Ly49H⁺ NK Cells Migrate to and Protect Splenic White Pulp Stroma from Murine Cytomegalovirus Infection¹

In immunocompetent individuals, acute infection with the β-herpes virus, CMV, is generally asymptomatic, but it causes morbidity and mortality in NK cell-deficient mice and humans (1, 2, 3). Mice infected with murine CMV (MCMV)³ are either susceptible or resistant to disease, depending on their genetic background (4). Resistance maps to a single genetic locus encoding the Ly49 family of NK lectin receptors (5, 6). Specific resistance is through Ly49H, expressed on a subset of NK cells, present in resistant C57BL/6 mice, but not in susceptible BALB/c (7, 8). This receptor directly recognizes the MCMV-encoded protein m157 on the surface of infected cells (9, 10), resulting in a MCMV-specific NK response (11). Although both perforin and IFN-γ expression by NK cells contributes to early protection from MCMV (12), NK perforin expression is crucial to the prevention of the immunopathology observed in the spleen (13).

Recent studies have indicated that the murine pathogen, lymphocytic choriomeningitis virus, infects the chemokine-expressing fibroreticular cells in the splenic white pulp, accounting for disorganization of lymphoid structures and consequent immunodeficiency observed in that disease (14). In this study, we demonstrate that infection of susceptible BALB/c mice with MCMV is also associated with white and red pulp MCMV infection, and that viral replication in both sites is suppressed in Ly49H⁺ C57BL/6 mice. Second, we show that MCMV can infect and replicate in podoplanin-expressing stromal cell lines to levels similar to that found in the fibroblasts that are used to propagate the virus in vitro. To investigate why MCMV replication was suppressed in C57BL/6 mice, we analyzed the location and activation of Ly49H⁺ NK cells. There was selective up-regulation of perforin and IFN-γ in Ly49H⁺ NK cells following infection. Furthermore, a significant fraction of activated Ly49H⁺ NK cells moved from the red pulp (their location in noninfected mice) to the T cell areas, where they were found specifically associated with podoplanin-expressing fibroreticular cells. These studies suggest a mechanism for the acquired T cell immunodeficiency found in humans that fail to control CMV infection, and demonstrate how NK cells protect T zone stroma.

All experiments were performed in accordance with United Kingdom laws and with the approval of the University of Birmingham ethics committee. All strains of mice were bred in our animal facility.

For infections, we used wild-type MCMV K181 that was grown in BALB/c mice and isolated from salivary glands. Each mouse was injected i.p. with 50 μl of 4 × 10⁵ PFU MCMV. For in vitro infections, the same stock of virus was applied for 4 × 10⁵ cells at multiplicities of infection (MOIs) 1.6, 3.2, 1.6 × 10⁻², and 3.2 × 10⁻⁴, and cells were harvested 6, 5, 10, or 9 days later, respectively. Otherwise, 4 × 10⁵ cells were infected at MOI 3.3 and harvested 2, 4, or 6 days later.

The procedure for preparing and staining mouse spleen tissue for light and confocal microscopy has been described before (15, 16). Ly49H⁺ cells were detected with a FITC (fluorescein)-conjugated mouse anti-Ly49H mAb (7). Pixel analysis and confocal image acquisition were performed, as previously described (16).

NK cells were depleted using a single i.p. injection of 50 μl of anti-asialo GM1 (Wako Chemicals) from a 1-ml stock. The efficiency of the depletion was checked by NK cell staining in the blood before infecting the mice (always 1 day after anti-asialo GM1 injection).

Total genomic DNA from frozen 15-μm-thick spleen sections, or laser-microdissected tissue or cells was prepared using the DNeasy blood and tissue kit (Qiagen), according to the manufacturer’s instructions. cDNA was prepared from cells using the μMACS One-step cDNA kit (Miltenyi Biotec), according to the manufacturer’s instructions, or from 8-μm-thick spleen sections using the RNeasy micro kit (Qiagen), according to the manufacturer’s instructions.

Quantitative real-time PCR (TaqMan) reactions were performed, as described previously (16). The target MCMV gene was glycoprotein L (5′→3′): forward, CCC-CGC-CGT-GTT-TTC-A; reverse, GCC-ATC-ACG-GTC-TCT-TTC-GT; probe, 6FAM-ACC-CGA-CAA-CAC-TAT-CGC-CCT-CAA-TAT-CA-BHQ1. The 18S rRNA was used as the control housekeeping gene in a duplex reaction (5′→3′): forward, GCC-GCT-AGA-GGT-GAA-ATT-CTT-G; reverse, CAT-TCT-TGG-CAA-ATG-CTT-TCG; probe, VIC-CCG-GCG-CAA-GAC-GGA-CCA-GA-TAMRA. The following genes were also detected with the same method (5′→3′): CCL21, forward, TCC-CGG-CAA-TCC-TGT-TCT-C; reverse, TTC-TGC-ACC-CAG-CCT-TCC-T; probe, 6-FAM-CCC-CGG-AAG-CAC-TCT-AAG-CCT-GAG-CTA-T-BHQ-1; CCL19, forward, CCT-TCC-GCT-ACC-TTC-TTA-ATG-AAG; reverse, ACA-GAG-CTG-ATA-GCC-CCT-TAG-TGT; probe, 6-FAM-TGC-AGG-GTG-CCT-GC-TAMRA; and CXCL13, forward, ACA-TCA-TAG-ATC-GGA-TTC-AAG-TTA-CG; reverse, TCT-TGG-TCC-AGA-TCA-CAA-CTT-CAG; probe, 6-FAM-CCT-GGG-AAT-GGC-TGC-CCC-AAA-TAMRA. To detect CCL19, lymphotoxin-β receptor (LTβR), and IL-7 in cultured white pulp stromal cells, we used the following primers (5′→3′): CCL19, forward, GGG-GTG-CTA-ATG-ATG-CGG-AA; reverse, CCT-TAG-TGT-GGT-GAA-CAC-AAC-A; LTβR, forward, GAG-CAG-AAC-CGG-ACA-CTA-GC; reverse, GAA-GGT-AGG-GAT-GAG-CAC-C; and IL-7, forward, TTC-CTC-CAC-TGA-TCC-TTG-TTC-T; reverse, AGC-AGC-TTC-CTT-TGT-ATC-ATC-AC. The control gene was β-actin: forward, ATC-TAC-GAG-GGC-TAT-GCT-CTC-C; reverse, CTT-TGA-TGT-CAC-GCA-CGA-TTT-CC. These reactions were performed using SYBR Green, as previously described (17). CXCL10 and CXCL11 were used as ready-made TaqMan gene expression assays (CXCL10, Mm00445235_m1; CXCL11, Mm00444662_m1).

Spleen tissue was cut at 5-μm-thick sections in PALM membrane slides (Zeiss) and then stained with a 1% w/v cresyl violet solution made from cresyl violet acetate (Sigma-Aldrich) dissolved in molecular grade ethanol (Sigma-Aldrich). Slides were dipped sequentially into 100, 70, and 50% solutions of ethanol and stained immediately with cresyl violet for 6 min, and then dipped in the ethanol solutions in the reverse order. Laser microdissection was performed using a Microbeam HT microscope (PALM Microlaser Technologies). Dissected areas were used to obtain genomic DNA.

Splenocytes were prepared, as previously described (16), and stained with anti-CD3ε PE (eBioscience), and either anti-DX5 FITC (eBioscience) or anti-NK1.1 allophycocyanin (eBioscience). CCR7 was detected with anti-CCR7 PE (BD Biosciences). CXCR3 was detected with anti-CXCR3 (rabbit; Invitrogen), followed by anti-rabbit FITC (Southern Biotechnology Associates). White pulp stromal cells were stained for podoplanin with hamster anti-podoplanin (clone 8.1.1), followed by anti-hamster FITC (Southern Biotechnology Associates) or isotype FITC (Southern Biotechnology Associates), for VCAM-1 with anti-VCAM-1 PE (Southern Biotechnology Associates) or isotype PE (Southern Biotechnology Associates), and for CD248 with rabbit anti-CD248 (gift from C. Isacke, Breakthrough Breast Cancer Research Centre, London, U.K.), followed by anti-rabbit PE (Southern Biotechnology Associates) or isotype PE (Southern Biotechnology Associates). For intracellular staining, spleen cells were cultured for 4 h in the presence of GolgiStop (BD Biosciences), and they were stained for intracellular IFN-γ PE (BD Biosciences) or perforin PE (eBioscience) using the Cytofix/Cytoperm kit (BD Biosciences), according to the manufacturer’s instructions. Samples were acquired using a BD FACScan flow cytometer, and the results were analyzed using FlowJo software.

To generate white pulp stromal cell lines, spleen cells from 7- to 10-day-old C57BL/6 mice were cultured for 2–3 days and then nonadherent cells were washed out. Adherent cells were cultured for an additional 5–6 mo, at which time they were positively MoFlo sorted for CD45⁻ podoplanin⁺ cells. Podoplanin⁺ stromal cells were kept in culture for an additional 5 mo before used for in vitro MCMV infection experiments.

All statistical analyses were performed with the nonparametrical Mann-Whitney U test using StatView 5.0 (p < 0.05 is considered significant).

CMV infection is associated with secondary T cell immunodeficiency (18). To investigate the role of NK cells in protection from MCMV, we compared day 0 and day 4 spleens from resistant Ly49H⁺ C57BL/6 or susceptible BALB/c mice (Fig. 1,A). Before infection, B and T cell white pulp areas were comparable in both strains (Fig. 1,A). However, 4 days postinfection, there were marked differences between the two strains of mice with regard to preservation of white pulp areas. In C57BL/6 mice, there was retention of CD11c⁺ dendritic cells (DCs) and CD3⁺ T cells (Fig. 1,A). In contrast, there was marked destruction of splenic white pulp areas in BALB/c mice (Fig. 1,A), with loss of both CD11c⁺ DCs (Fig. 1,A), and a ∼10-fold reduction in the number of T (Fig. 1, A–C) and NK (Fig. 1, B and D) cells. At this time point, there was preferential activation of Ly49H⁺ NK cells in C57BL/6 mice, shown by their expression of IFN-γ and perforin (Fig. 1 E). These data showed that absence of a specific NK response was linked to near complete loss of T cells and DCs, and disintegration of the structure of the splenic white pulp areas.

The splenic immunopathology observed was not dependent on nonspecific NK cell activation, because infection of NK-depleted BALB/c and C57BL/6 mice resulted in the same immunopathology (Fig. 2, A and B). Moreover, MCMV-infected BALB/c nude mice (few T cells) exhibited similar immunopathology (Fig. 2 C), suggesting that destruction of the white pulp is not T cell mediated. Taken together, the above data support the importance of specific NK activation for protection of white pulp areas.

Using quantitative PCR primers for glycoprotein L, a late expressed MCMV gene (19), we evaluated viral titers in whole spleen sections (Fig. 3,A). Viral titers were ∼10-fold greater on day 2 and ∼100-fold greater on day 4 in BALB/c compared with C57BL/6 mice (Fig. 3,A). In C57BL/6 mice, viral titers were reduced at day 4 compared with day 2, indicating NK-mediated protection (Fig. 3,A). In contrast, in BALB/c mice, viral titers were increased at day 4 compared with day 2 (Fig. 3,A). Depletion of NK cells in C57BL/6 mice resulted in comparable viral titers to BALB/c mice (Fig. 3 A).

The main site of MCMV viral replication has been identified as the splenic red pulp (20). To investigate whether the destruction of splenic white pulp areas in susceptible BALB/c mice was secondary to MCMV infection, we performed laser capture microdissection of splenic white and red pulp areas, and evaluated MCMV viral load. Cresyl violet-stained spleen sections identified white pulp areas as darker violet compared with red pulp (Fig. 3,B). At day 2 postinfection, MCMV could be detected in both the splenic red and white pulp areas of both C57BL/6 and BALB/c mice (Fig. 3, C and D). Red pulp titers at day 2 were significantly higher in BALB/c mice (Fig. 3,D). At this time point, MCMV was also readily detected in the white pulp areas: there was no significant difference in MCMV titers between C57BL/6 and BALB/c mice (Fig. 3,C). By day 4, the situation had changed substantially: red pulp titers were falling in C57BL/6 mice, whereas viral titers remained unchanged in BALB/c mice (Fig. 3,D). In the white pulp, there was a significant reduction in viral titers associated with preservation of white pulp areas in C57BL/6 mice, whereas titers had continued to rise in BALB/c mice (Fig. 3,C). NK-depleted C57BL/6 mice showed similar red and white pulp viral titers to BALB/c mice (Fig. 3, C and D), consistent with NK-mediated viral control in both locations. This series of experiments indicated that the immunopathology associated with white pulp areas could be attributable to MCMV replication at this site.

Both red and white pulp stroma express VCAM-1, although expression levels are much higher in the red pulp, where there is a high density of stromal cells (Fig. 4). In the white pulp, VCAM-1 expression is restricted to fibroreticular stromal cells that ensheath the splenic conduit (21, 22), which guides lymphocyte movement (23). Splenic T zone stroma also expresses the transmembrane mucin-like glycoprotein, podoplanin, which is strongly up-regulated following MCMV infection in both C57BL/6 and BALB/c mice (Fig. 4, A–D, and Benedict et al. (20)). Although the podoplanin network is increased in BALB/c mice, very few live cells could be identified at day 4 postinfection using a 4′,6′-diamidino-2-phenylindole (DAPI) nuclear counterstain, compared with uninfected and day 4 C57BL/6 mice (Fig. 4,E). Furthermore, whereas VCAM-1 expression was preserved in C57BL/6 red and white pulp stroma, it was greatly reduced in BALB/c mice in both red and white pulp areas (Fig. 4, B and D), consistent with infection and destruction of these cells by MCMV. Accordingly, the same stromal cell phenotype was exhibited by C57BL/6 and BALB/c mice that were depleted of NK cells and by T cell-deficient nude BALB/c mice (data not shown) in agreement with white pulp disintegration in these mice (see Fig. 2) and viral-, but not immune-mediated destruction.

There is evidence that viral infection can cause down-regulation of homeostatic chemokines, particularly CCL21 (20), and this can be secondary to immune activation through IFN-γ (24). We found that the destruction of the white pulp in BALB/c mice was associated with much lower levels of CCL21 and CCL19 compared with C57BL/6 (Fig. 5, A and B), whereas the difference in CXCL13 expression was significant, but not as marked (Fig. 5 C). This grossly reduced chemokine expression in BALB/c mice is consistent with destruction of the white pulp chemokine-expressing stroma.

To test directly whether white pulp podoplanin-expressing stromal cells could be infected with MCMV, we compared viral replication in mouse embryonic fibroblasts (MEFs), used to propagate MCMV in vitro, and podoplanin-expressing stromal cells derived from day 7–10 murine spleen (see Materials and Methods). The podoplanin-expressing stromal line was very similar to freshly isolated podoplanin-expressing stroma (Fig. 5, D and E) (25). Besides podoplanin, this cell line expressed VCAM-1, but lacked CD248 (Fig. 5,D), which specifically marks red pulp stroma (26). In addition, at the mRNA level, those cells expressed CCL19, IL-7, and LtβR (Fig. 5,E), all of which are associated with white pulp stroma (25). Following infection in vitro with MCMV, the virus replicated to comparable levels in both MEFs and the podoplanin-expressing cell line (Fig. 5 F), consistent with the idea that MCMV infected podoplanin-expressing stroma in vivo.

The implication of our data is that the specific resistance to MCMV provided by Ly49H⁺ NK cells acts not only by controlling viral replication (7), but by preserving the structure of the splenic white pulp as well. To try to understand how might Ly49H⁺ NK cells protect the white pulp, we investigated their response and localization properties following MCMV infection.

We found that this subpopulation was preferentially localized in the splenic red pulp in MCMV naive mice (Fig. 6, day 0). However, following infection in C57BL/6 mice, Ly49H⁺ NK cells relocated to the white pulp areas of the spleen (Fig. 6). A few Ly49H⁺ cells were found in the white pulp areas by day 2, and they were clustered in bridging channels of the marginal zone (Fig. 6), where CD11c⁺ DCs are located normally, and where they enter the white pulp areas after immunization. By day 4, significant numbers of Ly49H⁺ cells were found in the T zone areas, where they were engaged with podoplanin-expressing stroma cells (Fig. 6, bottom high magnification micrograph), illustrating how Ly49H⁺ NK cells might inhibit viral replication in T zone stroma. Consistent with the requirement of both IFN-γ and perforin expression in NK cells for protection from destruction (13), Ly49H⁺ NK cells selectively up-regulated expression of these molecules postactivation (Fig. 1 E).

Unlike T cells, murine NK cells do not express CCR7 (Fig. 7,A), and therefore cannot respond to CCL19 and CCL21, which are produced by T zone stroma. It has been shown before that CXCR3 can mediate NK cell migration to lymph nodes (27) and that CXCR3^−/− mice have elevated MCMV titers in spleen and liver (28). We found that NK cells express CXCR3 before and after MCMV infection (Fig. 7,A). Importantly, the CXCR3 ligands, CXCL10 and CXCL11, were up-regulated following infection in total spleen tissue (Fig. 7,B), white pulp (Fig. 7,C), and red pulp (Fig. 7,D), as well as podoplanin-expressing stromal cells (Fig. 7, E and F). This suggests a mechanism whereby, upon infection, NK cells can be recruited first to the spleen and second to distinct areas within the spleen, including the white pulp and T zone.

In this study, we report that MCMV infection in Ly49H-deficient BALB/c mice is associated with splenic white pulp destruction and subsequent loss of CD3⁺ T and CD11c⁺ DCs. Recent studies have indicated that viral infection nonspecifically leads to down-regulation of homeostatic chemokines (mainly CCL21, but also CXCL13) via a mechanism dependent on IFN-γ production from CD4 T cells (24). We found that down-regulation of CCL21 was exacerbated in the absence of a specific NK response, which could be explained if viral replication occurred in the CCL21-expressing T zone fibroreticular cells, as reported for lymphocytic choriomeningitis virus (14). It has already been shown that MCMV replicates in red pulp areas of the spleen in Ly49H⁺ C57BL/6 mice (20), and consistent with a cytopathic effect, we found extensive destruction of VCAM-1-expressing red pulp stromal cells in the absence of Ly49H⁺ NK cells. We also noted, however, that there was loss of VCAM-1 expression from the podoplanin-expressing white pulp stroma, consistent with viral infection of these cells in BALB/c, but not C57BL/6 mice. Using laser capture microscopy of red and white pulp areas, we demonstrate viral replication in white pulp areas. At day 2 postinfection, comparable levels of MCMV virus were found in both BALB/c and C57BL/6 mice, but by day 4 white pulp titers had fallen in C57BL/6 mice, but had risen substantially (∼10-fold) in BALB/c-infected mice.

The above data taken together were consistent with red and white pulp stromal cells being targets for MCMV replication. Although viral titers were higher in red pulp than white pulp areas, this is probably due to the difference in stromal cell density. The expression of the stromal marker, VCAM-1, is far more intense in the red pulp, because white pulp VCAM-1 expression is restricted to the B and T cell conduit. A striking observation was the preservation of podoplanin expression in the T zone, but staining of live nuclei with DAPI demonstrated profound loss of viable cells in susceptible mice.

Direct evidence that MCMV could infect white pulp stroma was provided by our observation that a podoplanin⁺ stromal cell line derived from day 7 neonatal spleen, which expresses CCL19, IL-7, and the LTβR, supported the replication of MCMV to levels comparable to those attained in mouse embryonic fibroblasts, normally used to propagate the virus in vitro. These data taken together imply that MCMV infects and destroys podoplanin-expressing white pulp stroma with consequent abrogation of expression of VCAM-1 and CCL21.

NK cells are normally found in the red pulp of spleen. To understand how white pulp areas were protected in C57BL/6 mice, we analyzed Ly49H⁺ NK cells before and after infection. As expected, preinfection Ly49H⁺ cells were located in the red pulp. By day 4 postinfection, Ly49H⁺ NK cells were preferentially expanded and expressed both perforin and IFN-γ, both linked with protection from MCMV (12, 13). Some Ly49H⁺ cells were clustered in marginal zone bridging channels by day 2, and on days 2 and 4, Ly49H⁺ cells were also found in white pulp areas, associated with podoplanin-expressing stromal cells.

The chemokine receptor CXCR3 has been implicated in the migration of NK cells in secondary lymphoid tissue (27). Accordingly, we found that NK cells express CXCR3 and that MCMV infection induces expression of its ligands, CXCL10 and CXCL11. The expression of CXCR3 by NK cells and the up-regulation of CXCL10 and CXCL11 by white pulp and podoplanin-expressing stromal cells suggest that migration into the white pulp is likely to be CXCR3 driven. Furthermore, overall up-regulation of CXCL10 and CXCL11 by splenic tissue indicates a mechanism for NK cell recruitment and thus protection from MCMV, consistent with previous findings that CXCR3^−/− mice have reduced ability to control MCMV (28).

Our data taken together suggest a simple mechanism whereby Ly49H⁺ NK cells activated in the splenic red pulp respond to CXCR3 signals and migrate to and protect the white pulp podoplanin-expressing stromal cells through expression of perforin and IFN-γ, allowing protective adaptive T cell-dependent immune responses to develop. They also offer an insight into why CMV-infected patients that fail to control this infection are then susceptible to infections, in which immunity is dependent on T cell help.

The authors have no financial conflict of interest.