Influenza vaccines promote humoral and cellular immune responses: a randomized, double-blind, phase 3 trial

Abstract

Annual vaccination is an effective strategy for preventing severe disease caused by seasonal influenza. Quadrivalent influenza vaccines (QIVs) protect against two strains of influenza A and two strains of influenza B, thereby enhancing the host antiviral neutralizing antibody response and inducing CD4⁺ and CD8⁺ T cell responses. Here, we report findings from a randomized, double-blind, active-controlled phase 3 clinical trial (NCT05431725) that includes 334 healthy adults aged 18–64 years and evaluates the humoral and cellular antiviral immune responses induced by two inactivated QIVs, Sinovac-QIV and Vaxigrip-Tetra™. The primary endpoint of the study is the specific antibody responses measured by hemagglutination inhibition (HAI) assays 28 days post-vaccination, while the secondary endpoint is virus-specific T cell responses. Both QIVs elicit significant increases in antibody titers 28 days after vaccination; Sinovac-QIV induces 9–10-fold increases in geometric mean titers, while Vaxigrip-Tetra™ elicits 7–8-fold increases (p < 0.05). Cellular immune responses using ELISPOT and supervised and unsupervised flow cytometry analyses show that both vaccines modulate the frequency of hemagglutinin-specific CD4⁺ and CD8⁺ T cell subsets, inducing distinct T cell response profiles. Although cellular analyses are evaluated in a subgroup of the cohort, the data indicate that both QIVs induce robust humoral and cellular immunity in adults, providing mechanistic insights into vaccine-induced protection.

Introduction

Seasonal flu outbreaks cause acute respiratory diseases in ~20% of the population worldwide, resulting in hundreds of thousands of deaths each year^1,2. Influenza virus infection is primarily transmitted through aerosols and droplets released by infected individuals³. Influenza viruses A, B, and C cause upper and lower respiratory diseases in humans, with influenza viruses A and B being the most prevalent compared to influenza C³. Influenza A viruses are further classified into different viral subtypes based on the composition of their membrane glycoproteins, hemagglutinin (HA) (18 HA subtypes) and neuraminidase (NA) (11 NA subtypes)⁴. On the other hand, influenza B viruses are classified in two subtypes: B/Victoria and B/Yamagata. For approximately the past 45 years, two influenza virus subtypes (A/H1N1 and A/H3N2) have co-circulated with both influenza B genetic lineages, causing seasonal outbreaks with varying degrees of morbidity and mortality⁵. It is important to note that the B/Yamagata lineage has not been conclusively detected in circulation by global surveillance data since early 2020, suggesting a possible eradication of this lineage⁶.

Currently, annual flu vaccination is the most effective strategy for preventing severe disease and mortality caused by influenza viruses⁷. Influenza vaccines are available as trivalent and quadrivalent formulations, which are either based on inactivated or live-attenuated virus. While trivalent influenza vaccines (TIVs) contain two strains of influenza A and one of influenza B⁸, quadrivalent vaccines (QIVs) contain two strains of influenza A and two of influenza B and are characterized by generating a wider immune response against influenza compared to TIVs⁸.

The primary parameter to determine vaccine-elicited protective immunity against influenza viruses is the generation of antibodies capable of neutralizing infectious virus, which correlates with the prevention of influenza-related disease^9,10. However, evidence suggests that cellular immunity induced by vaccination also plays a role in protecting the host during influenza virus infection, namely through influenza-specific CD8⁺ T cells and follicular T-helper cells that promote B cell responses^11,12,13. Nevertheless, determining influenza-specific cellular immune responses elicited by immunization is experimentally more challenging than measuring humoral responses due to the presence of pre-existing influenza-specific T cells in the population, which may have been elicited by prior viral exposure or previous vaccination.

A large, multicenter, phase 3, randomized, double-blind, active-controlled clinical trial using samples collected in this study demonstrated that the Sinovac-QIV showed an equivalent safety profile and antibody neutralizing response compared with Vaxigrip-Tetra™, as demonstrated through non-inferior humoral responses analyses between the two vaccines¹⁴. Vaxigrip-Tetra™ and Sinovac-QIV are quadrivalent inactivated influenza vaccines designed to protect against four different influenza virus strains, including two A subtypes (H1N1 and H3N2) and two B lineages (Victoria and Yamagata)^14,15. Additionally, both vaccines are administered through intramuscular injection and share similar storage requirements (2–8 °C). However, their regulatory status, age indications, and global use are key differences. Vaxigrip-Tetra™ is widely approved in many countries and has been recognized by the European Medicines Agency, being recommended for individuals from 6 months of age¹⁵. In contrast, Sinovac-QIV has been approved primarily in China and indicated for individuals aged 3 years and older¹⁴.

Here, we sought to extend the results obtained in the previous study to analyze the dynamics of the T cell response induced by vaccination with the two QIVs (Sinovac-QIV and Vaxigrip-Tetra™) in the samples collected in Chile. The primary endpoint of this study is the hemagglutination inhibition (HAI) antibody response 28 days post-vaccination, and the secondary endpoint is the virus-specific T cell responses. Using ELISPOT and multi-parametric flow cytometry, we comprehensively analyzed the frequency and phenotype of hemagglutinin-specific CD4⁺ and CD8⁺ T cell subsets following vaccination. Our findings reveal that both Sinovac-QIV and Vaxigrip-Tetra™ significantly modulate the T cell responses to influenza, and each vaccine induces a unique distribution of antigen-specific T cell subsets.

This study complements previous humoral immunogenicity analyses and provides insights regarding the induction of virus-specific T cell responses after immunization by integrating high-resolution immune profiling techniques. These findings contribute to a broader understanding of how inactivated QIVs elicit both humoral and cellular immune responses, which are crucial for protection against influenza infection.

Results

Demographic characteristics and safety profile

This study represents a subset analysis of a larger Phase 3, multicenter, randomized, double-blind, active-controlled clinical trial registered at clinicaltrials.gov (NCT05431725). A previous study reported humoral immunogenicity upon vaccination—measured as neutralizing response—and safety of Sinovac-QIV compared to Vaxigrip-Tetra™ in a diverse population from Chile and the Philippines¹⁴. The current study focuses on a subgroup of participants recruited in Chile to primarily assess the humoral and T cell responses induced after vaccination with influenza antigens.

A total of 334 volunteers participated in the PRO-QINF-3004 clinical trial during the 2023 influenza season in Chile (April–July) across five different clinical centers: Centro Marcoleta, Red de Salud UC Christus (CL01) (antibody and PBMC analyses), samples from Centro de Investigaciones Médicas de Enfermedades Respiratorias (CIMER) (CL04) (serum only), Clínica Alemana (CL05) (serum only), Hospital de Puerto Montt (CL07) (serum only) and Centro de Salud Universidad San Sebastián (CL08) (serum only) (Fig. 1A). From the total recruited individuals, 14 withdrew during the study. Although participants under 18 years of age were enrolled and blood samples were collected for immunogenicity assays, their data were not included in this manuscript due to methodological considerations and the study’s focus on evaluating immune responses in adults, whose immunological profiles and disease risk differ markedly from those of the pediatric population. Therefore, 263 adult subjects were included in the immunogenicity analyses. Among them, 132 received the Sinovac-QIV vaccine, and 131 received Vaxigrip-Tetra™ (Fig. 1A). Humoral responses were assessed in all recruited volunteers, regardless of the clinical center. At the same time, cellular immunity was analyzed in a subgroup of 56 volunteers from a single center (CL01) (Fig. 1A). Blood samples for both humoral and cellular studies were collected before (day 0) and 28 days post-vaccination (Fig. 1B).

Fig. 1: Study profile, cohort included in the study, vaccination schedule, and sample collection of enrolled subjects.

A Schematic representation and sample distribution of the study. This study included 263 out of 320 immunized individuals from the cohort enrolled across five different centers in Chile: CL01 (UC Christus-Marcoleta), CL04 (CIMER), CL05 (Clínica Alemana), CL07 (Hospital de Puerto Montt), and CL08 (Universidad San Sebastián). From a total of 263 volunteers included, 131 received Vaxigrip-Tetra^TM, and 132 received Sinovac-QIV. Cellular immunity was analyzed in PBMCs from a subgroup of 56 volunteers, all from center CL01 (27 received Vaxigrip-Tetra^TM and 29 received Sinovac-QIV) at each time point. Samples excluded due to quality control failure correspond to PBMCs, for which the frequency of AIM⁺ T cells was too low to allow adequate representation in the bh-SNE analysis. This is because the algorithm requires a minimum number of events to detect cellular patterns. B Timeline of the vaccination schedule and sample collection. Created in BioRender. González, P. (2025) https://BioRender.com/h4f93v9.

Demographic characteristics were similar across immunized subjects, regardless of the vaccine applied. The enrolled subjects’ average age (±standard deviation) was 42.6 ± 16.2 years; 65.4% were women, and 89.0% ranged from 18 to 64 years old. No significant differences between subjects vaccinated with Sinovac-QIV or Vaxigrip-Tetra^TM were shown in terms of sex (p = 0.861), age (p = 0.109), body mass index (BMI) (p = 0.473), and co-morbidities (p = 0.353) (Table 1). Likewise, demographic characteristics of the subject subgroups included in immunogenicity analyses were similar between subjects immunized with Sinovac-QIV and Vaxigrip-Tetra^TM.

Table 1 Demographic and clinical characteristics of recruited subjects

Demographic findings obtained are consistent with those reported by Yang et al.¹⁴, in which a larger study population analyzed exhibited comparable demographic distributions. Furthermore, our study confirms the safety data reported by Yang et al.¹⁴, as no significant differences in adverse event rates were observed between the two vaccine groups (p = 0.141). Adverse events were mainly mild to moderate in severity, and no serious adverse events were reported throughout the study (Table 2).

Table 2 Frequency of adverse events

QIVs induce robust humoral responses against influenza viruses

Hemagglutination inhibition assay (HAI) showed that both Sinovac-QIV and Vaxigrip-Tetra^TM induced a significant increase in the neutralizing capacity against all the influenza strain tested at 28 days post-immunization (Fig. 2). Specifically, subjects vaccinated with Sinovac-QIV showed 9- to 10-fold increases in geometric mean titer (GMT) against A/H1N1, A/H3N2, and B/Victoria viruses, and 3.4-fold increase against B/Yamagata (Fig. 2A). Similarly, immunization with Vaxigrip-Tetra^TM increased 7- to 8-fold the GMT against A/H1N1, A/H3N2, and B/Victoria viruses, and a 3.0-fold increase against B/Yamagata virus (Fig. 2B). However, subjects vaccinated with Sinovac-QIV increased GMT against A/H1N1 1.4-fold post-immunization (Fig. 2C).

Fig. 2: Humoral immune response against influenza before and post-immunization.

Reciprocal titer of antibodies against several strains of influenza virus (A/H1N1, A/H3N3, B/Yamagata, and B/Victoria) was measured by hemagglutination inhibition assay (HAI) in subjects vaccinated with (A) Sinovac-QIV and with (B) Vaxigrip-Tetra^TM. C The reciprocal titer of antibodies was compared between the two vaccines at 28 days post-immunization. The geometric mean titer of HAI antibodies against each strain was calculated by geometric mean and 95% CI. The data were transformed to Log₂. The numbers in black correspond to pre- and post-immunization GMT, and the blue numbers correspond to the geometric mean increase, and the red numbers correspond to the geometric mean decrease. n = 131 (volunteers immunized with Vaxigrip-Tetra^TM) and n = 132 (volunteers immunized with Sinovac-QIV). A, B Statistical significance was assessed using a two-tailed paired t-test (p < 0.0001 for all comparisons). C Data are presented as geometric mean values ± SEM. Two-tailed unpaired t-test (independent samples). p  <  0.05 was considered statistically significant. Source data are provided in a Source Data file.

Next, we determined the seroprotection rate, defined as the percentage of subjects with antibody titers ≥1:40 after immunization. Pre-immune seroprotection rates of subjects assigned to be immunized with Sinovac-QIV were 15.2% against A/H1N1, 43.9% against A/H3N2, 54.5% against B/Victoria, and 97.7% against B/Yamagata (Table 3). In volunteers assigned to receive Vaxigrip-Tetra^TM, the pre-immune seroprotection rate against A/H1N1 virus was 13.7%, 38.9% against A/H3N2, 60.3% against B/Victoria, and 98.5% against B/Yamagata (Table 3).

Table 3 Seroprotection and seroconversion rates for each vaccine strain

In addition, we quantified seroprotection rates at 28 days post-immunization. Our data shows that volunteers immunized with Sinovac-QIV and Vaxigrip-Tetra^TM displayed equivalent seroprotection rates for influenza A/H3N2, B/Victoria, and B/Yamagata. As for the influenza A/H1N1, volunteers vaccinated with Sinovac-QIV had a higher seroprotection rate at 28 days post-vaccination as compared to subjects vaccinated with Vaxigrip-Tetra^TM (p = 0.024) (Table 3). Furthermore, we analyzed the effect of immunization on seroconversion. Our results show that vaccination with Sinovac-QIV and Vaxigrip-Tetra^TM induced high seroconversion rates against influenza A/H1N1, A/H1N3, and B/Victoria. However, for the influenza A viruses, volunteers immunized with Sinovac-QIV presented a significantly higher seroconversion rate than those vaccinated with Vaxigrip-Tetra^TM (p = 0.0002 and p < 0.0001, respectively) (Table 3). These data indicate that immunization with both vaccines induces strong humoral responses against influenza A/H1N1, influenza A/H3N2, and B/Victoria and a moderate response against influenza B/Yamagata, presumably due to pre-existing immunity.

QIVs do not increase HA-specific IFN-γ− and IL-4−producing T cells

Cellular responses induced by vaccination were analyzed in a subgroup of subjects receiving one dose of Sinovac-QIV or Vaxigrip-Tetra™ at 28 days post-vaccination (Fig. 1A). Peripheral blood mononuclear cells (PBMCs) isolated at days 0 and 28 were stimulated with HA proteins derived from influenza A/H1N1, A/H3N2, B/Victoria, or B/Yamagata. IFN-γ- or IL-4-producing cells were quantified by ELISPOT, with spot-forming cells (SFCs) calculated per 300,000 cells. The data were normalized to the DMSO-treated control (untreated). Most subjects showed elevated baseline levels of IFN-γ⁺ SFCs before vaccination (Fig. 3A, B). The administration of Sinovac-QIV or Vaxigrip-Tetra™ did not significantly increase the number of IFN-γ⁺ SFCs compared to pre-immune levels for most of the stimuli (Fig. 3), with the only exception of A/H1N1-stimulated cells from Sinovac-QIV recipients, which exhibited an increased 2.1-fold increase of IFN-γ⁺ SFCs at day 28 post-vaccination (Fig. 3C). Regarding IL-4 production, no substantial increase in IL-4-producing cells was observed after vaccination with any vaccine (Supplementary Fig. 1). The representative ELISPOT images further illustrate the overall pattern of IFN-γ secretion before and after immunization (Fig. 3D, E).

Fig. 3: Immunization with Sinovac-QIV and Vaxigrip-Tetra^TM did not increase the number of IFN-γ-producing cells in response to HA.

PBMCs were stimulated for 48 h with 2.5 μg/mL of HA-derived proteins from A/H1N1, A/H3N2, B/Victoria, and B/Yamagata. The number of Spot Forming Cells (SFCs) secreting IFN-γ was quantified in PBMCs isolated from subjects vaccinated with (A) Sinovac-QIV and with (B) Vaxigrip-Tetra^TM, before and after immunization. C The number of SFCs secreting IFN-γ was compared between the two vaccines at 28 days post-immunization. The data were normalized with the untreated control. Representative ELISPOT images for (D) Sinovac-QIV and E Vaxigrip-Tetra™ are shown, where red spots correspond to IFN-γ-secreting cells and blue spots correspond to IL-4-secreting cells: n = 27 (volunteers vaccinated with Vaxigrip-Tetra^TM) and n = 29 (volunteers vaccinated with Sinovac-QIV). Numbers above dots represent the mean. The numbers in blue represent fold increases, and the numbers in red represent fold reductions. A, B Statistical significance was assessed using a two-tailed paired t-test. C Data are presented as mean values ± SEM. Two-tailed unpaired t-test (independent samples). p  <  0.05 was considered statistically significant. Source data are provided in a Source Data file.

QIVs induce differential HA-specific CD4⁺ T cell responses

The effects of Sinovac-QIV and Vaxigrip-Tetra^TM on the induction of HA-specific CD4⁺ T cell responses were evaluated by the quantification of the percentage of T cells expressing the activation-induced markers OX40⁺CD137⁺ (CD4⁺AIM⁺ T cells). PBMCs were stimulated for 24 h with the Hemagglutinin protein derived from A/H1N1, A/H3N2, B/Victoria, or B/Yamagata, and CD4⁺AIM⁺ T cells were quantified by flow cytometry using a traditional gating strategy (Supplementary Fig. 2A-2C).

Similarly to what we observed for IFN-γ− and IL-4-producing cells, at 28 days post-immunization with Sinovac-QIV (Supplementary Fig. 3A) or Vaxigrip-Tetra^TM (Supplementary Fig. 3B), no increase in the frequency of CD4⁺AIM⁺ T cells was observed when compared to pre-immunization levels, regardless of the stimulus. Subsequent characterization of CD4⁺AIM⁺ T cells (Supplementary Fig. 2A) showed no major changes in the frequency of CD4⁺AIM⁺ memory (Supplementary Fig. 3C-F), exhausted (Supplementary Fig. 3G-J), or follicular T cell subsets (Supplementary Fig. 3K-R) following vaccination, regardless of the vaccine administered. However, we observed high variability among volunteers in terms of both, magnitude and frequency, which may have masked subtle but biologically relevant changes when using conventional strategies.

To deeply study and characterize the CD4⁺AIM⁺ T cell-mediated responses, we carried out a strategy previously used by our group to study the T cell response induced by multiple immunizations against SARS-CoV-2¹⁶. Specifically, we used a multidimensional approach with the unsupervised Barnes-Hut stochastic neighbor embedding (bh-SNE) algorithm, which allowed us to visualize fine phenotypic shifts in CD4⁺OX40⁺CD137⁺ T cells that are not captured by traditional flow cytometry analyses. Functional markers including activation (CD137, CD69, OX40), memory subset distribution (CD45RA, CCR7), exhaustion (PD-1, Lag3, CD38) and follicular markers (CXCR5, PD-1, ICOS) were incorporated in the analysis (Supplementary Fig. 4A).

Given the high heterogeneity of CD4⁺AIM⁺T cells evidenced by the bh-SNE analysis, we performed unsupervised clustering analyses using the Phenograph algorithm (Euclidean distance, 350 neighbors). The resulting clusters were further validated using the Euclid plugin in FlowJo, through the ClustRCheck and the TaylorIndex algorithms, which compute and visualize Euclidean distance-based hierarchies across multidimensional datasets, determining how different clusters are. For CD4⁺AIM⁺ T cells, a Modularity score between clusters lower than 0.25 and a Taylor Score of 105.13 indicated a high degree of concordance between datasets, supporting the robustness and reproducibility of the identified clusters (Supplementary Fig. 4B).

Our unsupervised clustering analysis grouped the total CD4⁺AIM⁺T cells in 29 distinct clusters (C1 to C29), which exhibited a unique expression pattern for the analyzed markers (Fig. 4A, B). Notably, Lag3, a surface protein commonly linked to cell exhaustion, was broadly expressed among clusters (Fig. 4B). We further examined the differential effects of Sinovac-QIV and Vaxigrip-Tetra™ immunizations on the frequency of these clusters in response to different influenza HA proteins.

Fig. 4: CD4⁺AIM⁺ T cell clustering analyses after immunization with Sinovac-QIV or Vaxigrip-Tetra^TM.

A Clustering analyses with the Phenograph algorithm (Euclidean distance, 350 neighbors) grouped CD4⁺AIM⁺ T cells in 29 clusters. B Relative expression levels of CD137, ICOS, CD45RA, Lag3, PD1, CCR7, CD69, CD38, OX40, and CXCR5 for each cluster were quantified and represented in a heatmap. Phenotype profile and changes in frequency of CD4⁺AIM⁺ T cell clusters following vaccination with Sinovac-QIV and Vaxigrip-Tetra^TM in response to (C) A/H1N1, (D) A/H3N2, (E) B/Vitoria, and (F) B/Yamagata. n = 27 (volunteers immunized with Sinovac-QIV) and n = 23 (volunteers immunized with Vaxigrip-Tetra^TM). Numbers above dots represent the mean. For each strain, a two-way mixed-effects ANOVA was performed, followed by Holm–Šídák’s multiple comparisons test. Paired comparisons were conducted within each vaccine group (day 0 vs. day 28), and unpaired comparisons were performed between vaccine groups at day 28. p  <  0.05 was considered statistically significant. Source data are provided in a Source Data file.

Specifically, A/H1N1 stimulation induced an increase of C1 (TCM/TEM with Lag3⁺PD1⁺; Fig. 4C) and C16 (naïve/TCM without exhaustion or follicular marker expression) in subjects vaccinated with Vaxigrip-Tetra™. In contrast, Sinovac-QIV induced an increase of C15 and C22 (naïve/TCM with similar exhaustion but variable follicular marker expression) (Fig. 4C) at 28 days. Sinovac-QIV reduced C6 and C26 following A/H3N2 stimulation; both clusters of cells are composed mainly of naïve, TCM, and TEM cells with moderate to low co-expression of the exhaustion markers PD1 and Lag3 (Fig. 4D). Stimulation with B/Victoria increased C24 (naïve/TCM with minimal exhaustion/follicular markers; Suppl. Fig. 4E) frequency in volunteers immunized with Vaxigrip-Tetra™. Under the same condition, vaccination with Sinovac-QIV increased C13 (naïve with low exhaustion and minimal follicular marker expression; Fig. 4E). With B/Yamagata stimulation, Vaxigrip-Tetra™ increased C15 (naïve/TCM with low exhaustion and moderate follicular markers; Fig. 4F), while decreasing C23 (TCM lacking exhaustion/follicular markers). Sinovac-QIV increased approximately 1.5-fold C7 (TCM/TEM with low exhaustion/follicular markers) (Fig. 4F).

Altogether, our deep phenotypic characterization of the HA-specific CD4⁺AIM⁺ T cell population indicates that immunization with Sinovac-QIV and Vaxigrip-Tetra^TM affects the distribution of CD4⁺AIM⁺T cells, which is unique to each vaccine formulation.

QIVs induce differential HA-specific CD8⁺ T cell responses

No significant differences in the frequency of total CD8⁺AIM⁺(CD69⁺CD137⁺) T cells were observed after vaccination with Sinovac-QIV (Supplementary Fig. 5A) or Vaxigrip-Tetra^TM (Supplementary Fig. 5B) for each stimulus, using a traditional flow cytometry analysis. However, we did see a downward trend of CD8⁺AIM⁺T cells following stimulation with hemagglutinin protein derived from B/Yamagata (Supplementary Fig. 5A), A/H1N1, and A/H3N2, where CD8⁺AIM⁺ T cell frequencies were reduced by more than 2-fold (Supplementary Fig. 5A). Similar trends were observed in PBMCs from subjects immunized with Vaxigrip-Tetra^TM stimulated with A/H1N1, A/H3N2, B/Victoria, and B/Yamagata, although none of these reductions reached statistical significance (Supplementary Fig. 5B).

Memory CD8⁺AIM⁺T cell analyses (Supplementary Fig. 5C-F) showed that TEM cells and TEMRA cells were the most prevalent CD8⁺AIM⁺T memory subsets in circulation, both before and after vaccination with Sinovac-QIV or Vaxigrip-Tetra^TM in response to each stimulus (Supplementary Fig. 5C-F). Like CD4⁺AIM⁺ T cells, a high heterogeneity in the frequency of exhausted CD8⁺AIM⁺ T cells was observed, with no major changes in the responses between prior and post-vaccination, regardless of the stimulus (Supplementary Fig. 5G-J). Notably, volunteers vaccinated with Sinovac-QIV presented approximately 10-fold and 3-fold more cells expressing exhaustion markers, for A/H1N1 (Supplementary Fig. 5G) and B/Yamagata (Supplementary Fig. 5J), respectively, at day 28 compared with those who received Vaxigrip-Tetra™.

Subsequent clustering analyses using the Phenograph algorithm (Euclidean distance, 350 neighbors) grouped CD8⁺AIM⁺ T cells in 21 unique clusters (Fig. 5A) with differential expression of the analyzed surface markers (Fig. 5B). Further evaluation of frequency changes of each cluster showed that Sinovac-QIV and Vaxigrip-Tetra^TM induced differential changes in the frequency of each cluster evidenced after stimulation with HA protein derived from different influenza strains.

Fig. 5: CD8⁺AIM⁺ T cell clustering analyses after immunization with Sinovac-QIV and/or Vaxigrip-Tetra^TM.

A Clustering analyses with the Phenograph algorithm (Euclidean distance, 350 neighbors), and grouped CD8⁺AIM⁺ T cells in 21 clusters. B Relative expression levels of CD137, ICOS, CD45RA, Lag3, PD1, CCR7, CD69, CD38, OX40, and CXCR5 for each cluster were quantified and represented in a heatmap. Phenotype profile and changes in frequency of CD4⁺AIM⁺ T cell clusters following vaccination with Sinovac-QIV and Vaxigrip-Tetra^TM in response to (C) A/H1N1, (D) A/H3N2, (E) B/Vitoria, and (F) B/Yamagata. Numbers above dots represent the mean. n = 26 (volunteers immunized with Sinovac-QIV) and n = 19 (volunteers immunized with Vaxigrip-Tetra^TM). Numbers above dots represent the mean. For each strain, a two-way mixed-effects ANOVA was performed, followed by Holm–Šídák’s multiple comparisons test. Paired comparisons were conducted within each vaccine group (day 0 vs. day 28), and unpaired comparisons were performed between vaccine groups at day 28. p  <  0.05 was considered statistically significant. Source data are provided in a Source Data file.

Following A/H1N1 stimulation, Sinovac-QIV increased C20 (TEM/TEMRA with low-to-moderate exhaustion; Fig. 5C) and again reduced C16 as well. With A/H3N2 stimulation, Sinovac-QIV increased C2 (TEM/TEMRA with moderate exhaustion). In contrast, Vaxigrip-Tetra™ increased C11 (TEM/TEMRA with moderate exhaustion) and decreased C9 (TEM/TEMRA with low exhaustion) (Fig. 5D).

Upon B/Victoria stimulation, Sinovac-QIV reduced C12 (TEM/TEMRA with minimal exhaustion) (Fig. 5E). B/Yamagata stimulation led to increased frequencies of C14 and C20 in Sinovac-QIV-immunized volunteers. C14 was composed of TEM/TEMRA with high exhaustion (45–51%). C16 again showed reduced frequency following this stimulation (Fig. 5F).

The same cluster validation described for CD4⁺AIM⁺ T cells was performed in the cluster analyses for CD8⁺AIM⁺ T cells (Supplementary Fig. 6B). In contrast, a modularity score lower than 0.5 was obtained for each cluster using the ClustRCheck algorithm, a Taylor Score of 75.45 was reached, indicating a moderate-to-high quality and stability of the cluster analyses, and supporting their biological relevance (Supplementary Fig. 6B). Consistent with what we observed with CD4⁺AIM⁺ T cells, these data indicate that immunization with Sinovac-QIV and Vaxigrip-Tetra^TM has differential effects on the composition of CD8⁺AIM⁺ T cells specific against the influenza virus’s HA protein.

Discussion

Our data showed that both Sinovac-QIV and Vaxigrip-Tetra™ induced strong humoral immune responses, despite many of the recruited volunteers having preexisting immunity to the viral strains included in the vaccines, as evidenced by the elevated antibody titers of recruited volunteers. A higher seroprotection rate in pre-immune sera against influenza B viruses, especially against the B/Yamagata lineage, was observed, reaching levels of 97.7% and 98.5% in volunteers vaccinated with Sinovac-QIV or Vaxigrip-Tetra^TM, respectively. 89.0% of the recruited volunteers were young adults, and previous studies report a high prevalence of influenza B viruses in this age group^17,18. Two global studies strongly suggest that the Yamagata lineage is the most prevalent influenza B virus in countries with warm weather in the southern hemisphere, which could explain the high seroprotection rates observed in our cohort^17,19.

Additionally, according to the World Health Organization (WHO), the composition of the annual influenza vaccine has contained the same B/Yamagata lineage strain (B/Phuket/3073/2013) since 2018, which could explain the high seroprotection rates of our cohort against this lineage. Interestingly, high titers of pre-existing antibodies could lead to more rapid viral clearance, although leading to limited activation of pre-existing CD4⁺ T cells and memory B cells²⁰. However, high titers of pre-existing antibodies have also been proposed to improve the antibody response and their function, both for viral neutralization and for antibody-dependent cellular cytotoxicity^21,22. Our study observed that despite a high seroprotection rate against influenza B/Yamagata, immunization with either Sinovac-QIV or Vaxigrip-Tetra^TM induced 100% of seroprotection at 28 days post-immunization, suggesting the induction of de novo antibodies against this virus. Regardless, post-immunization seroprotection rates were similar between subjects immunized with both vaccines, reaching seroprotection levels to A/H1N1 viruses as high as 77.1% in a volunteer who received Vaxigrip-Tetra^TM and 87.9% in a volunteer who received Sinovac-QIV. These findings are consistent with those reported by the study of Yang et al.¹⁴, which showed non-inferior humoral responses for Sinovac-QIV compared to Vaxigrip-Tetra™ in a larger, multicenter cohort that included volunteers from the Philippines and Chile. While the study by Yang et al. primarily focused on SCRs and geometric mean titers (GMTs), our analysis extends these results by comprehensively characterizing T cell-mediated immunity against influenza antigens.

ELISPOT revealed an elevated presence of HA-specific IL-4 and IFN-γ-producing cells before vaccination that remained constant at 28 days post-vaccination with either Sinovac-QIV or Vaxigrip-Tetra^TM. During inflammatory processes, IL-4 suppresses the inflammatory responses and enhances the activation of humoral immune responses, being essential for germinal center formation and the production of protective antibodies during infection²³. Although our results did not show post-vaccination changes in the numbers of IL-4-producing cells, the existing numbers of cells that produce IL-4 could contribute to generating antibodies that neutralize the viral particles. Additionally, changes in the number of IL-4-producing cells may have occurred earlier in the individuals and not at 28 days after vaccination. Although the basal numbers of IFN-γ-producing cells were moderate, which is consistent with what has been reported in other studies^24,25, CD4⁺ but not CD8⁺ T cells produce IFN-γ in response to peptides, in low levels that are still enough to induce a protective response²⁶. These IFN-γ dynamics could explain the unchanged amount of IFN-γ−producing cells post-vaccination.

The stable frequency of circulating follicular T cells following immunization with Sinovac-QIV or Vaxigrip-Tetra™ suggests long-lasting immunity, likely due to prior influenza exposure or repeated vaccinations ^27,28. Previous studies indicate that pre-existing T cell responses may limit post-vaccination activation through a “ceiling effect” on immune response²⁹. Wild et al. highlighted the predominant role of pre-existing immunity in CD4⁺ T cells, which can mask the activation of newly induced influenza-specific CD4⁺ T cells in adults ³⁰. Since follicular T cell activation peaks around day 7 post-vaccination ³¹, the increased antibody titers observed in our study at day 28 may reflect early activation of these cells.

Traditional flow cytometry analyses are useful to identify the population of HA-specific AIM⁺CD4⁺ and HA-specific AIM⁺CD8⁺ T cells induced by vaccination. However, this type of analysis excludes the possible phenotype differences within the AIM⁺CD4⁺ and AIM⁺CD8⁺ T cell populations. Given that no differences were found in memory, exhaustion, and follicular subsets in AIM⁺CD4⁺ and AIM⁺CD8⁺ T cells using traditional flow cytometry analyses, we used bh-SNE analyses to deeply study the heterogeneity in AIM⁺CD4⁺ and AIM⁺CD8⁺ T cell phenotypes, as well as to study the effects of each vaccine on the T cell response. Multiparametric flow cytometry analyses have previously been shown to be useful for analyzing the complexity of CD4⁺ and CD8⁺ T cell responses induced by multiple vaccinations against SARS-CoV-2, that are not detected using traditional flow cytometry analyses¹⁶. Consistent with this previous study, our data showed that immunization with either Sinovac-QIV or Vaxigrip-Tetra^TM induced differential changes in the phenotype of CD4⁺AIM⁺ and CD8⁺AIM⁺ T cells that were specific for each stimulus used. In general, CD3⁺AIM⁺T cells presented an exhaustion phenotype characterized by the surface expression of PD1 and Lag3 in CD4⁺AIM⁺T cells and Lag3 and CD38 in CD8⁺AIM⁺T cells. Chronically stimulated T cells lose their functional capacity over time and are characterized by the upregulation of inhibitory receptors such as PD1, Lag3, CD38, and a decreased cytolytic function^32,33.

The unsupervised bh-SNE analysis of AIM⁺CD4⁺ T cells highlights distinct immune modulations, emphasizing the differential impacts of both vaccines. Sinovac-QIV reduces T cell exhaustion while promoting follicular T cell clusters, potentially enhancing long-term immunity. In contrast, Vaxigrip-Tetra™ favors TCM clusters, which may contribute to broader immune memory.

Our results demonstrate that Sinovac-QIV and Vaxigrip-Tetra™ vaccines induce discernible phenotypic shifts in CD8⁺AIM⁺ T cell populations in response to HA antigens. Unsupervised clustering revealed vaccine- and antigen-specific modulation of memory CD8⁺ subsets, including TEM and TEMRA cells, with heterogeneous expression of activation and exhaustion markers. While no significant increase in CD8⁺AIM⁺ cell frequencies was observed, the qualitative remodeling supports the hypothesis that each vaccine induced an HA-specific T cell response. These influenza-specific T cells may reflect pre-existing responses to viral proteins, with CD4⁺ T cells primarily recognizing surface antigens like HA, while CD8⁺ responses likely target internal proteins, such as M1³⁴. Consistent with Liao et al. and Krnic et al., who demonstrated HA-based vaccine induction of functional CD8⁺ T cells in human and murine models^35,36, our findings suggest that the functionality of these cells is preserved.

Though bh-SNE was used only for enhanced characterization of the CD3 + T cells, this study significantly improved immunophenotypic characterization. This type of analysis could be included in the evaluation of immune responses, due to its superior ability to preserve local and global data structure, which facilitates a deeper understanding of the cellular mechanisms triggered by vaccination, overcoming the limitations of traditional manual gating and providing a robust framework for assessing vaccine efficacy at the single-cell level.

Altogether, our study provides previously unreported insights into Sinovac-QIV’s immunogenicity and safety, highlighting its ability to elicit robust humoral responses while promoting distinct T cell memory profiles that may confer durable protection. While both vaccines achieved 100% seroprotection at 28 days, the application of advanced dimensionality reduction techniques such as bh-SNE allowed the identification of subtle but potentially significant differences in cellular immune responses, underscoring the added value of integrative immunological profiling beyond conventional assays.

Methods

Study design, volunteers, and randomization

This study includes a subpopulation of subjects recruited in a Phase 3, double-blind, randomized, active-controlled clinical trial (NCT05431725 https://clinicaltrials.gov/study/NCT05431725?locStr=Chile&country=CL&cond=influenza&rank=7, entitled “Immunogenicity and Safety Study of Influenza Vaccine (Split Virion), Inactivated, Quadrivalent”, date of registration 06/24/2022) that evaluated the antibody-induced response and safety of a split influenza virion¹⁴. Randomization lists were generated using SAS version 9.4 software and implemented via an Interactive Web Response System (IWRS). Upon enrollment, investigators accessed the IWRS to register participants and retrieve unique randomization codes linked to pre-labeled vaccine kits to maintain allocation concealment. The inactivated quadrivalent influenza vaccine developed by Sinovac Biotech Co., Ltd. (Sinovac-QIV) was compared to the licensed influenza vaccine Vaxigrip-Tetra™ (Sanofi Pasteur). The present study was focused on a cohort of healthy adults recruited in Chile. Volunteers were enrolled in 5 clinical centers: Centro Marcoleta, Red de Salud UC Christus (CL01), which was used for antibody and PBMC analyses, and samples from Centro de Investigaciones Médicas de Enfermedades Respiratorias (CL04), Clínica Alemana (CL05), Hospital de Puerto Montt (CL07), and Centro de Salud Universidad San Sebastián (CL08), which were used for antibody analyses only. Once enrolled, volunteers were randomized 1:1 to receive either one dose of 0.5 mL (containing 15 μg hemagglutinin for each of the four strains) of Sinovac-QIV or Vaxigrip-Tetra^TM at day 0. Blood samples were collected before immunization (day 0) and 28 days after vaccination to evaluate both humoral and cellular immune responses induced by the vaccines. All participants recruited for this study provided written informed consent at enrollment.

Investigational vaccine

Sinovac-QIV vaccine is a split-virion, inactivated, quadrivalent vaccine approved in China in June 2020, for people over three years of age³⁷. Vaxigrip-Tetra^TM is manufactured by Sanofi Pasteur and approved for use in Chile. Both vaccines are quadrivalent split-virion inactivated influenza formulations containing purified HA and NA antigens from the four corresponding influenza virus strains recommended by the WHO for the southern hemisphere winter season^38,39. These vaccines select appropriate splitting agents and conditions to disrupt the viral envelope and cleave open virion particles^38,39. The active antigenic components (HA and NA) can be concentrated to stimulate maximum antibody production^38,39. The Vaxigrip-Tetra^TM and Sinovac-QIV vaccines contain 15 µg of hemagglutinin per virus, including the following: A/H1N1, A/H3N2, B/Victoria, and B/Yamagata^37,39.

Sample collection and experimental procedures

For serum isolation, 5 mL of blood was collected in a tube without anticoagulant (BD Vacutainer Clot Activator tubes #367896). Blood samples were then incubated for 1 h at room temperature (RT) and centrifuged in a horizontal rotor at 1300 × g for 10 min at RT. The serum was collected and stored at –80 °C until use. To isolate peripheral blood mononuclear cells (PBMCs), blood was collected in three heparinized tubes (BD Vacutainer #367874, 10 mL) and stored at room temperature until processing. Samples were diluted with PBS 1X (Sigma, Catalog # P4417-50TAB) (1:1) and centrifuged for 10 min at 1200 × g (RT) in SepMate tubes (StemCell Technologies Catalog # 86460) containing Lymphoprep (StemCell Technologies Catalog # 07851) density-gradient medium. PBMCs were isolated into a clean tube and washed twice with sterile PBS. Cells were counted and cryopreserved in 90% FBS (Industrial Biologicals, cat. 04-1271 A) and 10% DMSO (Sigma-Aldrich, cat. D5879-500ML). Serum samples were used for hemagglutination inhibition assay (HAI), while PBMCs were used for enzyme-linked immunosorbent spot (ELISPOT) assay, and flow cytometry for subsequent bh-SNE analyses.

Hemagglutination inhibition assay (HAI)

Influenza-specific antibody titers were measured using a hemagglutination inhibition assay (HAI) performed at PPC Laboratories, Taiwan. Non-specific inhibitors were eliminated using a receptor-destroying enzyme (cholera filtrate, Sigma) at 37 °C for 16–18 h, and spontaneous agglutinins were adsorbed with Turkey red blood cells. Starting at 1:10, serum samples were 10 times 2-fold diluted and incubated with an influenza virus solution of 4 hemagglutination units/25 μL. The highest serum dilution at which hemagglutination was completely inhibited was reported as the hemagglutination inhibition titer. HAI titers were measured prior (day 0) and 28 days after vaccination and were recorded as the reciprocal of the dilution as previously described⁴⁰. The lower limit of quantitation was set as the reciprocal of the lowest dilution (10) and the upper limit of quantitation as the highest dilution (10,240) used in the assay. Seroprotection was defined as an HAI titer ≥1:40; seroconversion rate was defined as: 1. in subjects with pre-vaccination HAI antibody titer of <1:10, a post-vaccination HAI antibody titer of ≥1:40, or 2. in subjects with pre-vaccination HAI antibody titer of ≥ 1:10, a minimum of 4-fold increase in post-vaccination⁴⁰.

Enzyme-linked immunosorbent spot (ELISPOT) assay

PBMCs were resuspended 1:10 in fresh media to remove DMSO remnants from the freezing media. Then, cells were centrifuged for 10 min at 400 × g, resuspended in fresh media RPMI-1640 (Cytiva, cat SH30255.02), and counted in an automated cell counter (Logos Biosystems #L40001). Cells were adjusted to 6 × 10⁶ cells/mL and kept at 37 °C, 5% CO₂ for 15 min. ELISPOT plates (CTL) containing PVDF membranes were activated with 15 µL of ethanol 70% (Winkler, cat. #2223), washed three times with sterile PBS, and then coated with human IFN-γ and IL-4 capture antibodies (1:250 and 1:125, respectively, CTL, ImmunoSpot). After 2 h of activation at RT, plates were washed twice with PBS and twice with PBS-Tween 20 0.05%. HA protein derived from Influenza virus A/H3N2 (A/Darwin/9/2021, Sinobiological 40859-V08H) (2.5 µg/mL), Influenza B/Yamagata (B/PHUKET/3073/2013, Sinobiological 40498V08B) (2.5 µg/mL), Influenza B/Victoria (B/Austria/1359417/2021 (B/Victoria lineage)-like virus, Sinobiological 40862-V08B) (2.5 µg/mL) and Influenza A/H1N1 (A/Sydney/5/2021, Sinobiological 40944-V08B) (2.5 µg/mL) were used as specific stimuli. Recombinant SARS-CoV-2 Spike RBD protein (S-RBD) (2.5 mg/mL, Sinobiological), and phorbol 12-myristate 13-acetate (PMA, SIGMA) (500 ng/mL)/ionomycin (1 µg/mL, SIGMA) (PMA/I), were used as positive controls for antigen processing/presentation, and T cell activation, respectively. RPMI-1640 medium (vehicle) was included as a negative control (untreated control). 3 × 10⁵ cells in 50 µL of media were added to each well containing 50 µL of media with each stimulus and incubated for 48 h at 37 °C, 5% CO₂. After the incubation time, the number of T cells secreting IFN-γ or IL-4 was determined by ELISPOT using ImmunoSpot technology (ImmunoSpot, #hIFNgIL4-1M-10), following the manufacturer’s instructions. SFCs were counted in an ImmunoSpot S6 Micro Analyzer (CTL) using the ImmunoSpot^® Software. The results were normalized to the untreated control.

Flow cytometry

5 × 10⁵ PBMCs/well were plated in round-bottom 96-well plates and stimulated as described above, for 24 h. After stimulation, PBMCs were washed and incubated for 30 min with the following antibodies: CD3-AF700 (clone OKT4, Biolegend 317340), CD4-BV605 (clone RPA-T4, Biolegend 300556), CD8-BV650 (clone RPA-T8, Biolegend 301042), CD14-BV510 (clone M5E2, Biolegend 301842), CD19-BV510 (clone HIB19, Biolegend 302242), CD45RA-BV421 (clone HI100, Biolegend 304130), CCR7-FITC (clone G043H7, Biolegend 353216), CD69-PE (clone FN50, Biolegend 310906), CD137-APC (clone 4-1BB, Biolegend 309810), CD134 (OX40)-PE-cy7 (clone BER-ACT35, Biolegend 350012), CD38-PE/Dazzle^TM 594 (clone HIT2, Biolegend 303538), Lag-3-BV711 (clone 11C3C65, Biolegend 369320), ICOS-APC-Cy7 (clone C398.4 A, Biolegend 313530), CXCR5-PerCP/Cyanine5.5 (clone J252D4, Biolegend 356910), PD1-BV785 (clone EH12.2H7, Biolegend 329930), fixable viability dye-BV510 (BD-Horizon^TM, AB2869572). After incubation, samples were washed with PBS, fixed in PBS/PFA 2%, and analyzed in a BD LSR-FORTESSA X20 (BD FACSDiva v9.0 (BD Biosciences)) at the flow cytometry facility at the Pontificia Universidad Católica de Chile. Following the acquisition, samples were analyzed in Flowjo^® 10.10.1 (flowjo.com).

Barnes-Hut distributed stochastic neighbor embedding (bh-SNE) analyses

To comprehensively characterize the phenotypic diversity of antigen-specific T cells, we employed Barnes-Hut stochastic neighbor embedding (bh-SNE), an advanced dimensionality reduction algorithm optimized for analyzing high-dimensional single-cell data, such as flow cytometry. This method has been previously used to study T cell responses induced by vaccination against the respiratory virus SARS-CoV-2¹⁶ and is characterized by preserving local neighborhood relationships of complex, multi-parametric cellular data while projecting into an interpretable two-dimensional space, facilitating the identification of subtle phenotypic heterogeneity within immune cell populations⁴¹.

CD4⁺CD137⁺OX40⁺ (CD4⁺AIM⁺) and CD8⁺CD137⁺CD69⁺ (CD8⁺AIM⁺) T cell subsets were extracted from FlowJo® (version 10.10.0) as CSV files and subsequently converted to FCS format for bh-SNE analysis using Cyt (SightOf, version 2.0) implemented in Matlab R2022b (MathWorks®, Natick, MA). The dimensionality reduction input included a panel of markers relevant to T cell activation (CD137, CD69, OX40), memory differentiation (CD45RA, CCR7), follicular phenotype (CXCR5, PD-1, ICOS), and cellular exhaustion (PD-1, LAG-3, CD38). To ensure optimal resolution of phenotypic structures, best practices for BH-SNE implementation were applied, including adjusting early exaggeration parameters and an extended number of iterations to prevent cluster fragmentation and preserve rare cell subsets⁴¹. Unsupervised clustering of the bh-SNE output was performed using the Phenograph algorithm (Euclidean distance, 350 nearest neighbors) within the Cyt environment. For the initial dimensionality reduction bh-SNE and clustering analyses, data were concatenated by subject, vaccine, day pre- or post-vaccination, and stimulus to generate a global and comprehensive map of phenotypic diversity of CD4⁺AIM⁺ and CD8⁺AIM⁺ T cells, and subsequent analyses were performed between subjects that received Sinovac-QIV or Vaxigrip-Tetra^TM at each time point for each stimulus. Subsequent analyses were conducted separately for each vaccine group and stimulation condition to maintain the resolution of vaccine-specific effects and prevent pooling biases⁴². The main text and Supplementary Fig. 2 detail all gating strategies. Following bh-SNE and clustering, samples were concatenated by stimulation condition and exported as new FCS files for visualization in FlowJo® (version 10.10.0). Phenotypic characterizations of clusters exhibiting differential abundance after vaccination are shown in Figs. 4, 5.

To validate the quality of the clustering analyses made by Phenograph in both CD4⁺AIM⁺ (OX40⁺CD137⁺) and CD8⁺AIM⁺ (CD69⁺CD137⁺) T cells, clustering quality was analyzed using the Euclid plugin v0.9 in FlowJo v10.10.0 by two algorithms: ClustRCheck that assessed the heterogeneity within clusters by quantifying cluster modularity for each analyzed marker (CD137, ICOS, CD45RA, Lag3, PD1, CCR7, CD69, CD38, OX40, CXCR5); and through the TaylorIndex algorithm that evaluates the separation between clusters in an N-dimensional, determining how related the clusters are. For both algorithms, data is expressed in HeatMap plots showing the Modularity and TaylorIndex scores between the different clusters and parameters, either for CD4⁺AIM⁺ (Supplementary Fig. 4B) or CD8⁺AIM⁺ T cells (Supplementary Fig. 6B).

Statistical analyses

Demographic variables, safety outcomes, and seroconversion and seroprotection rates for each strain were compared between vaccination groups using the χ² test or Fisher’s exact test for categorical variables, and the Student’s t test for continuous variables.

For the analysis of humoral and cellular immune responses, we focused on specific comparisons of interest: within-group comparisons between day 0 and day 28, and between-group comparisons at day 28. Paired-sample tests were used for within-group comparisons, while independent-sample tests were applied for between-group comparisons. To evaluate the overall response of HA-specific CD4⁺ and CD8⁺ T cells across days, the vaccine was tested using a two-way mixed-effects ANOVA. Specific comparisons of interest were: Sinovac-QIV at day 0 vs Sinovac-QIV at day 28, Vaxigrip-Tetra^TM at day 0 vs Vaxigrip-Tetra^TM at day 28, and Sinovac at day 28 vs Vaxigrip-Tetra^TM at day 28. This was followed by Holm–Šidak’s multiple comparisons correction to control the family-wise error rate. All statistical tests were two-sided, with a significance threshold set at p < 0.05. Analyses were conducted using GraphPad Prism version 10.2.0.

Ethics statement

The study protocol was carried out according to the current Tripartite Guidelines for Good Clinical Practices, the Declaration of Helsinki, and local regulations. The trial protocol was reviewed and approved by the certified Institutional Scientific Ethical Committee of Health Sciences of the Pontificia Universidad Católica de Chile (#220311002). Additionally, execution of the trial was approved by the national health regulatory agency, Instituto de Salud Pública de Chile (EC1978503/2).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.