Mosaic Clade M Human Immunodeficiency Virus Type 1 (hiv-1) Gag Immunogens

(12)United States Patent(10)Patent No.: US 9,844,590 B2
  et al. (45) Date of Patent:Dec.  19, 2017

(54)Mosaic clade M human immunodeficiency virus type 1 (HIV-1) gag immunogens 
(75)Inventors: LOS ALAMOS NATIONAL SECURITY, LLC,  Los Alamos, NM (US); 
(73)Assignees:Duke University,  Durham, NC (US), Type: US Company;
Los Alamos National Security, LLC,  Los Alamos, NM (US), Type: US Company;
Beth Israel Deconess Medical Center,  Boston, MA (US), Type: US Company;
The UAB Research Foundation,  Birmingham, AL (US), Type: US Company
(*)Notice: Subject to any disclaimer, the term of this patent is extended or adjusted under 35 U.S.C. 154(b) by 0 days. 
(21)Appl. No.: 14/726,373 
(22)Filed: May  29, 2015 
(65)Prior Publication Data 
 US 2015/0265700 A1 Sep.  24, 2015 
 Related U.S. Patent Documents 
(60) .
Division of application No. 12/737,761, now Pat. No. 9,044,445 , which is a continuation of application No. 12/192,015, filed on Aug.  14, 2008, now Pat. No. 7,951,377 , which is a continuation-in-part of application No. 11/990,222, now Pat. No. 8,119,140 .
(60)Provisional application No. 60/739,413, filed on Nov.  25, 2005.
 Provisional application No. 60/710,154, filed on Aug.  23, 2005.
Jan.  1, 2013 A 61 K 39 21 F I Dec.  19, 2017 US B H C Jan.  1, 2013 A 61 K 39 12 L I Dec.  19, 2017 US B H C Jan.  1, 2013 C 07 K 14 005 L I Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 7 00 L I Dec.  19, 2017 US B H C Jan.  1, 2013 A 61 K 2039 53 L A Dec.  19, 2017 US B H C Jan.  1, 2013 A 61 K 2039 70 L A Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 2710 24143 L A Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 2740 16022 L A Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 2740 16034 L A Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 2740 16122 L A Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 2740 16134 L A Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 2740 16234 L A Dec.  19, 2017 US B H C Jan.  1, 2013 C 12 N 2740 16334 L A Dec.  19, 2017 US B H C
(51)Int. Cl. A61K 039/21 (20060101); C07K 014/005 (20060101); A61K 039/12 (20060101); C12N 007/00 (20060101); A61K 039/00 (20060101)

(56)References Cited
 5,580,859  A  12/1996    Felgner et al.     
 5,589,466  A  12/1996    Felgner et al.     
 5,703,055  A  12/1997    Felgner et al.     
 6,710,173  B1  3/2004    Binley et al.     
 7,655,774  B2  2/2010    Mullins et al.     
 7,951,377  B2  5/2011    Korber et al.     
 8,119,140  B2  2/2012    Korber et al.     
 9,011,873  B2  4/2015    Korber et al.     
 9,011,875  B2  4/2015    Korber et al.     
 9,044,445  B2  6/2015    Korber et al.     
 2002//0198162  A1  12/2002    Punnonen et al.     
 2003//0044421  A1  3/2003    Emini et al.     
 2003//0104011  A1  6/2003    Rios     
 2003//0147888  A1  8/2003    Haynes et al.     
 2003//0180314  A1  9/2003    DeGroot     
 2003//0194411  A1  10/2003    Rubinstein et al.     
 2004//0001851  A1  1/2004    Haynes et al.     
 2005//0137387  A1  6/2005    Mullins et al.     
 2006//0216305  A1  9/2006    Lal et al.     
 2006//0275897  A1  12/2006    Nabel et al.     
 2007//0178562  A1  8/2007    Haynes et al.     
 2009//0324631  A1  12/2009    Korber et al.     
 2011//0150915  A1  6/2011    Korber et al.     
 2011//0301328  A1  12/2011    Korber et al.     
 2012//0121631  A1  5/2012    Korber et al.     

       WO       2002/004493                         1/2002      
       WO       2005/001029       A2                1/2005      
       WO       2005/012502                         2/2005      
       WO       2005/028625                         3/2005      
       WO       2005/035555                         4/2005      
       WO       2007/024941                         3/2007      
       WO       2007/047916                         4/2007      
       WO       2010/019262                         2/2010      
       WO       2012/047267                         4/2012      

  Gag GenBank accession No. U52953.
  Nef core GenBank accession No. AF069670.
  Nef core GenBank accession No. K02083.
  Nef core GenBank accession No. U52953.
  Barugahare et al., “Human Immunodeficiency Virus-Specific Responses in Adult Ugandans: Patterns of Cross-Clade Recognition,” J. Virol. Apr. 2005; 79:7 4132-4139.
  Oxenius, et al., “HIV-Specific Cellular Immune Response is Inversely Correlated with Disease Progression as Defined by Decline of CD4+ T Cells in Relation to HIV RNA Load,” J Infect Dis. (2004) 189 (7): 1199-1208.
  Barouch et al., “Control of Viremia and Prevention of Clinical AIDS in Rhesus Monkeys by Cytokine-Augmented DNA Vaccination”, Science Oct. 20, 2000: vol. 290, Issue 5491, pp. 486-492.
  Schmitz et al., “Control of Viremia in Simian Immunodeficiency Virus Infection by CD8+ Lymphocytes,” Science Feb. 5, 1999: vol. 283, Issue 5403, pp. 857-860.
  Barouch et al., “Viral Escape from Dominant Simian Immunodeficiency Virus Epitope-Specific Cytotoxic T Lymphocytes in DNA-Vaccinated Rhesus Monkeys,” J. Virol. Jul. 2003; 77:13 7367-7375.
  Moore and Burton, “Urgently needed: a filter for the HIV-1 vaccine pipeline,” Nat Med. Aug. 2004;10(8):769-71.
  Altfeld et al., “Enhanced detection of human immunodeficiency virus type 1-specific T-cell responses to highly variable regions by using peptides based on autologous virus sequences,” J Virol. Jul. 2003;77(13):7330-40.
  Norris et al., “Fine specificity and cross-clade reactivity of HIV type 1 Gag-specific CD4+ T cells,” AIDS Res Hum Retroviruses. Mar. 2004;20(3):315-25.
  Jones et al. “Determinants of Human Immunodeficiency Virus Type 1 Escape from the Primary CD8+ Cytotoxic T Lymphocyte Response,” J Exp Med. Nov. 15, 2004;200(10):1243-56.
  Allen et al. “De Novo Generation of Escape Variant-Specific CD8+ T-Cell Responses following Cytotoxic T-Lymphocyte Escape in Chronic Human Immunodeficiency Virus Type 1 Infection” J. Virol. Oct. 15, 2005; 79:20 12952-12960.
  Feeney et al., “HIV-1 viral escape in infancy followed by emergence of a variant-specific CTL response,” J Immunol. Jun. 15, 2005;174(12):7524-30.
  Killian et al., “Clonal breadth of the HIV-1-specific T-cell receptor repertoire in vivo as determined by subtractive analysis,” AIDS. Jun. 10, 2005;19(9):887-96.
  Milicic et al., “CD8+ T Cell Epitope-Flanking Mutations Disrupt Proteasomal Processing of HIV-1 Nef,” J Immunol 2005 175:4618-4626.
  Kong et al., “Immunogenicity of Multiple Gene and Clade Human Immunodeficiency Virus Type 1 DNA Vaccines,” J. Virol. Dec. 2003; 77:23 12764-12772.
  Yusmin et al., “Clustering Patterns of Cytotoxic T-Lymphocyte Epitopes in Human Immunodeficiency Virus Type 1 (HIV-1) Proteins Reveal Imprints of Immune Evasion on HIV-1 Global Variation,” J. Virol. Sep. 2002; 76:17 8757-8768.
  Lee et al., “T Cell Cross-Reactivity and Conformational Changes during TCR Engagement,” J Exp Med 2004 200:1455-1466.
  Williamson et al., “Characterization and selection of HIV-1 subtype C isolates for use in vaccine development,” AIDS Res Hum Retroviruses. Feb. 2003;19(2):133-44.
  Frahm et al., “Consistent Cytotoxic-T-Lymphocyte Targeting of Immunodominant Regions in Human Immunodeficiency Virus across Multiple Ethnicities” J. Virol. Mar. 2004; 78:5 2187-2200.
  Lichterfeld et al., “HIV-1 Nef is preferentially recognized by CD8 T cells in primary HIV-1 infection despite a relatively high degree of genetic diversity,” AIDS. Jul. 2, 2004;18(10):1383-92.
  Hel et al., “Improved Vaccine Protection from Simian AIDS by the Addition of Nonstructural Simian Immunodeficiency Virus Genes,” J Immunol 2006 176:85-96.
  Blagoveshchenskaya et al., “HIV-1 Nef downregulates MHC-I by a PACS-1- and PI3K-regulated ARF6 endocytic pathway,” Cell. Dec. 13, 2002;111(6):853-66.
  Masemola et al., “Hierarchical Targeting of Subtype C Human Immunodeficiency Virus Type 1 Proteins by CD8+ T Cells: Correlation with Viral Load,” J. Virol. Apr. 2004; 78:7 3233-3243.
  Bansal et al., “CD8 T-cell responses in early HIV-1 infection are skewed towards high entropy peptides,” AIDS. Feb. 18, 2005;19(3):241-50.
  Seaman et al., “Multiclade Human Immunodeficiency Virus Type 1 Envelope Immunogens Elicit Broad Cellular and Humoral Immunity in Rhesus Monkeys,” J. Virol. Mar. 2005; 79:5 2956-2963.
  Singh et al., “The Role of T Cell Antagonism and Original Antigenic Sin in Genetic Immunization,” J Immunol 2002 169:6779-6786.
  André et al., “Increased Immune Response Elicited by DNA Vaccination with a Synthetic gp120 Sequence with Optimized Codon Usage,” J. Virol. Feb. 1998 ; 72:2 1497-1503.
  Haas et al., “Codon usage limitation in the expression of HIV-1 envelope glycoprotein,” Curr Biol. Mar. 1, 1996;6(3):315-24.
  Ammaranond et al., “A new variant cytotoxic T lymphocyte escape mutation in HLA-B27-positive individuals infected with HIV type 1,” AIDS Res Hum Retroviruses. May 2005;21(5):395-7.
  Kiepiela et al., “Dominant influence of HLA-B in mediating the potential co-evolution of HIV and HLA,” Nature. Dec. 9, 2004;432(7018):769-75.
  Altfeld et al, “HIV-1 superinfection despite broad CDS+ T-cell responses containing replication of the primary virus”, Nature, 420(6914):434-439 (2002).
  Barouch et al., “Mosaic HIV-1 Vaccines Expand the Breadth and Depth of Cellular Immune Responses in Rhesus Monkeys”, Nat Med., 16(3):319-323 (2010).
  Desrosiers, “Prospects for an AIDS Vaccine”, Nature Medicine, 10(3):221-223 (2004).
  Doria-Rose et al, “Human Immunodeficiency Virus Type 1 Subtype B Ancestral Envelope Protein is Functional and Elicits Neutralizing Antibodies in Rabbits Similar to Those Elicited by a Circulating Subtype B Envelope”, Journal of ViroloRJI, 79(17): 11214-11224 (2005).
  Fischer et al, “Polyvalent vaccines for optimal coverage of potential T-cell epitopes in global HIV-1 variants”, Nature Medicine, 13(1): 100-106 (2007).
  Gallo, “The end or the beginning of the drive to an HIV-preventive vaccine: a view from over 20 years,” The Lancet, 366:1894-1898.(2005).
  Gao et al, “Antigenicity and Immunogenicity of a Synthetic Human Immunodeficiency Virus Type 1 Group M Consensus Envelope Glycoprotein”, Journal o{ViroloKJI, 79(2): 1154-1163 (2005).
  Gaschen et al, “Diversity Considerations in HIV-1 Vaccine Selection”, Science, 296(5577):2354-2360 (2002).
  Go et al., “Giycosylation Site-Specific Analysis of Clade C HIV-1 Envelope Proteins”, Journal ofProteome Research, 8(9):4231-4242 (Author Manuscript). (2009).
  Hanke et al, “DNA multi-CTL epitope vaccines for HIV and Plasmodium falciparum: immunogenicity in mice”, Vaccine, 16(4):426-435 (1998).
  Haynes et al, “Aiming to induce broadly reactive neutralizing antibody responses with HIV-1 vaccine candidates,” Expert Rev. Vaccines, 5(3)347-363 (2006).
  International Search Report dated Apr. 6, 2010 issued in International Appln. No. PCTIUS2009/004664.
  International Search Authority Written Opinion dated Apr. 6, 2010 issued in International Appln. No. PCTIUS09/004664.
  International Search Report dated Jul. 3, 2008 issued in International Appln. No. PCTIUS06/32907.
  International Search Report dated Aug. 27, 2008 issued in International Appln. No. PCTIUS06/32907.
  International Search Report dated Apr. 18, 2012 issued in International Appln. No. PCTIUS2011/001664.
  International Search Authority Written Opinion dated Apr. 18, 2012 issued in International Appln. No. PCTIUS2011100166.
  International Search Report dated Aug. 27, 2008 issued in WO 2007/024941.
  International Search Report dated Aug. 27, 2008 in WO 2007/024941 (Korber).
  Kong et al., “Expanded Breadth of the T-Cell Response to Mosaic Human Immunodeficiency Virus Type 1 Envelope DNA Vaccination”, Journal ofViroloKJI, 83(5):2201-2215 (2009).
  Korber et al., “T-Cell Vaccine Strategies for Human Immunodeficiency Virus, the Virus with a Thousand Faces”, Journal ofViroloKJI, 83(17).8300-8314 (2009).
  Korber et al, “Evolutionary and immunological implications of contemporary HIV-1 variation”, British Medical Bulletin, 58:19-42 (2001).
  Letvin, N., “Progress and obstacles in the development of an AIDS vaccine” (2006).
  Liao et al., “A Group M Consensus Envelope Glycoprotein Induces Antibodies that Neutralize Subsets of Subtype Band C HIV-1 Primary Viruses”, Virology, 353:268-282 (Author Manuscript). (2006).
  McMichael, A. J., “HIV vaccines”, Ann. Rev. Immunol., 24:227-255 (2006).
  Nabel et al, “HIV vaccine strategies”, Vaccine, 20(15):1945-1947 (2002).
  Santra et al., “Mosaic Vaccines Elicit CD8+ T lymphocyte Responses in Monkeys that Confer Enhanced Immune Coverage of Diverse HIV Strains”, Nat Med., 16(3):324-328 (2010).
  Santra et al., “A centralized gene-based HIV-1 vaccine elicits broad cross-clade cellular immune responses in rhesus monkeys”, PNAS, 105(30):10489-10494 (2008).
  Shinoda et al, “Polygene DNA vaccine induces a high level of protective effect against HIV-vaccinia virus challenge in mice”, Vaccine, 22:3676-3690 (2004).
  Supplementary European Search Report dated Jul. 11, 2012 issued in EP 09 80 698.
  Supplementary European Search Report dated Oct. 25, 2012 issued in EP 06802155.
  Supplementary European Search Report dated Nov. 9, 2012 issued in EP 06802155.
  Tomaras et al., “Initial B-Cell Responses to Transmitted Human Immunodeficiency Virus Type 1: Virion-Binding Immunoglobulin M (IgM) and IgG Antibodies Followed by Plasma Anti-GP41 Antibodies with Ineffective Control of Initial Viremia”, Journal of ViroloRJI, 82(24): 14229-14263. (2008).
  Walker et al., “Toward an AIDS vaccine”, Science, 320:760-765 (2008).
  Weaver et al., “Cross-Subtype T-Cell Immune Responses Induced by a Human Immunodeficiency Virus Type 1 Group M Consensus Env. Immunogen”, Journal o/Virology, 80(14):6745-6756. (2006).
  Ferrari, G, et al; Identification of highly conserved and broadly cross-reactive HIV type 1 cytotoxic T lymphocyte epitopes as candidate immunogens for inclusion in Mycobacterium bovis BCG-vectored HIV vaccines; AIDS Res Hum Retroviruses; Sep. 20, 2000; 16(14); 1433-43.
  Robertson, DL, et al; Recombination in HIV-1; Nature; Mar. 9, 1995; 374(6518); 124-6.
  Sbai, H, et al; Use of T cell epitopes for vaccine development; Curr Drug Targets Infect Disord.; Nov. 2001; 1(3); 303-13.
  European search report and opinion dated May 11, 2017.
  Gag GenBank accession No. AF004885.
  Gag GenBank accession No. K03455.
  Gag GenBank Accession Nos. U52953, p. 1 of 4.
  GenBank accession No. AF530576.
  GenBank accession No. AF533131.
  GenBank accession No. AY173953.
  GenBank accession No. AY856956.
  GenBank accession No. AY857186.
  Net core GenBank accession No. AF069670.
  Net core GenBank accession No. K02083.
  NEF core GenBank Accession Nos. U52953, p. 2 of 4.
  Peng et al., “Replicating Ad-recombinants encoding non-myristoylated rather than wild-type HIV Nef elicit enhanced cellular immunity”, AIDS, 20:2149-2157 (2006).
     Primary Examiner —Jeffrey Parkin
     Art Unit — 1648
     Exemplary claim number — 1
(74)Attorney, Agent, or Firm — Baker Donelson



The present invention relates, in general, to an immunogenic composition (e.g., a vaccine) and, in particular, to a polyvalent immunogenic composition, such as a polyvalent HIV vaccine, and to methods of using same. The invention further relates to methods that use a genetic algorithm to create sets of polyvalent antigens suitable for use, for example, in vaccination strategies.
9 Claims, 109 Drawing Sheets, and 46 Figures

[0001] This application is a divisional of U.S. application Ser. No. 12/737,761, filed Feb. 1, 2012, which is the U.S. national phase of International Application No. PCT/US2009/004664, filed 14 Aug. 2009, which designated the U.S. and is a continuation of U.S. application Ser. No. 12/192,015, filed 14 Aug. 2008, now U.S. Pat. No. 7,795,377 and a continuation-in-part of U.S. application Ser. No. 11/990,222, filed Feb. 8, 2008, now U.S. Pat. No. 8,119,140, which is the U.S. national phase of International Application No. PCT/US2006/032907, filed Aug. 23, 2006, which designated the U.S. and claims the benefit of U.S. Provisional Application No. 60/710,154, filed Aug. 23, 2005, and U.S. Provisional Application No. 60/739,413, filed Nov. 25, 2005. The entire contents of each of the above-identified applications are hereby incorporated herein by reference.


[0002] The present invention relates, in general, to an immunogenic composition (e.g., a vaccine) and, in particular, to a polyvalent immunogenic composition, such as a polyvalent HIV vaccine, and to methods of using same. The invention further relates to methods that use a genetic algorithm to create sets of polyvalent antigens suitable for use, for example, in vaccination strategies.


[0003] Designing an effective HIV vaccine is a many-faceted challenge. The vaccine preferably elicits an immune response capable of either preventing infection or, minimally, controlling viral replication if infection occurs, despite the failure of immune responses to natural infection to eliminate the virus (Nabel, Vaccine 20:1945-1947 (2002)) or to protect from superinfection (Altfeld et al, Nature 420:434-439 (2002)). Potent vaccines are needed, with optimized vectors, immunization protocols, and adjuvants (Nabel, Vaccine 20:1945-1947 (2002)), combined with antigens that can stimulate cross-reactive responses against the diverse spectrum of circulating viruses (Gaschen et al, Science 296:2354-2360 (2002), Korber et al, Br. Med. Bull. 58:19-42 (2001)). The problems that influenza vaccinologists have confronted for decades highlight the challenge posed by HIV-1: human influenza strains undergoing antigenic drift diverge from one another by around 1-2% per year, yet vaccine antigens often fail to elicit cross-reactive B-cell responses from one year to the next, requiring that contemporary strains be continuously monitored and vaccines be updated every few years (Korber et al, Br. Med. Bull. 58:19-42 (2001)). In contrast, co-circulating individual HIV-1 strains can differ from one another by 20% or more in relatively conserved proteins, and up to 35% in the Envelope protein (Gaschen et al, Science 296:2354-2360 (2002), Korber et al, Br. Med. Bull. 58:19-42 (2001)).
[0004] Different degrees of viral diversity in regional HIV-1 epidemics provide a potentially useful hierarchy for vaccine design strategies. Some geographic regions recapitulate global diversity, with a majority of known HIV-1 subtypes, or clades, co-circulating (e.g., the Democratic Republic of the Congo (Mokili & Korber, J. Neurovirol 11(Suppl. 1):66-75 (2005)); others are dominated by two subtypes and their recombinants (e.g., Uganda (Barugahare et al, J. Virol. 79:4132-4139 (2005)), and others by a single subtype (e.g., South Africa (Williamson et al, AIDS Res. Hum. Retroviruses 19:133-144 (2003)). Even areas with predominantly single-subtype epidemics must address extensive within-clade diversity (Williamson et al, AIDS Res. Hum. Retroviruses 19:133-44 (2003)) but, since international travel can be expected to further blur geographic distinctions, all nations would benefit from a global vaccine.
[0005] Presented herein is the design of polyvalent vaccine antigen sets focusing on T lymphocyte responses, optimized for either the common B and C subtypes, or all HIV-1 variants in global circulation [the HIV-1 Main (M) group]. Cytotoxic T-lymphocytes (CTL) directly kill infected, virus-producing host cells, recognizing them via viral protein fragments (epitopes) presented on infected cell surfaces by human leukocyte antigen (HLA) molecules. Helper T-cell responses control varied aspects of the immune response through the release of cytokines. Both are likely to be crucial for an HIV-1 vaccine: CTL responses have been implicated in slowing disease progression (Oxenius et al, J. Infect. Dis. 189:1199-208 (2004)); vaccine-elicited cellular immune responses in nonhuman primates help control pathogenic SIV or SHIV, reducing the likelihood of disease after challenge (Barouch et al, Science 290:486-92 (2000)); and experimental depletion of CD8+ T-cells results in increased viremia in SIV infected rhesus macaques Schmitz et al, Science 283:857-60 (1999)). Furthermore, CTL escape mutations are associated with disease progression (Barouch et al, J. Virol. 77:7367-75 (2003)), thus vaccine-stimulated memory responses that block potential escape routes may be valuable.
[0006] The highly variable Env protein is the primary target for neutralizing antibodies against HIV; since immune protection will likely require both B-cell and T-cell responses (Moore and Burton, Nat. Med. 10:769-71 (2004)), Env vaccine antigens will also need to be optimized separately to elicit antibody responses. T-cell-directed vaccine components, in contrast, can target the more conserved proteins, but even the most conserved HIV-1 proteins are diverse enough that variation is an issue. Artificial central-sequence vaccine approaches (e.g., consensus sequences, in which every amino acid is found in a plurality of sequences, or maximum likelihood reconstructions of ancestral sequences (Gaschen et al, Science 296:2354-60 (2002), Gao et al, J. Virol. 79:1154-63 (2005), Doria-Rose et al, J. Virol. 79:11214-24 (2005), Weaver et al, J. Virol., in press)) are promising; nevertheless, even centralized strains provide limited coverage of HIV-1 variants, and consensus-based reagents fail to detect many autologous T-cell responses (Altfeld et al, J. Virol. 77:7330-40 (2003)).
[0007] Single amino acid changes can allow an epitope to escape T-cell surveillance; since many T-cell epitopes differ between HIV-1 strains at one or more positions, potential responses to any single vaccine antigen are limited. Whether a particular mutation results in escape depends upon the specific epitope/T-cell combination, although some changes broadly affect between-subtype cross-reactivity (Norris et al, AIDS Res. Hum. Retroviruses 20:315-25 (2004)). Including multiple variants in a polyvalent vaccine could enable responses to a broader range of circulating variants, and could also prime the immune system against common escape mutants (Jones et al, J. Exp. Med. 200:1243-56 (2004)). Escape from one T-cell receptor may create a variant that is susceptible to another (Allen et al, J. Virol. 79:12952-60 (2005), Feeney et al, J. Immunol. 174:7524-30 (2005)), so stimulating polyclonal responses to epitope variants may be beneficial (Killian et al, Aids 19:887-96 (2005)). Escape mutations that inhibit processing (Milicic et al, J. Immunol. 175:4618-26 (2005)) or HLA binding (Ammaranond et al, AIDS Res. Hum. Retroviruses 21:395-7 (2005)) cannot be directly countered by a T-cell with a different specificity, but responses to overlapping epitopes may block even some of these escape routes.
[0008] The present invention relates to a polyvalent vaccine comprising several “mosaic” proteins (or genes encoding these proteins). The candidate vaccine antigens can be cocktails of k composite proteins (k being the number of sequence variants in the cocktail), optimized to include the maximum number of potential T-cell epitopes in an input set of viral proteins. The mosaics are generated from natural sequences: they resemble natural proteins and include the most common forms of potential epitopes. Since CD8+ epitopes are contiguous and typically nine amino-acids long, sets of mosaics can be scored by “coverage” of nonamers (9-mers) in the natural sequences (fragments of similar lengths are also well represented). 9-Mers not found at least three times can be excluded. This strategy provides the level of diversity coverage achieved by a massively polyvalent multiple-peptide vaccine but with important advantages: it allows vaccine delivery as intact proteins or genes, excludes low-frequency or unnatural epitopes that are not relevant to circulating strains, and its intact protein antigens are more likely to be processed as in a natural infection.


[0009] In general, the present invention relates to an immunogenic composition. More specifically, the invention relates to a polyvalent immunogenic composition (e.g., an HIV vaccine), and to methods of using same. The invention further relates to methods that involve the use of a genetic algorithm to design sets of polyvalent antigens suitable for use as vaccines.
[0010] Objects and advantages of the present invention will be clear from the description that follows.


[0011] FIGS. 1A-1F. The upper bound of potential epitope coverage of the HIV-1 M group. The upper bound for population coverage of 9-mers for increasing numbers of variants is shown, for k=1-8 variants. A sliding window of length nine was applied across aligned sequences, moving down by one position. Different colors denote results for different numbers of sequences. At each window, the coverage given by the k most common 9-mers is plotted for Gag (FIGS. 1A and 1B), Nef (FIGS. 1C and 1D) and Env gp120 (FIGS. 1E and 1F). Gaps inserted to maintain the alignment are treated as characters. The diminishing returns of adding more variants are evident, since, as k increases, increasingly rare forms are added. In FIGS. 1A, 1C and 1E, the scores for each consecutive 9-mer are plotted in their natural order to show how diversity varies in different protein regions; both p24 in the center of Gag and the central region of Nef are particularly highly conserved. In FIGS. 1B, 1D and 1F, the scores for each 9-mer are reordered by coverage (a strategy also used in FIG. 4), to provide a sense of the overall population coverage of a given protein. Coverage of gp120, even with 8 variant 9-mers, is particularly poor (FIGS. 1E and 1F).
[0012] FIGS. 2A-2C. Mosaic initialization, scoring, and optimization. FIG. 2A) A set of k populations is generated by random 2-point recombination of natural sequences (1-6 populations of 50-500 sequences each have been tested). One sequence from each population is chosen (initially at random) for the mosaic cocktail, which is subsequently optimized. The cocktail sequences are scored by computing coverage (defined as the mean fraction of natural-sequence 9-mers included in the cocktail, averaged over all natural sequences in the input data set). Any new sequence that covers more epitopes will increase the score of the whole cocktail. FIG. 2B) The fitness score of any individual sequence is the coverage of a cocktail containing that sequence plus the current representatives from other populations. FIG. 2C) Optimization: 1) two “parents” are chosen: the higher-scoring of a randomly chosen pair of recombined sequences, and either (with 50% probability) the higher-scoring sequence of a second random pair, or a randomly chosen natural sequence. 2) Two-point recombination between the two parents is used to generate a “child” sequence. If the child contains unnatural or rare 9-mers, it is immediately rejected, otherwise it is scored (Gaschen et al, Science 296:2354-2360 (2002)). If the score is higher than that of any of four randomly-selected population members, the child is inserted in the population in place of the weakest of the four, thus evolving an improved population; 4) if its score is a new high score, the new child replaces the current cocktail member from its population. Ten cycles of child generation are repeated for each population in turn, and the process iterates until improvement stalls.
[0013] FIG. 3. Mosaic strain coverage for all HIV proteins. The level of 9-mer coverage achieved by sets of four mosaic proteins for each HIV protein is shown, with mosaics optimized using either the M group or the C subtype. The fraction of C subtype sequence 9-mers covered by mosaics optimized on the C subtype (within-clade optimization) is shown in gray. Coverage of 9-mers found in non-C subtype M-group sequences by subtype-C-optimized mosaics (between-clade coverage) is shown in white. Coverage of subtype C sequences by M-group optimized mosaics is shown in black. B clade comparisons gave comparable results (data not shown).
[0014] FIGS. 4A-4F. Coverage of M group sequences by different vaccine candidates, nine-mer by nine-mer. Each plot presents site-by-site coverage (i.e., for each nine-mer) of an M-group natural-sequence alignment by a single tri-valent vaccine candidate. Bars along the x-axis represent the proportion of sequences matched by the vaccine candidate for a given alignment position: 9/9 matches (in red), 8/9 (yellow), 7/9 (blue). Aligned 9-mers are sorted along the x-axis by exact-match coverage value. 656 positions include both the complete Gag and the central region of Nef. For each alignment position, the maximum possible matching value (i.e. the proportion of aligned sequences without gaps in that nine-mer) is shown in gray. FIG. 4A) Non-optimal natural sequences selected from among strains being used in vaccine studies (Kong et al, J. Virol. 77:12764-72 (2003)) including an individual clade A, B, and C viral sequences (Gag: GenBank accession numbers AF004885, K03455, and U52953; Nef core: AF069670, K02083, and U52953). FIG. 4B) Optimum set of natural sequences [isolates US2 (subtype B, USA), 70177 (subtype C, India), and 99TH.R2399 (subtype CRF15_01 B, Thailand); accession numbers AY173953, AF533131, and _AF530576] selected by choosing the single sequence with maximum coverage, followed by the sequence that had the best coverage when combined with the first (i.e. the best complement), and so on, selected for M group coverage FIG. 4C) Consensus sequence cocktail (M group, B- and C-subtypes). FIG. 4D) 3 mosaic sequences, FIG. 4E) 4 mosaic sequences, FIG. 4F) 6 mosaic sequences. FIGS. 4D-4F were all optimized for M group coverage.
[0015] FIGS. 5A and 5B. Overall coverage of vaccine candidates: coverage of 9-mers in C clade sequences using different input data sets for mosaic optimization, allowing different numbers of antigens, and comparing to different candidate vaccines. Exact (blue), 8/9 (one-off; red), and 7/9 (two-off; yellow) coverage was computed for mono- and polyvalent vaccine candidates for Gag (FIG. 5A) and Nef (core) (FIG. 5B) for four test situations: within-clade (C-clade-optimized candidates scored for C-clade coverage), between-clade (B-clade-optimized candidates scored for C-clade coverage), global-against-single-subtype (M-group-optimized candidates scored for C-clade coverage), global-against-global (M-group-optimized candidates scored for global coverage). Within each set of results, vaccine candidates are grouped by number of sequences in the cocktail (1-6); mosaic sequences are plotted with darker colors. “Non-opt” refers to one set of sequences moving into vaccine trials (Kong et al, J. Virol. 77:12764-72 (2003)); “mosaic” denotes sequences generated by the genetic algorithm; “opt. natural” denotes intact natural sequences selected for maximum 9-mer coverage; “MBC consensus” denotes a cocktail of 3 consensus sequences, for M-group, B-subtype, and C-subtype. For ease of comparison, a dashed line marks the coverage of a 4-sequence set of M-group mosaics (73.7-75.6%). Over 150 combinations of mosaic-number, virus subset, protein region, and optimization and test sets were tested. The C clade/B clade/M group comparisons illustrated in this figure are generally representative of within-clade, between-clade, and M group, coverage. In particular, levels of mosaic coverage for B and C clade were very similar, despite there being many more C clade sequences in the Gag collection, and many more B clade sequences in the Nef collection (see FIG. 6 for a full B and C clade comparison). There were relatively few A and G clade sequences in the alignments (24 Gag, 75 Nef), and while 9-mer coverage by M-group optimized mosaics was not as high as for subtypes for B and C clades (4-mosaic coverage for A and G subtypes was 63% for Gag, 74% for Nef), it was much better than a non-optimal cocktail (52% Gag, 52% for Nef).
[0016] FIGS. 6A and 6B. Overall coverage of vaccine candidates: coverage of 9-mers in B-clade, C-clade, and M-group sequences using different input data sets for mosaic optimization, allowing different numbers of antigens, and comparing to different candidate vaccines. Exact (blue), 8/9 (one-off; red), and 7/9 (two-off; yellow) coverage was computed for mono- and polyvalent vaccine candidates for Gag (FIG. 6A) and Nef (core) (FIG. 6B) for seven test situations: within-clade (B- or C-clade-optimized candidates scored against the same clade), between-clade (B- or C-clade-optimized candidates scored against the other clade), global vaccine against single subtype (M-group-optimized candidates scored against B- or C-clade), global vaccine against global viruses (M-group-optimized candidates scored against all M-group sequences). Within each set of results, vaccine candidates are grouped by number of sequences in the cocktail (1-6); mosaic sequences are plotted with darker colors. “Non-opt” refers to a particular set of natural sequences previously proposed for a vaccine (Kong, W. P. et al. J Virol 77, 12764-72 (2003)); “mosaic” denotes sequences generated by the genetic algorithm; “opt. natural” denotes intact natural sequences selected for maximum 9-mer coverage; “MBC consensus” denotes a cocktail of 3 consensus sequences, for M-group, B-subtype, and C-subtype. A dashed line is shown at the level of exact-match M-group coverage for a 4-valent mosaic set optimized on the M-group.
[0017] FIGS. 7A and 7B. The distribution of 9-mers by frequency of occurrence in natural, consensus, and mosaic sequences. Occurrence counts (y-axis) for different 9-mer frequencies (x-axis) for vaccine cocktails produced by several methods. FIG. 7A: frequencies from 0-60% (for 9-mer frequencies>60%, the distributions are equivalent for all methods). FIG. 7B: Details of low-frequency 9-mers. Natural sequences have large numbers of rare or unique-to-isolate 9-mers (bottom right, FIGS. 7A and 7B); these are unlikely to induce useful vaccine responses. Selecting optimal natural sequences does select for more common 9-mers, but rare and unique 9-mers are still included (top right, FIGS. 7A and 7B). Consensus cocktails, in contrast, under-represent uncommon 9-mers, especially below 20% frequency (bottom left, FIGS. 7A and 7B). For mosaic sequences, the number of lower-frequency 9-mers monotonically increases with the number of sequences (top left, each panel), but unique-to-isolate 9-mers are completely excluded (top left of right panel: * marks the absence of 9-mers with frequencies<0.005).
[0018] FIGS. 8A-8D. HLA binding potential of vaccine candidates. FIGS. 8A and 8B) HLA binding motif counts. FIGS. 8C and 8D) number of unfavorable amino acids. In all graphs: natural sequences are marked with black circles (λ); consensus sequences with blue triangles (σ); inferred ancestral sequences with green squares (ν); and mosaic sequences with red diamonds (♥). Left panel (FIGS. 8A and 8C) shows HLA-binding-motif counts (FIG. 8A) and counts of unfavorable amino acids (FIG. 8C) calculated for individual sequences; Right panel (FIGS. 8B and 8D) shows HLA binding motifs counts (FIG. 8B) and counts of unfavorable amino acids (FIG. 8D) calculated for sequence cocktails. The top portion of each graph (box-and-whiskers graph) shows the distribution of respective counts (motif counts or counts of unfavorable amino acids) based either on alignment of M group sequences (for individual sequences, FIGS. 8A and 8C) or on 100 randomly composed cocktails of three sequences, one from each A, B and C subtypes (for sequence cocktails, FIGS. 8B and 8D). The alignment was downloaded from the Los Alamos HIV database. The box extends from the 25 percentile to the 75 percentile, with the line at the median. The whiskers extending outside the box show the highest and lowest values. Amino acids that are very rarely found as C-terminal anchor residues are G, S, T, P, N, Q, D, E, and H, and tend to be small, polar, or negatively charged (Yusim et al, J. Virol. 76:8757-8768 (2002)). Results are shown for Gag, but the same qualitative results hold for Nef core and complete Nef. The same procedure was done for supertype motifs with results qualitatively similar to the results for HLA binding motifs (data not shown).
[0019] FIG. 9. Mosaic protein sets limited to 4 sequences (k=4), spanning Gag and the central region of Net optimized for subtype B, subtype C, and the M group.
[0020] FIG. 10. Mosaic sets for Env and Pol.
[0021] FIG. 11. This plot is alignment independent, based on splintering all M group proteins, (database and CHAVI, one sequence per person) into all possible 9-mers, attending to their frequencies, and then looking for matches and near matches in each vaccine antigen or cocktail with the database.
[0022] FIG. 12. Additional summaries of coverage.
[0023] FIG. 13. 9-mer coverage by position (Mos.3 vaccine cocktail).
[0024] FIGS. 14A-14D. Plots resorted by frequency of 9-mer matches for each vaccine proposed for use.
[0025] FIGS. 15A-15D. Plots mapping every amino acid in every sequence in the full database alignment.
[0026] FIG. 16. 3 Mosaic, M group Optimizations.
[0027] FIG. 17. Coverage of the HIV database plus CHAVI sequences (N=2020).
[0028] FIG. 18. Differences in acute infection patient sequences compared to patient consensus.
[0029] FIG. 19. The compromise and benefit in terms of coverage for Env M group versus subtype-specific design.
[0030] FIG. 20. Proposed vaccine mosaic coverage of Gag and Env.
[0031] FIG. 21. Gag, Nef and Env sequences.
[0032] FIG. 22. Mosaic gag and nef genes and M consensus gag and nef genes.


[0033] The present invention results from the realization that a polyvalent set of antigens comprising synthetic viral proteins, the sequences of which provide maximum coverage of non-rare short stretches of circulating viral sequences, constitutes a good vaccine candidate. The invention provides a “genetic algorithm” strategy to create such sets of polyvalent antigens as mosaic blends of fragments of an arbitrary set of natural protein sequences provided as inputs. In the context of HIV, the proteins Gag and Nef are ideal candidates for such antigens. To expand coverage, Pol and/or Env can also be used. The invention further provides optimized sets for these proteins.
[0034] The genetic algorithm strategy of the invention uses unaligned protein sequences from the general population as an input data set, and thus has the virtue of being “alignment independent”. It creates artificial mosaic proteins that resemble proteins found in nature—the success of the consensus antigens in small animals models suggest this works well. 9 Mers are the focus of the studies described herein, however, different length peptides can be selected depending on the intended target. In accordance with the present approach, 9 mers (for example) that do not exist in nature or that are very rare can be excluded—this is an improvement relative to consensus sequences since the latter can contain some 9 mers (for example) that have not been found in nature, and relative to natural strains that almost invariably contain some 9 mers (for example) that are unique to that strain. The definition of fitness used for the genetic algorithm is that the most “fit” polyvalent cocktail is the combination of mosaic strains that gives the best coverage (highest fraction of perfect matches) of all of the 9 mers in the population and is subject to the constraint that no 9 mer is absent or rare in the population.
[0035] The mosaics protein sets of the invention can be optimized with respect to different input data sets—this allows use of current data to assess virtues of a subtype or region specific vaccines from a T cell perspective. By way of example, options that have been compared include:
[0036] 1) Optimal polyvalent mosaic sets based on M group, B clade and C clade. The question presented was how much better is intra-clade coverage than inter-clade or global.
[0037] 2) Different numbers of antigens: 1, 3, 4, 6
[0038] 3) Natural strains currently in use for vaccine protocols just to exemplify “typical” strains (Merck, VRC)
[0039] 4) Natural strains selected to give the best coverage of 9-mers in a population
[0040] 5) Sets of consensus: A+B+C . . .
[0041] 6) Optimized cocktails that include one “given” strain in a polyvalent antigen, one ancestral+3 mosaic strains, one consensus+3 mosaic strains.
[0042] 7) Coverage of 9 mers that were perfectly matched was compared with those that match 8/9, 7/9, and 6/9 or less.
This is a computationally difficult problem, as the best set to cover one 9-mer may not be the best set to cover overlapping 9-mers.
[0043] It will be appreciated from a reading of this disclosure that the approach described herein can be used to design peptide reagents to test HIV immune responses, and be applied to other variable pathogens as well. For example, the present approach can be adapted to the highly variable virus Hepatitis C.
[0044] The proteins/polypeptides/peptides (“immunogens”) of the invention can be formulated into compositions with a pharmaceutically acceptable carrier and/or adjuvant using techniques well known in the art. Suitable routes of administration include systemic (e.g. intramuscular or subcutaneous), oral, intravaginal, intrarectal and intranasal.
[0045] The immunogens of the invention can be chemically synthesized and purified using methods which are well known to the ordinarily skilled artisan. The immunogens can also be synthesized by well-known recombinant DNA techniques.
[0046] Nucleic acids encoding the immunogens of the invention can be used as components of, for example, a DNA vaccine wherein the encoding sequence is administered as naked DNA or, for example, a minigene encoding the immunogen can be present in a viral vector. The encoding sequences can be expressed, for example, in mycobacterium, in a recombinant chimeric adenovirus, or in a recombinant attenuated vesicular stomatitis virus. The encoding sequence can also be present, for example, in a replicating or non-replicating adenoviral vector, an adeno-associated virus vector, an attenuated mycobacterium tuberculosis vector, a Bacillus Calmette Guerin (BCG) vector, a vaccinia or Modified Vaccinia Ankara (MVA) vector, another pox virus vector, recombinant polio and other enteric virus vector, Salmonella species bacterial vector, Shigella species bacterial vector, Venezuelean Equine Encephalitis Virus (VEE) vector, a Semliki Forest Virus vector, or a Tobacco Mosaic Virus vector. The encoding sequence, can also be expressed as a DNA plasmid with, for example, an active promoter such as a CMV promoter. Other live vectors can also be used to express the sequences of the invention. Expression of the immunogen of the invention can be induced in a patient's own cells, by introduction into those cells of nucleic acids that encode the immunogen, preferably using codons and promoters that optimize expression in human cells. Examples of methods of making and using DNA vaccines are disclosed in U.S. Pat. Nos. 5,580,859, 5,589,466, and 5,703,055. Examples of methods of codon optimization are described in Haas et al, Current Biology 6:315-324 (1996) and in Andre et al, J. Virol. 72(2):1497-1503 (1998).
[0047] It will be appreciated that adjuvants can be included in the compositions of the invention (or otherwise administered to enhance the immunogenic effect). Examples of suitable adjuvants include TRL-9 agonists, TRL-4 agonists, and TRL-7, 8 and 9 agonist combinations (as well as alum). Adjuvants can take the form of oil and water emulsions. Squalene adjuvants can also be used.
[0048] The composition of the invention comprises an immunologically effective amount of the immunogen of this invention, or nucleic acid sequence encoding same, in a pharmaceutically acceptable delivery system. The compositions can be used for prevention and/or treatment of virus infection (e.g. HIV infection). As indicated above, the compositions of the invention can be formulated using adjuvants, emulsifiers, pharmaceutically-acceptable carriers or other ingredients routinely provided in vaccine compositions. Optimum formulations can be readily designed by one of ordinary skill in the art and can include formulations for immediate release and/or for sustained release, and for induction of systemic immunity and/or induction of localized mucosal immunity (e.g, the formulation can be designed for intranasal, intravaginal or intrarectal administration). As noted above, the present compositions can be administered by any convenient route including subcutaneous, intranasal, oral, intramuscular, or other parenteral or enteral route. The immunogens can be administered as a single dose or multiple doses. Optimum immunization schedules can be readily determined by the ordinarily skilled artisan and can vary with the patient, the composition and the effect sought.
[0049] The invention contemplates the direct use of both the immunogen of the invention and/or nucleic acids encoding same and/or the immunogen expressed as indicated above. For example, a minigene encoding the immunogen can be used as a prime and/or boost.
[0050] The invention includes any and all amino acid sequences disclosed herein, as well as nucleic acid sequences encoding same (and nucleic acids complementary to such encoding sequences).
[0051] Specifically disclosed herein are vaccine antigen sets optimized for single B or C subtypes, targeting regional epidemics, as well as for all HIV-1 variants in global circulation [the HIV-1 Main (M) group]. In the study described in Example 1 that follows, the focus is on designing polyvalent vaccines specifically for T-cell responses. HIV-1 specific T-cells are likely to be crucial to an HIV-1-specific vaccine response: CTL responses are correlated with slow disease progression in humans (Oxenius et al, J. Infect. Dis. 189:1199-1208 (2004)), and the importance of CTL responses in nonhuman primate vaccination models is well-established. Vaccine elicited cellular immune responses help control pathogenic SIV or SHIV, and reduce the likelihood of disease after challenge with pathogenic virus (Barouch et al, Science 290:486-492 (2000)). Temporary depletion of CD8+ T cells results in increased viremia in SIV-infected rhesus macaques (Schmitz et al, Science 283:857-860 (1999)). Furthermore, the evolution of escape mutations has been associated with disease progression, indicating that CTL responses help constrain viral replication in vivo (Barouch et al, J. Virol. 77:7367-7375 (2003)), and so vaccine-stimulated memory responses that could block potential escape routes may be of value. While the highly variable Envelope (Env) is the primary target for neutralizing antibodies against HIV, and vaccine antigens will also need to be tailored to elicit these antibody responses (Moore & Burton, Nat. Med. 10:769-771 (2004)), T-cell vaccine components can target more conserved proteins to trigger responses that are more likely to cross-react. But even the most conserved HIV-1 proteins are diverse enough that variation will be an issue. Artificial central-sequence vaccine approaches, consensus and ancestral sequences (Gaschen et al, Science 296:2354-2360 (2002), Gao et al, J. Virol. 79:1154-1163 (2005), Doria-Rose et al, J. Virol. 79:11214-11224 (2005)), which essentially “split the differences” between strains, show promise, stimulating responses with enhanced cross-reactivity compared to natural strain vaccines (Gao et al, J. Virol. 79:1154-1163 (2005)) (Liao et al. and Weaver et al., submitted.) Nevertheless, even central strains cover the spectrum of HIV diversity to a very limited extent, and consensus-based peptide reagents fail to detect many autologous CD8+ T-cell responses (Altfeld et al, J. Virol. 77:7330-7340 (2003)).
[0052] A single amino acid substitution can mediate T-cell escape, and as one or more amino acids in many T-cell epitopes differ between HIV-1 strains, the potential effectiveness of responses to any one vaccine antigen is limited. Whether a particular mutation will diminish T-cell cross-reactivity is epitope- and T-cell-specific, although some changes can broadly affect between-clade cross-reactivity (Norris et al, AIDS Res. Hum. Retroviruses 20:315-325 (2004)). Including more variants in a polyvalent vaccine could enable responses to a broader range of circulating variants. It could also prime the immune system against common escape variants (Jones et al, J. Exp. Med. 200:1243-1256 (2004)); escape from one T-cell receptor might create a variant that is susceptible to another (Lee et al, J. Exp. Med. 200:1455-1466 (2004)), thus stimulating polyclonal responses to epitope variants may be beneficial (Killian et al, AIDS 19:887-896 (2005)). Immune escape involving avenues that inhibit processing (Milicic et al, J. Immunol. 175:4618-4626 (2005)) or HLA binding (Ammaranond et al, AIDS Res. Hum. Retroviruses 21:395-397 (2005)) prevent epitope presentation, and in such cases the escape variant could not be countered by a T-cell with a different specificity. However, it is possible the presence of T-cells that recognize overlapping epitopes may in some cases block these even escape routes.
[0053] Certain aspects of the invention can be described in greater detail in the non-limiting Examples that follow.

Example 1

Experimental Details

[0054] HIV-1 Sequence Data.
[0055] The reference alignments from the 2005 HIV sequence database (, which contain one sequence per person, were used, supplemented by additional recently available C subtype Gag and Nef sequences from Durban, South Africa (GenBank accession numbers AY856956-AY857186) (Kiepiela et al, Nature 432:769-75 (2004)). This set contained 551 Gag and 1,131 Nef M group sequences from throughout the globe; recombinant sequences were included as well as pure subtype sequences for exploring M group diversity. The subsets of these alignments that contained 18 A, 102 B, 228 C, and 6 G subtype (Gag), and 62 A, 454 B, 284 C, and 13 G subtype sequences (Nef) sequences were used for within- and between-single-clade optimizations and comparisons.
[0056] The Genetic Algorithm.
[0057] GAs are computational analogues of biological processes (evolution, populations, selection, recombination) used to find solutions to problems that are difficult to solve analytically (Holland, Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence, (M.I.T. Press, Cambridge, Mass. (1992))). Solutions for a given input are “evolved” though a process of random modification and selection according to a “fitness” (optimality) criterion. GAs come in many flavors; a “steady-state co-evolutionary multi-population” GA was implemented. “Steady-state” refers to generating one new candidate solution at a time, rather than a whole new population at once; and “co-evolutionary” refers to simultaneously evolving several distinct populations that work together to form a complete solution. The input is an unaligned set of natural sequences; a candidate solution is a set of k pseudo-natural “mosaic” sequences, each of which is formed by concatenating sections of natural sequences. The fitness criterion is population coverage, defined as the proportion of all 9-amino-acid sequence fragments (potential epitopes) in the input sequences that are found in the cocktail.
[0058] To initialize the GA (FIG. 2), k populations of n initial candidate sequences are generated by 2-point recombination between randomly selected natural sequences. Because the input natural sequences are not aligned, “homologous” crossover is used: crossover points in each sequence are selected by searching for short matching strings in both sequences; strings of c−1=8, were used where a typical epitope length is c=9. This ensures that the recombined sequences resemble natural proteins: the boundaries between sections of sequence derived from different strains are seamless, the local sequences spanning the boundaries are always found in nature, and the mosaics are prevented from acquiring large insertions/deletions or unnatural combinations of amino acids. Mosaic sequence lengths fall within the distribution of natural sequence lengths as a consequence of mosaic construction: recombination is only allowed at identical regions, reinforced by an explicit software prohibition against excessive lengths to prevent reduplication of repeat regions. (Such “in frame” insertion of reduplicated epitopes could provide another way of increasing coverage without generating unnatural 9-mers, but their inclusion would create “unnatural” proteins.) Initially, the cocktail contains one randomly chosen “winner” from each population. The fitness score for any individual sequence in a population is the coverage value for the cocktail consisting of that sequence plus the current winners from the other populations. The individual fitness of any sequence in a population therefore depends dynamically upon the best sequences found in the other populations.
[0059] Optimization proceeds one population at a time. For each iteration, two “parent” sequences are chosen. The first parent is chosen using “2-tournament” selection: two sequences are picked at random from the current population, scored, and the better one is chosen. This selects parents with a probability inversely proportional to their fitness rank within the population, without the need to actually compute the fitness of all individuals. The second parent is chosen in the same way (50% of the time), or is selected at random from the set of natural sequences. 2-point homologous crossover between the parents is then used to generate a “child” sequence. Any child containing a 9-mer that was very rare in the natural population (found less than 3 times) is rejected immediately. Otherwise, the new sequence is scored, and its fitness is compared with the fitnesses of four randomly chosen sequences from the same population. If any of the four randomly chosen sequences has a score lower than that of the new sequence, it is replaced in the population by the new sequence. Whenever a sequence is encountered that yields a better score than the current population “winner”, that sequence becomes the winner for the current population and so is subsequently used in the cocktail to evaluate sequences in other populations. A few such optimization cycles (typically 10) are applied to each population in turn, and this process continues cycling through the populations until evolution stalls (i.e., no improvement has been made for a defined number of generations). At this point, the entire procedure is restarted using newly generated random starting populations, and the restarts are continued until no further improvement is seen. The GA was run on each data set with n=50 or 500; each run was continued until no further improvement occurred for 12-24 hours on a 2 GHz Pentium processor. Cocktails were generated having k=1, 3, 4, or 6 mosaic sequences.
[0060] The GA also enables optional inclusion of one or more fixed sequences of interest (for example, a consensus) in the cocktail and will evolve the other elements of the cocktail in order to optimally complement that fixed strain. As these solutions were suboptimal, they are not included here. An additional program selects from the input file the k best natural strains that in combination provide the best population coverage.
[0061] Comparison with Other Polyvalent Vaccine Candidates.
[0062] Population coverage scores were computed for other potential mono- or polyvalent vaccines to make direct comparisons with the mosaic-sequence vaccines, tracking identities with population 9-mers, as well as similarities of 8/9 and 7/9 amino acids. Potential vaccine candidates based on natural strains include single strains (for example, a single C strain for a vaccine for southern Africa (Williamson et al, AIDS Res. Hum. Retroviruses 19:133-44 (2003))) or combinations of natural strains (for example, one each of subtype A, B, and C (Kong et al, J. Virol. 77:12764-72 (2003)). To date, natural-strain vaccine candidates have not been systematically selected to maximize potential T-cell epitope coverage; vaccine candidates were picked from the literature to be representative of what could be expected from unselected vaccine candidates. An upper bound for coverage was also determined using only intact natural strains: optimal natural-sequence cocktails were generated by selecting the single sequence with the best coverage of the dataset, and then successively adding the most complementary sequences up to a given k. The comparisons included optimal natural-sequence cocktails of various sizes, as well as consensus sequences, alone or in combination (Gaschen et al, Science 296:2354-60 (2002)), to represent the concept of central, synthetic vaccines. Finally, using the fixed-sequence option in the GA, consensus-plus-mosaic combinations in the comparisons; these scores were essentially equivalent to all-mosaic combinations were included for a given k (data not shown). The code used for performing these analyses are available at: ftp://ftp-t10/pub/btk/mosaics.


[0063] Protein Variation.
[0064] In conserved HIV-1 proteins, most positions are essentially invariant, and most variable positions have only two to three amino acids that occur at appreciable frequencies, and variable positions are generally well dispersed between conserved positions. Therefore, within the boundaries of a CD8+ T-cell epitope (8-12 amino acids, typically nine), most of the population diversity can be covered with very few variants. FIG. 1 shows an upper bound for population coverage of 9-mers (stretches of nine contiguous amino acids) comparing Gag, Nef, and Env for increasing numbers of variants, sequentially adding variants that provide the best coverage. In conserved regions, a high degree of population coverage is achieved with 2-4 variants. By contrast, in variable regions like Env, limited population coverage is possible even with eight variants. Since each new addition is rarer, the relative benefits of each addition diminish as the number of variants increases.
[0065] Vaccine Design Optimization Strategies.
[0066] FIG. 1 shows an idealized level of 9-mer coverage. In reality, high-frequency 9-mers often conflict: because of local co-variation, the optimal amino acid for one 9-mer may differ from that for an overlapping 9-mer. To design mosaic protein sets that optimize population coverage, the relative benefits of each amino acid must be evaluated in combination with nearby variants. For example, Alanine (Ala) and Glutamate (Glu) might each frequently occur in adjacent positions, but if the Ala-Glu combination is never observed in nature, it should be excluded from the vaccine. Several optimization strategies were investigated: a greedy algorithm, a semi-automated compatible-9 mer assembly strategy, an alignment-based genetic algorithm (GA), and an alignment-independent GA.
[0067] The alignment-independent GA generated mosaics with the best population coverage. This GA generates a user-specified number of mosaic sequences from a set of unaligned protein sequences, explicitly excluding rare or unnatural epitope-length fragments (potentially introduced at recombination breakpoints) that could induce non-protective vaccine-antigen-specific responses. These candidate vaccine sequences resemble natural proteins, but are assembled from frequency-weighted fragments of database sequences recombined at homologous breakpoints (FIG. 2); they approach maximal coverage of 9-mers for the input population.
[0068] Selecting HIV Protein Regions for an Initial Mosaic Vaccine.
[0069] The initial design focused on protein regions meeting specific criteria: i) relatively low variability, ii) high levels of recognition in natural infection, iii) a high density of known epitopes and iv) either early responses upon infection or CD8+ T-cell responses associated with good outcomes in infected patients. First, an assessment was made of the level of 9-mer coverage achieved by mosaics for different HIV proteins (FIG. 3). For each protein, a set of four mosaics was generated using either the M group or the B- and C-subtypes alone; coverage was scored on the C subtype. Several results are notable: i) within-subtype optimization provides the best within-subtype coverage, but substantially poorer between-subtype coverage—nevertheless, B-subtype-optimized mosaics provide better C-subtype coverage than a single natural B subtype protein (Kong et al, J. Virol. 77:12764-72 (2003)); ii) Pol and Gag have the most potential to elicit broadly cross-reactive responses, whereas Rev, Tat, and Vpu have even fewer conserved 9-mers than the highly variable Env protein, iii) within-subtype coverage of M-group-optimized mosaic sets approached coverage of within-subtype optimized sets, particularly for more conserved proteins.
[0070] Gag and the central region of Nef meet the four criteria listed above. Nef is the HIV protein most frequently recognized by T-cells (Frahm et al, J. Virol. 78:2187-200 (2004)) and the target for the earliest response in natural infection (Lichterfeld et al, Aids 18:1383-92 (2004)). While overall it is variable (FIG. 3), its central region is as conserved as Gag (FIG. 1). It is not yet clear what optimum proteins for inclusion in a vaccine might be, and mosaics could be designed to maximize the potential coverage of even the most variable proteins (FIG. 3), but the prospects for global coverage are better for conserved proteins. Improved vaccine protection in macaques has been demonstrated by adding Rev, Tat, and Nef to a vaccine containing Gag, Pol, and Env (Hel et al, J. Immunol. 176:85-96 (2006)), but this was in the context of homologous challenge, where variability was not an issue. The extreme variability of regulatory proteins in circulating virus populations may preclude cross-reactive responses; in terms of conservation, Pol, Gag (particularly p24) and the central region of Nef (HXB2 positions 65-149) are promising potential immunogens (FIGS. 1,3). Pol, however, is infrequently recognized during natural infection (Frahm et al, J. Virol. 78:2187-200 (2004)), so it was not included in the initial immunogen design. The conserved portion of Nef that were included contains the most highly recognized peptides in HIV-1 (Frahm et al, J. Virol. 78:2187-200 (2004)), but as a protein fragment, would not allow Nef's immune inhibitory functions (e.g. HLA class I down-regulation (Blagoveshchenskaya, Cell 111:853-66 (2002))). Both Gag and Nef are densely packed with overlapping well-characterized CD8+ and CD4+ T-cell epitopes, presented by many different HLA molecules (, and Gag-specific CD8+ (Masemola et al, J. Virol. 78:3233-43 (2004)) and CD4+ (Oxenius et al, J. Infect. Dis. 189:1199-208 (2004)) T-cell responses have been associated with low viral set points in infected individuals (Masemola et al, J. Virol. 78:3233-43 (2004)).
[0071] To examine the potential impact of geographic variation and input sample size, a limited test was done using published subtype C sequences. The subtype C Gag data were divided into three sets of comparable size—two South African sets (Kiepiela et al, Nature 432:769-75 (2004)), and one non-South-African subtype C set. Mosaics were optimized independently on each of the sets, and the resulting mosaics were tested against all three sets. The coverage of 9-mers was slightly better for identical training and test sets (77-79% 9/9 coverage), but essentially equivalent when the training and test sets were the two different South African data sets (73-75%), or either of the South African sets and the non-South African C subtype sequences (74-76%). Thus between- and within-country coverage approximated within-clade coverage, and in this case no advantage to a country-specific C subtype mosaic design was found.
[0072] Designing Mosaics for Gag and Nef and Comparing Vaccine Strategies.
[0073] To evaluate within- and between-subtype cross-reactivity for various vaccine design strategies, a calculation was made of the coverage they provided for natural M-Group sequences. The fraction of all 9-mers in the natural sequences that were perfectly matched by 9-mers in the vaccine antigens were computed, as well as those having 8/9 or 7/9 matching amino acids, since single (and sometimes double) substitutions within epitopes may retain cross-reactivity. FIG. 4 shows M group coverage per 9-mer in Gag and the central region of Nef for cocktails designed by various strategies: a) three non-optimal natural strains from the A, B, and C subtypes that have been used as vaccine antigens (Kong et al, J. Virol. 77:12764-72 (2003)); b) three natural strains that were computationally selected to give the best M group coverage; c) M group, B subtype, and C subtype consensus sequences; and, d, e, f) three, four and six mosaic proteins. For cocktails of multiple strains, sets of k=3, k=4, and k=6, the mosaics clearly perform the best, and coverage approaches the upper bound for k strains. They are followed by optimally selected natural strains, the consensus protein cocktail, and finally, non-optimal natural strains. Allowing more antigens provides greater coverage, but gains for each addition are reduced as k increases (FIGS. 1 and 4).
[0074] FIG. 5 summarizes total coverage for the different vaccine design strategies, from single proteins through combinations of mosaic proteins, and compares within-subtype optimization to M group optimization. The performance of a single mosaic is comparable to the best single natural strain or a consensus sequence. Although a single consensus sequence out-performs a single best natural strain, the optimized natural-sequence cocktail does better than the consensus cocktail: the consensus sequences are more similar to each other than are natural strains, and are therefore somewhat redundant. Including even just two mosaic variants, however, markedly increases coverage, and four and six mosaic proteins give progressively better coverage than polyvalent cocktails of natural or consensus strains. Within-subtype optimized mosaics perform best—with four mosaic antigens 80-85% of the 9-mers are perfectly matched—but between-subtype coverage of these sets falls off dramatically, to 50-60%. In contrast, mosaic proteins optimized using the full M group give coverage of approximately 75-80% for individual subtypes, comparable to the coverage of the M group as a whole (FIGS. 5 and 6). If imperfect 8/9 matches are allowed, both M group optimized and within-subtype optimized mosaics approach 90% coverage.
[0075] Since coverage is increased by adding progressively rarer 9-mers, and rare epitopes may be problematic (e.g., by inducing vaccine-specific immunodominant responses), an investigation was made of the frequency distribution of 9-mers in the vaccine constructs relative to the natural sequences from which they were generated. Most additional epitopes in a k=6 cocktail compared to a k=4 cocktail are low-frequency (<0.1, FIG. 7). Despite enhancing coverage, these epitopes are relatively rare, and thus responses they induce might draw away from vaccine responses to more common, thus more useful, epitopes. Natural-sequence cocktails actually have fewer occurrences of moderately low-frequency epitopes than mosaics, which accrue some lower frequency 9-mers as coverage is optimized. On the other hand, the mosaics exclude unique or very rare 9-mers, while natural strains generally contain 9-mers present in no other sequence. For example, natural M group Gag sequences had a median of 35 (range 0-148) unique 9-mers per sequence. Retention of HLA-anchor motifs was also explored, and anchor motif frequencies were found to be comparable between four mosaics and three natural strains. Natural antigens did exhibit an increase in number of motifs per antigen, possibly due to inclusion of strain-specific motifs (FIG. 8).
[0076] The increase in ever-rarer epitopes with increasing k, coupled with concerns about vaccination-point dilution and reagent development costs, resulted in the initial production of mosaic protein sets limited to 4 sequences (k=4), spanning Gag and the central region of Nef, optimized for subtype B; subtype C, and the M group (these sequences are included in FIG. 9; mosaic sets for Env and Pol are set forth in FIG. 10). Synthesis of various four-sequence Gag-Nef mosaics and initial antigenicity studies are underway. In the initial mosaic vaccine, targeted are just Gag and the center of the Nef protein, which are conserved enough to provide excellent global population coverage, and have the desirable properties described above in terms of natural responses (Bansal et al, Aids 19:241-50 (2005)). Additionally, including B subtype p24 variants in Elispot peptide mixtures to detect natural CTL responses to infection significantly enhanced both the number and the magnitude of responses detected supporting the idea that including variants of even the most conserved proteins will be useful. Finally, cocktails of proteins in a polyvalent HIV-1 vaccine given to rhesus macaques did not interfere with the development of robust responses to each antigen (Seaman et al, J. Virol. 79:2956-63 (2005)), and antigen cocktails did not produce antagonistic responses in murine models (Singh et al, J. Immunol. 169:6779-86 (2002)), indicating that antigenic mixtures are appropriate for T-cell vaccines.
[0077] Even with mosaics, variable proteins like Env have limited coverage of 9-mers, although mosaics improve coverage relative to natural strains. For example three M group natural proteins, one each selected from the A, B, and C clades, and currently under study for vaccine design (Seaman et al, J. Virol. 79:2956-63 (2005)) perfectly match only 39% of the 9-mers in M group proteins, and 65% have at least 8/9 matches. In contrast, three M group Env mosaics match 47% of 9-mers perfectly, and 70% have at least an 8/9 match. The code written to design polyvalent mosaic antigens is available, and could readily be applied to any input set of variable proteins, optimized for any desired number of antigens. The code also allows selection of optimal combinations of k natural strains, enabling rational selection of natural antigens for polyvalent vaccines. Included in Table 1 are the best natural strains for Gag and Nef population coverage of current database alignments.
[00001] [TABLE-US-00001]
  Natural sequence cocktails having the best available 9-mer coverage
  for different genes, subtype sets, and numbers of sequences
      Gag, B-subtype, 1 natural sequence
    Gag, B-subtype, 3 natural sequences
    Gag, B-subtype, 4 natural sequences
    Gag, B-subtype, 6 natural sequences
    Gag, C-subtype, 1 natural sequence
    Gag, C-subtype, 3 natural sequences
    Gag, C-subtype, 4 natural sequences
    Gag, C-subtype, 6 natural sequences
    Gag, M-group, 1 natural sequence
    Gag, M-group, 3 natural sequences
    Gag, M-group, 4 natural sequences
    Gag, M-group, 6 natural sequences
    Nef (central region), B-subtype, 1 natural sequence
    Nef (central region), B-subtype, 3 natural sequences
    Nef (central region), B-subtype, 4 natural sequences
    Nef (central region), B-subtype, 6 natural sequences
    Nef (central region), C-subtype, 1 natural sequence
    Nef (central region), C-subtype, 3 natural sequences
    Nef (central region), C-subtype, 4 natural sequences
    Nef (central region), C-subtype, 6 natural sequences
    Nef (central region), M-group, 1 natural sequence
    Nef (central region), M-group, 3 natural sequences
    Nef (central region), M-group, 4 natural sequences
    Nef (central region), M-group, 6 natural sequences
[0079] Summarizing, the above-described study focuses on the design of T-cell vaccine components to counter HIV diversity at the moment of infection, and to block viral escape routes and thereby minimize disease progression in infected individuals. The polyvalent mosaic protein strategy developed here for HIV-1 vaccine design could be applied to any variable protein, to other pathogens, and to other immunological problems. For example, incorporating a minimal number of variant peptides into T-cell response assays could markedly increase sensitivity without excessive cost: a set of k mosaic proteins provides the maximum coverage possible for k antigens.
[0080] A centralized (consensus or ancestral) gene and protein strategy has been proposed previously to address HIV diversity (Gaschen et al, Science 296:2354-2360 (2002)). Proof-of-concept for the use of artificial genes as immunogens has been demonstrated by the induction of both T and B cell responses to wild-type HIV-1 strains by group M consensus immunogens (Gaschen et al, Science 296:2354-2360 (2002), Gao et al, J. Virol. 79:1154-63 (2005), Doria-Rose et al, J. Virol. 79:11214-24 (2005), Weaver et al, J. Virol., in press)). The mosaic protein design improves on consensus or natural immunogen design by co-optimizing reagents for a polyclonal vaccine, excluding rare CD8+ T-cell epitopes, and incorporating variants that, by virtue of their frequency at the population level, are likely to be involved in escape pathways.
[0081] The mosaic antigens maximize the number of epitope-length variants that are present in a small, practical number of vaccine antigens. The decision was made to use multiple antigens that resemble native proteins, rather than linking sets of concatenated epitopes in a poly-epitope pseudo-protein (Hanke et al, Vaccine 16:426-35 (1998)), reasoning that in vivo processing of native-like vaccine antigens will more closely resemble processing in natural infection, and will also allow expanded coverage of overlapping epitopes. T-cell mosaic antigens would be best employed in the context of a strong polyvalent immune response; improvements in other areas of vaccine design and a combination of the best strategies, incorporating mosaic antigens to cover diversity, may ultimately enable an effective cross-reactive vaccine-induced immune response against HIV-1.

Example 2

[0082] Group M consensus envelope and trivalent mosaic envelopes (both of which were designed by in silico modeling and are predicted to be superior than wildtype envelopes) will be compared to a monovalent wild-type envelope and trivalent wild-type transmitted envelopes in a 4 arm immunogenicity clinical trial. The mosaic antigens have been designed based on the current Los Alamos database, a set that includes more full length envelopes sampled globally from more than 2000 individuals with a large set of sequences of transmitted viruses primarily from the CHAVI database.
[0083] The selection of the natural strains to be used for the comparison is based on the following criteria: For the monovalent natural antigen, use will be made of the single transmitted virus that is the best choice in terms of providing coverage of potential T cell epitopes in the global database. The database is biased towards B clade envelopes, so the single best acute Env is a B clade representative. One A, one B and one C subtype transmitted virus sequence is proposed for inclusion in the trivalent set, to compensate for the biases in sampling inherent in the global sequence collection, and to better reflect the circulating pandemic strains. The A and C natural sequences are those that optimally complement the best B clade sequence to provide potential epitope coverage of the database. Vaccine antigens have been selected from among available SGA sequenced acute samples, each representing a transmitted virus. Therefore, this study, although primarily a T cell study, will also provide important additional data regarding the ability of transmitted envelope vaccines to elicit neutralizing antibodies.
[0084] For a mosaic/consensus human trial, the following 4 arm trial is proposed, 20 people per group, with a negative control:
[0085] 1) Con S (a well studied consensus of the consensus of each clade, based on the 2002 database; Con S has been extensively tested in animal models, and has theoretical coverage roughly comparable to a single mosaic.)
[0086] 2) A 3 mosaic M group antigen set designed to, in combination, provide optimal global coverage of 9 amino acid long stretches in the database. Such 9-mers represent potential epitope coverage of the database. Unnatural 9-mers are excluded in mosaics, and rare variants minimized.
[0087] 3) The optimal single best natural protein selected from sequences sampled from acutely infected patients with SGA sequences available; these sequences should correspond to viable, transmitted sequences. As in (2), this sequence will be selected to be the one that provides optimal 9-mer coverage of the database. The B clade currently dominates sampling for the sequence database, so the sequence with the best database coverage will be a B clade sequence.
[0088] 4) The best natural strains from acute infection SGA sequences that in combination provide the best global coverage. (Note: the B and C dominate the M group sampling hence the code naturally selects one of each as the two best. Thus, the third complementary sequence was forced to be selected from an acute SGA A clade set, to counter this bias and better reflect the global epidemic).
[0089] 5) Negative control buffer/saline
[0090] The current M group alignment in the HIV database was combined with all of the newer CHAVI sequences—this includes a total of 2020 sequences:
[0091] 728 B clade
[0092] 599 C clade
[0093] 693 that are all other clades, circulating recombinant forms, and unique recombinants. This was used for the M group vaccine design.
[0094] This sampling is obviously skewed toward the B and C clade. As will be shown subsequently, the coverage of “potential epitopes” (9-mers) in other clades is still excellent.

The Sequences

[00002] [TABLE-US-00002]
    M consensus
    3 mosaics

Single Optimal Natural Sequence Selected from Available Acute SGA Sequences:
[00003] [TABLE-US-00003]

3 Optimal Natural Selected from Available Acute Samples, SGA Sequences:
[00004] [TABLE-US-00004]

Coverage Comparison of the Four Vaccine Antigens.
[0098] Mosaics and naturals are optimized for the first red bar on the left for each vaccine (the total). The “total” represents all sequences, database+CHAVI. The “B” is the subset that are B clade, “C” the subset that are C clade, and “N” the remaining M group sequences that are not B or C (all other clades and recombinants). As B is most common, the single best natural is of course a B, and B thus has the best coverage for Nat.1. Con S, as expected, provides much more even coverage for all clades, and provides better coverage for all the groups except B clade. (Note: in a Con S Macaque study, the natural B was not selected to be optimal, and Con S had better coverage even within B clade than the B vaccine strain that had been used; this was reflected in the number of detected responses to heterogeneous B's. A difference here is that the natural B was selected to be the natural B clade sequence from acute infection that provides optimal coverage). Nat.3 gives good broad coverage, Mos.3 better. (See FIG. 11.)
[0099] The mosaics will minimize rare 9-mers but in Env they cannot be excluded or it is not possible to span certain really variable regions to make intact proteins. For all other HIV proteins tested, it was possible to exclude 9-mers that were found at 3 times or less. Still, the 3 best natural Envs contain more than twice the number of rare 9-mer variants relative to the 3 Env mosaics.
[0100] FIG. 12 includes additional summaries of coverage; ConS gp160 contains quite a few conserved 9-mers that are missed in gp140DCFI, as one would expect. ConS provides slightly less coverage than a single mosaic, but it is already known that ConS works very well in macaques so serves as a good positive control. 1, 2, and 3 mosaics give increasingly better coverage, and Nat.3 is not as good as Mos.3.
[0101] FIG. 13 is alignment dependent, and based on the database alignment (the tow plots above this are alignment independent). Each position represents the 9-mer it initiates as one moves across the protein. The upper bound (black dashed lined) is the sum of the frequencies of the three most common 9-mers starting from each position; it represents the maximal limit that could be achieved for coverage with 3 proteins, and this is not quite achievable in practice because there can be conflicts in a given position for overlapping 9-mers, although the 3 mosaic combination very nearly achieves it. The reason the “total 9-mers” shown in grey varies is because of insertions and deletions in the alignment.
[0102] Only the Mos.3 vaccine cocktail is shown in FIG. 13. However, all four vaccines resorted by coverage is shown in FIG. 14, where those positions that start the 9-mers that are best covered by the vaccine are moved to the left. The exact match line is left in all four plots for a reference point. Not only does Mos.3 (red) approach the maximum, but the orange and yellow near-matches that have potential for cross-reactivity are also improved in this vaccine cocktail as compared to the others.
[0103] The plots shown in FIG. 15 map every amino acid in every sequence in the full database alignment. A row of pixels is a sequence, a column is an alignment position. White patches are insertions to maintain the alignment. All 9-mers that encompass an amino acid are considered. If every 9-mer that spans the amino acid has a perfect match in the vaccine cocktail, the pixel is yellow, so yellow is good. If one is off, light orange, two off, darker orange . . . through no spanning 9-mer matches represented by black. Note: lots of yellow for 3 mosaics, relative to the other vaccines. There is a big patch of the most yellow for the B clade in Nat.1 as the single best natural is a B clade. Note, all those dark bits: in these regions the sequences in the database are different than any 9-mer in the vaccine, so cross-reactivity would be several limited.

Optimization Using 9-Mers.

[0104] 9-mers were selected because that is the most common size of an optimal CD8+ T cell epitope. They range from 8-12, and optimal CD4+ T cell epitopes can be even be larger or smaller. As it turns out, coverage of 9-mers is best when optimized for 9-mer coverage, but if optimization on a different size yields very little decrease in coverage for 9-mers. The same goes for all lengths, 8-12, the peak coverage is for the size selected but the coverage is excellent for other lengths, as the solutions are related. 9-versus 12-mers are shown in FIG. 16, 12 being the most extreme value one might reasonably consider. The coverage is nearly identical for 9-mers optimized for 9 or 12, or for 12-mers optimized for 9 or 12; it is 1-2% higher for the length selected for optimization. Naturally, 12-mers have fewer identities than 9-mers in general, because they are longer so it is harder to find a prefect match. A more comprehensive study was made of this for HIV proteins showing that the loss was consistently larger for 12-mers when optimized on 9 rather than vice versa, and that, in other proteins, this difference could be up to 4-5%. Thus, for Env the selection of 9-mers is less of a problem. Given all of the above, 9-mers were selected since this is the most common optimal CTL epitope length, and since optimal coverage of 9-mers provides approaching optimal coverage of other lengths.

Options for the 3 Best Natural Strains: Acute Transmission Cases, SGA Sequences.

[0105] Use of all database sequences as a source for natural strains for vaccine cocktails was first explored, and then a comparison was made of that with selecting from a restricted group of just acute SGA sequences, essentially transmitted viruses. Essentially comparable coverage of the full database could be achieved by restricting to acute infection sequences. As these have other obvious advantages, they will be used for the natural sequences.
[0106] First, the exploration of coverage using the full database as a source for a natural cocktail. As noted above, the current M group Env one-seq-per-person data set is dominated by B clade infections, closely followed by C clade. Thus, the single best optimal natural selected by the vaccine design program to cover 9-mers in the (database+CHAVI) data set is a B. If one picks from among any sequence in the database, YU-2 comes up as the best single sequence. To get better representation of other clades, the best B was fixed, and then the next best sequence was added to complement YU-2, which is (logically) a C clade sequence, DU467. Those two were then fixed, and the third complement of the antigen was selected. (If the first two are not fixed, and the program is allowed to choose the third, it logically found a B/C recombinant, it has to be forced to select an A. It is believed that forcing the ABC set would improve global coverage, and partly counteract the B & C clade sampling bias among sequences.)
[0107] The optimal naturals from the database tend to harken back to older sequences; this is not surprising, as the older sequences tend to be more central in phylogenetic trees, and thus more similar other circulating strains. For this study, however, it is preferred to use more contemporary Envelope proteins sampled during acute infection and sequenced using SGA, as these sequences accurately reflect the transmitted virus. Given that constraint, it is still desired to optimize for 9-mer coverage, so that the cocktail of natural sequences is given the best chance for success in the comparison with mosaics. It turns out when this was done there was an extremely minor loss of coverage when comparing the trivalent cocktail selected from among acute SGA sequences to the trivalent antigen selected from the entire database, (in both cases optimizing for coverage the full database). Thus, by restricting the antigen cocktails to transmitted virus, coverage is not compromised. This alternative has several advantages. Most importantly, it enables a determination of the cross-reactive potential of antibodies generated from acute infection viruses used for the natural cocktail relative to consensus or mosaics as a secondary endpoint of interest, without compromising the primary endpoint focusing on a comparison of T-cell response breadth of coverage. A large set of B (113) and C (40) clade acute samples sequenced from CHAVI study is available, giving a large dataset from which to select an optimum combination. For the selection of the complementary sequence from the A clade, to complete the B and C in the trivalent vaccine. Several acute sequences were available.
[0108] Analysis of gp160 was undertaken that included the 8 subtype A gp160s, and also a subregion analysis was done with all 15 in V1-V4, to get an indication of whether or not more sequencing was required. Fortunately, one of the available full length sequences made an excellent complement to the B and C acutes, essentially as good as any of the others. This comparison indicated there was no particular need to do more sequencing at this time. It is believed that this is appropriate since with such a limited A baseline to select from, because the A sequence only needs to complement the choice of B and C clade strains, and many Bs and Cs were available from which to choose. Two of the patients from which the Nat.3 cocktail is derived are below. Nat.1 is just the first one.

B Patient 1059

Patient Sex=M


Sample country=USA

Sample city=Long Beach, Calif.

Patient cohort=CA-UCSF

Patient health status=Acute

Viral Load=2,800,000

Infection country=USA

Sample date=Mar. 26, 1998

C Patient 0393

Fiebig Stage=4

Infection country=Malawi

Sample date=17 Jul. 2003

Viral Load=12,048,485

Patient sex=F

CD4count=618 (measured 13 days after sequenced sample)

Patient age=23


[0109] FIGS. 17 and 18 illustrate the minimal loss of coverage in selecting from acute SGA sequences, and a highlighter plot of each of the 3 patients env sequences, that shows that the consensus of each patient is equivalent to the most common strains, and thus an excellent estimate of the actual transmitted virus.

Why M Group and not Clade Specific Coverage?

[0110] It is believed that it is important to strive for a global HIV vaccine, if at all possible, with exploratory methods such as these since many nations have multiclade epidemics, and people travel. While intra-clade coverage can definitely be gained by a within-clade optimized vaccine, the result of such a strategy would be dramatic loss of inter-clade coverage. The hope is that a multivalent mosaic could provide enough breadth to counter viruses of virtually any clades or recombinants. The compromise and benefit in terms of coverage for Env M group versus subtype-specific design is shown in FIG. 19.

Why Env?

[0111] This proof of concept study is well positioned to see differences in breadth of responses using Env as the test antigen. This is partly because of the theoretical considerations described herein (ENV has twice many conserved 9 mers in the mosaics relative to the best natural strain, and only half as many rare variants) and partly because of the prior animal studies. Env studies with a consensus versus natural in macaques showed a highly significant increase in breadth of responses: 3-4 fold more epitopes per Env protein were recognized (Santra et al, in press, PNAS). Env mosaics have shown an even more profound advantage in a mouse study (up to 10-fold over comparable numbers of natural antigens, manuscript in preparation in collaboration with the VRC). Based on this prior work, it makes sense to start with a small human trial testing the breadth of responses to Env. Ultimately, the hope is to apply the proof of concept gained with Env to a more conserved protein like a Gag where it may be possible to confer broadest protection. Gag gives outstanding coverage of the full M group. Tests of Gag and Nef are ongoing in macaque, using a 4 mosaic vaccine cocktail approach (see Example 3). A coverage comparison of macaque 4 mosaic Gag vaccine and proposed human Env 3 mosaic vaccine against the current database is in FIG. 20. There is more theoretical potential for cross-reactivity with the Gag vaccine, but more progress has been made with Env in the animal models to date, so Env has the best foundation to justify moving forward. The three mosaic Env sequences described above and the sequences used in Example 3 are shown in FIG. 21.
The DNAs to be used will be in the form of the full gp160 Env. The gp160 would be in the PCMVR plasmid (Gary Nabel) and will be the identical plasmid used in all VRC DNA immunization trials. Dose is anticipated to be 4 mg. The following DNA constructs will be used:
[0112] DNA optimal Wildtype Env transmitted/founder env (WT Env)
[0113] DNA group M consensus Env (ConS Env)
[0114] DNA Trivalent optimal wildtype transmitted/founder Env (WT Tri Env)
[0115] DNA Trivalent Mosaic Env
NYVAC (vP866) is a recombinant poxvirus vector which has an 18 gene deletion versus wild-type virus. The NYVAC vector will be licensed from Sanofi-Pasteur and manufactured by a third party contractor and will be propagated on a CEF cell substrate. The Env construct expressed in NYVAC will be gp140C (entire Env with transmembrane and cytoplasmic domain deleted and gp41/gp120 cleavage site mutated) or will be a full gp160. The choice of construct design will depend on the ability to make the NYVAC with gp160 forms vs gp140. The dose of NYVAC is anticipated to be ˜1×10^7 TCID50. The following NYVAC constructs will be used:
[0116] NYVAC WT Env
[0117] NYVAC ConS Env
[0118] NYVAC Trivalent Native Env
[0119] NYVAC Trivalent Mosaic Env
Vaccinations will be given by intramuscular injection.
[00005] [TABLE-US-00005]
  Protocol Schema
    Injection schedule in weeks
  Group   Number   Dose   0   4   20   24
  1   20     DNA WT   DNA WT   NYVAC WT   NYVAC WT
    4     Env   Env   EnvA   EnvA
        Placebo   Placebo   Placebo   Placebo
  2   20     DNA ConS   DNA ConS    
    4     Env   Env   NYVAC ConS   NYVAC ConS
        Placebo   Placebo   Placebo   Placebo
  3   20     DNA   DNA   NYVAC   NYVAC
    4     Trivalent   Trivalent   Trivalent   Trivalent
        Native Env   Native Env   Native Env   Native Env
        Placebo   Placebo   Placebo   Placebo
  4   20     DNA   DNA   NYVAC   NYVAC
    4     Trivalent   Trivalent   Trivalent   Trivalent
        Mosaic Env   Mosaic Env   Mosaic Env   Mosaic Env
        Placebo   Placebo   Placebo   Placebo
  Total   96 (80/16)          

Healthy, HIV-1-uninfected volunteers aged 18 to 50 years:
[0121] 80 vaccines
[0122] 16 control recipients
[0123] 96 total participants
Randomized, placebo-controlled, double-blind trial
Duration Per Participant:
Approximately 12 months
Estimated Total Study Duration:
Approximately 18 months

Example 3

Construction of the Plasmid DNA Vaccines and Recombinant Vaccinia (rVV)

[0124] Mosaic gag and nef genes, group M consensus gag and nef genes were generated by converting amino acid sequences of said Gag and Nef, group M consensus Gag and Nef CON-S to nucleotide sequences using a strategy for optimal gene expression. For use as a DNA vaccine, mosaic gag and nef genes, group M consensus gag and nef genes were subcloned into WLV0001-AM DNA vaccine vector. Endotoxin-free plasmid DNA preparation were produced by Puresyn, Inc. (Malvern, Pa.) for the immunization of rhesus monkeys. For boosting recombinant vaccinia viruses expressing the individual mosaic gag and nef genes, group M consensus gag and nef genes were generated. The methods used were as previously described (Liao et al, Virology 353:268-282 (2006); Earl, BioTechniques 23:1094-1097 (1997)).
[0125] Experimental Groups and Vaccination Schedule.
[0126] Three groups of rhesus monkeys were immunized with either 10 mg of the empty DNA vector plasmid (group 1, 6 monkeys), or 5 mg each of group M gag and nef plasmid DNA (group 2, 12 monkeys) or 1.25 mg each of 4 mosaic gag and 4 nef plasmid DNA (group 3, 12 monkeys) intramuscularly at Day 0 and Day 30. The monkeys will be boosted with the corresponding rVV expressing the initial immunizing immunogen (10^9 pfu/monkey) 5 month post-immunization with the 2nd DNA immunization.
[0127] Myristoylation of Gag and Nef has a potential down regulation effect on immune responses and thus the myristoylation of Gag and Nef has been mutated in the sequences used in this study.
[0128] All documents and other information sources cited above are hereby incorporated in their entirety by reference.


1. A polypeptide or protein comprising any one of the mosaic Gag polypeptide Gag M4.1-4Dmyr selected from the group consisting of SEQ ID Nos: 173-176.
2. A composition comprising at least one polypeptide or protein according to claim 1 and a carrier.
3. A method of inducing an immune response in a mammal comprising administering to said mammal an amount of at least one polypeptide or protein according to claim 1 or claim 2 sufficient to effect said induction.
4. The polypeptide according to claim 1, wherein the polypeptide comprises the amino acid sequence of SEQ ID NO: 173.
5. The polypeptide according to claim 1, wherein the polypeptide comprises the amino acid sequence of SEQ ID NO: 174.
6. The polypeptide according to claim 1, wherein the polypeptide comprises the amino acid sequence of SEQ ID NO: 175.
7. The polypeptide according to claim 1, wherein the polypeptide comprises the amino acid sequence of SEQ ID NO: 176.
8. The composition according to claim 2 comprising the four mosaic Gag M4.1-4Dmyr polypeptides comprising SEQ ID Nos: 173-176.
9. The composition according to claim 2 further comprising an adjuvant.

Download Citation

Sign in to the Lens