抗體免疫表型與細胞免疫聚型的結構免疫資訊應用

Structural Immunoinformatics Exploits on AMI-Epitopes and CMI-Agretopes

范振杰(Chen-Chieh Fan)

指導教授 ： 高成炎

共同指導教授 ： 張春梵(Chun-Fan Chang)

國立臺灣大學/電機資訊學院/資訊工程學研究所/博士(2012年)

https://doi.org/10.6342/NTU.2012.10917

摘要

鼻咽癌(Nasopharyngeal Carcinoma, NPC)屬於鼻咽部表皮發生的癌症，常見於台灣與中國大陸東南沿海。NPC治療目前主流療程皆以放射線治療暨合併化學治療，然而局部復發與遠端轉移仍有可能發生，並且治療的副作用常會影響患者的生活品質。於此，醫界對於傳統NPC主流療程仍然積極嘗試開發相關的輔助替代療程，其中包括免疫療法。
病毒、遺傳、與環境相關三項因子皆屬高度關連於鼻咽癌NPC發生。病毒方面，鼻咽癌NPC與EB病毒(Epstein Barr virus, EBV)相互密切關聯，多數NPC細胞皆可測得EBV存在。EBV攻擊B淋巴球與抑制細胞毒殺T細胞(cytotoxic T lymphocyte, CTL)；相異於典型病毒感染，EBV造成潛伏感染時期宿主細胞表面僅只呈現少數抗原(包括EBNA1與LMP1, LMP2)，其中功用相較重要關鍵的潛伏感染時期胞膜蛋白甲(latent infection membrane protein 1, LMP1)，助使潛伏感染細胞同時具備持續增生能力暨突變遁逃細胞免疫(cell-mediated immunity, CMI)的降低CTL細胞毒殺辨識，致使長期潛伏感染細胞持續增生癌化。遺傳方面，台灣族群相較歐美族群流行病學研究調查呈現較高的NPC發生機率，似與二個族群分別具有的人類白血球組織抗原(human leukocyte antigen, HLA)第一群亞型A*02:07與A*02:01呈現高度關連的統計要件，台灣族群NPC檢體的腫瘤細胞檢測分析發現，相較常見LMP存在些微差異的突變型N-LMP1似為免疫遁逃結果。
本論文結構免疫資訊學研究目標設定發展弱化EBV-LMP1細胞免疫遁逃相關NPC免疫療法，實施對策強化EBV潛伏感染細胞相關LMP1細胞免疫抗原的呈現效率，期能提升細胞免疫系統相關LMP1辨識毒殺清除效率。結構免疫資訊演算實作架構建立已知蛋白質結構片段資料庫，利用細胞免疫CMI相關細胞免疫抗原的歐米茄型胜肽表型與聚型(epitope & agretope)，並利用遺傳演算法預測抗原胺基酸序列之立體結構，據此進行蛋白質-蛋白質對接(protein-protein docking, PPD)演算，避免使用LMP1或LMP2完整蛋白質的致癌風險，進行蛋白質-蛋白質對接(protein-protein docking, PPD)演算，預測MHC與抗原胺基酸片段之結合強度(即細胞聚型之結合強度)，演算細胞免疫抗原與A*02:07功用複合體強化(Action complex enhancement, Ace)，相關佳化胜肽片段與篩選核可藥物，分別採取核酸疫苗轉染與直接藥物投用。
強化EBV-LMP1細胞免疫抗原的呈現效率方面，本研究篩選出LMP1中胜肽表型表現較佳之片段，根據其聚型表現，尋求其胺基酸序列之最佳化組合。弱化EBV-LMP1細胞免疫遁逃方面，本研究評估目前核可之藥物，篩選出在其存在的情況下，MHC與抗原胺基酸片段之結合強度會因而增強者(即細胞聚型之結合強度)。經結構免疫資訊演算後得知相關最佳化胜肽片段與篩選核可藥物，可分別採取核酸疫苗轉殖與直接藥物投用之方式以加強免疫療法的效果。
本論文結構免疫資訊學研究發展EBV LMP1免疫遁逃相關NPC免疫療法二項施行任務，涵蓋活體外(ex vivo)活化CTL與活體內(in vivo)強化CMI：前者分離個人免疫細胞的巨噬細胞或B細胞與CTL前驅細胞並以半透網隔分區細胞培養，併用核酸疫苗呈現與直接藥物使用確效活體外活化CTL；後者個案臨床療程僅採直接藥物使用強化抗原呈現效率暨植回活體外活化CTL避免核酸疫苗毒性，或許合併療程期間自體內活化CTL達成活體內強化CMI；預期藉由二項操作提升臨床療程NPC-CMI功用，繼而有效毒殺清理EBV-LMP1與A*02:07免疫遁逃暨持續增生的潛伏感染細胞，達成減低癌化風險展現免疫療法效用。

關鍵字

鼻咽癌 ； EB病毒 ； 細胞媒介免疫 ； 潛伏感染時期胞膜蛋白甲 ； 胜肽聚型 ； 功用/聚型複合體強化

並列摘要

Nasopharyngeal carcinoma (NPC) is a squamous cell carcinoma that occurs on the epithelium of nasopharynx. It is a common malignancy in south-east Asia countries including Taiwan, Indonesia, Singapore, Malaysia, and Vietnam in addition to Hong Kong and southern China. Environmental factors, Ebstein-Barr virus (EBV), and genetic susceptibility are thought to play important roles towards the development of NPC. The radiotherapy or concurrent chemoradiotherapy of NPC clinical treatment may still occur local pathologic failure and distant metastasis in many patients despite of some outcome improvements. Moreover, the radiotherapy with chemotherapy often accompanies with acute side effects and long-term sequelae including secondary malignancy. Pursue for novel approaches aiming at improving outcome and reducing demand for conventional cytotoxic therapy seems thus to indicate immunotherapy as of an attractive option under development. The crucial advantage of antigen-specific immunotherapy is the ability to evaluate and monitor immune responses against targeted antigens and to correlate the findings with clinical responses.
NPC shows strong association with EBV infection that attacks B-lymphocytes as primary target towards resulted lifelong latent infection while and as well reveals an observed inhibition on specified cytotoxic T lymphocyte (CTL) populations with EBV antigenecity specificity. Notably, NPC latent infection case expresses only limited EBV viral antigens with less immunogenicity including EBV-encoded nuclear antigen (EBNA1) and latent membrane protein 1 and 2 (LMP1 and 2) which is greatly unlike that regular EBV latent infection case with expression of many EBV viral antigens in symptomatic EBV-related diseases. Both LMP1 and LMP2 may serve potentially as better vaccine targets due to the poor processing efficiency over with EBNA antigen while in antigen-presenting cells (APC) as of the infected B lymphocytes. However, LMP1 and LMP2 are with main shortages both in risky oncogenicity and as well in weak immunogenicity by stringent class I major histocompatibility complex (MHC-I) presentation in the host cell of infected B lymphocytes in order for cytotoxic T lymphocyte (CTL) activations in which as a result shifts the balance towards flexible class II-MHC (MHC-II) presentation of infected B lymphocytes in order for T helper (Th) lymphocyte activations with subtle feedback network to enhance B lymphocyte proliferations towards aberrant tumorigenesis.
Immunotherapeutic vaccination strategy with immunogenic vaccine polypeptides of assembled multiple epitope set is thus preferred whereas the oncogenic full-length LMP1 and/or LMP2 are therefore not recommended. Promising progress in tumor growth controlling has been exemplified in animal model studies with polyepitope vaccines comprising MHC-I equivalent class I Human Leukocyte Antigen (HLA-I) restricted CTL epitope peptides from LMP1 and LMP2 despite of being with notable restriction in a relatively narrow spectrum of HLA-I alleles as of genetic susceptibility. Thus, prevalent HLA-I alleles in NPC endemic regions as of HLA-A11, A24, B27, and B57 should also be included in designing LMP-based vaccine polyepitopes along with most common HLA-I alleles such as HLA-A2.
Likely, the restricted HLA spectrum of genetic susceptibility may indicate that the overlooked anchoring agretopes of omega-shape vaccine peptide seems to be required for crucial docking onto the MHC-I pocket sub-zones towards adequate antigen presentation of peptide epitope on demonstrating immunogenicity. The design strategy of MHC-I vaccine peptide thus seemingly demands both optimized agretopes and immunogenic epitope to which additional peptide segments for improved APC proteasome processing are attached at both flanking sides. The intended vaccine peptide of epitope and agretope may be delivered in the format of “in silico DNA vaccine” which is constructed with expression DNA sequence deduced from the intended vaccine peptide sequence and as well with upstream control sequence of LMP1/2 promoter sequence. The developed “in silico DNA vaccine” with intended specific expression in EBV latent infection lymphocytes may be verified with NPC cell line of EBV-latent infected B lymphocytes for immunogenic induction in order to demonstrate the potential ability in shifting cell-mediated immunity (CMI) pathway towards MHC-I CTL while away from MHC-II Th cell.
In this thesis, we verified structure-based immunoinformatic algorithms of implemented in-house bmPDA tool in chapter 1 towards important application aspects of vasopressin bio-mimicry peptide design of known structure, MHC-I binding epitope peptide prediction of unknown structure, and EBV LMP1 related cancer vaccine peptide design of combined structure with adequate agretope and epitope for MHC-I presentation in designed delivery format likely as of DNA vaccine. The implemented algorithm comprises three sections including constructed peptide building blocks database, assembled peptide backbone model of building block candidates, and predicted peptide surface model of functional peptides.

Basically, with the concept of tri-peptide fragment assembly in chapter 2, we implemented an in-house tool of bio-mimicry peptide design algorithm (bmPDA-tool) for modeling given peptide structure. With the extracted penta peptides (penta-pep, PDB-5mer) from all entries of current protein data bank (PDB) in order for serving as basic bmPDA building blocks, the segmental backbone angles of the 3rd alpha carbon (defined as aC[3]) towards neighboring aC[2] and aC[4] as of the middle aC[2~4] in each aC[1~5] building blocks are analyzed and constructed into searchable tri-peptides structure string (TPSS-3mer) database which is based on the “structure alphabet” with putative 22 clusters according to the parameter values including defined theta angle and edge distance, rotation axis, and rotation angle in order for the k-mean clustering analysis with bootstrapping 10,000 data entries of tri-peptide structures. With structure string alphabet of TPSS database, the mining task for similar backbone structure of 9-mer vasopressin peptide simply takes less than 1 minute for searching exact matches in entire TPSS database transformed and indexed from entire PDB.
First, to model bio-mimicry peptide structures similar to reference peptide with known backbone structure in chapter 3, the matched aC[2~4] according to serial reference penta peptide structures are mined from in-house penta-pep TPSS database with bmPDA tool in order for assembling peptide structure contig. Specifically, two mined aC[2~4] building blocks exemplified with KAV and VYN are assembled towards KAVYN contig based on superimposing [NaC/C] co-plane of both [aa] tail-with-head amino acids between two mined aC[2~4] blocks in which the spatial rotation of mined blocks is accomplished by Quaternion-based approach along with the simple spatial shift to avoid potential structural hindrance. All fused peptide conformations in respective block combinations of bio-mimicry structures are evaluated based on minimal free energy (maximal stability) of each conformation or based on maximal structure similarity to reference structure in order for ranking optimal structures by Genetic Algorithm (GA) search strategy and/or third party program such as ProCheck for instability and Ramachandran plot analysis. The yield candidate peptide structures are converted to TPSS data in which vasopressin 9-mer peptide with known backbone structure may normally yield about 400 TPSS data entries. Second, the selected peptide model with surface structure is converted to quantitative structure–activity relationship (QSAR) model which is constructed with TPSS data and quantitative descriptors including peptide surface properties of amino acids such as exposed surface, accessibility, flexibility, hydrophilicity, charge, and so forth towards binary clustering based on structure similarity and/or binding affinity with support vector machine (SVM) according to the surface structure of reference peptide.
On the first algorithm validation in chapter 4 exemplified with known backbone structure of reference peptide vasopressin 9-mer [1YF4] CYF QNC PRG, our bmPDA tool mines bio-mimicry aC[2~4] building blocks from constructed TPSS database with qualified [theta/Ad] values in order for assembling candidate peptide combinations with highly mimicking reference structure. The bmPDA-designed bio-mimicry peptide backbone structures with different amino acid sequences from vasopressin are exemplified with annotated solution numbers (SN) of KGN SVL AIP (SN.12), DGN SVL AIP (SN.36), and DGN SVL ADS (SN.37) taken from pooled combinations of candidate peptides in which further requires massive computational optimization with GA search strategy. The yield candidate peptide structures are coded as TPSS data in which vasopressin 9-mer peptide with known backbone structure may normally yield about 400 TPSS data entries in addition to the larger epitope peptide TMB-355 with about 3,000 TPSS data entries. Thus, the structure similarity evaluation on respective assembled structure combinations according to reference structures of vasopressin backbone and surface is accomplished by evaluating parameters with GA search strategy in physiochemical property, energy stability, and docking fitness based on accounted reference peptide structures of vasopressin backbone and surface.
On the second algorithm validation in chapter 5 exemplified with unknown epitope structure of EBV LMP1/LMP2 peptide sequence, our bmPDA tool mines bio-mimicry aC[2~4] building blocks from constructed TPSS database with qualified [theta/Ad] values in order for predicting epitope peptide structure for which the pre-processing filtering applies GA search strategy and/or ProCheck analysis in order to preliminarily predict and select stable peptide structures from assembled massive candidate block combinations and subsequently to be used for assembling runs until completion. Again, the predicted candidate epitope backbone and surface structures of assembled peptides are coded as TPSS data in order for full-size immunogenic epitope structure evaluation by GA search strategy with grouped parameters including physiochemical property, energy stability, docking fitness, and so forth. In that, our predicted peptide structures of EBV LMP1/2 contain epitope structure regions which demonstrate high consistency with epitope antigenecity index measured with NetCTL server, Kolaskar and Tongaonkar antigenecity scale, and Bepipred program. Moreover, the peptide design application for NPC cancer vaccine of likely omega shape MHC-I vaccine peptide from EBV LMP1/LMP2 demands both immunogenic epitope of previous session and as well optimal anchoring agretopes onto which respective peptide segments for improving proteasomal processing in antigen presentation cell (APC) are attached at either flanking sides towards integrated exogenous peptide expression in DNA construct.
Along with prediction methods for LMP1/2 epitope structures in previous session, the additional interactions between the potential docking sub-zones in HLA-I antigen presentation pocket and the anchoring agretopes of predicted candidate vaccine peptides are evaluated with converted QSAR models for accurate docking analysis by Molegro Virtual Docker towards mining qualified HLA-specific agretopes. The binding affinity between HLA docking sub-zones and peptide anchoring agretopes is evaluated with SVM based on the correlations among the docking scores and the quantitative descriptors of amino acid properties. With the reference data set of the used epitopes of NPC vaccines in previous studies, the comparison on the predicted epitope and agretopes with our bmPDA tool of structural immunoinformatic approaches reveals high consistency between the candidate agretope segments and the predicted candidate epitope segments of EBV LMP1/LMP2. Moreover, the highly potential epitope segment without effective agretope segments maybe replaced with proper agretope segments in order to become highly immunogenic epitopes with improved antigenecity index when compared to the original peptide structure as of poor immunogenic epitopes.
In chapter 6, we collected approved drugs from Drugbank. Virtual screening was done by docking with MHC receptor. Drugs with better binding affinity with MHC receptor were collected as possible candidate for adjuvant immunotherapy. Epitopes with better performance of antigenecity were collected by the same procedure in chapter 5. Epitope structure prediction was done by modeling method in chapter 2. MHC receptor and candidate drugs were docked with candidate epitopes. Drugs which could enhance the binding affinity between epitope and MHC receptor were identified. We suggest drugs with ACE (action complex enhancement) to be adjuvant immunotherapy for NPC.
In conclusion, the in-house designed HLA-I cancer vaccine peptide of epitope and agretope flanking with proteasomal processing peptide can be delivered adequately in the likely practical format from “in silico DNA vaccine” which is constructed in chapter 6 with expression DNA sequence deduced from the designed vaccine peptide sequence and as well with upstream control sequence of active LMP1/2 promoter sequence. The developed “DNA vaccine of MHC-I cancer peptide in silico” with intended specific expression in EBV latent infection lymphocytes can be verified with NPC cell line of EBV latent infection for immunogenic induction which may demonstrate the potential CMI pathway shifting towards MHC-I Tc of CTL while away from MHC-II Th cell.

並列關鍵字

Nasopharyngeal carcinoma ； Epstein-Barr virus ； cell-mediated immunity ； class I human leukocyte antigen ； latent membrane protein 1 ； agretope ； action/agretope complex enhancer

參考文獻

1. Yang X, Yu X: An introduction to epitope prediction methods and software. Reviews in medical virology 2009, 19(2):77-96.

2. Wei WI, Sham JST: Nasopharyngeal carcinoma. The Lancet 2005, 365(9476):2041-2054.

3. Guigay J: Advances in nasopharyngeal carcinoma. Current opinion in oncology 2008, 20(3):264-269.

4. Thompson MP, Kurzrock R: Epstein-Barr virus and cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 2004, 10(3):803-821.

5. Jenson HB: Epstein-Barr virus. Pediatrics in review / American Academy of Pediatrics 2011, 32(9):375-383; quiz 384.