Production of immunoglobulins

Although immunoglobulins are produced by B cells, Ig synthesis and release depends on the reciprocal interaction of many different cell types of the immune system. Induction of the innate response is essential, as well as the activation and proliferation of T cells and B cells.

However, to evaluate the Ig response, we focused on T cells and B cells, mainly because the innate cellular response counteracts pathogens at a predominantly local level, and leads to the eventual elimination of them in a time frame probably longer compared to the fast neutralization of the humoral response.

The software suggested by professor Viola[1] required us to get in touch with the owner company, explaining our needs and not being sure if they could be satisfied on time. For this reason, we opted for other tools more quick to use, but still as reliable as possible.

For this purpose, we used the Immune Epitope Database & Tools (IEDP)[2] and the epitope prediction algorithm associated.

Using the IEDB database

First of all, we researched if there were any known linear epitopes of any M13 deposited protein in the database. The IEDB database had only G3P and G8P proteins registered, i.e. we could not assess the immunogenicity of the other proteins. The specified parameters of our query are listed below.

• Epitope: any
• Assay: T cell, B cell, MHC ligand;
• Outcome: positive;
• Organism: Inovirus (M13 bacteriophage);
• Antigen: any material entity of the particle;
• MHC restriction: any;
• Host: Human;
• Disease: any.

According to the software, the only immunogenic linear epitope known of the virus is TFMYVFSTF[3], which is a fragment (407-415aa) of the G3P protein.

Using prediction softwares

After the research among known data, we proceeded with a new query using prediction softwares listed in the IEDB website to look for the most probable not-known epitopes of the viral proteins. Since there were many different software, we selected a few of them based on our purpose and interests.

T cell epitopes prediction

Firstly, we looked into how the phage proteins would be processed by the immune system.

MHC-I processing prediction

The MHC class I complex is present on the surface of almost any human cell, except for immune cells. It binds intracellular pathogen peptides and exposes them to CD8+ T cells. Although MHC-I analysis is not very relevant for our project since the phage does not infect any human cell (i.e. it is very unlikely that a phage peptide would be exposed on MHC-I), we decided to look for MHC-I epitopes anyway because of the much higher accuracy of those predictions, compared to the MHC-II epitopes predictions.

NetCTL analysis

NetCTL[4][5] is a predictor of T cell epitopes along a protein sequence. This software requires a protein sequence as input data, giving in output any possible linear epitope with relative MHC binding affinity, cleavage affinity, TAP transport score and an overall prediction score. Thus, we use the software to select the most probable epitopes exposed on MHC-I proteins on the cell surface. Since the algorithm required to select an HLA supertype family, here is a table that lists HLA supertype classification[6]. We used the default A1 supertype value.

Parameters used in the queries:

• Prediction method: NetCTL;
• Weight on C terminal cleavage: 0.15;
• Weight on TAP transport efficiency: 0.05;
• Select Supertype: A1 supertype;
• Threshold: 0.75.

Explaining output values:

• #: residue position;
• Peptide Sequence: amino acid residue;
• Predicted MHC Binding Affinity: the value is given as 1–log50K(aff), where log50K is the logarithm with base 50.000, and aff is the affinity in nM units;
• Rescale Binding Affinity: the predicted binding affinity is normalized by the 1st percentile score;
• C Terminal Cleavage Affinity: predicted proteasomal cleavage score;
• TAP Transport Efficiency: predicted TAP transport efficiency;
• Prediction Score: overall prediction score.

G3P protein

Prediction score vs aa position, the red line represents the threshold.

G6P protein

Prediction score vs aa position, the red line represents the threshold.

G7P protein

Prediction score vs aa position, the red line represents the threshold.

G8P protein

Prediction score vs aa position, the red line represents the threshold.

G9P protein

Prediction score vs aa position, the red line represents the threshold.

MHC-II binding prediction

The MHC class II complex is present on the surface of antigen presenting cells. It binds internalized pathogen peptides and exposes them to CD4+ T cells. Since the phage tropism spares with high confidence any human cells, MHC-II antigen processing and binding affinity is the focus of our interest.

We analyzed the MHC-II processing of the protein, using MHC-II binding prediction software[7-17].

Since the MHC-II binding epitopes prediction tool is a bit more complex than the previous one, we add some more explanation on how it works for a more comprehensive reading the tool:

• breaks the sequence into all possible peptides of chosen lengths;
• predicts the binding affinity for each peptide based on the method;
• compares the predicted affinity to that of a large set of randomly selected peptides;
• assigns a percentile rank depending on individual predicted affinity;
• through the consensus method, picks the median rank of the methods used - consensus percentile rank.

Parameters used in the queries:

• Prediction method: NetMHClIpan 4.1 EL (recommended epitope predictor-2023.09);
• Select species/locus: Human, HLA-DR;
• Select alfa & beta chains separately if applicable: no;
• Select full HLA reference set: no;
• Select 7-allele HLA reference set: yes;
• Select length(s): default (15).

Explaining output values

The predicted output is given in units of IC50 (nM) for the combinatorial library and SMM_align. Therefore a lower number indicates higher affinity. As a rough guideline, peptides with IC50 values <50 nM are considered high affinity, <500 nM intermediate affinity and <5000 nM low affinity. Most known epitopes have high or intermediate affinity. Some epitopes have low affinity, but no known T-cell epitope has an IC50 value greater than 5000.

For each peptide, a percentile rank for each of the three methods (combinatorial library, SMM_align and Sturniolo) is generated by comparing the peptide's score against the scores of five million random 15 mers selected from SWISSPROT database. And the adjusted percentile rank is the percentile rank adjusted based on the frequency of peptide lengths.

A small numbered percentile rank indicates high affinity. The median percentile rank of the three methods were then used to generate the rank for consensus method. The data for percentile rank was updated in the release 2.22 by recalculating the percentile data with a consistent sample peptides datasets for all the methods.

The cutoff of percentile rank was set on 10.0 (percentile rank ≤ 10.0), as the tool instructions suggested. Since there are several possible alleles for MHC-II complex, the number of results is very high: each predicted epitope can bound more than one couple of alleles, and vice versa.

Proteins analyzed, with no. of results in brackets:

• G3P protein (246)
• G6P protein (53)
• G7P protein (1)
• G8P protein (33)
• G9P protein (2)

Immunogenicity prediction

CD4 T cell immunogenicity prediction software analysis

We then used another software to assess what linear epitopes of the phage proteins are immunogenic for T cells. For this purpose, we used the CD4 T cell immunogenicity prediction software[18][19], mainly because it allowed us to assess the immunogenicity for all seven most common alleles in the population[20] for each query. The software requires the protein sequence as input; the user can predict the T cell immunogenicity using the 7-allele method[20], immunogenicity method and combined method (IEDB recommended). The combined method predicts the final score that combines the predictions from the 7-allele method and immunogenicity method. In the results, the "Immunogenicity Score" ranges from 0 to 100, with low values identifying more immunogenic peptides and high values non-immunogenic peptides. And the "Combined Score" is given by combining both HLA binding and the immunogenicity prediction which presumably incorporate the capacity of being recognized by TCR. And low value identifying high capacity of being recognized by TCR.

Parameters used in the queries:

• Prediction: IEDB recommend (combined);
• Sort Peptides by: position in protein;
• Select maximum percentile rank threshold: 50.

Proteins analyzed, with no. of results in brackets:

• G3P protein (9)
• G6P protein (10)
• G7P protein (0)
• G8P protein (5)
• G9P protein (1)

The G7P protein query was negative, as it did not show any results.

Comparative analysis of epitopes

In order to reach the highest reliability possible, we then compared the results of the two softwares (MHC-Il binding prediction tool and CD4 T cell immunogenicity prediction tool), and thus the predicted epitope through their respective parameter (MHC-II binding prediction and T-cell immunogenicity prediction).

Our assumption was that the two software evaluate two different processes of the human immune response. On one hand we have MHC processing, on the other hand we have the binding affinity between epitopes and T cell receptors, thus the grade of probability to activate the T cells. In fact, it was our belief that a comparative analysis between the results of two software could lead to a more reliable prediction.

The comparative analysis was conducted as follows:

• we selected all positive 15mers and 9-mers (epitopes) peptides predicted by the MHC-II binding prediction software, each of which represented a possible epitope exposed on MHC-II complex;
• we selected all positive 15-mers and 9-mers (epitopes) peptides predicted by CD4 T cell immunogenicity prediction software;
• we then compared the two lists, looking for the no. of identical epitopes present in both lists.

G3P protein

G6P protein

G8P protein

G8P protein

B cell epitopes prediction

We used two different softwares to predict B cell epitopes namely DiscoTope and Elliscope. Whereas the former predicts only three dimensional epitopes, with the latter we were able to predict linear epitopes too.

DiscoTope

DiscoTope[21][22] is a method for predicting discontinuous epitopes from 3D structures of proteins in PDB format. The chain ID for all the viral proteins was the same: A. The used version of the software was 2.0.

Since the 3D protein structure is needed, we could analyze G3P and G8P only.

Explaining output values:

• Chain ID:The chain id of the protein chain used in prediction (specified by the user)
• Residue ID:PDB Residue id
• Residue Name:Name of the residue
• Contact Number:The residue contact number is the number of Cα atoms in the antigen within a distance of 10 Å of the residue's Cα atom. A low contact number correlates with localization of the residue close to the surface or in protruding regions of the antigen's structures.
• Propensity Score:This score tells about the probability/tendency of being part of an epitope for that particular residue. The propensity is reflected in amino acid epitope log-odds ratios, which were calculated on a set of 75 antigens. The propensity score is calculated by sequentially averaging epitope log-odds ratios within a window of 9 residues. Then the scores are summed up based on the proximity in the 3D structure of the antigen. For any given residue, the sequentially averaged log-odds scores from all residues within 10Å are summed to give the propensity score.
• Discotope Score:This score is calculated by combining the contact numbers with propensity score. DiscoTope score above the threshold value indicates positive predictions and that below the threshold value indicates negative predictions.

G3P protein

Discotope score vs aa position, the red line represents the threshold.

G8P protein

Discotope score vs aa position, the red line represents the threshold.

ElliPro

ElliPro[23] predicts linear and discontinuous antibody epitopes based on a protein antigen's 3D structure. ElliPro accepts as an input protein structures in PDB format.

Parameters used for the queries:

• Minimum score: 0.5 (default);
• Maximum distance (Angstrom): 6 (default).

G3P protein

Residue score vs position, the red line represents the threshold

Predicted linear epitopes:

Predicted discontinuous epitopes:

Linearized epitopes:

We also tried to analyze G8P, but the tool did not work for still unknown reasons, even after several attempts.

Allergic reactions

Allergic reactions can cause significant damage in individuals, especially if occure outside the hospital. For this reason, we designed an allergic reaction analysis of the phage proteins. Our idea was to use the data obtained from the immunogenicity essay to predict the probability of an allergic reaction to the phage.

Using the IEDB database

To do so, we first search in the IEDB database for any known allergies to the phage, i.e. if the only certain epitope of the phage (G3P). The query had no positive results.

Using prediction softwares

At this point, we tried to analyze the predicted epitopes with the previous softwares.

Predicting through T-cell linear epitopes

For T cell epitopes, we decided to search for the one we found in both the MHC-II binding prediction tool and the CD4+ T cell binding prediction tool, i.e. the ones resulted from the comparative analysis between the two softwares.

As for the parameters, we applied the following:

• Epitope: linear peptide (input);
• Homology rate: 80%, then exact match if 80% positive;
• Receptor: anytype;
• Assay: T cell, B cell, MHC ligand;
• MHC Restriction: any;
• Host: human;
• Disease: allergic.

G3P protein

G6P protein

G8P protein

G9P protein

Predicting through B-cell epitopes

Since DiscoTope results are shown as a DiscoTope score assigned to each amino acid, we could not rebuild the discontinuous epitopes easily. Thus, for B cell epitopes, we started analyzing the linear epitopes obtained from ElliPro. As a matter of fact, the prediction principle relies on two possible assumptions.

1. Compositional theory: we consider that every 3-dimensional epitope is continuous, i.e. every fragment of the protein can represent an epitope.
2. Positional theory: we consider that every 3-dimensional epitope can be either continuous or, more likely, discontinuous, due to the spatial structure and the sterical hindrance of the proteins.

Composition theory

Following the composition theory assumption, we search in the database every linear epitope obtained from ElliPro analysis of G3P protein.

G3P protein

Linear epitopes:

Position theory

Following the position theory assumption, we search in the database every discontinuous epitope obtained from ElliPro analysis of G3P protein. Since the IEDB database can only get linear epitopes as input, we treated discontinuous epitopes as they were, precisely, linear. We based our research on the belief that in 3-dimensional space the discontinuous epitopes behave like linear ones due to the proximity (d≤8Å) of their residues.

Linearized discontinuous epitopes:

Autoimmunity

Finally, we repeated the same process for the autoimmunity risk. One other important risk to evaluate when developing a therapeutic agent is, in fact, the phenomenon of molecular mimetism, i.e. the fact that the therapeutic agent can be similar (in terms of chemical composition, 3D structure, overall charge distribution) to a human protein, thus carrying the risk of generating an autoimmune disease.

Using the IEDB database

We made a research of both peptidic and non-peptidic material of M13 for homology in any human autoimmune disease epitopes registered in the database, with no results. We applied the same parameter as for the allergic reaction query, selecting “autoimmune” instead of “allergy”. The result was still negative.

Using prediction softwares

We searched in the database if there were any known autoimmune disease epitopes with a certain homology to our T and B cell predicted epitopes.

Predicting through linear T-cell epitopes

G3P protein

G6P protein

G8P protein

G9P protein

Predicting through B-cell epitopes

Linear epitopes:

Linearized discontinuous epitopes:

Further investigation

Professor Viola made another important statement. Even if the phage is immunogenic, there is the possibility to make it less visible to the immune system. Besides the many different delivery methods to hide phage particles before they reach the target site, there is another kind of mimetism which depends directly on the amino acid composition of the coat proteins. As a matter of fact, it is possible to slightly change the amino acid sequence of the most immunogenic epitopes to make them less immunogenic. One tool to achieve this is Deimmunization[24]. Due to a time limit of the project, we decided not to look into this software.

References

1. Moise L, Gutierrez AH, Bailey-Kellogg C, Terry F, Leng Q, Abdel Hady KM, VerBerkmoes NC, Sztein MB, Losikoff PT, Martin WD, Rothman AL, De Groot AS. The two-faced T cell epitope: examining the host-microbe interface with JanusMatrix. Hum Vaccin Immunother. 2013 Jul;9(7):1577-86. doi: 10.4161/hv.24615. Epub 2013 Apr 12. PMID: 23584251; PMCID: PMC3974887.
2. Vita R, Mahajan S, Overton JA, Dhanda SK, Martini S, Cantrell JR, Wheeler DK, Sette A, Peters B. The Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res. 2019 Jan 8;47(D1):D339-D343. doi: 10.1093/nar/gky1006. PMID: 30357391.
3. Mikkel Harndahl; Kasper Lamberth; Sune Justesen; Gustav Røder; Michael Madsen; Morten Nielsen; Claus Lundegaard; Mette Voldby Larsen; Sheila Tang; Søren Brunak; Ole Lund; Søren Buus. Large scale analysis of peptide-HLA class I interactions.
4. Larsen M.V., Lundegaard C., Lamberth K., Buus S., Lund O., and Nielsen M. 2007. Large-Scale validation of methods for cytotoxic T-lymphocyte epitope prediction. BMC Bioinformatics 8:424. PMID: 17973982.
5. Larsen M.V., Lundegaard C., Kasper Lamberth, Buus S,. Brunak S., Lund O., and Nielsen M. 2005. An integrative approach to CTL epitope prediction: A combined algorithm integrating MHC-I binding, TAP transport efficiency, and proteasomal cleavage predictions. European Journal of Immunology 35(8):2295-303. PMID: 15997466.
6. Sidney, J., Peters, B., Frahm, N. et al. HLA class I supertypes: a revised and updated classification. BMC Immunol 9, 1 (2008). https://doi.org/10.1186/1471-2172-9-1.
7. Wang P, Sidney J, Kim Y, Sette A, Lund O, Nielsen M, Peters B. 2010. Peptide binding predictions for HLA DR, DP and DQ molecules. BMC Bioinformatics. 11:568. PMID: 21092157.
8. Wang P, Sidney J, Dow C, Mothé B, Sette A, Peters B. 2008. A systematic assessment of MHC class II peptide binding predictions and evaluation of a consensus approach. PLoS Comput Biol. 4(4):e1000048.PMID: 18389056.
9. Jensen KK, Andreatta M, Marcatili P, Buus S, Greenbaum JA, Yan Z, Sette A, Peters B, Nielsen M. 2018. Improved methods for predicting peptide binding affinity to MHC class II molecules. Immunology 154(3):394-406.PMID: 29315598.
10. Nielsen M, Lund O. 2009. An artificial neural network-based alignment algorithm for MHC class II peptide binding prediction. BMC Bioinformatics. 10:296. PMID: 19765293.
11. Nielsen M, Lundegaard C, Lund O. 2007. Prediction of MHC class II binding affinity using SMM-align, a novel stabilization matrix alignment method. BMC Bioinformatics. 8:238. PMID: 17608956.
12. Sidney J, Assarsson E, Moore C, Ngo S, Pinilla C, Sette A, Peters B. 2008. Quantitative peptide binding motifs for 19 human and mouse MHC class I molecules derived using positional scanning combinatorial peptide libraries. Immunome Res 4:2. PMID: 18221540.
13. Sturniolo T, Bono E, Ding J, Raddrizzani L, Tuereci O, Sahin U, Braxenthaler M, Gallazzi F, Protti MP, Sinigaglia F, Hammer J. 1999. Generation of tissue-specific and promiscuous HLA ligand databases using DNA microarrays and virtual HLA class II matrices. Nat Biotechnol. 17(6):555-561. PMID: 10385319.
14. Kaabinejadian S, Barra C, Alvarez B, Yari H, Hildebrand WH, Nielsen M. 2022. Accurate MHC Motif Deconvolution of Immunopeptidomics Data Reveals a Significant Contribution of DRB3, 4 and 5 to the Total DR Immunopeptidome. Front Immunol. 13:835454. PMID: 35154160.
15. Reynisson B., Alvarez B., Paul S., Peters B., Nielsen M. 2020. NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Res. 48(W1):W449-W454. PMID: 32406916.
16. Jensen KK, Andreatta M, Marcatili P, Buus S, Greenbaum JA, Yan Z, Sette A, Peters B, Nielsen M. 2018. Improved methods for predicting peptide binding affinity to MHC class II molecules. Immunology 154(3):394-406. PMID: 29315598.
17. Andreatta M, Karosiene E, Rasmussen M, Stryhn A, Buus S, and Nielsen M. 2015. Accurate pan-specific prediction of peptide-MHC class II binding affinity with improved binding core identification. Immunogenetics.67(11-12):641-50. PMID: 26416257.
18. Dhanda et. al.: Prediction of HLA CD4 immunogenicity in human populations, Frontiers in Immunology, 2018, 9, 1369.
19. Paul et. al: Development and validation of a broad scheme for prediction of HLA class II restricted T cell epitopes, Journal of Immunological Methods, 2015, 422, 28-34.
20. Sinu Paul, Cecilia S. Lindestam Arlehamn, Thomas J. Scriba, Myles B.C. Dillon, Carla Oseroff, Denise Hinz, Denise M. McKinney, Sebastian Carrasco Pro, John Sidney, Bjoern Peters, Alessandro Sette. Development and validation of a broad scheme for prediction of HLA class II restricted T cell epitopes, Journal of Immunological Methods, Volume 422, 2015, Pages 28-34, ISSN 0022-1759, https://doi.org/10.1016/j.jim.2015.03.022.
21. P. H. Andersen, M. Nielsen and O. Lund. 2006. Prediction of residues in discontinuous B cell epitopes using protein 3D structures. Protein Science.15:2558-2567. [PMID: 17001032].
22. J. V. Kringelum, C. Lundegaard, O. Lund, M. Nielsen. 2012. Reliable B cell epitope predictions: impacts of method development and improved benchmarking. PLoS Comput Biol. 8:(12):e1002829. [PMID: 23300419]
23. Ponomarenko JV, Bui H, Li W, Fusseder N, Bourne PE, Sette A, Peters B. 2008. ElliPro: a new structure-based tool for the prediction of antibody epitopes. BMC Bioinformatics 9:514. PMID: 19055730.
24. Sandeep Kumar Dhanda, Alba Grifoni, John Pham, Kerrie Vaughan, John Sidney, Bjoern Peters, Alessandro Sette (2017). Development of a strategy and computational application to select candidate protein analogs with reduced HLA binding and immunogenicity. Immunology. PMID: 2883308.