Designing a therapeutic SARS-CoV-2 T-cell-inducing vaccine for high-risk patient groups

Here, we describe the preliminary results of an experimental vaccination of a self-experimenting healthy volunteer with eight SARS-CoV-2-derived peptides: five predicted to bind to HLA class I molecules (CD8 peptides) and three predicted to bind to HLA-DR molecules (CD4 peptides). The vaccine formulation also included one long and one short CMV-pp65-derived peptide that had previously been administered to the same individual and could thus act as positive controls. It further contained the new adjuvant XS15 and was administered as an emulsion in Montanide as a single subcutaneous (s.c.) injection. Peripheral blood mononuclear cells (PBMCs) isolated from blood drawn on day 36 before vaccination and day 19 after vaccination were assessed using an ex vivo Interferon- γ ELISpot assay. We detected strong vaccine-induced T-cell responses against all four CD4 peptides and against the recall CMV CD8 epitope, but found no immune responses against the five predicted SARS-CoV-2 CD8 peptides. Antibody reactivity against all the SARS-CoV-2 CD4 peptides, as detected using ELISA, was negative or marginal. We interpret these results in terms of the prospects of a therapeutic vaccine to be applied in symptomatic COVID-19 patients. An advantage of this approach is the possibility to assess efficacy or failure within a short time after vaccination. predicted to bind multiple HLA-DR self-administered healthy volunteer vaccination). Vaccination was performed with synthetic peptides solubilized in water and 20% DMSO including 50 µg XS15 as an adjuvant, emulsified in Montanide ISA51 VG in a total volume of 0.5 ml. The amount of peptides administered in the vaccine was 240 µg for each peptide.


Introduction
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which causes corona virus disease 2019 (COVID- 19), was first sampled and described in December 2019 in Wuhan, China 1 . It has a high basic reproduction number (R0), estimated at 3.28 2 , with person-toperson transmission even occurring between asymptomatic or presymptomatic individuals 3 . This virus has emerged as a global threat and brought many national health systems to the limit of their treatment capacities.
COVID-19 symptoms can be mild or even negligible and initial manifestations encompass nonspecific flu-like symptoms, such as fever, cough and fatigue 4 . Most infected individuals will recover without medical intervention, but a relevant number of patients require medical attention due to severe or even life-threatening symptoms, such as pneumonia, acute respiratory distress syndrome and a syndrome with similar features to a cytokine storm 5 . Increased cytokine levels (including IL-6, IL-10 and TNF), lymphopenia (affecting both CD4 + and CD8 + T cells) and a reduced expression of IFNγ in CD4 + T cells have been observed and reported to be associated with COVID-19 severity 4 .
While a severe clinical manifestation of SARS-CoV-2 infection may affect any person, risk groups for disease complications mainly include individuals with (multiple) co-morbidities and the elderly, who have a disproportionally high mortality 6 .
The international scientific community is working at an unprecedented pace to find therapeutic and prophylactic responses to COVID-19. This includes attempts to develop vaccines. Important strategies to pursue are: passive immunotherapies, such as plasma transfer from convalescent individuals 7,8 , as practiced for other infectious diseases since 1892; and active immunotherapies, intended to treat at-risk COVID-19 patients. To the best of our knowledge, the potential of therapeutic vaccines has yet to be widely considered.
We previously reported the results of an experimental approach based on a peptide vaccine that comprises multiple peptides consisting of various viral epitopes. Peptides were administered together with Montanide and a novel lipopeptide adjuvant, the TLR2/1 ligand XS15, to the same healthy volunteer 9 .
Building on this previous work, here we report the findings from self-experimentation involving the same volunteer. The previously tested protocol for vaccination was used, but this time the vaccine contained in silico predicted peptide sequences from a SARS-CoV-2 virus isolate. Relevant data that were available very early from a virus isolate from China were used for this purpose 10 .

Experimental procedures
The sequences described here are derived from a virus isolate obtained from a worker at the Wuhan fish market (https://www.ncbi.nlm.nih.gov/nuccore/MN908947 10 ). Using the SARS-CoV-2 nucleocapsid and envelope protein sequences from this isolate and the prediction software SYFPEITHI (www.syfpeithi.de; 11 ) we predicted candidate CD8 (HLA-A*01 or -B*08) and candidate CD4 T-cell epitopes (HLA-DRB1*11) based on the HLA allotype of the healthy volunteer. The resulting sequences were then run on other widely used more up to date software for confirmation. We chose the nucleocapsid protein, because nucleoprotein peptides from many other viruses are potent T-cell epitopes (see: www.iedb.org).
Peptide selection was completed on January 23, 2020. We synthesized eight peptide sequences that had the highest SYFPEITHI-scores and also appeared to be good candidates, based on our subjective experience, for the SARS-CoV-2 nucleocapsid or envelope proteins (i.e., they were predicted to be HLA ligands for the corresponding HLA molecules) and that had no cysteine residues (for chemical stability reasons). They were complemented by two CMV-derived peptides that had previously elicited immune responses in the same individual and could thus be used as positive controls 9 . These ten peptides ( Table 1) were synthesized in-house using automated peptide synthesis.
The personalized vaccine consisted of the peptides (240 µg or 720 µg of each; see Table 1.) solubilized in water and 20% dimethyl sulfoxide (DMSO). The vaccine contains two out of the ten candidate peptide sequences derived from the SARS-CoV-2 nucleocapsid predicted to bind to HLA-DR, shown in Table 2 (respective predictions were performed much later, when the idea of the potential merit of a therapeutic T-cell epitope-only vaccine had occurred to us). It further contained the lipopeptide adjuvant XS15 (50 µg) and was administered as an emulsion in Montanide™ ISA51 VG (Seppic, Paris, France). A total volume of 0.5 ml was selfadministered subcutaneously (s.c.) under the skin of the abdomen on March 6, 2020. The vaccinated individual did not report any flu-like symptoms or other symptoms consistent with SARS-CoV-2 infection during the relevant period. The individual did not report having visited any areas designated as high risk for virus transmission.
Heparin-anticoagulated peripheral venous blood was drawn 36 days before and 19 days after vaccination PBMCs were isolated via density centrifugation 12 and either frozen and stored in liquid nitrogen (pre-vaccination PBMCs) or used fresh (post-vaccination PBMCs). In addition, plasma diluted 1:1 with phosphate-buffered saline (PBS) was obtained from the vaccinated volunteer during PBMC isolation. Blood serum from three healthy blood donors (HDs) was obtained after informed consent for use as control samples.
PBMCs obtained pre-vaccination were thawed and rested overnight in culture medium (IMDM with 10% heat-inactivated human serum, 100 units/ml penicillin, 0.1 mg/ml streptomycin and 50 µM β-mercaptoethanol), containing 1 µg/ml DNase I. PBMCs obtained post-vaccination were used directly ex vivo after density gradient separation. In both cases, 300,000 cells per well were plated on a pre-coated ELISpot plate. The negative control (DMSO and water) and vaccine peptides were respectively tested in six and three replicates. For phytohemagglutinin (PHA-L) stimulation (positive control), duplicates were plated with 150,000 cells per well. Peptides were added at a concentration of 5 µg/ml (for HLA class I peptides) or 2.5 µg/ml (for HLA-DR peptides), and PHA-L was used at 10 µg/ml. The interferon-γ (IFNγ) ELISpot was performed as described previously 9,12 with incubation at 37°C in a 7.5% CO2 atmosphere for 26 h before development.
Antibody reactivity in the plasma obtained from the vaccinated volunteer and in serum from the three HDs was used in an in-house ELISA performed as described previously 12 . Briefly, 96well plates were incubated for 16 h with the respective peptides (produced in-house at the University of Tübingen, Department of Immunology, Tübingen, Germany) or with recombinant proteins (SinoBiological, Peking, China), utilized at a concentration of either 35 µg/ml or 1 µg/ml. Plasma or serum samples were diluted 1:500 (previously determined to be optimal for the analysis of peptide-antibody reactivity). To differentiate between IgG-and IgMantibodies, peroxidase-conjugated goat anti-human IgG-and IgM-antibodies were applied in parallel. Ortho-phenylene-diamine was used as a substrate. The reactivity was measured with an ELISA-reader at 450 nm and is expressed as the optical density multiplied by 1,000 (OD x 10 3 ). Cut off values were determined testing sera from healthy blood donors against the respective antigens/ peptides. Values above the mean of OD x 10 3 plus twice the standard deviation were defined as antibody positive.

Ethical and scientific considerations
As previously reported in a comparable self-vaccination study by the same human volunteer 9 , we consider this to be an ethically and legally legitimate form of experimentation 13 . Selfexperimentation is a special case of research and limited to individual subjects. It is essential that such self-experimentation does not contravene any questions of personal interest or ethical imperative. Coercion and dependency can be excluded here, ensuring decision-making autonomy. The volunteer in this case is a renowned expert in immunology and thus was able to understand any risks and implications of his own actions. It should therefore be clear that this study was permissible. In addition, relevant precedent exists, where human selfexperimentation has opened new avenues for research and contributed to medical progress 14 .
We are naturally aware that single case reports cannot provide conclusive evidence or any generalizable results. Rather these findings enable the formulation of new hypotheses. Case reports are attributed with a high sensitivity for detecting novelty and are deemed relevant for medical progress 15 . This study is intended as a starting point for intensified discussion and development, rather than a substitute for proper drug development and clinical trials.

Clinical aspects of the vaccination site
As expected with any s.c. vaccination using Montanide 16 or when further adding XS15 as an adjuvant 9 , a granuloma developed. It was palpable from day one and grew to a maximum size of approximately 3 x 5 x 1.5 cm by day 12. The skin surface temperature was 37.0°C at the center of the granuloma vs. 36.2°C on adjacent visually unaffected abdominal skin. The granuloma was described as a painless, slightly itchy induration that was sensitive to touch between days 9 and 14, and. It started to shrink in size from day 13. The CMV-pp65 HLA-DR epitope YQEFFWDANDIYRIF (amino acids 510-524), which was included as a positive control, was weakly recognized (mean: 16 spots/300,000 cells). Notably, three years prior, vaccination with this peptide resulted in a mean spot count of 525 four weeks after it had been administered, with the same adjuvant and protocol 9 . As expected from our previous experience, boosting with one additional vaccination more than one year later with the same peptide (CMV-pp65 510-524) gave a strong response (mean: 910 spots/300,000 cells). The CMV-pp65 HLA-A*01 epitope YSEHPTFTSQY (amino acids 363-373), which had induced a weak response in an ex vivo ELISpot three years prior with a mean spot count of only 12, and showed a negative response before the recent booster vaccination, now gave a mean spot count of 115. These results indicate that a three-year memory against these HLA class II-and class I-restricted CMV-pp65 epitopes prevails after a single vaccination. Importantly, the vaccinated volunteer had previously been tested as CMV seronegative.

T-cell responses
No other pre-existing T-cell responses were detectable prior to vaccination. The response for the five predicted SARS-CoV-2 CD8 peptides remained negative after vaccination. Based on previous findings, we speculate that some of these CD8 peptides may show positive results after a 12-day in vitro restimulation (ongoing experiment) or in ex vivo ELISpot with blood obtained later. We intend to test this with blood samples drawn about one month after vaccination.
By contrast, a strong T-cell response was induced against all three SARS-CoV-2 CD4 peptides by a single vaccination: a mean of 214 spots for the nucleoprotein-derived peptide ASAFFGMSRIGMEVT; 71 spots for the nucleoprotein-derived peptide IGYYRRATRRIRGGD; and 29 spots for the envelope protein-derived peptide FYVYSRVKNLNSSRV. Thus, all three SARS-CoV-2-derived CD4 peptides induced T cells to produce IFNγ, most probably representing a TH1 response. Phenotyping of the responding T-cell subsets and the production of further cytokines will be tested using intracellular cytokine staining (ongoing experiments).
Based on these promising results, a second vaccination attempt with a new peptide cocktail designed to be suitable for all individuals independent of their HLA type (see Table 3) was performed by the self-vaccinating individual on April 3, 2020.

Antibody responses
The antibodies contained in the plasma of the self-experimenting volunteer and serum of three healthy blood donors were tested using an ELISA. IgG and IgM antibodies against the three SARS-CoV-2 CD4 peptides were not detectable or showed negligible induction ( Table 4). We know from previous work that repeated vaccination with CD4 peptides in Montanide with or without additional adjuvants leads to the induction of antibodies 12 . We therefore speculate that antibodies against these epitopes may develop later on and will evaluate this with future blood samples. There seems to be a weak recall IgG response against the CMV-pp65 510-524 CD4 epitope YQEFFWDANDIYRIF (an 8-fold increase compared to the reactivity before vaccination). It is interesting that two out of the three healthy donors show high IgG reactivity against YQEFFWDANDIYRIF. Testing serum for antibodies against linear synthetic peptides from all proteins, not only from surmised neutralizing epitope-bearing ones, might be a useful complement to the use of recombinant proteins.
In addition, sera of the vaccinated volunteer and several healthy donors, obtained before the current pandemic, were tested with an ELISA covering SARS-CoV-2 proteins and peptides ( Table 5). Further, serum from one donor (HD CoV+-1), was included, who was tested SARS-CoV-2-positive according to medical routines (qRT-PCR) and had recovered from the infection.

Discussion
Ex vivo IFNγ-ELISpot results revealed strong T-cell responses against all three SARS-CoV-2derived CD4 peptides 19 days after a single s.c. vaccination with the peptides, Montanide and the toll-like receptor (TLR) 1/2 ligand XS15. Based on previous findings, we assume that these immune responses are most probably robust and durable 9 . There were no detectable T-cell responses against the five CD8 peptides and no measurable antibody responses.
We are currently considering the development of a therapeutic SARS-CoV-2 peptide vaccine that would induce the same profile of immune responses seen here in one self-experimenting human volunteer. The potential of such an approach is based on the following assumptions: 1. CD4 + TH1 cells should vigorously activate virus antigen-experienced B cells that should already pre-exist in most COVID-19-patients. An illustration for this assumption is provided in Fig. 2. These CD4 + T cells would also be expected to directly contribute to virus clearance and deliver strong T helper signals to the CD8 + T cells already primed during natural infection. The resulting enhanced activity could lead to more rapid virus clearance and/or transiently increased lung damage.
2. Vaccine-induced CD8 + T cells may appear later (based on our own experience with this vaccine approach), and should therefore not immediately induce or exacerbate a potential CD8 + T-cell-mediated damage of lung tissue. Note that all 10 proposed vaccine peptides contain embedded CD8 candidate epitopes predicted to bind to many HLA class I allotypes. Once activated, such CD8 + T cells should also contribute to faster virus clearance, but potentially also to temporarily increased lung damage, while the virus is present, in case this emerges as an issue with CD8 T cells.
3. Vaccine-induced antibodies against the viral peptides tested in this study may also appear much later, if at all. Thus, we assume that there is no immediate danger of vaccine-induced antibody-dependent enhancement (ADE) as described for anti-spike IgG in acute SARS-CoV infections 17,18 . However, there is a danger of vaccine-induced ADE in cases where the patient' s B cells have already been primed against epitopes from the regular seasonal coronavirus strains that infect humans, produce low amounts of antibodies, antibodies with low affinity or antibodies with the wrong class. This danger should also be considered and may be causal for ADE, triggered in the course of natural infection. In theory, vaccine-induced CD4 + T cells thus might also cause or exacerbate immunopathological effects indirectly.
4. Since we found IFNγ-producing T cells, it is very likely that TH1 CD4 + T cells are present. Therefore, there should be no disease enhancing-effects related to the induction of TH2-bias as described for other corona viruses 19 .
Of course, a vaccine designed for broad use must be designed to be suitable for all patients independent of their individual HLA allotypes. We propose that the ten SARS-CoV-2 nucleocapsid protein-derived peptides, carefully selected for this purpose to be promiscuous for most HLA-DR molecules ( Table 2) should be used for a first exploration of this vaccine concept that we called CovidFort. Together, these 10 peptides should cover the vast majority of the population. Additional peptides from the spike glycoprotein or other viral proteins not expected to contain any linear antibody epitopes could be added after careful examination of the risk of inducing ADE through non-neutralizing antibodies improving viral uptake as has been described for SARS-CoV-1 and dengue virus infection in humans and feline infectious peritonitis virus in cats [20][21][22] . There might also be strong antibody responses (high titers) against certain spike epitopes that are neutralizing, whereas the same antibodies at low titer result in enhancing 23 . The most rigid way to circumvent this difficult issue is to completely exclude potential linear antibody epitopes from early stage for therapeutic peptide vaccines. Thus, our strategy is to carefully select spike glycoprotein-derived CD4 candidate epitopes that are promiscuous for most HLA-DR allotypes, but that do not represent any potential B-cell epitopes, as far as can be judged using in silico analysis and analysis of serum antibodies from SARS-CoV-2-infected individuals (ongoing investigation on a vaccine approach called CovidFort). Only later, when we and others have gained additional experience, we would dare adding spike CD4 peptides, containing linear antibody epitopes confirmed as neutralizing, or determined as antibody targets in convalescent patients recovered from SARS-Cov-2infection, or even in seemingly healthy individuals with high titers of spike antibodies, since these individuals should have had an unnoticed SARS-CoV-2 infection that was cleared by particularly efficient immune responses. The relevant underlying theoretical considerations are illustrated in Fig. 2.
This strategy has the advantage of enabling quick assessment of vaccine efficacy as well as any adverse effects in terms of induction of immune responses and potential viral load reduction in SARS-CoV-2-infected patients with and without symptoms. We speculate that effects could be determined within one week after administration in the following manner: i) The percentage of SARS-CoV-2 PCR-negative patients on day seven (or later), would demonstrate the efficacy of virus clearance compared to the results for untreated patients.
ii) The vaccine-induced increase in the levels of antibodies against SARS-CoV-2 proteins, including the spike glycoprotein, could be easily measured in short intervals after vaccination.
In addition, the Montanide-induced granuloma, usually seen as a nuisance, here enables to quick termination of most of the vaccine-induced T-cell reactivity in patients that develop increased immunopathology in the first days after vaccination. Local injection of steroids into the granuloma or even its surgical removal should broadly terminate the induced immune responses, since most of the activated immune cells are initially located inside the granuloma within newly developing lymphoid structures.
Further considerations led us to the hypothesis that vaccination with T-cell epitopes alone might also prove useful for prophylactic vaccination. As the situation in zoonotic coronaviruses suggests, this strategy might even result superior to traditional vaccination approaches, where primarily B-cell responses against the vaccine antigens are to be induced, since in many such infections mainly memory T-cells and not antibodies seem to play the major role for long termimmunity against disease 24 . A vaccine solely consisting of T-cell epitopes would not be expected to completely prevent infection, as with traditional vaccines aiming for strong neutralizing antibody responses, but to help the patient's immune system to quickly resolve the commencing infection by fostering faster antibody production, enabled by the vaccineinduced CD4 + T cells. Overall, in contrast to many treatments currently introduced in COVID-19 and used on patients off-label, we expect therapeutic vaccines to have a particularly positive risk-benefit ratio.
This potential for a lack of prevention that may be seen as a disadvantage should be counterbalanced by the broader cross-reactivity of T cells, as seen in the example of influenza 25 . Thus, a T-cell epitope-based vaccine against influenza should be an efficient prophylaxis against several seasonal influenza strains, provided these strains contain the Tcell epitopes included in the vaccine (Fig. 2). In the case of SARS-CoV-2, this type of vaccine should also work for new viral serotypes and variants that already exist in so far unknown environments or could develop in the future.
Taken together, we propose a therapeutic immunization strategy targeting primarily T cells, but affecting B cells only indirectly, to be applied in at-risk patients after SARS-CoV-2 infection. We do not know at this time, whether such a strategy will be successful in eliminating the virus more rapidly or does comprise risks that may be harmful for the patients. An advantage of the approach is that it can be tested very quickly: Vaccinating infected elderly patients (e.g. >70 years old) as soon as possible after confirmed infection may yield results within only 7 days after vaccination. Lacking effects, improvement or deterioration of symptoms, an increase in antibody responses, or rapid virus clearance should guide us in the right direction.     26 . Only a selection of the resulting peptides will fit to the respective HLA molecules present based on their peptide specificity. The B cell will then present these peptides on the cell surface, with one peptide per HLA molecule. If a T cell is specific for exactly this peptide-HLA combination and if the B cell has been activated via its BCR-antigen contact, then the CD4 + T helper cell will deliver help to this B cell, both trough cellular interaction and cytokines. Since all viral proteins are presented on the B cell's HLA in this scenario, a nucleocapsid-specific T cell will also activate a spike-specific B cell. (B) If the virus-specific B cell takes up separate viral proteins via BCR-mediated phagocytosis, e.g., after previous destruction of viral particles by follicular dendritic cells, only peptides from these proteins will be presented on the B cell's HLA molecules. This is shown here for the spike glycoprotein. Thus, a nucleocapsid-specific T cell will not activate a spike-specific B cell in this constellation. (C) If the same B cells as described in B) are activated by a spike-specific CD4 + T helper cell, the B cell will now be activated. This constellation is described in the last part of the discussion. Simply replace "SARS-CoV-2 spike protein" with, e.g., "Influenza haemagglutinin" to make this congruent for the case of influenza infection. Abbreviations: HLA: human leucocyte antigen, SARS-CoV-2: severe acute respiratory syndrome coronavirus 2, TCR T-cell receptor.  Table 2. Selection of peptides from SARS-CoV-2 nucleocapsid predicted to bind to multiple HLA-DR molecules.

H.G. Rammensee has ownership interest in Immatics Biotechnologies
The genetic sequence known from a virus isolate obtained from a worker at the Wuhan fish market (https://www.ncbi.nlm.nih.gov/nuccore/MN908947 10 ), was used to predict promiscuous 15mer HLA-DR-binding candidate antigens from the SARS-CoV-2 nucleoprotein. Promiscuous means here that each of the sequences is predicted to bind to several different HLA-DR allotypes, as predicted with our in-house method described earlier 27 . The peptides included in the first vaccine provided in Table 1 are given in bold, those used for the second vaccine are given in italics. Note: Sequences of the currently prevailing SARS-CoV-2 isolates and reported variants, need to be carefully re-evaluated before use in vaccine development.