	4	8	Small fluorochrome
CD26	BV650	2	4	Small fluorochrome; titrated concentration doubled
CD45RA	BV570	2	1	Small fluorochrome
CD73	Biotin	4	20	Small tag, needs to be stained before streptavidin-QD525
Staining buffer		Add to 100 μL
Step 3
CD45	APC/Fire 810	1	1	High density antigen, large fluorochrome
CD14	Spark UV387	1	4	High density antigen
CD44	PerCP/Fire806	2	4	Large fluorochrome, high density antigen, titrated concentration doubled
IgD	FITC	1	2	High density antigen, titrated concentration doubled
CD16	RB744	1	1	High density antigen, best performance in step 3
CD8	Spark V538	2	2	High density antigen, best performance in step 3; titrated concentration doubled
CD57	BV421	1	1	Exceptionally bright conjugate
CD19	SB600	2	1	Best performance in step 3, titrated concentration halved
CD11c	mFluorV610	4	23.2	Best performance in step 3, titrated concentration doubled
CD148	RB780	2	4	Performed well in step 3
CD95	PE	1	1	Large fluorochrome, titrated concentration doubled
CD34	APC	2	2	Large fluorochrome, titrated concentration doubled
CD56	PE/Cy5	2	2	Large fluorochrome
CD27	PE/Fire810	4	8	Large fluorochrome
CD244	PE-Cy5.5	2	1	Large fluorochrome
CD11b	BUV805	2	4	High density antigen
CD94	R718	2	4	Performed well in step 3
Streptavidin	Qdot525	1.5	15 nM	Needs to be stained after CD73-biotin
Live–dead	Zombie NIR	2 of 100X stock	Not provided	Added to final staining step to account for staining-induced cell death
Staining buffer		Add to 100 μL

^∗unless indicated otherwise.

In this panel, we included several important lineage markers for T cells (CD3, CD4, and CD8) including gamma-delta T cells (γδTCR); mucosal-associated invariant T (MAIT) cells (CD161); B cells (CD19 and CD20); natural killer (NK) cells (CD56); monocytes (CD14); several dendritic cell (DC) subsets including CD123+ plasmacytoid DCs, CD141+ cDC1, CD1c+ cDC2; hematopoietic stem cells (CD34); and the pan-leukocyte marker CD45.³ Within the T cell compartment, several functional markers allow for the discrimination of differentiation (CD27, CD28, CD45RA, CD45RO, CCR7, CD62L, and CD95) or activation status (HLA-DR and CD38). Several markers of cytotoxic T cells, such as CD244, CD94, and CD159c, as well as the well-recognized markers of T cell senescence KLRG1 and CD57, were included. CD103 marks tissue-resident memory T cells, of which a small population may be found in circulation. Within the B cell compartment, the differentiation markers CD24, CD38, IgD, CD27, CD73, CXCR5, and CD10 were included. For NK cell differentiation, CD16, CD57, CD62L, CD94, CD244, CD62L, and CD159c were included, while the panel incorporated CD11b, CD11c, CD16, CD33, CD148, and ICAM1 for monocyte classification. Several markers are not restricted to a single cell type; for example, many are shared between T and NK cells, including CD56, KLRG1, CD62L, CD57, CD244, CD94, CD159c, and CD16; T and B cells share CD5, CD27, CD73, and CD38. The death receptor CD95 is expressed on many subsets, including monocytes and T cells, such as T memory stem cells (TSCMs). For instance, T, NK cells, and monocytes share CD16 and CD244. CD31 marks monocytes as well as recent thymic emigrant T cells. CD148 is found on monocytes and DC subsets. CD33 is expressed on monocytes as well as myeloid-derived suppressor cells. The human leukocyte antigen class II molecule HLA-DR is constitutively found on DCs, B cells, monocytes, and activated T cells. CD44 is as broadly expressed as the pan leukocyte marker CD45 but can also be found on non-leukocyte cells, such as mesenchymal stem or endothelial cells. In addition to the well-known senescence markers CD57 and KLRG1, several others implicated in aging and senescence were included in this panel to explore the potential differences in expression between older and younger peripheral blood mononuclear cells (PBMC) donors, including ICAM1,⁴ CD26,⁵ CD148,⁵ CD244,⁶ and CD73.⁷

Institutional permissions

In this protocol, only commercially purchased, de-identified PBMCs were used. Investigators aiming to reproduce this work must obtain appropriate ethical approvals from their own institutions.

Innovation

Spectral flow cytometry enables the design and use of very high-color panels increasing the complexity of manual gating strategies and complicating data analysis. In the analysis of high parameter panels, manual gating is user-dependent and therefore prone to bias reducing reproducibility and limiting the ability to fully leverage complex datasets. This protocol aims at enhancing the reproducibility and usability of high-parameter spectral flow cytometry panels. Using an ID7000 optimized 45-color spectral flow cytometry panel, this protocol guides through all major steps of high-quality data generation and analysis. The workflow includes an optimized 3-step staining protocol of human PBMCs, the acquisition of accurate spectral references (SR, beads or cells), minimal required manual SR adjustment (SRA) and essential data pre-processing steps such as optimization of axis parameters. As innovative feature, the data analysis pipeline enables consistent data cleaning, normalization, and clustering using the Spectral Flow Analysis (SFA) platform. The main difference from existing methods is the quantitative analysis of differences between groups (e.g., old and young PBMC donors) using BL-FlowSOM,⁸ an accelerated and batch-learning–assisted version of the FlowSOM⁹ algorithm that avoids human bias in the form of extensive, manual gating. Instead, gating is used only to define clean, contamination-free cell populations of interest (e.g., CD8⁺ T cells). Subsequently, concatenated unsupervised cluster identification is performed within these populations, and event numbers from the identified clusters and meta-clusters are exported for quantification and statistical analysis between groups.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
QD525-Streptavidin (1:67)	Thermo Fisher Scientific	Cat#Q10143MP
CD73-biotin (1:25)	Sony	Cat#2320085, RRID: AB_2564561
CD14-Spark UV™ 387 (1:100)	Sony	Cat#2596080, RRID: AB_2922626
CD28-BUV496 (1:50)	BD Biosciences	Cat#741168, RRID: AB_2870741
ICAM1-BUV563 (1:25)	BD Biosciences	Cat#741374, RRID: AB_2870873
CD4-BUV615 (1:50)	Thermo Fisher Scientific	Cat#366-0049-42, RRID: AB_2920982
CD3-BUV661 (1:50)	BD Biosciences	Cat#612965, RRID: AB_2916886
γδTCR-BUV737 (1:25)	BD Biosciences	Cat#748533, RRID: AB_2872944
CD11b-BUV805 (1:50)	Thermo Fisher Scientific	Cat#368-0118-42, RRID: AB_2896088
CD57-BV421 (1:100)	BioLegend	Cat#393326, RRID: AB_2860964
CD45RO-Pacific BlueTM (1:50)	Sony	Cat#2121080, RRID: AB_493659
CD20-BV480 (1:100)	BD Biosciences	Cat#566132, RRID: AB_2739531
CD62L-SBV515 (1:25)	Bio-Rad Laboratories	Cat#MCA1076SBV515, RRID: AB_3099935
CD8-Spark VioletTM 538 (1:50)	Sony	Cat#1375100, RRID: AB_2890706
CD45RA-BV570 (1:50)	Sony	Cat#2120660, RRID: AB_2563813
CD19-SB600 (1:50)	Thermo Fisher Scientific	Cat#63-0199-42, RRID: AB_2744834
CD11c-mFluorTM Violet 610 (1:25)	Novus Biologicals	Cat#NBP1-45018MFV610, RRID: AB_3210476
CD26-BV650 (1:50)	BD Biosciences	Cat#745394, RRID: AB_2742951
CD1c-BV711 (1:25)	Sony	Cat#2257680, RRID: AB_2629760
CD24-BV750 (1:25)	BD Biosciences	Cat#746890, RRID: AB_2871688
CXCR5-BV785 (1:25)	Sony	Cat#2384680, RRID: AB_2629528
IgD-FITC (1:100)	Sony	Cat#2341030, RRID: AB_10612567
CD123-RB545 (1:50)	BD Biosciences	Cat#569269, RRID: AB_3684923
CD33-SBB580 (1:50)	Bio-Rad Laboratories	Cat#MCA1271SBB580, RRID: AB_3100170
CD38-RB613 (1:50)	BD Biosciences	Cat#759239, RRID: AB_3691330
CD5-NFB660/120S (1:50)	Thermo Fisher Scientific	Cat#H034T03B08-A, RRID: AB_3098435
CD159c-RB705 (1:100)	BD Biosciences	Cat#756941, RRID: AB_3689107
CD103-NFY730 (1:50)	Thermo Fisher Scientific	Cat#H075T03Y07-A, RRID: AB_3098834
CD16-RB744 (1:100)	BD Biosciences	Cat#570475, RRID: AB_3685768
CD148-RB780 (1:50)	BD Biosciences	Cat#755640, RRID: AB_3687976
CD10-SBB810 (1:50)	Bio-Rad Laboratories	Cat#MCA1556SBB810, RRID: AB_3100312
CD44-PerCP/FireTM 806 (1:50)	Sony	Cat#1115410, RRID: AB_3083253
CD95-PE (1:100)	Sony	Cat#2128040, RRID: AB_314546
KLRG1-RY586 (1:100)	BD Biosciences	Cat#568493, RRID: AB_3684320
CD161-PE-eFluorTM 610 (1:100)	Thermo Fisher Scientific	Cat#61-1619-42, RRID: AB_2574585
HLA-DR-PE/FireTM 640 (1:50)	Sony	Cat#2138380, RRID: AB_2876604
CD56-PE/Cyanine5 (1:50)	Sony	Cat#2412580, RRID: AB_2564089
CD244-PE-Cyanine5.5 (1:50)	Thermo Fisher Scientific	Cat#35-5838-42, RRID: AB_2784679
CD31-SBY720 (1:100)	Bio-Rad Laboratories	Cat#MCA1738SBY720, RRID: AB_3100508
CCR7-PE/Cy7 (1:50)	Sony	Cat#2366130, RRID: AB_10913813
CD27-PE/FireTM 810 (1:25)	Sony	Cat#2114300, RRID: AB_2894557
CD34-APC (1:50)	Sony	Cat#2317550, RRID: AB_1877153
CD141-AF647 (1:50)	Sony	Cat#2320620, RRID: AB_2800918
CD94-R718 (1:50)	BD Biosciences	Cat#752211, RRID: AB_2917318
Zombie NIRTM Fixable Viability Kit (1:500)	Sony	Cat#2715530
CD45-APC/FireTM 810 (1:100)	Sony	Cat#2120380, RRID: AB_2860792
Experimental models: Cell lines
Normal Human PBMCs, Cryopreserved	Charles River Laboratories or iQ Bioscience	Cat#PB009C-2 or Cat#IQB-PBMC102
Software and algorithms
Spectral Flow Analysis (SFA) - Life Sciences Cloud Platform V1.1	Sony	https://www.sonybiotechnology.com/instruments/sfa-cloud-platform
ID7000 Software V2.1	Sony	https://www.sonybiotechnology.com/instruments/id7000-spectral-cell-analyzer/software
FlowJo V10.10.0	BD Biosciences	https://www.flowjo.com/
Prism V10.5.0	GraphPad Software	https://www.graphpad.com/
Other
Human TruStain FcXTM	BioLegend	Cat#422302
True-Stain Monocyte BlockerTM	BioLegend	Cat#426103
CellBloxTM blocking buffer	Thermo Fisher Scientific	Cat#C001T06F01
Phosphate-buffered saline (PBS) 1×	Thermo Fisher Scientific	Cat#10010031
Fetal calf serum (FCS)	Biowest	Cat#S1400-500
Roswell Park Memorial Institute (RPMI) medium 1640	FUJIFILM Wako	Cat#189-02025
Penicillin/Streptomycin	Sigma-Aldrich	Cat#P4333-100ML
UltraComp eBeads™ Plus Compensation Beads	Thermo Fisher Scientific	Cat#01-3333-42
DNase I	Thermo Fisher Scientific	Cat#18047019
ID7000TM Spectral Cell Analyzer, 6 laser model	Sony	Cat#LE-ID7000E
Purified water	Trusco Nakayama	Cat#W10-A2
Align Check beads	Sony	Cat#LE-AE700510
8 Peak beads	Sony	Cat#LE-AE700522
NucleoCounter® NC-202™	ChemoMetec	https://chemometec.com/nucleocounters/nc-202/
Corning® 15 mL PP Centrifuge Tubes, Bulk Packed with CentriStar™ Cap, Sterile	Corning	Cat#430791
Corning® 50 mL PP Centrifuge Tubes, Conical Bottom with CentriStar™ Cap, Bulk Packed, Sterile	Corning	Cat#430829
Falcon® 5 mL Round Bottom High Clarity PP Test Tube, with Snap Cap, Sterile	Corning	Cat#352063
Falcon® 5 mL Round Bottom Polystyrene Test Tube, with Cell Strainer Snap Cap	Corning	Cat#352235
Via2-Cassette™	ChemoMetec	Cat#941-0023

Materials and equipment

Preparation of buffers and reagents

Timing: 1 h

Staining and washing buffer

• •Under a sterile bench, add fetal calf serum (FCS) at a final concentration of 2% (v/v) to a bottle containing phosphate buffered saline (PBS, w/o Ca²⁺ and Mg²⁺), e.g., 20 mL FCS to 1000 mL PBS. Keep at 4°C. Aliquot required amounts of buffer.

Note: Keep the buffer sterile to avoid contamination; 0.1% (w/v) sodium azide or a non-toxic alternative, NeoCide PC-300, may be added as an anti-fungal and anti-bacterial agent.

CRITICAL: Sodium azide is highly toxic; hence, handle, store, and dispose of it according to your institutional guidelines.

Zombie NIR stock and working aliquots

• •Add 500 μL DMSO (supplied with the Zombie NIR dye) to reconstitute the lyophilized dye (500 tests), creating a 1000× stock solution.
• •Aliquot (e.g., 5 μL) and store at −20°C.
• •Prepare a single-use working dilution by diluting the stock solution in deionized water or PBS at 1:10. Avoid re-freezing after use for optimal staining performance.

Note: Make sufficiently small aliquots to avoid excessive freeze-thaw cycles of the stock solution, which can degrade the dye.

Complete RPMI for thawing PBMCs

• •Under a sterile bench, add 1× penicillin–streptomycin (1:100 of stock solution) and 10% (v/v) FCS to a bottle of RPMI 1640. Freshly add DNase (200 U/mL) when thawing PBMCs.

Complete RPMI for PBMC preparation

Reagent	Final concentration	Amount
RPMI 1640	N/A	500 mL
FCS	10%	55 mL
Penicillin–streptomycin (100×)	1×	5.5 mL
DNase I (add fresh to aliquoted medium)	200 U/mL	variable (depends on DNase I lot concentration)

Note: Thaw frozen FCS gently at 4°C to preserve serum quality.

Step-by-step method details

Acquisition of reusable spectral references

Timing: 2–3 h

This step details the acquisition of high-quality spectral references from single-stained beads or cells.

• 1.
  Preparation of unmixing controls (beads).

  • a.
    Prepare the required number of 5 mL polystyrene tubes for 45 single stains and one tube each for unstained beads and unstained cells (PBMCs or any other cell type).
  • b.
    Add 1 drop of compensation beads, e.g., UltraComp eBeads Plus, to each tube, except those designated for Zombie NIR and unstained cells.
  • c.
    Prepare single stains by adding 1 μL of antibody conjugate to the appropriate tube.
  • d.
    For CD73-biotin/streptavidin-Qdot 525 staining, first stain with CD73-biotin, wash, and re-stain with 1 μL streptavidin-Qdot 525 in 100 μL PBS.
  • e.
    Incubate for 15 min in the dark at 24°C.
  • f.
    Wash each single stain by adding 4 mL PBS.
  • g.
    Centrifuge at 500 g for 5 min at 24°C.
  • h.
    Carefully decant the supernatant (SN) and resuspend the beads in 500 μL PBS.
  • i.
    Keep on ice in the dark until data acquisition using the ID7000.

CRITICAL: Qdots contain toxic cadmium selenide. Handle with care and dispose of contaminated waste according to your institution's guidelines.

• 2.
  Alternatively to Step 1, prepare single-stained cells as unmixing controls.

  • a.
    Retrieve the cryotube containing PBMCs from the liquid nitrogen tank and thaw in a pre-warmed 37°C water bath for 1–2 min without agitation.
  • b.
    Under a sterile bench, wipe down the cryotubes with alcohol and mix gently with 10 mL pre-warmed RPMI/10% FCS/1× penicillin–streptomycin freshly supplemented with 200 U/mL DNase.
  • c.
    Aliquot a small volume (∼200 μL) into a 1.5 mL tube for viability and concentration measurements on a cell counter.

    Note: We used the ChemoMetec NucleoCounter NC-202 that uses single-use cassettes containing immobilized acridine orange (AO) and 4′,6-diamidino-2-phenylindole (DAPI and does not rely on manual pipetting. Perform cell viability and live cell concentration checks always at the same protocol step to maximize reproducibility.

  • d.
    Pellet cells at 500 g for 10 min at 24°C.
  • e.
    Carefully remove the SN and loosen the pellet by flicking the tube before adjusting the cells to 5 × 10⁶/mL in staining buffer plus freshly added 200 U/mL DNase at 24°C.
  • f.
    Add 5 μL of the following blocking reagents per 100 μL cells: Human TruStain FcX^TM Fc receptor blocking solution (BioLegend), True-Stain Monocyte Blocker^TM (BioLegend), and CellBlox^TM blocking buffer (Thermo Fisher Scientific).
  • g.
    Incubate for 5 min at 24°C.
  • h.
    Prepare labelled 5 mL tubes and add 50 μL cell suspension per tube.
  • i.
    Add the appropriate antibody to each tube at the previously titrated concentration (see Table 2; Figure 1C).
  • j.
    Incubate for 30 min at 24°C in the dark.
  • k.
    Wash each tube with 4 mL staining buffer.
  • l.
    Pellet 5 min at 24°C.
  • m.
    Carefully decant the supernatant and resuspend each pellet in 500 μL staining buffer, except for CD73-biotin.
  • n.
    For CD73-biotin/streptavidin Qdot 525 perform a second staining step with the streptavidin conjugate and wash as described above.

• 3.
  To prepare the viability dye control, Zombie NIR, heat-kill ∼2 × 10⁶ (primary or cultured) cells through incubation at 65°C for 5 min.

  Note: Alternatively, viability dye-binding beads such as ViaComp beads (Slingshot Biosciences) can be used.

  • a.
    Pellet the heat-killed cells at 500 g for 3 min at 24°C.
  • b.
    Remove the SN and resuspend the pellet in 1 mL PBS.
  • c.
    Divide the cells equally (500 μL each) into two 5 mL tubes: One should be assigned for the unstained control and left untreated; the other should receive 0.5 μL reconstituted Zombie NIR solution for a 1000× dilution.
  • d.
    Incubate for 15 min at 24°C in the dark.
  • e.
    Pellet at 500 g for 3 min at 24°C; carefully remove the SN and wash the pellet with 4 mL PBS.
  • f.
    Pellet again and resuspend in 500 μL PBS.
  • g.
    Keep on ice in the dark until analysis using the ID7000.

• 4.
  Acquire single stains.

  • a.
    Before starting up the analyzer, ensure the sheath tank is filled and the waste tank is empty.

    Note: The ID7000 Spectral Cell Analyzer supports the use of deionized water or isotonic buffers as sheath fluid. Please contact your local service team to switch the type of sheath fluid. For this study, deionized water was used.

  • b.
    Switch on the ID7000 (toggle the two main switches on the left side of the machine and press the “Power” button) and PC.
  • c.
    Log into the ID7000 software and perform priming once the priming screen pops up.
  • d.
    Prepare the Daily QC beads by adding two drops of Align Check Beads to 450 μL of deionized water or PBS in a 5 mL tube.
  • e.
    If Performance QC has not been performed within 1 week of the experiment, prepare the Performance QC beads by adding two drops of 8 Peak Beads to 450 μL of deionized water or PBS in a 5 mL tube.
  • f.
    After priming is completed, go to the QC tab and select “Daily and Performance QC,” place the tubes into the extra station as instructed by the QC wizard, select the correct lot for the QC beads, and click “Start”.

    Note: Daily QC is obligatory and is used to set up the system including the Standardization mode (https://resources.sonybiotechnology.com/Tech_Notes/Sony_ID7000_Improved_Flow_Cytometry_Standardization_Technical_Note.pdf). Performance QC checks the fluorescence detection performance. Ensure to properly mix beads before making the dilutions.

• 5.
  Create the experiment in the “Experiment Designer”:

  • a.
    In the “Experiment” tab, select “Experiment Designer” and add all colors of the 45-color panel under “Add New Spectral Reference”.

    Note: Some newer fluorochromes may not be part of the ID7000 Spectral Reference library. In that case, a window will pop up prompting the addition of a new fluorochrome. Select the correct excitation laser and emission channels and provide a color of your choice. In case laser or emission channels are incorrectly assigned, they can be changed later in the “Unmixing Settings” by right-clicking on the fluorochrome.

  • b.
    Select all lasers to be used for your panel and click “Next”.

    Note: For spectral flow cytometry, higher numbers of excitation lasers enhance performance. However, for very high-parameter panels that do not contain infrared-laser excited fluorochromes, when using a 7-laser ID7000, we recommend a 6-laser configuration with the 808 nm laser switched off to achieve maximum sensitivity, as the 808 nm laser requires notch filters that reduce the number of available detection channels.

  • c.
    Select the appropriate plate type, e.g., “24 Tube Rack 5mL”.
  • d.
    Optional: Place a checkmark in the “Use a common unstained control for single stains” box.

    Note: While using unstained beads or cells as a common negative control is convenient, best practice is using the negative population of each single stain as a negative control for unmixing SRs. However, an unstained control should always be manually added to the plate since some fluorochromes may non-specifically bind the negative population (i.e., beads that do not bind antibodies or cells that do not express the target antigen). In this case, the unstained beads or cells should be used for unmixing the negative population. Additionally, include an extra tube containing unstained, heat-killed cells to serve as a negative control for the viability dye as heat killing will alter the autofluorescence profile.

  • e.
    Check that “Standardization” is selected under mode.
  • f.
    In the next window, set sample acquisition parameters, including flow speed (we used 8.5), detector settings (FSC 5, SSC 4.78, all detectors 4.78) cleaning (inner & outer), agitation (Once & Short), and acquisition offset time (3 s).
  • g.
    Select your preferred sample acquisition order; click next and select “Create Experiment”.

    Note: Consider lowering the detector voltages if signal intensity exceeds 10⁵ to avoid oversaturation which affects the quality of the calculated spectral references.

    CRITICAL: Re-use of spectral references requires the acquisition of both single stains and fully stained samples in Standardization mode. Standardization mode sets the system to an optimized master specification (integer values) that allows to maintain instrument settings between experiments and across multiple instruments. Standardization mode adjusts the output of each channel so that the SSC and fluorescence detection sensitivities are aligned across multiple instruments. On the other hand, Normal mode sets detector voltages as a percentage of the maximum voltage for each detector. When Normal mode is selected, detector voltages between single-stains and multi-stain samples must match. Similar settings cannot be transferred to a different ID7000 but need to be optimized individually. Therefore, we strongly recommend Standardization mode for this protocol.

• 6.
  Acquire single stains and obtain spectral references.

  Note: In the newly created experiment, all samples are automatically added to the “Auto Acquisition” group and the worksheets contain a FSC vs. SSC plot with a gate and spectrum plots with positive and negative gates based on the excitation laser and emission channels and ribbon plots for those gates. Minor corrections of the size or positions of those gates may be necessary after data acquisition.

  • a.
    Add the single stains and set the number of events to be acquired in the “Stopping Conditions” and click “Auto Acquire”.

    Note: Acquisition of ca. 5–10,000 events per tube is sufficient for beads. If single-stained cells are used as unmixing controls, ∼20,000 events should be considered or more for very rare markers.

  • b.
    After completing the automatic sample acquisition, verify the correct positioning of the positive and negative gates (see Figure 2 for setting appropriate gates) for spectral unmixing.
  • c.
    For Zombie NIR, the heat-killing treatment eliminates the unstained population. Therefore, delete the negative gate and assign a singlet gate from the unstained heat-killed cells as the negative population by right-clicking on the gate and selecting “Assign Gate” > “Negative Gate” > “Zombie NIR”.
  • d.
    Go to the “Unmixing Settings” and select “Calculate”.
  • e.
    You might be asked if oversaturated events should be excluded before calculating the SRs—always check the box.
  • f.
    In the “Matrix” tab, you can opt to export the final unmixing matrix for use on a fully stained sample.

Figure 2.

Obtaining spectral references from single stains, using CD45RO-Pacific Blue-stained beads as an example

(A) Scatter plot with gate set on beads.

(B) Positive and negative gates are automatically placed on a spectrum plot using the “Experiment Designer”. Alternatively, gates used for unmixing the SR can also be placed on dot plots or histograms and assigned as positive or negative gates.

(C) Ribbon plots for positive and negative gates, as well as unstained beads. Confirm that the positive gate contains one clean spectrum and that the spectrum in the negative gate resembles the unstained beads. In case of nonspecific binding, select the unstained beads as the negative population.

(D) In the “Color Panel” tab of the “Unmixing Settings,” confirm that positive and negative gates have been set for all colors and calculate the SRs.

(E) In the “Matrix” tab of the “Unmixing Settings”, unmixing matrices can be exported or imported.

Preparation of fully stained PBMC samples

Timing: ∼3 h

This step details the three-step staining protocol for human PBMCs with the 45-color panel.

• 7.
  Prepare PBMCs.

  • a.
    Retrieve the cryotubes containing PBMCs of young (age range 18–21 years) and old donors (age range 68–75 years) from the liquid nitrogen tank and thaw in a pre-warmed 37°C water bath for 1–2 min without agitation.
  • b.
    Under a sterile bench, wipe down the cryotubes with alcohol and mix gently with 10 mL pre-warmed RPMI/10% FCS/1× penicillin–streptomycin freshly supplemented with 200 U/mL DNase.
  • c.
    Aliquot a small volume (∼200 μL) into a 1.5 mL tube for viability and live cell concentration measurements using a cell counter (e.g., the ChemoMetec NucleoCounter® NC-202™).
  • d.
    Pellet cells at 500 g for 10 min at 24°C.

    Note: To reduce losses, PBMCs are always pelleted at 500 g; however, the centrifugation time is adjusted to the approximate volume—small volumes (∼2 mL or less) are pelleted for 3 min, medium volumes (∼5 mL) are pelleted for 5 min, and large volumes (>5 mL) are pelleted for 10 min.

  • e.
    Carefully remove the SN and loosen the pellet by flicking the tube before adjusting the cells to 5 × 10⁶/mL in staining buffer plus freshly added 200 U/mL DNase at 24°C.

• 8.
  Prepare master mixes and stain cells.

  Note: Staining was optimal when cells were kept at ambient temperatures (∼22–24°C). For cells kept on ice during all staining and washing steps, several lymphocyte-expressed antigens, including CXCR5, CCR7, CD62L, γδTCR, CD26, CD28, CD45RO, CD20, CD4, CD16, CD161, CD244, CD31 and CD141 showed poor staining efficiency compared to staining at 24°C (see Figure S3).

  CRITICAL: Partial internalization of surface markers expressed on monocytes was observed when staining at ∼24°C. We hence recommend performing the staining at 4°C if monocytes are the main cell type of interest.

  • a.
    Prepare master mixes before adding blocking reagent to the cells (left at 24°C in staining buffer plus DNase until preparation is completed).
  • b.
    Prepare antibody cocktails for three staining steps in cooled 1.5 mL Eppendorf tubes; protect from direct light exposure.
  • c.
    Table 2 shows the composition and necessary volumes, respectively concentrations, for one donor; the conjugates and viability dye may be added in any order.

    Note: We recommend preparing a surplus, e.g., 8.5-fold master mixes for staining 8 donors. As many antibody concentrations are lot-specific, consider re-titrating new antibody lots.

    CRITICAL: Antibody stocks and master mix cocktails must always be cooled and protected from light exposure.

  • d.
    To remove aggregates, pellet each master mix for 1 min at 10 000 g at 4°C and transfer the SN to pre-chilled 1.5 mL tubes.
  • e.
    For each donor, aliquot 3 × 10⁶ cells, corresponding to 600 μL cell suspension, into a 5 mL polypropylene tube with snap-cap and add 30 μL of each blocking reagent (5 μL per 100 μL): Human TruStain FcX^TM Fc receptor blocking solution, True-Stain Monocyte Blocker^TM, and CellBlox^TM blocking buffer.
  • f.
    Block cells for 5 min at RT before pelleting at 500 g for 3 min.
  • g.
    Carefully remove the SN using a pipette and loosen the pellet by flicking the tube.
  • h.
    Add the appropriate amount (Table 2) of master mix 1 to each donor tube and estimate the volume by pipetting the solution carefully with a 200 μL pipette set to 100 μL. Add staining buffer (without DNase) to a final volume of ∼100 μL.
  • i.
    Incubate the cells for 30 min at RT in the dark.
  • j.
    Add 4 mL staining buffer to each tube, close the lid, and invert thrice before pelleting at 500 g for 5 min at 24°C.
  • k.
    Decant the SN carefully and pipette the remaining liquid off without disturbing the pellet.
  • l.
    Flick the pellet gently and add master mix 2 before adjusting the final volume to ∼100 μL.
  • m.
    Repeat steps i–k for master mixes 2 and 3.
  • n.
    After the final staining step, wash cells twice with 4 mL staining buffer.
  • o.
    Prepare an unstained control: Take ∼500 μL cell suspension, add 500 μL staining buffer, strain through 35 μm mesh caps into a fresh 5 mL tube and keep on ice until measurement.
  • p.
    Resuspend the final pellets in 1 mL staining buffer, strain through 35 μm mesh caps into fresh 5 mL tubes and transfer all samples into the pre-chilled sample acquisition chamber of the ID7000 Spectral Cell Analyzer equipped with a pre-chilled 24 x 5 mL tube rack.

Data acquisition and analysis

Timing: 2–3 h

This step describes data acquisition and analysis, including autofluorescence extraction, unmixing and biexponential axis optimization, of the fully stained PBMC samples using the ID7000 software.

• 9.
  Prepare the instrument and set up the experiment.

  • a.
    Refer to Step 4 above on how to start up the ID7000.
  • b.
    Prepare a new experiment: On the left side in the “Experiment” tab, select “Blank template “ > “24 Tube Rack 5 mL,” then click “Create Experiment”.
  • c.
    In the “Acquisition” tab, right-click on the newly generated experiment and check that “Standardization” is selected under “Properties” > “Settings: Mode”.
  • d.
    In the plate panel, select the number of required wells, right-click, and select “Move to Sample Group” > “New Sample Group”.
  • e.
    Switch to the list view in the plate panel and name your samples.
  • f.
    Add plots on the worksheet; by default, new experiments are set up in “Shared Worksheet” mode, where all plots and axis settings will be set in the same way for all samples.
  • g.
    Creating plots for the entire gating strategy before acquiring cells is not necessary, but one FSC vs. SSC and a ribbon-plot set on a cell singlet gate is required to optimize detector voltages prior to sample acquisition: Add an FSC-A vs. SSC-A density plot to the worksheet by selecting “Density” from the top panel. Add a polygonal gate to the plot and a downstream singlet gate (e.g., FSC-A vs. FSC-H) and generate a “Ribbon plot”.
  • h.
    Load the 24-tube sample rack onto the cooling station.

Note: The rack may not be loaded during QC. Ensure the experiment is prepared and rack is loaded before sample preparation is completed to allow for cooling.

• 10.
  Load pre-acquired spectral references.

  • a.
    In the “Unmixing settings” on the “Unmixing” tab, add all fluorochromes and dyes to the color panel.
  • b.
    In the “Unmixing” tab, click on “Load Spectral Reference” and select all previously recorded spectra for your panel from the Spectral Reference library.
  • c.
    Alternatively, export the unmixing matrix generated during unmixing of the single-stain samples and re-import it into the fully stained cell experiment (Figure 2E).

Note: Standardization mode allows for the reuse of spectral references and unmixing matrices acquired on a different day or instrument setup under different settings. However, the use of same conjugate lot numbers and regular updating of spectral references is highly recommended to reduce unmixing errors that will require manual SRA, especially for tandem dyes prone to degradation.

• 11.
  Define acquisition conditions and optimize PMT detector settings.

  • a.
    Load the sample and press acquire (green arrowhead) in the “Status” panel.
  • b.
    In the “Instrument Settings” under “Flow control,” set the flow speed, “Acquisition Offset” to ∼3–5 s, and “Agitation” to “Once” and “Short” for cells.

    Note: Beads are sturdier and settle faster than cells. Therefore, “Cyclic” and “Normal” mode agitation are recommended.

  • c.
    In the “Detector & Threshold” tab of the “Instrument Settings” (found in the top “Analysis” tab), optimize the FSC and SSC detector values and determine an appropriate FSC threshold to cut off smaller events such as debris (e.g., 11%).
  • d.
    Optimize detector values: From the starting value in “Standardization mode” (3.78), incrementally raise the value and monitor signal intensities on the singlet ribbon plot for each detector array until the highest intensity peaks for each laser reach ∼10⁵. For this panel the following detector settings were used: FSC: 5, SSC: 4.78, 320: 5.3, 355: 5.4, 405: 4.2, 488: 5.2, 561: 5.0, 637: 5.6.

    Note: Avoid higher detector settings (>10⁵) as oversaturation reduces unmixing accuracy.

• 12.
  Acquire samples.

  • a.
    After optimizing the detector settings, pause the sample acquisition, configure “Stopping condition,” and add all the fully stained samples to the “Auto Acquisition” group.

    Note: For a high-parameter panel covering many cell types, including rare cells, acquisition of at least 1,000,000 events is recommended for fully stained samples.

  • b.
    Once the set-up is complete, click “Auto Acquire” to automatically acquire all fully stained samples.
  • c.
    Acquire 100,000 events from the unstained control.

    Note: Flow speed can be adjusted during auto-acquisition if desired. All other parameters will be locked.

  • d.
    Perform AF extraction using the “Autofluorescence Finder” in the “Unmixing” tab. Confirm that the unstained control is selected. See Figure 3 for a typical AF unmixing workflow.

    Note: The AF Finder will not be selectable if “Unmixing” is toggled “On” in the upper right corner of the worksheet.

    Note: Autofluorescence spectra are very similar for different human PBMC donors; hence, acquisition of only one unstained sample is usually sufficient. In case, stronger differences between samples are found, AF can also be unmixed separately for each sample, respectively, donor. In this case, create a new sample group for each sample and the corresponding unstained control, import the unmixing matrix for the fluorochromes and unmix the AF from each unstained sample.

• 13.
  Perform manual SR adjustments (SRA).

  CRITICAL: Unmixing errors are usually caused by higher degrees of spectral overlap or mismatches between SRs obtained from single-stain controls and the actual SRs on fully stained cells on the day of the experiment. Manual adjustments should be performed sparingly and require a well-optimized panel that avoids assigning spectrally similar fluorochromes on co-expressed markers. Otherwise, true co-expression may be difficult to distinguish from under-unmixing (see Problem 4). While using cells for SR calculation may overall lessen required SRA, dimmer fluorochromes or markers with very few positive cells tend to produce SR errors (negative SRs or wrong peaks, see Figure S4C). Bead and cell-derived matrices deliver comparable SFA results with minimal required SRA on an ID7000 (Figures S5A–S5D).

  • a.
    Check the unmixing results in the “Unmixing Viewer” (Figure 4). Select a suitable gate, e.g., singlet or live cell gate, and inspect the 2-by-2 plots for all fluorochromes and the AF for unmixing errors.
  • b.
    Toggle the “Adjuster” mode “on” and drag over- or under-unmixed populations to their correct position.
  • c.
    Correct incidences of over- or under-unmixing by carefully considering the expression patterns. Start by prioritizing the correction of prominent unmixing errors (see Figure 4). Addressing these major issues (see Note below on pairs that may require SRA) will automatically fix some of the minor errors in other color pairs. Fix minor errors only, if still present, after all major incidents have been corrected.
  • d.
    Once adjustments are complete, toggle the “Adjuster” mode “off” or click “close” and save the adjusted matrix under a new name.

    Note: Pairs that may need SRA include fluorochromes that emit at a similar wavelength even if the main excitation laser is different, e.g., BV711/SBY720, mFluorV610/PEeF610, BUV661/NFB660-120S or Percp-Fire806/PE-Fire810. Other combinations include tandem dyes and fluorochromes that emit close to their tandem base peak e.g. SparkUV-387 against brilliant violet dyes or pairs with relatively high spectral overlap, e.g., FITC and RB545. Adjusted pairs when using either beads or cells as unmixing controls are documented in Figure S5E.

• 14.
  Optimize biexponential axis settings, apply gating strategy, and export data.

  • a.
    Adjust the biexponential axes by using the ‘’Auto Adjust” buttons or directly manipulating the axes through drag-and-drop. The double-negative population should ideally form a rounded population on the lower left corner of the 2-by-2 plots, while the positive populations should utilize the full dynamic range of each axis.

    Note: In parallel to the SR adjustments, the “Unmixing Viewer” can also be used to adjust the scaling of all axes. See Figure 5 for examples of before and after biexponential axis adjustment. For markers with either few positive or negative cells, auto-adjusted axes may need manual corrections. In this panel this included CD44, CD45, CD34 and CD141. All biexponential axis parameters used in this protocol are detailed in Table 1. The impact of scaling on unsupervised analysis is shown in Figure S6.

  • b.
    Finalize the gating strategy (see Figure 6): Optimize gating to generate pure cell populations for SFA-based cluster identification. Therefore, apply a second singlet gate on the SSC parameter and “clean-up” gates to remove contaminants, such as residual B cells and monocytes, from the T cell population.

    Note: The outcome of the SFA clustering algorithms does not rely on gating beyond defining cell populations of interest, reducing the need for a complete and extensive manual gating strategy that also becomes very cumbersome for high color panels. Instead, use gating to define cell populations of interest for analysis using the BL-FlowSOM cluster identification algorithm, e.g., CD4 or CD8 T cells.

  • c.
    Convert the shared worksheet into individual worksheets by selecting all samples, right-clicking, and selecting “Change Worksheet Mode > Individual”.

    Note: For large data files, changing the worksheet mode may take some time.

  • d.
    Adjust gate positions if necessary; for example, CD8 and CD33 levels may vary between donors. In addition, the cell gate on the FSC vs. SSC scatter plot may need minor, donor-dependent adjustments.
  • e.
    Export the Exdat file as follows: “File” > “Database” > “Export”.

    Note: All the data processing steps (e.g., unmixing, AF extraction, SR and biexponential adjustment), except for data acquisition, can also be performed in the SFA software.

Figure 3.

Autofluorescence (AF) unmixing workflow for the 45-color panel

(A) Ribbon plot showing the fluorescence spectrum of unstained PBMCs. Autofluorescence is especially excited by the 320 nm and 355 nm lasers.

(B) In the ID7000 Software, the “Autofluorescence Finder” tool is employed to identify and unmix the AF of two distinct populations within the AF root gate, excited by the 320 nm and 355 nm lasers.

(C) AF-A and AF-B are highlighted in the “Spectral Reference Viewer” among the 45-color + 2 AF spectra. The fluorochromes with the greatest AF overlap are BUV496 (turquoise) and SparkUV 387 (purple).

(D) AF unmixing improves population separation by extracting AF contributions to the negative population for CD28-BUV496 and CD14-SparkUV 387.

Figure 4.

Manual SR adjustments in “Unmixing Viewer” for IgD-FITC plotted against all other 44 colors and two AF spectra

(A) Before SR adjustment. Populations displaying unmixing errors are marked with arrows: Red arrows mark cases of over-unmixing; green arrows mark cases of under-unmixing. Note, not all affected color pairs need to be individually adjusted. Often, correcting the pair with the greatest contribution to unmixing issues will correct other plots with minor issues, for example, FITC (green SR) and Real Blue 545 (blue SR) show some degree of spectral overlap that causes over-unmixing. Fixing this can correct many of the other unmixing errors; thus, we recommend prioritizing pairs with higher spectral overlap. The figure also shows unmixing errors caused by other color pairs, not involving IgD-FITC, that can be seen, for example, in the 4th plot from the left in the third row from the top—the downward-bending double negative population. Those errors must be fixed in the “Unmixing Viewer” for the affected fluorochrome (in this case, APC displays unmixing errors with NovaFluor Blue 660/120S). 1: Selected fluorochrome displayed on the x-axis. 2: Selected parent gate. 3: “SR Adjuster” toggle switch. 4: “Auto Adjust” axes buttons. 5: Displayed number of events. 6: SR panel.

(B) After SR adjustment.

Figure 5.

Biexponential axis adjustments

CD45RA vs. CCR7 staining for the non-MAIT CD4 gate before (A) and after (B) optimization of the negative area and maximum values for both axes. (B) After optimization, the CD45RA–CCR7– T effector memory cells form a rounded population in the lower left corner of the right-hand plot, and the available free space on the plot area is optimally used.

Figure 6.

Gating strategy of 45-color aging panel optimized for analysis with the SFA software, with identification of T cells, B cells, NK cells, several dendritic cell populations, monocytes, and hematopoietic stem cells

Gated populations are named above each plot. Gates that have been used for high-dimensional reduction using FIt-SNE, UMAP, or meta-cluster identification using BL-FlowSOM are shown in red. Gating included a second singlet gate and several “clean-up” gates to define pure cell populations for precise analysis using the Sony SFA Life Sciences Cloud Platform. All parameters except for the forward and side-scatter parameters were plotted with biexponential scales. The negative biexponential component was optimized using “Auto Adjust” to position negative populations in the lower left corner of each dot plot. An extended version of the gating strategy showing all 45-colors as well as FMO controls can be found in Figure S7.

Comparative cluster identification and statistical analysis

Timing: Variable (depending on the number of donors, events, and cell populations)

Timing: 30 min (for step 15)

Timing: 1 h (for step 16)

Timing: 30 min (may differ depending on the number of donors, events, and populations) (for step 17)

Timing: 30 min (for step 18)

Timing: 30 min (for step 19)

Timing: 15 min (for step 20)

Timing: 1 h (for step 21)

This step describes the quantification of age-dependent changes through unsupervised cluster analysis performed in the SFA software (see Figure 7 for the SFA workflow).

• 15.
  Data import into SFA.

  • a.
    Open the SFA software and import the Exdat file containing the fully stained samples via the “File” tab (go to “Experiment” and select “Import”).
  • b.
    Select the data file from the directory and place a check mark in the “I confirmed that the import files do not contain any personal information of subjects” checkbox and click “Import”.

    Note: After a few minutes, you will receive an email notifying you that the data are ready for analysis.

  • c.
    Select the imported experiment in the “Experiment Explorer” on the left side of the “View” tab and click the blue “Start Analysis” button at the bottom of the worksheet.

• 16.
  Adjust event numbers: To analyze differences within a specific cell subset, such as CD4 or CD8 T cells, an equal number of events should be selected to avoid confounding donor-dependent differences.

  Note: Performing non-catenated clustering and dimensionality reduction for all donors first can elucidate how the percentages of major leukocyte subsets, e.g., monocytes, T or B cells, differ by donor and age.

  • a.
    For the first donor, select the “non-MAIT CD8” gate and right-click or select “DataCleaning” from the top panel, choose “Batch Other Samples” and include all donors. Select “flowCut” to remove noise and outliers and click “OK”.
  • b.
    After cleaning is completed, a new gate is generated, e.g., “non-MAIT CD8-flowCut”. Use this gate for BL-FlowSOM and dimensionality reduction analyses.

    CRITICAL: For data cleaning functions, flowAI, PeacoQC, flowCut and flowClean are available. If data was acquired in multiple rounds using the “Append” function in the ID7000 Software, DataCleaning may remove those appended data accidentally. If many events were removed, confirm by plotting against the “TIME” parameter if appended data was removed. Also, flowClean is not suited for high parameter panels (>32 colors) and is thus not recommended for this panel.

  • c.
    For each donor, generate a timestamp gate within the target population: for example, for CD8 T cells, right-click on the “non-MAIT CD8-flowCut” gate from the SFA-optimized gating strategy and select “Create a Child Plot” > “Density Plot”. For convenience, first identify the donor with the lowest CD8 T cell number and start from this worksheet.
  • d.
    Click on the label of the x-axis and change it to “TIME (seconds)”.
  • e.
    You can leave the y-axis label as is or change it to “SSC-A”.
  • f.
    Change the scaling of the TIME axis until all events are visible either by clicking on and dragging the axis itself or by right-clicking on the axis label and selecting properties, where the maximal and minimal values as well as the biexponential component of the x- and y-axes can be adjusted.
  • g.
    Create a new rectangular gate: Make sure all events are contained for the y-axis and adjust the x-axis component of the gate until the desired number of events is within the gate. Adjust the gate and check the event number in the “Gates and Statistics” window.
  • h.
    In the “Gate and Statistics” window, right-click on the created gate and select “Copy”.
  • i.
    Go to the next worksheet and select the “non-MAIT CD8“ gate, right-click, and select “Paste”.
  • j.
    Right-click on the pasted gate and select “Create a Plot” > “Density Plot,” which creates a new plot with the same axis settings as in the previous worksheet.
  • k.
    Empirically adjust the width of the gate to include approximately the same number of events. An exact match is not required; for example, if the lowest-yield donor has approximately 51,000 CD8 T cells, a ∼2% variance (50,000–52,000 events) is acceptable.
  • l.
    Repeat the above steps for all donors.

• 17.Run concatenated BL-FlowSOM analyses.
  The BL-FlowSOM⁸ algorithm is a consistent and accelerated version of FlowSOM. FlowSOM⁹ is a clustering and visualization method that uses Self-Organizing Maps (SOMs) to analyze flow cytometry data.

  • a.
    Return to the worksheet of the first donor (e.g., Young 1, Y1).
  • b.
    Select the new event-adjusted gate for CD8 T cells; right-click and select “BL-FlowSOM” or choose “BL-FlowSOM” from the top panel (Figure 7B).
  • c.
    In the BL-FlowSOM window (Figure 7C), select parameters to be included for cluster identification.

    Note: For CD8 T cells, select parameters that are not already gated besides the lineage marker CD8; for example, Zombie NIR_A, CD45-APC-Fire810_A, CD3-BUV661_A, CD4-BUV615_A, CD14-SparkUV 387_A, CD161-PE-eF610 and CD19-SB600_A can be excluded.

  • d.
    At the top of the BL-FlowSOM window, choose “Concatenate Other Samples” and select all donors.
  • e.
    Keep the default parameters for “SOM Grid Size,” “Number of Meta Clusters,” and “Number of Iterations,” and click “OK” to start the learning process.
  • f.
    Depending on the number of events to be concatenated, the cluster identification process may take a few minutes. The finished results are displayed automatically in a minimum spanning tree (MST) plot (Figure 7D) on all worksheets.

    Note: BL-FlowSOM uses a 2-step hierarchical approach same as FlowSOM, clustering and meta-clustering respectively. “SOM Grid Size” determines clustering granularity and “Number of Meta Clusters” ideally corresponds to the number of cell populations. “Number of Iterations” determines how many times the SOM model updates during clustering. A larger grid size may capture finer details in the dataset, and longer iterations may improve clustering stability, but may increase processing time. However, the BL-FlowSOM algorithm is optimized for fast execution, enabling clustering to complete much faster than with FlowSOM. For detailed information on parameter optimization, please refer to publications by Quintelier et al.¹⁰ and Tao et al.¹¹ In this protocol, we obtained good results using the default grid size and setting the “Number of Meta Clusters” to “Auto”.

• 18.Perform concatenated dimensionality reduction (FIt-SNE/UMAP).
  Specific cell populations, as defined in the gating strategy (Figure 6), were visualized in 2-dimensional reduction Fast Fourier transform-accelerated Interpolation-based t-SNE (FIt-SNE)¹² and Uniform Manifold Approximation and Projection (UMAP)¹³ plots using default settings. UMAPs are more suitable to judge similarities or differences between populations based on island distances. FIt-SNE plots have larger, less compacted islands that can make judging frequency changes easier when comparing concatenated plots. Using CD8 T cells as an example, concatenated FIt-SNE plots were generated for all donors as follows.

  • a.
    Return to the worksheet of the first donor and select the event-adjusted TIME gate for CD8 T cells.
  • b.
    Select “FIt-SNE” from the top panel.
  • c.
    Select the same set of parameters as for the BL-FlowSOM analysis, that is, only parameters that are not already gated.
  • d.
    Select all other donors from the concatenation menu. Use the default parameters and click “OK” to start the dimensionality reduction. After a few minutes, a concatenated FIt-SNE density plot is automatically created on the worksheets of all donors.

    CRITICAL: If the same gating strategy is applied to all samples and the event-adjusted gates are set correctly, variability in 2D reduction plots within donors should be due to donor-specific differences. If a donor’s FIt-SNE or UMAP plot looks very different—for example, showing overly dense or sparse islands—check whether differing biexponential axis settings between worksheets are the cause (e.g., if markers were inadvertently not adjusted), and correct them by using the “Sync Scale and Gate” option in the ID7000 or SFA software”.

• 19.
  Backgating of BL-FlowSOM meta-clusters on dimensionality reduction or 2-by-2 dot plots.

  • a.
    Meta-clusters can be visualized on FIt-SNE and UMAP dot plots. Start by copying the concatenated 2D reduction plot and changing the plot type to “Dot plot” in the top panel.

    Note: The gate colors of the meta-clusters of the last BL-FlowSOM run are automatically displayed on the dot plot. Using the “Gate Manager” accessible from the top panel, you can select or deselect gates, including manual or meta-cluster gates, to be visualized on the dot plots. A total of 24 colors are available for population painting (16 default and 8 custom). Gate names can be displayed or hidden in the backgated dot plots using the “Show Paint Gate Name” checkbox.

  • b.
    Backgate meta-clusters (e.g., those significantly different between young and old donors) on 2-by-2 dot plots gated on the non-MAIT CD8 cells by selecting the desired meta clusters in the “Gate Manager”.
  • c.
    Create dot plots (as shown in Figures 8F and 8L) for markers expressed in the meta-clusters and sequentially select significantly regulated meta-clusters in the “Gate Manager” after deselecting all other manual and meta-cluster gates.

• 20.Create color axis plots.
  For FIt-SNE and UMAP plots, create color axis plots for markers of interest, prioritizing differentially expressed markers in meta-clusters between old and young donors.

  • a.
    Copy an existing FIt-SNE or UMAP dot or density plot and change the plot type to “Color Axis” in the top panel.
  • b.
    In the upper right corner of the color axis plot, select the marker of interest to be displayed. Optionally, display or hide the scale bar by right-clicking on the plot and selecting “Properties”.

    Note: Optimizing the dynamic range of the plot space by adjusting axis maxima and biexponential settings will ensure a clear color distribution and contrast in color axis plots and heatmaps. To improve visualization, i.e. color gradation, one axis of the color axis plot can be changed to the marker of interest and axis settings optimized. However, changing the axis settings after BL-FlowSOM or 2D reduction was performed may alter the analysis outcome for additional analyses. Therefore, we recommend optimizing the axis settings before performing BL-FlowSOM, FIt-SNE, or UMAP analysis.

• 21.Display heatmaps and perform statistical analysis.
  As part of the BL-FlowSOM analysis, color-coded heatmaps of the identified clusters and meta-clusters can be displayed (see Figures 8B and 8H).

  • a.
    Generate the heatmap for each donor from the BL-FlowSOM MST plot by clicking on the “eye” symbol and selecting “Show Heatmap” in the MST plot toolbar (Figures 7D and 7E).

    Note: The color indicates the percentage of positive cells for a given marker (clusters can be displayed by clicking on “>” next to the meta-cluster name). The corresponding median intensity values and exact percentages are displayed when hovering the cursor over a color square.

  • b.
    Export and save the CSV files from the upper right corner of the heatmap (Figure 7E-5) containing the number of events and median intensity values for all clusters and meta-clusters for each donor. These files are used to quantify differences between old and young donors.
  • c.
    Combine data for all clusters and meta-clusters, as well as event numbers from all donors, into a single file in Microsoft Excel.

    Note: It may be necessary to adjust rows before performing statistical analyses, since empty clusters will not appear in the exported file. Therefore, align all clusters by rows and enter a “0” as the event number for absent clusters.

  • d.
    Perform an unpaired t-test using the formula T.TEST (data range young donors, data range old donors, 2,2) to determine the likelihood of a significant (P < 0.05) difference between the young and old donor groups.
  • e.
    Calculate the standard deviation within the young and old donor groups using the “STDEV” function in Excel.

    Note: For esthetic reasons, data are plotted in GraphPad Prism rather than Excel. Prism also supports a wider variety of plots, such as bar plots that also display individual data points in addition to the average value and multi-segmental axis settings.

  • f.
    Copy data into a Prism worksheet and create a grouped bar plot from the “Insert” tab, selecting “New Graph of Existing Data”.
  • g.
    To improve the visibility of the event-number scale, create a y-axis break by double-clicking the y-axis and selecting “Two segments ---//---” in the “Gaps and Direction” drop-down menu.
  • h.
    Optionally, modify color and fill patterns for the young and old data groups in the “Appearance” tab of the “Format Graph” window (accessible by double-clicking on a data column in the graph).
  • i.
    Export the finished graph as an image file.

Figure 7.

Workflow for comparative cluster identification

(A) Workflow in the SFA software.

(B) Gate tools for data cleaning, BL-FlowSOM and dimensionality reduction.

(C) Example of BL-FlowSOM analysis window.

(D and E) Example of cluster identification results. D: Minimum spanning tree (MST) plot showing identified clusters (circles) grouped into meta-clusters. E: Heatmap showing marker expression in all BL-FlowSOM meta-clusters and clusters. 1: Selection of parameters for cluster identification. 2: Concatenation function. 3: Option to use existing SOMs to visualize concatenation results for all donors. 4: BL-FlowSOM parameters. 5: Export event numbers and marker expression intensities for BL-FlowSOM results. 6: Markers selected for cluster identification. 7: Events contained per cluster and meta-cluster. 8: Heatmap.

Figure 8.

Significant differences between old and young donors found for CD4 and CD8 T cells

CD4 (A–F) and CD8 T cell (G–L) subpopulation differences between young and old donors. Using concatenated BL-FlowSOM, FIt-SNE and UMAP, we analyzed 85,000 events per donor from the non-MAIT CD4 gate for CD4 T cells and 51,000 events per donor from the non-MAIT CD8 gate for CD8 T cells. MST plots (A and G), heatmaps (B and H), FIt-SNE and UMAP dot plots (D and J). FIt-SNE color axis plot s (E and K) and backgated 2-by-2 dot plots for meta-clusters that differ between old and young donors (F, L), as exemplarily shown for young donor 1 (Y1). Concatenated FIt-SNE and UMAP dot plots with mapped meta-clusters are shown for all young (Y1–Y4) and old donors (O1–O4) for CD4 (D) and CD8 (J) T cells. Grouped bar plots show CD4 meta-clusters (P < 0.1, C) and CD8 meta-clusters (P < 0.05, unpaired t test, n = 4 per group, I) with different numbers of events between young (black) and old (red) donors. Error bars represent standard deviation. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

Expected outcomes

Figure S6 shows how different biexponential axis parameters (negative area and axis maximum) affect the compaction and island formation in the dimensionality-reduction plots, as well as the number of identified meta-clusters, using the ”live” gate from a single donor as an example. In this protocol, “Auto Adjust” was used for most markers (see Table 1 for used axis settings) except for markers with either small positive or negative populations (CD44, CD45, CD34, CD141) where manual corrections were performed to adjust the negative population in the lower left corner (see Figure 5 for guidance).

The gating strategy was optimized to define pure cell populations, i.e., double negative for non-related lineage markers, for downstream SFA analyses (see Figure 6 for condensed; Figure S7 for the extended gating strategy showing the staining performance for each marker and gateable populations). Cell, singlet, live and CD45+ gates were set. CD3+ events were selected for T lymphocytes. To remove contaminants from other cell populations, a “clean-up” gate was set on CD14– CD19– events. αβ T (CD3+ γδTCR–) and γδ T cells (CD3+ γδTCR+) were distinguished in the pan-T cell gate. Downstream of the αβ T cell gate, MAIT cells were differentiated from non-MAIT cells through CD161 expression. Within the non-MAIT compartment, CD4+ and CD8+ T cells were used for concatenated SFA analysis.

Table 3 summarizes concatenated SFA analyses of non-MAIT CD4 and CD8 T cell populations from four younger (18–21 yrs) and four older donors (68–75 yrs), and includes meta-cluster phenotypes, biological interpretations, population frequencies per donor, and statistical analyses. BL-FlowSOM–derived meta-clusters (MC) mapped to biologically meaningful cell populations at varying differentiation stages.

Table 3.

Summary of CD4 and CD8 BL-FlowSOM analyses for 8 donors

Meta-cluster	Phenotype	Biological interpretation	p-value (t test Young vs. Old)	Frequency per donor relative to parental gate in %
				Y1 (20 yrs)	Y2 (18 yrs)	Y3 (19 yrs)	Y4 (21 yrs)	O1 (75 yrs)	O2 (69 yrs)	O3 (68 yrs)	O4 (72 yrs)
CD4-MC1	CCR7+ CD27+ CD28int CD45RA+ CD45RO– CD62L+ CD57– KLRG1–	Naïve CD4	0.03607	45.3	40.2	48.7	60.4	12.5	30.0	36.5	38.2
CD4-MC2	CD27– CD28+ CCR7– KLRG1+ or KLRG1– CD57– CD95+ CD45RA– CD45RO+	Non-senescent effector memory CD4	0.36236	13.4	10.9	27.1	14.7	32.0	16.8	32.9	11.0
CD4-MC3	CD27– CD28– CD45RA– CD45RO+ CD57+ KLRG1+ CD95lo	Senescent TEMRA CD4	0.57223	5.5	0.1	0.2	3.6	11.3	0.1	4.7	0.2
CD4-MC4	CCR7+ or CCR7– CD45RA– CD45RO+ KLRG1lo or KLRG1– CD57– CD95+ CD27+ CD28+	Non-senescent central memory/early effector memory CD4	0.36773	21.4	41.4	20.4	17.2	37.0	27.6	23.5	38.9
CD4-MC5	CD27+ CD28+ CD45RA– CD45RO+ CD62L+ CD95+ CCR7lo	Early transitional memory/partially differentiated central-memory–like CD4	0.47067	14.4	7.4	3.5	4.2	7.2	25.4	2.4	11.7
CD8-MC1	CCR7–, or CCR7lo, CD45RA– CD45RO+ KLRG1lo, or KLRG1– CD57– CD95+, mostly CD27+ CD28+	Non-senescent, mostly effector-memory CD8	0.01175	21.9	11.3	19.0	6.7	21.6	39.4	40.9	46.3
CD8-MC2	CD57+ KLRG1+ CCR7– CD45RA– CD45RO+ CD95lo CD27lo CD28– CD244+	Senescent effector memory CD8	0.21978	6.9	0.6	1.5	1.0	16.3	18.8	0.9	1.2
CD8-MC3	CCR7+ CD45RA+ CD45RO– KLRG1– CD57– CD95lo CD27hi CD28lo CD62Llo	Naïve CD8	0.00019	53.6	79.0	72.5	74.4	3.7	13.8	5.4	20.8
CD8-MC4	CCR7– CD45RA+ CD45RO-lo/– KLRG1+, CD57+ or CD57–, CD27– CD28– CD244+	Mostly senescent TEMRA CD8	0.01136	17.7	9.1	7.0	18.0	58.4	28.0	52.8	31.7

Notably, age-induced changes in the CD4 and CD8 T cell compartments were observed. SFA BL-FlowSOM and backgating on traditional 2-by-2 dot plots showed that age-induced changes were overall less pronounced in the CD4 T cell compartment than in the CD8 T cell compartment. MC1 containing naïve CD4 T cells (shown in red in Figures 8D and 8F) was reduced in older donors. Additional age-related changes were observed in two clusters within MC4 (increased in old) and one cluster within MC2 (up in old). However, the entire MC2 and MC4 were not statistically different, indicating trends rather than drastic differences between older and younger donors. Interestingly, senescent TEMRA CD4 T cells (MC3) did not differ between older and younger individuals.

This indicates that CD4 T cell senescence is not strictly dependent on donor age and that the CD4 T cell compartment remains remarkably stable throughout life, aside from the age-associated decline in naïve CD4 T cells. Consistent with the literature, SFA results show that CD4 T cell aging is mainly associated with a decrease in the naïve-to-memory-cell ratio.¹⁴ This shift results from thymic involution, which starts in early childhood and peaks in puberty, leading to a shift in T cell immunity from naïve-like (naïve and central memory T cells) to an “experienced” profile,¹⁵ dominated by effector memory and terminally differentiated, mostly senescent, TEMRA cells.

Within the CD8 compartment, stronger differences between young and old donors were observed (Figures 8G–8L). Three out of four MC showed statistically significant differences between young and old donors (Figure 8I), including MC1 (dark blue; Figures 8J and 8L), up in Old, MC3 (bright green), down in Old, and MC4 (ochre), up in Old. MC1 contained non-senescent effector memory CD8 T cells. The most significant difference (P < 0.001) was found for MC3, representing naïve CD8 T cells. MC4 comprised senescent TEMRA CD8 T cells. Interestingly, MC2 (magenta; Figure 8J) also included senescent cells (effector memory), but was not statistically significant (P = 0.22). In the concatenated FIt-SNE dot plots, this cluster is clearly visible in donors O1 and O2, but almost absent in O3 and O4, while noticeably visible in Y1.

These SFA-based findings indicate that T cell senescence is not strictly correlated with donor age but may instead be influenced by biological aging shaped by lifestyle and health status, or by premature expansion in younger individuals due to factors such as chronic viral infections.¹⁶

Limitations

The above 45-color panel was designed for and optimized on a 6-laser ID7000. Using a different instrument or changing the panel may require re-optimization. Cluster identification using the SFA software requires Exdat files compatible with Sony spectral analyzers or sorters. However, a similar algorithm, FlowSOM, as well as dimensionality reduction analyses, including tSNE and UMAP, are available on other platforms, including, for example, FlowJo (BD Biosciences) or OMIQ (Dotmatics), where similar data pre-processing steps, especially the optimization of biexponential axis settings, are required for optimal outputs. The described staining protocol was optimized on frozen PBMCs. Using fresh cells from peripheral blood or apheresis products may require re-optimization of staining or addition of steps, such as red blood cell lysis. When acquiring samples on different days, such as in longitudinal studies involving many donors, different manual SR adjustments may be required. The originally acquired SRs can be reused, but they should be regularly updated to avoid unmixing errors that may necessitate manual SR adjustments, particularly when using a new antibody lot or after enough time has passed. Best practice is to update the SRs shortly (within 1 week) before a new batch of samples is analyzed to minimize unmixing errors. For the multi-stains, always use same detector settings in Standardization mode to achieve comparable signal intensities for each fluorochrome (not considering donor-specific variations). For combined dimensionality reduction and meta-cluster analyses, use the same biexponential axis settings before concatenating samples. Worksheets containing biexponential axis settings can be copied between samples or the original experiment used as template to preserve instrument and worksheet settings. Limits on reproducibility arise from manually pipetting master mixes for the 3-step staining potentially inducing batch effects. Consider automating master mix production, for example, by using the Cytek Orion™ Reagent Cocktail Preparation System or the Revvity Fontus Cytometry Cocktail Prep Workstation. Consider re-titrating new antibody lots.

Troubleshooting

Problem 1

No distinct clusters or meta-clusters are found (BL-FlowSOM analysis step 17 & Statistical analysis step 21).

Potential solution

Consider acquiring more events, especially when analyzing a rarer population, and include more donors in the analysis. Concatenating ≥50,000 events per donor and cell population is recommended. Also, consider adding more relevant markers that are expressed on your population of interest to the panel. Alternatively, try increasing the BL-FlowSOM grid size or setting the “Number of Meta Clusters” to a fixed number (e.g., number of expected cell populations or number of distinct islands on 2D reduction plots).

Problem 2

FIt-SNE or UMAP plots show poor clustering or cluster definition (dimensionality reduction, step 18).

Potential solution

Check the biexponential axis settings for all markers. Ensure that positive and negative populations are clearly distinguishable. If adjusting the axes does not yield satisfactory results, re-optimize the staining or re-assign the affected markers to a brighter fluorochrome for improved separation.

Problem 3

Data display major unmixing issues (manual SR adjustment, step 13).

Potential solution

The introduced fluorochrome combinations should unmix well on an ID7000. If the instrument or panel composition is altered, evaluate unmixing performance first by making a single-stain mix of all colors of your panel on beads or cells and check all 2-by-2 dot plots, e.g., in the Unmixing Viewer. Selectively delete fluorochromes from pairs showing strong over- or under-unmixing, such as tilted negative populations or too many double-positive events that should not exist in a single stain mix.

Problem 4

Cannot decide between co-expression or under-unmixing in the multi-stain sample for 2 fluorochromes (manual SR adjustment, step 13).

Potential solution

Check in the “Unmixing Viewer” by selecting the plot of question whether spectral overlap exists between the 2 fluorochromes as under-unmixing often occurs for fluorochromes with higher spectral overlap. If you still cannot decide, prepare a single stain-mix of both conjugates on PBMCs and acquire at similar detector settings used for the multi-stain sample and apply the 45c matrix before SR adjustments. Double positive events will indicate under-unmixing, respectively, no double positive events indicate that the markers are co-expressed or non-specifically co-stained (e.g., monocyte-expressed marker and non-specific monocyte staining).

Problem 5

How to best modify the panel to include different sets of markers or transfer to another spectral platform.

Potential solution

The introduced 45 fluorochromes unmix well in this combination on an ID7000. Therefore, when using the same platform, we recommend reassigning fluorochromes to different markers rather than introducing new ones. For high-parameter panels, pairing antigen density with fluorochrome brightness is very important. An ID7000-specific brightness chart for commonly used fluorochromes is provided in Figure S8B. Note that brightness is platform-dependent due to differing laser and detector configurations and should be assessed separately on other spectral platforms, which may require re-optimizing antigen–fluorochrome pairings. When scaling down to fewer than 45 colors, we recommend removing fluorochromes evenly across the spectrum (see Figure 1A) to maximize the reduction in panel complexity (this can be calculated using the “Spectral Viewer” provided by FluoroFinder). When modifying the marker combinations, the heatmap in Figure S8A, which shows the expression of all 44 surface markers across major lymphocyte subsets, can be consulted. We recommend reassigning fluorochromes from populations of no particular interest to your markers of interest.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Motohiro Furuki (motohiro.furuki@sony.com).

Technical contact

Questions about the technical specifics of performing the protocol should be directed to the technical contact, Claudia J. Krause (claudia.krause@sony.com).

Materials availability

This protocol does not create new reagents.

Data and code availability

Flow cytometry data will be made available upon request to the lead contact.

Acknowledgments

We thank Attiya Abbas (Harvard Medical School) for scientific advice on designing an aging panel for the human immune system, as well as Ana Leda F. Longhini and Yoshiharu Hayashi for insightful comments on the manuscript.

Author contributions

C.J.K. designed and optimized the panel, acquired and analyzed the data, and wrote the manuscript. A.H. and K.F. oversaw panel design and data analysis. F.O. and K.N. oversaw high-dimensional cluster analysis. S.T., T.Y., A.H., F.O., K.N., K.F., and M.F. reviewed and edited the manuscript. S.T., T.Y., K.F., and M.F. supervised and managed the project.

Declaration of interests

All authors are employees of Sony Corporation.