Intravital quantification reveals dynamic calcium concentration changes across B cell differentiation stages

Calcium is a universal second messenger present in all eukaryotic cells. The mobilization and storage of Ca2+ ions drives a number of signaling-related processes, stress–responses, or metabolic changes, all of which are relevant for the development of immune cells and their adaption to pathogens. Here, we introduce the Förster resonance energy transfer (FRET)-reporter mouse YellowCaB expressing the genetically encoded calcium indicator TN-XXL in B lymphocytes. Calcium-induced conformation change of TN-XXL results in FRET-donor quenching measurable by two-photon fluorescence lifetime imaging. For the first time, using our novel numerical analysis, we extract absolute cytoplasmic calcium concentrations in activated B cells during affinity maturation in vivo. We show that calcium in activated B cells is highly dynamic and that activation introduces a persistent calcium heterogeneity to the lineage. A characterization of absolute calcium concentrations present at any time within the cytosol is therefore of great value for the understanding of long-lived beneficial immune responses and detrimental autoimmunity.


Introduction
During generation of humoral immunity to pathogens, calcium-mobilizing events in lymphocytes can communicate such diverse outcomes as migration, survival, stress responses, or proliferation. An elevation of cytoplasmic calcium from external space is mostly mediated through ligand binding to surface receptors (Baba et al., 2014). Especially in the germinal center (GC), where B cells fine-tune their B cell receptor (BCR) in order to become positively selected by T cells, ligand density in the form of native antigen (AG), stimuli for toll-like receptors (TLR), or chemokine receptors, is high. Selected B cells that leave the GCs fuel the pool of memory B cells and long-lived plasma cells (LLPCs). These cells produce high-affinity antibodies granting up to lifelong protection against threats such as infectious diseases, but also can account for the persistence of an autoimmune phenotype, when selection within the GC is impaired (Berek et al., 1991;da Silva et al., 1998;Hiepe et al., 2011;Victora and Nussenzweig, 2012). B cell activation by AG uptake through the BCR promotes calcium influx into B cells (Tolar et al., 2009). Calcium mobilization eventually switches on effector proteins and transcription factors like nuclear factor kappa B, nuclear factor of activated T cells, or myelocytomatosis oncogene cellular homolog, thereby inducing differentiation events and remodeling of metabolic requirements (Crabtree and Olson, 2002;Jellusova, 2018;Luo et al., 2018;Saijo et al., 2002;Su et al., 2002). Dependent on the amount of AG taken up and the quality of major histocompatibility complex II (MHCII)-mediated presentation to T follicular helper cells, B cells receive additional, costimulatory signals (Gitlin et al., 2015). Interestingly, recent studies propose that costimulatory signals have to occur within a limited period of time after initial BCR activation and that the limit is set by a calcium threshold, eventually leading to mitochondrial dysfunction (Akkaya et al., 2018). Thus, quantification of changes in absolute cytoplasmic calcium concentration tolerated by GC B cells would help to understand how B cell selection in the GC is accomplished.
In contrast to qualitative description, absolute calcium measurements in B cells have not yet been performed in vivo, partly because of the lack of internal concentration standards. Two-fluorophore genetically encoded calcium indicators (GECI) relying on Fö rster resonance energy transfer (FRET) can take on a calcium-saturated (quenched) and calcium-unsaturated (unquenched) condition, overcoming this issue. However, intravital application of quantitative FRET has been hampered by light distortion effects in deeper tissue. The differential scattering and photobleaching properties of the two fluorophores would lead to a false bias towards a higher quenching state. We here introduce a single-cell fluorescence lifetime imaging (FLIM) approach for absolute calcium quantification in living organisms that is tissue depth-independent. Both time-domain and frequency-domain FLIM technologies have been employed in the past 30 years to sense changes in pH, ionic strength, pO 2 , metabolism, and many other cellular parameters within living cells. Due to the impact of these parameters on the chemical structure, the fluorescence lifetime of the analyzed fluorophores can change (Elson et al., 2004;Lakowicz et al., 1992;Le Marois and Suhling, 2017). Particularly, the versatility of FLIM to quantify FRET quenching has been demonstrated and applied in various biological contexts (Chen et al., 2003;Levitt et al., 2020;Mossakowski et al., 2015). The enhanced cyan fluorescent protein (eCFP)/citrine-FRET pair-GECI TN-XXL is able to measure fluctuations in cytoplasmic calcium concentration through the calcium binding property of the muscle protein troponin C (TnC) (Mank et al., 2006). Calcium binding to the fluorophore-linker TnC quenches eCFP fluorescence through energy transfer to citrine (FRET), linking decreasing eCFP fluorescence lifetime to increasing calcium concentration. Whereas eCFP fluorescence lifetime changes with refractive index (Strickler and Berg, 1962) and may change upon large shifts in pH value, ionic strength, oxygenation, or temperature, these parameters hardly vary in the cytosol of living cells. Thus, we expect only changes in cytosolic calcium concentration to have an impact on the fluorescence lifetime of eCFP as donor in the TN-XXL construct. In addition, phasor analysis of FLIM data elegantly condenses multicomponent fluorescent decay curves into single vector-based information (the phasor) (Digman et al., 2008). For calcium concentration analysis in microscopic images, we first took advantage of the previously published titration curve of TN-XXL by Geiger et al., which we also confirmed in our experimental setup (Geiger et al., 2012). We further adapted the phasor-based calibration strategy to quantify calcium levels in vivo proposed by Celli and colleagues to the TN-XXL construct expressed in B lymphocytes (Celli et al., 2010). With this method, we are able to describe shortand long-term changes in absolute calcium concentrations within B cells during affinity maturation and differentiation into antibody-producing plasma cells.
We here describe the calcium reporter mouse strain 'YellowCaB' (termed after energy transfer to the yellow fluorescent protein citrine in case of calcium present in the cytosol of B cells). These mice express cytosolic TN-XXL in all CD19-positive cells. Intravital FLIM of adoptively transferred Yellow-CaB cells shows that calcium concentrations are highly dynamic in B cells involved in the GC reaction. We describe different patterns of calcium fluctuation regarding amplitude and baseline within non-activated and AG experienced cells and plasma blasts. We observe the emergence of Ca 2+ -high differentiated B cells and plasma blast populations, which might point to cells undergoing metabolic stress.

YellowCaB: a system for FRET-based calcium analysis in B cells
Mice expressing a loxP-flanked STOP sequence followed by the TN-XXL-construct inserted into the ROSA26 locus were crossed with the CD19-Cre strain (Rickert et al., 1997). The offspring had exclusive expression of the GECI TN-XXL in CD19 + B lymphocytes, as confirmed by visualization of eCFP and citrine fluorescence by confocal microscopy after magnetic B cell isolation (Figure 1a, b). These YellowCaB cells were excited with a 405 nm laser that is capable of exciting eCFP but not citrine. The detection of yellow emission thus can be attributed to baseline FRET representing steady-state calcium levels. Expression of TN-XXL in YellowCaB mice was further confirmed by flow cytometry after excitation with the 488 nm laser and detection in a CD19 + GFP + (green fluorescent protein) gate that would also detect citrine fluorescence. Citrine was found to be present in a substantial part of CD19 + B lymphocytes and was not detectable in the CD19population ( Figure 1c). Cd19 cre/+ mice heterozygous for TN-XXL and Cd19 cre/+ mice homozygous for TN-XXL did not differ in the proportion of cells within the CD19 + GFP + population, nor did male and female mice (Figure 1-figure supplement 1). In addition, no differences in total cell numbers and B cell numbers between Cd19 cre/+ Tn-xxl +/-, Cd19 cre/+ Tn-xxl +/+ , and wild-type mice were detected (Figure 1-figure supplement 1). We next set out to test if we could induce a FRET signal change under calcium-saturating conditions in the cytoplasm. The ionophore ionomycin is commonly used as positive control in in vitro experiments measuring calcium concentrations as it uncouples the increase of calcium concentration from the physiological entry sites of Ca 2+ ions by forming holes in the cell membrane. When stimulated with ionomycin, a steep increase of the FRET level over baseline was recorded by flow cytometry in the GFP-channel after excitation with the 405 nm laser. Calcium-dependence of the signal increase was further independently confirmed by staining with the calcium sensitive dye X-Rhod-1, that shows a red fluorescence signal increase after calcium binding (Figure 1d; Li et al., 2003).
In preparation of our intravital imaging experiments, we first tested if the YellowCaB system is stable enough for time-resolved microscopic measurements and sensitive enough for subtle cytoplasmic calcium concentration changes as they occur after store-operated calcium entry (SOCE). In SOCE, stimulation of the BCR with AG leads to drainage of intracellular calcium stores in the endoplasmic reticulum (ER), which triggers calcium influx from the extracellular space into the cytosol through specialized channels (Baba et al., 2014). We established a customizable perfusion flow chamber system to monitor and manipulate YellowCaB cells over the duration of minutes to hours ( Figure 1e). Division of the fluorescence intensity of electron acceptor citrine by that of donor eCFP yields the FRET ratio (R), which is then put into relationship to baseline FRET levels. As expected, we detected a decrease of the eCFP signal, concurrent with an increased citrine fluorescence after the addition of 4 mg/ml ionomycin to continuous flow of 6 mM Krebs-Ringer solution. Overall, this resulted in a maximal elevation of DR/R of 50-55% over baseline ( Figure 1f). Analysis of >100 cells showed that approximately in three quarters of the cells we were able to detect FRET in response to ionomycin treatment, and that the majority of these cells showed 35-40% FRET signal change. According to the two populations visible in the histogram, we defined a change of 20% DR/R as a relevant threshold for the positive evaluation of responsiveness ( Figure 1g). In conclusion, we achieved the functional and well-tolerated expression of TN-XXL exclusively in murine CD19 + B cells for measurement of changes of cytoplasmic calcium concentrations.
Repeated BCR stimulation results in fluctuating cytoplasmic calcium concentrations SOCE in B cells can be provoked experimentally by stimulation of the BCR with multivalent AG, for example, anti-Ig heavy chain F(ab) 2 fragments. To test the functional performance of the GECI TN-XXL in YellowCaB cells, we stimulated isolated YellowCaB cells with 10 mg/ml anti-IgM F(ab) 2 fragments to activate the BCR. In an open culture imaging chamber, we induced an elevated FRET signal with a peak height of >30% that lasted over 3 min (Figure 2a). The signal declines after this time span, probably due to BCR internalization or the activity of ion pumps. We tested antibody concentrations at 2, 4, 10, and 20 mg/ml. An antibody concentration of 2 mg/ml was not enough to provoke calcium flux (data not shown), whereas at 4 mg/ml anti-IgM-F(ab) 2 we observed 20% an elevated DR/ R over baseline (Figure 2b). At 20 mg/ml anti-IgM-F(ab) 2 , we detected no further FRET increase (Figure 2-figure supplement 1a, b). Thus, we conclude a concentration dependency of the GECI TN-XXL and saturating conditions at 10 mg/ml BCR heavy chain stimulation. Interestingly, the reaction is not completely cut off after the FRET signal has declined, but a residual FRET signal of about 7% compared to baseline values was measured for approximately 3.5 additional minutes ( Figure 2a). Thus, B cells seem to be able to store extra calcium within the cytoplasm for some time. We therefore wondered if it is possible to stimulate YellowCaB cells more than once. For this purpose, we connected our imaging culture chamber to a peristaltic pump and took advantage of the fact that under continuous perfusion with Ringer solution the flow will dilute the antibody out of the chamber. This way, it is possible to stimulate B cells several times rapidly and subsequently, before BCRs are internalized, indicated by multiple peaks in DR/R (Figure 2b). In addition, stimulation of the BCR light chain using an anti-kappa antibody led to calcium increase within YellowCaB cells (Figure 2figure supplement 1a). Of note, the resulting FRET peak is shaped differently, and concentrations > 150 mg/ml antibody were needed in order to generate a response.
Since T cell engagement and the binding of microbial targets to innate receptors like TLRs have also been described to raise cytoplasmic calcium in B cells (Ojaniemi et al., 2003;Pone et al., 2015;Ren et al., 2014), we investigated the response of YellowCaB cells after incubation with anti-CD40 antibodies, as well as the TLR4 and TLR9 stimuli lipopolysaccharide (LPS) and cytosine-phosphate-guanine-rich regions of bacterial DNA (CpG), respectively. Within the same cells, we detected no reaction to anti-CD40 treatment alone, but observed three types of shapes in post-CD40 BCRstimulated calcium responses, which differed from anti-CD40-untreated cells (Figure 2a). These calcium flux patterns were either sustained, transient, or of an intermediate shape ( Figure 2c). Sustained calcium flux even saturated the sensor at a level comparable to that achieved by ionomycin treatment. Cells that showed only intermediate flux maintained their ability to respond to ionomycin treatment at high FRET levels, as demonstrated by the DR/R reaching 0.4 again after stimulation ( Figure 2c). Furthermore, integrated TLR and BCR stimulation affected the appearance of the calcium signal. The addition of TLR9 stimulus CpG alone had no effect on YellowCaB FRET levels; however, the subsequent FRET peak in response to anti-Ig-F(ab) 2 was delayed (Figure 2d TLR4 stimulation via LPS could elevate calcium concentration of B cells, but only to a minor extent ( Figure 2e). When TLR4 stimulation by LPS was performed before BCR stimulation, decreased FRET levels in response to anti-IgM-F(ab) 2 were observed. We conclude that, in order to become fully activated, B cells are able to collect and integrate multiple BCR-induced calcium signals and that signaling patterns are further shaped by innate signals or T cell help. BCR inhibition abolishes a FRET signal change in response to anti-IgM-F(ab) 2 ( Figure 2-figure supplement 1b). Of note, we excluded the possibility that measured signal changes were related to chemokine stimulation. In vitro, we detected no FRET peak after applying CXCL12, probably because of lacking GECI sensitivity to small cytoplasmic changes ( Figure 2-figure supplement 1c, d). Thus, the YellowCaB system is well suited for the detection of BCR-induced cytosolic calcium concentration changes.
Fluctuating calcium levels are observed as a result of sequential cell contacts in vivo We next set out to investigate if the ability of B cells to collect calcium signals sequentially is also shared by GC B cells. For two-photon intravital imaging, nitrophenyl (NP)-specific B1-8 hi B cells from YellowCaB mice were magnetically isolated and transferred into wild-type hosts, which were subsequently immunized with NP-chicken gamma globulin (CGG) into the right foot pad (Shih et al., 2002a). Mice were imaged at day 8 p.i. when GCs had been fully established. Activated TN-XXL + YellowCaB cells had migrated into the GC, as confirmed by positive PNA-and anti-FP-immunofluorescence histology (Figure 3a). At this time point, mice were surgically prepared for imaging as described before . Briefly, the right popliteal lymph node was exposed, moisturized, and flattened under a cover slip sealed against liquid drainage by an insulating compound. The temperature of the lymph node was adjusted to 37˚C and monitored during the measurement. Our experiments revealed that the movement of single YellowCaB cells is traceable in vivo. Calcium fluctuations can be made visible by intensity changes in an extra channel that depicts the FRET signal, as calculated from relative quenching of TN-XXL. Color coding of intensity changes in the FRET channel showed time-dependent fluctuations of the signal and, in some particular cases, a sustained increase after prolonged contacts between two YellowCaB cells ( Figure 3b, Video 1). Interestingly, FRET intensity seemed to be mostly fluctuating around low levels in moving cells, whereas sustained increase required cell arrest, as reported previously (Negulescu et al., 1996;Shulman et al., 2014), (Figure 3-figure supplement 1). The observed calcium fluctuations might therefore coincide with cell-to-cell contacts between follicular dendritic cells (FDCs) and B cells, resulting in AG-dependent BCR stimulation. To test for this, we first measured the colocalization between signals within the FDC channel and the citrine channel. The intensity of colocalization I coloc of all cells was plotted as a function of frequency and biexponentially fitted ( Figure 3c). We set the threshold for a strong and sustained colocalization of FDCs and B cells to an intensity of 150 AU within the colocalization channel. At this value, the decay of the biexponential fit was below 10%. We thus decided to term all cells with a colocalization intensity = 0 (naturally the most abundant ones) not colocalized, cells with a colocalization intensity between 1 and 150 transiently colocalized to FDCs ('scanning' or shortly touching the FDCs), and all cells above this intensity threshold strongly or stably colocalized. We compared the relative FRET intensity changes DR/R of two tracked cells (Figure 3b, cells 1 and 2), where baseline R is the lowest FRET intensity measured, and its contacts to FDCs. We detected several transient B-cell-FDC contacts in cell 1 that were followed by a gradual increase of DR/R, indicating an increase of cytoplasmic calcium concentration ( Figure 3d). Cell 2 kept strong FDC contact over the whole imaging time and maintained elevated, mostly stable values. These experiments confirmed that GC B cells are able to collect calcium as a consequence of repeated contact events, which are mediated by B cell-to-FDC contacts in vivo.

Calibration of the TN-XXL construct for in vivo quantification of cytosolic calcium in B lymphocytes using the phasor approach to FRET-FLIM
The comparison of calcium responses in different B cell subsets of multiple GCs requires normalization of TN-XXL FRET. Since this is hardly achievable in tissue due to its inherent heterogeneity, we aimed for the determination of absolute cytosolic concentration values in YellowCaB cells by calibration of TN-XXL FRET intensities. However, ratiometric calibration would require equal conditions for donor and acceptor fluorescence signals, especially in terms of scattering and photobleaching. Due to the aforementioned heterogeneous tissue composition, these requirements cannot be met in vivo . Therefore, donor FLIM is the appropriate solution as it circumvents comparative evaluation of different fluorescence signals. Fluorescence lifetime is defined as the mean time a fluorescent molecule stays in an elevated energetic state after excitation, before photon emission and relaxation to the ground state take place. As a fully calcium-quenched eCFP in the GECI TN-XXL would transfer its energy mainly to citrine, its fluorescence lifetime would be measurably shorter than that of unquenched eCFP. Time-correlated single-photon counting (TCSPC) devices offer the possibility of simultaneous photon detection and recording of the respective fluorescent lifetimes within a nanosecond scale, yielding a fluorescence lifetime decay histogram of photons. Deriving fluorescent lifetimes t from the eCFP decay curve of photon histograms requires fitting. We decided to use the phasor analysis as it virtually transfers time-resolved fluorescence data into phase domain data by discrete Fourier transformation (Digman et al., 2008). This approach overcomes the obstacles of multicomponent exponential analysis and yields model-free, readily comparable pixelor cell-based plots that assign a position within a semicircle to each data point, dependent on the mixture of lifetime components present (Leben et al., 2018).
Typically, data correction based on reference dyes is needed for reliable phasor analysis (Ranjit et al., 2018). We verified the reliability of our TCSPC setup to acquire high-quality fluorescence decays in an image to be evaluated using the phasor approach by measuring the instrument response function given by the second harmonic generation signal (SHG) of potassiumdihydro-phosphate crystals (laser excitation wavelength 940 nm) and the fluorescence decays of eGFP, expressed in HEK cells (laser excitation wavelength 900 nm). As shown with the phasor plots of the raw data ( Figure 4a), both the SHG signal and the eGFP fluorescence are located in expected positions on the semicircle (Murakoshi et al., 2008;Rinnenthal et al., 2013). Therefore, no further correction of the data is necessary in our system.
As the TN-XXL construct is exclusively expressed in the cytosol of B lymphocytes from YellowCaB mice, the following equilibrium holds true for Ca 2+ (free cytosolic calcium), TNXXL (the calcium-free FRET construct, i.e., the unfolded tertiary structure of TnC), and Ca 2+ TNXXL (the FRET construct saturated by calcium, i.e., the completely folded tertiary structure of TnC): characterized by the dissociation constant K d : As measured in lysate of YellowCaB B cells expressing TN-XXL, the fluorescence lifetime t of the donor eCFP of the FRET Ca-sensitive construct TN-XXL depends on the free calcium concentration [Ca 2+ ] following a sigmoidal function (Eq. 3).
with t free the fluorescence lifetime eCFP at 0 nM free calcium, and t FRET the fluorescence lifetime of completely FRET-quenched eCFP in the TN XXL-construct at 39 mM free calcium. By fitting the fluorescence lifetime of eCFP in TN-XXL excited at 850 nm and detected at 460 ± 30 nm acquired in time domain using our TCSPC system at various free calcium concentrations, we determined K d = 475 ± 46 nM and Hill slope = À1.43 ± 0.17 ( Figure 4b). Thus, we can calculate the free calcium concentration as Hill slope (4) Similar to the calculation in time domain, for the phase domain we can express each phasor vector p ! based on the formalism proposed by Celli and colleagues (Celli et al., 2010) as  with e free and e FRET the relative brightness (Chen et al., 1999;Müller et al., 2000) of eCFP in TNXXL at 0 nM and saturated (39 mM) free calcium, given by the following equations: with d CFP the effective two-photon absorption cross section of eCFP (independent of the pathways of relaxation from the excited state), and k F the fluorescence rate of eCFP in vacuum, that is, no quenching due to surrounding molecules. The phase vectors can be written also as complex numbers as given by the following equations: We determined the averages and median real and imaginary values of the phasor distributions obtained by performing FRET-FLIM in lysates of YellowCaB B cells at 0 nM and 39 mM free calcium to be Re free = 0.4035 (average), 0.40326 (median); Im free = 0.45801 (average), 0.45779 (median) and Re FRET = 0.82377 (average), 0.82203 (median); Im FRET = 0.3225 (average), 0.32093 (median), respectively, indicating that both distributions are symmetric, corresponding to normal distributions ( Figure 4c).
(2), the free cytosolic calcium concentration is given by Thus, the free calcium concentration depends only on the K d , t free , t FRET as determined from the calibration curve measured in lysate and the phase vectors, describing the extreme states of eCFP in the TN-XXL construct.
Since variations in refractive index, ion strength, pH value, or temperature in the cytosol of the B lymphocytes may additionally influence the fluorescence lifetime of eCFP, as well as the phase  (Jameson et al., 1984;Scott et al., 1970), we verified whether the FRET trajectory of the-TN-XXL construct changes in the cytosol of B lymphocytes in cell culture (Figure 4figure supplement 1) and under in vivo conditions, in lymph nodes (Figure 4d, e). We found that in all our measurements the phasor cloud lays on the trajectory determined in lysate solutions (Figure 4e). Measurements performed at different depths in lymph nodes led to the same result: there is no change in the slope of the trajectory at different tissue depths (Figure 4f).
To assess the impact of autofluorescence on the interpretation of the fluorescence signal in the phasor plot and, thus, on the cytosolic calcium levels, we also performed FLIM in B cell follicles of lymph nodes of non-fluorescent wild-type mice. While the acquired signal was extremely low, the phasor cloud in these measurements was located around position (0,0) in the plot, indicating that it mainly originates from detector noise (Figure 4g).
We compared the results of cytosolic free calcium concentration determined using the timedomain and the phase-domain approach and found deviations of max. 5% between the evaluation pathways using Eqs. (4) and (11) due to numerical uncertainty. We determined the calcium dynamic range of TN-XXL measured by phasor-analyzed FRET-FLIM to span between 100 nM and 4 mM free calcium, with the linear range of the titration curve in the range between 265 and 857 nM free calcium.

Comparative FLIM-FRET reveals heterogeneity of absolute calcium concentrations in B cells in vivo
Adoptively transferring AG-specific YellowCaB cells and non-AG-specific (polyclonal) YellowCaB cells (stained ex vivo) allowed us to divide GC B cells into five different populations based on their location in the imaging volume and their fluorescent appearance (Figure 5a, b). At day 8 p.i., polyclonal YellowCaB cells, identified by their red labeling, mostly lined up at the follicular mantle around the GC, with some of them having already entered into activated B cell follicles. AG-specific, citrine-positive B1-8 hi :YellowCaB cells were found clustered in the GC, close to FDCs, outside of GC boundaries, or as bigger, ellipsoid cells in the extrafollicular medullary cords (MC), probably comprising plasma blasts (Figure 5b, left). Color-coded 2D-and 3D FLIM analysis of these populations confirmed that calcium concentrations were fluctuating within all of those B cell populations, and that most B cells were maintaining relatively high mean eCFP fluorescence lifetimes and therefore low calcium concentrations on average, with only few exceptions (Figure 5b, middle and right, Video 2).
Bulk analysis of cells revealed additional calcium-intermediate and calcium-high cell subsets present among AG-specific cells and plasma blasts (Figure 5c, Video 3). Especially for plasma blasts this was somehow unexpected, given that these are thought to downregulate their surface BCR during differentiation.
Comparison of AG-specific cells inside GCs with those outside GCs and non-AG-specific cells inside GCs as well with those outside GCs showed that the distribution of calcium concentrations of these B cells was dependent on BCR specificity and rather independent from their location within the imaging volume, despite higher fluctuation seen among AG-specific populations ( Figure 5-figure supplement 1). These maxima were reached as transient fluctuation peaks, that is, periods shorter than 1 min, in which these concentrations seem to be tolerated. Calcium values exceeding the dynamic range of TN-XXL (>857 nM) were recorded for all measured subsets, but most cells >857 nM were found among intrafollicular AG-specific B cells and extrafollicular AG-specific B cells (Figure 5d). The heterogeneity in temporal calcium concentrations therefore is smallest among non-AG-specific B cells, increases with activation in AG-specific GC B cells, and is most prominent among MC-plasma blasts. The high cytoplasmic calcium levels observed in MC-plasma blasts were unlikely to be the result of chemokine-induced signaling since only a minor calcium increase was detected after CXCL12 stimulation in vitro (Figure 2-figure supplement 1d). Thus, a progressive heterogeneity of calcium signals within B cells can be seen alongside the process of activation and differentiation.

Functional relevance of increased calcium concentration among extrafollicular YellowCaB cells
We next wondered if high cytoplasmic calcium levels within certain B cell subsets could be the result of AG-mediated signals. We therefore intravenously (i.v.) injected NP-bovine serum albumine (BSA) (66 kDa) into mice that had been adoptively transferred with B1-8 hi (AG-specific) YellowCaB cells and recorded absolute calcium concentrations within the GC by intravital FLIM. Antigens up to 70 kDa have been reported to be transported into the follicles via conduits in less than 5 min (Roozendaal et al., 2009). Accordingly, AG-specific GC YellowCaB cells significantly upregulated cytoplasmic calcium after AG injection within minutes (Figure 6a, top panel, Figure 6-figure supplement 1). To test if this concentration increase could be explained via BCR-dependent AG recognition, we pre-injected the inhibitor of Bruton's tyrosine kinase (BTK) ibrutinib (Hendriks et al., 2014), which blocks BCR downstream activation, in a control group. BTK inhibition could abrogate the increase in mean cytoplasmic calcium after additional injection of NP-BSA (Figure 6a, bottom panel) and even seemed to downregulate baseline signals ( Figure 6-figure  supplement 1). In addition, we found that also in 48 hr LPS/ interleukin 4 (IL-4)-cultured B1-8-plasma  G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G G  blasts, an increase in calcium was detectable after addition of NP-BSA ( Figure 6-figure supplement 2), suggesting that stimulation via AG remains possible in at least a proportion of these differentiated B cells. These results led us to investigate the possible sources of AG abundance outside of GCs and their effect on calcium in B cells. Earlier studies proposed that one possible AG source in lymph nodes are subcapsular sinus macrophages (SCSM) Moran et al., 2018;von Andrian and Mempel, 2003). We tested if SCSM contacts could be the cause of elevated calcium levels in extrafollicular B cells. We intravitally imaged wild-type host mice that have been adoptively transferred with B1-8 hi :YellowCaB cells and received an injection of efluor660-labeled anti-CD169 antibody together with the usual FDC labeling 1 day prior to analysis. We concentrated on the area beneath the capsule, identified by second harmonic generation signals of collagen fibers in this area. Thresholds of colocalization between CD169 + macrophages and TN-XXL + Yel-lowCaB cells are described in Figure 6-figure supplement 3. Together, these methods led to a 3D visualization of the SCS with CD169 stained macrophages, lined up in close proximity (Figure 6b). AG-specific YellowCaB cells were detected clustering in GCs nearby. Extrafollicular YellowCaB cells crowding the SCS space were found to have multiple contact sites to SCSM. Some B cells were observed to migrate along the SCS, possibly scanning for antigenic signals (Video 4). Bulk analysis of YellowCaB cells and their colocalization with SCSM showed that the calcium concentration in YellowCaB cells with direct contact to SCSM reaches values that are more than doubled compared to values in cells that were not in contact, and that calcium concentration is positively cor-Video 2. Side-by-side depiction of fluorescence, fluorescence lifetime imaging, and cell-based phasor data of intravitally imaged germinal center (day 8 p.i.), single z-plane. Left: fluorescence data with antigen (AG)-specific YellowCaB cells (green) and stained non-AG-specific YellowCaB cells transferred 1 day prior to imaging (red; autofluorescence of capsule also visible in the same channel). 4 fps, scale bar 50 mm. Middle: false color-coded presentation of fluorescence lifetime t (0-3000 ps, see range scale in Figure 5b). 4 fps, scale bar 50 mm. Right: raw cellbased phasor plot with cells segmented according to fluorescence and spatial distribution, subsets indicated. 4 fps.
https://elifesciences.org/articles/56020#video3 Video 4. 3D projection of intravital imaging of germinal center and subcapsular sinus. YellowCaB cells (green, with track lines) and subcapsular sinus macrophages (red), stained by CD169 in vivo labeling. White circles highlight antigen-specific B cells migrating along subcapsular space. 4 fps, scale bar 50 mm. https://elifesciences.org/articles/56020#video4 related with contact strength (Figure 6c, top). Single-cell tracking and simultaneous analysis of absolute calcium concentration and colocalization intensity revealed that the increase of cytoplasmic calcium is a direct cause of B-cell-to-SCSM contacts (Figure 6c, bottom). We conclude that contacts of B cells to SCSM could induce elevation of B cell cytoplasmic calcium concentrations, presumably due to antigenic activation, with the absolute concentrations being dependent on the contact strength.

Discussion
Intravital imaging technologies have contributed greatly to a better understanding of the dynamic processes in GCs. AG-capture, cycling between zones, and development of clonality patterns have been made visible by two-photon microscopic techniques (Hauser et al., 2007;Suzuki et al., 2009;Tas et al., 2016;Victora et al., 2010). Furthermore, important functional in vivo data like signaling in T helper cells have been collected using a calcium-sensitive protein (Kyratsous et al., 2017;Shulman et al., 2014  However, calcium mobilization within GC B cells was mostly investigated via ex vivo analysis of sorted cells after adoptive transfer and immunization, or BCR activation was measured using a nonreversible BCR signaling reporter like Nur77, altogether neglecting calcium flux (Khalil et al., 2012;Mueller et al., 2015). These data suggested that BCR signaling in the GC is reduced. However, no statement was made about the dynamics and timely coordination of (even small) calcium pulses and the relation to their microenvironment. In fact, a recent study confirmed that BCR signals play a central role in positive selection and display a fragile interdependence with costimulatory events (Luo et al., 2018).
BCR-regulating surface proteins like CD22 or sialic acid-binding immunoglobulin-type lectins have been related to development of autoimmunity and point out BCR-mediated calcium flux as an important component, not only during B cell development but also in their differentiation to effector cells (Hoffmann et al., 2007;Jellusova et al., 2010;Müller and Nitschke, 2014;Nitschke and Tsubata, 2004;O'Keefe et al., 1999). Apart from regulating gene transcription, cytosolic calcium mobilized has been shown to be essential for F-actin remodulation and B cell spreading on antigen presenting cells (Maus et al., 2013). Furthermore, cytosolic calcium concentration is closely linked to metabolic reprogramming of activated B cells and their cell fate (Boothby and Rickert, 2017;Caro-Maldonado et al., 2014). It has been shown that SOCE is acting directly on the mitochondrial capability to import cytosolic calcium (Shanmughapriya et al., 2016). In mitochondria, calcium is regulating ATP production through increase of glycolysis and fatty acid oxidation, processes for which activated and GC B cells have high demands, although there is controversy as to which of the two metabolic pathways predominates in GC B cells (Griffiths and Rutter, 2009;Jellusova et al., 2017;Maus et al., 2017;Weisel et al., 2020).
For flexible analysis of calcium mobilization in cells of the CD19 + lineage, we developed a novel transgenic reporter system and image processing approach, enabling quantification of cytosolic calcium concentrations. The FRET-based GECI TN-XXL can be used stably in moving, proliferating, and differentiating lymphocytes, and the reversibility of the sensor makes it suitable for longitudinal intravital measurements. Switching from ratiometric acquisition of FRET-donor and FRET-acceptor fluorescence intensities to measuring FRET-donor fluorescence lifetime enabled quantification of calcium concentrations in absolute numbers.
A first advantage of FRET-FLIM in tissue is that different photobleaching or scattering properties of the fluorophores can be neglected. We further decided to perform all analyses based on the phasor approach that circumvents the problem of multiexponential fluorescence decays we encounter measuring a two-fluorophore FRET-based GECI in tissue. For titration of TN-XXL, we used lysate of YellowCaB plasma blasts. Besides TN-XXL affinity, also eCFP fluorescence lifetime itself may be influenced by large shifts in pH value, ionic strength, oxygenation, or temperature. We ensured that these parameters were similar in lysates and cells, except for the temperature, which was at room temperature for calibration. Temperature was reported to only slightly change the fluorescence lifetime of a CFP variant (cerulean) (Laine et al., 2012). However, for accuracy purposes and in order to exclude such artifacts when determining cytosolic calcium levels in B lymphocytes within lymph nodes of YellowCaB mice, we adapted the phasor-based calibration strategy proposed by Celli et al., 2010 for the use of Calcium Green in skin samples to our data. In this way, we were able to reliably determine absolute values of cytosolic calcium concentrations in B cells within lymph nodes.
In our set-up, we have shown that TN-XXL in B cells has suitable sensitivity and fast reversibility. This key factor made it possible to observe repeated and partially sustained calcium elevation in the cytoplasm, showing that B cells are able to collect sequential signals, possibly up to a certain threshold, which determines their fate.
In support of that, B cellular calcium concentration must not constitutively exceed a certain value in order to prevent mitochondrial depolarization (Akkaya et al., 2018;Bouchon et al., 2000;Niiro and Clark, 2002). Gradual calcium elevation could be a mechanistic link for that. For example, calcium levels of >1 mM over the duration of >1 hr have been reported to be damaging to other cell types, such as neurons Siffrin et al., 2015). Accordingly, stimulation of AG receptors via large doses of soluble AG can lead to tolerogenic apoptosis in GC B cells, which could be explained by uninhibited calcium influx (Nossal et al., 1993;Pulendran et al., 1995). Since apoptosis is the default fate for B cells in the GC reaction (Mayer et al., 2017), CD40 and TLR signaling might contribute to limiting cytoplasmic calcium concentrations, and thus promote survival of B cell clones with appropriate BCR affinity (Akkaya et al., 2018;Eckl-Dorna and Batista, 2009;Pone et al., 2015;Pone et al., 2012;Pone et al., 2010;Ruprecht and Lanzavecchia, 2006). For CD40 signaling in immature B cells, this has been confirmed (Nguyen et al., 2011). Our data does show that TLR signaling can attenuate calcium flux in stimulated B cells, while CD40 can either attenuate or augment the calcium response ( Figure 2). Whether the different outcomes of CD40 stimulation are dependent on the affinity of the BCR and its efficiency in presenting AG (Schwickert et al., 2011;Shulman et al., 2013) will be subject of further studies.
Measuring absolute calcium concentration in GC B cells after administration of soluble AG in vivo, we could detect an increase of B cell calcium that is attenuated by BCR inhibitor ibrutinib, showing that BCR-mediated calcium increase is substantially contributing to calcium heterogeneity in the GC. However, the interpretation of the data should not neglect other causes of calcium elevation, given the multifaceted role of this second messenger. Therefore, it is likely that apart from BCR signaling, also other events, like binding of non-AG ligands and stress-related calcium release from internal stores, contribute to an overall cytosolic calcium increase, which needs to be regulated in order to prevent a damaging calcium overload. Causes for stress-related cytosolic calcium elevations in cells can be hypoxia, a condition reported to be present within GCs (Jellusova et al., 2017); nutrient deprivation, which mostly will affect highly proliferative cells like GC B cells; or ER-calcium release as a result of the unfolded protein response that is indispensable in plasma cells (Díaz-Bulnes et al., 2020;Høyer-Hansen and Jäättelä, 2007;Lam and Bhattacharya, 2018). The complex interaction of factors makes an exact characterization of the absolute calcium concentration in various B cell subsets crucial in order to obtain information about their regulation and containment. This characterization should preferably be done intravitally since any manipulation of cells can result in enormous non-physiological variations of cytosolic calcium levels.
For the first time, we determined absolute values of B cell cytoplasmic calcium concentrations during the GC reaction within living mice. It appears that BCR AG specificity and state of differentiation are closely related to distinct degrees of heterogeneity of calcium concentrations. Notably, heterogeneity was also evident in extrafollicular B cells in the SCS region, as well as in plasma blasts. The latter actually reach the highest calcium concentrations within the B cell compartment of the lymph node. We also observed an increase of cytoplasmic calcium in plasma blasts after stimulation with specific AG in vitro. These data are in line with a report of residual BCR signaling occurring in antibody-secreting cells (Pinto et al., 2013), which challenges the finding that B cells completely downregulate their surface BCR during differentiation to plasma cells (Manz et al., 1998). Our experiments were done in short-lived plasma blasts, not LLPC, for which the situation could be different. Therefore, an investigation of possible BCR signaling in LLPC is of high interest. Stimulation with the chemokine CXCL12, which has previously been shown to induce migration of antibodysecreting cells Hauser et al., 2002), resulted only in a minor increase of cytoplasmic calcium in plasma blasts in our hands.
In B cells that have exited the GC, ongoing calcium flux might reflect reactivation. We confirmed that B cells in contact to SCSM had significantly higher cytosolic calcium concentrations. These are possibly attributed to BCR signaling since the SCS has been proposed as a site of reactivation of memory B cells via AG (Moran et al., 2018).
The YellowCaB system provides a tool for measuring calcium as ubiquitous, universal cellular messenger, integrating signals from various pathways, including chemokine receptor signaling and intrinsic calcium release or BCR-triggered activation. Importantly, changes in mitochondrial membrane potential and/or the integrity of the ER also lead to varying calcium concentrations within the cytoplasm since both act as major intracellular calcium buffering organelles (Kass and Orrenius, 1999) A close connection between mitochondrial calcium homeostasis, altered reactive oxygen speciesproduction, and the expression of plasma cell master transcription factor BLIMP1, as well as changes in metabolism, has been reported previously (Jang et al., 2015;Shanmugapriya et al., 2019). We have recently applied phasor-FLIM of endogenous NAD(P)H fluorescence for mapping of metabolic enzyme activities in cell cultures (Leben et al., 2019). The combination of this technique with FLIM-based intravital calcium analysis will help to further dissect immunometabolic processes in B cells, as well as in short-lived plasma cells and LLPCs in vivo.

Mice
Mice carrying a STOP cassette flanked by two loxP sites upstream of the region encoding for the TN-XXL biosensor (Mank et al., 2008) in the ROSA26 locus were obtained from F. Kirchhoff, Saarland University, Homburg. YellowCaB mice were generated by crossing those mice with the Cd19 cre/ cre strain (Rickert et al., 1997) and maintained on a C57Bl/6 background. Only YellowCaB mice heterozygous for Cre were used to avoid deletion of CD19. Mice with monoclonal NP-specific BCR (B1-8 hi :YellowCaB) were generated by crossing of YellowCaB mice with B1-8 hi mice (Shih et al., 2002b). All mice were bred in the animal facility of the DRFZ. All animal experiments were approved by Landesamt fü r Gesundheit und Soziales, Berlin, Germany, in accordance with institutional, state, and federal guidelines.

Cells
Primary splenocytes were isolated from whole spleens of YellowCaB mice or B1-8 hi :YellowCaB mice in 1Â PBS and erythrocytes lysed. B cells were negatively isolated using the Miltenyi murine B cell isolation kit via magnetic-assisted cell sorting (MACS), leaving B cells untouched in order not to prestimulate them.

Staining and flow cytometry
Single-cell suspensions were prepared and stained according to the guidelines for flow cytometry and cell sorting in immunological studies (Cossarizza et al., 2019). To simultaneously assess calcium influx with a dye-based method, we stained whole splenocytes or isolated B cells with the calciumsensitive dye X-Rhod-1 (Invitrogen). X-Rhod-1 is a single-fluorophore calcium reporter molecule that enhances its fluorescence intensity upon calcium binding in a range of 0-40 mM up to 100 times at a wavelength of 600 nm. Measurements were carried out at a BD Fortessa flow cytometer. TN-XXL expression was checked assessing positive fluorescence in a 525 ± 25 nm channel after 488 nm excitation on a MACSQuant flow cytometer. Cell transfers, immunization, and surgical procedures B cells from spleens of YellowCaB mice were negatively isolated using the Miltenyi murine B cell isolation kit via MACS. 5 Â 10 6 cells were transferred to a host mouse with a transgenic BCR specific for an irrelevant AG (myelin oligodendrocyte glycoprotein). When NP-specific B cells were analyzed, B cells from spleens of B1-8 hi YellowCaB mice were transferred to wildtype C57Bl/6 mice. Host mice were immunized in the right footpad with 10 mg NP-CGG in complete Freund's adjuvant 24 hr after B cell transfer. After 6-8 days p.i., FDCs were labeled with Fab-Fragment of CD21/35-Atto590 or CD21/35-Alexa647 (in-house coupling) into the right footpad. Polyclonal B cells from Yel-lowCaB mice were stained with a red fluorescent dye (CellTracker Deep Red, Thermofisher) and adoptively transferred. 24 hr later, the popliteal lymph node was exposed for two-photon imaging as described before . Briefly, the anesthetized mouse is fixed on a microscope stage custom-made for imaging the popliteal lymph node. The mouse is shaved and incisions are made to introduce fixators that surround the spine and the femoral bone. The mouse is thus held in a planar position to the object table. The right foot is fixed by a wire allowing to increase the tension on the leg to position the lymph node parallel to the imaging set-up. A small incision is made to the popliteal area. The lymph node is exposed after freeing it from surrounding adipose tissue. A puddle around the lymph node is formed out of insulating silicon compound, then filled with NaCl and covered bubble-free with a cover slide.

Perfusion chamber
For intravital application of AG, 500 mg NP-BSA were administered i.v. after acquiring 10 time steps of baseline FLIM signal (four mice). To check for BCR specificity of calcium elevation recorded after AG injections, BTK inhibitor ibrutinib was pre-injected i.v. before AG application at 3.75 mg/kg and recorded in a control group of three mice. Technically, measurements were paused for injections for about 5 min. For accurate comparison of baseline calcium levels with calcium levels after injections, we imaged the same GC before and after. Therefore, measurements are always to be treated as separate measurements and it is not possible to track individual cells before and after an injection. However, the mean calcium elevation in the presented subset of cells can be visualized. For better comparability, we have chosen to present a time course that virtually combines two consecutive measurements in the same GC into one ( Figure 5-figure supplement 1). Up to 22 individual cell tracks were randomly chosen after gating out overlapping signals from macrophages, filtering for maximum track duration and completeness of the series of events (absolute calcium value in the center of the segmented object).

Intravital and live cell imaging and image analysis
Imaging experiments of freshly isolated B cells were carried out using a Zeiss LSM 710 confocal microscope. Images were acquired measuring 200-600 frames with one frame/3 s frame rate while simultaneously detecting eCFP and citrine signals at an excitation wavelength of 405 nm.
For intravital two-photon ratiometric imaging, z-stacks were acquired over a time period of 30-50 min with image acquisition every 30 s. eCFP and citrine were excited at 850 nm by a fs-pulsed Ti:Sa laser, and fluorescence was detected at 466 ± 30 nm or 525 ± 25 nm, respectively. Fluorescence signals of FDCs were detected in a 593 ± 20 nm channel. For experiments including macrophage staining, the fluorescence data has been unmixed for a possible overlap of the TN-XXL-citrine signal with that of the injected marker to prevent false-positive colocalization analysis between the red efluor660 coupling of anti-CD169 and the green fluorescence of TN-XXL in the 525 ± 25 nm channel (Rakhymzhan et al., 2017).
For intravital FLIM experiments, eCFP fluorescence lifetime was measured with a time-correlated single-photon counting system (LaVision Biotec, Bielefeld, Germany). The fluorescence decay curve encompassed 12.4 ns (80 MHz laser repetition rate) with a time resolution of 55 ps. The pixel dwell time was 4 Â 5 ms, allowing to detect photons from 1600 laser pulses for the fluorescence decay. The fluorescence decay, while being multiexponential, may be approximated by a bi-exponential function containing the two monoexponential decays of unquenched CFP and of FRET-quenched CFP, respectively. The phasor approach allows us to display the data prior to data interpretation graphically, that is, prior to the decision on the multiexponentiality of the CFP decay function, and was primarily used for the FRET-FLIM data evaluation. Similar to fluorescence intensity two-photon experiments, we performed time-lapse FRET-FLIM measurements and repeated the described acquisition every 30 s.

Analysis of two-photon data
For ratiometric analysis of two-photon data, fluorescence signals were corrected for spectral overlap (the eCFP to citrine ratio in 525 ± 25 nm channel is 0.52/0.48) and refined by taking into account the sensitivity of photomultiplier tubes (0.37 for 466 ± 30 nm and 0.4 for 525 ± 25 nm). Ratiometric FRET for in vivo experiments was calculated accordingly as Evaluation of FLIM data was performed using the phasor approach (Digman et al., 2008;Leben et al., 2018). Briefly, the fluorescence decay in each pixel of the image is Fourier-transformed at a frequency of 80 MHz and normalized, resulting into a phase vector, with the origin at (0|0) in a Cartesian system, pointing into a distinct direction within a half-circle centered at (0.5|0) and a radius of 0.5. For pure substances, vectors end directly on the half-circle, for mixtures of two on a connecting segment between the respective pure lifetimes and within a triangle, if three substances are present, and so on. The distance between several fluorescence lifetimes on the half-circle is naturally distributed logarithmically, with the longest lifetimes closer to the origin. In the case of TN-XXL, the extremes are the unquenched CFP fluorescence (2312 ps) and the eCFP fluorescence completely quenched by FRET (744 ps). The location of the measured lifetime on the connecting line can directly be translated into the amount of either eCFP state and thus to a corresponding calcium concentration in each pixel of the image.
At low signal-to-noise ratio values, the FLIM signal of the donor with a large contribution of electronic noise is shifted towards the origin of coordinates in the phasor plot, indicative for the infinite lifetime of noise. In non-fluorescent medium, we measured electronic noise and Gaussian-fitted the histograms of real and imaginary parts. The Gaussian fit of each part gives the mean (distribution center) as well as the full distribution width at half maximum, FWHM = 2 ffiffiffiffiffiffiffiffiffi 2ln2 p s, which was the same for both parts. The width of electronic noise distribution gives the radius within which we expect only noise (Figure 4g). In order to increase the accuracy, we excluded all data points in an area within the radius of ¾FWHM = 0.3.

Titration of TN-XXL construct
Sensor calibration was performed using lysate of cultured, homozygous B1-8 hi :YellowCaB plasma blasts induced from isolated B1-8 hi :YellowCaB cells stimulated with LPS/IL-4 for 2 days. Briefly, cells were freeze-thawed in liquid nitrogen 3-4 times and treated with ultrasound for 15 min. Lysate was filtered and cell clumps separated by high-speed centrifugation. Lysis was done at two equal shares in a sufficiently small volume of calcium-free calibration buffer (Life Technologies) or calcium-saturated (39 mM CaEGTA), respectively. Calcium buffer concentrations measured were achieved by dilution of 39 mM-buffered cell lysate with 0 mM-buffered cell lysate. Sample concentrations were loaded into glass microscope slides with recess, covered, and their fluorescence was measured at the twophoton microscope in a time-resolved manner. Focus was adjusted to a z-position with maximal photon counting numbers, as ensured by acquisition of proper decay curves. In time domain, the eCFP mean fluorescence lifetimes t at various free calcium concentrations were determined by approximating the corresponding fluorescence decay curves F(t) acquired with our TCSPC-based FLIM detector to a monoexponential function containing background y 0 : F t ð Þ ¼ y 0 þ A Á e À t t . Fitting was performed iteratively using a Levenberg-Marquardt gradient algorithm (Rinnenthal et al., 2013).

Statistical information
Time-dependent FRET curve analysis shows representative graphs for the number of analyzed cells and independent experiments given. For multiple curve analysis, mean is shown and SD indicated in each data point. For column analysis, one-way ANOVA with Bonferroni multiple comparison test was applied with a confidence Interval of 95%.

Data availability
All raw data and analyzed data shown here are stored on institutional servers. Imaging source data and raw Excel files have been deposited at https://datadryad.org under DOI: 10.5061/dryad. cc2fqz63d.
supported by the Deutsche Forschungsgemeinschaft (DFG) TRR130, project 17 (to AEH and HR) and C01 (to AEH and RAN), and DFG SFB 1444, project 14 (to AEH and RAN). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Ethics Animal experimentation: The study was approved by the Berlin Landesamt fü r Gesundheit und Soziales under the registration # G00158/16. All surgeries and experimental procedures were conducted following the principle of minimization of suffering and 3R means were used where possible.