Research ArticleImmunology

B cell antigen receptors of the IgM and IgD classes are clustered in different protein islands that are altered during B cell activation

See allHide authors and affiliations

Science Signaling  15 Sep 2015:
Vol. 8, Issue 394, pp. ra93
DOI: 10.1126/scisignal.2005887

How BCRs mingle

The B cell antigen receptor (BCR) consists of a plasma membrane–bound antibody [immunoglobulin (Ig)] that is associated with a pair of signaling proteins. Antigen binding to the BCR stimulates B cells to differentiate into antibody-secreting cells. Maity et al. used high-resolution microscopy, electron microscopy, and proximity ligation assays to visualize the organization of IgM-BCRs and IgD-BCRs on mature B cells. Under resting conditions, the different BCRs were separated in relatively large clusters called protein islands. Antigen triggered the protein islands to become smaller and more disperse, reducing the distance between the different BCRs.

Abstract

The B cell antigen receptors (BCRs) play an important role in the clonal selection of B cells and their differentiation into antibody-secreting plasma cells. Mature B cells have both immunoglobulin M (IgM) and IgD types of BCRs, which have identical antigen-binding sites and are both associated with the signaling subunits Igα and Igβ, but differ in their membrane-bound heavy chain isoforms. By two-color direct stochastic optical reconstruction microscopy (dSTORM), we showed that IgM-BCRs and IgD-BCRs reside in the plasma membrane in different protein islands with average sizes of 150 and 240 nm, respectively. Upon B cell activation, the BCR protein islands became smaller and more dispersed such that the IgM-BCRs and IgD-BCRs were found in close proximity to each other. Moreover, specific stimulation of one class of BCR had minimal effects on the organization of the other. These conclusions were supported by the findings from two-marker transmission electron microscopy and proximity ligation assays. Together, these data provide evidence for a preformed multimeric organization of BCRs on the plasma membrane that is remodeled after B cell activation.

INTRODUCTION

B lymphocytes play a central role during the humoral immune response by generating antibodies against a vast range of molecules termed antigens. Each B cell senses a specific antigen with the help of its B cell antigen receptors (BCRs), which are found in large amounts (~120,000 to 200,000 copies per cell) on the cell surface (1, 2). A single BCR consists of a membrane-bound antibody, which has two identical membrane-bound heavy chains (mHCs) and two identical light chains (LCs), together with the signaling subunit, which consists of the heterodimer of immunoglobulin α (Igα) and Igβ (3). Depending on their developmental stage and the mHCs that they produce, B cells carry different BCR classes (namely, IgM, IgD, IgG, IgE, and IgA) on their cell surface (4, 5). Immature B cells have only IgM-BCRs on their surface (6, 7), whereas mature B cells have both IgM-BCR and IgD-BCR complexes, which have identical antigen-binding specificities (8, 9). The IgM-BCRs and IgD-BCRs differ from each other only in the constant parts of their respective mHCs (μm and δm). In particular, the Igα-Igβ signaling subunit is identical for the two BCR classes, apart from minor differences in the glycosylation of Igα (5, 10, 11).

The structural similarity of the IgM-BCR and IgD-BCR makes it difficult to understand the described differences in signal transduction between the two receptors (1214). Furthermore, studies of IgD transgenic and knockout mice failed to reveal a specific functional role for the IgD-BCR, suggesting that the two receptor classes have overlapping or partially redundant functions (1517). Both receptors seem to have preformed oligomeric structures on resting B cells that are distinct from each other (18). The oligomeric organizations of IgM-BCRs and IgD-BCRs have been detected by several different methods, namely, blue native polyacrylamide gel electrophoresis (3), the bimolecular fluorescence complementation assay (19), and different versions of the proximity ligation assay (PLA) (20, 21). PLA can detect the close proximity of two proteins on the surface of primary cells (22, 23). Together, these studies contradict one of the major assumptions of the cross-linking model of BCR activation, which states that BCR complexes on the resting B cell surface are freely moving monomers that are activated by aggregation or dimerization (24, 25). Rather, these findings support the dissociation activation model, whereby B cell activation is accompanied by an opening up of the preorganized BCR oligomers (2).

Evidence for a compartmentalization or lateral segregation of proteins and lipids into distinct domains on the plasma membrane has been found in numerous studies (2628). Nanoscale clusters of the high-affinity Fc receptor for IgE (FcεR1) or of the T cell antigen receptor (TCR) were detected by transmission electron microscopy (TEM) (26, 29). These findings were further supported by super-resolution microscopy techniques showing that the TCR, as well as the BCR, is clustered on the surface of resting lymphocytes (1, 30). Such structures are variably called “nanoclusters” or “protein islands,” and we use the latter term here. Given their width of 40 to 250 nm, the composition and distribution of protein islands on the cell surface cannot be easily studied by conventional light microscopy, which has a light diffraction barrier of ~250 nm (3032). Using Fab-based PLA (Fab-PLA), we previously studied the nanoscale organization of the BCR and its co-receptors on the surface of resting and activated B lymphocytes (20). This study suggests that the IgM-BCRs and the IgD-BCRs reside in different membrane compartments with a distinct protein and lipid composition. To analyze in greater detail the nanoscale organization of the IgM-BCRs and IgD-BCRs on the B cell surface, we have combined different super-resolution techniques, including two-color direct stochastic optical reconstruction microscopy (dSTORM) and two-marker TEM. Together, our results provide evidence that the IgM-BCRs and IgD-BCRs are indeed localized in separate protein islands, which become smaller and coalesce upon B cell activation.

RESULTS

IgM- and IgD-containing protein islands on resting B cells

To study the distribution of the two different BCR classes on the cell surface at nanoscale resolution, we used two-color dSTORM with fluorophore-labeled, high-affinity cognate antigens. We transfected RAG, λ5, and SLP-65 triply deficient (TKO) cells (33) to generate cell lines carrying IgM-BCRs and IgD-BCRs, each with a different antigen-binding specificity on their cell surface (Fig. 1A, top panels). To avoid mixed HC-LC pairing, we replaced the VH and CH1 domain of the μmHC with a VH-VL single chain using the VH and VL domains of the hen egg lysozyme (HEL)–specific antibody HyHEL10 (34). The IgD-BCR contains the VH domain and the λ1-LC of the B1-8 antibody (35) and thus binds with high affinity to the hapten 4-hydroxy-5-iodo-3-nitrophenylacetyl (NIP). We thus generated the TKO-HμNδ cell line with a HEL-specific single-chain IgM-BCR (scIgM-BCR) and a NIP-specific IgD-BCR (Fig. 1A, left). As a control, we also generated TKO-HμNμ cells that expressed both a HEL-specific scIgM-BCR and a NIP-specific IgM-BCR (Fig. 1A, right). The binding of each antigen to these cells was tested by flow cytometry with fluorophore-conjugated HEL and NIP molecules (Fig. 1A, bottom). The two different receptors were also detected on the surface of TKO-HμNδ with class-specific antibodies (fig. S1A). The HEL and NIP antigens were coupled with the dyes Atto-565 (A565) and Atto-647N (A647), respectively, which are suitable for two-color dSTORM imaging (36). The antigen HEL-A565 was prepared by chemical coupling of A565 to purified HEL, resulting in the conjugation of, on average, 1.43 fluorophores per protein molecule. The antigen 1NIP-A647 is a 9-mer peptide that contains one NIP group coupled to a lysine at the third position and the A647 group coupled to a C-terminal cysteine. The optimum concentrations of HEL-A565 and 1NIP-A647 required for the labeling of TKO-HμNδ cells were determined by measuring the mean fluorescence intensity (MFI) of binding (fig. S1B). Note that the binding of these monovalent antigens to their cognate BCRs did not activate the TKO-HμNδ B cells (fig. S1, C and D).

Fig. 1 Two-color dSTORM imaging of IgM- and IgD-containing protein islands.

(A) Top: Labeling strategy for two-color dSTORM imaging of TKO-HμNδ cells (left) coexpressing a HEL-specific scIgM-BCR and a NIP-specific IgD-BCR, as well as of control TKO-HμNμ cells (right) coexpressing a HEL-specific scIgM-BCR and a NIP-specific IgM-BCR. Bottom: Flow cytometric analysis of the binding of the fluorescently labeled antigens HEL-A565 and 1NIP-A647 to TKO (untransfected control), TKO-HμNδ, and TKO-HμNμ cells. Data are representative of four independent experiments. (B) Two-color dSTORM grouped images of TKO-HμNδ and TKO-HμNμ cells with HEL-A565 (green) and 1NIP-A647 (red). Reconstructed dSTORM images of TKO-HμNδ cells (top) and TKO-HμNμ cells (bottom) showing scIgM-BCR (green, first column), IgD-BCR (red, second column, top), or IgM-BCR (red, second column, bottom), and both channels merged (third column). Signals in the images are grouped by 50-nm radii and an off gap of 10 frames. Scale bars, 2 μm (green and red channels). The magnified regions (6 × 6 μm, fourth column) correspond to the white squares in the third column; scale bars, 1 μm. Data are representative of four independent experiments. (C) Pair autocorrelation analysis of the molecular localizations of HEL-A565 bound to scIgM-BCR (left) and 1NIP-A647 bound to IgD-BCR (middle) to estimate the clustering and overcounting frequency of the fluorophores (right). The quantification of the overcounting frequency of scIgM-BCRs and IgD-BCRs in TKO-HμNδ cells is shown on the right. graw(r), radial autocorrelation function of original image; ggroup(r), grouped autocorrelation; gpsf(r) = graw(r) – ggroup(r); PSF, point spread function that occurs due to overcounting and the finite localization precision of the dSTORM measurement. (D) Quantification of the number of scIgM-BCRs and IgD-BCRs per island (left) after normalization of the overcounting frequency and of the average radii of the islands (right). Medians are indicated by a blue line, and the corresponding values are given below the x axis. Data were compared by Mann-Whitney test. ***P < 0.001. (E) Quantification of the clustering tendency and distribution of scIgM-BCR (left) and IgD-BCR (right) with Ripley’s K function analysis. Top: Plots of H(r) function. Bottom: Plots of the first-order differential of the H function H′(r). The average radii of the islands [at H′(r) = 0] are indicated by arrows. The theoretical estimate of H(r) and H′(r) for a random distribution is represented by the black dashed line. (F) Cross-correlation analysis of the localization of HEL-A565 and 1NIP-A647 binding to their respective receptors on TKO-HμNδ cells (top) and TKO-HμNμ cells (bottom). The pair of blue lines represents the estimate of the 99% confidence interval of association and dissociation, and the dashed red line represents the random dispersion calculated by 30 random simulations of identical images. (G) Plot of the bivariate Ripley’s function [Lbiv(r) − r] of the binding of HEL-A565 and 1NIP-A647 to their respective receptors on TKO-HμNδ cells (top) and TKO-HμNμ cells (bottom). Data in (D) to (G) are derived from at least three independent experiments with a minimum of six cells per experiment and are means ± SEM (E to G) or complete distributions with the median of the numbers of BCRs per island and the median of the island radius (D).

For the dSTORM analysis, the cells were incubated on ice with the antigens HEL-A565 and 1NIP-A647 at concentrations (100 nM each) that ensured saturated binding (fig. S1B). The cells were then washed, placed on glass dishes coated with nonstimulatory poly-d-lysine, and fixed with 4% paraformaldehyde (PFA). For two-color dSTORM imaging, we analyzed the samples with an inverted total internal reflection fluorescence (TIRF) microscope with an attached electron-multiplying charge-coupled device (EMCCD) camera operating at sequential activation and repetition programs (37). To minimize the effect of chromatic aberration, we analyzed the cells located at the center of the imaging field (~51 × 51 μm2, at 100-nm pixel width). Indeed, by keeping the cells at the center of the field, we did not detect any substantial changes between the rendered images upon parabolic mode correction (see Materials and Methods for details). The reconstructed images showed that scIgM-BCRs and IgD-BCRs on TKO-HμNδ cells were concentrated in separate protein islands (Fig. 1B, upper panels). In contrast, dSTORM images of control TKO-HμNμ cells showed frequent overlapping and intermixing signals for HEL-A565 and 1NIP-A647 (Fig. 1B, lower panels). As an additional control, we exposed TKO-Hμ cells, which had only the HEL-specific scIgM-BCR on the surface, to a 1:1 mixture of HEL-A565 and HEL-A647 and found extensive costaining of the IgM-BCR–containing protein islands (fig. S2).

The imaging data from three independent experiments with at least six cells per experiment were analyzed, taking into account the photophysical properties of the fluorophores undergoing blinking during the dSTORM imaging procedure. The localization of the fluorophores was measured with software, and the data were quantified in two different ways. First, we studied receptor clustering and the range of overcounting by the pair autocorrelation method (38, 39) on different regions of the cells as described in Materials and Methods. The graph of the average radial autocorrelation function, graw(r), over radii of 0 to 500 nm centered around the fluorophore localizations showed a positive clustering tendency of BCR complexes, specifically at shorter distances where the curve deviates from the theoretical value of one (Fig. 1C). To estimate the overcounting value as a result of the multiple appearance of a single fluorophore, we calculated the grouped autocorrelation, ggroup(r), after grouping the localizations in the consecutive frames of the image sequences. To obtain the optimum grouping radii and off-state gap interval of the fluorophore (“off gap”), we tested several grouping radii and off gap values (see Materials and Methods for details). Finally, the correlation function of the localized BCRs excluding the multiple detection of fluorophores bound to the BCRs, gpsf(r) = graw(r) − ggroup(r), was calculated, which is expected to have a theoretical value of zero for any radius for randomly dispersed condition (Fig. 1C, left and middle). Analyzing the pair autocorrelation functions, we obtained overcounting factors of 3.1 ± 1.2 for scIgM-BCRs and 2.8 ± 1.2 for IgD-BCRs, suggesting that the fluorophores exhibited similar distribution and detection properties (Fig. 1C, right). The autocorrelation analysis normalized for the overcounting factors was then used to calculate the number of BCRs in each protein island, as well as the average radii of these islands (Fig. 1D). Assuming that the two antigen-binding sites of each BCR were equally accessible, we divided the fluorophore counts by two to obtain the number of BCRs in each protein island. As expected from visual inspection of the reconstructed images, the IgM-containing islands were smaller (median radius, 218 nm) than the IgD-containing islands (median radius, 290 nm). The IgM-containing islands also contained lower numbers of BCRs (median, 30) than did the IgD islands (median, 48). To verify the variability among different cells, we compared the BCR numbers and the radii of the BCR protein islands from four different segments (3 × 3 μm) on each of the six cells in one experiment. No statistically significant difference of the data was detected by Kruskal-Wallis one-way analysis of variance (ANOVA) (fig. S3).

We also analyzed the data in a different way, with Ripley’s spatial clustering analysis, and calculated the H function, H(r) = L(r) − r (Fig. 1E, upper plots), to determine the average clustering tendency, which was based on the numbers of molecules found within a given distance r (39, 40). The value of r at the Hpeak is an estimate of the average radius of the protein island and can be more precisely determined by plotting the H′(r) value, a discrete differential of the H function (40). For the IgM- and IgD-containing protein islands, the H′(r) function crossed the zero line at 151 and 240 nm, respectively (Fig. 1E, lower plots). These calculations support the notion that the IgD-containing islands were larger than the IgM-containing islands (1).

We then analyzed the colocalization of the scIgM-BCRs and the IgD-BCRs on resting TKO-HμNδ B cells by calculating the pair cross-correlation function, C(r), between the two separate channels of the dSTORM images. As a positive control in this analysis, we used resting TKO-HμNμ B cells because the colocalization and interaction of different channels in multicolor localization microscopy depends on a multitude of factors, including the precision of image acquisition, microscope stage drift, choice of fluorophore pairs, and fixation (41, 42). A value of 1 for C(r) at any point of radial distance indicates the absence of an interaction (38). For statistical analysis of the obtained cross-correlations, we measured the average values of random distribution of the same localizations within the identical coverage area with a minimum of 30 simulations and then calculated the 99% confidence interval of association and dissociation. There was no cross-correlation between the scIgM-BCR and IgD-BCR channels in the analysis of TKO-HμNδ B cells because the data points completely resided within the 99% confidence interval (Fig. 1F, top); however, the parallel analysis of TKO-HμNμ B cells showed the colocalization of scIgM and IgM over a large range of radial distances (Fig. 1F, bottom). Similarly, the analysis of the bivariate Ripley’s function [Lbiv(r) − r] also revealed no colocalization of the scIgM-BCR and the IgD-BCR channels in TKO-HμNδ B cells (Fig. 1G, top). In comparison, scIgM-BCRs and IgM-BCRs on TKO-HμNμ B cells remained colocalized over a large range of radial distances (Fig. 1G, bottom). Similarly, both the cross-correlation and bivariate Ripley’s function analyses of HEL-A565 and HEL-A647 localizations on resting TKO-Hμ cells showed statistically significant colocalization throughout the radial distance (fig. S2D). We compared the cross-correlation data with both a random distribution model of the same number of localizations and a random two-color BCR labeling process (fig. S2, D and E, panels v and vi). This analysis confirmed our previous Fab-PLA studies, indicating the close proximity of IgM-BCRs to each other (20). The different locations and dimensions of the IgM and IgD protein islands did not depend on the attachment surface because we obtained similar results from TKO-HμNδ cells that were adhered to either a poly-d-lysine– or a fibronectin-coated glass surface (compare Fig. 1 with fig. S4). In summary, the two-color dSTORM analysis showed that the IgM-BCRs and IgD-BCRs on resting B cells were compartmentalized in different membrane areas.

Reduced sizes of BCR islands upon B cell activation

To study the effect of B cell activation on the organization of these protein islands, we stimulated TKO-HμNδ B cells with either latrunculin-A (Lat-A), which inhibits F-actin formation, or a combination of anti-IgM and anti-IgD antibodies (hereafter referred to as anti-Ig treatment). The membrane-proximal actin cytoskeleton framework retains the BCRs in a closed, inactive state (43). Treatment with Lat-A disrupts this framework, and as a consequence, the BCRs become dissociated and activated. These activation conditions were chosen because our dSTORM protocol required free antigen-binding sites to label the BCRs. To treat the cells with anti-Ig, we used each antibody at a concentration of 10 μg/ml (44). We tested several concentrations of Lat-A from 0.25 to 2 μM to activate B cells by measuring the effects on intracellular Ca2+ influx, cell viability, BCR abundance, and cell morphology (fig. S5). In our test, 1 μM Lat-A resulted in detectable intracellular Ca2+ flux (fig. S5C) in B cells within 150 to 200 s of administration but did not alter cell viability (fig. S5A), BCR abundance (fig. S5B), or cell morphology (fig. S5, D and E). Therefore, we used 1 μM Lat-A for activating B cells in further experiments.

Stimulated TKO-HμNδ B cells were first incubated with 1NIP-A647 and HEL-A565 for 10 min on ice and then were settled on precooled poly-d-lysine–coated coverslips on ice for 15 min before being fixed. Two-color dSTORM images showed that, after B cell activation, the IgM- and IgD-containing protein islands were smaller and more dispersed than they were in cells before activation (Fig. 2A, compare panels resting with Lat-A and anti-Ig of columns i and ii). This finding was confirmed by the pair autocorrelation analysis of the acquired images (Fig. 2A, compare top panels with middle and bottom panels of columns iii and iv). The two-dimensional (2D) radial intensity plots of the pair autocorrelation functions for the scIgM-BCR and IgD-BCR channels also showed a decrease in the autocorrelation intensities of the two receptors on Lat-A– or anti-Ig–treated B cells compared to those of resting B cells. Furthermore, the overall molecule density of the acquired images and the overcounting factors did not change substantially between resting and activated TKO-HμNδ B cells (fig. S6, A to C).

Fig. 2 Two-color dSTORM imaging of IgM- and IgD-containing islands on activated B cells.

(A) Overview (i) and magnified (ii) two-color grouped dSTORM images of resting TKO-HμNδ cells (top), TKO-HμNδ cells stimulated with 1 μM Lat-A (middle), and TKO-HμNδ cells stimulated with a combination of anti-IgD and anti-IgM antibodies (10 μg/ml each) (anti-Ig; bottom). Signals in the images are grouped by 50-nm radii and an off gap of 10 frames. Scale bars, 2 μm (i). A magnification of the area in the white box (2 × 2 μm) is shown in (ii); scale bars, 0.5 μm. Representative intensity plots of the 2D autocorrelation function for both scIgM-BCR (iii) and IgD-BCR (iv) indicate the clustering tendency of receptors on the TKO-HμNδ cells by a normalized coloring index from −0.2 to +0.2. Data are representative of four independent experiments. (B) Normalized average numbers of receptors inside IgM- and IgD-containing islands on resting TKO-HμNδ cells (data from Fig. 1D are shown as gray symbols for comparison), anti-Ig–treated TKO-HμNδ cells, and TKO-HμNδ cells treated with 1 μM Lat-A. The average numbers of IgM-BCRs and IgD-BCRs on Lat-A–treated TKO-HμNδ cells were compared with those on TKO-HμNδ cells treated with 0.4% ethanol as a vehicle control. Data were compared by two-tailed Mann-Whitney test. Medians of the numbers of scIgM-BCRs and IgD-BCRs per island are indicated as green and red horizontal lines, respectively. ***P < 0.001; ns, not significant. (C) Pair cross-correlation analysis of the localization of HEL-A565–bound scIgM-BCRs and NIP-A647–bound IgD-BCRs in Lat-A–treated (top) and anti-Ig–treated (bottom) TKO-HμNδ cells. The dashed red line represents the random dispersion calculated by the random simulation of identical images, and the pair of blue lines represents the estimate of the 99% confidence interval of association and dissociation. Representative data from four experiments and a minimum of six cells per experiment are shown as means ± SEM. (D) Quantification of the cross-correlation values from 0 to 100 nm separation distance between localization peaks for Lat-A– and anti-Ig–treated TKO-HμNδ cells in comparison to those of untreated TKO-HμNδ and TKO-HμNμ cells. The results from the resting TKO-HμNδ cells were used for normalization, and data are presented as means ± SD. (E) Plot of the bivariate Ripley’s function [Lbiv(r) − r] of HEL-A565–bound scIgM-BCRs and NIP-A647–bound IgD-BCRs in Lat-A–treated (top) and anti-Ig–treated (bottom) TKO-HμNδ cells. Data in (B) to (E) are derived from at least three independent experiments with a minimum of six cells per experiment and are means ± SD (D), means ± SEM (C and E), or complete distributions with the median (B).

As described earlier, the median numbers of scIgM-BCRs and IgD-BCRs in each island in resting TKO-HμNδ cells were 30 and 48, respectively. After exposure of the cells to 1 μM Lat-A, these numbers were reduced to 7 and 3, respectively, whereas after the anti-Ig treatment, they were reduced to 15 and 11, respectively (Fig. 2B). Consistent with this decrease in BCR numbers per island, the radii of the protein islands also were reduced in response to either Lat-A or anti-Ig (fig. S6D). In the Ripley’s spatial clustering analysis, using the H(r) and H′(r) functions, we also found a decrease in the average radius of the IgM islands from 151 nm on resting cells to 100 nm (for Lat-A–treated) and 112 nm (for anti-Ig–treated) on activated TKO-HμNδ B cells (fig. S6E). The corresponding values for the IgD islands were 240 nm for resting cells, 107 nm for Lat-A–treated cells, and 161 nm for anti-Ig–treated cells (fig. S6F).

A parallel pair autocorrelation analysis of TKO-HμNμ cells carrying scIgM-BCR and an IgM-BCR (fig. S7, A and B), each with different antigen-binding sites, showed that the median number of these receptors per island was 16 and 17, respectively, on resting cells (fig. S7, C and D) and 4 and 6, respectively, on Lat-A–treated TKO-HμNμ cells (fig. S7, E and F). Consistent with this decrease in the numbers of BCRs per island, the radii of the protein islands in Lat-A–treated cells were also decreased compared to those in untreated control cells (fig. S7, D and F). Furthermore, in the acquired images, the molecular density of the bound antigens 1NIP-A647 and HEL-A565 remained unchanged in untreated and Lat-A–treated TKO-HμNμ cells (fig. S7B). The cross-correlation and bivariate Ripley’s spatial clustering analyses showed that, in comparison to the untreated control cells, the intensity of interaction of 1NIP-A647 and HEL-A565 in Lat-A–treated TKO-HμNμ cells partially decreased but that the proximity of 1NIP-A647 and HEL-A565 remained above the confidence interval (fig. S7, G and H). The same behavior was found in Lat-A–treated TKO-Hμ cells (fig. S2E). Together, these data suggest that B cell activation results in the dispersion of the protein islands.

As described earlier, a pair cross-correlation analysis showed no colocalization of scIgM and IgD protein islands on resting TKO-HμNδ B cells (Fig. 1F). However, upon exposure to Lat-A or anti-Ig, the cross-correlation between the scIgM and IgD channels was significantly increased above the 99% confidence interval (Fig. 2C). The quantification of the intensity of the cross-correlations in the range of 10- to 100-nm distance between scIgM and IgD showed a two- to threefold increase in the cross-correlation on Lat-A– or anti-Ig–treated TKO-HμNδ B cells compared to that on resting cells (Fig. 2D). As a positive control, we analyzed the cross-correlation between scIgM and IgM on TKO-HμNμ B cells and found that it was 8- to 10-fold greater than that between scIgM and IgD on resting TKO-HμNδ B cells (Fig. 2D). The bivariate Ripley’s analysis also showed an increase in the interaction between scIgM and IgD upon treatment of the TKO-HμNδ B cells with Lat-A or anti-Ig (Fig. 2E). Together, these data suggest that the IgD and IgM protein islands move closer together in response to B cell stimulation.

Class-restricted stimulation of BCR-containing protein islands

Stimulation of mature B cells with anti-IgD antibodies results in the phosphorylation of the IgD-associated, but not IgM-associated, Igα subunit (45). This finding suggests that upon class-specific activation of B cells, signals do not spread between protein islands. To test this insulation hypothesis, we performed dSTORM analysis of TKO-HμNδ cells when only one receptor class, either the scIgM-BCR or the IgD-BCR, was stimulated. Stimulation of TKO-HμNδ cells with an anti-IgM antibody resulted in a substantial decrease in the autocorrelation between scIgM-BCRs, as well as in the number of scIgM-BCRs in each protein island, whereas the autocorrelation between IgD-BCRs was unchanged (Fig. 3, A and B). Conversely, the stimulation of TKO-HμNδ cells with an anti-IgD antibody resulted in a substantial decrease in the autocorrelation between IgD-BCRs as well as in the numbers of IgD-BCRs but had no such effect on IgM-BCRs (Fig. 3, A and B). This class-specific alteration of the BCR islands was also revealed by the Ripley’s spatial clustering analysis of the acquired dSTORM images (Fig. 3, C and D). Together, these data suggest that the distinct IgM and IgD islands can be activated independently of each other.

Fig. 3 Two-color dSTORM imaging of IgM- and IgD-containing islands on B cells activated by either anti-IgM or anti-IgD antibodies.

(A) Overview (i) and magnified (ii) two-color dSTORM images of TKO-HμNδ cells stimulated with anti-IgM antibody (10 μg/ml; top) or anti-IgD antibody (10 μg/ml; bottom). Scale bars, 2 μm (i). Images in (ii) show a magnification of the area in the white box (2 × 2 μm). Signals in the images are grouped by 50-nm radii and an off gap of 10 frames. Scale bars, 0.5 μm (ii). Representative intensity plots of the 2D autocorrelation function for scIgM-BCRs and IgD-BCRs are shown in columns (iii) and (iv), respectively. (B) Average numbers of receptors within IgM- or IgD-containing islands on TKO-HμNδ cells stimulated with anti-IgM antibody or anti-IgD antibody. Data from the analysis of IgM- and IgD-containing islands on resting TKO-HμNδ cells from Fig. 1D are shown as gray symbols for comparison. Data were compared by two-tailed Mann-Whitney test. ***P < 0.001, *P < 0.05. (C) Quantification of the clustering tendency of scIgM-BCRs (left, green circles) and IgD-BCRs (right, red squares) in TKO-HμNδ cells treated with anti-IgM antibody (10 μg/ml) by Ripley’s K function analysis. Top: Means ± SEM of the H function. Bottom: Means ± SEM of H′(r). (D) Ripley’s K function analysis of scIgM-BCR– and IgD-BCR–containing islands in TKO-HμNδ cells treated with anti-IgD antibody (10 μg/ml). Top: Means ± SEM of the H function. Bottom: Means ± SEM of H′(r). The mean values of resting TKO-HμNδ cells in (C) and (D), shown as gray symbols for comparison, are taken from Fig. 1E. The estimated average radii of the islands at H′(r) = 0 are indicated by arrows in (C) and (D). Data in (B) to (D) are derived from three independent experiments with a minimum of six cells per experiment.

The calculations and conclusions derived from our dSTORM analysis are critically dependent on the photophysical properties of the fluorophores A565 and A647, as described previously by other groups (36, 38). These properties include brightness, duration of blinking, and number of appearances of the fluorophores as measured by photon count, on-off duty cycle, and number of switching cycles, respectively. We thus carefully determined the average photon count, on-off duty cycle, and number of switching cycles of A565 and A647 in dilute conditions as described in detail in Materials and Methods (fig. S8). The obtained on-off duty cycles for A565 and A647 were 0.006457 and 0.006905, respectively (fig. S8, F and G). The obtained switching cycles for A565 and A647 were 4.13 and 6.7 per fluorophore, respectively (fig. S8, H and I). We further analyzed the signals obtained from A565 and A647 for their optical PSF width (s, PSF SD), number of detected photons (N), and local background photons (b) recorded by the camera (fig. S9, A to C). Using these values of PSF width, photons, and background photons, we calculated the precision of our dSTORM imaging by two different methods (fig. S9, D and E). The obtained “blind estimate” and the Thompson’s estimate of localization precisions for A565 and A647 were 7.1 ± 1.4 and 8.7 ± 1.2 nm and 25 ± 10 and 29 ± 9 nm, respectively. In each method, we found similar localization precision range for A565 and A647, suggesting similar resolution of the two fluorophores, which enabled us to group them together. Note that the ELYRA PS.1 system that we used for the dSTORM imaging also used the Thompson’s estimate formula for individual localization. Furthermore, the drift in the accuracy of estimating localization during the imaging procedure for both the A565 and A647 channels remained very low (fig. S9F).

In addition to the photophysics and localization precision of A565 and A647, we analyzed the effect of different grouping radii and different number of image frames representing the off-state gap interval of the fluorophores corresponding to the transient dark state of the blinking fluorophores (fig. S9, G to J). Using the limiting Sparrow’s resolution formula (resolution = 2× precision), we obtained a resolution of 50 nm for A565 (46). We thus used a radius of 50 nm for grouping the localizations. With this setting, we eliminated 92 ± 6% of the overcount events (fig. S9G). In addition, the grouping of A565 localization at 10 frames of off gap removed 99.9% of the overcount events (fig. S9H). A similar grouping of A647 localization at a 50-nm radius and 10 frames of off gap eliminated 90 ± 3% and 99.9% of the overcount events, respectively (fig. S9, I and J). We also analyzed the percentages of A565 and A647 overcount events grouped at different radii of 50, 75, and 150 nm (fig. S9K). Furthermore, we showed that the overcount factor for both A565 and A647 at higher grouping radii of 75 and 150 nm remained unaltered compared to the overcount factors when grouping radii of 50 nm were used, as identified by the Mann-Whitney test (fig. S9L).

To learn more about the interactions between the two-color channels with the pair cross-correlation analysis, we performed a random two-color BCR labeling simulation and obtained the null hypothesis (fig. S10, A to E). With three different values of BCR surface density (fig. S10A), and assuming a 50% probability of dual-color occupancy of each of the binding arms of the BCR (fig. S10B), we simulated the images of random two-color BCR labeling (fig. S10C). With this simulation, we showed that the pair cross-correlation values remained similar to those of the random distribution model described earlier (compare fig. S10D with fig. S2D, panel v). In addition, we found that the increase in BCR surface density from 150 to 200 μm−2 had no effect on the cross-correlation function, except for lowering the initial cross-correlation value at a radius of zero (fig. S10E, panels i to iii). Furthermore, we simulated the two-color BCR labeling process in resting and activated B cells similarly to what was described earlier (fig. S2). With the input variables of BCR surface density, the density of receptors in clusters, and the two different cluster radii for resting and activated B cells as experimentally determined (fig. S10F, panels i to iii), we obtained values of 27 ± 9 BCRs and 8 ± 3 BCRs per cluster under resting and activated conditions, respectively (fig. S10G). These numbers of BCRs per cluster are similar to the experimentally measured numbers of IgM-BCR islands in resting and activated B cells (Fig. 2B), which suggests that the simulation process is valid. We then analyzed the cross-correlation function of the simulated images and found that under activated condition, the cross-correlation values decreased compared to those under resting conditions (fig. S10, H and I). In addition, in this simulation, we obtained a similar distribution and segregation of the three different colors (red, green, and yellow) as was observed earlier (fig. S2, D and E). Thus, our simulations and dual-labeling experiments led to similar conclusions about the distribution and clustering of BCRs.

Analysis of IgM- and IgD-containing protein islands by two-marker TEM

We then used two-marker TEM as an alternative method to study the organization of two different receptors on the cell surface at a distance below the diffraction limit of visible light (26, 47). We transfected 3046 pro-B cells (Igα−/−, SLP65−/−) to coexpress a μHC and a δHC, each with a different C-terminal tandem peptide tag, together with Igα and the λ1-LC. The resulting cell line 3046-MD contains a 2×Myc-tagged IgM-BCR and a 2×HA (hemagglutinin)–tagged IgD-BCR, with both receptors binding to the antigen NIP (Fig. 4A, left). The presence of the two BCR classes on the surface of 3046-MD cells was verified by flow cytometric analysis with the antigen NIP-A647 and anti-Ig antibodies (Fig. 4A, middle). The 3046-MD cells were placed on either nonstimulating, poly-l-lysine–coated or stimulating, NIP15-BSA (bovine serum albumin)–coated electron microscopy grids. The cells were then “ripped” to generate plasma membrane sheets attached to the grid surface. The exposed inner leaflet of the plasma membrane was incubated with anti-HA (12CA5) and anti-Myc (9E10) antibodies, which were followed by secondary antibodies coupled to gold nanoparticles (Fig. 4A, right). In this way, the relative positions of IgM (Fig. 4B, green arrows) and IgD (Fig. 4B, red arrows) on the plasma membrane were indicated by 10- and 6-nm gold particles, respectively. The two-marker TEM images from resting B cells showed that the IgM-BCRs and the IgD-BCRs were each clustered in different membrane areas (Fig. 4B, top). The BCR clusters that we detected are consistent with the results of a previous TEM study on the distribution of the IgD-BCR on the surface of fixed J558L B cells (48). The different BCR islands were also visible on the surface of antigen-exposed B cells, but they seemed to be placed in closer proximity to each other (Fig. 4B, bottom). A Ripley’s function analysis of data from a minimum of 10 images confirmed the clustering behavior of the different BCR classes in the 10- to 200-nm range (Fig. 4C). Similar to what was found in the dSTORM analysis, the two-marker TEM study showed that the IgD islands were in general larger than the IgM islands and that the dimensions of both types of islands became smaller after the cells were stimulated with antigen.

Fig. 4 IgM-BCR and IgD-BCR islands come in close proximity to each other upon stimulation with antigen.

(A) Labeling and detection strategy to visualize IgM-BCRs and IgD-BCRs in the plasma membrane sheet by two-marker TEM. Left to right: Production of a C-terminal tandem 2×Myc-tagged IgM-BCR and a tandem 2×HA-tagged IgD-BCR, both of which are specific for the NIP antigen. Flow cytometric analysis of the binding of fluorescently labeled NIP (1NIP-A647) to 3046 cells (untransfected controls) and 3046-MD cells. Flow cytometric analysis of untransfected 3046 cells and 3046-MD cells for the coexpression of IgM-BCRs and IgD-BCRs by surface staining. Detection of tags on ripped plasma membranes on an EM grid by labeling with anti-Myc and anti-HA primary antibodies followed by secondary antibodies conjugated to 10- and 6-nm gold particles. An alternate combination of secondary antibodies conjugated to 6- and 10-nm gold particles was also used to detect the Myc-tagged IgM-BCRs and the HA-tagged IgD-BCRs, respectively (see fig. S4). (B) Two-marker TEM images of Myc-tagged IgM-BCRs and HA-tagged IgD-BCRs detected by 10- and 6-nm gold particles, respectively. Representative 1 × 1 μm sections of representative TEM images on poly-l-lysine–coated (top) or NIP15-BSA–coated surfaces (bottom). Right: A magnified region (0.2 × 0.2 μm). Green and red arrowheads indicate gold particles detecting the location of IgM-BCRs and IgD-BCRs, respectively. Data are representative of three independent experiments. (C) Ripley’s K function analysis of the localizations of 10- and 6-nm gold nanoparticles in TEM images detecting IgM-BCRs and IgD-BCRs, respectively. Top: Plot of the H(r) of IgM-BCRs (green, left) and IgD-BCRs (red, right) on a nonstimulating poly-l-lysine–coated surface. Bottom: Plot of the H(r) of activated IgM-BCRs (green, left) and IgD-BCRs (red, right) on a NIP15-BSA–coated surface. The dashed line at H(r) = 1 represents random dispersion above which the particles are clustered. Black arrowheads indicate the distances at which the H(r) function decreases to below the random dispersion. (D) Plot of the bivariate Ripley’s function [Lbiv(r) − r] of the proximity of IgM-BCRs to IgD-BCRs in TEM images of resting (poly-l-lysine, top) and activated (NIP15-BSA, bottom) cells. Data in (C) and (D) are means ± SEM of a minimum of 10 images per experiment from at least two experiments.

The relative proximity of the IgM and IgD islands to each other was further analyzed by bivariate Ripley’s function analysis, which was used to determine the spatial association (upper envelope), dissociation (lower envelope), and random distribution of two different sets of samples from the same area. On resting B cells, the bivariate Ripley’s function analysis indicated a random distribution of the two different receptor classes throughout the measured distance (Fig. 4D, top). In contrast, on activated B cells, the relative proximity of the IgD-BCR and IgM-BCR clusters to each other was statistically significantly increased and, according to the Lbiv(r) − r function, was placed above the 99% confidence limit of a nonrandom association (Fig. 4D, bottom). To exclude a possible steric bias of labeling, we also exchanged the sizes of the gold particles coupled to the secondary antibodies so that the IgM-BCRs and IgD-BCRs on the plasma membrane were indicated by 6- and 10-nm gold particles, respectively. We found a similar receptor distribution (fig. S11). As a control experiment, we also analyzed 3046-MD cell expressing an IgD-BCR with the two different tags (fig. S12A). With both combinations of secondary antibodies and gold particles for detecting the tags, the Ripley’s function analysis indicated clustering of the IgD-BCRs (fig. S12). Furthermore, the bivariate Ripley’s function remained above the 99% confidence of association, which indicated a nonrandom distribution of IgD-BCRs (fig. S12, D and G).

We also analyzed the TEM images from 3046-MD cells with the help of a nearest-neighbor clustering method (1, 49). After varying the neighbor distance (ND), we chose an ND of 50 as the most appropriate for our further analysis (fig. S13A and see Materials and Methods). Applying this method on resting 3046-MD cells, we found that 49% of the probes detecting IgM-BCRs and 76% of the probes detecting IgD-BCRs resided inside larger clusters or protein islands (fig. S13B) and that the sizes of the clusters were reduced upon B cell activation (fig. S13C). The numbers of probes detected per cluster also changed upon activation of the cells (fig. S13D). Together with the results of the dSTORM experiments, these data suggest that the BCR protein islands became dispersed upon cellular activation.

Analysis of the proximity between IgM and IgD by Fab-PLA

We previously used Fab-PLA to monitor the proximity of receptors on resting B cells at a separation distance of 10 to 20 nm and found evidence for the existence of closed IgM-IgM and IgD-IgD dimers that are dissociated and opened up upon B cell activation (20). Furthermore, our dSTORM and TEM studies suggested that, upon activation, some of the opened IgM-BCRs and IgD-BCRs moved closer together. To verify this, we conducted an IgM:IgD Fab-PLA on resting and activated murine and human B cells (Fig. 5). As a source of murine B cells, we used the transfected pro-B cell line 3046-MD and naïve B cells from the B1-8 mouse, whose B cells all carry the NIP-specific BCR on their surface. On resting 3046-MD cells and on B1-8 B cells, we detected very little colocalization of IgM and IgD (Fig. 5, A and B, and fig. S14, A and B); however, upon activation of these B cells with either the antigen NIP15-BSA or Lat-A, Fab-PLA detected a close proximity of the two different BCR classes. An increased proximity between IgM-BCRs and IgD-BCRs was also detected on human peripheral blood B cells upon treatment with either PerV or Lat-A (Fig. 5C). These assays were repeated with three further samples of human peripheral blood B cells with the same results (fig. S14C). The findings of these Fab-PLA studies are thus consistent with the dSTORM and TEM data.

Fig. 5 Proximity of IgM-BCRs and IgD-BCRs on primary B cells from mice and humans.

(A) Fab-PLA images of the proximity of IgM-BCRs and IgD-BCRs on resting 3046-MD cells, 3046-MD cells stimulated with antigen [NIP15-BSA (50 ng/ml)], and 3046-MD cells treated with 1 μM Lat-A. The proximity of IgM-BCRs and IgD-BCRs was detected with a pair of anti-IgM Fab PLUS and anti-IgD Fab MINUS probes (IgM:IgD). The alternate combination of anti-IgD Fab PLUS and anti-IgM Fab MINUS probes gave similar results (see fig. S5). Scale bar, 5 μm. Graph shows the quantification of the PLA signals (red dots) per cell [DAPI (4′,6-diamidino-2-phenylindole), blue nucleus] in the indicated cells as compared by Mann-Whitney test. ***P < 0.001. The median of the number of PLA dots per cell is shown by a blue line. (B) Fab-PLA images of the proximity of IgM-BCRs and IgD-BCRs proximity in resting, antigen-stimulated [NIP15-BSA (50 ng/ml)], and Lat-A–treated B cells from B1-8 transgenic mice. Graph shows the quantification of the number of PLA signals per cell in the indicated cells as compared by Mann-Whitney test. ***P < 0.001. (C) Fab-PLA images of the proximity of IgM-BCRs and IgD-BCRs in resting, 0.5 mM pervanadate (PerV)–treated, and 1 μM Lat-A–treated naïve B cells isolated from human peripheral blood. The proximity of IgM-BCRs and IgD-BCRs was analyzed in naïve B cells from four individual samples (see fig. S5) with a pair of anti-human IgM Fab PLUS and anti-human IgD Fab MINUS probes. Graph shows the quantification of the number of PLA signals per cell in the indicated cells as compared by a Mann-Whitney test. ***P < 0.001. Data in (A) and (B) are representative of three independent experiments. Data in (C) are representative of four human blood samples.

DISCUSSION

We analyzed the distributions of IgM-BCRs and IgD-BCRs on resting and activated B cells with three different methods, namely, two-color dSTORM, two-marker immunogold-TEM, and Fab-PLA. Data from the dSTORM and TEM studies revealed that the two different receptor classes were organized in separate protein islands on resting B cells. On activated B cells, the BCR-containing protein islands were smaller than those on resting B cells. In addition, analysis of dSTORM images and data from TEM and Fab-PLA studies revealed that IgM-BCRs and IgD-BCRs increased their proximity to each other in activated B cells.

It is by now well appreciated that the plasma membrane has a high degree of compartmentalization. Most transport and signaling processes at the membrane occur inside or between membrane compartments having dimensions of 40 to 200 nm, which are below the diffraction barrier of visible light and thus are difficult to study by classical light microscopy techniques. A better understanding of these processes thus requires techniques that enable the localization of proteins and lipids at separation distances of 10 to 200 nm. The currently available super-resolution techniques are, however, still challenging, and each of them is associated with certain technical limitations. For example, the analysis of results obtained by dSTORM must deal with the intrinsic noise of a fluorescence image and the estimation of signal overcounting. On the other hand, TEM requires fixation and extensive sample manipulation, thus reducing the labeling efficiency. In contrast to these two techniques, Fab-PLA does not require the production of altered or fluorophore-tagged proteins in cell lines but can be used in studies of the surface organization of proteins on fixed primary cells. The disadvantage of PLA, however, is that as a sampling method, it can monitor only 10 to 40 interactions per cell. Given these technical problems, the analysis of the nanoscale organization of proteins on the cell membrane must be based on several techniques, as is the case in this study.

Our finding that IgM-BCRs and IgD-BCRs were localized in different protein islands is consistent with a previous study that showed that these receptors can be stimulated independently of each other (45). The functional insulation of the IgM and IgD protein islands was also seen in our dSTORM study of B cells stimulated with either anti-IgM or anti-IgD antibodies (Fig. 3, A and B). The IgM and IgD protein islands seem to differ not only in their location on the B cell surface but also in their protein and lipid composition. In our previous Fab-PLA study, we found that the IgD-BCR on the surface of resting B cells is in close proximity to the co-receptor CD19 and to tetraspanin proteins, such as CD20 (20). The IgD-BCR– and CD19-associated tetraspanins are enriched in ordered lipid domains (5052). Consistent with this, we previously showed that the IgD-BCR, but not the IgM-BCR, on resting B cells is in close proximity to the glycosylphosphatidylinositol-anchored protein CD52 and to GM1 gangliosides, both of which are part of ordered lipid domains (20). The IgM-BCR becomes associated with these proteins and lipids only after B cell activation. These data suggest that on resting B cells, the IgD-BCR is found together with the CD19 complex and several tetraspanins in an ordered lipid protein island, whereas the IgM-BCR protein island has a different lipid and protein composition. This class-specific membrane organization on the B cell surface is reminiscent of the protein island distribution on the T cell membrane (30). In T cells, the raft-associated membrane adaptor protein LAT (linker of activated T cells) localizes to a protein island that is different from the island that contains the TCR. At least on the basis of their lipid composition, the IgD and IgM islands on the B cell surface may thus correspond to the LAT and TCR islands on the T cell surface, respectively.

Our finding that both IgD-BCRs and IgM-BCRs were clustered on the surface of resting B cells is consistent with a one-color dSTORM study that showed that IgD, IgM, and CD19 exist in preformed nanoclusters (1). Both super-resolution studies also were consistent in the observation that the IgD-BCR nanoclusters are somewhat larger than the IgM-BCR nanoclusters. However, such a one-color approach may not sufficiently support conclusions about the separate localizations of all three proteins (IgD, IgM, and CD19). Our previous Fab-PLA study suggested that the detected IgD and CD19 nanoclusters are part of the same protein island (20). The finding of distinct IgD- or IgM-containing protein islands raises the question of how these structures are formed and maintained. To address this question, it would be advantageous to be able to follow the appearance and behavior of these protein islands on the surface of living B cells. Unfortunately, super-resolution live-cell imaging studies have not yet reached the required resolution. Thus, the techniques presently available give only a snapshot of the nanoscale membrane organization on the B cell surface at any given time.

The different nanoscale locations of IgM-BCRs and IgD-BCRs likely result from a sorting process, the rules of which are not well understood at present. IgM-BCRs and IgD-BCRs have identical cytosolic sequences but have different transmembrane (TM) and extracellular Ig domain sequences. In particular, the mHC TM sequences are conserved and are involved in the oligomerization of the BCR (3, 19). Furthermore, the TM sequence is implicated in the compartmentalization of a large group of plasma membrane proteins in yeast (53). We need a better understanding of the sorting signals, encoded within the TM sequence, as well as the subcellular location in which the sorting of proteins into protein islands takes place. We favor the possibility that the components of protein islands are assembled in the Golgi apparatus into specific vesicles and that protein islands result from the fusion of such vesicles with the plasma membrane in a manner in which the assembled proteins and lipids stay together as long as the cells remain in the resting state. Consistent with this scenario, LAT-containing transport vesicles may be the precursors of the LAT protein islands on T lymphocytes (5456).

The activation of B cells through their BCRs is accompanied by many events that also change the conformation and location of these receptors on the B cell surface. Biochemical studies have shown that the activated BCR is phosphorylated by and associates with the kinase Syk (57), and our previous Fab-PLA studies indicate that the binding of Syk to the receptor is required for the opening up of BCR oligomers (20). Furthermore, in these studies, we found that the CD19-CD20 module apparently moves from the IgD-BCR to the IgM-BCR. This suggests that the Ig protein islands are remodeled after B cell activation. Given these alterations, it is perhaps surprising that the one-color dSTORM study mentioned earlier did not find any alteration in the size of the detected Ig or CD19 nanoclusters (1). In contrast, our two-color dSTORM and two-marker TEM analyses revealed a reduction in the size of the IgD and IgM protein islands on activated B cells. The reason for this discrepancy is not clear but may be a result of the different labeling techniques used in the two studies. The one-color dSTORM study used fluorophore-conjugated Fab fragments of anti-Ig antibodies, whereas our two-color dSTORM experiments used small, high-affinity antigens that may have a more specific binding and higher accessibility to the resting and activated BCR. A reduction in the size of the Ig-containing protein islands is also consistent with the increased mobility of the BCR on activated B cells (58).

One of the most surprising findings of our nanoscale receptor study is the increased proximity of IgM-BCRs and IgD-BCRs on the surface of activated B cells. This juxtaposition of IgM-BCRs and IgD-BCRs was detected on B cell lines and on primary B cells of murine and human origin with three different activation protocols. B cell activation is tightly associated with the opening up of the BCR and the loss of IgM-IgM and IgD-IgD proximity on activated B cells (2, 19, 20). Apparently, the dispersed forms of the two BCR classes have an increased affinity for each other. Whether this is a result of a direct interaction or is mediated by an adaptor protein is not clear, nor is the function of the association between IgM-BCRs and IgD-BCRs. One possibility is that upon B cell activation, the IgM and IgD protein islands form a “nanosynapse” that enables the exchange of proteins and lipids with each other (21). On activated T cells, an increased juxtaposition of TCR- and LAT-containing protein islands has been found, but whether the two proteins also form tight complexes is not clear (30). Thus, nanoscale studies of the membranes of resting and activated lymphocytes, as we presented here, could reveal previously uncharacterized phenomena, and much remains to be learned about these processes.

MATERIALS AND METHODS

Antibodies

We used the following antibodies for the flow cytometric analysis of mouse splenic B cells, transfected TKO cells, and 3046 cells: anti-IgM–V450 (eB121-15F9) was from eBioscience and anti-IgD–Alexa Fluor 647 (11-26c.2a) was from BioLegend. The Fab-PLA probes for mouse samples were prepared from the following unlabeled antibodies: anti-IgD (LO-MD-6) was from AbD Serotec and anti-IgM (RMM-1) was from BioLegend. For flow cytometric analysis of human B cell isolates, the following fluorophore-conjugated antibodies were used: anti-IgD–phycoerythrin (PE) (IA6-2), anti-CD19–biotin (HIB19), and anti-CD27–V450 (M-T271), all from BD Biosciences; anti-IgM–PerCP (peridinin chlorophyll protein)–Cy5.5 (MHM-88) and anti-CD10–APC (allophycocyanin)–Cy7 (HI10a) from BioLegend; and anti-CD20–APC (2H7), anti-CD43–FITC (fluorescein isothiocyanate) (84-3C1), and streptavidin-PE-Cy7 for CD19-biotin detection from eBioscience. PLA probes for human samples were prepared from the following unlabeled antibodies: anti-IgD (IA6-2) was from BioLegend; anti-IgD (IADB6) and anti-IgM (SA-DA4) were from Acris Antibodies; and anti-IgM (Fc5u) was from GenWay Biotech.

Antigens and chemicals

All of the NIP chemicals, including NIP15-BSA and NIP15-BSA–biotin, were obtained from Biosearch Technologies. The 1NIP-A647N (1NIP-A647) conjugate was custom-synthesized by Biosyntan with the small peptide backbone K-S-K(ε-NIP)-G-E-S-A-G-C-A674. HEL, poly-d-lysine, and sodium orthovanadate were purchased from Sigma. HEL-A565 and HEL-A647 were prepared by the chemical conjugation of NHS-functionalized A565 and A647, both of which were purchased from Atto-Tec. Coupling and spectrophotometric detection of the coupling efficiency were performed according to the manufacturer’s protocol. The obtained fluorophore/protein ratios were 1.43 and 1.62 for HEL-A565 and HEL-A647, respectively. PerV was prepared freshly from sodium orthovanadate by reaction with H2O2 at an equimolar ratio. Lat-A (formula weight, 421.6) was purchased from Cayman Chemical. A stock solution of Lat-A was delivered in ethanol at a concentration of 100 mg/ml, which is equivalent to 240 μM. HEL multimerization (fig. S1C) was performed by reacting HEL with sulfo-S-4FB and sulfo-S-HyNic (Solulink Inc.), which was followed by a cross-linking reaction at a 1:5 ratio in 100 mM citrate buffer (pH 6.0) in the presence of 10 mM aniline. HEL multimers were size-separated with 30-kD cutoff Amicon Ultra centrifugal filters (Merck Millipore) to obtain trimers (3×HEL) or higher oligomers of HEL. The optimum concentrations of 1NIP-A647 and HEL-A565 to label TKO-HμNδ cells were determined by measuring the MFI of the binding of various concentrations of antigen. TKO-HμNδ cells and untransfected TKO cells were incubated with different concentrations of 1NIP-A647 and HEL-A565 and were analyzed by flow cytometry. The difference in MFI (ΔMFI) of HEL-A565 and 1NIP-A647 binding to TKO-HμNδ compared to that to untransfected TKO cells was measured and plotted against the concentration (nM) of HEL-A565 and 1NIP-A647 (fig. S1B). For the dSTORM experiments, the antigens HEL-A565 and 1NIP-A647 were used at a concentration of 100 nM, which resulted in 81 to 92% and 78 to 88% saturation of binding, respectively.

Cell culture and transfections

The 3046 (Ig-α−/−, SLP65−/−) and TKO (Rag2−/−, λ5−/−, SLP65−/−) cells were grown and transfected as described previously (7, 33). Briefly, cells were cultured in Iscove’s modified Dulbecco’s medium (Biochrom) supplemented with penicillin and streptomycin (each at 10 U/ml, Life Technologies) and mouse recombinant interleukin-7 (IL-7; 0.4 ng/ml, eBioscience) or with culture medium from J558L mouse plasmacytoma cells stably transfected with a murine IL-7 expression vector, 50 μM β-mercaptoethanol (Sigma-Aldrich), and 10% fetal calf serum (FCS; PAN-Biotech). scIgM-BCR was constructed by replacing the CH1 domain of HyHEL10 μHC with a polymerase chain reaction (PCR)–amplified VL domain of HyHEL10 κLC containing an 18–amino acid serine-glycine flanking linker at the VH-VL junction. The entire cassette was subcloned into a bicistronic retroviral vector encoding an internal ribosome entry site (IRES)–driven split yellow fluorescent protein (YFP) marker. To produce NIP-specific IgD, bicistronic retroviral vectors encoding the λ1-LC, the blasticidin section marker, and B1-8 δmHC encoding a split YFP marker were constructed. TKO cells were cotransfected with all three plasmids to generate TKO-HμNδ cells. TKO-HμNδ cells were reconstituted by cotransfection of plasmids encoding scIgM, λ1-LC, and B1-8 μmHC. For EM studies, the C-terminal 2×Myc-tagged B1-8 μmHC and 2×HA-tagged B1-8 δmHC constructs were generated by PCR amplification with primers containing flanking oligonucleotide sequences encoding 2×Myc or 2×HA tags, respectively. A short glycine linker (two amino acid residues) was inserted between the C terminus of the mHC and the tag sequences. For EM studies, 3046-MD cells were derived from 3046 cells by being transfected with plasmids encoding Igα and carrying the puromycin selection marker, λ1-LC, Myc-tagged B1-8 μmHC, and HA-tagged B1-8 δmHC. The B cells were transduced retrovirally with the viral supernatant obtained from transfected Phoenix retrovirus packaging cells 2 days after they were transfected with plasmid. Transduced cells were sorted for YFP that was assembled from the bicistronic retroviral plasmids encoding split YFP fragments, and then the cells were selected by being cultured in the presence of blasticidin or puromycin. All cell sorting was performed with a BD Influx fluorescence-activated cell sorting (FACS) machine.

Isolation of mouse splenic B cells

Total spleen cells were isolated from 4- to 6-week-old anti-NIP BCR transgenic mice B1-8KI, and the naïve B cell population was enriched by magnetic-activated cell sorting (MACS) depletion with anti-CD43 magnetic beads (Miltenyi Biotec) according to the manufacturer’s protocol. Before PLA was performed, purified B cells were cultured at 37°C and in 5% CO2 in RPMI 1640 medium (Life Technologies) supplemented with 10% FCS, 50 μM β-mercaptoethanol, 25 mM Hepes, and penicillin and streptomycin (each at 10 U/ml) for 12 to 18 hours to recover from the isolation procedure. To monitor the activation status of these cells, the amount of CD86 on the cell surface was analyzed with a FACScan flow cytometer and compared with that of NIP15-BSA–stimulated cells.

Isolation of human naïve B cells

Total peripheral blood mononuclear cells (PBMCs) were prepared from freshly drawn anti-coagulated peripheral blood (about 15 to 20 ml) from healthy donors with a 50-ml LeucoSep tube (Greiner Bio-One) and Pancoll (1.077 g/liter, PAN-Biotech) according to the manufacturers’ instructions. PBMCs were washed with phosphate-buffered saline (PBS), and the fraction of naïve B cells was enriched by MACS-based depletion with the Naïve B Cell Isolation Kit II (Miltenyi Biotec). To verify the purity and identity of cell subpopulations such as immature, plasma, and memory B cells, isolated fractions were analyzed by seven-color flow cytometry with different B cell surface markers including IgD, IgM, CD10, CD19, CD20, CD27, and CD43. Thereafter, but before the PLA experiments were performed, the B cells were cultured at 37°C and 5% CO2 in RPMI 1640 medium supplemented with 25 mM Hepes, penicillin and streptomycin (each at 10 U/ml), and 2% filter-sterilized self-plasma (separated during PBMC preparation) for 12 to 18 hours to recover from the isolation procedure.

Imaging and dSTORM data acquisition

Cells were incubated with 100 nM 1NIP-A647 and 100 nM HEL-A565 for 15 min on ice, washed twice with PBS, and then placed on the surface of a precooled nonstimulatory, poly-d-lysine–coated glass-bottom μ-dish (Ibidi) for 15 min. Unbound cells were removed by a wash with PBS, and the bound cells were fixed with 4% PFA at room temperature for 30 min. For stimulation, cells were incubated with either 1 μM Lat-A for 5 min or with anti-IgM or anti-IgD (each at 10 μg/ml) antibodies alone or combined for 3 min at 37°C. Treated cells were immediately cooled on ice and washed once with ice-cold PBS before being incubated with 1NIP-A647 and HEL-A565. Fixed cells were incubated with 10 mM mercaptoethylamine in PBS (36) and analyzed with the ELYRA PS.1 system (Carl Zeiss Microscopy) with the TIRF field conditions using the 100× 1.46-NA (numerical aperture) oil immersion objective lens. Through a sequential activation and repetition program to trigger the stochastic transition of fluorophores to the bright state from a metastable dark state, cells were analyzed in two different color channels. A total of 6000 frames, with an exposure time of 30 ms, were recorded on an EMCCD camera (iXon DU-897D, Andor). Image acquisition procedures were controlled and monitored by Zen 2010D software (black edition, Carl Zeiss Microscopy). With the same software system, 2D Gaussian–fitted, single-molecule localizations and localization precision were calculated after correcting the drift with the fluorescently labeled markers fiducials (TetraSpeck microspheres, 0.1 μm, Life Technologies), which were incorporated before fixation of the cells. We obtained a precision of 26.6 ± 9.2 nm (mean ± SD) and 29.5 ± 9.1 nm for the localization of HEL-A565 and 1NIP-A647, respectively. Reconstruction of the image and overlay of two-color channels were performed with the same Zen software. To minimize the chromatic aberration effect, we always analyzed the cells located nearest the center of the imaging field (~51 × 51 μm2; pixel width, 100 nm). To examine the chromatic aberration effect under this condition, we also tested the “parabolic mode” of channel alignment guided by at least three fiducials with the same software. The parabolic mode of channel alignment corrects for lateral (x, y), axial (z), and chromatic aberration by second-order stretching of the images of different channels. Indeed, by keeping the cells at the center of the field, we did not detect any substantial changes in the visualization of rendered images after parabolic mode correction. To obtain grouped localizations for overcounting analysis by pair-correlation, we specified “max on time,” “off gap,” and “capture radius” to be 6 frames, 10 frames, and 50 nm, respectively. The max on time is the maximum number of frames in which the peaks are found within the capture radius. Thus, the max on time closely represents the fluorophore switching cycle, which is the number of appearances of a single fluorophore during dSTORM imaging. The obtained switching cycles under current experimental conditions were 4.1 and 6.7 for A565 and A647, respectively (fig. S9, H and I). The off gap is the maximum number of frames in which the peaks could be missing within the capture radius. Grouped localizations of consecutive frames within a capture radius of 50 nm (~2× of the average precision as determined in fig. S9E) were also calculated with the same software. Finally, the localizations were rendered as “render best quality” and with a “pixel resolution” of 10 nm with the Zen tools. The render best quality option renders the data subpixel precise. That is, the Gaussian distribution of the peaks is used to render the event such that it will be right on the localized position, even if the peak of the event appears off-center in a pixel of the camera. Inside the Zen software tool, the size of the 2D Gaussian (x/y) filter used for uniform rendering of the peaks is automatically set at the correct precision by default.

dSTORM data analysis

The processing and statistical analysis of dSTORM localization data were performed in MATLAB with custom-written codes, image processing toolbox, curve fitting toolbox, and optimization toolbox. First, we analyzed the pair autocorrelation function for both raw localizations [graw(r)] and grouped localizations [ggroup(r)] with established MATLAB code from Veatch et al. (38). Briefly, pair autocorrelation functions were calculated from 0 to 500 nm separation distance between the localization peaks (Fig. 1C) at a step of 10 nm within a randomly selected mask of size 3 × 3 μm, avoiding the corners and borders of the cells. As mentioned earlier, we kept the cells located at the center of the field, and the average diameter of each cell was 10 to 20 μm. We placed the 3 × 3 μm mask at any location from 0 to 2.0 μm from the center of the field. In each case, we could avoid the border of the cell by visualizing the cropped image within the mask region. In addition, we tested the long-range (up to 4 μm) autocorrelation with a larger mask (Figs. 2A and 3A). To evaluate the overcounting factors, we calculated gpsf(r) = graw(r) − ggroup(r), and fitted using the equation gpsf(r) = {1/(4πσ2ρav)} × exp{−r2/(4σ2)} and graw(r) = 1 + {1/(4πσ2ρraw)} × exp{−r2/(4σ2)}, where σ is the effective precision of measurement, and ρav and ρraw are the true density and average density of the probe. Note that the correlation function of the average, gpsf(r), refers to a PSF that occurs because of overcounting and the finite localization precision of the dSTORM measurement, not the actual PSF of the optical system. The observed average density of probe ρrawav is a measure of overcounting (Fig. 1C). With the same model and radial averaged autocorrelation functions, we calculated the average number of molecules per cluster (Fig. 1D) corrected for overcounting. In addition, we used the combined model containing a Gaussian and an exponential part (59) to fit the radial, averaged autocorrelation functions, and calculated the domain radius, the density of the probes in the domain (fig. S6B), and the number of multiple localizations from the same molecule. Assuming that the two antigen-binding sites of each BCR are equally accessible, we divided the fluorophore counts by two, thus obtaining the numbers of BCRs in each protein island. Furthermore, fitting the obtained overall BCR densities from images (fig. S6A) with the previously described surface area of a primary B cell and considering a twofold increase in cell surface area because of membrane ruffles (60), our estimate for the total number of BCRs on the surface differed from the predicted value of Mattila et al. (1), which could be due to the variable BCR abundances of different B cells. Pair cross-correlation functions, C(r), were calculated (with 10-nm steps) for the separation distance from 0 to 500 nm between the localization peaks (Fig. 1F) within a randomly selected mask of size 2 × 2 μm on the localizations of the IgM-BCR and IgD-BCR channels according to the method described by Veatch et al. (38). To verify the statistical significance of C(r), we measured the average values of the random distribution of the same localizations within the identical coverage area with a minimum of 30 simulations, and we calculated the 99% confidence interval of association and dissociation. For each C(r) analysis, we measured the average of the random distribution and the 99% confidence intervals of association and dissociation, and plotted together with the obtained C(r) values. A custom-written MATLAB code was used to calculate the C(r) from the pair of dSTORM localization coordinates for the two channels calculated from each image. Analysis of the Ripley’s K function derived H(r) function [L(r) − r] and its first-order derivative H′(r) was performed (with 10-nm steps) with a custom-written MATLAB code, which was corrected for the edge effect (39, 40). The rectangular regions of 2 × 2 μm from each set of dSTORM localization coordinates were selected as mentioned earlier and analyzed by Ripley’s method for the separation distance from 0 to 500 nm between the localizations. Bivariate Ripley’s analysis was performed with the pair of coordinates of the same 2 × 2 μm region of each image and was plotted with a 99% confidence interval obtained from the simulation.

Analysis of on-off duty cycle, switching cycle, and the precision of A565 and A647

To evaluate the blinking properties of A565 and A647, we coupled A565 and A647 individually to HEL. The resulting HEL-A565 and HEL-A647 molecules were placed on poly-d-lysine–coated glass-bottom dishes at low concentrations (100 to 10 nM) to achieve a single-molecule sparse condition. The average conjugation efficiency of HEL-A565 and HEL-A647 was 1.43 and 1.62 fluorophores per protein molecule, respectively. The HEL-A565– and HEL-A647–containing dishes were incubated for 30 min at room temperature, washed, and fixed with 4% PFA. Samples were then analyzed by dSTORM in an ELYRA PS.1 microscope using the TIRF field. A total of 6000 frames, with an exposure time of 40 ms, were recorded on an EMCCD camera. From the dSTORM images, we analyzed the overall molecular density within the image and pair autocorrelations for HEL-A565 and HEL-A647 (fig. S8, A and B). As expected, the detected molecular density and autocorrelation decreased from 100 to 10 nM for both HEL-A565 and HEL-A647. Analyzing the pair autocorrelation functions from the localization coordinates of sparse HEL-A565 and HEL-A647, we calculated the average number of fluorophores per cluster and the radius of each cluster. The obtained cluster radii were plotted against the number of fluorophores per cluster for the different concentrations of HEL-A565 and HEL-A647 (fig. S8C, left and right), and the mean values were calculated. A plot of average cluster radius versus number of fluorophores per cluster (fig. S8D) showed that at 10 nM antigen, the average numbers of fluorophores per cluster were 1.9 for HEL-A647 and 1.6 for HEL-A565, which were similar to their conjugation ratio (1.62 for HEL-A647 and 1.43 for HEL-A565). Therefore, at concentrations of 10 nM, HEL-A565 and HEL-A647 were distributed in a sparse condition that enabled the analysis of the corresponding fluorophores for their blinking properties as single-molecule events. In the experiments that followed, we therefore used 10 nM HEL-A565 and HEL-A647. Furthermore, the median and the range of precision of detection of A565 and A647 remained statistically significantly unaltered at different concentrations (fig. S8E, top and bottom). First, we calculated the average number of photons emitted per single particle by analyzing the frames of the dSTORM images. A plot of the average number of detected photons per single particle remained unaltered throughout the frame numbers for both HEL-A565 (fig. S8F, top) and HEL-A647 (fig. S8G, top). Second, we analyzed the average unbleached fraction and on-off duty cycle of the fluorophores and plotted these against the frame number (fig. S8, F and G, bottom). The unbleached fraction decreased over the length of the image. The reported average on-off duty cycles of A565 (0.006457) and A647 (0.006905) were calculated between frames 2000 and 3000. The on-off duty cycles were calculated within a sliding window of 1000 frames (40 s) at a step of 25 frames (1 s) with the formula described by Dempsey et al. (36). The complete length of the movie is 6000 frames. Third, we analyzed the number of switching cycles for A565 and A647 and plotted these against the frame number (fig. S8, H and I). The reported average switching cycles of A565 (4.13) and A647 (6.7) were calculated between frames 2000 and 3000. To estimate the precision of the localizations of A565 and A647, we retrieved the optical PSF width (s, PSF SD), number of detected photons (N), and local background photons (b) recorded by the microscope (fig. S9, A to C). With these values, we first measured the blind estimate localization precisions (fig. S9D) with a standard method, σ (precision) = (s/√N), as suggested by Rees et al. (46, 61). Second, we measured the more refined estimate of localization precision by taking into account the background photons with Thompson’s method (fig. S9E) (62). Note that the reported precision from the ELYRA PS.1 system, which we mentioned earlier, was also calculated by the same Thompson’s estimate for individual localization.

Analysis of grouping parameters for A565 and A647

To evaluate the grouping radii and gap intervals (off gap) for A565 and A647, we plotted the percentage of remaining localizations from three independent images after grouping them variably (fig. S9, G to J). We fitted the data by one-phase decay kinetics with the formula Y = {(Y0 − plateau) × exp(−k × X)} + plateau, and calculated the plateau, rate constant (k), and span (Y0 − plateau). With the obtained values of rate constant (k) and span (Y0 − plateau), we measured the percentage of overcount events that had undergone grouping at a selected limiting Sparrow’s resolution (2× precision) of 50 nm for A565. Similarly, we measured the effect of different off gap values for grouping the localizations at a radius of 50 nm. Finally, we used an off gap of 10 frames, which resulted in elimination of 99.9% of overcounts for the localizations of both A565 and A647.

Simulation of two-color BCR labeling

To compare the pair cross-correlation functions, we used two different null hypothesis models. First, we measured the average values of the random distribution of the same localizations within the identical coverage area with simulations and then we calculated the average and 99% confidence intervals of association and dissociation. We used this model for most of our figures and analysis of cross-correlation data including bivariate Ripley’s analysis. Second, the two-color random labeling process in which each binding site of the BCR could be equally occupied by either color was described (fig. S10B). Each BCR was randomly labeled with a pair of colors at a distance of ~5 to 10 nm from each other, corresponding to the separation between two binding sites on the BCR arms. Therefore, 25% of the total BCRs were labeled with two red colors, 25% with two green colors, and 50% with one red color and one green color (to give yellow). With this model, we compared the dSTORM imaging data of scIgM-BCR islands with differently labeled HEL molecules (fig. S2, D and E, panel vi). Simulation of the two-color BCR labeling process in resting and activated B cells was performed with the following input variables: total BCR surface density (fig. S10F, panel i), BCR density in each cluster (fig. S10F, panel ii), and the radii of BCR clusters in resting and activated B cells (fig. S10F, panel iii). The input functions were established by a random number generation process in MATLAB. The average of the input variables was constrained by the experimental results from Fig. 2 and figs. S6 and S7. The simulation process assigned the number of clusters within a fixed 2 × 2 μm area and calculated the number of BCRs per cluster by a sampling method (fig. S10G). The sum of all of the BCRs in all of the clusters was equal to the total number of BCRs within the coverage area (BCR surface density × 4 μm2). Half of the total number of BCRs was sampled and allocated in both red and green channels to obtain 50% yellow BCRs. The remaining half of the total BCRs was divided into two groups for allocating 25% each of red and green BCRs. The simulated images (fig. S10, H and I, panels i and ii) were either directly analyzed by pair cross-correlation or analyzed after filtering with a 25-nm Gaussian filter (fig. S10, H and I, panels iii and iv).

TEM analysis of membrane sheets

3046 cells expressing Myc-tagged IgM-BCRs and HA-tagged IgD-BCRs were maintained as described earlier. Plasma membrane sheets on EM grids were prepared as previously described (47). Briefly, cells were placed either for 1 hour on poly-l-lysine–coated EM grids or for 10 min on streptavidin-coated EM grids in the presence of biotinylated NIP15-BSA and then were ripped off. The obtained plasma membrane sheets were fixed with a mixture of 4% PFA and 0.2% glutaraldehyde for 15 min. For TEM analysis, fixed plasma membrane sheets were labeled with mixtures of mouse anti-HA (12CA5) and rabbit anti-Myc (Millipore) antibodies followed by gold-conjugated secondary anti-rabbit and anti-mouse antibodies. Samples were post-fixed in 2% (w/v) glutaraldehyde and then stained in sequence with 1% (w/v) OsO4 in 0.1 M cacodylate buffer, 1% (w/v) aqueous tannic acid, and 1% (w/v) aqueous uranyl acetate. EM grids were air-dried and then were examined with the Zeiss LIBRA 120 PLUS Energy Filter TEM microscope (Carl Zeiss Microscopy). Mapping of gold particle distribution and statistical analyses were performed as described previously (63). Briefly, the localization of the gold particles was determined by their centroids and used to calculate the H(r) function for the separation distance from 0 to 200 nm (30, 64). Bivariate Ripley’s analysis was performed with the pair of coordinates of 6- and 10-nm gold particles from each image, and these were plotted with a 99% confidence interval obtained from simulations.

Clustering analysis of TEM data

We analyzed the centroids of gold particles of TEM images by a simple nearest-neighbor clustering process in which the group members of a cluster were included on the basis of a given minimum distance threshold called an ND (49). With this method, one could measure the percentage of probes present in a cluster, which was identified by a nonlinear arrangement of at least three members; however, the process largely depends on a set value of ND (fig. S13A). We analyzed the cluster distribution with an ND of 50 nm and at least three members per cluster as previously described by Mattila et al. (1) and Espinoza et al. (49). Those linear arrangements of members in clusters whose area could not be calculated were excluded from clustering. With this condition, 49% of probes detecting IgM-BCR and 76% of probes detecting IgD-BCR remained in clusters (or protein islands) in resting 3046-MD cells (fig. S13B). We also analyzed the distribution of clustering area (nm2) and the number of probes per cluster (fig. S13, C and D).

Proximity ligation assays

Fab-PLA experiments were performed as previously described (23). In brief, Fab fragments from anti-IgM or anti-IgD antibodies were prepared by the Fab Micro Preparation Kit and desalted with Zeba spin columns (Thermo Fisher Scientific) according to the manufacturer’s protocol. The purified Fabs (20 to 50 μg) were coupled with PLA probemaker PLUS or MINUS oligonucleotides according to the manufacturer’s instructions (Olink Bioscience) to generate Fab-PLA probes (PLUS and MINUS). To perform the in situ PLA assay, cells were settled on a poly-d-lysine–coated glass slide for 5 min or on polytetrafluoroethylene slides (Thermo Fisher Scientific) for 30 min. Cells attached to the glass slide were treated with different stimuli and immediately fixed with 4% PFA in PBS at room temperature for 30 min. The PLA was performed as described previously (22, 65). The fixed cells on glass slides were incubated for 1 hour with a blocking solution containing BSA (250 μg/ml), sonicated salmon sperm DNA (2.5 μg/ml), 5 mM EDTA, and 0.05% Tween 20 in PBS. Cells were then incubated with the respective PLA probe pairs (1 μg/ml) in blocking buffer. Cells labeled with PLA probes were then sequentially incubated with the Duolink II ligation kit and amplification mixtures according to the manufacturers’ instructions. Finally, the PLA results were analyzed in a Zeiss 510 Meta confocal microscope (Carl Zeiss) with a Zeiss (Plan-Apochromat) 63× oil immersion objective lens and were quantified by counting the PLA dots corresponding to DAPI-stained nuclei with “BlobFinder” software, as described previously (44, 66). To verify that the Fab-PLA was able to detect isotype-specific BCRs, we first titrated the Fab reagents on naïve B cells carrying either IgM-BCRs or IgD-BCRs on their surface. With these cells, a Fab-PLA was performed with different concentrations of the IgM-IgM and IgD-IgD Fab-PLA probe pairs and was quantified by counting the PLA dots with the Blobfinder tool. For further Fab-PLA analysis of the IgM-IgD interaction in 3046-MD cells and B cells from B1-8 mice, we used combinations of one anti-IgM and one anti-IgD Fab-PLA probe (50 μg/ml each). With the two possible combinations, anti-IgM Fab PLUS:anti-IgD Fab MINUS (Fig. 5, A and B, IgM:IgD) or anti-IgD Fab PLUS:anti-IgM Fab MINUS (fig. S14, A and B, IgD:IgM), the proximity of IgM-BCRs to IgD-BCRs was analyzed in 3046-MD and B1-8 B cells.

Data processing and statistical analysis

Flow cytometry data acquisition was performed on an LSR II instrument controlled by FACSDiva software (BD Biosciences). Plotting and analysis of FACS data were performed in FlowJo 8.7 (Tree Star Inc.). Plotting and statistical analysis of PLA data were performed in Prism 5 (GraphPad).

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/8/394/ra93/DC1

Fig. S1. The Atto-coupled antigens bind to, but do not activate, TKO-HμNδ cells.

Fig. S2. dSTORM imaging of scIgM-BCR islands with differently labeled HEL molecules.

Fig. S3. Variability in the numbers of BCRs per island and in island size.

Fig. S4. Two-color dSTORM imaging of TKO-HμNδ cells on a fibronectin-coated surface.

Fig. S5. Viability, BCR abundance, Ca2+ flux, and morphology of Lat-A–treated B cells.

Fig. S6. The sizes of scIgM-BCR and IgD-BCR islands decrease in activated TKO-HμNδ cells.

Fig. S7. Activation of IgM and scIgM co-islands in TKO-HμNμ cells.

Fig. S8. Analysis of the blinking properties of A565 and A647 under sparse conditions.

Fig. S9. Analysis of localization precision and estimation of grouping radii and off gap.

Fig. S10. Simulation of two-color BCR labeling in resting and activated B cells.

Fig. S11. TEM analysis of IgM-BCR and IgD-BCR islands in resting and activated B cells.

Fig. S12. TEM analysis reveals that differently tagged IgD-BCRs coexist in the same islands.

Fig. S13. Segmentation analysis of IgM-BCR and IgD-BCR islands in TEM images.

Fig. S14. Proximity of IgM-BCRs and IgD-BCRs in activated mouse and human B cells.

REFERENCES AND NOTES

Acknowledgments: We thank F. Batista for providing complementary DNA encoding HyHEL10, and J. Heil of the Carl Zeiss Demo Center and A. Akhtar for providing access to the ELYRA PS.1 system. We thank J. Fitzpatrick, M. Jones, and S. Dunn of the Waitt Advanced Biophotonics Core at the Salk Institute for their services and help in TEM imaging. We thank P. Nielsen, J. Yang, and E. Hobeika for their support and critical reading of this manuscript. Funding: This study was supported by the Excellence Initiative of the German Federal and State Governments (EXC 294), by ERC-grant 322972, and by the Deutsche Forschungsgemeinschaft through SFB746 and TRR130. Author contributions: The experiments were planned and interpreted by P.C.M. and M.R.; super-resolution imaging and analysis, Monte-Carlo simulations, TEM segmentation, and PLA experiments were conducted and analyzed by P.C.M.; data analysis in MATLAB was performed by P.C.M. and O.R.; EM studies were conducted and analyzed by A.B. and B.L.; H.J. provided the TKO cell line; and the manuscript was prepared by P.C.M. and M.R. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Signaling

Navigate This Article