1. Introduction
Antibody-dependent cell-mediated cytotoxicity (ADCC) by natural killer (NK) lymphocytes is a form of long-lasting immunity that provides protection against many viruses (reviewed, [
1]). Antibodies that support ADCC (IgG1 and IgG3 [
2,
3]) are sustained throughout an individual’s lifetime. IgG1 antibodies will support killing at concentrations as low as 0.1 ng/ml [
4], indicating that ADCC will persist even as antibody levels decrease with time. Effective antibodies include those that are unable to neutralize viral infectivity [
5]. These antibodies recognize many viral epitopes in addition to those of viral receptor-binding domains (RBDs) that are essential for viral entry into cells [
6]. The NK cells responsible for ADCC are abundant, circulate throughout the body, and constitute 10–20% of all the blood lymphocytes. CD16A expressed by NK cells will recognize the Fc region of IgG antibodies attached to viral proteins in the plasma membranes of infected cells. ADCC by NK cells kills immediately, unlike cytotoxic memory T cells that need a recall response before they can effectively kill [
7,
8]. In light of the potency of ADCC, quantification of ADCC-supportive antibodies is desirable to assess the strength of long-term anti-viral protection.
Straightforward quantification of antibody is unable to predict the ability of antibodies to support ADCC because of post-translational modifications of the Fc [
9,
10,
11,
12]. In particular, fucosylation of the antibody Fc reduces its affinity for CD16A [
13,
14] and thereby reduces the ability of fucosylated antibodies to support ADCC [
4,
15,
16]. Intra-donor variations in afucosylation range between 2% and over 24% of IgG1 molecules,
e.g., for antibodies to SARS-CoV-2 [
17,
18]. Currently, measurements of afucosylation for antigen-specific antibodies utilize affinity purification of the antibodies and mass-spectrometry; however, these techniques have limited clinical use because they are costly in both money and time. This approach is also impractical due to the limited volumes of clinical serum samples. Simpler and faster assays are needed to titrate the effective antibodies in serum samples. Here we report a fast NK cell-based assay for detection of specific antibodies that support ADCC which is independent of antibody purification and requires only small amounts of antibody.
The assay depends on NK cell externalization of lysosomal associated membrane protein 1 (LAMP-1), also designated as CD107a [
19,
20,
21]. CD107a lines the inner membranes of lysosomes and intracellular granules. In NK and CD8 cytotoxic T cells, the granules contain cytotoxic proteins (perforin and granzymes) that are released during killing [reviewed, [
22]]. During cytotoxic granule release, the granule membrane fuses with, and is incorporated into, the cytotoxic cell’s extracellular plasma membrane. In the process, CD107a becomes externalized [
23]. Externalization of CD107a is a hallmark of receptor engagement of ‘effector’ killer cells with ‘target’ cells for both cytotoxic T cells [
24,
25] and NK cells [
25]. CD107a has been used as a surrogate marker to indicate cytotoxic activity. A previous report of CD107a externalization by NK-92-CD16A cells monitored the development of antibodies towards influenza after vaccination or natural infection [
26]. This assay with microtiter-plated antigens required hours for NK cell externalization and titered rather than quantified the influenza-specific antibodies. Also, as we report here, CD107a externalization has limitations as a measurement of NK cellular activity. At least in the case of ADCC, CD107a is best viewed as an indicator of antibody-CD16A receptor engagement rather than as surrogate marker for ‘target’ cell death.
The sensitivity of CD107a externalization provided the basis for our design of an optimized assay to compare antibodies with post-translational modifications. Our ultimate goal is to monitor the effects of natural fucosylation of human anti-viral antibodies that support ADCC. There are three components to our assay. The first component is an invariant, renewable, and stable source of NK cells expressing CD16A. An NK-92-CD16A tumor cell line [
27,
28] meets this criterion by having the following properties: a) immortality of the parental line NK-92 [
29,
30]; b) excellent cytotoxic activity; and c) the expression of CD16A and GFP as a result of lentiviral transformation. The second component is an internal antibody standard for calibration to control for intra-experimental variations. The standard for reference is the monoclonal antibody (mAb) GA101-WT anti-CD20 that is ~10% afucosylated, comparable to human afucosylation of total circulating IgG1 [
31,
32]. The third component is ‘target’ cells with antigen expression. Here we used the Raji B cell tumor [
33] that expresses consistent CD20 and is a poor target for NK activity in the absence of antibody. In addition, production of the GA101 mAb in genetically modified CHO cells produces glyco-engineered anti-CD20 mAb (GA101-GE) [
34] with a protein identical in sequence with GA101-WT but 50% afucosylated and marketed as a therapeutic antibody Gazyva
®. The 40–50% afucosylation is the typical upper range for naturally occurring antiviral antibodies (30% afucosylated to the spike (S) protein of SARS-CoV-2 [
17,
35]; 40% to HBV and 80% to human CMV [
17]). Since our ultimate application is anti-viral antibodies, we included an evaluation of the effects of heat inactivation and formaldehyde that are used to inactivate viruses.
We describe an improved NK-92 CD16A CD107a-based biological assay for antibodies that will support ADCC. The NK cell CD107a externalization detects lower levels of antibodies than are directly detectable by flow cytometry. To the best of our knowledge, we are the first to quantify the antibody detection and to demonstrate that this cellular CD107a externalization detects differences in antibody fucosylation. We observed unexpected conditions that influenced CD107a externalization: the lower the effector NK cell to Raji target cell ratio (E:T), the greater the CD107a expression. Remarkably, the high E:T conditions that supported killing of Raji cells failed to support CD107a externalization. Overall, we report that the NK CD107a assay is (1) very effective for the measurement of antibodies that will support ADCC and (2) unsuitable as a surrogate assay to monitor target cell death by ADCC.
2. Materials and Methods
2.1. Cell Lines and Peripheral Blood Mononuclear Cells (PBMCs)
NK-92-CD16A cells were derived by the author BW from a cell line ATCC CRL-2407 and lentivirus-transformed to express CD16A AA158valine and green fluorescent protein [
28]. Cells were cultured per original ATCC instructions with alpha Minimum Essential Media containing L-glutamine and sodium pyruvate, (Gibco [Waltham, MA]/ThermoFisher), 0.2 mM 2-mercaptoethanol, 0.2 mM inositol, 0.02 mM folic acid, 12.5% horse serum (Gibco), 12.5% fetal bovine serum (FBS) (Biowest, Riverside MO), 1% pen-strep (Gibco), and 1000 U/ml Teceleukin recombinant interleukin 2 (Roche, Basel Switzerland). Cells were maintained at 5% CO
2 and 37 °C.
Raji cells (ATCC CCL-86) [
33] were cultured in RPMI media with L-glutamine (GenClone, El Cajon CA), 10% FBS and 1% pen-strep at 5% CO
2 and 37 °C.
Both cell lines were regularly tested for mycoplasma (Lonza MycoAlert, Basel, Switzerland) and were negative.
Peripheral blood mononuclear cells (PBMCs) from the healthy donors were isolated at UNR by ficoll-hypaque density gradient centrifugation [
36] as described in [
37] where the ADCC was originally reported for these donors as data points of Figure 2B. ADCC and EC
50 assays, FcγRIII genotyping, CD16A-positive NK cell TruCounts
® (Becton Dickenson no. 340334), and immunophenotyping, were determined in the original study. The use of human subjects was approved by the institutional review board of the University of Nevada, Reno School of Medicine. Written informed consent was obtained from the blood donors.
2.2. Antibodies for CD107a Exocytosis and ADCC
Humanized anti-CD20 monoclonal antibody was used to support ADCC. There were two glycosylated forms of one monoclonal antibody. One form was highly afucosylated obinutuzumab (brand name Gazyva
R that is used therapeutically; also reported as glyco-engineered GE GA101) [
34,
38,
39]. This antibody was produced in CHO cells that were genetically modified to reduce fucosylation of antibodies [
34]. The other WT antibody was ~10% afucosylated and produced in standard CHO cells. The extent of Fc-fucosylation for obinutuzumab/GE GA101 is ca. 50%, and for WT GA101 is greater than 90% (communicated by Christian Klein, Ph.D., Roche Innovation Center Zurich, Switzerland).
2.3. Detection of CD107a Externalization
When cytotoxic T [
40] or NK [
41] cells kill other cells, they release perforin and granzymes from intracellular granules. The membranes of the cytotoxic granules and the cells’ plasma membrane fuse during the release of the granule contents. Simultaneously, an inner membrane granule protein LAMP-1/CD107a is externalized and becomes part of the killer cell’s plasma membrane. Here we detected external CD107a of un-permeabilized NK cells with PE-mouse mAb anti-CD017a.
Specifically, Raji “target” cells (500,000 cells at 2.0 × 106/ml in 0.25 ml) were incubated for 30 minutes at RT with various concentrations of GA101-GE or GA101-WT antibody in duplicates. A control for exocytosis was 50 ng/ml phorbol myristic acid (PMA, Sigma, St. Louis MO) and 50 ng/ml ionomycin (StemCell, Vancouver BC). NK-92-CD16A cells to produce various effector-to-target cell ratios were added in 0.25 ml. (For example for an E:T of 1:2, 250,000 cells at 1.0 × 106/ml of NK-92 CD16Acells in 0.25 ml were added.) Tubes were centrifuged at 1000 rpm for three minutes to bring the effector and target cells together and then incubated at 5% CO2 and 37 °C for 0, 20, 40, 60, or 120 minutes. After incubation, the cells were placed on ice and ice-cold isotonic Na2EDTA in FACS buffer (Sheath fluid (BioSure, Grass Valley, CA), 1% FCS, and 0.09% NaN3) was added to a final concentration of ~3 mM EDTA to chelate calcium and stop exocytosis. The tubes were centrifuged at 1200 rpm for 10 minutes, the supernatants decanted, and the cells left in minimal volumes (~50 ul) to promote good labeling with fluorescent antibodies. Cells were stained for 30 minutes at RT with an antibody panel containing PE-αCD107a (clone H4A3, BD Bioscience, San Jose CA), APC-αCD19 (clone H1B19, BioLegend, San Diego CA), PacBlue-αCD45 (clone H130, BioLegend), and BV650-αCD56 (clone HCD56, BioLegend). The stained cells were washed twice with FACS-EDTA buffer and 0.5% formaldehyde fixed before the flow cytometric analysis. The instrument used was a BD Biosciences Special Order Research Product (SORP) LSR II analytical flow cytometer with a High Throughput Sampler. The data were analyzed using FlowJo version 10 (FlowJo, LLC, Ashland, OR).
Antibody EC50. The effective concentration of antibody needed to support 50% of maximal CD107a externalization or ADCC (EC
50, [
42]) is a measurement of NK cell receptor engagement of antibody on the target cells. For CD107a, the maximal externalization was the percent of cells with CD107a antibodies minus the CD107a externalization of NK cells without mAb. This subtraction was necessary because of variation in NK CD107a externalization without mAb. Half of this ADCC-specific externalization was added to the background NK CD107a externalization and used as the Y in the linear equation y = mx + b to solve for X the log
10 of antibody for the EC
50. Note, in case of EC
50s for
51Cr release, the NK cytotoxicity without mAb was negligible so that Y was half the maximal
51Cr release at high antibody concentrations.
2.4. Flow Cytometric Detection of Antibodies Bound to Raji Cells
This method was used to compare cellular CD107a with detection of Raji-bound antibodies that stimulated the CD107a externalization. Raji cells were pre-incubated with dilutions of GA101 WT or GE mAb for 30 minutes. After incubation, the cells were washed twice and then labeled for 30 at RT with AF594-conjugated affiniPure donkey anti-human IgG (H + L) (Jackson ImmunoResearch, West Grove PA). Using this procedure, we also verified that the CD20 antigen-binding properties of the WT and the GE antibodies were identical.
2.5. Cytotoxicity Assays
Targets release internalized
51Cr into the supernatant when they die. Target Raji cells were labeled with Na
51CrO
4 (Perkin Elmer, Waltham, MA) [
43]. Raji and effector NK-92-CD16A cell counts were determined by Trypan blue (MilliporeSigma) exclusion. Assays were in V-bottom plates (Costar 3894, 96 well) in 0.2 ml with 1 × 10
4 Raji cells per well in quadruplicate. There were two experimental formats, one for antibody EC
50s and another for the effects of different E:T ratios. For the antibody EC
50 determinations, GA101 WT or GE antibodies were diluted 2-fold. Radiolabeled Raji cells (10
4) were added to each well and incubated for 30 minutes at RT. NK-92-CD16A cells for a final E:T of 16:1 were added to each well. For evaluation of E:T effects, NK-92-CD16A cells were diluted 2-fold to create the E:Ts. Radiolabeled Raji cells were pre-incubated with GA101 WT or GE for 30 minutes at room temperature then added to the wells with the varying NK-92-CD16A cells. Plates were centrifuged at 1000 rpm for three minutes to bring the effector and target cells together and incubated at 5% CO
2 and 37 °C for 40 minutes, two, or four hours. After incubation, ~3 mM Na
2EDTA was added to each well then, the plates were centrifuged at 1200 rpm for 10 minutes. Half of the cell-free supernatant was removed for analysis in a Perkin-Elmer Wizard gamma counter. The spontaneous release was the average leak rate of target cells without effectors; the maximum release was the radioactivity released by the target cells lysed with 1% SDS. The calculated % specific release is a measure of the dead target cells. Percent specific release (SR) was calculated using the following formula:
2.6. Effects of Anti-Viral Biosafety Conditions
Heat inactivation of sera. Human male AB serum (Sigma Aldrich, Visalia CA) was heat inactivated at 56 °C for 30 minutes with and without addition of GA101 WT. The heated serum with GA101 was diluted with heated serum without mAb to make two mAb concentrations. The solutions were incubated with Raji cells for 30 minutes, then assayed with NK-92-CD16A cells for 40 minutes.
Formaldehyde treatment. The effects of formaldehyde could impair NK recognition of denatured anti-target antibodies and/or denature the epitope of CD107a. For effects on the bound mAb, Raji cells with antibody were washed once, then treated with 1.0% formaldehyde for 15 minutes, washed twice to remove formaldehyde, counted, and used as in section 2.3 to elicit CD107a externalization. For the formaldehyde effects on CD107a, cells were treated immediately after incubation with or without 0.5% formaldehyde for 15 minutes and washed twice before labeling with fluorescent antibodies.
2.7. Graphics
Graphics were made using GraphPad Prism 9 (V 9.5.1.733 for Windows, GraphPad Software, San Diego CA) and modified using Microsoft PowerPoint V2019.
2.8. Statistical Analyses
CD107a assessments. The FlowJo “compare population” tool was used for Overton subtractions [
44]. Excel Student’s T tests [
45] were applied to compare duplicate samples with other duplicate samples, using one-tailed and type 2 (two-sample equal variance [homoscedastic]) settings.
51Cr-cytotoxicity assays. Data were calculated with Microsoft Excel and the significance of comparisons was assessed with paired 4-well sets using Student’s T tests.
Comparisons of EC50s. For comparison of the linear regressions in Figure 1(B1,B2), the data were evaluated using analysis of variance with SPSS Statistics (IBM, version 28, Armonk, NY). For comparison of the EC50s for CD107a vs. 51Cr cytotoxicity in Figure 1(C1), the 95% confidence intervals were calculated for each EC50 and then the confidence intervals were compared for overlap.
5. Discussion
We designed and optimized a cell-based bioassay to quantify antibodies that can support ADCC. The assay used clonal anti-CD20 antibody with two levels of afucosylation as a proof of concept. We evaluated antibody concentrations, duration of assay, and effector-to-target ratios to determine the conditions to support the most CD107a externalization by NK-92 CD16A cells. The assay detected as little as 0.2 ng/ml (1.5 × 10−12 M) antibody, was optimal at 40 minutes and at E:Ts with excess target cells, and was sensitive to antibody afucosylation. CD107a signals differed by ~20-fold in response to differences in antibody afucosylation. We observed a negative correlation between the CD107a externalization and target cell death mediated by ADCC and propose a model below to address this conundrum. This assay, using immortalized NK-92-CD16A at E:Ts with excess targets, is a suitable basis to develop serum assays to characterize the ADCC potential of anti-viral antibodies to diverse viruses.
Details are important to optimize this assay. It is crucial to use a high concentration of fluorescent antibody to CD107a. GFP, translated after CD16A, helped identify healthy NK-92-CD16A cells. NK-92 CD16A lines developed by other investigators are likely to be comparably sensitive for use in CD107a assays [
56]or even more sensitive (
e.g., with CD16A as a fusion protein with domains of 41BB and CD3ζ to improve intracellular signaling [
57]). Regardless of the cell source, it will be important to have E:T ratios with excess targets. The effects of E:T ratios on CD107a detection were profound. At the high E:Ts that supported ADCC cytolysis, there was marginable CD107a externalization. This phenomenon has been observed and reported before for NK cell killing without antibodies (Figure 4, [
25]. In practice, for antibodies that can support ADCC, the E:Ts should be 1:1, 1:2, or 1:4 or even lower. To assess antiviral antibodies with infected cells, the assays will require careful attention to actual E:T ratios because only a fraction of the cells will be virally infected and able to become ADCC targets.
The unexpected disconnect between low CD107a and high ADCC has ramifications for interpretation of CD107a externalization in tumor and virally infected cell microenvironments. How can optimal killing occur with a low release of cytotoxic granules? We advance a model to explain this phenomenon (
Figure A4). At a low E:T (e.g., 1:4) one killer cell may attack multiple target cells without killing any of the targets. We postulate that the effectors released too few granules per target to be lethal. The high CD107a of each outnumbered killer cell indicates that the killer spent a lot of granule ammunition futilely, probably engaging in multiple sublethal attacks. At a high E:T (e.g., the inverse 4:1 ratio), multiple NK cells can attack a single target cell simultaneously. Together, the NKs deliver sufficient ammunition to kill this target cell and then they may halt their cytotoxic granule release after the target dies [
58]. The low CD107a of each of these identified killer cells indicates each one released only a small amount of its stored granule ammunition. NK-92 cells can kill with as few as three granule externalizations (detectable with a CD107a-GFP construct and confocal microscopy [
59]), which suggests that the threshold for detection of degranulation by flow cytometry may be too high to detect all the killers at high E:T ratios. Nonetheless, multiple killers, probably working together, spent sufficient granule ammunition to cause death while outnumbered killers failed. One ramification of our model is that low or undetectable NK or T cell extracellular CD107a in vivo may actually be associated with cytotoxic activity!
We contrast our assay with assays developed by other investigators. We focus on assays designed to detect differences in specific antibody afucosylation without use of mass spectrophotometry. For the assay presented, effects of afucosylation would be relative to an IVIG or a mAb standard for specific anti-viral antibodies. We compare three alternative assays by these criteria: sensitivity, simulation of viral antigens, structure of CD16A, laboratory time, and quantitation. Simulation of physiological antigens is important since infected cells will display multiple viral proteins. NK cell CD16A binds better to IgG than CD16A produced in HEK cells [
50]. Quantitation of the amounts of antiviral antibodies, provides insight into molar antibody concentrations and how far they might be able to drop before losing bioactivity.
One assay utilized infected cells as the source of viral antigens and a T cell tumor line transfected with CD16A as the sensor for anti-viral antibodies [
60]. After CD16A-recognition of cell-bound anti-viral antibodies, the T cells secreted IL-2 which was measured by ELISA. This design permitted detection of antibodies specific for many viruses that were able to support ADCC. This assay was sensitive to 640 ng/ml monoclonal anti-RSV IgG1 and has the advantage of viral antigens in physiological protein conformation and at infectious densities of proteins in the plasma membranes. It is a ‘universal’ assay and will detect antibodies to any virus that generates antigens in infected cells’ plasma membranes. A disadvantage of this assay is the time needed for the T cells to produce IL-2 and for the IL-2 ELISA.
Two other assays that detected afucosylated antiviral antibodies utilized recombinant (r-) viral protein antigens, either bound to enzyme-linked immunosorbent assay (ELISA) plates [
61] or to beads for flow cytometry [
62]. These assays relied on r-CD16A protein for antibody detection (rather than cellular CD16A). The assay called Fucose-sensitive Enzyme-linked ImmunoSorbent Assay (FEASI) [
61] used SARS-CoV-2 r-spike (S) protein as the antigen, with two ELISA read-outs: (1) for anti-human IgG for antibody quantification; and (2) for biotinylated monomeric r-CD16A followed by enzyme-linked avidin. A mAb IgG1 anti-S protein with varying percentages of afucosylation was used for calibration of afucosylation. The assay provided excellent assessment of afucosylation but required 100 ng/ml specific mAb when the antibody was 4% afucosylated. This assay is time-efficient because it is independent of tissue culture, quantitative, and is suitable when specific antibodies are elevated in serum. Its disadvantages are limited sensitivity and restriction to a single r-viral protein.
The flow cytometric ‘Fc-array’ assay [
62,
63,
64] employed color-coded beads bearing diverse r-viral antigenic proteins combined with different Fc-receptors that included CD16A. HIV and influenza proteins were coupled to the beads. Specific antibodies bound to the beads were detected with (a) PE-tagged antibodies to IgG subclasses and (b) PE-tagged avidin molecules containing four biotinylated r-CD16As per avidin. The CD16A tetramers could detect 1 × 10
−9 M mAb anti-HIV [
63], with less sensitivity than the NK-92 CD16A assay. PE-tagged, fucose-sensitive lectins were also used as probes. The Fc-array is suitable for calibration with an antiviral mAb with different levels of fucosylation and for quantification of the bound anti-IgG1 and IgG3. Its current disadvantages are dependence on ratios of qualitative MFIs for bead-bound anti-IgG1 and r-CD16A. The r-CD16A produced in HEK cells may have reduced sensitivity compared to NK cell CD16A.
The ‘NK-CD107′ assay described here will need further development in order to evaluate anti-viral antibodies. To quantify antiviral antibodies rather than titer them, one could measure the IgG 1&3 bound to infected cells by flow cytometry. An accepted method [
63] utilizes phycoerthrin (PE)-labeled monoclonal anti-human IgGs 1&3 (with one PE molecule per mAb molecule) and PE standards [
65] to quantify the bound antibody by flow cytometry. The viral innocula used for infections and the post-infection time of expression of plasma-membrane viral proteins will require careful attention. A reference antiviral antibody standard, such a humanized antiviral mAb or an intravenous immunoglobulin preparation IVIG [
63] will be needed to control for inter-assay variation. The advantages of this approach are viral proteins with native protein structures and in physiologically relevant concentrations and the ability to detect multiple viral proteins. For example, the coronavirus OC-43 has a hemagglutinin as well as a spike protein that will be found in plasma membranes. Intact viruses with their multiple envelope proteins, may be displayed for ADCC when the viruses are bound to infected cells by tetherin [
66,
67]. We have characterized a sensitive cell-to-cell based assay to detect antibodies that support ADCC that also has unique potential for physiological insights.
Author Contributions
The authors’ contributions were: conceptualization, D.H., J.C.A., V.C.L. and B.W.; methodology, J.C.A., D.H. and J.S.-G.; validation, D.H., J.C.A. and J.S.-G.; formal analyses, D.H., J.C.A. and J.S.-G.; investigation, D.H. and J.C.A.; resources, D.H.; data curation, D.H. and J.C.A.; writing—original draft preparation, J.C.A. and D.H.; writing—review and editing, D.H., V.C.L., J.S.-G. and B.W.; visualization, J.C.A.; supervision, D.H.; project administration, D.H.; funding acquisition, D.H. All authors have read and agreed to the published version of the manuscript.