Splenic Red Pulp Macrophages Cross-Prime Early Effector CTL That Provide Rapid Defense against Viral Infections

Cross-presentation allows APCs, in particular dendritic cells (DCs), to present extracellular Ags to CTL (1–3), important in the defense against malignant cells or infections with intracellular pathogens. The additional presence of inflammatory signals and/or T cell help allows cross-priming (i.e., the immunogenic activation of CTL by cross-presentation) (4, 5). Several CTL differentiation stages have been discriminated based on expression of killer-cell lectin-like receptor G1 (KLRG1) and IL-7R CD127, including CD127⁻ KLRG1⁻ early effector cells (EEC) (whose in vivo role is unclear), CD127⁻ KLRG1⁺ short-lived effector cells (SLEC) (which provide rapid antiviral protection), and CD127⁺ KLRG1⁻ memory precursor effector cells that give rise to effector or central memory CTL (6, 7) (which can be discriminated by the markers CD62L or CX₃CR1) (8).

DCs of the spleen are important for inducing systemic immune responses (9). Splenic DCs express MHC class II (MHC II) molecules and the marker CD11c and can be subdivided into XCR1⁺ CD8⁺ classical DCs type 1 (cDC1), the XCR1⁻ CD8⁻ CD11b⁺ classical DCs type 2 (cDC2), and plasmacytoid DCs. Among these subsets, the cDC1 are considered to be the main cross-presenting APCs (3, 10, 11). Their development, but not their function, depends on the transcription factor basic leucine zipper transcription factor ATF-like 3 (Batf3) (12, 13). The spleen also contains several subtypes of macrophages, including CD169⁺ metallophilic macrophages, SIGNR-1⁺ marginal zone macrophages, and F4/80⁺ red pulp macrophages (RPM) (9, 14). Some macrophage types also express CD11c and have been shown to cross-present in vitro or to support cross-presentation by cDC1 in vivo (15–19). A major in vivo role of cross-presentation by tissue-resident macrophage subsets has not been documented yet.

The cell biology of cross-presentation is still incompletely understood. In principle, endocytosed Ags must reach an organelle where they can be loaded on MHC class I (MHC I) molecules. Endogenous Ags are loaded in the endoplasmic reticulum, and certain endocytosed Ags can reach this organelle to be cross-presented (20). Several other molecular mechanisms of cross-presentation have been described, including phagosomal, endosomal, and cytoplasmic pathways (5, 21–27). We have previously reported that bone marrow–derived DCs (BM-DCs), which are widely used for cell biological studies on DCs, simultaneously use distinct endocytosis mechanisms and organelles for either MHC I–restricted cross-presentation to CTL or for MHC II–restricted Ag presentation to CD4⁺ T helper cells (28, 29). The mannose receptor (MR) endocytosed and routed the model Ag OVA into early endosomes, from where it was exported into the cytosol to be degraded by the proteasome (30). The resulting peptides were reimported into early endosomes by TAP and loaded on MHC I molecules. The endosomes were then routed in a primaquine-sensitive manner to the cell surface, to allow presentation of peptide–MHC I complexes to CTL (29).

The MR targets Ags to early endosomes that degrade Ag less rapidly than acidified late endosomes (28, 31, 32). There are other receptors that do so, among them Clec9a, which mediates uptake and cross-presentation of cell-associated Ag by cDC1 (33). However, cDC1 do not express the MR (15). Some macrophages do so, but cross-priming is considered a hallmark cDC1 function (34, 35). In this study, we reconcile these discrepancies by identifying splenic MR⁺ RPM as a splenic CD11c⁺ APC subset that can rapidly cross-prime CTL. Such cross-priming generates distinct CTL that seem to contribute to immunity early after viral infection.

All reagents were from Sigma-Aldrich unless specified otherwise. All mice had been backcrossed to C57BL/6 at least 10 times, were bred under specific pathogen-free conditions at the central animal facility of the University Clinic of Bonn, and were used at 8–12 wk of age. All animal experiments were approved by a governmental ethics board of the German state of Northrine Westfalia with approval of the Bezirksregierung Köln of the German state of North Rhine-Westphalia and were performed in strict accordance with the recommendations in the Federation of Laboratory Animal Science Associations.

Single-cell suspensions of splenocytes and lymph nodes were generated by digesting the tissue with 50 μg/ml DNAse and 400 U/ml collagenase at 37°C for 20 min and passing it through a metal sieve and a sterile 50-μm nylon mesh afterward. DCs were enriched from single-cell suspensions of murine splenocytes using magnetic cell separation (Miltenyi) according to the manufacturer’s protocol. When applicable, paramagnetic RPM were removed as described (36). If indicated, DCs from the CD11c⁺-enriched fraction were further sorted afterward with a BD FACS Aria III Cell Separator. cDC1 were identified as CD11c^high CD8α⁺ DEC205⁺, cDC2 as CD11c^high CD11b⁺ CD8⁻, and RPM as CD11c^int F4/80^high.

To purify RPM, single-cell suspensions of splenocytes were passed over LS Separation Columns (Miltenyi) without further treatment, as described. Before being removed from the magnet, the columns were washed three times with 2 ml of MACS buffer (Miltenyi). After removal of the magnet, the RPM were eluted from the separation column with 5 ml of MACS buffer.

Cells were stained with the following Abs from eBioscience unless stated otherwise: anti-B220 (clone C363.16A), anti-CD4 (clone GK1.5), anti-CD8α (clone 53-6.7; BD Biosciences), anti-CD11b (clone M1/70), anti-CD11c (clone N418; BioLegend), anti-CD25 (clone PC61.5), anti-CD45.1 (A20; BD Biosciences), anti-CD68 (clone FA-11; AbD Serotec), anti-CD69 (clone H1.2F3), anti-CD86 (clone GL1), anti-CD169 (clone 3D6.112; BioLegend), anti-DEC205 (clone 205yekta), anti–EEA-1 (clone H-300; Santa Cruz Biotechnology), anti-F4/80 (clone CI:A3-1; BioLegend), anti–GR-1 (clone RB6-8C5; BioLegend), anti-Lamp1 (clone 1D4B; BD Biosciences), anti–Lucifer yellow (LY) (polyclonal; Invitrogen), anti-MARCO (clone ED31; AbD Serotec), anti–MHC II (clone M5/114.15.2; BioLegend), anti-MR (clone C068C2; BioLegend), anti-TAP (clone N-19; Santa Cruz Biotechnology), anti-Ly6C (clone HK1.4; BioLegend), anti–Grz-B (clone GB11; Thermo Fisher Scientific), anti-TNF (clone MP-XT22; BioLegend), and anti–IFN-γ (clone XMG1.2; BioLegend). Fc receptors were blocked with 1.5 mg/ml human IgG (Privigen). Dead cells were excluded with Hoechst 33342 Dye and doublets by forward scatter width. A FACSCanto II and a FACSDiva (BD Biosciences) were used for flow cytometry, and data were analyzed with FlowJo Software (TreeStar).

Five-micrometer cryosections from shock-frozen spleens were prepared, fixed with iced acetone, and blocked 1 h with 1% (w/v) BSA in PBS. B cell zones were visualized by PE-conjugated anti-B220 staining (eBioscience), and the marginal zones were visualized with Alexa Fluor 647–conjugated anti-CD169 (BioLegend). Sections were viewed with an Olympus IX71 Microscope.

Cells were pulsed for 15 min with fluorochrome-labeled OVA (10 mg/ml) or LY (0.3 mg/ml) and were chased for another 15 min with medium. Staining experiments were done as described. Nuclei were visualized with the DNA-intercalating dye DAPI (1 μg/ml).

For tracking OT-I T cells in the spleen, 4 × 10⁶ GFP–OT-I T cells were adoptively transferred 1 d before immunization with 200 μg of OVA and 20 μg of CpG. Eight hours after immunization, the spleen was harvested and PLP fixed, and 30-μm cryosections from shock-frozen spleens were prepared. We stained marginal zone, red pulp, and activated OT-I T cells with anti-CD169 (clone 3D6.112 in AF594, 1:200; BioLegend), anti-F4/80 (clone BM8 in BV421, 1:200; BioLegend), anti-CD69 (polyclonal goat anti-mouse CD69, 1:50; R&D Systems), and AF647-donkey anti-goat (1:500; Invitrogen). Cells were analyzed with an LSM720 Confocal Microscope, with ZEN and Image J Software (Zeiss).

Cells were incubated for 20 min with 10 mg/ml labeled OVA. If indicated, pinocytosis was blocked with 500 μM dimethylamiloride (DMA), and MR-mediated endocytosis was blocked with 3 mg/ml mannan. For in vivo uptake experiments, mice were injected with 5 μg/g body weight labeled OVA.

OT-I cells from OT-I RAG 1–deficient mice and OT-II cells were isolated as described and were further purified with CD8⁺ and CD4⁺ T cell Isolation Kits, respectively (Miltenyi). For cross-presentation assays, 1 × 10⁵ DCs were incubated with OVA in the presence or absence of UL49.5-Trf (1 mg/ml), Mannan (3 mg/ml), DMA (500 μM), or the proteasome inhibitor epoxomicin (1 μM). After 2 h, cells were washed, fixed for 3 min with 0.008% (w/v) glutaraldehyde, and incubated together with 1 × 10⁵ OT-I or OT-II cells. The upregulation of the T cell activation markers CD25 and CD69 was determined after 18 h by flow cytometry.

For analysis of presentation of cell-endogenous OVA, 2 × 10⁷ RPM were cultured in 1 ml of hypertonic media (0.5 M sucrose, 10% polyethylene glycol, 10 mM HEPES, and 10 mg/ml OVA in RPMI 1640) at 37°C. Then 13 ml of prewarmed hypotonic media (40% H₂O, 60% RPMI 1640) was added, and cells were incubated for 2 min at 37°C, washed twice with PBS, fixed, and cultured together with OT-I cells as described above.

OT-I T cells, CD8⁺DC, and RPM were isolated and purified as described above. A total of 0.5 × 10⁵ APCs were cocultured with 1 × 10⁵ OT-I T cells for 2 d for analyzing surface marker expression and cytokine production. For measurement of in vitro cytotoxicity 4 d after of coculture, cells were washed with medium and transferred to a new cell culture plate. A total of 4 × 10⁴ target cells were added to every well for 16 h. Target cells were prepared as described below. Splenocytes were loaded for 15 min at 37°C with OVA peptide (SIINFEKL; 1 μg/ml) or left untreated. The unloaded cells were labeled with 1 μm CFSE (CFSE^hi), and the peptide-loaded cells were labeled with 0.1 μm of CFSE (CFSE^lo). The survival of target cells was analyzed by flow cytometry. Specific cytotoxicity was calculated with the following formula: % specific cytotoxicity = 100 − (100 × [CFSE^hi/CFSE^lo] primed/[CFSE^hi/CFSE^lo] control).

For analysis of in vivo cross-presentation, 2 × 10⁶ CFSE-labeled OT-I T cells were transferred into recipient mice as described (37). Eighteen hours later, mice were i.v. injected with 200–500 μl of a 400 μg/ml OVA solution (i.e., 80–200 μg per mouse) as indicated. To prepare cell-associated OVA, bm1 splenocytes mice (2 × 10⁸ cells/ml) were incubated with 10 mg/ml OVA for 10 min at 37°C, UV irradiated with 15 mJ for 5 min, and washed extensively. Forty-eight hours after Ag challenge, T cell proliferation was assessed via flow cytometric analysis. As controls, either 2 × 10⁶ CFSE-labeled OT-II T cells were transferred or mice were challenged with 10 μg/mouse SIINFEKL peptide instead of OVA.

Spleen cells were loaded for 15 min at 37°C with OVA peptide (SIINFEKL; 2 μg/ml) and labeled with 1 μM CFSE (CFSE^hi cells) or remained unloaded with peptide and were labeled with 0.1 μM CFSE (CFSE^lo cells). Both target cell types (10⁷ each) were injected i.v. After 16 h, the survival of target cells in the spleen was analyzed by flow cytometry. Specific cytotoxicity was calculated with the following formula: % specific cytotoxicity = 100 − (100 × [CFSE^hi/CFSE^lo] primed/[CFSE^hi/CFSE^lo] control).

We used E1-deleted and E3-deleted adenoviral vectors expressing fusion proteins of the enhanced GFP, OVA, and click beetle luciferase (38). AdLGO was propagated on human embryonic kidney (HEK 293) cells and purified by cesium chloride density-gradient centrifugation, and bioluminescence was measured with an IVIS 200 System after i.p. injection of luciferin (50 mM) (Caliper LifeSciences), as previously described (39). For acute infection, 1 × 10⁴ or 2 × 10⁵ OT-I T cells were adoptively transferred 1 d prior to i.v. infection with 5 × 10⁶ PFU of AdLGO. For measurement of TNF-α, IFN-γ, and IL-2 production, 1 × 10⁶ splenocytes were restimulated with 1 μg/ml SIINFEKL for 4 h at 37°C. Granzyme B staining was performed without restimulation. For memory experiments, mice were rechallenged with 5 × 10⁶ PFU of AdLGO 4 wk after immunization with 200 μg of OVA/20 μg of CpG. Data were analyzed with Living Image 2.50 Software (Caliper LifeSciences).

For analysis of MR protein levels, 1 × 10⁶ BM-DCs or sorted splenic APCs were lysed for 20 min at 4°C under agitation in lysis buffer (Cell Signaling) with 0.5 mM PMSF and protease inhibitor mixture (Roche). Protein concentrations were measured using a nanodrop spectrophotometer (Thermo Fisher Scientific). Twenty micrograms of protein/sample was separated on 8% polyacrylamide gels at 120 V for 2 h and blotted to a nitrocellulose membrane (Thermo Fisher Scientific) for 2 h at 25 V. Successful protein transfer was assessed with Ponceau staining. Membranes were blocked for 1 h at room temperature in TBST (10 mM Tris, pH 7.5, 100 mM NaCl, and 0.05% Tween-20) + 5% milk powder. Afterwards, membranes were washed with TBST and stained overnight at 4°C against either the MR or actin. Membranes were washed again with TBST and incubated for 1 h at room temperature with HRP-conjugated secondary Ab. Immunoreactive proteins were visualized by chemiluminescence using the ECL Western Blotting Substrate (Pierce) on a Curix 60 Film Processor (Agfa-Gevaert).

Results are expressed as mean ± SEM. Differences between multiple groups were assessed using a one-way ANOVA in combination with a Bonferroni multiple comparison test, and differences between two groups were assessed by using a two-tailed paired Student t test (Prism 4; GraphPad Software). The p values < 0.05 were considered significant. Tests were reported only where data met assumptions of tests. On the basis of preliminary experimental data, a power analysis of 0.8 with p < 0.05 indicates a minimum number of three samples per group, but in some cases, four samples per group were used.

BM-DCs can use the MR to endocytose Ags for in vitro cross-presentation (28, 40). To identify APC subsets that use this receptor for in vivo cross-presentation, we analyzed splenic CD11c⁺ cells by flow cytometry for MR expression. Such expression was detected neither on the cross-presenting CD8⁺ cDC1 nor on the non–cross-presenting CD11b⁺ cDC2 (Fig. 1A, 1B, subsets I and II). However, strong MR expression was seen on an APC subset expressing intermediate levels of CD11c, high F4/80 levels, and neither CD8 nor CD11b (Fig. 1A, 1B, subset III), indicating that these APC do not belong to the classical DC subsets. This subset was still present in Batf3^−/− mice, whereas cDC1 were much reduced (Fig. 1A). CD11c^int F4/80^high APCs from wild-type (WT), but not from MR-deficient, mice captured large amounts of OVA both in vitro (Fig. 1C) and in vivo (Fig. 1D). By contrast, cDC1 and cDC2 endocytosed much less OVA, and they did so independent of MR expression (Fig. 1C, 1D). Thus, only subset III APCs used the MR for OVA uptake.

The CD11c^int F4/80^high APCs displayed the phenotype CD11c^int CD4^– CD8^– CD11b^– CD80⁺ CD86⁺ MHC II⁺ F4/80⁺ Gr1⁻ MARCO⁻ (Supplemental Fig. 1A), identifying them as RPM (14). In support of this, they showed the strong autofluorescence characteristic of RPM (Supplemental Fig. 1B), which has been attributed to hemoglobin degradation products that these cells produce after phagocytosing and degrading aged or damaged erythrocytes (41). Fluorescence microscopy localized MR⁺ cells exclusively in the splenic red pulp in colocalization with F4/80 (Supplemental Fig. 1C).

Mice deficient for the transcription factor SpiC have been shown to selectively lack RPM, whereas DCs and CD169⁺ macrophage numbers were unchanged (14) (Supplemental Fig. 1D). Also, we found that CD11c^int F4/80^high APCs, but neither cDC1 nor cDC2, were reduced by >90% in the spleens of these mice (Fig. 1E) (WT: 1.07% ± 0.14; SpiC^−/−: 0.042 ± 0.006). Consistently, immunofluorescence microscopy showed no F4/80 and much reduced MR staining in the splenic red pulp of SpiC^−/− mice (Supplemental Fig. 1C). In vivo OVA uptake by the few CD11c^int F4/80^high APCs remaining in SpiC^−/− mice was strongly decreased (Fig. 1F). Thus, the splenic CD11c^int F4/80^high APCs that use the MR to endocytose OVA can be classified as SpiC-dependent RPM.

We next asked whether RPM were the APC type responsible for MR-dependent in vivo cross-presentation of soluble OVA (40). We tested this idea by sorting RPM, cDC2, and cDC1 from OVA-injected mice and coculturing them with CFSE-labeled OVA-specific transgenic CD8⁺ T cells (OT-I cells). cDC1 and RPM, but not cDC2, activated OT-I cells ex vivo (Fig. 2A, 2B), indicating that two cross-presenting APC subsets coexist in the spleen. To determine which APC type used the MR, we injected soluble OVA into MR^−/− mice. RPM isolated from these mice failed to stimulate OT-I cell proliferation, whereas cDC1 still did (Fig. 2A, 2B). Conversely, RPM from OVA-injected Batf3^−/− mice stimulated OT-I cells’ effectivity (Fig. 2C, 2D), consistent with previous data showing that RPM numbers are unaltered in these mice (12). We found in Batf3^−/− mice a few remaining cDC1, but these elicited a poor proliferative OT-I cell response (Fig. 2C, 2D).

Based on these findings, we predicted that in vivo cross-presentation of soluble OVA should still occur in Batf3^−/− mice. Indeed, CFSE-labeled OT-I cells did proliferate in these mice, albeit less intensely as in control WT mice (Fig. 2E, 2F). Consistent with previous studies showing that cross-presentation of cell-associated OVA is MR independent, we found that it was independent of RPM as well, as CFSE-labeled OT-I T cells proliferated in response to such Ag comparably well in SpiC^−/− mice and control mice (Fig. 2G). By contrast, OT-I T cells proliferated much less intensively in Batf3^−/− mice (Fig. 2G), indicating that cDC1 were necessary for cross-presentation of cell-associated Ag. Taken together, these experiments demonstrated that Batf3-dependent cDC1 and SpiC-dependent RPM simultaneously cross-presented soluble OVA in vivo, but only the latter used the MR for that purpose.

We next examined the intracellular mechanisms of MR-dependent cross-presentation by RPM. Although the strong autofluorescence of RPM complicated studies by immunofluorescence microscopy, we were able to demonstrate that fluorescently labeled OVA was transported to EEA-1⁺ endosomes, but not to LAMP1⁺ late endosomes, which instead contained the pinocytosis marker LY (Fig. 3A–C). Endocytosed OVA did not colocalize with LY but, instead, with the MR in early endosomes (Fig. 3D–F).

We next inhibited the MR with mannan and pinocytosis with DMA, as previously described, (28, 29) and found that RPM took up OVA through both pathways but took the majority through the MR (Fig. 4A). We next studied the ability to cross-present MR-endocytosed OVA by coculture experiments with OT-I cells. Their in vitro activation was completely inhibitable by mannan, but not by DMA (Fig. 4B), indicating that RPM only used MR-endocytosed OVA for cross-presentation. By contrast, cDC1 still cross-presented OVA after blocking the MR and after inhibiting pinocytosis (Fig. 4B), indicating that they used a different mechanism for cross-presentation that has yet to be identified.

Blocking the cytosolic proteasome by epoxomicin in RPM suppressed activation of cocultured OT-I cells (Fig. 4C, Supplemental Fig. 2A), suggesting that the proteasome produced the antigenic OVA peptides. Its cytosolic location suggested that these peptides might be imported into EEA-1⁺ endosomes by TAP for subsequent loading on MHC I molecules, as previously shown to occur in BM-DC (28, 29). In support of this, OVA-containing early endosomes stained positive for TAP (Fig. 4D). We blocked such endosomal TAP using the bovine herpesvirus 1–derived TAP-inhibitor UL49.5-Trf (42, 43), which we had covalently linked to transferrin, to target it to early endosomes but not to the endoplasmic reticulum (29). Such UL49.5-Trf inhibited cross-presentation of endocytosed OVA, but not of cytoplasmic OVA introduced into the cytoplasm via osmotic shock (Fig. 4E, Supplemental Fig. 2B), indicating that endosomal TAP was necessary for cross-presentation. We also found that cross-presentation by RPM was sensitive to primaquine (Fig. 4F, Supplemental Fig. 2C), previously shown to block routing of early endosomes containing cross-presented OVA to the cell surface (44).

We finally tested whether RPM can present OVA on MHC II to OT-II cells. We detected a minor OT-II response, which was inhibitable neither by mannan nor by UL49.5-Trf but instead by amiloride (Supplemental Fig. 2D–G), indicating that RPM used only pinocytosed OVA for MHC II–restricted presentation, as previously reported to be the case in BM-DCs (29). Taken toundergether, these findings indicated that RPM loaded antigenic peptides on MHC I within early endosomes of RPM on MHC I, spatially separated from the MHC II loading pathway.

To investigate the in vivo relevance of cross-presentation by RPM, we transferred CFSE-labeled OT-I cells into RPM-deficient SpiC^−/− mice and immunized them with OVA/CpG. OT-I cells still proliferated in these mice but less intensely as in SpiC-competent mice (Fig. 5A, Supplemental Fig. 3A). Notably, in SpiC^−/− mice, OT-I completed less cell cycles during the 48-h observation period (Fig. 5A), suggesting that their activation was delayed when RPM were absent.

We wished to directly visualize OT-I activation by RPM. To this end, we transferred GFP-expressing OT-I cells, injected OVA/CpG, and stained spleen sections after 8 h for F4/80 as an RPM marker and for the early activation marker CD69. Indeed CD69⁺ OT-I cells were seen in the red pulp adjacent to RPM, whereas no OT-I cells in the white pulp expressed CD69 at this early time point (Fig. 5B).

We next tested for synergy between cross-presentation by RPM and cDC1 by crossing SpiC^−/− mice with Batf3^−/− mice to clarify whether the absence of both APC types would reduce cross-presentation more than the absence of only one. SpiC/Batf3^−/− mice bred even more poorly than SpiC^−/− mice but otherwise appeared healthy and acted inconspicuously. Their spleens contained 80–90% less cDC1 and RPM, whereas most other immune cell types were present at normal numbers (Supplemental Fig. 3B). During this analysis, we noted that Batf3^−/− mice harbored more RPM (Fig. 4C), and cDC2 were only somewhat more numerous in Batf3^−/− mice (Supplemental Fig. 3D). The latter increase should not affect cross-presentation, whereas the former might imply that the soluble OVA cross-presentation response we had observed (Fig. 1) might somewhat overestimate their function in Batf3-competent mice. Additionally, histological analysis of cryosections demonstrated that the white and red pulp, marginal zone, and B cell follicles appeared microscopically similar to those in WT mice (Supplemental Fig. 3E), arguing against major anatomical alterations in the spleens of SpiC/Batf3^−/− mice.

We then tested the capacity for cross-presentation of OVA in these double knockout mice. Adoptively transferred OT-I cells proliferated less intensely than in both Batf3– and SpiC–single knockout mice (Fig. 5C), confirming that both APC types synergized at cross-presenting soluble Ag. In lymph nodes, which lack RPM, OT-I cell proliferation was similar in SpiC^−/− mice and WT mice and equally reduced in SpiC/Batf3^−/− mice and Batf3^−/− mice (Fig. 5D), verifying that SpiC deficiency affected cross-presentation only in the organ that harbors RPM, the spleen. Adoptively transferred OT-II cells proliferated equally well in WT and all knockout strains in response to OVA, excluding defects in Ag uptake and MHC II–restricted Ag presentation in double knockout mice (Fig. 5E). OT-I cells proliferated normally in response to injected SIINFEKL peptide that does not need to be intracellularly processed because it directly binds to surface MHC I (Fig. 5F), verifying that functional MHC I molecules were expressed by splenic APCs of SpiC/Batf3^−/− mice.

To clarify whether RPM can cross-prime CTL also in a normal T cell repertoire, we vaccinated mice with OVA plus CpG as adjuvant and determined OVA-specific cytotoxicity after 5 d. That was similarly reduced in both Batf3^−/− and SpiC^−/− mice but almost lost in SpiC/Batf3^−/− mice (Fig. 5G). The numbers of OVA-specific CTL were reduced in SpiC^−/− mice and very low in Batf3^−/− mice and SpiC/Batf3^−/− mice (Fig. 5H), indicating that the very few CTL in Batf3^−/− mice, which presumably were induced by RPM, must have possessed a high cytotoxic activity. Thus, SpiC-dependent RPM and Batf3-dependent cDC1 synergistically cross-prime CTL.

The findings above encouraged us to explore whether cross-presentation by RPM can be employed for improving antiviral vaccinations. We used the AdLGO adenovirus that expresses luciferase and OVA, allowing us to track the infection by in vivo luminescence imaging of the liver or by using OVA tools such as OT-I cells. We and others had previously used this viral infection model to compare the efficacy of vaccine adjuvants and to demonstrate that antiviral immunity in the first 2 wk hinges on CTL (39, 45). We immunized mice lacking RPM and/or cDC1 with OVA and CpG, infected them 5 d later with AdLGO, and then quantified the viral burden by in vivo bioluminescence imaging at different time points. Mice harboring both RPM and cDC1 were most effectively protected from viral infection (Fig. 6A, 6B). Our protocol did not induce measurable levels of type I IFN in the serum (data not shown), arguing against a potential role of such IFN I on the antiviral defense, consistent with our previous studies (39, 45). In SpiC^−/− mice lacking RPM (red line), an increase in the initial viral load was noted between days 7 and 10 after vaccination, but viral clearance was achieved after 10 d (Fig. 6A, 6B). By contrast, cDC1 deficiency in Batf3^−/− mice hardly changed the initial viral load, but viral clearance was delayed until after day 14 after vaccination (Fig. 6A, 6B). SpiC/Batf3^−/− mice showed both a higher initial viral load and delayed viral clearance (Fig. 6A, 6B). These findings suggested that RPM and cDC1 are both required for optimally protective vaccination against viral infection and that RPM contributed earlier after vaccination than did cDC1.

We next tested whether cDC1 and/or RPM were required for memory CTL induction by vaccinating mice with OVA/CpG and infecting them after 4 wk with AdLGO. The viral burden seemed somewhat higher in SpiC^−/− mice than in WT mice, but this was statistically insignificant, and both mouse lines had cleared the virus 5 d postinfection (Fig. 6C, 6D), indicating that effective memory CTL had been generated in the absence of RPM. By contrast, Batf3^−/− mice showed a significantly higher initial viral load and required 2 wk to clear the virus (Fig. 6C, 6D), comparable to the kinetics in nonvaccinated mice (Fig. 6A, 6B). The numbers of virus-specific CTL were reduced by 60% in SpiC^−/− mice, but were almost completely lost in Batf3^−/− mice (Fig. 6E), indicating that cDC1 were critical for memory CTL induction and that RPM played a supportive, but not an essential, role.

The findings above suggested that RPM contributed to the early phase of the antiviral CTL response. To understand the underlying mechanisms, we examined in vitro the Ag uptake and the OT-I T cell response elicited by RPM and cDC1, which had been isolated 3 and 18 h after OVA injection into WT mice. RPM endocytosed large OVA amounts and induced strong OT-I proliferation and cytotoxicity when isolated 3, but not 18, h after OVA injection, whereas cDC1 took up much less OVA yet induced robust OT-I cell proliferation and cytotoxicity after 18, but not after 3, h (Fig. 7A–C). This is consistent with the view that macrophages take up many Ags and degrade them swiftly, whereas DCs take up less but can store them for prolonged presentation to T cells (46).

We also examined the phenotype of CTL activated by RPM and cDC1. After 2 d of coculture, neither APC type had induced expression of the established differentiation markers CX₃CR1, CD127, and KLRG1 in activated OT-I cells (Fig. 7D, Supplemental Fig. 4A, 4B). However, almost 60% of RPM-activated OT-I cells, but hardly any of the cDC1-induced CTL, expressed the effector molecule Grz-B, and 10–20% produced TNF-α and IFN-ɣ (Fig. 7E, Supplemental Fig. 4C), consistent with the cytotoxic functionality of these CTL (Fig. 7C). This phenotype was incompatible with the CTL differentiation types SLEC (KLRG1⁺, CD127⁻), T effector memory cells (KLRG1⁻, CD127⁺, CX₃CR1⁺), and T central memory cells (KLRG1⁻, CD127⁺, CX₃CR1⁻) but matched CTL that had previously been termed EEC CTL (KLRG1⁻, CD127⁻, Grz-B⁺) (47). We noted that many of the RPM-activated OT-I cells expressed the markers Ly6A/E and Ly6C (Fig. 7D, Supplemental Fig. 4B), which has not yet been described on EEC (47). In conclusion, RPM induced distinct cytotoxic Ly6AE⁺ Ly6C⁺ effector CTL very early after immunization.

Finally, we wished to clarify whether RPM-induced CTL also contributed to the endogenous immune defense against a CTL-dependent viral infection. To this end, we infected nonimmunized WT and SpiC^−/− mice with AdLGO and tracked the viral burden by bioluminescence imaging. SpiC^−/− mice showed a higher burden early postinfection but cleared the virus after 9 d (Fig. 8A). By contrast, Batf3^−/− mice showed similar viral burden until day 5 postinfection but needed 1 wk longer to clear the adenovirus than WT and SpiC^−/− mice (Fig. 8A), consistent with the observations after antiviral vaccination (Fig. 6).

The increased viral load in SpiC^−/− until day 5 after adenoviral infection might be explained by the loss of protective RPM-induced CTL. To test this interpretation, we injected 2 × 10⁵ or 1 × 10⁴ OT-I cells into mice infected with 5 × 10⁶ AdLGO and examined their phenotype on days 4 and 7 after vaccination. On day 4, we found generally few activated OT-I cells in the spleen (Fig. 8B). Nevertheless, SpiC^−/−, but not Batf3^−/−, mice harbored significantly fewer OT-I cells compared with WT mice (Fig. 8B). At this early time point, nearly all OT-I cells presented the KLRG1⁻ CD127⁻ EEC–like phenotype and hardly any expressed KLRG1 that signifies SLEC (Fig. 8C). Likewise, CX₃CR₁ expression was more or less absent from OT-I cells (data not shown). Notably, in SpiC^−/− mice, much fewer OT-I cells expressed Grz-B compared with WT and Batf3^−/− mice (Fig. 8D), consistent with the poorer viral control in these lines (Fig. 8A). Almost all Grz-B⁺ OT-I cells were KLRG1⁻ CD127⁻, nearly all of them expressed Ly6AE, and about half of them expressed Ly6C (Fig. 8E, 8F), consistent with the OT-I phenotype resulting from in vitro activation by RPM (Fig. 7). Some OT-I cells expressed TNF-α and IFN-γ, and very few expressed IL-2 (Supplemental Fig. 4D, 4E). The proportion of OT-I cells that expressed Ly6A/E and Ly6C was higher in Batf3^−/− mice (Fig. 8E, 8F), likely because the absence of cDC1 prevented the activation of OT-I cells lacking these EEC markers.

On day 7 postinfection, OT-I cell numbers were almost 10-fold lower in Batf3^−/− mice and still∼50% reduced in SpiC^−/− mice (Fig. 8G), indicating that OT-I expansion depended mostly on cDC1 and that RPM supported this function. Likewise, SLEC differentiation on day 7 depended on cDC1, but RPM contributed to a certain degree (Fig. 8H), perhaps because SLEC differentiation from EEC (47) was reduced. Total EEC numbers did not differ (Fig. 8I). We also noted that PD-1 expression was somewhat higher on SLECs and EEC in SpiC^−/− mice compared with WT controls, (Supplemental Fig. 4F, 4G), suggesting that RPM can reduce CTL exhaustion. In summary, our findings indicated that RPM contribute to the early antiviral defense by rapidly cross-priming effector CTL that contain viral spread until cDC1 have achieved SLEC differentiation, which then can clear the viral infection.

Batf3-dependent cDC1 are widely considered the main cross-presenting cell type (3, 5, 10, 11). Our present findings identify RPM as a Batf3-independent APC type that can cross-present soluble Ags like OVA in vivo with comparable potency. For this function, RPM use the cytosolic Ag-processing pathway that operates also in BM-DCs (24), which are widely used for mechanistic analysis of DC functions and Ag presentation mechanism. Thus, RPM can be considered a functional in vivo correlate of BM-DCs with respect to cross-presentation of soluble Ag. Their location in the spleen allows them to quickly acquire circulating soluble Ags (i.e., proteins applied during i.v. vaccination and possibly viral proteins released into the circulation during infection) to cross-prime CTL.

Under inflammatory conditions, cDC2, monocytes, and macrophages are known to contribute to cross-presentation (15, 48–50). Some of these APCs express the MR, and it is possible that they use the cytosolic cross-presentation pathway operating in RPM and BM-DC, but this has yet to be formally shown. Our present findings demonstrate that this pathway operates not only under inflammatory conditions but also in homeostasis. There may be further APC types using this pathway. cDC1 lack the MR and thus use another molecular mechanism that has yet to be identified.

RPM have been historically classified as macrophages because of their phagocytic activity, their F4/80 expression, and their stationary location outside the splenic white pulp but not as DCs, despite their dendritic morphology and their expression of the murine DC marker CD11c. Our present study demonstrated that RPM can cross-prime naive CD8⁺ T cells and thus perform a hallmark DC function. Nevertheless, their kinetics of Ag uptake and cross-presentation matched established concepts of macrophage function (46). The discrimination between DCs and macrophages is still intensely debated, and recent studies support a much greater variety of mononuclear APC subtypes than previously thought (51–54). Our findings demonstrate that RPM possess characteristics of both cell types, which may aid future efforts to revise the mononuclear phagocyte classification system.

RPM may have escaped notice in previous studies for several reasons. First, because of their CD11c expression, as they are present in splenic DC isolates based on CD11c positivity. They may then perform ex vivo functions like cross-presentation that might be falsely attributed to DCs. Magnetic cell sorting is especially prone to RPM contaminations because RPM are intrinsically paramagnetic due to intracellular iron accumulation, a consequence of their physiologic function of degrading aged or damaged erythrocytes. In fact, positive magnetic splenocyte isolates have been reported to be routinely contaminated with RPM, unless special removal steps are performed (36). Second, the CD11c expression of RPM renders them depletable in CD11c-DTR mice (55). Thus, their functions are lost and might be falsely attributed to CD11c⁺ DCs. Third, they lack expression of the markers CD11b, CD8, and CD103 that are used to identify cDC1 and cDC2, respectively. Thus, their ability to cross-present will escape detection when only sorted cDC1 and cDC2 are analyzed (15). Finally, RPM do not survive adoptive transfer into recipient mice, limiting the experimental options to dissect their functions. These characteristics of RPM may help reconciling discrepancies on the role of APC subtypes and the MR in cross-presentation (15, 56).

In SpiC/Batf3^−/− mice lacking both cDC1 and RPM, in vivo cross-priming was much reduced but not completely abrogated. This may be due to contributions of further cross-presenting APC types, such as liver sinusoid endothelial cells, which express the MR and can cross-present in vivo (57), or CD169⁺ macrophages (17). However, splenic CD169⁺ marginal zone macrophages lack the MR and activated CTL either by transferring Ag to cDC1 or when infected themselves but not by cross-presentation (18, 19). We observed direct adenoviral infection of CD169⁺ marginal zone macrophages (data not shown), and this may result in CTL priming using the endogenous pathway, contributing to the residual CTL response we detected in SpiC/Batf3^−/− mice. However, this residual response was small compared with CTL cross-priming by cDC1 and RPM in our system. The presence of two cross-presenting APC types using different molecular mechanisms may impede viral immune escape strategies, but situations in which this is important have yet to be reported.

Many parameters may theoretically influence the relative contributions of cDC1 and RPM to cross-priming, for example, whether they express endocytosis receptors that can endocytose a given Ag (28, 58). In the current study, we found that these two APC types cooperated by cross-presenting at different phases after Ag exposure. Consistent with the rapid Ag-processing capacities of macrophages (46), cross-presentation by RPM was dominant very early after immunization or infection. It induced Grz-B⁺ CTL lacking the canonical differentiation markers KLRG1, CD127, and CX₃CR1 (6–8). Such CTL have been previously described in several bacterial and viral infections and examined regarding their differentiation potential into SLEC or memory CTL (47, 59). Our findings extend these studies by clarifying how such early effector CTL are activated and showing an in vivo functional role in the antiviral defense. The expression of the Ly6AE and Ly6C markers may be specific to the induction by RPM and may help tracking such cells in other mouse models of infection.

We found that RPM-induced EEC contained the viral spread early during the infection but could not eliminate the infection. Cross-presentation by cDC1 and ensuing SLEC and memory CTL differentiation required more time than EEC induction, suggesting that EEC serve as a rapid reaction force intended to limit collateral viral damage until more potent CTL subsets are available. EEC have been reported to give rise to both SLECs and memory CTL (47, 59), and RPM-induced CTL may do so, as well.

Another previous study showed that cross-priming of CTL against vaccinia virus is preceded by CTL priming by infected DCs, which preactivated the CTL and induced the expression of chemokine receptors that guided them to cross-priming cDC1 (60). Extrapolated to the situation in the current study, this may imply that RPM activation enabled CTL to enter the white pulp and encounter cDC1 for full activation and differentiation. The delayed proliferation of CFSE-labeled OT-I cells that we noted in SpiC^−/− mice supports the idea that cDC1 cross-priming took longer when RPM were missing. Formal evidence for this scenario would require extensive in vivo microscopy studies. CTL memory generation absolutely required cDC1, although the presence of RPM improved the memory recall response, supporting the possibility that RPM-induced EEC may differentiate into memory CTL.

In conclusion, we have identified RPM as a spleen-resident APC subset that perceptibly contributes to CTL cross-priming against soluble Ag. RPM are specialized at cross-priming very early after immunization or infection and generate Ly6AE⁺ Ly6C⁺ Grz-B⁺ EEC that can reduce the viral spread. Thereby, they may reduce tissue damage until cDC1 have induced SLEC that can eliminate the virus. Our findings suggest that RPM may be targeted, for example, through the MR via the i.v. route (61), to improve vaccination strategies aimed at inducing CTL against viral infections or tumors.

We acknowledge support by the Central Animal and Flow Cytometry Core Facilities of the Medical Faculty Bonn. We thank Jessica Gonyer for technical support.

The authors have no financial conflicts of interest.