GM-CSF increases cross-presentation and CD103 expression by mouse CD8+ spleen dendritic cells

Authors

  • Yifan Zhan,

    Corresponding author
    1. Immunology Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    2. Department of Medical Biology, University of Melbourne, Parkville, Victoria, Australia
    • Autoimmunity and Transplantation Division, The Water and Eliza Hall Institute of Medical Research, 1G Royal Parade, Parkville 3052, Victoria, Australia Fax: +61-03-93470852
    Search for more papers by this author
    • Co-senior authors.

  • Emma M. Carrington,

    1. Immunology Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    2. Department of Medical Biology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Annemarie van Nieuwenhuijze,

    1. Inflammation Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    Search for more papers by this author
  • Sammy Bedoui,

    1. Department of Microbiology and Immunology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Shirley Seah,

    1. Immunology Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    2. Department of Medical Biology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Yuekang Xu,

    1. Inflammation Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    Search for more papers by this author
  • Nancy Wang,

    1. Department of Microbiology and Immunology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Justine D. Mintern,

    1. Inflammation Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    2. Department of Medical Biology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Jose A. Villadangos,

    1. Inflammation Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    2. Department of Medical Biology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Ian P. Wicks,

    1. Inflammation Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    2. Department of Medical Biology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author
  • Andrew M. Lew

    1. Immunology Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia
    2. Department of Medical Biology, University of Melbourne, Parkville, Victoria, Australia
    Search for more papers by this author

Abstract

Resident CD8+ DCs perform several functions, including cross-presenting antigen and rapidly engulfing the Gram-positive intracellular pathogen Listeria monocytogenes. Little is known about how these functions of CD8+ DCs are modulated. Here, we show that granulocyte-macrophage CSF (GM-CSF), a cytokine that exists at low levels at steady state but is elevated during infection and inflammation, enhances cross-presentation and rapid uptake of L. monocytogenes by resident CD8+ DCs. This previously unrecognized functional enhancement of CD8+ DCs by GM-CSF was independent of promoting DC survival in vitro. Enhancement of these functions by GM-CSF was also marked by CD103 expression on CD8+ DCs that was strongly regulated by GM-CSF. Our findings not only identify GM-CSF as a key molecule regulating CD8+ DC function, but also as a factor responsible for functional heterogeneity of CD8+ DCs that is at least substantially demarcated by CD103 expression.

Introduction

DCs found in the spleen can be grossly separated into CD8+ DCs and CD8 DCs. The latter includes CD8CD4+ and CD8CD4 conventional DCs and CD45RA+ plasmacytoid DCs. Different DC subsets have different functions. Notably, the resident CD8+ DCs show superior capacity for ingesting cell-associated antigens 1, for cross-presenting antigens (i.e. presenting exogenous antigens to CD8+ T cells) 2–4, for the rapid uptake of certain bacterial pathogens 5 and for producing IL-12 6. It has been previously reported that a small proportion of CD8+ DCs express the integrin CD103 and these CD103-expressing CD8+ DCs are enriched in cells engulfing cellular antigens and cross-priming CD8+ T cells 7. However, it is unclear what regulates the above functions of CD8+ DCs and what controls CD103 expression on CD8+ DCs.

Differentiation of resident CD8+ DCs is mainly driven by Flt3L 8. Flt3L deficiency in vivo can result in a tenfold reduction in the numbers of resident DCs 9, reinforcing a critical role for Flt3L in the differentiation of resident DCs. Furthermore, certain transcription factors have been shown to influence the lineage development of resident DC subsets. For example, the absence of IRF4 results in a substantial deficiency in CD8 DCs 10, whereas the absence of IRF8 or Batf3 results in a substantial deficiency in CD8+ DCs 11, 12. Beyond lineage ontogeny, the conditions and factors that drive the functional differentiation of resident DCs have not been well-elucidated.

Granulocyte-macrophage CSF (GM-CSF) is a cytokine that exists at low levels at steady state but is elevated during infection and inflammation 13. It widely used to promote human hematopoiesis and to differentiate DCs in vitro 14, 15. Side-by-side comparisons concluded that GM-CSF mainly differentiates monocyte-derived DCs and Flt3L differentiates the equivalents of resident CD8+ DCs 16. Although exogenous GM-CSF seems dispensable for the development of CD8+ DCs 16, GM-CSF has been used in the cultures of resident DCs including CD8+ DCs for functional assays to promote DC survival 17. It was not known whether GM-CSF is required for the potentiation of function of CD8+ DCs in vivo; nor what the molecular target of GM-CSF is.

Here, we investigated whether GM-CSF might influence the function of CD8+ DCs and CD103 expression. Functionally, DCs from GM-CSF transgenic (GMtg) mice had enhanced cross-presentation, whereas in the absence of GM-CSF signaling, cross-presentation was reduced. Rapid uptake of Listeria monocytogenes by CD8+ DCs was also reduced in the absence of GM-CSF. Phenotypically, we showed that bacterial infection preferentially increased CD103 expression by CD8+ DCs and this increase was dependent on GM-CSF. Together, our results identify a critical role for GM-CSF in preferentially regulating expression of the integrin CD103 by CD8+ DCs and in regulating certain functions of CD8+ DCs.

Results

GM-CSF enhances cross-presentation by CD8+ DCs

GM-CSF is a cytokine that exists at very low levels at steady state but is greatly induced during infection and inflammation 13. To understand the role of GM-CSF in regulating the functions of CD8+ DCs, we chose mouse models with GM-CSF overexpression. We first compared the cross-presentation by CD8+ DCs from GMtg mice and control littermates. When soluble OVA was used as the antigen in vitro, CD8+ DCs from GMtg mice had enhanced capacity to stimulate a CD8+ T-cell (OT-I) proliferative response (Fig. 1A). Production of T-cell cytokines such as IFN-γ and IL-17 by CD8+ T cells was also enhanced in cultures with GMtg DCs (Fig. 1A). Exogenous GM-CSF (2 ng/mL) was included in all cultures to exclude potential difference in the levels of endogenous GM-CSF produced by DCs and T cells. Notably, enhanced proliferation of CD8+ T cells with GMtg DCs was not seen during direct presentation with OVA peptide (SIINFEKL) (Fig. 1B). Similarly, OVA-protein-induced proliferation of CD4+ T cells (OT-II) with GMtg CD8+ DCs was also not significantly enhanced, compared with DCs from littermates (Fig. 1C). As with soluble antigen, enhancement of OT-I proliferation was also evident with cell-associated antigen (Fig. 1D).

Figure 1.

GM-CSF enhances cross-presentation by CD8+ DCs. CD8+ DCs were isolated from GMtg and WT littermates. DCs (20 000/well in triplicates) were cultured with 50 000 (A, B, and D) CSFE-labeled OT-1 cells or (C) CSFE-labeled OT-II cells. (A–C) Soluble OVA and OVA peptide were used as stimuli; or (D) graded numbers of OVA-coated spleen cells (OCS) as stimuli. Cells were cultured for 3 days. Harvested cells were measured for cell proliferation. Harvested supernatants were assayed for cytokines. Mean and SEM are shown. *p<0.05 (Student's t-test by Graphpad Prism). Three similar experiments were performed with similar results.

Given that GM-CSF can promote cell survival, it is possible the above enhancement was due to differences in survival of the two types of DCs. To test this, we compared the survival of isolated CD8+ DCs from GMtg mice and littermate controls. We found that there was no difference in the numbers of recovered viable DCs between the two types of mice in the presence or absence of exogenous GM-CSF (Fig. 2A). Thus, enhancement of OT-1 cell proliferation with CD8+ DCs from GMtg mice compared with WT littermates is unlikely to be due to DC survival.

Figure 2.

CD8+ DCs from GMtg mice and WT littermates have similar survival rates and surface phenotypes. (A) Sorted CD8+ DCs from GMtg and WT littermates were cultured in the absence or presence of GM-CSF (2 ng/mL) for 36 h. Recovered viable cells were calculated by the inclusion of calibrating beads in the propidium iodide-negative cell fraction. Mean and SEM of triplicate wells are shown. (B) Spleen DCs from GMtg and WT littermates were stained for costimulation and MHC molecules. Three independent experiments were performed with similar results.

Apart from DC survival, we also investigated the expression of surface molecules that participate in T-cell priming, and found that DCs from GMtg mice and WT littermates expressed comparable levels of CD80 and CD86 (Fig. 2B). They also express similar levels of MHC class II and class I (Db and Kb) (Fig. 2B).

Despite low levels of GM-CSF during steady state, CD8+ DCs from WT mice stimulated a stronger response of CD8+ T cells than CD8+ DCs from GMKO (GM-CSF knockout) mice (numbers of proliferating OT-I cells halved in GMKO; Fig. 3A). Again, such differences were not found for MHC class II presentation (proliferation of CD4+ T cells; Fig. 3B) or direct MHC class I presentation with SIINFEKL peptide (Fig. 3C).

Figure 3.

Defective GM-CSF signaling reduces cross-presentation by CD8+ DCs. CD8+ DCs were isolated from GMKO mice and B6 mice. DCs were cultured with CSFE-labeled (A) OT-1 cells or (B) OT-11 cells. Means and SEM of proliferating T cells from triplicate cultures are shown. Briefly, 2 ng/mL of GM-CSF was also included in all cultures. (A and B) More than three independent experiments were performed with similar results. (C) CD8+ DCs were isolated from GMKO mice and B6 mice. DCs were cultured with CSFE-labeled OT-1 cells in the presence or absence of OVA peptide. Means and SEM of proliferating T cells from triplicate cultures are shown. (p<0.05 Student's t-test, compared with WT with the same doses of OVA).

Thus, CD8+ DCs from GMTg mice are better at cross-presentation than those from WT mice, which are better than those from GMKO mice. Moreover, the GM-CSF-mediated enhancement of antigen presentation appears to be specific for cross-presentation.

GM-CSF is involved in the rapid entry of L. monocytogenes into CD8+ DCs

If DC survival and costimulation could not account for enhanced cross-presentation by CD8+ DCs under GM-CSF influence, can antigen uptake by CD8+ DCs be influenced by GM-CSF? Interestingly, another unique functional feature of CD8+ DCs is the rapid uptake of L. monocytogenes5. We first tested here whether GM-CSF also regulates this function of CD8+ DCs. We made mixed bone marrow chimeric mice from WT and βcKO (deficient in GM-CSFR signaling) and infected them with CFSE-labeled L. monocytogenes. In this manner the experimental conditions for both types of DC were internally controlled. Three hours after injection, uptake of L. monocytogenes by DCs was assayed by flow cytometry. We consistently observed that WT DCs had about 50% higher uptake of bacteria than βcKO CD8+ DCs (Fig. 4A). As expected, CD8 DCs of both origins had very low uptake of L. monocytogenes.

Figure 4.

GM-CSF enhances uptake of L. monocytogenes but not other antigens by CD8+ DCs. (A) In vivo uptake of L. monocytogenes. Mixed BM chimera from WT (CD45.1) and βcKO (CD45.2) donors were injected with 108 viable CFSE-labeled L. monocytogenes for 3 h. Dot plots show the uptake of L. monocytogenes by CD8+ and CD8 DCs. Scatter graph shows the mean and SD of CD8+ DCs harboring L. monocytogenes from five mice. (p-value was generated by Student's t-test); (B) In vitro uptake of soluble OVA by DCs from GMKO and GMtg mice. (C) In vitro uptake of microbeads by DCs from GMtg mice. Numbers in dot plots indicate the percentages of cells with bead uptake. Bar graph shows mean+SEM of percentage with bead uptake. Two independent experiments were performed.

Apart from bacteria, we also tested the influence of GM-CSF on the uptake of several forms of antigens. Comparisons of uptake by DCs were made between WT and GMtg as well as between WT and GMKO mice. In vitro, uptake of soluble antigen (FITC-conjugated OVA) by CD8+ DCs was not significantly different between WT and GMKO, or between WT and GMtg mice (Fig. 4B). The same was also observed for 500 nm FITC microbeads (Fig. 4C). We also tested the in vivo uptake of soluble OVA, FITC microbeads as well as irradiated CFSE-labeled thymocytes by DCs. Uptake of all three forms of antigen by CD8+ DCs was not affected by GM-CSF deficiency viz. WT and GMKO DCs behaved similarly (Supporting Information Figure 1).

Thus, GM-CSF seems to regulate certain type of uptake (such as live bacteria) but not all forms of antigens by CD8+ DCs. Nevertheless, as the uptake of thymocytes and soluble OVA was not different, uptake cannot explain the enhancement of cross-presentation by GM-CSF for soluble or cell-associated antigen.

GM-CSF regulates expression of CD103 on CD8+ DCs

Currently, little is known about how GM-CSF modulates the function of CD8+ DCs at the molecular level. There is at least one report of CD103 expression on CD8+ DCs being associated with several functional features related to cross-presentation: uptake of cellular antigen, sensitivity to suicide by exogenous cytochrome c (indicative of cytosolic translocation), and superior ability to stimulate proliferation of CD8+ T cells 7. The proportion of CD103+CD8+ DCs appears to vary between different mouse colonies. For example, some have described that all splenic CD8+ DCs express high levels of CD103 18, whereas others have found that only about half express high levels of CD103 7. As we found that GM-CSF can enhance cross-presentation by CD8+ DCs, we wanted to investigate whether GM-CSF could regulate CD103. Hence, we first compared the CD103 expression on CD8+ DCs between WT and GMKO mice. At steady state, the proportion of CD8+ DCs with appreciable CD103 expression (termed CD103+) was reduced in GMKO spleens compared with WT spleens (4 versus 20%; p<0.01) (Fig. 5A), whereas CD8 DCs from both WT and GMKO mice had low CD103 expression. The absence of GM-CSF did not significantly reduce the pool size of splenic CD8+ DCs (Fig. 5A).

Figure 5.

GM-CSF enhances CD103 expression on resident CD8+ DCs. (A) Spleens from WT mice and GMKO mice were enriched for low density cells and stained for CD11c, CD11b, CD8, and CD103 for flow cytometry. Dot plots on left panel show expression profiles of CD11b versus CD8 on gated CD11c+ DCs. Middle and right panels show CD103 expression by gated CD8+ DC and CD8 DC subsets. Numbers in plots indicate the percentages of indicated populations. (B) Spleens from GMtg mice, WT mice, and CD103 KO mice were enriched for low-density cells and stained for CD11c, CD11b, CD8, and CD103 or isotype control. Dot plots show CD103 expression by gated DC subsets. Numbers indicate percentage of indicated populations. Histograms show overlays of CD103 expression by gated DC subsets. (C) Unseparated spleen cells from above mice were stained for TCRβ, CD4, CD8, and CD103 or isotype control. TCRβ+CD8+ cells and TCRβ+CD4+ cells were plotted for their CD103 expression. Histograms show overlays of CD103 expression by T-cell subset. More than three experiments were performed with similar results.

Considering that only about 20% of splenic CD8+ DCs of our WT mice had moderate to high CD103 expression, it implies that the basal level of GM-CSF has limited impact on CD103 expression. To determine whether increased GM-CSF levels would upregulate CD103 expression on CD8+ DCs, we examined GMtg mice; we observed that spleens from GMtg mice had consistently lower percentages of CD8+ DCs (mean±SE: 10±3 in GMtg versus 27±2 in WT, p<0.002 by Student's t-test); not surprisingly the absolute numbers of CD8+ DCs were not significantly reduced due to the twofold increase in total CD11c+ cells in the spleen of GMtg mice (data not shown). CD8+ DCs had greatly increased CD103 expression compared with sex- and age-matched littermates (mean fluorescence±SE: GMtg 2648±142 versus WT 516±24, p<0.0001 by Student's t-test) (Fig. 5B). The increased expression of CD103 on CD8+ DCs also resulted the CD130+ DCs increased from 20% (20±3) in WT to 86% (86±1) (p<0.0001), whereas resident CD8 DCs had a modestly detectable increase (Fig. 5B). In contrast to CD8+ DCs, CD103 expression by T-cell subsets was not affected by GM-CSF transgenic hyper-expression (Fig. 5C). First, the population of CD103+ CD8+ T cells was similar (44±2 in WT versus 40±1 in GMtg, not significant). Second, the levels of CD103 on CD8+ T cells were also similar (mean fluorescence±SE: GMtg 328±11 versus WT 354±18, not significant). There was a very small proportion of CD4+ T cells in both WT and GMTg mice (around 4%, data not shown). All data were consistent from two independent lines of GMtg mice.

GM-CSF is responsible for infection-induced CD103 upregulation on CD8+ DCs

As well as transgenically, GM-CSF hyperexpression can be induced naturally by infection with the intracellular bacteria L. monocytogenes13. Therefore, we tested here whether such infection can modulate CD103 expression on CD8+ DCs and we observed that both the proportion of CD103+ CD8+ DCs and the intensity of CD103 expression by CD8+ DCs were dramatically increased 3 days after listerial infection (Fig. 6A). Any elevation of CD103 levels was unremarkable for CD8 DCs (Fig. 6A). A small proportion of CD4+ and CD8+ T cells did express CD103 but their pattern of expression was not affected by infection (data not shown). The infection-induced upregulation of CD103 on CD8+ DCs observed in WT mice was abrogated in GMKO mice (Fig. 6A). Therefore, although many cytokines can be elicited by listerial infection (Fig. 6B), GM-CSF seems to be critical for CD103 upregulation during infection.

Figure 6.

GM-CSF is required for infection-induced CD103 upregulation in resident CD8+ DCs. (A) WT and GMKO mice were infected with 2×104L. monocytogenes for 3 days. Spleens from infected and uninfected mice were enriched for low density cells and stained for CD11c, CD4, CD8, and CD103. Histograms show CD103 expression by gated DC subsets. (B) Sera from 1-day-infected WT mice were measured for GM-CSF by cytokine Bio-plex according to the manufacturer's instruction and p-value was generated with Student's t-test. More than three experiments were performed.

Intrinsic GM-CSF signaling is required for the expression of CD103 on CD8+ DCs

To assign a direct role for GM-CSF for the regulation of CD103 on CD8+ DCs, we made mixed BM chimera by reconstituting lethally irradiated B6 mice with equal numbers of WT (Ly5.1) and βcKO (Ly5.2) mice. Four weeks after reconstitution, spleen DCs were analyzed for CD103 expression. Both types of cells reconstituted approximately equally in CD11c+ cells (Fig. 7). However, in all mice, the percentage of CD8+ DCs of βcKO origin was higher compared with that of WT origin (Fig. 7). This pattern is in accordance with the reduced percentage of CD8+ DCs in GMtg mice. WT CD8+ DCs expressed much higher levels of CD103 than βcKO CD8+ DCs (Fig. 7). Notably, the levels of CD103 expression by WT CD8+ DCs from mixed bone marrow chimera were higher than that of WT CD8+ DCs from B6 mice. This may be related to different GM-CSF concentrations available to WT CD8+ DCs between the two situations. CD8 DCs of both origins expressed very low levels of CD103.

Figure 7.

Direct GM-CSF signaling is required for CD103 upregulation. Lethally irradiated B6 mice were reconstituted with a mixture of 2×106 WT (Ly5.1) and βcKO (L5.2) BM cells. Spleens from mixed BM chimera after 4-wk reconstitution were harvested. Low-density cells were stained for CD45.1, CD11c, CD11b, CD8, and CD103. Dot plots show the DC subpopulation of WT and βcKO origin within total CD11c+ spleen cells. Histograms show CD103 expression by DC subsets. Bar graph shows the mean fluorescence+SEM of CD103 on DC subsets of WT and βcKO origin from six individual mice (indicated p-value was generated by Student's t-test). Two experiments were performed with similar results.

The above findings were also confirmed with in vitro-derived DCs. Using a two-step coculture system, we cultured a mixture of equal numbers of BM cells from WT and βcKO mice with Flt3L 8. After 6 days, cell cultures were split into two: one with added GM-CSF and the other without GM-CSF. Cells were then cultured for further 3 days before cells were harvested for phenotypic analysis. As in vitro-derived DCs do not express CD8, high expression of CD24 and low expression of CD11b were to identify CD8+ DC equivalents. In agreement with our in vivo data, the addition of GM-CSF significantly enhanced CD103 expression on CD8+ (CD24+) DC equivalents of WT origin but not of βcKO origin in the same culture (Supporting Information Fig. 2). Even without exogenously added GM-CSF, WT CD24+ DCs contained a proportion of CD103+ DCs, whereas βcKO DCs totally lacked the CD103+ population.

Thus, we conclude that direct signaling through the GM-CSF receptor on CD8+ DCs is required for upregulation of CD103. It remains to be established whether CD103 is functionally involved in regulating cross-presentation by CD8+ DCs.

GM-CSF enhances the differentiation of resident CD8 DCs and monocyte-derived DCs

As this study mainly aimed to investigate the influence of GM-CSF on the function and CD103 expression of CD8+ DCs, we also observed that GM-CSF influences spleen CD8 DCs. There were two distinct populations within CD8 DCs: monocyte-derived DC (mDCs) and resident CD8 DCs. mDCs expressed high levels of CD11b, intermediate levels of CD11c, and were positive for MHC class II and Ly6C; resident CD8 DCs expressed high levels of CD11c and CD11b but not Ly6C. Therefore, Ly6C can segregate two CD8 DC populations. mDCs existed in low numbers in the spleens of WT mice but they increased by more than ten-fold in the spleens of GMtg mice (Fig. 8A and B). Resident CD8 DCs increased about two- to threefold in the spleens of GMtg mice (Fig. 8A and B). mDCs from GMtg mice could stimulate OVA induced OT-1 proliferation. However, compared with CD8+ DCs from GMtg mice, they were less potent at doing so (Fig. 8C).

Figure 8.

GM-CSF enhances differentiation of CD8 DCs. Low-density spleen cells were prepared from three individual WT and GMtg mice and stained for surface markers. (A) Dot plots show gating of three populations of spleen DCs: CD8+ DCs, CD11chigh Ly6C CD8 DCs, and CD11cint I-A+ Ly6+ monocyte-derived DCs. The numbers in the plots indicate the percentage of the populations. (B) Bar graphs show the numbers of mDCs and total resident CD11chigh DCs. Data shown are mean and SEM. *p<0.05, **p<0.01 (compared with controls by Student's t-test). More than three experiments were performed with similar results. (C) Cross-presentation by CD8+ DCs and mDCs from GMtg mice. Purified DCs from WT and GMtg mice were cultured with CFSE-labeled OT-1 T cells. T-cell proliferation was evaluated by CFSE dilution. Bars in graph represent the mean+SEM of proliferating OT-I cells from triplicate cultures (*p<0.05 compared with CD8+DCs by Student's t-test).

Discussion

Flt3L, rather than GM-CSF, has been regarded as critical in promoting the generation of resident DCs including CD8+ DCs. Our study revealed that GM-CSF also has a potent impact on resident CD8+ DCs. GM-CSF enhances cross-presentation in CD8+ DCs but has little effect on MHC class II presentation or direct class I presentation by peptide. GM-CSF also selectively regulates CD103 on resident CD8+ DC expression: infection-induced GM-CSF hyperexpression (as well as transgenic hyperexpression) drives this CD103 upregulation and the GM-CSF receptor is required for this upregulation. CD103 expression provides further functional division within CD8+ DCs and this supports our previous contention that not all CD8+ DCs cross-present efficiently 19.

It has been reported previously that GM-CSF in vitro does not grossly influence the expression of costimulatory molecules on CD8+ DCs 20. We found here that overexpression of GM-CSF in vivo also does not enhance the expression of CD80 and CD86 or MHC class II expression. Thus, the impact of GM-CSF on cross-presentation is unlikely due to classical costimulation.

GM-CSF is a prosurvival factor at least in vitro for resident DCs 20, 21. Therefore, enhanced cross-presentation by DCs under GM-CSF overexpression might be explained by DC survival. However, we provide several lines of evidence to suggest that the influence of GM-CSF on cross-presentation extends beyond cell survival. First, the survival rate of CD8+ DCs from GMtg and control mice was comparable in the presence or absence of exogenous GM-CSF (2 ng/mL) although exogenous GM-CSF did enhance the survival of DCs either from GMtg or from control mice. CD8+ DCs, even from GMtg mice, produce <0.1 ng/mL GM-CSF under stimulation with CpG. Second, DCs from GMtg and control mice stimulated a comparable response when peptide antigen is used (therefore bypassing cross-presentation). Third, there was no difference in MHC class II-restricted responses between WT and GMtg DCs (Fig. 1). Thus, enhancement of cross-presentation by CD8+ DCs from GMtg mice or reduction in cross-presentation by CD8+DCs from GMKO mice is unlikely due to DC survival.

The effect on cross-presentation by GM-CSF was also not due to differential uptake. CD8+ DCs from GMtg mice did not show enhanced uptake of soluble and particle antigen in vitro. Conversely, GMKO mice are not defective in uptake of soluble particle and cell-associated antigen in vitro and in vivo. The one exception is the rapid uptake of L. monocytogenes, a function that is prominent in CD8+ DCs but not in CD8 DCs 5. This enhanced uptake corresponds with CD103 expression and it is tempting to speculate that CD103 as a β integrin might be aiding internalin B-mediated engulfment 22. It is not clear whether uptake of other bacterial pathogens by CD8+ DCs will be influenced in a similar fashion.

Given that uptake of antigen by CD8+ DCs is not grossly affected by GM-CSF, how then does GM-CSF impact on cross-presentation by CD8+ DCs? We have previously shown that injection of cytochrome c selectively kills cross-presenting cells in an Apaf-1 (apoptosome)-dependent way 19, based on the rationale that only cross-presenting cells would translocate cytochrome c from the endosome into the cytosol (where the apoptosome is formed). It has been shown that cytochrome c preferentially kills CD103hi CD8+ DCs 7. We also observed similar findings for CD8+ DCs from WT and GMtg mice (unpublished data). These data suggest that more exogenous cytochrome c assesses to cytosol of CD103-expressing CD8+ DCs. However, precise events at molecular level that are responsible for GM-CSF-enhanced cross-presentation by CD8+ DCs are currently unknown. There are many pathways leading to cross-presentation of antigens 23: phagosome-to-cytosol 24, endosome-to-ER, ER–phagosome fusion 25, 26, endosome-to-ER 27, as well as vacuolar pathway 28, 29. It remains a task to identify which pathway(s) is affected by GM-CSF. A related question is whether CD103 is involved in regulating the function of CD8+ DCs, given that GM-CSF selectively enhanced CD103 expression on CD8+ DCs. Although mice deficient in CD103 can be useful in providing the answer to the above question, CD103 expression is variable in WT mice. That is because CD103 is a “induced” molecule for CD8+ DCs: low on CD8+ DCs of GM-CSF-deficient mice, intermediate on CD8+ DCs of WT mice and high on CD8+ DCs when GM-CSF is elevated (by transgenesis or inflammatory states); therefore, the microbiological milieu may also affect CD103 expression. Currently, we compared cross-presentation by CD8+ DCs from CD103-deficient mice (on B6 background) and WT B6. We observed lower CD8+ T-cell proliferation when CD103-deficient CD8+ DCs were used (Supporting Information Fig. 3). It implies that CD103 may have a role in DC function. Further studies using CD103-deficient GM-CSF transgenic mice might be informative. Even if CD103 plays a direct role in regulating cross-presentation by CD8+ DCs, the molecular mechanism of its action can only be speculated; for example, it can be involved either in the interaction with CD8+ T cells or in the intracellular events occuring in CD8+ DCs after antigen uptake.

Although the molecular mechanisms for CD103 in mediating the function of CD8+ DCs require further investigation, we have firmly established that CD103 expression on CD8+ DCs is subject to regulation by GM-CSF. We have shown herein that infection can lead to increased GM-CSF levels that would lead to CD103 upregulation. This may explain the wide variation in the reporting of CD103 expression by WT CD8+ DCs. For example, in one report, all spleen CD8+ DCs express high levels of CD103 18. In another report, about half of the spleen CD8+ DC express high levels of CD103 7. We find that in our colony of B6 mice, only 10–30% of CD8+ DCs express appreciable levels of CD103. Therefore, it is highly plausible that the variation in CD103 expression is due to differing microbiological environments. Curiously, although injection of GM-CSF can enhance CD103 expression on CD8+ DCs to a certain degree (Supporting Information Fig. 2A), we observed very limited upregulation of CD103 when isolated CD8+ DCs were incubated with GM-CSF. It could be that signals (either soluble or surface molecules) from other cells may be required for CD103 expression. It is also plausible that GM-CSF is required at a particular developmental stage for CD103 expression. Nevertheless, we found that supplementing GM-CSF to BM cells that were cultured with Flt3 ligand for 6 days can still upregulate CD103 on CD8+ DC equivalents (CD24+CD11b DCs) within 3 days with GM-CSF (Supporting Information Fig. 2B). These observations suggest that GM-CSF can target late stages of CD8+ DC development for CD103 upregulation. It is somewhat difficult to dissect out whether exposure to exogenous GM-CSF included in T-cell assays can also continuously mature DC function (i.e. enhancing cross-presentation). As GM-CSF does increase DC survival, at least in vitro, it is difficult to compare the function of WT DCs with or without exogenous GM-CSF. If overexpression of prosurvival molecules such as Bcl-2/Mcl-1/A1 can enhance DC survival to bypass the requirement of GM-CSF for cell survival, DCs from these mice can be used to test whether exogenous GM-CSF can enhance cross-presentation.

CD103 is an integrin that is also expressed by mucosal T cells 30. Our analysis of CD103 expression on T cells concludes that GM-CSF does not grossly affect the CD103 expression on T cells. Apart from T cells and resident CD8+ DCs, certain mucosal DCs 31, 32 and dermal DCs also express CD103 33. GM-CSF deficiency did not affect the numbers of resident CD8+ DCs, but led to a severe reduction in the numbers of dermal CD103+ DCs 33. We also observe a similar reduction in the population in GM-CSF-deficient mice (Supporting Information Fig. 4). However, it has been questioned whether the reduction in the population reflects the loss of CD103 expression or loss of the cells. Increase in CD11bCD103+ DCs in lymph nodes of GM-CSF trangenic mice provides strong evidence that GM-CSF does positively regulate the pool of CD11b CD103+ DCs. As GM-CSF positively reguates the levels of CD103 expression on CD8+ DCs, the level of CD103 expression by dermal CD103+ DCs was not elevated (Supporting Information Fig. 4). Collectively, GM-CSF may have different influences on different CD103+ DC populations. Currently, it is unknown what endows T cells and tissue DCs to express CD103.

Apart from affecting CD8+ DCs and CD11b CD103+ DCs, GM-CSF also had profound effect on other DC subsets. In the spleen, two CD8 DC populations (resident CD8 and mDCs) increased in GM-CSF transgenic mice. The functional significance of these changes in CD8 DC populations remains to be revealed.

Taken together, we conclude that GM-CSF promotes cross-presentation and uptake of L. monocytogenes by resident CD8+ DCs. Thus, GM-CSF has potent functional effects on DCs beyond its well-known role as a hematopoietic hormone or as an in vitro tool for expanding/differentiating BM- or monocyte-derived DCs. It is also responsible for basal and infection-induced levels of the integrin CD103 on resident CD8+ DCs. The significance of this CD103 upregulation, apart from being marker of the functional maturation of CD8+ DCs, is currently being investigated.

Materials and methods

Mice

C57BL/6, GMKO mice 34 on C57BL/6 background, GM-CSF receptor common β chain deficient (βcKO) mice 35 on C56BL/6 background, GMtg mice on SJL×C57BL/6 mixed background 36 and littermate control mice were generated and maintained in the animal facility of the Walter and Eliza Hall Institute of Medical Research. Mice with targeted disruption of CD103 (CD103KO) were housed at the University of Melbourne 37.

Listeria infection

Mice were injected i.v. with 2×104–5L. monocytogenes for 1–2 days. Sera and spleen lysates were prepared from uninfected and infected mice and assayed for cytokines by ELISA. To produce lysates, spleens were homogenized in 1 mL culture medium (Polytron, Lucerne, Switzerland). Bacterial load in spleens and livers in infected mice was also determined by plating out serial-diluted homogenates on horse blood agar plates 34.

Mixed BM chimera

Irradiated (2×550 cGy) 6-wk-old B6 mice received 2 million BM cells from Ly5.1 B6 mice and βcKO mice. Four weeks later, spleens were harvested for DC analysis. Anti-Ly5.1 Ab was used to distinguish WT DCs from KO DCs.

DC enrichment and flow cytometry

Spleens from infected and uninfected mice were digested for 20 min at room temperature with collagenase-DNase and then treated for 5 min with EDTA to disrupt T cell–DC complexes. Low density cells (<1.077 g/cm3 at mouse osmolarity) were separated by centrifugation in Nycodenz medium (Nycomed Pharma AS, Oslo, Norway) and stained for CD11c and CD11b, washed, and subjected for analysis and sorting on FACSaria (BD Biosciences, San Jose, CA, USA).

For flow cytometry, spleen cells and low-density spleen cells were incubated with rat anti-mouse FcgRII/FcgRIII monoclonal Ab (2.4G2) for 15 min at 4°C, to block nonspecific binding of Abs, before staining with various combinations of monoclonal Abs to CD11c (N418), CD11b (M1/70), Mac-3 (M3/84), I-A/I-E (2G9), Ly6C (AL-21), CD103 (M290), CD80 (16-10A1), CD86 (GL1) (BD Biosciences). Cell analysis was performed on an LSR flow cytometer or FACSaria (BD Biosciences).

In vitro DC survival

Spleen CD8+ DCs were sorted from GMtg mice and littermate controls. DCs (50 000/well) were cultured in a U-bottom 96-well plate for 36 h in the presence and absence of recombinant GM-CSF (2 ng/mL). Cells were then harvested and stained for surface markers. The numbers of viable DC subsets were determined by calibrating beads and PI exclusion.

Cytokine production after in vitro stimulation

Sorted splenic DCs were resuspended at 0.5×106/mL in fresh media in the presence or absence of a single TLR agonist. The following panel of TLR agonists was used: CpG ODN 1826 (2 μg/mL) (Coley Pharmaceutical, Ottawa, Canada), Poly I:C (50 μg/mL, Invivogen, San Diego, CA, USA), and LPS (1 μg/mL) (Sigma, St. Louis, MO, USA). DCs were plated at 1 mL/well in 48-well plates and cultured for 20 h before supernatants were collected and analyzed for IL-12p40, IL-12p70, IL-23, IL-10 using ELISA according to the manufacturer's instructions (BD Biosciences).

Uptake of apoptotic cells, latex beads, and soluble OVA in vivo and in vitro

For in vivo uptake, B6 thymocytes were given 25 Gy irradiation and labeled with CFSE. After washing, 4×107 cells or 5×109 washed fluorescent latex beads (0.5 μm Fluoresbrite® Yellow Green Carboxylate Microspheres, Polysciences, Warrington, PA, USA) or 2 mg FITC-conjugated OVA were injected intravenously. Spleens were harvested 90–120 min after injection. DC-enriched spleen cells were stained for CD11c, CD11b, CD103, and CD8. Uptake of apoptotic (CFSE+) cells or fluorescent beads by DCs was analyzed by flow cytometry. For in vitro uptake, DC-enriched spleen cells were harvested. One million of cells were incubated with 2×108 FITC beads, 100 μg/mL FITC-OVA at 37°C or kept on ice for 45 min. After washing with cold 2% FCS-EDTA (0.02 mM) PBS, cells were stained for CD11c, CD11b, CD103, and CD8. Uptake of different forms of material was analyzed by flow cytometry.

Uptake of L. monocytogenes in vivo

A suspension containing 1010/mL viable L. monocytogenes in 1 mL Hanks solution was incubated with 2 μL 5 mM CFSE at 37°C for 30 min. After washing, viable bacteria were enumerated on horse agar plates after serial dilution. Mice were injected with 108 viable CFSE-labeled L. monocytogenes for 3 h. Uptake of L. monocytogenes by splenocytes was analyzed after surface staining for CD11c, CD11b, CD103, and CD8.

In vitro proliferative responses and cytokine production of OVA-specific CD8+ T cells

Purified OT-I cells (50 000) were labeled with CSFE and cultured together with purified DCs (10 000–20 000) with or without graded numbers of OVA-coated Kb−/−spleen cells. Replicate cultures were in 200 μL medium in the U-bottom wells of 96-well culture trays. The culture medium was modified RPMI 1640 containing 10% FCS. Where indicated, 2 ng/mL murine GM-CSF were included in the cultures. Culture was normally for 3 days at 37°C in a humidified 10% CO2-in-air incubator. Proliferation of T cells was assessed either by reduction in dye intensity of harvested or by incorporation of tritiated thymidine. Harvested supernatants were assayed for cytokines by Bio-Plex (Bio-rad).

Acknowledgements

The authors thank the technical assistance from Nicole Ashman and Manny Hancock.

This work was supported by the National Health and Medical Research Council of Australia (NH&MRC) program and project grants (♯575543, ♯637324 and 1007703), Diabetes Australia, Juvenile Diabetes Research Foundation, NHMRC Independent Research Institutes Infrastructure Support Scheme grant ♯361646, and Victorian State Government Operational Infrastructure Support grant. Y. X. was supported by a Peter Doherty Fellowship from NH&MRC.

Conflict of Interest: The authors declare no financial or commercial conflict of interest.

Ancillary