Mycobacterium tuberculosis genome-wide screen exposes multiple CD8⁺ T cell epitopes

Abstract

Mounting evidence suggests human leucocyte antigen (HLA) class I-restricted CD8⁺ T cells play a role in protective immunity against tuberculosis yet relatively few epitopes specific for the causative organism, Mycobacterium tuberculosis, are reported. Here a total genome-wide screen of M. tuberculosis was used to identify putative HLA-B*3501 T cell epitopes. Of 479 predicted epitopes, 13 with the highest score were synthesized and used to restimulate lymphocytes from naturally exposed HLA-B*3501 healthy individuals in cultured and ex vivo enzyme-linked immunospot (ELISPOT) assays for interferon (IFN)-γ. All 13 peptides elicited a response that varied considerably between individuals. For three peptides CD8⁺ T cell lines were expanded and four of the 13 were recognized permissively through the HLA-B7 supertype family. Although further testing is required we show the genome-wide screen to be feasible for the identification of unknown mycobacterial antigens involved in immunity against natural infection. While the mechanisms of protective immunity against M. tuberculosis infection remain unclear, conventional class I-restricted CD8⁺ T cell responses appear to be widespread throughout the genome.

Introduction

Tuberculosis (TB) is a devastating disease affecting most countries, but particularly those that are developing. It is estimated that about 8 million people develop disease and 2–3 million die each year because of it [1]. Based on tuberculin skin-test surveys it is believed that as many as a third of the world's population may be latently infected with the causative organism, Mycobacterium tuberculosis. Cellular immunity is known to be crucial for protection against disease [2–4] and during HIV-1-induced progressive immune deficiency the rate of development of TB disease is increased dramatically, from about 10% in a lifetime to around 8% per year [5,6].

Although a role for CD4 T cells in protection against M. tuberculosis is well reported, there is also a large body of evidence that supports an important role for CD8 T cells [7]. It has been suggested that while CD4 T cells are crucial during the early stages of infection, CD8 T cell responses are more important for control of chronic M. tuberculosis infection [8]. The effector mechanisms utilized by CD8 T cells are not clear but may include the production of macrophage-activating signals [9], cytokines such as interferon (IFN)-γ and tumour necrosis factor (TNF)-α[10–12] and cytolytic molecules such as perforin and granulysin [13,14].

To date, relatively few human CD8 T cell epitopes have been found by conventional methods in the 3924 predicted open reading frames (ORFs) of M. tuberculosis[15] compared, for instance, to the large body of available data on CD8 T cells against HIV-1 (e.g. http://www.hiv.lanl.gov). In order to prepare for natural history studies of M. tuberculosis infection and future intervention studies involving new TB vaccines, we have begun to identify human leucocyte antigen (HLA)-B*35-restricted mycobacteria-specific CD8 T cell responses [12,16]. This HLA type is common in West Africa [17,18], reaching about 30% in The Gambia, and T cell epitopes restricted by HLA-B*35 have been observed thus far in only six M. tuberculosis-derived antigens: Ag85 complex (Rv3804c–Rv1886c–Rv0129c) [16], SPASE I (Rv2903c) [12], 38-kDa lipoprotein (Rv0934) and MPT53 (Rv2878c) (Hammond and Klein, unpublished observations). Another adventitious circumstance of HLA-B*35 is that it is part of the so-called B7-supertype family, of which several other members are also common in West Africa and which have been shown to share similar peptide binding features [19,20]. Furthermore, with the recent publication of several genome sequences of mycobacterial species [21–24] we now have an unprecedented opportunity to evaluate the entire genome of tuberculous mycobacteria in relation to cell-mediated immune responses against M. tuberculosis, in particular the role of major histocompatibility complex (MHC) class-I restricted CD8 T cell responses in TB. Genome-wide scanning with assisted immunobioinformatics will allow for the rapid identification of novel T cell epitopes [25,26], which will be important to further dissect the role of CD8 T cells during M. tuberculosis infection and to help developing immunological tools for monitoring new TB interventions.

Previously we have used with success the publicly accessible internet-based algorithm HLA_BIND, that allows for the rapid identification of putative T cell epitopes, to find HLA-B*35 restricted T cell epitopes in selected M. tuberculosis antigens [12,16]. In the present study we used the algorithm HLA BIND [27] to evaluate the entire genome of M. tuberculosis H37Rv, and identified a series of novel CD8 T cell epitopes which can also be presented in the context of the HLA-B7 supertype family members.

Subjects, materials and methods

Human subjects

Healthy Bacille Calmette–Guérin (BCG)-vaccinated volunteers naturally exposed to pathogenic and non-pathogenic environmental mycobacteria were recruited from the blood bank of Royal Victoria Hospital (RVH) in Banjul, The Gambia. BCG vaccination was confirmed by the presence of a distinct scar or vaccination card. Heparinized venous blood was subsequently collected after written informed consent was obtained. These were processed and peripheral blood mononuclear cells (PBMC) were obtained by gradient centrifugation with Lymphoprep® medium 1·077 (Axis-Shield PoC AS Oslo, Norway) and subsequently cryopreserved in liquid nitrogen until use. DNA was extracted and HLA genotyping was performed on site for common class-I alleles by polymerase chain reaction (PCR) with sequence-specific primers [28]. Twenty-six subjects with HLA B*3501, two donors with HLA-B*07 and two donors with HLA-B*53 were selected for this study.

Synthetic peptides

We utilized computer algorithm HLA_BIND, designed by Dr K. Parker et al. (accessible on the internet at http://bimas.dcrt.nih.gov/molbio/hla_bind/) [27] and searched all 3924 predicted ORFs of M. tuberculosis H37Rv for potential HLA-B*35 restricted T cell epitopes and selected 13 peptides with a high score (see Table 1). In addition, we used as controls three HLA-B*35-binding peptides: HIV-1_SF2 Nef (aa. 72–80) (4T6R), FPVTPRVPL and Epstein–Barr virus (EBV; strain B95·8), large tegument protein (TEGU; aa 1974–82) HPLTNNLPL [29] and six M. tuberculosis-derived HLA-B*3501 restricted epitopes that we have published previously [12,16] (see Table 1).

Table 1.

In silico genome-wide screening for HLA-B*3501 restricted T cell epitopes.

			Start	Sequence	IC50 (µm)	t1/2 (h)	HLA_BIND d1/2 (min)	Ref.
Published epitopes
p39	EBV	TEGU	1974	HPLTNNLPL	n.d.	n.d.	20	[28]
p40	HIV-1	Nef	072	FPVTPRVPL	0·6–1·3	>7	20	HIV Immunology database ref.
M. tuberculosis
p43	Rv1886c	Ag85B	110	MPVGGQSSF†	0·2–0·7	>7	20	[14,15]
p52	Rv0129c	Ag85C	204	WPTLIGLAM†	0·1–0·4	>7	40	[14,15]
p54	Rv1886c	Ag85B	264	IPAEFLENF†	1·0–1·1	>7	40	[14,15]
p55	Rv3804c	Ag85A	267	LPAKFLEGF	0·7–1·1	6·4–7	20	[14,15]
p66	Rv0129c	Ag85C	268	IPAKFLEGL	7·1–7·3	<1	20	[14,15]
p58	Rv0934	38-kDa LP	045	TPASSPVTL	8·2–10·3	0·8	20	Hammond, unpubl.
p46	Rv0934	38-kDa LP	175	LPGTAVVPL†	1·3–1·6	>7	20	Hammond, unpubl.
p59	Rv0934	38-kDa LP	292	TPANQAISM	2·6–5·4	>7	40	Hammond, unpubl.
p49	Rv2903c	SPASE I	201	EPYLDPATM†	1·6–2·1	>7	60	[11]
p47	Rv2878c	MPT53	134	VPWQPAFVF†	0·1–0·5	>7	20	Hammond, unpubl.
Predicted epitopes, this study								Parent protein
M. tuberculosis
p99	Rv2823c		615	RPREATIIY	n.d.	n.d.	480	h. proteina
p100	Rv1461		324	IPRDEVRVM	n.d.	n.d.	360	h. proteina
p101	Rv0670		038	KPRDDAAAL	n.d.	n.d.	360	endonuclease IV
p102	Rv1641		091	RPKIDDHDY	n.d.	n.d.	360	initiation factor
p103	Rv3689		009	RPKPDTETY	n.d.	n.d.	360	h. proteina
p104a	Rv3378c		254	RPKPDYSAM	n.d.	n.d.	360	h. proteina
p105a	Rv2182c		020	RPKVEGLEY	n.d.	n.d.	360	h. proteina,b
p106a,b	Rv1280c		284	RPRLDSITY	n.d.	n.d.	360	oppA
p107	Rv2476c		785	RPRYEIFVY	n.d.	n.d.	360	h. proteina,c
p108a,c	multi hit-3			IPKLRQGSY	n.d.	n.d.	120	IS1081 element
p109	multi hit-4			KPGCDAPAY	n.d.	n.d.	120	REP 13E12 family
p110	multi hit-5			RPGCDAPAY	n.d.	n.d.	120	REP 13E12 family
p111	multi hit-6			SPKETWLRL	n.d.	n.d.	120	h. proteina

HLA_BIND was used to predict the HLA-B*3501 binding motif from the H37Rv genome. Predicted scores with a median half-life of dissociation (t_1/2) ≥ 120 min were much higher than the published HLA-B*3501 motif obtained through conventional methods.

^†9-mer HLA-B*3501 peptides from our previous published work.

^aHypothetical protein

^bcontains ATP/GTP-binding site motif A

^cNAD-specific dehydrogenase. Estimators of peptide binding: IC₅₀ and t_1/2 values were derived from our previously published data set and included for comparison [12,15,16].

These 9-mer predicted epitopes were synthesized using Fmoc chemistry and analysed by high performance liquid chromatography (HPLC) and mass spectrometry (Research Genetics Inc., Huntsville, AL, USA) indicating >80% purity. Lyophilized peptides (5–10 mg) were reconstituted in 400 µl dimethyl-sulphoxide (DMSO) and diluted with 0·9% saline to a stock concentration of 10 mm, aliquoted and stored at −70°C until use. All the peptides identified were able to bind HLA-B*3501 single transfectants (RMA-S B*3501) shown by staining with w6/32 (an anti-HLA class I monoclonal antibody) in flowcytometry, as reported previously [16].

In vitro expansion of peptide-specific CD8⁺ T cells

PBMC were restimulated in vitro as described previously [12,16,30]. Briefly, cells were thawed, washed and resuspended to a concentration of 1 × 10⁶ cells per ml in complete medium containing RPMI-1640 (Life Technologies, Paisley, UK) supplemented with 10% heat-inactivated AB serum (Sigma, St. Louis, MO, USA), 2 mm l-glutamine, 50 µg/ml ampicillin and 5 ng/ml recombinant human interleukin (IL)-7 (rHu IL-7) (Genzyme, Cambridge, MA, USA) added from the start. Cells were cultured in 96-well round-bottomed microtitre plates and individual peptides added to a final concentration of 20 µm per well and incubated for 10 days at 37°C 5% CO₂. Cultures were fed on day 3 by adding 50 µl culture medium containing all the above with the exception of rHu IL-7, which was replaced with 20 U/ml recombinant human IL-2 (rHu IL-2; generously provided by Dr Ronald Rombouts, Chiron Benelux BV, Amsterdam, the Netherlands). The cultures were again refreshed on day 7 by replacing culture supernatants with 100 µl complete media containing 10 U/ml rHu IL-2. Cells were harvested on day 10, washed twice in RPMI-1640 and resuspended in complete medium (RPMI-1640 supplemented with 10% heat-inactivated AB serum, 2 mm l-glu and 50 U/ml ampicillin) to a final concentration of 1 × 10⁶ cells/ml. CD8 T cells were depleted using MACS magnetic beads according to the manufacturer's instructions (Miltenyl Biotec, Baraisch-Gladbach, Germany). Besides peptide-stimulated T cells, we also tested unstimulated cells directly ex vivo.

Enzyme-linked immunospot (ELISPOT) assay for IFN-γ

The ELISPOT assays were performed essentially as described previously [16,31]. Briefly, 96-well nitrocellulose-backed MAIP S45 plates (Millipore, Bedford, UK) were coated overnight at 4°C with 10 µg/ml antihuman IFN-γ mouse monoclonal antibody (1-D1K; Mabtech, Stockholm, Sweden). The plates were washed six times with sterile phosphate buffered saline (PBS) to remove unbound coating antibody. Cell suspensions were prepared and four replicate wells were seeded with 10⁵ cells per well. Peptides were added in a final concentration of 20 µm. As positive controls, cells were stimulated with 10 µg/ml M. tuberculosis purified-protein-derivative (PPD RT49; Statens Serum Institut, Copenhagen, Denmark) or 5 µg/ml phytohaemaglutinin (PHA) (Sigma-Aldrich, UK). Cells were incubated with complete medium alone as negative control. After incubation for 18–20 h at 37°C and 5% CO₂ plates were washed with PBS containing 0·05% Tween-20. Biotinylated antihuman IFN-γ antibody 7-B6-1 (MABTECH) was added at 1 µg/ml and plates were incubated for 3 h at 37°C and 5% CO₂. After incubation, plates were washed and incubated further for 1 h with 1 µg/ml of streptavidin–alkaline phosphatase conjugate (Mabtech). Next, the plates were washed as before and developed with a chromogenic alkaline phosphatase substrate kit (Bio-Rad Laboratories, Hercules, CA, USA). Finally, plates were washed under running tap water and dried at room temperature. IFN-γ spot-forming cells (SPC) with fuzzy borders were enumerated using an ELISPOT reader (AID, Strassberg, Germany). For statistical analysis, we computed the mean, standard deviations (s.d.) and average plus 2 s.d. of spots from four replicates of each peptide response. A response was noted significant if P-value (t-test) was less than 0·05 and test response was more than background.

Results

Screening for novel CD8 T cell epitopes in M. tuberculosis genome

The computer algorithm HLA_BIND [27] was used to screen for the presence of potential T cell epitopes within the genome sequence of M. tuberculosis H37Rv. The algorithm calculates the half-time (t_1/2) of dissociation of peptide MHC molecules based on coefficient tables deduced from the published literature [3].

We first determined a threshold score for HLA-B*35 restricted T cell epitopes using the concatenated sequence containing all known HLA class-I restricted T cell epitopes and ligands [32]. To assess how well the HLA_BIND can predict unknown epitopes we also submitted 12 published epitopes, including six HLA-B*3501 restricted epitopes, which we have studied previously [12,16], and obtained t_1/2 values between 1 and 60 min, with a median score of 20 min. Next we used HLA_BIND to scan all 3924 translated open reading frames of M. tuberculosis H37Rv representing at least 1·4 million potential 9-mer peptides, and found 479 predicted peptides in the genome of M. tuberculosis H37Rv with a predicted t_1/2 value of ≥120 min. For the present study we selected nine peptides with the highest score (≥ 360 min), and an additional four peptides with a score of 120 min that occurred multiple times in the genome (Table 1).

Immunogenicity of predicted T cell epitopes

Ten HLA-B*35 positive BCG vaccinated healthy blood donors were screened for reactivity towards the predicted epitopes by generating CD8⁺ T cells in the presence of IL-7 and IL-2. The expanded cells were restimulated with peptide and the presence of IFN-γ-producing cells was assessed by ELISPOT assay. To all newly identified peptides we observed significant but heterogeneous responses. The frequencies of IFN-γ responding cells after stimulation varied among the donors from 80 to 1930 SFU/10⁶ PBMC. The number of peptides recognized varied per donor between 1 and 10 epitopes (Fig. 1). Most frequently recognized were peptides 105 and 106, for which we observed positive responses in 60% of individuals tested (Fig. 1). Results of the 16 individuals not shown had responses of less than 10 SFU/well and were considered not to be statistically significant.

Fig. 1.

Short-term (10-day) cultured IFN-γ ELISPOT screening for binding to HLA-B*3501 in BCG-vaccinated healthy donors. Significant responses are black bars above cut-off (horizontal dotted line), obtained from two standard deviations of unstimulated plus peptide. First six responses on the x-axis represents published epitopes obtained by conventional methods followed by 13 newly predicted epitopes obtained through genome-wide screening.

For selected peptides we were able to enumerate circulating effector T cells using pools of six predicted peptides in ex vivo ELISPOT assays for IFN-γ. The results confirmed our findings with the cultured ELISPOT approach, and observed similarly that frequencies varied considerably among different donors (data not shown). Such variation was not unexpected and probably represents a number of factors including dominance of the response in relation to other class I alleles expressed by an individual.

In order to verify the cellular source of IFN-γ we depleted CD8 T cells using MACS magnetic beads after short-term culture and stimulation with peptide. We obtained proof of principle for three selected peptides in two donors tested (Fig. 2). Upon depletion of CD8 T cells there was a significant drop in peptide-specific IFN-γ producing cells indicating that the majority of IFN-γ production we observed in our cultured ELISPOT approach is CD8 T cell-mediated.

Fig. 2.

Peptide specific IFN-γ secretion mediated by CD8 T cells. Cells were depleted by MACS CD8 beads using immunomagnetic separation methods after short-term culture and stimulation with responding individual peptides at a final concentration of 20 µm in the presence of rhIL-7 and IL-2. (a) Response to peptide 104 and 105; (b) response to peptide 108. Black bars represent PBMC, while grey bars are CD8 depleted. All donors were naturally exposed healthy BCG vaccinees typed by PCR for HLA-B*35. Each peptide was tested in four replicate wells in ELISPOT.

HLA B7-supertype restricted T cell epitopes

With the knowledge that some of our donors were also HLA-B53 and -B7 by PCR, we also tested the panel of HLA-B*35 predicted epitopes for recognition by B7-supertype positive BCG-vaccinated donors. Again the responses were heterogeneous and similar to what we observed for the group of HLA-B*35 positive individuals (Fig. 3). For simplicity, response to all epitopes of the so-called HLA-B*7 superfamily were combined into a Venn diagram (Fig. 4). Of the 13 predicted epitopes that responded to HLA-B*3501, seven were recognized permissively in both HLA-B*35 and HLA-B*53. Peptide p99 from Rv2823c could be recognized in the context of all three B7-supertype members tested. In addition, the HLA-B*35 and -B*07 groups shared recognition of peptide p47 (Fig. 4).

Fig. 3.

Screening for binding to other HLA-B alleles (HLA-B*7 and -B*53). Representative IFN-γ ELISPOT responses to healthy BCG vaccinees with binding motif to (a, b) HLA-B*53 and (c) HLA-B*07. Significant responses are above the cut-off shown as a horizontal line, obtained from 2 s.d. of unstimulated plus peptide. First six responses on the x-axis represents published epitopes obtained by conventional methods followed by 13 newly predicted epitopes obtained through genome-wide screening.

Fig. 4.

Venn diagram of promiscuous HLA-B*35 restricted CD8⁺ T cell epitopes are immunogenic across HLA-B*7 superfamily. Peptides in bold represents published epitopes obtained by conventional methods.

Discussion

We identified 13 CD8 T cell epitopes derived from a total genome-wide screen of M. tuberculosis H37Rv. The epitopes identified were all from uncharacterized mycobacterial antigens and are likely to be restricted through HLA-B*3501 or the HLA-B7 supertype family.

While T cell epitope mapping before publication of the M. tuberculosis genome tended to focus on known antigens [15], the approach we use is completely unbiased by antigen selection. We assumed that the M. tuberculosis genome, with its size and large number of predicted ORFs, would contain a myriad of T cell epitopes. Indeed, the data we present support this notion. Thus we opted to use a computer algorithm to scan the entire M. tuberculosis H37Rv genome sequence for potential HLA-B*35 restricted T cell epitopes. The HLA-B*35 class-I allele was selected because it is common in West Africa and is part of the HLA-B7 supertype family, which shares peptide binding features [17,33]. We therefore expect epitopes identified in this or similar studies to have particular value for impending major TB trials in West Africa.

The putative epitopes selected for further immunogenicity study were those with highest binding scores or at least those with scores higher than well-known HLA-B*35 epitopes. However, avidity of epitope binding may not necessarily correlate with immunogenicity. Indeed, the epitopes that bind most tightly may skew T cell responses towards a particular function. It has been suggested that natural T regulatory cells may tend to be specific for epitopes that bind particularly tightly in the HLA groove [34]. Much further work is needed to investigate the avidity of peptide binding to HLA and the subsequent immune response elicited.

Peptides responses were screened in both cultured and ex vivo ELISPOT assays using lymphocytes recovered from healthy BCG-vaccinated individuals who are also naturally exposed to environmental pathogenic and non-pathogenic mycobacteria. The data imply indirectly that the native antigens are being processed naturally and presented through a classical class I pathway, as well being capable of eliciting effector and memory CD8 T cell responses [35]. Surprisingly, the response profile between individuals was extremely diverse, indicating heterogeneity in response to individual antigens. While such heterogeneity might relate in part to the HLA type of a particular individual, the antigens themselves might be expressed differentially according to infection status, such as whether recent or latently infected [36]. Although a hierarchy of response might be expected for different antigens no obvious hierarchy was noticed at an epitope level for the new antigens we identified or those reported previously.

Another possibility is that of the many hundreds of antigens likely to be expressed by the bacillus at any one time the actual choice of which to mount a dominant immune response against might be somewhat random. Most probably, all the above factors might be involved. Regardless, the data show the CD8⁺ T cell responses to M. tuberculosis to be very broad and suggests the options for subunit vaccines against TB might be much greater than anticipated previously.

Although further testing is needed, this study is one of the first to report use of an M. tuberculosis genome-wide screen for identification of CD8 T cell epitopes. The peptides identified may be used to unfold completely uncharacterized immunogenic antigens capable of entering a class I pathway for presentation. Peptides may be used for defining the important role of CD8 T cells in protective immunity during the natural history of TB in an endemic setting such as West Africa. Peptides may also serve as monitoring tools for novel TB interventions in future subunit vaccines which harbour these epitopes [37,38]; and finally, they may be considered as vaccine candidates in peptide cocktails or in polyepitope constructs [39].