Intravital imaging reveals MZ B cell–RBC contacts
MZ B cells are distinguished from follicular B cells by high expression of CD21 and IgM and low expression of CD23 and IgD4,5. To enable intravital two-photon microscopy of MZ B cells, we used transferred B cells from ubiquitin-green fluorescent protein (Ub-GFP)-expressing mice to Cd19-knockout (KO) mice that lack an endogenous MZ B cell population (Fig. 1a)8. After 8–12 weeks, the MZ was occupied by the transferred B cells, and the majority of the transferred GFP+ B cells had a CD21hiCD23lo MZ B cell phenotype (Fig. 1b). Three hours before analysis, the mice were transfused with PKH26 (red) dye-labeled RBCs such that about 2% of RBCs in the recipient mice were labeled (Fig. 1c). The mice were also given transfers of CellTrace Violet (CTV)-labeled follicular B cells 1 or 2 d before to help identify lymphoid follicles. To guide expectations for the intravital imaging analysis, which can detect fluorescent cells at depths of up to 200 µm, thin sections were taken within 200 µm of the capsule and examined for the distribution of the transferred B cells and RBCs and for total IgD+ follicular B cells. This analysis showed that occasional small follicular structures could be detected at this depth, and these had GFP+ B cells superficially associated that were IgDlo and thus were most likely MZ B cells (Fig. 1d). Clusters of large GFP-bright cells were also detected in the red pulp, distant from the B cell follicles, likely corresponding to plasma cells.
The behaviors of MZ B cells were observed with intravital two-photon microscopy in Cd19−/− mice reconstituted with Ub-GFP+ B cells. CTV-labeled naive B cells were used to identify follicles. PKH26-labeled RBCs were intravenously injected 3 h before imaging, and 70-kD Texas Red dextran was introduced 30 min before imaging. a, Schematic diagram of the intravital imaging protocol. b, Representative flow cytometry profiles of MZ B cells among GFP+ B cells 8–12 weeks after reconstitution. c, A representative flow cytometry profile of PKH26-labeled RBCs in total splenic RBCs. d, Representative distribution patterns of PKH26-labeled RBCs (red), GFP+ B cells (green), CTV-labeled B cells (blue) and IgD+ B cells (white) in the spleen; scale bar, 100 μm. The section shown is representative of multiple cross-sections from at least three mice. e, Examples of contacts between RBCs and MZ B cells (blue arrowhead) in spleens. The region (120 μm × 120 μm) indicated with white box (left; scale bar, 40 μm) was zoomed in for the time series (right; scale bar, 20 μm). One RBC was tracked interacting with several different MZ B cells. The interaction starts at 9:32 and ends at 18:45. Time is indicated in minutes:seconds. See corresponding Supplementary Videos 2–4. Data are representative of five independent experiments.
Scanning different regions of the spleen using intravital two-photon microscopy allowed for the identification of occasional areas with GFP+ MZ B cells near clusters of CTV+ follicular B cells (Fig. 1e and Supplementary Video 1). Labeled RBCs were observed passing near MZ B cells, and transient contacts between MZ B cells and RBCs could be detected (Fig. 1e and Supplementary Videos 2–4). Some RBCs encountered different MZ B cells during their movement (Fig. 1e and Supplementary Videos 2 and 3), and some MZ B cells were observed interacting with several RBCs (Supplementary Videos 2 and 4). In some regions of the MZ, the RBCs appeared to be moving in an irregular manner, perhaps within a turbulent and semiconfined space, whereas in other regions, the RBCs passed through the MZ with uniform flow (Supplementary Videos 2–4). Although we could readily detect RBC–MZ B cell contacts in each 20- to 30-min video, it is important to appreciate that only ~2% of the RBCs were labeled, and thus the contact frequency is expected to be 50-fold greater than imaged. These data establish that MZ B cells regularly contact RBCs in regions of flow within the MZ.
CD97 is required for MZ B cell homeostasis
Because frequent interactions were observed between RBCs and MZ B cells, we asked if there were ligand–receptor pairs expressed on the two cell types that might be required for MZ B cell function. Notable among the few surface molecules expressed by RBCs that engage surface receptors is CD55, the decay-accelerating factor of complement that is also a ligand for CD97 (encoded by Adgre5)12,15. Flow cytometric analysis showed that MZ B cells were marked by high CD97 expression (Fig. 2a,b). We therefore investigated the function of CD97 on MZ B cells. Analysis of CD97-deficient (Adgre5−/−) mice (Extended Data Fig. 1a) revealed a significant reduction in MZ B cells (Fig. 2c,d). The frequencies of immature (T1 and T2) and mature follicular B cells were unaffected (Extended Data Fig. 1b). The frequency of B1 cells, another type of early-responder B cells4, was also unaffected in the spleen (Extended Data Fig. 1c). Immunofluorescence staining of tissue sections for IgM and IgD showed a reduction in the thickness of the IgMhi MZ in mice lacking CD97 (Fig. 2e and Extended Data Fig. 1d). Using mixed bone marrow (BM) chimeras, the CD97 requirement was established to be cell intrinsic (Fig. 2f and Extended Data Fig. 1e). Staining of tissue sections from IgHa:IgHb mixed BM chimeras showed a reduction in IgMhi MZ thickness selectively in the CD97-deficient IgMb MZ compartment (Extended Data Fig. 1f). By intravascular antibody labeling of blood-exposed spleen cells, about 55% of MZ B cells in wild-type (WT) mice were labeled; the remaining fraction was within follicles at the time of antibody injection and was protected from labeling7 (Extended Data Fig. 1g). CD97 deficiency did not change the fraction of MZ B cells labeled, consistent with the notion that the rapid shuttling of MZ B cells between the MZ and follicle keeps the proportion of cells in each compartment intact even with the overall drop in MZ B cells (Fig. 2g). MZ B cells in littermate-matched control and Cd97-KO mice showed the same extent of turnover as determined by staining for the cell cycle antigen Ki-67 (Extended Data Fig. 1h) and the apoptosis marker Annexin V (Extended Data Fig. 1i).
a, Structural components of CD97. Triangles indicate EGF domains. The ten amino acids (419–428) following the GPS are indicated. b, Representative histogram plot of CD97 on MZ and follicular (FO) B cells; ISO, isotype control. c,d, Flow cytometry profiles (c) and frequencies of MZ B cells in B cells (d; left) and total splenocytes (d; right) in Adgre5−/− (n = 16) and control (n = 14) mice. e, Distribution of IgMhi MZ (red) and IgDhi follicular (blue) B cells in indicated spleens; scale bar, 200 μm. Sections are representative of multiple cross-sections from at least three mice of each type. f, Flow cytometry profiles (left) and frequencies (right) of MZ B cells of the indicated genotypes in WT:Adgre5−/− (n = 9) and control (n = 9) chimeras. g, Frequencies of in vivo anti-CD45-phycoerythrin (PE)-labeled MZ B cells of the indicated genotypes in WT:Adgre5−/− (n = 8) and control (n = 8) chimeras. Lines connect data from the same animals. h, Flow cytometry profiles (left) and frequencies (right) of MZ B cells in blood in WT mice 3 h after treatment with anti-CD97 (n = 8) or saline (n = 8). i, Frequencies of MZ B cells in the spleen in WT mice after 4 d of treatment with anti-CD97 (n = 13) or saline (n = 12). j, Frequencies of ‘leaving cells’ (GFP+ B cells that enter large vessels) in mice reconstituted as in Fig. 1a with Adgre5−/− (n = 10) and Adgre5+/+ (n = 6) MZ B cells. See corresponding Supplementary Videos 5 and 6. k, BM chimeras were reconstituted with 10% non-transduced CD45.1 WT and 90% CD45.2 Adgre5+/+ or Adgre5−/− BM transduced with retroviral constructs encoding Adgre5 WT (n = 15) or its mutants (n = 8 in L424A, n = 5 in M425K, n = 10 in PBM and n = 8 in T419G) or empty vector (CON; n = 18 in Adgre5+/+ and n = 20 in Adgre5−/−). The graph shows the frequencies of MZ B cells in Thy1.1+ or Thy1.1− B cells. Data are pooled from four (d and i), two (f–h), five (j) or six (k) independent experiments. Each symbol represents one mouse, and lines denote means. Statistical significance was tested by two-tailed t-test (d, h, i and j) or two-way analysis of variance (ANOVA) followed by a Sidak’s multiple-comparisons test (f, g and k).
Source data
We speculated that the reduced frequency of MZ B cells in CD97-deficient mice might be a consequence of loss into blood circulation. However, the frequency of MZ B cells in Cd97-KO blood was too low to detect reliably. We considered the possibility that there might be loss of very small numbers of cells at any given moment that could amount to a large loss over periods of days. We therefore asked whether conditional blockade of CD97 ligand binding could lead to a more synchronized release of cells. Indeed, when mice were treated with a CD97-blocking antibody, a small population of MZ B cells could be detected in the blood 3 h later (Fig. 2h). At this time point, the treatment was not sufficient to measurably deplete MZ B cells from the spleen (Extended Data Fig. 2a). However, more prolonged treatment with CD97-blocking antibody led to a reduction in MZ B cells, consistent with gradual loss from the spleen (Fig. 2i). A slight increase in follicular B cell frequency was observed (Extended Data Fig. 2b), likely due to the reduction in MZ B cells.
Intravital two-photon microscopy of the splenic red pulp in mice that harbored GFP+ MZ B cells and PKH26+ RBCs revealed large numbers of labeled RBCs passing through the red pulp, some at high speed within vessels or sinuses and others more slowly that were likely traveling through the parenchyma (Supplementary Videos 5 and 6). We noted occasional GFP+ cells, possibly MZ B cells in the red pulp, in addition to the large GFPhi cells that are likely plasma cells (Supplementary Videos 5 and 6). While imaging spleens of mice that had been reconstituted with either WT or Cd97-KO Ub-GFP+ B cells, we noted occasional GFP+ Cd97-KO B cells being released into circulation (Fig. 2j and Supplementary Video 6). Although the trend for release of more Cd97-KO cells than WT cells did not reach statistical significance, this observation is consistent with the increased flow cytometric detection of MZ B cells in the blood after CD97 blockade. Integrin expression was intact in CD97-deficient mice and in WT mice after antibody blockade (Extended Data Fig. 2c–f).
To assess the fate of MZ B cells that have been released into blood circulation, we tracked GFP+ MZ B cells in the first 18 h after intravenous transfer of splenocytes. Compared to the total number of MZ B cells that were injected, most cells were lost from the blood within 10 min, and only ~0.25% could be recovered 1 h later (Extended Data Fig. 2g,h). Somewhat greater numbers were present in the recipient spleen at this time, and a gradual increase in number occurred, reaching a plateau by 6 h that corresponded to a recovery of about 5% of the transferred MZ B cells. Transferred GFP+ CD97-deficient MZ B cells showed a similar loss from circulation and similar initial appearance in the spleen, but these cells were not maintained and had largely decayed by 18 h (Extended Data Fig. 2i,j). Thus, a reduced ability to reseed the spleen may contribute to the overall splenic MZ B cell deficiency in CD97-deficient mice.
CD97-tethered ligand requirement in MZ B cells
We next characterized what features of CD97 were required for its in vivo function in MZ B cells. CD97 is composed of noncovalently attached NTF and CTF (Fig. 2a). Mutation of T419, the residue immediately following the GPS site, to glycine (T419G) prevents CD97 autoproteolysis but permits normal surface expression21. Using a Cd97-KO BM retroviral transduction and reconstitution approach, WT CD97 was able to restore MZ B cell accumulation, whereas the non-cleaved T419G mutant did not (Fig. 2k) despite comparable expression (Extended Data Fig. 2k). In these experiments, the transduced CD45.2+ cells were detected using a Thy1.1 reporter; untransduced CD45.2+ cells in the same BM chimeras were identified as Thy1.1− cells. The first ~10 amino acids following the GPS (Fig. 2a) are thought to function as a tethered ligand in CD97 and many other adhesion GPCRs12,13. In a cell line study, mutation of tethered ligand residue L424 to alanine (L424A) or M425 to lysine (M425K) reduced CD97 signaling in vitro22. When Cd97-KO mice were reconstituted with BM transduced with CD97 L424A or M425K, the MZ B cell compartment was not rescued (Fig. 2k) despite surface expression being comparable to that observed in WT mice (Extended Data Fig. 2k). CD97 has a C-terminal motif that can interact with PDZ domain proteins21. CD97 with a mutation in this PDZ-binding motif (PBM) had intact expression (Extended Data Fig. 2k) and was functional in restoring MZ B cells, although perhaps less efficiently than WT CD97 (Fig. 2k). Taken together, these data are consistent with a model where extraction of the CD97 NTF leads to activation of the receptor by a tethered ligand, and this signal promotes MZ B cell retention and homeostasis in the spleen.
CD97 in MZ B cells signals via Gα13 and ARHGEF1
Because prior studies in cell lines and type 2 conventional dendritic cells indicated that CD97 can signal via Gα13-containing heterotrimeric G proteins17,21,23,24, and other work showed a role for Gα12/Gα13 in MZ B cells25, we asked if Gα13 may be required for CD97 function in MZ B cells. Like CD97-deficient mice, animals lacking Gα13 selectively in B cells (Gna13fl/fl Mb1-Cre+ conditional knockout (cKO), labeled as Gna13cKO) showed a twofold reduction in MZ B cell frequency (Fig. 3a,b). There was no effect of Gα13 deficiency on immature or mature B cell frequencies in the spleen (Extended Data Fig. 3a). Splenic B1 cells were also unaffected (Extended Data Fig. 3b). Microscopy showed that Gna13cKO mice had reduced MZ thickness (Fig. 3c and Extended Data Fig. 3c). Using mixed BM chimeras, the Gα13 requirement for MZ B cell homeostasis was confirmed to be cell intrinsic (Fig. 3d). Importantly, chronic anti-CD97 treatment of Gα13-deficient mice did not cause any further reduction in MZ B cell frequencies, whereas it did reduce the MZ B cell compartment in WT mice, as expected (Fig. 3e). Gα12 deficiency did not affect MZ B cell frequencies (Extended Data Fig. 3d,e). These data confirm that CD97 and Gα13 function in the same pathway. ARHGEF1 (also known as p115RhoGEF or Lsc) is a Rho-activating guanidine nucleotide exchange factor and the best-defined effector of Gα13 (ref. 26). A similar series of experiments performed with ARHGEF1-deficient mice showed that this Gα13 effector is needed for MZ B cell homeostasis (Fig. 3f–h and Extended Data Fig. 3f), in agreement with a prior study27. ARHGEF1 deficiency did not affect immature B cell or B1 cell frequencies in the spleen, and it led to a slight increase in follicular B cell frequencies (Extended Data Fig. 3g,h). Mixed BM chimeras established that ARHGEF1 acts in a cell-intrinsic manner in MZ B cells (Fig. 3i). These findings are in agreement with CD97 signaling via Gα13 and ARHGEF1 in MZ B cells.
a,b, Representative flow cytometry profiles (a) and frequencies of MZ B cells in total CD19+ B cells (b; left) and in total splenocytes (b; right) in Gna13cKO (n = 12) and control (n = 13) mice. c, Representative distribution patterns of IgMhi MZ B cells (red) relative to IgDhi follicular B cells (blue) in spleens of mice of the indicated genotypes; scale bar, 200 μm. d, Mixed (50:50) BM chimeras were made with CD45.1 WT and CD45.2 Gna13WT or Gna13cKO BM cells. Frequencies of MZ B cells among the indicated genotype CD19+ B cells in WT:Gna13cKO (n = 6) and control (n = 6) mixed BM chimeras. e, Frequencies of MZ B cells in total CD19+ B cells (left) and in total splenocytes (right) in Gna13cKO (n = 5) and control (n = 4) mice 4 d after treatment with anti-CD97 or saline. f,g, Representative flow cytometry profiles (f) and frequencies of MZ B cells in total CD19+ B cells (g; left) and in total splenocytes (g; right) in Arhgef1−/− (n = 7) and control (n = 7) mice. h, Representative distribution patterns of IgMhi MZ B cells (red) relative to IgDhi follicular B cells (blue) in spleens of mice of the indicated genotypes; scale bar, 200 μm. i, Mixed (50:50) BM chimeras were made with CD45.1 WT and CD45.2 Arhgef1+/− or Arhgef1−/− BM cells. Frequencies of MZ B cells among the indicated genotype CD19+ B cells in WT:Arhgef1−/− (n = 6) and control (n = 6) mixed BM chimeras. Data are pooled from four (b) or two (d, e, g and i) independent experiments. Sections are representative of multiple cross-sections from at least three mice of each type (c and h). Each symbol represents one mouse, and lines denote means. Statistical significance was tested by two-tailed t-test (b and g) or two-way ANOVA followed by a Sidak’s multiple-comparisons test (d and i) or one-way ANOVA followed by a Tukey’s multiple-comparisons test (e).
Source data
CD55 on RBCs acts as the CD97 ligand
CD55 is well expressed on RBCs and is also present on various other hematopoietic cells15,17. Interestingly, MZ B cells expressed low amounts of CD55 compared to follicular B cells (Fig. 4a), perhaps ensuring limited cis interaction between CD55 and CD97 and maximal availability to interact with CD55 on other cells. To test the importance of CD55 for CD97-dependent functions in MZ B cells, we analyzed Cd55−/− mice. MZ B cell frequencies were reduced in Cd55−/− mice to a similar extent as in Cd97-KO mice, and the MZ showed a similar decrease in thickness in imaging (Fig. 4b–d and Extended Data Fig. 4a). Immature, follicular and B1 cell frequencies in the spleen were unaltered by CD55 deficiency (Extended Data Fig. 4b,c). The fractions of MZ B cells in the cell cycle based on Ki-67 staining (Extended Data Fig. 4d) or that were undergoing cell death based on Annexin V staining (Extended Data Fig. 4e) were not changed by CD55 deficiency. Mixed BM chimeras showed that CD55 was not required intrinsically by MZ B cells (Fig. 4e). Importantly, when cells lacked both CD55 and CD97 (double KO (dKO)), the deficiency in MZ B cells was of the same magnitude as for single-KO cells (Fig. 4f), consistent with these genes acting in the same pathway and with CD55 serving as the only CD97 ligand involved in MZ B cell maintenance.
a, Representative histogram plot of surface CD55 on MZ B cells and follicular B cells. b,c, Representative flow cytometry profiles (b) and frequencies of MZ B cells in total CD19+ B cells (c; left) and in total splenocytes (c; right) in Cd55−/− (n = 18) and control (n = 18) mice. d, Representative distribution patterns of IgMhi MZ B cells (red) relative to IgDhi follicular B cells (blue) in spleens of mice of the indicated genotypes; scale bar, 200 μm. Sections are representative of multiple cross-sections from at least three mice of each type. e, Mixed (50:50) BM chimeras were made with CD45.1 WT and CD45.2 Cd55+/+ or Cd55−/− BM cells. Frequencies of MZ B cells of the indicated genotype CD19+ B cells in WT:Cd55−/− (n = 6) and control (n = 6) mixed BM chimeras. f, Mixed (50:50) BM chimeras were made with CD45.1 WT and CD45.2 WT (n = 6), Adgre5−/−Cd55−/− (dKO; n = 6) or Cd55−/− and dKO (n = 5) BM cells. Frequencies of MZ B cells among the indicated genotype CD19+ B cells in mixed BM chimeras. g, RBC transfusions were performed from Cd55−/− or Cd55+/+ mice to the indicated recipient mice once per week, and analysis was performed after 4 weeks. Frequencies of MZ B cells in total B cells (left) and in total splenocytes (right) in mice with purified RBCs transfused as indicated (n = 10 or 11 recipients in each group) are shown. h,i, Mixed chimeras were made with 85% Mpl−/− and 15% Cd55+/+ or Cd55−/− BM cells. A representative histogram shows surface CD55 on platelets (left) and RBCs (right) in chimeras (h). Frequencies of MZ B cells in total splenocytes in Mpl−/−:Cd55−/− (n = 7) and control (n = 8) chimeras are shown (i). Data are pooled from five (c), two (e, f and i) or three (g) independent experiments. Each symbol represents one mouse, and lines denote means. Statistical significance was tested by two-tailed t-test (c and i) or two-way ANOVA followed by Sidak’s multiple-comparisons test (e and f) or one-way ANOVA followed by Tukey’s multiple-comparisons test (g).
Source data
In addition to being expressed by RBCs and some other hematopoietic cells, CD55 is expressed by various non-hematopoietic cells17,28. As a broad approach to determine the necessary CD55-expressing cell types for MZ B cell homeostasis, we generated reciprocal BM chimeras. In mice lacking CD55 in all radiation-sensitive BM-derived cells, there was a deficiency in MZ B cells (Extended Data Fig. 4f), whereas in mice lacking CD55 in radiation-resistant cells, including all stromal cells, the MZ B cell compartment remained intact (Extended Data Fig. 4g). Although follicular B cells abundantly express CD55, analysis of 85:15 Rag1−/−:Cd55−/− BM chimeras that lack CD55 on all B cells and most T cells revealed an intact MZ B cell compartment (Extended Data Fig. 4h–j). RBCs make up 99.9% of the cells in blood29. RBCs also express CD55, although at a lower surface level than B cells (Extended Data Fig. 4k). To test whether RBCs were the relevant source of CD55, WT or Cd55-KO RBCs were transferred into Cd55-KO mice. Because the MZ B cell compartment is slow to turn over4, the transfusion was performed weekly for 4 weeks. At the end of the transfusion, approximately 65% of RBCs were of donor origin (Extended Data Fig. 4l). Reconstitution of Cd55−/− mice with WT, but not Cd55-KO, RBCs rescued the size of the MZ B cell compartment (Fig. 4g). Platelets also express CD55 (ref. 30). To test for a possible contribution of platelet CD55 to MZ B cell homeostasis, 85:15 mixed BM chimeras between Mpl−/− BM (unable to generate platelets) and Cd55−/− BM were made such that all the platelets in these mice were CD55 deficient (Fig. 4h). CD55 deficiency on platelets did not lead to a reduction in the MZ B cell compartment (Fig. 4i). Thus, RBCs are the key CD55+ cell type needed for MZ B cell homeostasis.
CD55-dependent extraction of the CD97 NTF under shear stress
Flow cytometric analysis of MZ B cells for CD97 showed that surface levels were elevated in mice lacking CD55 (Fig. 5a) and returned to normal in mice receiving transfers of RBCs from WT donor mice (Fig. 5a). Under in vitro conditions, MZ B cell CD97 surface abundance was reduced after 45 min of co-incubation with WT RBCs in a shaker at 1,000 r.p.m., which generates a shear stress of approximately 14 dyne cm−2 (ref. 31; Fig. 5b). MZ B cells incubated with RBCs without shaking or incubated with Cd55−/− RBCs with shaking did not show any reduction in CD97 surface abundance (Fig. 5b). When MZ B cells were taken from CD97-deficient mice that had been reconstituted with BM transduced with a construct encoding the non-cleavable T419G form of CD97, incubation with RBCs with shaking did not lead to any change in CD97 surface abundance, consistent with the reduced expression being due to extraction of the NTF (Fig. 5c). To confirm that exposure to CD55+ RBCs under shear stress was causing NTF extraction and not CD97 degradation, we used MZ B cells from chimeras that had been reconstituted with a CD97–GFP fusion protein (Extended Data Fig. 5a). Exposure of MZ B cells expressing this construct to Cd55+/+ RBCs under shear stress conditions led to reduced CD97 surface staining but had no effect on GFP intensity (Extended Data Fig. 5b,c), confirming that the reduced staining was due to extraction of the NTF. Finally, we transferred splenic B cells from chimeras expressing the CD97–GFP fusion protein into WT or Cd55-KO recipients and analyzed the transferred cells in recipient blood 30 min later. Due to the rarity of transferred MZ B cells in the recipient blood, we instead tracked changes in CD97 and GFP abundance in transferred follicular B cells. Compared to B cells in Cd55-KO recipients, B cells in the blood of WT recipients had low CD97 surface staining, but GFP intensity was unaffected (Fig. 5d,e). Taken together, these data provide evidence that CD97 on MZ B cells undergoes NTF extraction following encounter with CD55+ RBCs under shear stress conditions.
a, Representative histogram (left) and geometric mean fluorescence intensity (MFI; right) of surface CD97 expression on MZ B cells in mice with blood transfusion as indicated (n = 8 recipients in each group). Line colors in the histogram are for mice of the types shown by color code beneath the graph. b,c, Splenocytes from Cd55−/− mice (b; n = 6 in each group) or CD97 T419G-expressing BM chimeras (c; n = 8 in each group) were cocultured for 45 min with Cd55−/− or Cd55+/+ RBCs on a shaker or not. Representative histograms (left) and MFIs (right) of surface CD97 expression on MZ B cells are shown. Line colors in the histograms are for conditions of the types shown by color codes beneath the graphs. Each symbol represents one incubation, and each donor contributes two incubations. d,e, MFI of surface CD97 expression (d) and GFP (e) on transferred MZ B cells from CD97–GFP fusion protein-expressing BM chimeras in blood collected from Cd55−/− (n = 4) or Cd55+/+ (n = 5) recipients after a 30-min transfer as indicated. Each symbol represents one mouse. f, Representative histogram plot of surface CD97 on naive B cells and MZ B cells from the human spleen. g, Representative histogram (left) and MFI (right) of surface CD97 expression on human MZ B cells after incubation for 45 min with human RBCs on a shaker or not (n = 5 in each group). Each symbol represents one incubation, and line colors in the histogram are for conditions of the types shown by color code beneath the graph. Lines in graph denote means. Data are pooled from two independent experiments (a–e) or are representative of two independent experiments (g). Statistical significance was tested by one-way ANOVA followed by Tukey’s multiple-comparisons test (a–c and g) or two-tailed t-test (d and e).
Source data
Human splenic MZ B cells are CD27hi, and they have variable levels of CD1c32,33. Analysis of published single-cell RNA-sequencing data of human spleen B cells showed higher ADGRE5 mRNA expression in the two MZ B cell clusters, although expression of transcripts did not appear abundant compared to that for CD27 or CR2 (Extended Data Fig. 5d)33. However, fluorescence-activated cell sorting analysis established that most MZ B cells express higher amounts of CD97 protein than follicular B cells (Fig. 5f and Extended Data Fig. 5e). Human RBCs were positive for CD55 expression (Extended Data Fig. 5f), as expected30. When RBC-depleted human splenocytes were incubated in the presence of human RBCs under shear stress, there was a marked reduction in CD97 surface staining compared to splenocytes incubated without RBCs or without shear stress (Fig. 5g). These data suggest that human MZ B cells experience CD97 NTF extraction during interactions with RBCs under shear stress conditions.
CD97 promotes cell membrane retraction
Activation of Rho-based signaling is established to promote retraction of cell membrane processes, and there is in vitro evidence of CD97 activation of Rho21,23,34. However, it remains unclear whether shear stress-mediated activation of CD97 is sufficient to engage Rho activity and membrane retraction. We examined the impact of CD97 expression (Extended Data Fig. 6a) on the morphology of adherent HEK293T cells. Cells expressing WT CD97 or the cleavage-resistant T419G CD97 variant showed a similar irregular morphology (Extended Data Fig. 6b). By contrast, cells expressing the CD97 CTF were more rounded (Extended Data Fig. 6b). The differences in shape were confirmed by comparing the longest dimension of the cells under each condition (Fig. 6a and Extended Data Fig. 6c).
a, HEK293T cells were transfected with CD97 (WT)–GFP, CD97 (T419G)–GFP or CTF–GFP fusion constructs. Longest diameters of individual WT- (n = 62), T419G- (n = 62) and CTF-transfected (n = 63) cells are shown. Each symbol represents one cell. One of three independent experiments with similar results is shown. b, Schematic diagram of an optically trapped ligand-coated bead applying pulling force while in contact with the cell membrane. The cell is also indicated to contain an anillin–GFP RhoA biosensor. c,d, HEK293T cells were cotransduced with anillin–GFP RhoA biosensor and CD97 (WT)–Scarlet, CD97 (T419G)–Scarlet or CD97 (PBM)–Scarlet fusion lentivirus and sorted. c, Cell trap forces of individual WT- (n = 70), T419G- (n = 64) and PBM-transduced (n = 62) cells. d, Normalized active RhoA biosensor fluorescence intensity of WT- (n = 27), T419G- (n = 20) and PBM-transduced (n = 32) cells as indicated. e–g, Examples of anillin–GFP active RhoA biosensor signals at the site of bead contact and membrane tether (blue arrowhead) of one HEK293T cell expressing CD97 (WT)–GFP fusion (e), CD97 (T419G)–GFP fusion (f) or CD97 (PBM)–GFP fusion (g) protein. The blue arrowhead indicates the site of bead pulling. The time series begins at the start time for pulling, which was about 30 s after the bead contacted the cell. See corresponding Supplementary Videos 7–12; scale bar, 5 μm (e and f); scale bar, 3 μm (g). Data are representative of three independent experiments (a) or a pool from two (c–g) independent experiments. Statistical significance was tested by one-way ANOVA followed by a Tukey’s multiple-comparisons test (a and c) or mixed-effects analysis followed by a Geisser–Greenhouse correction (d).
Source data
To directly test whether the CD55–CD97 interaction leads to membrane retraction, we turned to optical tweezers35. Beads coated with recombinant mouse CD55 were brought into contact with HEK293T cells expressing mouse CD97–Scarlet for 30 s to allow a membrane tether attachment. Subsequently, the bead was pulled away using the optical trap35. The displaced bead remains attached to the cell via a membrane tether, which exerts a restoring force that is proportional to the square of membrane tension35,36 (Fig. 6b). These measurements revealed that cells expressing WT CD97 exerted an almost twofold greater tether force than cells expressing the signaling defective T419G CD97 variant (Fig. 6c and Extended Data Fig. 6d). Similar analysis of the CD97 PBM variant mutated in the C-terminal PBM showed a partial reduction in the retraction response (Fig. 6c and Extended Data Fig. 6d). Using cells that also expressed the active RhoA biosensor anillin37 and WT CD97, induction of RhoA activation at the bead contact site was observed within a few seconds after applying a pulling force on the bead (Fig. 6d,e and Supplementary Videos 7 and 8). RhoA activation was not observed at the bead contact site in equivalently treated cells that expressed the T419G CD97 mutant (Fig. 6d,f and Supplementary Videos 9 and 10). RhoA activation was induced by the CD97 PBM mutant, although the activity was less sustained than for WT CD97, suggesting that there is a difference in the kinetics of recruitment and activation of RhoA (Fig. 6d,g and Supplementary Videos 11 and 12). These data are consistent with CD97 becoming activated to signal via RhoA and cause membrane retraction after binding CD55+ particles and exposure to pulling forces.
CD97 deficiency leads to reduced T cell-independent responses
MZ B cells have an established role in mounting rapid antibody responses against circulating T cell-independent antigens, such as bacterial capsular polysaccharides4. Trinitrophenol (TNP)-haptenated Ficoll, a large highly branched polysaccharide, is widely used as a model T cell-independent antigen. Like other polysaccharide antigens, Ficoll becomes rapidly coated with complement fragments following systemic injection, and the complex can bind to MZ B cells via complement receptors38,39. Forty minutes after injection, the amount of TNP–Ficoll bound to MZ B cells was lower in Cd97-KO mice than in littermate control mice (Fig. 7a). This defect was cell intrinsic, as Cd97-KO MZ B cells bound less TNP–Ficoll than WT B cells in mixed BM chimeras (Fig. 7b). Similar findings were made in mice lacking Gα13 in B cells (Fig. 7c). CD55-deficient and ARHGEF1-deficient mice also showed reductions in TNP–Ficoll capture (Extended Data Fig. 7a–c). These data indicate that the MZ B cells remaining in CD97 pathway-deficient mice are altered in their ability to capture a blood-borne antigen. When spleen B cells were injected intravenously just before treating mice with TNP–Ficoll, WT and Cd97-KO MZ B cells in blood circulation captured Ficoll with similar efficiency (Extended Data Fig. 7d). B1 cells can also contribute to T cell-independent antibody responses5. However, peritoneal B1 cells showed minimal capture of intravenously injected TNP–Ficoll, and binding was unaffected by CD97 deficiency (Extended Data Fig. 7e,f). To test the impact of B cell CD97 deficiency on the antibody response, IgHa:IgHb mixed BM chimeras were generated by combining Cd97-KO (or control) IgHb BM with BM from WT mice congenic for the IgHa locus. Flow cytometric analysis confirmed that, compared to their frequency in control mixed chimeras, Cd97-KO IgHb MZ B cells were reduced in WT IgHa:Cd97-KO IgHb mixed chimeras (Extended Data Fig. 7g). At day 6 after TNP–Ficoll immunization, serum was collected, and the amount of TNP binding IgMa and IgMb was measured. Although the production of TNP-specific IgMa from WT cells was equivalent in each type of mixed BM chimera, the production of TNP-specific IgMb by Cd97-KO cells was significantly reduced (Fig. 7d). Immunization of the same types of mixed chimeras with the T cell-dependent antigen NP-CGG led to IgM and IgG1 responses at day 12 that were unaffected by CD97 deficiency (Fig. 7e,f). Finally, we examined the IgM response to TNP–Ficoll in CD97-deficient mice and found that it was reduced (Fig. 7g). Thus, CD97 is required for mounting an intact early IgM response against a blood-borne polysaccharide antigen.
a–c, Mice were analyzed 40 min after intravenous TNP–Ficoll injection. a, Representative histogram (left) and MFI (right) of TNP–Ficoll levels on MZ B cells in Adgre5−/− (n = 8) and control (n = 8) mice. b, MFI of TNP–Ficoll levels on MZ B cells in WT:Adgre5−/− (n = 6) and control (n = 6) mixed BM chimeras. Lines connect data from the same animals. c, Representative histogram (left) and MFI (right) of TNP–Ficoll levels on MZ B cells in Gna13cKO (n = 8) and control (n = 7) mice. d–f, Enzyme-linked immunosorbent assay (ELISA) analysis of serum from 50:50 mixed IgHa WT (Adgre5+/+) with IgHb Adgre5+/+ or Adgre5−/− chimeras. d, Mixed chimeras were immunized intravenously with TNP–Ficoll. TNP-specific IgMa (left) and IgMb (right) in serum from the indicated time points (n = 11 or 12 chimeras for each condition). e,f, Mixed chimeras were immunized intraperitoneally with NP-CGG with Alum. NP-specific IgG1a (e; left), IgG1b (e; right; n = 9 for preimmunized (preimm) and n = 6 for postimmunized groups), IgMa (f; left) and IgMb (f; right; n = 6 for each group) in serum from the indicated time points. g, TNP-specific IgM in serum from Adgre5+/− (n = 13) or Adgre5−/− (n = 13) mice day 5 after TNP–Ficoll immunization. In d–g, RU indicates relative units. Data are pooled from two (a–c, e and f) or three (d and g) independent experiments. Each symbol represents one mouse, and lines denote means. Statistical significance was tested by two-tailed t-test (a, c and g) or two-way ANOVA followed by a Sidak’s multiple-comparisons test (b) or one-way ANOVA followed by a Tukey’s multiple-comparisons test (d–f).
Source data
link