MAPK-interacting kinase 2 (MNK2) regulates adipocyte metabolism independently of its catalytic activity

The mitogen-activated protein kinase (MAPK)-interacting kinases (MNKs) are serine/ threonine protein kinases that are activated by the ERK1/2 (extracellular regulated kinase) and p38 α / β MAPK pathways. The MNKs have previously been implicated in metabolic disease and shown to mediate diet-induced obesity. In particular, knockout of MNK2 in mice protects from the weight gain induced by a high-fat diet. These and other data suggest that MNK2 regulates the expansion of adipose tissue (AT), a stable, long-term energy reserve that plays an important role in regulating whole-body energy homeostasis. Using the well-established mouse 3T3-L1 in vitro model of adipogenesis, the role of the MNKs in adipocyte differentiation and lipid storage was investigated. Inhibition of MNK activity using speci ﬁ c inhibitors failed to impair adipogenesis or lipid accumulation, suggesting that MNK activity is not required for adipocyte differentiation and does not regulate lipid storage. However, small-interfering RNA (siRNA) knock-down of MNK2 did reduce lipid accumulation and regulated the levels of two major lipogenic transcriptional regulators, ChREBP (carbohydrate response element-binding protein) and LPIN1 (Lipin-1). These factors are responsible for controlling the expression of genes for proteins involved in de novo lipogenesis and triglyceride synthesis. The knock-down of MNK2 also increased the expression of hormone-sensitive lipase which catalyses the breakdown of triglyceride. These ﬁ ndings


Introduction
In animals, highly integrated systems have evolved to promote storage during periods of energy abundance and conversely to mobilise stores during periods of energy deprivation and demand [1]. Excessive nutrient intake relative to expenditure produces a metabolic state that promotes energy storage; in the context of adipose tissue (AT), this results in adipocyte hyperplasia and hypertrophy [2]; hyperplasia involves recruitment of stem cells and their differentiation to the adipocyte lineage (adipogenesis), whereas hypertrophy involves the expansion of existing adipocyte triglyceride (TAG) stores. Given AT is fundamental for regulating whole-body energy homeostasis, it is not surprising that dysregulation or perturbation of adipose function can cause disease (reviewed in [3]). Excess or dysregulated AT expansion leads to weight gain and largely underpins the development of many chronic disorders including cardiovascular disease, hypertension and type-2 diabetes. Understanding the pathways that regulate adipocyte hypertrophy and hyperplasia is, therefore, of paramount importance.
Established pre-adipocyte cell models have been indispensable for the study and identification of key transcription factors that regulate adipogenesis: 3T3-L1 pre-adipocytes are the most widely used model as they faithfully recapitulate the events that occur during in vivo adipogenesis [4]. Established protocols have been developed to initiate the differentiation of pre-adipocytes; inducers that activate the insulin [5], glucocorticoid and cyclic adenosine monophosphate (cAMP)-signalling pathways [4,6,7] initiate a phase of early transcription, which is followed by synchronous entry into the cell cycle and several rounds of mitosis. Upon exit from the cell cycle, cells lose their fibroblastic morphology, accumulate TAG and acquire the metabolic features and appearance of adipocytes [5,8]. TAG accumulation is closely correlated with an increased rate of de novo lipogenesis (DNL) and a co-ordinate rise in expression of the enzymes of fatty acid (FA) and TAG biosynthesis [9].
The program of adipogenesis involves the co-ordinated and temporal expression of several key transcriptional regulators, namely CCAAT/enhancer-binding protein (C/EBP) α, β and δ and peroxisome proliferatoractivated receptor-γ (PPARγ) [1]. PPARγ and C/EBPα function as the key transcriptional regulators of adipogenesis. These two factors induce the expression of each other [10], whilst binding sites for these proteins are strongly enriched in the vicinity of genes induced in adipogenesis and in FA and glucose metabolism [11].
We previously showed that mice in which the gene encoding mitogen-activated protein kinase (MAPK)-interacting kinase 2 (MNK2) had been knocked out (MNK2-KO) exhibit protection from high-fat diet (HFD) induced obesity [12]. MNK2-KO mice have fewer but larger adipocytes which do not increase further in size on a HFD. Metabolic phenotyping indicated there is no reduction in food intake to account for the decreased weight of MNK2-KO mice (unpublished data). MNK2 is the predominant isoform in mature adipocytes and considerably induced during adipogenesis [12,13]. Furthermore, inhibiting MNK activity substantially inhibits lipid accumulation and the expression of key adipogenic genes [12]. MNK2 is, therefore, hypothesised to regulate the development of new adipocytes or indeed metabolism.
In humans and mice, there are two MNK proteins, MNK1 and MNK2, which are encoded on separate genes, Mknk1 and Mknk2, respectively. The kinases exhibit similarity in their primary sequence (78% homology) and contain similar structural motifs: an N-terminal basic amino acid region that mediates localisation, a catalytic kinase domain; and an MAPK-binding domain in their carboxy-termini [14,15]. The MNKs are activated by phosphorylation, as catalysed by extracellular regulated kinase (ERK) or p38 MAPK (reviewed, [16]). MNK1 binds both ERK and p38 MAPK, which strongly activate its low basal activity [17][18][19]. In contrast, MNK2 binds strongly only to activated ERK and this is thought to be responsible for its constitutive high basal activity, which is not further enhanced by stimuli [20,21]. MNK1/2 are responsible for phosphorylating eukaryotic initiation factor 4E (eIF4E) [22] on its only phosphorylation site, Ser209 [23], which is the only in vivo validated substrate of the MNKs [16].
The aim of this study was to investigate the role of MNK2 in adipogenesis and adipocyte metabolism, and the mechanisms involved, to help explain the phenotype of MNK2-KO mice.

Differentiation of 3T3-L1 pre-adipocytes
For adipogenesis studies, 3T3-L1 pre-adipocytes were seeded at a density of 3.5 × 10 4 cells/cm 2 and differentiation was induced as described earlier [12] with minor modifications. Cells were grown for 2-days after they reached confluence ('day 0') at which point the medium was replaced with growth medium freshly supplemented with 350 nM insulin, 500 μM 3-isobutyl-1-methylxanthine, 0.5 μM dexamethasone and 2 μM rosiglitazone (differentiation medium). On day 3, the medium was replaced with growth medium supplemented with 350 nM insulin (maintenance medium). The maintenance medium was replenished every 3-days until the cells were utilised for experimentation or the end-point of differentiation was reached (day 9). When assaying the effect of MNK inhibitors on adipocyte differentiation, compounds were added at the time of cell seeding ( pretreatment) and supplemented at each change of medium, unless indicated otherwise. The MNK inhibitor eFT508 [24] was purchased from Jomar Life Research. MNK-I1 was kindly provided by Professor Jiang Tao School of Medicine and Pharmacy, Ocean University of China, 5 Yushan Road, Qingdao, China.

siRNA knock-down
SilencerSelect siRNA (ThermoFisher) were used to silence protein expression in 3T3-L1 cells (Table 1). siRNA was transfected using Lipofectamine RNAiMAX Transfection Reagent (Invitrogen; 13778150) in a 1 : 3 ratio, according to the manufacturer's instructions, except that reagent and siRNA dilutions were made in serum and antibiotic-free DMEM (Invitrogen; 11995-065), not Opti-MEM. Where more than one siRNA was directed against a single gene, the molar ratio of each siRNA was proportioned so that the total molar amount (25 pmol) remained the same. When assaying the effect of siRNA-mediated knock-down on adipocyte differentiation, cells were transfected 24 h after seeding, then re-transfected 3-days post-induction.

Oil red O (ORO) staining
A 0.35% (w/v) solution of ORO (Sigma-Aldrich) in isopropanol was prepared and used to stain differentiated adipocytes, according to the manufacturer's recommendations. Stained cells were imaged on a ZEISS Axio Vert A1 inverted microscope (ZEISS) using a 5X objective lens; six representative images were captured per well and quantified using a macro script in Image J (NIH). Following imaging, the amount of lipid staining ( per well) was assayed semi-quantitatively by eluting ORO in 100% isopropanol for 5 min and measuring the absorbance at 490 nm.

Metabolic assays
For all metabolic assays, 3T3-L1 cells were seeded in 96-well assay plates, transfected with or without siRNA as appropriate, and subjected to adipogenic differentiation, as described above.
Glucose uptake was measured using the Glucose Uptake-Glo Assay (Promega; J1341) and following the manufacturer's recommendations. Briefly, cells were starved of serum prior to the assay by incubation in serum-free DMEM (SF-DMEM) overnight. On the day of the assay (day 9), cells were treated with SF-DMEM without glucose (Invitrogen; 11966-025, with 1 mM sodium pyruvate), supplemented with insulin (0.01, 0.1, 1, 10 or 100 nM) for 1 h. Cells were then treated with 1 mM 2-deoxyglucose (2DG), diluted in PBS for 10 min at room temperature. Luminescence, corresponding to the amount of 2-DG uptake, was recorded using the GloMax Discoverer (Promega) with 0.3-1 s integration.
Lipolysis was measured using the Lipolysis Assay Kit (Abcam; ab185433). On day 9, cells were washed twice with Lipolysis Wash Buffer, then stimulated with or without 100 nM isoproterenol (in sterile water) for 3 h. The assay was conducted according to the manufacturer's recommendations. Absorbance at 570 nm was measured using the GloMax Discoverer (Promega) and glycerol concentration was determined using a standard curve.
Lactate levels were assayed using the L-Lactate Assay Kit (Abcam; ab65330) in medium collected from differentiating adipocytes on day 6. Medium samples were first diluted 1 : 1000 in assay buffer, then 2 μl of this dilution was assayed in duplicate in black polystyrene microplates (Corning; CLS3631). The assay was conducted according to the manufacturer's recommendations for a fluorometric assay. Fluorescence output (ex/em:520/ 580 nm) was measured using the GloMax Discoverer (Promega). L-Lactate concentration was determined using a standard curve and adjusted to the relevant dilution factor.

RNA isolation
RNA was routinely harvested from cells using TRI Reagent (Sigma; T9424) and cDNA synthesised using the QuantiNova Reverse Transcription Kit (Qiagen; 205413), according to the manufacturer's recommendations. Quantitative gene expression analysis was performed using TaqMan Fast Advanced MasterMix (Applied Biosystems) on a 7500 Fast Real Time PCR System (Applied Biosystems). RT-qPCR reactions were performed using pre-designed TaqMan assay probes (Table 2), whilst Mknk1, Mknk2 and Gapdh expression analysis was performed using Fast SYBR Green Master Mix (Applied Biosystems; 4385617) with gene-specific primers ( Table 3). All reactions were performed in triplicate. Relative gene expression was determined using 2 −ΔΔCt , with the geometric mean of Nono and Hprt used as the internal reference. Acacb Mm01204671_m1 Acly Mm01302282_m1 Cd36 Mm00432403_m1 Chrebpa Mm02342723_m1 Dgat1 Mm00515643_m1 Dgat2 Mm00499536_m1 Fasn Mm0062319_m1 Hprt Mm03024075_m1 Table 3. Primers used for RT-qPCR gene expression analyses

Immunoblot analyses
Cell lysates were heated at 95°C for 5 min in Laemmli Sample Buffer and then subjected to sodium-dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE; [25]) followed by electrophoretic transfer to 0.45 μm nitrocellulose membranes (Bio-Rad; 1620115). Membranes were blocked in PBS-0.05% Tween20 (PBST) containing 5% (w/v) skim milk powder for 60 min at room temperature. Membranes were probed with the primary antibody in PBST with 5% BSA (w/v) ( Table 4) overnight at 4°C. Membranes were then washed (3 × 5 min) in PBST and incubated with fluorescently tagged secondary antibody, diluted in PBST, for 1 h. Membranes were washed again in PBST (3 × 5 min) and fluorescent signals were visualised using the Odyssey Quantitative Imaging System (LI-COR, Lincoln, NE) and quantified using Image Studio software (v5.2.5).

Results and discussion
To study the individual contributions of MNK1 and MNK2 to adipocyte differentiation, siRNAs directed against Mknk1 (siMknk1 #1/#2) and Mknk2 (siMknk2 #1/#2) were employed. The siRNAs were assayed individually (#1/#2) or in combination (#1+#2) for their efficacy in knocking down MNK expression in 3T3-L1 pre-adipocytes. Both siRNAs directed against Mknk1 reduced the expression of the Mknk1 mRNA and MNK1 protein ( Figure 1A,B). Both Mknk2-directed siRNAs decreased expression of the Mknk2 mRNA, individually and in combination, as expected ( Figure 1C). Interestingly, knock-down of Mknk2 tended to increase the levels of the Mknk1 mRNA ( Figure 1A), hinting at possible cross-talk/feedback regulation. This may be a compensatory response to the loss of MNK2 (the main MNK species in this cell type); however, as there is yet no published information on the transcriptional control of MNK expression, it is not clear what mechanisms might be involved. All siRNAs appeared to be isoform-specific with the exception of siMknk1 #2, which partially reduced Mknk2 levels ( Figure 1B). siRNA for Gapdh did not affect MNK expression but did very markedly decrease levels of the Gapdh mRNA and protein ( Figure 1B,D), suggesting that the transfection/knock-down procedure was effective. As no suitable commercial MNK2 antibody is available [26,27], knock-down of MNK2 was assessed by measuring the phosphorylation of eIF4E ( p-eIF4E), a readout of MNK activity. Both Mknk2-directed siRNAs almost completely abolished p-eIF4E; with the residual phosphorylation being attributable to MNK1 since siMknk1 and siMknk2 used in combination completely abolished p-eIF4E ( Figure 1D). Mknk1-directed siRNAs did not markedly affect p-eIF4E levels ( Figure 1B) suggesting that MNK2 is the predominant MNK isoform in these cells.
In light of this, it was important to establish the relative activities of the MNK isoforms in 3T3-L1 preadipocytes. Serum-starved 3T3-L1 cells, silenced for MNK1, MNK2 and MNK1/2 expression by siRNA transfection, were treated with DMSO or the phorbol ester tetradecanoylphorbol acetate (TPA), which activates ERK by stimulating an upstream kinase ( protein kinase C), to monitor basal and induced levels of p-eIF4E, respectively. Under basal conditions, p-eIF4E was almost completely abolished in MNK2-, but not MNK1-knock-down cells ( Figure 1E,F). The residual P-eIF4E seen in siMknk2 cells was likely attributable to MNK1 since it was lost when siMknk1 and siMknk2 were combined ( Figure 1E,F). TPA stimulation of MNK2 knock-down cells elevated p-eIF4E beyond basal levels, indicating that MNK1 activity is significantly increased in response to TPA ( Figure 1E,F), consistent with the known properties of MNK1 [14,18]. However, TPA stimulation was unable to elevate p-eIF4E levels above basal in MNK1-knock-down cells, indicating that, as expected, the high basal activity of MNK2 cannot be further enhanced by stimuli ( Figure 1E,F). These data suggest that MNK2 is principally responsible for maintaining basal p-eIF4E levels, whilst MNK1 is exclusively responsible for inducible p-eIF4E in 3T3-L1 pre-adipocytes. These findings are consistent with the established properties of MNK1/2 [14,17,18].
Given that MNK2 is the major isoform in mouse AT and its expression is considerably induced during adipocyte differentiation [12], MNK2 was proposed to play a role in adipogenesis. To investigate this, MNK2 expression was knocked-down in 3T3-L1 cells using siMknk2 (#1+#2) which were subject to adipogenic induction. As expected, siMknk2 transfection decreased both Mknk2 mRNA and p-eIF4E levels, indicating the successful knock-down of MNK2 (Figure 2A,B). Notably, the expression of C/EBPβ, C/EBPδ, C/EBPα and PPARγ was not affected by the knock-down of MNK2, nor was the level of p-C/EBPβ (a pre-requisite for dimerisation and DNA binding), indicating that the DNA binding activity of C/EBPβ is not regulated by MNK2 ( Figure 2B). C/EBPα and PPARγ expression was normal, as were the levels of proteins encoded by downstream target mRNAs: PLIN1, ATGL and FABP4 (except for small decreases on day 3, which recovered at later times), suggesting that MNK2 does not regulate, and is, therefore, not required for adipogenesis itself ( Figure 2B).
Intriguingly, despite this, lipid accumulation was significantly reduced by the knock-down of MNK2, as quantified by ORO staining and absorbance of eluted ORO ( Figure 2C-E). Since the reduced lipid content could not be directly attributed to a defect in adipogenesis (adipocyte differentiation; Figure 2B), it was apparent that MNK2 may regulate aspects of lipid metabolism in adipocytes. In terms of metabolism, a decrease in lipid accumulation may be attributable to decreased lipogenesis (lipid synthesis), increased lipolysis (lipid breakdown) or a combination of both.
LPIN1 is induced during 3T3-L1 adipogenesis and is thought to be critical for adipocyte development in vivo and in vitro, as it is required for the expression of PPARγ and C/EBPα [28,29]. LPIN1 is also responsible for converting phosphatidate to diacylglycerol (DAG), a key step in the synthesis of TAG [30,31], and acts as a transcriptional activator of genes involved in lipid metabolism [28,32]. Interestingly, LPIN1 expression was data are mean ± SEM (n = 2); ordinary one-way ANOVA (*P < 0.05 **P < 0.01 ***P < 0.001).
reduced in MNK2 knock-down cells, most notably during terminal differentiation ( Figure 2B). The mRNA for LPIN1 is alternatively spliced, thereby producing two protein isoforms designated LPIN1α and LPIN1β. These isoforms are expressed differentially and exhibit distinct functions: LPIN1α is involved in adipocyte differentiation, whereas LPIN1β has a pronounced effect on the induction of lipogenic genes [32]. Given that LPIN1β is the predominant isoform during the later stages of adipogenesis [32], MNK2 may, therefore, promote the expression of LPIN1β either through regulation of alternative splicing or selective mRNA translation.
To examine the robustness and validity of the reduced lipid phenotype, 3T3-L1 cells were treated individually with either siMknk2 #1 or #2. As expected, siMknk2 #1 and #2 each decreased p-eIF4E levels, confirming the successful knock-down of MNK2 (Supplementary Figure S1A,B). The expression of C/EBPβ, C/EBPδ, C/EBPα and PPARγ was not affected by either siMknk2 #1 or #2, nor was the level of p-C/EBPβ (Supplementary Figure S1A,B), as would be expected from the above data. Furthermore, the expression of downstream proteins induced during terminal differentiation (ATGL, FABP4) was also unaffected, as expected. The expression of PLIN1 (a lipid droplet protein) was partially reduced by siMknk2 #1 on days 3 and 6 but not by siMknk2 #2; it is not clear why. Notably, LPIN1 expression was reduced by both siMknk2 #1 and #2 and each significantly reduced lipid accumulation, although the effect was greater with siMknk2 #2 (∼40%) than siMknk2 #1 (∼30%), as quantified by ORO staining (Supplementary Figure S1C,D); this may be due to the fact that siMknk2 #2 decreases Mknk2 expression more strongly than siMknk2 #1 ( Figure 1B). Nevertheless, these findings clearly implicate MNK2 in the control of adipocyte lipid metabolism, which may reflect effects on LPIN1 expression.
To understand which aspect(s) of lipid metabolism was specifically affected by the knock-down of MNK2 and identify pathways that might be subject to its regulation, the expression of genes associated with adipocyte metabolism were assessed by RT-qPCR in differentiating 3T3-L1 pre-adipocytes in which Mknk2 expression had been knocked down.
In DNL, ATP citrate lyase (encoded by Acly) catalyses the conversion of citrate to acetyl-CoA which is then converted by acetyl-CoA carboxylase 1 (Acaca) to malonyl-CoA, the essential precursor for FA synthesis, catalysed by fatty acid synthase (Fasn) (schematised in Supplementary Figure S2). Expression of Acly, Acaca and Fasn, the key enzymes of DNL, was significantly reduced by siMknk2 on days 6 and 9 ( Figure 3A-C). This suggests that the capacity for DNL was significantly reduced by MNK2 knock-down. It should be noted that similar reductions in mRNA gene expression were observed when each of the siRNAs against Mknk2 was used individually (see Supplementary Figure S3).
In the pathway of TAG synthesis, FAs are covalently linked to a glycerol backbone (Supplementary Figure S2); glycerol-3-phosphate (G3P) acyltransferase (GPAT) enzymes catalyse the addition of the first FA to G3P, while the second and third FAs are added by enzymes known as acyl-glycerophosphate acyltransferases (AGPATs) and diacylglycerol-acyltransferases (DGATs), respectively [33]. Dgat1, and particularly Dgat2, are highly expressed in AT and catalyse the final rate-limiting step of TAG synthesis [34,35]. The levels of Dgat1 and Dgat2 were significantly reduced by the knock-down of MNK2 on day 6 ( Figure 3D,E). Lpin1 levels were also significantly reduced by siMknk2 on day 6 ( Figure 3F). As alluded to previously, LPIN1 is also a transcriptional activator of lipogenic gene expression; reduced LPIN1 expression may impair DNL in two ways: a reduction in the availability of substrate for TAG synthesis and impaired transcription of genes for enzymes of lipogenesis.
Acetyl-CoA carboxylase 2 (Acacb) is expressed at high levels in AT [36] and is localised to mitochondria where it regulates FA oxidation [37], since its product, malonyl-CoA, inhibits the rate-limiting enzyme of FA oxidation, carnitine palmitoyl-transferase I. Acacb −/− mice have increased rates of β-oxidation because they are unable to produce sufficient malonyl-CoA and are thus protected from diet-induced obesity [38]. In MNK2 knock-down adipocytes, Acacb expression was significantly reduced on day 9 ( Figure 3G), suggesting the rate of FA oxidation may be increased, which could contribute to their reduced lipid content. This may have broader implications in other FA-oxidising tissues, such as the heart and muscle, and may be particularly shown (n = 4) (C) Day 9 differentiated adipocytes were stained with ORO to assess lipid accumulation; representative images are shown, scale bar 0.2 mm. (D) Total stained area was quantified from (C) or from the absorbance of eluted ORO at 490 nm (E); data are mean ± SEM (n = 2); unpaired t-test (*P < 0.05 **P < 0.01 ***P < 0.001).
relevant for understanding the in vivo MNK2-KO phenotype in which mice are protected from HFD-induced weight gain [12]. Among enzymes involved in lipolytic pathways, MNK2 knock-down significantly increased the levels of hormone-sensitive lipase (HSL) (Lipe), the key enzyme of DAG hydrolysis ( Figure 3H). Interestingly, the levels of adipocyte triglyceride lipase (ATGL) (Pnpla2) and monoacylglycerol lipase (Mgll), which catalyse the hydrolysis of TAG/DAG and MAG, respectively, were not significantly affected by the knock-down of MNK2 ( Figure 3I,J).
The lipid droplet surface is complexed with one or more members of the Perilipin family of proteins which help to stabilise the droplet and regulate lipolysis [39]; PLIN1 (Plin1) is regarded as the major lipid droplet protein and is central to the regulation of lipolysis in adipocytes. Although lipid content was reduced in MNK2 knock-down adipocytes, Plin1 levels were not significantly affected ( Figure 3K), suggesting the expression of PLIN1, and presumably its control of lipolysis at the lipid-droplet interface, are not regulated by MNK2.
Lipoprotein lipase (Lpl) is expressed in the endothelium and is responsible for the breakdown of TAG in circulating chylomicrons and lipoprotein particles [40]. Lpl levels were not significantly affected by the knockdown of MNK2, although they did trend lower ( Figure 3L). Furthermore, the level of the FA transport protein, Cd36, which is responsible for the uptake of exogenous FAs into adipocytes [41] was not significantly affected by MNK2 knock-down but also trended lower ( Figure 3M). This suggests the uptake of exogenous FAs is unlikely to be regulated by MNK2, which is consistent with the fact that MNK2-KO mice have normal circulating levels of FAs [12].
Lipid metabolism is in a constant state of flux, where during periods of lipolysis, a considerable proportion of FAs is re-esterified back into TAG [42]. This 'recycling' of FAs is thought to be important for regulating lipid homeostasis and FA release during fasting [43]. G3P is required for the re-esterification of FA, and indeed the synthesis of TAG from lipogenic pathways. However, under conditions where lipolysis is activated, the major source of G3P, glycolysis, is markedly reduced [43]. Glycerol kinase is also expressed at a very low level in AT, and so after complete hydrolysis of TAG, very little of the glycerol backbone can be phosphorylated back to G3P. A metabolic pathway, known as glyceroneogenesis, is required for the synthesis of G3P from noncarbohydrate substrates, such as pyruvate and lactate, when glucose levels are low. Phosphoenolpyruvate carboxykinase 1 (Pck1) is the major enzyme regulating glyceroneogenesis [44]. The knock-down of MNK2 significantly reduced the level of Pck1 ( Figure 3N), suggesting that the availability of G3P for TAG synthesis and FA recycling may be reduced.
Secreted frizzled-related protein 5 (Sfrp5) antagonises Wnt signalling and has been widely implicated in adipocyte biology [45]. Multiple studies report that Sfrp5 is highly induced with genetic or diet-induced obesity [46][47][48], although one study reported that Sfrp5 is actually suppressed under these conditions [49]. In MNK2 knock-down adipocytes, Sfrp5 levels were significantly reduced ( Figure 3O). Sfrp5 expression is one of the best a priori predictors of adiposity in mice prior to HFD exposure [46] which is consistent with the protection of MNK2-KO mice from HFD-induced weight gain [12].
Given the changes in metabolic gene expression largely affected genes in lipogenic pathways, it seemed possible that a transcription factor(s) responsible for the regulation of lipid metabolism was regulated by MNK2. Sterol regulatory element-binding protein 1 (Srebf1) enhances the transcriptional activity of PPARγ and thereby promotes the expression of genes involved in FA and lipid metabolism [50,51]. The level of Srebf1 was not, however, significantly affected by the knock-down of MNK2, although it did trend lower ( Figure 3P). Carbohydrate response element-binding protein (ChREBP) is a lipogenic transcription factor that is activated by carbohydrate feeding [52] and increases during adipocyte differentiation [53][54][55].
In the context of 3T3-L1 adipocytes, which are cultured in high-glucose medium, ChREBP is thus particularly relevant. In MNK2 knock-down adipocytes, on days 6 and 9, the level of ChREBP protein was significantly reduced as was Chrebpa mRNA (Figure 4A-C). Notably, three genes among our data set that are considered as targets for ChREBP [56,57], i.e. Acly, Acaca and Fasn were very substantially decreased by the knock-down of MNK2 ( Figure 3A-C), consistent with the lower levels of ChREBP. Notably, the levels of FAS were significantly decreased by the knock-down of MNK2 on day 9 ( Figure 4A,D). This suggests that MNK2 may regulate lipogenic gene expression by regulating the levels of ChREBP. Consistent with these findings, ChREBP depletion in precursor adipocytes has previously been demonstrated to reduce lipid accumulation and partially inhibit adipocyte differentiation [55]. LPIN1 expression was also significantly reduced in MNK2 knock-down adipocytes on days 3, 6 and 9 ( Figure 4A,E), despite Lpin1 mRNA being only significantly reduced on day 6 ( Figure 3F). The discrepancy between mRNA and protein levels suggests MNK2 may regulate LPIN1 expression, as mentioned earlier. HSL expression was significantly increased by the knock-down of MNK2 on days 6 and 9 ( Figure 4A,F), consistent with Lipe mRNA levels ( Figure 3H). The level of ATGL was not significantly affected by MNK2 knock-down whilst the abundance of PLIN1, which regulates lipolytic activity at the lipid droplet interface, was unchanged ( Figure 4A,G,H), consistent with the lack of a significant change in Plin1 mRNA levels ( Figure 3K).
The increased levels of HSL prompted us to test the rate of lipolysis under basal conditions following stimulation of differentiated 3T3-L1 cells by the adrenergic agonist isoproterenol. As expected, isoproterenol markedly increased the production of glycerol (generated by lipolysis) and this effect was greater in cells where MNK2 was knocked down ( Figure 4I). This suggests that knock-down of MNK2 moderately increases the rate of stimulated lipolysis perhaps due to elevated HSL expression. The rate of basal lipolysis was not significantly increased, consistent with the fact that access to the lipid droplet surface is blocked under these conditions by PLIN1 [58].
Because ChREBP is activated by glucose metabolites, its reduced expression could be explained by altered intracellular glucose levels. To assess whether glucose transport was regulated by MNK2, glucose uptake was assayed quantitatively in response to insulin dosage in MNK2 knock-down adipocytes. Unexpectedly, glucose uptake was significantly elevated by the knock-down of MNK2, even under basal conditions ( Figure 5A), suggesting that MNK2 may negatively regulate glucose uptake or insulin signalling more broadly. However, the expression of insulin-responsive glucose transporter GLUT4 (Slc2a4) at both the mRNA ( Figure 3Q) and protein levels ( Figure 5B-D) were not statistically different.
The increase in glucose uptake observed in MNK2 knock-down adipocytes was likely to have implications for glycolytic metabolism. Using phenol red as an indicator of the acidity of the medium, the rate of acidification in differentiating adipocytes in which MNK2 was knocked down was greater than in adipocytes with normal MNK2 levels (data not shown). This could reflect faster glycolysis in MNK2 knock-down adipocytes, since increased glycolytic flux would generate more L-lactate and thereby lower the pH of the growth medium more quickly (see scheme in Supplementary Figure S2). In support of this, L-lactate levels were higher in MNK2 knock-down adipocytes suggesting that these cells do indeed exhibit moderately increased glycolytic flux ( Figure 5F).
The nuclear translocation of ChREBPα is an important regulatory event that is necessary to induce ChREBPβ and a program of lipogenic and glycolytic gene transcription. The physical interaction of ChREBP with other proteins has been shown to regulate its nuclear translocation: for instance, over-expression of HSL impedes ChREBPα translocation and thereby reduces transcription of lipogenic genes [27]. In siMknk2 adipocytes, the reduction in ChREBP was associated with reduced lipogenic gene expression ( Figures 2B and 3A-F), increased β-adrenergic stimulated lipolysis ( Figure 4I) and consequently, reduced lipid accumulation ( Figure 2C-E). Given MNK2, like HSL, is cytoplasmic, it was possible that MNK2 may positively regulate the nuclear translocation of ChREBPα by physical interaction. However, we were consistently unable to detect an interaction between ChREBP and MNK2 (data not shown).
Given the metabolic effects described thus far are a result of siRNA-mediated knock-down of MNK2, it was important to assess whether inhibition of the kinase activity of MNK2 exerted similar effects. To assess this, 3T3-L1 pre-adipocytes were subjected to adipogenic differentiation in the presence or absence of the selective MNK1/2 inhibitor eFT508 [24], and the expression of key metabolic genes was assessed by RT-qPCR. ACTIN were monitored by Western blot at the indicated times as quantified in (C-H); representative immunoblots are shown (n = 3); one-way ANOVA (*P < 0.05 **P < 0.01 ***P < 0.001). (B) The mRNA level of Chrebpa was measured by RT-qPCR (n = 3); 2-way ANOVA (*P < 0.05 **P < 0.01 ***P < 0.001). (I) Glycerol levels were measured in medium from day 9 adipocytes treated with or without 100 nM isoproterenol (n = 3); one-way ANOVA (*P < 0.05 **P < 0.01 ***P < 0.001).
To assess the required dose of eFT508, which inhibits MNK1 and MNK2 [24], 3T3-L1 cells were treated with a range of concentrations of this compound, and p-eIF4E was monitored as a readout of MNK activity. eFT508 substantially blocked eIF4E phosphorylation even at 0.01 μM (Supplementary Figure S4). At a  concentration of 0.3 μM, eFT508 completely eliminated p-eIF4E and this concentration was selected for further studies. At this concentration, eFT508 did not affect any of the other signalling pathways tested (i.e. phosphorylation of ERK, PKB/Akt, or ribosomal protein S6 or eIF4E binding protein 1, 4E-BP1, both of which lie downstream of mTORC1), except for a minor impairment of ERK or PKB phosphorylation at the highest concentration tested (5 mM, i.e. >15× higher than that used in our studies here).
Importantly, in contrast with the effects of knock-down of MNK2, the levels of Acaca, Acacb, Acly, Fasn, Dgat1, Dgat2 and Sfrp5, were not significantly reduced by eFT508 (Supplementary Figure S5A-G); in fact, some mRNAs were, at some time-points, marginally or significantly increased. Furthermore, the level of Lipe was not significantly increased by eFT508, suggesting that the regulation of HSL expression is independent of the catalytic activity of MNK2 (Supplementary Figure S5L). The levels of key metabolic proteins were assessed in adipocytes differentiated in the presence of eFT508. As expected, and in contrast with the knock-down of MNK2, the expression of FAS, HSL, ATGL, LPIN1 and PLIN1 was not affected by eFT508 ( Figures 6A,B and 7A-F). eFT508 also had no effect on the expression of ChREBP protein or mRNA (Figures 6B and 7A,G). Furthermore, eFT508 had no effect on lipid accumulation as assessed by ORO staining (Figure 6C-E) or on basal or isoproterenol-induced lipolysis (glycerol production) ( Figure 7H). eFT508 did not significantly affect Srebf1, Slc2a4 or Cd36 mRNA levels (Supplementary Figure S5), consistent with the knock-down of MNK2, nor did eFT508 affect the expression of Mknk1 and Mknk2, as expected (Supplementary Figure S5P,Q). Plin1 levels were significantly increased by eFT508 on day 6, however, this discrepancy was unlikely to have any major physiological effect given that PLIN1 protein levels are more critical for their regulatory function than its mRNA. Therefore, given eFT508 did not elicit the same changes in gene expression as observed in MNK2 knock-down adipocytes, the observed metabolic changes associated with MNK2 knock-down are suggested to be mediated by a non-catalytic function(s) of MNK2.

Concluding comments
These studies were prompted by our striking observation that MNK2-KO mice are substantially protected against the weight gain that is usually induced by consuming a HFD [12]. Those findings suggested that MNK2, the major MNK in AT, might play a role either in the differentiation of adipocytes (adipogenesis) or in AT lipid metabolism.
Using the well-established 3T3-L1 cell model of adipocyte differentiation, we have explored the role of MNK2 in these processes. Our data clearly show that knocking down the expression of MNK2 did not affect adipogenesis as judged by the unaltered expression of markers of this process. However, knock-down of MNK2 did impair the accumulation of lipids (TAGs, derived in this system from DNL), and the expression of genes and enzymes involved in lipid synthesis/storage.
Our data point to lower expression of the transcription factor ChREBP, which drives expression of many genes that encode enzymes needed for DNL, in cells in which MNK2 has been knocked down. Its mRNA levels were also reduced. These data indicate that MNK2 positively modulates the expression of ChREBP, an important regulator of lipogenic genes, and thus lipid accumulation in adipocytes. Knock-down of MNK2 also decreased the levels of Lipin-1, another important regulator of lipid synthesis (via its effects on the transcription factor SREBP, see e.g. [59]).
An unexpected finding of our study is that these effects were not recapitulated by treating 3T3-L1 cells with a potent and specific MNK inhibitor, eFT508. This indicates that the consequences of knocking down MNK2 are not due to loss of its catalytic activity, but rather to loss of the MNK2 protein itself. Presumably, this is because MNK2 engages in, so far unknown, interactions with other proteins to promote lipogenesis. Interestingly, another target for p38 MAP kinase, the protein kinase MK2, also exerts effects that are independent of its catalytic activity [60].
Thus, our findings show that MNK2 plays a role in promoting lipid accumulation, but not adipocyte differentiation, and help to explain our earlier data [25] showing lower fat accumulation in MNK2-KO mice fed a HFD. Part 2 of 2 differentiated adipocytes were stained with ORO to assess lipid accumulation; representative images are shown, scale bar 0.2 mm. (D) Total stained area was quantified from (C) or from the absorbance of eluted ORO at 490 nm (E); data are mean ± SEM (n = 3); unpaired t-test (*P < 0.05 **P < 0.01 ***P < 0.001).