High-Sensitivity Assessment of Phagocytosis by Persistent Association-Based Normalization

Phagocytosis is an ancient receptor-driven cellular process that involves the engulfment of a particular prey by a cell with phagocytic capability (1). In metazoans, it is essential for the immune system as well as for homeostasis by cleaning up cellular debris and waste (2). As it is such a ubiquitous process, it is natural that many research fields study or use phagocytosis as a functional readout, and ever since Metschnikoff discovered the phenomenon (3), researchers have come up with different ways to assess and quantify phagocytosis.

Phagocytosis has typically been measured either through direct observation with electron or light microscopy or through various indirect methods in which the prey is labeled with a marker such as radiation (4) or fluorescence (5). Several approaches are able to discriminate between prey that are on the outside or the inside of the phagocyte, typically by differential double labeling (6, 7), quenching (8), or by fluorescent color change induced by the intracellular environment (9). Arguably, the most common instrument used today is the flow cytometer, in which fluorescently labeled prey and phagocytes can be quantified in a high-throughput manner (10) and sometimes in combination with fluorescence microscopy for qualitative confirmation (11). Imaging flow cytometry combines the high-throughput benefits of flow cytometry with the ability to do single-cell image analysis and provides an option for detailed analysis of phagocytosis (12). Another common approach is the CFU-based assay (13, 14), in which phagocytosis is indirectly assessed through phagocyte-mediated killing and typically employed when the question is more focused on the outcome of the phagocytic process than the process itself.

Ideally, we would like to be able to draw general conclusions about phagocytosis and compare results across different biological systems. We know that phagocyte characteristics differ significantly, including for the two most common human ones, neutrophils and macrophages (15). We also know that prey characteristics, such as shape and size (16), as well as surface properties, including opsonization density (17) and localization, vary even more (18, 19) and that the microenvironment plays a vital part in changing the prey characteristics (20). Besides the variability that biology will impart, there are also well-known physical and experimental factors that impact phagocytosis assays (21), including but not limited to temperature, time, volume, and the ratio of prey to phagocyte. Independent from method principle and regardless of the specific assay, there is currently no standardized way to reduce the impact of experimental factors. There is also no standard way to define a phagocytic index or to describe phagocytosis assays. This lack of standards, including inconsistent terminology, leads to difficulty comparing across experiments, experimental systems, and laboratories, ultimately affecting assay reproducibility and sensitivity. Besides the obvious benefits of higher quality data, more reproducible, higher sensitivity assessment of phagocytosis would allow for more detailed mechanistic questions regarding this ancient process.

We hypothesize that the determination of the actively prey-associated proportion of phagocytes allows us to normalize for the many factors causing experimental variation and, in the process, increase the measurable biological signal (Fig. 1A). If we regard the phagocytes and prey as particles that can react with each other, we can infer from collision theory that the critical event is a persistent collision event, in which two particles interact in such a way that a reaction occurs and they remain together (21, 22). Association between prey and phagocyte can then be described in terms of a persistent association (PA) event, and the level of the association will be related to collision frequency (Fig. 1A). During PA, the phagocyte can either have prey adhered to the surface, prey that have been internalized, or both. The type of prey–phagocyte interaction is not considered when a PA event is defined. If a similar level of association across various experimental conditions results in a similar level of phagocytosis, this shows that normalization is possible for experimental variation as long as the degree of prey–phagocyte association can be determined.

In this study, we present a systematic approach to assess phagocytosis, irrespective of the chosen experimental method. From collision theory, which concerns the likelihood that two particles collide and react, we can conclude that the phagocytic interaction step will have a considerable impact on experimental variation. We show that this can be normalized by employing dose-response curve analysis and comparing conditions at the same level of persistent prey–phagocyte association. Such normalization leads to increased assay robustness, reproducibility, and sensitivity. PA normalization also allows for comparison across completely different experimental settings and laboratory conditions. Finally, we provide experimental tools and guidelines for different degrees of phagocytosis assessment so that the researcher can match the method to the question being asked.

Staphylococcus aureus Cowan-1 and Escherichia coli DH5alpha (Invitrogen) were cultured under shaking conditions at 37°C with 5% CO₂ atmosphere in Todd Hewitt Broth (Bacto Laboratories) complemented with 0.2% (w/v) yeast (BD Biosciences) or Luria–Bertani (Sigma-Aldrich), respectively. The bacteria were harvested at OD₆₂₀ = 0.4.

THP-1 cells (23) (TIB-202, male; American Type Culture Collection) were maintained in RPMI 1640 medium (Sigma-Aldrich) supplemented with 10% FBS (Life Technologies), 1% penicillin–streptomycin (Thermo Fisher Scientific) and 2 mM GlutaMAX (Life Technologies). Cells were grown at 37°C in a 5% CO₂ atmosphere. The cell density was kept between 0.2 and 1.0 × 10⁶ cells/ml, with a viability over 95% and were kept up to 3 mo before thawing a new aliquot of fresh cells.

Flash Red–labeled 1-μm microspheres coated with streptavidin (catalog number CFFR004; Polysciences Europe, Bangs Laboratories) were washed three times in 1 ml of Na medium (5.6 mM glucose, 127 mM NaCl, 10.8 mM KCl, 2.4 mM KH₂PO₄, 1.6 mM MgSO₄, 10 mM HEPES, 1.8 mM CaCl₂; pH adjusted to 7.3 with NaOH) at 20,000 × g for 2 min and kept protected from light at 8°C prior to opsonization. All chemicals were from Sigma-Aldrich.

After harvesting the bacteria, they were washed three times with 1 ml of Na medium at 2000 × g for 2–5 min. Heat killing was induced at 80°C for 5 min in a heat block, vigorously shaking, followed by cooling the bacteria rapidly on ice. Bacteria not used directly were stored in a refrigerator up to 2 wk.

Heat-killed bacteria were stained at 37°C with gentle shaking protected from light for 30 min, either with 2 μg/ml DyLight650 (Thermo Fisher Scientific) or 5 μM Oregon Green 488-X succinimidyl ester (Invitrogen), depending on whether fixation of phagocytosis was planned; if so, the former staining was used. The samples were washed once with Na medium (1 min, 5000 × g, swing-out rotor) and, if needed, sonicated for up to 4 min (VialTweeter; Hielscher Ultrasound Technology) to disperse any large aggregates of bacteria that were confirmed by microscopy. The samples were then gently spun down (200 × g, 2 min, swing-out rotor), and only the supernatant was used to avoid any remaining bacterial aggregates.

An i.v. Ig (IVIG; OCTAGAM; Octapharma Plasma) was centrifuged 2 min at ∼20,000 × g to precipitate any aggregates. The prey were opsonized with 0–10 mg/ml IVIG at 37°C with gentle shaking and protected from light for 30 min (Supplemental Fig. 2D). Any unbound IVIG was washed away with 1 ml of Na medium five times (1 min, 5000 × g for bacteria and 20,000 × g for beads, swing-out rotor) for all experiments, except when evaluating the effect IgG density has on phagocytosis.

On the day of experiments, the cell density was kept between 0.5 and 0.7 × 10⁶ cells/ml. Medium was changed to Na medium (5 min, 500 × g, fixed rotor), and cells were kept on ice until the start of the phagocytosis assay. In experiments without fixation, dead cells were labeled with 2 μM impermeable cell membrane dye DRAQ7 (Abcam). In experiments with fixation, cells were instead labeled according to the manufacturer’s protocol with the LIVE/DEAD Fixable Violet Dead Cell Stain Kit for 405 nm excitation (Thermo Fisher Scientific), except when evaluating the effect IgG density has on phagocytosis, when the prey is live E. coli, or the cytochalasin D inhibition test.

The concentration of prey and THP-1 cells was measured prior to phagocytosis by flow cytometry (Accuri C6; BD Biosciences or CytoFLEX; Beckman Coulter); if needed, prey were sonicated in advance. For gating strategies see Supplemental Fig. 4. The concentration of THP-1 cells was set for each experiment and ranged between 1 × 10⁵ and 2 × 10⁶ cells per sample. The cells were added on ice to their prepared samples; at least seven different multiplicities of prey (MOP) with the range 0–200 were prepared for each experiment. Incubation was performed in Na medium at 37°C on a heating block with moderate shaking protected from light. Phagocytosis was halted by either putting the samples on ice (Supplemental Fig. 2A), adding of 1 μM cytochalasin D (Supplemental Fig. 2B, 2C) (Thermo Fisher Scientific) for 15 min, or by fixing them in 1% paraformaldehyde (Thermo Fisher Scientific) for at least 30 min.

If not specified otherwise, the incubation time and volume were 30 min and 150 μl, respectively. Phagocytes and prey were centrifuged (3000 × g, 1 min, including 30 s of acceleration) together prior to incubation when evaluating the effect of centrifugation on association. When possible, 96-well plates, eight-channel multipipet and low-binding pipet tips, and Eppendorf tubes were used.

For cytochalasin D inhibition test, 1 × 10⁵ cells were suspended in 50 μl of Na medium, after which cytochalasin D (0.1 mM) was added to a final concentration of 1, 5, and 10 μM. The cells were then incubated at room temperature for 20 min while shaking. After the incubation, the cells were placed on ice for 10 min before the phagocytosis assay was performed in the same manner as previously described.

Phagocytosis of live E. coli was performed as previously described. Assessment of bacterial association and internalization was done by gating for THP-1 cells, and a further gate selected only single cells. Allophycocyanin-positive events signified cells with associated bacteria (E2-Crimson–expressing E. coli), whereas FITC positive events represented bacteria that had been taken up into acidic intracellular phagosomes (pHrodo Green pH STP Ester).

Imaging flow cytometry was performed in a similar manner as previously described, with the exception that in this study the bacteria and Na medium were added to 6 × 10⁵ cells in a total volume of 600 μl. After phagocytosis and cool down, the mixtures were concentrated by centrifugation for 5 min at 600 × g and 4°C. The supernatants were removed, and the pellet was resuspended in 25 μl of Na medium prior to acquisition.

THP-1 cells were labeled with mouse Ab anti-human CD18 (PE; BioLegend or BV421; BD Biosciences) (1 μl per 200,000 cells) for at least 10 min at room temperature. The extracellular bacteria were labeled with 1:1000 Fab-specific DyLight 488–conjugated AffiniPure F(ab′)2 Fragment Goat Anti-human IgG (Jackson ImmunoResearch Laboratories). Fluorescent Atto 488 nm biotin (0.008 μg/sample; Sigma-Aldrich) was used to label extracellular beads.

For fixed samples, fixation was performed in 1% paraformaldehyde (Thermo Fisher Scientific) for at least 30 min on ice protected from light. Postfixation samples were incubated with 50 mM glycine and 5% BSA for at least 10 min at room temperature, except when evaluating the effect IgG density has on phagocytosis when only glycine was used.

For live prey experiments, DH5alpha E. coli were labeled using the plasmid pUCP20T-E2Crimson (24). pUCP20T-E2Crimson was a gift from Mariette Barbier (Addgene plasmid 78473; http://n2t.net/addgene:78473). They were grown as stated above under constant antibiotic selective pressure (100 μg/ml) to optimize fluorescent protein expression. After reaching the desired OD₆₂₀ the bacteria were washed twice in 1× PBS followed by a final wash step in 1 ml of 0.1 M Na₂CO₃ (pH 9), whereby the bacterial suspension was transferred to a low-binding 1.50-ml tube (Eppendorf). Finally, the pellet was resuspended in 95 μl of Na₂CO₃ and 5 μl of the pH-sensitive fluorescent dye pHrodo Green STP Ester (Invitrogen) was added. The bacteria were incubated at room temperature for 45 min, after which they were washed once in 1× PBS and resuspended in Na medium at a final volume of 1 ml. Bacteria were quantified in a CytoFLEX flow cytometer (Beckman Coulter), and the pHrodo Green staining quality was assessed by acidifying the bacterial suspension with 3 M C₂H₃NaO₂ (pH 6) and subsequently evaluated for a gain in fluorescent signal.

For imaging flow cytometry, heat-killed bacteria were stained with 3 μM Oregon Green 488-X succinimidyl ester for 30 min in the dark at 37°C while rotating. After the first staining, the bacteria were washed once with Na medium (10 min, 4000 × g) and resuspended in 1 ml Na2CO3 (pH 9). In a second staining step, 20 μl of 1 μg/μl CypHer5E NHS Esters (Cytiva) was added to the bacterial suspension followed by an incubation at room temperature for 2 h while being protected from light and rotating. After the second staining, the bacteria were washed two times with Na medium (10 min, 4000 × g) and resuspended to a final volume of 1 ml of Na medium.

Flow cytometric acquisition was performed using three different flow cytometers: Accuri C6 (BD Biosciences) equipped with 640- and 488-nm lasers, CytoFLEX (Beckman Coulter) with 405-, 488-, and 638-nm with 450/45 PB450 and 525/40 KO25, respectively, 525/40 FITC, 585/42 PE, and 660/10 allophycocyanin, 780/60 allophycocyanin–A750 filters; and FACSVerse (BD Biosciences) with 640- and 488-nm lasers with filters 527/32 FITC, 586/42 PE, 660/10 allophycocyanin, and 783/56 allophycocyanin–Cy7. For each experiment threshold, gain and velocity was set. Acquisition was set to analyze at least 5000 events of the target population. ImageStream (EMD Millipore) acquisition was done using 405-, 488-, and 642-nm brightfield, and side scatter was done with 785-nm lasers using 60× magnification and low fluidics setting.

Flow cytometry data were analyzed using FlowJo version 10.2 (Tree Star). Compensation was performed using inbuilt matrices in the CytoFLEX or FACSVerse when needed. The general gating strategy was similar for all experiments (see Supplemental Fig. 3). Dead cells were excluded by being extremely positive (a population at least two logs higher than all other signals) for DRAQ7 or for the LIVE/DEAD violet signal. Doublets were excluded by gating on forward scatter height versus forward scatter area. THP-1 cells were gated on forward and side scatter and CD18 when used. We defined association as a THP-1 cell being positive fluorescence corresponding to a prey, Oregon Green, DyLight 650, or Flash Red.

Imaging flow cytometry data were analyzed using IDEAS Software (Amnis) with the gating strategy as described in Fig. 2.

To aid analysis using PA-based normalization (PAN), we have provided a GraphPad Prism template that automatically makes calculations and graphs (see Supplemental Fig. 3 for a guide to the template). The user provides observed PA values and mode fluorescence intensity values from method of choice.

Association, the percentage of phagocytes positive for at least one prey, was plotted versus MOP, resulting in curves that were analyzed using a Prism 8.2.1 (GraphPad) inbuilt nonlinear regression analysis tool named “Agonist vs. response – Variable slope (four parameters) or Find ECanything.” Least squares regression was used with no special handling for outliers and no weighting. Parameters of the curves were constrained as follows: bottom, more than 0; top, between 0 and 100; and EC₅₀, larger than 0. If technical replicates were used, the mean was used to generate one curve; for each replicate, one curve was generated, but the experiment is visualized with one average for the whole experiment. The characteristics of the curves are generated when the curve is created, except for (AUC), which is based on the prey median fluorescence intensity (MFI) for the whole MOP range. When comparing different prey or labeling techniques, the signal was first normalized by dividing it by the signal for a single prey and then compared at the same MOP range.

The fluorescent intensity for association and adhesion was plotted separately versus MOP expressed in mean, median, or mode. Normalization was performed by interpolating the MFI corresponding to the MOP, which generated by the above mentioned analysis of half of the maximal MOP response (MOP₅₀). Interpolation was done depending on the best fit, either by nonlinear regression or linear regression between the two MOPs that were between MOP₅₀ or when needed with a manual read out.

The fluorescent signal of a prey unit was determined by measuring the single prey unit by flow cytometry. This could then be used to determine the ratio between the fluorescent staining, thus converting the adhesion signal to the same unit as the association signal by multiplying it with the ratio, allowing us to estimate the internalized signal through subtraction when no intracellular marker was used. Thus, the total prey signal subtracted by the adhered prey signal results in internalized prey signal. The number of prey per phagocyte (PxP), could be quantified by dividing each signal with the corresponding prey unit signal. Observe that one prey unit may not always correspond to one single prey; depending on the type of prey, a unit can be singlets or include multiple prey because some bacteria prefer to form chains, whereas others and beads are more prone to be singlets.

The statistical details of the experiments can be found in each figure legend; mean and SD were used if not otherwise specified. All statistical analysis performed in this study was done using Prism 8.2.1 (GraphPad). Correlation was evaluated using nonparametric Spearman, two-tailed with α = 0.05, and nonlinear regression was performed using semilog fit with the mean of the y-replicates. The sensitivity was quantified for each assessment type by dividing each different IgG density signal with the corresponding background; in this study, the nonopsonized samples were set as background. An average sensitivity of four experiments was plotted over the range MOP 10–150 for each IgG concentration and type of assessment; for MOP₅₀, instead of the MOP range, the distribution of each experiment was plotted.

The GitHub page with up-to-date templates is provided through the laboratory homepage link to Quantitative Immunobiology (https://nordlab.med.lu.se/?page_id=34). The templates provide a simple way to generate phagocytosis analysis data. The PA data and the corresponding prey signal are entered in the template, which then calculates MOP₅₀ and interpolates at MOP₅₀ the PxP by performing a linear regression between the two closest data points to MOP₅₀. All the data are plotted in predefined graphs.

Our aim was to establish a robust and standardized method to assess phagocytosis with high sensitivity. The method should also be able to discriminate between association, adhesion, and internalization. To achieve this, the method had to normalize for experimental factors modulating effective association (following a collision) between phagocyte and prey. If a phagocytosis experiment could be regarded as a reaction vessel with two reactants, collision theory describes the likelihood for a reaction to occur, based on collision frequency and what parameters it depends on. For phagocytosis, we define such a reaction as a PA, with the critical parameters being reaction volume, reactant concentration, and reaction time (Fig. 1A). To have terminology that is relatable to phagocytosis, we introduce the term MOP to describe the ratio of particulate prey to phagocytes. The term MOP followed by a subindex (MOP_X) is to indicate the effective MOP to reach X percentage of PA, which is the result of an effective collision and is similar to EC₅₀ used in pharmacological studies (25). Finally, the term PxP is used to indicate number of prey units per phagocyte. This results in the following equation to describe MOP₅₀, which is a function of prey concentration, [MOP]:

where PA is the observed PA, PA_max is the maximal PA at high prey concentrations, and where n^H, the Hill coefficient (26), describes the slope of the curve. The Hill coefficient is an indication of cooperativity in a binding process, in which a number above 1 indicates positive cooperativity.

To analyze how PA varies, we systematically tested the most important parameters derived from collision theory: volume, MOP, and time. To reduce the number of variables, we opted to use the THP-1 cell line, which is commonly used to measure phagocytosis, for all experiments in this study, (27). They have a consistent expression of Fc receptors when kept at certain cell densities [<0.5 million cells/ml (10)]. As prey, we used Oregon Green–labeled S. aureus opsonized with pooled polyclonal human IgG (IVIG). The results are summarized as heat maps in Fig. 1B and show that several variables have potentially dramatic effects on the outcome. A few individual experiments are highlighted with their corresponding flow cytometry scatter plots and quantified data. The effect of changing volume is clear at MOP 50, where the PA increases from 30 to 70% when decreasing the volume. Similar for time, which is well-known (28), at MOP 20, increasing incubation time results in a doubling of the PA from 34 to 60%. Not surprisingly, the clearest effect can be found when phagocytes and prey are centrifuged together prior to incubation. At MOP 2, the interaction increases several fold.

To evaluate how the dynamic range was affected by experimental factors we plotted volume with respect to MOP in linear-logarithmic (lin-log) dose-response curves and compared PA and corresponding cell-associated prey (Fig. 1C, 1D). The percentage of phagocytes that persistently associated with prey resulted in a sigmoidal relationship centered around MOP₅₀ (Fig. 1C), and this corresponded to a region with a linear relationship in terms of prey that associate with phagocytes (Fig. 1D). This trend could also clearly be seen when analyzing the whole range of MOPs and volumes, visualized as heat maps in Fig. 1C and 1D. Because the dose-response relationship was consistent across such a large range of conditions, we postulated that interaction differences due to nonbiological experimental factors, such as volume, could be normalized by using the characteristics of these types of curves.

In Fig. 1E, we summarize our proposed new way of assessing phagocytosis, PAN, which works by normalizing interaction effects estimated through persistent prey–phagocyte association. In principle, a set number of phagocytes are incubated with a range of prey concentrations (MOPs). Data are acquired using any method that can distinguish between noninteracting and interacting phagocytes, such as flow cytometry, imaging, or other methods of choice. A lin-log function is then fitted to the data to estimate at what MOP half of the maximal PA is evoked (MOP₅₀). Subsequently, the rest of the analyses are performed at MOP₅₀ where PA-modulating experimental factors are normalized. This approach will allow for normalized assessment of phagocyte association with prey and can also include analysis of both adhesion and internalization frequencies.

To validate that an increase in PA is coupled with phagocytosis we labeled E. coli with Oregon Green and a pH-sensitive dye (CypHer-5E) to be able to detect internalized prey. The data were acquired using an imaging flow cytometer to allow us to view images of single cells in the selected cell populations. The gating strategy for selecting persistently associating cells and cells that had internalized bacteria is shown in Fig. 2A. Images from indicated cells are shown to confirm that the gating is correct (Fig. 2B). PA analysis shows an expected increase in phagocytes that become associated (associating gate in Fig. 2A) with bacteria as the MOP is increased (Fig. 2C). This association is coupled with a linear increase (R² = 0.99) in associated green bacteria (Fig. 2D), whereas the subpopulation of cells that have completed internalization of at least one green bacterium (internalizing gate in Fig. 2A) are associated with overall more bacteria. However, the linearity is not as clear (R² = 0.81), when only analyzing internalizing cells (Fig. 2D).

To compare PA with internalization, we analyzed the internalizing subpopulation of phagocytes as measured by the pH-sensitive dye (Fig. 2E). The proportion of cells that had internalized at least one bacterium (red signal) followed a similar curve as for associating cells, whereby not all cells were able to complete internalization (PA_max = 80.3%; Fig. 2E). The red fluorescence intensity was linearly associated with the percentage of internalizing cells (R² = 0.97; Fig. 2F), correlating well with the number of green bacteria in persistently associating cells (Fig. 2D). This indicates that using the total prey-associating population (green bacteria signal in this example) for MOP₅₀ analysis will better reflect the overall phagocytosis trend rather than restricting it to cells that have completed internalization of at least one prey.

To evaluate whether we could use MOP₅₀ to normalize the effect of experimental factors on collision frequency, we designed an experiment in which we could look at the change in PA separated from other factors. Given that the biology should be unaffected, we hypothesized that by changing experimental volume and nothing else, comparison at MOP₅₀ should give the same PxP (see Fig. 3A).

Fig. 3B shows the theoretical PA outcome of increasing (right shift) or decreasing (left shift) volume from the starting volume. The experimental data show similar results to the predicted outcome (Fig. 3C), with the association curves moving left or right, depending on the corresponding change in volume. The corresponding shifts in MOP₅₀ are also indicated in the Fig. 3C. The prey signal analysis shows a large range in observed prey signal per phagocyte, with up to a 6-fold difference (Fig. 3D), which converges to very similar levels after applying MOP₅₀-based normalization (Fig. 3E). Taken together, this suggests that experimental factors that alter relative collision frequency can be normalized across experiments by estimating MOP₅₀.

To address the question of whether normalizing for PA is an improvement for assay sensitivity and robustness, we wanted to compare the PAN method to the current state of the art. However, there are many ways to analyze phagocytosis, and there is no single gold standard. We did a meta-analysis based on a random selection of 100 articles from 2013 to 2018 that included phagocytosis assessment of their studies (Supplemental Fig. 1). We found that we could group the methods into one indirect approach and three bona fide phagocytosis assessment levels (Fig. 4E; levels A–D, Table I) and then proceeded to compare our method with them. The first level, A, represents whole phagocytic population analysis, whereas the second level, B, distinguishes between interacting and noninteracting phagocytes. The third level, C, in addition to the second level, separates adhesion from the internalization of the prey. We place the approach described in this study as a fourth level, D, where the usage of MOP₅₀ estimation to normalize for experimental factors can improve the detection of differences in association, adhesion, and internalization.

To compare these levels of phagocytosis assessment and evaluate whether MOP₅₀-based normalization provides any benefits, we analyzed the same data set according to the different levels. In this study, we kept the experimental parameters the same, and we varied a biological one, IgG density. Fluorescent streptavidin beads (Far-Red) were opsonized with 0–1 × 10³ μg/ml IgG, and after phagocytosis, the extracellular beads were coated with fluorescent Alexa 488 Biotin (green). The data were acquired through flow cytometry, in which adhered beads were detected in both Far-Red and green channels, whereas internalized only in Far-Red.

In Fig. 4A, we show how the prey signal varies with IgG density, depending on which population of phagocytes is being analyzed. When all cells are included, there is almost no difference across the IgG densities, whereas more information is resolved when the analysis is focused on persistently associated phagocytes as well as the further subdivision into adhered and internalized prey signal. This indicates that phagocytosis assessment gains sensitivity by increasing the level of analysis as described in Table I.

We hypothesize that the biggest gain in sensitivity will come from MOP₅₀ normalization. To exemplify the large effect MOP has on PA, we compare two common MOPs used in the literature, 10 versus 100. At MOP 10, a clear difference in the association can be seen, and association increases with IgG density. This is nullified when the cells are overloaded at MOP 100 (Fig. 4B), resulting in the same range of PA for all IgG densities. Also, the resulting associated prey data are not consistent with expected results and vary depending on whether MOP 10 or MOP 100 is used. Consequently, if the analysis is performed at a fixed MOP independent of condition, which is most often the case in the literature, this will affect assay sensitivity. If the phagocyte association is instead normalized to a comparable MOP, the internalized prey signal increases (Fig. 4C). To quantify the sensitivity of different types of analysis, we looked at the detection level of the analysis, both at different MOPs as well as when using MOP₅₀ normalization. Because we knew that an increase in IgG opsonization should result in an increase in prey association, we defined sensitivity as a fold increase in prey association as compared with the baseline association of nonopsonized beads (Fig. 4D). This resulted in very low sensitivity for level A and B analysis, with the only significant signal being detected at the highest degree of opsonization and only at one selected MOP, which happened to be close to MOP₅₀. Level C analysis worked well when the selected MOP was close to MOP₅₀ but not at all when it was closer to MOP₉₀. Additionally, MOP₅₀-based analysis had a significantly higher sensitivity at all conditions, making it possible to already detect a signal at the lowest degree of opsonization (1 μg/ml). To summarize, the PA-normalized method represents a clear improvement compared with other standard methods in the field, with increased robustness and assay sensitivity.

Determining association curves provides additional possibilities besides the increased robustness and sensitivity (Fig. 5A). The association curves can also give information about what to expect at low or high MOP situations, such as when few (MOP₁₀) or many (MOP₉₀) of the phagocytes are persistently associated with prey. Additionally, the saturation value of the curve (PA_max), and the slope of the curve (Hill coefficient) is related to cooperative effects (26). This can reveal biological differences of phagocytosis between conditions because the PAN approach normalizes for physical parameters. In this study, we apply these types of analyses to the simple scenario of changing the IgG density on the surface of prey during phagocytosis.

The relationship between opsonin density and Fc receptor-based internalization is not completely clear, with conflicting results in the literature (17, 29, 30), and the effect on PA has not been reported at all. Fig. 5B and 5C and Supplemental Fig. 2E show PA curves, prey association with phagocytes, and examples of flow cytometry scatter plots, respectively, at different MOPs. This data show an expected increase in association for both prey and phagocytes, which we analyze more carefully in Fig. 5D, based on curve parameters. The top row indicates that the maximum association value is barely affected by IgG density, whereas the curve slope (n^H) gets flatter with higher IgG density, and the overall amount of prey associated with phagocytes is only significantly increased by increasing IgG density (AUC panel). The middle row explores at what MOPs there are 10, 50, or 90% persistently associated phagocytes. In this study, the trend looks similar at the lower association scenarios, with expectedly increased IgG density leading to lower MOPs required for reaching a given PA. At 90% association, the IgG density appears to have little effect on the outcome. In the lower panel, we analyzed the number of PxP at MOP₁₀, MOP₅₀, and, MOP₉₀, which showed similar levels for each degree of association. There are small differences within each group, but in general, the PxP seems more dependent on association than on IgG density. Overall, the association curve characteristics indicate that the primary effect of increased IgG density lies in a higher PA, not primarily affecting the fate of prey that has been associated with phagocytes.

To further explore the effects of prey IgG density, we analyzed PA and the fate of associated prey (Fig. 5E). First, we looked at the probability of interaction with at least one prey, phagocytic interaction, which exponentially increases with IgG density (R² = 0.98). That means, that to evoke 50% of maximal PA, only one third of the number of beads was needed when opsonized with 1000 μg/ml compared with 1 μg/ml. However, the average number of associated beads per interacting phagocyte was two to three beads independent of IgG density (Fig. 5E). Still, internalization appears to increase exponentially as well but not as strongly as phagocytic association (R² = 0.68). From 1 to 1000 μg/ml, the average number of beads internalized increased 2-fold. As expected, adhesion had an inverted relationship compared with internalization. Overall, our findings indicate that the association capacity of a phagocyte population increases exponentially with IgG density combined with a smaller but increased internalization capacity of prey-associated phagocytes.

A conclusion from this study is that it would normally be very difficult to compare phagocytosis across different types of prey, including species and strains if they are expected to have an effect on phagocyte association. However, by normalizing for PA using the PAN method, it would be possible to actually compare quite different types of prey in a quantitative manner. In this study we compared the Gram-positive cluster-forming cocci S. aureus with the Gram-negative bacilli E. coli and streptavidin-coated beads, all opsonized with saturation >1 mg/ml IgG. To facilitate the comparison, we used heat-killed prey when comparing across prey, but we also compared phagocytosis of live versus heat-killed E. coli.

Phagocytosis was markedly higher when using heat-killed bacteria as compared with live E. coli. The PA was increased with a shift in MOP₅₀ from 31.7 to 5.3 and a shift in PA_max from 100 to 59% (Fig. 6A). Interestingly, the Hill coefficient indicated that once a phagocyte had engaged with live bacteria, they were more likely to do so with additional bacteria; this was more pronounced with live bacteria (n^H = 1.08 for heat-killed versus 1.43 for live). The degree of internalization was otherwise similar for those cells that did engage in phagocytosis, with similar fluorescence intensity levels for both heat-killed and live (Fig. 6B).

Overall, phagocytosis appears to be different when changing the prey. The PA curves either have similar slopes but different top levels or a different slope compared with the others (Fig. 6C). The number of associated prey is also different, with the beads at a much lower level than the others (Fig. 6D), even when normalized for potential differences in fluorescence signal between prey. The top right panel in Fig. 6D clearly shows that the overall amount of prey associated with phagocytes is very low for the beads compared with bacteria. This is also consistent with the fact that far more beads are required to reach a similar level of PA, whereas the bacteria behave similarly in terms of MOP for a given association (Fig. 6E). The differences in surface properties between the prey might be a natural reason why beads are not as good a phagocytic target as bacteria. However, interestingly enough, beads have the same maximum association value as E. coli; around 100% of the phagocytes will associate with them at high levels of MOPs, whereas S. aureus is saturated around 80%. Looking at PxP reveals that E. coli are associating better at lower levels of interaction (MOP₁₀ and MOP₅₀), whereas S. aureus are more prone to be associated at MOP₉₀ than the other prey. At this point, we do not know what the differences in PA_max, Hill coefficient, or association characteristics mean mechanistically, but we know that these differences are indicative of a biological difference in terms of phagocytosis. This shows that the sensitivity and comparative quantification that PAN-based analysis offers allows for further mechanistic studies or careful quantitative comparisons that might explain inherent differences in phagocytosis across prey.

Phagocytosis is measured as a functional outcome in many fields (Supplemental Fig. 1) and could even be considered a standard assay for many research areas. Despite this, there is no gold standard and no established way of doing a robust comparison across experiments or laboratories. In the present work, we have, to our knowledge, provided a new framework for the measurement of phagocytosis, including theory, terminology, and analysis guidelines and templates (Supplemental Fig. 3). A major benefit to the method we introduce in this study is increased robustness and sensitivity and also the fact that it can be readily implemented into most existing methods to measure phagocytosis. Our hope is that this will allow for both better data when measuring phagocytosis as well as providing the option for cross-experiment and cross-laboratory standardization.

The PAN method is fundamentally based on dose-response curve analysis, commonly applied in a number of fields, particularly in pharmacology (25). There, it has become an essential way of doing analysis, with EC₅₀ or IC₅₀ numbers commonly used to evaluate the pharmacological effect, and has provided an analytical tool to compare cross-laboratory, across different compounds, and over time. The MOP₅₀ used in PAN could potentially be used in a similar fashion. Additionally, translating individual data points into a mathematical function also comes with a number of benefits, where curve characteristics can be quantified and compared. MOP₁₀, MOP₅₀, and MOP₉₀ allow for the exploration of scenarios where phagocytes are encountering various amounts of prey and what the physiological effect could be. The slope, or Hill coefficient, offers intriguing ramifications, where a coefficient different from 1 indicates either positive or negative feedback (26). Moreover, IC₅₀ studies on HIV, in which the effect of resistance mutations have been overlooked because of a lack of slope analysis, highlight the importance of this metric (31). The PAN method is especially valuable when looking at different bacterial mutants or strains that would affect surface properties (32) and thus potentially directly affect PA. A key aspect that should be emphasized is that the MOP curves not only serve as an analysis in of themselves but can also provide a guide to the dynamic range of the system and reveal at which MOP more detailed analysis, such as microscopy, should be performed. This can improve sensitivity and potentially save unnecessary experimental set up time. The PAN method is fully compatible with microscopy-based phagocytosis assays, albeit at a slower throughput than with flow-based techniques but with potentially higher quality data.

A comparative meta-analysis is feasible with a standardized approach. That way, it would be possible to create a database with phagocytosis characteristics for phagocytes and prey from many different sources. This could, for instance, be PxP and MOP₅₀ numbers for different phagocytes and strains of bacteria under various opsonizing conditions. Such a database could then be used for the mining of phagocytosis data to identify important preclinical as well as clinical data. Systems with limited, sensitive data and potentially high heterogeneity, such as different clinical isolates and patient material, would benefit not only from increased assay sensitivity but also from generating comparable results that are stable over time and that could be preserved in such a database. In the infection medicine field, preclinical and clinical researchers come together for the important undertaking to develop alternative therapeutics to antibiotics. Identifying functional mAbs that can be used to treat infections would be an obvious use for the PAN method.

In summary, we have established a universal analytical method that can be used across different systems to normalize for factors affecting the association between phagocyte and prey, in which adhesion and internalization can be analyzed separately with improved robustness and assay sensitivity.

We acknowledge the Oonagh Shannon Lab for help and access to their flow cytometer.

The authors have no financial conflicts of interest.