DUSP5 functions as a feedback regulator of TNFα-induced ERK1/2 dephosphorylation and inflammatory gene expression in adipocytes

Adipose tissue inflammation is a central pathological element that regulates obesity-mediated insulin resistance and type II diabetes. Evidence demonstrates that extracellular signal-regulated kinase (ERK 1/2) activation (i.e. phosphorylation) links tumor necrosis factor α (TNFα) to pro-inflammatory gene expression in the nucleus. Dual specificity phosphatases (DUSPs) inactivate ERK 1/2 through dephosphorylation and can thus inhibit inflammatory gene expression. We report that DUSP5, an ERK1/2 phosphatase, was induced in epididymal white adipose tissue (WAT) in response to diet-induced obesity. Moreover, DUSP5 mRNA expression increased during obesity development concomitant to increases in TNFα expression. Consistent with in vivo findings, DUSP5 mRNA expression increased in adipocytes in response to TNFα, parallel with ERK1/2 dephosphorylation. Genetic loss of DUSP5 exacerbated TNFα-mediated ERK 1/2 signaling in 3T3-L1 adipocytes and in adipose tissue of mice. Furthermore, inhibition of ERK 1/2 and c-Jun N terminal kinase (JNK) signaling attenuated TNFα-induced DUSP5 expression. These data suggest that DUSP5 functions in the feedback inhibition of ERK1/2 signaling in response to TNFα, which resulted in increased inflammatory gene expression. Thus, DUSP5 potentially acts as an endogenous regulator of adipose tissue inflammation; although its role in obesity-mediated inflammation and insulin signaling remains unclear.

While it is well recognized that MAPKs are activated by phosphorylation of the conserved threonine and tyrosine residues within the T-X-Y motif in the activation loop by upstream MAPK kinases, less is known about their deactivation by phosphatases in the regulation of adipose inflammation 6,13,14 . Dual specificity phosphatases (DUSPs) constitute a subclass of protein tyrosine phosphatases that specifically dephosphorylate (i.e. deactivate) MAPKs. DUSPs are inducible protein phosphatases that have been shown to regulate cellular inflammation and more recently have been shown to play a role in obesity and insulin resistance, in particular through the regulation of MAPK deactivation [14][15][16] .
In this report, we show that TNFα-mediated phosphorylation of ERK1/2 and JNK stimulated expression of the ERK-specific phosphatase, dual-specificity phosphatase 5 (DUSP5) resulting in feedback inhibition of ERK signaling and ERK-dependent inflammatory gene expression in adipocytes. Consistent with these findings in vitro, DUSP5 deletion in mice exacerbated TNFα-mediated ERK1/2 phosphorylation and inflammatory protein expression in adipose tissue. Our findings highlight the prospective role for DUSP5 in the regulation of adipocyte inflammation and suggest that phosphatases offer potential therapeutic targets to regulate MAPK-dependent inflammation.

DUSP5 positively correlates with TNFα expression in adipose tissue of diet-induced obese mice.
Adipose tissue is a predominate site for the development of obesity-associated inflammation leading to systemic metabolic dysfunction 10,12 . As such, we examined DUSP5 and TNFα gene expression in response to body weight gain in mice over time. For these studies, C57BL/6 J male mice were given ad libitum access to a high fat diet (HFD; 60% kcal) or control diet (CD) starting at six weeks of age. Studies were conducted at 18 wks (stage I) and 24 wks (stage II) of age, representing 12 wks and 18 wks of specialized diet, respectively. As shown in Table 1, 18 wk and 24 wk old mice fed a HFD presented with a 13% and 26% increase in body weight, relative to lean mice fed the CD.
To assess the inflammatory status at each stage, relative mRNA expression was determined for TNFα, a critical inflammatory marker in the development of chronic adipose tissue inflammation 10,12 . TNFα gene expression was increased in adipose tissue from epididymal depots of diet-induced obese (DIO) mice, with a progressive increase in TNFα abundance correlating with increased obesity (Fig. 1B). DUSP5 gene expression proportionally increased with increased obesity (Fig. 1A) and this positively correlated to an increase in TNFα (Fig. 1C).
TNFα regulates DUSP5 expression in adipocytes. Adipose tissue is predominantly composed of stromal vascular cells that contain preadipocytes and mature adipocytes. Therefore, we examined DUSP5 mRNA expression in preadipocytes and adipocytes. The 3T3-L1 preadipocyte cell line can be stimulated to differentiate into mature adipocytes, with greater than 90% conversion to mature fat cells. We found that DUSP5 mRNA was expressed to similar levels in preadipocytes and differentiated mature day 8 adipocytes (Fig. 2A); protein expression mirrored mRNA expression ( Fig. 2B; Supplemental Fig. 1A). Based on these results, and since TNFα expression positively correlated with DUSP5 expression in vivo, we next examined DUSP5 mRNA expression in 3T3-L1 preadipocytes and differentiated adipocytes stimulated with 100pM TNFα. This dose of TNFα has previously been shown to suppress insulin signaling and closely approximates TNFα concentrations in obese humans and animals 9,10 . Preadipocytes and mature adipocytes were stimulated in parallel with 100pM TNFα and RNA collected over time post-TNFα stimulation prior to qPCR analysis for DUSP5 gene expression. TNFα failed to stimulate DUSP5 mRNA expression in preadipocytes (Fig. 2C). In contrast, 100pM TNFα transiently (<1hr) increased DUSP5 mRNA expression in mature adipocytes (Fig. 2D).
DUSP5 is an inducible ERK-specific phosphatase [17][18][19][20] . Therefore, we examined ERK and JNK phosphorylation in mature 3T3-L1 adipocytes stimulated with 100pM TNFα. Cells were lysed for protein over time and immunoblotted with antibodies against total and phosphorylated ERK and JNK. TNFα-induced phosphorylation of ERK and JNK, which peaked at 15 minutes. This was followed by rapid dephosphorylation by 30 minutes (Fig. 2E; Supplemental Fig. 1B). DUSP5 mRNA (Fig. 2D) and protein expression ( Fig. 2E; Supplemental Fig. 1B) increased concomitant to ERK dephosphorylation, suggesting that ERK phosphorylation is inversely proportional to DUSP5 expression. DUSP5 functions to dephosphorylate ERK1/2 in adipocytes in response to TNFα. Based on findings above, we postulated that DUSP5 regulates ERK dephosphorylation in adipocytes in response to TNFα. To test this postulate, adipocytes were transiently transfected with short interfering RNAs (siRNAs) targeting two different regions of the DUSP5 gene (siDUSP5) for 48 hrs prior to stimulation with 100 pM TNFα. A non-targeting control siRNA (siControl) was also transiently transfected under the same conditions. Total RNA was collected 30 minutes post TNFα stimulation and whole cell lysates harvested over time post-TNFα and analyzed for relative DUSP5 mRNA abundance as well as total and phosphorylated ERK and JNK. Targeted siRNA significantly suppressed TNFα-induced DUSP5 gene expression compared to siControl (Fig. 2F). Transfection with non-targeting siControl did not affect TNFα-induced DUSP5 expression (Fig. 2F). Loss of DUSP5 gene expression via siRNA was mirrored by prolonged ERK phosphorylation over time (   2H). While loss of DUSP5 function exacerbated ERK phosphorylation (Fig. 2H), JNK phosphorylation remained unchanged. Lastly, transient transfection with siRNAs did not impact total ERK or JNK protein expression, demonstrating that DUSP5 specifically targeted ERK for dephosphorylation in adipocytes.

DUSP5 regulates ERK-dependent inflammatory gene expression in adipocytes.
Data presented above demonstrate that DUSP5 regulates ERK dephosphorylation. To determine if changes in ERK phosphorylation resulted in changes in adipocyte function, we next examined pro-inflammatory gene expression. Previous screening efforts from our lab (data not published), identified nuclear factor kappa B-and MAPK-dependent inflammatory genes. We report that TNFα stimulation resulted in a significant increase in inflammatory gene expression of C-C motif chemokine ligand 2/monocyte chemoattractant protein-1 (CCL2/ MCP-1), cyclooxygenase-2 (Cox-2) and interleukin-16 (IL-16) (Fig. 3A). To determine a role for ERK activity 3T3-L1 adipocytes were stimulated without TNFα (100pM) and protein harvested over time. Whole cell lysate was subjected to immunoblotting for the detection of DUSP5, phospho-ERK and JNK as well as total ERK and JNK. Adipocytes were transiently transfected with small, non-targeting control (100 nM) siRNAs or siRNAs (100 nM) targeting independent regions of the DUSP5 gene for 48 hrs prior to stimulation with TNFα. (F) RNA was isolated 30 min post-treatment and qPCR used to examine DUSP5 expression. (G) Whole cell lysate was collected over time post-TNFα stimulation and immunoblotting used to assess phospho-ERK and JNK as well as total ERK and JNK. (H) Adipocytes (n = 3/group) were transiently transfected with small, non-targeting control (100 nM) siRNAs or siRNAs (100 nM) targeting independent regions of the DUSP5 gene for 48 hrs prior to stimulation with TNFα. Protein was collected 15 minutes post-TNFα stimulation and immunoblotted for phospho-and total-ERK. Image J software was used to quantify ERK phosphorylation to total ERK expression and one-way ANOVA with Tukey's post-hoc analysis used to determine significance p < 0.05. regarding changes in gene expression, adipocytes were pretreated with U0126, a potent and selective inhibitor of the upstream ERK kinase, MEK, for 30 minutes prior to stimulation with TNFα. Total RNA was harvested after 1 hr post-TNFα stimulation. TNFα-induced CCL2, Cox-2 and IL-16 was significantly suppressed by ERK inhibition (Fig. 3A). Based on findings above, we postulated that loss of DUSP5 would increase expression of ERK-dependent pro-inflammatory genes. To test this, adipocytes were transfected with two independent siRNAs targeting DUSP5 for 48 hrs prior to stimulation with TNFα. Knockdown of DUSP5 significantly exacerbated CCL2, COX-2 and IL-16 compared to siControl in response to TNFα stimulation; demonstrating biological significance for DUSP5-mediated regulation of ERK phosphorylation.
ERK and JNK activity are necessary for TNFα-induced DUSP5 gene expression. It has previously been reported that ERK regulates DUSP5 expression creating a feedback loop 21 . To determine if MAPKs regulate TNFα-induced DUSP5 gene expression, adipocytes were pretreated with U0126, SP600125, and SB203850, representing inhibitors of ERK, JNK, and p38, respectively, for 30 minutes prior to stimulation with TNFα. RNA was harvested 30 minutes post-TNFα stimulation, while whole cell lysates were collected 15 minutes post-TNFα and immunoblotted for total and phosphorylated ERK, JNK and p38. ERK phosphorylation was inhibited when cells were treated with the MEK inhibitor U0126 (Fig. 4A). Of note, JNK phosphorylation was inhibited with treatment of SP600125 (Fig. 4A). This is interesting as the JNK and p38 inhibitor act as ATP-competitive inhibitors and thus are thought to primarily interfere with downstream phosphorylation. Indeed, inhibition of JNK and p38 with SP600125 and SB203850, led to attenuation of c-Jun and MAPKAPK2 phosphorylation, respective downstream targets of JNK and p38 (Fig. 4A). Inhibition of both ERK and JNK significantly suppressed DUSP5 gene expression (Fig. 4B).
To validate these findings, we used two independent inhibitors of the ERK and JNK pathway. Adipocytes were pretreated with U0126 and PD98059 to inhibit ERK as well as SP600125 and JNK-IN-8 to inhibit JNK for 30 minutes prior to TNFα stimulation. TNFα-induced ERK phosphorylation was completely blocked with both ERK inhibitors (Fig. 4C). Moreover, inhibition of ERK significantly blocked DUSP5 expression in which gene expression was reduced back to vehicle treated levels (Fig. 4D). Similarly, TNFα-mediated JNK phosphorylation was attenuated with SP600125 and JNK-IN-8 (Fig. 4E). Consistent with these findings, TNFα-induced DUSP5 gene expression was significantly attenuated in the presence of JNK inhibition (Fig. 4F). While it should be noted that we observed no cross talk with the JNK or ERK inhibitors SP600125 and U0126 on MAPK signaling, we did not address actions of these inhibitors regarding other signaling pathways. Collectively, however, these data suggest that JNK and ERK regulate DUSP5 gene expression. For all MAPK inhibitor experiments, complete blots with molecular weight markers are shown (Supplemental Fig. 2).

DUSP5 regulates ERK dephosphorylation and inflammation in vivo.
To determine if DUSP5 regulates ERK phosphorylation and inflammatory gene expression in adipose tissue in vivo, 1.25 µg TNFα or vehicle control was administered by intraperitoneal (i.p.) injections to 10-week old, male DUSP5 null mice and wild-type (WT) littermates. Animals were sacrificed 1 hr post-treatment and white adipose tissue collected from epididymal fat pads. Total RNA was subsequently isolated. Primers were designed to amplify exon 2 and exon 3, consisting of the phosphatase catalytic domain, to demonstrate loss of the DUSP5 gene. Using these primers, amplification of DUSP5 was not detected in the WAT of DUSP5 null mice but was abundantly expressed in WT littermates (Fig. 5A). Protein was also isolated from WAT and immunoblotted for total and phosphorylated ERK and JNK. Consistent with 3T3-L1 adipocytes, ERK phosphorylation was significantly increased in DUSP5 null mice compared to WT littermates stimulated with TNFα (Fig. 5B,C). ERK phosphorylation trended higher with vehicle treatment in DUSP5 null mice compared to WT littermates although not significant. No differences were noted with total ERK or JNK, nor were differences noted in phosphorylated JNK. Based on these results, we next looked at inflammatory gene expression. Similar to the increase in ERK phosphorylation, Cox-2 and CCL2 was significantly increased in DUSP5 null mice compared to WT mice in response to TNFα (Fig. 5D). Moreover, loss of DUSP5 protein expression (Fig. 5E) led to a significant increase in Cox2 and CCL2 protein expression (Fig. 5E,F) in response to TNFα. Of note, basal expression of these two inflammatory molecules was also increased with loss of DUSP5 function. Complete immunoblots for in vivo data can be found in Supplemental Fig. 3.

Discussion
The findings of this study demonstrate that DUSP5 acts as a negative regulator of adipocyte inflammatory gene expression that functions by regulating the magnitude and duration of ERK phosphorylation. We propose a model in which ERK and JNK-dependent induction of DUSP5 acts in a reciprocal feedback loop to regulate ERK signaling in a manner that links TNFα to pro-inflammatory gene expression in adipocytes (Fig. 5G).
Obesity is associated with chronic inflammation that is critical for the development of most obesity-mediated comorbidities including insulin resistance 12 . In response to metabolic stress, adipocytes produce and secrete chemokines (e.g. CCL-2/MCP-1) and cytokines (e.g. TNFα) that promote macrophage infiltration; this exacerbates adipose tissue inflammation and promotes insulin resistance 2,10,12 . We report that TNFα expression increased in adipose tissue dependent on the developmental stage of obesity (Fig. 1). As obesity increased TNFα expression increased suggesting infiltration of inflammatory immune cells and exacerbation of adipose tissue inflammation. Recent attention has been given to DUSPs, which have been shown to act as critical regulators of inflammation 14,22 . Similar to TNFα expression, DUSP5, an ERK1/2 specific phosphatase, increased with the development of obesity (Fig. 1). Furthermore, DUSP5 mRNA expression increased concomitant with increased TNFα expression (Fig. 1C), suggesting that TNFα regulates DUSP5 mRNA expression. Consistent with this hypothesis, we report that TNFα induced DUSP5 expression in adipocytes (Fig. 2D). These data suggest that MAPKs such as ERK1/2 are essential regulators of adipose tissue inflammation, which link TNFα signaling to activation of downstream transcriptional programs that promote pro-inflammatory gene expression 3,7 . Indeed, genetic and pharmacological inhibition of ERK signaling has been shown to protect animals from diet-induced obesity, adipose tissue inflammation and insulin resistance 7,23,24 . We report that TNFα increased ERK1/2 phosphorylation in adipocytes (Fig. 2E). Moreover, we demonstrated that ERK activation was necessary for inflammation in adipocytes, where pharmacological inhibition of ERK1/2 attenuated inflammatory gene expression (i.e. IL-16, CCL2, and COX-2) (Fig. 3A).
MAPK-specific DUSPs have been shown to act as critical downstream regulators of MAPK activation and thus inflammation 14,22 . Our observations that ERK1/2 activation (i.e. phosphorylation) was transient (Fig. 2E), suggesting rapid deactivation i.e. dephosphorylation by phosphatases, support the postulate that phosphatases act as critical regulators of ERK signaling. Consistent with this postulate, induction of DUSP5 expression was inversely proportional to ERK1/2 phosphorylation in adipocytes in response to TNFα (Fig. 2D,E). DUSP5 has previously been implicated as a regulator of ERK1/2 signaling in response to pathological cardiac hypertrophy and fibrosis 21,25 . Indeed, we report that genetic loss of DUSP5 exacerbated ERK1/2 phosphorylation in adipocytes in vitro (Fig. 2G,H) as well as in adipose tissue in vivo (Fig. 5B,C).
Emerging literature demonstrates that DUSP5 is a critical regulator of inflammation and immune cell function 26,27 . Consistent with these reports, genetic loss of DUSP5 exacerbated TNFα-induced inflammatory gene expression in adipocytes (Fig. 3B) as well as inflammatory gene and protein expression in epididymal white adipose tissue (Fig. 5D,E and F). It should be noted that secretion of pro-inflammatory cytokines from preadipocytes, adipocytes as well as resident adipose tissue macrophages contribute to obesity-induced inflammation and its related comorbidities 2,12 . Thus, it would be important to delineate the role for DUSP5 in each cell type, as well as the impact of cellular cross-talk (e.g. adipocyte-macrophage) on DUSP5 regulation and function. While our in vitro data, suggests that loss of DUSP5 perpetuates inflammation in the adipocyte, we do not know how macrophage-adipocyte interactions might change this dynamic. While we can postulate that exacerbated adipose tissue inflammatory gene expression is a consequence of adipocyte signaling dysfunction (i.e. ERK1/2 phosphorylation), studies designed to knockdown DUSP5 in macrophages or tissue-specific knockout studies in mice would provide valuable insight into the actions of this phosphatase.
Canonical DUSP regulation involves feedback inhibition, in which MAPKs induce DUSPs that feedback inhibit MAPK phosphorylation 21,28 . Consistent with this notion, we report that DUSP5 is downstream of ERK1/2 and JNK activation, in which ERK or JNK inhibition attenuated TNFα-induced DUSP5 expression (Fig. 4). While pharmacological inhibitors can elicit off-target actions, it should be noted that others have reported that ERK1/2 regulates DUSP5 gene expression 21,29 . In addition, the use of independent compound inhibitors to target JNK and ERK for inhibition likely precludes off-target overlap and strongly supports the findings that JNK and ERK regulate DUSP5 message. Collectively therefore, these data suggest cross-talk between the JNK and ERK1/2 signaling cascades, in which JNK-mediated induction of DUSP5 can inhibit ERK1/2 phosphorylation. Others have reported similar cross-talk between signaling cascades, in which JNK-dependent induction of DUSP10/16 blocks ERK1/2 activation in COS-7 cells 30 . Together, these findings argue that intracellular signaling cross-talk is mediated, in part, by inducible DUSPs, such as DUSP5.
To conclude, our findings highlight the importance of DUSP5 as an endogenous regulator of ERK1/2 activation in adipocytes as well as regulation of inflammatory gene expression. Our data suggests that ERK1/2-and JNK-dependent regulation of DUSP5 creates cross-talk and/or feedback inhibition for ERK1/2 signaling in adipocytes in response to TNFα. While our findings suggest that DUSP5 potentially regulates ERK-mediated activation of obesity-induced inflammation, studies examining DUSP5 in obesity-mediated metabolic dysfunction have yet to be performed. Loss of DUSP5 function studies in rodent models of diet-induced obesity would highlight the role for this phosphatase in the regulation of ERK signaling, adipose tissue inflammation and insulin resistance. Studies outlined in our report examined the acute impact for DUSP5 loss of function on adipocyte ERK signaling and inflammatory gene expression, in which cells or mice were transiently exposed to inflammatory stimuli and in which DUSP5 responded to TNFα stimulation in a transient manner. As mice on a high fat diet had increased DUSP5 expression (Fig. 1) however, these data would suggest that long-term loss of DUSP5 function can greatly impact adipocyte function and metabolic outcomes.

Methods
Experimental animals. C57BL/6 J mice were rendered obese by diet. The supplier (Jackson Laboratories, Bar Harbor, Maine) housed and treated the mice until shipment 1 week prior to tissue harvest. Diet-induced obese C57BL/6 J mice were given free access to a high fat diet (HFD) consisting of 60% of total kcal from fat (Research Diets Inc. D12492) from 6 weeks of age, while lean controls were fed a low fat diet (LFD) consisting of 10% total kcal from fat (Research Diet Inc. D12450B). 10% kcal was from protein in both diets. The balance in caloric value was provided by differences in carbohydrate. Mice were given free access to food. Mice were shipped at 18 weeks and 24 weeks of age prior to euthanasia by CO 2 gas asphyxiation. Epididymal white adipose tissue (WAT) was collected for total RNA.
Ten week-old, male, wild-type (WT) and DUSP5 null mice (sv/129 strain) were for TNFα experiments. WT and DUSP5 null animals were subjected to intraperitoneal (i.p.) injection with vehicle (PBS + 0.1% BSA) or 1.25 µg TNFα for 1 hr prior to asphyxiation, and WAT collected for protein and total RNA. Animal care and use was in compliance with the Institute of Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Use and Care Committee at the University of Colorado Denver. All experiments were in accordance with the NIH guidelines.

Real-Time qRT-PCR.
Total RNA was isolated from WAT, preadipocytes or adipocytes via Qiazol (Qiagen) following the manufacturer's protocol. Total RNA was quantified with a Nanodrop ND-1000 spectrophotometer and reverse-transcribed to cDNA using the Verso cDNA Synthesis Kit (Thermo Scientific; AB-1453). The reverse transcription (RT) master mix containing RT buffer, deoxyribonucleotide triphosphate (dNTP) mix, RT random primers, RNase inhibitor (1.0 U/µl), and RT enzyme was added to 500 ng RNA and RNase-free water.
PCR amplification was run using the BioRad CF96X qPCR instrument (BioRad) that consisted of enzyme activation at 95 °C for 15 minutes, followed by 40 cycles of denaturation at 95 °C for 15 sec combined with annealing at 55 °C for 30 sec and extension at 72 °C for 1 minute to examine DUSP5, CCL2, Cox-2 and IL-16 gene expression (Table 2). TNFα was examined using TaqMan TM primers purchased from Applied Biosystems (ABI; Mm99999068_m1). Data were recorded and analyzed with Sequence Detector Software (BioRad) and graphs visualized with GraphPad Prism software. All data were presented as mean ± standard error of the mean (SEM). Data were normalized to 18 S, previously validated as a suitable reference gene 31 .
RNA Interference. Short interfering RNAs (siRNAs) for DUSP5 (Sigma; SASI_Mm01_00262935; DUSP5#1 and SASI_mM02_00300305; DUSP5#2) specific sequences as well as non-targeting siRNA control (siControl) sequences (MISSION siRNA Universal Negative Control #1; Sigma-SIC001) were transfected using Lipofectamine 3000 transfection reagent according to manufacturer's (Thermo Scientific) protocol. Briefly, 3T3-L1 adipocytes were differentiated in 6-well culture dishes. Growth medium was then replaced with DMEM under serum-deprived conditions prior to addition of 3 µl of Lipofectamine 3000 reagent and either 100 nM DUSP5 specific siRNAs or non-targeting siRNA control for 3 hr. Growth medium was subsequently switched to fresh growth medium with FBS for 48 hr incubation prior to stimulation with TNFα. Cells lysate was harvested for RNA or protein as described above.
Statistical analyses. Statistical analysis was completed by ANOVA followed by post-hoc testing (Tukey's test was used for post-hoc analysis unless otherwise noted) using GraphPad Prism software. Statistical significance (defined as p < 0.05) is reported as applicable.
Data Availability. Data will be made freely available upon request.