Altered effector functions of virus-specific and -cross-reactive CD8⁺ T cells in mice immunized with related flaviviruses

Summary

Memory cross-reactive CD8⁺ T cell responses may induce protection or immunopathology upon secondary viral challenge. To elucidate the potential role of T cells in sequential flavivirus infection, we characterized cross-reactive CD4⁺ and CD8⁺ T cell responses between attenuated and pathogenic Japanese encephalitis virus (JEV) and pathogenic West Nile virus (WNV). A previously reported WNV NS4b CD8⁺ T cell epitope and its JEV variant elicited CD8⁺ T cell responses in both JEV- and WNV-infected mice. The peptide variant homologous to the immunizing virus induced greater cytokine secretion and activated higher frequencies of epitope-specific CD8⁺ T cells. However, there was a virus-dependent, peptide variant-independent pattern of cytokine secretion; the IFNγ⁺-to-IFNγ⁺ TNFα⁺ CD8⁺ T cell ratio was greater in JEV- than in WNV-infected mice. Despite similarities in viral burden for pathogenic WNV and JEV viruses, CD8⁺ T cells from pathogenic JEV-immunized mice exhibited functional and phenotypic profiles similar to those seen for the attenuated JEV strain. Patterns of KLRG1 and CD127 expression differed by virus type, with a rapid expansion and contraction of short-lived effector cells (SLECS) in JEV infection and persistence of high levels of SLECs in WNV infection. Such cross-reactive T cell responses to primary infection may affect the outcomes of sequential flavivirus infections.

Introduction

The arthropod-borne Flaviviruses co-circulate in different geographic regions worldwide and include important human pathogens. The Japanese encephalitis serogroup includes Japanese encephalitis virus (JEV), the leading cause of viral encephalitis among children in Southeast Asia, and West Nile virus (WNV), which causes neuroinvasive disease in adults in temperate regions [1]. A live-attenuated JEV vaccine, SA14-14-2, has been licensed in China, but currently, there is no licensed WNV vaccine for humans [2]. The flavivirus genome encodes 3 structural (C, prM, E) and 7 nonstructural genes (NS1, NS2a, NS2b, NS3, NS4a, NS5). Both the humoral and cellular arms of the immune system are vital to protect mice from JEV and WNV encephalitis [3–6]. Protective CD8⁺ and CD4⁺ T cell epitopes residing in the WNV NS4b and NS3 proteins, respectively, play an important antiviral role through cytokine production and cytotoxic activity [7–9].

Heterologous immunity to related or unrelated viral pathogens induces protection or immunopathology upon a secondary viral challenge due to cross-reactive memory CD8⁺ T cell responses [10, 11]. Immunization with live or inactivated JEV vaccine protects against lethal WNV challenge in animals, whereas WNV immunization only reduces disease severity against JEV challenge, suggesting that the sequence of infection impacts disease outcome [12–14]. Cross-reactive memory CD4⁺ T cells affect CD8⁺ T cell responses to secondary dengue infections in mice [15]. Therefore, JEV/WNV cross-reactive CD4⁺ T cell epitopes may also play an important role in heterologous protection of JEV-immunized rodents from WNV infection [12].

We investigated JEV-WNV cross-reactive CD4⁺ and CD8⁺ T cell responses following primary JEV and WNV infection as a first step in elucidating the role these cells may play in heterologous immunity. We characterized effector functions elicited by a previously identified immunodominant WNV NS4b CD8⁺ T cell epitope and its JEV variant in both JEV- and WNV-infected mice and found that the homologous peptide variant to the immunizing virus induced higher levels of cytotoxic activity and cytokine responses. However, there were striking virus-dependent differences in the quality of the response; the ratio of IFN–γ⁺ CD8+ T cells to IFN-γ⁺ TNF-α⁺ CD8⁺ T cells was greater in JEV-infected mice compared to WNV-immunized mice. To further understand these differences, we compared epitope-specific CD8+ T cell responses (cytokine profile, epitope hierarchy, phenotype) as well as the effect of virus burden in mice immunized with a low or high dose of pathogenic JEV and compared these responses to those seen in attenuated JEV and pathogenic WNV infection.

Results

Identification of Japanese encephalitis-West Nile virus cross-reactive CD4⁺ and CD8⁺ T cell epitopes

To identify cross-reactive CD4⁺ and CD8⁺ T cell epitopes, we stimulated splenocytes harvested on day 7 from JEV SA14-14-2 immunized mice with peptide pools corresponding to each of the 10 WNV proteins. We found that the JEV-WNV cross-reactive CD4⁺ T cell IFN-γ responses, as assessed by intracellular cytokine staining, were mainly directed at peptides in the NS4b, NS2a, NS3 and E proteins (Supplementary Figure 1A and Supplementary Table 1). In contrast, the majority of the JEV-WNV cross-reactive IFN-γ-producing CD8⁺ T cells was induced by a single peptide pool corresponding to the WNV NS4b protein.

Deconvolution of the positive peptide pools identified three peptides, WNV NS1 A, WNV NS3 B and WNV NS4b_209-226, that consistently induced the highest responses in splenocytes from JEV-immunized mice (Table 1). WNV NS3B and WNV NS4b_209-226 have previously been identified as epitopes in WNV-infected C57Bl/6 mice [8, 9, 16]. WNV NS1 A and WNV NS3 B and their corresponding truncations (NS1 A-1 and NS3 B-2) induced IFN-γ production by splenocytes from both H2-D^b−/− and H2-K^b−/− mice suggesting that these might be CD4⁺ T cell epitopes. We confirmed that NS1 A-1 and NS3 B-2 are JEV-specific CD4⁺ T cell epitopes that are cross-reactive to WNV by intracellular cytokine staining (Figure 1A, Table 1).

Table 1. IFN-γ production induced by peptide stimulation of JEV-immune splenocytes from C57Bl/6, H2-D^b−/− and H2-K^b−/− mice in ELISPOT assay.

Mice were immunized with JEV SA14-14-2 and splenocytes prepared on day 7. Values represent spots per million splenocytes.

			Mouse strain
Peptide name	Amino Acid #	Sequence	C57Bl/6	Db−/−	Kb−/−
WNVa NS1 A	132-149b	TFVVDGPETKECPTQNRA	197c	136	347
WNV NS1 A-1	132-145	TFVVDGPETKECPT	174	84	309
JEV NS1 A-1d	132-145	TFVVDGPETKECPD	N.T.e	N.T.	N.T.
WNV NS1 A-2	134-147	VVDGPETKECPTQN	0	4	35
WNV NS1 A-3	136-149	DGPETKECPTQNRA	7	5	0
WNV NS3 B	560-574	DRRWCFDGPRTNTIL	70	139	201
WNV NS3 B-1	560-571	DRRWCFDGPRTN	52	35	71
WNV NS3 B-2	563-574	WCFDGPRTNTIL	93	153	408
JEV NS3 B-2	563-574	WCFDGPRTNAIL	N.T.	N.T.	N.T.
WNV NS3 B-3	565-575	FDGPRTNTIL	3	4	0
WNV NS4b	209-226	LWENGASSVWNATTAIGL	N.T.	N.T.	N.T.
WNV NS4b A	209-221	LWENGASSVWNAT	14	0	0
WNV NS4b B	211-224	ENGASSVWNATTAI	1835	29	1704
WNV NS4b C	213-226	GASSVWNATTAIGL	415	7	2028
WNV NS4b G9f	213-221	GASSVWNAT	498	5	58
WNV NS4b A9	214-222	ASSVWNATT	2264	0	1934
WNV NS4b A10	214-223	ASSVWNATTA	2718	9	2020
WNV NS4b S9	215-223	SSVWNATTA	1403	43	1734
WNV NS4b S10	215-224	SSVWNATTAI	2063	0	1482
JEV NS4b C	213-226	GASAVWNSTTATGL	N.T.	N.T.	N.T.
JEV NS4b A9	214-222	ASAVWNSTT	N.T.	1.2	2390
JEV NS4b S9	215-223	SAVWNSTTA	N.T.	5	2412

^aWNV 3356 GenBank Accession number AF404756;

^bNumbers represent amino acid position within each protein;

^cIFN–γ spot forming cells/million splenocytes;

^dJEV GenBank Accession number AF315119 (JEV SA14-14-2) and L48961 (JEV Beijing);

^eNot tested;

^fRepresents amino acid in position 1 followed by peptide length.

Figure 1. Identification of cross-reactive CD4⁺ and CD8⁺ T cell epitopes in JEV SA14-14-2 immunized mice.

(A) Identification of CD4⁺ T cell epitopes. C57Bl/6J mice were immunized i.p. with 1×10⁶ pfu of JEV SA14-14-2. Splenocytes were cultured in the presence of peptides (1μg/ml) or media alone for 5 hours in the presence of Brefeldin A. Cells were gated on CD3⁺ CD4⁺ T cells (upper panels) or CD3⁺CD8⁺ (lower panels) with IFN-γ displayed on the y-axis. (B) In vivo cytotoxic activity was measured in a naïve mouse (left panels) or 8 days after JEV-immunization (right panels). DDAO and CFSE-labeled target cells coated with 10⁰–10⁻² μg/ml of JEV S9 (top panels) or WNV S9 (bottom panels) were injected i.v. and spleens were harvested 2 hours later. Target cells pulsed with 1μg/ml of influenza NP 366–374 peptide (flu) serve as a negative control. CFSE is shown on x-axis. Percent specific lysis values are indicated on top of each peak and were calculated as described in Materials and Methods. (C) Summary of in vivo lysis of peptide-pulsed target cells in JEV-immunized mice. JEV S9-pulsed target cells (●); WNV S9-pulsed target cells (○). Horizontal bars represent median specific lysis. Flu peptide and unpulsed target cells not shown. Data are compiled from 2 independent experiments (n=2–3 mice per group).

Stimulation with WNV NS4b_209-226 and its truncations of splenocytes from JEV SA14-14-2-immunized H2-K^b−/−, but not from H2-D^b−/− mice induced IFN-γ, confirming H2-D^b restriction [7, 8]. Ex vivo peptide dose response CTL analysis of bulk cultures from JEV SA14-14-2-immunized mice confirmed WNV NS4b S9 (WNV S9) to be the optimal cross-reactive epitope (data not shown).

In order to determine whether the cross-reactive WNV S9 epitope was recognized in vivo, we assessed cytotoxicity during acute JEV SA14-14-2 infection. Splenocytes pulsed with decreasing doses of JEV NS4b S9 (JEV S9) were lysed to a similar extent in each of the JEV-immunized mice (Figures 1B and 1C). In contrast, the mean percent specific lysis of WNV S9-pulsed target cells was consistently lower than that seen for the JEV S9 variant for all dose ranges of peptide. Target cells pulsed with a H2-D^b-restricted influenza NP epitope (Figure 1B) and unpulsed splenocytes were not lysed in JEV-immunized or naïve mice (data not shown). These in vivo findings support ex vivo cytotoxicity studies demonstrating the higher cytotoxic activity of the JEV S9 variant compared to the WNV S9 variant in JEV-immunized mice (data not shown).

Characterization of JEV/WNV cross-reactive and JEV-specific CD8⁺ T cell variant epitopes in JEV-immunized mice

Functional avidity, defined as T cell responsiveness to a given epitope and its variants, may be influenced by the infecting virus, resulting in an altered outcome upon secondary heterologous virus infections [17–19]. Dose response experiments revealed that at higher peptide concentrations (1–0.1 μg/ml), the JEV S9 and WNV S9 peptide variants stimulated similar frequencies of IFN–γ⁺ CD8⁺ T cells in JEV-immunized mice. At lower peptide concentrations (0.01 μg/ml), the JEV S9 variant stimulated a greater proportion of IFN–γ⁺ CD8⁺ T cells than did the WNV S9 variant, suggesting a higher functional avidity for the homologous JEV variant (Figure 2A). The pattern for TNF-α production was similar to that seen for IFN–γ (data not shown).

Figure 2. Characterization of effector responses to JEV NS4b S9 and WNV NS4b S9 peptides by CD8⁺ T cells from JEV- and WNV-immunized mice.

(A) Intracellular cytokine staining of pooled splenocytes from 2 C57Bl/6J mice (n=3 independent experiments) obtained 7 days after immunization with JEV SA14-14-2 (left panel) or WNV (right panel). Splenocytes were stimulated with 10⁰–10⁻³ μg/ml of the JEV S9 (left panels) and WNV S9 (right panels) variants and stained for production of IFN-γ. Frequencies of IFN–γ⁺ CD3⁺ CD8⁺ T cells in media alone were subtracted from those of peptide-stimulated cells. JEV S9-stimulated cells (●); WNV S9-stimulated cells (○). (B–D) Cytokine profiles of virus-specific and cross-reactive CD8⁺ T cells during WNV and JEV infections. (B) CD8⁺ T cells from JEV-immunized and WNV-infected mice were analyzed for their ability to produce IFN–γ and TNF-α after stimulation with 1 μg/ml JEV S9 and WNV S9. Values represent the percentage of IFN–γ⁺, TNF-α⁺ and IFN–γ⁺ TNF-α⁺ CD8⁺ T cells. Representative data for one mouse per group immunized with JEV SA14-14-2 (top row), 1×10³ pfu JEV Beijing (second row), 1×10⁶ pfu JEV Beijing (third row) or WNV (bottom row). (C) Cumulative data (3 experiments, 2 spleens pooled per experiment). Bars represent mean ± SEM of the percent of CD8⁺ T cells producing IFN–γ, TNF-α or both IFN–γ and TNF-α from mice immunized with JEV or WNV following stimulation with JEV S9 (black bar) or WNV S9 (white bar). (D) Ratios of IFN–γ⁺ CD8⁺ T cells to IFN–γ⁺ TNF-α⁺ CD8⁺ T cells from JEV- and WNV-infected mice upon stimulation with JEV S9 (black bar) and WNV S9 (white bar) peptides. (*) indicates p<0.05 between WNV-infected group and all JEV groups, Mann-Whitney U test.

In WNV-infected mice, at higher peptide concentrations, the homologous WNV S9 variant induced higher frequencies of IFN–γ⁺ CD8⁺ T cells compared to the JEV S9 variant but frequencies declined rapidly at lower peptide concentrations (Figure 2A). In contrast, the frequency of IFN–γ⁺ CD8⁺ T cells induced by the heterologous JEV S9 variant was maintained at lower peptide concentrations (mean ± SEM % IFNγ⁺ CD8⁺ T cells at 0.01 μg/ml: JEV S9 = 1.63 ± 0.31% vs. WNV S9 = 0.45 ± 0.26%). Again, the pattern for TNF-α was similar to that seen for IFN-γ (data not shown).

Altered cytokine profiles depend upon the infecting flavivirus

We next examined the frequency of CD8⁺ T cells that secrete both IFN-γ and TNF-α in the context of the specific stimulating variant as well as infecting virus (JEV vs. WNV), in order to determine the contribution of each factor to CD8⁺ T cell cytokine profiles. In both JEV SA14-14-2 and WNV-infected mice, we found that stimulation by either the JEV S9 or WNV S9 variant induced both IFN-γ⁺ and IFN–γ⁺ TNF-α⁺ CD8⁺ T cells while single positive TNF-α⁺ CD8⁺ T cells were not detected in either JEV SA14-14-2 or WNV-infected mice (Figures 2B and 2C). In JEV SA14-14-2-immunized mice, stimulation with the JEV S9 or WNV S9 peptides induced higher frequencies of IFN-γ⁺ CD8⁺ T cells than IFN–γ⁺ TNF-α⁺ CD8⁺ T cells. In contrast, in WNV-infected mice, stimulation with either variant induced a higher frequency of IFN–γ⁺ TNF-α⁺ CD8⁺ than IFN–γ⁺ CD8⁺ T cells. The ratio of the frequencies of IFN–γ⁺ CD8⁺ T cells to IFN–γ⁺ TNF-α⁺ CD8⁺ T cells was significantly higher after JEV SA14-14-2 immunization compared to WNV infection for JEV S9 and WNV S9 (p<0.05, Mann-Whitney U) (Figure 2D). No significant difference in this ratio was detected between the JEV S9 and WNV S9 variants in either JEV SA14-14-2 immunized or WNV-infected mice. Of note, IFN–γ⁺ TNF-α⁺ CD8⁺ T cells from WNV-infected mice produced more TNF-α on a per cell basis than those from JEV SA14-14-2 immunized mice, while levels of IFN–γ from this population were similar for JEV and WNV (Supplementary Figure 2).

Since JEV SA14-14-2 is an attenuated virus, we used a pathogenic JEV (Beijing strain) to determine if differences in cytokine profiles between JEV and WNV could be explained on the basis of the pathogenicity of the infecting virus. We infected mice with a low dose (10³ pfu – comparable dose to WNV) or high dose (10⁶ pfu – comparable dose to JEV SA14-14-2) of JEV Beijing. Similar to JEV SA14-14-2, infection with either low or high dose JEV Beijing induced a significantly higher frequency of IFN-γ⁺ CD8⁺ T cells than IFN-γ⁺ TNF-α⁺ CD8 T cells compared to WNV infection (p < 0.05, Mann-Whitney U) (Figures 2B and 2C). These findings indicate that the infecting virus (JEV vs. WNV) determined the altered cytokine profile.

Frequency of epitope-specific CD8⁺ T cells in JEV-immunized and WNV-infected mice

To ascertain whether the differences in the cytokine profiles are related to different CD8⁺ T cell kinetics, we measured epitope-specific dimer⁺ CD8⁺ T cells 5, 7 and 10 days post-infection. Rapid expansion of CD44^hi dimer⁺ CD8⁺ T cells occurred between days 5 and 7 with peak levels occurring at day 7 for all infections with the exception of high dose JEV Beijing, which peaked at or before day 5 post-infection (Figure 3 and Supplementary Figure 3A). For JEV SA14-14-2 and low dose JEV Beijing, an approximately 4–8 fold contraction in frequency and absolute cell number (data not shown) of JEV S9 dimer⁺ CD8⁺ T cells occurred between days 7 and 10 while only a 1–2-fold contraction in frequency and absolute cell number (data not shown) of WNV S9 dimer⁺ CD8⁺ T cells occurred in WNV-infected mice. Similar to the pattern seen for cytokine production, infection with JEV induced a higher proportion of cross-reactive WNV S9 CD8⁺ T cells than cross-reactive JEV S9 CD8⁺ T cells seen in WNV infection. Although the peak CD8⁺ T cell response for high dose JEV Beijing occurred earlier, there was no difference in the frequency of IFN-γ⁺ and IFN-γ⁺ TNF-α⁺ CD8⁺ T cells at day 7 for all JEV infections. These results suggest that the kinetics of epitope-specific cells are not related to the altered cytokine profiles seen.

Figure 3. Kinetics of epitope-specific CD8⁺ T cell expansion and contraction.

Mice were immunized with (A) JEV SA14-14-2; (B) 1×10³ pfu JEV Beijing; (C) 1×10⁶ pfu JEV Beijing; or (D) WNV. Splenocytes were harvested on days 5, 7 and 10 post infection and stained with anti-CD3, -CD8, -CD44 and either JEV S9 (●) or WNV S9 dimer (○). Cells were gated on CD3⁺, CD8⁺, CD44^hi T cells as shown in Supplementary Figure 3A. Percent CD44^hi dimer⁺ CD8⁺ T cells are shown after subtraction of non-specific staining of JEV S9 and WNV S9 in naïve mice was subtracted from each value. Horizontal bars represent mean epitope-specific CD8⁺ T cell frequency.

Phenotype of epitope-specific cells during acute JEV and WNV infection

Effector CD8⁺ T cell activation depends on many factors, including antigen stimulation and inflammatory conditions [20]. To determine whether the differences in functional characteristics reflect differences in the activation state of CD8⁺ T cells during acute JEV and WNV infections, we examined the phenotype of epitope-specific cells over the course of acute infection. Splenocytes from infected mice were harvested on day 5, 7 and 10 post-infection, and CD62L, KLRG1 (killer cell lectin-like receptor G1) and CD127 (IL-7Rα) expression was measured on CD44^hi dimer⁺ CD8⁺ T cells (Figure 4A, Supplementary Figure 3A and 3B). At day 5, low-level expression of CD62L on dimer⁺ CD8⁺ T cells was seen in all infections indicating similar levels of CD8⁺ T cell activation (Figures 4A and 4B). By day 10, re-expression of CD62L was detected on JEV and WNV S9 dimer⁺ CD8⁺ T cells in all JEV groups. However, on day 10 after WNV infection, CD62L expression for the cross-reactive JEV S9 population increased while the WNV S9 dimer⁺ population had a persistent CD62L^lo phenotype (p<0.05, Mann-Whitney U).

Figure 4. Cell surface phenotype of epitope-specific cells differ following JEV and WNV infection.

(A) Mice were immunized with JEV SA14-14-2 (top row), 1×10³ pfu JEV Beijing (second row), 1×10⁶ pfu JEV Beijing (third row), or WNV (bottom row). Splenocytes were harvested on days 5, 7 and 10 and expression of CD62L, KLRG1 and CD127 was determined on CD3⁺ CD8⁺ CD44^hi JEV S9 (black line) and WNV S9 (gray line) dimer⁺ T cells. Shaded plot represents expression in naïve mice (representative from 1 naïve mouse per day). Each row shows a single representative mouse from each group (n=3–4 infected mice per group per day). (B) Phenotype analysis summary by immunization group. Percent of CD62L^lo (left panels), KLRG1^hi (middle panels) and CD127^hi (right panels) expressing CD44^hi CD8⁺ T cells bound to JEV S9 dimer (●) or WNV S9 dimer (○). Median values shown as horizontal line. #p<0.05 between WNV and JEV SA14-14-2 or 1×10⁶ pfu JEV Beijing on day 10, p>0.05 between WNV and 1×10³ pfu JEV Beijing; *p<0.05 between WNV and each JEV group on day 7 and day 10, respectively, Mann Whitney U. The mean background level for dimer staining was 0.12% for JEV A (range: 0.01% to 0.21%) and 0.19% for WNV A (range: 0.08% to 0.3%).

The pattern of KLRG1 and CD127 expression on effector CD8⁺ T cells define CD8⁺ T cell subsets that differ in their survival following an acute viral infection [20]. KLRG1 expression was upregulated on WNV S9 and JEV S9 dimer⁺CD8⁺ T cells for all groups as early as day 5, but progressively decreased in all JEV groups (Figures 4A and 4B). In contrast, KLRG1 expression increased between days 5 and 7 and persisted at high levels through day 10 in WNV-infected mice (median day 10 %CD44^hi WNV S9 dimer⁺ KLRG1^hi = 65.5% in WNV vs. %CD44^hi JEV S9 dimer⁺ KLRG1^hi 20.8%, 26.5%, 22.9% for 1×10³ pfu, 1×10⁶ pfu JEV Beijing, and JEV SA14-14-2, respectively; p<0.05, Mann-Whitney U). An inverse pattern was seen for CD127 expression; uniform downregulation of CD127 was seen by day 5 in all groups; re-expression of CD127 on dimer⁺ CD8⁺ T cells occurred by day 10 for all JEV groups but remained low in WNV-infected mice (median %CD44^hi CD127^hi WNV S9 dimer⁺ CD8⁺ T cells = 32.1% in WNV vs. 61.7%, 62.4%, and 64.8% for 1×10³ pfu, 1×10⁶ pfu JEV Beijing, and JEV SA14-14-2, respectively; p<0.05, Mann-Whitney U).

KLRG1^hi CD127^lo CD8⁺ T cells are defined as short-lived effector T cells (SLECs) that die off during the contraction phase while KLGR1^lo CD127^hi CD8⁺ T cells are memory precursor effector cells (MPECs) that survive contraction and differentiate into long-lived memory cells [21, 22]. Upregulation of KLRG1 and SLEC generation began by day 5 post infection in all groups but peaked on different days (Figures 5A and 5B). For JEV SA14-14-2 and high dose JEV Beijing, the highest frequency of SLECs occurred at day 5 (median 25.8% for SA14-14-2 and 40.2% for 10⁶ Beijing) (Figure 5B). For low dose JEV Beijing and WNV, the frequency of SLECs increased between days 5 and 7. By day 7, 32.2% of dimer⁺ CD8⁺ T cells were KLRG1^hi CD127^lo during low dose JEV Beijing infection compared to 58.3% of the dimer⁺ CD8⁺ T cells after WNV infection (p<0.05 between WNV and all JEV groups, Mann Whitney U). At day 5, frequencies of MPECs were low for all groups. At day 7, MPECs increased only in JEV infection and by day 10, 56.0 – 58.0% of the dimer⁺ CD8⁺ T cells were KLRG1^lo CD127^hi (Figure 5C). In contrast, during WNV infection, a majority of the dimer⁺ CD8⁺ T cells maintained a SLEC phenotype (KLRG1^hi CD127^lo) with a low frequency of MPECs on days 7 and 10 post-infection (p<0.05 between WNV and all JEV groups, Mann Whitney U).

Figure 5. KLRG1 and CD127 expression on dimer positive CD8⁺ T cells.

(A) Mice were immunized with JEV or WNV and splenocytes were harvested on the indicated days. Representative data for one mouse per group staining for KLRG1 and CD127 are shown. Summary data for (B) CD44^hi dimer⁺ KLRG1^hi CD127^lo CD8⁺ T cells and (C) CD44^hi dimer⁺ KLRG1^lo CD127^hi CD8⁺ T cells. Data represent the homologous dimer staining for each infection from individual spleens (JEV S9 (●) for JEV infections or WNV S9 (○) for WNV infection). *p<0.05 between WNV and all JEV groups on day 7 and day 10, respectively, Mann-Whitney U.

Viral Replication of JEV and WNV viruses in tissues

Differences in cytokine profiles and phenotype of effector CD8⁺ T cells may be related to differences in viral replication. Therefore, we measured viral titers by plaque assay in spleen, serum and brain 3 and 7 days post-infection with JEV and WNV to determine whether there were differences in peripheral (spleen and serum) and CNS (brain) replication. On day 3, between 6×10³–1.3×10⁵ pfu/ml and 2×10⁴–6×10⁴ pfu/g WNV was detected in the serum and spleen, respectively (Figures 6A and 6B). In contrast, we detected low titers (500 pfu/g) of JEV in spleens from 1 mouse in each of the low and high dose JEV Beijing groups. We were unable to detect virus in serum on day 3 from any of the JEV groups. At day 7 post-infection, we detected high titers of virus in brains from mice infected with 10⁶ pfu of JEV Beijing and WNV, but not from low dose JEV Beijing or JEV SA14-14-2 infected mice (Figure 6C). As expected, virus was not detectable in serum on day 7 or in brains on day 3 from any group (data not shown). These results suggest that overall virus burden may not be responsible for the altered cytokine profiles and altered phenotype responses measured between JEV and WNV but rather reflect differences in peripheral replication.

Figure 6. Virus burden in tissues following JEV and WNV infection.

Virus titer as assessed by virus plaque assay in (A) serum and (B) spleen at 3 days and (C) brain at 7 days post-infection with JEV or WNV. Each symbol represents a single mouse. Horizontal line represents geometric mean titer.

Discussion

Altered responses to flavivirus cross-reactive T cell epitopes can affect the outcome upon heterologous virus challenge. Our model system utilizes two viruses in the JEV serogroup, JEV and WNV, which have different clinical outcomes on sequential virus infection [14]. Overall, our results demonstrate that variant peptides that are homologous to the immunizing virus induce a greater frequency of epitope-specific CD8⁺ cells and higher levels of cytokine production and cytolytic activity. However, distinct CD8⁺ T cell functional responses arise depending on the infecting virus (JEV or WNV) independent of pathogenicity or peptide variant.

We identified a novel immunodominant JEV NS4b H-2D^b restricted CD8⁺ T cell epitope that is a variant of a recently published WNV epitope [7, 8]. We found that both the JEV and WNV variants induced cytokine secretion and stimulated lysis of peptide-coated targets in JEV-immunized mice. Regardless of the infecting virus, we found that the epitope hierarchy was higher for the variant peptide corresponding to the infecting virus. In addition, a greater proportion of CD8⁺ T cells were cross-reactive by dimer staining in JEV vs. WNV-infected mice. Dose response analyses suggested that although the frequency of WNV S9-specific cells was higher in WNV-infected mice, there was a greater functional avidity for the JEV S9 variant in both JEV-immunized and WNV-infected mice. These lower frequency but higher avidity cross-reactive CD8⁺ T cells may preferentially expand upon secondary heterologous JEV challenge and contribute to virus clearance as seen in vaccinia virus infection of lymphocytic choriomeningitis virus (LCMV)-immune mice [11].

Subdominant T cell epitopes have previously been shown to mediate heterologous immunity in the murine LCMV model, but immunodominant epitopes may also play a role. This has been suggested in studies of humans in whom immunodominant HLA-A2-restricted influenza M1-specific CD8⁺ T cells found to be cross-reactive to Epstein-Barr virus BMLF-1 expand during acute infectious mononucleosis and are thought to contribute to lymphoproliferation [23]. Similarly, in our model, CD8⁺ T cells specific for the immunodominant epitope are cross-reactive in both JEV and WNV-infected mice. In both JEV- and WNV-infected mice, higher frequencies of IFN–γ⁺ CD8⁺ T cells were detected compared to frequencies of TNF-α⁺ CD8⁺ T cells on day 7 post-infection, as has been seen after acute LCMV infection, independent of stimulating peptide variant [24]. However, we detected a significantly higher proportion of IFN–γ⁺ TNF-α⁺ -producing CD8⁺ T cells in mice infected with WNV compared to those immunized with both attenuated and pathogenic JEV strains (Figure 2B–D), as well as higher TNF-α production on a per cell basis (Supplementary Figure 2). The role of TNF-α in WNV infection is pleiotropic and may lead to resolution of the infection or to immunopathology depending on the concentration of TNF-α. Wang et al. demonstrated decreased mortality from WNV infection in TLR3^−/− mice, which they related to a decrease in TNF-α production and subsequent diminution in blood-brain permeability resulting in reduced WNV neuroinvasion [25]. However, Shrestha et al. demonstrated that neutralization of TNF-α in WNV-infected mice decreased their survival due to lower numbers of CD8⁺ T cells and macrophages trafficking to the brain [26]. CD8⁺ T cell production of TNF-α during acute WNV infection may contribute to their own trafficking into the central nervous system resulting in control of virus infection or increased immunopathology.

The qualitative disparity in cytokine profiles during acute infection with closely related viruses may be due to one of several factors: (1) differences in the kinetics of the response; (2) differences in activation state in different virus infections; (3) differences in viral burden and/or tissue tropism between attenuated JEV and WNV. In order to further delineate whether these differences are related to virus family versus viral virulence, we investigated responses to a pathogenic JEV virus strain, Beijing, at similar doses and clinical outcome to those of attenuated JEV SA14-14-2 and virulent WNV. At 1×10³ pfu of JEV Beijing, no mortality was seen in 6–7 week old mice, which is similar to what was seen after the attenuated JEV SA14-14-2 at 1×10⁶ pfu. In contrast, 1×10⁶ pfu of JEV Beijing resulted in 78% mortality in these mice, slightly higher than the 50% mortality observed with 1×10³ pfu WNV (data not shown). However, the differences in the CD8⁺ T cell responses between WNV and JEV did not correlate with mortality or inoculum dose, because all JEV strains, whether attenuated or pathogenic, induced similar CD8⁺ T cell responses. These results suggest that differences in the cytokine profiles is due to intrinsic differences between JEV and WNV infections

Kinetic analysis of JEV S9 and WNV S9-specific CD8⁺ T cell responses demonstrated that peak CD8⁺ T cell responses occurred on day 7 post-infection for all viruses with the exception of responses to 1×10⁶ pfu JEV Beijing which peaked on or before day 5. Activation state, as demonstrated by downregulation of CD62L, was similar for all groups at days 5 and 7 post-infection. The increase in SLECs during JEV infection was much shorter in duration than what has been reported for acute LCMV infection [27]. However, a significantly higher proportion of KLRG1^hi CD127^lo SLECs was detected after WNV infection on day 7 compared to all JEV virus infections, and these differences persisted to day 10 post-infection. These findings are in contrast to those reported by Brien et al in which WNV S9 dimer⁺ CD127^hi CD8⁺ T cells predominated at day 7 after WNV infection [7]. That study utilized a different WNV strain, a lower dose of virus (20–600 pfu) and a different route of administration (subcutaneous), which may have impacted the kinetics of virus replication and subsequent effector CD8⁺ T cell generation. We also found that the frequency of KLRG-1^lo CD127^hi CD8⁺ T cells was higher at day 10 post-infection in JEV-infected mice compared to WNV-infected mice.

As expected, replication of the attenuated JEV SA14-14-2 strain in peripheral tissues was below the level of detection in viral plaque assay (Figure 6) [28]. However, unexpectedly, infection with low or high dose JEV Beijing also resulted in minimal peripheral virus replication on day 3, whereas high dose JEV Beijing infection resulted in very high titers of virus in brains on day 7 post-infection. In contrast, WNV was easily detectable in serum and spleen on day 3 as well as in brains at day 7. The ability of WNV to replicate in the spleen early during infection may influence programming of the CD8⁺ T cell response. However, it is also possible that peripheral replication of JEV peaked at an earlier time point. These differences in viral replication may influence inflammatory signals generated during the acute immune response. IL-12 and IFN–γ are two inflammatory cytokines known to influence the generation of SLECs and the levels of these cytokines may differ in JEV and WNV infections [27, 29]. The persistence of KLRG1^hi CD127^lo SLECs in WNV infection may reflect prolonged antigenic stimulation or increased inflammatory responses due to persistent virus as has been described in other WNV animal models [30, 31].

Previous studies of heterologous immunity identified memory cross-reactive CD8⁺ T cells following secondary heterologous viral challenge [10, 11]. These cross-reactive T cells were found to be subdominant during the primary response, and the sequence of infection influenced the hierarchy of these subdominant cross-reactive T cells after secondary heterologous challenge [32, 33]. In our model, the immunodominant CD8⁺ T cell epitope was found to be cross-reactive, but to differing degrees, following either JEV or WNV infection. Our detailed characterization of these epitope-specific responses did not demonstrate an alteration in epitope hierarchy, but rather differences in cytokine profiles and T cell phenotype As previous studies have elucidated a role for subdominant cross-reactive CD4⁺ and CD8⁺ T cells in protection as well as immunopathology, future experiments will address the role of the two cross-reactive CD4⁺ T cell epitopes we identified and subdominant cross-reactive CD8⁺ T cell epitopes along with the immunodominant cross-reactive CD8⁺ T cell epitope in secondary heterologous JEV and WNV infections [10, 11].

Here we have shown that primary infections with JEV and WNV give rise to functionally and phenotypically distinct CD8⁺ T cell responses. These differences are due to the infecting virus (JEV vs. WNV) rather than the stimulating variant (WNV S9 vs. JEV S9) or viral pathogenicity. The JEV-WNV cross-reactive CD4⁺ and CD8⁺ T cell epitopes we have identified will be useful tools to study the pathogenesis of sequential heterologous flavivirus infections. Flaviviruses continue to emerge into new geographic regions of the world, giving rise to the possibility of new patterns of sequential infection with unknown outcomes (e.g. WNV into dengue- and yellow fever virus-endemic regions of South America). Altered CD8⁺ T cell effector functions between flaviviruses may to lead to immunopathology or protection upon a secondary flavivirus infection. Additional experiments are needed to determine whether cross-reactivity occurs between other members of the flavivirus family and its possible impact on disease outcome.

Materials and Methods

Viruses and Cell Lines

JEV strain SA14-14-2 was provided by Dr. Thomas Monath (Acambis, Inc.). JEV strain Beijing was provided by Dr. Alan Barrett (University of Texas Medical Branch, Galveston, TX). WNV strain 3356 was provided by Dr. Kristen Bernard (Wadsworth Center, Albany, NY). Flaviviruses were propagated and titered in Vero cells (ATCC). The EL-4 T cell lymphoma cell line (H-2^b) served as target cells.

Peptides

Peptide (15–19mer) arrays corresponding to the entire proteome of WNV were obtained through the NIH Biodefense and Emerging Infections Research Resources Repository, NIAID, NIH (BEI Resources, Manassas, VA). Peptide truncations (>70% or >90% purity) were obtained from AnaSpec, Inc. (San Jose, CA) and 21^st Century Biochemicals (Marlborough, MA).

Mice and immunizations

Female C57Bl/6J mice aged 6–11 weeks (Jackson Laboratories, Bar Harbor, ME) and C57BL/6Ji-Kbtm1 N12 (H2-K^b−/−) and C57BL/6Ji-Dbtm1 N12 (H2-D^b−/−) mice (Taconic Farms, Germantown, NY) were housed in specific pathogen-free conditions in the University of Massachusetts Medical School Biocontainment facility and were cared for according to guidelines approved by the University of Massachusetts Medical School’s Institute Animal Care and Use Committee. Mice were infected i.p. with JEV SA14-14-2 (1×10⁶ pfu), JEV Beijing (1×10³ or 1×10⁶ pfu), or WNV (1×10³ pfu).

Generation of Bulk Culture Cell Lines

Spleens were harvested one week following JEV boost and splenocytes were prepared as previously described [34]. Splenocytes were stimulated with 10 μg/ml peptide in RPMI-1640 containing 10% FBS, 1% penicillin/streptomycin, 5×10⁻⁵ M β-mercaptoethanol and recombinant human IL-2 (rhIL-2; BD Biosciences) (25 U/ml) at 37° C. At day 14 and every 14 days thereafter, γ-irradiated naïve C57Bl/6J splenocytes were pulsed with 10 μg/ml peptide, washed, and added to the bulk cultures at a stimulator-to-responder ratio of 5:1.

IFN-γ ELISPOT

ELISPOT assays were performed as described [34]. Freshly isolated day 7 splenocytes from 2 naïve or JEV-immunized mice were pooled and plated on anti-mouse IFN-γ coated 96-well plates in duplicate or triplicate (2.5 × 10⁵ per well) and stimulated with WNV or JEV peptides (2 μg/ml), Con A (2.5 μg/ml), or media overnight at 37°C. After PBS wash, anti-mouse IFN-γ biotinylated mAb was added for 2 hours followed by streptavidin-HR. Spots were developed with NovaRed substrate kit (Vector Laboratories, Burlingame, CA) and counted with a CTL reader. The number of spot forming cells per million was calculated as [(mean spots in experimental wells – mean spots in medium control) × 4] × 10⁶. The average number of spot forming cells per million in media alone was 21 ± 22. A positive response was ≥2 times media background.

Intracellular cytokine staining

Splenocytes (1×10⁶ cells) were stimulated either with peptide (1 μg/ml), peptide pools (5 μg/ml), PMA (50 ng/ml) and ionomycin (250 ng/ml) (positive control) or without peptide (negative control) in the presence of brefeldin A (BD GolgiPlug) for 5 hours. Cells were washed in PBS supplemented with 2% FBS and 0.05% Sodium azide and incubated with 1 μg anti-CD16/32 (2.4G2). Cells were surface stained with anti-CD3 (145-2C11; eBioscience, San Diego, CA), -CD4 (L3T4), or -CD8 (Ly-2; eBioscience). After permeabilization (BD CytoFix/CytoPerm), and wash with BD Perm/Wash, cells were stained with anti-IFN–γ (XMG1.2) and anti-TNF-α (MP6-X522; eBioscience) and fixed in 1% paraformaldehyde. Samples were acquired on a FACSCalibur (BD Biosciences) and data were analyzed using FloJo software (Tree Star, Inc.). The percentage of CD4⁺ or CD8⁺ T cells producing IFN-γ in response to media was subtracted from peptide-stimulated cells. Reagents were obtained from BD Bioscience unless otherwise noted.

⁵¹Chromium Release Assay

⁵¹Chromium release assay were performed as previously described [34]. In brief, ⁵¹Cr-labelled EL-4 cells were incubated with peptide or media alone. Effector cells were added in triplicate and incubated for 4 hours at 37°C. Renex or media alone were added to target cells for determination of maximum target cell lysis and spontaneous lysis, respectively. Supernatants were harvested and counted on an automated gamma counter. Percent specific lysis was calculated as [(sample ⁵¹Cr release – spontaneous ⁵¹Cr release)/(maximum ⁵¹Cr release – spontaneous ⁵¹Cr release)] × 100.

In vivo cytotoxicity

In vivo cytotoxicity experiments were performed as described with modifications [35]. Naive splenocytes (target cells) were pulsed with 1, 0.1 or 0.01 μg/ml of JEV NS4b S9, WNV NS4b S9 peptide or control influenza NP 366-374 peptide (1 μg/ml) for 45 min at 37°C. Cells were stained with 1 μM Cell Trace Far Red 7-hydroxy-9H- (1,3-dichloro-9,9-dimethylacridin-2-one)-SE (DDAO-SE; Invitrogen, Carlsbad, CA) and serial dilutions of CFSE (5 μM, 1.5 μM, 0.4 μM, 0.1 μM; Invitrogen). Target cells in PBS (2 × 10⁷ cell/ml) were injected i.v. into JEV-immunized or naïve mice 8 days post immunization. Splenocytes were harvested 2 hours later and analyzed using a FACS Aria. Percent specific lysis was calculated by the formula 1-(Ratio Immune/Ratio Naïve) × 100, where Ratio=(# events of JEV or WNV peptide/# events of control influenza peptide).

Phenotype Analysis of epitope-specific CD8⁺ T cells

Recombinant H-2D^b:Ig fusion protein (4μg; BD Biosciences) was loaded with variant peptides (>90% purity) at 640 molar excess peptide in PBS (pH=7.2) at 37°C overnight according to manufacturers guidelines. Peptide-loaded dimer was incubated with 2.4 μg APC-anti-mouse IgG (BD Biosciences, mAb A85-1) followed by incubation with purified mouse IgG isotype control (4μg; BD Biosciences; mAb A111-3). Splenocytes were resuspended in PBS, stained with Live/Dead Aqua, and incubated with anti-CD16/32 (2.4G2; BD Bioscience), followed by staining with 4 μg of peptide-loaded dimer. Cells were surface stained with anti-CD44, -CD62L, -KLRG1 and -CD127 conjugated with FITC, PE-Cy7, and PerCP-Cy5.5, washed and resuspended in BD Stabilizing Buffer. Peptide-loaded dimer staining levels in naïve mice were subtracted from experimental values in infected mice. The gating strategy is shown in Supplementary Figures 3A and 3B.

Plaque Assay of tissues from JEV and WNV-infected mice

On days 3 and 7 post JEV or WNV infection, spleen, brain and serum were obtained and flash frozen at −70°C. Tissues were homogenized to give a 10% (spleen) or 20% (brain) homogenate based on tissue weight using a Qiagen mixer mill. Serial dilutions were made in MEM and titers were determined on Vero cells as described [36]. Plates were incubated for 2 (WNV) or 4 days (JEV Beijing and SA14-14-2) prior to second agar overlay. The limit of detection was 50 pfu/ml for serum, 250 pfu/g for brain and 500 pfu/g for spleen.

Statistics

Means, medians and standard errors were calculated using GraphPad Prism (GraphPad Software, Inc., LaJolla, CA). Comparisons of variables between JEV and WNV infection groups were performed with log transformed data using the Mann-Whitney U test on STATA software (StataCorp, College Station, TX). P < 0.05 was considered significant.

Abbreviations

WNV

West Nile virus

JEV

Japanese encephalitis virus

LCMV

lymphocytic choriomeningitis virus

KLRG1

killer cell lectin-like receptor G1

E

envelope

NS

nonstructural

SLEC

short-lived effector cell

MPEC

memory precursor effector cell