A major challenge in vaccine design is stimulating the potency and duration of the immune responses.1,2 The immune responses to infection or vaccination are temporal sequences of events, which depend on the ordered exposure of antigenic components to the immune system,3,4 as well as the coordinated actions of the lymph nodes, immunocytes, cytokines, etc.5,6 Thus, potent vaccines are expected to harness spatial and temporal control over sequential immune activation.7

To address this, nano- and micro-delivery systems with controllable physicochemical properties and multi-level nanostructures are engineered to deliver multiple vaccine components.8,9 Additionally, since pathogens are the perfect vehicles of natural selection, there is a trend to mimic their structures or physiochemical properties.10,11 Increased lymph node accumulation of antigen, antigen uptake, and antigen cross-presentation have been witnessed in previous attempts to replicate live pathogens’ sizes, shapes, charges, and softness.12,13 As for the delivery kinetics, it is thought to replicate the natural dissemination of multiple antigenic components, which may dictate the exposure sequence for subsequent immune activation in a biomimetic manner.14

Nonetheless, pathogens usually evolve to escape the immune system rather than to provoke it.15,16,17 In the case of enveloped RNA viruses, genome replication results in the accumulation of pathogen-associated molecular patterns, which can lead to a strong host anti-viral response.18,19 To circumvent this, immunogenic components, such as viral genes and proteins (e.g., nucleocapsid protein, NP), are tightly bound and hidden inside.20,21 Subsequently, the embedded NP is delayed in its exposure to immune surveillance, leading to suppressed type I interferon (IFN-I) expression, as well as impeded anti-viral effects.22 Accordingly, the exact replicas of natural dissemination may not be an optimal solution. As a preliminary test, we treated bone marrow-derived dendritic cells (BMDCs) with the surface antigen and NP of H1N1 influenza virus (A/Puerto Rico/8/1934)23 and SARS-CoV-2 (hCoV-19/China/CAS-B001/2020),24 respectively. In the presence of the surface antigen, higher doses of NP resulted in the up-regulated expression of IFN-α, suggesting a robust anti-viral effect (Supplementary Fig. 1).

Under these circumstances, we anticipated that it would be the natural packaging of NPs on the inside and surface antigens on the outside, which delays the exposure of NPs to immune surveillance. Instead, the inside-out assembly of the viral antigens, which enables the exposure of the core antigens before the surface antigens (reversed delivery), may potentiate the immune responses. Compared to the exact replicas of the natural dissemination, the inside-out strategy may trigger a more robust IFN-I-mediated innate immune response in advance, cultivating an immune stimulatory environment for enhanced potency and duration of the immune responses. To this end, the delivery system is expected to offer a multi-level landing spot for the ordered and inside-out assembly of viral antigens with high loading efficiency, which may offer a tunable release at the specific location and the right time, thus dictating IFN-I signaling. Moreover, to maintain the protein structure and immunogenicity, it is also imperative to provide a facile and mild loading method to avoid the involvement of high-shear stress or organic solvents.

To achieve this, we developed a multi-layered alum-stabilized emulsion (MASE) to harness the delivery kinetics of the surface and core antigens. Through the co-assembly of alum and antigen at the oil/water (o/w) interface, the core antigen was trapped within the nanocage formed by the alum and o/w interface. Subsequently, another layer of alum was deposited, which further shielded the inner antigen and provided adsorption sites for the outer antigen. As such, the embedded antigen was only released after the detachment of the deposited alum, thus constituting the sequential delivery system. On the o/w interface, the layer-by-layer assembly may bypass the multiple encapsulation procedures and the involvement of organic reagents, assuring the epitope integrity of the proteins and the consecutive loading of surface antigen and NP in a facile and moderate way. To demonstrate the natural dissemination, surface antigen and NP were assembled consecutively on the outside and inside of the multi-layered droplets (iMASE). In contrast, the inside-out assembly reversed the delivery of surface antigen and NP (rMASE), thereby exposing the “soft spot” of the viruses (depicted as the caterpillar in Fig. 1). Consequently, it is anticipated that the inside-out strategy can cultivate the reversed encounter of the surface and core antigens to the immune system, which may strongly stimulate the anti-viral host immune responses. In this manner, IFN-I-mediated innate immune response may be activated for enhanced adaptive immune responses.

Fig. 1
figure 1

Schematic illustration of the inside-out assembly of the viral antigens. Compared with the natural dissemination (iMASE, left), the reversed delivery (rMASE, right) allowed for the prior encountering of NP to the immune surveillance, as hidden caterpillars (pathogens) were forced to expose their “soft spot” to the predators (immune systems), provoking IFN-I secretion in advance. Along with the delivered surface antigen, rMASE promoted the innate immune response for enhanced antibody secretion and antigen-specific T cell immune response against viral challenge


Tailoring multi-layer alum-stabilized emulsions for the inside-out assembly of the antigens

Here, the inside-out strategy was first tested to deliver hemagglutinin (HA) and NP of the H1N1 influenza virus (A/Puerto Rico/8/1934, Supplementary Table 1). rMASE was prepared according to the schematic illustration (Fig. 2a). By adjusting the pH and type of the continuous phase, alum and HA were co-assembled on the o/w interface, forming alum/HA-assembled droplets (Supplementary Fig. 2). Circular dichroism (CD) analysis revealed that the secondary structure of HA remained unchanged after co-assembly with alum (Supplementary Fig. 3a). Subsequently, as evidenced by the quartz crystal micro-balance with dissipation monitoring (QCM-D), another layer of alum was attached via the interaction with the alum/HA-assembled droplets (Supplementary Fig. 3b). After optimization of the outer layer alum concentration, multi-layered droplets were prepared with no excess alum in the continuous phase, but with enough alum to cover the inner antigen (Supplementary Fig. 4a–c). Unlike the alum/HA-stabilized droplets, scanning electron microscopy (SEM) images demonstrated an increased padding morphology, and stimulated emission depletion microscopy (STED) indicated an additional alum layer (red) on the alum/HA-stabilized droplets, thereby entrapping the inner HA (green) within the layer-by-layer nanostructures (Fig. 2b). In addition, a large surface area was exposed for NP adsorption. Based on the changes in the zeta potentials and elemental compositions, rMASE was prepared with HA and NP loaded inside and outside, respectively (Fig. 2c and Supplementary Fig. 4d, e).

Fig. 2
figure 2

Tailoring rMASE for the inside-out strategy. a Schematic illustration on rMASE preparation for the inside-out assembly of HA and NP. Stepwise formation of rMASE. (i) Co-assembly of alum and HA on the o/w interface; (ii) another layer of alum deposition to shield the inner HA; (iii) NP adsorption to constitute the sequential loading of HA and NP to finally obtain rMASE. b SEM (scale bar: 2 µm) and STED (scale bar: 1 µm) images of co-assembly alum/HA (i) and alum deposition (ii). c Zeta potential and structure illumination microscopy (SIM) images of rMASE. HA, NP, and alum were labeled with Cy3 (green), Cy5 (blue), and lumogallion (red), respectively. Scale bar: 1 µm. Data were shown as mean ± s.e.m (n = 3, from 3 independent experiments). d Verifying the thorough coverage of the inner HA and the surface display of NP for the inside-out strategy. The droplets were treated with 4% (v/v) FBS solution to avoid non-specific interaction and then with a mixture of anti-HA antibody (green) and anti-NP antibody (red), followed by confocal imaging. Scale bar: 5 µm. e XRD analysis on surface residual stress of droplets. The presence of residual stress (σφ) reflected the force tendency of the inner and outer antigens. The data were analyzed by regressing each data point to a straight line, and the linear slope M was obtained. Measuring and calculating the modulus of elasticity and Poisson’s ratio to calculate K, and the stress can be calculated from σφ = KM. Residual stress (σφ) indicated the tendency of the antigen towards (σφ > 0, compressive stress) or backwards (σφ < 0, tensile stress) the o/w interface. f DSC studies on the varied cooperated states of the inner HA. The right-shifting of the thermal peak indicated a greater energy for antigen to escape from the droplets, implying a more impeded release tendency

To test whether the inner antigen was completely shielded, the droplets were treated with a mixture of monoclonal antibodies against HA and NP. The images illustrated that alum/HA-stabilized droplets were strongly bonded with anti-HA antibodies (green). Subsequently, the addition of outer layer alum showed an evident reduction in the fluorescent intensity, suggesting the covering of the inner HA to avoid pre-exposure during antigen delivery. Additionally, after NP adsorption, the strong signal of a single fluorescence (red) indicated the dense display of NP on the rMASE surface (Fig. 2d). In a similar fashion, iMASE was prepared to load NP on the inside and adsorb HA on the outside of the droplets, which was determined with similar size and antigen loading efficiency but reversed antigen distribution compared to rMASE (Supplementary Fig. 5). Accordingly, the multi-layered alum-stabilized emulsion was developed for the inside-out assembly of NP and surface antigens. Through the layer-by-layer procedure on the o/w interface, iMASE and rMASE achieved consecutive loading of HA and NP in a facile and moderate manner, demonstrating the natural and reversed antigen distributions of the H1N1 influenza virus (Supplementary Table 2).

Dictating the release tendency of the antigens

Next, we investigated whether consecutive loading could affect the release kinetics of the outer and inner antigens. To verify the release tendency, the residual stress was evaluated via X-ray diffraction (XRD) analysis (Fig. 2e). Compared to their antigen-adsorbed counterparts (pink line), the co-assembled antigen and alum demonstrated decreased tensile stress. Whereas the attachment of the outer alum layer changed the residual stress from tensile stress to compressive stress, suggesting that the inner antigen was more likely to be entrapped within the nanocage formed between the close-binding alum and the o/w interface. Moreover, the right shift of the thermal peak in differential scanning calorimetry (DSC) demonstrated that a high thermal energy was required for the antigen to escape from the alum/HA-assembled droplets. These results indicated that the inner antigen had an impeded release tendency (Fig. 2f).

Then, rMASE was co-incubated with a 10% (v/v) fetal bovine serum (FBS) solution to test the release profile after administration. To better simulate the interstitial fluid, macromolecules larger than 30 kDa were removed using a centrifugal concentrator (30 kDa MWCO).25,26 As shown in Supplementary Fig. 6a, b, a limited amount of antigen was discharged from the droplets, suggesting that the antigens were only released after cellular internalization. With the macromolecules in the system, the outer NP was released before the entrapped HA (Supplementary Fig. 6c–e). Moreover, the release rate increased with increasing FBS concentrations, suggesting that the antigen may be emitted by ligand exchange with fluidic macromolecules (Supplementary Fig. 6f, g).

Reversed delivery of surface and core antigens

To further assess the release profiles, the intracellular distribution of rMASE was evaluated within BMDCs. As illustrated in the transmission electron microscopy (TEM) images, the droplets were first wrapped by the membrane to increase the contact area, triggering phagocytosis (Fig. 3a, i–ii). With an increase in the specific surface area, the multi-layered droplets stimulated cellular uptake and reached maximum internalization after 6 h (Supplementary Fig. 7a, b). As macromolecular proteins increased within the cytoplasm, the surface alum gradually fell off, along with the release of NP (Fig. 3a, iii). After 24 h, the apparent dissociation of alum occurred, indicating discharge of the inner antigen (Fig. 3a, iv).

Fig. 3
figure 3

Reversed delivery for the potent IFN-I-mediated immune activation in vitro. a Intracellular transfer of rMASE traced by TEM. Arrows indicated the dischargement of the alum. Scale bar: 2 μm. b Intracellular release of the antigens monitored via high content screen microscopy. HA and NP were labeled with FITC (green) and Cy5 (red), respectively. Scale bar: 20 µm. The release profile could be assessed by the fluorescence decaying of the loaded antigens. c IFN-α mRNA expression levels of the treated DCs at 24 h. d IFN-α secretion of the treated DCs at 48 h. e The expression of CD40, CD80, and CD86 of the treated DCs at 48 h. All data in the graphs were presented as the arithmetic mean ± s.e.m. from three independent experiments. For statistical analysis, a one-way analysis of variance was conducted with Tukey’s correction for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001

Subsequently, the intracellular release of the antigen was monitored using a high content imaging system (Operetta CLS, PerkinElmer). First, HA (FITC-labeled, green) and NP (Cy5-labeled, red) were sequentially loaded. The droplets were subsequently co-incubated with BMDCs for 6 h to achieve the maximum uptake. The fluorescence intensities of the fluidic antigens were too weak to be detected, in contrast to the fluorescence enrichment via micro-sized droplets. Accordingly, the release profile could be assessed by comparing the fluorescence decay of the loaded antigens. In the case of rMASE, a decrease in Cy5 fluorescence intensity demonstrated NP release. After 12 h, the fluorescence intensity of HA started to decline with a primarily constant but gentler slope, representing its subsequent dischargement at a relatively slow rate. Additionally, a reversed trend in iMASE-treated cells suggested the prior release of HA before NP (Fig. 3b and Supplementary Fig. 7c, d). To further verify this, NP-specific immunoglobulin M (IgM) was evaluated. Serum from mice immunized with rMASE exhibited a decreased level of NP-specific IgM on day 14. However, iMASE induced a higher titer on day 14 than on day 7, implying that the release of the inner antigen was delayed for immune recognition (Supplementary Fig. 7e). Thus, by consecutive loading via multi-layered droplets, the inside-out strategy dictated the delivery kinetics of the viral antigens.

To explore the immune effect, the inside-out strategy was tested in BMDCs. Compared with iMASE, rMASE significantly boosted the expression of IFN-α mRNA, with a 138% increase in IFN-α cytokine secretion, indicating the robust activation of IFN-І signaling (Fig. 3c, d). Subsequently, rMASE-treated DCs showed elevated expression of CD40, CD80, and CD86 by 174%, 128%, and 180%, respectively (Fig. 3e). Additionally, both iMASE and rMASE were detected with limited endotoxin levels and cytotoxicity, suggesting that the increased DC activation was attributed to the exposure of NP before HA, instead of potential material contamination or cell damage (Supplementary Fig. 8). Consequently, simply reversing the delivery of HA and NP can promote IFN-I-mediated immune responses.

Boosting humoral and cellular immune responses against H1N1 influenza

We postulated that rMASE may improve the immune response to H1N1 influenza infection. BALB/c mice were intramuscularly injected once with the formulations indicated in Supplementary Table 3, and the antigen depot was traced over time using an in vivo imaging system. As shown in Supplementary Fig. 9a, an evident antigen depot was observed, comparable with the HA and NP co-adsorbed alum (term “Alum”), and persisted for longer than 3 d. In contrast, the fluidic mixture of HA and NP (term “Antigen”) was cleared from the injection site within 12 h. With the elevated antigen repertoire, DCs were evidently attracted to the injection site for higher antigen uptake (Supplementary Fig. 9b–e). Notably, both iMASE and rMASE demonstrated similar trends in the antigen depot and DC internalization, indicating that the sequential release of viral antigens occurred intracellularly.

After immunizing BALB/c mice intramuscularly twice (three weeks apart), rMASE-induced antigen-specific immune responses were investigated. Notably, rMASE induced significantly higher HA-specific IgG titers compared with Alum (10-fold increase, P < 0.001) and iMASE (3-fold increase, P < 0.001) after 28 d (Fig. 4a). Additionally, elicitation of a potent serum antibody was also observed after 49 d. To test the cross-reactivity, the serum was also tested on other strains, including A/California/07/2009 (H1N1), A/Hong Kong/3039/2011 (H3N2), and A/Shanghai/4664 T/2013 (H7N9). As shown in Fig. 4b, compared with H1N1 (1934), the antibody titers of iMASE-treated mice showed 220%, 340%, and 800% decreases in H1N1 (2009), H3N2 (2011), and H7N9 (2013) subtypes, respectively. As for rMASE, the reductions were alleviated, with 110%, 210%, and 350% decreases in H1N1 (2009), H3N2 (2011), and H7N9 (2013) subtypes, respectively, suggesting the increased humoral immune responses.

Fig. 4
figure 4

Potent adaptive immune response against H1N1 influenza virus. a Serum HA-specific IgG titer. b Cross-reactive antibody responses to HA antigens, which derived from A/California/07/2009 (H1N1), A/Hong Kong/3039/2011 (H3N2), and A/Shanghai/ 4664 T/2013 (H7N9), respectively. c ELISPOT assay on IFN-γ spot-forming cells among the splenocytes, following stimulation with surface antigen (HA, A/Puerto Rico/8/1934). d Serum HA-specific IgG1 and IgG2a levels. e Body weight loss after virus challenge. For H1N1 vaccinations, BALB/c mice (n = 6) were administrated with the indicated formulations at prime-boost manner (3-week interval) and challenged by influenza virus (A/PR/8/34/1934 (H1N1)) on day 35, 2 times of LD50/mice. f Survival rate after viral challenge. Mice that lost ≥20% of their initial body weight were euthanized and counted as dead. g Viral loads in the lung by RT-qPCR. h Pathological changes in the lung. The sections were stained with H&E. Arrows indicated the perivascular and interstitial infiltration of inflammatory cells and lung consolidation. All data in the graphs were presented as the arithmetic mean ± s.e.m. from three independent experiments. For statistical analysis, a one-way analysis of variance was conducted with Tukey’s correction for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001

Moreover, reversed delivery of HA and NP promoted cellular immune responses. Compared to iMASE, rMASE increased the presence of IFN-γ-secreting T cells by 162% among splenocytes (P < 0.001, Fig. 4c). However, no significant differences were observed in IL-4 secreted cells (Supplementary Fig. 10a). Meanwhile, rMASE elicited a more than 6-fold higher HA-specific IgG2a titer (P < 0.001) than iMASE (Fig. 4d). With potent secretion of IL-2, IL-12, and TNF-α, the cytokine profiles further demonstrated that the inside-out strategy stimulated a Th1-biased immune response (Supplementary Fig. 10b). Furthermore, among the rMASE-treated splenocytes, IFN-γ-secreting T cells demonstrated increased cross-reactivity against H1N1 (2009), H3N2 (2011), and H7N9 (2013) subtypes, indicating the enhanced cross-protection against viral mutations (Supplementary Fig. 10c).

To further test immune protection in mice, we challenged the animals with H1N1 strain A/PR/8/34/1934. Weight loss and survival of the animals were monitored for 21 d post-challenge. rMASE-treated mice experienced a slight decrease in mean body weight on day 10, but quickly increased back to normal weight in less than 7 d. All mice in the Antigen- and Alum-treated groups experienced ≥20% weight loss within 13 d. Comparatively, the mean weight loss in the rMASE-treated group was approximately 15.7% of the original weight 10 days post-challenge (Fig. 4e). Notably, the survival rate in the rMASE-treated group was 100%. In contrast, the survival rate was only 50% in the iMASE-treated group, indicating that the reversed delivery of HA and NP increased immune protection against the H1N1 influenza viruses (Fig. 4f). Moreover, reverse transcription quantitative real-time polymerase chain reaction (RT-qPCR) analysis revealed that rMASE-immunized mice had significantly lower amounts of viral RNA in the lung tissues than those immunized with iMASE on day 9 (Fig. 4g). To evaluate pulmonary inflammatory damage, pathological examination was performed using Hematoxylin and Eosin (H&E) staining. As shown in Fig. 4h, no significant infiltration was observed in rMASE-immunized mice. However, iMASE-treated mice developed perivascular and interstitial infiltrates. Next, inflammatory cytokines in the lung were tested. Compared with iMASE, rMASE significantly decreased the levels of MCP-1, IL-8, IL-1β, and IL-6, indicating the alleviation of inflammation (Supplementary Fig. 10d–g). Collectively, the delivery of NP before HA provoked antigen-specific adaptive immune responses against viral infection.

Robust IFN-I-mediated immune response and lymph node activation

It remains challenging to enhance the long-term immune response and neutralization capabilities against the prevailing mutant strains of SARS-CoV-2. After finding an increase in adaptive immune response and cross-reactivity in H1N1 influenza vaccines, we postulated that the inside-out strategy may also improve the immune potency and duration of SARS-CoV-2 vaccination. Here, the multi-layer alum-stabilized emulsion shielded the surface antigen (RBD) on the o/w interface and adsorbed the core antigen (NP) on the outside (rMASE), allowing for the prior release of NP before RBD (Supplementary Fig. 11a–c and Supplementary Table 4). By successive loading of the viral antigens, iMASE achieved higher release concentrations of RBD before NP (Supplementary Fig. 11d, e).

To test whether the inside-out assembly of RBD and NP also provoked IFN-I-mediated immune responses, their impact on the transcriptome profile was assessed. The gene ontology (GO) term enrichment analysis revealed that rMASE significantly increased the IFN-I-related signaling pathway (under the criteria of P ≤ 0.05; Fig. 5a). Furthermore, comparative gene signature analysis revealed that interferon regulatory factor 7 (Irf7) was differentially expressed among rMASE-treated DCs, triggering the activation of a series of IFN-stimulated genes (Jak1, Stat1, Stat2, Irf9, and Isg15; Fig. 5b).

Fig. 5
figure 5

Provoking IFN-I activation in advance for the robust local reaction and lymph node activation. a GO term enrichment of the differentially expressed genes between iMASE and rMASE. GO analysis of differentially expressed genes within clusters identified the top associated enriched GO terms with corresponding enrichment P values. b Transcriptome analysis of DCs after co-culture with iMASE and rMASE. Representative heatmap showed differentially expressed genes relevant to the IFN-I signaling pathway. c IFN-α concentrations at the injection site over time. d Frequencies of CD40 expressions among the recruited DCs. e CCR-7 expressions among the recruited DCs, indicating the LN tropism. f The DC subsets within lymph nodes. g The bubble plot displays the engagement of the CD40L, germinal center: follicular T helper cells and GC B cells in LN. h Representative images of ICOS and CXCR5 immunofluorescence staining in LN. Sections were stained for anti-mouse CD4 antibody (green) and anti-mouse ICOS (red) antibody. The other sections were stained for anti-mouse CD4 antibody (green) and anti-mouse CXCR5 antibody (pink). Scale bar: 50 μm. i Memory B cell populations within the LN. All data in the graphs were presented as the arithmetic mean ± s.e.m. from three independent experiments. One-way and two-way analyses of variance were conducted with Tukey’s correction for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001

After intramuscular administration (Supplementary Table 5), rMASE significantly boosted the secretion of IFN-α (P < 0.001) and IFN-β (P < 0.001) at the injection site compared with iMASE (Fig. 5c and Supplementary Fig. 12a). As a consequence of IFN-I-mediated innate immune responses, higher levels of IL-2 (P < 0.001, Supplementary Fig. 12b) and TNF-α (P < 0.001; Supplementary Fig. 12c) secretions were observed in the early stage after administration, cultivating a more robust immunogenic microenvironment.27,28 In response to this, the expression of CD40 and CD86 in the recruited DCs was evidently increased, suggesting enhanced DC activation (Fig. 5d and Supplementary Fig. 12d, e). Notably, no evident inflammation or abnormal levels of IL-6, IL-17A, or MPC-1 were observed, indicating acceptable biosafety and well-controlled immunogenicity (Supplementary Fig. 13). Thus, the inside-out strategy stimulated IFN-I-mediated pathways, which cultivated an immune-stimulatory environment for the onset of an anti-viral state.29,30

Then, the activation of the draining lymph nodes (LNs) was probed.31 At the injection site, rMASE increased the expression of CCR-7 on the recruited DCs by 195% after 7 d post-administration compared to iMASE, indicating that the recruited DCs have a high potential for migration to LNs (Fig. 5e).32 Accordingly, rMASE showed a lower elevated LN-resident DCs proportion (CD8α+ CD11c+), but a more noticeable increase in the number of migrated DCs (CD103+ CD11c+ and CD11b+ CD11c+) within the LNs (Fig. 5f and Supplementary Fig. 14a, b). This indicated that rMASE promoted potent DCs migration from the injection site to the LN to achieve higher LN-accumulation of the antigens, rather than the direct delivery of antigens. Regarding DC activation within the LNs, CD40+ DCs were boosted by 150% on the 7th day after the administration of rMASE, compared with the iMASE-treated mice (Supplementary Fig. 14c, d). Furthermore, a significantly higher CD40L expression (P < 0.001) among the LN-residing CD3+ T cells suggested the increased interactions between DCs and T cells.33 Consequently, the rMASE-treated group exhibited a notable expansion of CXCR5+ ICOS+ CD3+ T cells and FAS+ GL-7+ B220+ B cells with an approximate 150% and 200% elevation, respectively (Fig. 5g and Supplementary Fig. 15a–d). Meanwhile, the LN immunofluorescence staining also demonstrated a similar trend, suggesting robust activation of the germinal center (GC; Fig. 5h). Furthermore, we also found that rMASE induced 140% more CD27+ B220+ cells, compared to iMASE, demonstrating an increase in memory B cells (Fig. 5i and Supplementary Fig. 15e). Thus, the inside-out strategy induced a higher IFN-I-mediated immune response at the early stage of vaccination, which subsequently led to potent LN activation for the onset of a strengthened adaptive immune response.

Activations on the long-term immune protection against SARS-CoV-2

Next, RBD-specific humoral response was evaluated (Fig. 6a). rMASE-adjuvanted formulations induced significantly higher RBD-specific IgG titers than Alum (16-fold increase, P < 0.001) and iMASE (2-fold increase, P < 0.001) after 35 days, and this persisted for longer than 3 months. Then, antibody affinity to the RBD antigen was determined using bio-layer interferometry (BLI). As shown in Fig. 6b and Supplementary Table 6, the association rate constant of rMASE (Kon = 3.63 × 104 Ms−1) was significantly higher than that of iMASE (Kon = 1.06 × 104 Ms−1). Additionally, rMASE elicited a lower value of equilibrium dissociation constant (KD = 0.26 ± 0.03 nM) compared with iMASE (KD = 22.4 ± 4.7 nM), indicating that reversed delivery of RBD and NP may enhance the antibody affinity against the viral infections. Neutralizing activity against the pseudovirus was subsequently probed. Compared with iMASE, rMASE elicited ~180% and ~275% higher neutralization titers (NT90) on days 28 and 49, respectively (Fig. 6c), which suggested efficient binding capabilities against viral infections. Notably, no differences in NP-specific antibody secretion were observed in either iMASE or rMASE, implying a limited effect of NP on the enhanced neutralizing capability (Supplementary Fig. 16a). The prior delivery of NP was further validated by the elevated cytokine profiles of IL-2, IL-12, TNF-α, and granzyme B (Supplementary Fig. 16b). Furthermore, central memory T cells (CD3+ CD8+ CD44high CD62Lhigh) and effector memory T cells (CD3+ CD8+ CD44high CD62Llow) in response to rMASE were significantly increased by 150% (P < 0.001) and 148% (P < 0.001), respectively (Supplementary Fig. 16c–e), suggesting an increased long-term immune response.34

Fig. 6
figure 6

Boosting humoral and cellular response for the persistent protection against SARS-CoV-2. a Serum RBD-specific IgG titer over time. Arrows illustrated the time points for vaccinations. The yellow line indicated the highest level of RBD-specific IgG titer induced by Alum. b The binding affinity of antibodies with RBD antigen measured by BLI. Post-administration, the antibodies were purified from the mouse serum collected on day 28. The association signals at different concentrations of RBD were monitored and fitted to obtain the kinetic parameters. c Measurement of SARS-CoV-2 pseudotyped virus 90% neutralizing titer (NT90) of serum samples from mice) on the 28th and 49th day after the first immunization. The SARS-CoV-2 pseudoviruses were developed by inserting full-length S protein (Wuhan-Hu-1) into vesicular stomatitis virus (VSV) G pseudotyped virus (G*ΔG-VSV). d Schematic illustration of the experimental design. BALB/c mice (n = 6) were administrated with the indicated formulations at two-week intervals. To assess the short-term immune protection, mice were transduced with 8 × 108 pfu of Ad5-hACE2 via i.n. route on day 23. For long-term evaluations, mice were transduced on day 44. After 5 d, the transduced mice were challenged with 5 × 105 TCID50 of SARS-CoV-2 (hCoV-19/China/CAS-B001/2020, GISAID No. EPI_ISL_514256-7) via the i.n. route, following the harvest of lung tissues to test the viral load and pathology 3 d later. e Virus titers in lung. SARS-CoV-2 titration from lung tissue by RT-qPCR probing virus gRNA. f Histopathology analysis of the harvested lung tissue. Tissue sections were stained with H&E. The black arrows indicated the infection-related symptoms, including the thickened alveolar walls, vascular congestion, and inflammatory cell infiltration. Scale bar: 625 µm (left) and 100 µm (right), respectively. g Serum neutralizing activity evaluated by authentic SARS-CoV-2 mutants, illustrated by the serum half-maximal neutralizing titer (NT50) against live SARS-CoV-2 WT, Delta, and Omicron. The serum was collected on day 28 post-administration. The number represented the fold decrease in neutralizing antibody titer. All data in the graphs were presented as the arithmetic mean ± s.e.m. from three independent experiments. For statistical analysis, a one-way analysis of variance was conducted with Tukey’s correction for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001

To explore the protective efficacy, a SARS-CoV-2-sensitive animal model was constructed using intranasal transduction of Ad5-hACE2-expressing adenovirus. After 5 d, the transduced mice were intranasally challenged with 5 × 105 TCID50 of SARS-CoV-2 (Fig. 6d).35,36 With high serological neutralizing antibody (NAb) titer, rMASE significantly reduced the viral loads in the lung compared with iMASE, with an approximate 1000-fold decrease compared with the sham group (Fig. 6e). No detectable subgenomic RNA (sgRNA) was observed in the rMASE-treated group, indicating that the rMASE vaccine was able to inhibit viral replication in lung tissue (Supplementary Fig. 16f). Furthermore, lung sections from the sham group exhibited thickened alveolar walls, vascular congestion, and inflammatory cell infiltration, which was consistent with viral pneumonia. In addition, iMASE-vaccinated mice exhibited moderate vascular congestion and inflammatory cell infiltration. In the case of rMASE, milder lesions were observed with substantially less infiltration of inflammatory cells (Fig. 6f). Furthermore, pulmonary histopathology was scored based on thickening of the alveolar septa, pulmonary alveolar congestion and inflammatory cell infiltration in the alveoli and trachea. As shown in Supplementary Fig. 16g, high lung lesion scores were found in the control animals, and the scores of lung lesions were reduced in the rMASE-vaccinated animals compared to the iMASE-treated groups. In particular, all six control animals (Sham) showed severe pulmonary alveolar congestion. In contrast, the rMASE-vaccinated group demonstrated almost no signs of alveolar congestion. These results indicated that rMASE can cultivate protective immune responses to diminish SARS-CoV-2-induced infections and lung injuries in mice.

To evaluate the efficacy of rMASE against the prevailing variants, we tested its neutralizing abilities against the live wild-type (WT) and variants of concern Delta (B.1.617.2) and Omicron (B.1.1.529). As shown in Fig. 6g, iMASE experienced a 230% decrease in Delta and a 1710% decrease in Omicron in the NAb titers compared with WT. In the case of rMASE, NAb titers decreased by approximately 2-fold and 8-fold against Delta and Omicron variants, which was less evident than that of iMASE, indicating the enhanced immune protection against the variant infections.


In summary, we developed a multi-layer alum-stabilized emulsion for the inside-out assembly of viral antigens, which led to a higher concentration of NP before the release of the surface antigen. In this manner, rMASE reversed the delivery of the surface antigen and NP and stimulated increased IFN-I-mediated innate immunity, in contrast to the natural package (iMASE). Furthermore, prior engagement of IFN-I signaling boosted adaptive immune responses against the influenza A (H1N1) and SARS-CoV-2 viruses. Thus, without any additional adjuvant components, simply altering the delivery sequence of the surface antigen and NP significantly increased the immune potency and duration against enveloped RNA viruses. Through multi-layer alum-stabilized emulsion, the inside-out strategy may offer a facile and efficient platform to elicit potent vaccine efficiency and broad immune protection.

In addition, the increased immune activations were attributed to the reversed delivery of the core and surface antigens, instead of the positive charges on NP. To test it, RBD and protamine, a commonly employed cationic protein, were loaded sequentially via MASE. As a result, the co-delivery of protamine and RBD failed to increase the immune activations by changing the release sequence of the cargos, which indicated that the inherent positive charges of NP scarcely contributed to the enhanced immune response (Supplementary Fig. 17). Moreover, the adjuvant effect of the inside-out strategy was not restrained by the source or sequence of the NP and surface antigens. Through the sequential delivery with NP, SARS-CoV-2 S1 protein or RBD-monomer as surface antigens also elicited a similar trend via MASE. Additionally, the prior release of NPs from Escherichia coli (E. coli) and eukaryotic cell lines (baculovirus-insect cells) also induced increased antibody secretion and T cell-mediated immune response (Supplementary Fig. 18).

To further demonstrate the efficacy, rMASE was compared with the commercial adjuvant (AddaVax™), a surfactant-stabilized emulsion in a formulation similar to that of MF59®.37 As shown in Supplementary Fig. 19, the AddaVaxTM-adjuvanted formulation failed to elicit comparable IgG titers (P < 0.001) and cellular immune responses, indicating increased immune potentiation against the commercial emulsion adjuvant. Subsequently, rMASE was compared with SARS-CoV-2 RBD-mRNA@ lipid nanoparticle (LNP), which shared a similar antigen sequence with the protein RBD antigen employed in this manuscript and formulated with SM-102 as the ionizable lipid, 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC) as the phosphate lipid, 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2000) as the PEGylated lipid, along with the cholesterol.38 As shown in Supplementary Fig. 20, rMASE induced slightly lower antigen-specific antibody titers and neutralizing antibody titers against WT than RBD-encoded mRNA@LNP. Additionally, rMASE showed comparable responses to Delta (B.1.617.2) and Omicron (B.1.1.529) variants, and a similar level in the engagement of the IFN-γ-producing T cells. Interestingly, rMASE elicited a higher frequency of central memory T cells (CD3+ CD8+ CD44high CD62Lhigh; P < 0.001), suggesting the higher immune memory may be activated by the reversed delivery of the core antigens and surface antigens. By contracting the mRNA-based strategy, rMASE induced comparable immune responses, which were evidently higher than the natural exposure of the antigens (iMASE), indicating that the reversed delivery of antigens may offer an enhanced strategy for recombinant protein vaccinations.

Notably, we do not imply that the inside-out strategy can be applied to all viral vaccines. As a preliminary attempt, we employed the multi-layer alum-stabilized emulsion to deliver the surface antigen (HBsAg) and core antigen (HBcAg) of the hepatitis B virus (HBV, a type of DNA virus). Intriguingly, the inside-out assembly failed to elicit higher immune potency. Instead, it is the natural distribution pattern that entrapped the core antigen inside, but displayed the surface antigen on the outmost layer, which significantly boosted the antigen-specific IgG titers, induced higher levels of IFN-γ+ T cells, and increased populations of the central memory T cells (CD44high CD62Lhigh) among the CD8+ T and CD4+ T cells (Supplementary Fig. 21a–d). This may be attributed to the more advanced potency of HBsAg to induce IFN-α expression compared with HBcAg (Supplementary Fig. 21e), which may subsequently potentiate the adaptive immune engagement. Presumptively, for the rational delivery of multi-component vaccines, prior delivery of immunogenic ingredients may result in the enhanced immune effect. As such, the prior delivery of adjuvants, such as Toll-like-receptor agonists and STING activators, before the co-delivered antigens may better boost the immune responses for enhanced vaccinations.

Collectively, the exact replication of live virus may not always offer an optimal solution. In the case of vaccines against H1N1 and SARS-CoV-2, reversed delivery of surface antigens and NPs (core antigens) were proved to potentiate the anti-viral effects. Besides modeling the steric structures of pathogens, it is also imperative to dictate the delivery kinetics of the vaccines, in view of spatiotemporal dynamics during immune activation.