Single-cell deep phenotyping of IgG-secreting cells for high-resolution immune monitoring

Klaus Eyer, Raphaël C L Doineau, Carlos E Castrillon, Luis Briseño-Roa, Vera Menrath,
Guillaume Mottet, Patrick England, Alexei Godina, Elodie Brient-Litzler, Clément Nizak,
Allan Jensen, Andrew D Griffiths, Jérôme Bibette, Pierre Bruhns & Jean Baudry

Studies of the dynamics of the antibody-mediated immune
response have been hampered by the absence of quantitative,
high-throughput systems to analyze individual antibody-
secreting cells. Here we describe a simple microfluidic
system, DropMap, in which single cells are compartmentalized
in tens of thousands of 40-pL droplets and analyzed in two-
dimensional droplet arrays using a fluorescence relocation-
based immunoassay. Using DropMap, we characterized
antibody-secreting cells in mice immunized with tetanus
toxoid (TT) over a 7-week protocol, simultaneously analyzing
the secretion rate and affinity of IgG from over 0.5 million
individual cells enriched from spleen and bone marrow.
Immunization resulted in dramatic increases in the range
of both single-cell secretion rates and affinities, which
spanned at maximum 3 and 4 logs, respectively. We observed
differences over time in dynamics of secretion rate and affinity
within and between anatomical compartments. This system
will not only enable immune monitoring and optimization of
immunization and vaccination protocols but also potentiate
antibody screening.

B cells and the antibodies they secrete are a critical arm of the immune
response. Antibodies are classified into several isotypes; IgG is the
main type of antibody found in blood and extracellular fluid and enables control of infection. The levels of antigen-specific IgG in circulating blood correlate with an individual’s immunological history, and
the average affinity of antibodies for their target tends to increase during an immune response. Mass spectrometry of circulating IgG coupled with next-generation sequencing has recently been used to gain
further insights at the molecular level. However, this approach
does not capture a complete picture of the complexity of the humoral
immune response and serum IgG composition, which is a function
of the frequency of IgG-secreting cells (IgG-SCs), their IgG secretion
rates, and the affinity of the secreted IgGs for the antigen. The pool
of IgG-SCs is heterogeneous and consists mainly of plasmablasts and short-lived plasma cells in the spleen and long-lived plasma cells in
the bone marrow, but each individual IgG-SC is unique and produces
only one IgG variant. Therefore, a direct and statistical analysis of
individual IgG-SCs would provide a more profound understanding
of the immune response.

Functional analysis of cloned, immortalized IgG-SCs and myeloma
cells has allowed quantitative measurement of average secretion rates,
affinities and specificities of antibodies, but usually only from a small
number of clones (typically <102). In contrast, next-generation
sequencing allows the molecular identification of large numbers of
expressed antibody genes, but subsequent cloning and expression
are needed to determine phenotypic properties such as the affinity and specificity of the antibodies, a process that is laborious and
low-throughput. For the large-scale phenotypic characterization of
IgG-SC populations, tens of thousands of non-immortalized IgG-
SCs can be screened using enzyme-linked immunospot (ELISPOT)
or related assays, or by compartmentalization in microfabricated arrays, to detect antibody secretion and antigen binding.
Compartmentalization of individual cells in picoliter droplets has
also been used for high-throughput screening and sorting of hybridoma cells based on the binding or inhibitory activity of secreted
antibodies. However, none of these methods allows quantitative,
sensitive, and simultaneous real-time kinetic measurement of antibody secretion rate and affinity at the single-cell level, rendering
impossible a deep and quantitative assessment of the phenotypical
complexity and variability of IgG-SC populations following immunization or vaccination.

DropMap technology

We describe here a droplet-based microfluidic technology (DropMap) that overcomes previous limitations. The combination of immobilizing (‘parking’) picoliter-size droplets and a fast and ultrasensitive bioassay allows massively parallel kinetic analyses of
single IgG-SCs, with simultaneous measurement of antibody secretion rate, specificity, and affinity for the antigen. We used this system
to characterize the humoral response in mice immunized with TT over a 7-week protocol, producing a quantitative, statistical analysis
of the distribution of secretion rates and affinities over the course of
an immune response.

The DropMap technology is based on kinetic analysis of droplets
immobilized in a two-dimensional (2D) array. Single cells were compartmentalized in monodisperse 40-pL aqueous droplets in inert fluorinated
carrier oil by hydrodynamic flow focusing in a microfluidic
system. Cells and assay reagents were brought together on the microfluidic
chip just before droplet creation, with the cells diluted to give an average of 0.2–0.4 cells per droplet for the encapsulation of up to one cell
per drop. Droplets were immobilized by a slight physical confinement
in the z-direction within the chamber (<15%), allowing them to be
observed over time. In this study, 40,000 droplets (i.e., 8,000–16,000
individual cells) were imaged over an area of 8.6 x 6.5 mm.

To quantify IgG, we devised an in-droplet sandwich immunoassay based on paramagnetic nanoparticles. The nanoparticles were pre-coated with an anti-mouse kappa light-chain nanobody to capture secreted immunoglobulin. The 1,300 nanoparticles in each drop were induced to form an elongated, observable
agregate, termed a beadline, by applying a magnetic field. Compared
to previous approaches relying on the co-encapsulation of a single cell and a single 6-Mm bead, the use of nanoparticles increased
the antibody binding capacity and lowered the number of droplets
containing no beads, which improved the performance of the assay. Each droplet also contained Alexa647-labeled F(ab')2 specific for mouse IgG Fc (termed
anti-IgG(Fc), red fluorescent) and Alexa488-labeled antigen (here TT) (green fluorescent). IgG and anti-IgG(Fc) were captured onto the beadline,
and the concentration of IgG was determined from the ratio of red
fluorescence on the beadline to mean fluorescence within the droplet. Our calibration curve was based on a purified anti-TT
IgG (TT7) with Kd of 8 nM. In this calibration, increasing concentrations of TT7 resulted in an increase in red fluorescence (anti-IgG(Fc)) relocation on the beadline up to a concentration of around 75 nM, then a decrease as the binding capacity of the nanoparticles was exceeded. The same general behavior was seen with green fluorescence (Alexa488-labeled TT) relocation, although saturation occurred at a slightly lower concentration (40 nM).

Secretion rates were calculated by measuring the change in IgG concentration over time. IgG secretion in droplets of pL volume reached nM range concentrations in about 30 min, allowing quantitative measurement via the relocation immunoassay. Secretion rates were precisely determined from ~4 to 10,000 IgG/s, a dynamic range that
is in accordance with secretion rates reported in the literature. Cells were defined as IgG-SCs if the secretion rate exceeded 4 IgG/s, which was determined to be the lower limit of detection within the timeframe of an experiment.

If the secreted IgG recognizes the antigen (here TT), the ratio of antigen (green fluorescent) bound to IgG on the beadline and the mean fluorescence within the droplet can be used to calculate the strength of the interaction (Kd), because the concentration of IgG is measured and that of antigen is imposed. The Kd is determined from the slope of the line defined by
plotting the relocation of red fluorescent anti-IgG(Fc) versus the relocation of green fluorescent antigen to the beadline at different concentrations of IgG. These curves are plotted using the experimental data at IgG concentration ~50 nM for the anti-TT antibody TT7 and an anti-CD34 antibody, which does not bind TT. Slopes were determined for a panel of seven purified anti-TT IgGs and plotted versus the Kd values, which ranged from 300 to 0.02 nM (determined using surface plasmon resonance (SPR)). The results are in agreement with the theoretical model and show that the assay has a dynamic range of ~4 logs, spanning the typical range of affinities found for humoral responses.

Five of those anti-TT IgGs, with Kd values ranging from 150 to 0.6 nM, were also expressed from individual Chinese hamster ovary (CHO) cells and human embryonic kidney (HEK293T) cells. The mean IgG secretion rate measured in droplets was similar for the three stably transfected CHO lines (173 ± 12 IgG/s per cell) and close to that measured as an average in bulk, using homogeneous time-resolved fluorescence (192 ± 21 IgG/s per cell). However, IgG secretion rates from single cells varied widely (~10 to ~900 IgG/s, CV 1.99), consistent with the wide variance in
protein (including antibody) expression at the single-cell level previously described for CHO and other cell lines. The mean IgG secretion rates measured for the transiently transfected HEK293T cells were lower than for the CHO cells, but similar for the different antibodies (74 ± 51 IgG/s per cell). For all five IgG-secreting cell lines, the mean Kd values measured in droplets closely approximated the values determined by bulk measurements using SPR (gold standard), with relatively similar coefficient of variation (CV; mean CVs 0.5 and 0.25, respectively). No significant correlation between the measured secretion rates and the measured Kd was observed.

Next, we applied the DropMap system to follow anti-TT responses over time after primary (day 0) and secondary (day 14) immunizations in adjuvant, and a boost (day 42) without adjuvant. We harvested splenocytes and bone marrow cells from the same mouse, with three mice per time point. We analyzed a total of 575,553 B-cell lineage-enriched cells ( ~12,800 cells/mouse), simultaneously measuring IgG secretion rate and affinity for TT for each IgG-SC.

Before immunization, the frequency of IgG-SCs was low (~0.1%), but increased 80-fold in the spleen and 45-fold in the bone marrow, peaking 4–7 d after each immunization. Notably, the frequency of IgG-SCs with and without detectable TT-binding evolved similarly, with, on average, 17 ± 3% of IgG-SCs producing anti-TT antibodies in both anatomical compartments. The use of complete Freund’s adjuvant may contribute to the high frequency of cells secreting IgGs without detectable TT-binding after the primary immunization, but it does not explain their high frequency after secondary immunization and boost. Cells secreting polyspecific IgG (antibodies that non-specifically stick to the beadline and non-related antigens) were identified and excluded from data analysis. The differences in frequencies of cells secreting non-TT-binding IgG and anti-TT IgG were minor between the three mice at each time point, indicating high reproducibility.

Figure 1 DropMap technology principle and calibration. (a) Overview of the workflow and timing. IgG-SCs were isolated and enriched from immunized
mice, and compartmentalized in 40-pL droplets using a microfluidic device. Two aqueous phases (I and II), one containing cells or calibration antibodies (I) and the other containing paramagnetic nanoparticles and fluorescently labeled detection reagents (II). After creation, droplets were loaded into an observation chamber to form a 2D droplet array, flow was stopped, and the IgG secretion and affinity was measured over time.

Figure 2 Ex vivo measurement of spleen and bone marrow single-cell IgG secretion rates and affinities using DropMap. (a) - (f) Time course analyses of B-cell-enriched cell suspensions from the spleen (orange) and bone marrow (blue) of TT-immunized mice (cell suspensions from three individual mice analyzed per time point). Frequency of IgG-secreting cells (a) or anti-TT IgG-secreting cells (b) in B-cell-enriched populations. (c) Relative frequency of anti-TT IgG-secreting cells. (d) Time course of mean serum IgG affinity for TT and specific IgG as percentage of the peak response at day 28, measured by SPR. Mean secretion rates (e) and mean Kd for TT (f) of non-polyspecific anti-TT IgG-secreting cells.

We also compared spleen IgG-SCs from three TT-immunized and three bovine serum albumin (BSA)-immunized mice (identical immunization schedule) at day 18. We found
similar frequencies of IgG-SCs with similar secretion rate distributions.
The frequency of IgG-SCs producing non-polyspecific anti-TT IgG in BSA-immunized mice was lower than in TT-immunized mice, but only by eightfold. The frequency of IgG-SCs producing non-polyspecific anti-BSA IgG in TT-immunized mice was in the same range, and is over tenfold higher than the frequency of false positives calculated by analyzing a CHO cell line expressing an antibody against TT for binding to BSA. Thus, a substantial number of IgG-SCs produce cross-reactive antibodies that bind an antigen unrelated to the immunogen.

In the TT-immunized mice, the frequency of anti-TT IgG-SCs declined rapidly after the peak following secondary immunization and boost in spleen and bone marrow. In contrast, serum titers of anti-TT IgGs remained high after peaking following the secondary immunization and boost, including during an 18-d period (days 24–42) where very few cells were secreting anti-TT IgGs. These results were confirmed using SPR, where TT-specific IgG became detectable at only day 10, and the TT-specific IgG peaked 14 d after the secondary immunization, remaining high until the last measurement.

DropMap captured the full dynamics of the humoral immune response that were only visible with quantitative, time-resolved, and high-throughput single-cell analysis.

After immunization, average IgG secretion rates and affinities increased only modestly, but dramatic increases in the distribution of secretion rates and affinities were observed at the single-cell level. There were large increases in both the frequency of IgG-SCs and single-cell secretion rates, whether the IgGs were antigen-specific or not, pointing to extensive activation and/or stimulation of bystander B cells. Activation of bystander cells may also contribute to the substantial frequency of IgG-SCs producing cross-reactive IgGs that bind to an antigen other than the immunogen. METHODS Methods, including statements of data availability and any associated accession codes and references, are available in the online version of the paper.

ACKNOWLEDGMENTS We would like to thank the Institut Pierre-Gilles de Gennes for use of clean room facilities and the laser engraver, and Pfizer for the generous gift of TT-Alexa488, and CHO cell lines secreting TT4, TT7, and TT10. We thank Pfizer and HiFiBiO Team for identification, rapid cloning, and validation of TT11 and TT27 antibodies. This work received support from the French Investissements d’Avenir program under the grant agreements, the French Agence Nationale de la Recherche, from Région Ile-de-France, and by the Institut Carnot Pasteur Maladies Infectieuses.