Multiple Mechanisms Downstream of TLR-4 Stimulation Allow Expression of NKG2D Ligands To Facilitate Macrophage/NK Cell Crosstalk Free

One strategy for immune surveillance is via recognition of proteins that are usually not expressed on cells but are upregulated in response to cellular stress (1, 2). The activating receptor NKG2D is expressed on NK cells and subsets of T and NKT cells, and recognizes a group of such inducible “stress” ligands (3, 4). In humans, NKG2D ligands are encoded by two MIC (MHC class I chain gene family related molecules) genes, MICA and MICB, and six ULBP/RAET (UL16-binding protein; also known as retinoic acid early transcript) genes (5). Cell surface expression of ligands for NKG2D is almost completely absent on most cell types, but is upregulated in response to stress stimuli, such as heat shock, UV irradiation, viral infection, or genotoxic/oncogenic stress (6). As a consequence, ligation of NKG2D results in NK cell activation that leads to the release of cytotoxic granules and subsequent killing of target cells, which is an important defense against many infections and tumor transformation (7).

LPS, a cell wall component of Gram-negative bacteria, is a ligand for TLR 4 (8), one of the pattern recognition receptors that recognize pathogen-associated molecular patterns released during infections. Ligation of TLRs at the site of infection triggers the release of inflammatory mediators, including cytokines and chemokines, initiating a response that should ultimately result in pathogen clearance and healing of damaged tissue (9). Nonetheless, it is important to limit immune responses in time and space to avoid excessive inflammation, which itself can cause pathological conditions, as, for example, during sepsis (10). Thus, cell surface expression of NKG2D ligands to trigger killing of macrophages may be an important regulatory component of the inflammatory response (11). Indeed, macrophages exposed to high amounts of LPS upregulate ligands for NKG2D and can be consequently killed by NK cell-mediated lysis (12, 13). Aberrant expression of NKG2D ligands is also associated with autoimmune diseases (14), an observation that underscores the importance of tight regulation of NKG2D ligand expression.

To date, several different mechanisms have been identified for the regulation of NKG2D ligand expression in response to cellular stress stimuli. Genotoxic stress agents, for example, induce expression of the NKG2D ligand MICA by a mechanism involving the DNA damage pathway kinases ataxia telangiectasia mutated (ATM)/ataxia telangiectasia and Rad3 related (ATR) and their downstream targets CHK-1 and -2 (15). Posttranslational regulation of NKG2D ligands can occur by ubiquitination and by enzymatic shedding (16–18). In addition, it has recently been shown that cellular microRNAs (miRNAs) target the 3′UTR of MICA and MICB (19, 20). Importantly, stress stimuli increased expression of MICA and MICB without lowering expression of these miRNAs. It has been suggested instead that miRNAs set a fixed threshold in the level of mRNA expression that is required to allow protein translation (20). This is consistent with MICA and MICB mRNA transcripts often being constitutively expressed in cells under normal conditions without detectable protein expression (13, 21, 22).

In this paper, we set out to probe the mechanisms that control NKG2D ligand expression in human macrophages in response to TLR stimulation. Surprisingly, we found that expression of MICA in human macrophages was controlled at multiple levels. Of particular interest, we found that TLR-4 ligation led to decreased expression of miRNAs that target MICA, expanding the role of these miRNAs to more than merely establishing a threshold in the level of mRNA needed for protein expression.

Human macrophages were isolated and cultured as described (13) and stimulated with LPS (Salmonella Minnesota; Sigma-Aldrich, St. Louis, MO), CL097, or polyinosinic-polycytidylic acid [poly(I:C)] (InvivoGen, San Diego, CA). Human NK cells were isolated and cultured as described (23). THP-1 were cultured in standard RPMI-based media. Abs were used as follows: anti-MICA (clone 159227), anti-MICB (236511), control IgG_2b (all R&D Systems, Minneapolis, MN), anti-MICA [BAM195, a kind gift from D. Pende (24)], control IgG₁, anti-CD54 (HA58), PE-conjugated goat anti-mouse (all BD Pharmingen, San Diego, CA), anti-MHC class I (W6/32, American Type Culture Collection), Cy5-conjugated goat anti-mouse (Jackson ImmunoResearch Laboratories, West Grove, PA). THP-1 were transfected by electroporation (Microporator; Labtech International, Ringmer, U.K.) and macrophages using INTERFERin (Polyplus, Illkirch, France). Sequences of antagomirs (Integrated DNA Technologies, Coralville, IA) were as described previously (20). Inhibitors used were actinomycin D, caffeine (both Sigma-Aldrich), and ATM inhibitor (KU55933; Calbiochem, San Diego, CA). Cytokine secretion was analyzed by ELISA in culture supernatants according to the manufacturer’s instructions (Autogen Bioclear, Calne, Wiltshire, U.K.; and BD Biosciences, San Jose, CA). For Western blotting, cells were lysed in 0.5% Triton X-100 lysis buffer, and proteins were separated by SDS-PAGE, transferred onto a polyvinylidene difluoride membrane, and analyzed using mAb anti-MICA (M673, a kind gift from D. Cosman and J. Chalupny, Amgen, Thousand Oaks, CA) or rabbit polyclonal anti-actin (Sigma-Aldrich), followed by secondary HRP-conjugated mAb.

Macrophages were collected by gentle scraping and incubated with 10% autologous human serum in PBS for 45min on ice. Ab staining was performed by incubating cells with the respective primary and secondary Abs in PBS/2% BSA/0.01% azide for 45 min on ice. Samples were acquired on a FACSCalibur (BD Biosciences), and data were analyzed using FlowJo software (Tree Star, Ashland, OR). For analysis of MICA expression, anti-MICA mAb BAM195 was used unless stated otherwise. Live/dead cell discrimination was performed using propidium iodide (Invitrogen, Carlsbad, CA).

Lysis was assessed in standard 5 h [³⁵S]Met release assays. For blocking experiments, NK cells were incubated with 10 μg/ml of the respective mAb for 30 min before and during coincubation with target cells.

mRNA was isolated (RNeasy Mini kit; Qiagen, Valencia, CA) and reverse transcribed (High Capacity Reverse Transcription kit, Applied Biosystems, Foster, City, CA). Quantitative real-time PCR was performed using TaqMan Gene Expression assays Hs00741286_m1 MICA and Hs00792952_m1 MICB (also using TaqMan Universal PCR master mix; all Applied Biosystems). miRNAs were isolated from total RNA (mirVana PARIS kit) and reverse transcribed and quantified (individual TaqMan miRNA assays and TaqMan Universal PCR Master mix, No AmpErase UNG; all Applied Biosystems). Cycle thresholds were normalized to 18S rRNA or miR-16 [found to be constant after LPS stimulation (25)]. All tests were conducted in triplicate (7500 Fast Real-Time PCR System, Applied Biosystems), and relative expression was calculated using the comparative C_T method. To determine the stability of mRNA, actinomycin D (10 μg/ml) was added to macrophages stimulated for 4 h with LPS, after which cells were collected at different times and mRNA levels were determined.

A database of predicted miRNA targets in the human genome (MicroCosm, European Molecular Biology Laboratory–European Bioinformatics Institute) was analyzed to generate a list of distinct miRNAs predicted to target each gene (using pivot-table analysis, Prism). When more than one transcript for a gene was listed, the one with the highest predicted number of targeting miRNAs was included, and when a miRNA had more than one predicted target site in a gene’s 3′ UTR, it was counted only once. A histogram showing the number of distinct miRNAs predicted to target each gene was plotted. This analysis identified 21,090 genes that are predicted to be targeted by at least 1 miRNA, and the current build of the National Center for Biotechnology Information human genome contains 32,470 genes. Specific genes could then be compared with distribution across the whole genome.

Previously it has been reported that stimulation of macrophages with LPS results in surface expression of NKG2D ligands (12, 13). However, it has not been tested whether ligation of different TLRs on human macrophages could also upregulate NKG2D ligands. To test this, we stimulated primary human macrophages for 48 h with CL097 (2.5 μg/ml), a ligand for TLR-7 and TLR-8 (26); with the TLR-3 ligand poly(I:C) (50 μg/ml) (27); or with LPS (200 ng/ml), a ligand for TLR-4. All three TLR ligands upregulated CD54, a marker for macrophage activation (28), confirming that the doses of TLR ligands used were effective in causing cell activation. Interestingly, whereas LPS preferentially upregulated MICA and not MICB, CL097 stimulation resulted in upregulation of both MICA and MICB, and poly(I:C) upregulated neither (Fig. 1A, 1B). Thus, TLRs vary in their ability to trigger expression of different NKG2D ligands.

The extent to which macrophages upregulated MICA and/or MICB varied (Fig. 1B). To test whether the response was consistent for a given donor, we repeated isolations of macrophages from the same donors and analyzed MICA and MICB upregulation in response to LPS stimulation. Isolation, maturation, and analysis of cells were performed in parallel with cells from the same donors. This revealed that macrophages matured from cells isolated from the same donor on the same day responded similarly to LPS (Fig. 1C), which confirms that the heterogeneity in response was not merely due to variation in the assay itself. Analysis of donors at different times showed that the heterogeneity in response to LPS was, at least to some extent, intrinsic to the donor (Fig. 1D). To attempt to understand this variation, both between donors and between different TLR stimulations, we also analyzed the levels of secreted cytokines in the supernatant. We identified a significant correlation between the level of MICA upregulation and the levels of IL-12 and TNF-α secretion (Fig. 1E). However, stimulation of macrophages with exogenous IL-12 and TNF-α is not sufficient to upregulate MICA (data not shown), suggesting there is not a simple causal relationship where the proinflammatory cytokines drive MICA upregulation. Focusing on TLR-4 stimulation, with LPS as a prototypical example, we next set out to study how TLR signaling leads to expression of NKG2D ligands.

To probe how LPS regulates expression of MICA, we first compared how protein and mRNA levels of MICA and MICB were affected by stimulation of macrophages with LPS. Surface expression of MICA and MICB proteins was assessed by flow cytometry at different times after stimulation with LPS (200 ng/ml) (Fig. 2A). Surface expression of MICA protein did not increase within 4 h of LPS stimulation, but a modest increase was detectable after 10 h and expression was further increased after 24 h and 48 h of stimulation. In contrast, MICB protein was not expressed at any time (Fig. 2A). Surface expression of CD54 also increased over time (Fig. 2A), but was already detectable within 4 h of LPS stimulation, demonstrating that the inability to detect MICA or MICB at this early time point was not caused by macrophages not responding. When LPS was washed away after 4 h, surface expression of MICA was reduced in comparison with cells cultured for 48 h in the continuous presence of LPS (Fig. 2B). Hence, prolonged stimulation with LPS was required for maximal expression of MICA.

Next we assessed expression of MICA and MICB mRNA by quantitative real-time PCR at different times after LPS stimulation. In contrast to the time course of protein surface expression, we detected increased levels of both MICA and MICB mRNA rapidly (i.e., 4 h after LPS stimulation) (Fig. 2C). At 48 h poststimulation MICA mRNA expression remained elevated, whereas the level of MICB mRNA had dropped back to that found in unstimulated cells (Fig. 2C), suggesting that sustained high levels of mRNA expression favor protein surface expression.

To test how LPS increased steady-state levels of MICA or MICB mRNA, we next set out to test whether LPS stimulation influenced the stability of MICA and/or MICB mRNA. For this, macrophages were stimulated with LPS (200 ng/ml), and after 4 h, further transcription was stopped by adding the inhibitor actinomycin D (29). Levels of MICA and MICB mRNA were then analyzed by quantitative real-time PCR at different times. This revealed that the stability of MICA and the stability of MICB mRNA are dramatically different but that LPS did not influence either mRNA stability (Fig. 2D, 2E). This finding suggests that the increase in MICA mRNA expression after LPS stimulation is caused by de novo transcription. Interestingly, however, although MICA mRNA was very stable (Fig. 2D) and maximum mRNA levels were induced 4 h after LPS stimulation (Fig 2C), sustained LPS stimulation was necessary for optimal MICA expression (Fig 2B). This implies that MICA protein expression is additionally controlled posttranscriptionally.

It has been previously established that ATM and ATR kinases are involved in the regulation of MICA and MICB expression in human fibroblasts and tumor cells (15, 30). To ascertain the role of these kinases in TLR-4–mediated expression of MICA in macrophages, we first tested whether caffeine, a widely used inhibitor to ATM/ATR (31), influenced expression of MICA triggered by LPS. Macrophages were either left unstimulated or stimulated with LPS for 24 h, either with or without first being treated for 2 h with caffeine. Cells stimulated with LPS upregulated expression of CD54, indicating that macrophages were activated irrespective of whether they were pretreated with caffeine. However, when cells were pretreated with caffeine, LPS did not induce surface expression of MICA, whereas the same cells did upregulate MICA without caffeine treatment (Fig. 3A). To determine whether caffeine affected expression of surface or total MICA protein, we also analyzed MICA expression by Western blotting the macrophage lysates. This revealed that pretreatment with caffeine reduced expression of total MICA protein in LPS-stimulated cells (Fig. 3B), demonstrating that the reduction of MICA surface expression produced by caffeine is not caused by intracellular retention of MICA but by a decrease in the total MICA protein.

The effects of caffeine can be broad (32), and thus we also tested KU55933, a more specific ATM inhibitor (33). Again, cells were stimulated with LPS (200 ng/ml) for 24 h, with or without pretreatment of either 1 μM or 10 μM KU55933 for 2 h. CD54 upregulation was not affected by KU55933, indicating that macrophages were efficiently activated by LPS irrespective of treatment with KU55933. However, upregulation of MICA was reduced in a dose-dependent manner when cells were treated with KU55933 (Fig. 3C). This observation suggests that ATM/ATR-mediated pathways are involved in upregulation of MICA protein expression in human macrophages.

We next analyzed MICA mRNA levels by quantitative real-time PCR in macrophages that were stimulated with LPS, with or without 2 h pretreatment of either caffeine or KU55933. Neither caffeine nor KU55933 reduced expression of MICA mRNA (Fig. 3D, 3E). Thus, although ATM/ATR kinases are involved in MICA protein expression, they do not act by controlling MICA mRNA levels. This finding implies that ATM/ATR kinases control MICA protein expression via translational or posttranslational mechanisms, whereas MICA mRNA levels are controlled through a separate signaling pathway downstream of TLR-4.

It was recently reported that several miRNAs, most of which are expressed in specific cell types and tissues, can restrain MICA expression (19, 20). Indeed, bioinformatic analysis of the complete database of miRNAs and their predicted targets [MicroCosm (34)] revealed the surprising result that MICA lies in the top 1% of genes ranked according to the number of miRNAs targeting each gene (Fig. 4A). This finding is consistent with miRNAs playing a major role in regulating MICA expression. Importantly, the expression levels of miRNAs proven to target MICA remained unchanged when MICA protein expression was triggered by cellular stresses, such as viral infection or heat shock (20). Thus, miRNAs targeting MICA are thought to establish a threshold in mRNA expression that has to be surpassed for expression of MICA protein. Four of the miRNAs proven to target MICA, namely, miR–17-5, miR-20a, miR-93, and miR-106b, are expressed in human CD14-positive monocytes, as determined by a miRNA “atlas” (35). We therefore compared expression of these four miRNAs in macrophages with and without LPS stimulation, using quantitative real-time PCR (Fig. 4B, 4C).

Strikingly, the expression of all four miRNAs could be reduced upon LPS stimulation. However, the extent to which expression of individual miRNAs changed varied between donors (Fig. 4B). Of 16 donors tested, macrophages derived from 4 of the donors had decreased expression of all miRNAs, 7 of the donors had decreased expression of three of the four miRNAs tested, 2 had decreased expression of two miRNAs, 1 had decreased expression of one miRNA, and no decreased expression of any miRNA was detected in two populations (examples shown in Fig. 4B). Expression of miRNA-20a and miRNA-93 was decreased with statistical significance across all donors, with decreased expression of miRNA–17-5 being very close to statistical significance (Fig. 4C). Thus, in contrast to invariant levels of miRNAs previously reported in other circumstances, these data suggest that LPS downregulates miRNAs that could directly facilitate MICA protein expression.

To test the ability of miRNAs to influence MICA expression in myeloid cells, we transfected the cell line THP-1, which endogenously expresses MICA, with synthetic modified oligonucleotides complementary to specific miRNAs (36) (i.e., antagomirs, to individually target miR–17-5, miR-20a, miR-93, and miR-106b). Transfection of THP-1 with all four antagomirs increased MICA surface expression in comparison with a control antagomir, while not affecting MHC class I expression (Fig. 5A). Each antagomir alone was able to increase MICA surface expression in THP-1 (Fig. 5B), implying that regulation of any of these miRNAs alone influences MICA protein expression. Moreover, MICA could also be upregulated in primary human macrophages transfected with antagomirs (Fig. 5C), showing that MICA expression can be directly controlled by miRNAs in primary human macrophages.

An important question is whether the relatively small increases in MICA expression caused by downmodulated miRNA activity (Fig. 5A–C) would be sufficient to influence NK cell activation. Thus, we tested whether cells treated with antagomirs would become susceptible to increased NK cell-mediated lysis. Untreated THP-1 cells are already lysed by NK cells to some extent, but importantly in this study, treatment of THP-1 with the four antagomirs resulted in increased lysis (Fig. 5D). The increase in lysis was modest but consistent with previous observations by others testing how inhibition of miRNAs by antagomirs can functionally influence NK cell-mediated lysis (20, 37). Moreover, the increased lysis could be abolished by addition of an anti-NKG2D–blocking mAb, but not an isotype-matched control mAb (Fig. 5E), which is consistent with an important role for NKG2D recognition in mediating increased lysis caused by antagomirs. This clarifies that control of MICA expression by miRNAs can influence NK cell recognition. Hence, in summary, multiple mechanisms link TLR-4 stimulation to the expression of NKG2D ligands, which can in turn influence NK/macrophage cross-talk, as schematically summarized in Fig. 6.

Reciprocal regulation between human myeloid and NK cells has received much attention recently (38–41). Human macrophages were shown to induce NK cell proliferation and cytokine secretion, whereas exposure to high amounts of LPS rendered them susceptible to lysis by NK cells via upregulation of ligands for NKG2D. In this study, we extended these findings by showing that ligation of TLR-4 with LPS or ligation of TLR-7 or -8 with CL097, but not ligation of TLR-3, upregulated NKG2D ligands. Now, a central gap in this field is how NKG2D ligand expression is linked to TLR stimulation, because mechanisms to control upregulation of NKG2D ligands have only previously been studied for tumor transformation or infection (42), or after T cell activation (43).

First, our data imply that signaling cascades downstream of the adapter MyD88 mediate surface expression of NKG2D ligands, as signaling through TLR-4, -7 and -8, but not TLR-3, is coupled to this adapter. This is consistent with previous research showing impaired expression of NKG2D ligands after TLR stimulation in MyD88 knockout mice (12).

Second, we found an involvement of the ATM/ATR pathway in controlling NKG2D ligand expression following TLR ligation in macrophages (as suggested in the schematic diagram in Fig. 6). The kinases ATM and ATR are best characterized for their role in sensing breaks in double- and single-strand DNA, respectively, and in subsequent initiation of the DNA damage pathway. Recently, this pathway was also implicated in expression of mouse and human NKG2D ligands in nontumor cells following genotoxic stress and stalled DNA replication (15). Furthermore, upregulation of NKG2D ligands in response to TCR engagement and viral infection involves activation of this pathway (43, 44). In this study, we demonstrated that inhibition of the ATM/ATR pathway by two different inhibitors, caffeine and KU55933, impaired LPS-stimulated MICA surface and total protein expression, while not affecting the upregulation of MICA mRNA. Interestingly, this implies that in macrophages the ATM/ATR pathway regulates MICA by a translational or posttranslational mechanism, and pathways that do not involve ATM/ATR kinases are important in upregulation of MICA mRNA. In contrast, previous research has demonstrated an involvement of the ATM/ATR pathway in expression of NKG2D ligand mRNA. For example, fibroblasts from mice with a deletion of the ATR gene displayed reduced mRNA expression of the mouse NKG2D ligands MULT1 and Rae1 after exposure to the genotoxic agent aphidicolin (15). Also, increased MICA mRNA expression in T cells stimulated with PHA was inhibited by caffeine (43), and MICA and MICB mRNA expression in Dicer knockout cells was reduced when cells were treated with inhibitors to ATM/ATR or when DNA damage pathway components were targeted by siRNA (45). In contrast, consistent with the ATM/ATR pathway not being involved in regulating NKG2D ligand mRNA levels, caffeine had no effect on the expression of ULBP-1 mRNA induced by proteasome inhibition (46). Together these data suggest that the involvement of the ATM/ATR pathway in controlling expression of NKG2D ligand mRNA can depend on the particular ligand, the specific cell type, and/or the specific stimulus.

Our data suggest that elevated levels of MICA mRNA by LPS are the result of de novo transcription, although we did not test for this directly, as we were unable to detect changes in MICA or MICB mRNA stability after LPS stimulation at times when mRNA levels were already elevated. However, it is interesting that we observed the stability of MICA and MICB to be dramatically different, MICB being more unstable. Differences in MICA and MICB mRNA stability could have been driven, for example, in response to specific pathogens (47). Some viruses interfere with mRNA stability (48), and hence differences in the stability of NKG2D ligand mRNAs could be advantageous for the host to ensure some ligands are expressed if others are interfered with.

Gene expression is regulated by cellular miRNAs at both the level of mRNA stability and translation (49). Importantly, previous research showed that the levels of miRNAs that target NKG2D ligands remained unchanged in response to stress stimuli or viral infection (20). In contrast, we found that TLR-4 stimulation correlates with reduced expression of relevant miRNAs that can directly increase MICA protein expression. At least four miRNAs—miR–17-5, miR-20a, miR-93, and miR-106b—can control MICA expression in human macrophages, as shown by using antagomirs to inhibit their function. Interestingly, we found that targeting any of these miRNAs alone was sufficient to change surface expression of MICA. This implies that changes in the expression level of one relevant miRNA could affect MICA expression. Indeed, we found that macrophages from 14 of 16 donors tested downregulated at least one of the four miRNAs after exposure to LPS. It is well established that most miRNAs, including those regulating MICA (20), target multiple transcripts (50). It could thus be speculated that flexibility for macrophages to use different miRNAs to regulate expression of MICA is important for allowing modulation of MICA expression when macrophage gene expression is in different network states (51).

In summary, we set out to study how TLR signaling linked to NKG2D ligand expression. Surprisingly, we found that rather than a simple linear sequence of events, several different mechanisms are involved in the expression of NKG2D ligands by human macrophages in response to TLR-4 stimulation. Furthermore, the correlation between secretion of proinflammatory cytokines and the level of MICA upregulation on macrophages suggests an important role for MICA expression in the regulation of macrophage/NK cell crosstalk during inflammation. The complexity of the regulation may reflect the importance for ensuring appropriate expression of NKG2D ligands that are capable of triggering potent immune responses while being highly targeted by pathogens.

We thank B. Askonas, L. Bugeon, F. Culley, N. Guerra, J. Harris, and members of our laboratory for discussions and critical reading of the manuscript.

Disclosures The authors have no financial conflicts of interest.