Nucleotides Released From Palmitate-Challenged Muscle Cells Through Pannexin-3 Attract Monocytes

Nutrient excess is a major factor contributing to the alarming incidence of obesity worldwide (1). Obesity leads to whole-body insulin resistance, a leading cause of type 2 diabetes and cardiovascular complications (1). A paradigm shift in our understanding of the inflammation accompanying obesity was the discovery that a high-fat diet (HFD) increases the number of immune cells in adipose tissue, and a growing body of evidence now suggests that obesity and type 2 diabetes are inflammatory diseases (2). Surprisingly, equivalent studies in skeletal muscle had been relatively scant despite this tissue being responsible for the majority of postprandial glucose use (3).

We and others recently demonstrated an increase in macrophage number and inflammatory phenotype within skeletal muscle from HFD-fed mice and obese subjects (4,5), but, strikingly, the factors responsible for immune cell infiltration in obese muscle (or other metabolically relevant tissues) are largely unknown. Skeletal muscle secretes several cytokines, recently renamed “myokines” (6), but the specific soluble mediators, channels, and receptors involved in the crosstalk between skeletal muscle and immune cells are virtually undefined. Interestingly, in addition to myokines, selective stimuli induce the release of small molecules from muscle, such as prostanoids, lactate, and nucleotides (7–12). The mechanism of release of these small molecules is unclear, but likely occurs through channels expressed at the plasma membrane. Pannexins are recently discovered channel-forming proteins that allow the release of cytoplasmic molecules to the extracellular space (13,14). Whereas pannexin-1 and its role in the physiological release of small molecules have been widely studied (15–17), the functions of pannexin-2 and pannexin-3 remain elusive. Moreover, the contribution of pannexin channels and small molecule release during metabolic inflammation remains unexplored.

Although the initial trigger of inflammation in vivo is debated, excess lipids contribute to the induction of insulin resistance and proinflammatory genes in metabolic tissues (18). Compellingly, palmitic acid (also called hexadecanoic acid, 16:0), a major dietary saturated fatty acid in blood, promotes a proinflammatory phenotype in several cell types in vitro (19–22). On the other hand, unsaturated fatty acids such as the monounsaturated fatty acid palmitoleic acid ([Z]-9-hexadecenoic acid, 16:1Δ9) are either innocuous or able to suppress inflammation (22,23).

Given the diversity of cells coexisting within metabolic tissues (parenchymal, endothelial, and myeloid), it is difficult to dissect the particular role of myocytes in the immune cell infiltration of muscle tissue during HFD feeding. Only cell culture paradigms allow for the establishment of vectorial communication between distinct cell types and their response to defined hyperlipidic environments. Here we show that when myotubes are challenged with palmitate—but not with palmitoleate—they release nonpeptidic factors that attract monocytes. We provide evidence that nucleotides released through pannexin-3 are major factors in this palmitate-induced crosstalk between muscle and immune cells.

Myeloid differentiation factor-88 (MYD88) inhibitory peptides were from InvivoGen (San Diego, CA). Small interfering RNA (siRNA) oligonucleotides for connexin (Cx)-43, Cx45, and Toll-like receptor (TLR) 4 were from GenePharma (Shanghai, People’s Republic of China), and for pannexin-3 from Qiagen (Chatsworth, CA). Other chemicals were from Sigma-Aldrich (St. Louis, MO).

Palmitate or palmitoleate (P9767 and P9417; Sigma-Aldrich) stock solutions (200 mmol/L) were prepared in 50% ethanol by heating at 50°C. Fatty acid–free, low-endotoxin BSA (A8806; Sigma-Aldrich) was dissolved in serum-free α-minimum essential medium (αMEM) to 10.5%. Fatty acid stocks were diluted 25× in the BSA solution and conjugated under agitation at 40°C for 2 h. This solution (lipid/BSA ratio 5:1) was further diluted in culture media. Palmitate and palmitoleate solutions thus coupled to BSA are denoted as PA and PO, respectively.

L6 muscle cells were grown and differentiated as described previously (24). THP1 monocytes were grown in RPMI 1640 medium containing 5% FBS. Cellular viability was assessed from lactate dehydrogenase (LDH) activity and MTT reduction (Cytotoxicity Detection Kit and Cell Proliferation Kit I; Roche Applied Science, Indianapolis, IN). Oligonucleotide siRNAs were transfected with JetPRIME (Polyplus Transfection, Illkirch, France). Myotubes were treated with 200 nmol/L siRNA for 24 h then stabilized in fresh media for 24 h before performing experiments.

L6 myotubes were treated in αMEM (2% FBS) for doses and times indicated. Supernatants were centrifuged at 10,000 rotations per minute for 5 min to pellet debris, aliquoted, and frozen immediately at −80°C. Supernatants (conditioned media [CM]) from PA-, PO-, and BSA-treated myotubes are herein denoted as CM-PA, CM-PO, and CM-BSA, respectively.

Nonesterified fatty acids were quantified in the media of fatty acid–incubated myotubes using the NEFA-HR(2) R2 Set (Wako Chemicals USA, Richmond, VA). Uptake was indirectly estimated from the fatty acid content in the myotube supernatant at the beginning and the end of the incubation.

In Boyden chambers (Transwell, 6.5 mm diameter, 5 μm pore diameter; Corning, Lowell, MA), 600 μL of attractant was added to the lower chamber. THP1 monocytes (100 µL of 5 × 10⁶cells/mL) in αMEM supplemented with 2% FBS were placed in the upper chamber. After 3 h at 37°C, cells were dislodged from the filter by gentle shaking, the upper chamber was discarded, and monocytes that transmigrated to the bottom well were counted using a Z2 Coulter Counter (Beckman Coulter Canada, Mississauga, ON, Canada).

ATP was specifically measured using the luciferase-based ENLITEN ATP Assay (Promega, Madison, WI). Other nucleotides (monophosphate, diphosphate, and triphosphate) and nucleosides were measured by hydrophilic interaction liquid chromatography coupled to mass spectrometry, as previously reported (25). The method was modified by limiting the monitored metabolites to nucleosides and nucleotides. Supernatants were analyzed by a Nexera UPLC (Shimadzu Corporation, Kyoto, Japan) coupled to AB/SCIEX Triple Quad 5500 mass spectrometer (AB SCIEX, Framingham, MA) using an electrospray ionization technique operating in multiple reactions monitoring mode. Parent-to-product transitions used for each detected metabolite are presented in Supplementary Table 1. Calibration curves (12.5–500 ng/mL) were generated for each detected metabolite. Uridine triphosphate (UTP), ATP, thiamine diphosphate, thiamine triphosphate (TTP), cytidine diphosphate, and guanosine triphosphate (GTP) were undetectable (limit of detection 1 ng/mL). The standard of uridine diphosphate (UDP) was not available at the time of analysis, and consequently results were reported in relative units by comparing the UDP peak areas.

All reagents were from Life Technologies (Carlsbad, CA). RNA was isolated using Trizol, and cDNA was synthesized using the SuperScript VILO cDNA kit. Ten nanograms per reaction were used for quantitative RT-PCR using predesigned Taqman probes for target genes and hprt1 or eef2 (housekeeping references).

Cells were scraped in lysis buffer, and protein content was measured using the bicinchoninic acid–based assay. Samples were boiled in Laemmli buffer, separated by SDS-PAGE, and transferred onto nitrocellulose. Membranes were blotted using primary and peroxidase-coupled secondary antibodies, then developed using chemiluminescence (ECL Kit; Bio-Rad, Hercules, CA), and analyzed using ImageJ software (National Institutes of Health, Bethesda, MD).

YO-PRO (1 µmol/L) and Texas-red dextran (10 kDa, 0.1 mg/mL) from Life Technologies were added to cells for 15 min. Cells were then washed with PBS and fixed (3% paraformaldehyde for 10 min). Images were acquired with a Leica DMIRE2 fluorescent microscope with a 10× air objective. Total green and red fluorescence were measured on 15 random fields per condition using ImageJ software.

Cytokines were determined with a rat Milliplex MAP Magnetic Bead Panel (Millipore, Hellerup, Denmark) on a Bio-Plex-200 System (Bio-Rad Laboratories, Copenhagen, Denmark).

The study was approved by The Hospital for Sick Children Animal Care Committee. Male C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME), singly caged and maintained at 21–22°C with light from 0600 to 1800 h, were fed a standard chow diet (5P07 Prolab RMH 1000; LabDiet, St. Louis, MO) or an HFD (60% by kcal) (D12492; Research Diets, New Brunswick, NJ) for 18 weeks. After a 4-h fast, mice were killed via cervical dislocation, and tissue was collected, flash frozen in liquid nitrogen, and preserved at −80°C.

Analyses were performed using Prism software (GraphPad Software, San Diego, CA). Results from dose responses were compared by two-way ANOVA (dose, treatment) followed by Tukey post hoc tests. One-way ANOVA was used to test differences between groups with equal variances. Statistical significance was set at P < 0.05.

Palmitate is the most abundant saturated fatty acid in Western diets, and it is widely used to challenge cells in culture. L6 myotubes were treated with BSA-conjugated PA or with the equivalent 16-carbon chain–length monounsaturated PO. Thereafter, supernatants (i.e., CM) were collected, and their chemoattracting activity toward THP1 monocytes was determined (Fig. 1A and B). CM-PA from PA-treated myotubes had a marked chemoattracting effect that was not reproduced by CM-PO or CM-BSA from PO- or BSA-treated myotubes, respectively. Maximal chemoattraction was observed after treatment with 0.5 mmol/L PA for 18 h. As the CM still contain fatty acids, monocyte migration was also measured toward regular medium containing 0.5 mmol/L fatty acids (Fig. 1C). Unlike the chemoattracting chemokine (C-C motif) ligand 2 (CCL2)/MCP1 (100 ng/mL), neither PA nor PO on its own significantly affected monocyte migration. The uptake of PA and PO by myotubes was similar (Fig. 1D), and an equivalent reduction in fatty acid and glucose content in the CM was observed in all conditions (Supplementary Tables 1 and 3). These results indicate that monocyte chemoattraction by CM-PA was not mediated by the fatty acid itself or by differences in myotube uptake of either fatty acids or glucose.

TLRs couple with adaptor MYD88 to activate the nuclear factor-κB (NF-κB) transcription factor, a critical step in palmitate-induced inflammatory response (18,21). In myotubes, PA evoked a dose-dependent decrease in inhibitor of κB (IκB), the canonical NF-κB repressor (Fig. 2A). Preventing IκB degradation in myotubes with parthenolide or a cell-permeant MYD88 inhibitory peptide (Fig. 2B and C) blunted the increase in monocyte migration induced by CM-PA. The involvement of this pathway was further confirmed by gene silencing. With an achieved reduction in the expression of TLR4 and MYD88 of 65% and 80%, respectively, the CM-PA–evoked monocyte chemoattraction was abolished (Fig. 2D). Together, these experiments demonstrate that PA engages the TLR/MYD88/NF-κB pathway in myotubes to attract monocytes.

The expression of chemokines and cytokines, which are classic monocytes attractants, was analyzed using quantitative PCR (qPCR) arrays (Fig. 3A). In myotubes exposed to PA, but not to PO, there was a significant rise in the gene expression of several chemokines (Cxcl2, Ccl2, and Cxcl1) and cytokines (Il1α, Tnfα, and Il6), but, surprisingly, that did not translate into a corresponding elevated secretion for them into the medium, as measured by Luminex Multiplex Immunoassay (Fig. 3B). This was confirmed using a membrane cytokine array and ELISAs for tumor necrosis factor-α and CCL2 (Supplementary Fig. 1). To explore whether the chemoattracting factor would be affected by conditions altering protein stability, CM-PA and CM-BSA were either heated or treated with proteinase K (Fig. 3C). These manipulations did not prevent the monocyte migration induced by CM-PA, suggesting that the chemoattractant is unlikely to be a polypeptide. Furthermore, upon filtering the CM through a Vivaspin column, only the fraction containing molecules of <3,000 Da displayed monocyte-chemoattracting activity (Fig. 3C).

As TLR4 signaling in myotubes was involved in the monocyte-chemoattracting ability of CM-PA (Fig. 2), we explored whether the activation of TLR4 signaling with lipopolysaccharide (LPS) could elicit similar effects. CM from LPS-treated myotubes (CM-LPS) also induced monocyte migration, but, interestingly, and unlike CM-PA, the monocyte-chemoattracting activity of CM-LPS was heat sensitive and proteinase K sensitive, and was retained in the fraction containing molecules >3,000 Da (Fig. 3D). Blocking antibodies neutralizing the chemokine CCL2/MCP1 did not affect CM-PA–induced monocyte migration (Fig. 3E) but completely prevented monocyte migration toward CM-LPS (Fig. 3F). Together, these results suggest that in response to palmitate, the myotube-derived factors responsible for monocyte migration is a low-molecular-weight compound. In contrast, CCL2/MCP1 is the monocyte chemoattractant released by myotubes in response to LPS.

Small molecules such as eicosanoids and nucleotides are bona fide regulators of the immune response, and can potentially affect monocyte migration (26,27). We consequently tested the chemoattractant effect of a variety of such small molecules. Eicosanoids (prostaglandins E₂ and F_2α, and arachidonic acid) and histamine were monocyte repellents, urate and lactate had no effect, and only CCL2/MCP1, formyl peptides (WKYMVdM), and ATP were able to attract monocytes (Fig. 4).

Nucleotides were the only small molecules that induced higher monocyte transmigration than CM-PA, with the potency order UTP > ATP > ADP > UDP > TTP (Fig. 5A and B). The attracting ability of CM-PA was eliminated when nucleotides were hydrolyzed with the nucleotidase apyrase (Fig. 5C). Notably, using an ATP-specific luciferase–based assay, ATP levels were detected and found to be higher in CM-PA compared with CM-PO or CM-BSA, and, consistent with the blunted monocyte attraction, apyrase treatment eliminated any measurable amounts of ATP in the CM (Fig. 5D).

Most cells express surface extracellular ecto-nucleotidases that cleave nucleotides from the interstitial space (28). Accordingly, we explored whether endogenous ecto-nucleotidases would tonically reduce the amount of chemoattractant released by the myotubes. The inhibition of myotube ecto-nucleotidases with ARL67156 augmented the levels of ATP (Fig. 5E) and concomitantly potentiated the monocyte-chemoattracting activity of CM-PA (Fig. 5F), without affecting either parameter in CM-BSA or CM-PO. Compellingly, the ATP concentration in CM-PA correlated strongly with monocyte chemoattraction (r = 0.747, P < 0.0001) across treatments and conditions (Fig. 5G), buttressing the proposition that nucleotides are responsible for the enhanced monocyte transmigration toward CM-PA.

Blocking the P2 family of receptors on monocytes using the broad-spectrum nonselective antagonists suramin and pyridoxal-phosphate-6-azophenyl-2′,4′-disulfonate (PPADS) also prevented CM-PA–induced monocyte chemoattraction (Fig. 5H). However, since P2 receptors can be activated by several nucleotides (29) and apyrase can cleave all nucleotide triphosphates and diphosphates (Fig. 5I), the results suggested that, in addition to ATP, other nucleotides may also be involved. Indeed, using hydrophilic interaction liquid chromatography coupled to mass spectrometry, we found elevated concentrations of ADP, UDP, and several monophosphate nucleotides in CM-PA compared with CM-BSA and CM-PO (Fig. 5J and K).

Since cytotoxic effects of fatty acids have been described in several cell types, and cells can release ATP after membrane damage, we assayed myotube viability. MTT reduction, LDH release, and caspase-3 cleavage were not significantly affected in myotubes treated with PA compared with BSA or PO (Supplementary Table 1 and Supplementary Fig. 2). Moreover, after mechanically damaging myotubes by scraping and vortexing to induce necrosis, LDH release increased 15-fold, indicating that >95% of the cells were undergoing necrosis. However, contrary to palmitate treatment, the supernatant from these scraped cells did not attract monocytes. In addition, no correlation was found between LDH release and monocyte chemoattraction (Supplementary Fig. 2). Thus, it is unlikely that the monocyte-chemoattracting activity of CM-PA was due to sporadic myotube cell death.

The route of ATP release from myotubes was next examined. Pannexins are channels that allow the release of cytoplasmic molecules into the extracellular space (14). In addition, gap junction molecules (i.e., Cxs) can form hemichannels, also allowing the release of small molecules (13). In L6 myotubes, Cx43 and Cx45 were the most abundant Cxs (Fig. 6A), while Cx40 and pannexin-2 were undetectable. Notably, only the expression of pannexin-3 significantly rose with PA treatment (Fig. 6B), and this effect was blocked by inhibiting NF-κB with parthenolide (Fig. 6C).

To mediate ATP release, channels must be open at the plasma membrane. The small dye YO-PRO-1 has a molecular weight (630 g/mol) close to that of ATP (507 g/mol), readily diffuses through open Cx/pannexin channels, and fluoresces upon binding to nucleic acids; hence, it has been used to ascertain the presence of open channels at the cell surface (15). YO-PRO was administered to myotubes treated with PA, PO, or BSA, along with a large Texas-Red-dextran (10 kDa) polysaccharide that cannot go through Cx/pannexin channels but enters cells with damaged membranes (Supplementary Fig. 3). In myotubes treated with PA, the YO-PRO/Texas-Red-dextran fluorescence ratio was significantly higher than that in PO-treated or BSA-treated myotubes (Fig. 6D), indicating that PA increases the number of open channels at the myotube plasma membrane.

Next, we reduced the expression of channels using cognate siRNA sequences. Only the knockdown of pannexin-3 eliminated the CM-PA–induced monocyte transmigration (Fig. 6E). Consistently, the ATP content significantly diminished in CM-PA derived from pannexin-3–depleted myotubes (Fig. 6F). These results demonstrate that pannexin-3 is required for nucleotide release by PA-treated myotubes into the media and for the consequent monocyte chemoattraction.

Finally, in a pilot experiment we measured the expression of pannexins in epididymal white adipose tissue (eWAT) and quadriceps muscle from a small cohort of mice fed an HFD for 18 weeks. Pannexin-1 was highly expressed in both tissues and increased in response to HFD feeding in eWAT but not in quadriceps muscle (Fig. 7A and B). The expression of pannexin-2 and pannexin-3 was lower, but HFD feeding induced a significantly higher expression in eWAT. In particular, pannexin-3 expression in this tissue was undetectable in chow-fed mice and was noticeably induced by HFD feeding (Fig. 7B). In quadriceps muscle, the trend was very limited, and more mice will be needed for a proper statistical analysis of the results. Of note, in human primary myotubes, pannexin-2 expression rose significantly in response to palmitate (Fig. 7C). These results demonstrate that, although pannexin isoforms may differ, palmitate-induced expression of pannexin channels is common across species and relevant during HFD feeding in mice.

Recent studies document a rise in macrophages within skeletal muscle of HFD-fed mice and obese individuals (4,5,30–33), but a key unresolved issue is whether and how skeletal muscle is capable of attracting monocytes. Here, we present evidence that CM from myotubes exposed to the saturated fatty acid PA, but not the unsaturated PO, attract monocytes in culture. Myotubes challenged in this manner activate endogenous inflammatory programs leading to the expression of pannexin-3 channels, and we identify nucleotides as new potential factors in muscle-to-monocyte crosstalk during metabolic inflammation.

In vivo, downregulation of CCL2/MCP1 (34) or its receptor CCR2 (35) only partially prevented the gain in inflammatory macrophages in adipose tissue and skeletal muscle of HFD-fed mice (5), suggesting that additional factors contribute to macrophage infiltration of tissues. In addition, two important studies (36,37) recently ascribed the gain in adipose tissue macrophages to in situ proliferation, which was dependent on CCL2/MCP1. Hence, the chemoattracting function of CCL2/MCP1 in obesity is debatable. Even when exerting a chemoattracting effect, CCL2/MCP1 might have been produced in response to endotoxin-like stimuli. Finally and importantly, it is not possible to ascertain from the above in vivo studies whether CCL2/MCP1 arose from myocytes/adipocytes, or whether instead it was contributed by endothelial cells or myeloid cells inside the tissues. It is precisely to approach these questions that we here explored the ability of a reconstituted cellular system to deconstruct the potential crosstalk between muscle and immune cells in the presence of fatty acids. We show that L6 myotubes selectively challenged with PA can attract monocytes, and that this does not rely on any chemokine. Instead, the chemoattracting factors are nucleotides released through pannexin-3 channels. By contrast, LPS-treated myotubes provoked a CCL2/MCP1-dependent monocyte chemoattraction. This suggests that both nucleotides and CCL2/MCP1 might contribute to the immune cell infiltration of skeletal muscle in vivo that occurs during HFD feeding. Finally, primary human myotubes also showed increased pannexin expression in response to palmitate, as did tissues from high fat-fed mice.

Dietary fats, in particular saturated fats, confer a state of low-grade inflammation to skeletal muscle. Either TLR activation or intracellular lipid intermediates can trigger an inflammatory response through activation of stress kinases (e.g., Jun NH₂-terminal kinase, extracellular signal–related kinase), generation of reactive oxygen species, and stimulation of NF-κB signaling, classically enhancing the expression of proinflammatory cytokines (38,39). However, there is no previous evidence of inflammatory cues regulating the expression of channels for small molecules. We report here that activation of the TLR4/MYD88/NF-κB pathway in myotubes challenged with palmitate significantly elevates the expression of pannexin-3. This response was not observed when exposing myotubes to the unsaturated fatty acid palmitoleate, and therefore represents a nutrient-specific response. In particular, we observed a significant rise in several nucleotides in the CM from palmitate-challenged myotubes, and silencing the expression of pannexin-3 abolished this gain.

The importance of extracellular nucleotides in cell-to-cell communication is evident in the immune system, where released ATP acts as a “find-me” signal for immune cells to promote phagocytic clearance of damaged or apoptotic cells (27). Specific release of ATP also occurs in skeletal muscle during physiological exercise as well as during pathological situations such as sepsis or myopathies (for review, see Pillon et al. [40]). Attempts to measure interstitial nucleotides using microdialysis have been scant, but have estimated ATP concentrations at ∼1 µmol/L in skeletal muscle from anesthetized cats (41) and ∼ 0.1 µmol/L in humans (42). Studies in other tissues reported interstitial ATP in the nanomolar range (43,44), but none of these experiments could take into account the ATP concentration in the unstirred layer covering the cell surface (45). Overall, the basal interstitial concentration of ATP can be estimated in the range of 1–100 nmol/L, but in pathological situations, extracellular ATP concentration can rise markedly, reaching up to 10 µmol/L, as was recently observed in a tumor microenvironment (46). In contrast, it is unknown whether extracellular/interstitial nucleotide levels change during obesity-associated inflammation.

Although significantly higher than the level of ATP detected in CM-PO and CM-BSA, the concentration of ATP in CM-PA reached 10 nmol/L, a concentration that was insufficient to induce monocyte migration. However, the monocyte-chemoattracting activity of CM-PA was abolished by apyrase, and we corroborated that apyrase cleaves all nucleotide triphosphates. While the vast majority of studies on physiologically released nucleotides have focused on ATP, UTP is also released during cardiac ischemia (47) and is a potent inducer of cell migration (48). Since cytidine triphosphate, GTP, and TTP could not be detected in CM-PA (assayed with a detection limit of 2 nmol/L), and >1 µmol/L each is required for effective chemoattraction (Fig. 5), these nucleotides are unlikely to participate in the CM-PA–induced chemoattraction. Hence, we surmise that a combination of ATP, ADP, UDP, and/or UTP constitute the chemoattracting find-me signal for monocytes that is broadcast toward palmitate-challenged muscle cells. Consistent with this assertion, monocytes express several P2 receptors that selectively recognize nucleotides. Even though all of them respond to ATP, P2Y₂ and P2Y₄ have high affinity for UTP; P2Y₆, and P2Y₈ have high affinity for UDP; and P2Y₁, P1Y₁₂, and P2Y₁₃ have high affinity for ADP (29). As two broad-spectrum nonselective antagonists of the P2 receptors, suramin and PPADS, reduced the CM-PA–induced monocyte attraction, it is conceivable that nucleotides released by PA-challenged myotubes enact chemoattracting activity by activating one or more P2X/P2Y receptors on monocytes.

The results described in this study bring a new understanding to the lipotoxic inflammatory response of myotubes. Along with our previously reported inflammatory polarization of macrophages conferred by palmitate-treated muscle cells (22), these results illustrate the bidirectional crosstalk that occurs between muscle and immune cells in the context of hyperlipidic environments. Such bidirectional communication could be ascertained only through the described use of defined cell culture paradigms.

In addition to the palmitate-induced increase in pannexin-3 in L6 myotubes and pannexin-2 in human myotubes, our pilot in vivo results suggest that pannexin-2 and pannexin-3 are upregulated in eWAT from obese mice (and there might be a trend toward an increase in the quadriceps muscle). The response in tissues is more complex than in cells, as muscle and adipose tissue are composed of various cell types and express different pannexin isoforms. Moreover, increasing the expression of pannexin is not the only way to induce nucleotide release, because opening of the channels can also be regulated. Animal models will be needed in the future to provide deeper insight into the role of nucleotides and pannexin channels in diet-induced obesity.

Irrespective of whether changes in pannexin expression occur in muscle with obesity, our results show a previously unrealized principle, that treatment of myotubes with PA results in the expression of pannexin-3 through the TLR/NF-κB pathway and a consequent release of nucleotides to the medium to chemoattract monocytes (Fig. 8). Whether increases in pannexin expression and/or pannexin opening are required needs to be explored, but our siRNA results show that the existence of pannexin channels is required for nucleotide release and monocyte chemoattraction.

In conclusion, these findings constitute a proof of concept of muscle-to-monocyte communication in hyperlipidic environments, and raise the possibility that, in vivo, muscle fibers might also release nucleotides through pannexins to promote macrophage infiltration of skeletal muscle. Our findings would also predict that targeting the pathways responsible for chemokine production may be insufficient to reduce macrophage infiltration of muscle, because other factors such as nucleotides may play a significant role in immune cell chemoattraction. Finally, our studies point to pannexins as interesting targets for tapering the recruitment of tissue inflammatory macrophages during metabolic disease.

Acknowledgments. The authors thank Dr. Sheila Costford (The Hospital for Sick Children, Toronto, Ontario), for providing the RNA from obese animals, and Dr. Michael Salter (The Hospital for Sick Children), for helpful discussion around nucleotides.

Funding. This project was supported by grants from the Canadian Diabetes Association and the Canadian Institutes of Health Research (grant MT12601) to A.K. N.J.P. was supported by postdoctoral awards from the Research Training Centre at The Hospital for Sick Children and from the Banting and Best Diabetes Centre of the University of Toronto. Y.E.L. was supported by a summer studentship from the Banting and Best Diabetes Centre of the University of Toronto.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. N.J.P. participated in the design of the study, coordinated and carried out the majority of the experiments, performed the statistical analysis, wrote the manuscript, and approved the final manuscript. Y.E.L. helped with experiments and analysis concerning muscle signaling and monocyte migration, and read and approved the final manuscript. L.N.F. performed the Luminex multiplex analysis, and read and approved the final manuscript. J.T.B., A.N., and M.-S.K. performed the liquid chromatography mass spectrometry analysis, and read and approved the final manuscript. P.J.B. participated in the design of the study, helped to perform the experiments, helped to write the manuscript, and read and approved the final manuscript. A.K. conceived the study, participated in its design and coordination, participated in the writing of the manuscript, and read and approved the final manuscript. A.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.