Palmitoleic acid prevents palmitic acid-induced macrophage activation and consequent p38 MAPK-mediated skeletal muscle insulin resistance

Highlights • Palmitate-treated macrophage-conditioned medium causes myotube insulin resistance.• This involves activation of myotube p38 mitogen activated protein kinase.• Conditioned medium effects are mediated by tumour necrosis factor-α.• These effects are prevented by addition of palmitoleate.• Palmitoleate treatment of macrophages is insulin sensitising for myotubes.

It is as yet unclear whether FAs or other lipid derivatives are responsible for direct activation of macrophages or if macrophages are activated by signals from other tissues or cells, given the extensive cross-talk that occurs between adipocytes and infiltrating inflammatory cells (Lumeng et al., 2007a,b;Prieur et al., 2011).
Whereas some studies have demonstrated that macrophages accumulate in tissues other than adipose during obesity, including skeletal muscle ((Fink et al., 2013;Hevener et al., 2007;Nguyen et al., 2007); NAT, CWJ and MEC, unpublished data) little attention has been paid to the potential impact of this upon muscle insulin sensitivity. However, a recent study suggested that macrophages in muscle from obese individuals may also contribute to the increase in pro-inflammatory cytokine release and thus muscle IR (Varma et al., 2009). Furthermore, after incubation with the SFA palmitic acid, pro-inflammatory pathways are activated in macrophages, leading to IR in adipocytes through a paracrine loop involving tumour necrosis factor (TNF)a (Suganami et al., 2005), which has been repeatedly shown to induce IR in muscle cells in vitro and in vivo (de Alvaro et al., 2004;Hotamisligil et al., 1994;Liang et al., 2008;Plomgaard et al., 2005;Uysal et al., 1997). The effects of TNFa may be mediated through p38 MAPK, as inhibition or silencing of this kinase in vitro ameliorated TNFa-induced skeletal muscle IR (de Alvaro et al., 2004).
Recently, palmitic acid-treated macrophages were shown to generate conditioned medium (CM) that reduced glucose uptake and PI3K signalling and increased inflammatory signalling in GLUT4-overexpressing L6 myoblasts (Samokhvalov et al., 2008), an effect that was mediated through induction of protein kinase C (PKC) h and e isoforms (Kewalramani et al., 2011). In addition, CM from FA-treated macrophages caused IR in L6 myotubes that was TLR2/4-dependent . Conversely, CM from palmitic acid-treated myoblasts was capable of causing a proinflammatory switch in macrophage phenotype (Pillon et al., 2012). However, it is unclear whether UFAs might be able to alleviate these effects. Here, we aimed to further interrogate the mechanisms involved in the impairment of insulin sensitivity in differentiated skeletal muscle cells generated by CM derived from SFA-treated macrophages and to establish whether UFA treatment would alleviate these effects.

Cell culture
C2C12 myoblasts and J774 macrophages were cultured in DMEM containing 4.5 mM glucose, 10% foetal bovine serum (FBS) and 1% antibiotic anti-fungal (ABAF) mixture. Before study, differentiation of myoblasts into myotubes was achieved by switching to DMEM containing 2% horse serum for 5 days. Macrophages were treated with 200 ng/ml phorbol myristate acetate (PMA) for 3 days before use (Karten et al., 1999). Macrophage treatment medium was generated by coupling DMEM containing 10% FBS, 1% ABAF and 2% bovine serum albumin (BSA) with 0.75 mM palmitic acid (SFA), 0.75 mM palmitoleic acid (UFA, chosen because of its identical acyl chain length), a combination of both, or 10 ng/ml of lipopolysaccharide (LPS) as positive control. This was added to J774 cells for 8 h, before being aspirated and the cells washed in PBS x3. Absence of carry-over of FAs into the CM was confirmed by measurement using a kit (Wako Chemicals, Neuss, Germany). Fresh DMEM was then added for 16 h and the CM generated transferred to C2C12 myotubes for a further 16 h. Myotubes were then serum-starved for 2 h and selected wells stimulated with 100nM insulin (Novo Nordisk, Crawley, UK) prior to measurement of glycogen synthesis or lysis and western blotting.

Use of inhibitors
Where pharmacological inhibitors were used, myotubes were pre-treated for 1 h with 1 lM SB203580 and 0.1 lM BIRB796, 1 lM JNK V inhibitor, or vehicle (DMSO), before being treated with CM containing the same substances for 16 h. Where siRNA was employed, C2C12s were transfected with 50 nM nonsense or p38a MAPK-targeting siRNA pool using Dharmafect 3 (Dharmacon, Fisher Scientific, Loughborough, UK) on day 2 of differentiation and then left for 72 h before treatment with CM. Where TNFa blockade was undertaken, half of the CM for each treatment group contained 10 lg/ml of blocking antibody (eBioscience, Hatfield, UK), added before and during myotube incubation with control, palmitic acid or LPS-treated macrophage-CM for 16 h.

SDS-PAGE and immunoblotting
Myotubes were lysed in radioimmunoprecipitation assay (RIPA) buffer, homogenised using an Ultra-Turrax (IKA; Staufen, Germany) and denatured in Laemmli buffer for 10 min at 65°C. Proteins were resolved by SDS-PAGE, electro-transferred and immunoblotted as previously described (Patel et al., 2012). Specific bands were detected using chemoluminescence (Western Lightning Plus, Perkin Elmer) on Fuji Super RX film (Bedford, UK), scanned and quantified using Quantity One software (BioRad Laboratories, Hemel Hempstead, UK). Loading controls of corresponding total protein immunoreactivity or b-actin were utilised. All treatment groups were represented on each blot on which bands were quantified.

Real-time PCR analysis
Macrophages were collected in Trizol (Invitrogen, Paisley, UK) and homogenised using the Ultra-Turrax. Total RNA was extracted as per the manufacturer's instructions and resuspended in nuclease-free water. RNA concentration was determined using a Nanodrop 1000 (Wilmington, DE) and integrity confirmed by visualisation of rRNA bands after agarose gel electrophoresis. RNA preparations were DNAse-digested and cDNA was generated using an Omniscript kit (Qiagen, Manchester, UK). Real-time PCR analysis was performed using Fast Start SYBR Green reagent (Roche Diagnostics, Burgess Hill, UK) on an Opticon 2 detector (Bio-Rad Laboratories, Hemel Hempstead, UK). Reaction mixtures contained 20 ng of cDNA, 1.5 M each primer, 2.5 mM MgCl 2 , and were subjected to a 10 min hot start, followed by 37 cycles of 15 s at 95°C, 60 s at 55-62°C and 30 s at 72°C, with a final 5 min extension. Primer pairs (Invitrogen) were designed using the Primer3 program (http://bioinfo.ut.ee/primer3-0.4.0/). Sequences were TNFa: GTAGCCCACGTCGTAGCAA and GTGGGTG AGGAGCACGTAGT, MCP-1: ACCAGCCAACTCTCACTGAA and ACAG CTTCTTTGGGACACTT, nitric oxide synthase-2 (NOS2): TGACCTGA AAGAGAAAAGGA and TCCAGGGATTCTGGAACATT, 36B4: ACAGC TTCTTTGGGACACTT and ATCTGCTGCATCTGCTTG. The relative abundance of duplicate cDNA aliquots was quantified using a standard curve plotted from amplification of a 10-fold dilution series of DNA generated by conventional PCR from the same primer pairs and gel purified. Results are quoted after normalisation to expression of 36B4, which was unchanged by the treatments. Generation of a single appropriate PCR product was confirmed by melting curve analysis and periodic agarose gel electrophoresis.

Glycogen synthesis
Incorporation of glucose into glycogen by myotubes was measured as previously described (Cazzolli et al., 2001). C2C12 myotubes were serum starved and then 2 lCi D-[U-14 C]-glucose (NEN/ Perkin-Elmer; Waltham, MA, USA) was added per well ±100 lM insulin for one hour. The reaction was terminated by washing Â3 with ice cold PBS. Myotubes were lysed in RIPA buffer at 100°C for 10 min. Protein content was measured using a bicinchoninic acid assay (Pierce Biotechnology, Rockford, IL). Glycogen was precipitated in ethanol at 4°C overnight, then at À20°C for 1 h, before centrifugation at 13,000g for 20 min. The pellet was dissolved in water at 60°C, mixed with scintillant (Ultima Gold, Perkin-Elmer) and counted using a LS6500 beta counter (Beckman Coulter, UK). Results were calculated as pmol/min/mg of protein.

ELISA for secreted peptides
CM was centrifuged at 13,000g for 10 min and the cell-free medium aliquoted into collection tubes and stored at À80°C until analysis. Quantikine colormetrix sandwich ELISAs (R&D Systems, Minneapolis, MN, USA) were used to measure levels of TNFa, IL1b, Chemokine (C-X-C motif) ligand 2 (CXCL2) and monocyte chemoattractant protein (MCP)1. Samples were diluted where necessary and the procedure carried out according to the manufacturer's instructions. Absorbance was measured on a microplate reader (Tecan, Reading, UK) at 450 nm and corrected by the absorbance at 540 nm.

Arginase assay
J774 cells were treated with FA-containing medium for 16 h overnight and then washed in PBS before being collected and lysed in 150 ll buffer containing 0.1% Triton X-100/10 mM MnCl 2 / 25 mM Tris-HCl. Lysates were frozen overnight, defrosted and arginase activity was measured as previously described (Corraliza et al., 1994). 50 ll of lysate was incubated with 50 ll of 0.5 M L-arginine (pH 9.7) for 120 min at 37°C. The reaction was stopped by the addition of H 2 SO 4 /H 3 PO 4 /H 2 O (1:3:7) and then 25 ll 9% a-Isonitrosopropiophenone in ethanol was added and the mixture heated to 100°C for 45 min. Samples were then cooled on ice for 30 min before measurement of absorbance at 550 nm.

Griess assay
Inducible nitric oxide synthase (iNOS) activity was assessed by measurement of NO release into the medium. J774 cells were activated with PMA for 3 days and then incubated with phenol redfree DMEM containing BSA-coupled palmitic acid, palmitoleic acid, a combination of both, LPS or vehicle for 16 h. Samples of medium were collected at the end of this experiment, aliquoted into collection tubes and stored at À20°C before analysis. The Griess assay was performed as a 96 well microplate assay (Molecular Probes, Life Technologies, UK) according to the manufacturer's instructions and absorbance was read on the plate reader at 548 nm.

Statistics
Statistical analysis of data was performed using GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA). Student's t-test, oneway and two-way ANOVAs were utilised, accompanied as appropriate by post hoc testing using Fisher's least significant difference test. Statistical significance was accepted at p < 0.05.

Treatment of macrophages with palmitoleic acid generates conditioned medium that insulin sensitises myotubes
After macrophages were simultaneously incubated with both palmitoleic acid and palmitic acid, the generated CM (comb-mac-CM) did not cause a significant impairment in insulin-stimulated glycogen synthesis in myotubes, as the palm-mac-CM-induced deficit was reduced by 66% (76% higher than palm-mac-CM, p = 0.0139; Fig. 1A). Furthermore, CM generated from macrophages treated with palmitoleic acid alone (palmitoleic-mac-CM) increased both basal and insulin-stimulated glycogen synthesis versus the equivalent controls (by 92% and 40%, p = 0.008 and p = 0.0054 respectively). Similarly, the impairments in insulinstimulated phosphorylation of PI3K signalling pathway intermediates were largely abrogated when palmitoleate was added to the palmitate-BSA used to treat macrophages. Myotubes incubated with comb-mac-CM showed a 90% increase in pY612-IRS1 compared to palm-mac-CM (p = 0.0017), thus restoring phosphorylation to the level of insulin-treated control CM-incubated myotubes (Fig. 1B). Phosphorylation of Akt, GSK3b and AS160 was also variably restored in myotubes (p = 0.0026-0.060; Fig. 1C-E). In addition, consistent with the elevated glycogen synthesis,  Palmitic acid-treated macrophage-conditioned medium impairs glycogen synthesis and insulin signalling in C2C12 myotubes, while these defects are rescued by palmitoleic acid. C2C12 myotubes were incubated with conditioned medium derived from macrophages treated with LPS, palmitic acid, palmitoleic acid, a combination of the two or vehicle (control group) for 16 h, before being serum starved and (A) incubated with D-[U-14 C]-glucose tracer ± 100nM insulin to measure glycogen synthesis, or collected in RIPA buffer and SDS-PAGE and immunoblotting conducted to assess phosphorylation and total protein levels of (B) IRS1 (pY612), (C) Akt (pS473), (D) GSK3b (pS9) and (E) AS160 (pT642). Representative blots and summary data (mean ± SEM of 3-6 individual experiments) are shown. All treatment groups were represented on each blot on which bands were quantified, but basal and insulin-stimulated samples are shown separately here for clarity. Total protein levels of all intermediates and b-actin were unchanged by any treatment. Selected post hoc significance is shown to simplify interpretation: ## p < 0.01, ### p > 0.001 and #### p < 0.0001 vs. basal control; * p < 0.05, ** p < 0.01, *** p < 0.001 and **** p < 0.0001 vs. control, insulin treated cells; £ p< 0.05, ££ p < 0.01 and ££££ p < 0.0001 versus palmitate treated, insulin treated cells. C -Control (+ insulin treatment); P -Palmitic acid (+ insulin treatment); L -LPS (+ insulin treatment); PO -Palmitic and palmitoleic acid (+ insulin treatment); O -Palmitoleic acid (+ insulin treatment). palmitoleic-mac-CM increased pY612-IRS1 (by 43%, p = 0.0062; Fig. 1B), pS9-GSK3b (by 39%, p = 0.0007; Fig. 1D) and pT642 (by 64%, p = 0.0006; Fig. 1E) versus insulin-stimulated control-mac-CM-treated myotubes, while this effect was not detected with respect to pS473-Akt. Total protein levels of each intermediate were again unchanged. Thus, palmitoleic acid treatment of macrophages results in CM that insulin sensitises myotubes and can rescue the defect in glycogen synthesis caused by palm-mac-CM treatment of myotubes.
3.3. Palmitic acid and palmitoleic acid-treated macrophageconditioned media have contrasting effects on inflammatory signalling in C2C12 myotubes To determine whether palm-mac-CM might activate inflammatory/stress signalling pathways in myotubes, immunoblotting was performed initially for phosphorylated and total MAPKs and Inhibitor jBa (IjBa), which causes cytoplasmic retention of NFjB.
Although most of these pathways were activated by palm-mac-CM and LPS-mac-CM, interestingly these effects were only apparent under insulin-stimulated conditions. Both LPS-and palmmac-CM caused increases in pY182-p38 MAPK (p = 0.0003 and p = 0.0024; Fig. 2A), accompanied by increases in pY185/pT183-p46 JNK (p = 0.0034 and p = 0.010) and pY185/T183-p54 JNK (p = 0.0001 and p = 0.0010; Fig. 2B&C), indicative of greater activation of each kinase. A similar trend was observed for ERK phosphorylation after palm-mac-CM but not LPS-mac-CM treatment (Fig. 2D). Total protein levels of each were unaltered. Insulin treatment increased IjBa (by 512%, p = 0.0010) but palm-mac-CM especially reversed this effect (58% decrease, p = 0.0041; Fig. 2E), implying activation of the NFjB pathway. Thus multiple proinflammatory pathways are activated in myotubes and may be responsible for the palm-and LPS-mac-CM-induced IR. Strikingly, phosphorylation of p38 MAPK, JNK1 and JNK2 were all normalised by palmitoleic-mac-CM and comb-mac-CM ( Fig. 2A-C). In contrast, all FA-palm-CMs tended to similarly activate ERK1/2 (not significant by ANOVA; Fig. 2D), suggesting that this MAPK does not mediate the differential effects of SFA and UFA treatment of macrophages on myotube insulin sensitivity. In addition, combmac-CM reduced degradation of IjBa compared to palm-mac-CM treatment (91% increase, p = 0.049; Fig. 2E). Thus palmitoleic acid treatment of macrophages results in CM that does not generate an inflammatory response involving the MAPK and NFjB signalling pathways in myotubes, in contrast to the effects of palm-mac-CM alone. Furthermore, abrogation of the palm-mac-CM-induced MAPK activation might underpin the insulin-sensitising effects of palmitoleic acid.

Macrophage-conditioned medium-induced defects in glycogen synthesis and insulin signalling are partially restored by pharmacological MAPK inhibition
As previous findings have suggested that activation of the NFjB pathway did not contribute to the generation of IR in muscle in vivo (Hommelberg et al., 2011, Polkinghorne et al., 2008, next we aimed to determine whether pharmacological inhibition of the two MAPKs that were activated by SFA and LPS treatment of macrophages could restore the impaired myotube glucose disposal. Glycogen synthesis was measured with or without pre-incubation with the p38 MAPK inhibitors SB203580 (1 lM) and BIRB796 (0.1 lM) (Bain et al., 2007;Kuma et al., 2005) or JNK V inhibitor (1 lM) (Bain et al., 2007). Both palm-and LPS-mac-CM induced increases in pY182-p38 MAPK, pT183/pY185-p46 and p54 JNK, although in the case of LPS-mac-CM, these increases did not reach significance by post hoc testing on this occasion, perhaps due to variation in batches of LPS used. Nevertheless, these trends were all abolished by the corresponding inhibitor treatment (Fig. 3A-C), confirming the efficacy of the inhibition. Each MAPK inhibitor alone showed similar but smaller effects when used alone (data not shown).
The effects of MAPK and JNK inhibition on myotube glycogen synthesis and signalling were investigated next. Inhibition of neither of these kinases had effects in non-insulin treated cells, as would be expected based on the data presented in Fig. 2 (data not shown). However, pre-incubation with SB203580 and BIRB796 together had dual effects on insulin-stimulated glycogen synthesis. The 45% reduction in insulin-stimulated glycogen synthesis in palm-mac-CM-treated cells (p = 0.0038) was nearly abolished by inhibitor treatment, with a lesser effect in LPS-mac-CM-treated myotubes (Fig. 4A). This was apparently contributed to by both ameliorations of the effects of the mac-CM treatments and an inhibitor-induced reduction in control-mac-CM insulin-stimulated glycogen synthesis. Similarly, p38 MAPK inhibition caused a 53% increase in insulin-stimulated pY612-IRS1 versus that in palmmac-CM-treated myotubes (p = 0.0183; Fig. 4B), substantially restoring the defect. This was accompanied by partial restorations of pS473-Akt (increased by 46%, p = 0.025), pS9-GSK3b (increased by 51%, p = 0.0028) and pT642-AS160 (increased by 48%, p = 0.032) (Fig. 4C-E), with p38 MAPK inhibition also limiting the effects of insulin on phosphorylation of PI3K signalling intermediates in control cells. Interestingly, the LPS-mac-CM-induced defects in insulin signalling were not ameliorated by p38 MAPK inhibition (Fig. 4B-E). Indeed, there was a further 44% reduction in pT642-AS160 compared to LPS-mac-CM treated myotubes (p = 0.044; Fig. 4B).
JNK inhibition appeared to partially restore the palm-mac-CMinduced defects in PI3K signalling in insulin-stimulated cells (Fig. 4B-E), although these effects were of lower magnitude than that induced by p38 MAPK inhibition and did not appear to be reflected in improved glycogen synthesis (Fig. 4A). However, inhibition of JNK did cause a 41% increase in pS9-GSK3b versus palmmac-CM (p = 0.013; Fig. 4D). Instead, JNK inhibition was more effective in restoring the defects induced in glycogen synthesis and phosphorylation of signalling intermediates caused by LPSmac-CM, in particular with regard to pS473-Akt (54% increase over LPS-mac-CM alone; p = 0.0067; Fig. 4C), which may be the result of a different secretory profile of macrophages induced by LPS versus palmitic acid. Nevertheless, it appears that JNK is less important than p38 MAPK in the palm-CM-induced defect in myotube glycogen synthesis.
3.5. The palmitic acid-treated macrophage CM-induced defect in myotube PI3K signalling is also partially restored by siRNA-induced silencing of p38a As pharmacological inhibition of p38 MAPK had more striking effects than that of JNK and these effects were more obvious with regard to insulin signalling than glycogen synthesis, we sought to corroborate this finding by investigating the effects of siRNA-mediated silencing of the principal pro-inflammatory p38 MAPK isoform in skeletal muscle (p38a) (Kuma et al., 2005) on myotube insulin signalling. A 74% reduction in p38a protein was obtained (p < 0.0001; Fig. 5A) and this was sufficient to reduce collective phosphorylation of all the p38 MAPK isoforms. Specifically, palmmac-CM had an additive effect with insulin (Antonescu et al., 2005) to increase pY182-p38 MAPK, but both these increases were abolished by siRNA. A 34% decrease in total pY182-p38 MAPK occurred (p = 0.0007), reducing this to the level measured in control myotubes (Fig. 5B), a trend that was replicated in LPS and control-CM-incubated myotubes, again demonstrating a dual effect of manipulating p38 MAPK on insulin signalling. Similarly, total p38 MAPK protein was reduced by a mean of 45% across all groups (ANOVA p = 0.0003; Fig. 5B), reflecting residual expression of other MAPK isoforms.
The siRNA-mediated reduction in p38 MAPKa ameliorated most of the observed signalling defects to varying extents, as indicated by the abolition of the significant differences in phosphorylation between insulin-treated control and palm-or LPS-mac-CM-treated myotubes ( Fig. 6A-D). Most clearly, there was a significant increase in phosphorylation of Akt after palm-mac-CM treatment (58%; p = 0.034; Fig. 6B) and a similar trend for GSK3b phosphorylation (35% increase, p = 0.10; Fig. 6C), confirming that this was not purely due to an inhibition of control insulin-stimulated phosphorylation, although this likely contributes, as with the effects of p54 JNK (normalised to control) p46 JNK (normalised to control) pERK1/2 (normalised to control) Fig. 2. Palmitic acid and palmitoleic acid-treated macrophage-conditioned media have contrasting effects on inflammatory signalling in C2C12 myotubes. C2C12 myotubes were incubated with conditioned medium derived from macrophages treated with LPS, palmitic acid, palmitoleic acid, a combination of the two or vehicle (control group) for 16 h, before being serum starved and lysed in RIPA buffer, and SDS-PAGE and immunoblotting conducted to assess phosphorylation and total protein levels of (A) p38 MAPK (pY182), (B) ERK1/2 (pT202), (C) p46 JNK (pT183/pY185), (D) p54 JNK (pT183/pY185) and E) IjBa. Densitometry was performed and sample blots and summary graphs are shown, representing the mean ± SEM of 3-4 individual experiments. All treatment groups were represented on each blot on which bands were quantified, but basal and insulin-stimulated samples are shown separately here for clarity. Total protein levels of MAPKs and b-Actin were unchanged by the treatments. Selected post hoc significance is shown to simplify interpretation: * p < 0.05, ** p < 0.01, *** p < 0.001 vs. insulin-treated control; ### p < 0.001 vs. basal control; £ p < 0.05, and ££££ p < 0.0001 versus palmitate treated, insulin treated cells. C -Control (+ insulin treatment); P -Palmitic acid (+ insulin treatment); L -LPS (+ insulin treatment); PO -Palmitic and palmitoleic acid (+ insulin treatment); O -Palmitoleic acid (+ insulin treatment). pharmacological inhibition. In particular, this could provide an explanation for the changes observed in AS160 phosphorylation (Fig. 6D), or indeed this could be the result of effects upon activity of an alternative upstream kinase, such as AMP-activated protein kinase (AMPK) (Thong et al., 2007). However, the data are broadly consistent with those generated using p38 MAPK inhibitors, confirming a role for p38 MAPK in the mechanism.

Macrophages treated with palmitic acid show a pro-inflammatory polarisation that is abrogated by co-incubation with palmitoleic acid
We next investigated the effects of each FA treatment on macrophage polarisation and secretion of potential mediators of the effects observed in the myotubes. Macrophage activation is classically defined as either M1 (pro-inflammatory) or M2 (anti-inflammatory), according to cellular iNOS and arginase activities (Lumeng et al., 2007a,b), therefore NOS2 expression in, NO release by and arginase activity of treated macrophages were measured. Macrophages treated with LPS demonstrated robust respective 44 and 5.6-fold increases in NOS2 and NO release over control (p < 0.0001 and p = 0.0002; Fig. 7A and B). However, in addition, palmitic acid-treated macrophages also showed an 18-fold increase in NOS2 mRNA (p = 0.0074) and produced 164% more NO than control macrophages (p = 0.012), while palmitoleic acid and the combined treatment did not affect NOS2 mRNA or NO release. Palmitic acid-treated macrophages also produced 44% less urea per mg of protein than control macrophages (p = 0.034), although the reduction caused by LPS treatment was not significant (Fig. 7C). In contrast, palmitoleic acid or palmitic/palmitoleic acid-treated cells produced a similar amount of urea to control cells. Thus palmitic acid-treated macrophages demonstrate a classical pro-inflammatory M1 phenotype on the basis of these two assays, while this shift is prevented in the presence of palmitoleic acid.
M1-activated macrophages secrete a range of pro-inflammatory cytokines/chemokines that activate inflammatory pathways in adjacent cells, therefore the levels of 2 pro-inflammatory cytokines and 2 pro-inflammatory chemokines that could be implicated in muscle IR (De Paepe et al., 2012;Hotamisligil et al., 1993;Sell et al., 2006;Tardif et al., 2011) were measured in mac-CM. LPS treatment of macrophages caused the expected substantial increases in production of TNFa, MCP1 and CXCL2 (all p < 0.0001; Fig. 7D-H). However, there were also respective 308 and 296% increases in TNFa macrophage mRNA and peptide in palm-mac-CM versus control (p = 0.0101 and p = 0.001; Fig. 7D and E), changes that were normalised by concurrent palmitoleic acid incubation. MCP1 mRNA was increased by 213% in macrophages (p = 0.0074; Fig. 7F) but the increase in peptide concentration in palm-mac-CM did not reach significance (79% increase, p = 0.15; Fig. 7G), and this was not substantially affected by addition of palmitoleic acid. No effects on CXCL2 secretion were noted (Fig. 7H), while IL1b levels were below the level of detection for treatment groups other than LPS-mac-CM (data not shown). This suggests that neither CXCL2 nor IL1b are likely to be responsible for the effects of CM in myotubes, while the evidence for a role of MCP1 is equivocal.

Addition of a TNFa blocking antibody leads to partial restoration of the palmitic acid-treated macrophage CM-induced defect in myotube insulin signalling
As palm-mac-CM contained elevated levels of TNFa and this increase was prevented by co-incubation with palmitoleic acid, we aimed to establish whether TNFa might be responsible for the induction of myotube IR by palm-mac-CM. Preliminary experiments confirmed that 2.5-100 ng/ml recombinant TNFa decreased pS473-Akt in myotubes, confirming that TNFa is capable of having this effect, while the TNFa blocking antibody did not affect Akt phosphorylation. In addition, 1 ng/ml recombinant TNFa * p < 0.05, ** p < 0.01 and *** p < 0.001 vs. no inhibitor control; £££ p < 0.001 vs. no inhibitor palmitic acid; $ p < 0.05 and $$ p < 0.01 vs. no inhibitor LPS. C -control conditioned medium (+ insulin treatment); P -palmitic acid-treated macrophage-conditioned medium (+ insulin treatment); L -LPS-treated macrophage-conditioned medium (+ insulin treatment); p38 -SB205380 and BIRB796 inhibitors used; JNK -JNK V inhibitor used.
Insulin caused the expected increases in phosphorylation of all signalling intermediates in control-mac-CM myotubes (all p < 0.0001), while the addition of blocking antibody had no significant effect, although there did tend to be small reductions in each ( Fig. 8A-D). The expected reductions in phosphorylation of signalling intermediates caused by palm-mac-CM and LPS-mac-CM (p = 0.012 to p < 0.0001; Fig. 8A-D) were attenuated in the presence of blocking antibody. There were 30-62% increases in pY612-IRS-1, pS473-Akt, pS9-GSK3b and pT642-AS160 (range p = 0.0079 to p = 0.0271; Fig. 8A-D) in myotubes incubated with palm-mac-CM, which were sufficient to restore phosphorylation of each intermediate to near normal levels. In addition, blocking antibody lead to a 30% increase in pY612-IRS1 (p = 0.019; Fig. 8A) and a 29% increase in pS9-GSK3b (p = 0.042; Fig. 8C) when added to LPS-mac-CM, while Akt and AS160 phosphorylation did not change significantly. Thus TNFa is a major mediator of the induction of myotube IR by palm-mac-CM, while LPS-mac-CM likely has effects mediated through other cytokines in addition to TNFa.

Discussion
In this study, we have utilised a model involving macrophage treatment with an SFA, a UFA or a combination of these and application of the CM generated to differentiated C2C12 myotubes in culture to determine whether and how increased local infiltration of macrophages into skeletal muscle during obesity ( (Fink et al., 2013;Hevener et al., 2007;Nguyen et al., 2007); NAT, CWJ and MEC, unpublished data) might impact upon local insulin sensitivity and the differential effects of FA types in this context. The principal findings were that palmitate treatment induces an M1-type polarisation of macrophages and increased secretion of TNFa, which causes activation of inflammatory/stress signalling pathways in myotubes and thus inhibition of insulin-stimulated glucose uptake and incorporation into glycogen, while treatment of macrophages with palmitoleic acid results in myotube insulin sensitisation and is capable of preventing the effects of palmitate in this model. Importantly, we verified that there was no detectable carry-over of FAs into the CM after cell washing and that there was no effect of control-mac-CM on myotube glycogen synthesis versus medium that had not been incubated with macrophages (data not shown).
Our data are generally consistent with a recent study in which CM generated by palmitate treatment of RAW macrophages caused reductions in glucose uptake and insulin-stimulated signalling in GLUT4-overexpressing L6 myoblasts (Samokhvalov et al., 2008). However, arguably our model utilising unmodified differentiated myotubes and macrophages from the same species (mouse) reflects the in vivo situation more closely. Interestingly, this previous study found that LPS-mac-CM actually had positive effects on insulin signalling, while we found its effects to be similar to those of palm-mac-CM, perhaps due to these differences in the models used or contrasting secretory profiles of the macrophage lines. Instead, our data are consistent with the proposal that both LPS and SFAs activate macrophages through binding to Toll-like Receptor-4 (Hwang, 2001;Lee et al., 2001;Nguyen et al., 2007). Our findings also clearly implicate a role for TNFa-induced activation of p38 MAPK, a mechanism that has been shown to impair insulin sensitivity in muscle cells before (de Alvaro et al., 2004). The effect of TNFa on myotubes involved activation of IKK and thus NFjB downstream of p38 MAPK in that study, which is consistent with the reduced levels of IjBa we observed. The Klip group have identified a role for novel PKC isoform activation in causing mac-CMinduced effects in myoblasts (Kewalramani et al., 2011). Since previously the effects of hyperglycaemia in monocytes have been ascribed to conventional PKC-mediated p38 MAPK activation (Devaraj et al., 2005), it is possible that novel PKC activation may occur upstream of p38 MAPK and IKK activation in our model.
Numerous publications have suggested that stress kinases such as JNK, p38 MAPK and IKK serine phosphorylate IRS1, leading to reduced tyrosine phosphorylation and thus reduced downstream signalling to glucose disposal (Aguirre et al., 2002;Fujishiro Fig. 5. Effects of siRNA-mediated p38 MAPKa silencing on conditioned medium-induced changes in p38 MAPK phosphorylation and protein levels in C2C12 myotubes. Western blots were generated using lysates derived from myotubes incubated with a siRNA pool targeting p38 MAPKa or a nonsense (scrambled) control for 72 h. (A) Silencing of p38a MAPK protein in 100nM insulin-treated C2C12 myotubes (mean 74% reduction). (B) C2C12 myotubes were incubated with conditioned medium generated by macrophages treated with palmitic acid, LPS or vehicle (control group) ± p38a siRNA pool for 16 h. Cells were serum starved for 1 h and then treated ± 100nM insulin for 15 min. Total phosphorylation and total protein levels of p38 MAPK (all isoforms) were assessed. Summary graphs show the mean ± SEM of 4 individual experiments, accompanied by representative blots. Selected post hoc significance is shown to simplify interpretation: ## p < 0.01 vs. scrambled siRNA, basal *** p < 0.001 and **** p < 0.0001 vs. scrambled siRNA, insulin-treated; £££ p < 0.001 vs. scrambled siRNA palmitic acid; $$ p < 0.01 vs. scrambled siRNA LPS. C -control conditioned medium (+ insulin treatment); P -palmitic acid-treated macrophage-conditioned medium (+ insulin treatment); L -LPS-treated macrophage-conditioned medium (+ insulin treatment). BA -basal (no insulin); p38a -p38 MAPKa siRNA-treated; D3 -transfection reagent-treated only. Hemi et al., 2011;Hirosumi et al., 2002;Yuan et al., 2001), although IRS1-independent mechanisms have also been proposed (Cleasby et al., 2007;Hoehn et al., 2008). We found that activating phosphorylation of both JNK and p38 MAPK were increased in myotubes, whereas this trend did not reach significance in the case of ERK, as was found in a human cell co-culture model (Varma et al., 2009). Pharmacological inhibition of p38 MAPK lead to partial restoration of palm-mac-CM-induced defects in glucose utilisation and signalling, data that were mostly recapitulated using p38 MAPKa siRNA. However, the abolition of the palm-mac-CM effects was due at least in part to the dual effects of p38 MAPK inhibition on insulin signalling: not only can p38 MAPK impair glucose utilisation (de Alvaro et al., 2004), but there is some evidence that it is also required for full insulin stimulated glucose uptake and PI3K signalling (Antonescu et al., 2005). The lesser effects of the siRNA-mediated gene silencing may be ascribed to the <80% reduction in p38a MAPK produced and/or the targeting of only the a-isoform, although this is the principal pro-inflammatory isoform in muscle (Boppart et al., 2000). Thus, targeting of p38 MAPK in muscle may be of therapeutic benefit for obesity-associated IR. In support of this, pharmacological inhi-bition of p38 MAPK improved basal glucose disposal in cultured adipocytes (Carlson and Rondinone, 2005), reduced obesity and whole body IR induced by high fat diet-feeding in rodents (Maekawa et al., 2010) and alleviated oxidative stress-associated IR in soleus muscle (Diamond-Stanic et al., 2011). In contrast, the pharmacological attenuation of JNK activity did not significantly prevent the palm-mac-CM-induced defects. Instead, this treatment showed some efficacy in ameliorating the similar defects induced by LPS-mac-CM, suggesting that LPS and palmitate treatment of macrophages may result in the release of a different range of soluble mediators that induce muscle IR through activation of alternative MAPKs. Indeed, Samokhvalov et al. showed that LPS treatment of macrophages elicited secretion of the anti-inflammatory cytokine IL-10, which might account for the similar effects of LPSmac-CM and palm-mac-CM on myotubes, despite the far larger macrophage TNFa secretion induced by the former treatment.
Of the four cytokines/chemokines measured here, only TNFa secretion was significantly increased by palmitic acid treatment, but the importance of this finding was confirmed by repeating myotube incubation with palm-mac-CM in the presence of a TNF-a blocking antibody. This cytokine also played a role in Fig. 6. The palmitic acid-treated macrophage-CM-induced defect in myotube PI3K signalling is partially restored by siRNA-induced silencing of p38a. Conditioned medium from macrophages treated with palmitic acid, LPS or vehicle (control group) was used to incubate C2C12 myotubes ± a p38 MAPKa siRNA pool for 16 h. Cells were serum starved for 1 h and treated ± 100nM insulin for 15 min. Phosphorylation and protein levels of (A) IRS-1 (pY612), (B) Akt (pS473), (C) GSK3b (pS9) and (D) AS160 (pT642) were assessed by western blotting of lysates. Densitometry was performed on blots and summarised as graphs showing the mean ± SEM of 4 individual experiments, alongside representative blots. Total protein levels were unchanged by any treatment. Selected post hoc significance is shown to simplify interpretation: * p < 0.05 and ** p < 0.01 vs. scrambled siRNA insulin-treated control; £ p < 0.05 vs. scrambled siRNA palmitic acid; # p < 0.05, ### p < 0.001 and #### p < 0.0001 vs. basal (no insulin) sample from the same siRNA treatment group. C -control conditioned medium (+ insulin); P -palmitic acid-treated macrophage-conditioned medium (+ insulin); L -LPS-treated macrophageconditioned medium (+ insulin). BA -basal (no insulin); p38a -p38 MAPKa siRNA-treated. mediating the effects of palm-mac-CM on L6 myoblasts (Samokhvalov et al., 2008), while TNFa is generally wellestablished as a mediator of IR, as infusion leads to reduced whole body insulin-mediated glucose uptake in humans (Plomgaard et al., 2005), while knockout of TNFa or its receptors protect mice from obesity (Hotamisligil et al., 1994;Uysal et al., 1997). In support of our proposed cellular mechanism, TNFa has been shown to induce p38 MAPK-dependent serine phosphorylation of both the insulin receptor and IRS1, reducing insulin-stimulated PI3K signalling, glucose uptake and GLUT4 translocation in myotubes (de Alvaro et al., 2004). However, an additional role for MCP1 (Sell et al., 2006) or other cytokines/chemokines in the mechanism has not been ruled out.
We have shown that macrophages treated with palmitoleic acid alone generated CM that was insulin sensitising, as well as increasing basal glycogen synthesis, while addition of palmitoleic acid to a palmitic acid incubation is also capable of preventing the IR associated with the palm-mac-CM-induced activation of MAPKs, suggesting an additional mechanism whereby UFAs have insulin sensitising effects in skeletal muscle. These effects may be mediated either by Fig. 7. Macrophages treated with palmitic acid show a pro-inflammatory polarisation that is abrogated by the addition of palmitoleic acid. J774 Macrophages were incubated with 0.75 mM palmitic acid, 0.75 mM palmitoleic acid, a combination of both, LPS, or vehicle (control group) for 8 h, washed with PBS and then conditioned medium (CM) generated for 16 h. After this period, CM and macrophages were collected and (A) macrophage mRNA expression of NOS2 was measured, (B) content of nitric oxide (lM) in culture medium was measured using the Griess assay, (C) arginase activity was assessed based on the amount of urea (lg) produced per mg of protein, (D) TNFa mRNA and (E) TNFa peptide were measured in macrophages/CM respectively by real-time PCR/ELISA, (F) MCP1 mRNA and G) MCP1 peptide were measured in macrophages/CM respectively and (H) CXCL2 peptide was measured in CM by ELISA. Values are the mean ± SEM of 5-6 individual experiments. Post hoc: * p < 0.05, *** p < 0.001 and **** p < 0.0001 vs. control.
anti-inflammatory factors secreted by the macrophages that have insulin sensitising effects in myotubes or by modifying activation of intracellular signalling pathways in macrophages, although further work will be needed to assess these possibilities. Similar effects of palmitate and palmitoleate on bone marrow-derived macrophage M1/M2 polarisation have been observed previously (Prieur et al., 2011), while UFAs have been shown to prevent SFA-induced NFjB activation  in macrophages.
Palmitate and palmitoleate have distinct direct effects upon insulin-stimulated glucose disposal in L6 myotubes (Dimopoulos et al., 2006), but the effects of palmitoleate were not shown to be mediated through a protective effect on insulin signalling previously. However, others have shown that oleate protects against the effects of palmitate by reducing the impairment in PI3K signalling (Coll et al., 2008;Gao et al., 2009) and through anti-inflammatory effects (Coll et al., 2008), including prevention of MAPK activation (Kadotani et al., 2009). This is likely mediated through reduced synthesis of ceramide and/or diacylglycerol in palmitate-treated myotubes (Chavez and Summers, 2003), rather than accumulation of the more inert triacylglycerol and this mechanism may also result in increased pro-inflammatory cytokine production by macrophages (Schilling et al., 2013).
Our data imply that palmitoleate has dual effects in muscle when mediated through the altered secretory milieu generated by macrophage treatment, on both basal and insulin-stimulated glucose disposal, associated with activation of the PI3K signalling pathway. Importantly, the beneficial effects of palmitoleic acid have also been demonstrated in vivo, as IR is improved by the feeding of a diet high in UFAs (Tardif et al., 2011) and by infusion of palmitoleate (Cao et al., 2008) in rodents. Indeed, Cao et al. proposed that palmitoleate represented a specific novel lipid signal, or ''lipokine'', which is not found in large amounts in the diet, but instead is released in more substantial amounts from adipose tissue and regulates muscle metabolism (Cao et al., 2008;Pinnick et al., 2012).

Conclusions
We have demonstrated that IR in muscle may at least in part result from the effects of SFAs on local macrophages, causing release of pro-inflammatory cytokines, paracrine activation of stress signalling in adjacent muscle fibres and thus impaired insulin-stimulated glucose incorporation into glycogen. This phenomenon is likely to be of most significance when increased numbers of macrophages accumulate in muscle during obesity. Further work must aim to establish whether this represents a quantitatively important mechanism of generation of muscle IR in vivo. The presence of palmitoleate is sufficient to prevent these effects, although it is unclear whether this reflects a universal property of UFAs or a ''lipokine'' effect. Nevertheless, these findings serve to reinforce the likely health benefits of a diet rich in UFAs rather than SFAs and β Fig. 8. Addition of a TNFa blocking antibody leads to partial restoration of the palmitic acid-treated macrophage CM-induced defect in myotube insulin signalling. C2C12 myotubes were incubated with conditioned medium derived from macrophages treated with palmitic acid, LPS or vehicle (No blocking group), with or without TNFa blocking antibody (10 lg/ml) for 16 h, before being serum starved and lysed in RIPA buffer, and SDS-PAGE and immunoblotting conducted to assess phosphorylation and total protein levels of (A) IRS1 (pY612), (B) Akt (pS473), (C) GSK3b (pS9) and (D) AS160 (pT642). Representative blots and summary data (mean ± SEM of 6 individual experiments) are shown. Protein expression levels of all intermediates were unchanged by any treatment. Selected post hoc significance is shown to simplify interpretation: * p < 0.05, ** p < 0.01, *** suggest that therapeutic targeting of TNFa action and p38 MAPK activity remain potential approaches for the treatment of T2D.