Saturated Fatty Acids Promote GDF15 Expression in Human Macrophages through the PERK/eIF2/CHOP Signaling Pathway

Growth differentiation factor-15 (GDF-15) and its receptor GFRAL are both involved in the development of obesity and insulin resistance. Plasmatic GDF-15 level increases with obesity and is positively associated with disease progression. Despite macrophages have been recently suggested as a key source of GDF-15 in obesity, little is known about the regulation of GDF-15 in these cells. In the present work, we sought for potential pathophysiological activators of GDF15 expression in human macrophages and identified saturated fatty acids (SFAs) as strong inducers of GDF15 expression and secretion. SFAs increase GDF15 expression through the induction of an ER stress and the activation of the PERK/eIF2/CHOP signaling pathway in both PMA-differentiated THP-1 cells and in primary monocyte-derived macrophages. The transcription factor CHOP directly binds to the GDF15 promoter region and regulates GDF15 expression. Unlike SFAs, unsaturated fatty acids do not promote GDF15 expression and rather inhibit both SFA-induced GDF15 expression and ER stress. These results suggest that free fatty acids may be involved in the control of GDF-15 and provide new molecular insights about how diet and lipid metabolism may regulate the development of obesity and T2D.


Cell Viability Assay
Cell proliferation reagent WST-1 (#05015944001, Roche Applied Science, Penzberg, Germany) was used to assess cell proliferation, viability and toxicity according to the manufacturer's instructions. 1.5 × 10 5 THP-1 cells were differentiated with PMA in 96-well plate followed by treatment for 16 h before addition of WST-1 reagent.

Luminex Assay
GDF-15 was measured in culture supernatants from MDMs by Luminex assay (R&D Systems, Minneapolis, USA) according to manufacturer's instructions. Beads were read on a Bio-Plex 200 system (Bio-Rad, Hercules, USA) or on a Bio-Plex MAGPIX system (Bio-Rad, Hercules, USA).

RT-qPCR Analysis
Total RNAs were extracted with high pure RNA isolation kit (Roche Applied Science, Penzberg, Germany) or TRIzol reagent (Ambion, Waltham, USA) according to the manufacturer's recommendations. DNase treatment was performed on column for RNA isolation kit or after resuspension of RNA pellet for TRIzol extraction by using Dnase I (#EN0521, Thermo Scientific, Waltham, USA). Purified RNAs were reverse-transcribed to complementary DNA (cDNA) by using the high capacity cDNA reverse transcription kit (#4368813, Applied Biosystems, Waltham, USA). qPCR was performed by using Brilliant II SYBR Green QPCR Master Mix (#600828, Agilent, Santa Clara, USA) and ran on a Mx3000P qPCR system (Agilent, Santa Clara, USA) or on a LightCycler 480 (Roche Applied Science, Penzberg, Germany). Gene expressions were calculated using the 2 -∆∆CT method. OAZ1, a highly stable gene in human [37], was chosen as housekeeping gene. Primers were designed with Primer-BLAST (National Center for Biotechnology Information (NCBI)) to amplify all the isoforms of the target gene, except for XBP1-s primers that were designed to only amplify the spliced isoform (isoform 2, NM_001079539). Primer sequences are provided in Supplementary Table S2.

Statistical Analyses
All statistical analyses were carried out using GraphPad Prism 7 for Windows (GraphPad Software, Inc., San Diego, USA) and presented as the means ± standard error of the mean (SEM). When one independent variable was involved, two-tailed Student's t-test was performed to compare two groups and one-way ANOVA with Dunnett's multiple comparisons test to compare more than two groups. Two-way ANOVA with Sidak's multiple comparisons test was performed when the experiment involved two independent variables. The statistical test used and the number of biological replicates (n) are described in each figure legend. Only independent biological replicates, MDM from different donors or independent cell line experiments, were used to draw graph. Biological replicates were obtained from a single measurement, such as western blot, or from the mean of multiple measurements (technical replicates), such as gene expression by RT-qPCR that was assessed in duplicate or cell viability assay in triplicate. Paired statistical tests were used. Statistical significance was set at p < 0.05.

ER Stress, but Not p53, Is Involved in SFA-Induced GDF15 Expression
GDF-15 was extensively studied for its role in cancer and shown to be mainly regulated by the transcription factor p53 [10] and the ER stress pathway [29]. SFAs did not induce common p53 target genes such as CDKN1A, encoding p21, nor MDM2 (Figure 2A,B), but strongly increased the expression of ER stress-related markers such as HSP5A, encoding BIP, and the spliced form of XBP1 mRNA in MDMs ( Figure 2C,D). Inhibition of p53 with pifithrin-α did not reverse C18:0-induced GDF15 expression in MDMs ( Figure 2E), while ER stress inhibition by the chemical chaperone 4-phenylbutyric acid (PBA) ( Figure S2A,B) decreased GDF15 expression ( Figure 2F) and secretion induced by C18:0 ( Figure 2G).
On the other hand, ER stress induction with tunicamycin, a glycosylation inhibitor leading to the accumulation of misfolded proteins in ER, increased ER stress marker ( Figure 2H) and GDF15 expression ( Figure 2I), confirming that ER stress is able to induce GDF15 expression in macrophages.

PERK/eIF2/CHOP Pathway Regulates SFA-Induced GDF15 Expression
To better understand how ER stress leads to GDF15 expression, we inhibited each of the three UPR branches by silencing IRE1α, ATF6 or PERK through siRNA transfection in PMA-differentiated THP-1 cells. The human monocytic cell line THP-1, differentiated in macrophage-like cells by using PMA, was used for its better transfection efficiency compared to primary macrophages. SFAs induced GDF15 expression and ER stress in PMA-differentiated THP-1 cells as observed in primary MDMs ( Figure S3A,B). While the expression of each ER stress sensor was significantly downregulated after transfection of the corresponding siRNA ( Figure S4A-C), only PERK silencing decreased C18:0-induced GDF15 expression ( Figure 3A and Figure S4D,E). IRE1α and ATF6 silencing did not decrease GDF15 expression but significantly reduced their downstream targets ( Figure S4F,G). Moreover, pharmacological inhibition of PERK with GSK2606414 or GSK2656157 reduced C18:0-induced GDF15 expression in PMA-differentiated THP-1 cells ( Figure 3B). PERK inhibition also decreased GDF15 expression and secretion following stimulation with C18:0 in primary MDMs ( Figure 3C,D). In addition to C18:0-induced GDF15 expression, tunicamycin-induced GDF15 expression was also prevented by PERK inhibitors (Figure 3E), demonstrating that the PERK pathway is involved in ER stress-induced GDF15 expression in macrophages. As a consequence of ER stress, PERK phosphorylates eIF2α that activates the ISR and increases the expression of ATF4, ATF3 and DDIT3, the latter encoding CHOP [33]. In agreement with our results showing that SFAs induce an ER stress response, SFAs increased eIF2α phosphorylation ( Figure S5A) and expression of ATF4, ATF3 and DDIT3 ( Figure S5B-D). Treatment with trans-ISRIB, an eIF2 complex inhibitor, decreased both C18:0-and tunicamycin-induced GDF15 expression ( Figure 3F). In basal conditions and under ER stress, eIF2α is dephosphorylated by PP1/CReP and PP1/GADD34 complexes driving a negative regulation of eIF2 activity. Interestingly and as anticipated, preventing dephosphorylation of eIF2α by inhibiting these phosphatase complexes with salubrinal exacerbated C18:0-and tunicamycin-induced GDF15 expression ( Figure 3G). Altogether, these results show that the eIF2 signaling pathway, namely the ISR, regulates GDF15 expression in macrophages.
ATF3 and CHOP are two key transcription factors involved in ISR which are both transcriptionally induced by SFAs ( Figure S5C,D). To restrain ATF3 and CHOP induction, we transfected siRNA for 8 h before treatment to prevent the increase of ATF3 or DDIT3 mRNA level following C18:0 treatment ( Figure S6A,B). CHOP silencing decreased GDF15 induction by C18:0 ( Figure 4A), while ATF3 knock down did not ( Figure 4B). Analysis of the GDF15 promoter region revealed seven potential CHOP response elements that we grouped in 4 sites ( Figure 4C). To determine whether CHOP is recruited to the GDF15 promoter in response to C18:0 treatment, we performed a ChIP-qPCR assay. BSA treatment alone did not enriched CHOP at any sites suggesting no or few basal bindings of CHOP to the GDF15 promoter ( Figure 4D), an observation corroborated by the undetectable CHOP protein levels under basal conditions ( Figure 4E). C18:0 increased CHOP protein levels ( Figure 4E) and CHOP binding to the GDF15 promoter ( Figure 4D). These results suggest that CHOP regulates GDF15 expression in macrophages by direct binding to its promoter. It is worth noting that all the previous treatments restraining GDF15 expression such as PBA, PERK siRNAs, PERK inhibitors and trans-ISRIB also decreased DDIT3 expression ( Figure 5A-E), while treatments exacerbating GDF15 expression, such as with the p53 inhibitor pifithrin-α or salubrinal, increased DDIT3 expression (Figure 5F,G). Finally, IRE1α and ATF6 silencing by siRNA, which did not modulate GDF15 expression, had no effect on DDIT3 expression ( Figure S7A,B). Collectively, these data demonstrate that C18:0 promotes GDF15 expression in macrophages following the induction of an ER stress through the PERK/eIF2/CHOP signaling pathway.

UFAs Inhibit SFA-Induced ER Stress and GDF15 Expression
Unlike SFAs, UFAs did not induce GDF15 expression nor secretion ( Figure 1B,C). It has been previously reported that UFAs can prevent SFA-induced cytokine production such as IL-1β or IL-6 [36,41,42]. To investigate whether UFAs prevent C18:0-induced GDF15 expression, PMA-differentiated THP-1 cells were treated with C18:0 alone or in combination with equimolar concentrations of UFAs. Co-treatments with C16:1, C18:1 and C18:2 prevented C18:0-induced GDF15 expression ( Figure 6A). Since CHOP is involved in SFA-induced GDF15 expression, the ER stress response upon co-treatment with C18:0 and UFAs was analyzed. UFA treatments restrained C18:0-induced ER stress markers including the splicing of XBP1 mRNA ( Figure 6B), eIF2α phosphorylation ( Figure 6C) and the expression of HSPA5 and DDIT3 ( Figure 6D,E). In contrast, tunicamycin-induced GDF15 expression ( Figure 6F) and ER stress ( Figure 6G-J) were not inhibited by UFAs co-treatment, suggesting that UFAs have no general effect on ER stress but specifically restrain the effect of SFAs on ER stress. The protection conferred by UFA treatment against C18:0-induced ER stress and CHOP expression likely account for the inhibition of GDF15 expression. Taken together, our results show the complex action of FFAs on GDF15 regulation in macrophages with SFAs promoting GDF-15 production and UFAs counteracting SFAs effect.

Discussion
The role of GDF-15 and its receptor GFRAL in the development and progression of obesity and T2D is well described [12][13][14][16][17][18][19][20], but the molecular mechanisms accounting for the high expression of GDF-15 in these conditions is poorly understood. Obesity increases GDF15 expression in liver and adipose tissue [28]. Parenchymal cells from these tissues (i.e., hepatocytes, preadipocytes and mature adipocytes) have been proposed as sources for the increased blood levels [22,23,28]. Macrophages are the main immune cell population in both adipose tissue and liver and expand with obesity [26]. Recently, it was shown that Gdf15 deletion in macrophages induces a more severe obesity and IR upon high fat diet feeding [15], suggesting that macrophages also participate to the obesity-associated increase of GDF-15.
To better understand how GDF15 expression may be regulated in macrophages, we investigated the effect of several physiological stimuli and identified SFAs as promoters of GDF15 expression and secretion. C18:0 (stearate) induced a stronger GDF15 expression and secretion than C16:0 (palmitate) at the same concentration ( Figure 1B,C). C18:0 is the second most important SFA, representing 12.5% of plasma FFAs, while C16:0 represents 28% [40]. Several studies reported a positive association for both C16:0 and C18:0 with obesity and T2D [43,44].
We showed that C18:0-induced GDF15 expression is ER stress-dependent and involves the UPR/ISR via the PERK/eIF2/CHOP signaling pathway and direct binding of the transcription factor CHOP to the GDF15 promoter. In basal conditions, macrophages express no or low levels of CHOP ( Figure 4E), likely explaining the absence of CHOP binding to the GDF15 promoter ( Figure 4D). Moreover, ER stress, UPR and ISR inhibitors have no effect in untreated cells, suggesting that basal GDF15 expression and secretion in macrophages is ER stress-independent. However, upon SFA treatment, the PERK/eIF2/CHOP signaling pathway is activated leading to CHOP expression and binding to the GDF15 promoter, GDF15 expression and secretion. The role of ER stress and ISR on GDF15 regulation was previously described in cancer cells [29], myocytes [13], gut epithelial cells [24] and hepatocytes [23]. These studies revealed that various ISR stimuli, including the non-steroidal anti-inflammatory compound sulindac sulfide [29], mitochondrial UPR [13], enteropathogenic Escherichia coli [24], tunicamycin [23,28] and thapsigargin [23,28], induce GDF-15. In addition to those different ER stress stimuli known to promote GDF15 expression, SFAs represent interesting pathophysiological stimuli linking obesity and GDF-15. Indeed, obesity and IR increase FFA levels and a disequilibrium in the SFA/UFA balance is commonly associated with progression of obesity and the development of complications such as T2D [43,44]. The transcription factor CHOP has also been involved in obesity development and CHOP-deficient mice are more susceptible to obesity [34], similarly to Gdf15-deficient mice [18]. The molecular mechanism accounting for the higher weight gain of CHOP-deficient mice is not elucidated but, given the role of CHOP in the regulation of GDF15 and the role of GDF-15 in the control of food intake, a decrease of GDF-15 level in these mice may play a role.
Previous studies have demonstrated that UFAs can prevent cytokines production by SFAs, such as IL-6 [42] and IL-1β [36,41]. We observed that equimolar concentrations of UFAs also prevent C18:0-induced GDF15 expression in macrophages ( Figure 6A). UFAs also suppress ER stress and CHOP expression, probably accounting for the inhibition of GDF15 expression. Inhibition of ER stress by UFAs is specific to SFAs as UFAs were unable to protect against tunicamycin-induced ER stress and GDF15 expression. These results reveal the complexity of FFA-mediated regulation of inflammatory process and provide evidences than the ratio SFAs/UFAs may be more important than the absolute value of SFAs.
GDF-15 is protective against obesity and T2D [12][13][14][16][17][18][19][20]. However, GDF-15 shows a positive correlation with obesity progression and the development of complications such as T2D [1][2][3][4]. This puzzling discrepancy may result from SFAs properties. In vitro, SFAs activate macrophages to produce pro-inflammatory cytokines such as TNF-α and IL-1β [36,45]. In obese adipose tissue, macrophages develop a pro-inflammatory phenotype in line with the presence of metabolic stimuli such as SFAs [46]. Since SFAs induce a pro-inflammatory phenotype in macrophages [45], GDF-15 may be produced besides other cytokines and be considered as an independent marker of macrophage activation by SFAs. The positive association between GDF-15 level and obesity progression may therefore result from the pro-inflammatory status of SFA-activated macrophages, given that pro-inflammatory macrophages are involved in the development of obesity complications [26]. In conclusion, we show that SFAs promote GDF15 expression and secretion in macrophages. SFA-induced GDF15 expression involves ER stress and the PERK/eIF2/CHOP signaling pathway. UFAs prevent SFA-induced GDF15 expression and ER stress. These new findings provide new molecular insights about how diet and lipid metabolism may regulate the development of obesity and T2D, but also suggest that GDF-15 may be a marker of the metabolic activation of macrophages during obesity.