Abstract
Infection of mice with murine rotaviruses induces life-long immunity, characterized by high levels of IgA in the intestine and large numbers of rotavirus (RV)-specific Ab-secreting cells in gut-associated lymphoid tissues. Lymphocyte trafficking into gut-associated lymphoid tissues is mediated by interaction of the α4β7 integrin on lymphocytes with the vascular mucosal addressin cell adhesion molecule-1. To determine whether B cell memory for RV correlates with α4β7 expression, we transferred sorted B220+ phenotypically defined memory (IgD−α4β7high and IgD− α4β7−) and naive (IgD+α4β7+) splenocytes into recombination-activating gene-2 knockout mice (B and T cell-deficient) that were chronically infected with RV. Only mice receiving α4β7high memory (IgD−) B cells produced RV-specific IgA in the stool, cleared the virus, and were immune to reinfection. α4β7high (but not α4β7−) memory B cells from donors boosted as much as 7 mo previously also cleared the virus, indicating that α4β7high memory B cells maintain long term functional immunity to RV. Although only α4β7high memory cells provided mucosal immunity, α4β7− cells from recently boosted donor animals could generate RV-specific serum IgG, but, like naive (IgD+) B cells, were unable to induce viral clearance even 60 days after cell transfer. These data indicate that protective immunity for an intestinal pathogen, RV, resides in memory phenotype B cells expressing the intestinal homing receptor, α4β7.
Rotavirus (RV)4 is a segmented dsRNA virus of the family Reoviridae responsible for up to a million infant and child deaths per year worldwide (1). RV infection occurs in and is largely limited to the villus epithelium of the small intestine (2). The specificity of viral replication ensures that the immunologic response to RV is focused in the intestinal compartment. In infected mice, RV-specific Ab-secreting cells (ASCs) are found predominantly in the lamina propria (LP) and Peyer’s patches (PP), and express primarily the IgA isotype (∼90% IgA+ in LP, ∼60% IgA+ in PP) (3, 4, 5, 6, 7). In both neonatal and adult mice, large amounts of RV-specific IgA are found in stool samples following infection with murine virus, and stool IgA persists for up to 1 yr following primary infection (8, 9). Anti-RV IgA in stool and serum correlates with protection from reinfection (8, 10). In contrast, while high serum IgG levels are also observed (10, 11, 12), their role in immunity is unclear. Experiments using immunodeficient mice have demonstrated that local Ab is sufficient to clear viral infection in the absence of T cells (3, 13) and is the principal determinant of protection from reinfection (3, 14, 15).
The preferential accumulation of RV-specific ASCs in PP and intestinal LP suggests that these cells are able to traffic specifically to these sites. Lymphocytes that have previously encountered Ag display tissue-selective trafficking patterns (16, 17, 18, 19, 20). Mesenteric lymph node (MLN) B blasts, for example, bind preferentially to LP endothelium in vitro (21) and preferentially traffic to mucosal (gut, cervix, vaginal wall, uterus, mammary gland, and MLN) as opposed to peripheral organs and tissues (22, 23). Selective localization is thought to be directed in part by the differential expression of homing receptors on the surface of circulating cells and endothelial cell adhesion molecules in the target tissues (17, 19, 21, 24). The integrin α4β7 is essential in lymphocyte homing to intestinal LP and PP (25, 26, 27). Moreover, recent studies demonstrate that memory phenotype T lymphocytes expressing α4β7 recirculate selectively through the intestine (28), and that α4β7high, but not α4β7−, memory T cells home to the intestinal lymphoid organs, PP (16). Homing of these α4β7 defined subsets of memory cells to other mucosal-associated lymphoid tissue (cervix, vaginal wall, uterus, and mammary gland) has not been examined.
Circulating B cells can be divided into naive and memory/effector cells based on the expression of cell surface markers such as IgD in mice (29, 30) and humans (31, 32). Memory phenotype (IgD−) B cells in human peripheral blood subdivide into β7high and β7− populations (L. S. Rott, manuscript in preparation), of which only β7high cells bind selectively to the mucosal vascular addressin MAdCAM-1 (33). In humans, memory phenotype (IgD−) B cell blasts in LP express high levels of α4β7 and are primarily IgA+ (34). Moreover, recent studies of B cells responding to typhoid vaccination and bacterial enteric pathogens have shown that in comparison with systemic vaccination, mucosal presentation of Ags (by oral or rectal vaccination) is associated with increased numbers of circulating Ag-specific B cells expressing α4β7 (35, 36, 37, 38).
To determine whether memory for an intestinal virus correlates with expression of α4β7, we have assessed the response of memory phenotype (α4β7high IgD− and α4β7− IgD−) and naive phenotype (α4β7+IgD+) B220+ cells to RV. Using in vitro ELISPOT assays with sorted cells, we show that only the IgD− cells expressing high levels of α4β7 produce RV-reactive Abs. Moreover, in vivo cell transfer experiments demonstrate that critical effector functions, including anti-RV IgA production, clearance, and protection against reinfection correlate only with transfer of memory B cells expressing high levels of α4β7. Interestingly, mice receiving α4β7− memory B cells from recently immunized donors produced RV-specific IgG in serum, but continue to shed RV in the stool. Thus, following oral immunization, both α4β7high and α4β7− memory B cells can transfer RV-specific Ab responses, but only the α4β7high memory subset is capable of providing protective humoral immunity.
Materials and Methods
Mice, viruses, and virus inoculation
Stocks of wild-type murine EC RV (EDIM, Cambridge) were prepared as intestinal homogenates and titrated in mice as previously described (8). Stocks of tissue culture-adapted rhesus rotavirus (RRV) were prepared and titrated to determine plaque-forming units (pfu) per milliliter as previously described (39).
Six-week-old C57BL/6 mice were obtained from the Charles River Laboratories (Hollister, CA) and were bred in the Palo Alto Veterans Administration breeding facility to be used as donors for cell transfer experiments. Mice were orally gavaged with 105 diarrheal dose 50 (DD50) (14) of EC RV after receiving 100 μl of 1.33% sodium bicarbonate to neutralize stomach acid. Viral shedding was monitored by ELISA to assess the progress of infection (8). Following initial infection, immunized mice were boosted by oral gavage with a second dose of virus (105 DD50). Between 1 and 2 mo after boosting, splenocytes and PP were harvested for FACS analysis. Also, between 1 and 2 mo after oral boosting, splenocytes from donor mice were harvested for sorting experiments. For systemic inoculation, mice were injected i.p. with live RRV at a dose of 106 pfu/mouse. One month later, a second dose of 106 pfu/mouse was administered. For cell transfer experiments examining earlier boost time points, 6-wk-old C57BL/6 mice were obtained from the Palo Alto Veterans Administration breeding facility, orally gavaged, and boosted following standard procedures (as described above). Splenocytes from these inoculated mice were harvested for sorting 7 mo after boosting.
Recipient RAG-2 mice (40) were obtained from GenPharm International (Mountain View, CA), bred in the Palo Alto Veterans Administration breeding facility, and infected with 105 DD50 of the murine EC virus as previously described (14). Stool samples from infected RAG-2 mice were assayed for viral Ag 2 wk postinoculation and again before cell transfer to confirm the establishment of a chronic infection. RAG-2 mice used in this study were chronically infected for 1 to 3 mo before study.
Isolation of lymphocytes from blood, spleen, lymphoid organs, LP, intraepithelial lymphocytes (IELs), and bone marrow (BM)
Lymphocytes were isolated from mouse MLN, peripheral lymph nodes, PP, and spleens as previously described (16). LP lymphocytes were isolated by collagenase or by dispase digestion of intestines from which PP had been removed, as previously described (3, 41). IELs were isolated as previously described (14). BM was obtained from the femurs of mice by removing the ends of the bone with a razor blade and flushing the inside of the bone with RPMI supplemented with 10% FCS (RPMI 10) through a 22-gauge needle. BM was dissociated by repeated resuspension through the needle and was washed with 10 ml of RPMI 10.
ELISPOT assay
The ELISPOT assay for detection of ASCs was performed as previously described (3) with the following modifications. Millipore 96-well filtration plates with Immobilon-P membranes (Millipore, Bedford, MA) were coated overnight at 4°C with 1 μg/ml goat anti-mouse Ig (Kirkegaard & Perry Laboratories, Gaithersburg, MD) for detection of total ASCs. For detection of anti-RV ASCs, plates were coated with a 1/1000 dilution of rabbit anti-RRV hyperimmune serum (produced in the laboratory of H. Greenberg); washed with 10 mM Tris (pH 7.5), 100 mM NaCl, and 5 mM CaCl2; and incubated with a 1/5 dilution of RRV stock virus overnight at 4°C. Just before the addition of lymphocytes, plates were washed with 10 mM Tris (pH 7.5), 100 mM NaCl, and 5 mM CaCl2 and blocked for 30 min at 37°C with RPMI 10. Cells were added to plates at various dilutions in RPMI 10 and cultured overnight at 37°C. Secreted Abs were detected with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG, anti-IgA, or anti-Ig(A+G+M) (Kirkegaard & Perry Laboratories) and visualized by reaction with the precipitating substrate, 3-amino-9-ethylcarbazole. Spots formed by precipitated substrate were counted using a Bausch and Lomb stereomicroscope at 1.5-fold magnification. Under the conditions used in these assays, the efficiency of capture of ASCs was essentially 100%, since equal total ELISPOTs were obtained when assaying the RV-specific hybridoma, IE11, on plates coated with RV and on plates coated with anti-Ig mAb (data not shown; n = 15).
In the ELISPOT assays described in Figure 4⇓, VP2 and VP6 virus-like particles (made from baculovirus recombinants expressing heterologous bovine RF rotavirus) (42) were used as Ag for coating plates directly. Since in mice >90% of the anti-RV Ab response is directed to VP2 and VP6 (5) (M. Franco, unpublished observations), the results using the two protocols (reported in Tables I, II, and III vs Fig. 4⇓) were similar. As a control for the specificity of the ELISPOT assay, cells were also incubated in wells from which Ag had been omitted (and resulted in no detectable spots; n = 27).
Detection of viral Ag and Abs
Viral Ag detection was performed by ELISA as described previously (8). Briefly, microtiter plates (Dynatech, McLean, VA) were coated with diluted hyperimmune guinea pig anti-RRV serum and blocked with 5% nonfat milk. Stool samples were made to a 10% suspension and added to plates for 2 h at 37°C. Ag was detected with rabbit anti-RRV serum followed by HRP-conjugated goat anti-rabbit serum (Kirkegaard & Perry Laboratories). Visualization was performed by incubation of plates with 2,2′-azino-di-[3-ethyl-benzthiazoline sulfonate] (6) substrate (Kirkegaard & Perry Laboratories), followed by quenching of the reaction with 10% SDS. Plates were read using an EIA Autoreader (Bio-Tek Instruments, Burlington, VT) at 405 nm. Clearance was defined as the point at which ELISA values fell below 0.08 OD. To determine background absorbance levels, stool samples from mice that had not been infected with RV (OD, ≤0.08) in addition to wells with no stool sample (OD, ≤0.04) were assayed by ELISA. Also, samples were assayed in the absence of rabbit anti-RRV serum (OD, ≤0.04) in the absence of HRP-conjugated goat anti-rabbit serum (OD, ≤0.05) as well as on sample wells that were not coated with guinea pig anti-RRV serum (OD = −0.05 to −0.06).
Analysis of RV-specific Abs was performed as described previously (8). Briefly, RV-specific Abs were captured on ELISA plates that were first coated with hyperimmune guinea pig anti-RRV serum and were subsequently treated overnight with a 1/5 dilution of RRV stock virus at 4°C. After three washes, 5% stool samples were added to the plates at 37°C. Ab was detected with HRP-conjugated anti-mouse IgA or IgG (Kirkegaard & Perry Laboratories) and visualized as described above. To determine background absorbance levels, stool samples from animals that were not infected with RV were assayed (OD = 0.011–0.030; n = 6). For determination of total Ig, plates were coated with goat anti-mouse Ig(A+G+M) (Kirkegaard & Perry Laboratories) and blocked with 5% nonfat dry milk.
Analysis of serum Ab was performed as described previously (8). Briefly, serum samples were diluted serially in 1% nonfat dry milk and added to plates at 37°C. After 2- to 4-h incubation, Ab was detected with HRP-conjugated anti-mouse IgA or IgG (Kirkegaard & Perry Laboratories) and visualized as described. Serial threefold dilutions of serum were performed. The titer of the serum for each isotype of Ab was defined as the maximum dilution of serum that had an OD405 ≤0.1. Background was defined as the mean OD of three wells incubated without serum and was typically between 0.047 and 0.067. The serum titers were log3 transformed to calculate the geometric mean titer.
Abs and reagents
For sorting, biotinylated mAb to B220 (CD45R/B220; RA3-6B2), PE-conjugated mAb to α4β7 (DATK32), FITC-conjugated mAb to IgD (217-170), FITC-conjugated mAb to CD4 (RM4-5), and biotin-conjugated mAb to CD8 (53-6.7) were purchased from PharMingen (La Jolla, CA). Biotinylated mAb to human CD44 (43) (Hermes-1, produced in the laboratory of E. C. Butcher) and PE-conjugated anti-rat-IgG2a (PharMingen) were used as isotype control Abs. Streptavidin Red 613 (SA-Red 613) was purchased from Life Technologies (Gaithersburg, MD). For staining of IELs of recipient mice, PE-conjugated mAb to the TCR (H57-597), FITC-conjugated mAb to CD4 (RM4-5), and biotinylated mAb to CD8 (53-6.7) were purchased from PharMingen (La Jolla, CA). Streptavidin-conjugated peridinin chlorophyll protein was purchased from Becton Dickinson (San Jose, CA).
Staining, cell sorting, and FACS analysis
Splenocytes to be sorted from donor RV-infected C57BL/6 mice were harvested, disaggregated by teasing between glass slides, and pressed through stainless steel screen mesh. Pooled cells were washed with 5 to 10 ml of DMEM supplemented with 10% FBS (DMEM-10) and then treated with 5 ml of lysis buffer (0.15 M NH4Cl, 1 mM KHCO3, and 0.1 mM Na2EDTA) for 5 min to lyse erythrocytes. After lysis, splenocytes were washed twice and resuspended in DMEM-10. The yield was typically 1 to 2 × 108/harvested spleen. To maximize purity, CD4+ splenocytes were isolated with anti-CD4 beads (Dynabeads, Dynal, Lake Success, NY) and subsequently detached from beads following the manufacturer’s protocol.
Splenocytes were stained for sorting as follows. Dynabead-isolated CD4+ splenocytes were incubated on ice with anti-CD4-FITC mAb (0.5 μg/106 cells) for 20 min, washed with DMEM-10, stained with SA-Red 613 (0.5 μl/106 cells) for 20 min on ice, washed, resuspended in DMEM-10, and then sorted. To isolate B cells, the fraction that did not adhere to anti-CD4 beads was resuspended in DMEM-10 and stained with anti-B220-biotin (0.5 μg/106 cells), anti-α4β7-PE (DATK32; 0.2 μg/106 cells), and anti-IgD-FITC (0.5 μg/106 cells). The cells were then washed, stained with SA-Red 613 (0.5 μl/106 cells), and sorted. In experiments in which donor mice were boosted 7 mo before transfer, cells that did not bind the anti-CD4 beads were further depleted of CD8+ cells by incubation with anti-CD8 beads (Dynabead) following the manufacturer’s instructions. To isolate CD8 cells, splenocytes were stained with anti-CD8-FITC (0.5 μg/106 cells), and CD8+ cells were sorted for transfer into selected recipients (see below).
Cells were sorted on a modified FACStar (Becton Dickinson, San Jose, CA) with a single 488-nm argon laser and three fluorescence detectors. Filters for the three fluorochromes were 530/30 for FITC detection, 585/42 for PE detection, and 630/22 for Red 613 detection. In two experiments (in which donor mice were boosted 1 to 2 mo before transfer) the α4β7− memory and naive populations were sorted twice, and the α4β7high memory population was sorted once. In one experiment (in which donor mice were boosted 7 mo before transfer and CD8+ cells were depleted with anti-CD8 beads before sorting; see above), the α4β7− memory, α4β7high memory, and naive populations were sorted once. The level of purity after sorting once (98.7 ± 1.3) was determined using the FACStar. Because of the scarcity of cells, the purity could not be determined on the transferred cells that had been sorted twice; however, in parallel experiments, we determined by FACStar analysis that the purity after sorting twice was typically at least 99.7%.
All experiments involved cotransfer of CD4+ cells. Because of the possibility that low levels of CD8+ cells might contaminate single-sorted α4β7high (but not double-sorted α4β7−) B cells in the experiment described in Figure 2⇓ and might affect viral clearance, in three animals we transferred 300 double-sorted CD8+ cells from orally inoculated donors (representing a 3% contamination level, the maximum level of impurity we observed in single-sorted cells; see above) along with 10,000 single-sorted α4β7− IgD− B220+ cells and 20,000 double-sorted CD4+ cells. We also transferred 10,000 single-sorted α4β7− IgD− B220+ cells and 20,000 double-sorted CD4+ cells (and no CD8+ cells) in three additional animals (see Table V⇓). Intentional addition of CD8+ contaminants to transferred B cells had no effect on the clearance of RV (data not shown). CD8+ and CD4+ cells were sorted twice. Sorted donor cells were resuspended in sterile saline solution at 105/ml and injected i.p. into chronically infected RAG-2 mice.
In the experiments in which donor mice were boosted 7 mo before transfer and CD8+ cells were depleted, IELs and splenocytes were isolated from recipient mice 22 to 25 days after transfer; stained with mAbs to CD8 (0.5 μg/106 cells), CD4 (0.5 μg/106 cells), and TCR (0.5 μg/106 cells); and visualized by FACS. There were no TCR+CD8+ cells in any mice with transferred sorted cells. (Since RAG-2 mice have a limited population of CD8+ cells (∼3%) (40), but no TCR+ cells, we gated on double-positive (TCR+ CD8+) cells to visualize transferred CD8+ cells.)
Results
RV-reactive B cells populate mucosal tissues after oral, but not after systemic, inoculation
To compare the localization of RV-specific ASCs after oral vs parenteral immunization, we inoculated adult mice either orally with homologous murine RV or i.p. with the heterologous simian RRV. Heterologous RRV was selected for i.p. inoculation because it shares major antigenic epitopes with the mouse RV and has very limited capacity to infect intestinal epithelium or proliferate in the mouse host (and thus is expected to yield a systemic immune response by not invoking mucosal presentation) (2). We then compared ASCs in spleen and GALT of these mice as well as examined localization of ASCs in BM, a site that has not previously been examined in studies of RV infection. The results of ELISPOT analyses of tissues from orally and i.p. inoculated mice are shown in Table I⇓. Consistent with previous reports (4, 6), oral inoculation led to the production of large numbers of RV-specific ASCs in LP, moderate levels in mucosal tissues (PP and MLN), and low levels in spleen. High levels of RV-specific ASCs were also present in BM, but no detectable RV-specific ASCs were found in PLN (superficial inguinal lymph nodes; data not shown). In contrast, i.p. inoculation failed to generate RV-specific ASCs in mucosal tissues, as there were no detectable RV-specific ASCs in LP, MLN, or PP. Instead, RV-specific ASCs were found only in spleen and BM. In one of four mice, RV-specific ASCs (8.5 RV-specific ASCs/100 total ASCs; see Table I⇓) were also detected in the superficial inguinal lymph nodes (possibly reflecting Ag drainage from the abdominal site of i.p. injection). These data suggest that at these doses of homologous and heterologous virus, an efficient intestinal response is generated using only the oral inoculation protocol.
Tissue localization of RV-specific ASCs in orally versus systemically inoculated micea
Moreover, orally inoculated mice had RV-specific IgA in the stool (data not shown) and RV-specific IgA and IgG in serum (Table II⇓). In contrast, i.p. inoculated mice had only RV-specific IgG in the serum (Table II⇓) and no detectable RV-specific Ab in the stool (data not shown). These results agree with previous observations of the isotypes of secreted Abs of RV-reactive cells in stool and serum samples of oral vs parenterally inoculated mice (4, 44). A small, but significant, number of RV-specific IgA-secreting cells was present in BM of i.p. inoculated mice (Table II⇓) even though anti-RV IgA was not detected in serum.
Ig isotype of anti-RV Abs in BM and serum in orally versus systemically inoculated micea
Distribution of α4β7 on splenic and PP B cells, and analysis of RV-reactive ASCs among α4β7-defined memory subsets
Since oral inoculation generates RV-specific ASCs primarily in intestinal sites, we hypothesized that there could be an association between effector function and expression of the intestinal homing receptor, α4β7, on B cells. To test this hypothesis, we first examined the expression of α4β7 on splenic and PP B220+ cells isolated from mice orally inoculated with RV (Fig. 1⇓). To distinguish memory/activated-phenotype from naive phenotype B220+ cells, we stained lymphocytes with Abs to IgD (29, 30). FACS dot plots show three main populations of B220+ cells: naive phenotype cells that express low levels of α4β7 (α4β7+ IgD+) and two subsets of memory/activated phenotype cells, one of which expresses high levels of α4β7 (α4β7hi IgD−) and the other of which expresses negligible levels of α4β7 (α4β7− IgD−).
B220+ cells in PP and spleen of RV-immunized mice. FACS dot plot shows gated B220+ cells in PP (a) and spleen (b) from mice orally inoculated with murine RV (see Materials and Methods). Lymphocytes were stained with Abs to B220, IgD, and α4β7. B220+ small lymphocytes were gated and analyzed for IgD (x-axis) and α4β7 (y-axis) expression. The FACS dot plot shows three populations (delineated on the plot) of B220+ small lymphocytes: two memory phenotype populations (α4β7high IgD− and α4β7− IgD−) and one naive phenotype population (α4β7+ IgD+). Typical gates used to sort α4β7high IgD− B220+ (A), α4β7− IgD− B220+ (B), and α4β7+ IgD+ B220+ (C) small lymphocytes for transfer experiments are shown on the plot. The sorted α4β7high IgD− B220+, α4β7− IgD− B220+, and α4β7+ IgD+ B220+ cells with these gates represented 1.3 ± 0.5, 8.1 ± 4.3, and 26 ± 7% of the total small B220+ cells, respectively (mean ± SE; n = 3).
We sorted memory/activated phenotype (IgD−) splenocytes from orally inoculated mice according to expression of α4β7 (for typical sort gates, see Fig. 1⇑b) and analyzed the number of RV-reactive APCs in the sorted populations using ELISPOT assays (see Materials and Methods). As shown in Table III⇓, RV-reactive cells represented 2.3% of the total Ig-secreting cells in the sorted α4β7high memory population, but no detectable RV-reactive cells were in the α4β7− memory population. Thus, mature RV-specific ASCs are significantly enriched in the α4β7-expressing subset after oral RV inoculation. Due to the limited time of incubation with Ag (4–12 h), the ELISPOT assay only detects cells actively secreting Abs and therefore would not be expected to detect germinal center or resting memory phenotype cells (18, 45, 46) that could potentially exert effector function after transfer (see below).
RV-reactive ASCs among splenocyte subsets sorted from mice orally infected with EC RVa
Memory (IgD−) B cells expressing α4β7 clear RV infection
We next assessed the ability of α4β7-defined B cell subpopulations to clear RV. We sorted B220+ spleen cells from orally inoculated mice into α4β7high IgD− (2% of B220+ cells), α4β7− IgD− (8% of B220+ cells), and α4β7+ IgD+ (10% of B220+ cells) subsets for injection into immunodeficient RAG-2 (B and T cell-deficient) mice that were chronically infected with RV (for typical sort gates, see Fig. 1⇑b) We gated on small B220+ lymphocytes to separate presumptive memory B cells from the blast cell population (which has a larger forward scatter) (18, 47) and transferred 10,000 sorted small B220+ cells. In addition to sorted B cells, sorted CD4+ T cells (20,000) from orally immunized mice were cotransferred to provide T cell help for reactivation of memory B cells (45).
Figure 2⇓ shows the effects of the transferred cell populations on viral shedding in chronically infected RAG-2 mice. Mice receiving α4β7high IgD− and CD4+ T cells clear the virus 12 days after transfer, whereas mice receiving α4β7− IgD− or naive phenotype α4β7+ IgD+ B cells in conjunction with CD4+ T cells failed to clear the virus for up to 60 days after transfer (when the experiment was terminated). Chronically infected RAG-2 mice receiving CD4+ cells alone also failed to clear the virus.
Viral Ag shedding in RAG-2 mice chronically infected with murine RV: effect of transfer of sorted B cell subsets. Ten thousand sorted B cells and 20,000 sorted CD4+ T cells from C57BL/6 mice orally inoculated with murine RV were resuspended in sterile saline and injected i.p. into chronically infected recipient RAG-2 mice. C57BL/6 donor mice were orally boosted with murine RV 1 to 2 mo before transfer of sorted cells. Viral shedding was monitored by ELISA of stool samples (see Materials and Methods). The plot shows mean absorbance values (A405); error bars represent the SE of absorbance values from two separate experiments with totals of five (α4β7high IgD− B220+), six (α4β7− IgD− B220+), and four (α4β7+ IgD+ B220+ and CD4 only) mice receiving each sorted, transferred population, respectively (noted in the upper right corner of the plot).
Anti-RV Abs (IgA) became detectable in the stool of mice receiving α4β7high IgD− cells beginning just before viral clearance (day 10) (A405 = 0.25 ± 0.04; mean ± SE; n = 4). By day 13, stool IgA Ab levels (A405 = 1.1 ± 0.1; n = 4) were comparable to those observed in immunocompetent, orally infected C57BL/6 mice (A405 = 1.4 ± 0.2; n = 5). High Ab levels were still observed at 18 days (A405 = 1.2 ± 0.03; n = 5) after transfer of sorted cells and remained consistently high for up to 60 days after transfer (when the experiment was terminated). These mice were completely protected from reinfection with 105 DD50 of EC at 1 mo after clearance (not shown; n = 3), presumably because of high levels of anti-RV Ab, which clears the virus before it can replicate in the animal. Low to negligible levels of RV-specific Ig were detected in the stool of all recipients of other transferred populations (A405 ≤0.04 for mice receiving α4β7− IgD− and CD4+ cells (n = 6), α4β7+ IgD+ and CD4+ cells (n = 4), and CD4+ cells alone (n = 4)).
We examined serum from each of these recipient mice (Table IV⇓). Mice receiving α4β7high IgD− cells had significant levels of both RV-specific IgA and IgG in serum. In contrast, no RV-specific IgA was produced in recipients of α4β7− IgD− B cells; these α4β7− B cells did, however, give rise to significant levels of RV-specific serum IgG, even though, as stated above, these animals did not clear the virus. Furthermore, titers of total IgG in these mice exceeded those of α4β7high IgD− recipient mice. Mice receiving naive (α4β7+ IgD+) B cells and mice receiving only CD4+ cells did not have significant levels of anti-RV Ig in serum. No Ab was present in the sera of RAG-2 mice receiving only CD4+ cells, confirming the absence of B cells in this sorted population. Only orally infected immunocompetent C57BL/6 controls and RAG-2 mice receiving naive (α4β7+ IgD+) cells had detectable levels of total IgM. Total serum IgG was detected in all RAG-2 mice receiving B220+ cells, whereas total IgA was only detected in orally infected C57BL/6 mice and RAG-2 mice receiving α4β7high IgD− cells.
Serum Ab responses in RV-infected Rag-2 mice after transfer of sorted lymphocytes from donor C57BL/6 mice orally boosted with EC RV 1 to 2 mos prior to transfera
Memory (IgD−) B cells expressing α4β7high maintain long term immunity to RV
To examine long term immunity provided by RV-specific B cells, we transferred sorted B220+ splenocytes from donor C57BL/6 mice that had been orally boosted with RV at an earlier time point (7 mo before transfer, as opposed to 1–2 mo in the experimental transfers depicted in Fig. 2⇑ and Table IV⇑). As in the previous experiments, we transferred 10,000 sorted B220+ cells along with 20,000 double-sorted CD4+ lymphocytes into chronically infected RAG-2 recipients. Again, only mice receiving α4β7high IgD− cells cleared RV infection (Fig. 3⇓a), although clearance was slightly delayed compared with that in previous experiments (day 16 rather than day 12). As in previous experiments, only mice that received α4β7high IgD− sorted B220+ cells produced RV-specific IgA in the stool (Fig. 3⇓b).
Viral Ag shedding and RV-specific intestinal IgA in RAG-2 mice chronically infected with murine RV after transfer of sorted B220+ cells from donor mice that had been orally boosted 7 mo before transfer. a, Ten thousand sorted B cells and 20,000 sorted CD4+ cells from C57BL/6 mice orally inoculated with murine RV were resuspended in sterile saline and injected i.p. into chronically infected recipient RAG-2 mice. C57BL/6 donor mice were orally boosted with EC virus 7 mo before transfer. Viral shedding was monitored by ELISA of stool samples (see Materials and Methods). The plot shows mean absorbance values (A405), and error bars represent the SE of absorbance values from three mice receiving each sorted, transferred population (noted in the upper right corner). b, Stool samples from RAG-2 mice that received sorted cells were analyzed for RV-specific Ig (IgA) by ELISA (see Materials and Methods). The plot shows mean absorbance values (A405) on day 0 (the day of transfer) and days 6 to 20 after transfer; error bars represent the SE of the mean absorbance values from three mice receiving each sorted population (noted in the upper right corner.)
Mice receiving α4β7high IgD− sorted B220+ cells had RV-specific ASCs in spleen, BM, and LP 22 to 25 days after transfer of sorted cells (Fig. 4⇓). There were higher levels of RV-specific IgA than IgG-producing ASCs in spleen, BM, and LP (∼4-, ∼15-, and ∼100-fold, respectively). Also, there was a significantly higher frequency of RV-specific IgG and IgA ASCs in LP (per resident cell) than in spleen and BM (∼60- and ∼230-fold more ASCs, respectively), reflecting the localization of ASCs to the intestinal compartment where RV proliferates (13) and where RV Ag is retained for at least 20 days after inoculation (7). (Viral Ag within intestinal APCs has been evaluated up to but not beyond 20 days after infection (7).) Mice receiving naive phenotype (α4β7+ IgD+ B220+) cells had low levels of RV-specific ASCs in LP, BM, and spleen (by days 22–25, when ASCs were assessed; Fig. 4⇓). Also, naive B cell recipients had low levels of RV-specific IgA in the stool by day 18 (Fig. 3⇑b), indicating an initial response to RV. No other recipient mice, including α4β7− B cell recipients, had detectable RV-specific Ab in the stool (Fig. 3⇑b) or RV-specific ASCs in BM, spleen, or LP (data not shown; n = 3 for each group of recipient mice).
RV-specific ASCs in spleen, BM, and LP of RAG-2 mice receiving sorted B220+ cells. RAG-2 mice were sacrificed 22 to 25 days after transfer of sorted B220+ cells from donor mice that had been orally boosted 7 mo before transfer. RV-specific IgG and IgA ASCs in spleen, BM, and LP were analyzed by ELISPOT (see Materials and Methods). Bar graphs depict RV-specific ASCs in RAG-2 mice that received α4β7high IgD− B220+ or α4β7+ IgD+ B220+ sorted cells. Values represent the mean number of ASCs per 106 total cells, and error bars represent the SE (n = 3 for each value shown). There were no detectable RV-specific ASCs in spleen, BM, or LP of mice receiving any sorted population of cells other than those shown in the figure.
Levels of total IgA ASCs (per resident cell) were higher in LP than in BM or spleen of α4β7high recipient mice (Table V⇓). Interestingly, total IgG ASCs in spleen and BM of mice transferred with α4β7− IgD− or α4β7+ IgD+ cells were comparable to or greater than those of mice receiving α4β7high IgD− cells. We detected serum Abs to RV only in α4β7high recipients and not in recipients of memory α4β7− or naive α4β7+ B cells (data not shown.) This contrasts with experiments (presented in Table IV⇑ and Fig. 2⇑) with sorted cells from donor mice that had been boosted within 1 to 2 mo of transfer, in which recipients of memory α4β7− B cells had RV-specific IgG in serum. As in our previous experiments, there was no detectable Ab in the serum of RAG-2 recipients of CD4+ cells alone (data not shown), confirming that no B cells were transferred in this sorted population.
Total ASCs in spleen, BM, and LP of RV-infected Rag-2 mice transferred with sorted B220+ cellsa
Discussion
To determine whether B cell memory for an intestinal virus correlates with expression of the intestinal homing receptor, α4β7, we transferred sorted memory B splenocytes expressing high vs negligible levels of this integrin into RAG-2 mice chronically infected with RV. For these analyses, B-lineage cells lacking surface IgD were sorted as the presumptive memory population, and surface IgD+ B cells were assumed to be predominantly naive. We show that recipients of 10,000 α4β7high memory B cells, and not α4β7− memory or α4β7+ naive B cells, clear RV. Moreover, recipients of α4β7high memory B cells are protected from reinfection, presumably due to clearance of RV by Ab at intestinal surfaces before the virus can infect enterocytes. Interestingly, we detected RV-specific Ig in the sera of both α4β7high and α4β7− recipient mice (when donor mice were boosted 1–2 mo before transfer). Thus, although both α4β7high and α4β7− memory B cells from orally immunized mice produce RV-specific Ig, only α4β7high cells provide protective humoral immunity. This is the first direct demonstration that the ability to confer functional immunity at intestinal surfaces segregates among memory B cells according to homing receptor expression.
To examine the B cell response to RV in orally vs i.p. inoculated mice, we compared ASCs in spleen, GALT, and BM in mice inoculated via each route. Nonreplicative simian RRV was used for i.p. immunization because murine RV would spread rapidly to infect intestinal epithelia. We confirmed previous reports showing that mice orally inoculated with RV produce virus-specific ASCs in spleen and GALT (MLN, PP, and LP) (4, 6, 7). We also showed that significant numbers of RV-specific ASCs localize to BM, a site that has not previously been examined in studies of RV infection. We demonstrated that in contrast to oral inoculation, if virus was injected i.p., RV-reactive cells localized to spleen and BM, but not to intestinal lymphoid tissue (PP, MLN, and LP). This agrees with previous reports demonstrating that the majority of the response to RRV in i.m. inoculated mice localizes in peripheral lymph nodes (inguinal lymph nodes and draining ipsilateral lymph nodes), with significantly less response in GALT (12, 44).
The route of RV inoculation also plays a role in determining the isotype and titer of anti-RV Abs. We confirmed the findings of previous studies showing that high titers of serum anti-RV IgG and IgA as well as stool anti-RV IgA are observed following oral infection of mice with both murine and heterologous viruses (9, 10). No anti-RV IgG is detected in the stool, although anti-RV IgG ASCs can be detected in the LP of orally infected mice (4, 5). Our studies also confirm that parenteral administration of heterologous viruses results in high titers of primarily anti-RV IgG, and not IgA, in serum (12). In this case, some anti-RV IgG is found in the lumen and is thought to be due to transport from circulating serum IgG (12, 44, 48). Importantly, the isotypes of RV-specific Ig in serum and stool in transfer experiments with memory α4β7high vs α4β7− cells mimic the isotypes observed at these sites in orally vs peritoneally inoculated normal mice. α4β7high cells (from orally inoculated donors) generate IgA and IgG in serum and RV-specific IgA and IgG ASCs in LP, but only IgA, and not IgG in stool. On the other hand, α4β7− cells never give rise to IgA in serum and produce serum IgG only when donor animals were recently boosted (1–2 mo before transfer). It is possible that the α4β7− B cells responsible for this IgG production are B cells primed outside the intestine, as viral Ag has been detected in the spleen after infection with RV (7). Alternatively, it is possible that these α4β7− cells responsible for RV-specific IgG production are not mature memory cells, but rather, are residual germinal center B cells that lack surface homing receptor expression (49), including β7.
We have also assessed the anti-RV ASC response in BM of orally vs parenterally inoculated mice and in RAG-2 recipients of sorted B cell subsets. As for stool, LP, and serum, our analyses revealed similarities in the isotypes of BM ASCs in orally inoculated immunocompetent C57BL/6 mice and in RAG-2 recipients of α4β7high B cells and, on the other hand, between i.p. inoculated normal mice and RAG-2 recipients of α4β7− B cells (even though the transferred populations were both from orally inoculated donors). We observed primarily IgA ASCs in BM of mice orally inoculated with RV as well as in RAG-2 recipients of α4β7high B cells. Conversely, systemically inoculated normal mice and RAG-2 recipients of α4β7− cells produced primarily IgG ASCs in BM. Furthermore, these data suggest that in intestinal infection, ASCs localize efficiently not only to LP but also to BM, which is believed to be a source of long term Ab production (50). The predominance of IgA ASCs in BM in orally inoculated normal RV-infected mice contrasts with the predominance of IgG ASCs in the marrow compartment in most reported models, perhaps reflecting the unusual degree of site (intestine) selectivity of RV Ag presentation.
Based on studies of adult SCID mice hosting transplanted hybridomas secreting RV-specific IgA vs IgG, we have hypothesized that for Abs to mediate effective blocking and clearance of RV infection, the Abs must be transcytosed by the polyimmunoglobulin receptor (pIgR) into villous epithelium, where the virus replicates, and then into the lumen, where neutralization occurs (13). The pIgR transports J chain-containing dimeric IgA and pentameric IgM from the intestinal LP across the epithelial barrier into the lumen (51, 52, 53, 54, 55). IgG cannot enter epithelial cells through direct transport by pIgR. In our transfer experiments, RV-specific serum IgG, observed in mice transferred with α4β7− cells, is ineffective at clearing the virus, most likely due to the lack of transport of IgG into or through villous epithelium. On the other hand, it has been shown that in suckling mice and neonatal calves, IgG can protect against infection by RV and other viruses (such as influenza virus) whose site of replication is confined to the mucosal surface (56, 57, 58, 59). However, experiments with neonatal animals may not be relevant to our transfer model. Unlike our model, which uses adult mice as both recipients and donors, oral ingestion of Ab (in mother’s milk) results in localization of substantial amounts of IgG in the neonatal intestine. Similarly, when rabbits are injected parenterally with heterologous RV, it has been shown that IgG and not IgA correlates with protection (60). In these studies high levels of virus-specific IgG (with negligible levels of IgA) are observed in the intestine of i.m. inoculated animals. In contrast, we observe no virus-specific IgG in the stool of any recipient mice, including mice transferred with α4β7− cells, which may account for differences in our observations and those with neonatal mice and i.m. inoculated rabbits.
α4β7high IgD− B cells maintain long term memory for RV, as mice transferred with sorted α4β7high cells from donors boosted as much as 7 mo before sacrifice produce RV-specific Ab and clear the virus. In contrast to experiments in which donors received a recent oral boost (1–2 mo before transfer), we did not detect RV-specific serum Ig in RAG-2 recipients of α4β7− cells from donors boosted 7 mo before transfer. Some researchers have argued that Ag must be intermittently sampled for maintenance of long term memory (45, 46), and while α4β7high memory cells responded and were capable of clearing the virus, there may have been too few RV-reactive α4β7− cells in the transferred population to produce an anti-RV response as the delay from the oral boost increased. While long term immunity to RV was maintained by α4β7high memory cells from animals boosted 7 mo before transfer, clearance of RV in α4β7high recipients is slightly delayed compared with that in mice receiving donor cells from animals boosted only 1 to 2 mo before transfer (clearance on day 16 compared with day 13), perhaps reflecting a decrease in the number of RV-specific cells in this population as well. RV Ag is retained in the intestine (PP and MLN) of normal mice for at least 20 days postinfection (7), and probably much longer. The increased exposure of recirculating RV-specific α4β7high cells to localized Ag in PP and intestine may enhance their survival within RAG-2 recipients compared with RV-specific memory cells that lack the α4β7 intestinal homing receptor. In fact, the majority of the ASCs were RV reactive in α4β7high recipients. Thus, our data are consistent with the hypothesis that the selective homing of lymphocyte subsets to specialized microenvironments can help control lymphocyte life span, function, and population dynamics (or homeostasis) by determining access to regulatory factors (in this case, residual RV Ag) that support differentiation and/or survival (20).
α4β7− memory B cells from donors boosted 7 mo before transfer do not mount a detectable anti-RV response. In contrast, α4β7+ IgD+ (naive) B cells isolated from these same donors produce anti-RV ASCs in LP, spleen, and BM 22 to 25 days after transfer. Also, production of virus-specific intestinal IgA in recipients of α4β7+ naive B cells begins to increase on days 18 and 19 after transfer, while anti-RV intestinal IgA does not change in recipients of α4β7− IgD− B220+ splenocytes. These data indicate that α4β7+ naive B cells, apparently unlike α4β7− memory B cells, have the potential to develop immunity to the virus. It should be pointed out that we did not detect RV-specific Ab in all mice transferred with α4β7+ naive B cells, indicating some variability in the response. Also, while some of these animals produce virus-specific Ab and ASCs, none clears the virus 60 days after transfer; this may simply reflect the presence of inadequate numbers of RV-specific B cells among the naive cells transferred.
What are the implications of these data for RV vaccine development? It is clear from these and previous studies (10, 14, 61) that production of anti-RV (IgA) Ab at intestinal surfaces is the most effective means of clearing and preventing RV infection in the mouse model. Our studies show that following oral infection, the majority of RV-specific B cells populate intestinal sites and BM, and that only transferred α4β7high memory phenotype B cells can mediate clearance of RV in our RAG-2 model. Interestingly, B cells making Ab to RV persist in intestinal sites, particularly the LP, for periods up to 1 yr (6, 9). While this persistence is not observed in humans (62), immunization strategies optimized to stimulate a strong anti-RV response in α4β7high memory B cells will likely be most effective at preventing or attenuating RV infection.
In conclusion, our results demonstrate that B cell memory following oral RV infection correlates with α4β7 expression, and that while α4β7− IgD− B cells can give rise to RV-specific Ig in vivo, these cells primarily produce serum IgG and are ineffective at clearing the virus. Only cells with intestinal homing phenotype produce significant levels of stool and serum IgA and can function in viral clearance and active prevention of reinfection. Furthermore, we show that α4β7high IgD− B cells maintain long term functional immunity to RV, as much as 7 mo after reinfection with the virus. These results extend and complement our studies that have demonstrated that CD8+ cells expressing α4β7high also carry memory to RV and clear RV in chronically infected, immunodeficient mice (63). We conclude that humoral immunity can be segregated according to homing receptor expression, and that α4β7 expression is a critical correlate and determinant of humoral as well as cellular immune responses in the intestinal tract.
Acknowledgments
We thank members of the Butcher and Greenberg laboratories for helpful discussions, and E. O’Hara and K. Youngman for critical reading of the manuscript.
Footnotes
-
↵1 This work was supported by National Institutes of Health National Research Service Award AI08872 from the National Institute of Allergy and Infectious Diseases (to M.B.W.), an Arthritis Foundation Postdoctoral Fellowship (to M.B.W.), Microbiology and Immunology Training Grant 5T32A107328–09 (to J.R.R.), the FACS Core Facility of the Stanford Digestive Disease Center under Grant DK38707 (to L.S.R.), a Walter and Idun Berry Fellowship (to M.A.F.), National Institutes of Health Grants R37AI21632 (to H.B.G.) and AI37832 (to E.C.B.), and Merit Review Awards from the Department of Veterans Affairs Palo Alto Health Care System (to E.C.B. and H.B.G.).
-
↵2 M.B.W. and J.R.R. contributed equally to this manuscript, and E.C.B. and H.B.G. contributed equally as senior authors.
-
↵3 Address correspondence and reprint requests to Marna B. Williams, Ph.D., Department of Pathology, L235, Stanford University School of Medicine, Stanford, CA 94305-5324. E-mail address: mwilliam{at}cmgm.stanford.edu
-
↵4 Abbreviations used in this paper: RV, rotavirus; ASC, antibody-secreting cell; LP, lamina propria; PP, Peyer’s patches; MLN, mesenteric lymph node; MAdCAM-1, mucosal addressin cell adhesion molecule-1; ELISPOT, enzyme-linked immunospot; RRV, rhesus rotavirus; pfu, plaque-forming units; DD50, diarrheal dose 50; RAG-2, recombination-activating gene-2; IEL, intraepithelial lymphocyte; BM, bone marrow; HRP, horseradish peroxidase; PE, phycoerythrin; SA-Red 613, streptavidin Red 613; DMEM-10, Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum; GALT, gut-associated lymphoid tissues; pIgR, polyimmunoglobulin receptor.
- Received March 20, 1998.
- Accepted June 22, 1998.
- Copyright © 1998 by The American Association of Immunologists