Development of a standardized and validated flow cytometry approach for monitoring of innate myeloid immune cells in human blood

Innate myeloid cell (IMC) populations form an essential part of innate immunity. Flow cytometric (FCM) monitoring of IMCs in peripheral blood (PB) has great clinical potential for disease monitoring due to their role in maintenance of tissue homeostasis and ability to sense micro-environmental changes, such as inflammatory processes and tissue damage. However, the lack of standardized and validated approaches has hampered broad clinical implementation. For accurate identification and separation of IMC populations, 62 antibodies against 44 different proteins were evaluated. In multiple rounds of EuroFlow-based design-testing-evaluation-redesign, finally 16 antibodies were selected for their non-redundancy and separation power. Accordingly, two antibody combinations were designed for fast, sensitive, and reproducible FCM monitoring of IMC populations in PB in clinical settings (11-color; 13 antibodies) and translational research (14-color; 16 antibodies). Performance of pre-analytical and analytical variables among different instruments, together with optimized post-analytical data analysis and reference values were assessed. Overall, 265 blood samples were used for design and validation of the antibody combinations and in vitro functional assays, as well as for assessing the impact of sample preparation procedures and conditions. The two (11- and 14-color) antibody combinations allowed for robust and sensitive detection of 19 and 23 IMC populations, respectively. Highly reproducible identification and enumeration of IMC populations was achieved, independently of anticoagulant, type of FCM instrument and center, particularly when database/software-guided automated (vs. manual “expert-based”) gating was used. Whereas no significant changes were observed in identification of IMC populations for up to 24h delayed sample processing, a significant impact was observed in their absolute counts after >12h delay. Therefore, accurate identification and quantitation of IMC populations requires sample processing on the same day. Significantly different counts were observed in PB for multiple IMC populations according to age and sex. Consequently, PB samples from 116 healthy donors (8-69 years) were used for collecting age and sex related reference values for all IMC populations. In summary, the two antibody combinations and FCM approach allow for rapid, standardized, automated and reproducible identification of 19 and 23 IMC populations in PB, suited for monitoring of innate immune responses in clinical and translational research settings.

Innate myeloid cell (IMC) populations form an essential part of innate immunity. Flow cytometric (FCM) monitoring of IMCs in peripheral blood (PB) has great clinical potential for disease monitoring due to their role in maintenance of tissue homeostasis and ability to sense micro-environmental changes, such as inflammatory processes and tissue damage. However, the lack of standardized and validated approaches has hampered broad clinical implementation. For accurate identification and separation of IMC populations, 62 antibodies against 44 different proteins were evaluated. In multiple rounds of EuroFlow-based design-testing-evaluation-redesign, finally 16 antibodies were selected for their non-redundancy and separation power. Accordingly, two antibody combinations were designed for fast, sensitive, and reproducible FCM monitoring of IMC populations in PB in clinical settings (11-color; 13 antibodies) and translational research (14-color; 16 antibodies). Performance of pre-analytical and analytical Introduction Monocytes, dendritic cells (DCs) and granulocytes, together also called innate myeloid cells (IMCs), play key roles in multiple different processes related to maintenance of tissue homeostasis, including sensing of tissue damage, induction and/or resolution of inflammation, antigen presentation and pathogen eradication (1)(2)(3)(4)(5)(6)(7)(8)(9). While some of these cell populations, such as mast cells and macrophages, are merely tissue-resident, others like monocytes, DCs, basophils, eosinophils and neutrophils circulate via peripheral blood (PB) with the ability to sense micro-environmental changes (such as inflammatory processes) and migrate to tissues where they modulate local responses in both physiological and disease conditions (10)(11)(12). This great plasticity and functional heterogeneity of IMCs renders them into ideal candidates for monitoring disturbances in body homeostasis at the systemic level, e.g. in PB. Consequently, the clinical utility of monitoring IMCs in PB for diagnosis, staging, prognostic assessment and/or evaluating response to treatment in multiple disease conditions has been demonstrated previously (9, 13-24).
However, monitoring IMCs for translational research and diagnostic patient care is currently hampered by the lack of standardized approaches. This includes the absence of immunophenotypic consensus criteria for the definition of the distinct IMC subsets, due to their great heterogeneity and plasticity (25-31) and the limited availability of lineage-specific proteins, which have led to the introduction of e.g., marker cocktails for lineage exclusion and highly variable strategies and/ or extensive sets of markers for correct identification of the target populations (25-27, 29, [31][32][33][34]. Additionally, new monocytes and DCs have been identified, leading to progressively more complex antibody panels and data analysis procedures. For example, new subsets of classical (cMo) and non-classical (ncMo) monocytes have recently been defined based on the expression pattern of CD9, CD62L, CD93 and/or FcϵRI and CD9, CD36 and Slan, respectively (35-39). Likewise, CD1c + myeloid dendritic cells (myDCs) are now known to include different functional subsets, that can be identified based on CD14 expression (CD14noninflammatory and a CD14 lo pro-inflammatory CD1c + myDC population) (40) and CD5: CD5 hi CD1c + myDCs with higher ability to migrate to the lymph nodes and induce T cell proliferation, and CD5 -CD1c + myDCs with a closer functional profile to monocytes (41, 42). In addition, the new subset of Axl + and SIGLEC6 + DC (Axl + DCs) has been described recently, which was previously included in the plasmacytoid dendritic cell (pDC) population and exhibits mixed gene expression and functional profiles between pDCs and myDCs (40). In parallel, a new population of DC precursors has been described, co-expressing CD34 int and CD100 hi , with the ability to generate in vitro both CD1c + and CD141 + myDCs (31,40,43).
In recent years, different 8-12 to 38 color panels have been designed and proposed for monitoring monocytic and DC populations in PB by flow cytometry (FCM) and mass cytometry, respectively (26, [44][45][46]. However, careful analysis of these FCM antibody panels shows that they typically include multiple redundant markers for defining IMC populations (e.g., CD123, CD303 and/or CD304 for identification of pDCs) and/or they require the use of antibody cocktails for exclusion (e.g., "dump channel") of e.g. lymphoid cells, which prevent the addition of other relevant markers (29,45). In contrast, a previously described 38-color mass cytometry antibody panel allows identification and characterization of virtually all monocyte, monocyte-derived macrophage, DC and myeloidderived suppressor cell populations (47). However, mass cytometry is not readily available in many centers and, most importantly, has a very low throughput (250-350 cells/sec) and limited levels standardization, which limit its use in clinical settings. Furthermore, none of the previously reported antibody panels allow identification of the recently described DC and monocyte populations. At the same time these antibody panels did not use standardized and validated procedures for antibody panel design and data analysis in a multicentric setting, and failed to provide age-matched related ranges for the IMC populations (26, 44-46, 48, 49).
Here we designed and validated two (11-and 14-color) antibody panels for standardized, automated, and reproducible identification of 19 to 23 IMC populations in human blood by FCM, and provide age and sex-matched reference values for more objective interpretation of altered IMC profiles in multicentric clinical settings. Ultimately, the antibody panels developed will allow to set a new benchmark for IMC in both clinical and translational research settings.

Samples
For this study, 261 PB samples (195 ethylenediaminetetraacetic acid -EDTA-and 66 sodium heparin-anticoagulated) obtained from 205 healthy donors (HD) were evaluated. From them, 242 samples from 197 donors were used for antibody panel development and evaluation (72 men, 118 women and 7 donors lacking sex information, with median age of 32 years -y-ranging from 5y to 99y). For assessment of the technical performance of the antibody panels, construction & validation of the reference database for automated gating (50) 57 samples from 48 donors were used (20 men, 25 women; median age 38y; range: 5y -85y; of note, sex information was not available for 3 donors). A total of 116 samples from 67 women and 45 men (unknown sex in 4 donors) with median age 30y (range: 8y-69y) were processed for defining ageand sex-related normal reference ranges. Additionally, 4 cord blood (CB) samples collected in EDTA were also included for the study of infrequent populations in steady-state PB, which are reported to be increased in CB (e.g., myeloid-derived suppressor cells -MDSC-, immature neutrophils). All samples were collected after informed consent was provided by each donor according to the Declaration of Helsinki and the guidelines of the local ethics committees and review boards. Of note, this study includes pre-vaccination samples collected and processed in the context of the Dutch 'BERT study', which was initiated by the Innovative Medicines Initiative (IMI)2 PERISCOPE consortium (51, 52) and was approved by the Medical Research Ethics Committees United (MEC-U, NL60807.100.17-R17.039) and registered at the EU Clinical trial registry (EudraCT number 2016-003678-42).

Immunophenotypic studies
Samples were processed within 4 hours (h) after collection, according to the EuroFlow bulk lysis and sample preparation and staining standard operating procedures (SOP) (52, 53) for surface membrane (Sm) only and Sm plus cytoplasmic (Cy) labeling of 10 7 cells, employing the antibodies (Supplementary Table 1) and antibody combination depicted in Table 1 and Supplementary  Tables 2, 3. Protocols are described in detail in the Supplementary Methods section and on the EuroFlow website (www.EuroFlow.org ).
Stained cells were stored at 4°C and measured within 1h by FCM. Absolute counts were assessed employing a double platform method based on quantitation of nucleated cells obtained in the Sysmex XP-300 automated hematological analyzer (Sysmex Europe GmbH, Norderstedt, Germany).

In vitro activation assay of monocytes and DCs
Short-term in vitro activation assays were performed using sodium heparin anti-coagulated PB, as described elsewhere (53). Briefly, 500 ml of PB diluted 1/1 (vol/vol) with RPMI 1640 medium (Sigma-Aldrich, Zwijndrecht, The Netherlands) were incubated for 6h at 37°C in a sterile environment containing 5% CO 2 in the presence of 100 ng/ml of lipopolysaccharide (LPS) (Sigma-Aldrich). For those experiments in which intracellular detection of cytokines was performed, 10 mg/ml of Brefeldin A (Sigma-Aldrich) was added to block cytokine secretion. For each condition, an unstimulated aliquot of the same sample was processed in parallel in an identical way. Stimulated PB samples were then stained with a panel of monoclonal antibodies (MoAb) (Supplementary Table 2) using the EuroFlow bulk lysis and sample preparation and staining SOPs (www.EuroFlow.org) as previously described (54).

Sample acquisition and analysis
For each sample evaluated, 2.5 to 5 x 10 6 cells were measured using LSR Fortessa (Becton Dickinson Biosciences (BD), San Jose, CA) instruments equipped with 4 lasers (405nm, 488nm, 561nm and 640nm) or a 3-laser (405nm, 488nm, 640nm) Aurora (Cytek, Fremont, CA) instrument. BD Fortessa flow cytometers were set-up at each center according to the EuroFlow guidelines (www.EuroFlow.org) and calibrated daily by use of BD ™ Setup and Tracking (CS&T) beads (BD Biosciences), their performance being checked daily by acquisition of SPHERO ™ Rainbow calibration particles (Cytognos S.L., Salamanca, Spain). Calibration and daily quality control of the Aurora flow cytometer was performed according to the SOP recommended by the manufacturer. For data analysis, the Infinicyt ™ software (version 2.0.2.d.000; Cytognos S.L., Salamanca, Spain) was used. Gates were defined based on internal negative and fluorescenceminus-one (FMO) controls, for general population identification and immunophenotypic characterization, respectively.

Antibody evaluation and selection for the EuroFlow IMC tubes
To design accurate and reproducible antibody combinations for IMC detection in PB, 62 antibodies against 44 proteins were stepwise evaluated in several rounds of EuroFlow-based designtesting-evaluation-redesign (Table 1, Supplementary Table 1). In a first step, 8 antibodies were used as backbone to accurately identify the major monocytic populations (CD14, CD16, CD45, CD300e, HLA-DR) and their subsets (CD36, CD62L and Slan) ( Table 1) (9, 37, 39). Selection of different reagents was carried out for each target antigen, based on discrimination between positive and negative reference populations, employing stain index values [calculated as (MFI PRP -MFI NRP )/2 x rSD NRP ; where MFI, median fluorescence intensity; PRP, positive reference population; NRP, negative reference population; rSD, robust standard deviation], as previously described (55).
In a subsequent step, selection of the minimum set of the most informative markers for identification of additional subsets of IMC was performed per cell population, e.g., pDCs, myDCs, Axl + DCs, CD100 + preDCs, myeloid-derived suppressor cells (MDSCs) ( Table 1), using counter-staining with the backbone markers for the major population identification (CD14, CD16, CD45, CD300e, HLA-DR). Selection of individual markers and marker combinations was based on specificity, redundancy, population discrimination and lack of cross-contamination by other cell subsets, as assessed by principal component analysis (PCA) and canonical multivariate analysis (CA) using Infinicyt ™ (56). For Axl + DCs, accuracy of the set of markers used for their identification was further validated (Supplementary Table 2  to improve the discrimination of leukocytes from debris and platelets, and better identify immature neutrophils, the CD45 antibody reagent was replaced with a brighter conjugate in the 11-color version of the tube (version 3; Table 1). At a later stage, an extended 14-color version (version 4, Table 1) was designed, which also included i) CD5 for further subsetting of CD1c + CD14 -myDCs; ii) CD34 for identification of hematopoietic precursor cells (HPCs) and CD100 + CD34 int pre-DCs; and iii) CD192 for identification of M-MDSCs. Additionally, the fluorochrome conjugated to CD62L was changed to minimize its spread on the FceRI channel, as required for clear cut subsetting of cMos (Table 1).

Intra-and inter-assay reproducibility
Intra-assay variation of the EuroFlow IMC tube, expressed by the intra-assay coefficient of variation (%CV) was determined in duplicates of 5 EDTA-anticoagulated PB samples, processed in parallel (version 3; Table 1) and measured in a Fortessa X20 (BD) flow cytometer. In addition, inter-center reproducibility was also evaluated via analysis of PB samples from HD locally collected, processed (version 3; Table 1) and measured at 4 centers: Leiden University Medical Center (LUMC) (n=9), University of Salamanca (USAL) (n=5), National Institute for Public Health and the Environment (RIVM) (n=3), and University of Turku (UTU) (n=4), using five different instruments (2 LSR Fortessa and 3 Fortessa X20). For this purpose, the %CV of the median fluorescence intensity (MFI) obtained for each marker in predetermined positive reference cell populations was first calculated per center (intra-center variation), and the technical variability between centers (inter-center variation) estimated based on the median MFI of each marker per center.

Reproducibility of manual data analysis
To evaluate the inter-operator reproducibility of manual analysis, flow cytometry standard (.FCS) sample files from 6 adult HD (stained with version 3 of the EuroFlow IMC tube) were independently analyzed in parallel by an experienced (expert 1 -E1) and a novice (expert 2 -E2) flow cytometrist. Intra-operator variability was assessed for E1, who analyzed the files twice within a time lapse of ≥2 months.

Database construction for automated data analysis
For construction of the database for automated analysis of the 11-color version of the EuroFlow IMC tube (version 3, Table 1), 18 PB samples from healthy adults were processed and measured in Fortessa X20 and LSR Fortessa instruments, at the four different sites mentioned above, within the framework of the Horizon 2020/IMI multicenter PERISCOPE consortium (http://periscope-project.eu/): LUMC (n=5), USAL (n=8), RIVM (n=2), and UTU (n=3). Flow cytometry data files from those 18 samples that fulfilled all the selection criteria (described in detail in Supplementary Methods) were then merged into a single data file used as database tool, implemented in Infinicyt ™ (Cytognos) (57). For validation of the database vs. manual analysis performed by an experienced flow cytometrist (E1), a second set of PB samples from HD (n=6), processed and acquired at LUMC (n=3) and USAL (n=3), was prospectively used. For these samples, analysis was repeated at two different timepoints set ≥2 months apart from each other.

Selection of markers for identification of dendritic cell populations
Based on earlier work (9, 37-39), a set of eight markers (CD14, CD16, CD36, CD45, CD62L, CD300e, HLA-DR and Slan) that allows for identification and subsetting of monocytic cells, was pre-selected as backbone (Table 1), based on antibody clones that had previously shown to perform well technically (Supplementary Table 1) (9, 37, 39, 58). Three markers (CD123, CD303 and CD304) were evaluated for the specific identification of pDCs (25,27,31) in combination with the backbone combination required for identification of the major populations of monocytes and granulocytes (CD14, CD16, CD45, CD300e and HLA-DR). All three markers individually allowed clear identification of pDCs within the HLA-DR + /CD14 -/CD16cell compartment ( Figures 1A-F). However, whereas the CD303 and CD304 expression was highly specific for pDCs, CD123 was also van der Pan et al. 10.3389/fimmu.2022.935879 Frontiers in Immunology frontiersin.org   Figure 1N). Importantly, since CD303 and the backbone marker CD300e are not expressed on the same cells (i.e., monocytes/myDCs vs. pDCs, respectively) (Supplementary Figure 1), both antibodies could be used in the later versions of the antibody combination with the same fluorochrome.

Detection of CD100 + preDCs in PB does not need a CD100 antibody
Even though the definition of the myDC precursor is still elusive, a PB population identified based on a CD34 int CD100 hi immunophenotype (Supplementary Figure 3), with the ability to differentiate to both CD1c + and CD141 + myDC, has been described in PB (40). PCA performed on PB cells stained with CD34 and CD100 in combination with CD14, CD16, CD33, CD45, CD300e, CD303 and HLA-DR exhibited a clear separation between CD34 + HPC and CD100 + DC precursors mostly due to their different pattern of expression of HLA-DR hi and CD34 int (Supplementary Figure 3B), showing that CD100 is not critically required, as a similar discrimination power was observed when CD100 was excluded (Supplementary Figures 3C-D). Based on these results, CD34, but not CD100, was included in the extended 14-color version 4 of the IMC antibody panel ( Table 1).

Detection of Axl + DCs does not need an Axl antibody
In 2017, Villani et al. (40) described a new population of DCs that overlaps with pDCs, when classical identification markers are used, but that could be accurately discriminated based on the expression of Axl. In order to identify Axl + DC, Axl was combined with CD1c, CD14, CD16, CD33, CD45, CD141, CD303, CD300e and HLA-DR. Overall, inclusion of Axl in the antibody combination proved not to be critically required for identification of this DC population, as the expanded backbone combination allowed for the separation of the Axl + DCs from pDCs and myDC populations based on its unique pattern of expression of CD33 lo , CD141 + and CD303 lo ( Figures 3A-G). This was further confirmed by multivariate analysis, which revealed similar population discrimination patterns, independently of the presence or absence of Axl (Figures 3H, I), associated with similar Axl + DC counts ( Figure 3J). Axl + DCs have been reported to display a mixed gene expression signature between myDC and pDC, with shared immunophenotypic features with pDCs (e.g., CD123 and CD303 expression) and functional characteristics of myDCs (e.g., response to LPS) (40). Therefore, we further validated the functional identity of the Axl + DC population, identified based on the restricted set of markers selected for evaluation of DC populations, (Table 1;  Supplementary Table 2). Our results showed that in unstimulated samples, expression of CD11b was restricted to myDCs, whereas CD33 was also (dimly) expressed on Axl + DCs, but not on pDCs. In turn, steady-state Axl + DCs displayed a higher frequency of pro-inflammatory cytokine producing cells vs. myDCs CD1c + CD14 -(p<0.02 for IL1b and IL12) and pDCs (p<0.003 for IL1b and IL8) (Supplementary Figures 4,5). Upon exposure to LPS, CD1c + myDCs populations displayed a strong response to LPS, while pDCs and CD141 + myDCs were mostly unresponsive, Axl + DCs exhibited an overall intermediate activation pattern, associated with a unique profile for those markers that showed significant differences in steady-state and/ or in LPS-stimulated samples (CD33, CD62L, CD63, CD69, CD83, CD86, IL1b, IL6, IL8, IL12 and TNFa) (Figures 3K, L).

Selection of markers for identification of immature vs. mature neutrophils
In order to determine whether additional markers are required for accurate identification of immature vs. mature neutrophils, PCA-based evaluation of the performance of the IMC tube extended backbone (i.e., backbone markers plus the markers required for identification of DCs) vs. the extended backbone plus CD11b, CD15 and CD66b, for identification of different polymorphonuclear (PMN) cells, including immature neutrophils, was performed. Of note, combined usage of cell size (forward scatter -FSC-) and internal complexity (side scatter -   Figures 4A,B). This was also confirmed by the expression pattern of markers known to be associated and/or modulated during neutrophil maturation ( Figure 4C), as cells identified based on an HLA-DR -CD14 -CD16 -/lo CD33 + CD45 lo CD300ephenotype in fact correspond to immature (CD11b -/+ , CD15 + , CD66b + , CD244 -/lo ) neutrophils. Interestingly, neutrophils could be further subclassified based on expression of CD16 and CD62L ( Figure 4D) as mature neutrophils (CD16 hi CD62L + ), a phenotype previously reported to be associated with segmented neutrophils (63), immature neutrophils CD16 lo CD62L + , compatible with band neutrophils (63), and other, even more immature subsets of CD16 -/lo CD62Lneutrophils, that might include an admixture of promyelocytes (CD11b -), myelocytes (CD11b + ) and metamyelocytes (CD16 lo/+ ) (Figures 4C, D) (64). As expected (65), significantly higher frequencies of immature neutrophils were observed in CB samples vs. adult PB with the extended backbone ( Figure 4E), with similar immature neutrophil counts in the presence vs. absence of additional neutrophil-associated markers ( Figure 4F).

Selection of markers for identification of monocytic myeloid-derived suppressor cells
Monocytic M-MDSCs have been classically identified as CD14 + CD11b + (or CD33 + ) CD15 − and HLA-DR −/lo cells (32). This combination relies on the expression of HLA-DR as the discriminating marker vs. cMos, which requires FMO or internal negative controls for accurate identification of this cell population.
To specifically identify markers that would allow for an improved identification of M-MDSCs, we evaluated the pattern of expression of monocyte and M-MDSC-related markers on cMos vs. CD14 + HLA-DR -/lo cells from CB and/or adult PB samples (Supplementary Figure 6A). Our results confirmed the absence of CD15 together with expression of CD11b on CD14 + HLA-DR -/lo M-MDSCs, and showed significant (p=0.03) up-regulation of CD16 and down-regulation of CD123 and CD192 on CD14 + / HLA-DR -/lo cells vs. cMos (Supplementary Figures 6B, C). PCA revealed that only CD16, HLA-DR and CD192 had significant (independent) impact on the discrimination between the two populations (Supplementary Figure 6D), with addition of CD123 having negligible value for identification and quantification of the population (Supplementary Figure 6E). When comparing the frequency of M-MDSC in CB vs. adult PB, defined based on a CD14 + HLA-DR -/lo or CD14 + HLA-DR -/lo CD192 -/lo phenotype, lower frequencies were overall observed with the latter, more stringent, criteria (Supplementary Figure 6F). Importantly, statistically significantly higher frequencies of M-MDSCs in CB vs. adult PB were only observed when the CD14 + HLA-DR -/lo CD192 -/lo criteria was used (Supplementary Figure 6F), suggesting that the addition of CD192 could allow for a more accurate identification of M-MDSCs.

Impact of the anticoagulant, delayed sample preparation and freezing on identification of IMC populations
Since the performance of the EuroFlow IMC tube was evaluated on PB collected in EDTA and, in some settings, sodium heparin (e.g., for functional assays) is required, which might affect the staining patterns and quantification of IMC populations (66), staining of samples collected with EDTA vs. sodium heparin was compared. Except for CD300e that showed lower expression on monocytes from heparin samples (median stain index reduction in heparin vs. EDTA of 38.4%; range: 14.2%-71.1%; p=0.02), no significant differences were observed in the stain index of individual markers between samples collected with these two anticoagulants (data not shown). However, despite the lower CD300e expression on heparinanticoagulated samples, multivariate PCA analyses revealed no significant impact on the overall discrimination of the distinct populations of IMCs ( Figure 6A). Likewise, no significant differences were observed on the absolute counts of the populations between the two anticoagulants with exception of a lower absolute count of CD1c + CD14 lo myDC observed in heparin samples (p=0.02) (Figures 6B-V). Regarding immediate vs. delayed sample preparation and staining for 6h, 12h and 24h with the EuroFlow IMC tube, similar stain index values were observed for all markers evaluated (data not shown), except for CD16 (median decrease in stain index of 25.3%, 38.5% and 41.9%, respectively) and Slan (median decrease in stain index of 43.2%, 5.4% and 8.1%, respectively), also confirmed by PCA analyses, as all populations evaluated for all timepoints tested clustered within one standard deviation of the 0h staining pattern ( Figure 7A).
Impact of the anticoagulant on the staining patterns and IMC population (absolute) counts in blood. Peripheral blood samples (n=7) were collected into K3 ethylenediaminetetraacetic acid (EDTA) and sodium heparin (HEP) tubes and stained with versions 2 (n=3) and 3 (n=4) of the EuroFlow immunemonitoring innate myeloid tube.  To determine the assay reproducibility, duplicates of the same EDTA-anticoagulated PB samples (n=5) were stained and measured in the same instrument and analyzed manually by an expert cytometrist. Overall, an average intra-assay %CV of 5.0% ± 4.5% was observed across the 26 populations evaluated, with 80.8% (21/26) of the populations displaying an intra-assay %CV <10% and only CD36 -Slan -ncMo exhibiting a median intra-assay %CV >15% (Supplementary Table 5).
Comparison of the performance of the EuroFlow IMC tube between different instruments with distinct detector/optical Impact of delayed sample processing on the overall performance of the innate myeloid panel for population identification and quantification of innate myeloid cell ( Frontiers in Immunology frontiersin.org configurations (conventional vs. spectral, and 3vs. 4-laser flow cytometers) was evaluated. Overall, a significant correlation (R 2 >0.90; p<0.05) was observed for virtually all (92%; 23/25) IMC populations identified, with no significant differences and a limited bias (absolute mean normalized bias <15%) being detected between instruments. The only exceptions were CD62L + FceRI + cMos and CD36 + Slan + ncMos which were overestimated (bias: +17.5%) and underestimated (bias: -32%) in the data files generated in the Aurora vs. Fortessa X20 instruments, respectively (Table 2). To further evaluate the feasibility of using the EuroFlow IMC tube in multicentric settings, 21 samples were locally collected, processed, and measured at 4 distinct facilities (LUMC, USAL, RIVM and UTU) using 5 distinct instruments. PCA revealed fully comparable and reproducible results for all centers/ instruments ( Figure 8A). Furthermore, when comparing the assay %CV for MFI values of predefined positive reference IMC populations (PRP) for the different markers evaluated, the inter-center assay %CV was within the range of the observed biological variability (i.e., intra-assay %CV) within individual centers ( Figure 8B)

Reproducibility of expert-based manual analysis
Reproducibility of expert-based manual analysis of the EuroFlow IMC tube was evaluated by experienced (E1) and For determination of the comparability between samples measured using different types of instruments (conventional vs. spectral cytometers) regarding the relative distribution of the populations, a linear regression was performed to evaluate the direction and strength of the relationship between the two conditions, a Wilcoxon test was performed to compare the differences observed between the two conditions and a Bland-Altman analysis was done in order to determine the potential bias. MNB, mean normalized bias (calculated as % of difference between the relative frequencies obtained with the Aurora compared to the results obtained with the Fortessa X20); N.S., not significant (p<0.05); cMo, classical monocytes; iMo, intermediate monocytes; ncMo, non-classical monocytes; myDC, myeloid dendritic cells; pDC, plasmacytoid dendritic cells.  (Table 3). However, a lower correlation and degree of agreement were observed for populations identified based on a limited number of heterogeneously expressed markers (e.g., cMo populations defined based on CD62L and FceRI expression and ncMo populations, defined based on expression of CD36 and Slan) and infrequent (<0.05% of total leukocytes) IMC populations (e.g., Axl + DCs). To establish the intra-operator variability, expert E1 repeated the analysis of the files with a ≥2-month interval. Of note, even though the overall degree of correlation increased compared to expert E1 vs. E2 (significant correlation of 78.6%; 22/28 vs. 71.4%; 20/28) and agreement (absolute MNB <15% of 82.1%; 23/ 28 vs. 71.4%; 20/28), the same patterns for populations with lower degree of agreement (i.e., population defined based on limited and heterogeneous markers and infrequent subsets), were observed (Table 3 and Supplementary Table 6).

Database construction and automated data analysis
Comparison of manual expert-based vs. database-guided automated gating showed a better degree of correlation (85.7%; 24/28) and agreement (82.1%; 23/28), compared to intra-and inter-operator manual analysis (Table 3), with an improved identification of some IMC populations defined based on the expression of heterogeneous markers (i.e., most of the ncMo subsets). Despite this, low correlation and/or degree of agreement was still observed for cMo subsets, defined based on the expression of CD62L and FceRI, and IMC populations present at low frequency (<0.05%) such as Axl + DC or CD62L + immature neutrophils. Of note, database-guided automated gating and identification (AGI) performed at two different timepoints displayed a 100% correlation and degree of agreement for the 28 (IMC and non-IMC) populations tested, which clearly improves reproducibility compared to both intraand inter-operator manual analysis.

Discussion
Monitoring of IMC populations for diagnostic patient care has been historically hampered by the lack of standardized For determination of the comparability between analysis performed by two distinct experts (E1 vs. E2), at two distinct timepoints (2 months apart; 1 st round vs. 2 nd round) and between conventional manual and automated database-guided analysis, a linear regression was performed to evaluate the direction and strength of the relationship between the two conditions (high agreement defined by R 2 >0.9 and p<0.05). Additionally, a Bland-Altman analysis was done in order to determine the potential bias (high agreement defined as -15% > mean normalized bias (MNB) < +15%). *Median % of cells as identified by expert 1 (E1) (1 st round). identification of 4 additional, less frequently reported, IMC populations. For fast translation to diagnostic laboratories, we evaluated the impact on both IMC population phenotypes and counts in PB, of different anticoagulants, immediate vs. delayed sample preparation and the usage of distinct types (conventional vs. spectral) of FCM instruments in single vs. multicenter settings. Finally, we developed a database-guided automated analysis approach for standardized data analysis and provided normal age-and sex-matched reference values as a basis for future immune-monitoring in patient care. A backbone previously identified and validated by the EuroFlow and TiMaScan consortia for immune-monitoring of major granulocytic and monocytic (sub)populations (9, 37, 39, 67), was employed as a basis for panel design. This combination already allowed for identification of eosinophils, mature neutrophils, two populations of cMos (CD62L + and CD62L -), iMo and four populations of ncMos (defined based on CD36 and Slan expression). Of note, previous reports suggested that CD9 instead of CD36 might also be used for ncMo subsetting within the Slan + compartment (35). However, the expression of the two markers is redundant within Slan + cells (35) and CD36 further allows for identification on an additional Slan -ncMo population and at the same time, it is more specific for monocytes and DCs than CD9.
In a second step, markers classically employed for identification of pDCs (i.e., CD123, CD303 and CD304) (25, 27, 31, 41) and myDCs (i.e., CD11c and CD33) (38,41,42,45,68) were tested. CD303 and CD33 showed the best performance for clear discrimination of pDCs and myDCs, respectively, overcoming the need for an exclusion cocktail of lymphoidassociated markers. This is due to the fact that CD303 is highly specific for pDCs (69), and CD33 cross-contamination would result mainly from monocytes (70), which can be excluded based on counterstaining with the backbone markers. Other markers, e.g., CD11c are also expressed on B cells (71), and would require the inclusion of an exclusion B-cell marker. Although a splicing polymorphism has been reported for CD33, leading to loss of epitopes recognized by anti-CD33 antibodies (72), the usage of a bright fluorochrome (i.e., PE Cy7) in combination with other markers in the panel (e.g., FceRI, CD14, CD16, CD1c, CD141, CD303) still allowed for accurate identification of myDCs, also in individuals displaying CD33 lo expression (data not shown).
Recent reports have highlighted the great heterogeneity of the myDC compartment (40,42). For example, CD1c + myDCs (or cDC2) are comprised of functionally distinct subsets that can be discriminated based on CD14 expression (CD14 lo inflammatory myDCs vs. CD14 -myDCs) (40). Likewise, Yin et al. (42) reported two populations of CD1c + myDCs with distinct gene expression, cytokine production, migration potential, antigen presentation and T-cell polarization profiles, identified based on the expression of CD5 hi vs. CD5 lo . Combining both markers allowed identification of three distinct populations of CD1c + myDCs with the EuroFlow IMC tube: i) CD1c + CD14 lo myDCs, ii) CD1c + CD14 -CD5 myDCsand iii) CD1c + CD14 -CD5 + myDCs. Since both the CD14 lo and CD14 -CD5subsets of CD1c + myDCs have been recently shown to display gene expression patterns closer to monocytes (40,42), further transcriptomics, proteomics and/or functional comparative analyses are required to better understand the relationship among these subsets.
Classical gating strategies for pDCs identification have been associated with cross-contamination with the recently described Axl + DCs (40). As these cells show myDC and pDC mixed transcriptomic and functional profiles, this could lead to potentially inaccurate data interpretation (40). Here we identified CD303 + Axl + DC vs. pDCs and myDCs, in the absence of an anti-Axl antibody, based on a distinctive immunophenotypic profile (HLA-DR + CD33 lo CD141 + CD303 lo ). This CD303 lo Axl + DC population also showed unique functional features both at steady-state and in response to LPS. As described by Villani et al (40), the Axl + DCs, here identified employing the above-mentioned combination, displayed higher CD86 and CD5 baseline expression vs. pDCs and produced IL6, IL8 and TNFa in response to TLR4 stimulation, with an intermediate degree of response between pDCs and CD1c + myDCs, further supporting that Axl + DCs can be identified based on the HLA-DR + CD33 lo CD141 + CD303 lo phenotype. In addition to the pDC-like Axl + DCs (CD11c -/lo , CD123 + , Axl + ), another Axl-expressing DC population has been reported in the literature (CD11c + CD123 lo Axl + DCs), which exhibits an immunophenotypic profile (CD11c + CD14 -CD5 + ) (25, 40) similar to CD1c + CD14 -CD5 + myDCs. In line with this, both populations have also been reported to induce strong CD4 + T-cell proliferation (40,42), suggesting that these two DC populations might be (at least partly) overlapping subsets. Further studies are required to confirm these observations. While the nature of the myDCs precursor in PB is still a matter of debate (31,40,43), a population defined by a CD100 hi CD34 int phenotype, ability to proliferate and differentiate into CD1c + myDCs and CD141 + myDCs has been reported (40). Remarkably, CD100 was not critically required for its identification since the HLA-DR hi CD34 int phenotype showed a high discrimination power vs. other CD34 + cells. Interestingly, several recently described preDC populations, based on different antibody combinations, show significant overlapping features. For example, CD45RA + CD33 + CD123 + HLA-DR + preDCs described by See et al. (31) in fact correspond to Axl + DC as proposed by Villani et al. (40) Altogether, these findings highlight the need for a standardized nomenclature of IMC populations for more direct comparison of data derived from different panels and studies.
Discrimination of M-MDSCs from cMos frequently depends solely on the pattern of expression of HLA-DR, which ultimately requires FMO or internal controls to set the gates for their arbitrary identification (32). While several studies have reported markers with the potential to improve the discrimination from cMos (e.g., CD64, CD86, CD124, CD163, S100A9) (32,81), no comprehensive evaluation of the expression of high numbers (n>30) of proteins in cMos vs. M-MDSCs has been previously performed. In line with earlier reports (32,81,82), a trend for lower expression of CD32, CD64, CD86 and CD163 and increased expression of CD124 and S100A9 was observed in M-MDSCs vs. cMos. Despite this, only CD16, CD123 and CD192 showed overall statistically significant different expression in M-MDSCs vs. cMos. This might be due to the fact that normal CB and healthy adult PB samples were tested in our study, whereas other reports evaluated these markers in cancer, infection and/or inflammatory conditions (81,82), that can potentially lead to more pronounced distinct phenotypes. Multivariate analysis further revealed that only CD192 was of additional value for discrimination of the two populations and therefore, only this marker was included in the extended version of the EuroFlow IMC tube. Interestingly, when CD192 was used, a significantly higher frequency of M-MDSC was observed in CB vs. adult PB, a pattern previously reported for PMN-MDSCs but not M-MDSCs (78), suggesting that the more restricted CD14 + , HLA-DR -/lo , CD192 -/lo phenotype could potentially more accurately identify CD14 + HLA-DR -/lo M-MDSCs. Further Tcell proliferation inhibition assays are required to confirm this hypothesis. Based on all the above, we can conclude that the number of markers required to identify all distinct target populations of IMC was optimized in the EuroFlow IMC combinations.
For increased flexibility, two versions of the EuroFlow IMC tube were designed. A more limited, smaller 11-color antibody combination (13 antibodies), aimed for the clinical setting, in which available IVD-certified instruments frequently have the ability to detect fewer parameters, and an extended 14-color tube (16 antibodies), that further allows identification of less frequent and/or more recently discovered IMC populations (e.g., M-MDSCs and preDCs), mostly aimed at the discovery/research settings, in which instruments allowing simultaneous detection of >12 colors are more commonly available.
In line with previous reports (66, 83), both antibody combinations can be used in EDTA vs. sodium heparin anticoagulated samples, although slightly lower counts of CD1c + CD14 lo myDCs might be detected in heparin samples. Similarly, no significant impact on the overall staining patterns and individual marker resolution was observed for samples stored at RT for up to 24h prior to staining, except for lower CD16 and Slan levels, according to previously reported findings for CD16 (66). However, an increasing time lapse between sample collection and sample processing had a significant impact on the absolute counts of specific IMC populations, already at >12 hours and particularly at ≥ 24h, when >60% of all IMC populations evaluated exhibited some degree of altered (>10% variation vs. 0h) cell counts, in line with previous studies (66, 77). However, it should be noted that delayed sample preparation mainly affected infrequent populations (e.g., Axl + DCs, CD1c + CD14 dim myDCs, CD141 + myDCs), leading to an overestimation of their counts, which might be due to the lower viability of more frequent populations, as supported by an increased percentage of cell debris, particularly at 24h. Conversely, underestimation of populations of (particularly CD36 -) ncMos was observed after 12h, probably because ncMos have been reported to be more prone to spontaneous apoptosis (84). Interestingly, CD62L -cMos were more sensitive to delayed processing than CD62L + cMos. Downregulation of CD62L by mechanisms such as cleavage from the cell surface membrane has been shown in apoptotic mature neutrophils (85). A similar process might occur in monocytes. Of note, our time course experiments were performed at RT, aiming at mimicking transportation of the samples between centers. However, the performance of the EuroFlow IMC tube could be improved by storage/transportation of samples at 4°C in sodium heparin-anticoagulated tubes, as recommended by Diks et al. (66) who reported good stability of major myeloid populations up to 24h under these conditions. A frequently employed alternative approach for the study of samples that cannot be evaluated within a short period upon collection is freezing. However, while the overall staining resolution of samples with the EuroFlow IMM combination was not significantly affected by the freezing process, and still allowed for identification of all IMC populations present in the sample, a clear impact on the relative frequency of populations was observed. Overall, this suggests, that despite the combination can be employed for characterization of frozen PBMCs, in the context of comparison of samples processed with the same method, the interpretation and reporting of the results on relative frequency of populations should consider the bias vs. freshly obtained samples induced by freezing procedure.
Further evaluation of the EuroFlow IMC tube showed a very good reproducibility both in single center, multi-instrument, and multi-center settings. Of note, the highly comparable results obtained in conventional vs. spectral instruments support the possibility of employing the EuroFlow IMC tube as a basis for e x p a n s i o n wi t h ad d i t i o n a l ap p l i c a t i o n -d e p e n d e n t characterization markers, when high-end (>20 colors) instruments are used in a research setting. Noteworthy, as the frequency of some of the IMC population can be as low as 0.1 cells/mL in healthy donors, to reliably and reproducibility identify and quantify these populations also in situations in which a significantly decreased frequency is observed, staining of 10 7 cells is recommended. While the EuroFlow IMC combination can be employed for processing of lower numbers of cells, in case of limited sample availability, the limits of detection (LOD) and quantitation (LOQ) (≥30 and ≥50 events to define a cell population, respectively) should be taken into account for data analysis and reporting.
A high correlation between automated vs. expert-based manual analysis was observed for population identification and quantification, in line with previous reports (56, 57, 86). The higher reproducibility observed for repeated databaseguided AGI procedures vs. expert-based manual analysis, together with the faster (approximately 5min vs. 20min for analysis of one sample, respectively) and less labor-intensive features of AGI, further support the potential of database-guided automated analysis to reduce operator-related variability and allow for more efficient and reproducible data analysis. These features become particularly relevant in the diagnostic clinical setting and in cases where a high number of parameters and/or IMC populations are investigated (57, 86, 87). Interestingly, less than optimal performance observed for database-guided automated analysis was restricted to the analysis of minimally represented IMC populations (<0.05% of all leukocytes) close to the limit of quantification (LOQ) of the tube, and populations defined by a limited number of gating markers with heterogeneous expression patterns (e.g., ncMo or cMo subsets). Improvement of the performance of the databaseguided analytical procedures might be potentially achieved by staining and acquisition of higher numbers of cells (e.g., 10 million) and fine-tuning of Wanderlust trajectory-based automated gating on heterogenous markers (50,86).
The frequency of IMC populations has been previously shown to be modulated throughout life (37). Therefore, knowledge of the normal age-related distribution of the populations is crucial for clinical translation of the data. Overall, three major patterns were observed for the absolute counts of PB IMC populations in relation with age: i) stable cell counts, ii) modulation during adolescence and iii) changes in older (>55y) adults. Since the distribution of several immune cell populations has been reported to occur within the first 2 years of life most prominently (37, 86), it is possible that earlier kinetic changes in populations have been missed, as our cohort only includes children >8 years. Further inclusion of samples from younger infants would allow for applicability of the reference values in the pediatric settings. Overall, several populations displayed clear kinetics around adolescence (10-17y) (e.g., pDCs, Axl + DCs and CD36 -/Slan -ncMos), most likely associated with the physiological changes observed in puberty (e.g., hormonal variations and increased tissue remodeling). In turn, other IMC populations showed modulation in older adults (e.g. eosinophils, immature neutrophils, CD62L + FceRI -cMos and CD36 + Slan + ncMos), potentially as a result of a skewing of hematopoiesis towards myeloid vs. lymphoid lineages, decrease in the function of neutrophil, monocytes and DCs and possibly also low-grade inflammation also known as "inflamm-aging" (88,89).
In contrast to age, limited sex-related differences were observed, except for the more mature neutrophils (more frequent in women), immature neutrophils and CD62L -cMos (more frequent in men), similarly to what has been previously reported for neutrophils and lymphocytes (90, 91).
In summary, we developed two standardized, and highly reproducible versions of the EuroFlow IMC tube, which are suitable for clinical and research/discovery studies, even in multi-instrument and multi-center settings, allowing for robust and accurate identification and quantitation of 19 to 23 IMC populations in blood. By addressing distinct (i.e., pre-analytical, analytical and post-analytical) variables that might impact the reproducibility of laboratory testing, and providing normal ageand sex-related reference ranges, our study sets the basis for standardized immune-monitoring of IMC in distinct disease and treatment conditions, in the context of clinical trials and/or patient care such as in inflammatory diseases, various forms of tissue damage as well as for monitoring immune responses to infectious diseases, vaccination or immunotherapy.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The studies involving human participants were reviewed and approved by Medical Research Ethics Committees United