Toll-like receptor expression in murine DC subsets: lack of TLR7 expression by CD8α+ DC correlates with unresponsiveness to imidazoquinolines

Authors

  • Alexander D. Edwards,

    1. Immunobiology Laboratory, Cancer Research UK, London Research Institute, London, GB
    Search for more papers by this author
    • The first two authors contributed equally to this work.

  • Sandra S. Diebold,

    1. Immunobiology Laboratory, Cancer Research UK, London Research Institute, London, GB
    Search for more papers by this author
  • Emma M. C. Slack,

    1. Immunobiology Laboratory, Cancer Research UK, London Research Institute, London, GB
    Search for more papers by this author
  • Hideyuki Tomizawa,

    1. Pharmaceuticals and Biotechnology Laboratory, Japan Energy Corporation, Saitama, Japan
    Search for more papers by this author
  • Hiroaki Hemmi,

    1. Department of Host Defense, Research Institute for Microbial Diseases, Osaka University, Osaka, Japan
    2. Solution-oriented Research for Science and Technology, Japan Science and Technology Corporation, Osaka, Japan
    Search for more papers by this author
  • Tsuneyasu Kaisho,

    1. Department of Host Defense, Research Institute for Microbial Diseases, Osaka University, Osaka, Japan
    2. Solution-oriented Research for Science and Technology, Japan Science and Technology Corporation, Osaka, Japan
    3. RIKEN Research Center for Allergy and Immunology, Kanagawa, Japan
    Search for more papers by this author
  • Shizuo Akira,

    1. Department of Host Defense, Research Institute for Microbial Diseases, Osaka University, Osaka, Japan
    2. Solution-oriented Research for Science and Technology, Japan Science and Technology Corporation, Osaka, Japan
    Search for more papers by this author
  • Caetano Reis e Sousa

    Corresponding author
    1. Immunobiology Laboratory, Cancer Research UK, London Research Institute, London, GB
    • Immunobiology Laboratory, Cancer Research UK, London Research Institute, Lincoln's Inn Fields Laboratories, 44 Lincoln's Inn Fields, London WC2A 3PX, GB Fax: +44-20-7269-2833
    Search for more papers by this author

Abstract

Toll-like receptors (TLR) recognize microbial and viral patterns and activate dendritic cells (DC). TLR distribution among human DC subsets is heterogeneous: plasmacytoid DC (PDC) express TLR1, 7 and 9, while other DC types do not express TLR9 but express other TLR. Here, we report that mRNA for most TLR is expressed at similar levels by murine splenic DC sub-types, including PDC, but that TLR3 is preferentially expressed by CD8α+ DC while TLR5 and TLR7 are selectively absent from the same subset. Consistent with the latter, TLR7 ligand activates CD8α DC and PDC, but not CD8α+ DC as measured by survival ex vivo, up-regulation of surface markers and production of IL-12p40. These data suggest that the dichotomy in TLR expression between plasmacytoid and non-plasmacytoid DC is not conserved between species. However, lack of TLR7 expression could restrict the involvement of CD8α+ DC in recognition of certain mouse pathogens.

Abbreviations:
DN:

Double-negative

PDC:

Plasmacytoid DC

TLR:

Toll-like receptor

1 Introduction

Recognition of infection by APC is the primary trigger for the initiation of adaptive immune responses. Among several strategies to sense infection, APC use cell-bound pattern-recognition receptors (PRR) that recognize conserved microbial structures and signal the cells for activation 1. Toll-like receptors (TLR) have emerged as a major family of PRR, recognizing a wide spectrum of patterns ranging from bacterial and yeast cell wall components through nucleic acids to protozoan lipoproteins 1. For example, TLR2 in co-operation with TLR1 or TLR6can recognize bacterial lipoproteins and peptidoglycan, TLR3 can mediate innate activation by viral double-stranded RNA, TLR4 is the primary LPS receptor, TLR5 senses bacterial flagellin and TLR9 is the receptor for bacterial DNA 1. In addition, TLR can act as receptors for synthetic pharmaceuticals previously identified by their immunostimulatory properties. Thus, TLR7/– mice have recently been shown to be completely unable to respond to the imidazoquinoline compounds imiquimod and R-848 2, and TLR4 can act as a receptor for the anti-cancer drug taxol 3.

It is clear that TLR expression is not uniform among leukocytes, raising the possibility that different cell types may be specialized to recognize distinct classes of pathogens. For example, an early systematic analysis of expression of TLR1–5 revealed that TLR1 is expressed widely among human peripheral blood leukocytes, but TLR2–5 are expressed much more selectively 4. The only cell type expressing all five TLR in that study was the monocyte-derived dendritic cell (DC). DC are a family of APC encompassing several subtypes prominent in the initiation of immune responses to infection 5. Broad expression of TLR among DC could mean that APC are under pressure to maintain a large repertoire of PRR to enable them to respond to any pathogen 4. However, subsequent work revealed important differences in TLR expression among subtypes of human DC 610. In particular, TLR9 message was shown to be restricted to plasmacytoid DC (PDC) and not to be expressed on monocyte-derived or CD11c+ blood DC 610. In addition, some studies 8, 9, but not others 7, 11, have suggested that TLR7 expression is also restricted to PDC. Conversely, human PDC were found not to express most other TLR, including the LPS receptor, TLR4 610. This pattern of expression correlates with the observation that CpG DNA activates PDC but not CD11c+ or monocyte-derived DC, while the latter but not the former respond strongly to LPS 69. These observations have generally been linked to the idea that recognition of different pathogens by different DC could allow generation of alternative Th1 or Th2 responses to infection 12. However, the role of DC subtypes in Th1/Th2 determination remains unclear and a large body of data suggests that a simple link between DC ontogeny and function is unwarranted 13.

Although extensive analysis of TLR expression in the mouse has not been reported, analysis of genomic sequence shows species-specific differences in promoter regions that suggest that expression patterns may differ between mouse and human 14. Various DC types have been identified in mouse secondary lymphoid tissues, including three CD11chi subsets [CD8α+, CD4+ and double-negative (DN) CD4CD8α DC] 15 and one subset of CD11clo plasmacytoid DC 1619. We measured expression of mRNA for TLR1–9 in all these DC subsets after direct isolation from mouse spleen. We find that although TLR1, 2, 4, 6, 8 and 9 are expressed by all DC subsets studied, significant differences in expression of TLR3, 5, and 7 are seen between CD8α+ and CD8α DC, but not between conventional CD11chi DC and PDC. Consistent with the lack of TLR7 mRNA expression in CD8α+ DC, TLR7 ligand did not activate that subset. These results confirm that certain DC subtypes may be preferentially involved in recognition of distinct classes of pathogens across species, but suggest that the dichotomy between PDC and non-plasmacytoid DC observed in humans does not extend to mouse.

2 Results and discussion

Splenocyte preparations were enriched for CD11c+ and for Ly6C+ cells using magnetic selection, and the enriched populations were used for the purification of the four known murine spleen DC subsets using five-color high speed cell sorting. Four populations were routinely isolated with greater than 95% purity (data not shown) and included CD11chi Ly6CB220 (classical) DC and CD11cloLy6ChiB220+ PDC. The former were further subdivided into CD4+, CD8α+ and DN subsets 15. The cDNA from each of the four DC populations was used for semi-quantitative PCR amplification of TLR message using specific primers. As shown in Fig. 1A, by this method, most TLR appeared to be expressed by all murine DC subsets, and there were only three TLR for which there was obvious differential expression: TLR3 appeared to be expressed at highest levels by CD8α+ DC while TLR5 and TLR7 appeared to be selectively absent from the same population (Fig. 1A). This was confirmed by quantitative PCR for TLR3, TLR5 and TLR7 (Fig. 1B): TLR3 levels were highest in CD8α+ DC, intermediate in DN DC and at least tenfold lower in PDC and CD4+ DC; TLR5 levels were highest in CD4+ DC and lowest in CD8α+ DC; TLR7 mRNA levels were highest in PDC and were tenfold lower in CD8α+ DC than in either CD8α DC population. In contrast, levels of TLR9 message varied by less than fivefold among subsets (Fig. 1B), in marked contrast to results obtained in human 610. These data suggest that gross variation in the TLR repertoire of murine DC subsets is confined to TLR3, TLR5 and TLR7. However, we cannot exclude the possibility that there are further differences at the levels of TLR protein expression, which cannot be assessed due to lack of appropriate antibodies.

To analyze the functional significance of differential TLR mRNA expression, we chose to confine our analysis to a comparison of the response of DC subsets to ligands for TLR7 and TLR9. There were two reasons for this choice: (1) Of all the TLR, TLR7 mRNA showed the greatest discrepancy among subsets, being expressed at 55-fold lower levels in CD8α+ DC than in CD4+ DC; in contrast, TLR9 only differed by 3-fold in a similar comparison although it shows the greatest differential expression between subsets in the human system 610. (2) TLR7 and TLR9 are two TLR for which synthetic ligands are available, thus, precluding the issue of cross-contamination that sometimes confounds the interpretation of experiments done using TLR ligands such as bacterial flagellin purified from microbial organisms. TLR3 is the second TLR showing the greatest discrepancy in expression among subsets (28-fold higher expression in CD8α+ DC than in CD4+ DC) and one for which a synthetic ligand is also available. We compared the responses of all DC subsets to synthetic dsRNA (data not shown). However, we found that DC responses to synthetic double stranded RNA can be TLR3-independent (Diebold et al, manuscript in preparation) and, therefore, were not able to use that stimulus to assess the significance of differential TLR3 expression in DC subsets.

TLR triggering leads to production of Th1-promoting cytokines such as IL-12 and type I interferons (IFN-I) 1. Among murine DC subsets, CD8α+ DC have been reported to be the primary producers of IL-12 20. However, PDC, CD4+ DC and DN DC can all make IL-12p40 in response to an appropriate stimulus, even though CD4+ DC are selectively unable to produce bioactive IL-12p70 21, 22, and unpublished observations]. Thus, IL-12p40 (but not IL-12p70) may be used as a universal marker for activation of all DC subsets in response to TLR triggering. When splenocyte suspensions enriched for either conventional DC and PDC were stimulated with an oligonucleotide-containing unmethylated CpG motifs (CpG), a ligand for TLR9 23, both CD8α+ and CD8α DC as well as PDC (Ly6Chi) stained for IL-12p40 (Fig. 2). In contrast, when the same cells were stimulated with a specific TLR7 ligand, the imidazoquinoline R-848 2, IL-12p40 staining was limited to the CD8α and PDC subsets (Fig. 2). IL-12p40 production was not detectable if the cells were taken from MyD88–/– mice (Fig. 2), demonstrating the involvement of TLR signaling in the response to either stimulus. Thus, the differential production of IL-12p40 by CD8α+ DC in response to CpG vs. R-848 is consistent with the absence of TLR7 and presence of TLR9 mRNA in that subset (Fig. 1B).

To analyze responsiveness to the stimuli in the absence of cross-regulation by other DC or non-DC contaminants in the cultures, CD8α+ DC, CD8α DC and PDC were purified by cell sorting and stimulated with a dose-range of CpG or R-848 in the absence of other cell types. As shown in Fig. 3, there was a correlation between TLR7 mRNA levels and the ability of individual DC subsets to make IL-12p40 in response to R-848. Thus, PDC were much more sensitive to low doses of R-848 than CD8α DC and, consistent with their apparent lack of TLR7 expression, CD8α+ DC did not make IL-12p40 in response to any dose of the stimulus (Fig. 3). In contrast to TLR7 ligand, all subsets produced IL-12p40 in response to CpG. Dose-response analysis revealed that PDC were the least sensitive of all subsets to CpG despite expressing the highest amount of TLR9 mRNA, although they did produce higher absolute levels of IL-12p40 than other subsets in response to saturating doses of ligand (Fig. 3). Interestingly, CD8α+ DC expressed the lowest amounts of TLR9 message (Fig. 1B), but displayed the highest sensitivity to CpG (Fig. 3). This analysis underscores the danger of extrapolating from small differences in TLR mRNA levels to predictions about responsiveness to individual ligands. Differences in responsiveness to CpG could reflect alternative pathways for ligand uptake by different DC subsets, as TLR9 signaling requires access of CpG to endocytic compartments 24.

To confirm that the lack of IL-12p40 production was representative of a general failure of CD8α+ DC to respond to TLR7 ligand, we assessed other parameters of DC activation. Purified DC rapidly undergo apoptosis ex vivo unless activated by innate signals, cytokines or CD40L (25, and unpublished observations). When purified CD8α+ DC were cultured in medium alone, viability dropped to 15% overnight (Fig. 4A). Addition of CpG or GM-CSF to the cultures rescued a significant proportion of the cell from apoptosis and increased their viability by three to fourfold (Fig. 4A). In contrast, R-848 failed to rescue CD8α+ DC from apoptosis (Fig. 4A). Therefore, by this additional criterion, R-848 again failed to act as a CD8α+ DC activator.

Up-regulation of surface markers such as CD86 and CD40 is a further parameter of DC activation. Although culturing primary DC in medium alone is sufficient to initiate up-regulation of these markers, higher expression levels are achieved in the presence of microbial activators 22. Indeed, CpG addition to the medium led to an increase in CD40 expression by both CD8α+ and CD8α DC compared to control cells cultured in medium alone (Fig. 4B). Similarly, R-848 caused an increase in expression of CD40 by CD8α DC (Fig. 4B). However, even when CD8α+ DC were cultured in GM-CSF to maintain viability, addition of R-848 failed to promote up-regulation of CD40 (Fig. 4B) or CD86 (data not shown). Similarly, rescue by GM-CSF or CD40L did not result in IL-12p40 synthesis in response to R-848 (data not shown). We conclude that CD8α+ DC selectively fail to respond to TLR7 ligands.

The notion that DC subsets evolved to respond to different pathogens has gained support from the discovery that human blood PDC differ from CD11c+ DC and monocyte-derived DC by the expression of TLR9 and absence of expression of TLR2, 3, 4 and 8 611. Here, we show that this dichotomy in TLR repertoire does not extend to the mouse. Other than TLR3, murine PDC purified from spleen express all known TLR, at least at the message level. Interestingly, human PDC also fail to express TLR3 26. In contrast, there is gross variation in TLR3, 5 and 7 mRNA expression among the conventional CD11chi DC subsets. For TLR7, this translates into differential responsiveness of CD8α+ and CD8α DC to a synthetic TLR7 ligand. Some reports have suggested that, like murine CD8α+ DC, human CD11c+ DC also do not express TLR7 mRNA 8, 9. However, other groups have found TLR7 expression in human CD11c+ DC 7, 11 and shown that this correlates with responsiveness to imidazoquinolines 11, although the interpretation of the latter finding is complicated by the fact that, in human (but not in mouse), imidazoquinolines have recently been shown to also signal via TLR8 27. DC in mouse are generally purified from tissues, whereas in humans, they are taken from blood, making it hard to compare mouse and human data. It remains to be determined whether a DC subset unresponsive to TLR7 ligands also exists in man and, conversely, whether other murine DC subsets, yet to be found, lack TLR9 expression as in the human case.

The significance of differential TLR expression and, by inference, selective pathogen recognition by DC subsets remains undetermined. It has often been linked to the idea that subsets are ontogenetically programmed to produce different cytokines and induce alternative forms of adaptive immunity 12. However, it is important to recognize that differential pathogen recognition by APC need not be related to ‘Th’ induction. Indeed, differential TLR expression by DC subsets may be of little consequence for ‘Th’ differentiation, as it is believed that TLR triggering is primarily involved in Th1 rather than Th2 responses 28. Whether selective pathogen recognition by individual DC subsets relates to the induction of differential Th responses to infection will, instead, require comparison of the distribution of TLR vs. that of PRR for Th2-promoting parasites (if such receptors exist). However, it is increasingly clear that ontogeny is not the primary factor dictating DC subset Th-directing ability. Both mouse and human DC are flexible in their cytokine response to activation and can induce Th1 or Th2 differentiation in response to signals from microbes or other exogenous stimuli 13. Thus, it remains possible that differential TLR expression by DC subtypes is related to other specializations offunction. A microbial ligand for TLR7 has yet to be identified, but the prediction from data presented here is that CD8α+ DC will be unresponsive to it. This may offer the opportunity to test the significance of differential TLR expression by DC subsets in a real infection model.

Figure 1.

 Analysis of TLR expression in splenic DC subsets. (A) DC subsets were purified by cell sorting as detailed in Sect. 4, and cDNA was amplified by semi-quantitative PCR using sets of primers specific for the murine TLR1–9 genes. Data are representative of 3–5 PCR reactions using 2–5 sets of independently-sorted cells. (B) TLR3, 5, 7 and 9 expression was further analyzed by quantitative PCR. Data represent mean ± SD of four independently sorted DC samples measured in three different PCR reactions.

Figure 2.

 MyD88-dependent production of IL-12p40 in response to R-848 by CD8α DC and PDC, but not by CD8α+ DC. Spleen cells from C57BL/6 or MyD88–/– mice were enriched for Ly6C or for CD11c and were cultured for 1 h with the indicated stimuli. Brefeldin A (5 μg/ml final concentration) and GM-CSF were added, and the cells were fixed 15 h later. Cells were permeabilized and stained for IL-12p40 and either Ly6C or CD11c and CD8α. Events shown are gated on scatter (Ly6C staining) or on CD11chi cells (CD11c/CD8α staining). The Ly6ChiIL-12+ cells shown in the top dot plots correspond to PDC as assessed independently by staining for B220 (not shown). Data are representative of three independent experiments. Similar patterns of staining were seen at various time points (not shown). Note the lower frequency of CD8α+ DC in the samples cultured in medium alone or with R-848 compared to the sample with CpG. This is due to selective death of CD8α+ DC in the absence of stimulation (see Fig. 4).

Figure 3.

 Differential response of CD8α+ DC versus other DC subsets to TLR7, but not TLR9 ligands. Sorted PDC and CD11chiCD8α+ and CD8α DC subsets were cultured overnight with the indicated concentrations of CpG or R-848. Levels of IL-12p40 in supernatants were measured by ELISA. Data are representative of four independent experiments.

Figure 4.

 R-848 does not activate CD8α+ DC. CD8α+ and CD8α DC were purified by cell sorting and cultured overnight with 2 μg/ml CpG or 300 nM R-848 in the presence or absence of GM-CSF. (A) Scatter plots of cultured CD8α+ DC illustrate that significant apoptosis occurs upon culture in medium alone, which is prevented either by stimulation with an appropriate TLR ligand (CpG) or by addition of GM-CSF (lower panel). Numbers represent the percentage of events falling within the indicated live cell gate. (B) Both CpG and R-848 induce up-regulation of CD40 on CD8α DC, whereas only CpG induces CD40 up-regulation on CD8α+ DC. Cells from the GM-CSF-containing cultures in (A) were analyzed for CD40 expression. Dotted line: medium alone; solid line: stimulus. Data are representative of two experiments.

3 Materials and methods

3.1 Animals

Male and female BALB/c or C57BL/6 mice (8–12 weeks old) were obtained from Harlan UK (Bicester, Oxon) or from the breeding unit of Cancer Research UK (Clare Hall Laboratories, South Mimms, GB). BALB/c mice were used unless otherwise stated. MyD88–/– mice 29 on a C57BL/6 background were bred at Cancer Research UK.

3.2 Reagents

The CpG-containing phosphorothioate-linked oligonucleotide 1668 5′-TCCATGACGTTCCTGATGCT-3′ 30 was made by the Cancer Research UK oligonucleotide synthesis service. R-848 was synthesized by the Pharmaceuticals and Biotechnology Laboratory, Japan Energy Corporation (Saitama, Japan). Recombinant GM-CSF was made by the Cancer Research UK protein purification service. Each GM-CSF batch was titrated for generation of bone marrow-derived DC and used at an optimal concentration (generally around 1 μg/ml).

All mAb used were from PharMingen (BD Biosciences, San Jose, CA) or Caltag (San Francisco, CA): HL3, a hamster IgG mAb against CD11c; RM4–5, RA3–6B2, 3/23, GL1and 53–6.7, rat IgG2a mAb againstCD4, B220, CD40, CD86 and CD8α, respectively; AL-21, a rat IgM mAb against Ly6C; C15.6, a rat IgG1 anti-IL-12p40.

3.3 Cells

Spleen cell suspensions were prepared by LiberaseTM CI (Roche Diagnostics Ltd., Lewes, GB) and DNaseI digestion. Splenocytes were incubated with a mixture of FITC- and biotin-conjugated mAb to Ly6C, followed by a mixture of streptavidin-coated and anti-CD11c-coated MACS® magnetic beads (Miltenyi Biotec Ltd., Bisley, GB). Positive magnetic selection, using the MACS system (Miltenyi Biotec Ltd.), resulted in a population highly enriched for conventional DC as well as PDC (SD and CRS, unpublished observations). DC-enriched preparations were then stained with PE-anti-CD11c, TriColor-anti-B220 and APC-anti-CD8α and APC/Cy7-anti-CD4 and were sorted in a MoFlo cytometer (Cytomation, Fort Collins, CO) using 5-colour gating: conventional DC were defined as CD11chiLy6CloB220 and were separated based on differential expression of CD8α and CD4, as described 22; PDC were defined as CD11cdimLy6ChiB220+ cells. Sorted cells were >95% pure upon re-analysis. In some experiments, CD11chi DC were sub-divided only on the basis of CD8α expression into CD8α+ and CD8α subsets by omitting the anti-CD4 antibody.

For intracellular cytokine staining experiments, splenocytes were first enriched for Ly6C+ cells, and the depleted fraction was subsequently enriched for CD11c+ cells. The former population contains Ly6ChiB220+CD11cdim PDC, as well as other cells that are Ly6C+B220CD11c.

3.4 PCR

Total RNA from sorted DC subsets was isolated using the RNeasy minikit from Qiagen (Crawley, GB). Total cDNA was synthesized with the Superscript II pre-amplification system (GIBCO BRL). PCR primer pairs are described in Table 1. Semi-quantitative PCR was carried out using standard amplification conditions. Samples were removed and analyzed by standard agarose gel electrophoresis after 20, 30 or 35 amplification cycles. The data shown for each primer pair correspond to the cycle number at which the amplicons could be detected but were not yet at saturation. The specificity of the primers for the intended target was confirmed by restriction digest analysis of the PCR product (data not shown). Quantitative PCR was carried out on a ABI PRISM 7700 detection system (PE Applied Biosystems, Warrington, UK) using SYBR Green (Molecular Probes, Europe BV, Leiden, The Netherlands). In each sample, message levels were calculated by comparison with a standard curve generated using serial dilutions of a reference cDNA sample. These levels were then expressed relative to levels of GAPDH as recommended by the manufacturer (PE Applied Biosystems). In order to average the data across multiple independent PCR using different reference cDNA, the data were expressed relative to the levels found in CD4+ DC, the most abundant DC population in mouse spleen.

3.5 Flow cytometry

IL-12p40, CD86 and CD40 staining was performed as described 22. Briefly, for IL-12p40 staining, cells were fixed in paraformaldehyde, resuspended in PBS/EDTA containing 1% FCS, 0.02% sodium azide and 0.1% saponin (Sigma) and stained with appropriate mAb conjugated to different fluorophores. After 30 min at 4°C, cells were washed twice in PBS/EDTA containing 1% FCS and 0.02% sodium azide without saponin. A similar procedure was used for CD40 and CD86, omitting the fixation procedure and saponin and substituting anti-CD40 or anti-CD86 for anti-IL-12p40. Cell acquisition was performed on a FACSCalibur® flow cytometer (BD Biosciences, San Jose, CA), and data were analyzed using FlowJo software (Tree Star Inc., San Carlos, CA).

Table 1. PCR primers used for TLR amplification
original image

Acknowledgements

This study was supported by Cancer Research UK. We thank Derek Davies, Gary Warnes, Cathy Simpson and Ayad Eddaoudi for cell sorting. We are grateful to members of the Immunobiology Laboratory, Cancer Research UK, for discussions. S.S.D. is supported by an EMBO fellowship and a Cancer Research UK postdoctoral fellowship.

Footnotes

  1. 1

    WILEY-VCH

  2. 2

    WILEY-VCH

  3. 3

    WILEY-VCH

  4. 4

    WILEY-VCH

Ancillary