Prime-and-Trap Malaria Vaccination To Generate Protective CD8⁺ Liver-Resident Memory T Cells

This article is featured in In This Issue, p.1811

Plasmodium parasites cause ∼250 million malaria infections and 429,000 deaths annually, mostly in children under 5 years old (1). A long-lived protective vaccine is urgently needed to reduce morbidity and mortality in endemic regions. Vaccination against the pre-erythrocytic (PE) stage is a particularly attractive approach because the number of Plasmodium-infected cells is relatively low compared with other life cycle stages. Complete protection at the PE stage would directly benefit vaccinated persons (e.g., no disease) and would aid elimination efforts (e.g., no transmission potential). The PE stage commences with the bite of an infected Anopheles mosquito, which releases sporozoites (spz) into the skin and blood, where they migrate to the liver to infect hepatocytes. In hepatocytes, exoerythrocytic forms develop progeny called merozoites that exit the liver and infect erythrocytes. The Plasmodium falciparum PE stage completes in ∼6.5 d (2, 3), whereas the PE stage of mouse-infecting species (P. yoelii, Plasmodium berghei) completes in 2 d (4).

Repeated i.v. immunizations with radiation-attenuated spz (RAS) reliably protect mice, monkeys, and humans against subsequent spz challenge (5–10). Mechanistically, both Abs and CD8⁺ CTL contribute to protection at the PE stage by blocking invasion (11) and by killing infected hepatocytes (12), respectively. In experimental models, either means can be adequate to achieve complete PE protection if responses are sufficiently robust [e.g., expression of mAb (13) or transfer of specific cloned T cells (14)]. However, CD8⁺ T cell responses are emerging as the most important factor in long-term protection following RAS immunization. CD8⁺ T cell responses are essential for RAS-induced protection in mice (15, 16) and primates (17), whereas RAS-induced Abs are not likely responsible for long-term protection against challenge in humans (18).

Recent data suggest that long-term PE stage protection results from liver-targeted CD8⁺ T cells called tissue-resident memory CD8⁺ T (Trm) cells (10, 18–20). Because Trm cells reside in the experimentally inaccessible liver, they cannot be easily measured in human volunteers. Instead, nonhuman primate studies have been used to measure their frequency and showed that such cells were 10–100 times more abundant in the liver than in PBMCs (18). This finding helped explain why RAS-vaccinated human volunteers with low-frequency, spz-specific CD8⁺ T cells in PBMCs (<0.5% of total) go on to be protected from challenge 1 year postvaccination (when spz-specific Ab titers were low and nonprotective) (18).

Tissue-resident memory T cells are increasingly appreciated in a wide array of organs, including the lung (21), liver (19, 22), CNS (23), gastrointestinal tract (24), and genital tract (25). Positioned at the site of infection, Trm cells are able to act immediately, rapidly clearing nearby infected cells and initiating a larger immune response. Because they are not easily or routinely measured, it is unknown which vaccination methods effectively induce liver Trm cell formation, and it is possible that widely used PBMC- and spleen-based immunogenicity measurements may not reflect Trm cell frequencies or correlate with protection. Indeed, non-spz vaccines that achieve high-frequency CTL in peripheral blood often fail to completely protect against the Plasmodium PE stage (26, 27), suggesting that a tissue-specific approach is needed to initiate liver Trm cell formation.

In this study, we sought to design a vaccine using components known to be safe in humans. Working on the basis of the “prime-and-trap” vaccine strategy (19), we demonstrate that complete PE stage protection can be achieved if circulating spz-specific CD8⁺ T cells are repositioned to the liver using a single dose of RAS. In this heterologous prime-and-trap regimen, spz-specific CD8⁺ T cells are primed using gene gun–delivered DNA and boosted with RAS to attract spz-specific CD8⁺ T cells to the liver, where they become liver Trm cells. Vaccinated in this manner, BALB/c mice were completely and reliably protected against challenge.

Female BALB/cj mice (4–6 wk old) were obtained from The Jackson Laboratory (Bar Harbor, ME), housed in an Institutional Animal Care and Use Committee–approved animal facility at the University of Washington, and used under an approved Institutional Animal Care and Use Committee protocol.

The P. yoelii circumsporozoite protein (CSP)–minigene DNA vaccine encoding the known MHC epitope for P. yoelii CSP, SYVPSAEQI (full minigene-encoded peptide: GNNNNNGNNNEIDFSYVPSAEQIGLSEHQELPCN), was codon optimized for mice and constructed in the NTC9384R vector (Nature Technology, Lincoln, NE) with an N-terminal ubiquitin tag to increase the likelihood of class I MHC presentation downstream of proteasome degradation (28). For all vaccinations, a plasmid encoding the Escherichia coli heat-labile lymphotoxin (LT) (29) was used as an adjuvant in a 1:10 ratio with the CSP-minigene vector. DNA was purified using an endotoxin-free purification kit (Qiagen), loaded onto gold beads (1–2-μm diameter; InBio Gold), and coated on tubing as cartridges for helium gas–propelled bombardment via gene gun on trimmed abdominal skin as described previously (30). Mice were vaccinated using a PowderJect-style gene gun by cluster priming for the first DNA vaccination using two cartridges (0.5 μg DNA per cartridge), each 2 d apart as shown in the figures (hereafter called “cluster primed” and denoted in figures with a “⁺”), and were boosted a single day later as indicated using two cartridges. This overall method of administering P. yoelii CSP-minigene/LT-encoding plasmids via gene gun is referred to as ggCSP. The P. yoelii malate dehydrogenase (MDH) and P. yoelii L3 ribosomal protein (L3) gene gun vaccines were constructed and administered identically to ggCSP (codon-optmized minigenes with relevant CD8 epitopes). In the text and figures, these are abbreviated ggMDH and ggL3 and generally as ggDNA.

Hydrodynamic transfection (HDT) of DNA by large-volume tail vein injection was performed under isoflurane anesthetic and adapted from published protocols (31, 32). Briefly, mice were weighed, and 10% v/w PBS containing 20 μg of endotoxin-free plasmid DNA (Qiagen) was injected into the tail vein in 5–10 s using a 27.5-gauge needle. HDT was performed 5 d after the last gene gun DNA administration as described in the text and Fig. 3. Plasmids for HDT were constructed using a modified pcDNA3.1(+) (Addgene) DNA backbone [removed SV40 Poly(A) signal and G418/kanamycin resistance cassettes] (SM014 plasmid). RE9h red-shifted luciferase (luc; kindly provided by Dr. B. Branchini, Connecticut College), a more tissue-penetrating luc, was inserted in the multiple-cloning site downstream of the CMV promoter of SM014 for the control plasmid (SM022 luc plasmid). This plasmid was additionally modified to include the P. yoelii CSP-minigene, as in the gene gun DNA-vaccine above. The P. yoelii CSP-minigene was linked to a ribosomal skipping element, T2A (SM017 csp-luc plasmid). The T2A element allows the ribosome to translate the P. yoelii CSP-minigene up to that sequence and then cleave the transcript, liberating the peptide chain, and continue translation of luc, thus translating two distinct products (33).

Female Anopheles stephensi mosquitoes infected with wild-type P. yoelii 17XNL were reared at the Center for Infectious Disease Research. Spz were freshly obtained by salivary gland dissection followed by gradient purification (34). RAS were generated from wild-type P. yoelii by x-ray exposure (10,000 rad; Rad Source Technologies, Suwanee, GA). All spz were administered i.v. in a volume of 100 μl.

Liver lymphocytes were isolated by mechanical dissociation and Percoll density gradient, adapted from (35). Briefly, livers were perfused with 10 ml PBS with 2 mM EDTA by injection into the portal vein, with outlet drainage via the inferior vena cava. Gall bladder was removed, and livers were placed in 5 ml RPMI 1640 supplemented with glutamine with 5% FBS on ice to ensure cell survival. Livers were mashed through a 200-μm mesh filter (pluriSelect, San Diego, CA) with the back of a 3-ml syringe plunger. The mesh filter and plunger were washed with FBS-/glutamine-supplemented RPMI 1640. Cell suspension was spun at 80 × g for 1 min at 4°C without braking; supernatants were collected and transferred to a clean 50-ml conical tube where they were spun at 500 × g for 8 min at 4°C. The cell pellet was resuspended in 10 ml room temperature 35% Percoll (GE Healthcare Life Sciences) in HBSS (Life Technologies) supplemented with 100 U heparin and spun at room temperature at 900 × g for 25 min with no brake. The final cell pellet containing intrahepatic lymphocytes was resuspended in 2 ml ammonium-chloride-potassium lysis buffer for 2–3 min, quenched with 8 ml MACS buffer (PBS, 1 mM EDTA, 0.5% FBS), and then spun at 450 × g at 4°C for 5 min. Final pellets were resuspended in 100 μl MACS buffer and moved to a 96-well plate for treatment with an Fc block for 30 min (anti-CD16/32, clone 2.4G2; BD Biosciences), Ab staining for 45 min (see Ab materials listed below), fixation for 20 min (Cytofix/Cytoperm reagent; BD Biosciences), and analysis by flow cytometry on an LSR II (BD Biosciences).

The following Abs were used to assess liver Trm cells: CD3e-BUV395 (clone 145-2C11; BD Biosciences), B220-BV711 (clone RA3-6B2; BioLegend), CD4–Alexa Fluor 700 (clone GK1.5; BioLegend), CD8α-BV421 (clone 53-607; BD Biosciences), CD69-BV510 (clone H1.2F3; BD Biosciences), CD44–Alexa Fluor 488 (clone IM7; BioLegend), CD62L-PE-Cy7 (clone MEL-14; BD Biosciences), KLRG1-PerCP-Cy5.5 (clone 2F1/KLRG1; BioLegend), CXCR6-PE (clone 221002; R&D Systems), and CSP tetramer (CSP epitope SYVPSAEQI provided by the National Institutes of Health Tetramer Core) conjugated to streptavidin-allophycocyanin (ProZyme) per standard protocols. Cells were gated for CD8⁺ T cells (CD3e⁺, B220⁻, CD4⁻), CD44^hi by CD62L^lo and then assessed as liver Trm cells by either KLRG1^lo by CD69⁺ or by CXCR6⁺ by CD69⁺. Ag specificity was then assessed by CSP tetramer. Cell count per gram of tissue was calculated based on a known concentration of counting beads per sample (product 18328-5; PolyScience, Warrington, PA) to normalize data.

Splenocytes were purified as above, and 1 × 10⁶ cells were incubated at 37°C with SYVPSAEQI peptide (10 μM final) (Genemed Synthesis, San Antonio, TX) for 1 h. Brefeldin A was added (BD Biosciences), and cells were incubated for an additional 4 h. Cells were washed; resuspended in Fc block for 30 min; stained with CD3e-BUV395 (clone 145-2C11; BD Biosciences), B220-BV711 (clone RA3-6B2; BioLegend), CD4-allophycocyanin (clone GK1.5; BioLegend), and CD8α-BV421 (clone 53-607; BD Biosciences) for 45 min; washed; and fixed/permeabilized using Cytofix/Cytoperm kit (BD Biosciences). Cells were resuspended in Perm/Wash Buffer (BD Biosciences) containing 1:100 IFN-γ-allophycocyanin-Cy7 (clone XMG1.2; BioLegend) and incubated on ice in the dark for 1 h. Cells were washed and analyzed by flow cytometry using an LSR II (BD Biosciences).

For ELISPOTs, peptides (1 μg/ml final) (Genemed Synthesis) were combined with 1 × 10⁵ murine splenocytes by murine IFN-γ ELISPOT (eBioscience), cultured for 18 h at 37°C as reported previously (36), and developed following manufacturer guidelines. The Percentage of CD8⁺ activated T cells was calculated based on the spot-forming units counted in each well divided by the total number of splenocytes applied to each well.

Luc-based in vivo imaging of liver burden was performed as described previously (37). Briefly, mice were gene gun immunized, and 5 d after boosting, they were infected with 2 × 10⁴ luc-expressing purified P. yoelii spz. Luc luminescence was evaluated 48 h later by s.c. administration of luciferin (30 mg/ml, 100 μl/mouse; Gold Biotechnology, St. Louis, MO) followed by isoflurane anesthesia and IVIS imaging. Thin blood smears were taken by tail vein bleeds, stained with Giemsa, and evaluated for patent parasitemia. Sterile protection was defined as being blood smear negative during days 5–14 after challenge with live spz.

Most comparisons of flow cytometry cell counts and IVIS liver burden assessments were done by using the nonparametric two-tailed Mann–Whitney U test. Where indicated in each figure legend, Welch t test was performed when the Mann–Whitney U test could not be performed reliably (small sample sizes). Protection was evaluated using Fisher exact test. Error bars in all figures are reported as the SD of the mean.

Initial vaccination challenge studies conducted in mice were intended to assess the efficacy of the DNA gene gun for multiantigen vaccine delivery against the PE stage. Gene gun vaccines were designed to maximize CD8⁺ T cell activation and reduce Ab responses. The P. yoelii CSP-minigene DNA vaccine encoded the known class I (H2-K^d) epitope SYVPSAEQI and was constructed in the NTC9384R plasmid with an N-terminal ubiquitin tag designed to enhance class I MHC processing, maximize CTL responses, and minimize Ab responses (38, 39). The Ag-coding plasmid was mixed with a plasmid encoding LT as adjuvant (29), and mice were cluster primed and boosted with the mixture on their abdominal skin (hereafter called ggCSP in text and figures). Ag-specific, IFN-γ–producing CD8⁺ T cells were induced at high frequencies by DNA vaccination (Fig. 1A, 1B, Supplemental Fig. 1A, 1B). When mice were challenged at an acute time point (5 d after the booster administration of ggCSP), sterile protection was achieved (Fig. 1B, Supplemental Fig. 1C, 1D). Protection was Ag-specific because high-frequency IFN-γ responses to other known P. yoelii immunogenic proteins such as L3 (36) and MDH (30) failed to protect (Fig. 1B, Supplemental Fig. 1C, 1D). However, at a memory time point (4 wk after booster), ggCSP-immunized mice were not sterilely protected from challenge despite having relatively high numbers of CSP-specific CD8⁺ T cells in their peripheral blood as determined by CSP tetramer staining (Fig. 1C). Taken together, these data suggested that high-frequency CSP-specific T cells induced by ggCSP vaccination were capable of protecting mice at an acute time point and are functional effector cells based on IFN-γ production after peptide restimulation. However, at a memory time point, these data indicated that the declining T cell frequency was inadequate to completely protect against PE infection.

Because low frequencies of parasite-specific T cells in the blood are not always a reliable indicator of protection at challenge (18), we wanted to determine the location of ggCSP-generated CD8⁺ T cells at a memory time point. We compared CSP-specific CD8⁺ T cells in spleen and Trm cell formation in the liver after DNA-only ggCSP vaccination as compared with three doses of RAS vaccination, a regimen known to be protective (40–42). CSP tetramer–stained cells were defined as Trm cells by either CD69⁺/KLRG1^lo expression or, more stringently, as CD69⁺/CXCR6⁺ expression as previously described (19) (Supplemental Fig. 2). Total CSP-specific CD8⁺ T cells in the spleens of ggCSP-vaccinated mice were much higher than in RAS-vaccinated mice (Fig. 2A). However, whereas CSP-specific liver Trm cells (both KLRG1^lo- and CXCR6⁺-marked Trm cells) were abundant in mice immunized three times with RAS, they were significantly lower in ggCSP-immunized mice (Fig. 2B, Supplemental Fig. 2A). Thus, these data suggested that the lack of protection in the DNA-only vaccine regimen stemmed from limited formation of liver Trm cells after contraction of acute T cell responses. With such high frequencies of CSP-specific CD8⁺ T cells in ggCSP-immunized mice at boosting, we sought to determine the feasibility of repositioning these CSP-specific CD8⁺ T cells in the liver so that they could act immediately upon challenge at a memory time point.

To determine if CSP-specific CD8⁺ T cells produced by ggCSP vaccination were capable of giving rise to liver Trm cells, we performed HDT on ggCSP-vaccinated mice to directly express the P. yoelii CSP epitope in hepatocytes and to assay in vivo CTL responses against Ag-expressing hepatocytes and subsequent liver Trm cell formation. In HDT, mice are anesthetized and injected in the tail vein with a volume of DNA-containing saline equal to 10% body weight over 5–10 s; the mice tolerate the procedure without incident, recover within minutes, and suffer no apparent long-term side effects (31, 43). The sudden increase in the oncotic pressure gradient drives injected plasmid DNA into hepatocytes, where plasmid-encoded genes are expressed. HDT plasmids were constructed to coexpress the P. yoelii CSP-minigene and luc linked through a T2A element (csp-luc) or luc only as a control. Mice were ggCSP immunized by either cluster priming only or cluster priming and boosting as in the previously described vaccination regimen, and HDT was performed 5 d later at the peak T cell response (Fig. 3A). When csp-luc HDT was performed in mice previously cluster primed and boosted with ggCSP, luminescence was rapidly ablated (99.2% reduction by 3 d and 99.99% by 7 d, Fig. 3B, 3C), indicating rapid killing of CSP-expressing cells. CSP DNA vaccination (by either HDT or ggCSP) triggered dose-dependent increases in CSP-specific T cell frequencies based on tetramer staining of splenocytes 28 d after the final ggCSP dose (23 d after HDT administration) (Fig. 3D). HDT expression of the CSP epitope in the liver also drove formation of liver Trm cells as predicted, which were most abundant in mice previously cluster primed and boosted with ggCSP 28 d after the final ggCSP dose, indicating a contribution of the high-frequency ggCSP-induced precursors for optimal Trm cell–lodgment in the liver (Fig. 3E). Notably, ggCSP vaccination followed by luc-only HDT failed to induce any detectable CD69⁺/CXCR6⁺ liver Trm cells. Overall, these experiments demonstrated that pre-existing ggCSP-induced T cells can reach the liver and are functionally cytotoxic for CSP-encoding DNA expressed de novo in the liver.

Synthesizing what we learned from the above experiments, we next wanted to combine the potent T cell immunogenicity of ggCSP with the liver specificity of attenuated spz. Mice were cluster primed with ggCSP and boosted 4 wk later with ggCSP combined with 5 × 10⁴ i.v. administered RAS (Fig. 4A). Control groups consisted of ggCSP-only cluster prime and booster vaccinations or a single dose of 5 × 10⁴ RAS. Responses were also assessed in naive animals. As expected, regimens with ggCSP vaccination yielded easily measurable CSP-specific T cell frequencies in the spleen and blood at 28 d after the final vaccination, whereas no detectable responses were found in those tissues in mice receiving a single dose of RAS (Fig. 4B). ggCSP alone and RAS alone yielded low but measurable CSP-specific liver Trm cell formation. However, by comparison, the combined prime-and-trap regimen yielded significantly higher frequencies of CD69⁺/KLRG1^lo liver Trm cells (Fig. 4C, top panel) and was the only condition that induced significant increases in CD69⁺/CXCR6⁺ liver Trm cells (Fig. 4C, bottom panel). The difference was also apparent for total T cells (no tetramer staining): all vaccinations favored a central memory T cell phenotype in the spleen (Supplemental Fig. 3A) and a Trm cell phenotype in the liver, especially for the prime-and-trap regimen (Supplemental Fig. 3B). When parallel groups of immunized mice were challenged with 500 wild-type spz 28 d after the final immunization, the only regimen that achieved sterile protection in all mice was the prime-and-trap approach (Fig. 4D). As expected from earlier experiments, ggCSP-only vaccination was completely nonprotective. A single dose of 5 × 10⁴ purified RAS protected most but not all animals (10 out of 13 mice), whereas lower single doses of RAS (2 × 10⁴) were less protective (2 out of 5 mice, data not shown). Our experiments rely on gradient-purified RAS, which are more immunogenic than unpurified spz because mosquito Ags are largely eliminated (37). Despite increased RAS immunogenicity, sterile protection in all animals was never seen, even for a single high dose of purified RAS, which is consistent with earlier literature reported for unpurified RAS (40, 42, 44). However, complete protection was achieved in mice by prime-and-trap vaccination using a lower trapping dose of 2 × 10⁴ RAS at the boost (data not shown). Because we thought that circulating T cells needed to be at high frequencies prior to the trapping dose, a single concurrent administration of ggCSP and RAS was not initially tested in this study as a control. However, this condition and others were addressed below when modifications to the vaccination regimen were later evaluated. Overall, this study provides a simplified two-dose vaccination regimen that significantly increases CSP-specific liver Trm cell frequencies and provides complete protection of all animals in a mouse model of malaria vaccination.

Next, a variety of modified prime–boost regimens were assessed to determine if the prime-and-trap regimen could be simplified and/or accelerated (Fig. 5A, patterned bars in all graphs are reproduced from Fig. 4 to demonstrate statistical comparisons). When the day −28/−26 ggCSP cluster priming dose was omitted and ggCSP and RAS were administered on a single day (day 0), high-frequency CSP-specific T cells were detected in the spleen a month later, as seen with all DNA vaccinations (Fig. 5B), but liver Trm cell formation was not increased significantly compared with a single dose of RAS (Fig. 5C). Accelerating the vaccination by administering ggCSP cluster priming on day 0 and 2 and RAS 5 d later to boost expanding T cells only moderately increased Trm cell formation to frequencies comparable to RAS alone. Giving a cluster prime and boost of ggCSP (on day −28, −26, and 0) and then giving RAS on day 5 was no more immunogenic (Fig. 5B, 5C) and was not protective (Fig. 5D). These data indicate that Trm cell formation may be suboptimal when a liver-specific immunogen (RAS) is provided in the setting of an activated, expanding T cell response. However, a cluster priming dose of ggCSP followed 1 mo later by 5 × 10⁴ RAS alone (no ggCSP boost) induced high-frequency Trm cells that were comparable to those seen for the completely protective prime-and-trap regimen described in Fig. 4. ggCSP cluster priming followed 1 mo later by RAS achieved sterile protection in most but not all mice (Fig. 5D). Thus, the data indicate that optimal Trm cell formation and protection following this vaccination benefits from a 4-wk interval between priming and trapping (boosting) vaccinations with RAS.

Liver CD8⁺ Trm cells were recently shown to be critical for memory-stage CD8⁺ T cell–mediated protection against PE Plasmodium infection (19), and our study provides a translational pathway for development of prime-and-trap vaccines designed to harness such Trm cells. The window for achieving complete protection is extremely short for the malaria PE stage because this stage completes and blood stage infection commences 2–3 d after spz exposure in mice and 5–6 d after spz exposure in humans. This short window of opportunity is in contrast to many other pathogens in which control by T cell responses over a matter of weeks confers a clinical benefit. The large size and minimally immunogenic nature of the liver as well as the low number of infected hepatocytes following a typical spz exposure present additional challenges for development of a successful T cell–based PE vaccine. With the discovery that liver Trm cells are critical for long-term protection against Plasmodium (10, 18, 19) and recognizing the compressed time frame of the Plasmodium liver stage, an optimal way to accelerate parasite clearance is to pre-position protective CD8⁺ T cells in the liver. We tested a DNA vaccine that primes potent CTL responses against CSP and then directed these cells to become Trm cells in the liver using attenuated parasites that naturally traffic to the liver. Such vaccinated mice were completely protected against challenge 1 mo later, a level of protection never achieved for DNA-only vaccination and not reliably achieved for one dose of attenuated spz. The prime-and-trap strategy is therefore capable of reliably clearing 100% of infected hepatocytes before the end of the short-lived PE stage.

Our approach combines two strategies that each have shortcomings of their own but is administered in a way that exploits their advantages. First, although attenuated spz vaccines induce liver Trm cells (18, 19), at this time, they can only be manufactured in mosquitoes, and they require multiple i.v. doses to achieve 100% protection in humans and rodent models (10, 44, 45). Manufacturing was historically considered to be a major obstacle, but production has progressed, and both irradiated PfSPZ and wild-type PfSPZ Challenge products made under Good Manufacturing Practice are available for injection in humans (45, 46). Vialed spz are shipped and stored on liquid nitrogen, which is suitable for use in endemic settings (47). These advances have improved scalability and use on a wider scale (48). However, the requirement for three or more i.v. doses adds complexity and prolongs the time required to complete a vaccine regimen. Some cost, delivery, and logistics obstacles could be overcome if an alternative strategy such as DNA prime-and-spz trap vaccination was developed. A reduction in doses may also improve T cell repertoire diversity. There is an apparent lack of peripheral CTL boosting after the first RAS immunization (49), and T cell frequencies at boosting are consistently lower than after the first RAS vaccination (18, 50). These observations are limited by the inability to obtain liver lymphocytes from human volunteers, but preclinical work has shown that each dose of RAS immunization reduces the liver burden of the next RAS vaccine dose, in turn reducing T cell response frequencies against de novo–expressed liver-stage Ags (36, 51). From these data, we hypothesize that the most immunogenic and effective dose of attenuated spz in naive individuals is the first dose.

With regard to DNA vaccines, the gene gun–delivered DNA-only vaccines used in this study induced high-frequency CD8⁺ T cell responses but failed to protect mice at memory time points. The lack of ggCSP protection at the memory time point was initially perplexing because ggCSP-vaccinated T cell frequencies were high in the spleen, exceeding frequencies measured in mice immunized repeatedly with RAS. The spleen is one of the most commonly tested tissues in rodent immunogenicity studies, but splenocyte-based tests proved misleading in this case. Further study showed that ggCSP vaccination failed to produce high numbers of CSP-specific liver Trm cells. The ggCSP-vaccinated animals that were protected at an acute time point were likely protected by extremely high numbers of circulating effector T cells, which then contracted after the last DNA booster. The lack of memory-stage protection by DNA-only vaccine–induced CD8⁺ T cells is consistent with the previous literature (52, 53). Although DNA vaccines for malaria have shown some success in mice (54, 55), such vaccines have historically been considered to be poorly immunogenic in humans (56–58). However, earlier studies were often limited by inefficient routes of DNA delivery (e.g., i.m. injection), and much of the earlier research was focused on Ab responses over T cell responses. Currently, newer technologies such as gene gun and electroporation devices are improving delivery (59, 60), and this has translated to increased immunogenicity in humans (61) and larger animals (62, 63). Other modifications such as codon optimization for some Ags (64), addition of genetic adjuvants (reviewed in Ref. 60), and addition of ubiquitin or other tags (39) have further enhanced DNA vaccination efficacy. Ubiquitin-tagged inserts specifically serve to maximize CD8⁺ T cells, minimize Ab responses, and provide strong protection in other immunization challenge models (39). Such modifications may be beneficial because induction of anti-CSP Ab during the priming phase is likely to reduce the percentage of RAS in the subsequent dose capable of reaching the liver at boosting. However, even with increases in effector T cell frequencies, our data suggest that DNA-only vaccination of CD8⁺ T cells fail to protect at memory time points because of inadequate formation of liver Trm cells. For the most part, prior literature has not addressed formation of liver Trm cells by DNA vaccination alone and instead has focused on more easily measured peripheral responses, which have been shown to be unreliable indicators of future protection. For instance, although LSA-1–specific, IFN-γ–producing CD8⁺ T cells were induced by electroporation-delivered DNA plasmids in livers of BALB/c mice, the T cells were not immunophenotyped further to determine if Trm cells were affected (65). In that same study, Ag-specific CD8⁺ T cells were detected in PBMCs of DNA-vaccinated rhesus macaques, but it remains unknown whether liver-specific Trm cells in rhesus macaques were induced because only PBMCs were evaluated. With the recent advances in our understanding of the contribution of Trm cells to malaria protection, some past studies will need to be re-evaluated in this light, and future studies focusing on T cell–mediated protection will need to consider liver Trm cells in efficacy assessments.

Prime-and-pull and prime-and-trap vaccines are generally heterologous prime–boost vaccines modified to direct immune cells to specific tissues. Prime-and-pull was developed to establish local tissue Trm cells at target sites by parenterally vaccinating to elicit systemic T cell responses (prime) and then to recruit activated T cells to specific tissues with a topical chemokine (66) or with nonspecific inflammation (67) (pull), where such T cells could become Trm cells and mediate protection. Methods have been reported for the liver (19), urogenital tract (66, 68, 69), and lungs (70, 71). The recent malaria prime-and-trap report showed that TCR transgenic cells adoptively transferred into naive recipient mice could be primed with peptide-pulsed DCs and then trapped with a liver-specific, Ag-expressing recombinant adeno-associated virus (19). This experimental approach increased Trm cells and led to protection against challenge. An unrecognized priming and trapping phenomenon may also explain the short-term protection achieved in an earlier paper in which mice were vaccinated with a P. yoelii CSP–expressing yellow fever virus vaccine strain followed by P. yoelii RAS boosting, although Trm cells were not specifically studied (72). In this study, we use the potent peripheral CTL-boosting abilities of the DNA vaccine with the inherent tissue-targeting properties of attenuated spz, thereby exploiting the advantages of each to overcome their shortcomings.

Although this approach was extremely successful in mice, further development may need to focus on diversifying the DNA-primed T cell repertoire beyond CSP. Whereas CSP-specific CTL are protective in the rodent model (73–75), CSP-only priming may be inadequate to achieve protection in humans because of MHC diversity. Non-CSP Ags as a class have shown the potential to mediate protection (74, 76–79). For example, vaccination against the thrombospondin-related anonymous protein/sporozoite surface protein-2 (TRAP/SSP2) induces high-frequency TRAP/SSP2-specific CD8⁺ T cells in mice (80) and humans (81) that can kill infected liver cells (82). However, beyond CSP, TRAP/SSP2, and relatively few partially protective Ags (83–90), most of the thousands of PE proteins remain unstudied as vaccine Ags. Thus, a renewed interest in discovery of protective CD8⁺ T cell Ags may be warranted. With respect to such Ag discovery, our study additionally demonstrated a novel method for rapidly evaluating candidate Ags and for making Go/No Go decisions about their protective potential. Immunogenicity testing is fraught with misleading outcomes because many Ags are immunogenic in immunized animals but are ultimately nonprotective (30, 36, 91). In the DNA-only acute challenge model described in this article (Fig. 1, Supplemental Fig. 1), when gene gun–immunized mice are challenged with spz at the peak of the effector CTL response (day 5–6 following a boost in mice), protection can be rapidly assessed without establishment of protective liver Trm cells. This Ag discovery approach uses the extremely high effector T cell frequencies attained by gene gun vaccination at this time point to enable selection of parasite proteins that are in the right place at the right time to be protective PE stage Ags. Such a method can accelerate our understanding of which proteins have truly CTL-mediated protective potential while leaving issues of vaccine optimization to maximize Trm cell formation for later studies on downselected Ags.

In this study, the initial DNA vaccination was given by cluster priming, whereby gene gun–delivered DNA was administered twice 2 d apart to rapidly expand naive CD8⁺ T cells. This limitation could be addressed by evaluating simplified vaccination regimens. Future studies will also need to address the role of pre-existing or vaccine-induced Ab formation on liver-specific trapping of T cells. In this study, the DNA vaccine vector encoded a small linear epitope linked to an ubiquitin tag to minimize Ab formation and maximize class I MHC processing. Thus, the CSP vaccine used in these studies is unlikely to induce potent Ab responses. However, future vaccine products will likely encode full- or partial-length Ags, and the role of Ab will need to be evaluated. Finally, use of this strategy in humans would require the use of cryopreserved, attenuated spz, and some of these spz will be nonviable. It will be important to evaluate the effect of nonviable parasites on tissue-specific immunogenicity and overall efficacy.

In conclusion, we have demonstrated that prime-and-trap vaccination elicits critical liver-specific Trm cells. The described approach marries the potent CD8⁺ T cell–forming capacity of DNA vaccines with the liver specificity of RAS to seed liver Trm cells that can act immediately upon challenge and contribute to sterile protection.

We thank Heather Kain, Matt Fishbaugher, Will Betz, Brandon Sack, Nana Minkah, Alexis Kaushansky, and Stefan Kappe (Center for Infectious Disease Research) for assistance and support of P. yoelii–infected mosquito production; Deb Fuller and James Fuller (University of Washington) for use of the gene gun; Bruce Branchini (Connecticut College) for the RE9h red-shifted luc plasmid; Rolendia Goshu (University of Washington) for assistance with animal care; Brian Welt (University of Washington) for project administration; and James Kublin (Fred Hutchinson Cancer Research Center) for critical review of the manuscript.

S.C.M. and B.C.S. filed a provisional patent on selected aspects of this prime-and-trap concept through the University of Washington. S.C.M. and B.C.S. have founding equity in a startup company (Sound Vaccines, Inc.) that is negotiating with the University of Washington for rights to this intellectual property. The relationship between the authors and Sound Vaccines, Inc., has been reviewed by the University of Washington and complies with all University and State of Washington policies on such activities. The other authors have no financial conflicts of interest.