CD8⁺ T Effector Memory Cells Protect against Liver-Stage Malaria

Abstract

Identification of correlates of protection for infectious diseases including malaria is a major challenge and has become one of the main obstacles in developing effective vaccines. We investigated protection against liver-stage malaria conferred by vaccination with adenoviral and Modified Vaccinia Ankara (MVA) vectors expressing pre-erythrocytic malaria antigens. By classifying CD8⁺ T cells into effector (T_E), effector/memory (T_EM) and central memory (T_CM) subsets using CD62L and CD127 markers, we found striking differences in T cell memory generation. While MVA induced accelerated T_CM generation, which could be efficiently boosted by subsequent adenoviral administration, it failed to protect against malaria. In contrast, adenoviral (Ad) vectors, which permit persistent antigen delivery, elicit a prolonged T_E and T_EM response that requires long intervals for an efficient boost. A preferential T_EM phenotype was maintained in liver, blood and spleen following Ad/MVA prime-boost regimens and animals were protected against malaria sporozoite challenge. Blood CD8⁺ T_EM cells correlated with protection against malaria liver-stage infection, assessed by estimation of number of parasites emerging from the liver into the blood. The protective ability of antigen-specific T_EM cells was confirmed by transfer experiments into naive recipient mice. Thus, we identify persistent CD8 T_EM populations as essential for vaccine-induced pre-erythrocytic protection against malaria, a finding that has important implications for logical vaccine design.

Introduction

Major efforts are currently being made at developing an effective malaria vaccine, as infection by Plasmodium falciparum continues to be the most common cause of mortality in children under five in a number of countries, mainly in sub-Saharan Africa(1). Vaccines based on the pre-erythrocytic antigens circumsporozoite protein (CSP)(2) and thrombospondin-related adhesion protein (TRAP)(3) are the leading immunization strategies currently under development. Some vaccine regimes inducing large numbers of CD8⁺ T cells specific for pre-erythrocytic antigens can protect humans against experimental challenge(4); (Ewer, K. et al., submitted). However, despite having vaccination schedules inducing extensive malaria-specific T-cell responses, such as those employing adenoviral and poxviral vaccine vectors(5), our knowledge on how to use these tools efficiently to protect against malaria is hampered by an incomplete understanding of how they work. For example, it is known that circulating T cell frequencies, measured by ELIspot, recognising liver-stage antigens following vaccination did not predict protection in individual subjects(4). Indeed, there is ongoing uncertainty about natural and vaccine induced mechanisms of protection in pre-erythrocytic malaria (6, 7). For these reasons, there is an increased interest in studying the phenotype and circulatory patterns of T cells to develop protective correlates (8, 9). Defining correlates of protection for T-cell vaccines is a challenging task that could help optimizing prime-boost regimes and efficiently modulate the quality of memory T cells (10).

Vaccination with irradiated sporozoites is one of the most effective ways to induce sterile protection both in animal models and humans (11, 12). This approach can be substantially more efficacious with the use of live sporozoites in recipients under concomitant treatment with chloroquine (13, 14). It has been demonstrated that persistent antigen presentation of sporozoite-derived antigens is important for optimal induction of CD8⁺ responses against the malaria liver-stage parasites(15). Viral vectors expressing liver-stage antigens are another promising malaria vaccine approach able to induce sterile protection in animal models(16, 17). Of these, adenoviral vectors have the ability to persistently express a transgene in vivo for at least two years to maintain an activated cohort of antigen-specific T cells(18), while MVA expresses a transgene only for some hours and becomes undetectable within 48 hours(19).

Differences in longevity or persistence of transgene expression have an impact on the generation of memory T cell responses(20). Three major subsets of antigen-experienced CD8⁺ T cells have been identified, based on the expression of CD62L (L-selectin) and CD127 (IL-7R α-chain): central memory T cells (T_CM) (CD62L⁺CD127⁺), effector memory T cells (T_EM) (CD62L⁻CD127⁺) and effector T cells (T_E) (CD62L⁻CD127⁻). Bachmann et al. have shown that the protective ability of each subset varies depending on the context of the infection(20): T_E and T_EM have a greater ability to protect against peripheral organ infection with vaccinia virus than T_CM due to the distinct anatomical distribution of these cell types (20).

In this study, we investigate T_E/T_EM/T_CM generation by viral vectors. We show that different regimes differ in T cell induction profiles, and unequivocally demonstrate for the first time that T_EM induction is critical for protection against sporozoite challenge.

Results

MVA vaccination induces accelerated development of memory CD8⁺ T cells

Following a single immunization with either adenovirus (Ad) or MVA encoding the ME.TRAP antigen, we found a striking difference between the two viral vectors in the kinetics of antigen-specific CD8⁺ T cell expansion, contraction and generation of memory phenotype (Figure 1). Antigen (Pb9 epitope)-specific CD8⁺IFN-γ responses in spleen after MVA vaccination peaked within a week following immunization and contracted rapidly by week 3. In contrast, immunization with adenovirus resulted in delayed kinetics, with the peak of CD8⁺IFN-γ⁺ cell expansion at three weeks, and a slow contraction phase (Figure 1a). We further analyzed the phenotype of the CD8⁺IFN-γ⁺ Ag-specific T cells using CD27, CD62L and CD127 as memory T cell markers and CD43 as an effector cell marker. These markers enabled us to distinguish three main subsets of antigen-experienced CD8⁺ T cells: effector T_E (CD62L⁻CD127⁻), effector memory T_EM (CD62L⁻CD127⁺) and central memory T_CM (CD62L⁺CD127⁺)(20).

Figure 1. Accelerated CD8 T cell memory generation by MVA and delayed memory induction by AdC9.

BALB/c mice were immunized intradermally (i.d.) with AdC9 (5×10⁹ vp) or MVA (1×10⁶ pfu) expressing ME.TRAP. (a) – (f) Spleen T cells were assessed for expression of IFNg and effector/memory cell surface markers in response to Pb9 peptide (CS_252-260) stimulation, at different intervals following immunization with AdC9 () or MVA () viral vectored vaccines. (a) Kinetics of antigen specific CD8⁺ T cells expressing IFN-γ⁺. This cell subset was further characterized for the presence of the effector T cell (T_E) marker CD43^hi (b), and memory cell markers CD27^hi (c), CD62L (L-selectin) (d), IL-2 (e), CD127 (IL-7Rα) (f), at the same time-points. ** Indicates statistically significant differences between the AdC9 and the MVA immunizations at the given time-point (**=p<0.01, ***=p<0.001). (g) Unstimulated antigen-speciffic CD8⁺ T cells in the spleen were identified using the H-2K^d Pb9 tetramer at the indicated days post-immunization and analysed for effector (T_E), central memory (T_CM) and effector memory (T_EM) phenotype, based on the expression of memory markers CD62L or CD127. Numbers in quadrants represent the percentage of Pb9 Tet⁺ cells classified as T_E: CD62L⁻CD127⁻, T_EM: CD62L⁻CD127⁺, T_CM: CD62L⁺CD127⁺. Data are representative of two independent experiments with between three and six mice per experiment.

Vaccination with adenovirus induced a high proportion (>90%) of antigen-specific CD8⁺IFN-γ⁺CD43^hi cells which gradually declined over time, with >40% of the IFN-γ⁺ cells still retaining the CD43^hi effector status by week 8 (figure 1b). A concomitant decrease or absence of memory markers CD27 and CD62L was observed on antigen-specific CD8⁺ T cells early after Ad administration (Figure 1c-f). In contrast, the initial peak of around 80% of Ag-experienced CD8⁺IFN-γ⁺ cells expressing CD43^hi at week 1 post-immunization with MVA was followed by a sharp loss of CD43 expression and an early and steady increase in expression of all memory markers assessed: CD27^hi (Fig 1c), CD62L (Fig 1d), IL-2 (Fig 1e) and CD127 (Fig 1f). The CD127 (IL-7Rα) was the only memory marker that increased in expression after Ad immunization and followed a similar trend to the MVA responses, albeit lower at all time points tested. These results suggest that, in contrast to adenovirus, a single injection of a poxviral vector MVA can result in a fast transition to CD8⁺ T-cell memory responses, which develop as early as two weeks after immunization.

To investigate whether these observations remain true in the absence of in vitro antigen stimulation, we used a H-2K^d Pb9 tetramer and analyzed ex vivo the kinetics of memory T cell generation in the spleen. Immunization with adenovirus induced a substantial number of CD8⁺ T_E cells shortly after vaccination (47% of the Pb9-tetramer positive cells) and these slowly contracted, leaving a residual population of 15% of tetramer positive cells 60 days after vaccination (Figure 1g). The T_EM compartment gradually expanded over time from 52%, reaching 78% of tetramer-positive cells by day 60. The compartment containing the lowest proportion of cells was the T_CM, which reached a maximum of 5.7% on day 60 post-vaccination. In contrast, immunization with MVA induced a striking expansion of T_CM cells, which by day 60 represented a third of all tetramer positive cells. This was paralleled by a complete contraction of the early T_E population by day 60 (Figure 1g). The proportion of T_EM remained stable during the time tested.

An MVA boost 8 weeks after Ad prime provides the highest degree of protection against malaria challenge but does not correlate with frequency of CD8⁺IFN-γ⁺ T cells in spleen

We asked whether the increased CD8⁺ memory generation by MVA could be translated into a more efficient vaccination regime by altering the order of the viral-vectored vaccines and the prime-boost interval. To this end, we chose several vaccination regimes and assessed their protective efficacy against malaria two and eight weeks after the last vaccination, using the malaria parasite P. berghei in a mouse challenge model (Table 1). This stringent challenge model consists of an intravenous (i.v.) administration of 1,000 sporozoites and protection is usually measured as a complete absence of parasites in blood (sterile protection). Frequencies of IFN-γ-producing antigen-specific CD8⁺ T cells in the mouse spleen were used as a measure of the immunogenicity of the different regimes.

Table 1. Antigen-specific CD8⁺ responses and protective efficacy against sporozoite challenge by Ad-MVA and MVA-Ad vaccination regimes with different prime-boost intervals.

TOP: BALB/c mice (n=16 per group) were immunized i.d. with AdC9 ME.TRAP (5×10⁹ vp/mouse) or MVA ME.TRAP (1×10⁶ pfu/mouse). A heterologous boost was administered 1, 2, 4 or 8 weeks later using the same viral dose for each vector. Subsets of vaccinated mice (n=3 per group) were sacrificed on weeks 2 and 8 after the last vaccination and their spleens harvested to assess the number of antigen-specific CD8⁺ T cells producing IFN-γ⁺. In the remaining animals, antigen-specific IFN-γ⁺ production in peripheral blood was measured by ex vivo PBMC ELISpot at weeks 2 and 8 post last vaccination and mice subsequently challenged with malaria parasites. Challenge was performed by i.v. administration of 1000 sporozoites of Plasmodium berghei. Numbers in the two columns on the right represent the percentage of animals that had sterile protection. Statistical analyses were performed using a logrank test as described in methods. Statistical differences are indicated as: * p<0.05, ** p<0.01, *** p<0.001, and show comparison of individual regimes with the naïve control.

BOTTOM: BALB/c mice were immunized i.d. with AdC9 ME.TRAP (5×10⁹ vp/mouse) or MVA ME.TRAP (1×10⁶ pfu/mouse). A heterologous boost was administered 8 weeks later using the same viral dose for each vector. Antigen-specific T-cell responses were assessed 14 weeks later in blood by flow cytometry after stimulation of PBMCs with Pb9 peptide. % indicate cells producing IFN-γ within the CD8⁺ compartment and statistical differences are indicated as ** p<0.01. Mice went through a challenge with 1,000 sporozoites of Plasmodium berghei per mouse and sterile protection was assessed.

Regime	Prime-boost Interval (weeks)	Antigen-specific CD8+ / spleen (×106) (n=3)		Difference in Antigen-specific CD8+ / spleen (× 106) Between MVA-AdC9 & AdC9-MVA (mean, 95% Confidence interval)		% Sterile protection against P. berghei sporozoite challenge (n=5)
		2 wks post boost	8 wks post boost	2 wks post boost	8 wks post boost	2 wks post boost	8 wks post boost
No - AdC9	-	0.49	0.90	N/A	N/A	20	0
MVA-AdC9	1	2.47	1.54	N/A	N/A	40	20
MVA-AdC9	2	2.11	1.83	N/A	N/A	40 **	0
MVA-AdC9	4	1.39	1.54	N/A	N/A	40	20
MVA-AdC9	8	2.69	2.36	N/A	N/A	50 ** (n=20)	10 (n=10)
No - MVA	-	0.07	0.07	N/A	N/A	0	0
AdC9-MVA	1	1.67	0.90	−0.8 [−2.4 to 0.79]	−0.64 [−1.2 to −0.098]	100 **	20
AdC9-MVA	2	1.75	1.00	−0.36 [−1.7 to 1.0]	−0.83 [−2.9 to 0.012]	80 **	40 *
AdC9-MVA	4	1.93	1.18	0.54 [−1.2 to 2.2]	−0.36 [−1.4 to 0.67]	40 *	20
AdC9-MVA	8	2.37	1.38	−0.31 [−1.8 to 1.1]	−0.97 [−2.6 to 0.60]	96 **** (n=27)	59 *** (n=17)
Naïve	-	0	0			0	0
Regime	Prime-boost Interval (weeks)	% Antigen-specific CD8+ in blood		Difference in Antigen-specific CD8+ in blood Between MVA-AdC9 & AdC9-MVA (95% Confidence interval) p value		% Sterile protection against P. berghei sporozoite challenge
			14 wks post boost		14 wks post boost		14 wks post boost
MVA-AdC9 (n=24)	8		30.18 **		[1.866-12.02] p=0.0086		9.5 (n=14)
AdC9-MVA (n=17)	8		23.24	6.9 [1.866 to 12.02]			21.43 (n=21)

Regardless of the length of the prime-boost interval, vaccination regimes involving Ad prime and MVA boost showed a trend towards a more sustained protection against malaria as compared to MVA only or MVA primed Ad-boosted regimes (Table 1): the highest degree of long-term sterile protection - 96% survival of animals challenged at 2 weeks and 59% survival of animals challenged at 8 weeks - was achieved when animals were primed with adenovirus and boosted with MVA 8 weeks later (Table 1). This protection was significantly higher than the MVA-Ad regime on a challenge performed both, at 2 weeks (Hazard ratio of 12.02 95% CI[3.336-43.34] p=0.0001) and 8 weeks after the last vaccination (Hazard ratio of 7.794 95% CI[1.794 −33.87] p=0.0062) Importantly, we observed that this protective efficacy was not associated with higher antigen-specific T cell responses in the spleen prior to challenge (Table 1). For most prime-boost intervals the MVA-Ad regime was superior to Ad-MVA at eliciting strong antigen-specific IFN-γ production in the spleen, but indicated a trend towards less protection against malaria challenge at both 2 and 8 weeks after vaccination. To further confirm these results, BALB/c mice (n=17-24) were vaccinated with both regimes (AdC9-MVA and MVA-AdC9) in an 8-week interval and antigen-specific responses were assessed long after vaccination -14 weeks- to determine the memory immune responses of every regime using blood as a different compartment (Table 1, bottom). We observed a significantly higher immunogenicity by the MVA-Ad regime than the Ad-MVA (30.18% vs 23.24% of CD8⁺IFN-γ⁺, 95% CI [1.866-12.02] p=0.0086). Despite inducing higher Ag-specific frequencies of T cells, protection induced by MVA-Ad in these mice indicated a trend towards less protective efficacy.

These data indicate that high numbers of antigen-specific CD8⁺ IFN-γ-producing T cells are not a predictor of the vaccine’s protective ability and highlight the importance of dissecting vaccination-induced immune responses to find better correlates of protection.

The highly protective Ad-MVA regimen predominantly induces CD8⁺ cells with T_EM phenotype in the spleen

To identify the mediators of vaccine induced protection, we investigated the phenotypic and functional profile of the antigen-specific T cells in the spleen using two of the most protective regimes consisting on prime-boost intervals of 8 weeks (Table 1): Ad-MVA and MVA-Ad regimes. Two weeks after priming with adenovirus, the majority of Ag-experienced T cells were T_EM (61.8%), with an overall hierarchy T_EM>>T_E>T_CM (Figure 2a). In contrast, priming with MVA induced a predominant (56.1%) T_CM phenotype and a small proportion of effector T cells: T_CM>T_EM>>>T_E (Figure 2a). Two weeks after the boost injection, the proportion of Ag-specific cells with a T_CM phenotype was comparable between the two regimes (mean for AdC9-MVA=7.9 vs MVA-AdC9=9.2, 95%CI [−5.463 to 2.939]. However, the T_EM population was significantly higher in the Ad-MVA regime (Mean of 86.1%, compared to 77.6% for MVA-Ad; 95%CI [4.4-12.5], p=0.0022 by t-test); whereas the opposite was true of the T_E population (Mean of 5.9%, compared to 13.0% for MVA-Ad, 95%CI [−12.13 to −1.995], p=0.014 by t-test) (Figure 2a).

Figure 2. Ad/MVA prime/boost regimes induce a predominant CD8⁺ T_EM phenotype in spleen and strong liver T-cell responses.

BALB/c mice were immunized with AdC9-MVA or MVA-AdC9 regimes using the same doses as in Figure 1. Antigen-specific immune responses to the H2-K^d immunodominant Pb9 peptide were assessed in the spleen 2 weeks after the boost using the Pb9 tetramer (a, b) and in the liver upon stimulation with Pb9 peptide (c, d, e). (a) Relative frequencies of CD8⁺ T_E (effector), T_EM (effector memory) and T_CM (central memory) subsets after a viral-vectored prime or boost shown in representative plots for each vaccination regime. Ex vivo splenocytes were stained with anti-CD8, CD62L, CD127 and H2-K^d Pb9 tetramer. Levels of CD62L and CD127 expression within the Pb9 tet⁺ population are shown (gates were set using isotype controls). (b) Bar graph representation of the plots in (a) following prime-boost, showing the group averages for the three T cell sub-populations (***p<0.001, **p<0.01). (c) Antigen-specific CD8⁺ IFNγ⁺ responses in the liver generated by prime or prime/boost immunizations; representative plots are shown. Mononuclear cells were isolated from perfused livers 2 weeks after the last vaccination and stimulated for 5 hours with Pb9 peptide to identify antigen-experienced T cells induced by vaccination and homing to the liver. (d) Kinetics of the Pb9-specific IFN-γ responses by CD8⁺ T cells in the liver following prime-boost vaccination. BALB/c mice (n=33) were vaccinated with either Ad-MVA or MVA-Ad regimes and liver responses were assessed at the indicated time points (3-4 mice per time point). (e) Antigen-specific CD8⁺ T-cell responses in the liver generated by Ad-MVA and MVA-Ad regimes. Cells were stained intracellularly for IL-2, TNF-α and IFN-γ cytokines and a Boolean analysis performed using FlowJo. Data was exported and analyzed using SPICE software (M. Roederer, VRC, NIH). The pie charts display relative proportions of cells producing one (yellow), any two (orange) or any three (red) cytokines and bars show the frequencies of CD8⁺ T-cells from the liver compartment producing the indicated cytokines, in response to antigen stimulation. Significant differences were calculated using the t-test provided by the SPICE software; * indicates p<0.05.

It is recognized that different subsets of memory CD8⁺ T-cells vary in their ability to acquire effector functions and circulatory patterns. T_EM cells, for instance, rapidly acquire effector functions(21) and have a preferential circulatory pattern to the peripheral organs, such as the liver and lung(22). To further confirm the predominant induction of antigen-specific T_EM cells by Ad-MVA, we analyzed in the same mice the circulatory properties of antigen-specific CD8⁺ T-cells in the liver - a peripheral organ, as well as the target organ for our malaria vaccines (Figure 2c). A single immunization with an adenoviral vector induced a stronger Ag-specific CD8⁺ response in the liver compared to MVA prime (Mean of Ad 45.4%, 95%CI [30.96 to 45.56] compared to 7.1%, for MVA; p<0.0001 by t-test) two weeks post immunization. The T-cell responses in the liver were also higher on week two after the boost in the Ad-MVA regimen compared to MVA-Ad (Mean of 35.9%, 95%CI [2.459 to 31.44] compared to 18.9%, 95% for MVA-Ad; p=0.0287 by t-test). Thereafter, analysis of the full kinetics of the antigen-stimulated responses in the liver revealed that, eight weeks post-boost (week 16) onwards, the proportion of CD8⁺IFN-γ⁺ Pb9-specific cells in the liver is the same in both Ad-MVA and MVA-Ad vaccination regimes (Figure 2d). The capacity of the Ag-specific CD8⁺ T cells in the liver to produce multiple immune defense cytokines was also assessed for the two vaccination regimes (Figure 2e). However, phenotypic differences in vaccine induced hepatic T cells were also observed when measuring antigen-induced IFN-γ⁺/TNF-α⁺/IL2⁺ secretion by hepatic T cells with significantly higher IFN-γ⁺/ TNF-α⁺, (Mean of 20.4% compared to 11.4% for MVA-Ad; 95%CI [1.240 to 16.69], p=0.0296 by t-test). (Figure 2e) with the protective regime, suggesting a persistent alteration in phenotype induced by the Ad-MVA.

In summary, vaccination with the Ad-MVA regime, when compared to the less protective MVA-Ad, preferentially induced CD8⁺ T_EM (CD8⁺IFN-γ⁺CD127⁺CD62L⁻) cells in the spleen and high frequencies of antigen-specific CD8⁺ cells in the liver, a key organ required by the sporozoite for replication and a target for our pre-erythrocytic vaccine.

Blood CD8⁺ T_EM responses correlate with protection against liver-stage malaria

Our results suggested that CD8⁺ T_EM responses could play a role in protection against liver-stage malaria, so we set out to explore whether a preferential induction of blood antigen-specific T_EM cells constitutes a correlate of protection at the individual level. We employed the prime/boost interval of eight weeks, followed by a challenge with 1,000 malaria sporozoites at week 22 (Figure 3a). This late time-point was chosen in order to ensure optimal conditions for comparing the influence of T cell phenotype, as we showed earlier that the proportion of antigen-specific CD8⁺IFN-γ⁺ T cells in the liver at week 22 is similar in the two regimes (Figure 2d). We also tested the kinetics of peripheral blood Pb9-specific responses between the two regimes and at this time-point found lower levels in the protective Ad-MVA regime compared to the less protective MVA-Ad (Mean of 21.89%, compared to 30.16% for MVA-Ad; 95%CI [−13.69 to −2.862], p=0.0036 by t-test) (Figure 3b), making enhanced protection by the Ad-MVA regime due to higher CD8⁺IFN-γ⁺ T cell frequencies in the blood unlikely.

Figure 3. Enhanced pre-challenge T_EM numbers in peripheral blood correlate with protection.

(a) Diagram showing the experimental design. BALB/c mice (n=42) were vaccinated at weeks 0 (prime) and 8 (boost) and challenged with 1,000 P. berghei sporozoites at week 22. (b) Relative proportion of Pb9-specific IFNγ CD8⁺ T cell responses in peripheral blood, following prime-boost vaccination. (c) Distribution of the T_E, T_EM and T_CM subsets in blood following Ad-MVA and MVA-Ad vaccination regimes, measured at the time of challenge. (d) Growth curves of P. berghei blood-stage parasites from patent animals; animals were sampled from day 5 onwards (onset of the disease blood stage). (e) Protection against P. berghei disease by the two vaccination regimes, measured as time taken to reach 0.5% parasitaemia. (f, g, h) Survival curves for animals with the highest (red), middle (blue) and lowest (green) tertile of each cell phenotype: (f) T effector memory (T_EM); (g) T effector (T_E); (h) T central memory (T_CM).

In an analysis of the relative sizes of the T_E, T_EM and T_CM subsets in peripheral blood at the time of malaria challenge, we found that, similar to our observations in the spleen, the Ad-MVA regime induced significantly higher T_EM cell proportions than MVA-Ad (54.4% compared with 47.1% for MVA-Ad, p=0.003). T_E proportions were significantly lower in the Ad-MVA regime (p=0.02, t-test) while the T_CM subsets were similar (Figure 3c).

Following sporozoite challenge, the mice were sampled on days 5, 6 and 7 for the analysis of parasitaemia. These three-day serial blood counts were available on all animals (n=35). 5 animals showed no parasitaemia at any time point and therefore growth curves for the 30 parasitaemic animals are shown. As expected for a vaccine with no impact on blood stage growth, the parasitaemias exhibited exponential blood stage expansion with similar growth rates (Figure 3d). As a measure of disease outcome, we used the time taken to reach 0.5% parasitaemia, which reflects parasite numbers erupting from the liver, a recognized measure for evaluating pre-erythrocytic vaccine efficacy(23). The Ad-MVA regime showed a trend towards an enhanced protection compared to the MVA-Ad, which was not significant at this time point (Figure 3e). Next, having available several parameters calculated in blood samples from every mouse, we used a mathematical model to investigate whether pre-challenge T_E, T_CM and T_EM proportions in blood correlate with protection against liver-stage malaria. After plotting the highest, middle and lowest tertiles for each of the cell subsets, we found a significant difference in survival for the T_EM cell subset (test for equality of survival between groups, χ²=7.44, p=0.02), with the lowest T_EM numbers associated with lowest survival (Figure 3e). By contrast, neither T_E nor T_CM tertiles were significantly associated with survival (test for equality of survival, χ²=2.59, p=0.27 and χ²=3.68, p=0.14, respectively.)

In conclusion, using a mathematical model to calculate parasite eruption from the liver, we found that CD8⁺IFN-γ⁺ T_EM cells in blood are a correlate of protection against malaria sporozoite infection.

Parasite eruption from hepatocytes coincides with a CD8⁺ T cell phenotype change in the liver

The phenotype of antigen-specific CD8⁺ lymphocytes in the liver of the vaccinated animals was analyzed before and on day 7 after challenge (Figure 4a). At both time-points, fewer than 5% of all antigen-experienced CD8⁺ cells had the T_CM phenotype. However, a striking difference was observed between the relative proportions of the T_EM and T_E subsets before and after challenge (p=0.001 for both comparisons, Student’s t-test): At the time of challenge, 68.2% of the Pb9-specific CD8⁺ T cells in the liver were from the T_EM subset, whereas 7 days after challenge the majority (65.1%) of the Pb9 specific CD8 T cells displayed the T_E phenotype (Figure 4a). The causes of the altered phenotype are unclear, but probably include a combination of proliferation and differentiation of T_EM into T_E cells(20) and an influx of T_E cells into the liver.

Figure 4. Liver CD8⁺ T cell phenotype change correlates with parasite eruption from liver.

(a) Phenotype of the liver antigen-specific CD8 lymphocytes in vaccinated animals prior to, and 7 days post, challenge. (b) Association between liver CD8 T cell phenotype and parasite eruption from the liver, calculated as described in Methods.

To better understand the effect of parasite burden on the CD8⁺ T cell phenotype, we examined the relationship between the proportion of hepatic CD8⁺ T_EM cells 7 days after challenge and the number of parasites that erupt from the liver, estimated by linear regression analysis on sequential blood counts (see Materials and Methods). Interestingly, we found that the proportion of liver T_EM cells decreased in challenged mice when compared to unchallenged mice (Figure 4b). Furthermore, we noted that this phenotype switch was lowest in mice without parasitaemia (i.e. high proportion of T_EM), and increased with the rising number of parasites escaping from the liver, compatible with parasite burden driving the change of CD8⁺ phenotype in the liver.

CD8 T_EM protect against sporozoite challenge

We adoptively transferred CD8⁺ T_EM and T_E cells into naïve mice and evaluated protection against malaria challenge. The donor mice were immunised using the Ad-MVA regime with an 8 week prime-boost interval and two weeks after the last immunization PBMCs were isolated, stained with anti-CD8, -CD127 and -CD62L to be sorted in three populations: T_CM (CD62L⁺, CD127⁺), T_EM (CD62L⁻, CD127⁺) and T_E (CD62L⁻, CD127⁻). Part of the PBMC sample was stained with the H-2K^d Pb9 tetramer in addition to the phenotypic markers, in order to quantify the Ag-specific T cells and ensure identical absolute numbers of Pb9-specific cells were transferred for both T_EM and T_E cell populations. The harvested donor samples enabled transfer of 43,300 Ag-specific cells per mouse into 3 recipient animals for both the T_EM and T_E subsets (Figure 5a). Due to very low number of T_CM in the harvested donor sample it was not possible to assess their protective efficacy. However, our earlier results suggested that this CD8⁺ T cell subset plays no role in protection against malaria in this model (Figure 3h). Nevertheless, we would not rule out the potential of the T_CM cells to contribute to protection as this phenotypic cell possess the best proliferative ability and has been shown to protect mice in certain disease models, such as infection with LCMV that replicates in lymphoid organs(20). Therefore, it will be important in the future to find an immunization approach that induces high levels of T_CM to allow a cell transfer and determine their protective efficacy against pre-erythrocytic malaria.

Figure 5. CD8⁺ T effector memory cells protect against malaria upon transfer into naive recipient mice.

(a) BALB/c mice (n=15) were immunised with the Ad prime - MVA boost regimen 8 weeks apart. Mice were terminally bled two weeks after the last immunization, blood PBMCs were stained with anti-CD8, CD127 and CD62L and populations of T_E, T_EM and T_CM sorted on a MoFlo sorter. A separate sample was used to quantify Pb9-tetramer positive cells in each subset and the number of transferred cells adjusted to normalise the number of Pb9-specific CD8⁺ cells in each recipient mouse. The T_CM cells were not transferred due to an insufficiently low yield. (b) Growth curves of P. berghei blood-stage parasites from patent animals; animals were sampled at the onset of the blood stage. Parasite growth was exponential and similar in all mice sampled. (c) Kaplan-Meier survival analysis comparing the growth rates between the naïve controls and the mice receiving the T_E and T_EM cells.

The recipient mice were challenged with malaria at the time of the transfer (1,000 sporozoites per mouse, administered i.v.). Mice were screened for development of parasitaemia in peripheral blood on days 5, 6 and 7 after challenge and counts analyzed as described above. Parasite growth curves in each recipient mouse confirmed similar growth rates for all parasites (Figure 5b) and were used to calculate the initial parasite eruption from the liver. Using a Kaplan-Meier survival analysis, the protective efficacy for each transferred population was assessed using the proportion of mice not reaching a defined parasitaemia (Figure 5c). Survival was enhanced in the T_EM group (T_EM vs others: χ²=3.90, p=0.048). From these results, we can conclude that the previously observed association between T_EM cells and survival (Figure 3 f, g, h) is causal and that protection against malaria is not mediated by T_E cells. Nevertheless, with so few T_CM cells induced by our approach, so far we cannot rule out the contribution of these cells to protection against pre-erythrocytic malaria and further studies will be required to assess this.

Validation of the mathematical model using real-time in vivo imaging of luciferase transgenic parasites

Mouse models permit to assess the parasite burden directly in liver after a sporozoite challenge while quantitation of blood-stage parasitaemia is usually considered a surrogate endpoint for estimating the efficacy of pre-erythrocytic vaccines. The most adequate technique to accurately assess liver parasite burden in rodent models is the quantitative real-time PCR (qPCR)(24). We validated our mathematical model by correlating blood counts with liver parasite burden using an in vivo imaging system to visualize the parasite in the liver. Our mathematical model requires blood samples taken on days 5 to 7 after challenge and a correlation with liver qPCR is incompatible due to the necessity to sacrifice the mice to take the liver 2 days post-challenge. Recently, transgenic P. berghei parasites expressing the reporter gene luciferase have been used to visualize and quantify the parasite development in the mouse liver using real-time luminescence imaging and this technique has been shown to correlate very well with established quantitative RT-PCR methods(25). We used this non-invasive technique to quantify the liver parasite burden at 40h post-challenge while maintaining the mice alive for further quantification of parasites in blood on days 5 to 7 thus permitting a correlation to validate our mathematical model. We visualized and quantified the parasites by challenging naive and vaccinated mice with luciferase transgenic sporozoites to induce different levels of protection. Vaccinated mice showed low levels of luciferase expression (Figure 6a) while higher expression levels were observed in mock-vaccinated control mice (Figure 6b). Luciferase expression correlated well with parasite blood counts on day 5 (R=0.78, p=0.01) (Figure 6c) and an even better correlation was observed using the time required to reach a 0.5% parasitaemia, a variable calculated by a linear regression using % of parasites in blood on day 5, 6 and 7 (R=−0.79, p=0.008) (Figure 6d).

Figure 6. Validation of the statistical model: correlation of liver parasite burden with % parasites in blood.

(a) C57BL/6 mice (n=10/group) were vaccinated with (a) an adenovirus expressing the P. berghei CS (AdC63 PbCS) or mock-vaccinated with (b) an empty AdC63. Two weeks later, both groups were challenged with transgenic lucP. berghei sporozoites. Parasite burden was assessed by detecting the luciferase signal in liver after 40h and by blood screening for the presence of parasites from day 4 to 7 after challenge. The image shows an overlay of grayscale photo and luciferase bioluminescence image of 20 mice. Parasite counts on day 5 (c) and time to reach 0.5% parasitaemia (d) were correlated with the bioluminescence signal expressed as a total flux of photons per second of imaging time.

These results permitted a validation of the mathematical model that we have used using a real-time in vivo imaging system that permits quantification of the liver parasite burden without the requirement of qPCR.

Discussion

We show here that antigen-specific CD8⁺ T effector memory cells (T_EM) constitute a correlate of protection against liver-stage malaria that can be measured in peripheral blood and liver. Moreover, we demonstrate a preferential induction of a T_EM after immunization with adenovirus-MVA regime, which can efficiently deploy CD8⁺ T cells to the liver, the first potential target-organ for sporozoite arrest and prevention of their egress into the bloodstream(26, 27). Our results suggest that the induction of the protective phenotype can be tailored by the choice of the viral vector used to prime or boost responses, since MVA tends to induce fast formation of the central T cell memory response while the persistent nature of adenoviral antigen expression(18) drives the adaptive immune system to maintain a T_E/T_EM response over a longer period of time.

It has been demonstrated that prolonged antigen presentation is required for the development of protective CD8⁺ T cell responses in a mouse malaria liver-stage model(15). In fact, this is considered an important mechanism behind the efficacy of the immunization with irradiated sporozoites in protection against liver-stage malaria(15). Cockburn et al. showed that prolonged antigen-presentation of malaria liver-stage antigens can be beneficial for protection as it increases the magnitude of the memory cell compartment and promotes development of T cells that can replicate upon further encounter with the pathogen in a challenge(15).

Viral vectored vaccines are a leading vaccination strategy against malaria, and various regimens have been successfully tested in clinical trials, demonstrating various degrees of efficacy(3-5, 28). Little is known about either optimal routes for their administration, or the kinetics of in vivo antigen expression and how this can be used to maximize the generation of T cell responses and protection. Using a real-time in vivo photon imaging technique, Giben-Lynn et al. showed a long-term expression of an adenovirus-delivered transgene. In contrast, transgene expression by recombinant vaccinia was lost after 4 days and MVA was undetectable by day 2(19). We have in the past established the concept of persisting vaccines by using replication-deficient adenoviral vectors of both human and simian origin(18). In that study, we demonstrated that adenoviral vectors have the ability to persist in vivo, similar to a replication-competent adenovirus(18).

All these previous observations are important to understand the results described here regarding the phenotypic changes of CD8⁺ T cells induced by either adenoviral vector or MVA. While a single immunization with Ad vector induces predominantly CD8⁺ T_E and T_EM phenotypes, MVA rapidly generates a T_CM phenotype, possibly as a consequence of the limited antigen availability/expression by the poxvirus. Moreover, responses initially elicited by adenovirus give rise preferentially to a T_EM population following an MVA boost. Upon antigen encounter, the balance of these populations is determined by antigen availability and time. If the antigen is available the cells further differentiate into T_E, or in the absence of antigen, they become T_CM(20).

The three cell subsets efficiently produce effector cytokines, such as IFN-γ, but both T_E and T_EM cells are more efficient in protecting against peripheral infection(20). This has been explained by the distinct anatomical distribution of these three populations. For instance, CD62L⁻ (T_E and T_EM) cells have the ability to enter peripheral tissues, such as the ovaries(20) or the liver, and confer protection at the time of vaccinia challenge. However, CD62L⁺ T_CM cells primarily home to secondary lymphoid tissues where they would need to be re-activated in order to be able to migrate to peripheral tissues(20, 29).

For malaria sporozoite challenge, a time-delay in reaching the liver can be a fundamental factor in determining protection against liver-stage parasites. Hepatocyte infection represents a critical stage where the parasite is susceptible to recognition and elimination by CD8⁺ T cells. Its duration varies among different hosts and parasite strains, from approximately two days for rodent strains to 5-7 days for P. falciparum in humans(27). With such a short time to fight the parasite, it is easy to infer that previous or immediate presence of Ag-specific CD8⁺ T cells in the liver is necessary to protect against the disease. As shown by our results, the protective Ad-MVA regime induces a preferential CD8⁺ T_EM response and this translates into high frequencies of liver CD8⁺ T cells. Using immunization with radiation-attenuated sporozoites (RAS), Schmidt et al. recently demonstrated in mice that resistance or susceptibility to sporozoite infection does not result from differences in CD8⁺ frequencies in the host but is instead related to T_CM and T_EM phenotypic differences among different mouse strains(30). Our results provide an explanation for these observations, as our model establishes a correlation with protection of T_EM cells and a lack of correlation with either T_E or T_CM CD8⁺ cells. The study by Schmidt et al., however, define T_EM cells by the lack of expression of CD62L in secondary CD8 cells long after the immunization, without taking into consideration the co-expression of CD127. We have shown that such CD62L^−/lo population can consist of a mixture of T_EM/T_E cells that can be defined by the expression or lack of expression of CD127, respectively. These two phenotypes, as shown in our study, possess different abilities to protect, with only T_EM (CD62L⁻CD127⁺) responsible for protection.

Our study provides the first direct and compelling demonstration of CD8⁺ T_EM as mediators of protection against pre-erythrocytic malaria and a vaccine correlate of protection measurable in blood. It also raises further questions about the mechanism by which the CD8⁺ T_EM operate to induce protection, as well as a potential involvement of T_EM cells in infection by other hepatic intracellular pathogens, such as viral hepatitis. Our results open a route to logical vaccine design, suggesting that other regimes inducing potent, persistent CD8⁺ T_EM populations will induce protection against this important disease.

Materials and Methods

Mice and immunizations

Female BALB/c mice 6 to 8 weeks of age were purchased from Harlan, UK. All animals and procedures were used in accordance with the terms of the UK Home Office Animals Act Project License. Procedures were approved by the University of Oxford Animal Care and Ethical Review Committee. Viral vectors were administered intradermally in endotoxin-free PBS [Sigma] at a concentration of 1×10⁶ for MVA.ME.TRAP (MVA) and 5×10⁹ vp for AdC9.ME.TRAP (AdC9).

Viral vectors

Vectors expressing the transgene ME.TRAP have been previously described(31, 32). The insert ME.TRAP is a hybrid transgene of 2,398 bp encoding a protein of 789 amino acids. The ME string contains the BALB/c H-2K^d epitope Pb9 amongst a number of other human B- and T-cell epitopes(33). AdC9 was constructed and propagated as described previously(34). AdCh63PbCSP was constructed by cloning the full length CS gene from P. berghei ANKA strain which was synthesized by Geneart (Germany). The gene was codon-optimized for expression in mammalian system. To facilitate both, cloning into the adeno shuttle vector and gene translation within mammalian cells, the gene at 5′ and 3′ ends was flanked with Acc65I restriction site followed by the KOZAK sequence, and 2 stop codons followed by NotI site, repectively. Expression of the gene within the AdCh63 backbone was driven by a long CMV promoter containing a tet-repression system cassette.

Cell staining and Flow cytometry

For intracellular cytokine staining (ICS), ACK buffer treated splenocytes were incubated for 5 hours in the presence of 1 μg/ml Pb9 and 1μl/ml Golgi-Plug (BD). To assess the hepatic T cell responses, livers were initially perfused with PBS (Sigma) to eliminate the circulating blood. T-cells were then isolated from livers by mechanical disruption and incubation for 1 hour at 37°C in FCS-free MEM media supplemented with Glutamine (4mM) and Penicillin/Streptomycin (100U penicillin/100μg streptomycin), containing DNase at a final concentration of 0.03 mg/ml (Sigma, UK) and collagenase at 0.7 mg/ml (Sigma, UK). The reaction was stopped using MEM with 10% FCS and after washing, mononuclear cells were purified with Ficoll-Paque Premium (GE Healthcare, UK) and stimulated as described above.

Phenotypic analysis of CD8⁺ T cells was performed by intracellular cytokine staining (ICS) using previously described antibody clones(17) (eBioscience, UK), specifically anti-CD8 (clone 53-6.7), anti-IFN-γ (clone XMG1.2), anti-CD127 (clone A7R34). Non-specific binding of antibodies was prevented by incubating with anti-CD16/CD32 Fcγ III/II Receptor (2.4G2, BD/Pharmingen) prior to staining. When using anti-CD62L (clone MEL-14, eBioscience, UK), stimulated cells were incubated with TAPI-2 peptide (Peptides International, USA) at a final concentration of 250μM to prevent CD62L shedding from the cell surface. The Pb9 tetramer was produced by the NIH tetramer facility (MHC tetramer core facility, Emory University Vaccine Center, Atlanta, USA) using the peptide SYIPSAEKI (Proimmune, UK). Flow cytometric analyses were performed using a FACSCanto and LSRII (BD Biosciences). Data were analyzed with either FACSDiva (BD) or Flow Jo (Tree Star) software. Analysis of multifunctional CD8⁺ T-cell responses was performed using a Boolean analysis in FlowJo software, Pestle and SPICE 4.0 kindly provided by M. Roederer (NIH, Bethesda). Cell sorting was performed using a MoFlo High speed Sorter (DAKO) after staining cells with anti-CD8, CD127 and CD62L antibodies.

Adoptive cell transfer

For adoptive transfer of T_E and T_EM cells into naive recipient mice, donor BALB/c mice (n=15) were immunized with AdC9 ME.TRAP (5×10⁹ vp) and then boosted with MVA ME.TRAP (1×10⁶ pfu) eight weeks later. Mice were bled a week after the final immunization and T cells sorted using a MoFlo cell sorter (DAKO-Cytomation) after staining with anti-CD8 FITC; anti-CD62L PE and anti-CD127 PE-Cy7 using similar clones to those previously described(17) to obtain three separate populations: CD8 T_E, T_EM and T_CM. A separate staining with the Pb9 tetramer was used to calculate the total number of T_E, T_EM and T_CM in every sorted subpopulation and these were adjusted to contain equal numbers of Pb9-specific cells. A total of 43,300 Pb9-specific either T_E or T_EM cells were transferred to individual mice (n=3). The yield of T_CM cells was very low and it was technically not possible to transfer them into recipient mice. Cells were transferred immediately after challenging mice with 1,000 sporozoites. Blood smears were stained with Giemsa on days 5, 6 and 7 after challenge and percentage of parasitaemia calculated in these three consecutive samples.

Parasite challenge

Plasmodium berghei (ANKA strain clone 234) sporozoites (spz) were isolated from salivary glands of female Anopheles stephensi mosquitoes. Parasites were resuspended in RPMI-1640 media with each mouse receiving a total of 1,000 spz intravenously. Blood samples were taken daily from day 5 to day 20; blood smears were stained with Giemsa and observed under a light microscope for the presence of parasites within the red blood cells. In Table 1, survival was defined as complete absence of parasites in blood. For investigation of a correlate of protection, a Kaplan-Meier analysis was conducted to compare the parasite growth rate and protection was measured as a delay in reaching 0.5% parasitaemia, as described below.

The transgenic P. berghei parasites (PbGFP-Luc_con) used in the study expressed fusion GFP (mutant 3) and firefly luciferase (LucIAV) genes under the control of constitutive EF1a promoter (35). The parasites were generated by a stable double-crossover homologous integration of the transgene into P230p locus in the reference line of P. berghei ANKA line cl15cy1. The transgenic parasites were kindly provided by Dr. Oliver Billker from Wellcome Trust Sanger Institute, Hinxton, UK.

In vivo imaging after challenge

Bioluminescent luciferase signal was detected through imaging the whole animals using the in vivo IVIS 200 imaging system (Caliper Life Sciences, USA) as described before (25). Briefly, 44 hrs after the intravenous injection with 2000 transgenic P. berghei sporozoites the C57BL/c mice were anaesthetized in batches of three using isofluorane, bellies shaved and D-luciferin (Synchem Laborgemeinschaft OHG, Germany) was injected into the neck at a concentration of 100 mg/kg using sterile PBS (Sigma, US) as a diluting agent. Animals were imaged for 120 seconds at binning value of 8 and FVO of 12.8 cm, 8 minutes after the injection of D-luciferin. Mice were kept anaesthetized throughout the whole procedure. Quantification of bioluminescence signal was performed using Living Image 4.2 software (Caliper Life Sciences, USA). ROI were set around the liver area of mouse body and kept constant for all of the animals. The measurements were expressed as a total flux of photons per second of imaging time.

Statistical model for protection

Relationships between CD8 phenotype and protection were computed after challenging mice with 1,000 Plasmodium berghei sporozoites. Blood parasite counts were obtained every day for 3 days from day 5 after challenge, blood smears stained with Giemsa, and percentages of parasitaemia calculated in all animals. Relationships between log (percent parasitaemia) and time after challenge were plotted for mice developing parasitaemia. Potential influence of vaccine on blood stage growth was assessed visually. As expected for a vaccine containing only pre-erythrocytic antigens, all infected mice exhibited similar exponential blood stage growth regardless of vaccination regime (see Results), however, not all mice became parasitaemic. We used survival analysis to assess vaccine efficacy, using time to 0.5% parasitaemia (although any other level of parasitaemia could also be used with equivalent results) as an outcome. This approach has been previously used; since time to parasitaemia reflects number of parasites erupting from the liver provided there is no blood stage immunity efficacy(23). Differences between strata were assessed using log-rank or trend tests. We considered the ‘time at risk’ to start on day 5 (when counting started) and to end on day 7 (when counting stopped). We obtained maximal-likelihood estimates of time each mouse reached 0.5%parasitaemia by modelling Log(b) = kt+ c_i where k is the growth rate (which is assumed to be constant for all mice, and estimated from the data), t is the time following eruption from the liver and c_i is an intercept for each mouse, and is proportional to the number of parasites erupting from the liver. Statistical analyses to determine differences in protection after prime-boost regimes (table 1) were performed using a Kaplan-Meier survival plot and survival curves were compared using the log-rank test in Prism 5 (GraphPad software). Survival was considered as the complete absence of parasites in blood.

Statistical analysis of cell phenotypes

We compared proportions of cells expressing a particular phenotype after a single immunization using either unpaired homoscedastic t-tests (for example, T effector phenotype) for two groups or using ANOVA and a Bonferroni post-test. Exceptions (described in Results) were small populations with highly non-Gaussian distribution (as assessed visually, or with Shapiro-Wilk testing), where we used the Mann-Whitney U non-parametric tests. Analyses used GraphPad Prism or Stata 9 software.