Kindlin‐3 maintains marginal zone B cells but confines follicular B cell activation and differentiation

Integrin‐mediated interactions between hematopoietic cells and their microenvironment are important for the development and function of immune cells. Here, the role of the integrin adaptor Kindlin‐3 in B cell homeostasis is studied. Comparing the individual steps of B cell development in B cell‐specific Kindlin‐3 or alpha4 integrin knockout mice, we found in both conditions a phenotype of reduced late immature, mature, and recirculating B cells in the bone marrow. In the spleen, constitutive B cell‐specific Kindlin‐3 knockout caused a loss of marginal zone B cells and an unexpected expansion of follicular B cells. Alpha4 integrin deficiency did not induce this phenotype. In Kindlin‐3 knockout B cells VLA‐4 as well as LFA‐1‐mediated adhesion was abrogated, and short‐term homing of these cells in vivo was redirected to the spleen. Upon inducible Kindlin‐3 knockout, marginal zone B cells were lost due to defective retention within 2 weeks, while follicular B cell numbers were unaltered. Kindlin‐3 deficient follicular B cells displayed higher IgD, CD40, CD44, CXCR5, and EBI2 levels, and elevated PI3K signaling upon CXCR5 stimulation. They also showed transcriptional signatures of spontaneous follicular B cell activation. This activation manifested in scattered germinal centers in situ, early plasmablasts differentiation, and signs of IgG class switch.


INTRODUCTION
The correct differentiation of B cells requires their appropriate localization to distinct organs, zones, and signals in a sequential process. Integrins support leukocyte development and differentiation by orchestrating cell extravasation from the blood circulation and retention within the hematopoietic organs. The two major integrins expressed on B cells are the heterodimers alpha4beta1 (α4β1, a.k.a. very late Ag-4 (VLA-4), CD49d/CD29) and αLβ2 (lymphocyte functionassociated antigen-1 (LFA-1), CD11a/CD18). 1  The ligand-binding function of integrins requires their transition from a low into a high-affinity conformation. This transition, triggered by the so-called inside-out integrin activation cascade, allows rapid and dynamic changes in integrin-mediated cell adhesion. The process is initiated by cell surface receptors such as chemokine receptors, the B cell receptor (BCR), or CD44 that bind chemokines, Ags, or hyaluronic acid, respectively, [7][8][9] and subsequently involves several kinases to activate the integrin. The high-affinity conformation of the integrin is stabilized by at least two types of intracellular integrin adaptors, talins and kindlins, which bind independently to different motifs in the cytoplasmic tail of β integrin subunits. 10,11 The kindlin family comprises three members, namely Kindlin-1, Kindlin-2, and Kindlin-3 (K3), with K3 being exclusively expressed in the hematopoietic system. 12 Hematopoietic stem-and progenitor cells, T-cells, neutrophils and platelets require K3-mediated integrin activation for their correct positioning and function. [13][14][15][16] It has not yet been studied, whether K3 plays a role in the differentiation of normal B cells or their malignant counterparts, which are highly dependent on VLA-4 activation, 9,17 and elucidation of the function of K3 in B cells might bring new insights into mechanisms not only of lymphomagenesis but also of autoimmunity.
Here, we used the mb1 Cre transgene to create mouse strains harboring a K3 or CD49d/VLA-4 deletion in pre-pro B cells and all later stages. 18

Study approvals and mice
All mouse experiments were approved by the animal ethics com- Purity was confirmed to be >90% by flow cytometry. For all experiments, fl/fl littermates expressing no Cre were used as wildtype (wt) controls.

Immunofluorescence
Spleens were frozen in TissueTek (Miles Laboratories) and stored at −80 • C before cutting them to 10-μm cryostat sections. Sections were fixed in methanol, blocked using 1% BSA, and incubated with primary Abs at 4 • C, followed by respective secondary Abs. Primary Abs were directed against laminin (Novus), B220 (BD), Ki67 (R&D systems),

Real-time analysis of VLA-4 affinity
Real-time flow cytometry of VLA-4 affinity was performed as described. 23 The autofluorescence baseline of purified splenic B cells was determined using an Accuri flow cytometer (BD Biosciences).
After 1 min measurement, the small molecule VLA-4 ligand LDV-FITC was added to the sample and the fluorescence intensity was recorded for an additional 2 min. Then, CXCL12 (100 ng/ml), CXCL13 (500 ng/ml) or anti-IgM F(ab) 2 (10 μg/ml) was added. After another 2 min, unlabeled LDV was added in excess. Measurement continued for further 3 minutes. The dissociation rate constant (K off ), was obtained by fitting the data to a one-phase exponential decay equation. The ligand residence time was calculated as 1/K off as described. 24

In vitro shear flow assay
Shear flow assays were performed as previously described. 17,22 Plates were coated with protein A followed by immobilization of VCAM-1-Fc or ICAM-1-Fc and subsequent overlay with chemokine (CXCL12).
Shear stress was generated with an automated syringe pump attached to the outlet side of a parallel laminar flow chamber (6-channel μslides, Ibidi). Purified splenic B cells were perfused in the flow chamber, the entire perfusion period was recorded and digitalized. Anal-ysis of the video-recorded segments was performed using a customized software (WIMASIS). Frequencies of adhesive categories were determined as percentages of cells flowing immediately over the substrates.
Blood was taken 1 week after tamoxifen administration; mice were sacrificed 2 weeks after administration. Marginal zone and follicular B cells were quantified by flow cytometry.

Chemotaxis assay
Freshly isolated splenocytes were subjected to Boyden chamber transwells (Corning, 5 μm pore size) for 2 h, with the lower well-containing medium alone (control) or medium containing CXCL12 (100 ng/ml) or CXCL13 (500 ng/ml). Chemotaxis was cytometrically quantified using flow counting beads; B cells were identified by CD19 staining.

Calcium mobilization assay
Whole splenocytes were labeled with Fluo-3-AM (Thermo Fisher) and subsequently stained on ice with Abs against B220, CD24, and CD21, allowing for the identification of follicular B cells. Prior to measurement, cells were warmed to 37 • C. Fluorescence was acquired for 3 min on BD Cytoflex at 37 • C. After one minute, the stimuli (anti-IgM (10 μg/ml), anti-kappa (10 μg/ml), anti-IgD (10 μg/ml), CXCL12 (100 ng/ml), or CXCL13 (500 ng/ml) were added. Fluorescence over time was plotted using FlowJo, and the area under the curve was determined graphically using ImageJ.

Polarization analysis
B cells from spleens were isolated by negative selection to purity >90% and seeded on chamber slides coated with either Collagen I (Ibidi) alone or co-coated with CXCL12 or CXCL13 and allowed to adhere in HBSS containing 0.5% BSA, 1 mM CaCl 2 , and MgCl 2 for 30 min. The cells were then fixed and stained with Phalloidin-AF488 (Thermo Fisher).
Images were taken by an MRm mono-color camera at 20× PL NeoFluar objective (numerical aperture 0.5) of an AxioObserver.Z1 (Zeiss) using Zen software (Zeiss). Cellular polarization was classified according to the pattern of phalloidin staining and analyzed using Image J (NIH).

Statistical analysis
Statistical analysis was performed with GraphPad Prism 5. Normal distribution was tested for all datasets. Normally distributed groups were compared using unpaired or paired t-test, not-normally distributed data sets using Mann-Whitney or Wilcoxon matched-pairs signed-rank test. Significance threshold was P < 0.05, with values at P < 0.05 marked as *, P < 0.01 as **, and P < 0.001 as ***.

K3 as well as CD49d expression are required to maintain immature and mature B cells in the bone marrow
We intercrossed mice expressing the Cre-recombinase under the control of the mb1 (CD79a) promoter 18

Upon Kindlin-3 deletion, marginal zone B cells are lost while follicular B cells are expanded
Having observed that both CD49d integrin and its intracellular adaptor

Upon Kindlin-3 deletion B cells home preferentially to spleen
We next delineated the mechanisms of MZ B cell abrogation and follicular B cell expansion upon K3 loss. First, to test the recirculation and homing capacity of splenic cells, we crossed K3ΔB mice with mice expressing a fluorescent reporter for Cre activity. 31   Transitional (E i ) and follicular (E ii ) B cells were defined by expression levels of CD21 and CD24 as shown. Groups were compared using unpaired t-test. ns: P > 0.05; *: P < 0.05; **: P < 0.01; ***: P < 0.001 F I G U R E 3 K3 deficient B cells show altered homing and homing receptor expression, and MZ B cell retention in the spleen is dependent on K3. Splenocytes were isolated from K3 fl/fl Cre reporter mice ± mb1 Cre and injected into the tail vein of wt C57BL/6 mice. After 3 h, spleen, blood, and BM of the recipients were analyzed by flow cytometry. B cells were identified by CD19 and T cells by CD5 expression, transferred cells were identified by red/green fluorescence. The normalized cell recovery rate was calculated as transplanted cells per million of total cells per million cells injected (n = 9 per group) (A). Receptor expression was analyzed by flow cytometry on whole splenocytes of K3ΔB mice and control littermates. B cells were defined by CD19 expression (B) and (C). Groups were compared using unpaired t-test. ns: P > 0.05; *P < 0.05; ***P < 0.001

Chemokine-and BCR-mediated integrin activation is highly dependent on K3
To confirm the presumable defect in chemokine-induced integrin activation in K3-deficient B cells, we used a real-time binding assay based on the small monovalent VLA-4 ligand LDV, 23 coupled to a fluorophore (FITC), as described. 22 The rate at which bound LDV-FITC is replaced by unlabeled competing LDV is a direct measure of the affinity state of the VLA-4 integrin and allows for calculation of dissociation constants and the reciprocal ligand residence time. 23,24 Stimulation of the CXCR4 receptor by CXCL12 ( Figure 4A), or the CXCR5 receptor by CXCL13 ( Figure 4B) doubled ligand residence time from about 12 s to 24 s, reflecting a switch from low to medium affinity VLA-4. 23 Both chemokines failed to increase VLA-4 affinity in K3-deficient B cells.
Integrin activation of B cells interacting with follicular dendritic cells might involve B cell receptor stimulation by Ag, which is presented by the follicular dendritic cells, along with integrin ligands and chemokines. In wt but not K3-deficient cells, B cell receptor stimulation by anti-IgM increased VLA-4 affinity comparable to chemokinemediated activation ( Figure 4C). To demonstrate that not only VLA-4 but also LFA-1 activation on B cells is defective upon K3 loss, we used a reductive assay under shear flow mimicking homing processes, as previously described. 17,22 Thereby, inside-out activation of VLA-4 or LFA-1 by CXCL12 takes place in vitro when cells are allowed to interact with adhesive surfaces coated with VCAM-1/ICAM-1 and CXCL12. 35 In this assay, in the absence of the stimulatory chemokine signal, only a few cells arrested (Figure 4D, E). Arrests were strongly increased in presence of the chemokine and clearly required the presence of K3, suggesting its contribution for VLA-4 and LFA-1 inside-out activation to high-affinity states.
Integrin-mediated adhesion and retention are most relevant to cell populations at tissue-circulation interfaces, as is the spleen MZ.
However, inhibition or alteration of normally long-lasting integrinmediated signals could also contribute to the shifted B cell populations in K3 knockout mice. To differentiate between rapidly induced and slowly developing defects in the absence of K3, we used a tamoxifeninducible K3 knockout mouse strain (Fermt3fl/fl RosaCreERT2 mice). 13 One week after knockout induction, a significant population of MZ B cells was found in the blood of tamoxifen-treated, but not controltreated animals. 2 weeks after treatment, this population was no longer detectable ( Figure 4F). In the spleen, the loss of K3 reduced the MZ B cell population to about 20% of the numbers found in control mice ( Figure 4G). In summary, the data demonstrate that K3 is indispensable for inside-out activation of VLA-4 and LFA-1 integrins on splenic B cells, which in turn are required for homing and retention of shear force-experiencing B cells, particularly MZ B cells. In contrast, follicular B cells were unaltered in this short-term inducible K3 knockout mouse model (Figure 4H), indicating a long-term cause for follicular B cell expansion.

F I G U R E 4 K3 is indispensable for VLA-4 inside-out activation.
To determine VLA-4 affinity state, real time kinetic measurements were performed on B cells isolated by negative selection as described in. 23 CXCL13 (B), or anti-IgM F(ab) 2 (C) were added to stimulate VLA-4 affinity upregulation (n = 5). Negatively selected B cells were perfused over a VCAM-1 (D) (n = 5) or ICAM-1 (E) (n = 3) coated surface with or without co-immobilized CXCL12 K3 fl/fl -Rosa26CreERT2 mice were treated with 5 mg Tamoxifen (Tam) on two consecutive days. Control mice were treated with oil (n = 5 per group). One (1w) and 2 weeks (2w) after treatment, MZ B cells were detected in the peripheral blood by flow cytometry (F). Two weeks (2w) after the treatment, the spleen was analyzed by flow cytometry for the content of MZ (G) and follicular (H) B cells. Groups were compared using unpaired t-test. ns: P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001.

K3 deficient follicular B cells display a signature of increased activation and are prone to proliferate
To get insights into the causes for the increase in follicular B cell numbers, we sorted this population from the spleens of K3ΔB and control mice (Supplemental Figure 4A) and performed a transcriptome array. A linear based-model differential analysis showed 263 genes to be significantly (P < 0.05) up-regulated in Kindlin-3 deficient compared to control B cells, while only 41 genes were significantly downregulated (Supplemental Figure 4B). Among the 50 most significantly regulated genes, shown in Figure 5A

K3∆B mice show increased germinal center B cell and plasmablast numbers
Next, we characterized the possibility of a K3 loss-induced GC induction. Indeed, the absolute number of GC B cells (defined as CD95 high CD38 low , Figure 7A) was significantly increased in K3∆B mice compared to wt littermates. We also observed an increase in plasmablasts (TACI + /CD138 + /B220 high /CD19 low , Figure 7B Figure 7C-E). We observed a significant increase in PNA + Ki67 + GC numbers in K3∆B mice compared to wt littermates, with an intermediate GC number in K3∆B mice (around 2 GCs/mm 2 , Figure 7C). Thereby, Ki-67 positive clusters overlapped with GL7 ( Figure 7D); IgG positive areas were located adjacent to the main proliferative areas ( Figure 7E).
Further, proliferating B cells partially expressed high IgM levels (Supplemental Fig. 6C). Our data suggest that absence of K3 triggers an expansion of follicular B cells and enhances their commitment to differentiate into GC B cells and plasmablasts. F I G U R E 6 K3ΔB mice show an increase in proliferation centers in splenic follicles, CD40 and CD69 expression of follicular B cells of K3ΔB and control mice was determined by flow cytometry (n = 5 wt; n = 7 K3ΔB for CD40 and n = 3 in both groups for CD69) (A). Cryosections of spleens from K3ΔB and control mice were stained with Abs against Laminin (blue) to mark white pulp boundaries, B220 (red) to define B cell zones and Ki67 (green) as a marker for proliferating cells. Scale bars are 50 μm (B). Quantification was done using ImageJ by manually defining B-and T-cell zones based on laminin and B220 staining. The total and Ki-67 positive areas of B-or T-cell zones were measured and the Ki-67 positive fractions were calculated (one representative of 3 analyzed pairs of littermates is shown, 7 images were counted per spleen) (C). Groups were compared using unpaired t-test. ns: P > 0.05; *P < 0.05; ***P < 0.001

K3 deletion specifically increases the response of follicular B cells to CXCL13 and IgD activation
To identify the causes for the observed proliferation and lineage skewing of K3 deficient follicular B cells, we assessed the expression of the BCR IgD and IgM isotypes and found significantly elevated IgD but unaltered IgM on these cells, compared to K3 proficient cells (Figure 8A). We analyzed calcium (Ca2+) mobilization upon BCR as well as CXCR4 and CXCR5 (compare Figure 3B) stimulation. An example of the analysis is depicted in Supplemental Fig. 7A. While we found comparable calcium responses to anti-IgM, anti-kappa light chain and IgD activation in K3 deficient and wt follicular B cells (Supplemental Fig.   7B), the Ca2+ response to CXCL13, but not CXCL12 was increased in K3 deficient B cells ( Figure 8B). These data proposed an increase in CXCR5 responsiveness, in line with the increased CXCR5 but retained CXCR4 expression ( Figure 3B). To further corroborate this finding, we next analyzed activation of the PI3K pathway in response to CXCL13 and anti-IgD stimulation, using a phosphoflow approach for the PI3K target Akt. In accordance with the Ca2+ data, Akt phosphorylation in response to CXCL13, but not α-IgD was increased in K3 deficient follicular B cells ( Figure 8C). Along this line, CXCL13-induced chemotaxis was retained in K3 deficient follicular B cells despite their lower general and CXCL12-induced motility ( Figure 8D). We confirmed this observation on a collagen substrate with or without co-immobilized CXCL12 or CXCL13. In line with this finding, polarization of the actin cytoskeleton during migration on a collagen substrate, as analyzed by phalloidin staining, was significantly increased in the presence of CXCL13 but not CXCL12 in K3 deficient B cells ( Figure 8E, Supplemental Fig. 7C). In summary, we observed a clear preference of K3 deficient B cells for CXCR5 over CXCR4 signals and signs of altered motility, consistent with GC formation. 36,39 F I G U R E 7 K3ΔB mice show an increase in GC B cells and plasmablasts, GC B cells were flow cytometrically defined as CD95 high CD38 low (pre-gated on CD19 positive) in the spleens of K3ΔB and control mice as shown (n = 7) (A). Plasmablasts numbers were determined by flow cytometry as TACI + /CD138 + /B220 high /CD19 low (n = 7) (B). Cryosections of spleens from K3ΔB and control mice were stained for PNA/ GL7 (red) to mark germinal centers (C, D), or IgG (red) to mark class-switched B cells (E), together with Ki67 (green) defining proliferating cells. Scale bars are 50 μm. PNA + germinal centers (GCs) were quantification in K3ΔB and control mice (C). Groups were compared using unpaired t-test. ns: P > 0.05; *P < 0.05; ***P < 0.001

DISCUSSION
In this study, we analyzed B-cell intrinsic functions of the integrin adaptor K3 and one of the major B cell integrins, VLA-4 (CD49d), on basis of highly specific conditional knockout mouse models. We found that MZ B cell retention relies entirely on functional K3 and partially on VLA-4.
We also provide evidence that K3-mediated signaling serves as a moni-  One possible explanation is that the total loss of integrin-mediated signals and adhesion sites in the K3 KO influences the cytoskeleton (see Figure 8E), thereby altering membrane compartmentalization and BCR-co-receptor cooperation. 49 Lack of K3 also results in unoccupied focal adhesion molecules, which could be sequestered to and function in the nucleus. 50,51 In addition, K3 might also fulfill adaptor functions independent of integrin cytoplasmic tails. 52 The K3 interactome and K3-controlled signaling pathways in follicular B cells remain, however, to be elucidated in future projects.