Induction of SARS-CoV-2 Protein S-Specific CD8+ T Cells in the Lungs of gp96-Ig-S Vaccinated Mice

Given the aggressive spread of COVID-19-related deaths, there is an urgent public health need to support the development of vaccine candidates to rapidly improve the available control measures against SARS-CoV-2. To meet this need, we are leveraging our existing vaccine platform to target SARS-CoV-2. Here, we generated cellular heat shock chaperone protein, glycoprotein 96 (gp96), to deliver SARS-CoV-2 protein S (spike) to the immune system and to induce cell-mediated immune responses. We showed that our vaccine platform effectively stimulates a robust cellular immune response against protein S. Moreover, we confirmed that gp96-Ig, secreted from allogeneic cells expressing full-length protein S, generates powerful, protein S polyepitope-specific CD4+ and CD8+ T cell responses in both lung interstitium and airways. These findings were further strengthened by the observation that protein-S -specific CD8+ T cells were induced in human leukocyte antigen HLA-A2.1 transgenic mice thus providing encouraging translational data that the vaccine is likely to work in humans, in the context of SARS-CoV-2 antigen presentation.


INTRODUCTION
The rapid spread of the global COVID-19 pandemic has put pressure on the development of a SARS-CoV-2 vaccine to address global health concerns. We generated a gp96-Ig-secreting vaccine expressing full-length spike or "S" glycoprotein of SARS-CoV-2 via a cell-delivered platform. Targeting SARS-CoV-2 spike (S) protein remains the favorable vaccine choice as it is one of the most abundant and immunogenic proteins translated from the SARS-CoV-2 genome (1). Antibodies targeting S protein aim to neutralize mammalian host-cell interaction, thereby minimizing viral multiplicity of infection, however, recent studies have shown that "antibodies are not enough" to protect against COVID-19 for a variety of reasons, including S-protein glycosylation, which shields the antibody from eliciting an optimal neutralization response (2). Antibody decay has also been detected in individuals after recovery from COVID-19, and this decline was more rapid than reported for the first SARS infection in 2003 (3,4).
T-cell immunity plays a pivotal role in generating a durable immune memory response to protect against viral infection. Prior studies have shown that memory B-cell responses tend to be short lived after infection with SARS-CoV-1 (5,6). In contrast, memory T-cell responses can persist for many years (7). Recent data confirm that SARS-CoV-2-specific memory CD8+ T cells are present in the vast majority of patients following recovery from COVID-19 (7)(8)(9)(10), and their protective role has been inferred from studies in patients who have had both SARS and MERS (11)(12)(13). Recent reports show that patients who have recovered from a severe SARS-CoV-2 infection have T-cell responses against viral spike protein and other structural and nonstructural proteins; in some patients, Tcell responses were present regardless of symptoms or antibody seropositivity (14)(15)(16). Here, we generated a COVID-19 vaccine based on the proprietary secreted heat shock protein, gp96-Ig vaccine strategy, that induces antigen-specific CD8+ T lymphocytes in epithelial tissues, including lungs.
Tissue-resident memory (TRM) T cells have been recognized as a distinct population of memory cells that are capable of rapidly responding to infection in the tissue, without requiring priming in the lymph nodes (17)(18)(19)(20). Several key molecules important for CD8+ T cell entry and retention in the lung have been identified (21)(22)(23)(24)(25)(26) and recently CD69 and CXCR6 (20,(27)(28)(29) have been confirmed as core markers that define TRM cells in the lungs. Furthermore, it was confirmed that CXCR6-CXCL16 interactions control the localization and maintenance of virus-specific CD8+ TRM cells in the lungs (20). It has also been shown that, in heterosubtypic influenza challenge studies (30)(31)(32), TRM were required for effective clearance of the virus. Therefore, vaccination strategies targeting generation of TRM and their persistence may provide enhanced immunity, compared with vaccines that rely on circulating responses (32).
Our platform technology consists of a genetically engineered construct of gp96, fusion protein gp96-Ig, wherein the Cterminal KDEL-retention sequence was replaced with the fragment crystallizable (Fc) portion of immunoglobulin G1 (IgG1), and then encoded within a plasmid vector that is transfected into a cell line of interest. The cell serves as the antigen supply to secreted gp96-Ig. Complexes of gp96-Ig and antigenic peptides lead to specific cross-presentation of cellderived antigens by gp96-Ig in vivo (33,34). A crucial advantage offered by this gp96-based technology platform is that it allows for any antigen (such as SARS-CoV-2 S peptides) in the complex with gp96 to drive a potent and long-standing immune response. Over the last 2 decades, we have established that gp96-Ig, secreted from allogeneic or xenogeneic cells containing selected infectious antigens, generates potent, disease antigen specific, polyepitope, multifunctional CD8+ T cells in epithelial tissues (33)(34)(35)(36)(37)(38)(39). Here we generated a COVID-19 vaccine based on the secreted heat shock protein gp96-Ig vaccine strategy, and demonstrated vaccine-induced SARS-CoV-2 protein S-specific CD8+ and CD4+ T lymphocytes in epithelial tissues, including lungs and airways. The secreted gp96-Ig-COVID-19 vaccine has the potential to elicit robust long-term memory T-cell responses against multiple SARS-CoV-2 antigens and is designed to work cohesively with other treatments/ vaccines (as boosters or as second-line defense) with largescale manufacturing potential.
The slides were left in blocking media [5% bovine serum albumin (BSA) in PBS] at room temperature for 2 h. Rabbit anti-SARS-CoV-2 spike glycoprotein antibody (Abcam ab272504) and Donkey anti-rabbit IgG FITC, (BioLegend Cat# 406403) fluorescent antibody-were added in 1/50 and 1/100 dilutions of the antibodies combined in 5% BSA in PBS and/or rabbit isotype control (Abcam Ab172730 diluted 1/50), and incubated overnight at 4°C in a dark moisture chamber. The next day, slides were washed 3 times for 5 min with PBS and mounted with A B C D FIGURE 1 | Schematic of gp96-Ig and SARS-CoV-2 protein S constructs used to generate vaccine cells HEK-293-gp96-Ig-S and AD-100-gp96-Ig-S. (A) Each panel presents the protein expressed by the DNA (black outline) for the gp96-Ig and SARS-CoV-2 protein S vaccine antigen. Gp96-Ig and SARS-CoV-2-S DNA were cloned into the mammalian expression vectors B45 and pcDNA 3.1, which are transfected into HEK-293 and AD100. Stably transfected vaccine cell clones (1A, 1A6, 1D6) were generated after selection with L-Histidinol and Neomycin; (B) One million 293-gp96-Ig-S and AD-100-gp96-Ig-S (1D6) cells were plated in 1 ml for 24 h and gp96-Ig production in the supernatant was determined by ELISA using anti-human IgG antibody for detection with human IgG1 (0.5 ug/ml) as a standard; (C) Cell lysates were analyzed under reduced conditions by SDS-PAGE and Western blotting using anti protein S antibody and recombinant protein S1 as a positive control; (D) IF for protein S (in green) expressed in AD100-gp96-Ig-S cells using rabbit anti-SARS-CoV-2 S antibody and anti-rabbit Ig-AF488 as secondary antibody. AD100 was used as a negative control and b-actin for protein quantification. Original magnification 40× with DAPI nuclear staining shown in blue. DNA, deoxyribonucleic acid; ELISA, enzyme-linked immunosorbent assay; IgG, immunoglobulin G; N, amino terminus; C, carboxy terminus; IF, immunofluorescence; TM, transmembrane domain; KDEL, retention signal; CH2 CH3 gamma 1, heavy chain of IgG1. See text for explanation.
Prolong Gold antifade reagent with DAPI from Invitrogen (Catalog #36935), covered with a coverslip, and allowed to cure. The slides were then sealed with nail polish and taken to the KEYENCE microscope for examination. The following filter cubes were used: DAPI (for nuclear stain), FITC (for protein S), and images were acquired on KEYENCE microscope (BZ-X Viewer).

Animals and Vaccination
Mice used in this study were colony-bred mice (C57Bl/6) and human leukocyte antigen ( Equivalent number of 293-gp96-Ig-protein S and AD100-gp96-Ig-protein S cells that produce 200 ng gp96-Ig or PBS were injected via the subcutaneous (s.c.) route in C57Bl/6 and HLA-A2.1 transgenic mice. Mice received vaccination on days 0 and 30 and were sacrificed 5 and 30 days after vaccination. Spleen, lungs, and bronchoalveolar lavage (BAL) were collected and processed into single-cell suspension.

BAL and Lung Harvest and Cell Isolation
For mouse samples, spleens were collected, and tissues processed into single-cell suspension. Leukocytes were isolated from spleen by mechanical dissociation and red blood cells were lysed by lysing solution. BAL was harvested directly from euthanized mice via insertion of a 22-gauge catheter into an incision into the trachea. Hanks' Balanced Salt Solution (HBSS) was injected into the trachea and aspirated 4 times. Recovered lavage fluid was collected and BAL cells were gathered after centrifugation.
To isolate intraparenchymal lung lymphoid cells, the lungs were flushed by 5 ml of prechilled HBSS into the right ventricle. When the color of the lungs changed to white, the lungs were excised avoiding the peritracheal lymph nodes. Lungs were then removed, washed in HBSS and cut into 3-mm pieces, and incubated in Iscove's Modified Dulbecco's Medium containing 1 mg/ml collagenase IV (Sigma) for 30 min at 37°C on a rotary agitator (approximately 60 rpm). Any remaining intact tissue was disrupted by passage through a 21-gauge needle. Tissue fragments and majority of the dead cells were removed by a 250micrometer mesh screen, and cells were collected after centrifugation.

Ex Vivo Stimulation and Intracellular Cytokine Staining
Spleen and intraparenchymal lung lymphocytes from immunized and control animals were analyzed for protein S-specific CD8+ T cell responses. 1-1.5 × 10 6 cells were incubated for 20 h with 2 protein S peptide pools (S1 and S2, homologous to vaccine insert) (JPT Peptide Technologies, Berlin, Germany; PM-WCPV-S1). Peptide pools contain pools of 15-meric peptides overlapping by 11 amino acids covering the entire S protein (UniProt: P0DTC2) of SARS-CoV-2. Pool of 315 peptides (delivered in two subpools of 158 and 157 peptides) derived from a peptide scan through Spike glycoprotein. Peptide pools were combined (S1+S2) and used at a final concentration of 1.25 ug/ml of each peptide, followed by addition of Brefeldin A (BD GolgiPlug ™ ; BD Biosciences, San Diego, CA, USA) (10 ug/ ml) for the last 5 h of the incubation. Stimulation without peptides served as background control. The results were calculated as the total number of cytokine-positive cells with background subtracted. Peptide stimulated and non-stimulated cells were first labeled with live/dead detection kit (Thermo Fisher Scientific, Waltham, MA, USA) and then resuspended in BD Fc Block (clone 2.4G2) for 5 min at room temperature prior to staining with a surface-stain cocktail containing the following antibodies purchased from BioLegend ® ( Data were collected on Spectral analyzer SONY SP6800 instrument (Sony Biotechnologies, Inc, San Jose, CA, USA). Analysis was performed using FlowJo ™ software version 10.8 (Tree Star Inc, Ashland, OR, USA). Cells were first gated on live cells and then lymphocytes were gated for CD3+ and progressive gating on CD8+ T cell subsets. Antigen-responding CD8 (cytotoxic) T cells (IFNg, or IL-2, or TNFa-producing/ expressing cells) were determined either on the total CD8+ T cell population or on CD8+ CD69+ cells. ex vivo stimulation and intracellular staining) were fixed and permeabilized with Cytofix/Perm solution (BD) and then stained for intracellular cytokines: IFNg, and IL-2. Cells were acquired on SP6800 Sony instrument and data analyzed using FlowJo software version 10.8. Data were analyzed using forward sidescatter single-cell gate followed by CD45, CD3, and CD8 gating, then pentamer gating within CD8+ T cells. These cells were then analyzed for expression of markers using unstained and overall CD8+ population to determine the placement of the gate. Singlecolor samples were run for compensation and fluorescence minus 1 control sample were also applied to determine positive and negative populations, as well as channel spillover.

STATISTICS
All experiments were conducted independently at least 3 times on different days. Comparisons of flow cytometry cell frequencies were measured by the 2-way analysis of variance (ANOVA) test with Holm-Sidak multiple-comparison test, *p<0.05, **p<0.01, and ***p<0.001, or unpaired T-tests (2-tailed) were carried out to compare the control group with each of the experimental groups (alpha level of 0.05) using the Prism software (GraphPad Software, San Diego, CA, USA). Welch's correction was applied with the unpaired T test, when the p-value of the F test to compare variances were ≤0.05. Data approximately conformed to Shapiro-Wilk test and Kolmogorov-Smirnov tests for normality at 0.05 alpha level. Data were presented as mean ± standard deviation in the text and in the figures. All statistical analysis was conducted using GraphPad Prism 8 software.

AD100 and HEK-293 Express gp96-Ig and Protein S
Cell-based secreted heat shock protein technology has been previously validated in numerous animal models and in humans (37)(38)(39)42). The secreted form of gp96 protein (gp96-Ig) was generated by replacing the c-terminal, KDELretention sequence of human gp96 gene, with hinge region and constant heavy chains (CH2 and CH3) of human IgG1 (43) ( Figure 1A). The pcDNA 3.1(-) vector was used to express SARS-CoV-2 spike (S) protein (in this manuscript referred as protein S) ( Figure 1A), due to its propensity to constitutively express large amounts of the protein in mammalian cells. Complementary (c) DNA encoding the full-length SARS-CoV-S glycoprotein included Kozak sequence (GCCACC) to optimize expression in eukaryotic cells and the open-reading frame contained endogenous leader sequence, transmembrane, and cytosolic domains.
Vaccine cells, 293-gp96-Ig-S and AD100-gp96-Ig-S, were generated by cotransfection of AD100 and HEK293 cells with plasmids encoding gp96-Ig (B45) and protein S (pcDNA 3.1) and selection with G418 and L-histidinol as described in Methods. We confirmed by ELISA that both stable transfected cell lines secreted gp96-Ig into culture supernatants at a rate of 125 ng/ml/ 24 h/10 6 vaccine cells ( Figure 1B). Our previous data indicate that gp96-Ig accumulation in cell culture supernatant is linear and time dependent (35,43).
Protein S expression by the vaccine cells was confirmed by analyzing vaccine cell lysates on SDS page, blotting with anti-S A R S -C o V -2 S a n t i b o d y ( F i g u r e 1 C ) and by immunofluorescence ( Figure 1D). We observed expression of full-length protein S (250 kDa) only in AD100 transfected cell lines (lanes 2-4) but not in nontransfected AD100 cell line (lane 1). In addition, we observed molecular weight bands of 120 and 130 kDa that could represent cleavage products of full length protein S (protein S1 and S2) and/or gp96-Ig fusion protein chaperoning the protein S peptides. The expected molecular weight of gp96-Ig fusion protein is 116 kDa. Additional bands, of~70 kDa, were found to be expressed only in the transfected cell line and were not observed in the nontransfected AD100 cells. However, nonspecific bands of 100, 60, and 40 kDa were observed in the AD100 parental cell line. Recombinant protein S1 130 kDa was used as a positive control. We calculated the ratio of protein S to b-actin expression ( Figure 1C) and confirmed the expression of protein S by immunofluorescence ( Figure 1D). We observed cytoplasmic and transmembrane distribution of protein S in AD100-gp96-Ig-S cell line. We therefore confirmed the expression of gp96-Ig and S protein in our AD100 cell line and used it for immunogenicity studies as described below.

Secreted gp96-Ig-S Vaccine Induces CD8+ T Cell Effector Memory Responses in the Lungs
Our vaccination strategy is based on the quantity of gp96-Ig-S secreted by the vaccine cells to stimulate CD8+ CTL responses via APC cross-presentation. The vaccination dose, is therefore, standardized to a set amount of gp96-Ig secreted by 10 6 vaccine cells within 24 h. It has been well established from our previous vaccine immunogenicity studies that the optimal dose for induction of CD8+ T cell specific responses in mice is 200-500 ng/ml (33,35,38,39,43). Here, we used 200 ng/ml to immunize mice with AD100-gp96-Ig-S vaccine. Mice were vaccinated via the s.c. route and, after 5 days, the frequency of T cells within spleen, lungs (lung parenchyma), and BAL cells (lung airways) was determined. We observed significant increase in the frequencies of CD8+ T cells in the spleen and lungs, but not within the BAL of vaccinated mice (Figure 2A). Frequency of CD4+ T cells was unchanged between vaccinated and control mice in all analyzed tissues. It is well established that vaccination with gp96-Ig induces CD8+ TEM differentiation (33,37,39). Here, we confirmed that gp96-Ig-S vaccine primes strong effector memory CD8+ T-cell responses as determined by analysis of CD44 and CD62L expression ( Figure 2B). Whereas the frequency of naïve (N), CD44-CD62L+CD8 T cells and central memory (CM), CD44+CD62L+ CD8+ T cells was unchanged, we found statistically significant increase of TEM CD44+ CD62L-CD8+ T cells within the spleen and lungs ( Figure  2B). In addition, we observed a trend of more TEM CD8+ T cells within the CD8+ T cells in the BAL ( Figure 2B). TRM cells are a distinct memory T cell subset compared to CM and EM cells (44) that are uniquely situated in different tissues, including lungs (30,31). One of the canonical markers of TRM T cells is CD69 (20,44,45). We found that there was a significant increase in the frequency of CD8+CD69+ T cells in vaccinated mice compared to control, non-vaccinated mice in both spleen and lungs ( Figure 2C). Even though the frequency of CD8 +CD69+ T cells was the highest in the BAL compared to spleen and lungs, we did not observe a difference in their frequencies between vaccinated and control mice. Overall, vaccination with AD100-gp96-Ig-S induced robust TEM and CD69+ CD8+ T cell responses in both spleen and lungs. Our vaccine can therefore successfully elicit both systemic and tissue-specific immune response, which is pivotal in conferring robust immunity against infection such as against SARS-CoV-2.

Both Protein S-Specific CD8+ and CD4+ T Helper 1 T Cell Responses Are Induced by gp96-Ig-S Vaccine
To evaluate polyepitope, protein S-specific CD8+ and CD4+ T-cell responses induced by gp96-Ig-S vaccination, we used pooled S peptides (S1+S2) and a multiparameter intracellular cytokinestaining assay to assess Th1 (IFNg+, IL-2+, and TNFa+), CD8+ and CD4+ T cells (Figure 3). Spleen and lung cells were tested for responses to the pool of overlapping protein S peptides (S1 + S2) and all of the vaccinated animals showed significantly higher magnitude of the protein S-specific T cell responses against S1 and S2 epitopes compared with non-vaccinated controls ( Figures  3A-E). Increase in the vaccine-induced Th1 CD8+ T cell responses (IFNg+, IL-2+, and TNFa+) was noted in both spleen and lungs ( Figures 1A, B), whereas Th1 CD4+ T cell responses (IFNg+, IL-2+, and TNFa+) were induced only in lungs ( Figures 3C, D).
The proportion of the protein S-specific CD8+ T cells that produce | Secreted gp96-Ig-S vaccine induces protein S specific CD8+ and CD4+ T cells in the spleen and lung tissue. Five days after the vaccination of C57Bl6 mice, splenocytes and lung cells were isolated from vaccinated and control mice (PBS) and in vitro restimulated with S1 and S2 overlapping peptides from SARS-CoV-2 protein in the presence of protein transport inhibitor, brefeldin A for the last 5 h of culture. After 20 h of culture, ICS was performed to quantify protein Sspecific CD8+ and CD4+ T-cell responses. Cytokine expression in the presence of no peptides was considered background and it was subtracted from the responses measured from peptide pool stimulated samples for each individual mouse. (A, B) CD8+ T cells from spleen and lungs expressing IFNg, TNFa, and IL-2 in response to S1 and S2 peptide pool; (C, D) CD4+ T cells from spleen and lungs expressing IFNg, TNFa, and IL-2 in response to S1 and S2 peptide pool; (E) Representative dot plot of gated CD8+ T cells in lungs expressing indicated cytokines (IFNg; IL-2, and TNFa) in vaccinated and non-vaccinated (PBS, control) HLA-A2 mice at day 5 (F) Proportion of antigen (protein S)-experienced CD8+ and CD4+ T cells isolated from spleen and lung tissue expressing IFNg, TNFa, or IL-2 after o/n stimulation with S1 + S2 peptides. Pie charts corresponding to cytokine profiles of CD8+ and CD4+ T cells isolated from spleen and lung tissue; (G) Pie charts corresponding to cytokine profiles of CD8+ CD4+ T cells isolated from spleen and lung tissue after o/n stimulation with S1 + S2 peptides. Assessment of the mean proportion of cells making any combination of one to three cytokines (IFN-g, TNFa, IL-2). Data represent at least two technical replicates with three to six independent biologic replicates per group. *p<0.05, ***p<0.001. Kruskal-Wallis ANOVA with Dunn's multiple comparisons tests were applied. Asterisks (*) above or inside the column denote significant differences between indicated T cells producing cytokines in vaccine versus control (PBS) at 0.05 alpha level. ANOVA, analysis of variance; ICS, intracellular cytokine staining; IFN, interferon; IL, interleukin; PBS, phosphate-buffered saline; TNF, tumor necrosis factor. IFNg (26.8%) was significantly reduced in the lungs (6.2%), while both TNFa and IL-2 productions were increased in the lungs (41% and 52%, respectively) compared to spleen (27.4% and 45.7%, respectively) ( Figure 3F). We found that the proportion of the protein S-specific CD4+ T cells that produce IFNg was higher in the spleen than in the lungs [55.5% (spleen) versus 30.3% (lungs)], whereas IL-2 production was higher in the lungs than in the spleen [33% (lungs) versus 11.1% (spleen)] ( Figure 3F). Further assessment of protein S-specific CD8+ and CD4+ T cells in the spleen and lungs revealed that the vast majority of protein Sspecific CD8+ and CD4+ T cells, irrespective of their location, synthesized only 1 cytokine ( Figure 3G). It was therefore confirmed that a polyepitope, S-specific, CD4+, and CD8+ T cell response was generated in the spleen and lungs to different extents, providing a strong vaccine-induced Th1 cellular immune responses.

Induction of SARS-CoV-2 Protein S Immunodominant Epitope-Specific CD8+ T Cells in the Lungs and Airways of Vaccinated HLA-A2.1-Transgenic Mice
Recently, it was reported that SARS-CoV-2-specific memory CD8+ T cell responses generated against cognate antigens positively correlate with a number of symptom-free days after infection (14,16). Therefore, it is important to develop vaccines that can elicit SARS-CoV-2-specific CD8+ T cells. Having identified overall T-cell responses to SARS-CoV-2 protein S (Figure 3), we wanted to determine whether gp96-Ig-S vaccine induced HLA class I-specific cross-presentation of immunodominant SARS-CoV-2 protein S epitopes. In order to do this, we used transgenic HLA-A2.1 mice and HLA class I pentamers as probes to detect CD8+ T cells specific for two immunodominant SARS-CoV-2 protein S epitopes: YLQPRTFLL (YLQ) (aa 269-277) and FIAGLIAIV (FIA) (aa 1220-1228) in vaccinated mice ( Figures 4A-C). We found that the vaccine effectively induces both YLQ+CD8+ T cells, as well as FIA+CD8+ T cells in the spleen, lungs, and BAL ( Figures 4A-C). Interestingly, we found the highest frequency of YLQ+CD8+ T cells in the BAL of vaccinated mice and the lowest frequency of YLQ+ and FIA+ CD8+ T cells was observed in the lungs 5 days after primary vaccination. Importantly, we observed that after contraction of vaccine induced YLQ+CD8+ and FIA+CD8+ T cell responses in spleen, lungs and BAL, frequency of S-specific CD8+ T cell responses is preserved and it is significantly higher in BAL compared to non-vaccinated controls (Figures 4A-C).
Upon further phenotype analysis of YLQ+CD8+ T cells, it was confirmed that they express both CD69 and CXCR6 ( Figure 4D). Particularly, we found that during primary response (day 5) all YLQ+CD8+ T cells in the BAL were also CD69+ and CXCR6+, and the frequency of YLQ+CD8+CXCR6+ cells was significantly higher in the BAL compared to the lungs. However, during contraction phase (day 30) YLQ+CD8+ cells (Figures 4A,  B and Supplementary Figure 1) in the SPL, lungs and BAL express both, CD69 and CXCR6 ( Figure 4D).
We confirmed that S-specific CD8+ and CD4+ T cells are present in significantly higher frequencies compared to unvaccinated controls 30 days after single dose vaccination ( Figures 5A-E). The proportion of the protein S-specific CD8+ T cells that produce IFNg was comparable in the spleen and lungs (32.7% and 28%, respectively) while we observed increased proportion of the cells that produce TNFa in the lungs (47.3%) compared to spleen (24.5%) ( Figure 5C). In addition, we found that the proportion of the protein S-specific CD8+ T cells that produce IL-2 was higher in the spleen than in the lungs (42.6% and 24.5%, respectively) while the proportion of S-specific CD4+ T cells that produce IL-2 was higher in lungs than in spleen (44.4% and 33.3%, respectively) ( Figure 5C).
Furthermore, we observed increase in the frequency of YLQ+ CD8+ T cells in the spleen, lungs and BAL in all animals after receiving second dose of gp96-S vaccine at day 30 ( Figure 5F). Despite the similar magnitudes of primary and secondary CD8+ T cell expansion, contraction of vaccine induced Ag-specific CD8+ T cell responses was different during primary and secondary response. We observed more rapid contraction of CD8+ T cells after primary vaccination, with 68% loss of Ag-specific CD8+ T cells from day 5 to 30 ( Figure 5G), compared to 25% decrease in Ag-specific CD8+ T cells during secondary response ( Figure 5G). Thus, the program of contraction in CD8+ T cells that are responding to second dose of vaccine was prolonged, resulting in increased frequencies of memory S-specific CD8+ T cells.

DISCUSSION
Our vaccine approach is based on the gp96-Ig platform technology that elicits potent, antigen-specific CD8+ T-cells. This proprietary secreted heat shock protein platform has been successfully used to induce immunogenicity against tumors, HIV/SIV, Zika, and malaria in different animal models (37)(38)(39)(46)(47)(48). Importantly, this vaccine strategy has shown success in delaying virus acquisition, as well as in improving the survival of NSCLC patients in clinical trials (38,42).
The principle of a cell-based vaccine relies on the ability of gp96-Ig to chaperone antigenic proteins to be efficiently endocytosed and cross-presented by activated dendritic cells (DC) to CD8+ T cells, thereby stimulating an avid, pathogenspecific T-cell response (33,34,(37)(38)(39). We adapted this cell-based technology to create a vaccine that delivers SARS-CoV-2 spike (S) protein directly to DCs, so that primed and activated SARS-CoV-2 protein S-specific CD8+ T cells can identify and kill SARS-CoV-2 infected lung epithelial cells. We generated vaccine cells by coexpressing secreted gp96-Ig and full-length protein S. Gp96-Ig is an endoplasmic reticulum chaperone that, together with TAP (transporter associated with antigen processing) and calreticulin in the endoplasmic reticulum, is thought to constitute a relay line for antigenic peptide transfer from the cytosol to MHC class I molecules in a concerted and regulated manner (49,50). The gp96antigenic peptide complexes are predominantly internalized by subsets of APCs through cell surface receptor CD91. Internalized gp96 can effectively present the associated peptides to MHC class I and II molecules and thus activate specific CD8+ and CD4+ T-cell responses (34,39,51,52). We expressed full-length protein S in the vaccine cells to ensure broad representation of all immunodominant protein S peptides (S1-and S2-derived peptides) by secreted gp96-Ig. Since coronaviruses assemble in the compartment between the endoplasmic reticulum and Golgi apparatus (53,54) and the S leader directs it to the endoplasmic reticulum, the native leader sequence of protein S was retained, as well as transmembrane and cytosolic domain. We confirmed in previous studies that secreted gp96-Ig provides immunologic specificity for the antigenic repertoire expressed inside the cells, including surrogate antigen ovalbumin, as well as numerous tumor or infectious antigens, but does not cross-immunize to different cell-derived antigens (35,(37)(38)(39). Our data are consistent with the explanation that S1 and S2 peptides associated with secreted gp96-Ig are transferred to and presented by class I and II MHC and stimulate a S1-and S2-specific CD8+ and CD4+ T cell response. We confirmed that vaccination with AD100-gp96-Ig-S induces CD8+ T cells specific for S1-and S2-immunodominant epitopes in both lungs and airways. Most importantly, this is a proof-of-concept study that will be applied to other structural proteins such as nucleocapsid protein, membrane protein, and nonstructural proteins such as NSP-7, NSP-13 of ORF-1 that all have been reported to be important in induction of SARS-CoV-2specific CD4+ and CD8+ T cell responses in convalescents (7,10,11,55).  In agreement with our previous findings (33,38,39), the gp96-Ig vaccine resulted in the preferential induction of CD8+ T cell responses systemically and in epithelial compartments. However, this is the first report about the increase in the frequencies of vaccine induced CD8+ T cells in the lungs ( Figure 1A). TEM CD8+ T cells are considered to constitute the frontline defense within the different epithelial compartments including lungs and airways, which promptly recognize and kill infected cells. Our data suggest that after a single dose of AD100-gp96-Ig-S immunization there is preferential compartmentalization of TEM and TCM immune responses in the lungs and BAL compared to the spleen where a majority of cells are naïve CD8+ T cells ( Figure 1B). However, additional memory cells without migratory potential such as TRM CD8+ T cells, exist within the tissues including lungs and airways (20,44,45,56). Since TRM are uniquely situated in the lungs to immediately respond to reinfection, by inducing the protein-S-specific CD8+ T cells that home to the lungs, gp96-Ig-S vaccine provides an ideally balanced generation of both arms, TRM and TEM, of the memory response in the lungs. To further gauge the effect of gp96-Ig vaccination on the induction of epitope specific immunogenicity, we used pentamers to detect S1 and S2 epitope-specific CD8+ T cell responses. We found that gp96-Ig induced the highest frequencies of S1-and S2-epitope specific CD8+ T cells in the airways. In light of the new findings about exclusive highly clonally expanded SARS-CoV-2-specific CD8+ T cells with preferentially expressed tissue-resident genes (XCL1, CXCR6, and ITGAE) in the BAL of moderate COVID-19 cases (57) and not in the critical/severe COVID-19 patients, induction of SARS-CoV-2 specific CD8+ T cells that home to airway epithelium emphasizes the importance of developing vaccination strategies that induce TRM antigen-specific CD8+ T cells that will improve efficacy of vaccination against respiratory pathogens including SARS-CoV-2.
It is well appreciated that the antigen presenting cells at the site of immunization direct the imprinting of the ensuing T-cell response and control the expression of trafficking molecules (58). Priming of CD8+ T cells by CD103+ DC was found to promote TRM CD8+ T cell differentiation and migration into peripheral epithelial tissues, including lungs (56,59). Our previous studies indicated that gp96-Ig immunization increases frequency of CD11c high MHC class II high CD103+ cells at the vaccination site (33). In light of our previous findings and findings of Bedoui et al. (60) that CD103+ DCs are the main migratory subtype with dominant cross-presenting ability, induction of CD103+ DCs by gp96 represents an ideal vaccination strategy for priming effective and durable immunity in the epithelial tissues. We also hypothesize that secretory gp96-Ig can be captured by lung DC upon s.c. inoculation and in that way directly induce tissue memory responses and even more importantly support their longevity in the tissue. It was previously shown that, based on differences in the localization and functions, there are two different subsets of lung TRM cells: airway TRM and interstitial TRM (61)(62)(63). CXCR6-CXCL16 interactions are crucial in controlling the localization of virus-specific TRM CD8+ T cells in the lungs and maintaining the airway TRM cell pool (20).
Moreover, blocking of CXCR6-CXCL16 interactions significantly decreases the steady-state migration of TRM cells into airways, so vaccine induced SARS-CoV-2 S specific CD8+ T cells that express CXCR6 fulfill one of the major requirements for continued CXCR6 signaling in maintaining the airway TRM pool (20). In vivo anti-CD3 labelling of the lymphocytes is the best method for distinguishing intravascular vs extravascular antigen specific cells. Future studies should include in vivo labelling to distinguish if vaccine induced S-specific immune responses are circulating/effector or local/tissue resident cells.
We have confirmed in different infectious vaccine models that gp96-Ig carries all peptides of a cell that are selected in the recipient/vaccinee for MHC I, having the broadest, theoretically possible antigenic epitope-spectrum for cross-priming of CD8+ T cells by any MHC I type. Here, we showed that AD100-gp96-Ig-S resulted in the polyepitope and polyfunctional protein Sspecific CD8+ and CD4+ T cell responses (Figure 3). When stimulated in vitro with S1+S2 peptides, spleen and lung CD8+ T cells produce IFNg, TNFa, and IL-2 cytokines ( Figure 3) with CD8+ T cells in the lungs producing significantly less IFNg than the CD8+ T cells in the spleen. It is known that enhanced activation, resulting from high levels of inflammation, induces CD8+ T cells entering the lungs to produce regulatory cytokines (64) that initiate "dampening" of the immune response in order to prevent any excessive damage of the lung tissue. In addition, we report that CD4+ T cells in the lungs produce all 3 Th1 cytokines in equal ratio ( Figure 3F). This finding is in line with our previous report discussing gp96-Ig induced SIV-specific CD4+ T cells in the lamina propria which were almost the exclusive producers of IL-2 (33). Further studies will therefore be required to evaluate the role of protein-S specific CD4+ T cells in the induction of B cell and antibody responses. Previously, we have shown that gp96 is a powerful Th1 adjuvant for CTL priming and for stimulation of Th1 type antibodies in mice and nonhuman primates (37,38). In addition, we will evaluate memory responses after a single and booster dose to establish the best vaccination protocol for future challenge studies.
In summary, we provide a paradigm for a novel vaccine development approach capable of induction of cellular immune responses in epithelial tissues such as the lungs. Structure-guided SARS-CoV-2 S protein combined with a safe and efficacious gp96-Ig vaccine platform can pave the way for a protective and durable immune response against COVID-19. This is a first demonstration of the utility and versatility of our proprietary secreted gp96-Ig SARS-CoV-2 vaccine platform that can be rapidly engineered and customized based on other and future pathogen sequences. Furthermore, the platform is proof of concept for the prototype vaccine approach for similar pathogens that require induction of effective TRM responses in epithelial tissues.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by University of Miami Institutional Animal Care & Use Committee.

ACKNOWLEDGMENTS
This manuscript has been released as a pre-print at bioRxiv (65). We dedicate this work to the late Dr. Eckhard Podack. We are grateful to all members of Strbo laboratory, Heat Biologics, Inc. CEO Jeff Wolf, Chief Scientific and Operating Officer, Jeff Hutchins, and Director Discovery Sciences, Eric Dixon, for their overall support, advice, and editorial contributions. We are grateful to Mr. William Strong for his generous gift for vaccine development. We thank Sylvester Comprehensive Cancer Center Flow Cytometry Shared Resource (SCCC FCSR) staff for their help.