Immune defence against intracellular pathogens or tumours is the domain of CD8+ cytotoxic T lymphocytes (CTLs). These cells recognize antigenic peptides associated with MHC class I molecules, which are expressed on the surface of all cells in the body. When pathogen- or tumour-specific effector CTLs detect antigen being presented by a cell, they destroy the cell to avert the spread of infection or cancer. To avoid the killing of healthy bystander cells that have endocytosed viral or tumour debris, this endocytosed material does not usually enter the MHC class I-loading machinery in the endoplasmic reticulum. The endogenous MHC class I pathway is reserved for peptides derived from intracellularly synthesized proteins. Therefore, CTL cytotoxicity is focused on cells in which microorganisms replicate or that are malignant.

However, this restriction of MHC class I loading is insufficient in one situation: naive CTLs need to be activated by professional antigen-presenting cells (APCs), usually dendritic cells (DCs), before they can exert their cytotoxic effector functions. So when an intracellular pathogen does not infect APCs or compromises their endogenous MHC class I pathway, or when a tumour is not APC-derived (which is the case for most tumours), CTLs can only be activated if APCs present extracellular antigen on their MHC class I molecules. This process, termed cross-presentation, was named after the phenomenon of cross-priming that was discovered in the 1970s1, in which antigens from intravenously injected cells 'crossed' into the MHC class I pathway of host APCs for CTL priming. Cross-priming has since been shown to be required for defence against many viruses and tumours2,3, and it is essential for vaccinations with protein antigens, which must be cross-presented to activate CTLs4. Self antigens are also cross-presented but rather than cross-priming, this normally results in deletion of autoreactive CTLs, in a process termed cross-tolerance5.

Technical and conceptual advances in recent years have provided insights into the underlying mechanisms of cross-priming, especially the associated intracellular antigen processing mechanisms and DC subsets, which have being reviewed elsewhere4,6,7. The physiological importance of cross-priming and cross-tolerance in health and disease is less well understood. Here, we summarize recent advances in our understanding of the regulation of cross-priming and its implications for defence against infections, CTL-mediated diseases and tumours.

Basic mechanisms of cross-priming

Roles of APC types in cross-priming. In vivo depletion or functional inhibition of DCs compromises both cross-priming and cross-tolerance8,9, identifying these cells as the most relevant cross-presenting APCs in mice6,7,10,11. Among the DCs, only those expressing the surface markers CD24, CD8α and CD103 (also known as αE integrin) usually cross-present antigen6,7,10,11. Cross-presenting DC subsets are usually identified by ex vivo analysis of DCs isolated from antigen-injected mice, because selective in vivo targeting of DC subsets is difficult. The targeting of DC subsets has become possible by the discovery that the transcription factor BATF3 (basic leucine zipper transcriptional factor ATF-like 3) is expressed by CD8α+ and CD103+ DCs and is necessary for cross-presentation12. Using BATF3-deficient mice, it has been proposed that these DCs constitute a unified subset with a common lineage13. Under inflammatory conditions, in mucosal tissues and in tumours, DCs that lack expression of CD24, CD8α and CD103 can also cross-present14,15,16,17,18. However, the human counterparts of cross-presenting mouse DC subsets remain to be identified.

Classifying DC subsets on the basis of cell surface markers is convenient for flow-cytometric analysis, but does not mechanistically explain why a cell can cross-present. Recent studies have proposed two mechanistic explanations, which are not mutually exclusive. The first hypothesis proposed that only cross-presenting DCs possess the antigen processing machinery that loads endocytosed antigen onto MHC class I molecules19. Therefore, based on this hypothesis, cross-presentation must depend on specific antigen processing mechanisms that can be regulated separately, so that MHC class I-restricted presentation of endogenous antigens remains operational. Insulin-regulated aminopeptidase (IRAP; also known as cystinyl aminopeptidase), a trimming peptidase located in endosomal compartments, has been recently identified as a processing molecule with such specificity20; however, its general role in cross-presentation remains to be established. The second hypothesis proposed that cross-presentation depends on distinct endocytosis mechanisms that introduce antigen directly into the organelle (or organelles) in which cross-presentation occurs21 (Box 1). In support of this, uptake of antigen for cross-presentation is restricted to distinct endocytosis receptors, such as Fc receptors and certain members of the C-type lectin receptor family, such as C-type lectin 9A (CLEC9A; also known as DNGR1), CLEC7A (also known as dectin 1), DC-SIGN (also known as CD209), DEC205 (also known as CD205) and mannose receptor 1 (also known as CD206)19,22,23,24. These receptors are expressed by CD8α+ DCs21, providing a possible mechanistic explanation for why this DC subset can cross-present.

If CTL-mediated immune responses depend on cross-priming, one might ask why it is restricted to only a few DC subsets. One explanation is that DCs capable of cross-presentation automatically become targets for CTL killing after endocytosis of viral or tumour debris. Hence, during a systemic infection, all DCs might be killed if they all could cross-present. Although DCs possess mechanisms to avoid CTL-mediated killing to some degree25, it is clear that cross-presenting DCs can be effectively killed during viral infections26. However, this killing does not seem to affect the ensuing immune response as CTLs need less than one day of antigen presentation by DCs for activation and a little longer for developing cytolytic functionality27,28, and they can then carry out their tasks without continual stimulation, at least for some days. DCs that cannot cross-prime can maintain other immune responses that depend on CD4+ T helper (TH) cells, such as antibody production and macrophage stimulation, because they will not be killed by CTLs, unless they are virally infected themselves. Consistent with this notion, non-cross-presenting DCs are superior at activating TH cells19,21.

Cross-priming CTLs specific for peripheral tissue antigens involves the cooperation between the different DC types. Tissue DCs transport antigen from tissues to secondary lymphoid organs and transfer it to resident CD8α+ DCs, which then cross-prime CTLs29,30. This antigen transfer mechanism provides several theoretical advantages7: first, migratory DCs may distribute antigen to several resident DCs, which together can prime more CTLs; second, antigen endocytosed by non-cross-presenting DCs can also be used for cross-priming; third, non-cross-presenting DCs carrying viral antigens would not be killed by viral-specific CTLs before reaching draining lymph nodes; fourth, viruses may compromise endogenous antigen processing pathways in infected migratory DCs, but not in the antigen-receiving DCs; and fifth, the CTL will have to inspect only the few antigen-receiving resident DCs rather than a large number of migratory DCs, which may allow faster location of their cognate antigen. However, this concept of antigen transfer also contains some conceptual problems: even if antigen is transported by non-cross-presenting DCs, eventually it must reach a cross-presenting DC, which then might be killed by CTLs. Furthermore, the information obtained from direct encounters with microbial molecular patterns must somehow be transferred to the cross-priming DCs. Finally, the mechanism of antigen transfer between DCs is unclear.

In addition to DCs, macrophages, neutrophilic granulocytes and mast cells can cross-present antigen but cannot cross-prime CTLs31. B cells only cross-present antigen recognized by their immunoglobulin receptors and can cross-prime CTLs when Toll-like receptors (TLRs) are also engaged32. Resting B cells can tolerize CTLs when externally loaded with antigenic peptide33, but evidence for a functional role of these cells in cross-tolerance in vivo is lacking. By contrast, liver sinusoidal endothelial cells (LSECs) can efficiently cross-tolerize CTLs that are specific for food antigens or commensal bacteria34. In summary, distinct APC types have different roles in cross-priming and cross-tolerance.

DC activation and DC licensing. Immunogenic responses require that DCs encounter antigen that is associated with molecular patterns indicative of the presence of microorganisms or other 'danger' signals35. Sensing such patterns (for example, by activation of TLRs) leads to the upregulation of co-stimulatory molecules and enhances cross-priming; the enhancement of cross-priming being especially effective after TLR3 and TLR9 stimulation36,37. Furthermore, TLR signalling enhances peptide loading onto MHC class I molecules by inducing the relocation of components of the peptide-loading machinery to the endosomes in which cross-presentation occurs38. Many C-type lectin receptors preferentially recognize foreign glycosylation structures characteristic of microorganisms and thereby allow some degree of self–foreign discrimination21,39. Engaging these receptors may synergize with TLR signals to promote upregulation of co-stimulatory molecules because distinct signalling pathways are used40, and may theoretically enhance cross-priming. However, this cannot be extrapolated from studies targeting antigens to C-type lectin receptors, because such targeting enhances cross-priming by increasing antigen uptake, which is difficult to discriminate from C-type lectin receptor signalling. Also, activation of RIG-I-like cytoplasmic receptors for nucleic acids and the inflammasome boost CTL responses41,42and indirect evidence suggests that these cytoplasmic sensors can stimulate cross-priming43,44, but this has not yet been formally shown.

Although co-stimulatory signals are necessary for immunogenic cross-priming, they are insufficient in some models to break cross-tolerance45,46, unless specific TH cells are present45,47. This has led to the realization that DCs require signals from specific TH cells for immunogenic cross-priming. CTLs activated without T cell help (termed 'helpless' CTLs) lack expression of specific anti-apoptotic molecules, have a short life-span and cannot carry out cytotoxic effector functions5,48,49,50. The necessity for specific T cell help can be viewed as requesting a 'second opinion' from TH cells before programming CTLs for cytotoxic functions. Both autoreactive TH cells and CTLs are present at a low frequency in a normal repertoire, but the likelihood that clones specific for the same autoantigen coexist will be much lower. Hence, autoimmune responses become less likely when antigen must be recognized by two lymphocyte subsets that had been negatively selected for self-reactivity. An analogous safety strategy is implemented in antibody production, in which specific B cells and TH cells need to both recognize the foreign antigen. However, in contrast to B cells, CTLs lack MHC class II molecules and cannot directly receive specific help from TH cells. Therefore, cross-presenting DCs act as a temporal bridge between these cells. The TH cells convert the DCs to a transient state in which they can programme the CTL for sustained cytotoxic effector functions and memory differentiation (Fig. 1a)51,52. When such DCs are adoptively transferred into naive mice, they retain the ability to cross-prime CTLs, showing that they store the information received from the TH cells. This has been termed 'DC licensing'.

Figure 1: Molecular mechanisms of licensing dendritic cells for classical cross-priming.
figure 1

a | The molecular mechanisms involved in classical cross-priming are illustrated. Dendritic cells (DCs) take up antigen by distinct endocytosis mechanisms (not shown) and present it to CD4+ T helper (TH) cells through MHC class II molecules and cross-present it to CD8+ cytotoxic T lymphocytes (CTLs) through MHC class I molecules. Activated CD4+ TH cells can stimulate CTLs through the production of interleukin-2 (IL-2) and license DCs for cross-priming through CD40 ligand (CD40L)–CD40 interactions. Licensed DCs upregulate expression of co-stimulatory molecules, such as CD70, CD80 and CD86, and downregulate inhibitory molecules, such as programmed cell death ligand (PDL1). Toll-like receptor (TLR) ligands further activate DCs and increase their cross-presentation activity. Cross-primed CTLs are programmed for survival and cease TNF-related apoptosis-inducing ligand (TRAIL) production. 'Helpless' CTLs activated by unlicensed DCs die following secondary encounter with antigen in their effector phase (not shown). b | Chemokine-mediated regulation of cross-priming is illustrated. CD4+ TH cells, and the DCs they license, produce CC-chemokine ligand 3 (CCL3), CCL4 and CCL5 in the presence of TLR ligands, which recruit naive CTLs for classical cross-priming. Alternatively, DCs licensed by natural killer T (NKT) cells produce the CC-chemokine receptor (CCR4) ligand CCL17, and NKT cells themselves produce the CCR4 ligand CCL22, which recruit naive CCR4+ CTLs for cross-priming. The CCR4- and CCR5-mediated recruitment pathways are synergistic. In this figure, dashed arrows indicate antigen routing for cross-presentation. α-GalCer, α-galactosylceramide; TCR, T cell receptor.

DC licensing is usually facilitated by TH cell-expressed CD40 ligand (also known as CD154), which interacts with CD40 on DCs31. It remains unclear to what extent signalling events induced by TH cells or by TLRs overlap in cross-priming DCs. Some differences must exist, because TH cells and pattern-recognition receptors synergistically activate DCs and both are necessary for optimal cross-priming4,45.

CTL programming. Licensed cross-presenting DCs provide signals that dictate whether cross-priming or cross-tolerance ensues, a process termed 'CTL programming'27,28,53. Understanding the identity and regulation of these programming signals is important for optimizing vaccinations and understanding autoimmunity, and hence they have been intensely studied. As candidate signals, altered antigen presentation and modified co-stimulatory and/or co-inhibitory signals have been proposed. The duration of T cell receptor (TCR) signalling affects whether activated CTL survive or not27,28, whereas the TCR affinity dictated the level of CTL expansion and ensured that only high affinity CD8+ T cell clones develop into memory CTLs54. It is unknown, however, how these qualities of TCR signalling may be regulated by the DCs and by licensing TH cells.

CD70 is a candidate co-stimulatory signal for CTL programming. This activation-induced tumour necrosis factor (TNF) receptor family molecule interacts with CD27, which is expressed by CTLs, and increases their survival17,55. However, the expression of CD70 by DCs is not regulated by TH cells56. Downregulation of programmed cell death ligands (PDLs), which bind to the co-inhibitory receptor PD1 of the CD28 family and promote CTL tolerance, might also contribute to CTL programming57. It is unclear whether these ligands are regulated by TH cells. The expression of interleukin-12 (IL-12) by DCs is regulated by TH cells and TLR ligands, and promotes CTL effector functions58; however, its ability to programme CTLs for survival is unclear.

CTL survival can be programmed by IL-2 during priming59. However, IL-2 is not thought to be produced by DCs and therefore cannot be the elusive survival signal that licensed DCs provide. Theoretically, it may originate from TH cells, but this would require their presence during CTL priming, which is inconsistent with DCs storing the licensing information. Furthermore, TH cell-derived IL-2 seems to have a distinct, probably antigen-nonspecific, role in CTL memory maintenance53. Alternatively, IL-2 may be produced by the CTL themselves. Indeed, prolonged TCR signalling during priming induces autocrine IL-2 production60, and such IL-2 can sustain CTL survival, at least when the CTLs are present at high numbers61. Also, CD70–CD27 signals upregulate autocrine IL-2 secretion by CTLs62, whereas it is downregulated by PD1 ligation63, suggesting that these two candidates for the DC-derived CTL survival signal may operate indirectly by regulating autocrine IL-2 production. A converse effect on CTL survival has been reported for autocrine TNF-related apoptosis-inducing ligand (TRAIL), which CTLs produce by default and which is shut off by specific TH cells64. Thus, helpless CTLs may use TRAIL to kill themselves or each other if they also express the TRAIL receptor. However, IL-2 does not inhibit TRAIL production65, and causal connections to CD70 and PD1 ligands have not yet been established. In conclusion, many mechanistic steps important in cross-priming have been described, but a comprehensive theory of the molecular mechanisms by which licensed DCs programme CTLs for survival, cytotoxic effector function and memory generation that links these steps is lacking.

The cellular encounters for cross-priming

CTL activation in lymphoid organs. The need for DC licensing by TH cells has an important limitation: three rare immune cells — the cross-presenting DC, the antigen-specific TH cell and the CTL — need to all interact. However, because the licensing information is stored within DCs, these interactions may not necessarily occur at the same time. The CTL may theoretically interact with the licensed DC after the DC has interacted with the licensing TH cell51. Nevertheless, relying on random encounters during normal T cell recirculation would probably be too time-consuming to guarantee timely defence against rapidly replicating microorganisms. Recent studies showed that the encounters necessary for CTL activation are greatly enhanced by chemokines. In particular, CC-chemokine receptor 5 (CCR5) ligands are produced following DC–TH cell interactions under inflammatory conditions, which were mimicked by the injection of TLR ligands66 (Fig. 1b). These chemokines allowed CCR5-expressing naive CTLs to be attracted to licensed DCs, where they could scan for cognate antigen. Although TH cells and DCs first need to find each other, the subsequent attraction of the third rare cell type is thereby greatly enhanced and this decreases the time spent by naive CTLs scanning unlicensed DCs. Once activated, CTLs also produce CCR5 ligands67, which may guide other naive CTLs towards DCs that have successfully cross-primed. CTLs also produce XC-chemokine ligand 1 (XCL1; also known as lymphotactin) that binds to XC-chemokine receptor 1 (XCR1), which is selectively expressed by CD8α+ DCs, and this interaction supports cross-talk between these cell types and enhances cross-priming68.

Natural killer T (NKT) cells, which recognize glycolipid antigens presented by CD1d molecules, can also enhance cross-priming, for example by upregulating co-stimulatory molecules on DCs69. This process has been recently shown to involve cognate DC licensing70 and consequently a tripartite immune cell encounter. Chemokines, specifically CCR4 ligands and especially CCL17 (also known as TARC), produced by cross-priming DCs are instrumental in this setting70 (Fig. 1b). Thus, at least two chemokine systems recruit naive CTLs towards DCs that have been licensed by TH cells or NKT cells. The CTL chemotaxis mediated by CCR4 and CCR5 is synergistic, providing a mechanistic rationale for combining ligands for NKT cells with TLR ligands plus TH cell epitopes in vaccination strategies. However, it remains unclear how naive CTLs are induced to express CCR4 or CCR5 before they interact with DCs that were licensed by NKT or TH cells, respectively.

Regulation of effector CTLs in tissues. DCs can also regulate the CTL effector phase in tissues50,71,72. If DCs are infected, they may do so by using the endogenous MHC class I pathway, unless it has been compromised by the infection. If tissue DCs are not infected, they would need to cross-present the antigen to ensure the antigen specificity of this regulation. However, tissue DCs are usually of the CD8αCD11b+ subset, which has been shown to be unable of cross-presentation, at least under steady-state conditions and in vitro6,7. If such DCs could acquire this ability by inflammation-induced redifferentiation into cross-presenting cells (for example CD8α+ DCs), the ability to cross-present would essentially be a functional state that can be regulated, rather than a property of DC subsets as entities with distinct lineage. Recently, CD8αCD103+ DCs have been detected in certain tissues, which are developmentally related to CD8α+ DCs13 and can cross-present tissue antigens after migrating into draining lymph nodes11. However, it is unclear if these cells can cross-present antigen while in tissues. Thus, it is currently undecided whether DCs within non-lymphoid tissues can regulate effector CTLs in an antigen-specific manner.

An alternative mechanism by which tissue DCs could regulate infiltrating effector CTLs in an antigen-specific, but non-MHC class I-restricted, manner involves the presentation of antigen on MHC class II molecules to TH cells, which results in the production of cytokines and chemokines that recruit and/or regulate CTLs (Fig. 2). This has been described in mouse models of mesothelioma73, immune-mediated kidney disease50 and genital herpes simplex virus (HSV) infection72.

Figure 2: Recruitment of cross-primed effector cytotoxic T lymphocytes into non-lymphoid tissues.
figure 2

Viral infection of tissue cells leads to their secretion of pro-inflammatory cytokines and interferons that upregulate the expression of adhesion molecules, such as intercellular adhesion molecule 1 (ICAM1), by endothelial cells. Effector cytotoxic T lymphocytes (CTLs) attach to these molecules and are nonspecifically recruited from the bloodstream into non-lymphoid tissues. In addition, recent studies have revealed two antigen-specific recruitment mechanisms: first, endothelial cells in certain tissues, such as the liver, pancreatic islets or the brain, can cross-present microbial antigen, which allows them to selectively recruit antigen-specific CTLs. Second, when specific CD4+ T helper (TH) cells detect microbial (or self) antigen on non-cross-presenting tissue DCs, ligands for CC-chemokine receptor 5 (CCR5) or CXC-chemokine receptor 3 (CXCR3) are produced that recruit CTLs into the infected tissue. LFA1, lymphocyte function-associated antigen 1; TCR, T cell receptor.

There is also an antigen-specific and MHC class I-restricted CTL recruitment mechanism. This is not facilitated by DCs, but instead by cross-presenting endothelial cells. Such cells have been detected in pancreatic islets74, the central nervous system75 and the liver76, and have been shown to establish firm T cell contact and activation, which in turn will increase lymphocyte function-associated antigen 1 (LFA1; also known as αLβ2 integrin)–intercellular adhesion molecule 1 (ICAM1)-mediated transmigration thorough endothelial barriers into tissues. Subsequently, tissue-resident DCs may then regulate the incoming CTLs; for example, by cross-talk with TH cells and chemokine production. However, cross-presenting endothelial cells might be killed by CTLs, causing unwanted side-effects. This may be avoided by high levels of co-inhibitory molecule expression, such as PDL1, by the endothelial cells, which could help prevent CTL-mediated damage77. Thus, in addition to the classical paradigm of CTL recruitment into tissues by attachment to adhesion molecules on inflamed endothelial cells, two specific pathways exist that are facilitated by cross-presenting endothelial cells or by chemokines resulting from specific TH cell–DC crosstalk (Fig. 2).

Cross-priming during infections

Viral infections. Cross-priming is necessary for CTL-mediated immune responses against microorganisms that do not target DCs, such as Epstein–Barr virus (EBV), hepatitis B virus (HBV) or poliovirus (Table 1). Furthermore, cross-priming is important for DC-infecting viruses that evade the endogenous MHC class I pathway, such as members of the Herpes virus family, especially murine cytomegalovirus (MCMV). However, many viruses infect DCs, suggesting that avoiding DCs is not a survival advantage for the virus, perhaps because cross-priming exists.

Table 1 Cross-priming of CTLs in viral infections, tumours and CTL-mediated diseases

MCMV infects various cells, including DCs, and produces immune evasins that inhibit MHC class I-restricted antigen presentation78. Hence, MCMV-infected DCs fail to induce a virus-specific CTL-mediated immune response79. Despite escaping from the endogenous antigen presentation pathway, MCMV-specific CTLs eventually develop, demonstrating that cross-priming by uninfected DCs must occur80. This pathway also induces CTL-mediated immunity against other immune evasin-producing viruses, such as EBV81 and HSV-1 (Ref. 7). But even when effective cross-priming occurs, MCMV-infected cells still escape CTL-mediated killing because these cells present viral epitopes that are different from those that the CTLs are cross-primed against80. Interestingly, ablation of immune evasins from MCMV results in decreased cross-priming and antiviral immunity82. This paradoxical observation can be explained by the rapid control of viral replication and decreased viral antigen burden in the absence of these molecules. Other viruses also directly avoid cross-priming: vaccinia viruses escape DC cross-priming by restricting late viral antigens to a particular cellular compartment that is inaccessible for cross-presentation but accessible for MHC class II-restricted antigen presentation83.

Type I interferons (IFNs) are released from virus-infected cells or sentinel cells such as plasmacytoid DCs, and they can induce DC maturation and support cross-priming. This may be important when neither viral nor microbial-derived signals are available to induce the maturation of cross-priming DCs84. Type I IFNs are efficiently induced by triggering TLR3, which recognizes double-stranded viral RNA and effectively stimulates cross-priming37. TLR3-induced type I IFN expression is associated with control of viral infection, albeit at the expense of immune-mediated organ damage85. Despite these mechanisms, some viruses still do not elicit efficient CTL responses despite abundant antigen production. For example, hepatitis C virus (HCV) is recognized by RIG-I, but blocks the subsequent induction of type I IFNs, and HBV induces the production of IL-6 but not type I IFNs by infected cells86,87. The efficacy of type I IFN treatment in patients with chronic viral hepatitis may involve improved cross-priming, in addition to any direct antiviral effect.

Viruses such as HIV, influenza virus or HCV can escape established CTL responses by mutating their antigens (Table 1). Furthermore, during the course of an infection, different viral antigens are expressed, such as the early and late antigens of HSV. Thus, continuous delivery of viral antigens is necessary to cross-prime CTL clones specific for new antigens in lymph nodes. Such antigen delivery can be mediated by migratory DCs29 or by the endocytosis of viral antigens that drained though lymph vessels by lymph node-resident DCs88. In HSV infection of the skin or lung, antigen is transported by migratory CD103CD11bhi tissue DCs (or by Langerhans cells in the skin) to draining lymph nodes and transferred to cross-presenting CD8α+ DCs17, or is conveyed by migratory CD103+CD11b DCs that directly cross-present in draining lymph nodes7. By contrast, migratory CD103CD11bhi DCs transported viral antigen during influenza virus infection from the lung to the bronchial lymph nodes and directly cross-primed CTLs without the need to transfer antigen to CD8α+ DCs17. The role of these DC subsets in other viral infections is unclear.

Bacterial infections. The importance of cross-priming for bacterial infection has been mostly studied using Listeria monocytogenes. Following infection, L. monocytogenes gains access to the cytoplasm of splenic macrophages and hepatocytes, which subsequently undergo apoptosis. DCs acquire antigen for cross-priming following uptake of debris of infected cells, which is crucial for defence against L. monocytogenes infection9. Cross-priming can also occur in the absence of protein synthesis, on the condition that the innate immune stimulatory function of L. monocytogenes is preserved89. Surprisingly, cross-priming DCs are also important for bacterial dissemination into the spleen, as they take up circulating bacteria and allow their initial intracellular survival, thereby promoting subsequent bacterial proliferation and spread to other cell populations90. This may localize infections to sites where DCs can later orchestrate innate and adaptive immune responses91. Shortening the time of bacterial exposure curtailed specific immunity by decreasing TH cell responses92, showing the importance of DC licensing, which requires time-consuming establishment of physical contact with licensing TH or NKT cells. Terminating CTL immunity against infections that do not provide stimulation for a minimum time (that is, are eliminated by innate immune mechanisms) may focus adaptive immune responses to those infections requiring CTL immunity but may also allow pathogen escape mechanisms to establish persistent infection.

Acquisition of bacterial antigen by DCs has been most intensively studied during mycobacterial infection, in which mycobacteria persist in phagosomal vesicles and dispersion of bacterial antigens into the cytoplasm does not occur. Here, apoptosis of infected cells, facilitating the uptake of bacterial antigens, is indispensable for cross-priming by DCs and induction of efficient CTL-mediated immunity93. Similar to viral infections, transport of bacterial antigens by DC into lymph nodes was also required for cross-priming94. However, processing of bacterial antigens from apoptotic vesicles in DCs differs from that of viral antigens by requiring saposins, which open vesicles and release bacterial antigens into the cytoplasm for processing94. Taken together, DC cross-priming supports the induction of CTL-mediated immunity against pathogens with a narrow host cell tropism that avoid infecting DCs or possess immune escape mechanisms.

Cross-priming in immune-mediated diseases

Although central tolerance is an efficient process, some autoreactive CTLs escape negative selection and enter the circulation. In secondary lymphoid organs, cross-tolerance serves as a second checkpoint that can eliminate such escapees5,8. However, cross-tolerance occurs only when the autoantigen dose and/or CTL affinity are sufficiently high31,49. If not, then autoreactive CTLs may escape cross-tolerance and cause disease when cross-primed95.

CTLs are the main effector cells of human type 1 diabetes96. Cross-priming is necessary and sufficient for islet infiltration and destruction in transgenic diabetes models31,49 and in non-obese diabetic (NOD) mice97, which have multigenic disease susceptibility similar to human illness, including DC defects that compromise cross-tolerance98. However, the identity of the antigens that are cross-presented is not clear. Theoretically, these antigens could be pancreatic self antigens or microbial antigens mimicking such antigens. The suggestion of a role for microbial antigens is supported by studies with transgenic mice expressing viral components in islets, which developed CTL-mediated diabetes after virus infection99. However, the lymphocytic choriomeningitis virus used in these studies infects DCs, which then directly prime CTLs, and therefore cross-priming becomes unnecessary. Antigenic mimicry was suggested to be supported by the epidemiological association of diabetes with coxsackie virus infections, but this is now thought to result from changes in the micromilieu of infected islets rather than viral antigens99,100. Thus, the evidence for antigenic mimicry as a cause of type 1 diabetes is unconvincing101.

If so, then diabetogenic CTLs are likely to be cross-primed by pancreatic islet autoantigens. One candidate is an MHC class I-binding peptide from the insulin B chain102, an antigen also relevant in NOD mice103. Pancreatic autoantigens are cross-presented in the pancreatic lymph nodes and this normally leads to cross-tolerance5,104, unless specific TH cells are present31. Notably, TH cells specific for pancreatic autoantigens are also activated at this site105 and might license DCs for cross-priming. This may be important in patients with type 1 diabetes, because many of them possess particularities in MHC class II-restricted antigen presentation and in the responding TH cells106. Diabetes has long been known to be closely associated with the MHC class II loci107, although MHC class I-restricted CTLs are the effectors of disease. The abnormal TH cells of patients with diabetes might function by licensing DCs in the pancreatic lymph node, which then cross-prime diabetogenic CTLs.

Most concepts on the role of CTLs and cross-priming in autoimmunity have been extrapolated from type 1 diabetes models. However, the rapid destruction of pancreatic islets has hampered investigating intra-organ cross-talk between immune cells during the CTL effector phase. Work in autoimmunity models of the kidney and skin, which are more resistant to CTL-mediated killing of tissue cells, has provided new insights. Autoreactive CTLs specific for kidney glomerular antigens are cross-primed in the renal lymph node and are recruited by chemokines into the kidney, where they inflict damage50. These chemokines were produced following cross-talk of autoreactive TH cells with tubulointerstitial DCs presenting renal autoantigen, explaining the long-known ability of TH cells to facilitate kidney access of autoreactive CTLs; for example, in interstitial nephritis108. TH cell-dependent recruitment of autoreactive effector CTLs was also noted in psoriasis109.

In multiple sclerosis research, CTLs are being rediscovered as important effectors110. Myelin-specific CTLs that escape thymic tolerance and cross-tolerance mechanisms can damage oligodendrocytes111,112. Recent work in humanized mice has shown that several classes of bacteria contain an MHC class I-binding peptide that mimics myelin basic protein, a candidate autoantigen in multiple sclerosis113. But as with type 1 diabetes, the exact autoantigens in multiple sclerosis are unknown, hampering elucidation of the role of cross-priming and of antigenic mimicry.

CTL-mediated responses against EBV have been linked with ankylosing spondylitis114, long known to be associated with HLA-B27 and with autoimmune hepatitis115. Hepatic autoantigens are also recognized by HLA-A-restricted CTLs specific for HCV116. In organ transplantation, host CTLs cross-primed against graft antigens will be restricted to the host haplotype, but nevertheless can damage graft cells, either by alloreactivity or by killing host endothelial cells growing in the graft117. There are further examples for associations between disease entities and CTLs (Table 1), but in nearly all cases, a causal role of cross-priming remains to be formally shown.

Cross-priming and tumours

There are notionally three different types of tumour antigens: 'self' antigens, to which a person is partially tolerant (for example, normal embryonic or differentiation antigens that also happen to be expressed in tumours), 'neo-antigens' to which the host is not tolerant (for example, those due to mutations and viruses) and 'potential' antigens, which are epitopes that may be unmasked by treatments such as chemotherapy, which increases antigen delivery into the cross-presentation pathway118. Tumours contain thousands of potentially strongly immunogenic neo-antigens caused by mutations119. Several studies have shown that cross-presentation of these antigens is an efficient and continuous process120; therefore, other reasons as to why the tumour cells are not rejected must exist. Several experimental procedures have been used to analyse cross-priming in tumour models2,120,121,122,123, and these studies have confirmed that cross-presentation of tumour antigens occurs, it persists for the dominate antigen and, interestingly, tends to remain localized to the tumour-draining lymph node120. Within these lymph nodes, both CD8α+ and CD8α DCs cross-present tumour antigens12,123, although whether this reflects multiple mechanisms for delivery of tumour antigen into the cross-presentation pathway, such as through Fc receptor-mediated uptake of endogenous antibodies that have opsonized tumour cells, or other aspects of tumour antigen delivery, is not known.

Although in some tumour models cross-tolerance occurs, in others a weak, ineffective and localized CTL response is induced124. A strong CTL response rarely results from tumour antigen cross-priming without further manipulation125. These weak responses are understandable, as many tumour tissues lack intense inflammation and/or danger signals and do not contain pathogen-associated molecular patterns that usually drive a strong CTL response126. The efficiency of cross-priming to tumour antigens probably depends on the cell line and/or tissue of origin plus other factors, such as antigen dose127 (Table 1).

In summary, tumour antigens are cross-presented efficiently in draining lymph nodes but fail to elicit a strong antitumour CTL response.

Cross-priming and cancer treatment

It is in cancer therapy that cross-priming has triggered the most interest. For example, one of the main mechanisms whereby chemotherapy works is by stimulating antitumour immunity by increased cross-presentation and cross-priming of antitumour CTLs128. Importantly, different cancer chemotherapies kill tumour cells by different mechanisms, not all of which are immunogenic. Certainly it seems that the induction of apoptosis by cancer chemotherapies can increase the amount of tumour antigen delivered into the cross-presentation pathway128. The outcome, however, depends on the effects of chemotherapy on the host immune response and importantly, the way in which chemotherapy kills the tumour cell129,130,131. Increased cross-priming of tumour-specific CTLs induced by chemotherapy has been shown to facilitate long term cures of advanced solid, non-immunogenic tumours when combined with appropriate immunotherapies132 (Fig. 3).

Figure 3: Cross-priming and immunotherapy in an effective antitumour immune response.
figure 3

The first step in any successful antitumour immune attack usually involves cross-presentation of antigens released from tumours, a process that can be boosted by chemotherapies, monoclonal antibodies and vaccines (with or without dendritic cells (DCs) or heat shock proteins). However, this cross-presentation of tumour antigen is not sufficient to eradicate tumours; DC activation is required to turn cross-presentation into a cross-priming event, a process boosted by agents such as CD40 agonists and Toll-like receptor 7 (TLR7). Additional steps are then required; for example, expansion and circulation of antitumour-specific effector cytotoxic T lymphocytes (CTLs), entry of those effector cells into the tumour site and attack of the tumour — each step being subject to regulation. Therefore, boosting cross-presentation and/or cross-priming alone would be insufficient for tumour eradication: scientifically validated therapeutic combinations will be required to turn cross-priming into successful anticancer immunotherapy.

Our knowledge of cross-presentation will also affect cancer surgery. In almost all published studies, the cross-presentation of tumour antigens is restricted to the lymph node that drains the tumour. Indeed, this may generate a level of CTL activity within the local 'sentinel' lymph node that represents a first 'checkpoint' in stopping tumour spread124. Therefore, it has been proposed that the removal of tumour-draining lymph nodes, even if they do not show evidence of tumour spread, may have a negative effect on antitumour immunity.

It has been shown that stronger antitumour effects are seen if cross-priming of CTLs to tumour antigens is increased, for example by using adjuvants such as iscomatrix133, viral vectors134, heat shock proteins135, endogenous danger signals such as high-mobility group box 1 (HMGB1)136 or uric acid crystals43,44, or by using monoclonal antibodies. Indeed, it is possible that one way in which monoclonal antibodies work in cancer immunotherapy is by opsonizing tumour cells and increasing their delivery into the cross-priming pathway137. New strategies are also being developed for linking antigens with DC-targeting molecules which favour particular uptake receptors and hence deliver the tumour antigen into the appropriate cross-presentation pathway19,21,35.

One new concept that evolved from studies of tumour antigen cross-priming is the use of a tumour as its own vaccine132. Because the multitude of tumour-specific mutations cannot be easily identified, strategies that destroy tumours in a way that delivers increased antigen doses into the cross-presentation pathway may well increase the loading of these neo-antigens and boost cross-priming128. This will not generate a strong and prolonged antitumour response; if that were the case, then partial responses to chemotherapy would lead to immune-based tumour eradication after chemotherapy has ceased. However, it does predict that immunotherapies that target DCs following chemotherapy are more likely to be effective than those that do not.

In cases in which tumour antigens or oncogenic viruses have been identified, various vaccination strategies have been used to induce antitumour responses through the cross-presentation pathway. One promising approach is the use of synthetic long peptides. These seem to be more efficient than full-length proteins at entering the cross-presentation pathway138. Synthetic long peptides can induce measurable CTL and TH cell responses against potentially oncogenic viruses in humans139,140.

Concluding remarks

Antiviral vaccinations have long been taking advantage of cross-priming4. Recent findings have clarified some of the underlying mechanisms, such as the endocytosis mechanisms in DCs that facilitate cross-presentation, the organelles in which cross-presentation occurs, DC licensing, CTL programming, chemokine-mediated CTL recruitment and the interplay of cross-priming with chemotherapeutic agents. Translating these advances into the clinic may allow for the design of more effective vaccination strategies, for example by optimized antigen targeting to endocytosis receptors that engage the cross-presentation machinery in suitable DC subsets. Adjuvants can be chosen that not only augment co-stimulation, but also improve the cross-presentation of antigen. Furthermore, the use of adjuvants that induce different chemokines might further improve cross-priming. These chemokines may be viewed as a discrete signal that profoundly affects CTL priming by acting before the antigen- and co-stimulation-derived signals (also termed signals 1 and 2, respectively, for T cell activation)141 and DC-derived cytokines (proposed to be a third signal)58. Hence, chemokines have been interpreted as 'signal 0' in T cell priming142.

Improvements are especially needed in tumour vaccination, as recent clinical studies on DC-based tumour vaccination have been disappointing132,143. Future studies need to focus on how to harness cross-presentation to generate chemo-immunotherapy strategies for therapeutic benefit. Understanding which current chemotherapies work with, rather than against, the anticancer immune response may facilitate new cancer immunotherapies.

We are only beginning to understand in which infections and in which CTL-mediated diseases cross-priming is involved or crucial. Certainly, clarifying its role in human disease is far more difficult than doing so in mouse disease models. In vitro models have long been insufficient (Box 2), but are now improved and supplemented by knockdown techniques, and hence will probably provide further mechanistic insight into the role of cross-priming in many disease entities.