Intermediate maturation of Mycobacterium ... - Wiley Online Library

Cellular Microbiology (2007) 9(6), 1412–1425

doi:10.1111/j.1462-5822.2006.00881.x First published online 23 January 2007

Intermediate maturation of Mycobacterium tuberculosis LAM-activated human dendritic cells Nicolas Dulphy,1,2 Jean-Louis Herrmann,3* Jérôme Nigou,4 Delphine Réa,1 Nicolas Boissel,1 Germain Puzo,4 Dominique Charron,1,2 Philippe H. Lagrange3 and Antoine Toubert1,2* 1 INSERM, U662, Institut Universitaire d’Hématologie, Université Paris VII, Paris, F-75010 France. 2 Laboratoire d’Immunologie et d’Histocompatibilité, Centre d’Investigation Biomédicales, AP-HP, Hôpital Saint-Louis, Paris, F-75010, France. 3 Equipe d’Accueil EA3510, Faculté Lariboisière-Saint Louis, Université Paris VII, Paris, F-75005 France; Hôpital Saint Louis, Service de Microbiologie, Paris, F-75010, France. 4 CNRS, UMR 5089, Institut de Pharmacologie et de Biologie Structurale, Département ‘Mécanismes moléculaires des infections mycobactériennes’, Toulouse, F-31077, France. Summary Contrasting observations raise the question of the role of mycobacterial derived products as compared with the whole bacterium Mycobacterium tuberculosis on maturation and function of human dendritic cells (DCs). DC-SIGN has been identified as the key DC receptor for M. tuberculosis through its interaction with the mannosylated lipoarabinomannan (ManLAM). Although ManLAM is a major mycobacterial component released from infected antigen-presenting cells, there is no formal evidence yet for an effect of ManLAM per se on DC maturation and function. DCs activated with purified ManLAM displayed an intermediate maturation phenotype as compared with lipopolysaccharide fully matured DCs with reduced expression of MHC class I and class II molecules, CD83 and CD86 and of the chemokine receptor CCR7. They were sensitive to autologous natural killer (NK) lysis, thus behaving like immature DCs. However, ManLAM-activated DCs lost phagocytic activity and triggered priming of naive T-cells, confirming their intermediate maturation. Partial maturation of Received 24 November, 2006; accepted 27 November, 2006. *For correspondence. E-mail [email protected]; Tel. (+33) 1 42 49 46 40; Fax (+33) 1 42 49 46 41; E-mail: [email protected]; Tel. (+33) 1 47 10 79 50; Fax (+33) 1 47 10 79 49. © 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd

ManLAM-activated DCs was overcome by triggering the CD40/CD40L pathway as a second signal, which completed maturation phenotypically and abolished autologous NK lysis susceptibility. Altogether, these data provide evidence that ManLAM may induce a partial maturation phenotype on non-infected bystander DCs during infection suggesting that ManLAM released from infected cells might impair adaptive immune response towards M. tuberculosis. Introduction Interactions between antigen-presenting cells (APC) and Mycobacterium tuberculosis are key features determining the resistance of the host towards tuberculosis and further development as an active or latent disease. The first APCs encountered by the tubercle bacilli are the host alveolar macrophages (Mfs), within which they can replicate and interact with other leucocytes (such as T lymphocytes and dendritic cells, DCs). During infection, DCs are present all along the bronchial tree and throughout the lung parenchyma and play a role as sentinels (Russell, 2001). Dendritic cells are crucial members of the immune system, playing a finely tuned role in both adaptive immunity by the priming of pathogen-specific T lymphocytes, and innate immunity through cross-talk with natural killer (NK) cells (Degli-Esposti and Smyth, 2005). DCs, depending on their maturation status, influence the quality and the magnitude of the T-cell response (Martin-Fontecha et al., 2003). Indeed, immature DCs (iDCs) in peripheral tissues are highly phagocytic cells but are poorly efficient in antigen processing and T-cell priming. They express low levels of MHC class I and class II molecules. The costimulatory molecules CD80 and CD86 as well as the activation marker CD83 are poorly expressed. They also present pathogen-recognition receptors including the C-type lectins DC-SIGN (CD209) and mannose receptor (CD206) (Cambi and Figdor, 2003) and different Toll-like receptors (TLR) depending on the DC subpopulations (Iwasaki and Medzhitov, 2004). DC maturation is stimulated by the detection of an invading pathogen, cell necrosis or cytokine release (Guermonprez et al., 2002). Endocytic capacity is then downregulated, and MHC class I and class II molecules as well as CD80 and CD86 are overexpressed. DCs also acquire the expression of the

ManLAM activation of human DCs 1413 lymph node (LN)-homing chemokine receptor CCR7 and migrate to secondary lymphoid organs where they prime naive T-cells (Guermonprez et al., 2002). In addition, DCs are able to activate innate immunity by reciprocal interactions with NK cells at the site of inflammation or in secondary lymphoid organs (Moretta, 2002; Ferlazzo et al., 2004; Degli-Esposti and Smyth, 2005). Mature monocyte-derived DCs can activate resting NK cells, which then proliferate, produce cytokines (such as IFNg) and acquire cytolytic activity (Borg et al., 2004). In turn, NK cells are able to induce maturation of iDCs (Gerosa et al., 2002; Piccioli et al., 2002). In some cases, NK cells are also able to negatively regulate the pool of DCs by killing iDCs. Mature DCs are protected against such killing by virtue of the increased expression of the MHC class I molecules which are ligands of the inhibitory Killer Immunoglobulin-like receptors (KIRs) present on NK cells (Ferlazzo et al., 2002). Mycobacterium tuberculosis is able to infect DCs through the ligation of the C-type lectin receptor DCsSIGN (Geijtenbeek et al., 2003; Tailleux et al., 2003a) and can then persist within. Contrasting with infected Mfs, M. tuberculosis in DC is confined to an early endosomelike compartment, disconnected from cellular biosynthetic or recycling pathways (Tailleux et al., 2003b). Whether M. tuberculosis replicates inside DCs is still a matter of debate although it has been shown that several other intracellular pathogens like Mycobacterium bovis BCG or Listeria monocytogenes (Pron et al., 2001; Jiao et al., 2002) do not replicate. Ligation of M. tuberculosis to DC-SIGN is likely to be mediated by the mannose-capped lipoarabinomannan (ManLAM) (Maeda et al., 2003; Koppel et al., 2004; Pitarque et al., 2005). ManLAM is the surface component specific to slow-growing mycobacteria such as M. tuberculosis or the vaccine strain M. bovis BCG. In contrast, the fast-growing saprophytic mycobacteria Mycobacterium smegmatis or Mycobacterium fortuitum present a phosphoinositide-capped lipoarabinomannan (PILAM) which poorly binds to DC-SIGN (Maeda et al., 2003; Pitarque et al., 2005). It is now clear that mycobacterial antigens, including ManLAM, are released and delivered from infected Mfs to non-infected bystander Mfs or DCs (Beatty et al., 2000; 2001; Schaible et al., 2000; 2003; Fischer et al., 2001; Neyrolles et al., 2001). The release of ManLAM from infected Mfs is mediated by at least two described mechanisms. On the one hand, lipid-containing moieties of the mycobacterial cell wall have been shown to actively traffic out of the mycobacterial vacuole, transported to endosomes/lysosomes and ultimately found in extracellular vesicles isolated from the culture medium (Beatty et al., 2000; 2001; Fischer et al., 2001). On the other hand, it has been recently reported that mycobacteria induce apoptosis of the Mfs they infect, causing the

release of apoptotic vesicles carrying ManLAM (Schaible et al., 2003). An impact of ManLAM on human Mfs function (Nigou et al., 2001) or on lipopolysaccharide (LPS)induced DC maturation has been suggested (Nigou et al., 2001; Geijtenbeek et al., 2003) but there is not yet formal evidence for an effect of ManLAM per se on DC maturation and function. These observations raise an important question as to the effects of the whole bacterium as compared with mycobacterial derived products like ManLAM, on APCs maturation and function. By setting up a model of infection to differentiate between infected versus non-infected DCs using GFPexpressing M. tuberculosis H37Rv bacteria (Tailleux et al., 2003b), we analysed the maturation phenotype of infected and non-infected bystander DCs. This phenotype was detailed with purified ManLAM in contact with monocyte-derived DCs. Functional consequences on NK cell lysis and T-cell priming were assessed in an autologous setting. The observations support the hypothesis that ManLAM, reproducing the phenotype of bystander and non-infected DCs, could be a key factor in the deletion of bystander DCs at the infection site and consequently result in a deficient immune response during M. tuberculosis infection.

Results Effect of M. tuberculosis infection and purified ManLAM on DC phenotypes Immature DCs were obtained from healthy donor monocytes grown in GM-CSF and interleukin (IL)-4. In these conditions, the preparation was reproducibly at least 98% pure according to CD14 and CD1a stainings. To study the maturation phenotype of DCs in M. tuberculosis infection, we used GFP-expressing M. tuberculosis H37Rv to distinguish between infected (GFP-positive) and noninfected (GFP-negative) cells. iDCs were cultured for 3 h with GFP-expressing H37Rv and extracellular bacteria were then removed. Cell surface staining of HLA class I and class II molecules (HLA-I, -II), CD83, CD86 and CCR7 were used as markers of DC maturation during a 48 h post-infection period (Fig. 1). Upregulation of CD86, HLA-I and -II molecules occurred for DC with integrated GFP-positive mycobacteria 3 h after the beginning of infection (Fig. 1). Expression levels were delayed for all markers in non-infected bystander GFP-negative DCs. At 48 h, a similar expression was achieved for HLA-I, -II and CD86 whereas CD83 and CCR7 still remained at a lower level in non-infected compared with infected DCs raising the issue of an incomplete maturation of bystander noninfected GFP-negative DCs. ManLAM is known to be a key factor in M. tuberculosis–DC interactions through its receptor

© 2007 The Authors Journal compilation © 2007 Blackwell Publishing Ltd, Cellular Microbiology, 9, 1412–1425

Fig. 1. Kinetic of DC maturation markers on infected and non-infected DCs. GFP-positive M. tuberculosis H37Rv bacteria and DCs were left in contact for 3 h and infected (GFP-positive) and non-infected (GFP-negative) DCs were monitored for up to 48 h in flow cytometry analysis to investigate DC maturation. Levels of expression of HLA-I and -II molecules, CD83, CD86 and CCR7 were analysed on the two DC populations. After 48 h, non-infected (GFP-negative) DCs showed a reduced expression of CCR7 and CD83 compared with GFP-positive infected cells. One out of five experiments is shown.

1414 N. Dulphy et al.


ManLAM activation of human DCs 1415

Fig. 2. ManLAM binding on DC. A. GFP-expressing H37Rv/DCs coculture experiments with the same settings as in Fig. 1. ManLAM was detected on DCs with a ManLAM-specific monoclonal antibody after 48 h of infection. GFP-positive infected cells showed a high surface detection of ManLAM but GFP-negative non-infected DCs presented also ManLAM on their membrane. B. Fixation of LAM molecules and DC-SIGN cell surface expression were detected by specific monoclonal antibodies with increasing doses of ManLAM or PILAM. Increasing H37Rv-purified ManLAM concentrations reduced the detection of DC-SIGN concurrently to an increase in the detection of ManLAM at DC surface. C. Binding of soluble purified ManLAM and DC-SIGN detection on iDC quantified by flow cytometry in a 72 h DC culture. ManLAM is compared with medium (t = 0 h) or PILAM as negative controls.

DC-SIGN on the surface of DC (Geijtenbeek et al., 2003; Tailleux et al., 2003a). Together with other mycobacterial compounds, ManLAM is present in extracellular vesicles released from mycobacteria-infected APCs (Beatty et al., 2000; 2001; Fischer et al., 2001). It was considered important to evaluate whether ManLAM was captured by bystander and non-infected DCs. We looked at the presence of ManLAM on infected and non-infected DC surface with anti-ManLAM antibody. ManLAM was detected 48 h after infection not only on GFP-positive infected-DCs but also on non-infected DCs after staining on the cell surface with ManLAM-specific antibodies (Fig. 2A). In experiments using purified soluble ManLAM, we observed a dose-dependent decrease of the DC-SIGN signal on DC surface conversely related to ManLAM detection on

the cell surface. By contrast, soluble PILAM from M. smegmatis was detectable on DC surface only at very high concentrations of the soluble product and did not affect the DC-SIGN expression level (Fig. 2B). ManLAM association to DC-SIGN on iDCs was progressive and reached a maximum at 48 h together with the lowest DC-SIGN detection (Fig. 2C). Although we cannot formally exclude that ManLAM binding to DC-SIGN may compete with the DC-SIGN antibody staining, these observations underline the interactions between ManLAM and DC-SIGN (Geijtenbeek et al., 2003; Maeda et al., 2003; Tailleux et al., 2003a). As ManLAM was present at the surface of non-infected bystander DCs after M. tuberculosis infection, we further investigated whether the phenotype observed for non-


1416 N. Dulphy et al. infected bystander DCs could be observed with purified M. tuberculosis H37Rv ManLAM. The maturation process engaged by ManLAM was delayed and incomplete as compared with LPS. ManLAM-mediated maturation was observable after 24 h whereas CD83 overexpression was induced by LPS after 3 h of incubation only. By contrast, DCs stimulated with PILAM did not show any phenotypic change as compared with iDCs (data not shown). At 48 h, we found that purified ManLAM-activated DCs expressed a reduced level of HLA-I molecules as well as CD83 with a minor decrease of CD86 and HLA-II in comparison with LPS fully matured cells (Fig. 3A). ManLAM-activated DCs did not express the chemokine receptor CCR7 after 2 days of incubation (data not shown). Based on CD83 and CD86 expressions, we observed a dose-dependent effect of ManLAM on DC maturation (Fig. 3B). However, even at 10 mg ml-1 of ManLAM in the culture medium a full maturation was not achieved at 48 h compared with LPS at 10 ng ml-1 (Fig. 3B and data not shown). Interestingly, the addition of LPS in the culture medium together with increasing doses of ManLAM resulted in a terminal maturation phenotype of DC for all settings tested (Fig. 3B and data not shown). This indicates that in these experimental conditions ManLAM does not affect the maturation signal provided by LPS to iDCs. In conclusion, non-infected bystander DCs showed differences in maturation as compared with infected DCs predominantly by a reduced expression of CD83 and CCR7 and a delayed expression of antigen-presenting of HLA-I molecules. A partial maturation phenotype of DCs was observed with purified M. tuberculosis H37Rv ManLAM which is one of the bacterial components released after cell infection. These results led us to further study the effect of purified ManLAM on DC functions in both innate and adaptive immunity. Functional consequences of the intermediate DC maturation induced by soluble H37Rv-purified ManLAM Immature DCs undergo rapid maturation when in contact with different stimuli such as bacteria. Mature DCs release activating cytokines, decrease phagocytic activity, interact with NK cells and activate naive T-cells after migration to LN. Any maturation defect induced by ManLAM might result in an impairment of one of these properties. Indeed, interactions between DCs and NK or T-cells depend upon the maturation status of DCs: iDCs are sensitive to NK killing and are inefficient in the priming of naive-T cell responses. In contrast, mature DCs become resistant to NK cytotoxicity as a consequence of MHC class I overexpression through the interaction with inhibitory NK receptors such as KIRs (Ferlazzo et al., 2002) and are fully prepared to prime naive T-cells. These different properties were analysed on DCs stimulated with soluble ManLAM

comparing responses to iDCs (with medium alone or in the presence of PILAM) and to fully matured LPSstimulated DCs. Antigen uptake activity was investigated using FITClabelled BSA. In contrast with immature or PILAMstimulated DCs but comparably to LPS-matured DCs, ManLAM-activated DCs were unable to capture antigens (Fig. 4). Regarding cytokine production, ManLAMactivated DCs produced the proinflammatory cytokines IL-6 and IL-8 but at a lower amount compared with fully matured cells (Fig. 5). In contrast, they did produce strongly reduced amounts of IL-10 and the functional IL-12p70 heterodimer as previously described (Nigou et al., 2001; Geijtenbeek et al., 2003). TNFa secretion reached a plateau 3 h after LPS-mediated stimulation of iDCs but was strongly decreased or absent with ManLAMactivated DCs 48 h after stimulation (Fig. 5). Type I IFN were absent in all settings (data not shown). Interestingly, extremely low levels of TNFa, IL-10 and IL-12 after 48 h ManLAM-stimulation were observed in all samples (n = 6). To take into account the variations in raw data observed between DC preparations, we calculated the ratio of cytokines produced by ManLAM-activated DCs relatively to LPS for each sample. All cytokines except IL-8 were significantly reduced compared with LPS (P < 0.0001). We thus addressed the issue of the functional consequences of ManLAM-activated DCs interaction with T and NK cells. We set up an autologous killing assay using monocyte-derived DCs as a target for IL-2-activated NK cells (Fig. 6A). As expected, LPS-activated DCs became resistant to NK lysis, in keeping with a fully matured phenotype. By comparison, ManLAM-activated DCs were lysed by NK cells, as efficiently as iDCs or PILAMactivated DCs. We thus concluded that ManLAMactivated DCs were sensitive to autologous NK lysis, in concordance with their lower level of HLA-I expression as seen in Fig. 3A. In another set of experiments, the interaction of ManLAM-activated DCs with CD8+ T lymphocytes was investigated. CD8 T-cell responses offer the opportunity to easily quantify antigen-specific immune responses by using HLA-I tetramers and to compare the ability of DCs to expand T-cells in various conditions of maturation. The priming of naive CD8+ T lymphocytes was investigated using Melan-A as a model. Melan-A is a tumour antigen which includes the HLA-A*0201-restricted immunodominant peptide 26–35 (Romero et al., 1997). Melan-A-specific T-cells with the characteristics of a naive population are detected with HLA-I tetramers in the peripheral blood of healthy individuals (Zippelius et al., 2002) and can be efficiently primed only by professional APCs such as mature DCs (Salio et al., 2001; Zippelius et al., 2002). To analyse memory T-cell


Fig. 3. Effect of ManLAM on DC maturation phenotype. A. Phenotyping of DCs stimulated with H37Rv-purified ManLAM as compared with non-stimulated (medium) or LPS fully matured DCs. HLA-I and -II molecules as well as CD83 and CD86 expression was reduced on ManLAM-activated DCs by contrast to LPS fully matured DCs. The LN addressing chemokine receptor CCR7 was absent on ManLAM-activated DCs (data not shown). One out of six experiments is shown. B. CD83 and CD86 cell surface expression in median of fluorescence observed with (open bars) or without LPS (10 ng ml-1, black bars) and increasing doses of ManLAM (from 0 to 10 mg ml-1). ManLAM did not abrogate the maturation induced on DCs by a low dose of LPS. Increasing doses of ManLAM enhanced the maturation profile of DCs.



1418 N. Dulphy et al. Melan-A-specific naive T-cells were expanded by ManLAM-activated DCs, although less efficiently than by LPS-matured DCs in all experiments performed. They were not efficiently primed by iDCs or PILAM-DCs as controls. In summary, ManLAM-activated DCs shared some features with mature DCs such as a lower antigen uptake, the ability to prime naive autologous T-cells and to produce inflammatory cytokines although markedly less efficiently than fully mature DCs. In contrast, ManLAMactivated DCs were still sensitive to NK lysis. Taken together, our results support the hypothesis that ManLAM-activated DCs have a partial or intermediate maturation profile by contrast to LPS-matured or M. tuberculosis-infected DCs. This leads us to ask the question of how a second signal might complete DC maturation and functional responses.

A second signal fulfils complete ManLAM-activated DC maturation

Fig. 4. Phagocytic activity of ManLAM-activated DCs. Phagocytic activity of DCs after treatment was quantified by the capture of FITC-labelled BSA with or without bacterial products. ManLAM-activated or LPS-matured DCs did not phagocyte FITC-labelled BSA at 37°C by contrast to iDCs or PILAM-DCs. One out of three experiments is shown.

responses, we used the CMVpp65495-503 peptide presented by HLA-A*0201 (Gillespie et al., 2000). We stimulated HLA-A*0201 T-cells with peptide-loaded DCs in an in vitro autologous assay and compared priming of naive T-cells or stimulation of memory T-cells according to the DC maturation status (Fig. 6B). CMVpp65-specific memory T-cells were expanded by DCs in all experimental conditions with a comparable effect. By contrast,

We postulated that T-cells attracted to the site of inflammation, including activated CD4+ T lymphocytes expressing CD154 (so-called CD40L), could mediate DC maturation through CD40 ligation, a signal which more closely mimics an in vivo situation of M. tuberculosis infection than LPS. Activation of DCs by ManLAM induced the overexpression of CD40 as efficiently as LPS (Fig. 7A). We tested whether a combination of ManLAM and soluble trimeric CD40L (sCD40L) could induce a complete maturation of DCs. To this end, sCD40L was used in suboptimal conditions to evaluate a synergistic effect with ManLAM (Fig. 7B). The addition of sCD40L to ManLAM-activated DCs induced expression of HLA molecules, CD83 and CD86 to levels similar to DC maturation using LPS and consistent with a terminal DC maturation. A cytotoxic assay with autologous NK lymphocytes confirmed that DCs in the presence of ManLAM and sCD40L became resistant to NK lysis in agreement with the complete maturation of the targets (Fig. 7C). These observations demonstrate that a signal provided through the CD40/CD40L pathway is not inhibited by ManLAM binding and may to the opposite sustain the final maturation of ManLAM-DCs.

Discussion Mycobacterium tuberculosis is a highly successful pathogen, with a low infectious dose and a huge human reservoir. Under normal circumstances, infection is controlled by the immune system of the host, and in 90% of the cases, the infection does not give rise to overt


ManLAM activation of human DCs 1419 Fig. 5. Cytokines production of ManLAM-activated DCs. IL-6, IL-8, IL-10, IL-12p70 and TNF-a were quantified in supernatants of DCs cultured for 48 h with or without bacterial products (medium 䉭; LPS 䊐; ManLAM 䊉; PILAM 䉫). Results show the kinetics of cytokine production (in pg ml-1) as a mean of three samples at 0 h, 3 h and 24 h and a mean of six samples at 48 h. The histograms show cytokine amounts produced by ManLAM- (open bars) or PILAM-DCs (black bars) at 48 h as percentages relative to amounts secreted by LPS-activated DCs ⫾ SD. Statistical analysis was performed between LPS and ManLAM with the Student’s t-test, *P < 0.0001.

disease. Despite control of infection, the immune response is stopped prior to the elimination of remaining bacteria, leaving the host with a 10% lifetime risk of developing active disease. This persistence happens despite both innate and adaptive immunity, implying that an ‘immune status quo’ has been established after the acute response (Young et al., 2002). Our current understanding has established a crucial role for DCs at the interface between innate and adaptive immune responses. An efficient early response is clearly essential in shaping the subsequent adaptive response, by triggering expansion of T-cell populations, or by activating NK cells as part of a protective response. By examining in vitro the pathogen–DCs interactions, we have gained a clearer understanding that not all the DCs are infected with the pathogen (see Fig. 1). Most assays of cellular activation measure changes in bulk properties within cultures rather than individual cells according to their infection status. Our infection model based on the individual interactions between bacteria and/or bacterial products and DC required the setting of precise experimental conditions. Mycobacteria were cultured and prepared before infection in Tween-free medium in order to maintain mycobacterial cell wall integrity (Glickman et al., 2000; Tailleux et al., 2003b). DCs were differentiated from monocytes purified according to the expression of the marker CD14 and after culture iDC were defined as CD14– CD1a+ populations. This criterion ensured a bona fide homogeneous DC population was studied as compared with Mfs-like populations (see controls Figs 1 and 2B). In this way, experiments using M. tuberculosis H37Rv expressing GFP revealed that bystander non-infected DCs have a partial matured phenotype. The whole bacterium released several key

compounds after infection as previously described (Herrmann et al., 1996; Beatty et al., 2000; 2001; Fischer et al., 2001), which ultimately lead to the delayed but maximal increase in HLA-I, -II and CD86 molecules. However, at no time the increases in CD83 and CCR7 were induced at similar levels on bystander non-infected DCs as compared with infected DCs. This phenotype was also more strikingly illustrated with purified ManLAM, which was detected at the surface of non-infected DCs throughout a 48 h infection (Fig. 2A). ManLAM is one of the main mycobacterial components present at the surface of the mycobacterium playing a key role in DCs–M. tuberculosis interaction. ManLAM has been shown to be a ligand for DC-SIGN, a DC-specific lectin binding receptor, allowing M. tuberculosis entry into DCs (Tailleux et al., 2003a). It has also been shown to be actively released from infected APCs and transferred to non-infected bystander APCs (Beatty et al., 2000; 2001; Schaible et al., 2000; 2003; Fischer et al., 2001). In addition, it is now clear that direct cell to cell interactions through nanotubules network are responsible for the transmission of signals between cells (Watkins and Salter, 2005) and such observations may be relevant for ManLAM uptake by non-infected bystander DCs in our model. The incomplete maturation of DCs was matched by a reduced expression of surface markers such as HLA and costimulatory molecules, and by sensitivity to NK lysis. ManLAM-activated DCs produced small amounts of inflammatory cytokines (essentially IL-6 and IL-8), did not induce the chemokine receptor CCR7 cell surface expression, and were unable to phagocytose antigens. Functionally, ManLAM-activated DCs were able to prime or boost efficient antigen-specific CD8 T-cell proliferation in the presence of synthetic peptide despite the reduced expres-


1420 N. Dulphy et al.

Fig. 6. Interaction of ManLAM-activated DCs with NK and CD8+ T lymphocytes. A. Sensitivity of DCs to NK lysis: iDCs were cultivated for 48 h with or without bacterial products and tested in a 4 h 51Cr release assay with autologous IL-2-activated NK cells. ManLAM- or PILAM-DCs were susceptible to NK lysis as iDCs (medium). By contrast LPS-matured DCs were resistant to NK cytotoxicity. One experiment out of three is shown. B. DC-induced CD8+ T-cell antigen-specific responses. In HLA-A*0201 healthy donors, Melan-A peptide was detected by naive T-cells. Melan-A-specific naive T lymphocytes were expanded by LPS-matured DCs and ManLAM-activated DCs. CMVpp65 was used to study the expansion of memory T lymphocytes. All experimental settings induced the expansion of CMVpp65-specific memory populations. One experiment out of three is shown.

sion of HLA-I and costimulatory molecules such as CD86. Sensitivity to NK lysis together with the ability to prime peptide-specific naive T-cells illustrates the incomplete DC maturation profile in the presence of ManLAM. An initial consequence of the interaction between soluble ManLAM and DCs could be a defective M. tuberculosis-specific T-cell-mediated response due to the poor expression of CCR7 and subsequently an impaired migration of ManLAM-activated DCs to LNs. The inhibition of phagocytosis might also participate in this process. Indeed, bystander DCs in contact with released mycobacterial products would be deficient in phagocyto-

sis of apoptotic or necrotic bodies. In this way, ManLAMactivated DCs would become inefficient at carrying out bacterial antigen presentation at the infection site and in the draining LN, adding to the defect of low expression of HLA and costimulatory molecules. ManLAM-activated DCs may participate in the recruitment of NK cells at the infection site through the release of IL-6 and IL-8 in an environment already enriched by cytokines produced by infected Mfs and DCs (Giacomini et al., 2001). This inflammatory environment might attract the CXCR-1-positive perforin-enriched NK cells (Chiesa et al., 2005) to the site of infection. This, combined with



Fig. 7. CD40L/CD40 stimulation fulfils ManLAM-activated DC maturation. A. CD40 expression on DCs surface upon ManLAM stimulation as observed with LPS. B. Phenotypes of DCs population stimulated by suboptimal dose of sCD40L, LPS or ManLAM for 48 h were compared with DCs cultivated for 24 h with ManLAM and then 24 h with suboptimal dose of sCD40L. The addition of sCD40L on ManLAM-activated DCs completed DC maturation as compared with LPS-matured DCs. ManLAM or sCD40L alone were not able to fully mature DCs. C. DCs incubated with the same settings than in B were tested for their sensitivity to autologous NK cytotoxicity. Lysis was totally abrogated after DCs were stimulated with ManLAM + sCD40L by contrast to sCD40L or ManLAM alone, or PILAM + sCD40L.

the sensitivity of DCs to NK lysis, might as a consequence lead to a depletion of ManLAM-activated DCs. Interestingly, ManLAM-DCs were not able to produce type I IFN nor IL-12. These results are in accordance with previously published data describing an autocrine loop with type I IFN in the regulation of bioactive IL12p70 secretion by myeloid DCs (Gautier et al., 2005). This absence of IL12p70 might inhibit the development of a protective immune response (Nigou et al., 2001; Geijtenbeek et al., 2003). We confirmed here with purified ManLAM the results obtained by Buettner et al. suggesting that the defect in TNF-a production would prevent maturation of infected DCs (Buettner et al., 2005). Taken

together, these observations might favour the persistence of quiescent mycobacteria in the granuloma and the establishment of an immune ‘status quo’. DC–T lymphocyte interaction might induce a final maturation of ManLAM-activated DCs via the CD40/CD40L pathway, as demonstrated in vitro by the surface phenotype and resistance to NK lysis. This has to be interpreted together with data describing an inhibition by ManLAM of the LPS/TLR4-mediated DC maturation and IL-12 production (Nigou et al., 2001; Geijtenbeek et al., 2003). This suggests that ManLAM/DC-SIGN interaction may be able to inhibit only some of the various pathways inducing DC maturation.


1422 N. Dulphy et al. These results indicate that soluble ManLAM abundantly produced during M. tuberculosis infection might play a key role in the modulation of the immune response by affecting bystander DC maturation and function. ManLAM interaction with bystander DCs might at the end result in exhaustion of partially mature DCs by NK lysis and an inability of these APCs to efficiently capture antigens and migrate to LN. In contrast, infected DCs would fully mature but retain the ability to suppress intracellular bacterial growth (Tailleux et al., 2003b). As a final consequence, this might lead progressively to a comfortable niche for M. tuberculosis, including persistence in infected DCs migrating to LN, and the loss of DCs at the infection site. The global effect would be the downregulation of the inflammatory and protective immune responses within the granuloma.

Experimental procedures Cells and bacteria Peripheral blood mononuclear cells were isolated from freshly collected blood samples obtained from healthy voluntary blood donors (Blood Bank, Hôpital Saint-Louis, Paris, France) by density gradient centrifugation using a lymphocyte separation medium (Eurobio, Les Ulis, France). Monocytes were purified by positive selection using anti-CD14-coated magnetic microbeads (Miltenyi Biotec, Bergish Gladsbach, Germany). Sorted cells were > 98% CD14+ as assessed by flow cytometry staining. Monocytes were differentiated into DCs for 7 days in DC-medium described as RPMI 10% FCS medium supplemented with 800 UI ml-1 GM-CSF and 1000 UI ml-1 IL-4 (R&D Systems, Abingdom, UK). Fresh DC-medium is added to culture at day 4. For assays using stimulating products, at day 7, cells are harvested and cultured for two additional days in fresh culture medium supplemented with either E. coli LPS (250 ng ml-1, Sigma, Saint-Louis, MI, USA), ManLAM (5 mg ml-1), PILAM (5 mg ml-1) or sCD40L [200 ng ml-1, provided by Amgen (Berard et al., 2000)]. Natural killer cells were sorted after monocyte purification by negative selection using the NK cell isolation kit II from Miltenyi Biotec. Sorted cells were > 98% CD56+. NK cells were cultured for 7 days in RPMI 10% FCS with IL-2 (150 UI ml-1) before use in chromium-release assays. The GFP-expressing strain of M. tuberculosis H37Rv was generated by transformation with the GFP-encoding plasmid pEGFP and propagated in medium supplemented with 50 mg ml-1 hygromycin B (Roche Diagnostics, Mannheim, Germany) as previously described (Tailleux et al., 2003a).

Bacterial viability was above 90% (BacLigth; Molecular Probes). Before infection, bacteria were washed three times in fresh RPMI without serum and passed through a needle syringe to remove clumps as previously described (Tailleux et al., 2003b). Although we cannot exclude minute amount of mycobacterial compounds released during cell lysis, extensive washings before and after infection and low moi allow to minimize this carry-over of mycobacterial compounds. Mycobacterial suspension is mixed for a 3 h coculture with DCs in DC-medium. At the end of incubation, DCs are washed three times in RPMI in order to remove extracellular bacteria and kept in culture for up to 48 h in DC-medium.

Lipoarabinomannan purification Lipoarabinomannans were purified from M. tuberculosis H37Rv and M. smegmatis mc2 155, as previously described (Nigou et al., 1997; Ludwiczak et al., 2002). Briefly, cells were delipidated by several extractions with CHCl3/CH3OH (1:1, v/v). Delipidated cells were disrupted by sonication. Lipoglycans were further extracted by refluxing the broken cells in 50% ethanol at 65°C. Contaminating proteins and glucans were removed by enzymatic degradation using proteases and a-amylase treatments followed by dialysis. The resulting extract was submitted to hydrophobic interaction chromatography onto an octylsepharose CL-4B column, allowing the separation of glycans and lipoglycans. The resulting lipoglycans, LAM, lipomannans and phosphatidyl-myo-inositol mannoside, were separated according to their size by gel permeation using a Bio-Gel P-100 column. The endotoxin content of the LAM preparations was < 30 pg mg-1 of LPS as measured in both a chromogenic Limulus lysate assay (Cambrex, Verviers, Belgium) and a LPS detection assay based on the activation of TLR4 (Cayla/Invovogen, Toulouse, France).

Antibodies and immunofluorescence analysis Immunofluorescence analysis used the FITC-, PE- or PerCPlabelled monoclonal antibodies specific for: HLA-I molecules, HLA-DR, CD14, CD40, CD80, CD83 and CD86 produced by BD Biosciences Pharmingen (San Diego, CA, USA). The unlabelled anti-DC-SIGN antibody was provided by R&D Systems. The unlabelled CS35 anti-LAM antibody was a gift from Dr J.S. Spencer (Colorado State University, USA). For the immunofluorescence analysis on infected DCs, we used the unlabelled antiHLA-I molecules W6/32 produced in the laboratory (Brodsky and Parham, 1982). The secondary PE-labelled goat anti-mouse IgG antibody was provided by Jackson ImmunoResearch (West Grove, PA, USA). At least 2000 cells were acquired on a BD Biosciences LSR independently of the infection status. Usually, up to 15% of DC were infected according to the GFP signal (i.e. 300 infected cells acquired for 2000 total cell events).

Tetramer production Infection of DCs For infection assay, human DCs were harvested and resuspended in fresh culture medium containing or not M. tuberculosis H37Rv expressing GFP at 1 to 1 multiplicity of infection (moi) as previously described (Tailleux et al., 2003b). Briefly, frozen aliquot of M. tuberculosis H37Rv was resuspended in a 30 ml volume of 7H9 glycerol, without Tween, 24 h before infection.

Soluble MHC-peptide tetramers were produced as previously described (Altman et al., 1996). Briefly, recombinant soluble part of HLA-A*0201 (Gerry Gillespie and Chris Hourigan, Weatherall Institute of Molecular Medicine, University of Oxford, UK), modified by the addition of the birA substrate sequence, and b2-microglobulin were produced in Escherichia coli. Folding of the monomeric MHC was realised by mixing together the


ManLAM activation of human DCs 1423 b2-microglobulin, the heavy chain and the appropriate synthetic peptide (Epytop, Nimes, France). Soluble MHC complexes were first purified by gel filtration on Superdex 75 (Pharmacia, Guyancourt, France) and biotinylated overnight using birA enzyme (Avidity, Denver, USA). Biotinylated complexes were further purified using gel filtration again and anion exchange Resource Q column (Amersham Pharmacia Biotech, Orsay, France). Tetramerization was realized by adding gradually tetramer grade PE-labelled Streptavidin (Molecular Probes, Leiden, Netherlands) to the MHC at a 1:4 molecular ratio.

Endocytosis assay The ability of DCs to capture antigen was determined by measuring uptake of FITC-labelled BSA (Sigma, St Quentin Fallavier, France). Cells (5 ¥ 105 per sample) were pre-incubated in RPMI1640 medium for 15 min at 4°C or 37°C, before addition of FITC-BSA (50 mg ml-1) for 30 min. After incubation, cells were washed twice with cold PBS and fixed with PBS 2% paraformaldehyde. Uptake of FITC-BSA was then determined by flow cytometry.

Cytotoxic assay Dendritic cells used as targets were washed twice to remove all bacterial products remaining in culture medium. Targets were then loaded with 51Cr for 90 min at 37°C in RPMI supplemented with 10% FCS. Targets were distributed into a 96 well plate U bottom at a concentration of 5000 cells per well. Effector NK cells were resuspended in RPMI-10% FCS and distributed on targets according to the E/T ratio. After 4 h of incubation at 37°C, 51Cr release was measured in supernatants. Spontaneous 51Cr release was evaluated with targets incubated without NK cells and Maximum release was obtained by lysis of targets in 2% Triton X-100 (Merck Eurolab, Briare Le Canal, France).

Peptide-specific stimulation Mature and immature DCs were washed three times in RPMI1640 and incubated in 500 ml at 37°C for 2 h with or without peptide (5 mM). Peptides (Epytop, Nimes, France) used in experiments were CMVpp65495-503 [NLVPMVATV (Gillespie et al., 2000)] and the modified epitope Melan-A/MART-126-35 with an improved affinity for HLA-A*0201 [ELAGIGILTV (Romero et al., 1997)]. After one wash, DCs were cultivated with autologous PBL at a target/effector ratio of 1/5 for 10 days in RPMI 10% human serum supplemented with IL-2 (10 UI ml-1). At the end of the culture, cells were harvested and stained for flow cytometry analysis with FITC-labelled anti-CD8 antibody, PE-labelled peptide-specific tetramer and propidium iodide to exclude dead cells.

Cytokines quantification Immature DCs were stimulated with or without bacterial products and supernatants were harvested at 0 h, 3 h, 24 h and 48 h. IL-6, -8, -10, -12p70 and TNF-a were quantified using the Cytometric Bead Array inflammation kit (BD Biosciences, San Diego, CA)

according to the manufacturer’s instructions. IFN-a and IFN-b were quantified by ELISA with kits provided by PBL Biomedical Laboratories (Piscataway, NJ).

Acknowledgements We warmly thank R. Steinman (The Rockefeller Institute, New York), B. Marshall (Southampton University Hospital, Southampton) for carefully reading the manuscript and helpful discussions. The antibody CS35 was produced through funds from the National Institutes of Health, National Institute of Allergy and Infectious Diseases, Contract NO1-AI-75320, entitled ‘Tuberculosis Research Materials and Vaccine Testing’.

References Altman, J.D., Moss, P.A., Goulder, P.J., Barouch, D.H., McHeyzer-Williams, M.G., Bell, J.I., et al. (1996) Phenotypic analysis of antigen-specific T lymphocytes. Science 274: 94–96. Beatty, W.L., Rhoades, E.R., Ullrich, H.J., Chatterjee, D., Heuser, J.E., and Russell, D.G. (2000) Trafficking and release of mycobacterial lipids from infected macrophages. Traffic 1: 235–247. Beatty, W.L., Ullrich, H.J., and Russell, D.G. (2001) Mycobacterial surface moieties are released from infected macrophages by a constitutive exocytic event. Eur J Cell Biol 80: 31–40. Berard, F., Blanco, P., Davoust, J., Neidhart-Berard, E.M., Nouri-Shirazi, M., Taquet, N., et al. (2000) Cross-priming of naive CD8 T cells against melanoma antigens using dendritic cells loaded with killed allogeneic melanoma cells. J Exp Med 192: 1535–1544. Borg, C., Jalil, A., Laderach, D., Maruyama, K., Wakasugi, H., Charrier, S., et al. (2004) NK cell activation by dendritic cells (DCs) requires the formation of a synapse leading to IL-12 polarization in DCs. Blood 104: 3267–3275. Brodsky, F.M., and Parham, P. (1982) Monomorphic antiHLA-A,B,C monoclonal antibodies detecting molecular subunits and combinatorial determinants. J Immunol 128: 129–135. Buettner, M., Meinken, C., Bastian, M., Bhat, R., Stossel, E., Faller, G., et al. (2005) Inverse correlation of maturity and antibacterial activity in human dendritic cells. J Immunol 174: 4203–4209. Cambi, A., and Figdor, C.G. (2003) Dual function of C-type lectin-like receptors in the immune system. Curr Opin Cell Biol 15: 539–546. Chiesa, M.D., Sivori, S., Castriconi, R., Marcenaro, E., and Moretta, A. (2005) Pathogen-induced private conversations between natural killer and dendritic cells. Trends Microbiol 13: 128–136. Degli-Esposti, M.A., and Smyth, M.J. (2005) Close encounters of different kinds: dendritic cells and NK cells take centre stage. Nat Rev Immunol 5: 112–124. Ferlazzo, G., Tsang, M.L., Moretta, L., Melioli, G., Steinman, R.M., and Munz, C. (2002) Human dendritic cells activate resting natural killer (NK) cells and are recognized via the NKp30 receptor by activated NK cells. J Exp Med 195: 343–351.


1424 N. Dulphy et al. Ferlazzo, G., Pack, M., Thomas, D., Paludan, C., Schmid, D., Strowig, T., et al. (2004) Distinct roles of IL-12 and IL-15 in human natural killer cell activation by dendritic cells from secondary lymphoid organs. Proc Natl Acad Sci USA 101: 16606–16611. Fischer, K., Chatterjee, D., Torrelles, J., Brennan, P.J., Kaufmann, S.H., and Schaible, U.E. (2001) Mycobacterial lysocardiolipin is exported from phagosomes upon cleavage of cardiolipin by a macrophage-derived lysosomal phospholipase A2. J Immunol 167: 2187–2192. Gautier, G., Humbert, M., Deauvieau, F., Scuiller, M., Hiscott, J., Bates, E.E., et al. (2005) A type I interferon autocrineparacrine loop is involved in Toll-like receptor-induced interleukin-12p70 secretion by dendritic cells. J Exp Med 201: 1435–1446. Geijtenbeek, T.B., Van Vliet, S.J., Koppel, E.A., SanchezHernandez, M., Vandenbroucke-Grauls, C.M., Appelmelk, B., and Van Kooyk, Y. (2003) Mycobacteria target DC-SIGN to suppress dendritic cell function. J Exp Med 197: 7–17. Gerosa, F., Baldani-Guerra, B., Nisii, C., Marchesini, V., Carra, G., and Trinchieri, G. (2002) Reciprocal activating interaction between natural killer cells and dendritic cells. J Exp Med 195: 327–333. Giacomini, E., Iona, E., Ferroni, L., Miettinen, M., Fattorini, L., Orefici, G., et al. (2001) Infection of human macrophages and dendritic cells with Mycobacterium tuberculosis induces a differential cytokine gene expression that modulates T cell response. J Immunol 166: 7033–7041. Gillespie, G.M., Wills, M.R., Appay, V., O’Callaghan, C., Murphy, M., Smith, N., et al. (2000) Functional heterogeneity and high frequencies of cytomegalovirus-specific CD8(+) T lymphocytes in healthy seropositive donors. J Virol 74: 8140–8150. Glickman, M.S., Cox, J.S., and Jacobs, W.R., Jr (2000) A novel mycolic acid cyclopropane synthetase is required for cording, persistence, and virulence of Mycobacterium tuberculosis. Mol Cell 5: 717–727. Guermonprez, P., Valladeau, J., Zitvogel, L., Thery, C., and Amigorena, S. (2002) Antigen presentation and T cell stimulation by dendritic cells. Annu Rev Immunol 20: 621– 667. Herrmann, J.L., O’Gaora, P., Gallagher, A., Thole, J.E., and Young, D.B. (1996) Bacterial glycoproteins: a link between glycosylation and proteolytic cleavage of a 19 kDa antigen from Mycobacterium tuberculosis. EMBO J 15: 3547– 3554. Iwasaki, A., and Medzhitov, R. (2004) Toll-like receptor control of the adaptive immune responses. Nat Immunol 5: 987–995. Jiao, X., Lo-Man, R., Guermonprez, P., Fiette, L., Deriaud, E., Burgaud, S., et al. (2002) Dendritic cells are host cells for mycobacteria in vivo that trigger innate and acquired immunity. J Immunol 168: 1294–1301. Koppel, E.A., Ludwig, I.S., Hernandez, M.S., Lowary, T.L., Gadikota, R.R., Tuzikov, A.B., et al. (2004) Identification of the mycobacterial carbohydrate structure that binds the C-type lectins DC-SIGN, L-SIGN and SIGNR1. Immunobiology 209: 117–127. Ludwiczak, P., Gilleron, M., Bordat, Y., Martin, C., Gicquel, B., and Puzo, G. (2002) Mycobacterium tuberculosis phoP

mutant: lipoarabinomannan molecular structure. Microbiology 148: 3029–3037. Maeda, N., Nigou, J., Herrmann, J.L., Jackson, M., Amara, A., Lagrange, P.H., et al. (2003) The cell surface receptor DC-SIGN discriminates between Mycobacterium species through selective recognition of the mannose caps on lipoarabinomannan. J Biol Chem 278: 5513–5516. Martin-Fontecha, A., Sebastiani, S., Hopken, U.E., Uguccioni, M., Lipp, M., Lanzavecchia, A., and Sallusto, F. (2003) Regulation of dendritic cell migration to the draining lymph node: impact on T lymphocyte traffic and priming. J Exp Med 198: 615–621. Moretta, A. (2002) Natural killer cells and dendritic cells: rendezvous in abused tissues. Nat Rev Immunol 2: 957– 964. Neyrolles, O., Gould, K., Gares, M.P., Brett, S., Janssen, R., O’Gaora, P., et al. (2001) Lipoprotein access to MHC class I presentation during infection of murine macrophages with live mycobacteria. J Immunol 166: 447–457. Nigou, J., Gilleron, M., Cahuzac, B., Bounery, J.D., Herold, M., Thurnher, M., and Puzo, G. (1997) The phosphatidylmyo-inositol anchor of the lipoarabinomannans from Mycobacterium bovis bacillus Calmette Guerin. Heterogeneity, structure, and role in the regulation of cytokine secretion. J Biol Chem 272: 23094–23103. Nigou, J., Zelle-Rieser, C., Gilleron, M., Thurnher, M., and Puzo, G. (2001) Mannosylated lipoarabinomannans inhibit IL-12 production by human dendritic cells: evidence for a negative signal delivered through the mannose receptor. J Immunol 166: 7477–7485. Piccioli, D., Sbrana, S., Melandri, E., and Valiante, N.M. (2002) Contact-dependent stimulation and inhibition of dendritic cells by natural killer cells. J Exp Med 195: 335–341. Pitarque, S., Herrmann, J.L., Duteyrat, J.L., Jackson, M., Stewart, G.R., Lecointe, F., et al. (2005) Deciphering the molecular bases of Mycobacterium tuberculosis binding to the lectin DC-SIGN reveals an underestimated complexity. Biochem J 392: 615–624. Pron, B., Boumaila, C., Jaubert, F., Berche, P., Milon, G., Geissmann, F., and Gaillard, J.L. (2001) Dendritic cells are early cellular targets of Listeria monocytogenes after intestinal delivery and are involved in bacterial spread in the host. Cell Microbiol 3: 331–340. Romero, P., Gervois, N., Schneider, J., Escobar, P., Valmori, D., Pannetier, C., et al. (1997) Cytolytic T lymphocyte recognition of the immunodominant HLA-A*0201-restricted Melan-A/MART-1 antigenic peptide in melanoma. J Immunol 159: 2366–2374. Russell, D.G. (2001) Mycobacterium tuberculosis: here today, and here tomorrow. Nat Rev Mol Cell Biol 2: 569–577. Salio, M., Shepherd, D., Dunbar, P.R., Palmowski, M., Murphy, K., Wu, L., and Cerundolo, V. (2001) Mature dendritic cells prime functionally superior melan-A-specific CD8+ lymphocytes as compared with nonprofessional APC. J Immunol 167: 1188–1197. Schaible, U.E., Hagens, K., Fischer, K., Collins, H.L., and Kaufmann, S.H. (2000) Intersection of group I CD1 molecules and mycobacteria in different intracellular compartments of dendritic cells. J Immunol 164: 4843–4852. Schaible, U.E., Winau, F., Sieling, P.A., Fischer, K., Collins, H.L., Hagens, K., et al. (2003) Apoptosis facilitates antigen


ManLAM activation of human DCs 1425 presentation to T lymphocytes through MHC-I and CD1 in tuberculosis. Nat Med 9: 1039–1046. Tailleux, L., Schwartz, O., Herrmann, J.L., Pivert, E., Jackson, M., Amara, A., et al. (2003a) DC-SIGN is the major Mycobacterium tuberculosis receptor on human dendritic cells. J Exp Med 197: 121–127. Tailleux, L., Neyrolles, O., Honore-Bouakline, S., Perret, E., Sanchez, F., Abastado, J.P., et al. (2003b) Constrained intracellular survival of Mycobacterium tuberculosis in human dendritic cells. J Immunol 170: 1939–1948.

Watkins, S.C., and Salter, R.D. (2005) Functional connectivity between immune cells mediated by tunneling nanotubules. Immunity 23: 309–318. Young, D., Hussell, T., and Dougan, G. (2002) Chronic bacterial infections: living with unwanted guests. Nat Immunol 3: 1026–1032. Zippelius, A., Pittet, M.J., Batard, P., Rufer, N., de Smedt, M., Guillaume, P., et al. (2002) Thymic selection generates a large T cell pool recognizing a self-peptide in humans. J Exp Med 195: 485–494.
