DOCK2 and phosphoinositide-3 kinase δ mediate two complementary signaling pathways for CXCR5-dependent B cell migration

Naive B cells use the chemokine receptor CXCR5 to enter B cell follicles, where they scan CXCL13-expressing ICAM-1+ VCAM-1+ follicular dendritic cells (FDCs) for the presence of antigen. CXCL13-CXCR5-mediated motility is mainly driven by the Rac guanine exchange factor DOCK2, which contains a binding domain for phosphoinositide-3,4,5-triphosphate (PIP3) and other phospholipids. While p110δ, the catalytic subunit of the class IA phosphoinositide-3-kinase (PI3K) δ, contributes to CXCR5-mediated B cell migration, the precise interdependency of DOCK2, p110δ, or other PI3K family members during this process remains incompletely understood. Here, we combined in vitro chemotaxis assays and in vivo imaging to examine the contribution of these two factors during murine naïve B cell migration to CXCL13. Our data confirm that p110δ is the main catalytic subunit mediating PI3K-dependent migration downstream CXCR5, whereas it does not contribute to chemotaxis triggered by CXCR4 or CCR7, two other chemokine receptors expressed on naïve B cells. The contribution of p110δ activity to CXCR5-driven migration was complementary to that of DOCK2, and pharmacological or genetic interference with both pathways completely abrogated B cell chemotaxis to CXCL13. Intravital microscopy of control and gene-deficient B cells migrating on FDCs confirmed that lack of DOCK2 caused a profound migration defect, whereas p110δ contributed to cell speed and directionality. B cells lacking active p110δ also displayed defective adhesion to ICAM-1; yet, their migration impairment was maintained on ICAM-1-deficient FDCs. In sum, our data uncover two complementary signaling pathways mediated by DOCK2 and p110δ, which enable CXCR5-driven naïve B cell examination of FDCs.


Introduction
Naïve follicular B cells are highly motile cells, which scan ICAM-1 + VCAM-1 + follicular dendritic cells (FDCs) for the presence of microbial antigen and the initiation of humoral responses. The chemokine receptor CXCR5 is critical for naïve B cell access to follicles, where FDCs, together with other stromal cells such as marginal reticular cells, produce its only ligand CXCL13 (1,2). Furthermore, CXCR5 promotes together with the ICAM-1 receptor LFA-1 dynamic B cell surveillance of FDCs (3,4). The lymphocyte-expressed guanine exchange factor (GEF) DOCK2 is a key signaling molecule for Rac activation and F-actin polymerization downstream of chemokine receptors in lymphocytes. In the absence of DOCK2, in vitro T and B cell migration towards homeostatic chemokines is strongly compromised, although residual migration persists (5). Accordingly, direct observation of peripheral lymph nodes (PLN) using intravital twophoton microscopy (2PM) uncovered that follicular accumulation and interstitial motility are substantially reduced but not completely abolished in DOCK2 -/deficient B cells (6).
DOCK family proteins contain two DOCK homology regions (DHR), of which DHR1 is involved in phospholipid binding for membrane localization and DHR2 mediates the GEF activity (7)(8)(9). The DHR1 domain of DOCK2 binds the phosohoinositide-3-kinase (PI3K) product phosphoinositide-3,4,5-triphosphate (PIP3) as well as phosphatidic acid (PA). In B cells, the relation between DOCK2 and PI3K activity remains unclear to date. Whereas DOCK2 activity is not required for PI3K activation (5) and PI3K inhibition does not affect DOCK2mediated migration in T cells (10), neutrophil-expressed DOCK2 regulates migration through PIP3-dependent membrane translocation and Rac activation (11). Along the same line, the class IA p110d catalytic subunit is involved in B cell chemotaxis towards CXCL13 not but CCL19, CCL21 and CXCL12 (12), and regulatory subunits of class IA are required for basal B cell motility in vivo (13). A potential participation of class I catalytic subunits besides p110d during CXCR5-mediated B cell chemotaxis has not been examined yet, despite evidence for activation of additional class I PI3K family members downstream of G-protein coupled receptors (14).
Here, we examined the migratory behavior of B cells carrying mutations in DOCK2 and the catalytic site of p110d (p110d D910A/D910A ), in combination with PI3K-specific pharmacological inhibitors, to dissect their contribution for CXCL13-elicited motility. Among class I PI3K catalytic subunits, we confirm a key contribution of p110d to CXCR5but not CXCR4 and CCR7-dependent migration. DOCK2 and p110d activity comprised two complementary pathways for CXCR5-triggered B cell migration, and inhibition of both factors completely abolished chemotaxis. We corroborated our in vitro findings using intravital imaging of interstitial B cell scanning of FDCs. Finally, we found that while LFA-1 activity is reduced in the absence of catalytically active p110d, the interstitial migration defect of p110d D910A/D910A B cells is maintained on ICAM-1-deficient FDCs. In sum, our study sheds light on intracellular signaling pathways governing CXCR5-driven follicular B cell motility, a prerequisite for the unfolding of humoral immune responses.

Results
p110d is the dominant class I PI3K mediating B cell chemotaxis to CXCL13 The class I PI3K family member p110d contributes to directed B cell migration towards CXCL13 (12). Using Transwell assays, we confirmed a role for the catalytic activity of p110d for in vitro chemotaxis of primary murine B cells towards CXCL13, which was particularly evident at lower chemokine concentrations (reduction of 48% at 100 nM and 33% at 250 nM CXCL13 for p110d D910A/D910A B cells as compared to WT B cells, respectively; Figure 1A). To address whether additional catalytic subunits might contribute to WT and p110d D910A/D910A B cell migration, we performed chemotaxis assays in presence of the p110b inhibitor TGX221, the p110g inhibitor AS604850, the p110a/b/d/g inhibitor PI-103 and, as control, the p110d inhibitor IC-87114. These data uncovered a decrease of WT B cell chemotaxis towards 100 nM CXCL13 only with PI-103 and IC-87114 (52% and 54% inhibition, respectively), while none of the inhibitors had a significant effect on p110d D910A/D910A B cell migration ( Figure 1B). These findings suggest that other class IA and IB subunits do not substantially contribute to primary B cell migration towards CXCL13. The promigratory signaling function of p110d was restricted to CXCR5, since B cell migration to CCR7 and CXCR4 ligands remained unchanged by genetic or pharmacological inhibition of its activity ( Figures 1C, D), as reported (12). Similarly, CCR7-mediated primary T cell chemotaxis was not reduced by genetic or pharmacological inhibition of the catalytic activity of p110d ( Figure 1E).

DOCK2 and p110d comprise two complementary pathways for CXCR5mediated B cell migration
We next examined the potential relationship of DOCK2 and p110d during in vitro B cell chemotaxis towards CXCL13, given that DOCK2 contains a PIP3 binding domain. In a first set of experiments, we treated WT B cells separately or in combination with IC-87114 and CPYPP, which blocks the GEF activity of DOCK2 by binding to its catalytic DHR2 domain (15). These data showed that DOCK2 and p110d comprised two complementary pathways for CXCR5-mediated chemotaxis, since only simultaneous treatment with both inhibitors completely abolished migration (Figure 2A). A blocking effect of CPYPP and IC-87114 was also observed for CXCL13-induced migration of p110d D910A/D910A and DOCK2 -/-B cells, respectively ( Figure 2A).
Since inhibitors are often not entirely specific, we generated p110d D910A/D910A x DOCK2 -/mice to corroborate our findings in a genetic model. Double-deficient mice were born at sub-mendelian ratios and showed growth retardation (not shown). Owing to the difficult breeding, we could isolate cells from these mice for only limited amounts of chemotaxis assays. In these experiments, residual migration of DOCK2 -/-B cells to 250 nM CXCL13 was abolished when p110d activity was additionally compromised ( Figure 2B). Taken together, these data suggest that DOCK2 and p110d act in largely non-overlapping pathways downstream of CXCR5 signaling. p110d activity contributes to B cells speed and directionality during follicular migration CXCR5 is required for B cell entry to B cell follicles (1), where it contributes to fast motility (4). This motility is in large part driven by DOCK2-mediated Rac activation, since DOCK2 -/-B cells show substantially reduced interstitial movement (6). Using 2PM of popliteal PLN containing adoptively transferred B cells (4), we confirmed a substantial drop in mean speeds in DOCK2-deficient B cells (from 7.9 ± 4.7 to 4.0 ± 2.9 µm/min for WT and DOCK2 -/-B cells, respectively; Figure 3A). This decline in speed was accompanied by broader turning angles and a low motility coefficient (MC), a proxy for a cell's ability to scan an area (20.7 and 3.6 µm 2 /min for WT and DOCK2 -/-B cells, respectively; Figures 3B, C), in line with our previous observations (6).
We then examined whether p110d contributed to B cell scanning of B cell follicles in vivo. In contrast to DOCK2 -/-B cells, WT and p110d D910A/D910A B cells accumulated efficiently in B cell follicles ( Figure 3D; Supplemental Movie 1). However, p110d D910A/D910A B cells moved with decreased speeds and less directionality compared to WT B cells, as measured by meandering index and turning angle distribution (Figures 3E-G). As a result, p110d D910A/D910A B cells had an approximately 50% reduction of their MC compared to WT B cells ( Figure 3H). In contrast, interstitial p110d D910A/D910A T cell migration speeds were similar to those of WT T cells ( Figure 3I). These data support a contribution of p110d activity to B cell motility along the FDC network inside B cell follicles. In addition to CXCR5, LFA-1 contributes to B cell motility on ICAM-1 + VCAM-1 + FDCs, whereas a4 integrins play no detectable role (16). In line with this, b2 integrin-dependent in vitro leukocyte migration requires Syk-mediated p110d translocation to the leading edge (17). Given the comparable impact of defective LFA-1 and p110d activity on dynamic B cell motility parameters, we examined whether p110d activity mediated its promigratory effect via LFA-1 activation. In support of this, an analysis of CXCL13-triggered in vitro adhesion to FDC-expressed adhesion molecules uncovered a reduction in p110d D910A/D910A B cell binding to ICAM-1 but not VCAM-1 ( Figures 4A, B). Again, this adhesion defect was restricted to B cells, since p110d D910A/D910A T cell adhesion to ICAM-1 was not impaired ( Figure 4C). We transferred WT and p110d D910A/D910A B cells into ICAM-1 -/recipients, the main stromal LFA-1 ligand used by B cells in lymphoid tissue (16). We hypothesized that WT and p110d D910A/D910A B cells would show similar migration speeds if p110d exerted its promigratory effect via LFA-1. However, we still observed reduced migration speeds, meandering index and increased turning angles in p110d D910A/ D910A B cells compared to WT B cells (Figures 4D-F). As a consequence, their MC remained lower than the one of WT B cells ( Figure 4G). These data suggest that the migration defect of p110d D910A/D910A B cells is largely independent of LFA-1mediated adhesion to the FDC network. In sum, our data uncover a role for p110d activity during B cell migration in lymphoid tissue, which is less pronounced than the effect caused by absence of DOCK2.

Discussion
CXCR5-driven B cell chemotaxis to CXCL13 is critical for the development of humoral immune responses, as it enables efficient surveillance of FDCs and the proper formation of germinal centers (1,18,19). Here, we examined the intracellular wiring of CXCR5 that transmits biochemical input into a promigratory response. Our in vitro chemotaxis assays confirmed a critical role for the Rac GEF DOCK2 in mediating robust B cell chemotaxis to CXCL13, while p110d participates in a complementary signaling module. These observations were recapitulated in vivo, suggesting the existence of two signaling pathways underlying CXCL13mediated motility. The requirement of a PI3Kd-dependent signaling module appears restricted to CXCR5, since migration to CXCR4 and CCR7 ligands was not impaired.
Intravital imaging has uncovered that B cell adhesion in PLN high endothelial venules (HEV) is more strongly attenuated by the absence of DOCK2 as compared to adhesion in Peyer's patch (PP) HEV, although in both cases there is a significant reduction in B cell attachment (10). In contrast, lack of PI3Kd activity mainly affects B cell homing to mesenteric lymph nodes (MLN) and PP, while these cells show normal homing to PLN (12). This may be due to the fact that CXCR5 plays a more prominent role for B cell homing to MLN and PP as compared to PLN, where CCR7 and CXCR4 play compensating roles (20,21). Thus, the cooperative action of DOCK2 and PI3Kd activity appears to extend to CXCR5-driven B cell entry into secondary lymphoid organs.
The parallel occurrence of a major, DOCK2-dependent pathway and a minor PI3K-dependent pathway in B cells mirrors observations made in naïve T cells. In T cells, the class IB p110g isoform mediates DOCK2-independent migration via a pathway involving the PIP3-binding pleckstrin homology (PH)domain containing Tec family kinase Itk (6,10,12,22). Accordingly, DOCK2 -/x p110g -/-T cells show no residual migration to CCL21 (10). In combination with the lack of p110d involvement during naïve T cell migration in vitro and in vivo, our data support a model where p110g and p110d catalytic subunits contribute to T and B cell motility in a subset-specific manner. Of note, during CD4 + T cell differentiation to follicular helper T cells (T FH ), p110d signals downstream ICOSL induce T FH precursor migration into the B cell follicles (23), suggesting context-specific roles for PI3K family members during lymphocyte positioning within lymphoid organs.
It remains incompletely understood how p110d signaling contributes mechanistically to B cell migration downstream CXCR5, although Rac activation is likely to be required (24). A conceivable scenario is that PI3Kd activates the B cell homologue of Itk, the PH-domain-containing Btk (25). In c h r o n i c l y m p h o c y t i c l e u k e m i a ( C L L ) c e l l l i n e s , pharmacological blockade of either p110d or Btk reduces migration to CXCL13 (26,27). Btk is linked to Vav phosphorylation, leading to downstream WASP activation and F-actin remodeling (28).
Unexpectedly, we found that the defect of p110d D910A/D910A B cell was maintained in lymphoid microenvironment lacking stromal ICAM-1, despite the known involvement of Syk-p110d signaling during b2-integrin-mediated migration on 2D surfaces (17). A plausible explanation is that akin to naïve T cell migration within lymph node parenchyma, the main role for LFA-1 might be for generation of traction forces without inducing substantial adhesion (29). In the 3D confined environment of lymphoid tissue, substrate adhesion is externally enforced by juxtaposed cells, thus compensating for reduced LFA-1 activity.
The robust DOCK2-driven migration of p110d D910A/D910A B cells to CXCL13 raises the question whether PI3K-mediated signaling has additional roles beyond promoting cell motility.
Another open point is whether PI3Kd signaling might be involved in signal transduction downstream GPR183, although this receptor appears to have an inhibitory effect on CXCR5mediated migration (4). In T cells, Itk contributes to homeostasis, suggesting a role for PI3K-dependent signaling in maintaining peripheral T cell numbers (22). Similarly, it is conceivable that CXCR5-mediated PI3K activation contributes to B cell homeostasis, in line with the well-documented role of this pathway for survival (30). In addition, the selective integration of p110d signaling downstream CXCR5, but not other receptors for homeostatic chemokines, might facilitate B cell activation by feeding into the BCR-triggered PI3K-Btk signaling axis. A similar costimulatory signaling pathway was reported for CCL21 during T cell activation (31).
In sum, our data uncover dual signaling pathways mediating physiological CXCR5-triggered B cell motility that underpins rapid detection of cognate antigens presented on FDCs. Given that small tyrosine kinase inhibitors targeting p110d and Btk are widely used in the treatment of leukemias (32)(33)(34), it is of clinical interest to understand potential implications on the patients' immune system. Chemotaxis CCL21 and CXCL12 were from Peprotech, and CXCL13 was purchased from R&D systems. Chemotaxis assays were carried out using Transwell chambers (5 µm pore size; CoStar) adding 100 µl cell suspension (5 x 10 6 cells/ml) in complete medium (RPMI/10% FCS/standard supplements) to the top chamber and indicated amounts of chemokine in the bottom chamber. After 2 h at 37°C, 7% CO 2 , the percentage of migrated cells was calculated by flow cytometry after comparing with a precalibrated bead standard (Sigma-Aldrich) and correcting for variations in input concentrations. The DOCK2 inhibitor CPYPP (Selleck) was used at 40 µM throughout the chemotaxis assay (15). The isoformspecific PI3K inhibitors TGX221 (0.1 µM final conc.; Tocris), PI-103 (1 µM; Tocris), AS604850 (1 µM; Selleck),and IC-87114 (0.5 µM; Selleck) were present throughout the chemotaxis assay.

Adhesion assay
Adhesion assays were performed as described (10). In brief, purified B or T cells were allowed to settle on 8-well-slides coated with 1.5 µg/ml murine ICAM-1 or 2.5 µg/ml VCAM-1 (R&D Systems). Chemokine was added at a final concentration of 1 µM for 3 min. Slides were rinsed with PBS to wash off unbound cells, fixed in glutaraldehyde, and the number of adherent cells was determined at the site of chemokine addition.

Twophoton intravital microscopy
Fluorescently labeled WT and genetically modified B cells were adoptively transferred into WT or ICAM-1 -/recipients 12-48 h before 2PM recording. In some experiments, PE-conjugated anti-CD35 mAb (0.5 µg in 10 µl PBS/mouse) was injected into the footpad 12 h before 2PM to label the FDC network of the draining popliteal PLN. Recipient mice were surgically prepared to expose the right popliteal PLN, which was kept at 36-38°C. Mice were then transferred to an Olympus BX50WI fluorescence microscope attached to a 2PM scanner (TrimScope system, LaVision Biotec, Bielefeld, Germany) equipped with an 20X objective (Olympus, NA 0.95). For four-dimensional analysis of cell migration, 8-16 zstacks (spacing 4 µm) of 200-300 µm x-y sections were acquired every 20 s for 20 to 30 min, with typically 3-4 distinct areas recorded per preparation. Image sequences were transformed into volume-rendered four-dimensional movies using Volocity (Perkin Elmer) or Imaris (Bitplane), which was also used for semiautomated tracking of cell motility in three dimensions. From x, y and z coordinates of cell centroids, parameters of cellular motility were calculated as described previously. In brief, the track speed is depicted as average speed, with each dot representing one track. Owing to the large number of tracks, they are shown as box and whisker plots with whiskers covering 1-99% of data points. For turning angles and motility coefficients, we used MatLab scripts kindly provided by Dr. Sarah Henrickson and Prof. Ulrich H. von Andrian (Harvard University, Boston, USA). In some experiments, purified WT and p110d D910A/D910A T cells were transferred into WT recipients and their migratory behavior was analyzed in the T cell area as above.

Statistical analysis
The student's t-test or ANOVA were used to determine statistical significance (Prism, GraphPad). Statistical significance was set at p < 0.05.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The animal study was reviewed and approved by Canton of Fribourg and the Canton of Bern.