Nanocrystalline alloy-mediated delivery of mosaic epitope peptides for universal influenza vaccine

Abstract

Seasonal influenza infection poses great threat to public health systems. The flu vaccine remains the most effective method to reduce transmission and mortality. However, its effectiveness is limited due to the challenges in protecting against all influenza variants, as well as the weaker immune response observed in the adult population. Here, combining machine learning, synchrotron small angle X-ray scattering, we design an adjuvanted influenza vaccine composing mosaic epitope peptides selected from the hemagglutinin proteins of influenza A and B virus. These epitopes share similar physiochemical properties cognate to host antimicrobial peptides (AMPs) allowing them to form supramolecular assembly with poly(I:C), a synthetic toll-like receptor 3 (TLR3) agonist, through electrostatic interaction. The poly(I:C) is arranged into columnar lattice with the average inter-poly(I:C) distance commensurate with TLR3 and thereby capable of inducing multivalent TLR3 binding and hyperactivating the downstream inflammatory pathway. Interestingly, multiple AMP-like epitopes (Ampitopes) with compatible lattice parameter can co-crystalize into the same lattice to form 'alloy'-like composite with better poly(I:C) arrangement which allows the co-delivery of mosaic Ampitopes. The designed Ampitope-poly(I:C) nanocrystalline (and alloy) successfully activates interferon regulatory factor 3 (IRF3)-mediated pathway in antigen presenting cells. The intramuscular delivery of the nanocrystalline to the mice strongly trigger IL-6 and IFN-α release, which well-mimics the cytokines release pattern in influenza infected patients. After the third boost, the antigen-specific T cell response is 55 times higher compared to the free Ampitopes treatment group. Together, this vaccine offers a versatile way of eliciting strong and broad anti-flu protection.

---

Yue Zhang

Dr Yue Zhang is an Assistant Professor at Westlake University (China). She earned her PhD degree from the University of California San Diego and completed postdoctoral training at the University of California, Los Angeles. Her lab focus on the intersection of microbiology and nanomedicine. She is dedicated to uncovering how pathogenic microbes subvert immune surveillance and applying these natural strategies to guide material design for disease treatment. She is the co-chair of Communication committee of Chinese Association for Biomaterials and serves as the junior editor for Bioactive Materials and Asian Journal of Pharmaceutical Sciences.

---

Introduction

Influenza viruses are the causative viral pathogen of seasonal flu outbreaks. Each year approximately a billion people are infected globally and 3–5 million of the infected patients develop into severe illness.¹ In the flu season, multiple influenza variants may circulate simultaneously: data collected at the end of 2023 shows 79.1% of the detected influenza strains were of type A, among them 72.9% is H1N1 and 27.1% is H3N2.² The type B influenza, majority of which belong to the Victoria lineage, accounted for rest 21.5%.³ To mitigate transmission and reduce mortality, flu vaccines are designed to prime the immune system. The vaccine is usually formulated with either the inactive influenza virus or the recombinant influenza proteins, so that the immune system can quickly detect and clear the real virus once being exposed.^4,5 These vaccines often need to be designed based on a candidate influenza virus that is predicted to circulate in the upcoming flu season. However, the accuracy of prediction is far from satisfactory as the epitope profiles of influenza virus can vary significantly among the major circulating variants.^6–8 Efforts have been devoted to improving the prediction accuracy when designing the candidate strain, but the effectiveness of current influenza vaccines still fluctuates markedly, ranging from as low as 20% in years of antigenic mismatch to 40–60% when the antigenic profile matches well.^9,10 Moreover, the vaccination effectiveness can be even lower in the aged population, which may require proper adjuvants to boost the immunogenicity.^11–13

Thus, a successful flu vaccine design should include epitope information collected from the major influenza variants, so that the host immune system can elicit broad protective immunity in the flu season.¹⁴ Studies have tried using nanoemulsion approach to deliver recombinant viral proteins/peptides to cover as many major epitopes as possible.¹⁵ Due to the low immunogenic nature of protein/peptide-based vaccines, adjuvant is co-encapsulated to solve this issue. However, a good emulsion formulation usually requires vigorous shear force and the use of surfactant as a stabilizer, which requires further purification to obtain an injectable formulation. These series of steps largely increase the time and cost of vaccine production. Moreover, such methods suffer from low peptide loading yield and fast release, especially for the water-soluble epitope peptides studied here.

Recently, we found peptides derived from SARS-CoV-2 structural proteins can self-assemble with the adjuvant poly(I:C) into a nanocrystalline complex, a spontaneous process driven by the electrostatic interactions and hydrophobic forces.¹⁶ Machine learning analysis reveals that these viral peptides share the same physiological features as host antimicrobial peptides (AMP),¹⁷i.e., cationic and amphiphilic. The poly(I:C) in the complex is not randomly oriented but is rather organized into highly ordered liquid crystal with the average inter-chain distance around 3.4–3.6 nm, allowing them to bind Toll-like receptor 3 (TLR-3) multivalently, which leads to the recruitment and hyperactivation of the downstream inflammatory pathway.¹⁸ The transcriptome in the nanocrystalline-treated cells matches well with the global gene expression pattern in COVID-19, although only <0.3% of the viral proteome is being encapsulated in the nanocrystalline.

Here, in this work, we build the hypothesis that such viral peptide-encapsulating nanocrystalline can be designed into an influenza vaccine to induce strong and broad immune protection against major influenza variants. To test this hypothesis, the following questions need to be answered: (1) if such AMP-like peptide sequences are also prevalently present in influenza variants and how much these identified sequences overlap with the predicted major histocompatibility complex class I (MHC-I) and MHC-II epitopes, a critical criterion for evaluating the intrinsic immunogenicity of a given peptide sequence; (2) if such self-assembly process allows the co-encapsulation of multiple AMP-like epitopes (Ampitopes) with distinct physiochemical properties into the same nanocrystalline structure; (3) whether the nanocrystalline (and alloy) strongly activate the immune responses in vivo. Addressing these questions allows us to decipher the physiochemical features of the viral peptides that allow the formation of pro-inflammatory nanocrystalline (and alloy), and distill the design principles of these peptide-based vaccine with broader epitopes coverage and better protection efficacy.

Materials and methods

Materials

Fetal bovine serum (FBS), trypsin-EDTA solution, Dulbecco's modified Eagle's medium (DMEM), penicillin–streptomycin-fungizone (PSF) and other reagents for cell culture were purchased from VivaCell (Denzlingen, Germany). Poly(I:C) (HMW), Lucia^TM ISG cell, QUANTI-Luc™ 4 Lucia/Gaussia were purchased from InvivoGen (San Diego, USA). Mouse spleen lymphocyte isolation kit (P8860) was purchased from Solarbio (Beijing, China). Mouse interleukin-6 (IL-6, MM-0163M1) and interferon-α (IFN-α, MM-0182M1) enzyme-linked immunosorbent assay (ELISA) kits were purchased from MEIMIAN (Jiangsu, China). Mouse IL-2 (3321-4HPW-2) and IFN-γ (3441-4HPW-2) ELISpot kits were purchased from BioLegend (San Diego, USA). The mice were purchased from the Laboratory Animal Resources Center (LARC) at Westlake University.

Machine learning-based identification of Ampitopes

A total of 20 878 HA sequences were sourced from UniProt, comprising 10 393 H1N1, 10 107 H3N2, and 378 IBV sequences. After eliminating duplicates, a pre-trained SVM-based AMP classifier was employed to identify AMP-like peptides within the HA proteins. Each protein sequence underwent an initial analysis using a 24-amino-acid-long moving window. The classifier assessed the sequences, generating a sigma score (σ) reflecting the likelihood of the sequence being AMP-like. Positive and negative σ scores denote probabilities (P (+1)) of the sequence being an AMP or not, respectively. A σ score > 0 corresponds to P (+1) > 0.5. Details of the SVM model, encompassing the training dataset, physiochemical descriptors, and training methodology, have been previously documented.¹⁹ The outcomes pertaining to HA proteins from the three types of IFV are elaborated upon. Subsequently, individual toxic strain sequences were respectively selected from each subtype as target sequences for subsequent investigations, namely, A/California/07/2009 (H1N1, GCA_001343785.1), A/Hong Kong/1-11-MA21-1/1968 (H3N2, ACF41779.1), and B/Hong Kong/22/2001 (ABL77090.1). And epitope prediction platform (https://services.healthtech.dtu.dk/) was utilized to further screen epitope information from the aforementioned findings, wherein sequences binding to the MHC class I were predicted using NetMHCcons, and sequences binding to the MHC class II were predicted using NetMHCII.

Peptides

All peptides (H1a, H1b, H1c, H3a, H3b, H3c, Ba, Bb and Bc) were custom-synthesized (GL Biochem (Shanghai) Ltd, purity > 95% confirmed by HPLC). The net charge of the peptide was calculated with Prot pi (https://www.protpi.ch/). The linear charge density was calculated using the formula net charge/total amino acid of a given sequence. The hydrophobicity of each sequence was calculated with the Eisenberg scale.

Small angle X-ray scattering (SAXS)

The SAXS samples were prepared following the previously published method.¹⁷ Briefly, the peptide stock solution was prepared by dissolving the lyophilized powder in DMSO as 10 mg mL⁻¹. Poly(I:C) (HMW) was dissolved in pH 7.4, 10 mM HEPES buffer (with 140 mM NaCl) as 5 mg mL⁻¹. The Ampitope-poly(I:C) complexes were made by mixing the Ampitope with poly(I:C) at isoelectric ratio. The complexes were equilibrated at room temperature over 1 week before being transferred to 1.5 mm quartz capillaries (Quarzkapillaren, Mark-tubes). SAXS measurements were conducted at Shanghai Synchrotron Radiation Facility (SSRF, Beamline BL19U2) using monochromatic X-rays with a threshold of 6 keV. A Pilatus 1M, Dectris detector (detector plane 0.33 mm (H) × 0.05 mm (V)) was used to measure the scattered radiation. Independent identical samples were prepared and measured over several time points to ensure data consistency. The two-dimensional (2D) powder diffraction patterns were integrated with BioXTAS RAW. SAXS data were analyzed by plotting integrated scattering intensity I(q) against the momentum transfer q using BioXTAS RAW.

Peak positions were measured by fitting the diffraction peaks to a Lorentzian. Complex structures were figured out based on ratios between the q-positions of all recorded peaks and comparing them with the allowed reflections for known liquid-crystalline phases. The position of the first diffraction peak was used to calculate the average inter-poly(I:C) distance by the formula d = 2π/q.

Transmission electron microscopy (TEM)

The nanocrystalline samples were dispersed in ddH₂O, then a drop of this suspension was deposited onto a carbon-coated copper grid (300 mesh) incubated with a drop of ammonium molybdate negative stain solution for several seconds, then remove the liquid and air-dried for 1 h at room temperature. TEM analysis was performed using a Talos L120C G2 microscope operated at 120 kV, with the images captured by a Ceta CMOS camera. Then the TEM images were analyzed using ImageJ.

In vitro immune stimulation experiment

RAW-Lucia™ ISG cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% FBS and 1% PSF. The cells were seeded into a 96-well plate (3 × 10⁴ cells/well) with complete medium one day prior to stimulation. Subsequently, the cells were stimulated for 24 h in serum-free medium containing poly(I:C) (1.25 μg mL⁻¹, 2.5 μg mL⁻¹, 5 μg mL⁻¹, 10 μg mL⁻¹) either alone or complexed with the peptides at the isoelectric ratio. To assess the immunoreactivity of the stimulation, luciferase expression in the cell supernatant was quantified after 24 h using luciferase detection reagent, and the results were expressed in terms of relative light units (RLUs).

Animal experiment

Eight-to-ten-week-old female C57BL/6J mice were administered poly(I:C) (100 μg/mouse) and/or peptides (344 μg/mouse) in a final volume of 200 μL by intramuscular injection.¹⁵ Blood samples (200 μL) were obtained via inferior vena cava puncture into anticoagulant tubes 1 hour post administration of poly (I:C) and/or peptides. Plasma was prepared by centrifuging citrated whole blood at 6,000 × g for 5 minutes at 4 °C. Concentration of IL-6 and IFN-α in the plasma were quantified using ELISA. To evaluate the antigen-specific T cell response, the vaccine was administered once a week for three weeks. Mice were sacrificed and the spleens were collected. The lymphocytes were isolated with mouse spleen lymphocyte isolation kit and were seeded into a 96 well plate at 1 × 10⁵ cell/well. The lymphocytes were stimulated with the epitope peptides at the same concentration of in vitro immune stimulation experiment. The IL-2 and IFN-γ release in lymphocytes were quantified using ELISpot. All experimental animals were conducted following approval from the Institutional Animal Care and Use Committee (IACUC) at Westlake University.

Statistical analyses

The data were graphed and analysed using GraphPad Prism 10 software. Each test was conducted in triplicate to validate data precision. Statistical significance was determined employing two-sided unpaired t-test. A confidence level of 95% (CI) was applied, with significance set at P < 0.05, indicating a notable difference. Results were expressed as mean ± SD.

Results

Hemagglutinin (HA) protein is one of the surface proteins that allow influenza attaching to the host cells and open the fusion pore for viral entry. Thus, HA protein is often considered a major target for vaccine design. Integrating the mosaic epitopes from the major HA mutants into one vaccine formulation is promising for cultivating universal immune protection against major influenza variants. To test this idea, the HA proteins from the following three representative influenza strains are selected: A/California/07/2009 (H1N1) that caused the 2009 H1N1 influenza pandemic, A/Hong Kong/1-11-MA21-1/1968(H3N2), which was responsible for the 1968 influenza pandemic, and B/Hong Kong/22/2001 as a representative type B influenza virus strain identified in seasonal flu. For simplicity, the HA protein from the strains mentioned above will be named as flu-H1, flu-H3, and flu-B. The candidate HA epitope peptide sequences included in this vaccine formulation are screened based on two criteria: (1) they are predicted to bind with MHC-I or MHC-II, two proteins on the plasma membrane of antigen-presenting cells (APCs) responsible for stimulating T cell immunity or B cell immunity, respectively. (2) share the same physiochemical features with natural AMPs, which allows them to assemble poly(I:C) into a nanocrystalline complex for hyperactivating TLR3-mediated inflammatory response.

To predict T cell and B cell epitopes on HA proteins, we used a tool from the DTU Health Tech webserver, which predicts binding affinity of a query sequence with HLA-C1 (MHC-I) or HLA-DR (MHC-II) molecules. An epitope binding strongly with HLA-C1 is likely to be a T cell epitope and elicit strong cellular immunity, while an epitope binding strongly with HLA-DR is likely to be a B cell epitope and elicit strong humoral response (Fig. 1(A)–(F)). The results suggest both T cell and B cell epitopes are abundant and evenly distributed throughout the HA proteins.

Fig. 1 Machine learning-based identification of AMP-like epitopes (Ampitopes) in three representative Influenza hemagglutinin (HA) proteins. (A)–(C) Prediction of HA epitopes that binds (strong binding and weak binding) to HLA-C1 molecule. Twelve most common HLA-C1 allotypes are included. (D)–(F) Prediction of HA epitopes that binds (strong binding) to HLA-DR molecule. Six most common HLA-DR allotypes are studied. (G)–(I) HA proteins from three representative influenza strains (H1N1, H3N2, and Victoria) are scanned with AMP classifier for peptide sequences sharing the physiochemical features with natural AMP. Each query sequence is assigned a σ score. The first amino acid in each sequence is colored based on the score to reflect AMP-ness of this query sequence. A large, positive σ score suggests a high probability of being AMP. Sequences with σ < 0 (non-AMP) are labeled as deep blue. (J)–(L) Based on σ scores and epitope predictions, three Ampitopes are selected from each HA protein. Arrows are placed above the HLA-C1 prediction graph to locate the selected sequence. “μ” is the averaged σ scores of all homolog sequences in this influenza subtype, which is used to assess the conservation of AMP-ness at that site. The “hit rate” of each sequence is calculated by dividing the number of HLA-C1/HLA-DR allotypes that this sequence binds over the number of allotypes included in the prediction. Results are summarized in the last two columns.

Next, we proceed to evaluate if the predicted epitopes share the same physiochemical features with known AMPs that allows them to condense poly(I:C) into nanocrystalline. We first scanned the whole HA protein with the previously trained machine learning-based AMP classifier. This classifier takes the input of the 12 physiochemical features (net charge, amphiphilicity, amino acid pairs, etc.) calculated from the query sequence and then output a sigma score (σ score) to evaluate its probability (P (+1)) of being an AMP. A sequence with a σ score greater than 0 indicates strong probability of being an AMP-like sequence (P (+1) > 0.5). Results show the AMP-like sequences are clustered into hotspots and distribute quite differently among the selected HA proteins: In flu-H1, two major AMP hotspots present in the receptor binding site (RBS) in HA1 and the helix region in HA2 (Fig. 1(G)). In contrast, in flu-H3 and B, the two strong AMP hotspots previously seen in flu-H1 are depleted. But a new hotspot emerges at the C terminal of the HA1 protein for both influenza strains (Fig. 1(H) and (I)). In flu-B, two new hotspots emerge at the N-terminal of the HA1 subunit. The hotspot at the HA-2 is not completely depleted but is distinctly less bright (Fig. 1(I)). A large-scale screening of 4,935 HA proteins collected from each subtype suggests distribution of AMP hotspot and the AMP score are relatively conserved in strains classified within one subtype (Fig. S1, ESI†), meaning the AMPness of these hotspots is highly resistant to the mutations. Combining the epitope prediction and machine learning-based AMP prediction, three AMP-like epitopes (Ampitopes) from each strain are selected for further study (Fig. 1(J)–(L)).

Our previous study has shown that AMP-like peptide sequences can organize nucleic acids into nanocrystalline complexes, enabling multivalent nucleic acid presentation to the membrane bound nucleic acid sensor, i.e., TLR3 and TLR9 etc.^20,21 The inter-nucleic acid spacing (sweet spot is 3–4 nm) determines the amplitude of TLR-mediated immune activation. A spacing deviate from sweet spot all leads to less immune activation.¹⁷ To investigate if the selected Ampitope can condense poly(I:C), a synthetic TLR3 agonist, into the nanocrystalline with the spacing that falls into the sweet spot, and if the Ampitopes from different influenza strains can synergistically assemble poly(I:C) into a highly ordered nanocrystalline alloy, we seek to answer this series of questions with composite readout from Dynamic Light Scattering (DLS) and SAXS experiments.

We first designed poly(I:C) nanocomplexes that contain either single-Ampitope, dual-Ampitopes or tri-Ampitopes, and studied how the Ampitopes’ AMP-related physiochemical properties (net charge, linear charge density, hydrophobicity) affect the size and the size distribution of Ampitope-poly(I:C) complex, and whether the assembly mechanism allows the formation of highly ordered crystal structure. The DLS results show all of the single Ampitopes can condense poly(I:C) into nanosized complexes with the diameters ranging from 199 to 3485 nm. Five out of nine complexes are smaller than 400 nm, a preferable size range for cellular uptake, and for endosomal TLR3 access (Table 1). Among the selected Ampitopes, H1a, H1b, Bb, and Bc are the most capable ones of producing nanocomplexes with good size and size distribution. Pearson correlation analysis suggests Ampitopes with lower hydrophobicity, higher net charge and linear charge density tend to condense poly(I:C) into smaller complexes with narrower size distribution. The results highlight the cationic charge and hydrophilicity of the Ampitopes are the two key parameters that control the sizes of the electrostatic nanocomplex (Fig. 2(A)–(F)).

Table 1 The hydrodynamic size of the poly(I:C) nanocomplex comprising single-, dual- and triple-Ampitopes

Note: Z-average is the mean diameter of the nanocomplex. PDI reflects the size distribution in the sample. Ampitope(s)-poly(I:C) complexes with small Z-average and PDI are chosen for later study. Colour blocks represent the constituent Ampitope(s) within the complex.

---

Fig. 2 Correlation analysis between the key physiochemical features of Ampitopes and the size of the nanocomplex. Ampitope and poly(I:C) are mixed at charge ratio = 2. The diameter and polydispersity index (PDI) are measured with DLS. A nanosized complex with small PDI is desired for optimal cell uptake. The diameter or PDI is plotted against three relevant physiochemical properties (net charge, linear charge density and hydrophobicity). The correlation index (CI) is calculated with Pearson correlation function (r). Values ranging between 0 and 1 indicate correlation strength, with positive values indicating positive correlation and negative values indicating negative correlation. A trend line is added to better visualize the correlations. Ampitopes with higher net charge, charge density and lower hydrophobicity tend to condense poly(I:C) into a nanocomplex with smaller size and narrower size distribution.

Next, we proceeded to analyze the crystal structure of the Ampitope-poly(I:C) nanocomplex using SAXS. In the single-Ampitope system, a single diffraction peak is observed in the complex made with H1b, Bb or Bc, which can be indexed into a two-dimensional (2D) columnar lattice. The first diffraction peak positions between 0.142 and 0.179 Å⁻¹ corresponding to the inter-poly(I:C) distance 3.5 to 4.4 nm (Fig. 3(A)). No diffraction peak is observed with H1a, which sequence might be too short to induce long range ordering of poly(I:C). In the dual-Ampitopes system, we first studied if Ampitopes with varied poly(I:C) condensation activity can work synergistically in promoting the growth of immune stimulative nanocrystalline alloy. Based on DLS and SAXS measurements, H3a alone cannot effectively condense poly(I:C) into nanocrystalline, however when we started introducing H1b (a good poly(I:C) condensing Ampitope), these two Ampitopes began to co-crystalize with poly(I:C) into an ‘alloy’-like nanocrystalline structure with tighter spacing (3.8 nm in the alloy vs. 4 nm in the H1b-poly(I:C) complex), which approaches the optimal spacing for TLR3 activation (maximum activity is reached when inter-poly(I:C) spacing at 3.4 nm).¹⁷ The diffraction peak is also sharper than the H1b-poly(I:C) complex, suggesting a longer-range ordering start to form due to the synergistic effect among Ampitopes (Fig. 3(B)). A similar trend is also observed in the tri-Ampitopes alloy comprising H1a, Ba and Bc. Among these three Ampitopes, H1a and Ba fail to condense poly(I:C) into the nanocrystalline when they act alone. However, when we introduced Bc, a good poly(I:C) condensing Ampitope, to the mixture of H1a and Ba before incubating with poly(I:C), the resulted nanocrystalline alloy also shows a more ordered structure than the nanocrystalline formed by Bc alone (Fig. 3(C)). The synergy among these Ampitopes is important in the context of encapsulating diverse epitopes into the vaccine formulation that are capable of eliciting strong and broad protection against different influenza variants, given that not all crucial epitopes possess AMP-like physiochemical properties necessary for effective complexation with poly(I:C) (Fig. 1(J)–(L)).

Fig. 3 Structural characterization of Ampitope-poly(I:C) nanocomplex with small angle X-ray scattering (SAXS). The diffraction pattern of the nanocomplex made with single-, dual- and tri-Ampitopes are presented in (A), (B) and (C) respectively. A single, strong diffraction peak is observed in most of the samples, suggesting that poly(I:C) is organized by Ampitopes into two-dimensional columnar lattice. The average inter-poly(I:C) spacing can be calculated with d = 2π/q. The transmission electron microscopy (TEM) image shown in Panel D, the size of the Bc-poly(I:C) complex is about 200 nm. The cross section of the lattice is shown in Panel E. The periodic ordering of poly(I:C) in the nanocrystalline (alloy) allows the multivalent presentation of poly(I:C) to TLR3. The TLR3-mediated immune activation maximizes when the inter-poly(I:C) spacing is commensurate with the steric size of TLR3 (PDB ID:7C76 & 1CGC).

The forces driving the formation of the Ampitope-poly(I:C) nanocrystalline may involve the following ones: (1) the electrostatic force between the positively charged Ampitopes and the negatively charged poly(I:C). (2) The hydrophobic interaction between the Ampitopes that allows the peptides to form a stable protein core that glues the poly(I:C) around them. One thing to be noted is that condensing the poly(I:C) into a nanosized complex is not the only criteria to obtain an immune-active vaccine formulation, the spatial arrangement of poly(I:C) in the complex is also important. SAXS analysis on the structure of the representative electrostatic complex shows that the Ampitopes not only glue the poly(I:C) together, but also pack the poly(I:C) into a periodic crystal structure. The average distance (3.5–4.4 nm) between the poly(I:C) chains matches well with the steric size of TLR3, a membrane-bound dsRNA sensor on the host cells, which leads to the multivalent binding and hyperactivation of TLR3. Such mechanism is not often seen in traditional poly(I:C)-adjuvanted vaccines, as poly(I:C) is either administered as free solution²² or physically absorbed onto a nanocarrier.²³ In both cases, poly(I:C) is not spatially ordered for multivalent TLR3 binding and hyperactivation.

To test the APC activation induced by the nanocrystalline, we treat RAW-Lucia ISG cells with either monomeric poly(I:C), Ampitope-poly(I:C) nanocrystalline (or alloy), and Ampitope-poly(I:C) nanocomplex. Compared with the Ampiyope-poly(I:C) nanocrystalline, the poly(I:C) in the Ampiyope-poly(I:C) nanocomplex is not spatially ordered. The RAW-Lucia ISG cells are murine macrophage cells integrated with the luciferase reporter gene that allows direct assessment of the activation of interferon-signalling pathway, which is highly upregulated in viral infection. The results show Ampitope itself barely activates macrophage, suggesting the immunogenicity of the epitope peptides is quite low (fold of activation ∼1.4). In contrast, poly(I:C) treatment drastically activates macrophage by 126-fold comparing with the PBS treatment (Fig. 4(A)). We seek to study how crystallinity affects the macrophage activation. We focus on the four Ampitopes: H1b, Bb and Bc, which all organize poly(I:C) into nanocrystalline, while H1a only condense poly(I:C) into nanocomplex but without long-range ordering. Results show single-Ampitopes nanocrystalline formulated with H1b, Bb or Bc induces 238-, 263- and 222-fold activation, while H1a-poly(I:C) nanocomplex induces 186-fold (Fig. 4(A)), which is 22.8% lower than the average activation level induced by the three nanocrystalline formulation, suggesting the multivalent presentation of poly(I:C) enabled by these highly ordered nanocrystalline strongly upregulate TLR3-mediated immune activation. Moreover, the dual-Ampitopes nanocrystalline alloy formulated with H1b + H3a induces 234-fold activation (Fig. 4(B)), while the tri-Ampitopes nanocrystalline alloy formulated with H1a + Ba + Bc induces 221-fold activation (Fig. 4(C)). Both formulations activate the macrophage much stronger than poly(I:C) control, but at the same level as the single-Ampitope nanocrystalline.

Fig. 4 Ampitope-poly(I:C) nanocrystalline and alloy induces immune activation in macrophage. Ampitope-poly(I:C) nanocrystalline (single Ampitope) or alloy (mosaic Ampitopes) is incubated with RAW-Lucia™ ISG cells, a murine macrophage cell line integrated with a luciferase reporter gene to directly assess the activation of the interferon signaling pathway. Ampiyope-poly(I:C) nanocomplex (without long range poly(I:C) ordering) and monomeric poly(I:C) are used as control. The supernatant is collected after 24 hours of incubation and the luciferase activity is quantified. The fold of activation (FOA) is calculated by normalizing the luciferase signal in the sample-treated cells against the PBS-treated cells. The nanocrystalline alloy studied in panel B is formulated with a mixture of two Ampitopes: H1b + H3a. The nanocrystalline alloy studied in panel C is formulated with a mixture of three Ampitopes: H1a + Ba + Bc. All experiments are performed in triplicate. Data are presented as mean ± SE. Statistical analysis is performed using t-test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Finally, to investigate the immune stimulative activity of nanocrystalline alloy in vivo, we tested the release of two critical cytokines (IL-6 and IFN-α). IL-6 is an early-stage inflammatory marker that can be released by macrophages and myocytes in the skeletal muscle fibres, serving as a recruitment and activation signal for other immune cells. In comparison, IFN-α is mainly secreted by plasmacytoid dendritic cells, which activates interferon stimulated genes (ISG) in broad cell types to induces expression of downstream antiviral effector molecules. The dual-Ampitopes nanocrystalline alloy (H1b + H3a) and tri-Ampitopes nanocrystalline alloy (H1a + Ba + Bc) were administered to mice intramuscularly. PBS, Ampitopes and poly(I:C) were given as controls. Because of the short life of IL-6 and IFN-α in the inflammatory response,²⁴ plasma samples were collected 1-hour post-injection and analyzed for these two cytokines levels (Fig. 5(A)). Consistent with the in vitro readout, administrating Ampitopes shows negligible effect in triggering cytokine release. In contrast, poly(I:C) injection induced 40-fold higher IL-6 release and 92-fold higher IFN-α release, suggesting a strong adjuvant effect. However, the immune activation in the poly(I:C)-treated mice is quite heterogenous, ranging from 12 to 2589 pg/mL for IL-6 and 29 to 5426 pg/mL for IFN-α. This phenomenon might be attributed to heterogenous cell uptake and tissue retention of the free poly(I:C). When poly(I:C) is condensed by Ampitopes into nanocrystalline alloy, the data is distinctively less heterogenous. More importantly, nanocrystalline alloy boosts the IL-6 release by 124-fold, which is much stronger than the Ampitopes treatment alone (Fig. 5(B) and (C)). In the case of IFN-α, we find that tri-Ampitopes nanocrystalline alloy treatment leads to 48-fold increase (Fig. 5(E)), which is much higher than the Ampitopes treatment alone. However, the dual-Ampitopes treatment only leads to 21-fold increase (Fig. 5(D)), which is lower than immune activation observed in poly(I:C) treatment group. The distinct impact of nanocrystalline alloy in the IL-6 and IFN-α production might be attributed to the difference in the engaged immune cell population when the nanocrystalline alloy is processed locally. Overall, these in vivo acute inflammation data suggest nanocrystalline alloy trigger stronger and more consistent inflammatory response to the mosaic influenza epitopes, which might be beneficial to the formation of long-term antigen-specific adaptive immunity.

Antigen-specific T cell response plays critical role in antiviral adaptive immunity, as they participate in the activation of B cells and scavenging the infected cells to terminate infection. We isolated the lymphocytes from the immunized mice and quantify the activation level upon stimulate with corresponding free Ampitopes. IL-2 and IFN-γ positive lymphocytes were quantified with ELISpot. IL-2 is secreted by activated CD4⁺ T cells to supported the activation and differentiation of other T cells and B cells subsets. IFN-γ is also produced by activated T cells which upregulated MHC I and MHC II expression to enhance antigen presentation, and boost the activity of effector T cells. In this experiment (Fig. 5(F)), mice immunized with nanocrystalline alloy (both dual- and tri-Ampitopes) produces significantly higher amount of IL-2 or IFN-γ positive lymphocytes compared with the mice immunized with Ampitopes control and poly(I:C) control. The latter two groups have barely dateable Ampitopes-specific lymphocytes (Fig. 5(G)–(J)). Together, the nanocrystalline alloy significantly increases the cellular immunity response against the target Ampitopes in the mice.

Fig. 5 Ampitope-poly(I:C) nanocrystalline alloy activates the inflammatory response in C57BL/6J mice. Mice are intramuscularly injected with Ampitopes, poly(I:C), or Ampitope-poly(I:C) nanocrystalline alloy. Plasma is collected 1-hour post injection. Plasma collected from untreated mice is used to access the baseline level of IL-6 and IFN-α; (A) illustration of the sample injection, plasma collection, and detection procedures. (B) and (D) IL-6 and IFN-α level in the mice treated with dual-Ampitopes nanocrystalline alloy (H1b and H3a). (C) and (E) IL-6 and IFN-α level in the mice treated with tri-Ampitopes nanocrystalline alloy (H1a, Ba, and Bc). (F) illustration of the sample injection, lymphocytes isolation, and detection procedures. (G) and (H) enumeration of dual-Ampitopes (H1b and H3a)-specific IL-2 and IFN-γ SFU using ELISpot assay. (I) and (J) Enumeration of tri-Ampitopes (H1a, Ba, and Bc)-specific IL-2 and IFN-γ SFU using ELISpot assay. All experiments are performed with n = 5. The statistical analysis is performed with t-test (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Discussion

This study explores a strategy for developing a universal influenza vaccine by utilizing epitope information from various influenza strains. The core of this approach is the self-assembly of viral epitope peptides with nucleic acids, where the latter acts as a natural adjuvant to enhance immune responses. Our focus is on influenza epitopes that share physiochemical properties with host antimicrobial peptides (AMPs)—cationic and amphiphilic peptides—known for their ability to condense nucleic acids into nanocrystalline complexes, thereby significantly inducing TLR recruitment and activation, which boosts the immunogenicity of the protein/peptide-based vaccine and cultivate long term cellular immunity.

The periodic arrangement of poly(I:C) within the nanocrystalline structure enables the multivalent binding with endosomal TLR3. The activation is maximized when the spacing between poly(I:C) commensurate with the steric size of TLR3.¹⁶ The observation that different Ampitopes produce poly(I:C) nanocrystalline with varied spacing is crucial, allowing us to fine-tune the level of immune activation to meet specific clinical needs.

The observation that multiple Ampitopes can co-crystallize into one crystal structure with better ordering is interesting. Despite that the peptide alloy is a relatively new observation, there exist a few theories in the field of metal alloy that highlight the structure requirements of the involved elements. One related theory is ‘crystal structure compatibility’: the elements should ideally have the same or similar crystal structures. This similarity allows the atoms to fit into each other's crystal lattice without causing significant lattice distortion. In the exemplar poly(I:C) nanocrystalline alloy made with H1b, H3b, and Bc, although all three can organize poly(I:C) into a columnar lattice, the inter-poly(I:C) distance is different. The distances in the H1b-poly(I:C) complex, H3b-poly(I:C) complex and Bc-poly(I:C) complex are 4.4, 3.9 and 3.67 nm, respectively. The mismatch of the inter-poly(I:C) distance may disrupt the final crystal structure. In contrast, in the poly(I:C) nanocrystalline alloy made with H1a, Ba and Bc, only Bc organizes poly(I:C) into a columnar lattice with inter-poly(I:C) distance of 3.67 nm, but the rest two Ampitopes cannot. Thus, the crystal structure with the tri-Ampitopes system is quite similar to the one made with Bc. The inter-poly(I:C) distance is 3.76 nm. Together, the results suggest that the crystal structure compatibility among the participating Ampitopes when condensing poly(I:C) is critical to form a better alloy structure.

T-cell immunity plays a pivotal role in clearing the virus-infected cells. Thus, successful delivery of T cell epitopes is important for developing effective antiviral vaccine. However, T cell epitopes are usually short, hydrophobic and carry a low net charge, which is less optimal to encapsulate into the poly(I:C) nanocrystalline. But the technology developed here provide alternative for solving this issue: (1) co-crystallize the short T cell epitopes with longer Ampitopes (e.g., B cell Ampitopes) to form the nanocrystalline alloy; (2) select a longer Ampitope sequence covering T cell epitopes to aid its encapsulation into the nanocrystalline. These strategies might facilitate the development of balanced cellular and humoral immunity.

The ability to co-crystallize peptides with varied physiochemical characteristics, including those not typically classified as strong AMPs, into a nanocrystalline alloy offers a gateway for functionalization the nanocomplex. For example, introducing peptides that specifically bind the surface marker of the desired immune cells allows the targeted orchestration to the immune activation among different cell types, which may help us understand why the current tri-Ampitopes nanocrystalline alloy strongly induces IL-6 release but less on the IFN-α release.

While our research presents a promising approach for designing a universal flu vaccine, we recognize several inherent limitations in our study that serves as important direction for future research. A key limitation is the potential mismatch between our chosen human epitopes and the validation model used (mice), which might not accurately represent the establishment of adaptive immunity in humans. Therefore, immunization efficacy results obtained from non-human primates, which are genetically closer to humans, would be more valuable and potentially more applicable to human populations.

The formulation principles underscored here hold significant clinical relevance. The nanocrystalline alloy formulated with multiple epitopes allows the exposure of epitopes originated from different influenza strains to the immunized host all at once, strongly enhancing the breadth of anti-influenza immunity. Secondly, condensing the adjuvant poly(I:C) into a nanocrystalline induces stronger and more consistent immune activation than monomeric poly(I:C), which is critical to minimize the heterogeneity in vaccine protection among different vaccine recipients. Thirdly, the strong Ampitope-specific T cell response in the immunized mice allows the mice to quickly generate strong protective immune response when the host exposes to the real influenza infection. All these aspects are critical for developing a safe and effective flu vaccine.

Conclusions

In summary, our study introduces an innovative approach for creating a universal influenza vaccine by harnessing the self-assembly of AMP-like epitopes with adjuvant molecules. We demonstrate that nanocrystallinity strongly affects the amplitude of immune activation, which can be fine-tuned for optimal immune response. While focusing on the flu vaccine, this methodology holds potential for broader applications, including vaccines against other infectious diseases and cancers, offering both breadth and depth in vaccine-induced protection.

Author contributions

H. Y. W. and Y. Z. designed research; H. Y. W., H. F., L. Y. Z., J. Q. L., B. H. G., Y. Y. Z. and C. L. J. performed research; H. Y. W., H. F., L. Y. Z., J. Q. L., B. H. G., Y. Y. Z., C. L. J. and D. P. L. contributed reagents/analyzed tools; H. Y. W., H. F., J. Q. L. and Y. Z. analyzed data; and H. Y. W., D. P. L. and Y. Z. wrote the paper.

Data availability

All data are presented in figures. The raw data that support the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We thank the funding support from the National Natural Science Foundation of China (92356310, to Y. Z.), the Zhejiang Natural Science Foundation (XHD24E1301, to Y. Z.), the Westlake Education Foundation (to Y. Z.), the Research Center for Industries of the Future (to Y. Z.) and the Hangzhou Postdoctoral Research Foundation (to H. Y. W.). This work was carried out with the support of Shanghai Synchrotron Radiation Facility, Beamline BL19U2.The photos featured in the Table of Contents were contributed by Danyang Zhu at Westlake University.

References

1. F. Krammer, G. J. D. Smith, R. A. M. Fouchier, M. Peiris, K. Kedzierska, P. C. Doherty, P. Palese, M. L. Shaw, J. Treanor, R. G. Webster and A. García-Sastre, Influenza, Nat. Rev. Dis. Primers, 2018, 4(1), 3.
2. Centers for Disease Control and Prevention, Information for the 2023-2024 Flu Season, (https://www.cdc.gov/flu/season/faq-flu-season-2023-2024.htm).
3. Centers for Disease Control and Prevention, Weekly U.S. Influenza Surveillance Report, (https://www.cdc.gov/flu/weekly/index.htm).
4. K. B. Wiggins, M. A. Smith and S. Schultz-Cherry, The Nature of Immune Responses to Influenza Vaccination in High-Risk Populations, Viruses, 2021, 13(6), 1109.
5. R. F. Nuwarda, A. A. Alharbi and V. Kayser, An Overview of Influenza Viruses and Vaccines, Vaccines, 2021, 9(9), 1032.
6. J. A. Lewnard and S. Cobey, Immune History and Influenza Vaccine Effectiveness, Vaccines, 2018, 6(2), 28.
7. W. Shao, X. Li, M. U. Goraya, S. Wang and J. L. Chen, Evolution of Influenza A Virus by Mutation and Re-Assortment, Int. J. Mol. Sci., 2017, 18(8), 1650.
8. C.-J. Wei, M. C. Crank, J. Shiver, B. S. Graham, J. R. Mascola and G. J. Nabel, Next-generation influenza vaccines: opportunities and challenges, Nat. Rev. Drug Discovery, 2020, 19(4), 239–252.
9. Centers for Disease Control and Prevention, Seasonal flu vaccines (https://www.cdc.gov/flu/prevent/flushot.htm).
10. Centers for Disease Control and Prevention, Vaccine effectiveness: how well do flu vaccines work? (https://www.cdc.gov/flu/vaccines-work/vaccineeffect.htm).
11. Centers for Disease Control and Prevention, GRADE: Higher Dose and Adjuvanted Influenza Vaccines for Persons Aged ≥65 Years, (https://www.cdc.gov/vaccines/acip/recs/grade/influenza-older-adults.html).
12. M. P. Yeung, S. K. Ng, E. T. Tong, S. S. Chan and R. Coker, Factors associated with uptake of influenza vaccine in people aged 50 to 64 years in Hong Kong: A case-control study, BMC Public Health, 2015, 1, 617.
13. S. Oakley, J. Bouchet, P. Costello and J. Parker, Influenza vaccine uptake among at-risk adults (aged 16–64 years) in the UK: A retrospective database analysis, BMC Public Health, 2021, 21, 1734.
14. J. S. Heitmann, T. Bilich, C. Tandler, A. Nelde, Y. Maringer, M. Marconato, J. Reusch, S. Jäger, M. Denk, M. Richter, L. Anton, L. M. Weber, M. Roerden, J. Bauer, J. Rieth, M. Wacker, S. Hörber, A. Peter, C. Meisner, I. Fischer, M. W. Löffler, J. Karbach, E. Jäger, R. Klein, H. G. Rammensee, H. R. Salih and J. S. Walz, A COVID-19 peptide vaccine for the induction of SARS-CoV-2 T cell immunity, Nature, 2022, 601(7894), 617–622.
15. Q. Long, Y. Yang, M. Yang, H. Bai, W. Sun, X. Yang, W. Huang, D. Li and Y. Ma, Recombinant VLPs empower RBM peptides showing no immunogenicity in native SARS-COV-2 protein to elicit a robust neutralizing antibody response, Nanomedicine, 2022, 41, 102527.
16. E. Y. Lee, T. Takahashi, T. Curk, J. Dobnikar, R. L. Gallo and G. C. L. Wong, Crystallinity of Double-Stranded RNA-Antimicrobial Peptide Complexes Modulates Toll-Like Receptor 3-Mediated Inflammation, ACS Nano, 2017, 11(12), 12145–12155.
17. Y. Zhang, V. Bharathi, T. Dokoshi, J. de Anda, L. T. Ursery, N. N. Kulkarni, Y. Nakamura, J. Chen, E. W. C. Luo, L. Wang, H. Xu, A. Coady, R. Zurich, M. W. Lee, T. Matsui, H. Lee, L. C. Chan, A. A. Schepmoes, M. S. Lipton, R. Zhao, J. N. Adkins, G. C. Clair, L. R. Thurlow, J. C. Schisler, M. C. Wolfgang, R. S. Hagan, M. R. Yeaman, T. M. Weiss, X. Chen, M. M. H. Li, V. Nizet, S. Antoniak, N. Mackman, R. L. Gallo and G. C. L. Wong, Viral afterlife: SARS-CoV-2 as a reservoir of immunomimetic peptides that reassemble into proinflammatory supramolecular complexes, Proc. Natl. Acad. Sci. U. S. A., 2024, 121(6), e2300644120.
18. K. Sakaniwa, A. Fujimura, T. Shibata, H. Shigematsu, T. Ekimoto, M. Yamamoto, M. Ikeguchi, K. Miyake, U. Ohto and T. Shimizu, TLR3 forms a laterally aligned multimeric complex along double-stranded RNA for efficient signal transduction, Nat. Commun., 2023, 14(1), 164.
19. M. W. Lee, E. Y. Lee, A. L. Ferguson and G. C. L. Wong, Machine learning antimicrobial peptide sequences: Some surprising variations on the theme of amphiphilic assembly, Curr. Opin. Colloid Interface Sci., 2018, 38, 204–213.
20. N. N. Kulkarni, A. M. O’Neill, T. Dokoshi, E. W. C. Luo, G. C. L. Wong and R. L. Gallo, Sequence determinants in the cathelicidin LL-37 that promote inflammation via presentation of RNA to scavenger receptors, J. Biol. Chem., 2021, 297(1), 100828.
21. X. Chen, X. Yang, J. de Anda, J. Huang, D. Li, H. Xu, K. S. Shields, M. Džunková, J. Hansen, I. J. Patel, E. U. Yee, D. T. Golenbock, M. A. Grant, G. C. L. Wong and C. P. Kelly, Clostridioides difficile Toxin A Remodels Membranes and Mediates DNA Entry into Cells to Activate Toll-Like Receptor 9 Signaling, Gastroenterology, 2020, 159(6), 2181–2192.
22. J. V. Pérez-Girón, A. Belicha-Villanueva, E. Hassan, S. Gómez-Medina, J. L. G. Cruz, A. Lüdtke, P. Ruibal, R. A. Albrecht, A. García-Sastre and C. Muñoz-Fontela, Mucosal Polyinosinic-Polycytidylic Acid Improves Protection Elicited by Replicating Influenza Vaccines via Enhanced Dendritic Cell Function and T Cell Immunity, J. Immunol., 2014, 193(3), 1324–1332.
23. S. Renu, N. Feliciano-Ruiz, F. Lu, S. Ghimire, Y. Han, J. Schrock, S. Dhakal, V. Patil, S. Krakowka, H. HogenEsch and G. J. Renukaradhya, A Nanoparticle-Poly(I:C) Combination Adjuvant Enhances the Breadth of the Immune Response to Inactivated Influenza Virus Vaccine in Pigs, Vaccines, 2020, 8(2), 229.
24. S. M. L. Salem, A. N. Kadima, D. J. Cole and W. E. Gillanders, Defining the antigen-specific T-cell response to vaccination and poly(I:C)/TLR3 signaling: evidence of enhanced primary and memory CD8 T-cell responses and antitumor immunity, J. Immunother., 2005, 28(3), 220–228.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4tb00742e

---