Cross-Talk with Myeloid Accessory Cells Regulates Human Natural Killer Cell Interferon-γ Responses to Malaria

Data from a variety of experimental models suggest that natural killer (NK) cells require signals from accessory cells in order to respond optimally to pathogens, but the precise identity of the cells able to provide such signals depends upon the nature of the infectious organism. Here we show that the ability of human NK cells to produce interferon-γ in response to stimulation by Plasmodium falciparum–infected red blood cells (iRBCs) is strictly dependent upon multiple, contact-dependent and cytokine-mediated signals derived from both monocytes and myeloid dendritic cells (mDCs). Contrary to some previous reports, we find that both monocytes and mDCs express an activated phenotype following short-term incubation with iRBCs and secrete pro-inflammatory cytokines. The magnitude of the NK cell response (and of the KIR− CD56bright NK cell population in particular) is tightly correlated with resting levels of accessory cell maturation, indicating that heterogeneity of the NK response to malaria is a reflection of deep-rooted heterogeneity in the human innate immune system. Moreover, we show that NK cells are required to maintain the maturation status of resting mDCs and monocytes, providing additional evidence for reciprocal regulation of NK cells and accessory cells. However, NK cell–derived signals are not required for activation of accessory cells by either iRBCs or bacterial lipolysaccharide. Together, these data suggest that there may be differences in the sequence of events required for activation of NK cells by non-viral pathogens compared to the classical model of NK activation by virus-infected or major histocompatibility complex–deficient cells. These findings have far-reaching implications for the study of immunity to infection in human populations.


Introduction
After years of neglect, in which the adaptive immune response to malaria was investigated in great depth with a view to rational development of antimalarial vaccines, the innate immune response to malaria has recently come under scrutiny. In part, this is driven by the realisation that innate immune responses initiate and mould the subsequent adaptive response and that understanding the innate immune response may aid the design of novel adjuvants and novel vaccine delivery systems. Furthermore, data from experiments in rodents and from clinical studies indicate that innate immune responses play a crucial role in controlling malaria infections [1] and can significantly alter the outcome of infection, especially in non-immune individuals. The key element in the cell-mediated effector response to malaria is the rapid induction of interferon (IFN)-c [2][3][4][5][6]. We have shown that human natural killer (NK) cells can be an important early source of IFN-c after exposure of leukocytes to Plasmodium falciparum-infected red blood cells (iRBCs) [7][8][9], although this response is extremely heterogeneous [8,9].
Emerging data suggest that signals from accessory cells such as macrophages, monocytes, and dendritic cells (DCs) are absolutely required for optimal NK cell responses to many infectious agents [10][11][12]. Consistent with this, we have previously demonstrated that cloned NK cells are unable to respond to iRBCs in the absence of autologous peripheral blood mononuclear cells (PBMCs) [8]. However, an essential role for accessory cells in promoting robust pro-inflamma-tory responses to malaria is somewhat controversial given previous observations suggesting that accessory cell function is seriously impaired by exposure to malarial parasite ligands. For example, following exposure to iRBCs, both human DCs and monocytes secrete the immunosuppressive cytokine interleukin-10 (IL-10) [13,14], have an impaired ability to mature in response to lipopolysaccharides (LPS) and other ligands [13,[15][16][17], and are unable to present antigen to T cells [13]. Moreover, DCs from children with acute malaria infection have reduced human leukocyte antigen (HLA)-DR expression [18].
Here we have assessed the precise contribution of various subsets of accessory cells to NK cell activation by iRBCs. We demonstrate that cells of the myeloid lineage are not only activated by iRBCs, but also play a crucial role in controlling the magnitude of the early burst of NK cell-derived IFN-c via both cytokine-and contact-dependent signals. Furthermore, we provide evidence that reciprocal interactions between NK cells and myeloid accessory cells regulate levels of activation of both cell populations, with major implications for the ability of the innate immune system to recognise and respond to infection.

Materials and Methods Blood Donors
Adult blood donors were recruited at the London School of Hygiene and Tropical Medicine through an anonymous blood donation system. All donors were malaria naïve and healthy and gave fully informed consent for their blood to be used in this study. Ethical approval was given by the London School of Hygiene and Tropical Medicine Ethics Committee, application number 805. As we have previously shown that NK cells from only some donors make a strong IFN-c response to iRBCs [7][8][9], cells from known high-responding donors were used for the majority of experiments described here.

PBMC Preparation and Culture
Venous blood was collected into sodium heparin (10 IU/ml blood; CP Pharmaceuticals, http://www.wockhardt.co.uk) and PBMCs were isolated by Histopaque 1077 (Sigma-Aldrich) density gradient centrifugation as described previously [9]. Cells were resuspended at a concentration of 10 6 cells/ml and cultured in flat-bottomed 24-well plates for 24 h. Schizontinfected (iRBCs) or uninfected red blood cells (uRBCs) were added at a ratio of three red blood cells per mononuclear cell.

NK Cell Purification
CD56 þ CD3 À NK cells were enriched from PBMCs by magnetic cell separation (NK Cell Enrichment kit; StemCell Technologies, http://www.stemcell.com) according to the manufacturer's instructions and using LS separation columns (Miltenyi Biotec, http://www.miltenyibiotec.com). B cells, T cells, monocytes, and erythrocytes were retained in the column, and the effluent containing unlabelled NK cells was collected. NK cells were counted, tested for viability by trypan blue exclusion, and routinely checked by flow cytometry for a purity of greater than 95%.

Transwell Cultures
PBMCs (1 3 10 6 ) in a volume of 800 ll of complete medium were placed in a tissue culture well. Purified NK cells (1 3 10 5 ) in a volume of 200 ll were placed in a Transwell with a 0.4-lm microporous polycarbonate membrane (Corning, http://www. corning.com) and lowered into the culture well so that all cells were submerged in culture medium. iRBCs (3 3 10 6 , bottom well, or 0.5 3 10 6 , top well) were added as indicated. Plates were cultured for 24 h, and NK cells from the top well

Synopsis
The outcome of infection is determined both by the ability to limit the initial phase of pathogen colonisation and by the ability to mount an effective adaptive immune response. Both of these processes are influenced by innate immune responses, of which a crucial component can be the ability of natural killer (NK) cells to secrete pro-inflammatory cytokines. Studies in both humans and mice indicate that the magnitude of the early (innate) interferon (IFN)-c response is a crucial determinant of the outcome of malaria infection. In this study, Newman et al. show that activation of human NK cells by Plasmodium falciparum-infected red blood cells to produce IFN-c is strictly dependent upon, and regulated by, contactmediated and soluble (cytokine) signals from two accessory cell populations (myeloid dendritic cells and monocytes). Furthermore, the magnitude of the human NK cell IFN-c response to P. falciparum-infected red blood cells is highly correlated with levels of expression of co-stimulatory molecules on resting accessory cells. These findings suggest that it might be possible to predict the magnitude of the innate cytokine response, and possibly even susceptibility to malarial disease, from the phenotype of resting monocytes. In addition, these data contribute to the development of a new model of NK activation by non-viral pathogens in which activation of accessory cells precedes, rather than follows, NK activation. (E) MACS-purified, CTG-labelled NK cells were added to autologous PBMCs. IFN-c production in CTG þ (MACS-purified) and CTG À (''untouched'') NK cells was assessed by flow cytometry after 24 h of stimulation with iRBCs. A representative example of the software gating strategy to analyse IFN-c responses of CTG-labelled NK cells is shown. FACS plots were gated on CD3 À CD56 þ NK cells within the lymphocyte population (first panel), and NK cells were analysed for CTG staining (second panel). CTG þ CD3 À CD56 þ NK cells (third panel) or CTG À CD3 À CD56 þ NK cells (fourth panel) were selected and analysed for intracellular IFN-c production. Numbers indicate the proportion of IFN-c þ NK cells. doi:10.1371/journal.ppat.0020118.g001 and/or PBMCs from the bottom well were harvested, stained, and analysed by flow cytometry as previously described.

Cytokine Blocking
The following neutralising antibodies were used to block cytokines or their respective receptors: polyclonal rabbit anti-human IL-2 (Serotec), polyclonal rabbit anti-human IL-15 (Biosource, http://www.biosource.com), monoclonal mouse anti-human IFN-a/b receptor chain 2 (Chemicon, http://www.chemicon.com), and polyclonal chicken anti-TGF-b (R&D Systems). The antibodies or their respective isotype-matched controls were added to PBMC cultures prior to the addition of iRBCs and at the concentrations indicated.

Quantification of Cytokine mRNA by Real-Time Quantitative PCR
For each donor, 10 7 PBMCs were cultured for 24 h in the presence or absence of 3 3 10 7 iRBCs. Non-adherent cells were discarded and adherent cells harvested after 15 min incubation on ice followed by scraping. Total RNA was isolated (RNeasy Micro kit; Qiagen, http://www1.qiagen.com) and cDNA was produced using oligo(dT) 12-18mer primers and SuperScript III reverse transcriptase (Invitrogen, http://www. invitrogen.com) according to the manufacturer's instructions.

T G G G A C A C A T G G A T C T A A G A G
A -3 9 a n d 5 9-G G A T C C T T T G C A A G C A G A A C -3 9) , I F N -a ( 5 9-G C C A T C T C T G T C C T C C A T G A -3 9 a n d 5 9-C A T -GATTTCTGCTCTGACAACCT-39), and IFN-c (5 9-T T C G G T A A C T G A C T T G A A T G T C C -3 9 a n d 5 9-TCTTCGACCTTGAAACAGCA-39).

TGF-b ELISA
The concentration of total (bioactive and latent) and spontaneously bioactive TGF-b in culture supernatants was assessed using the Human TGF-b 1 DuoSet system (R&D Systems) according to the manufacturer's instructions. Luminex Bead-Based Cytokine Quantification Multiplex quantification of cytokine concentrations in tissue culture supernatants was performed using the Luminex 100 system (Luminex, http://www.luminexcorp.com) with Bio-Plex upgrade (Bio-Rad Laboratories, http://www.bio-rad.com). Levels of IL-2, IL-10, and IL-12p70 were determined using the Beadlyte Human Multi-Cytokine Detection System 3 (Upstate, http://www.upstate.com) according to the manufacturer's instructions. Data analysis was performed using the Bio-Plex Manager software (Bio-Rad).

Statistical Analysis
All statistical analyses were performed using Prism 4 software (GraphPad Software, http://www.graphpad.com). All r 2 values given refer to a second-order polynomial fit of data shown. If data are plotted using log scale, analysis was carried out using log values. All t tests carried out were two-tailed paired t tests with 95% confidence intervals. Comparisons were made between actual data even where log axes have been used to plot data.

Optimal NK Cell Response to iRBCs Is Accessory Cell-Dependent
We have previously shown that cloned NK cell lines are unable to respond to iRBCs in the absence of accessory cells that are present among PBMCs [8]. To confirm and extend this observation, NK cells were purified from PBMCs of seven different donors by magnetic cell sorting (MACS) and stimulated for 24 h with IL-12 and IL-18, uRBCs, or iRBCs or left unstimulated and analysed for surface expression of CD3, CD56, CD69, and intracellular IFN-c; unsorted PBMCs from the same donors were stimulated and analysed in an equivalent fashion ( Figure 1). In PBMC cultures, as expected, IFN-c þ (CD3 À CD56 þ ) NK cells were detected in both cytokine-stimulated and iRBC-stimulated cultures but not in unstimulated or uRBC-stimulated cultures ( Figure 1A). We deliberately selected donors who were known, from previous studies, to be relatively high responders in this assay; nevertheless, the magnitude of the IFN-c response to iRBCs varied between donors, with between 5% and 50% of NK cells making IFN-c in response to iRBCs after 24 h ( Figure 1C). In complete contrast, although purified NK cells made IFN-c in response to cytokine stimulation ( Figure 1B), no IFN-c was produced by purified NK cells in response to iRBCs ( Figure  1B and 1C).
We have shown that up-regulation of surface CD69 expression is a marker of partial (early) activation of NK cells [9], and thus, as expected, up-regulation of CD69 was consistently seen on NK cells within the iRBC-stimulated PBMC population ( Figure 1D). Surface CD69 expression tended to be higher on unstimulated purified NK cells than on unstimulated NK cells within PBMCs, presumably because of partial activation during cell separation. However, CD69 surface expression was not significantly up-regulated on purified NK cells after stimulation with iRBCs, and NK CD69 expression was significantly higher among iRBC-stimulated PBMC NK cells than among purified NK cells ( Figure 1D).
Importantly, the inability of purified NK cells to respond to iRBCs is not an artefact of the NK cell purification process as MACS-purified NK cells labelled with CellTracker Green (CTG) and added back to autologous PBMC cultures respond as well to iRBC stimulation as do ''untouched'' NK cells within PBMCs ( Figure 1E).
Thus, these data indicate that accessory cells within PBMCs are required for both IFN-c and CD69 expression among iRBC-stimulated NK cells.

Depletion of CD1c þ and CD14 þ Accessory Cells Inhibits the NK Cell Response
To identify the accessory cell population(s) required for NK cell activation by iRBCs, PBMCs of known high-responding donors were depleted (by MACS) of cells expressing specific cell surface markers prior to incubation with iRBCs ( Figure 2). PBMCs of all the donors tested in these experiments responded to iRBC stimulation, with 5%-40% of all NK cells producing IFN-c after 24 h. Depletion of accessory cell populations did not affect the ability of NK cells to produce IFN-c following stimulation with positive control stimuli (IL-12 and IL-18; Figure S1). Depletion of HLA-DR þ cells completely abolished the NK cell IFN-c response to iRBCs in all donors, but depletion of BDCA-4 þ plasmacytoid DCs (pDCs) had no consistent effect on the NK IFN-c response. Likewise, depletion of CD19 þ B cells did not reduce the NK cell IFN-c response. However, depleting PBMCs of either CD1c þ myeloid DCs (mDCs) or CD14 þ monocytes significantly reduced NK cell IFN-c responses (monocytes, In the one donor where depletion using either CD14 or CD1c did not substantially reduce the IFN-c response (relative response, monocyte-depleted ¼ 0.84, mDC-depleted ¼ 0.90), a double depletion using both antibodies together was performed. In this case the IFN-c response was substantially reduced (relative response ¼ 0.32).
These data indicate that the NK cell response to iRBCs is not dependent on a single population of accessory cells; rather, both mDCs and monocytes can provide activating signals for NK cells, and, in most donors, optimal NK activation requires both of these cell types to be present.

Monocytes and mDCs Are Activated by iRBCs
Given the numerous published reports indicating that monocyte and DC function is inhibited by exposure to iRBCs [13,14], it was important to determine whether the accessory cells required for NK cell activation were themselves activated by exposure to iRBCs. PBMCs from ten different donors were therefore stimulated with uRBCs, iRBCs, or E. coli LPS (or left unstimulated, growth medium alone [gm]) for 24 h, and then mDC and monocyte populations were analysed for expression of the activation markers CD40, HLA-DR, and CD80 ( Figure 3; isotype control data to exclude non-specific binding of labelled antibodies by activated accessory cells are shown in Figure S2).
Significant up-regulation of CD40, CD80, and HLA-DR was observed on monocytes from all donors following 24 h incubation with iRBCs (CD40, mean fold increase of median fluorescence intensity [MFI] ¼ 6.3, paired t test, t ¼ 3.59, p ¼ 0.006; HLA-DR, mean fold increase of MFI ¼ 1.7, paired t test, t ¼ 3.42, p ¼ 0.008; CD80, mean fold increase of MFI ¼ 3.0, paired t test, t ¼ 7.56, p , 0.0001). Activation of mDCs was also observed, with significant up-regulation of CD40 and HLA-DR expression (CD40, mean fold increase of MFI ¼ 2.3, paired t test, t ¼ 3.60, p ¼ 0.006; HLA-DR, mean fold increase of MFI ¼ 2.4, paired t test, t ¼ 3.16, p ¼ 0.012). There was, however, no significant change in expression of CD80 on mDCs after 24 h co-incubation with iRBCs (mean fold increase of MFI These data indicate that both mDCs and monocytes are activated after 24 h co-incubation with iRBCs but that the nature of activation differs between the two accessory cell populations. Up-regulation of CD40 is more pronounced in monocytes and up-regulation of HLA-DR is more pronounced in mDCs, whilst up-regulation of CD80 occurs exclusively in monocytes.

Accessory Cells Express Both Pro-and Anti-Inflammatory Cytokines in Response to Activation by iRBCs
To characterise further the activation of accessory cells by iRBCs, we analysed the cytokine response profile after culturing PBMCs for 24 h with iRBCs by quantitative real-time PCR analysis of adherent cells and by Luminex cytokine bead array analysis of culture supernatants (Figure 4).
Unstimulated adherent cells expressed detectable levels of mRNA for IL-15, IL-18, and, at somewhat lower levels, IL-10, IL-12, and IL-23, indicating either constitutive expression of these cytokines or their induction during the cell separation process. Levels of IFN-a, IL-2, and IFN-c mRNA were below the limits of detection of the assay ( Figure 4A). However, after iRBC stimulation there was very marked up-regulation of mRNA for IFN-a and IL-2 and, to a somewhat lesser extent, for IL-10 and IFN-c but not for IL-12, IL-15, IL-18, or IL-23 ( Figure 4B). Taken together, these data indicate constitutive or induced expression of IL-12, IL-15, IL-18, IL-23, IFN-a, IL-2, IL-10, and IFN-c by adherent mononuclear cells stimulated with iRBCs. Production of IL-2, IL-10, and IL-12 in response to iRBC stimulation was confirmed at the protein level ( Figure 4C); the discrepancy between mRNA and protein levels for IL-12 is most likely explained by the timing of the assays (all of which were conducted after 24 h).
Type 1 IFN and IL-2, but Not IL-15, Are Required for NK Cell Activation by iRBCs whilst TGF-b Suppresses This Response NK cell activation by pathogens is typically dependent on accessory cell-secreted cytokines [19][20][21][22], and we have shown that IL-12 and, to a somewhat lesser extent, IL-18 are required for optimal NK responses to iRBCs [7,8]. To determine whether other accessory cell-derived cytokines are also required for NK activation, the response to iRBCs was analysed in the presence or absence of neutralising antibodies to individual cytokines or their receptors. Of particular interest were those cytokines that we had found (as described above) to be specifically and strongly upregulated by iRBCs, namely IFN-a and IL-2. Also, since TGF-b has recently been reported to be a powerful inhibitor of NK cell activation [23], and since we have shown that induction of TGF-b in vivo in both human and rodent malaria infections is associated with lack of early IFN-c responses [24,25], it was of considerable interest to determine whether TGF-b might regulate the NK cell IFN-c response to iRBCs in our in vitro assay system where platelets and mononuclear cells provide an abundant source of latent TGF-b. Representative fluorescence activated cell sorting (FACS) plots showing the effect of blocking cytokines on IFN-c response to iRBCs are shown in Figure S3.
In the presence of a blocking antibody to the IFN-a/b receptor, there is a significant and dose-dependent inhibition of the NK cell IFN-c response to iRBCs (highest antibody dose, paired t test, t ¼ 6.42, p ¼ 0.023) ( Figure 5A) and a similar, but less marked, reduction in CD69 expression (highest antibody dose, paired t test, t ¼ 5.26, p ¼ 0.034) ( Figure S4). Similarly, neutralisation of IL-2 markedly inhibited NK cell IFN-c responses in a dose-dependent manner (highest antibody dose, paired t test, t ¼ 8.79, p ¼ 0.001) ( Figure 5B), and CD69 surface expression was also reduced, but to a lesser degree (highest antibody dose, paired t test, t ¼ 4.43, p ¼ 0.021) ( Figure S4). On the other hand, neutralisation of IL-15 had only a minimal effect on NK cell IFN-c responses (highest antibody dose, paired t test, t ¼ 1.83, p ¼ 0.209) ( Figure 5C) or CD69 expression (highest antibody dose, paired t test, t ¼ 3.36, p ¼ 0.078) ( Figure S4).
When PBMCs were incubated with iRBCs for 24 h in the presence or absence of increasing doses of neutralising antibody to TGF-b, there was a highly consistent, dosedependent increase in the proportion of NK cells synthesising IFN-c, with up to a 3-fold increase in IFN-c þ NK cells in the presence of 100 lg/ml antibody (paired t test, t ¼ 7.98, p ¼ 0.004) ( Figure 5D) and a slight, but statistically significant, increase in CD69 expression (highest antibody dose, paired t test, t ¼ 4.00, p ¼ 0.028) ( Figure S4). Conversely, addition of recombinant human TGF-b led to a very marked, dosedependent suppression of the NK cell IFN-c response to iRBCs (highest dose, 1 ng/ml; paired t test, t ¼ 12.8, p ¼ 0.001; unpublished data).

Optimal NK Cell Response to iRBCs Requires Contact between NK Cells and Accessory Cells
For some microbial stimuli it has been reported that optimal NK cell activation requires contact with accessory cells [12,22,26,27]. We thus sought to determine whether such contact-dependent signals were also required for NK activation by iRBCs. Purified NK cells (in the upper chamber of a Transwell culture system) were separated from PBMCs (in the lower chamber) by a microporous membrane (0.4-lm diameter), which allowed diffusion of small molecules including cytokines, but not migration of leukocytes, iRBCs, or extracellular malarial merozoites. When a soluble positive control stimulus (phytohaemagglutinin) was added to the lower well, NK cells in the inner (top) well responded normally, indicating that diffusion of soluble signals was not impeded ( Figure S5). When iRBCs were added to the lower chamber, NK cells within the PBMC population synthesised IFN-c ( Figure 6A, PBMC control) and upregulated CD69 ( Figure 6B, PBMC control). However, addition of iRBCs to the upper well, the lower well, or both the upper and lower wells together was insufficient to induce optimal IFN-c production among the NK cells in the upper well, suggesting that contactmediated interactions between NK cells and iRBCs and/or between accessory cells and iRBCs are insufficient for full NK activation and that direct cell-cell contact between the NK cells and the accessory cells is required.
Contact between NK cells and accessory cells was also required for optimal up-regulation of CD69, although partial up-regulation of CD69 was observed in some cases when purified NK cells were in contact with iRBCs and/or when accessory cells were in contact with iRBCs. The up-regulation of CD69 by purified NK cells in the Transwell system ( Figure  6), but not in regular cultures of purified NK cells (Figure 1), is presumably due to the presence of soluble, cytokinemediated signals derived from PBMCs in the lower well.

NK Cell-Derived Signals Maintain the Maturation Status of Accessory Cells but Are Not Required for Accessory Cell Responses to iRBCs or LPS
In recent years it has become apparent that interactions between NK cells and DCs can be bidirectional. NK cells, especially those activated by prior exposure to IL-2, enhance DC maturation, inflammatory cytokine secretion, and antigen presentation in allogeneic situations and when exposed to virus-infected cells [28][29][30]. On the other hand, the degree of reciprocity between NK cells and monocytes, and the potential for NK cells to augment accessory cell responses to non-viral pathogens, has not to our knowledge been investigated. We therefore examined whether the interactions that we had observed to be crucial for the activation of NK cells by iRBCs led to reciprocal activation of either monocytes or mDCs.
Whole PBMCs or PBMCs depleted of NK cells were cultured for 24 h without stimulation (gm) or with LPS or iRBCs. After 24 h, cells were stained for HLA-DR, CD14, either CD40 or CD80, and either CD11c or CD11b ( Figure  7A). Both mDCs and monocytes expressed significantly lower levels of activation markers when cultured (without additional stimulation) in the absence of NK cells ( Figure 7B-7D), confirming a previous report of a role for NK cells in promoting DC maturation [29] and extending this observation to monocytes. On the other hand, when cells were cultured with either LPS or iRBCs, there was no significant effect of the presence or absence of NK cells on expression levels of any of the activation markers in either mDCs or monocytes ( Figure 7B-7D). Thus, while NK cells play a role in maturation of both monocytes and mDCs in resting conditions, the response of either of these cell types to LPS or iRBC stimulation is largely NK cell-independent. These data suggest that, at this early time point, the major inducers of myeloid cell activation are the pathogens themselves, rather than NK-derived IFN-c.

Degree of Monocyte Activation Correlates with Magnitude of NK Cell IFN-c Response
We have previously shown that human NK cell responses to iRBCs are heterogeneous [8,9]. Here, we noticed a considerable degree of heterogeneity of accessory cell surface marker expression ( Figure 3D) and of cytokine production ( Figure 4) among donors, in terms of both resting levels and responses to iRBCs. Given the absolute dependence of the NK cell response to iRBCs on accessory cells, we sought to determine whether heterogeneity of NK cell responses correlated with heterogeneity of accessory cell responses. Since NK cell IFN-c responses can vary between CD56 bright and CD56 dim NK cells-presumably because of differential expression of NK cell receptors including the killer cell immunoglobulin-like receptors (KIRs), which are expressed by CD56 dim but not CD56 bright NK cells-we analysed the correlation between accessory cell and NK cell responses for both of these NK cell populations.
Highly significant correlations were observed between iRBC-induced concentrations of IL-12 ( Figure 8A) and IL-2 ( Figure 8B) and the magnitude of the NK cell IFN-c response. A particularly close correlation was observed between IL-12 concentrations and the proportion of IFN-c þ CD56 bright (KIR À ) cells, whilst, for several donors (see arrows in Figure  8), a significant discrepancy was seen in the response of the CD56 dim population. Somewhat surprisingly, there was no correlation between iRBC-induced up-regulation of HLA-DR, CD40, or CD80 on either monocytes or mDCs and NK IFN-c production (unpublished data). However, resting levels of expression of both CD40 and HLA-DR were tightly correlated with the NK cell IFN-c response ( Figure 8C and 8D). For resting levels of CD40 expression there was again a much closer correlation with the IFN-c response of CD56 bright NK cells than with that of CD56 dim NK cells. One plausible interpretation of these data is that the NK cell IFN-c response is very sensitive to accessory cell CD40 expression and IL-12 production, but that in CD56 dim cells this response is moderated by KIR/major histocompatibility complex (MHC) class I-mediated signals.
Our observation that TGF-b is able to markedly suppress NK cell IFN-c responses ( Figure 5) led us to consider whether the heterogeneity of NK cell responses was due to differential production and/or activation of TGF-b. There was no consistent increase or decrease in the concentration of bioactive TGF-b in the supernatants of iRBC-stimulated PBMCs; a modest increase in bioactive TGF-b concentration was observed for two donors, but these were, in fact, the donors with the highest NK cell IFN-c responses to iRBCs, suggesting that suppression by TGF-b does not explain the heterogeneity of NK cell IFN-c response. Rather, regulation of bioactive TGF-b appears to be positively correlated with NK cell IFN-c expression ( Figure 8E), and furthermore, the correlation with CD56 dim NK cells was higher than that with CD56 bright NK cells. This suggests that TGF-b is not causal in NK cell activation, but rather that the total amount of IFN-c produced determines the amount of TGF-b. Thus, activation may lead to induction of negative feedback mechanisms in order to prevent excessive pro-inflammatory responses.

Discussion
By carefully dissecting the requirements for activation of human NK cells by iRBCs, we have revealed that the early NK cell-derived IFN-c response is absolutely dependent upon contact with, and cytokine production by, accessory cells. We have identified the requisite accessory cells as monocytes and mDCs and have shown that both cell types become highly activated during the first 24 h of exposure to iRBCs. Furthermore, we have demonstrated that whilst NK cells provide essential signals that maintain the maturation status of resting monocytes and mDCs, NK cells are not required for these same cells to respond fully to iRBCs. Finally, we have shown that the capacity of NK cells to secrete IFN-c is highly correlated with both the resting phenotype of the accessory cells and the concentration of iRBC-induced monokines. Thus, functional heterogeneity of monocytes and mDCs among human populations appears to play a crucial role in determining the magnitude of the innate immune response to iRBCs.
A wealth of evidence suggests that, after interaction with malarial parasites, both monocytes and mDCs secrete immunosuppressive cytokines and have an impaired ability to respond to a variety of stimulatory ligands [13][14][15][16][17]31]. These data have led some to conclude that monocyte and DC function is universally impaired by contact with malarial parasites and that this underlies an apparent suppression of immune responses seen during malaria infections. However, there have been several reports of activation of monocytes and DCs by Plasmodium-derived ligands [32][33][34]. Studies of rodent malaria infections have begun to resolve this debate, showing that in the first 24 h of infection, splenic DCs are activated, secrete pro-inflammatory cytokines including IL-12 and tumour necrosis factor-a, and present antigen to T cells, but that they subsequently become refractory to re-stimulation and secrete anti-inflammatory cytokines such as IL-10 [35]. This sequence of events allows infection to be controlled whilst simultaneously avoiding immune-mediated pathology [25]. Here we show that human monocytes and mDCs are highly activated after 24 h of co-incubation with iRBCs. The myeloid cell response occurred in the presence or absence of NK cells, indicating that direct interaction with parasites-as opposed to NK-derived IFN-c-drives the initial activation of these cells. Furthermore, this activation initiates and regulates the crucial early burst of IFN-c from NK cells. However, our observation that NK activation goes hand in hand with IL-10 and TGF-b production, and that TGF-b can suppress NK IFN-c production, is consistent with the notion that proinflammatory responses to malaria are carefully regulated by homeostatic mechanisms to prevent immune-mediated pathology.
Our observations regarding the accessory cell requirements for NK cell activation by malaria parasites are in partial agreement with a previous report invoking a role for monocytes in this response [36]; however, we show that mDCs play an equally important role. The fact that NK activation is suboptimal but not absent in the presence of only one or the other accessory cell population suggests that although there is some redundancy in their roles, they also provide complementary signals. Our data are consistent with reports of leukocyte activation by malarial glycosylphosphatidylinositol signalling through toll-like receptor (TLR) 2 and TLR4 [37], which are both expressed on monocytes and mDCs but not pDCs [38,39], although a role for other receptors cannot be ruled out. The reported activation of pDCs by a malarial hemozoin-TLR9 pathway [40,41] does not appear to be required for NK IFN-c production, consistent with previous reports that NK cytotoxic activity but not IFN-c secretion is dependent on pDC-derived signals [42,43]. In fact, depletion of pDCs increased the IFN-c response of NK cells in two out of three donors, suggesting that pDCs may have a net anti-inflammatory effect in this system, perhaps mediated by either IL-10 or TGF-b. Further experiments are required to test this possibility.
The data presented here reveal that accessory cell-dependent activation of NK cells by iRBCs is multifactorial and, as in other systems, involves multiple cytokines (IL-2, IL-12, IL-18, and IFN-a) [44][45][46] as well as direct contact between NK cells and accessory cells [12,22,26,27]. Whether accessory cells are the source of IL-2, as recently reported for human pDCs and mDCs [47], is not clear, and we cannot rule out that IL-2 is produced by a small contaminating population of rapidly responding T cells. Contact-dependent signals might include CD40-CD40L signalling [12], delivery of IL-12 from mDCs via an immune synapse [48], or binding of MHC-peptide complexes by activating KIR [49]. We believe that indirect, accessory cell-mediated signals are likely to be the major mechanism of NK cell activation by iRBCs, but we cannot rule out that the NK cell receives additional signals via direct interaction with iRBCs. Stable contacts between iRBCs and NK cells have been reported by ourselves and by others [9,36], and we have observed reorganisation of the actin cytoskeleton at the point of iRBC-NK contact [9]; the functional consequence of these interactions are, however, not yet fully understood.
Our observation that NK cells are not required for optimal activation of monocytes and mDCs by pathogen-derived ligands (iRBCs and LPS) runs somewhat counter to previous reports that NK cells augment DC maturation, cytokine secretion, and antigen presentation [28][29][30]. One plausible explanation for this discrepancy is that the sequence of events leading to activation of NK cells and accessory cells depends on the nature of the stimulus. In the case of virally infected or neoplastic cells, altered expression of MHC or MHC-like ligands on the target cell will lead to direct activation of NK cells either through engagement of natural cytotoxicity receptors or changes in the balance of activating and inhibitory signals delivered by KIR receptors (reviewed in [50]). In this case, amplification of the innate response and induction of adaptive immunity would be facilitated if NK cells were able to provide activating signals to DCs and other antigen-presenting cells, which are much less sensitive to changes in MHC expression. On the other hand, in the case of bacterial and protozoal pathogens, which do not grossly alter MHC class I expression or induce MHC-related stress ligands on host cells, direct activation of NK cells is unlikely. Rather, DCs and monocytes are likely to be directly activated by binding of pathogen-associated molecular patterns to TLRs and other pattern recognition receptors and may not require additional signals from NK cells.
Given the importance of the innate response to malaria in determining the outcome of disease (reviewed in [1]), dissecting the basis for the heterogeneity of the human NK cell response to iRBCs could be crucial in understanding disease susceptibility. The close correlations shown here between accessory cell maturation, accessory cell responses to iRBCs, and induction of IFN-c production by NK cells suggest that heterogeneity of NK cell responses to iRBCs [8,9] is a reflection of deeply rooted heterogeneity of innate responses within the human population. Ongoing studies of the immunogenetics of susceptibility to human malaria should begin to identify the underlying basis for this heterogeneity, but linking genotype to immunological phenotype may not be straightforward. Multiple genes, including various IFN regulatory factor proteins and components of the colony stimulating factor 1 signalling pathway, influence the maturation status of accessory cells [51][52][53]. Moreover, the much stronger correlation of accessory cell status with activation of CD56 bright (KIR À ) NK cells than with activation of CD56 dim (KIR þ ) NK cells suggests that NK cell activation by pathogeninduced accessory cell signals may be modulated by NK cell receptor expression. Since KIR genes, along with MHC genes, are some of the most highly polymorphic loci in the human genome [54], dissecting the genetic basis of diversity in the human innate response is a daunting task. Nevertheless, as CD56 dim cells are the major population of NK cells in peripheral blood and spleen [55], our data suggest that NK cell receptor expression is likely to have a significant influence on overall NK cell responsiveness and thus potentially affect the severity of malarial disease and other systemic infections. Despite the difficulties associated with carrying them out, population-based studies of disease risk and diversity in innate response genes could have farreaching implications, and we have recently begun such studies in malaria-endemic populations. Since the antigenpresenting and co-stimulatory functions of accessory cells are pivotal for induction of adaptive immunity, and the magnitude of the early IFN-c response affects adaptive responses [56], heterogeneity of expression of CD40, HLA-DR, and NK cell IFN-c is likely to have a profound impact on the generation of acquired immune responses.

Accession Numbers
Online database accession and ID numbers for genes and proteins mentioned throughout this article can be found in Table 1.