Recent advances have revolutionized the oldest high-throughput single-cell analytical tool, flow cytometry. Fluorescent analyzers and sorters with up to seven lasers and the potential to detect up to 50 parameters are changing the way flow cytometry is used, but old school practices which are inadequate for new technologies remain alive. This chapter summarizes recent advances, explains the most salient new features and offers a step-by-step guide to develop and successfully execute high-dimensional fluorescent flow cytometry experiments.
Keywords: Antibody, High-dimensional flow cytometry, Immunofluorescence, Immunology, Single-cell analysis
The introduction of the universal cell theory [16] based on earlier microscopy observations revolutionized biological research. The need to improve upon low-throughput light microscopes led to the development of the first truly high-throughput cell analytical method, flow cytometry. Flow cytometers analyze single cells in suspension at high speed. The cells are illuminated by laser and the scattered or emitted fluorescent lights collected through photoelectric devices. The photons are converted to electric current, recorded, and analyzed with dedicated computer software. Flow cytometers accurately measure the intensity (count) and wavelength (color) of the photons, but in contrast to microscopes, spatial visual information is lost.
Although reportedly laser pointers can serve as light source [6], most instruments use high-quality lasers with ever-increasing power. Fluorescent emissions are detected with photomultiplier tubes (PMT) or avalanche photo diodes (APD) that convert photons into electric current in a voltage-dependent fashion. It is important to keep in mind that all cells exhibit autofluorescence due to excitation of intracellular metabolic compounds; therefore increasing detector sensitivity may enhance both specific and non-specific, background fluorescent signals.
A major challenge of multicolor cell analysis is that fluorescent emission produces multiple wavelength photons with both shorter and longer wavelength than the emission maximum. This limits nonoverlapping spectrum reagents to a handful within the typical 350–850 nm range of the spectrum. The use of multiple lasers and narrow band pass optical filters greatly enhance color discrimination and expand the number of distinguishable fluorochromes, but spectral overlap cannot be avoided and must be compensated for [1]. Modern instruments can calculate spillover matrices rapidly and precisely as long as appropriate single-stained compensation controls are available (see Subheading 3.1). An alternative approach of collecting fluorescent signals is based on real-time monitoring of the entire emitted spectra of fluorochromes [17]; see below Subheading 2.2).
Compensation-free flow cytometers have been developed based on measuring fluorescent lifetime of signal decay [8], but current instruments are limited in fluorochrome selection. The alternative, nonfluorescent high-dimensional flow cytometry technology of mass cytometry is not discussed here [2, 3]. Although mass cytometry pioneered 50 plus parameter flow analysis, it has remained a niche application of a single manufacturer’s technology (Fluidigm’s CyTOF) and has not become widely available for most researchers. We also do not discuss imaging flow cytometry (Amnis Corp.), which can capture fluorescent images of up to ten different colors on single cells moving in flow at high speed (for review see [12].
With thousands of fluorescent reagents available, advanced high-dimensional experiments using 40 colors have been performed successfully [14]. This chapter offers the end user a simple guide on how to plan and execute a high-dimensional fluorescent analysis experiment. First, we briefly describe the basic technology of traditional and spectral high-dimensional fluorescent analyzers. Second, we discuss the most common mistakes and misunderstandings of experiment setup and fluorescence spillover compensation. Once we clarify these mistakes, we explain how to design a high-dimensional fluorescent analysis panel. We provide a step-by-step guide on how to set up high-dimensional acquisition on a conventional flow analyzer as well as a general guide highlighting important differences in setup on a spectral flow cytometer. Further, more detailed discussion of the principles of high-dimensional flow analyses is available in recent reviews [4, 11].
Here, we define conventional high-dimensional flow analyzers as instruments that detect photons of different wavelength (color) with individual photodetectors associated with specific optical filters. Most analyzers also use a photodiode to record forward light scatter. Photodetectors are highly sensitive to report photon count; however, due to the stochastic nature of the photoelectric effect, they are subject to photon-counting statistical errors. This error can be significant at the low end of the detection limit and also contributes to spillover spreading error during fluorescent compensation (see Subheading 3.7). The number of colors (parameters) distinguished by a high-dimensional flow analyzer is determined by the number of photodetectors and associated band pass filters. A major limitation of expanding colors in a traditional flow cytometer is the trade-off between the optical width of a narrow band pass filter and the diminishing number of photons that the detector can count. Most band pass filters are 20–50 nm wide which means that the maximum ~500 nm total spectrum of fluorescent light emission (350 nm–850 nm) can be divided among only 10–12 different detectors for the shorter wavelength (UV, violet and blue) lasers and even fewer for the longer wavelength (green, red) lasers. Conventional flow cytometers are equipped with 5–7 spatially separated lasers with adjustable laser power which theoretically raises discrimination power to a total of 40–50 colors. Most of these analyzers allow the user to change the PMT voltage or APD gain settings to increase/decrease detector sensitivity. While most users do not alter laser power, detector sensitivity is commonly and frequently incorrectly manipulated (see Subheading 3.1). These instruments offer automatic setup and quality control including establishment of correct laser delay which is critical for analyzers with spatially separated lasers. Currently BD Biosciences’ FAC-Symphony A5, Beckman Coulter’s Cytoflex LX, and Bio-Rad’s ZE5 (originally designed by Propel Labs) models are capable of more than 20-color traditional analysis. Of these manufacturers, BD Biosciences also offer the FACSymphony S6 cell sorter for more than 20-color cell-sorting applications.
Spectral flow cytometry is based on real-time collection of photons across the entire spectrum for each fluorochrome instead of relying only on peak emission data [5]. Nonetheless, spectral analyzers still use PMT or APD photodetectors, usually arranged in a linear array of 10–32 detectors per lasers. Fluorescent light emitted from the target particle is split into multiple wavelength components through a light dispersion device and detected simultaneously with the PMT or diode arrays. The main challenge with multicolor spectral flow cytometry is spectral unmixing, i.e., the correct, quantitative identification of the fluorochromes that contribute to the overall spectrum of emitted light. Spectral analyzers can also measure and report cellular autofluorescence, an inherent and potentially useful characteristic of cells. While theoretically spectral flow cytometry offers nearly unlimited color discrimination, in practice it still depends on measurable differences in emission maxima which limits it to the discrimination of 40–50 colors. Improved spectral unmixing algorithms with faster computers may increase these numbers in the future. Currently Sony Biotechnology’s ID7000 and Cytek Biosciences’ Aurora models are capable of more than 20-color spectral analysis, and BD Biosciences are also beginning to offer spectral upgrade on their FACSymphony A5 models. Cytek’s Aurora CS and Thermo Fisher’s Bigfoot (originally designed by Propel Labs) sorters offer spectral cell sorting on more than 20-color panels.
Most particles, including all live and fixed cells exhibit autofluorescence. Cells contain metabolic compounds that emit a variety of fluorescence in response to different wavelength of excitations, especially in the shorter UV, violet and blue laser light range. Autofluorescence is a characteristic property of cells and can potentially obscure weak fluorescent signals. Therefore, many users consider autofluorescence an inconvenience and wish to minimize it by reducing detector sensitivity to a very low level to make unstained/negative control cells appear with minimum background signals. This view is incorrect even if autofluorescence does cause potential difficulties. The adequate approach is to design the experiment such that the desired specific fluorescence can be well distinguished from background noise. On traditional flow cytometers, it is possible that certain fluorescent channels exhibit too high background in which case the reagent must be changed, typically to a green or red laser-excited fluorochrome, where the autofluorescence is acceptable for the given application. Detector sensitivity should be adjusted such that autofluorescence is clearly distinguished from the background noise of the detector (see Subheading 4.3). As a rule of thumb, detector sensitivity should be increased rather than decreased while keeping the brightest fluorochrome within the linear range of detection (see Subheading 4.3).
The specificity of antibody staining depends on the unique interaction of the antibody with its cognate epitope. Therefore, unrelated antibodies should not bind to cells, regardless of their isotype. Fc receptor-mediated, nonspecific binding can result in false positive signal which can be uncovered with the use of isotype-matched control staining. However, a better approach is to use Fc receptor blockade with commercial reagents every time there is a possibility of Fc receptor interactions. Once Fc receptor interactions are excluded, any positive signal with isotype control antibodies becomes uninterpretable and irrelevant to the analysis. Instead, for proper gating of low abundance or poorly characterized antigens, fluorescence minus one (FMO) controls should be used [10]. FMO controls also offer the most accurate assessment of false positive signals derived from fluorescent spillover spreading error (see Subheading 3.7) and allow setting the most accurate gate positions for weak staining signals in multicolor flow cytometry panels.
Every fluorescent reagent should be used at an optimal and not higher concentration. This requires careful titration of the reagents ideally on the target cells of interest. Antibodies should be used at saturating, but not higher concentrations. Suboptimal concentration will reduce sensitivity, while supraoptimal concentration will increase nonspecific background signal which may also compromise sensitivity. The optimal concentration is determined by finding the best stain index value of the fluorescent reagent on its target cells (see Subheading 4.4). While titration of antibody staining is the most common practice, other exogenous fluorescent reagents, such as viability, cell proliferation, DNA-binding, and metabolic sensor dyes must also be titrated, since many of these reagents bind stoichiometrically to abundant cellular targets which results in extremely bright fluorescent signals.
Fluorescent light emission is a stochastic event which produces variable wavelength photons. Light emitted from many fluorochromes results in spectral spillover into detector(s) other than the primary detector dedicated to that reagent, causing false positive signals. Spectral compensation is the arithmetic subtraction of this spillover signal to ensure that only the dedicated detector records signals higher than background autofluorescence. Spillover compensation can be avoided by using reagents with completely separate fluorescent emission spectra, but this limits the number of reagents to less than four or five at most. True high-dimensional flow analyses will inevitably require fluorochromes with overlapping emission spectra and therefore must include fluorescent spillover compensation.
Both cells and commercially available compensation beads can be used for fluorescent spillover compensation. More importantly, cells and beads can be included in the same compensation setup. The objective with correct compensation is to determine the magnitude of spillover of fluorescent signal of a fluorochrome between its primary detector and any one of the other detectors. The main criterium is to identify and distinguish the true positive fluorescent signal from background autofluorescence. The absolute level of autofluorescence is not relevant, only the accurate measurement of specific fluorescence above background. Therefore, for each fluorochrome a mix of stained and unstained particles of the same type must be used. Whether cells or beads are used does not matter to spillover calculations. Relying only on cells carries several disadvantages: the experimental cells may be limiting; the positive fluorescent signal may be present only on a very small subset of the cells or present at very low levels, not clearly distinguishable from background. If the experimental cells express endogenous fluorescent signals (e.g., from fluorescent proteins), they cannot be used as compensation controls with different color reagents. In any of these situations, the use of compensation beads is strongly recommended. When using beads, it is important to keep in mind several factors. Only beads with the ability to bind the experimental fluorescent reagents should be used, not chemically colored beads with seemingly similar but invariably different spectral properties. Stained and unstained beads of similar numbers must be included (many commercial preparations offer premixed beads for this purpose). Compensation beads may bind reagents at much higher avidity than natural antigens do; therefore beads may appear brighter than the limit of linear detection range. In this case the reagent must be titrated separately on beads for compensation purposes and may need to be used at higher dilutions than in the experimental staining of cells. Finally, compensation beads usually bind only antibodies of certain species and isotypes; therefore some antibodies and most other fluorescent compounds may not be used with beads and will require the use of cells with distinct positive and negative populations.
Many fluorescent non-antibody reagents, particularly endogenously expressed fluorescent proteins, may not be readily available as single-color control. While light emission of such reagents may colloquially be considered of a certain color (green, yellow, red); they cannot be replaced with similar color fluorescent compounds for compensation purposes. As explained above, if the exact experimental cells are not available, a different cell type expressing the same fluorescent protein may be used as long as positive and negative populations of that cell type can be distinguished. A related and often overlooked mistake arises when using antibodies conjugated with a tandem fluorescent dye. For base fluorescent dyes (e.g., FITC, PE, APC), the spectral emission should be generally similar enough between different preparations, or even manufacturers, such that they can be used interchangeably. Tandem dyes are covalent compounds of two fluorescent reagents where excitation of the base donor dye induces Förster fluorescence resonance energy transfer (FRET) to the acceptor dye. The emitted light will be a mix from the acceptor dye as well as some from the originally excited donor dye [7]. The efficiency of FRET in tandem dye antibody conjugates is unique and characteristic of the individual preparation. Therefore, another antibody conjugate, even with the same specificity, will likely emit significantly different combination of donor and acceptor dye-specific light which will result in incorrect measurement of spectral spillover and compensation. The use of the exact same preparation of tandem dye antibody conjugate is necessary to avoid this type of compensation error.
Fluorescent spillover is calculated with a relatively straightforward algorithm [15]. Theoretically, spillover can be predicted from the spectral properties of fluorochromes which is readily available from public “spectra viewer” websites. Initial staining panel designs placed the emphasis on the percentage values of fluorescent spillover. However, the biggest impact of spectral overlap on practical analysis derives from its effect of reduced sensitivity in spillover channels due to spectral spreading error. Spectral spreading is the consequence of photon-counting errors and errors introduced by logarithmic amplification of the PMT signals. The magnitude of this spreading error depends on the fluorochrome and the particular detectors used in the flow cytometer and must be determined empirically. Spreading error is proportional to the intensity of the fluorescent signal in the primary detector, even though the compensation value is independent of signal intensity within the linear range of detector sensitivity. Spreading error must be visualized in a biexponential display in order to properly assess correct compensation. A common mistake is to assume that compensation is wrong because of the appearance of elevated signal in the spillover channel, when in fact with proper display the symmetrical “butterfly” pattern of data distribution confirms correct compensation. Spreading error is real and unavoidable in multicolor flow cytometry and can impair detection of weaker signals in the spillover channels. This loss of sensitivity is a more critical factor in multicolor flow analysis than the mere percentage value of spectral spillover [13]. Spillover spreading error matrices (SSM) are now routinely calculated and published by instrument manufacturers and are essential tools in designing optimal staining panels (see Subheading 4.2).
The journal Cytometry Part A in 2010 launched an ambitious project to collect peer-reviewed antibody staining protocols which would present extensively tested and optimized multicolor immunofluorescent panels (OMIP). To date 82 such panels have been published ( Table 1 ). Though somewhat misleading, as four of the panels are for mass cytometry analyses, the majority have been designed for conventional fluorescent flow cytometers and in one case for spectral flow cytometers. Most of the panels are developed for the study of human hematopoietic, typically peripheral blood cell types, but nonhuman primate and rodent panels have been also published, and a few protocols deal with tumor cells and other non-hematopoietic cell types. Table 1 lists all published panels by number and categorizes them according to their species and predominant cell type subject. While it is unlikely that any single panel will match perfectly most researchers’ specific needs, they are very helpful tools to start building individual panels, especially for high-dimensional immunofluorescent analyses.
Summary of 82 OMIPs published to date. OMIPs are classified according to species and cell types
| Cell type | Human | NHP | Mouse/rat | Other |
|---|---|---|---|---|
| Immune phenotype | 012 , 023 , 024 , 033 , 034 , 042 , 058 , 062 , 063 , 069* , 077 , 078 , | 032 , 054 , 076 | ||
| Peripheral T-cell differentiation | 001 , 002 , 009 , 013 , 017 , 022 , 030 , 060 , 067 , 071 , 080 | 016 , 075 | 048 , 079 | 065 (dog) |
| T cytokine and chemokine | 008 , 014 , 018 , 025 , 056 , | 005 , 052 | ||
| T regulatory | 004 , 006 , 015 , 053 | |||
| T checkpoint | 036 , 037 , 050 | 031 | ||
| Thymocyte | 073 | |||
| NK cell | 007 , 027 , 029 , 039 , 064 , 070 | 028 , 035 | ||
| gdT-cell, iNKT, MAIT | 019 , 020 , 021 , 046 , 082 | 057 | ||
| ILC | 055 , 066 | |||
| B cell | 003 , 043 , 047 , 051 , 068 , 074 | 026 | ||
| Myeloid | 038 , 044 | 041 , 061 | ||
| Hematopoiesis | 049 , | 059 | ||
| Leukemia Lymphoma | 010 , 072 , 081 | |||
| Other | 011 (endothelial cells), 040 (prostate cells), 045 (HNC) |
All OMIPS are designed on conventional flow cytometers, except those marked in red (mass cytometers) and with asterisk (spectral flow cytometers). Italicized OMIPs represent more than 20-color high-dimensional panels
ILC innate lymphoid cells, HNC head and neck cancer cells, NK natural killer cells, MAIT mucosal-associated invariant T cells
Designing a high-dimensional flow cytometry experiment is principally not different from designing any other scientific experiment which tests a defined hypothesis. Once it is established that high-dimensional flow cytometry is needed to test the hypothesis, one can follow a simple flow chart ( Fig. 1 ). It is important to define the target cell populations (e.g., lymphocytes, myeloid cells, tumor cells) and their phenotype (e.g., naïve, memory, regulatory) as identified by well-characterized markers. It is also important to establish a list of unknown or poorly defined markers (e.g., rare antigens, novel antibodies, or other fluorescent reagents). Finally, rank marker expression levels and patterns based on data from published reports or technical data sheets from manufacturers.
Flow chart of preparing high-dimensional fluorescent flow cytometry experiment. Individual steps shown in the chart are discussed in the text in Subheadings 4 and 5
