Intravital quantification of absolute cytoplasmic B cell calcium reveals dynamic signaling across B cell differentiation stages

Development, function and maintenance of lymphocytes largely depend upon the cellular mobilization and storage of Ca2+ ions. In B lymphocytes, the absolute amount of calcium mobilized and retained after cell signaling remains unknown, athough it is a crucial part of their selection within germinal centers and differentiation into plasma cells. Here, we introduce the novel reporter mouse strain YellowCaB that expresses the genetically encoded calcium indicator TN-XXL in CD19+ lymphocytes. The construct consists of the electrondonor fluorophore eCFP and the acceptor citrine, linked by a calcium sensitive domain. Its conformation and therefore donor quenching is directly linked to cytosolic calcium concentrations. By combining intravital two-photon fluorescence lifetime microscopy with our numerical approach for phasor-based analysis, we are able to extract absolute cytoplasmic calcium concentrations in activated B cells for the first time in vivo. We show that calcium concentrations in B cells are highly dynamic and fluctuations persist in extrafollicular B cells with functional relevance.


Introduction 28
The capacity of the immune system to produce a variety of different antibodies and to further 29 fine-tune their affinity to bind antigen (AG) upon pathogen challenge is one of the pillars of 30 adaptive humoral immunity.
Fine-tuning is achieved by somatic hypermutation of 31 immunoglobulin genes, followed by a T cell-aided selection process, which B cells undergo requirements 5-10 . BCR-affinity dependent AG-capture has been thought to serve solemnly 40 the processing and MHCII-dependent presentation to follicular helper T cells and that 41 signaling is dampened 11 . However, newer studies show that BCR activated calcium signaling 42 has to precede T cell derived signals and that the latter have to occur within a limited period 43 of time after initial BCR activation 12 . Changes in cytoplasmic calcium concentration thus 44 could provide a mechanistic link between BCR signal strength, the switch-on of downstream 45 effector processes and their temporal regulation. 46 In contrast to qualitative description, absolute quantification of cytosolic calcium has not been 47 achieved yet, partly because of the lack of internal concentration standards. Two-fluorophore 48 Förster resonance energy transfer (FRET)-GECI, that can take on a calcium-saturated 49 (quenched) and calcium-unsaturated (unquenched) condition, could overcome this, however, 50 its intravital application has been hampered by light distortion effects in deeper tissue. The 51 differential scattering and photobleaching properties of the two fluorophores would lead to a 52 false bias towards a higher quenching state. We here introduce a single-cell fluorescence 53 lifetime imaging (FLIM) approach for absolute calcium quantification in living organisms that 54 is tissue depth-independent. The eCFP/citrine-FRET pair-GECI TN-XXL is able to measure 55 fluctuations in cytoplasmic calcium concentration through the calcium binding property of the 56 muscle protein Troponin C (TnC) 13 . Calcium binding to the fluorophore-linker TnC quenches 57 eCFP fluorescence through energy scavenging by citrine, linking decreasing eCFP 58 fluorescence lifetime to increasing calcium concentration. In addition, phasor analysis of 59 FLIM data elegantly condenses multicomponent fluorescent decay curves into single vector-60 based information (the phasor) 14 . For calcium concentration analysis in microscopic images, 61 we took advantage of this by projecting the phasor value in each pixel onto a given 62 calibration 15 . With this method, we are able to describe short-and long-term changes in 63 absolute calcium concentrations within B cells during affinity maturation and differentiation 64 into antibody-producing plasma cells.  could induce an elevated FRET signal with a peak height of >30% that lasted over three 137 minutes (Fig 2a). The signal declines after this time span, probably due to BCR 138 internalization or the activity of ion pumps. We tested antibody concentrations at 2 µg/ml, 139 4 µg/ml, 10 µg/ml and 20 µg/ml. An antibody concentration of 2 µg/ml was not enough to 140 provoke calcium flux (data not shown), whereas at 4 µg/ml anti-IgM-F(ab) 2 we could observe 141 20% elevated ΔR/R over baseline (Fig. 2b). At 20 µg/ml anti-IgM-F(ab) 2 we could see no 142 further FRET increase ( Fig S2). Thus, we conclude a concentration dependency of the GECI 143 TN-XXL and saturating conditions at 10 µg/ml BCR heavy chain stimulation. Interestingly, 144 the reaction is not completely cut off after the FRET signal has declined, but a residual FRET 145 signal of about 7% compared to baseline values can be measured for approximately 3.5 146 additional minutes (Fig. 2a). Thus, B cells seem to be able to store extra calcium within the 147 cytoplasm for some time. We therefore asked if it is possible to stimulate YellowCaB cells 148 more than once. Indeed, we could stimulate YellowCaB cells in vitro repeatedly with F(ab) 2 -149 fragments of anti-IgM-antibody. For this purpose, we connected our imaging culture chamber 150 to a peristaltic pump and took advantage of the fact that under continuous perfusion with 151 Ringer solution, the flow will dilute the antibody out of the chamber. This way, it is possible to 152 stimulate B cells several times rapidly and subsequently, before BCRs are internalized (Fig.  153 2b), indicated by multiple peaks in ΔR/R. We repeated this procedure several times and 154 could observe this type of repetitive response up to five times. Also, stimulation of the BCR 155 light chain using an anti-kappa antibody leads to calcium increase within YellowCaB cells 156 (Fig. S2). Of note, the resulting FRET peak is shaped differently, and concentrations 157 >150µg/ml antibody were needed in order to generate a response. This might be in part due 158 to the fact that monovalent AG is not sufficient to drive BCR activation and the light chain has 159 different conformational properties. 160 Since T cell engagement and the binding of unspecific microbial targets to innate receptors 161 like Toll like receptors (TLRs) have also been described to raise cytoplasmic calcium in B 162 cells 19-21 , we investigated the response of YellowCaB cells after incubation with anti-CD40 163 antibodies, as well as the TLR4 and TLR9 stimuli Lipopolysaccharide and cytosine-164 phosphate-guanin-rich regions of bacterial DNA (CpG), respectively. Within the same cells, 165 we could detect no reaction to anti-CD40 treatment alone, but observed three types of 166 shapes in post-CD40 BCR-stimulated calcium responses, that differed from anti-CD40-167 untreated cells (Fig. 2a). These calcium flux patterns were either sustained, transient or of an 168 intermediate shape (Fig. 2c). Sustained calcium flux even saturated the sensor at a level 169 comparable to that achieved by ionomycin treatment. Cells that showed only intermediate 170 flux maintained their ability to respond to ionomycin treatment at high FRET levels, as 171 demonstrated by the ΔR/R reaching 0.4 again after stimulation (Fig. 2c). Furthermore, 172 integrated TLR and BCR stimulation affected the appearance of the calcium signal. The 173 addition of TLR9 stimulus CpG alone had no effect on YellowCaB FRET levels, however, the 174 subsequent FRET peak in response to anti-Ig-F(ab) 2 was delayed (Fig 2d+e). TLR4 175 stimulation via LPS could elevate calcium concentration of B cells, but only to a minor extent 176 (Fig 2e). Pre-BCR TLR4 stimulation by LPS lead to decreased FRET levels in response to 177 anti-IgM-F(ab) 2 . We conclude that, in order to get fully activated, B cells are able to collect 178 and integrate multiple BCR-induced calcium signals and that signaling patterns are further 179 shaped by innate signals or T cell help. BCR-inhibition abolishes a FRET signal change in 180 response to anti-IgM-F(ab) 2 ( Fig S2). Of note, we excluded the possibility that measured 181 signal changes were related to chemokine stimulation. In vitro, we could detect no FRET 182 peak after applying CXCL12, probably because of lacking GECI sensitivity to small 183 cytoplasmic changes (Fig. S2). Thus, the YellowCaB system is well suited for the detection 184 of BCR-induced cytosolic calcium concentration changes. immunofluorescence histology (Fig 3 a). At this time point, mice were surgically prepared for 208 imaging as described before 23 . Briefly, the right popliteal lymph node was exposed, 209 moisturized and flattened under a cover slip sealed against liquid drainage by an insulating 210 compound. The temperature of the lymph node was adjusted to 37°C and monitored during 211 the measurement. Our experiments revealed that the movement of single YellowCaB cells 212 can be tracked in vivo. Calcium fluctuations can be made visible by intensity changes in an 213 extra channel that depicts the FRET signal, as calculated from relative quenching of TN-XXL. 214 Color-coding of intensity changes in the FRET channel showed time-dependent fluctuations 215 of the signal and, in some particular cases, a sustained increase after prolonged contacts 216 between two YellowCaB cells (Fig 3 b, movie S1). Interestingly, FRET intensity seemed to be 217 mostly fluctuating around low levels in moving cells, whereas sustained increase required 218 cell arrest ( Fig S3). We already showed that signal changes in FRET of TN-XXL are 219 reflecting BCR activation. The observed calcium fluctuations might therefore coincide with 220

cell-to-cell contacts between FDCs and B cells, resulting in AG-dependent BCR stimulation. 221
To test for this, we first measured the colocalization between signals within the FDC-channel 222 and the citrine channel. The intensity of colocalization I coloc of all cells was plotted as a 223 function of frequency and biexponentially fitted (Fig 3 c). We set the threshold for a strong 224 and sustained colocalization of FDCs and B cells to an intensity of 150 AU within the 225 colocalization channel. At this value, the decay of the biexponential fit was below 10%. We 226 thus decided to term all cells with a colocalization intensity = 0 (naturally the most abundant 227 ones) not colocalized, cells with a colocalization intensity between 1 and 150 transiently 228 colocalized to FDCs ("scanning" or shortly touching the FDCs) and all cells above this 229 intensity threshold strongly or stably colocalized. When we compared the relative FRET 230 intensity changes ΔR/R of a tracked cell (Fig 3 b, cell 1), where baseline R is the lowest 231 FRET intensity measured, and its contacts to FDCs, we could indeed detect several transient 232 B-cell-FDC contacts that were followed by a step-wise increase of ΔR/R and thus an 233 increase of cytoplasmic calcium concentration (Fig 3 d). These increases in B cells were not 234 only restricted to contacts with FDCs, but also occurred between B cells: A visible contact of 235 cell 1 (Fig 3 b, cell 1) to a fellow B cell (Fig 3 b, cell 2) caused a sustained boost of the 236 calcium concentration in the tracked cell 1 (Fig 3d). Cell 2 itself kept strong FDC contact over 237 the whole imaging period and maintained elevated, but mostly stable ΔR/R values. These 238 experiments confirmed that GC B cells are able to collect calcium as a consequence of 239 repeated signaling events mediated by FDC-to-B cell contacts and, surprisingly, also by B-to-240

B cell contacts in vivo. 241
Next, we asked if the ability to perform BCR signaling is dependent on BCR affinity. Thus, we 242 adoptively transferred stained polyclonal, non-AG-specific YellowCaB cells one day prior to 9 intravital imaging and compared FRET-signal changes of several tracked cells over time 244 within the same measurement. Non-AG-specific YellowCaB cells showed a rather 245 homogenous distribution of calcium concentration with low intensity fluctuations in the FRET 246 channel around a mean of 13.88 AU, whereas AG-specific YellowCaB cells showed 247 heterogeneous signaling patterns with higher FRET intensities of 14.25 AU on average (Fig.  248 3e). We compared ratiometric FRET histograms of non-AG-specific and AG-specific cells out 249 of five different measurements. To do so, we normalized FRET values by the mean 250 fluorescence intensity averaged over all non-AG-specific YellowCaB cells. This further 251 confirmed a positive correlation of B cell calcium concentration and BCR-affinity ( Fig S4).

282
Employing our adoptive cell transfer set up described above (Fig 4 a) comprising plasma blasts (Fig. 4 b). Color-coded 2D and 3D FLIM analysis of these 290 populations confirmed that calcium concentrations were fluctuating within all of those B cell 291 populations, and that most B cells were maintaining relatively high fluorescence lifetimes and 292 therefore low calcium concentration on average with single-cell exceptions (Fig. 4 b, movie  293   S2). It should be noted that autofluorescence of the capsule or macrophages contributed to 294 tau values <800ps, indicated by dark red color-coding and is not to be attributed to high 295 calcium values. Phasor analysis and plotting of the B cell-wise segmented lifetime data 296 further confirmed the presence of cell clones with quenched TN-XXL, somewhat surprisingly 297 also among plasma blasts, which are thought to down-regulate their surface BCR (Fig 4c,  298 movie S3). To translate the measured lifetime values into absolute calcium concentrations, 299 we projected the data points onto the phasor connecting quenched and unquenched eCFP 300 (Fig. 4d). In this way, we corrected for artifacts acting on the phasor vectors and, implicitly, 301 on the fluorescence lifetimes caused by contribution of background noise (Fig S5). 302 Comparison of AG-specific cells inside GCs with those outside GCs, and non-AG-specific 303 cells inside GCs as well with those outside GCs showed that the distribution of calcium 304 concentrations of these B cells were dependent on BCR specificity and rather independent 305 from their location within the imaging volume, despite higher fluctuation seen among AG-306 specific populations (Fig S5). However, we noted the emergence of a cell subset that is high 307 in calcium and therefore located on the right half of the plot, in AG-specific cells and most 308 prominent among cells within the MC, as compared to non-AG-specific YellowCaB cells. 309 Overlay of an imaging snapshot shows that only few cells are in a state of elevated calcium 310 (>800 nM) at one given time point (Fig. 4e). Accordingly, these maxima were reached as 311 transient fluctuation peaks, i.e. periods of below one minute, in which these concentrations 312 seem to be tolerated. Calcium values exceeding the dynamic range of TN-XXL (>857nM) 313 were recorded for all measured subsets, but the most cells >857nM were found among 314 intrafollicular AG-specific B cells and extrafollicular AG-specific B cells. (Fig. 4f). The 315 heterogeneity in temporal calcium concentrations therefore is smallest among non-AG 316 specific B cells, increases with activation in AG-specific GC B cells, and is most prominent 317 among plasma blasts. Thus, a progressive heterogeneity of calcium signals within B cells 318 can be seen alongside the process of activation and differentiation. 319  directly leads to a decrease in calcium concentration (Fig 5b). This was further confirmed by 340 bulk analysis of all detected YellowCaB cells over the whole imaging period (Fig 5c). Relating 341 calcium concentration and colocalization showed that the calcium concentration in 342 YellowCaB cells with direct contact to SCSM reaches values that are more than doubled 343 compared to that in cells that were not in contact Furthermore, calcium concentration   Intensity density values of the raw citrine signal were divided by the intensity density values 501 of the raw eCFP signal and related to the baseline ratio of the signals before stimulation. 502 Cell transfers, immunization and surgical procedures 503 B cells from spleens of YellowCaB mice were negatively isolated using the Miltenyi murine B 504 cell isolation kit via MACS. 5x10 6 cells were transferred to a host mouse with a transgenic B 505 cell receptor specific for an irrelevant antigen (myelin oligodendrocyte glycoprotein). B cells 506 from spleens of B1-8 hi :YellowCaB mice were transferred to WT C57/Bl6 mice. Host mice 507 were immunized in the right footpad with 10 µg NP-CGG in complete Freund's adjuvant 24h 508 after B cell transfer. After six to eight days p.i., FDCs were labeled with Fab-Fragment of 509 CD21/35-Atto590 or CD21/35-Alexa647 (inhouse coupling) into the right footpad. 24h after 510 labelling, the popliteal lymph node was exposed for two-photon-imaging as described 511 before 23 . Briefly, the anesthetized mouse is fixed on a microscope stage custom-made for 512 imaging the popliteal lymph node. The mouse is shaved and incisions are made to introduce 513 fixators that surround the spine and the femoral bone. The mouse is thus held in a planar 514 position to the object table. The right foot is fixed by wire allowing to increase the tension on 515 the leg to position the lymph node parallel to the imaging set up. A small incision is made to 516 the popliteal area. The lymph node is exposed after freeing it from surrounding adipose 517 tissue. A puddle around the lymph node is formed out of insulating silicon compound, then 518 filled with NaCl and covered bubble-free with a cover slide. 519

Intravital and live cell imaging and image analysis 520
Imaging experiments of freshly isolated B cells were carried out using a Zeiss LSM 710 521 confocal microscope. Images were acquired measuring 200-600 frames with 1 frame/3s 522 frame rate while simultaneously detecting eCFP and citrine signals at an excitation 523 wavelength of 405 nm. 524 For intravital two-photon ratiometric imaging, Z-stacks were acquired over a time period of 525 30-50 min with image acquisition every 30 seconds. eCFP and citrine were excited at 850nm 526 by a fs-pulsed Ti:Sa laser and fluorescence was detected at 466±30 nm or 525±25 nm, 527 respectively. Fluorescence signals of FDCs were detected in a 593±20 nm channel. Analysis of two-photon data 538 For ratiometric analysis of two-photon data (here the exciting wavelength of 850 nm is also 539 stimulating citrine fluorescence directly), fluorescence signals were corrected for spectral 540 overlap (the eCFP to citrine ratio in 525±25 nm channel is 0.52/0.48) and refined by taking 541 into account the sensitivity of photomultiplier tubes (PMTs, 0.37 for 466±30 nm and 0.4 for 542 525±25 nm). Ratiometric FRET for in vivo experiments was accordingly calculated as 543 = 1,2• ℎ2 2,7• ℎ1+2,5• ℎ2 • 100 (Eq. 1) 544 Evaluation of FLIM data was performed using the phasor approach 14,24 . Briefly, the 545 fluorescence decay in each pixel of the image is Fourier-transformed at a frequency of 546 80 MHz and normalized, resulting into a phasor vector, with the origin at (0|0) in a cartesian 547 system, pointing into a distinct direction within a half-circle centered at (0.5|0) and a radius of Gaussian-fitted the histograms of real and imaginary part after the transformation to the 563 frequency domain. The Gaussian fit of each part gives a mean, which indicates the center of 564 the noise distribution, as wells as the width ( = 2√2 2 ), which was the same for 565 both parts (real and imaginary axis) and gives us the radius (½FWHM = 0.2 (green, solid arc 566 in Fig. S5) within which we expect only noise, i.e. the signal-to-background-ratio is unreliable 567 small. In order to increase the accuracy, we enlarged the radius to ¾FWHM = 0.3 (green, 568 dashed arc in Fig. S5) and all data points within that radius were excluded from further 569 analysis. A second filtering was applied by determining the signal-to-noise ratio (SNR) from 570 the summed TCSPC signal for all segmented cells. In Imaris, a sphere of radius=20µm was 571 determined around each cell to establish a reference value for background noise. The SNR 572 was then calculated as follows (pixel by pixel): 573 20 − (Eq. 3) 574 With I sig being the intensity of TCSPC signal, I bg the intensity of background noise and STD bg 575 the background noise standard deviation 69 AG specific signals with SNR<2 and non-Ag 576 specific signals with SNR<1 were excluded from the analysis. All other phasor data points 577 be 265nM-857nM. All values below or above these margins are subject to uncertainty and 583 therefore simply referred to as <265nM or >875nM, respectively. 584

Statistical information 585
Time dependent FRET curve analysis shows representative graphs for the number of 586 analyzed cells and independent experiments given. For multiple curve analysis, mean is 587 shown and SD indicated in each data point. For column analysis, One-way ANOVA with 588 Bonferroni Multiple Comparison Test was applied with a confidence Interval of 95%. 589

Data availability 590
All raw data and analyzed data shown here are stored on institutional servers and may be 591 accessed upon request to the corresponding author. 592

Code availability 593
Python-based code for phasor analysis can be provided upon request.