The solid-phase peptide synthesis (SPPS) described by Merrifield and modified by Houghten (18) was used for peptide synthesis. Briefly, methylbenzhydrylamine hydrochloride (MBHA) resin was swollen with dichloromethane (DCM) and neutralised with 10% triethylamine (TEA) in DCM. Coupling reaction involved activating alpha-L-Boc aa with dicyclohexilcarbodiimide (DCC) and 1-hydroxy-benzotriazole (HOBt) in N,N-dimethylformamide (DMF) for 1 h; coupling reaction verification was followed by the Kaiser test. The resin was then washed and treated with 40% trifluoroacetic acid (TFA) and 0.01% anisole solution in DCM for 30 min to remove t-Boc protecting group. Successive cycles of coupling, deprotection, and neutralisation followed until each peptide’s final sequence was obtained. Peptides were cleaved from the resin by low–high hydrogen fluoride (HF) procedure. Peptides were extracted with TFA, washed with ethyl ether and dried for further characterisation.

The peptides were analysed by reverse-phase high-performance liquid chromatography (RP-HPLC and Chromolith RP-18e column) and by matrix-assisted laser desorption/ionization-time-of-flight (MALDI-TOF) mass spectrometry (MS). Peptides were purified by semi-preparative chromatography. Selected fractions were collected and lyophilised for obtaining pure peptides for monkey and human immunisation.

A JASCO J-815 circular dichroism (CD) spectrometer was used for determining peptide CD (JASCO Corp., Tokyo, Japan) in the far ultraviolet range (190–250 nm) at 27°C and 37°C, using 0.1-cm path length quartz cuvettes and 1-nm bandwidth. Each spectrum was recorded as the average of three repeat scans in continuous scanning mode (100 nm/min scanning speed and 2-s response time). Molar ellipticity was calculated for each peptidestock concentration of 2 mg/ml in 1.3 mM phosphate-buffered saline (PBS) with 6.7 mM potassium fluoride (KF) and 30% (v/v) 2,2,2-trifluoroethanol medium. Spectra Manager software (version 2.0) was used for analysing the data and the Contin algorithm and SP37A database were used for deconvolution (19).

The Universidad de Antioquia’s Immunovirology group produced SARS-CoV-2 (wild-type B.1D/614G, α, β, γ, and δ) variant-infected VERO cell cultures. Once fixed on eight-well slides, they were preserved at 4°C and left to dry at room temperature (RT) to evaluate native state viral protein recognition. They were blocked with 30 µl per well 10% bovine serum albumin for 15 min, washed five times for 5 min with PBS and left to dry at RT. Immunised Aotus sera (diluted 1:40 to 1:640 in PBS) was placed on slides at 15 µl per well. They were incubated in a humid chamber for 30 min, washed six times with PBS (5 min each) and left to dry. Fluorescein-5-isothiocyanate (FITC)–conjugated goat immunoglobulin G (IgG) (goat IgG-FITC) against purified Aotus IgG (produced in our Institute) was then added at 1:100 dilution in PBS, 1:100 Evans blue; plates were incubated in a humid and dark chamber for 30 min and washed six times with PBS and left to dry. A drop of 50% glycerol in saline solution was added between the slides and microarray coverslips before observing them by fluorescence microscope (100× oil immersion objective lens).

Enzyme-linked immunosorbent assay (ELISA) involved 96-well ELISA plates (F16 MAXISORP plates) was covered with 100 µl per well test of native unmodified parental peptide at 10 µg/ml in PBS and incubated for 1 h at 37°C and overnight at 4°C. The sensitised plates were incubated at 37°C for 1 h, washed five times with PBS, 0.5% Tween-20, followed by a final wash with distilled water. They were blocked with 200 µl per well blocking solution (PBS, 0.5% Tween-20, 5% non-fat milk) and incubated at 37°C for 1 h. Then, 100 µl per well immunised Aotus sera were placed in wells at 1:100 dilution, incubated at 37°C for 1 h, and washed again; 100 µl per well of goat anti-Aotus IgG peroxidase conjugate was then added at 1:1,000 dilution in blocking solution. Plates were incubated at 37°C for 1 h and washed five times, and 100 µl per well developing solution was added (TMB substrate peroxidase and hydrogen peroxide: 1:1 ratio). Microplates were read (LabSystems Multiskan MS) 30 min later at 620 optical density (OD).

The Universidad de Antioquia’s Immunology Group’s BSL3 laboratory used Vero E6 cells for a 50% ± 3% plaque reduction neutralisation test (PRNT50), using monkey sera having the B.1 D614G variant. Briefly, Vero E6 cells (1.1 × 10⁵ cells per well) were seeded onto 24-well tissue culture plates the day before infection. The next day, 80 plaque forming units (PFU) of SARS-CoV-2 were incubated with or without serially diluted Aotus heat-inactivated sera (56°C, 30 min) at 200-μl volume in microcentrifuge tubes for 60 min at 37°C in 5% CO₂. The mixtures were then added to Vero E6 monolayers and incubated at 37°C for 60 min; the inoculum was removed, 1 ml of semisolid medium added (1.5% carboxymethyl cellulose, 2% FBS, 1% streptomycin, and DMEM).

Mixtures were cultured at 37°C for 72 h; semisolid medium was removed and monolayers were washed twice with PBS, fixed and stained with 4% formaldehyde/1% crystal violet for 30 min, and washed twice with PBS. A 50% ± 3% reduction in plaque count (PRNT50) was used as neutralising end-point. Viral control (VC) in the absence of serum and a serum control without SARS-CoV-2 were included in the tests (20, 21).

Venous blood was collected in ACD vacutainers (Becton Dickinson, United Kingdom) before vaccination and 6 weeks after the second immunisation dose. Ficoll-Paque PLUS sterile density media (GE Healthcare, United States) was used for isolating peripheral blood mononuclear cells (PBMCs) from whole blood.

Aotus B-21 and C-21 PBMC were seeded at 10⁵ cells per well density. Group I was stimulated with a 50-peptide mixture (10 μg/mL) of S, M, and E proteins and Group II with modified peptides derived from previously reported T-cell activating sequences: CD4⁺ T cells with native parental peptides 42992 (152–171), 43016 (369–388), 42964 (402–421), 42956 (446–464), 42962 (460–479), 42948 (485–500), and 43036 (1,110–1,129). CD8⁺ T cells were used with 43002 (386–405), 42970 (977–996), 42954 (1,027–1,046), 43040 (1,027–1,046), and 42960 (1,176–1,194) (10 μg/ml). The medium was used as negative control and phorbol myristate acetate (100 ng/ml) and ionomycin (400 ng/ml) as positive control for both groups (20).

All stimuli were run in triplicate and all treatments were analysed after 5-day coculture. A FACS Canto II cytometer (BD Sciences) was used for reading positive CD4+/CD45RO+ and CD8+/CD45RO+ cell percentages; FlowJo V10 (Ashland, Oregon, USA) was used for data analysis. A BD cytometric bead array (CBA) NHP Th1/Th2 cytokine kit (BD Biosciences) was used for cytokine analysis. Mann–Whitney and Student’s tests were used for assessing statistically significant differences between groups.

Aotus MHC DRβ class II exon 2 was amplified by PCR; primers were designed from previously reported A. nancymaae and A. vociferans allele lineages using intron 1 and intron 3 sequences (23).

A 317–base pair (bp) MHCII DRB gene exon 2 fragment was amplified. The Barcosel web tool was used for selecting an optimal barcode set for high-throughput sequencing for assigning each primer a 6-bp-long molecular identifier (24, 25).

Phusion Hot Start II High-Fidelity PCR Master Mix was used for amplification, using the touchdown PCR technique. A Wizard SV Gel and PCR Clean-Up System Kit (Promega) was used for purifying PCR products, following the manufacturer’s recommendations. A NanoDrop 2000 full-spectrum UV-Vis spectrophotometer was used for quantifying products that were sent for Illumina NovaSeq 6000 sequencing.

Sequence data were processed using Amplicon Sequencing Analysis Tools (AmpliSAT) suite; AmpliCLEAN was used for removing low quality (<30 Phred) reads and non-amplicon sequences. AmpliCHECK was used for analysing MHC DRB exon II using a 1% base substitution error rate, a 0.001% indel error rate (base insertion or elimination) and 3% minimum per-amplicon frequency (PAF) (26). The maximum number of alleles per amplicon was set and reading depth was limited to ~80,000 reads for all amplicons analysed by AmpliSAS. Minimum read depth per amplicon was set at 100 (all amplicons having less than 100 reads were excluded); the software identified and removed chimeras (variants arising from parenteral sequences in the same amplicon). The ACACIA pipeline was used with default parameters and 0.03 proportion threshold (low-poor) for allele calling (code available under https://gitlab.com/psc_santos/ACACIA) (27). Sequences were deposited in GenBank ( Supplementary Table S1 gives accession numbers).

The Allele Frequency Net Database (AFND) was used for data mining and choosing HLA-DRβ1* alleles (24) occurring with greater than 1% frequency in the database-determined ethnic groups (69 alleles). Previous reports were used for Aotus MHC- DRβ and current typing for establishing allele frequency (28, 29).

NetMHCIIpan-4.1 was used for evaluating and selecting proposed peptides’ ability to bind to selected HLA-DRβ1* alleles, i.e., having strong BA and/or presentation ability (≤2.5 percentile range), being non-binders and/or lacking presentation ability (≥10). The server was also used for calculating Aotus MHC-DR BA and sequence typing. A set of 200,000 non-redundant UnitProtKB/Swiss-Prot UniProt-derived peptides (13- to 21-aa-long) (30) was used for calculating the predicted peptides’ percentile rank distribution. Seq2Logo (31) was used for constructing peptide binding motifs and scoring matrices for each allele based on stringent binding cores (<2.5 percentile rank). The Kullback–Leiber approach from the previous step was adopted for calculating an SI index for comparing each allele’s pair, using MS eluted ligand (EL) scoring matrices, defined as Pearson distance [centre: 1 − corr(x,y)] with min-max normalisation.

The set of epitopes’ potential coverage resulted from the sum of the alleles’ frequencies in the target human population for which each designed peptide had significant affinity. The frequency of the 69 previously selected HLA-DRβ1* alleles calculated from the AFND (24) was used for estimating world population coverage.

The SI, based on MHC-peptide affinity predictions, determined strong similarity between selected HLA-DR and genotyped Aotus MHC-DR (average 83.7 ± 8.5).

ELISA, IFA, and NAb tests were performed several times with at least two independent preparations of each sample.

Our previous studies on malaria peptide structure has provided the basis for suggesting that chemically synthesised peptide structure (according to native protein aa sequence) means that they maintain their secondary structure, whether this be α-helix, β-turn or random coil. This was assumed for SARS-CoV-2 synthetic peptide production.

Peptides selected as vaccine candidates had to comply with two, well-established, fundamental requirements:

First, they had to have been 20-aa-long and preferably had random coil or β-sheet secondary structure in the native protein. UCSF Chimera (32) was used for verifying such structures, i.e., the phi and psi angles in the 3D structure of already reported S, E, and M proteins [protein data bank (PDB): S protein: 6X6P, 6VXX, and 6VYB; E protein: 5X29 (33–36)]. Whereas the algorithms included in Swiss Institute of Bioinformatics (SIB) Resource Portal’s Expasy Tools (37) were used for predicting secondary structure for M and ORF9c proteins. Regions containing random coils or beta sheets were selected.

Second, it was preferred that most fragments selected by the secondary structure criteria fulfilled relevant biological functions in the target proteins. Selected S protein peptides were involved in angiotensin-converting enzyme 2 (ACE2) binding in the receptor binding domain (RBD) (38) and protein interaction with linoleic acid (LA) (39), cleavage sites S1 and S2’, membrane fusion sequence (40), linkers, heptad regions 1 and 2 (HR1 and 2) (41), and the polybasic binding site (PBBS). Selected E- and M-derived peptides were also involved in biological functions, e.g., ion and nutrient transport in the E protein and the PDZ binding motif (PBM) (42).

However, some helical fragments were chosen, because they fulfilled relevant functions in the protein, for example, protein fusion to target cell.

The selected peptides were called native or parental peptides; these were modified in some aa, which were critical regarding peptide binding to the MHCII-PBR [thereby obtaining modified peptides mHAPs], according to previously reported rules and principles, i.e., changing aa polarity whilst maintaining volume and mass) for obtaining immunogenic peptides, which were capable of inducing protection (11, 13, 43).

It should be mentioned that the S protein has an antibody-dependent-enhancement region (ADE) located between aa 597 and 630; it is known that Ab-based therapies and vaccines involve a risk of increasing COVID-19 severity through this Ab-dependent enhancement (ADE) mechanism (44), peptides in this aa region were thus not selected.

Parental/native aa sequences were selected for modifying and producing SM-COLSARSPROT. They had to have random coil or β-sheet structures, perform relevant functions or mediate fundamental inter- or intra-molecular interactions, and have little genetic variability (some mutated peptides were exempt from the last two requirements due to their relevant invasion-related role).

S (PDB code: 6X6P, 6VXX, and 6VYB) (33, 34), E (PDB code: 5X29) (35), and M protein (36) predicted structures were used for structural selection ( Figure 2 ). Trial B-21 involved 44 S protein peptides, 3 M protein-derived ones, 2 E protein peptides, and 1 ORF9c protein-derived peptide for determining the 50 modified peptides’ immunogenicity and immunity duration (180 days). Trial C-21 involved 53 molecules: 47 S protein, 38 β-sheet, and random coil or β-turn structures; 27 had relevant functions, plus another 20 for which no functional activity has yet been demonstrated. Nine of the 47 S-derived peptides had α-helix structures (7 being very relevant during invasion). Both E protein peptides had α-helixes ( Figure 3 ). Eleven functionally relevant peptides identified in S protein’s S1 region and cleavage sites S1/S2 were directly involved in intermolecular interactions with angiotensin converting enzyme 2 (ACE2) (38) and linoleic acid (LA) (39) host cell attachment and invasion-mediating receptors ( Figure 3 ).

Hinge and trimer stabilisation regions (45) previously recognised by other groups (46) were identified and modified in the ten peptides involved in fundamental S protein peptide NTD1-CTD1 and CTD1-CTD2 intra-molecular interactions. Two were membrane fusion peptide (FP) components, one was involved in fusion peptide proximal region (FPPR) formation (pH-dependent structure and function) (47), two in HR1-HR2 intersection, involved in six-helix bundle formation (40) and one in trimer stabilisation ( Figure 3 ). E protein (viroporin) peptides were directly involved in ion and nutrient transfer-related channel formation ( Figures 2F , 3 ); two of three M protein peptides (lacking 3D structure determination) were predicted to have random coil structures theoretically interacting with other SARS-CoV-2 proteins: one α-helix ( Figure 2E ) (36) and one ORF9c protein-derived peptide.

NetMHCpanII 3.2/4.1 (48, 49) strikingly predicted that 44/53 (83%) SM-COLSARSPROT native parental peptides in C-21 lacked HLA-DRβ1* and/or Aona-DRB allele-binding capability [elution (EL) or binding activity (BA)] ( Figure 4 ). This partially explained their poor/weak immunogenicity and/or antigenicity (50); the remaining nine peptides (17%) bound to a limited amount of alleles due to inappropriate TCR-contacting residues, suggesting the need for further structural modification to make them highly immunogenic and NAb-inducers.

Seven prior trials (A-20 to F-20 and A-21) were designed to determine the modifications needed to make them highly immunogenic, proline’s role in their conformation, selecting the most immunogenic ones, the amount of peptides to cover >80% of the world’s population and the use of Al(OH)₃ as adjuvant. This involved 30–50 modified peptides, each used in ~30 Aotus monkeys per trial. B-21 involved 50 modified peptides in 32 Aotus monkeys ( Figure 5 ), C-21 involved 53 peptides in 30 monkeys ( Figure 6 ). Figure 7 shows these peptides’ supramutational modifications to make them highly immunogenic. Figures 8D–G list peptides common to B-21 and C-21 (green), 1–3 aa differences between them (yellow), some having newer modifications (colourless) due to residues having VOC and VOI mutations (9, 11–13, 43, 51). Small coloured squares in Figures 5 – 8 represent peptides in which mutations were replaced, according to colour code.

SM-COLSARSPROT peptide design involved determining PPII_L conformation propensity by CD spectra ( Figure 9 ); selected peptides were modified to ensure PPII_L structural propensity (9, 51, 52). Most C-21 (40/53) native peptides had residues disrupting PPII_L structure, i.e., I, V, and T (β-branched aa) and aromatic aa, such as Y, F, and W ( Figure 3 , grey). These native parental peptides also contained polar aa having a short side chain promoting H-bond formation between side chain–backbone (SC-BB) ( Figure 3 , highlighted in green), as well as promoting n→ π*-type interactions (53).

All PPII_L-disrupting residues had to be replaced by others having similar physicochemical characteristics, such as mass, surface, and volume but opposite polarity (9, 11–13, 43, 51, 52), as well as aa having a strong propensity to form PPII_L-type (P>>>L>R>A/K>M/D>Q/E>H>>G/N/S/C/T/V/I/F/Y/W) (54). Ile was categorically avoided and replaced by Leu having similar mass and volume, but different conformation and a slight difference regarding polarity (54). Figures 3 , 7 show native peptide sequences, having the most relevant mutations (highlighted in red) and their locations (modified peptide sequences shown below in bold).

As examples of the role of these PPII_L conformation disrupting capacities, non-binding HLA-DRβ1* peptide 42954-derived (1,028–1,046) modified peptides 43240 and 43452 comprising disruptor aa V1031 and V1033 ( Figures 5 , 6 ) having strong HLA-DRβ1*1301 and Aona-DRB1*03N3 BA did not induce Ab due to β-branched aa’s negative role ( Figures 8D–G ). 43436 (NA:460-477) having F465 bound strongly to HLA-DRβ1* 0701 whilst Aona-DRB W300102 did not induce Ab production. 43186 (42992: 152–171) and 43420 (42992: 153–171) having F157 and Y160 and 43270 (43024: 502–518) having F505 and Y508 designed for strongly binding HLA-DRβ1*1502, Aona-DRB W10901 and W2908 did not induce any Abs. 43424 (43284: 338–353) having strong HLA-DRβ1*1303 and Aona-DRB1*031701 BA was also incapable of inducing Ab production ( Figures 8D–G ). These peptides provided clear examples of β-branched and aromatic (Aro) residues’ negative role when inside the PBR, despite many monkeys having such Aona alleles in the trials.

Further physicochemical intra-peptide electron interaction-related principles were identified during this study, as most theoretical SC-BB H-bonds were disrupted (whenever possible due to HLA-DRβ1* or Aona-DRB allele restrictions) for a better fit in the MHCII-PBR.

The main SC-BB motifs to be disrupted were NP, PD, PN, DP, TP, and SP, occurring with 14% to 32% frequency in proteins (55, 56). As they are found in native peptides, this makes them candidates for replacement to avoid such H-bond formation so that the structure of the nine aa fitting into the PBR-MHCII is as similar as possible to PPII_L. This gives a sufficiently stable MHCII-peptide bimolecular complex to ensure TCR recognition, establishing the MHCII–peptide–TCR trimolecular complex, thus triggering an immune response.

SC-BB interactions’ negative role was better observed when peptides were modified to ensure HLA-DRβ1* BA (DRβ1*0701, 1201, 1401). Such compulsory requirement insisted on small polar residues (N, T, D, and S) in Pockets 4 or 6 (P4 and P6), due to their small volume capacity and polarity, followed by P, forming the aforementioned motifs blocking P’s critical role in position 5 or 7 (p5 and p7), making them unequivocally negative for inducing Ab production. This mainly occurred when NP and SP motifs were preceded by positively charged, long side chain (K or R) or aliphatic residues (L), as in 43436 (0701), 43270 (1502), 43218 (0405), and 43414 (1401) (HLA-DRβ1* in parenthesis); this suggested SC-BB interaction’s negative impact on PPII_L propensity, having negative consequences regarding the immune response. Alternately, some of these Aotus alleles may not have occurred in the aforementioned trials ( Figures 5 , 6 ), i.e., 43436 was discarded.

Modifying physicochemical conditions led to dramatic improvement regarding elution (EL), BA and HLA-DRβ1* and Aona-DRB allele-binding ability ( Figures 5 , 6 ); most modified peptides thus became strong-binders, leading to significant immunogenicity (high Ab production rate) in Aotus monkeys, thereby strongly contrasting with native parental peptides having poor or no BA (compare Figure 4 ) and poor or no immunological activity.

The n→ π* delocalisation effect (stabilising α_L and β-turns in protein structures) was also disrupted, mainly when associated with neutral polar aa N and Q (57, 58) (both having different propensities to form SC-BB H-bonds in electron interactions); it has been shown that they contribute to H-bond electron density delocalisation (59). The carbonyl oxygen donor (C=O) n orbital became delocalised in carbonyl BB antibonding π* orbitals (H-N), bringing both carbonyl groups close to form a single C=O···HN bond in Asn, when i→ i-3 and i→ i+3 interactions were formed (60, 61); i→ i+2 is preponderant in β-turns and i→ i+3 in α left-handed (α-_L) helical regions.

Such interaction was analysed regarding peptides involved in B-21 and C-21 trials, mainly N and Q side chains interacting with their neighbours’ backbones in positions i−3 to i+3 (but located within the MHCII-PBR). We related this to an immune response (for the first time) and found that peptides involved in such interactions induced very low Ab titres compared to those not involved in such interactions. This suggested that a peptide having a structure with these characteristics does not adopt a PPII_L-type extended structure in the MHCII PBR, i.e., a requirement for obtaining stable MHCII-peptide complex formation. Such peptides therefore had to be modified.

The UCSF Chimera package (32) was used for analysing this effect regarding SM-COLSARSPROT peptides for Gln n→ π* interactions represented by H-bond formation between C=O···HN in i → i-4 or i → i-3, modified peptide 43238 (894–913) where Q904 had n→ π * interaction with L900 (62). β-turn type IV₃–derived 43424 (338–353) involved N343 having n→ π * interaction with G339 and peptide 43182 (105–124) where Q114 interacted with D111 ( Figure 10A ). All induced very low Ab titres in B-21 and C-21, highlighting n→ π* delocalisation’s negative effect on the immune response.

Most of the 50 modified peptides had strong class II molecule BA (percentile rank 0.0 to 2.5) in B-21 ( Figure 5 ); contrasting with their native parental peptides ( Figure 4 ), only one of 50 (43242, HLA-DRβ1*1401) modified peptides had ≥2.5 percentile purified HLA-DRβ1* EL capability (threshold level) and 11 of 50 had ≥2.5 percentile HLA-DRβ1* live homozygous cell BA (significance level). The core probability for these modified peptides to bind class II molecules (close to 1.0: data not shown) suggested a significant probability of binding predicted HLA-DRβ* or Aona-DRB alleles.

NetMHCIIpan4.1’s more specific and accurately predicted EL ranking for Aotus allele frequency was preferred due to few Aona-DRB genotyped monkeys (~900) compared to HLA-DRβ1* (>4,000,000 people) ( Figures 5 , 6 ). The server classified human HLA-DRβ1*1401, 1201, 1202, and 0802 equivalent alleles as not found (NF) in Aotus B-21 and C-21 because their Aona-DRB alleles did not occur in such relatively small group of monkeys or had very low frequency (some <1%). The difference between trials stressed the importance of carrying out several trials (as reported here) to identify relevant alleles having strong aa sequence SI and ID. DNA genotyping enabled accurate identification of genetic traits because up to six Aona-DRB alleles can be identified in Aotus monkeys ( Figure 11 ), unlike two HLA-DRβ1* in humans; such difference enabled more potentially relevant SM-COLSARSPROT peptides to be recognised.

Residues having ≥1,000 VOC and VOI variations were selected from Global Initiative on Sharing Avian Influenza Database (GISAID) ‘s EpiCoV database (63); 235,111 sequences were shown to have residues having mutations (March 10^th 2021, when this work began) ( Figure 7 ). GISAID was used for identifying aa not used by mutants in all SARS-CoV-2 variants and for replacing mutant residues by aa having very close or similar mass, volume, and surface but opposite polarity ( Figures 5 – 7 ). This is the essence of the SUPRAMUTATIONAL concept; the virus cannot use these aa during thousands of mutations, suggesting the genetic impossibility of using them.

Figures 3 , 5 – 7 show mutant residue replacements, serial numbers, location in a protein (in parenthesis), and the HLA-DRβ1* allele they bound to after modification. Figure 10B shows X-ray crystallography-determined S protein location in native peptides and residues (grey: backbone and residues) interaction with ACE2 receptor (pale blue, fuchsia, and red: backbone and residues) during invasion and the shift in orientation adopted by modified peptides where replaced residues (colour) disrupted original H-bonds (black dots) in intermolecular interaction with target receptors (UCSF Chimera package) (32).

Figure 10B also shows X-ray crystallography-determined native parental peptide location and that of modified peptides modelled from the structure adopted by the matching protein 3D structure fragment using UCSF Chimera (32). Immunogenic modified (bold) peptides in the receptor-binding domain (RBD) are shown, along with their location in aa sequences (small superscript numbers) and modifications (in brackets). 43428 (402–421) [L417K], 43434 (446–465) [G446P, Y449F, and Y453K], 43438 (485–500) [N487P and Q493E] and 43270 (502–518) [F515R], and 43200 (386–405) [F505F] (64). Figures 3 and 10F show 43424 (338–353) [V341E], 43428 (402–421) [R408H and Q409L], and 43430 (418–437) [C432T] hydrophobic interactions with linoleic acid and residues having intra-molecular interactions in the hinge region (46) or trimer stabilisation (45) during host cell membrane invasion. Helical peptides 43448 (FRPP), 43276 (trimer stabilisation), and 43462 (HR1-HR2) had intra-molecular stabilisation interactions ( Figures 10C – E ).

All peptide replacements compensated for the lack of antigenicity and/or poor immunogenicity due to poor HLA-DRβ1* binding or lack of it (compare Figures 5 – 7 to Figure 4 ) and VOC and VOI region mutability, making SM-COLSARSPROT the first SUPRAMUTATIONAL, highly immunogenic peptide mixture targeting SARS-CoV-2.

Sixty-one monkeys were bled for DNA genotyping; however, only 55 proved to be appropriate for DNA sequencing (29 in B-21 and 26 in C-21), 26 A. nancymaae and 3 A. vociferans in B-21.

Figures 5 , 6 show HLA-DRβ1* allele frequency, organised by frequency in the human population. Matching Aotus allele frequency (parenthesis) in these trials is shown ( Figures 5 , 6 , 8 ), suggesting the best-modified peptides for inclusion. This gave 43266 (W300102) and 43418 (W3802) in HLA-DRβ1*0701; 43200, 43272, and 43244 in (1*031701) in HLA-DRβ1*0301’ 43180 (W2908) and 43250 (W1806) in HLA-DRβ1*1501; 43230 and 43434 (1*0328) in HLA-DRβ1*1101; 43438 (W300102) in HLA-DRβ1*0101; 43456 (W2908) in HLA-DRβ1*1302; 43174 and 43274 (W470401) in HLA-DRβ1*0401; 43232 (W2910) in HLA-DRβ1*1104; 43190 (3*0627) in HLA-DRβ1*0404; and 43224 (W2908) in HLA-DRβ1*0901. Numbers boxed in green in Figure 8 represent peptides common to B-21 and C-21, stressing these results’ reproducibility and the long-lasting memory induced by them in B-21. Strikingly, human alleles were found where Aotus allele frequency rate was high, clearly suggesting that this methodology can be used for supramutational peptide identification regarding SARS-CoV-2 and other diseases, i.e., a universal methodology.

Practically, all peptides having properly modified supramutational modifications were capable of inducing Ab production able to recognise native sequences in native peptides (determined by ELISA reactivity and IFA recognition in infected cells). This clearly suggested that the methodology described here promoted appropriate Ab-inducing responses, supported by the immunological results.

CD studies for ascertaining PPII_L propensity involved 51/53 C-21 peptides due to the onerous task of ¹H-NMR 3D structure determination (3) for recognising modified SM-COLSARSPROT peptides’ secondary structure elements. CD data collected at 27°C and 37°C provided striking results. Figure 9 indicates random coil peptides, similar to ¹H-NMR determination of highly protective, long-lived COLSARSPROT peptides (3). CD revealed that most (~85%) SM-COLSARSPROT peptides inducing high Ab titres (>0.3, 450-nm OD; Figures 8F, G ) targeting C-21 mixture native parental peptides had strong negative ellipticity at 190 nm (starting below 0 at 27°C and 37°C). Another six peptides inducing very low Ab titres had strong negative ellipticity at 190 nm (same range) (<0.2, 450-nm OD, 43420, 43270, and 43188 permanently and 43274, 43430, and 43426 before II₂₀), suggesting a temperature-dependent physicochemical and immunological pattern ( Figures 8 , 9 ). Four peptides had weird structures not associated with specific biological or immunological activity; eight more were insoluble.

The small local maximum around 218–220 nm (characteristic of PPII_L propensity) was not observed for modified peptides. Strikingly, no SM-COLSARSPROT had α-helix tendency and only three of 50 had β-turn type tendency despite native parental structure, clearly suggesting that appropriate modifications had made them firm MHCII binders and strong immunogenic anti-SARS CoV-2 variant peptides.

The Universidad de Antioquia’s Immunovirology Group isolated SARS-CoV-2 variants in a biosafety level 3 laboratory (BSL-3). Briefly, Corpogen Laboratory (Bogotá, Colombia) used nanopore technology-based, next-generation sequencing (NGS) for identifying samples carrying SARS-CoV-2 variants following ARTIC SARS-CoV-2 sequencing protocol (65).

Variants were deposited in the GISAID database: wild-type (B.1/D614G) (EPI_ISL_536399), Alpha (EPI_ISL_4549188), Gamma (EPI_ISL_4926393), Delta (EPI_ISL_5103929), and Mu (EPI_ISL_4005445).

Figure 8D gives ELISA data showing that all 32 immunised monkeys developed Ab reacting with the 50 native parental peptides in B-21, i.e., wild-type (B.1/D614G) or VOC- and/or VOI-derived (small squares above peptide numbers, following established colour code, Figures 8D, E ) 20 days after the second dose (II₂₀). HLA-DRβ1*0301 (DP, KPE, PD SC-BB motifs) 0405 (KNP) had low reactivity (<0.2 absorbance at 450 nm at 1:100 sera dilution) after II₂₀, which increased in II₁₆₀. Peptides 1502 (RNP), 1401 (SPR), and 1201(APG) consistently had low Ab titres (9/50), some having SC-BB H-bond formation motifs (55, 56). Most peptides (42 of 50, 84.0%) in B-21 induced very high Ab titres (≥0.3 absorbance at 450 nm) in monkeys having the other 26 HLA-DRβ1* alleles, suggesting this peptide mixture’s widespread immunogenicity and coverage.

Strikingly, similar to IFA titres, Ab levels remained very high in B-21 180 days after the first or 160 after the second (II₁₆₀) dose; this increased in most peptides compared to II₂₀ ( Figures 8D, E ). These results clearly showed that SM-COLSARSPROT induced strong, long-lasting high Ab levels against specifically modified, native, mutated aa sequences, recognising/identifying VOC and VOI aa sequences, strongly supporting SUPRAMUTATIONAL methodology.

C-21 ELISA concerning peptides common to B-21 ( Figures 8D–G , green peptide numbers) confirmed this methodology’s reproducibility in the Aotus trials and highlighted strong reactivity specificity regarding newly VOC- and VOI-modified peptides ( Figures 8D–G , yellow and uncoloured boxed numbers) and that such reactivity remained high 75 days after the second dose or 105 days after the first (I₁₀₅).

It is worth stressing that SM-COLSARSPROT-induced reactivity was strong regarding specific Alpha (B.1.1.7, UK) 43216, 43276, 43246 43272, 43274, 43438, and 43456; Beta (B.1.351) 42846, 43190, 43224, 43258, 43428, 43438, and 43446; Gamma (B.1.1.28.1 or P.1) 43172, 43188, 43254, 43258, 43418, and 43428; Delta (B.1.617.2 and B.1617.1) 43264, 43266, 43272, 43274, 43418, 43434, and 43438; Epsilon (B.1.427/429) 43264, 43266, 43272, 43274, 43434, 43438, and 43446; and Mu (B.1.621) 43272 and 43274 variant mutations. This clearly suggested that a supramutational approach can deal with many genetic variations and their combinations (70).

Replacing mutant residues and inducing high Ab titres against their native, functionally relevant aa sequences (ACE2 and LA-bdg, cleavage sites S1/S2, S2’ trimer stabilisation, 6-helix bundle formation: Figure 3 ) strongly suggested that these Ab could directly block or impede (by steric hindrance) such functions, partially explaining their functional immunological activity.

D-21 using a peptide mixture, the QS-21 immunopotentiator and most of C-21, involved 42 Aotus and showed that SM-COLSARSPROT peptides could induce NAb activity against VOC and VOI mutants targeting B.1/D614G variant ( Figure 8 ), but that such activity was short-lived (<30 days).

Such highly specific Ab reactivity with the panel of modified peptide-induced native VOC and VOI variant specific peptides strongly suggested ELISA as a pertinent methodology for determining supramutational activity, better than or complementing other immunological methods like NAbs.

Each peptide’s potential coverage was determined in the main ethnic groups as the SM-COLSARSPROT synthetic peptide mixture was derived from genetic VOC and VOI originally identified in China, UK, South Africa Republic (SAR), Brazil, India, USA, and Colombia, but which nowadays are extensively spread around the world. Caucasians, Hispanics, Asians, Orientals, Arabs, Blacks, and Mestizos (Colombians) have different HLA-DRβ1* allele frequencies (fuchsia in Figures 5 , 6 , when ≥3× the world’s population), grouping 78.5% of the world’s population. SM-COLSARSPROT’s predicted minimum potential coverage is 77.6% for Asians and a maximum 83.1% for Orientals and Arabs (most being ~80%). Such coverage is quite similar to that recognised by IFA (80% to 90%) regarding wild-type (B.1/D614G) genetic variants (α, γ, δ, and μ) and ELISA titre Aotus sera (83.2%) reacting with peptides developed for specific HLA-DRβ1* alleles for VOC and VOI mutants and NAb targeting the Wuhan wild-type (B.1/D614G) variant (83%).

Such correlation emphasises SM-COLSARSPROT’s potential as a highly immunogenic (probably protectogenic) peptide mixture for providing desperately needed/sought-after “collective or herd immunity” against most VOC and VOI mutations with a single immunising agent: SM-COLSARSPROT synthetic peptide mixture.

Antigenicity- (6, 7) and immunogenicity-related structural analysis (9, 11–13, 43, 51, 52) has meticulously demonstrated that peptides must have polyproline II-like left-handed (PPII_L) structural propensity during the first critical step: antigen presentation. In-depth physicochemical analysis of 3D structure is therefore needed to ensure that non-immunogenic peptides, non- or low-MHCII binders become immunogenic, i.e., hence the extraordinary relevance of PPII_L formation propensity for highly immunogenic peptide development.

PPII_L propensity enables peptides to establish H-bonds between specifically located peptide BB atoms and HLA-DRβ1* or Aona-DRB SC atoms to stabilise such binding (6, 7, 9, 43, 51) ( Figure 1D ), unlike secondary structures like PPI, inverse β turns, left and right α-helixes and β-turns.

Immunogenicity is thus induced via PPII_L formation being substantially restricted by local interactions, particularity side chains having adjacent carbonyls from backbone (SC-BB) n→π* interactions, represented by H-bond formation (71, 72).

Peptides’ highly immunogenic PPII_L formation propensity must follow clearly defined rules and constraints (6, 7, 9, 11–13, 51, 52); residues outside the PBR not involved in such interactions were only modified in −p2, −p1, p10, and p11 based on specific physicochemical rules for stabilising PPII_L propensity.

Often-ignored, predicted intra-peptide electron interactions also had to be modified/replaced by others involving similar mass and volume but different SC charge or polarity ( Figure 4 ) to induce immunogenicity, because short polar SC residues (T>S>D>N) can induce inter- and/or intra-residue H-bonds with BB atoms (32%, 29%, 26%, and 19% frequency in proteins). They form special motifs with neighbouring P (i.e., TP, TG, SP, DG, DP, NP, PD, and DE), being α-helix and β-turn components, disturbing, and reducing PPII_L propensity (55, 56).

Aro residues were only allowed in specific positions to avoid Aro-Pro and Pro-Aro sequences as P was the main residue in PPII_L formation. Its propensity is cis-amide bond formation with neighbouring Aro (W>Y>F) residues, the last on the propensity scale (54, 73); however, Aro (F, Y, and W, in that order) were allowed in specific places because they were the most frequent (60%–75%) residues fitting into MHCII-PBR HLA-DRβ1* and Aona-DRB P1 (6, 7, 48, 49) ( Figures 1E, F ). They were modified inside the PBR due to their very strong (π-CH) interaction (53), being replaced by non-aromatic residues having the same surface and volume but opposite polarity.

Aro residue proximity of 1 to 3 residues apart was also avoided (whenever possible) because they can establish π-π face-to-face, T-shaped, parallel, or antiparallel structures (74, 75); DRβ1*15 and 16 peptides were the only exceptions, where Aro residues in P1 were also mandatory in P4 (74, 75), having dramatically negative consequences for Ab production in 43186, 43270 (HLA-DRβ1*1502).

Several modified peptides from the same alleles were synthesised and included in SM-COLSARSPROT to maximise coverage of small HLA-DRβ1*allele differences between different ethnic groups.

The previously mentioned, upwardly orientated soluble side-chain rotamer orientation, volume, and charge (76) are also critical features for MHC–peptide–TCR complex formation; residues in p3 and p7 were replaced by aa having similar mass, surface, and volume by aliphatic or P aa (the latter preferred because P in these positions induced or increased PPII_L propensity and T-cell activation) ( Figures 1B, C ) (3, 77, 78).

It has also been shown the PPII_L structures have been more stable when preceded by negatively charged or OH-displaying residues (D, E, T, and S) in their N-terminal (in −p1 or −p2) and C-terminal regions (p10 or p11, residues), followed by positively charged residues like preferred R or K (79). Therefore, besides the nine properly orientated residues fitting into the MHCII PBR–peptide–TCR complex, the four to five extra residues in N- and C-terminal regions provided the complex with more stability, making them highly immunogenic (80, 81) and protection- or NAb-inducing. Our peptides were thus 16- to 20-mer-long rather than short 9- to 10-mer-long predicted to fit into the PBR (having very little immunogenicity).

A major problem regarding COVID-19 vaccine development concerns high SARS-CoV-2 genetic variability (reported from September 2020 onwards). This mainly occurs concerning the trimer S protein mediating viral attachment to host cells and membrane fusion (82), pentamer E viroporin-mediated ion channel formation, assembly and virus budding, and the less studied M protein interacting with S, E, and N proteins (36) (little genetic variability reported to date).

Interestingly, many mutations were located in the S protein’s functionally relevant regions (mainly in the RBD, but most in the RBM), especially those directly interacting with the ACE2 receptor, the linoleic acid binding region and the S1/S2 cleavage site; very few were located in other regions. Blocking such relevant viral functions makes them excellent targets for inducing functionally directed, immune protection-inducing efficacy.

It was quite striking that most S protein mutations (very few exceptions) occurred in regions having β-sheet or random coil secondary structures ( Figure 3 , red boxes). Our selection involved inter- and intra-molecular interactions (e.g., receptor-ligand cleavage S1/S2 sites) being mediated by β-sheet or random coil structures, whereas S and E protein surface-exposed regions having α-helix structure had different functions, i.e., membrane fusion (FP, FRPP, and HR1-HR2) and ion channel formation in E. This confirmed a clear, structural-functional dichotomy, partially explaining the tremendous immunological pressure against receptor-binding and protein-processing functions and the very efficient viral escape mechanism, as a single aa mutation knocked out a very effective immune response. The impossibility or difficulty of using some aa for SARS-CoV-2 variants at these mutation sites was quite striking, a viral disadvantage taken advantage of by our institute for this SUPRAMUTATIONAL approach.

The SM-COLSARSPROT mixture (specifically C-21) theoretically covers ∼80% of the world’s population; it should be stressed that our strategy was to include peptides for the human population’s main HLA-DRβ1* alleles (≥1.0% global frequency), trying to cover ~80% of it and the main ethnic groups. The C-21 mixture can thus provide the so desperately needed “community or herd immunity”, whereas B-21 highlighted SM synthetic peptide-induced, long-lasting memory, many being present in C-21.

Amazonian Aotus monkeys having matching human alleles were used to test the aforementioned developments. Some allele equivalences occurring with very low frequency (<1.0%) were not found in them; they were thus classified as NF because our trials involved 30–50 monkeys per trial and the probability of finding these alleles was very low ( Figures 4 and 5 , Aona-DRB column).

All these principles or rules must be considered when developing highly immunogenic, protection- or NAb-inducing peptides as antigen presentation by the MHCII–peptide–TCR complex is the first and most critical step in immune activation. Figure 13 summarises the entire procedure used in this study, i.e., selecting the relevant fragments having genetic variations (first panel), production, characterisation, and prediction in HLA-DRB molecules (second panel) and immunological studies (third panel).

Chemically synthesised peptides are extremely pure, contaminant-free, according to the HPLC and MS, very stable at RT, do not require a cold chain for storage and transportation, and can be produced very quickly and very cheaply in not very sophisticated laboratories anywhere in the world, especially in the most needy regions. They have very high yield for immunising many people with pure, reproducible synthetic peptides (i.e., ~1,000,000 people could therefore be immunised with 10g [as produced for SPf-66 (83) if 10 micrograms of each peptide are administered per person (500 µg per dose)].

Chemists, decision-makers, ourselves, and others working in the field must assume that minimal subunit-based, multiepitope, multiprotein chemically synthesised, highly immunogenic SM-COLSARSPROT represents a tremendous advantage and solution for the currently raging threat to human health and welfare: the continuing COVID-19 pandemic. C-21 is thus an excellent example following the rules and principles expounded here, as it can be used against many other infectious diseases, including future ones, because it provides a logical, rational, and soundly established methodology for synthetic vaccine development.