Panel Optimization for High‐Dimensional Immunophenotyping Assays Using Full‐Spectrum Flow Cytometry

Technological advancements in fluorescence flow cytometry and an ever‐expanding understanding of the complexity of the immune system have led to the development of large flow cytometry panels reaching up to 43 colors at the single‐cell level. However, as panel size and complexity increase, so too does the detail involved in designing and optimizing successful high‐quality panels fit for downstream high‐dimensional data analysis. In contrast to conventional flow cytometers, full‐spectrum flow cytometers measure the entire emission spectrum of each fluorophore across all lasers. This allows for fluorophores with very similar emission maxima but unique overall spectral fingerprints to be used in conjunction, enabling relatively straightforward design of larger panels. Although a protocol for best practices in full‐spectrum flow cytometry panel design has been published, there is still a knowledge gap in going from the theoretically designed panel to the necessary steps required for panel optimization. Here, we aim to guide users through the theory of optimizing a high‐dimensional full‐spectrum flow cytometry panel for immunophenotyping using comprehensive step‐by‐step protocols. These protocols can also be used to troubleshoot panels when issues arise. A practical application of this approach is exemplified with a 24‐color panel designed for identification of conventional T‐cell subsets in human peripheral blood. © 2021 Malaghan Institute of Medical Research, Cytek Biosciences. Current Protocols published by Wiley Periodicals LLC.


INTRODUCTION
Technological advancements in fluorescence flow cytometry and an ever-expanding understanding of the complexity of the immune system have led to the development of large panels reaching 40 fluorophores in the Optimized Multicolor Immunofluorescence Panel (OMIP; Park, Lannigan, & Jaimes, 2020) and 43 colors in a technical note (Sahir, Mateo, Steinhoff, & Siveen, 2020). In contrast to conventional flow cytometry, which primarily measures the peak emission of each fluorophore in a target detector, full-spectrum flow cytometry uses a larger number of detectors with narrow band-pass filters. This allows the entire emission spectrum for every fluorophore to be captured across all laser lines, creating a detailed signature of each fluorophore. This makes it possible to distinguish fluorophores with very similar emission maxima but unique overall spectral fingerprints, increasing the flexibility in fluorophore selection. This feature, coupled with an instrument designed to maximize the detection of emitted light and highly efficient avalanche photodiodes, provides improved detection efficiencies that translate to better detection limits and higher signal resolution (Feher et al., 2016). While these features, unique to full-spectrum flow cytometry, provide the high-quality signals and low noise needed for successful high-dimensional panels, many of the same panel design considerations from conventional flow cytometry still apply. Common characteristics, together with those specific to full-spectrum flow cytometry, have been previously described (Ferrer-Font, Pellefigues, et al., 2020;Park et al., 2020). Now that hardware limitations hindering the use of highly overlapping dyes have been overcome, the main limitations for successful large panel design are the spillover-spreading error inherent to the use of fluorescence and the number of fluorophores available with unique spectral signatures. Many companies have recently begun developing spectrally distinct fluorophores, thus advancing the number of markers available to be analyzed in a single experiment.
Increasing the number of markers in a panel consequently increases the probability of compromising the resolution of the markers and populations of interest. The theoretical approach to panel design aims to avoid issues would that prevent resolution of every marker in the panel; however, in practice it is challenging to perfectly predict the impact of co-expression, variations in marker expression levels, and the performance of each specific reagent. Furthermore, it is difficult to anticipate the accuracy of reference controls for optimal unmixing results. For example, certain fluorophores emit slightly different spectra when bound to compensation beads or in the presence of different buffers. autofluorescent tissues such as brain, lung, skin, intestine, and tumor (Schmutz, Valente, Cumano, & Novault, 2016). It is therefore highly recommended to characterize the AF spectrum of an unstained sample from the tissue or cell type of interest prior to panel design. This will provide useful information during fluorophore selection, ensuring that fluorophores are not allocated to areas of the spectrum where AF dominates.

STRATEGIC PLANNING
Once a high-dimensional full-spectrum flow cytometry panel has been optimally designed (Ferrer-Font, Pellefigues, et al., 2020), the step-by-step protocols presented here provide a series of practical steps for full and successful optimization. As outlined in Figure 1, the main procedures include evaluation of spectral reference controls (SRCs; see Basic Protocol 1), evaluation of unmixing of the fully stained (FS) sample (see Basic Protocol 2), evaluation of marker resolution (see Basic Protocol 3), and assessment of data quality (see Basic Protocol 4). Before any of these methods are performed, it is essential to titrate the antibodies used as well as any viability dye used. This is described in Support Protocol 1. An additional protocol describes changes that can be made to the instrument settings (see Support Protocol 2). Methods for evaluating and mitigating autofluorescence are included in Basic Protocol 3 and Support Protocol 3. Together, these protocols can also be used when troubleshooting a panel to identify sources of problems and provide insights into fixing them.
A 24-color panel optimization for identification of conventional T-cell subsets in human peripheral blood is provided to illustrate these procedures. The protocols were developed using the five-laser (5L) Cytek Aurora (Cytek Biosciences), but should be adaptable to any spectral flow cytometer. The protocols were designed for new full-spectrum flow cytometry users. Once familiarity and experience with specific tissue types is achieved, it is possible to modify the steps and order of the protocols to reduce the overall time

of 36
Current Protocols spent evaluating the panel. If a modified approach is taken, it is recommended that all the overall goals of the protocols (as outlined in Fig. 1) still be carried out.

PREPARATION AND EVALUATION OF OPTIMAL SPECTRAL REFERENCE CONTROLS
This protocol is divided in two sections. The first is for preparation of SRCs and ensures generation of the high-quality controls required for accurate unmixing. The steps describe the staining of polystyrene compensation beads and cryopreserved PBMCs with surface-labeling antibodies. They can be adapted for other tissues or staining procedures such as intracellular staining. It is important to mention that the preparation procedure will also guide users to prepare a fully stained (FS) sample and fluorescence minus one (FMO) controls as the protocols are very similar; these will be used later for evaluating the unmixing of the FS sample (see Basic Protocol 2). If preferred, FS and FMO samples can be stained separately, but treatment of samples should be kept identical. Importantly, to successfully complete this protocol, SS and FS cells should also be treated the same (antibody concentration, incubation time/temperature, fixed/unfixed, etc.).
The second section of the protocol is for evaluation. This aims first to check the quality of the acquired SRCs and then to evaluate whether there are any spectral mismatches between beads and cells for each fluorophore, which is accomplished by assessing how well the beads unmix the cells using N × N plots for all markers. Unmixing accuracy must be assessed on the actual sample (cells) to be used in the assay. In some cases beads will be acceptable as SRCs and in other cases it may be necessary to use cells. When this protocol is completed, the optimum SRCs (cells or beads) will be determined for future unmixing. If desired, well-characterized and high-quality controls can be stored for future use.  As the thawing procedure can be critical, we recommend user-specific optimizations to ensure high sample viability (>80%) (Disis, dela Rosa, Goodell, & Ling-Yu, 2006).

CAUTION:
Steps 1-2 should be performed in a biosafety hood.
2. Centrifuge 5 min at 500 × g, room temperature, and carefully flick off the supernatant.
3. Resuspend cells in PBS with 2% FBS and count viable cells on a hemocytometer.
4. Centrifuge as before and carefully flick off the supernatant.
Ferrer-Font et al.

of 36
Current Protocols 5. Resuspend cells in PBS with 2% FBS to a concentration of 5 × 10 6 viable cells/ml. 6. Distribute 100 μl suspension per well of a 96-well U-bottom plate, allocating one well per fluorophore in the panel and one well for an unstained sample. 12. Centrifuge as above and flick off the supernatant.
14. Prepare SS, FS, and FMO control antibody mixes by diluting stocks in FACS staining buffer according to titration results (see Support Protocol 1). Be sure to pipette from the top of the liquid to avoid centrifuged aggregates.
Prepare enough volume of each antibody mix that two samples can be stained from the same mix. This will reduce differences due to antibody preparation when comparing bead versus cell SRCs. Allow for pipetting errors by making excess antibody mix (e.g., n + 1). Additionally, when more than one Brilliant polymer dye is used at the same time (FS and FMOs), Brilliant Stain Buffer Plus (or equivalent) should be added per manufacturer's instructions to decrease interaction between the dyes. Do not add Brilliant Stain Buffer Plus to compensation beads, as it is known to alter the spectral profile of some beads (Ferrer-Font, Pellefigues, et al., 2020).

of 36
Current Protocols 15. Stain all cells and beads with 100 μl of the appropriate antibody mix and incubate in the dark using the incubation time and temperature that will be used in the final assay.
16. Centrifuge as above and flick off the supernatant.
At this stage, cells may be fixed (e.g., 1% paraformaldehyde) if required for biosafety or sample longevity. If cells are fixed, beads must also be fixed to ensure equal treatment of fluorophores and to mimic any change to the spectral signature caused by fixation. The fixative must be removed by washing twice with FACS staining buffer before acquisition.
19. Acquire on a Cytek Aurora, taking care to meet the following acquisition criteria: a.
Cytek assay settings are used as a starting point for instrument setup. b. The scatter profiles of cells and beads are on scale and the FSC area scaling factor (ASF) is optimized (see example in Fig. 2B). c. All fluorescence signals are on scale (<4 × 10 6 ). This can be assessed in the full-spectrum plot or in individual plots for every detector. d. All tubes are recorded with the same fluorescence gain settings for each detector. e. Sufficient events are recorded to find a clear positive signal.
A minimum of 300 events is needed for each positive and negative population. A good starting point is 5000 total events for beads and 30,000-50,000 total events for cell controls, although it may be necessary to record >50,000 cells to get at least 300 positive events of similar fluorescence intensity for rare markers.
Evaluate results (see Video 1) 20. In the SpectroFlo software, check the raw reference control data and verify that the acquisition criteria in the previous steps have been met: a. The scatter profiles of the cells and beads are on-scale, clean, and easily gated, and the FSC ASF has been optimized ( Fig. 2B). b. Cytek assay settings or a close alteration (also see Critical Parameters discussion of Complex Samples and see Support Protocol 2) have been used, and all tubes have been recorded with the same settings.
Video 1 Evaluation of optimal spectral reference controls (Basic Protocol 1).
Ferrer-Font et al.

of 36
Current Protocols 21. Ensure that the unstained sample has no contamination from other fluorophores. 31. Evaluate the unmixing accuracy of the marker on the x axis by visually inspecting whether the positive and negative populations are well aligned horizontally along the x axis. If unmixing errors are seen in the N × N matrix plots (Fig. 2D, top), make a note of which control was being viewed.
32. Repeat steps 29-31 until the unmixing of all fluorophores has been evaluated.
33. Note which fluorophores require cells to be used for the SRC (i.e., those with identified unmixing errors in step 31). All other controls can remain as beads.
Ferrer-Font et al.

Current Protocols
The most likely explanation for bead SRCs not unmixing the cell SRCs correctly is that the signature of the fluorophore on the beads did not match that of the cells. This phenomenon is known to happen but cannot be easily predicted. Another possible cause is that the bead SRC is dimmer than the cell SRC (Fig. 2E,

F). If all recommendations for treatment of SS cells have been followed (i.e., identical treatment to FS cells, use of the same number of cells), then the cell SRC should have identical brightness to the FS cells and thus be an appropriate control.
34. Go through the Unmixing Wizard a second time using the controls identified in step 33 as the optimal controls. Unmix again using live unmixing.
35. Follow steps 23-26 for the modified controls to ensure the best unmixing outcome is achieved.
When setting gates using cells as SRCs, the scatter gate should be placed only on the cells expressing the marker of interest for each control. In cases where there is no negative staining in this population, a universal negative may be used as a surrogate. To ensure the observed negative signal is due to AF rather than contamination, compare it to the matching unstained tube.
36. Repeat steps 29-32 to evaluate the new unmixing for the fluorophores that did not unmix optimally before (Fig. 2D, bottom) and note which fluorophores require cells or beads to be used for the SRC (Fig. 2G). Tables 2-5 for more potential issues and how to fix them).

ANTIBODY TITRATION
Antibody titration is the crucial first step in developing high-dimensional flow cytometry panels. Using the incorrect antibody concentration can increase spread, decrease resolution, increase aggregation of reagents, and give rise to nonspecific binding (Stewart & Stewart, 1997), all of which result in poor panel performance and/or inaccurate results. Ideally, titrations should be carried out in the tissue that will be used in the assay. This is not always feasible, however, if the tissue in question is rare or difficult to work with, or the cells of interest are found at low frequencies within the tissue. In such cases, it is suggested to first add a lineage marker to the mix to help identify the cells that express the rarer marker or to use a surrogate tissue in which the marker is more abundant and simpler to process. Results from this type of titration should always be validated using the tissue of interest.
It is important that readout functional markers are titrated under the maximally activated conditions that will be used in the assay. Staining conditions should also be identical between titration and experiment to prevent spectral pattern mismatches and poor unmixing results.
This protocol describes the preparation of cryopreserved peripheral blood mononuclear cells (PBMCs) for titration of extracellular antibodies. It can be easily adapted and used as a general guideline for antibody titration using other tissues or staining procedures. 2. Centrifuge 5 min at 500 × g, room temperature, and carefully flick off the supernatant.

Resuspend cells in PBS with 2% FBS and count on a hemocytometer.
4. Centrifuge as before and carefully flick off the supernatant.

Resuspend cells in PBS with 2%
FBS to a concentration of 5 × 10 6 cells/ml. 6. Distribute 100 μl suspension per well of a 96-well U-bottom plate, allocating six wells to each antibody being titrated. Include additional wells for unstained and live/dead controls (see example plate plan in Fig. 3A).
The number of titer points is based on the range of concentrations being tested. The number of cells required will differ by tissue type; for PBMCs, ∼5 × 10 5 cells is sufficient.
8. Stain titration samples and live/dead controls with a viability dye that will not cause significant spillover into the fluorophore being titrated.
Viability dyes should also be titrated for use in high-dimensional spectral cytometry panels. When titrating a viability dye, skip steps 8-12 for that sample only (adding PBS instead if other samples are being processed simultaneously). If the viability dye has not yet been titrated, follow manufacturer guidelines for concentration and incubation procedures.
9. Centrifuge as above and flick off the supernatant.
10. Block Fc receptors by applying 100 μl of a 1:40 dilution of Human TruStain FcX and incubating 10 min at 4°C.
11. Add one drop of vortexed compensation beads per well for the appropriate single stain (SS) controls.
12. Centrifuge as above and flick off the supernatant.
13. Remove aggregates from antibody stocks by centrifuging the vials 5 min at 16,000-18,000 × g, 4°C, or by filtering using a pore size of 0.65 μm or smaller. 14. Create a dilution series for each antibody being titrated by diluting stock solutions in FACS staining buffer. Be sure to pipette from the top of the liquid to avoid the spun-down aggregates. 22. Generate standard deviation (SD) values for the negative population for each concentration.

Calculate the stain index (SI) using the equation SI
The stain index is a very useful metric of resolution, with a higher value indicating greater resolution. If a lower concentration of antibody produces a reduced positive population MdnFI compared to a higher concentration with no change to the negative population, its stain index is going to decrease, as the separation between the positive and negative populations will be reduced (see example in Fig. 3B). If a higher concentration of antibody gives rise to a similar positive population MdnFI compared to a lower concentration, but has a larger SD of the negative population, its stain index is going to be decreased, as spreading of the negative population reduces the resolution from the positive (see example in Fig. 3C).
24. Select the concentration that gives rise to the best stain index without giving rise to a positive shift of the negative population. Fig. 3B-D).

CHANGING INSTRUMENT SETTINGS
In some cases an adjustment of fluorescent gain settings away from the optimized settings cannot be avoided, for instance, if a bright reporter protein (e.g., eGFP) is off-scale. Such adjustments must be carried out carefully to minimize impact on all spectral signatures within the panel. Adjustments can be made detector by detector or to a whole detector array at one time, where all detectors of a given laser line are changed simultaneously. Reduction of the whole array is the recommended approach, as manual alteration of single detectors is prone to distortion of spectral signatures, where the ratio of brightness from detector to detector is not maintained. Regardless, reducing whole detector arrays can lead to peak emissions of spectral signatures being shifted to an incorrect detector (potentially on a different laser line). Therefore, all changes must be checked to ensure Ferrer-Font et al.

of 36
Current Protocols that signatures remain as expected based on Cytek's Full Spectrum Viewer, published fluorophore guidelines, or historical data.
1. Observe which fluorescent channel(s) has an off-scale signal.
2. Use trial and error to determine the gain setting that is appropriate to bring the emission peak on-scale (i.e., <4 × 10 6 on the Aurora).
Remember that gain and fluorescence have a linear relationship, so that a 50% reduction in gain will give rise to a 50% reduction in fluorescence signal.
3. Reduce secondary detector arrays with off-scale signals by the same percentage as the reduction made in step 2.

Use
Cytek's Full Spectrum Viewer, published fluorophore guidelines, or historical data to check that the altered spectral signature retains the correct overall pattern and that the primary emission peak has not been reduced below any secondary emission peaks.
If the overall pattern is correct, no further steps are required. Record all samples at these adjusted settings. If the overall pattern is not correct, continue to step 5.
5. If adjustments to other laser lines are required, continue reducing the respective detector arrays by the same percentage as in step 2.
6. Repeat steps 4 and 5 until the spectral signature appears as expected or all arrays have been reduced by the same percentage.
7. Record all samples at the adjusted settings.

UNMIXING EVALUATION OF FULLY STAINED SAMPLE
Before any analysis can be undertaken, the FS sample must be checked to ensure that clean data can be obtained through removal of artefacts such as doublets, dead cells, and aggregates. It is also necessary to verify that there is positive staining for all markers in the panel, taking into consideration the biology of each marker. Once this has been asserted, it must be determined whether the SRCs selected (either beads or cells) successfully unmix the SS cells in Basic Protocol 1 can also be used to successfully unmix the FS sample (Video 2).
Video 2 Unmixing evaluation of fully stained sample (Basic Protocol 2). Ferrer-Font et al.

Materials
Cytek SpectroFlo software FCS files generated in Basic Protocol 1 N × N worksheet template 1. In the SpectroFlo software, work through the Unmixing Wizard and select the SRCs chosen from Basic Protocol 1.
2. In the unmixed workspace, gate on time to remove any events collected during an unstable flow period by plotting time vs. scatter (e.g., FSC-A or FSC-H; Fig. 4A).
3. Gate out doublets by sequentially plotting FSC-H vs. FSC-A and SSC-H vs. SSC-A, and include only the events found in the diagonal population (Fig. 4A).
If working with whole blood or PBMC samples where incomplete RBC lysis was achieved, unlysed erythrocytes need to be excluded. One method is the blue/violet SSC gating method described by Petriz, Bradford, & Ward (2018).
4. Exclude aggregates through inspection of N × N plots for incorrectly unmixed super-bright events and gate these out of all further analyses (Fig. 4B).
5. Gate out dead cells by including only viability dye negative events (Fig. 4A).
Viability dyes also need to be titrated (see Support Protocol 1).
6. Gate on cells of interest using FSC-A vs. SSC-A, including only the events required for analysis (Fig. 4A).

If working with a sample that contains multiple cell types (such as PBMCs), include only the cells that will be involved in downstream analysis. For example, when analyzing a panel that contains only lymphocyte markers, gate out monocytes (validating that T cells are not excluded).
7. Create as many pseudocolor or dot plots as there are fluorophores in the panel. Set the y axis to SSC-A and each x axis to a different fluorophore (Fig. 4C).
8. Verify that a positive signal can be found for all markers (Fig. 4C).
For rare or dim markers, FMOs can assist in determining the gating boundaries (Fig.  4D). If no clear signal is observed, take note of that reagent and check the titer and staining protocol. For additional troubleshooting, see Table 3.
9. Open the N × N worksheet template created previously and evaluate how well unmixing was performed for each marker. Inspect the FS sample by assessing whether super-negative events are present (example in Fig. 4E), which can be an indication of unmixing issues.
10. Select all plots making up the matrix. In the Plot Properties window, change the x axis option to the next fluorophore down the list so that all plots change simultaneously.
12. If there are no major unmixing errors (i.e., the positive and negative populations of each marker in the N × N plots are well aligned), continue to evaluation of marker resolution (see Basic Protocol 3).
13. If there are unmixing issues, even once the unmixing has been performed correctly and optimally, it is possible to make small adjustments (<3%) to the compensation matrix (found in the Tube Properties window). Adjustments to the spillover matrix should be fully justified and the integrity of the data must not be impacted (made with clear knowledge of the biology, such as expression and pattern characteristics). Under-unmixing and over-unmixing issues can be manually adjusted (Ashhurst, Smith, & King, 2017

EVALUATION OF MARKER RESOLUTION
Once the best possible unmixing has been achieved using Basic Protocols 1 and 2, the resolution of each marker in the FS sample must be compared to the SS cell controls.
Assessing the spread of the negative population and/or shifts in the positive signal will provide an indication of whether there is any loss of resolution of markers in the panel when fully stained. This assessment is best achieved by overlaying each marker in the SS cell sample onto the FS sample. If any reduction in resolution is seen, the impact of this can be further investigated to identify if it will impact the ability to identify populations of interest using an established gating strategy. Ideally, the same number of cells from the same tissue type should be stained and acquired for all samples. In practice, however, this is not often feasible, and downsampling can be used to achieve identical cell numbers across samples during analysis. In theory, the only difference between the SS and FS cells should be the number of antibodies in the tube (Video 3). Additionally, AF extraction should be tested to determine if marker resolution can be improved (Video 4).

Video 3 Evaluation of marker resolution (Basic Protocol 3).
Video 4 Deciding on an approach for mitigating autofluorescence (Basic Protocol 3) and Support Protocol 3. Ferrer-Font et al.

of 36
Current Protocols

Materials
Data analysis software (e.g., FlowJo or FCS Express) Cytek SpectroFlo software FCS files generated in Basic Protocol 2 N × N worksheet template

Evaluate marker resolution (Video 3)
1. Open the unmixed FS and SS cell samples in a data analysis software package for analyzing FCS files (e.g., FlowJo or FCS Express).
3. Create a histogram for each fluorophore in the panel (Fig. 5A). If the positive population is rare, use a dot plot for visualization purposes instead of a histogram (Fig. 5B).
4. Overlay the same number of FS and SS cell sample events onto a histogram or dot plot (Fig. 5A,B).

To ensure that each sample contains equal events, downsample the cells of interest population from all samples to the same number of events based on the sample with the fewest cells of interest. In FlowJo, for example, select the population of interest, then go to the Workspace tab and select Plugins and then DownSample. Specify the number of events to be downsampled such that all samples can meet the criteria.
5. Determine whether any spreading of the negative population has occurred in the FS sample (Fig. 5A, right).
6. Determine whether such spreading has impacted the resolution of positive and negative populations, either visually or by calculating stain index (SI; see Support Protocol 1, step 22).
7. Create a spillover spreading matrix (SSM) to assess which fluorophores are likely to be introducing spread to the fluorophore of interest (in Fig. 6, find the SSM for the panel used to illustrate the protocol).

In FlowJo, for example, using the SS cells SRCs, navigate through the Compensation
Wizard defining positive and negative populations for each fluorophore. The SSM will be automatically generated alongside the compensation matrix and can be displayed or exported from the SSM button.

If the spread of the negative population results in significant loss of resolution such that separation of positive and negative populations becomes difficult, and the marker is
Ferrer-Font et al.

Figure 7
Determination of AF signature type. AF signatures of (A) unstained PBMCs, (B) unstained JAWS cells (immortalized immature dendritic cell line), and (C) unstained skin cells. The latter shows a highly heterogeneous signature that can be split into three unique spectral signatures.
co-expressed in the same cell type as the marker causing the spread, the panel design should be reconsidered and a different fluorophore may be needed to improve resolution.
8. Determine whether any loss in fluorescence intensity of given marker(s) has occurred in the FS when compared to the SS (Fig. 5C). Jalbert, Shikuma, Ndhlovu, & Barbour (2013) and Park et al. (2020), where sequential staining of chemokine receptors was required to achieve positive signal in the FS sample comparable to that of the SS cell controls. Sequential staining can also be applied to non-chemokine receptors (Fig. 5C).

Choose approach for mitigating AF (Video 4)
Two scenarios should be considered regarding AF: homogeneous AF, where the whole unstained sample has a single low spectral signature (Fig. 7A) or single bright spectral signature (Fig. 7B), or heterogeneous AF, where there are multiple AF signatures (Fig.  7C). For PBMCs, where the unstained sample has a relatively low homogeneous AF signature, the Simple AF Protocol can be followed to assess whether AF extraction improves marker resolution. AF extraction is particularly helpful in resolving low-expressed markers by lowering the background of the negative population with minimal effect on the positive signal. If samples have heterogeneous or homogeneous but very bright AF, it is likely that AF extraction must be used to obtain unmixing that looks correct in Basic Protocols 1 and 2. In this case, the decision to use AF extraction will be based on unmixing quality instead of improvement of marker resolution. The AF protocol for 11. If there are other cells of interest in the sample that have a brighter spectral signature, try to unmix with the scatter gate on this population in step 10b.
12. Determine the best unmixing outcome comparing the N × N plots by evaluating changes in resolution for the dyes that overlap with the range of emission of AF (e.g., BUV496, BV510, Pac Orange, BV570, FITC, PE).

MANAGING HETEREOGENEOUS AUTOFLUORESCENCE (VIDEO 5)
The following protocol is divided into three sections: Discover, Distinguish and Designate.

Discover
1. Observe N × N plot permutations of raw channel data.
a. On the Aurora, draw as many pseudocolor or dot plots as channels. b. Leave the x axis unchanged. c. Set each y axis to the different channels, so each plot shows a different raw combination.
An example of a SpectroFlo N × N raw template is shown in Figure 8.
2. Find a combination that separates the greatest number of populations from one another.

of 36
Current Protocols   4. Place gates around all unique populations and observe the scatter properties and spectral signatures on adjacent spectral plots and select only the populations that have unique spectral characteristics.
5. Find the peak fluorescent channel for each unique population and then display said population on a histogram using this channel.
6. Gate the brightest 300-500 cells only, in order to derive the brightest AF SS control.
7. Right-click on the gate and export events as a new FCS file.
8. Repeat for all unique populations, producing an FCS file for each.
9. Export an AF dim population (dimmest AF spectral signature found in the sample) to be used as the negative control for the previously exported populations.
Designate 10. Designate these new AF signatures as fluorescence tags in the SpectroFlo Library.
a. Create a new AF group.
This will facilitate filtering and exporting these fluorophores for offline analysis, as well as identification when designing experiments.
b. Assign the AF tag a name, choose the excitation laser, and assign an emission wavelength (use the optical configuration in Figure 9 to determine wavelength based on peak emission channel).
11. Open the experiment of interest and add the new AF tag(s) as if it were a fluorophore in the experiment.
12. In the reference controls tab, add an additional negative for the AF tags that will be the AF dim.
Ferrer-Font et al.

of 36
Current Protocols 13. Import the populations exported during the Distinguish steps as SRCs for the corresponding tags, and import the AF dim population as an additional negative for the AF SRCs.
14. Use the Unmixing Wizard QC tools to assess AF signature similarity.
In our experience, a similarity of 0.96 or less indicates that the AF signatures are generally considered to be unique.

Unmix the experiment and evaluate the unmixed N × N matrix.
If unmixing errors are still present, refer to the troubleshooting guide in Table 3.

ASSESS DATA QUALITY USING EXPERT GATING AND DIMENSIONALITY REDUCTION ALGORITHMS
Informed gating is a useful way to check panel quality. It is necessary to check that all populations of interest can be identified and to investigate how well they can be resolved (Fig. 10A, Video 6). The biology of the sample should be considered to ensure that populations seen in the sample are as expected and that markers are expressed on all the expected cell types. This protocol indicates whether the panel is ready for use on experimental samples. Additionally, dimensionality reduction algorithms can be useful tools for investigating panel quality, as they simplify the data for visualization while also exposing artefacts that may be missed through expert gating. Using t-distributed stochastic neighbor embedding (t-SNE; van der Maaten & Hinton, 2008; Fig. 10B), markers that are usually co-expressed should be checked to see they are found in similar regions of the t-SNE plot (Brummelman et al., 2019).

Materials
Data analysis software (e.g., FlowJo, FCS Express, or any other software that allows gating of data) FCS files generated in Basic Protocol 3 Perform expert gating 1. Open the unmixed FS sample in data analysis software used for analyzing FCS files.
2. Follow steps 2-6 of Basic Protocol 2 to gate on live single cells of interest.
3. Gate the rest of the markers based on panel design, prior knowledge, and published literature.
4. Check the gating strategy for unexpected marker combinations or cell populations (Fig. 10A).
For example, when analyzing PBMCs, a CD3 + CD19 + double-positive population is not expected. If one is found, the source of this staining must be investigated and corrected before the panel is ready for use.

5.
Evaluate whether all populations of interest have clear positive signals that can be easily resolved from the negative.
6. Ensure readout markers (used for experimental readout) can be quantified in each of the cell types of interest.

Apply dimensionality reduction algorithm (t-SNE)
7. Using the cleaned data from step 2, run a t-SNE at the default settings (iterations: 1000; perplexity: 30) utilizing all fluorescence parameters. Exclude non-informative parameters such as those used in the data cleaning steps (e.g., viability and phenotypic markers for identifying cells of interest). 8. Assess marker co-localization one by one in viSNE plots by coloring the t-SNE based on the expression of each marker (Fig. 10B).

viSNE is a visualization tool for high-dimensional single-cell data based on t-SNE.
Ferrer-Font et al.

Current Protocols
Video 6 Assessment of data quality using expert gating and dimensionality reduction algorithms (Basic Protocol 4).
9. As in step 4, check for the appearance of unexpected marker combinations on specific cell types.
See Figure 11 for a simple checklist to help ensure that all steps have been followed to obtain a high-quality and reliable immunophenotyping panel.

COMMENTARY Background Information
The first step of building a successful large multicolor panel is good theoretical panel design. Full-spectrum flow cytometry panel design has been previously described (Ferrer-Font, Pellefigues, et al., 2020;Park et al., 2020). In summary, in the context of successful theoretical panel design, it is very important to have a clear experimental question and to know the biology of the markers that are included in the assay, including their expression and co-expression patterns. It is important to understand the instrument configuration (i.e., number of lasers on board) to know what the spectral signatures for each fluorophore will look like with the given configuration, as fluorophore brightness will vary depending on the excitation wavelength available. This information, coupled with the amount of spread (both given and received) of each fluorophore in the panel (detailed in Nguyen, Perfetto, Mahnke, Chattopadhyay, & Roederer, 2013), allows fluorophores to be optimally assigned to the different markers of the immunophenotyping panel.
After the panel has been theoretically designed, performance of QC steps is strongly recommended. This process is necessary to theoretically validate the panel before proceeding to a full experiment in order to reduce preventable issues as much as possible. It is advised to review the panel on a markerby-marker and population-by-population basis, making sure that fluorophores inducing considerable amounts of spread (which could impair marker resolution) are allocated to nonco-expressed markers, and that dim markers receive a minimal amount of spread, while fluorophore brightness and antigen expression levels are well matched (high-expressing antigens with dim fluorophores and lowexpressing antigens with bright fluorophores). To address potential issues, markers that are available with multiple fluorophores can be substituted to see if spillover spread can be minimized. Additionally, fluorophores that create (but do not receive) the most spillover can be designated to dump or viability channels. It is strongly recommended to complete panel optimization before working with an actual biological experiment, as it is a wise investment in terms of time, effort, and cost, without jeopardizing precious experimental samples.

Critical Parameters
Multiple factors influence the success of this protocol. The importance of good sample preparation-addressing sample complexity issues and the use of high-quality controls-must be emphasized.

Working with complex samples
When working with more complex samples than PBMCs (e.g., skin, tumor, fat tissue), the steps for panel optimization are the same, but there are some extra considerations.
Quality of single-cell suspension. An important consideration that is sometimes underestimated is the quality of the single-cell suspension that will be used to run the immunophenotyping panel. As different immune cell subsets have varying susceptibility to cell death, the single-cell suspension should have a viability of 80% or more to ensure proportional representation of the original sample (Costantini et al., 2003). Cell death can occur for different reasons, including how the sample has been treated before staining. Cryopreservation and harsh digestion protocols can affect sample quality and these procedures also need to be optimized. Indeed, when working with digested tissue, epitopes for the markers of interest should be verified to ensure the digestion protocol has not negatively impacted them. For example, if the same epitopes of interest exist in the spleen, one should compare marker staining and epitope preservation between digested and undigested spleen (Ferrer-Font, Mehta, et al., 2020).
Assessing instrument setup. Wherever possible, the instrument settings for fluorescence detector gains should remain unaltered from the Cytek Assay Settings (CAS). When using fluorescently labeled antibodies, this is achieved by carefully pairing expression levels to fluorophore brightness during panel design, and then by antibody titration to ensure fluorescence signals remain on-scale at the optimized settings. In certain applications where fluorescence intensity is not tunable, such as fluorescent reporter protein expression within Ferrer-Font et al.

of 36
Current Protocols the cells of interest, the instrument settings may need to be adjusted to accommodate offscale fluorescence signal. Any such change will impact not only the spectral signature being accommodated, but also other signatures with emission in the same area. Reducing fluorescence gains may lead to increased similarity between spectral signatures, which can give rise to more spillover spreading error and thus negatively impact marker resolution (Ferrer-Font, Pellefigues, et al., 2020). See Support Protocol 2 for steps required to alter fluorescence gain settings to accommodate off-scale signals. If using something other than the Cytek Aurora, it will be necessary to optimize the settings based on manufacturer's recommendations.
Autofluorescence. Samples with heterogeneous (Fig. 7C) or very bright homogeneous AF signatures (Fig. 7B) can make unmixing challenging. To improve the accuracy of the unmixing and to improve marker resolution in the case of heterogeneous AF samples, multiple AF reference controls can be created for each of the different AF signatures present in the sample (as if they were individual fluorophores included in the panel). The steps required for this process are divided in three sections called Discover, Distinguish and Designate. The goal is to Discover unique spectral signatures within the unstained control through use of a raw N × N plots; Distinguish these unique signatures into separate SRCs by exporting each population as a new FCS file which can then be reimported; and Designate each signature as unique fluorophores within the software to be unmixed as if they were part of the original panel. Care must be taken to ensure only clearly unique signatures (with similarity index <0.98) with at least 300 events of similar fluorescence intensity are distinguished to ensure high quality SRCs are generated (for a detailed protocol, see Support Protocol 3 and Video 4). For very bright homogeneous AF, the steps outlined in Basic Protocol 3 and Video 4 can be followed to improve unmixing and markers resolution outcomes.

Controls
The quality of controls will directly translate to the quality of unmixing and the data obtained. It is therefore worth investing the time to optimize them. It is of utmost importance to include all necessary controls from the beginning of a project. A complete overview about controls can be found in Maecker & Trotter (2006). Below is a summary of the categories of controls that should be considered when optimizing a flow cytometry panel.
Unstained controls. The unstained control is meant for AF assessment and should have a clean signature with no contamination from other fluorophores and match the tissue or sample type being analyzed. If, for example, different tissue types are used in an experiment, multiple unstained sample controls must be used for each tissue type. Similarly, if samples are being treated differently (e.g., fixed/fresh or stimulated/unstimulated), an unstained control should be included for each condition. It is not advised to mix samples from different conditions and collect only one unstained control, as it may not be possible to have enough events to generate clear AF signatures.
Spectral reference controls. Appropriate single-stained SRC samples are required for optimal unmixing of the fully stained sample. The purpose is to provide a signature of each fluorophore to be used by the unmixing algorithm. Briefly, SRCs should have positive and negative populations that are clearly separated or a universal negative; positive populations should be brighter than the fully stained sample; the negative and positive populations should have identical AF characteristics; sufficient events for both populations should be collected; and the fluorescence spectrum of the positive control needs to be identical to the one in the fully stained sample. To fulfill these best practices, cells and beads should be compared and the best option should be used. Special considerations for viability staining are important, as live and dead cells have different AF signatures, which means the positive and negative controls will not have the same AF signature. To overcome this issue, it is possible to kill all cells in the viability SRC (e.g., heat-killing at 55°-70°C for 5-10 min), stain only half of them, and mix these with the unstained cells. In this case, all the cells will be dead and the AF will be the same for the positive and negative controls.
SS cell controls. SS cell controls are used not only as reference controls but also to assess the performance of each marker compared to the FS tube. SS cells are gold-standard for each antibody performance and are used to quantify the spread of the negative population and/or shifts in the positive signal and thus any subsequent loss of resolution of markers in the panel.
Gating controls. A fluorescence minus one (FMO) control is a sample stained with all Ferrer-Font et al.

of 36
Current Protocols fluorophores used in the experiment except one. Analyzing the FMO control for each fluorophore in the panel is not required for panel optimization, but FMOs can be used as a guide to set the boundary between positive and negative events if it is ambiguous. FMOs also aid in the assessment of spread between positive and negative events and are an important tool for assessing panel performance. A good alternative for large 20+ color spectral panels is the use of fluorescence minus multiple (FMM) controls, as recently described by Jensen & Wnek (2020).

Core versus FS panel
When optimizing a high-dimensional immunophenotyping panel, it is recommended to first complete Basic Protocols 1-4 with a core panel. A core panel is limited to only the essential lineage markers required to identify the cell types of interest. This will reduce confounding factors when attempting to understand the source of errors. Once a core panel has been optimized, additional markers can be added (including readout functional markers, intracellular markers, etc.) and optimized with the knowledge that issues are not originating from the core panel.

Fixation/permeabilization buffers
Intracellular staining procedures also require special consideration. It is important to source the fixation/permeabilization reagent that is most appropriate for the markers being detected (e.g., transcription factors, cytokines, or intracellular proteins) while also considering their relative locations (cytosolic or nuclear). The type of fixation buffer can influence the staining of both intracellular and surface markers. The type of fixative can have an impact in many aspects of staining: damaging the epitopes, altering the fluor stability and resulting in a different optimal titer, altering the background fluorescence without altering the positive signal (leading to reduced resolution between positive and negative populations), and altering the fluorescence spectral signatures (causing spectral signature mismatches between FS samples and SRCs, resulting in unmixing issues). It is recommended that antibodies be titrated using the same fixation/permeabilization buffer conditions used for the final staining of FS samples and SRCs.

Troubleshooting
The term troubleshooting is used when one or more issues are found in the panel and the source of these issues needs to be identified and steps included to rectify them. In this regard, the exact steps provided for validating the panel can also be applied for troubleshooting. By following these clear steps, the user will gain a better understanding of the quality of the panel and identify issues prior to acquiring experimental samples.
In general, 80%-90% of unmixing issues that arise can be traced to suboptimal controls. Therefore, it is important to use controls that are well characterized, high quality, and appropriate for the experiment (i.e., matched to the conditions of the experimental sample). This is particularly vital if they are going to be reused.
The checklist provided in Figure 11 aims to help users follow clear steps to optimize and troubleshoot their panels. By following these steps, users should be able to discover issues if they exist and have a pathway and alternatives to address and resolve them. Additional troubleshooting for general issues that may arise during panel optimization can be found in Tables 2-5.

Understanding Results
The example panel optimized using this protocol and presented here aims to identify conventional T cell subsets in peripheral blood of healthy participants. The participants were infected with low-dose human hookworm as part of a longitudinal study. It is well established that during early parasitic infection there is an increase in T-helper Type 2 immune responses (Th2) in the T cell compartments due to the primary role these cells play in host responses to parasites (McSorley & Loukas, 2010). However, the wider effects on the human immune system of long-term chronic parasitic infection with the gut-residing hookworm Necator americanus (Na) have not been investigated. Therefore, a panel was designed and optimized to look at the following conventional and unconventional T cell subsets of interest: CD4 + T cells (Th1, Th2, Th17, Th9, Th22, and Tregs), CD8 + T cells (Tc1 and Tc2), and innate-like T cells (γδ T, mucosal associate invariant T [MAIT], and natural killer T [NKT] cells). This panel can assess the frequency of these cell subsets as well as their memory and activation phenotypes, using both expert gating and high-dimensional data analysis.
Some characteristics of the fluorophore selection for a certain panel including the      Use Brilliant Stain Buffer when more than one polymer dye is added in the same tube following manufacturer recommendations similarity and complexity indices of the panel (Fig. 12A) and the spectral signatures of fluorophores used (Fig. 12B) have been included. Although some fluorophores have a high similarity index, they have been allocated to different cell types and therefore the impact of the expected spreading error should be minimal (Fig. 12). A panel distribution table has also been included (Fig. 12C) to show the peak emission wavelengths for the fluorophores and markers assigned to the panel.
After following all the steps of Basic Protocols 1-4 for our panel shown here, and resolving any issues that arose (such as one tandem degradation or implementing the sequential staining for some of the markers), high-quality full-spectrum flow cytometry data as defined by clear resolution of all expected populations was achieved. The optimized staining protocol for this panel is detailed in the Supplementary Material. Sequential staining was applied, following the order given in the protocol, for markers that showed a reduction in positive signal in the FS compared to the SS control, starting at the marker with the greatest reduction and working toward the one with the least reduction. This approach provided satisfactory results and did not require the testing of other combinations. To reduce staining procedure time, further optimization would be required to determine whether some markers could be added at the same time instead of using all of them sequentially.
A clean and clear positive population can be detected for all markers in the panel, with no signal resolution loss when all antibodies are combined. In some cases, such as for the activation markers, the use of FMOs is necessary to assist in evaluating gates to determine positive staining with confidence. All populations of interest could be found using expert gating, and the populations resemble the expected expression patterns and frequencies.
Additionally, high-dimensional data analysis algorithms were successfully used without the appearance of artefacts, confirming the high quality of the data.

Time Considerations
The time needed for designing, optimizing, and analyzing a high-dimensional flow cytometry panel can be highly variable and can depend on the assay complexity, number of markers, wait-time required for reagents (geographically dependent), sample access and frequency of sample delivery (particularly for patient samples), duration of the disease model being investigated, and more. As an example, it took five months to optimize the digestion, design, and optimization of a 23-color spectral flow cytometry panel in gut tissue (Ferret-Font, Mehta, et al., 2020), whereas one month may be sufficient for a PBMC panel where digestion does not need optimization. It is therefore advisable to develop certain core panels that can be applied to several experimental questions and models.