Immune Activation Resulting from NKG2D/Ligand Interaction Promotes Atherosclerosis

Abstract

Background

The interplay between the immune system and abnormal metabolic conditions sustains and propagates a vicious feedback cycle of chronic inflammation and metabolic dysfunction that is critical for atherosclerotic progression. It is well established that abnormal metabolic conditions, such as dyslipidemia and hyperglycemia, cause various cellular stress responses that induce tissue inflammation and immune cell activation, which in turn exacerbate the metabolic dysfunction. However, molecular events linking these processes are not well understood.

Methods and Results

Tissues and organs of humans and mice with hyperglycemia and hyperlipidemia were examined for expression of ligands for NKG2D, a potent immune activating receptor expressed by several types of immune cells, and the role of NKG2D in atherosclerosis and metabolic diseases was probed using mice lacking NKG2D or by blocking NKG2D with monoclonal antibodies. NKG2D ligands were upregulated in multiple organs, particularly atherosclerotic aortae and inflamed livers. Ligand upregulation was induced in vitro by abnormal metabolites associated with metabolic dysfunctions. Using ApoE^-/- mouse models we demonstrated that preventing NKG2D functions resulted in a dramatic reduction in plaque formation, suppressed systemic and organ inflammation mediated by multiple immune cell types, and alleviated abnormal metabolic conditions.

Conclusions

The NKG2D/ligand interaction is a critical molecular link in the vicious cycle of chronic inflammation and metabolic dysfunction that promotes atherosclerosis and might be a useful target for therapeutic intervention in the disease.

Introduction

Along with other risk factors such as metabolic dysfunction, immune cells are critically involved in promoting atherosclerosis^{1, 2}. Abnormal metabolic conditions induce stress responses in cells of artery walls to initiate vascular inflammation and result in the recruitment of immune cells that enhance the inflammation and promote plaque formation. Conversely, chromic inflammation exacerbates metabolic dysfunction, creating a vicious cycle of chronic inflammation and metabolic dysfunction that propagates atherosclerosis³.

Macrophages are the most abundant immune cells in plaques¹, where they accumulate lipid contents and secrete pro-inflammatory cytokines and chemokines. Various lymphocytes, including T and natural killer (NK) cells, are also detected in plaques and involved in atherosclerosis ^4-6. A role for T/B cells in atherosclerosis was initially suggested by the findings that chow-fed Rag1^-/-ApoE^-/- mice, which lack T/B cells, developed atherosclerotic lesions slower than ApoE^-/- control mice did⁷. Subsequent experiments suggested a role for CD4⁺ T cells in promoting the disease, by responding to antigens associated with abnormal metabolites and enhancing production of interferon-γ (IFN-γ)⁸. Additional studies found a modest decrease in plaque formation in mice lacking NKT cells, a T cell subset reactive against specific lipids^{6, 9}. On the other hand, regulatory T cells inhibit atherosclerosis by suppressing immune activation¹⁰. Innate lymphocytes such as NK cells have been also implicated in promoting atherosclerosis^6,11.

Involvement of immune cells in atherosclerosis is thought to be initiated by abnormal metabolic conditions and the resulting stress to arterial endothelial cells. Abnormal metabolic components activate endothelial cells and macrophages of susceptible regions such as aortic arches through Toll-like receptors 2/4 to promote vascular inflammation and atherosclerosis^12,13. Stressed endothelial cells also upregulate chemokines and adhesion molecules (including CCL5, CXCL1, macrophage migration inhibitory receptor, E-selectin, and vascular cell-adhesion molecule 1) to recruit atherogenic immune cells that promote vascular inflammation (reviewed in reference¹⁴). Immune cells, including NK and NKT cells, are also activated in other inflamed organs, such as the liver, to contribute indirectly to the atherosclerosis by increasing systemic inflammation and metabolic dysfunction. However, molecular events that link activation of different immune cells and abnormal metabolic conditions in atherosclerotic progression are not well understood.

NKG2D (also called killer cell lectin-like receptor subfamily K, member 1 or KLRK1) is a potent activating receptor expressed by several types of immune cells involved in atherosclerosis, including NK, NKT, γδ T and subsets of activated αβT cells^{15, 16}. Multiple different membrane proteins have been identified as ligands for NKG2D. In humans, they include two major histocompatibility complex class I chain-related family members (MICA and MICB) and five UL16-binding protein family members (ULBP1-4 and RAET1G). In mice, they include five retinoic acid early transcripts 1family members (Rae-1α, β, γ, δ and ε), three histocompatibility 60 family members (H60a, b and c) and MULT-1 (UL16-binding protein-like transcript 1)¹⁵. These ligands are generally expressed at low levels or not at all on healthy cells but are upregulated under various pathological conditions such as tumorigenesis, viral infection and genotoxic stress. Interaction of the ligands with NKG2D stimulates, or co-stimulates, the NKG2D-expressing immune cells to proliferate, produce cytokines (including IFN-γ, GM-CSF, MIP-1α and IL-2) and/or lyse target cells that express the ligands^15-20. While the upregulation of NKG2D ligands plays an important role in host defense against tumors and viral infections^{19, 21} their abnormal upregulation could contribute to immune-mediated inflammatory diseases such as arthritis, celiac disease and type 1 diabetes^22-25. We report herein that the NKG2D ligands were upregulated in humans and mice with hyperglycemia and/or hyperlipidemia and play a critical role in atherosclerosis, liver inflammation and the associated metabolic disease.

Methods

Mouse models and human samples

ApoE^-/- mice on C56BL/6 background were purchased from Jackson Lab. NKG2D knockout mice (Klrk1-/- mice) on the C56BL/6 background were recently described²⁶. ApoE^-/- and Klrk1-/- mice were intercrossed to generate Klrk1-/-ApoE^-/- mice. The Klrk1 and ApoE knockout genotypes were determined by genomic PCR, as exemplified in Supplementary Fig. 1.

In the Western diet (WD)-accelerated model of atherosclerosis, male ApoE^-/- and Klrk1-/-ApoE^-/- mice were switched from normal chow to a high fat diet (35.5% Fat, Bio Serv) at the age of six weeks and analyzed for plaques, blood cytokines and metabolites 8-9 weeks later. In the streptozotocin (STZ)-induced diabetic model of atherosclerosis²⁷, six-week old male ApoE^-/- and Klrk1^-/-ApoE^-/- mice were injected intraperitoneally with STZ (55mg/kg in 0.05M citrate buffer, pH 4.5) daily for 6 days. The treated mice were confirmed diabetic (blood glucose >350mg/dL) and fed on chow for additional 8-9 weeks before analysis. In both models, the Klrk1-/-ApoE^-/- and ApoE^-/- mice were always treated and analyzed in parallel.

In experiments to test the effect of NKG2D antibody blockade on atherosclerosis, six-week old male ApoE^-/- mice were injected intraperitoneally with monoclonal anti-NKG2D (clone MI-6, rat IgG2a)¹⁶ or isotype-matched control antibodies one week and three days before the STZ treatment and weekly afterwards for eight weeks (200μg/1^st injection and 100μg/subsequent injection). The mice were analyzed 8-9 weeks after the beginning of STZ treatment. All procedures were approved by the IACUC of The Pennsylvania State University. Usage of bio-samples of humans was approved by the IRB of The Pennsylvania State University and Harvard Medical School.

Antibodies and flow cytometry

The NKG2D antibody (MI-6) was prepared in house. Pan-Rae-1 antibody was purchased from R&D Systems. PE-anti-CD31, PECy5-CD11b, PECY7-anti-CD3, PE-anti-TCRβ, PE-anti-F4/80 and Alexa647-anti-NKG2D antibodies were purchased from eBioscience. FITC-anti-NK1.1, PECy5-anti-TCRβ, Alexa488-anti-IL-6, PE-anti-IL4 and PECY7-anti-IFNγ antibodies and the mouse IL-6 ELISA set were from BD Bioscience. For flow cytometry, cells were stained with proper combination of antibodies (indicated in the figures) and analyzed on a flow cytometer FC500 (Beckman Coulter).

En face Oil red O staining and quantification of plaque formation

Aortae beginning at the root outside the heart and stopping at the thoracic diaphragm were collected, opened up longitudinally, fixed in 10% formalin and stained with Oil red O dye as described¹². Sizes of total aortic arch and neighboring regions, and of Oil red O staining plaque areas within the regions, were calculated based on digital pictures of the stained aortae using Adobe Photoshop, as reported²⁸. The extent of plaque formation was expressed as the percentage of Oil red O stained area of the total area of the aortic arch and neighboring region.

Immunohistochemical (IHC) staining

Cryosections of mouse atherosclerotic plaques were stained with polyclonal goat anti-mouse Rae-1 or H60 antibody (R&D Systems or Santa Cruz Biotechnology respectively). Cryosections of human aortic plaques were stained with monoclonal anti-MICA/B (clone 6D4, eBioscience), anti-Mac-3 or anti-CD31 antibodies, performed similarly as previously described²⁹.

Cellular isolation

Mononucleocytes, endothelial cells and hepatocytes were isolated from aortae or livers according to published procedures^{30, 31}. Briefly, the liver was perfused with 3ml 50U/ml Heparin (Sigma-Aldrich) and 3ml 0.05% Collagenase A /0.05% Dispase II (Roche) via the portal vein and then pressed through a cell strainer (PGC Scientific). Liver mononucleocytes and hepatocytes were separated by 40/80% Percoll solution (Sigma-Aldrich). For isolation of cells from normal aortae and atherosclerotic aortic plaques, mouse aortae were first perfused with 50U/ml heparin and then digested with an enzyme cocktail dissolved in PBS (125 U/ml collagenase type XI, 60 U/ml hyaluronidase type I-s, 60 U/ml DNase1, 450 U/ml collagenase type I and 20 mM Hepes, Sigma-Aldrich) at 37°C for 1 hr. The isolated cells were counted and used for analyses.

Real-time RT-PCR analysis

RNA was analyzed for transcripts of Rae-1δ, Rae-1ε and H60b by quantitative real-time RT-PCR, as previously described^{18, 19, 32}.

In vitro treatment of macrophages

Resident cells were flushed out of the peritoneal cavity of naïve C56BL/6 mice with PBS and plate-adhered overnight. The adherent macrophages were cultured in 10%FBS-supplemented DMEM media containing 10 μg/ml human oxidized low-density lipoprotein (oxLDL, medium-oxidized with copper (II) sulfate), natural LDL (nLDL) (Kalen Biomedical), 200μg/ml advanced glycation end products (AGE) (Fitzgerald Industries Intl), or lipopolysaccharides (LPS) for two days³³, and then analyzed for expression of NKG2D ligands.

Western blot analysis of total Rae-1 proteins

The macrophages were lysed, separated on PAGE gels, transferred to PEGF membrane, blotted with anti-Rae-1 antibody (C20, Santa Cruz Biotechnology) and developed with an ECL (enhanced chemiluminescence) system (GE Healthcare Life Sciences).

Multiplex analysis of serum cytokines

Mouse sera were analyzed directly on a mouse proInflammatory-7plex kit (IL-6, TNF–α, IFN–γ, IL-12p70, IL-10, IL-1β and KC/CXCL1) (Meso Scale Discovery, Gaithersburg, MD).

Assessment of cholesterol and triglyceride levels in serum

Levels of cholesterol and triglycerides were analyzed on VetTest® Chemistry Analyzer (IDEXX Laboratories, Inc. Maine).

Alanine aminotransferase (ALT) activity analysis

ALT activities in the serum were determined using a kit from Bioo Scientific (Austin, TX) according to the manufacturer’s instruction.

Ex vivo culture of liver explants

PBS-perfused livers were excised from WD-fed ApoE^-/- or Klrk1-/-ApoE^-/- mice. Equal weights of excised livers were cut into about 1mm³ cubes and cultured in media for 1 or 3 days. The culture media were recovered and analyzed for cytokines by ELISA.

Intracellular cytokine staining

Mononucleocytes isolated from livers were cultured in media overnight in the presence of brofeldin A and analyzed by intracellular cytokine staining for the production of IL-6 and IFN–γ in gated subsets of immune cells according to the manufacturers’ instructions (eBioscience and Biolegend).

ELISA (enzyme-linked immunosorbent assay) detection of soluble MICA

MICA proteins in sera were assayed using an ELISA kit (R&D Systems, Minneapolis, MN).

Statistical analyses

Data are expressed as means ± standard errors. Two-tailed student T tests were used to determine statistical significance for two-group comparison. The ANOVA test was used for analysis of combined results of repeated experiments or multiple-group comparison (with Tukey adjustment). The Fisher’s exact test was used to compare proportions of diabetic and healthy people with detectable levels of soluble MICA proteins (sMICA) in sera. P < 0.05 is considered significant.

Results

Upregulation of NKG2D ligands in sera and atherosclerotic plaques of humans with metabolic dysfunctions

To investigate whether NKG2D or its ligands are involved in atherosclerosis associated with metabolic dysfunction, we first sought evidence that NKG2D ligands were upregulated in type 2 diabetes patients (Supplemental Table 1) by assessing sMICA in their sera. sMICA are enzymatically cleaved products of membrane MICA and were detected abundantly in patients with various cancers and immune disorders but not healthy people^{22, 34, 35}. Of 22 diabetes patients tested, 45% (10) had detectable sMICA (>25 pg/ml) and 27% (6) had high levels (400-12250 pg/ml) while in the healthy controls, only 11% (2/19) had detectable levels and 5% (1/19) had high levels of sMICA in sera (Fig. 1A). These results suggest that a high percentage of type 2 diabetes patients have upregulated MICA expression. We tried to associate the sMICA detection in the patients with other parameters (ages, years with diabetes and glucose levels) but found no correlation (Supplemental Table 1), suggesting that these factors might not be directly associated with the MICA upregulation.

Figure 1.

Upregulation of NKG2D ligands in sera and aortic plaques of human type 2 diabetes patients. (A) A high percentage of type 2 diabetes patients have elevated levels of sMICA in the sera. The dashed line indicates the detection limit. One dot represents one sample and the short flat lines indicate average concentrations. Fisher’s exact test was used to compare proportions of healthy vs. diabetic people with detectable MICA expression. *p<0.05; **p<0.01; ***p<0.001 (applied for all figures). (B-J) Detection of MICA/B proteins by IHC staining on human aortic plaques. Cryosections of the plaques were stained with anti-MICA/B antibodies (B, D, F, H) while adjacent sections were stained with anti-Mac-3 or CD31 antibodies to reveal macrophages or endothelial cells (C, E, G, I). A section of normal aorta was used as a control (panel J).

We also performed IHC staining of atherosclerotic aortic sections of humans to assess expression of NKG2D ligands in the tissues. In contrast to plaque-free aortae (Fig. 1J), atherosclerotic aortae contained many cells that stained positive with an anti-MICA/B antibody (Fig. 1B, D, F and H). The overlapping staining patterns of MICA/B with Mac-3 or CD31 in adjacent sections suggested that both macrophages and endothelial cells of the atherosclerotic aortae expressed MICA/B (Fig. 1B-I). The high expression of NKG2D ligands in aortic plaques suggested the potential involvement in atherosclerosis of immune activation mediated by NKG2D-ligand interactions.

Preferential upregulation of NKG2D ligands in tissue and organs of ApoE^-/- mice where abnormal metabolites accumulate

To understand the NKG2D ligand upregulation process and its role in atherosclerosis, we used the ApoE^-/- mouse model of atherosclerosis. ApoE^-/- mice are defective in lipid metabolism and develop plaques when they age, particularly in susceptible regions such as aortic arch areas, and the disease is markedly exacerbated when the mice are fed the high fat Western diet (WD).

We first assessed the amounts of transcripts for NKG2D ligands in RNA isolated from atherosclerotic aortic arch regions of WD-fed ApoE^-/- mice. Compared to those of either WD- or chow-fed wild-type C57BL/6 (WT) mice, samples from WD-fed ApoE^-/- mice had much higher (10-30 fold) amounts of transcripts for Rae-1δ, Rae-1ε and H60b, which are mouse ligands for NKG2D (Fig. 2A). These results suggest that the ligands are upregulated in the atherosclerotic aortae. Consistent with this conclusion, the atherosclerotic aortic arch regions of older chow-fed ApoE^-/- mice also had much higher amounts of transcripts for the NKG2D ligands than the atherosclerosis-resistant thoracic aortic region of the same mice (Supplementary Fig. 2).

Figure 2.

Preferential upregulation of NKG2D ligands in atherosclerotic aortae and livers of ApoE^-/- mice. (A) Real-time RT-PCR analysis of Rae-1δ, Rae-ɛ and H60b transcripts in RNA samples isolated from aortic arch regions of WD-fed ApoE^-/- mice. Samples of age-matched WD- and chow-fed WT mice were used as controls. The experiment included 3 mice/group and was representative of two repeated experiments of total 6 mice. (B) IHC staining of sections of atherosclerotic aortic arch regions (top and middle) and “plaque-free” thoracic regions (bottom) of the same WD-fed ApoE^-/- mice with Rae-1 (left) or H60 antibodies (right). The pictures in the middle are higher amplifications (400X) of the boxed areas of the top panels (100X). (C) Flow cytometry of cell surface Rae-1 expression on macrophages (gated on the CD45⁺CD11b⁺F4/80⁺ population) and endothelial cells (CD45^-CD31⁺ population) isolated from atherosclerotic aortae of the WD-fed ApoE^-/- mice or normal aortae of WT mice. The gray areas indicated isotype-matched control antibody staining. (D) Real-time RT-PCR analysis of Rae-δ transcripts in indicated organs of WD-fed ApoE^-/- mice and WT mice. *** indicates statistical significance compared to values of corresponding tissues of WT mice. N=4. (E) Real-time RT-PCR analysis of Rae-δ in purified macrophages of livers of WD-fed ApoE^-/- mice and WT mice. N=2. (F and G) Flow cytometry of Rae-1 expression on CD11b⁺F4/80⁺ liver macrophages (F) and hepatocytes (G) of WD-fed ApoE^-/- mice. The experiment included 3-4 mice/group and was performed twice with similar results. (H) Flow cytometry of Rae-1 expression on peritoneal macrophages of WT mice cultured in vitro in presence of oxLDL and AGE as in the panel C except that the dashed lines depict staining in the presence of a 10-fold excess of unlabeled anti-pan-Rae-1 antibodies. The experiment was performed 3 times, with pools of cells from 3 mice for each, with similar results. (I) Western blot analysis of Rae-1 proteins in macrophages cultured in vitro in presence of oxLDL or nLDL. The LPS treatment was included as a positive control³³. The blot was performed 3 times with similar results.

IHC staining with Rae-1 and H60 antibodies confirmed the preferential expression of NKG2D ligands in plaques. Extensive Rae-1 and H60 staining was detected within plaques of WD-fed ApoE^-/- mice of the atherosclerotic aortic arch regions (Fig. 2B, top and middle panels). In contrast, Rae-1 or H60 staining was nearly undetectable in plaque-free aortic regions of the same mice (Fig. 2B, bottom). To determine the types of cells that express the ligands, we performed flow cytometric analysis of cells isolated from atherosclerotic aortae of WD-fed ApoE^-/- mice. In contrast to those of normal aortae, many macrophages and endothelial cells of the atherosclerotic aortae had appreciable cell surface Rae-1 expression (Fig. 2C), consistent with the findings in humans.

Since metabolic dysfunctions are systemic conditions, we tested whether other organs in the WD-fed ApoE^-/- mice had upregulated NKG2D ligands. Rae-1 transcripts were elevated dramatically in the liver and somewhat less so in the kidney, whereas other organs tested showed lower elevations (Fig. 2D). Similar to the results with macrophages from aortic plaques, liver macrophages of ApoE^-/- mice contained significantly more Rae-1 transcripts than WT controls (Fig. 2E), and cell surface expression of Rae-1 was also significantly, although modestly, increased (Fig. 2F). Rae-1 was also markedly increased on the surface of hepatocytes of the ApoE^-/- mice, though WT hepatocytes exhibited a higher basal level of Rae-1 than did liver macrophages (Fig. 2G).

Considering that abnormal metabolites accumulate in both atherosclerotic plaques and livers, they may play a role in the induction of NKG2D ligands. Indeed, WT macrophages could be reproducibly induced to upregulate NKG2D ligands when cultured in vitro in presence of oxLDL or AGE, two abnormal metabolites associated with dyslipidemia and hyperglycemia (Fig. 2H-I and Supplementary Fig. 3). The Rae-1 staining was competitively inhibited by un-labeled Rae-1 antibodies, confirming the specificity of the staining (Fig. 2H). Together, these results suggest that abnormal metabolic conditions induce upregulation of NKG2D ligands in various tissues and organs, particularly those where abnormal metabolites accumulate, such as aortic plaques and livers. Considering that immune cell subsets such as NK cells and NKT cells in those tissues express NKG2D (Supplementary Fig. 4) and promote atherosclerosis^{6, 9, 11}, the NKG2D/ligand axis might play an important role in promoting atherosclerosis through activation of these or other NKG2D-expressing cells.

Preventing NKG2D/ligand interactions suppresses plaque formation in ApoE^-/- mice

To directly determine the role of NKG2D in atherosclerosis, we tested whether interfering with NKG2D function suppresses plaque formation in ApoE^-/- mice. Compared to WD-fed ApoE^-/- mice, similarly fed NKG2D-deficient Klrk1-/-ApoE^-/- littermates exhibited dramatically smaller plaques in the aortae (Fig. 3A-C). On average, the percentage of lipid-deposited plaque area in the total aortic arch region, as assessed by en face Oil red O staining, was reduced 5-6 fold in Klrk1^-/-ApoE^-/- mice (Fig. 3B and C). Oil red O staining of cross-sections of aortic roots also revealed significantly reduced plaque formation in Klrk1^-/-ApoE^-/- mice compared to the ApoE^-/- controls but the effect was less dramatic than seen in the aortic arch regions (Supplementary Fig. 5A). The different effects of NKG2D knockout on the plaque reduction in the aortic arch region and aortic roots are consistent with other reports showing that immune-mediated mechanisms differentially influence plaque formation in different anatomical sites^{36, 37}.

Figure 3.

NKG2D plays a critical role in plaque formation in WD-fed and STZ-diabetic ApoE^-/- mice. (A) Visibly different plaque sizes in aortic arch regions of representative WD-fed ApoE^-/- and Klrk1^-/-ApoE^-/- mice. The whitish plaque areas are indicated by arrows. The pictures shown are representative of 16-18 mice of each genotype. (B) En face Oil red O staining to detect lipid-deposited plaque areas in aortae of representative ApoE^-/- (n=12) and Klrk1^-/-ApoE^-/- (n=10) mice. Plaque areas stain red. The regions analyzed for percentage of plaque are boxed. (C and D) Quantitative comparison of percentages of average Oil red O staining positive plaque areas in the total aortic arch regions of Klrk1^-/-ApoE^-/- and ApoE^-/- mice that were fed on a WD (C) or treated with STZ (D). (E) Quantitative comparison of percentages of average Oil red O staining positive plaque areas in STZ-induced diabetic ApoE^-/- mice treated with NKG2D or control antibodies. Total numbers of mice analyzed in panels C, D and E are indicated in the graphs, which are compiled from data of 2-4 experiments for each model.

To test whether the NKG2D/ligand axis is broadly involved in atherosclerosis that accompanies diabetic conditions, we also examined STZ-treated ApoE^-/- mice. Because it is directly toxic for insulin-secreting pancreatic β-cells, STZ causes severe hyperglycemia and hyperlipidemia in mice by a mechanism similar to that of type 1 diabetes²⁷. As in the WD-fed model, STZ-treated Klrk1-/-ApoE^-/- mice had much smaller plaque sizes than similarly treated ApoE^-/- mice in aortic arch regions (Fig. 3D). The NKG2D knockout also significantly, although less dramatically, reduced the amount of plaque in the aortic root cross sections in the STZ-diabetic ApoE^-/- mice (Supplementary Fig. 5B).

As an additional approach to test the role of NKG2D in atherosclerosis, which might be also relevant for therapeutic applications, we tested whether the induction of atherosclerosis in STZ-diabetic ApoE^-/- mice was inhibited by injection of NKG2D antibodies that are known to block interactions of NKG2D with its ligands^{16, 24}. Compared to ApoE^-/- mice injected with control antibodies, those that received the NKG2D antibodies had markedly reduced plaque areas (Fig. 3E). Together, these results demonstrate a critical role of the NKG2D/ligand axis in atherosclerosis induced under various abnormal metabolic conditions.

Preventing NKG2D/ligand interactions reduces the systemic production of multiple cytokine markers associated with immune activation

Since the NKG2D/ligand interaction activates NKG2D-expressing immune cells, which produce inflammatory cytokines, preventing the interaction likely suppresses atherosclerosis by inhibiting immune activation. Supporting this conclusion, compared to ApoE^-/- mice, NKG2D-knockout or NKG2D antibody-treated ApoE^-/- mice exhibited significant reductions in the amounts of multiple cytokines in the serum, including inflammatory cytokines (IL-6, IFN–γ) and an immune regulatory cytokine (IL-10) (Fig. 4A–C). IL-6 and IFN–γ are known to be involved in atherosclerosis³⁸, and were markedly reduced in all the models in which NKG2D function was prevented, suggesting that NKG2D/ligand-mediated immune cell activation associated with atherosclerotic progression might function at least partly through the increased production of these cytokines. IL-12 was also reduced in all the models, though the difference was not significant in the STZ model. The other cytokines studied showed more variable results.

Figure 4.

Preventing NKG2D/ligand interactions reduces serum levels of inflammation-associated cytokines in WD-fed and STZ-diabetic ApoE^-/- mice. (A) Serum levels of multiple cytokines in WD-fed ApoE^-/- and Klrk1^-/-ApoE^-/- mice. N=12 each. (B) Serum levels of multiple cytokines in STZ-diabetic ApoE^-/- and Klrk1^-/-ApoE^-/- mice. N=4 each. (C) Serum levels of multiple cytokines of NKG2D antibody (n=9) and control antibody (n=6) treated STZ-diabetic ApoE^-/- mice. Note that scales for different cytokines in the different models are not the same.

NKG2D-mediated immune activation aggravates abnormal metabolic conditions by promoting liver inflammation and dysfunction

Preventing NKG2D interactions also considerably alleviated the abnormal metabolic conditions in the WD-fed ApoE^-/- model. Notably, the sera of WD-fed Klrk1^-/-ApoE^-/- mice were much less cloudy than the sera of similarly fed ApoE^-/- mice (Supplementary Fig. 6), suggesting reduced lipid content. Direct analysis showed that cholesterol and triglycerides were significantly reduced in the Klrk1^-/-ApoE^-/- mice (Table 1). The glucose levels were also significantly, albeit modestly, reduced in the WD-fed Klrk1^-/-ApoE^-/- mice (Table 1).

Table 1.

Metabolic parameters of NKG2D-knockout and anti-NKG2D antibody treated ApoE^-/- mice in Western diet and STZ models of atherosclerosis

	Western Diet			STZ			STZ-ApoE-/- + Antibody
	ApoE-/- (n=6)	ApoE-/-Klrk1-/- (n=6)	P	ApoE-/- (n=6)	ApoE-/-Klrk1-/- (n=6)	P	Control Ab (n=5)	NKG2D Ab (n=4)	P
CL	706±89	337±48	0.005	1284±339#	1446±402	0.78	1537±528	1001±198	0.38
TG	98.3±7.6	57.5±6.8	0.002	85.0±8.0	57.5±13.5	0.22	84.8±16.3	81.5±19.7	0.9
GL	372±27*	308±14†	0.03	852±202§	550±85	0.3	794±94	578±53	0.1

Note: CL, Cholesterol (mg/dL); TG, Triglyceride (mg/dL); GL, Glucose (mg/dL).

^#P <0.05 compared to the cholesterol concentrations of WD-fed ApoE^-/- mice.

^*14 samples were analyzed.

^†12 samples were analyzed.

^§P < 0.001 compared to the glucose concentrations of WD-fed ApoE^-/- mice.

Compared to WD-fed ApoE^-/- mice, STZ treated ApoE^-/- mice had more severe hyperglycemia and hyperlipidemia, neither of which were significantly corrected by interfering with NKG2D (Table 1). These findings suggest that the severe metabolic dysfunctions in the STZ model are not dependent on NKG2D-mediated immune activation, consistent with the notion that STZ induces diabetes through its direct toxic effect on pancreatic β–cells and other cells³⁹. Nevertheless, interfering with NKG2D in the STZ model suppressed production of the cytokines associated with immune activation and atherosclerosis (Fig. 3D), suggesting that preventing NKG2D-mediated inflammation reduces atherosclerosis without requirements of alleviating overall metabolic dysfunctions. However, in the WD model, the role of NKG2D in aggravating metabolic conditions may also contribute to the atherosclerosis.

Since the liver is the major organ of lipid metabolism and exhibited highly upregulated expression of NKG2D ligands in the ApoE^-/- mice, we tested whether the NKG2D/ligand axis contributes to liver inflammation that results in liver dysfunction. Compared to WD-fed ApoE^-/- mice, WD-fed Klrk1^-/-ApoE^-/- mice exhibited significantly lower levels of serum ALT activities (Fig. 5A), indicating that their liver dysfunction was alleviated. Furthermore, liver explants from Klrk1^-/-ApoE^-/- mice, cultured ex vivo, produced significantly less IL-6 than those from ApoE^-/- mice, indicative of reduced inflammation (Fig. 5B). In addition, the numbers of macrophages, NKT cells and NK cells were also significantly lower in livers of Klrk1^-/-ApoE^-/- mice than ApoE^-/- mice (Fig. 5C). These data suggest that NKG2D-mediated immune activation is directly involved in liver inflammation and dysfunction that is known to promote abnormal metabolic conditions.

Figure 5.

Preventing NKG2D/ligand interactions alleviates abnormal metabolic conditions by reducing liver inflammation in the WD-fed ApoE^-/- model of atherosclerosis. (A) Reduced serum ALT activity in WD-fed Klrk1^-/-ApoE^-/- mice. The experiment included 5 mice/group and was representative of 2 repeated experiments of total 10 mice. (B) Reduced IL-6 production by ex vivo cultured liver explants of Klrk1^-/-ApoE^-/- mice. Experiment includes 3 mice/assay and was representative of 2 repeated experiments of total 6 mice. (C) Reduced numbers of immune cells in livers of Klrk1^-/-ApoE^-/- mice. The numbers are of each type of immune cells per liver, calculated based on total mononucleocytes isolated from livers and percentages of each type. The experiment included 3 mice/assay and was representative of 3 repeated experiments of total 9 mice.

NKG2D/ligand interaction implicates multiple immune cell populations in the liver inflammation

To gain insight into cellular mechanisms that underlie NKG2D-mediated liver inflammation, we determined the influence of NKG2D-deficiency on activation of different immune cell populations in the WD-fed ApoE^-/- mice. The cell types most likely to be affected are those that express NKG2D, including NK and NKT cells in the liver (Supplementary Fig. 4). Indeed, NKT cells, and to a lesser extent NK cells, from livers of the ApoE^-/- mice exhibited considerable production of IFN–γ, which was markedly attenuated when the mice lacked NKG2D (Figure 6A-B). In contrast, few conventional αβ T cells of the liver expressed IFN–γ (not shown) and the liver macrophages produced high levels of IL-6 in ApoE^-/- mice whether or not the mice expressed NKG2D (Fig. 6C). These results suggest that the production of IL-6 by macrophages is independent of NKG2D, consistent with the fact that they did not express NKG2D (Supplementary Fig. 4). However, since liver macrophages were markedly reduced in Klrk1^-/-ApoE^-/- mice and they expressed NKG2D ligands in atherogenic conditions (Fig. 2E-F and 5C), these cells likely stimulate NKG2D-mediated liver inflammation, and at the same time are influenced by it.

Figure 6.

Multiple immune cell types are involved in the liver inflammation mediated by NKG2D interactions. (A) Comparison of the production of IFN–γ by liver NKT cells of WD-fed Klrk1^-/-ApoE^-/- and ApoE^-/- mice. Cells are gated on TCRβ⁺NK1.1⁺ cells. (B) Comparison of IFN–γ production by liver NK cells of WD-fed Klrk1^-/-ApoE^-/- and ApoE^-/- mice. Cells are gated on CD3^-NK1.1⁺ cells. (C) On a per cell basis, NKG2D deficiency did not affect IL-6 production by liver CD11b⁺F4/80⁺ macrophages of WD-fed ApoE^-/- mice. The experiment in panels A-C included 3 mice/group and is representative of 3 repeated experiments of total 9 mice. The number below the gate in each graph is the mean percentage ± SE of the positive cells of total cells in the plot. The isotype-control staining was similar in WT, ApoE^-/- and Klrk1^-/-ApoE^-/- mice. *** and * indicate statistical significance compared to the values from ApoE^-/- mice.

Discussion

The studies presented here identify the NKG2D/ligand axis as a molecular link between chronic inflammation and metabolic dysfunction that is critical for atherosclerosis and contributes to type 2 diabetes progression. Considering that the NKG2D ligands were upregulated in multiple tissues, NKG2D-mediated immune activation likely promotes atherosclerosis through several mechanisms. In aortae, preferential upregulation of the NKG2D ligands is associated with enhanced plaque formation in arch regions, suggesting that it is directly involved in atherosclerotic progression. NKG2D/ligand-mediated immune activation could enhance the inflammation in atherosclerotic aortae, which in turn might impact recruitment, survival and function of ligand-expressing macrophages that are directly involved in plaque progression⁴⁰. Although upregulation of NKG2D ligands is most profound in developed plaques, we found that enhanced expression of the ligands was apparent in aortae of ApoE^-/- mice as early as eight weeks of age, before plaques were detected (data not shown), suggesting that the NKG2D/ligand axis might be involved in early stages of atherosclerosis.

NKG2D/ligand mediated immune activation could also promote atherosclerosis by exacerbating metabolic dysfunctions and causing systemic inflammation in the liver and potentially other tissues. In this regard, the abnormal upregulation of NKG2D ligands in the pancreas of non-obese diabetes (NOD) mice, an autoimmune type 1 diabetes model, reportedly contributes to immune destruction of β-cells and development of diabetes²⁴. Although the underlying mechanisms are different, NKG2D also plays a role in progression of type 2 diabetes in the studies presented here. Therefore, the NKG2D/ligand axis is extensively involved in multiple aspects of diabetes and atherosclerosis. It will be interesting to determine whether the NKG2D/ligand axis is involved in the development of other diabetes-associated complications, such as retinopathy, nephropathy and neuropathy.

It is likely that NKG2D exerts its disease-causing effects in multiple cell types. We observed a marked reduction in NK and NKT cell activation in livers of WD-fed ApoE^-/- mice, suggesting that NKG2D-ligand interactions may be necessary for activating these cells, and that such activation may promote metabolic dysfunctions and atherosclerosis. Due to the very limited numbers of cells we could not perform a similar analysis of cells isolated from atherosclerotic plaques. Both NK cells and NKT cells have been reported to contribute to the severity of atherosclerosis in similar mouse models. Although it is difficult to compare different studies directly, the published studies suggest that the contribution of NK cells or NKT cells individually to atherosclerosis is less marked than the contribution of NKG2D shown here. Therefore, it is possible that both populations, along with even other NKG2D⁺ cells such as γδT cells, contribute to the disease in an NKG2D-dependent fashion, and that NKG2D deficiency therefore has a greater effect on the disease than either of the NKG2D+ cell populations do separately.

We also found that the NKG2D ligands are upregulated in blood and plaques of type 2 diabetes patients, suggesting the potential clinical application of blocking NKG2D/ligands in treatment of atherosclerosis or metabolic disease in patients. However, additional preclinical studies are necessary before this approach can be considered for the clinical usage. It remains to be determined whether blockade of NKG2D affects established plaques, for example. Another issue that needs further study is whether binding of NKG2D antibody to NKG2D induces global inhibition of the functions of NKG2D-expressing cells, which was suggested by some studies⁴¹ but refuted by others⁴². Although we cannot rule out that this occurs in our antibody blockade studies, the NKG2D knockout experiments provide a direct and compelling line of evidence that NKG2D is involved in the disease. In conclusion, our findings provide new insights into the mechanisms underlying association of inflammation with atherosclerosis and type 2 diabetes and suggest possible new approaches to therapy of these diseases. In addition, the observation that sMICA is elevated in sera of diabetic patients suggests that the sMICA measurements may have utility as a biomarker of atherosclerotic progression.