AMPK-independent inhibition of human macrophage ER stress response by AICAR

Obesity-associated insulin resistance is driven by inflammatory processes in response to metabolic overload. Obesity-associated inflammation can be recapitulated in cell culture by exposing macrophages to saturated fatty acids (SFA), and endoplasmic reticulum (ER) stress responses essentially contribute to pro-inflammatory signalling. AMP-activated protein kinase (AMPK) is a central metabolic regulator with established anti-inflammatory actions. Whether pharmacological AMPK activation suppresses SFA-induced inflammation in a human system is unclear. In a setting of hypoxia-potentiated inflammation induced by SFA palmitate, we found that the AMP-mimetic AMPK activator 5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside (AICAR) potently suppressed upregulation of ER stress marker mRNAs and pro-inflammatory cytokines. Furthermore, AICAR inhibited macrophage ER stress responses triggered by ER-stressors thapsigargin or tunicamycin. Surprisingly, AICAR acted independent of AMPK or AICAR conversion to 5-aminoimidazole-4-carboxamide-1-β-D-ribofuranosyl monophosphate (ZMP) while requiring intracellular uptake via the equilibrative nucleoside transporter (ENT) ENT1 or the concentrative nucleoside transporter (CNT) CNT3. AICAR did not affect the initiation of the ER stress response, but inhibited the expression of major ER stress transcriptional effectors. Furthermore, AICAR inhibited autophosphorylation of the ER stress sensor inositol-requiring enzyme 1α (IRE1α), while activating its endoribonuclease activity in vitro. Our results suggest that AMPK-independent inhibition of ER stress responses contributes to anti-inflammatory and anti-diabetic effects of AICAR.


AICAR inhibits hypoxia-enhanced palmitate-induced inflammation in human macrophages.
AMPK activation is considered to be anti-inflammatory 1 , but the ability of pharmacologic AMPK activators to block inflammatory responses is still unclear. We used an experimental system of hypoxia-enhanced palmitate-induced inflammation in primary human macrophages 8 to test distinct classes of AMPK activators for their effect on palmitate-induced ER stress and inflammatory markers. This experimental system reflects a pro-inflammatory, hypoxic milieu of hypertrophic adipose tissue in obesity. Analysing phosphorylation of AMPK at T172 as a marker of AMPK activation, we noticed that palmitate moderately activated AMPK, while hypoxia reduced this activity (Fig. 1A), suggesting that AMPK is not activated under palmitate/hypoxic exposure. As observed previously 8 , c-Jun phosphorylation, a readout of an inflammatory response, increased under hypoxia/ palmitate (Fig. 1A). We then chose drugs activating AMPK allosterically (A769662, salicylate), changing the adenylate energy charge (phenformin, R419), or mimicking AMP (AICAR) 24,25 at concentrations causing similar AMPK activation as followed by phosphorylation of the AMPK substrate acetyl-CoA carboxylase (ACC) (Fig. 1B). In parallel, we analysed the markers of ER stress, i.e. phospho-IRE1, and inflammation (phospho-cJun). Of all AMPK activators only AICAR consistently inhibited palmitate-induced IRE1 and cJun phosphorylation (Fig. 1B). mRNA expression of ER stress-responsive genes Grp78 and CHOP as well as inflammatory cytokines IL-1β, TNFα, and IL-6 confirmed that AICAR was the most potent inhibitor (Fig. 1C). Among other AMPK activators only phenformin inhibited ER stress responses and cytokines, whereas R419 and salicylate reduced only IL1β and IL6 mRNA expression. A769662 was without any effect at all. None of the AMPK activators inhibited the expression of hypoxia-sensitive GLUT1 mRNA, suggesting that signalling through hypoxia-inducible factor (HIF) remains intact. Analysis of IL1β , TNFα and IL-6 protein secretion after hypoxia/palmitate treatment revealed divergent effects of AMPK activators on different cytokines (Fig. 1D). Remarkably, only AICAR consistently inhibited the secretion of all three cytokines. Our data indicate that AMPK activators suppress SFAtriggered inflammatory responses to a variable degree, suggesting that AICAR may block palmitate-induced ER stress and inflammation through mechanisms unrelated to AMPK activation.
We went on to investigate mechanisms how AICAR inhibits SFA-induced inflammatory responses. Although previous studies suggested the involvement of Sirt1 and FAO in attenuating SFA-induced inflammation by AMPK 18,19 , the AICAR effect in our study was neither reversed by the Sirt1 inhibitor Ex527, nor by the FAO blocker etomoxir (Supplementary Figure S1A). Active AMPK also interferes with mechanistic target of rapamycin complex 1 (mTORC1), which may attenuate inflammatory cell responses 1 . However, treatment of macrophages with the mTORC1 inhibitor rapamycin did not influence the expression of inflammatory and ER stress markers after hypoxia/palmitate (Supplementary Figure S1A). Furthermore, AICAR did not affect triglyceride levels in palmitate-treated hypoxic macrophages (Supplementary Figure S1B), suggesting that changes in fatty acid metabolism are unlikely to explain AICAR effects.

AICAR inhibits ER stress responses in macrophages in AMPK-independent manner.
With no indication towards the previously described anti-inflammatory mechanisms of AMPK in our system, we followed the observation that only AICAR suppressed both, IRE1 phosphorylation and ER stress markers in palmitate-treated macrophages. Thus, we questioned whether AICAR acts as a general inhibitor of ER stress responses in macrophages. Therefore, we used the prototype ER stress inducers thapsigargin and tunicamycin in cells pre-treated with AICAR. AICAR largely attenuated induction of ER stress mRNA markers by thapsigargin and tunicamycin ( Fig. 2A).
ER stress responses are executed by three major branches initiated by the ER stress sensors PERK, IRE1, and ATF6. To evaluate whether AICAR inhibits all three branches we examined PERK activation by following phosphorylation of the eukaryotic initiation factor 2α (eIF2α ), a known PERK substrate. IRE1 activation was assessed by analysing pIRE1 and protein amounts of spliced X-box binding protein 1 (XBP1) as a readout of IRE1 activity. In addition, we assessed the ATF6 branch by following accumulation of cleaved ATF6 in nuclear lysates. AICAR blocked thapsigargin and tunicamycin-induced IRE1 phosphorylation, accumulation of spliced XBP1  (D) mRNA expression of ATF4 and ATF6α in macrophages pre-exposed for 1 h to AICAR and treated with thapsigargin or tunicamycin for 6 h. (E) Western analysis of macrophages pre-exposed for 1 h to 0.5 mM AICAR in the absence or presence of ABT-702 and treated with thapsigargin for 6 h. (F) mRNA expression of CHOP and Grp78 in macrophages pre-exposed for 1 h to indicated concentrations of AICAR in the absence or presence of ABT-702 and treated with thapsigargin or tunicamycin for 6 h. (G) Protein expression of AMPK 72 h post-transfection with control siRNA or AMPKα 1 siRNA. (H) mRNA expression of CHOP, Grp78 and AMPKα 1 in macrophages transfected with control siRNA or AMPKα 1 siRNA for 72 h prior to treatments with thapsigargin for 6 h with or without 1 h AICAR pre-exposure. *p < 0.05. Data represent mean values ± SE of at least three independent experiments. ns, non-specific band. Western Blot images are cropped scans either of the same or of the duplicate membranes probed with different antibodies.
Scientific RepoRts | 6:32111 | DOI: 10.1038/srep32111 (Fig. 2C). We also noticed that AICAR blocked ATF4 mRNA induction by ER stressors (Fig. 2D), providing a possible explanation for the lack of ATF4 protein accumulation. Furthermore, nuclear levels of cleaved ATF6α protein as well as ATF6α mRNA expression after ER stress were also diminished in AICAR-treated macrophages (Fig. 2C,D). Thus, all three branches of the ER stress response are inhibited by AICAR.
To further investigate mechanisms how AICAR inhibits the ER stress response we first questioned whether AICAR demands AMPK activity or relies on AICAR conversion to ZMP, a step critical for AICAR to become an AMPK activator. AICAR phosphorylation to ZMP is catalysed by adenosine kinase. Addition of a potent adenosine kinase inhibitor ABT-702 prevented AICAR-induced AMPK activation as seen by the lack of ACC phosphorylation in AICAR-treated cells (Fig. 2E). Furthermore, AICAR-triggered inhibition of mTOR, analysed by following the phosphorylation of ribosomal S6 protein, was abolished by ABT-702 (Fig. 2E). However, ABT-702 did not prevent the impact of AICAR on IRE1 phosphorylation or XBP1 splicing (Fig. 2E). In the presence of ABT-702, the inhibitory effect of AICAR on IRE1 phosphorylation and XBP1 splicing even appeared to be enhanced. Indeed, ABT-702 potentiated the ability of low AICAR concentrations to inhibit mRNA expression of ER stress markers (Fig. 2F). Collectively, these data suggested that AICAR acts upstream of its conversion to ZMP and independently of AMPK. Supporting these data, a knockdown of AMPK α 1 catalytic subunit ( Fig. 2G) did not prevent inhibition of CHOP and Grp78 gene expression by AICAR (Fig. 2H). Dissociation of AICAR inhibition of ER stress and inflammatory response from AMPK activation was also observed in concentration-dependence experiments in palmitate-treated hypoxic macrophages (Supplementary Figure S2).

Nucleoside transporters involved into AICAR trafficking into macrophages. Although we
observed pronounced effects of AICAR on ER stress responses in macrophages, these effects could not be reproduced in different cell lines, such as THP-1 monocytic cells or HEK293T cells (data not shown). We reasoned that differences in intracellular AICAR accumulation may explain cell type-specific effects of AICAR. Previous studies suggested that most of the AICAR trafficking into cells is mediated by dipyridamole-sensitive adenosine transporters of equilibrative nucleoside transporter (ENT) family (SLC29A1-4), but also concentrative nucleoside transporter (CNT) CNT3 (SLC28A3) can transport AICAR into cells 26,27 . Assessing mRNA levels of individual SLC28 and SLC29 family members in human primary macrophages revealed abundant expression of SLC29A3 and SCL28A3 mRNAs ( Fig. 3A; note the logarithmic scale of the graph), moderate amounts of SLC29A1 mRNA, whereas SLC29A2, SLC28A1, and SLC28A2 were marginally expressed. To question the roles of individual nucleoside transporters in AICAR -ER stress responses we inhibited SLC29A1/A2 with dipyridamole alone or in combination with silencing SLC28A3 using RNA interference. This was followed by the analysis of mRNA expression of ER stress markers CHOP and Grp78 in response to thapsigargin. Silencing efficiency was verified by western analysis of SLC28A3 protein (Fig. 3B). Combining dipyridamole treatment with SLC28A3 silencing abolished the ability of AICAR to inhibit ER stress responses, whereas single treatments were only partly effective (Fig. 3C). Furthermore, overexpression of SLC28A3 in HEK293T cells allowed inhibition of ER stress responses to thapsigargin or tunicamycin. This became evidenced by an attenuated CHOP and Grp78 expression, whereas AICAR showed limited efficacy in the absence of overexpressed SLC28A3 (Fig. 3D). In contrast, silencing of SLC29A3 did not influence the inhibition of ER stress marker mRNA expression by AICAR (data not shown). These data suggest that entry of AICAR into cells through SLC29A1 and SLC28A3 transporters is necessary to interfere with ER stress in macrophages.

Effects of AICAR on IRE1α activity in vitro.
With the information that AICAR prevented accumulation of phosphorylated IRE1 and spliced XBP1 proteins in ER-stressed macrophages we investigated the effects of AICAR on the IRE1-XBP1 branch of the UPR in more detail. Contrary to our expectations, analysis of XBP1 splicing using primers specific to spliced and unspliced XBP1 mRNA revealed that the ER stress-induced splicing activity of IRE1 was not affected by AICAR (Fig. 4A,B). However, we observed lower mRNA levels of spliced XBP1, suggesting that XBP1 mRNA expression was attenuated by AICAR. IRE1 endoribonuclease activity also participates in a process of regulated IRE1-dependent decay (RIDD) of select mRNAs. To evaluate the influence of AICAR on RIDD, we analysed the mRNA expression of a well-characterized RIDD target mRNA BLOC1S1. Indeed, ER stress reduced BLOC1S1 mRNA expression in macrophages, which was unaffected by AICAR (Fig. 4C). Finally, we analysed the interaction of AICAR with recombinant IRE1α in vitro. AICAR bound to the ATP-binding site of IRE1α as evidenced by the Lanthascreen assay (Fig. 4D), and inhibited the IRE1α autophosphorylation in a DELFIA assay (Fig. 4E). However, when assessing endoribonuclease activity of IRE1α , AICAR appeared to activate IRE1α (Fig. 4F), which is consistent with the properties of reported type I kinase inhibitors of IRE1α 28,29 . Thus, AICAR binds IRE1α and inhibits IRE1α kinase, but not its endoribonuclease activity in vitro. This is in agreement with the observations in cells.
Collectively, our observations indicate that AICAR, while not affecting initiation steps of the UPR, prevents accumulation of transcriptional effectors of the UPR, i.e., ATF4, sXBP1, and ATF6α both at mRNA and protein level. Remarkably, this effect is independent of AICAR to be converted to ZMP and to activate AMPK.

Discussion
The novel information of this study is the ability of AICAR to inhibit ER stress responses in an AMPKindependent manner. This may add to the potency of AICAR to prevent SFA-induced inflammation as compared to other AMPK activators, since ER stress is a key factor in metabolic overload-triggered inflammatory responses. Remarkably, other AMPK activators are either less potent or completely unable (A769662) to inhibit palmitate-induced inflammation. Although cumulative evidence from mouse genetic models of AMPK deficiency strongly supports anti-inflammatory roles of AMPK, it is still unclear, whether AMPK is similarly active in the human vs. mouse innate immune system. Furthermore, the phenotype of genetic AMPK deficiency is definitely influenced by compensatory responses to AMPK loss, whereas pharmacological AMPK activators are supplied to unperturbed cellular systems, making a comparison of the responses in these two settings difficult. Regardless of their impact on AMPK, several AMPK activators are known for AMPK-independent anti-inflammatory activities. This comprises mitochondrial complex I inhibition by biguanidines 21 , or targeting the NFκ B activation pathway by salicylate 30 . Although AICAR was known to inhibit inflammation in an AMPK-independent manner 22,23 , molecular mechanisms remained unclear. The novelty of our data implies that under conditions of metabolic overload, inhibition of ER stress responses may contribute to the spectrum of anti-inflammatory AICAR effects. It is likely that the ability of AICAR to interfere with other aspects of pro-inflammatory macrophage signalling, as reported by others 19,22,23 , generates multi-pronged effects, resulting in the profound inhibition of inflammatory cytokine expression as observed in Fig. 1C.
Interestingly, AICAR does not attenuate initiation of ER stress responses, but inhibits mRNA and protein expression of all major transcriptional ER stress effectors, i.e. ATF4, spliced XBP1, and ATF6. Although initial results indicated a more selective response to the IRE1 branch of the ER stress, detailed examination of the IRE1 cellular and in vitro activities revealed that AICAR may act as a type I kinase inhibitor of IRE1 28 , inhibiting IRE1 autophosphorylation and at the same time activating its RNase activity. The downstream signalling of IRE1 is still blocked by AICAR, since it apparently inhibits expression of total mRNA, and as a consequence, spliced XBP1. Although splicing activity of IRE1 is not inhibited by AICAR, its effect on IRE1 kinase activity may affect IRE1 interaction with JNK 15 and thus, contributes to attenuated JNK activity and subsequent inflammatory responses in AICAR-treated cells.
How AICAR inhibits ER stress-induced expression of ATF4, XBP1, and ATF6 remains unclear. To explore detailed mechanisms is complicated by fact that ER stress transcriptional effectors induce each other 31,32 . Furthermore, it is likely that AICAR also attenuates translation of ER stress effectors, as we observed that AICAR-treated macrophages contained reduced amounts of polysome-associated RNA (data not shown). Although AICAR may affect translation through AMPK-mediated inhibition of the mTORC1 pathway and phosphorylation of elongation factor 2 33 , persistence of AICAR inhibitory effects in the absence of its conversion to ZMP, despite blocking mTORC1 activity and elongation factor 2 phosphorylation (Fig. 2E and data not shown), suggests that other mechanisms of translational attenuation may operate. This is similar to our observations on AICAR-mediated inhibition of the peroxisome proliferator activated receptor γ translation in the absence of its conversion to ZMP 34 . Interestingly, mTORC1 inhibition induced either by rapamycin or AMPK activators is inferior to AICAR in alleviating palmitate-induced ER stress responses, arguing that the potency of AICAR is likely to represent a combination of transcriptional and translational effects. Our data also suggest that cellular levels of the CNT and ENT transporters as well as adenosine kinase activity critically determine the outcome of the AICAR effect. Corroborating previous findings that modulating CNT and ENT expression influences intracellular ZMP accumulation and hence the magnitude of ZMP-mediated intracellular responses 26 , our study is the first to show that the endogenous CNT3 expression controls AICAR actions in macrophages. Of note, ENTs and CNTs are important drug targets to modulate the responses to nucleoside-based chemotherapies 35 . Similarly, we suggest that strategies aimed at enhancing ENT1 and CNT3 expression may promote AICAR anti-inflammatory and anti-ER-stress effects. Second, adenosine kinase activity may also determine the outcome of AICAR effects, switching it from ZMP-dependent (predominantly AMPK-mediated) to ZMP-independent pathways. Remarkably, in some cell contexts, such as acute myeloid leukemia, AMPK activation actually promotes ER stress 36 . Thus, it is interesting to test how adenosine kinase inhibitors will interact with AICAR systemically in the context of inflammatory diseases. We suggest that this combination can be more effective to suppress ER stress and inflammation and ameliorate metabolic disease. Considering that combinations with adenosine kinase inhibitor may allow to lower the range of AICAR concentrations inhibiting ER stress towards levels observed in plasma of AICAR-treated volunteers 37 , our mechanistic finding may be thus translatable into human clinical setting.
Studies conform to the principles outlined in the Declaration of Helsinki and were approved by the ethics committee of the Faculty of Medicine at Goethe-University Frankfurt. The ethics committee waived the necessity of written informed consent when using the buffy coats from anonymized blood donors. siRNA transfection of macrophages. Silencing of AMPKα 1 and SLC28A3 in human primary macrophages was achieved using corresponding siGENOME SMARTpools (Thermo Fisher Scientific) at 50 nM and Hiperfect transfection reagent (Qiagen). Cells were treated 72 h post-transfection.
Quantitative PCR. Total RNA of primary human macrophages was isolated using PeqGold RNAPure kit (PeqLab) and transcribed using cDNA Synthesis kit (Fermentas). Quantitative PCR was performed with iQ SYBR green Supermix (Bio-Rad) using the CFX96 system from Bio-Rad. Primer sequences for quantitative PCR can be obtained upon request. Expression was normalized to β -microglobulin. XBP1 splicing analysis. To analyse XBP1 splicing 1 μ g of RNA was reverse-transcribed. The cDNA was PCR amplified using primers designed to encompass the splicing site 38 with DNA Taq polymerase (26 cycles at 94 °C 30 s, 60 °C 30 s, 72 °C 30 s). PCR products were resolved on 2% agarose gel, stained with ethidium bromide, and visualized using GenoSmart2 gel documentation system (VWR International).

IL-1β detection in supernatants.
Macrophage supernatants were analysed for IL-1β using a Cytometric Bead Array (CBA) (BD Biosciences) according to the manufacturer's recommendations. Samples were measured using a LSRII/Fortessa flow cytometer (BD Biosciences) and analysed by FCAP Array software V1.0 (Soft Flow).
Triglyceride determination. Cell pellets were lysed in PBS containing 1% Triton X-100. Determination of triglycerides in lysates was carried out using a commercial kit from Roche Diagnostics according to manufacturer instructions.
Statistical analysis. Data are presented as means ± SEM of at least three independent experiments.
Data were analysed by one-way ANOVA with Bonferroni post hoc means comparison using Prism software (GraphPad). Differences were considered statistically significant for P < 0.05.