Identification of H-2Kb Binding and Immunogenic Peptides from Human Papilloma Virus Tumour Antigens E6 and E7

Peptides can be used to induce MHC class I restricted cytotoxic T cells (CTL) tbrough in vivo immunization. This approach may enable the development of peptide vaccination schemes for immunization against viral infection in humans. Human papillomavirus (HPV) is one of a few viruses associated with human cancer and the development of an anti-cancer vaccine seems possible. As a model approach, we searched the E6 and E7 proteins of the human papillomavirus type 16 for possible murine MHC class I restricted peptide epitopes. We utilized the mouse H2-K^b peptide binding motif which consists of phenylalanine or tyrosine at position five and leucine at the carboxy-terminus with the modification that leucine could be replaced by other aliphatic but non-aromatic amino acids. Four peptide sequences from E6 and two from E7 were selected. These peptides were tested for their ability to bind and stabilize K^b and for their immunogenicity in vivo . It was shown tbat one peptide from E6, E6.1 (50–57), bound K^b, but was not able to prime mice in vivo . In contrast, the two selected E7 peptides E7.1 (21–28) and E7.2 (48–55) bound K^b and were immunogenic in vivo . The peptide induced CTL lysed syngeneic EL-4 cells transfected with the open reading frame of E7 but not vector only transfectants. This implies tbat both peptides were naturally processed and presented by K^b on the surface of target cells. MHC class I peptide binding motifs therefore appear to be an effective and useful tool to predict peptide epitopes of proteins associated with cancer.     

TLR9 Activation Increases TAP-Independent Vesicular MHC Class I Processing In vivo

Cross-presentation of soluble protein antigens on major histocompatibility complex (MHC) class I by dendritic cells (DC) can occur in vesicular, endolysosomal compartments and be either dependent or independent of TAP peptide transporters. Here we investigate if an immunostimulatory CpG oligodeoxynucleotide can increase the activity in a TAP-independent endolysosomal vesicular pathway (el-VP) in vivo as we have earlier found in in vitro cultured DC. We use the in vivo response of CFSE labelled OT-1 T cells, transgenic for a T-cell receptor (TCR) that recognizes an ovalbumin (OVA)-derived peptide (SIINFEKL) presented by H-2K^b, transferred into TAP1^−/− mice, as a functional read-out for activity in the el-VP. We have found a poor OT-1 T-cell response to soluble OVA which, however, could be strongly enhanced by the simultaneous administration of CpG. This increased responsiveness required both the endolysosomal cathepsin S (CatS) and Toll like receptor (TLR)9, the CpG receptor, both of which are present in the el-VP. Confocal microscopy demonstrated a co-localization of H-2K^b/SIINFEKL and the endolysosomal marker LAMP1 in CD11c positive DC which was markedly increased by CpG administration. No complexes were found in the ER and cis-Golgi compartments in TAP1^−/− mice, indicating the lack of classical MHC-I processing. In DC isolated from CatS^−/− mice the opposite was found, complexes were present in the ER but not in the el-VP. We conclude that in vivo activation of TLR9 by CpG increases the efficiency of TAP independent el-VP and that this might contribute to the potent adjuvant activity of this type of compound. The cellular mechanisms remain to be established.     

Non-Diabetogenic Insulitis in NOD↔BIO. GD Allophenic Mice in Spite of Permissive Class I MHC Antigens

Allophenic mice (embryo aggregation mouse chimeras) enable us to dissect the process of spontaneous autoimmunity under physiological conditions. Our previous experiments showed that the autoimmune process in allophenic mice of the NOD↔C57Bl/6 strain combination does not progress from insulitis to diabetes. One possible explanation for this protection is that H-2 K^d-restricted CD8⁺ T cells kill only NOD ß cells (K^d, D^b) in the chimeric islets, while the B6ß cells (K^b, D^b) are spared from destruction. To test this hypothesis we analysed 22 NOD↔B10. GD chimeras in which the class I MHC are shared by both parental strains. Therefore all the ß cells in these chimeras express H-2 K^d molecules. Ten allophenic mice were killed at 7 weeks and studied for early pathology. No evidence for intra-islet infiltration was obtained at this age, suggesting that the autoimmune process in NOD↔B10. GD chimeras is slower than in NOD mice. Twelve chimeras were followed up for 1 year for disease development and all failed to progress to full-blown diabetes, despite the occurrence of intra-insulitis in six out of 12 mice. The lack of disease in NOD↔B10. GD chimeras demonstrates that class I MHC chimerism does not account for diabetes resistance in NOD-allophenic mice.     

A simplified procedure for the preparation of MHC/peptide tetramers: chemical biotinylation of an unpaired cysteine engineered at the C-terminus of MHC-I

Recently, a powerful approach for the detection of MHC/peptide-specific T cells has been made possible by the engineering of soluble-tetrameric MHC/peptide complexes, consisting of singly biotinylated MHC/peptide molecules bound to fluorescent-labeled streptavidin. These tetrameric molecules are thought to compensate for the low affinity and relative fast dissociation rate of the TCR/MHC-peptide interaction by increasing the avidity of this interaction, thus allowing the stable binding of MHC/peptide tetramers to TCR expressing cells. Here we describe a new more simplified procedure for obtaining MHC/peptide tetramers using the well-characterized H-2K(b)/VSV system. This procedure consists of the incorporation of an unpaired cysteine residue at the C-terminus of the H-2K(b) molecule, allowing site-specific biotinylation by a -SH-specific biotinylating reagent. The H-2K(b)/VSV tetramers bound only to hybridomas expressing H-2K(b)/VSV-specific TCRs. When coated on a plate, these tetramers were able to induce IL-2 release by those hybridomas. Furthermore, H-2K(b)/VSV tetramers bound to CTL populations obtained from mice immunized with VSV-peptide. The specificity of the binding was further refined by studying cross-recognition of VSV by CTL populations obtained from mice immunized with single amino acid substituted VSV peptide variants. H-2K(b)/VSV tetramers bound only to those CTL populations that cross-reacted with the wild-type VSV peptide. Our method provides a simple, efficient and inexpensive procedure for making MHC/peptide tetramers, a highly specific and very useful reagent with a number of important applications in basic and clinical T cell research.     

Interaction of In Vitro- and In Vivo-Generated Cytotoxic T Cells with SV40 T Antigen: Analysis with Synthetic Peptides

Virus-specific cytotoxic T cells recognize antigens in the form of peptides (8 or 9 amino acids long) bound to MHC class-I molecules. Exposure of unprimed murine splenocytes to synthetic peptides of viral antigens elicits primary CTL in vitro. The fme specificity of such CTL as well as the correlation between binding affinity of peptides to class-I molecules and CTL induction was analysed using synthetic peptides corresponding to overlapping and distinct amino-acid residues in SV40 T antigen (Tag) D_b-restricted T-cell epitopes I, II-III, and V. The peptides induced cross-reactive CD8₊ primary CTL in spienocytes of naive C57 BL/6 mice. This reactivity was seen regardless of the peptides allelic anchor motifs or their abilities to stabilize empty class-I molecules. However, none of the primary CTL and CTL lines lysed Tag-expressing cells. In contrast, CTL generated in vivo by immunizing mice with Tag-expressing cells recognized endogenously processed Tag as well as synthetic peptides. The peptides recognized by these CTL depended on the intracellular concentration of Tag antigen in the immunizing cells. The reactivity of these CTL was peptide specific as shown by a functional peptide competition assay. Moreover, three peptides bound to and were recognized in the context of both K^b and D^b molecules. These results have revealed a flexible disposition of MHC class-I molecules with regard to peptide binding and also reflected lack of correlation between binding affinity to class-1 molecules and the capacity of peptides to induce primary CTL or to serve as potential targets. The significance of these findings in relation to identifying major T-cell epitopes using allele specific peptide motif and in vitro maintained CTL clones is discussed.     

Private specificity of H-2Ldx molecule detected serologically by a surface antigen redistribution method (capping)

The presence of H-2·1 positive-H-2·63 negative molecules, H-2L^dx, in the products of the ^dx region was originally detected in H-2^dx inbred strains GRS/A and LIS/A (Snoek et al. 1979).
This study confirms the existence of H-2L^dx molecules in 2 congeneic strains with a H-2^dx haplotype on C3H (C3H·LG strain) and B10 (B10·GR strain) background. Besides anti-H-2·1-like antibodies present in anti-D^dx allo-antiserum, monoclonal antibody d anti- k (H100-5/28, H-2·m3) was shown to react with H-2L^dx molecule. Furthermore, evidence is shown that H-2L^dx molecule has a unique specificity, which is detected serologically in a capping experiment.     

Altered Allogeneic Response in Neonatally Anti-MHC Class I-Treated Mice

Neonatal treatment of C3H mice (H-2^k) with anti-K^k monoclonal antibodies results in altered cytotoxic responses against allogeneic targets. After 2–3 weeks of antibody treatment, no difference in the number of CD4⁺8⁻ or CD4⁻ 8⁺ T cells was observed between the antibody- and saline-treated mice. However, antibody-treated mice had a significantly reduced cytotoxic response against various allogeneic major histocompatibility complex (MHC) class I-expressing targets. The strongest reduction was observed in very young mice (up to 2 weeks of age). As the mice got older, the allo MHC-specific responses reached control levels. No significant changes in T-cell receptor (TCR)-V-region usage was observed even in young antibody-treated mice. The results suggest that the reduction in the number of positively selecting elements reduces alloreactivity and most likely also the diversity of TCR-repertoire. However, the reduced alloresponsiveness was not restricted to either allogeneic K- or D-encoded molecules, suggesting that self MHC-D-region encoded molecules can mediate positive selection of T cells able to react against both K and D region-encoded allogeneic MHC class I molecules.     

Role of strong anchor residues in effective binding of 10-MER and 11-MER peptides to HLA-A2402 molecules.

Analysis of the structural requirements for the interaction of antigenic peptides with HLA-A24 molecules are very important for studies of T cell recognition of various antigens, because HLA-A24 (A^*2402) is most common HLA-A allele in the world, especially in Oriental population. In order to precisely investigate the interaction of peptides with HLA-A24 molecules beyond previous analysis of self-peptides eluted from HLA-A24 molecules, we examined the A^*2402 interaction of 172 chemically synthesized 8-mer to 11-mer peptides carrying two residues (Try and Phe) at P2 and four residues (Phe, Trp, Leu and Ile) at their C-terminus by the use of stabilization assay. The results were statistically analyzed to assess the influence of anchor residues on peptide binding. The length of peptides (9- to 11-mer) did not affect A^*2402 binding except 8-mer peptides. Peptides possessing the aromatic residues at their C-terminus bound to A^*2402 molecules stronger than those bearing the aliphatic hydrophobic residues. These results indicate that two aromatic hydrophobic anchor residues permit the binding of longer peptides to A^*2402 molecules. Compared to our recent studies of B^*3501 and B^*5101 binding peptides, the present study suggested that both B and F pockets of A^*2402 molecules might be large and deep because these pockets favored bulky aromatic residues.     

A viral peptide can mimic an endogenous peptide for allorecognition of a major histocompatibility complex class I product.

Alloreactive class I-restricted T cells may recognize the class I structure alone, in association with a specific peptide, or with any stabilizing peptide. We have tested the role of endogenous peptides in the recognition of H-2Kb molecules by two alloreactive cytolytic T lymphocyte (CTL) clones using the mutant tumor line RMA-S, which expresses its surface H-2b molecules devoid of peptides and is not lysed by these two CTL clones. Empty H-2b molecules on RMA-S cells can be stabilized by binding exogenously added peptides. H-2Kb-specific recognition of the RMA-S cells by one of the CTL clones was restored by endogenous peptide extracts which only minimally stabilized H-2Kb on the surface of RMA-S cells, indicating the requirement for a specific peptide on a limited number of H-2Kb molecules. In addition, one out of three peptides which greatly enhance the expression of H-2Kb, the nucleoprotein peptide 52-59 from vesicular stomatitis virus (VSV), was also able to restore the lysis of RMA-S cells by the clone. The recognition of a common motif by an alloreactive clone (H-2k anti-H-2Kb) and virus-specific Kb-restricted clones suggests that both H-2k and H-2b thymic environments allow selection of T cells capable of recognizing H-2Kb+VSV and that tolerance to self, as would be the case in the (H-2k x H-2b)F1 mice, would partially delete the repertoire of antiviral T cells.     

Binding of Helix Pomatia A Hemagglutinin to Human Erythrocytes and their Cells. Influence of Multivalent Interaction on Affinity

The binding of ¹²⁵I-labelled intact (hexavalent) and partially reduced (divalent) Helix pomatia A hemagglutinin to human A.₁, A.₂, A.₃, A₁B and A₂B erythrocytes, human lymphocytes, human lymphoblastoid and other tumor cell lines was investigated. The essential finding was that the association constants calculated for the interaction between hemagglutinin and A-erythrocytes were many orders of magnitude higher than the intrinsic association constant for the interaction between the hemagglutinin and the blood group A determinant. The latter value was 5–10³1/mole. The K-values for intact hemagglutinin and A and AB erythrocytes were in the order of 10¹⁰1/mole at 18–22 °C and pH 7.3. For partially reduced hemagglutinin the K-values were in the order of 5–10⁷ 1/mole. Multivalent interaction would seem to be the essential factor responsible for the high K-values in the cell binding experiments. Intact hemagglutinin reacted against A₁ and A₂ erythrocytes or against a human osteogenic sarcoma cell tine (2T) gave homogeneous binding curves in Scatchard's plot. Human lymphocytes and most lymphoblastoid cell lines lacked hemagglutinin receptors.     

Structural and kinetic basis for low affinity cross-reactivity in T cell allorecognition

The alloreactive BM3.3TCR interacts with high affinity with H-2Kb loaded with the endogenous peptide pBM1 (INFDFNTI), and shows low affinity cross-reactivity for H-2Kb loaded with a viral peptide VSV8 (RGYVYQGL), CTL activity requiring 10(3)-fold higher peptide concentration and being highly sensitive to inhibition by anti-CD8 monoclonal antibody. VSV8 peptides substituted with pBM1/TCR contact residues (N6 and T7) retained low affinity characteristics and among pBM1 peptides substituted with residues Q6 and/or G7 present in VSV8, only pBM1(G7) was recognized, albeit with characteristics akin to those of VSV8. Despite the difference in KD values and the faster dissociation rate of multimeric VSV8/H-2Kb as compared to pBM1/H-2Kb complexes, similar TCR occupancy could be achieved with both multimers either at 4 or 37 degrees C. Only TCR engagement with pBM1/H-2Kb, however, resulted in early (Ca2+ flux) and late (CD69 expression) activation events in naive BM3.3TCR CD8 T cells. CD8 coreceptor, essential for binding of the weak agonists, was dispensable for binding of pBM1/H-2Kb multimers and their induction of signaling in naive T cells. Hence, high number of TCR and coreceptor engagement by weak agonists fail to substitute for strong agonist TCR engagement that can be coreceptor-independent and involve a limited number of TCR.     

An allo-specific peptide derived from HLA-A2 is presented by HLA-DQ7

The peptide motifs of two HLA class II molecules, DR11 and DQ7, were determined from natural peptides. The EBV transformed B cells JVM (HLA-A2-DRB1^*1102-DQA1^*0501-DQB1^*0301) were cultured to a final yield of 2 10¹⁰ cells. DR11 and DQ7 molecules were immunopurified and peptides were extracted after acid elution and separated by reverse-phase HPLC. Five peptides from DR11 and five from DQ7 were sequenced using Edman degradation and other peptides were analysed by pool sequencing. Peptides were from 11 to 15 amino acids in length and P often occurred at the second residue for both DR5 and DQ7. The peptide motif for DR11 was I at position i and K or R at position i + 4. For DQ7 the most significant signal was A in the middle of the peptide as also described by Falk et al. The source of one peptide eluted from DQ7 was a polymorphic part of the HLA-A2 heavy chain (from 56th to 69th amino-acid). It was also exactly the same peptide than the synthetic peptide used by Krensky et al in the 10th international workshop to modulate lysis by HLA-A2-specific cytotoxic T lymphocytes. The biochemical characterisation of this peptide from HLA-DQ7 strongly supports functional tests showing an indirect presentation of alloantigens by MHC molecules.     

Expression of the H-2 Antigenic Complex on Murine Haematopoietic Stem Cells

We used hybridoma-derived monoclonal antibody to H-2K^k antigens and xenoantibodies to β₂-microglobulin (β₂) to study the expression of the H-2 antigenic molecular complex on murine hematopoietic stem cells and the effect of long-term culture in vitro on the expression of these antigens. Monoclonal anti-H-2K^k antibody produced potent complement-dependent inhibition of pluripotent (CFU-S), myeloid (CFU-C), and erythroid (CFU-E) stem cells from the bone marrow of C3H and AKR (H-2^k) mice and was without effect on stem cells from Balb/C (H-2^d) mice. Anti-β₂m xenoantibodies inhibited stem cells from all three strains. Stem cells could not be 'rescued' from the inhibitory effects of the antibodies by the addition of thymocytes to marrow cells after antibody treatment. Both the anti-H-2K^k monoclonal antibody and the anti-β₂m xenoantibodies produced potent inhibition of AKR (H-2^k) CFU-C that had been maintained in culture for up to 6 weeks. These results indicate that murine CFU-S, CFU-C, and CFU-E express H-2 antigens and that the expression of these antigens by CFU-C is not altered during long-term culture.     

Genetic Control of the Murine Cell-Mediated Immune Response In Vivo

Various inbred and congenic strains of mice were immunized with the linear terpolymer L-glutamic acid⁶⁰-L-alanine³⁰-L-tyrosine¹⁰ (GAT). Using a radioisotopic footpad assay to measure cell-mediated immunity in vivo, mice with H-2^a, H-2^b, H-2^d, and H-2^k histocompatibility alleles showed a positive reaction, whereas mice with H-2^P and H-2^S alleles failed to respond. The ability of 'responder'. lymphocytes to show an immune response resides in the thymus-derived (T) lymphocyte population. Unfractionated spleen cells from H-2^q nonresponder mice, on transfer, showed no reactivity, whereas T-cell-enriched preparations were active.     

Both Immunization with Protein and Recombinant Vaccinia Virus Can Stimulate CTL Specific for the E7 Protein of Human Papilloma Virus 16 in H-2d Mice

The transforming protein E7 of human papilloma virus type 16 can stimulate cytotoxic T lymphocytes (CTL) which can protect experimental animals against growth of E7 expressing tumour cells. In this study we compared CTL responses in mice immunized with either E7 protein in MF59 adjuvant or with recombinant vaccinia virus expressing E7 (Vac-E7). We have chosen H-2^d mice because no E7-specific CTL responses have been described in this MHC haplotype. Immunization of these mice with Vac-E7 generated CTL which lysed target cells infected with Vac-E7 or transfected with the E7 gene. CTL from mice immunized with E7 protein in MF59 adjuvant showed specificity for the same target cells. Antibody blocking experiments revealed that both immunization with Vac-E7 and E7 protein stimulated CD8⁺ effector CTL. The find specificity of CTL induced by the two immunization protocols was similar. A major CTL epitope was mapped to the carboxyl terminal amino acids 48–98 of the E7 protein. Peptide isolation from E7 expressing cells followed by HPLC separation indicated that CTL induced by immunization with protein and Vac-E7 recognized the same HPLC purified peptide fractions. Together, the study suggests that vaccines based on protein can activate CTL with similar fine specificity to CTL induced by vaccines based on recombinant vaccinia virus.     

Crystal structure of HLA-A*2402 complexed with a telomerase peptide

HLA-A*2402 is the most commonly expressed HLA allele in oriental populations. It is also widely expressed in the Caucasian population, making it one of, if not the most abundant HLA I types. In order to study its structure in terms of overall fold and peptide presentation, a soluble form of this HLA I (alpha1, alpha2, alpha3 and beta(2)m domains) has been expressed, refolded and crystallized in complex with a cancer-related telomerase peptide (VYGFVRACL), and its structure has been solved to 2.8 A resolution. The overall structure of HLA-A*2402 is virtually identical to other reported peptide-HLA I structures. However, there are distinct features observable from this structure at the HLA I peptide binding pockets. The size and depth of pocket B makes it highly suitable for binding to large aromatic side chains, which explains the high prevalence of tyrosine at peptide position 2. Also, for HLA binding at peptide position 5, there is an additional anchor point, which allows the proximal amino acids to protrude out, providing a prominent feature for TCR interaction. Finally, pocket F allows the anchor residue at position 9 to be bound unusually deeply within the HLA structure.     

T Cell-Mediated Cytotoxicity Against p53-Protein Derived Peptides in Bulk and Limiting Dilution Cultures of Healthy Donors

The p53 tumour suppressor gene product plays an important role in the development of most human cancers. Point mutations in the p53 gene are common in malignant states and results in over-expression of wild type and mutant determinants of the p53 protein. This process might generate MHC-I restricted epitopes for T cell recognition and p53-derived peptides have been suggested as targets for tumour-specific cytotoxic T lymphocytes (CTL). Our primary aim was to estimate the frequencies of p53-peptide reactive CTL precursors (CTLp) in peripheral blood from healthy young individuals. We selected wild type and mutated peptides derived from the p53 sequence with a binding motif for HLA-A2.1 molecules. Peripheral blood mononuclear cells (PBMC) from healthy HLA-A2 donors were stimulated in vitro in bulk cultures as well as in limiting dilution cultures using autologous cells pulsed with p53 peptides as stimulator cells. T cell reactivity was observed towards both wild type and mutated p53 peptide epitopes with CTL precursor frequencies varying from 1: 2 × 10⁴ to 1: 1.5 × 10⁵. These results might suggest the presence of an ongoing immune response in normal individuals against cells expressing increased levels of p53 protein.     

Compression of functional space in HLA-A sequence diversity

The major histocompatibility complex (MHC) is highly polymorphic and more than 1500 human MHC alleles are known to date. These alleles do not bind to a given peptide with identical affinity. Although MHC alleles are functionally related, it is difficult to quantify the functional variation between them. Three-dimensional structures of known MHC-peptide (MHCp) complexes suggest that specific peptide residues bind selectively to functional pockets in the binding groove. From a set of known MHCp structures we identified 21 critical polymorphic functional residue positions (CPFRP) that significantly reduced functional pocket variability to just 189 among 212 HLA-A alleles. Interestingly 101 HLA-A alleles clustered into 29 clusters such that the six functional pockets formed by the CPFRPs are identical within the cluster.     

Rabbit Antisera with Specificity for Isolated Mouse T-Lymphocyte Receptors

Rabbit antibodies obtained after Immunization of mouse Immunoglobulin (Mig)-tolerant rabbits with B6 anti-CBA IgG and having specificity for B6 anti-CBA IgG and T-cell receptors (antiserum 5936) were used to isolate 5936-reactive molecules from B6 anti-CBA mixed lymphocyte culture supernatants. Such 5936-reactive molecules were produced by the B6 T cells, and they did not react with rabbit anti-MIg antisera. They had a mol. wt of 50,000–75,000, and were single-chain polypeptides that did not react with concanavalin A (Con A) Sepharose. These molecules were in turn injected into rabbits, and the antisera thus obtained had the following characteristics: (1) they reacted against B6 anti-CBA T-cell receptor material but not against B6 anti-CBA IgG; (2) they reacted with about 35% of B6 (H-2^b. Ig-1^b) anti-CBA T cells, 25% of B6 Con A blasts and 0 10%; of normal B6 T cells but not with B6 lipopolysaccharide (LPS) blasts, C3H.B10 (H-2^b, Ig-I) anti CBA or CBA anti-B6 T cells, CBA Con A blasts or normal CBA T cells: and (3) they reacted with the same 50,000–75.000 mol. wt, T-cell-derived molecules as did antiserum 5936. The implications of these findings are discussed in relation to the nature of T-cell receptors.     

HLA-A*0201-restricted cytotoxic T-cell epitopes in three PE/PPE family proteins of Mycobacterium tuberculosis

CD8⁺ T cells are thought to play an important role in protective immunity against tuberculosis. We report the identification of three peptides derived from Rv1818c, Rv3812 and Rv3018c proteins of Mycobacterium tuberculosis that bound to HLA-A*0201 molecules and their ability to induce in vitro T-cell response in peripheral blood lymphocytes from HLA-A*0201-positive healthy individuals (PPD+) and patients with TB. The peptide-specific cytotoxic T lymphocytes (CTL) generated were capable of recognizing peptide pulsed targets. Three 9-mer peptides bound with high affinity to HLA-A*0201 and displayed low dissociation rates of the bound peptide from HLA. Epitope-specific recognition was demonstrated by the release of perforin and γ-interferon. Overall, our results demonstrate the presence of HLA class I-restricted CD8⁺ CTL against proteins from PE and PPE proteins of M. tuberculosis and identify epitopes that are strongly recognized by HLA-A*0201-restricted CD8⁺ T cells in humans. These epitopes thus represent potential subunit components for the design of vaccines against tuberculosis.