Designing Peptide-Based HIV Vaccine for Chinese

CD4+ T cells are central to the induction and maintenance of CD8+ T cell and antibody-producing B cell responses, and the latter are essential for the protection against disease in subjects with HIV infection. How to elicit HIV-specific CD4+ T cell responses in a given population using vaccines is one of the major areas of current HIV vaccine research. To design vaccine that targets specifically Chinese, we assembled a database that is comprised of sequences from 821 Chinese HIV isolates and 46 human leukocyte antigen (HLA) DR alleles identified in Chinese population. We then predicted 20 potential HIV epitopes using bioinformatics approaches. The combination of these 20 epitopes has a theoretical coverage of 98.1% of the population for both the prevalent HIV genotypes and also Chinese HLA-DR types. We suggest that testing this vaccine experimentally will facilitate the development of a CD4+ T cell vaccine especially catered for Chinese.


Introduction
Over 30 million people have died from HIV/AIDS related illnesses since HIV was discovered in the 1980s. There are currently 33 million of HIV carriers [1]. The rate of new infection is still on the rise globally. In China, HIV infection is a great concern, especially in southern part of China, for example, Yunnan, Sichuan, Guangxi, and Xinjiang Provinces, where a large number of infected people are drug users. Additionally, in the regions of Henan, Hubei Provinces where people were infected through illicit blood collection, the rate of infection reached up to 60% of blood donors [2]. Highly active antiretroviral therapy (HAART), a combination of three or more antiretroviral drugs, is routinely used to treat individuals with HIV infection [3]. It significantly extends the lifespan and improves the quality of life of people infected with HIV but cannot eradicate the virus [4]. The course of treatment is life-long and the medicines are expensive. In developing countries, available antiretroviral drugs are still limited. Therefore, a preventive HIV vaccine is especially needed.
HIV genome is comprised of nine structural (Env, Gag, and Pol) and regulatory (Tat, Rev, Nef, Vif, Vpr, and Vpu) genes. The pol gene encodes for reverse transcriptase which is error prone. This leads to high mutation rate, 15-20% divergence between the nucleic acid sequences of different HIV clades, and 7-12% variability within each clade [5]. Although the base composition of HIV genome is stable [6], host immune response further increases the HIV nucleotide diversity.
Due to the extreme sequence diversity and high mutation rate of HIV, it has been difficult to develop an efficacious HIV vaccine. A successful HIV vaccine requires inducing neutralizing antibodies and cytotoxic T cell responses, both of which can only be optimally induced and maintained in the presence of a concurrent CD4+ T helper cell response [7]. Despite many years of basic and clinical research, to date, there are only three major human HIV vaccine clinical trials completed. Set up in 1998, AIDSVAX gp120 protein vaccine is the first HIV vaccine going through Phase III trial in human and targeted to induce neutralizing antibody activity. Although antibodies to homologous virus were elicited, they failed to neutralize heterologous viruses [8]. In 2004, a Phase IIb trial with Merck's MRKAd5, which is a trivalent vaccine including gag, pol, and nef genes in an adenovirus 5 vector, is designed for inducing cytotoxic T cell responses [9,10]. Despite the induction of significant level of IFN gammaproducing T cells, the MRKAd5 has increased the risk of HIV acquisition in vaccine recipients and failed to reduce viral load after HIV infection [11]. Later in 2009, a Phase III trial of RV144 HIV-1 vaccine was completed in Thailand, which is a vaccine combination comprised of ALVAC (a vaccine containing genetically engineered versions of gag, env, and pol inserted in canarypox vector) and AIDSVAX (a bivalent gp120 envelope protein vaccine). These vaccines are theoretically capable of eliciting both CD8+ T cell response and neutralizing antibody response. Despite neither vaccine worked alone, in the combination, they unexpectedly lowered the HIV incidence by 31.2% in vaccine recipients; however, they did not reduce viral load [12]. These large clinical trials have opened new questions and revealed new opportunities for HIV vaccine research, including a rethinking of the need for a vaccine for CD4+ T helper cells.
In order to stimulate a CD4+ T helper cell response, antigens need to be processed and presented through MHC class II molecules. The form of antigen could be either whole protein or peptide epitopes. A previous study with a subunit vaccine comprised of 18 CD4+ T helper cell epitopes has demonstrated an efficient induction of robust helper T cell response in a Phase I clinical trial in Caucasian population [13]. Whether a similar strategy works in Chinese population requires to be tested.
To select antigenic epitopes for a vaccine, one must address several issues. One, HIV exhibits high mutation rates, and thus conserved sequences may be needed to cover a given population. Two, the human leukocyte antigen (HLA) is highly polymorphic, and it restricts the proportion of individuals who will respond to a particular antigen [14,15]. To overcome these problems, promising T cell epitopes that bind to several HLA alleles for maximal population coverage should be selected [16], and a large variety of HIV sequences should be considered in the design of a HIV vaccine.
MHC class II is a heterodimer that is comprised of a monomorphic and a highly polymorphic chain. There are over 400 class II alleles identified, spreading among HLA-DM, HLA-DO, HLA-DP, HLA-DQ, and HLA-DR loci. Among them DRB1 is the most polymorphic gene, consists of 221 alleles; followed by DPB1 and DQB1 that has 84 and 39 alleles, respectively. Whereas other gene loci may have only 1 or 2 alleles [17]. Therefore, DRB1 is the best choice to optimize MHC II coverage. The frequency of HLA-DRB1 serotype differs among ethnic groups. Within DRB1 allotype, DRB1 * 11 and 13 serotypes present in 16% and 14% of black population, whereas, in Caucasoid and Chinese, DRB1 * 07 and DRB1 * 11 and DRB1 * 12 and DRB1 * 15 appear in the highest percentage [17]. The above evidences support the development of a new HIV vaccine specifically for Chinese population. Such a vaccine should have higher probability in dealing with circulating HIV serotypes in China.
To overcome these complex issues of vaccine design, bioinformatics methods may help to determine common features of vaccine antigens that have potential to deal with divergent population and HIV quasispecies. Specifically, bioinformatics-based approach is the most feasible method in screening a large set of peptide epitopes and selection of promising vaccine antigens. In this study, we extracted 821 HIV sequence and 46 Chinese DRB1 alleles from public information and compiled a database. A combination of 7 public available epitope prediction algorithms was used to screen the database and identify CD4+ T cell epitopes as HIV vaccine antigens. We selected a set of 20 epitopes, which in combination could cover more than 98% of our target population.

Epitope Selection.
All epitopes are 15 amino acids in length. To be a potential epitope, it must have a MHC binding affinity threshold of IC 50 = 500 nM or below. A selected epitope was removed from the epitope pool before the next prediction. The process is repeated until all epitopes were selected. All calculations of epitope selection process were conducted in INFORSENSE Knowledge Discovery Environment (KDS) software platform [36]. The mathematical model used to calculate the predictive score for each DR allele of known coverage (as listed in Table 1) is the following equations: if in the combination of HIV sequence and DR allele 0, Otherwise.

(II)
In the first equation (I), represents the epitope; ( ) is the percentage coverage of number DRB1 allele; ( ) is the function to indicate whether the epitope exists in the combination of HIV sequence and DR allele, existence scored 1, and absence scored 0. ( ) is the sum of number of times of the binding of HIV sequence and DR allele after being standardized to the proportion of DR allele in all DRB1 We selected epitopes from a combined pool of epitopes through KDS platform using 7 prediction methods from IEDB with a dataset that consisted of 821 circulating HIV genome sequences in China and 46 Chinese HLA-DRB1 alleles. The epitopes bind to MHC class I molecules that were removed first, and then the value of IC 50 was considered. Next, we ranked all epitopes based on the coverage score (the higher the better coverage in HIV genome and DR-HLA alleles). After an epitope has been selected, it was removed from the database before next selection. This process was repeated until 20 epitopes were selected. The workflow diagram of this procedure was illustrated in Figure 1.

The Coverage Distribution of HLA-DR of Chinese Population.
A total of 46 HLA-DR alleles were identified from AFND ( Table 1). The alleles were listed and its coverage in Chinese population was given. The table showed the coverage ranged from 0.1% (DRB1 * 08:09) to 6.77% (DRB1 * 07:01) and in a total of 95.2% of the Chinese population. The sample population comprises 1704 individuals of the Han ethnicity. This information was obtained from ten regions within the mainland, China, and two other regions, Hong Kong and Singapore, where Chinese ethnicity dominates. Among them, the DRB1-02, -05, and -06 genotypes were not detected.

The Diversity of Epitope Coverage.
With a combination of 7 existing epitope prediction methods in IEDB, using database comprised of 46 different DRB1 alleles and 821 full genome sequences of HIV isolates circulating in China, we then predicted 38,460,402 potential epitopes. After duplicates were removed, 21,007,527 potential epitopes remained. We scored these epitopes based on the allele coverage and total coverage score, which was in general normally distributed. As shown in Figure 2, most epitopes displayed low coverage scores, 0.1 or lower; the highest epitope count reached approximately 3000.

HIV Epitopes
Specifically for Chinese Population. By using the methods described above, we obtained 20 epitopes, which in theory covered all 46 DRB1 allelic genotypes and 821 Chinese HIV sequences ( Table 2). All 20 epitopes were selected for binding to MHC class II and absence of binding to MHC class I. Table 2 listed the amino acid sequences of the 20 epitopes, their location in HIV-1 gene, their percentage of coverage in HIV-1 genome sequences from 4% to 43%, the proportion in the HLA-DR allele sequences between 52% and 100%, and the total coverage in both sequences as low as 4% and the highest at 41%. One single epitope WIILGLNKIVRMYSP covered 41% of both DRB1 and Chinese specific HIV-1 genome sequences, which is of note. This epitope had been reported before [18]. In fact, 4 other predicted epitopes (LNKIVRMYSPT-SILD, GFPVRPQVPLRPMTY, VDRFYKTLRAEQASQ, and LYKYKVVKIEPLGVA) have also been published previously [22][23][24]26] and 4 peptide sequences (PVVSTQLLLNGSLAE,   LRIIFAVLSIVNRVR, ILDLWVYHTQGYFPD, and YKRWI-ILGLNKIVRM) were reported in patents before [19][20][21]25], whereas the remaining 11 epitopes have never been reported. All 20 epitopes together provided 98.1% coverage in HIV genome and HLA-DR alleles. These predicted epitopes were found in HIV-1 gag, env, pol, and nef genes. Six of them were in gag gene, 6 in env, 2 in pol, and 6 in nef.
We then applied the new method to a previously published HIV vaccine comprised of T helper epitopes and tested in clinical trial [13]. The table listed 17 epitopes, from gag, pol, env, and vpu genes. One published epitope that has a HIA binding IC 50 above our threshold of 500, Env 566 IKQFINMWQEVKAMY, was not listed. For these epitopes, HIV coverage is from 2% to 43%, DR coverage is between 35% and 98%, and specific coverage is at highest of 41% and in sum of 69%.

Discussion
In this paper, we described a novel method for designing a peptide-based T helper cell vaccine for HIV, which is specific for Chinese HIV strains and Chinese MHC class II genotypes. The current method has several advantages. First, our methodology of epitope prediction is easily accessible to public use. In fact, it is a combination of all seven existing methods publically available in IEDB. The IEDB database comprises a series of most up-to-date and evidence based methods specifically created for the prediction of MHC restricted T cell epitopes. In contrast to other studies that only used one of the methods, we used them all for more accurate prediction of MHC class II restricted T helper epitopes.
So far, there are three major types of bioinformatics methods for the prediction of MHC class II restricted T helper cell epitopes. One is called matrix alignment algorithm, and these are SMM, ARB, and Sturniolo methods. This algorithm uses published T cell epitopes and their respective binding affinity to MHC class II, in terms of the IC 50 value, to determine epitopes. The other relies on machine learning, and NNalign and NetMHCIIpan methods belong to this category. New sequences are subjected to computer simulated models to predict whether any epitopes can bind to a particular MHC II to high enough affinity. The third type combines several methods together to predict epitopes. These include Consensus method and ComLib method.
Consensus method was reported to provide highest true positive rate, followed by NN-align and ARB [37]. NetMHCIIpan performed the best among all other panspecific methods for MHC class II with varied experimental settings [38]. NN-align performs especially well in handling large dataset among all other machine learning methods and in combination with ARB outperforms the use of NN-align alone [30].
In this study, we used all above seven methods simultaneously, scored the potential epitopes independently, and then used IC 50 value as a filter to select T cell epitopes that have the broadest population coverage. Our method did not use all 8 IEDB recommended methods but integrated 7 of the IEDB methods because the 8th IEDB method is an integration of the other seven and thus not an independent measurement. The method we used could be considered as "greedy" algorithm in the bioinformatics field, which predicts the best epitope among all in a pool of potential epitopes. Thus, we believe an integrated method that uses a combination of all seven original algorithms might be the best to predict more accurately MHC class II epitopes.
Another unique feature of our study is that we designed candidate helper T cell vaccine targets specifically to the  [27,39,40]. We extracted all 821 subtypes of HIV-1 strains which are mostly subtypes B and C for developing a highly specific vaccine for Chinese population. As T helper cell epitopes are recognized through MHC class II, and that Chinese exhibit divergence DRB1 alleles, we also included 46 published Chinese HLA-DRB1 genotypes into our prediction. In comparison to a previous paper that selected MHC class II binders according to the binding affinity to multiple HLA-DR subtypes [13], we focused on DRB1 alleles which are most polymorphic among human MHC class II loci and thus directed our study to be more specific and increased possibility to induce T cell responses specifically for Chinese.
One limitation in our study, as shown in Table 1, is that DRB1 genotypes 2, 5, and 6 were not included. This is due to a lack of publication of any information on DRB1 * 02, 05, and 06. Therefore, our dataset represents what is currently available; that is, there are only 46 DRB1 alleles in Chinese population.
By using our method, we obtained 20 helper T cell epitopes which covered 98.1% of HIV strains known to have been circulating in China and all Chinese HLA-DR genotypes. There are limited studies that have tested designed peptide T helper vaccine in humans. In a published paper that contains 18 T helper epitopes [13], our combination of epitope predication methods found that these epitopes covered 69% of Chinese HIV genomes (Table 3). In a different population that is predominantly Caucasian, these epitopes combined have a 100% coverage. Thus, the difference in the coverage may suggest our predicting method is more specific for Chinese population, and our epitopes are better potential HIV vaccine candidate for Chinese. Furthermore, 9 epitopes we obtained have been published before and 11 are not. Thus, we both have the empirical evidence to support that our allelic specific peptides have the potential to stimulate T cell responses and new epitopes to suggest that our prediction is innovative. Total specific coverage 5 69% 1 The epitopes were selected from a published paper.
Data in columns 2-5 were calculated using the same method as in Table 2.
There was one core epitope WIILGLNKIVRMY, appeared in both studies, showing very high HIV, HLA-DR, and specific coverage. The Gag epitope with two amino acids modification WIILGLNKIVRMYSP was reported to stimulate strong CD4+ T responses [27]; another variant of the same epitope KRWIILGLNKIVRMY exhibited superior HLA-DR binding capacity [13]. Another difference between our study and that published is that our epitopes consisted of those in nef gene but not vpu gene, whereas Walker's study did not cover nef but vpu. These comparisons suggest that a vaccine designed predominantly for Caucasian may not be optimal for Chinese population. One epitope, for instance, Env 566 (IKQFINMWQEVKAMY) [13], given in Walker's paper, was not picked up in our study.
Our method predicted epitopes, in theory, together covered 98.1% of HIV-1 genome and Chinese specific DRB1 alleles. In comparison, Walker's study reported 18 T helper cell epitopes that cover 100% of the global population. By using a prediction algorithm which based mostly on HLA supertypes [13]. However, when submitted to our new prediction method, the same epitopes only achieved 69% of coverage of the Chinese population. The discrepancy in methods for prediction may give different results. Further experimental evidence is required to find out whether our method is more accurate.
The allele coverage of DRB1 for Chinese was based on 1704 subjects of whom 1569 were from mainland China and 135 were from Hong Kong and Singapore. All Chinese allele data regarding DRB1 frequencies were extracted from AFND, and all 1704 subjects were Chinese Han ethnics. There is no information on other minor national groups in China available. This may lead to inaccuracy in prediction of helper T cell epitopes for the Chinese. Larger sample size may improve the quality of our prediction.

Conclusions
In this study, we report a novel bioinformatics method for designing peptide epitope based T helper vaccine for HIV. We suggest further in vitro and in vivo experiments to be performed to test the immunogenicity of this vaccine and improvement of method of prediction to be made when necessary.