Plasmacytoid dendritic cells: the key to CpG¹

Abstract

The vertebrate immune system has established TLR9 to detect microbial DNA based on unmethylated CG dinucleotides within certain sequence contexts (CpG motifs). In humans, the expression of toll-like receptor 9 (TLR9) is restricted to B cells and plasmacytoid dendritic cells (PDC). The PDC is characterized by the ability to rapidly synthesize large amounts of type I IFN (IFN-α and IFN-β) in response to viral infection. In contrast to other dendritic cell subsets which express a broad profile of TLRs, the TLR profile in PDC is restricted to TLR7 and TLR9. So far, CpG DNA is the only defined microbial molecule recognized by PDC. An intriguing feature of PDC is its ability to simultaneously produce the two major Th1-inducing cytokines in humans, IFN-α and IL-12, both at high levels. The ratio of IFN-α versus IL-12 and the quantity of these cytokines are regulated by T helper cell-mediated costimulation via CD40 ligation. The ratio also depends on the differentiation stage of the PDC at the time of stimulation and the type of CpG ODN used. We propose a model in which the establishment of Th1 responses in vivo is improved by appropriately stimulated PDC that otherwise – in the absence of CpG DNA – support Th2 or Th0 responses and thus have been called DC2.

Introduction

Three primarily independent tracks of investigations converged in the late 1990s, which lead to the identification of the plasmacytoid dendritic cell (PDC) as one defined immune cell subset. Historically, the first evidence for this cell type was published in 1958 by K. Lennert at the University of Kiel in Germany [1]. He described these cells as lymphoblasts occurring mostly in clusters in the lymph node pulp [1]. Later the same group termed this cell type T-cell associated plasma cells because it turned out that, in electronmicroscopic studies, these lymphoblasts have a morphology similar to plasma cells and are closely associated with the T cell areas of lymphatic tissues 2, 3. The function and the lineage remained elusive, but it was suggested early on that these cells due to their prominent rough endoplasmic reticulum might produce large amounts of lymphokines [4]. Later, the name was switched from plasmacytoid T cells to plasmacytoid monocytes based on immunohistochemical evidence of a monocyte/macrophage origin [5].

A second line of investigations came from the dendritic cell field established by early studies of Steinman at the Rockefeller University in New York in 1973 [6]. In 1994 this group described two distinct subsets (CD11c⁺ and CD11c⁻) of primary dendritic cells among CD4⁺ dendritic cells in peripheral blood [7]. Liu’s group earned credit for the finding that one of these dendritic cell subsets in peripheral blood (CD4⁺CD11c⁻) is identical to plasmacytoid T cells in lymphoid organs [8]. These results were supported by a study demonstrating that a large subset of DC in the T-cell dependent areas of human lymphoid organs is characterized by high levels of the IL-3 receptor α chain (CD123) [9].

A third independent line of investigations dates back to the early 1980s when it became clear that only a small fraction of cells in peripheral blood is responsible for the production of type I IFN in response to viral infection 10, 11. This cell subset could not be assigned to any of the established lineages such as monocytes, T cells or B cells, but it was found to express high levels of major histocompatibility complex (MHC) class II 12, 13. Important information came from Alm’s group who demonstrated that cells producing type I IFN in response to herpes simplex virus (by then called IFN-α producing cells, IPC) are found within a small population (1.4% of all peripheral blood mononuclear cells [PBMC]) of CD3⁻CD4⁺ cells in peripheral blood 14, 15. These cells were further characterized by the group of Fitzgerald-Bocarsly 16, 17. IPC responding to HIV were reported to belong to the fraction of CD4⁺ blood dendritic cells in peripheral blood [18]. Finally, in 1999 the groups of Liu and Colonna, by using different techniques, independently found that the IL-3 dependent dendritic cell subset in peripheral blood and plasmacytoid monocytes in lymphoid tissues are identical with the IPC 19, 20.

Different terms that are currently used to describe this cell type include plasmacytoid dendritic cells (PDC), interferon-producing cells (IPC) and DC2 (in contrast to Th1-inducing myeloid dendritic cells; DC1). We propose to use the term PDC for three reasons: the function of this cell is not limited to type I IFN production (the cells are not just IPC); if this cell is provided the appropriate stimulation, it is capable of inducing an IL-12 dependent Th1 response (the term Th2-inducing DC2 may not be suitable); and this cell is very potent at priming of allogeneic T-cell responses (this supports the term DC, although we are certainly aware that evidence for antigen processing and for antigen-specific priming of T-cell responses is scarce).

Plasmacytoid dendritic cells are stimulated by viruses or bacterial lysates, but they do not respond to LPS or poly I:C which are routinely used to stimulate APC. Small molecules such as imiquimod or resiquimod activate PDC via TLR7 [21], but the natural counterparts of these molecules have not been identified. To date the only defined microbial stimulus recognized by PDC is CpG DNA. What is CpG DNA?

The first evidence for immunostimulatory activity of bacterial DNA came from a series of elegant studies performed by Tokunaga 22, 23, 24. Later, a number of studies in the antisense field indicated that single stranded oligodeoxynucleotides (ODN) can have immunostimulatory properties 25, 26, 27, 28, 29. It was the achievement of Krieg to identify unmethylated CG dinucleotides within certain sequence contexts, so-called CpG motifs, to be responsible for the immunostimulatory properties of bacterial DNA 30, 31. Synthetic ODN containing such CpG motifs strongly supported immune responses in vitro and in vivo in mice 32, 33, 34, 35, 36, but the CpG specificity of such effects in the human immune system remained obscure at first [37]. Later, it turned out that mice and humans differ in the optimal CpG motif recognized. A distinct CpG motif for the human system was identified based on the activation of human primary B cells 38, 39. CpG ODN with this motif were potent vaccine adjuvants to elicit humoral immune responses in nonhuman primates 39, 40, 41 and in humans [42]. The potent adjuvant activity of these CpG ODN suggested that, similar to findings in mice 43, 44, also human dendritic cells might be sensitive to CpG ODN.

A convenient way to generate a large number of human dendritic cells is to incubate monocytes in the presence of GMCSF and IL-4 as established by Schuler’s group in 1994 [45]. In an attempt to study the effects of CpG ODN on human dendritic cells we first examined the response of human monocytes upon exposure to CpG ODN. To our surprise, unlike murine macrophages, human monocytes showed only a weak and delayed onset of the response upon stimulation with CpG ODN indicating that monocytes were activated within PBMC by an indirect mechanism [46]. Consistent with this finding, monocyte-derived DC did not respond to CpG ODN [47]. Since the Th2 cytokine IL-4 used to generate monocyte-derived dendritic cells may interfer with the stimulatory potential of the Th1 adjuvant CpG, we tested the effect of CpG ODN on primary DC isolated from peripheral blood. Indeed, CpG ODN strongly supported survival, activation, maturation, and T-cell stimulatory capacity of lineage-CD4⁺HLA-DR⁺ primary DC isolated from peripheral blood [47]. Within CD4⁺ blood DC, both the CD11c⁺ and CD11c⁻ subset of CD4⁺ blood DC responded to CpG ODN. However, further analysis on purified DC subsets revealed that only the CD11c⁻ DC were directly sensitive to CpG ODN, while purified CD11c⁺ DC were unresponsive. At that time, others found that the CD11c⁻ subset of CD4⁺ blood DC is identical with IPC and plasmacytoid T cells 19, 20.

The innate immune system has shaped the family of toll-like receptors (TLR) to detect different pathogen-derived microbial molecules [48]. It turned out that TLR9 is required for the recognition of CpG motifs within DNA 49, 50. The expression profile of the CD11c⁻ subset (PDC) and the CD11c⁺ subset (MDC) of CD4⁺ blood DC revealed that TLR expression in PDC is restricted to TLR7 and TLR9 (TLR1 is ubiquitous), whereas MDC express a broad profile of TLRs (TLR2, TLR3, TLR4, TLR5, TLR6, TLR8) but do not express TLR9 51, 52, 53. As a consequence, PDC were found to be sensitive to CpG ODN 52, 54, 55, 56, while MDC respond to other stimuli such as LPS but not to CpG ODN 51, 52, 53. These results supported the concept that DC subsets may have developed through distinct evolutionary pathways to recognize different microbial molecules and to elicit immune responses which are appropriate to defeat the corresponding pathogens.

CpG ODN strongly stimulated PDC to upregulate costimulatory molecules, HLA DR, the maturation marker CD83 resulting in an increased ability to stimulate allogeneic T cells [52]. PDC activated by CpG ODN produce inflammatory cytokines such as TNF-α and IL-6 52, 56. PDC rapidly upregulate CCR7 and IP-10 upon recognition of CpG ODN, whereas constitutive expression of CXCR3 is not affected. The inflammatory chemokine IP-10 is known to recruit CXCR3-expressing cells such as activated T cells 57, 58. Expression of CCR7 is known to mediate homing of immune cells to lymph nodes by binding of CCR7 to the chemokines SLC or ELC expressed on high endothelial venules and lymphatic endothelium [59]. Due to CCR7 expression, PDC may colocalize with naive T cells and central memory T cells [60] in the T cell areas of secondary lymphatic tissues, but this needs to be confirmed in vivo.

Although PDC produced measurable amounts of IFN-α in response to the CpG ODN sequences used in different studies 52, 54, 56, the quantity of IFN-α secretion upon stimulation with CpG alone in these studies was surprisingly low as compared to the extremely large amounts of IFN-α PDC are capable to produce upon viral stimulation [19].

Based on the ability to induce very large amounts of IFN-α in PDC we identified CpG ODN with particular sequences which not only induced activation and maturation of PDC but also stimulated the production of high amounts of IFN-α (up to 5 pg per single PDC; prototype sequence ODN 2216) [55]; the PDC was the only cell type within PBMC to produce IFN-α upon stimulation with ODN 2216. The only stimulus of PDC to date which induces comparable amounts of IFN-α in PDC are viruses (1 to 2 IU IFN-α per single PDC; corresponding to approximately 4 to 7 pg IFN-α) [61]. Thus, the structure provided by this group of CpG ODN seems to be interpreted by PDC as a molecular pattern for viral infection.

The CpG ODN which induce the highest levels of IFN-α have a chimeric backbone in which the 5′ and 3′ ends are phosphorothioate-modified, and the center portion is phosphodiester [55]. These CpG ODN share the following three features: (1) poly G sequences at both ends; (2) a central palindromic sequence; and (3) CG dinucleotides within the palindrome. All of these three structural elements contribute to the IFN-α inducing capacity of these CpG ODN. Control ODN which were unable to form G-tetrads (structure based on four strands of poly G) due to 7-deaza-guanosine substitutions within the poly G tails were inactive. Furthermore, the replacement of CG dinucleotides by GC or a non-palindromic sequence abolished the effect. The exact mechanisms by which these three components contribute to virus-like recognition by PDC remain to be established.

CpG ODN with these structural characteristics, and which induce very high amounts of IFN-α and IFN-β in PDC but are weak at activating B cells, are now termed CpG-A ODN (for example ODN 2216 31, 55). On the other hand, CpG ODN which selectively activate and mature PDC but are weak at inducing IFN-α and IFN-β are termed CpG-B ODN (for example ODN 2006). Unlike CpG-A ODN, CpG-B ODN are weak at activating NK cells but strongly stimulate B cells [31]. Similar types of CpG ODN were described by Klinman and coworkers based on NK cell activation and IFN-γ production in PBMC (D and K type CpG ODN in their studies correspond to CpG-A and CpG-B type ODN, respectively) 62, 63.

It is well established that PDC are capable of producing IFN-α 58, 61. In contrast, the ability of PDC to synthesize IL-12 is controversial 64, 65. We demonstrated that in PDC CpG ODN and CD40L synergistically induce large amounts of IL-12 including the bioactive form IL-12 p70, and that PDC stimulated with CpG ODN and CD40L are capable to drive naïve allogeneic CD4 T cells toward Th1 in an IL-12-dependent manner [52]. Interestingly, the ratio of IFN-α and IL-12 induced by CpG ODN and CD40L shifts toward IL-12 as PDC differentiate over time in culture with IL-3. The later in this differentiation process PDC encounter CpG ODN the more IL-12 and the less IFN-α is produced. The biological consequence of this is that the predominant production of IFN-α at the beginning of an immune response stimulates the proliferation of CD8⁺ memory T cells in vivo 66, 67, 68 and promotes innate effector mechanisms including activation of NK cells [69] which represent a first barrier to the pathogen. Later predominant IL-12 production supports priming of antigen specific T cell responses. Besides other sources, IL-3 can be produced by monocytes/macrophages and by TCR-triggered T cells 70, 71. Activated T cells may thus promote the development of IL-12-producing Th1-inducing PDC in two ways, via production of IL-3 and via expression of CD40L.

The CD40 pathway has been thought to play a greater role than microbial stimulation in eliciting IL-12 production by DC, implying that the secretion of this cytokine by DC is primarily regulated by feedback signals from activated T cells 72, 73, 74, 75, 76, 77. However, there is recent evidence that, unlike in vitro, in vivo the production of bioactive IL-12p70 heterodimer by DC requires both microbial and T cell-derived stimuli [78]. Human blood monocytes [79] as well as murine splenic DC express neither the p35 nor the p40 subunit of IL-12 constitutively and both IL-12 subunits are regulated separately by microbial stimuli and CD40L [78]. In vitro, mechanical stress during isolation of cells or cell culture conditions may substitute for a second signal 78, 80. In agreement with this, MDC exhibits spontaneous activation and maturation in cell culture and release significant amounts of IL-12 upon stimulation with either LPS or CD40L alone. The situation is different in PDC which in the absence of IL-3 show no activation but rather spontaneous apoptosis in cell culture. CpG ODN rescued PDC from apoptosis and markedly increased the expression of CD40, but high levels of IL-12 including bioactive IL-12 p70 were only produced after PDC received a second signal through CD40 ligation [52]. Thus IL-12 production of PDC is strictly controlled by two independent pathways, the presence of an exogenous microbial stimulus, and CD40 ligation which is provided endogenously by activated T cells.

Several early reports suggested that human monocytes, NK cells, and T cells are able to sense CpG ODN. When we used highly purified cell subsets, we made the observation that purified PDC and B cells but not monocytes, NK cells and T cells were sensitive to CpG ODN 81, 82. The quantitative analysis of the TLR expression profiles in these subsets revealed that the expression of TLR9 is associated with the sensitivity to CpG ODN. Although PDC (154 copies TLR9 per 1000 copies of cyclophilin B) and B cells (36 copies) expressed relatively high amounts of TLR9, freshly isolated human monocytes, T cells and NK cells expressed only low amounts (< 10 copies) that seem to be beyond the threshold of functional significance. Like monocytes, monocyte-derived DC and myeloid DC in peripheral blood are not directly activated by CpG ODN. It cannot be excluded that monocytes are capable to upregulate TLR9 expression along distinct differentiation pathways they enter and in response to certain cytokines or microbial molecules they encounter. However, so far we have no evidence for this. Of note, low numbers of “84contaminating” PDC (0.1%) in otherwise pure cell subsets (> 99%) are sufficient to cause CpG-mediated indirect activation of otherwise non-responding cell types. This might explain some of the conflicting results seen in earlier studies.

It is important to note that TLR9 expression differs between mice and humans. In contrast to humans, in mice cells of the myeloid lineage such as monocytes and myeloid dendritic cells express TLR9 and respond directly to CpG ODN 83, 84. This leads to differences in the target cells that are activated and the cytokines that are induced by CpG ODN in mice and humans, and calls for caution in extrapolating results from mice to primates and humans.

In humans, direct activation of PDC and B cells leads to strong indirect activation of other immune cell subsets (Figure 1). We observed striking differences of CpG type A and CpG type B regarding secondary stimulation of monocytes, T cells and NK cells reflecting distinct activation of PDC and B cells by these CpG ODN. Via PDC-derived IFN-α/β, CpG type A ODN induce an antigen-independent partial activation of memory CD8 T cells [81], strongly activate NK cells (IFN-γ production and lytic activity) and promote proliferation, IFN-γ production and lytic activity of γδ T cells [82]; in contrast CpG type B ODN are less potent activators of NK cells and γδ T cells. Both CpG type A and CpG type B indirectly upregulate costimulatory molecules in monocytes and myeloid dendritic cells. Together these indirect effects of CpG ODN emphasize the potent regulatory role of PDC in the human immune system and highlight the pivotal role of PDC in the functional profile of different types of CpG ODN.

To study PDC and CpG DNA, a number of important methodological issues have to be addressed. Despite the relatively low frequency of PDC in peripheral blood (0.2% to 0.4%), stimulated PDC strongly impact on other immune cell subsets. As a consequence, in CpG DNA experiments, small numbers of PDC (e.g., 0.1%) in otherwise pure cell populations (even 99% purity) can lead to the false conclusion of a direct effect of CpG DNA on this cell population. This has been a problem in the past for monocytes, NK cells and T cells for which direct effects of CpG ODN have been claimed but none of which directly responds to CpG ODN when PDC are completely absent. Depletion of PDC prior to the routine isolation protocol of the corresponding cell type is recommended. Of note, since IL-4 is known to induce apoptosis in PDC, such an indirect PDC-mediated effect is not seen when monocyte-derived dendritic cells are generated in the presence of GMCSF and IL-4 and then stimulated by CpG ODN.

Valid identification of stimulated PDC within peripheral blood mononuclear cells is usually limited to early time points (up to 24 hours), because activated PDC downregulate CD123 and increase in size and background staining. This is not a factor for experiments with purified PDC. The PDC-specific antibody BDCA-2 used for isolation of PDC interfers with CpG-induced (and virus-induced) type I IFN production; the PDC-specific antibody BDCA-4 does not (own unpublished observations). Upon irradiation (as used in some allogeneic MLR protocols), PDC loose their ability to produce type I IFN. For intracellular IFN-α staining in PDC, the routinely used blockade of cytokine secretion by brefeldin A may interfer with the IFN-a-mediated autocrine feedback loop [55].

For testing of PDC in T cell assays it has to be considered that IFN-α at high concentrations inhibits cell proliferation in general. Nuclease stability of a CpG ODN depends on the backbone of the ODN (phosphorothioate > phosphodiester) and determines the concentration of ODN to be used. The stimulatory activity of a CpG ODN usually decreases at higher concentrations; thus the optimal concentration for each experimental setting has to be determined. CpG ODN with phosphorothioate backbone but not with unmodified backbone (phosphodiester) may show CpG-independent immunostimulatory activity which may not be predictable by the sequence and therefore has to be determined experimentally.