IKKε Is Key to Induction of Insulin Resistance in the Hypothalamus, and Its Inhibition Reverses Obesity

In the hypothalamus, insulin and leptin are potent anorexigens (1). Insulin activates the insulin receptor (IR), leading to tyrosine phosphorylation of IR substrate 1 (IRS-1) and 2 (IRS-2). Once activated, IR substrates bind to and activate the enzyme phosphatidylinositol 3-kinase (PI3K), increasing protein kinase B (Akt) phosphorylation. These events decrease food intake (1–11). Leptin signaling in the hypothalamus occurs through leptin receptor, Janus kinase 2 (JAK2), and signal transducer and activator of transcription 3 (STAT3) activation. In addition, leptin may act through the JAK2/IRS/PI3K pathway (12). Both leptin and insulin increase the transcription of proopiomelanocortin (POMC), an anorexigenic neuropeptide (6,13–15), and inhibit the transcription of Agouti-related peptide and neuropeptide Y (NPY), which are orexigenic neuropeptides (8).

High-fat feeding affects insulin and leptin signaling, contributing to dysregulation of hypothalamic energy homeostasis control (16). Activation of serine kinases c-jun N-terminal kinase and inhibitor of κB kinase (IKKβ) induces inhibitory IRS-1 serine phosphorylation, trigging insulin resistance (17,18). It is well known that IKKβ is activated in hypothalamic neurons of obese mice, inducing insulin and leptin resistance. Inhibition of hypothalamic IKKβ improves insulin and leptin sensitivity, preventing diet-induced obesity (DIO) (18).

IKK epsilon (IKKε), also known as inducible IKK, is mainly expressed in immune cells, thymus, and spleen. Also a serine kinase, IKK is induced by the activation of nuclear factor-κB (NF-κB) (19). That IKKε might regulate the p65 NF-κB subunit and amplify inflammatory signals through the activation of other transcription factors is of interest (20–23).

IKKε levels and activity have been shown to be higher in the liver and adipose tissue of animals fed a high-fat diet (HFD) (24). Whole-body IKKε knockout (KO) mice were protected from DIO and insulin resistance, suggesting an important role of IKKε in mediating insulin resistance induced by HFD (24). Of note, despite the leanness on an HFD, IKKε KO mice had higher food intake, Vo₂, body temperature, and uncoupling protein 1 (UCP-1) expression in white adipose tissue, suggesting an alteration in the regulation of energy balance (24).

The enhanced IKKε expression in adipose tissue, liver, and adipose tissue macrophages of DIO mice, together with the complex IKKε KO phenotype, indicate that this serine kinase is key to the regulation of metabolism. However, the regulation of IKKε in the hypothalamus and its role in the energy balance of DIO mice have not yet been investigated. Thus, the aim of the current study was to investigate the expression and regulation of IKKε in the hypothalamus of obese mice and whether IKKε plays a role in regulating energy balance and hepatic glucose metabolism in the liver.

All experiments were approved by the Ethics Committee of the State University of Campinas. Eight-week-old male C57BL/6J and db/db mice obtained from the State University of Campinas, São Paulo, were assigned to receive a standard rodent chow or an HFD with 55% calories from fat as previously described (25,26) and water ad libitum.

Anesthetized mice underwent stereotaxic implantation (Model 963 Ultra Precise Small Animal Stereotaxic Instrument; Kopf) of 26-gauge stainless steel cannulas (Plastics One) in the right-side lateral ventricle as described previously (27).

To inhibit IKKε in the hypothalamus, we used the pharmacological inhibitor CAY10576 (Cayman Chemical Company, Merck KGaA, Darmstadt, Germany) 60 μmol/L diluted in 10% DMSO in saline (vehicle). CAY or vehicle were injected through intracerebroventricular (ICV) cannulation in obese mice twice a day (0800 h and 1700 h) for 5 days. In another set of animals, we inhibited the expression of IKKε by small interfering RNA (siRNA): RNA sense: 5′ [Phos] rCrCrGrGrCrArGrArArGrGrUrGrCrUrArArUrCrArU 3′ and RNA antisense: 5′ [Phos] rGrArUrUrArGrCrArCrCrUrUrCrUrGrCrCrGrGrCrU 3′ and its control green fluorescent protein (scramble): RNA sense: 5′ [Phos] rCrArGrGrCrUrArCrUrUrGrGrArGrUrGrUrArUdTdT 3′ and RNA antisense: 5′ [Phos] rArUrArCrArCrUrCrCrArArGrUrArGrCrCrUrGdTdT 3′, which were infused continuously by an ICV micro-osmotic pump (Alzet 1007D; DURECT Corporation, Cupertino, CA) during the 5 days.

Body weight, fat mass, UCP-1 protein expression in adipose tissue, PEPCK protein expression in liver, and pyruvate were measured after 5 days of ICV treatments. Free-feeding mice were singly housed in a metabolic cage (3701M081; Tecniplast, Buguggiate, Varese, Italy) for acclimatization for 5–7 days. Cumulative food intake was calculated by summing the values of 24 h of food intake during the 5 days of treatment. In another set of mice, 4 and 8 h of food intake were recorded in response to ICV injection of human recombinant insulin 2 μg (Eli Lilly and Co., Indianapolis, IN) or leptin 10 ng (Calbiochem, San Diego, CA).

Twelve-hour–fasted mice were injected with sodium pyruvate 2 g/kg i.p. Blood samples were collected from the tail immediately before and at 15, 30, 60, 90, 120, 150, and 180 min after the pyruvate injection.

Vo₂, Vco₂, and respiratory exchange ratio (RER) were measured in fed mice through an indirect open-circuit calorimeter (Oxymax Deluxe System; Columbus Instruments, Columbus, OH) as described previously (27). Mice were allowed to adapt 2 days before. Measurements were done on the last day of IKKε inhibition.

Hypothalami and blood were collected after decapitation from fasted mice treated with CAY or vehicle. We determined interleukin (IL)-1β, IL-6, and tumor necrosis factor-α (TNF-α) levels in serum and NF-κB in cellular nuclei of lysates obtained from hypothalami by using commercially available ELISA kits (Pierce Biotechnology Inc., Rockford, IL) following the manufacturer’s instructions.

Immunofluorescence was done as described before (28). We used anti-IKKε (D20G4-Rabbit mAb; Cell Signaling Technology, Danvers, MA) 1:1,000 dilution, anti-neuronal nuclei (NeuN), a neuronal marker (MAB377-Mouse; Millipore Corporation, Billerica, MA) 1:1,000 dilution, and anti-S100 β-antibody, an astrocyte marker (ab41548; Abcam). The negative controls were done by omitting the primary antibodies.

Hypothalami were removed and homogenized in buffer as described previously (27). Phospho-IR, UCP-1, PEPCK, phospho-JAK2, phospho-STAT3, and phospho-IKKα/β antibodies were from Santa Cruz Technology (Santa Cruz, CA). IKKε, phospho-Akt, JAK2, Akt, and TBK1 were from Cell Signaling Technology.

Hypothalami were collected from C57BL/6J mice fed chow or HFD treated with vehicle, CAY, or IKKε siRNA, homogenized with lysis buffer (50 mmol/L Tris [pH 7.5], 150 mmol/L NaCl, 2 mmol/L EDTA, 5 mmol/L NaF, 25 mmol/L B-glycerophosphate, 1 mmol/L sodium orthovanadate, 10% glycerol, 1% Triton X-100, 1 mmol/L dithiothreitol, and 1 mmol/L phenylmethylsulfonyl fluoride) in the presence of protease inhibitors (Sigma-Aldrich), and immunoprecipitation was performed as described previously (19). The pellets containing purified kinases (IKKε or TBK1) were incubated in kinase buffer containing 25 mmol/L Tris (pH 7.5), 10 mmol/L MgCl₂, 1 mmol/L dithiothreitol, and 10 μmol/L ATP and in 1 μg myelin basic protein (MBP) (Millipore) per sample as a substrate for 30 min at 30°C. The kinase reaction was stopped by adding 4× SDS buffer and warmed for 5 min at 95°C. Supernatants were resolved by SDS-PAGE, transferred to nitrocellulose, and incubated with antiphosphoserine (ab1603; Abcam). The bands were analyzed by autoradiography and quantified using UN-SCAN-IT software.

Twenty-four-hour–fasted mice treated for 5 days with CAY10576 or vehicle were injected ICV with saline or insulin. The hypothalami were harvested after 6 h, quickly frozen in liquid nitrogen, and stored at −80°C until RNA processing. Total RNA was obtained with an RNeasy Mini Kit (Cat. no. 74106; Qiagen, Alameda, CA). Quantitative PCR was done using TaqMan PCR Master Mix (Applied Biosystems) as previously described (25). Primer and probe sequences were purchased from Applied Biosystems and were NPY, Mm00445771_m1; POMC, Mm00435874_m1; and GAPDH, Mm00475829_g1 for mouse.

Results are expressed as mean ± SD. The significance was determined by two-tailed Student t test, one- or two-way ANOVA with Bonferroni posttest, as appropriate, and differences were considered significant if P < 0.05. We used GraphPad Prism software (GraphPad Software, San Diego, CA) for all statistical analyses.

IKKε protein expression was detected in the hypothalamus of control mice by immunoblotting, using the spleen as a positive control (Fig. 1A). Mice on an HFD and db/db mice had increased protein levels of IKKε in the hypothalamus compared with mice on a chow diet (Fig. 1B). These data were supported by immunofluorescence that showed higher staining for IKKε in the hypothalamus of mice on HFD compared with control mice (Fig. 1C). Using double staining for IKKε and NeuN markers or IKKε and astrocytes (s100 β-marker), we showed that IKKε is expressed predominantly in neurons (Fig. 1C), and there is no colocalization of IKKε with astrocytes in the hypothalamus (Supplementary Fig. 1A). The negative controls for the immunostaining for IKKε, NeuN, and s100 β-antibodies are shown in the Supplementary Fig. 1B. The activity of IKKε in the hypothalamus of mice on an HFD treated with vehicle was increased compared with mice on chow (Fig. 1D). This activity was blunted in the hypothalamus after 5 days of CAY or IKKε siRNA treatment. TBK1, which is also a serine kinase inhibited by CAY, was not inhibited by IKKε siRNA treatment (Fig. 1D). IKKα/β phosphorylation was not altered by CAY or siRNA treatment (Fig. 1E).

To evaluate whether the inhibition of IKKε affected energy homeostasis, we treated the mice on an HFD with ICV CAY for 5 days. ICV CAY injections decreased the body weight and epididymal and retroperitoneal fat mass of mice fed an HFD compared with the vehicle group (Fig. 2A–C). ICV CAY treatment decreased cumulative food intake compared with vehicle treatment (Fig. 2D). Accordingly, mRNA levels of the orexigenic neuropeptide NPY were lowered, and mRNA levels of the anorexigenic neuropeptide POMC were elevated in the hypothalamus of fasted mice fed an HFD and treated with CAY compared with vehicle-treated mice on HFD (Fig. 2E and F). In addition, ICV CAY treatment enhanced Vo₂ compared with mice treated with vehicle. Vco₂ and RER were not different (Fig. 2G–I). Furthermore, we observed elevated UCP-1 protein levels in the brown adipose tissue of mice on an HFD that received CAY treatment compared with the vehicle group (Fig. 2J).

Long-term treatment with ICV CAY decreased fasting blood glucose, insulin, and leptin levels (Fig. 3A–C). To investigate whether the inhibition of IKKε interfered with proinflammatory signals, we measured serum cytokine levels of obese mice treated with CAY or vehicle. There was no difference in the serum IL-1β and IL-6 levels of CAY- and vehicle-treated mice. Serum TNF-α levels were decreased in mice on an HFD treated with CAY (Fig. 3D–F). CAY-treated mice showed improved insulin sensitivity compared with vehicle-treated mice during an insulin tolerance test (Fig. 3G). To evaluate hepatic glucose production, we performed a pyruvate test, which indicated that ICV CAY treatment decreased glycemia after intraperitoneal pyruvate injection, suggesting lower hepatic glucose production compared with vehicle-treated mice (Fig. 3H). This result was associated with lower PEPCK protein levels in the liver of mice on HFD treated with CAY (Fig. 3I).

As expected, there was no change in food intake in response to insulin in mice on an HFD treated with vehicle. However, CAY treatment decreased food intake after 4, 8, and 24 h of insulin injection (Fig. 4A–C). Next, we investigated whether CAY treatment for 5 days affected insulin signaling in the hypothalamus of mice on an HFD. There was an improvement in IR and Akt phosphorylation and IRS-1/PI3K association in response to insulin in the hypothalamus of obese mice treated with CAY compared with vehicle-treated mice (Fig. 4D–F). Accordingly, inhibition of IKKε with CAY decreased IRS-1^Ser307 phosphorylation in the hypothalamus (Fig. 4G). The p65 NF-κB levels in the hypothalamus were decreased in mice treated with CAY compared with the vehicle group (Fig. 4H). As expected, there was no change in food intake in response to leptin in mice on an HFD treated with vehicle. However, CAY treatment decreased food intake after 4, 8, and 24 h of leptin injection (Fig. 4I–K). These results were accompanied by increased JAK2 and STAT3 phosphorylation, IRS-1/PI3K association, and IRS-2/PI3K association in response to leptin in the hypothalamus of CAY-treated mice (Fig. 4L–O).

To exclude the interference of adiposity and food intake on insulin and leptin signaling in the hypothalamus of mice treated with CAY, we performed pair-feeding experiments. Thus, animals received the same amount of an HFD during ICV treatment with vehicle or CAY (Fig. 5A). Of note, CAY treatment induced a severe weight loss and a reduction of epididymal and retroperitoneal fat mass compared with the vehicle group, even though they received the same amount of HFD (Fig. 5B–D). IR and Akt phosphorylation or JAK2 and STAT3 phosphorylation, in response to insulin and leptin, respectively, were higher in the hypothalamus of mice treated with CAY than in vehicle-treated mice (Fig. 5E–H). ICV CAY treatment increased Vo₂ consumption. Vco₂ and RER were not different in mice treated with CAY or vehicle (Fig. 5I–K). In addition, we observed elevated UCP-1 protein levels in the brown adipose tissue of mice on an HFD that received CAY treatment (Fig. 5L).

To definitively exclude the interference of adiposity on insulin and leptin signaling in the hypothalamus of mice treated with CAY, we performed pair-weight experiments. Thus, animals received different amounts of HFD to maintain the same body weight during CAY and vehicle treatments (Fig. 6A). At the end of the treatments, there was no difference in epididymal and retroperitoneal fat mass between the two groups (Fig. 6B and C). In this context, vehicle-treated mice received less HFD than the CAY group (Fig. 6D). The results of insulin and leptin signaling were similar to the results obtained in the pair-feeding experiments. IR and Akt phosphorylation or JAK2 and STAT3 phosphorylation in response to insulin or leptin, respectively, were higher in the hypothalamus of mice treated with CAY than in vehicle-treated mice (Fig. 6E–H).

To increase the specificity of IKKε inhibition in the hypothalamus, we used ICV injections by micro-osmotic pump with IKKε siRNA or scramble siRNA as a control for 5 days. IKKε siRNA decreased IKKε mRNA levels in mice on an HFD (Fig. 7A). Accordingly, there was a decrease in hypothalamic IKKε protein expression in mice treated with siRNA compared with the scramble group (Fig. 7B). Weight loss was increased and epididymal and retroperitoneal fat mass decreased in siRNA-treated mice (Fig. 7C–E). Food intake was lower and Vo₂ higher in siRNA-treated mice compared with the scramble group (Fig. 7F and G). Vco₂ was not different between the groups. In contrast, RER was lower in the siRNA group than in the scramble group (Fig. 7H and I). UCP-1 expression was increased in the brown adipose tissue of siRNA-treated mice (Fig. 7J). IR phosphorylation and Akt phosphorylation were enhanced in response to insulin in the hypothalamus of obese mice treated with siRNA. IRS-1 serine phosphorylation was decreased after siRNA treatment (Fig. 7K–M). Fasting blood glucose and PEPCK protein expression in liver were decreased in the IKKε siRNA group (Fig. 7N and O).

IKKε expression was higher in the hypothalamus of db/db mice (Fig. 1B). Thus, we used ICV CAY injections for 5 days to inhibit IKKε activity in this animal model. The activity of IKKε in the hypothalamus of db/db mice was lower after CAY injections compared with vehicle-injected mice (data not shown). CAY treatment induced a greater weight loss and a reduction of epididymal and retroperitoneal fat mass compared with the vehicle-treated db/db mice (Fig. 8A–C). This result was accompanied by a reduction in food intake and increased Vo₂. Vco₂ was similar in the two groups (Fig. 8D–F). RER was not different between the groups (Fig. 8G). UCP-1 protein expression was elevated in brown adipose tissue of db/db mice treated with CAY (Fig. 8H). In addition, CAY treatment decreased food intake after 4 and 8 h of insulin injection, and IR phosphorylation and Akt phosphorylation were enhanced in response to insulin in the hypothalamus of db/db mice treated with CAY (Fig. 8I–L). Furthermore, we observed a decrease in fasting blood glucose and PEPCK protein levels in the liver (Fig. 8M and N).

In the current study, we showed that in the hypothalamus of DIO and genetic obese mice, there was an increase in IKKε protein expression and activation. In addition, IKKε inhibition reduced adiposity and food intake, increased energy expenditure, and improved inflammation and glucose metabolism.

Obesity is associated with low-grade inflammation in various tissues, including the hypothalamus (17,28,29). Activation of the IKK/NF-κB pathway is a marker of inflammatory signals in the hypothalamus (18). It was shown before that hypothalamic neuronal IKKβ plays an important role in controlling energy homeostasis (18). In the current study, we demonstrate that IKKε is predominantly expressed in neurons not in astrocytes and may affect the neuronal regulation of energy balance.

The results showed that the inhibition of hypothalamic IKKε reduced NF-κB activity in the hypothalamus, suggesting alleviation of the inflammatory environment in this tissue. The molecular link between inflammation and insulin resistance remains to be completely established, although it is known that the NF-κB pathway plays an important role (30). Disruption of the NF-κB pathway by deletion of the canonical IKKβ gene or by pharmacological inhibition reverses insulin resistance in obese models (18). However, the role of the noncanonical IKKε is poorly understood.

Insulin and leptin resistance occurs before weight gain in rodents on an HFD, suggesting that abnormalities in these hormones are likely to trigger the development of obesity rather than the opposite (31). In the current study, we observed impairment of insulin and leptin action and signaling in the hypothalamus of mice on an HFD. High-fat feeding promotes an inflammatory environment, leading to activation of serine kinases, inducing IRS-1 serine phosphorylation. This event triggers hypothalamic insulin resistance. IKKε blockage with CAY or siRNA blunted IRS-1^Ser307 phosphorylation without altering the phosphorylation of IKKα/β. This observation reinforces that hypothalamic IKKε is important in blunting insulin signaling. Together with a reduction in IRS-1 serine phosphorylation in the hypothalamus of CAY- and siRNA-treated mice, we observed an enhanced IRS-1/PI3K association in response to insulin, indicating an improvement of insulin resistance in this tissue. In addition, we showed that CAY or siRNA treatment was able to reverse leptin resistance. This effect may be explained, at least in part, by the reduction of IRS-1 serine phosphorylation.

The inhibitor CAY blocks another serine kinase, TBK1, in peripheral tissues (19). TBK1 and IKKε have been proposed as key regulators of metabolic function in mice (19). Their inhibition with the drug amlexanox was able to improve metabolic dysfunction in obese mice (19). In our experiments, we observed similar effects on energy metabolism and insulin signaling in the hypothalamus by using CAY or siRNA. We argue that IKKε is the major player impairing energy and glucose metabolism in obese mice because siRNA treatment is very specific, inhibiting IKKε but not TBK1, and we observed enhanced TBK1 activity in the hypothalamus of siRNA-treated mice. The increased TBK1 activity may be due to direct activation by the canonical IKKs IKKα and IKKβ in response to inflammatory inputs (32). In fact, TBK1 phosphorylation and activation is increased in peritoneal macrophages from IKKε KO mice, suggesting that in the absence of IKKε, TBK1 may be upregulated (19). In the current study, it is unknown which hypothalamic neuron displayed enhanced TBK1 activity because we used the whole hypothalamus samples in our experiments.

CAY treatment was associated with hypophagia in response to insulin and leptin, suggesting a reversion of insulin and leptin resistance in the hypothalamus. These effects persisted even in pair-feeding and pair-weight experiments, suggesting that the inhibition of IKKε and not differences in body weight or food intake accounted for the improvement in energy metabolism. The reduction in food intake may also be a reflection of improved insulin and leptin action and signaling. Indeed, we observed decreased NPY and increased POMC gene expression, which may contribute to hypophagia. Chiang et al. (24) showed that IKKε KO mice on chow or HFD had elevated food intake. The difference in food intake from IKKε KO and inhibition of IKKε with CAY or siRNA may be due to the different timing of IKKε inhibition. IKKε KO mice had the IKKε gene deleted early in life and might have increased food intake to compensate for the elevated energy expenditure.

Despite elevated food intake, IKKε KO mice were lean. The leanness was mainly due to enhanced energy expenditure and thermogenesis (24). In the current study, we observed that CAY or siRNA was associated with increased energy expenditure and thermogenesis, which may have been caused by improved leptin signaling and action in the hypothalamus. We emphasize that the inhibition of IKKε increased energy expenditure and was probably an important contributor to the leanness. In the pair-feeding experiments, control animals ingested the same amount of food as mice treated with CAY. Nevertheless, CAY-treated mice had lower adiposity than the vehicle group. This phenomenon suggests that IKKε is important in the regulation of energy expenditure and deserves further investigation.

TNF-α, IL-6, and IL-1β are major enhancers of the inflammatory response (33). Evidence has demonstrated that these cytokines play a role in regulating energy metabolism. The present data show a reduction in serum TNF-α levels, suggesting a decrease in inflammatory signals. Decreased serum TNF-α levels might also have been due to the reduction of fat mass observed in the CAY-treated mice.

Knockdown of IKKε in the whole body decreased fasting blood glucose levels and decreased blood glucose in response to pyruvate (24,34). In contrast, overexpression of IKKε in hepatoma cells enhanced expression of selected gluconeogenic genes, suggesting direct and cell-autonomous effects in the liver. The present data show a similar result of inhibiting IKKε in the hypothalamus of HFD and db/db mice. This result on blood glucose was accompanied by lower PEPCK expression in the liver, suggesting a reduction of hepatic glucose output. Studies have shown that the action of insulin and leptin in the hypothalamus affects hepatic glucose production (35–37). Because db/db mice are leptin insensitive, we suggest that an improvement of insulin action in the hypothalamus induced by inhibition of IKKε might have been responsible for those phenotypes.

In summary, the data provide evidence that IKKε is expressed in neurons and upregulated in the hypothalamus of obese mice and contributes to insulin and leptin resistance. Blocking IKKε in the hypothalamus of obese mice, at least for a short period, decreases NF-κB activation, reducing IRS-1 serine phosphorylation in the hypothalamus. Inhibition of IKKε activity but not TBK1 decreases insulin and leptin resistance by improving the IR/IRS-1/Akt and JAK2/STAT3 pathways in the hypothalamus of obese mice. These improvements were independent of body weight and food intake. Increased insulin and leptin action and signaling in the hypothalamus may contribute to the reduction in adiposity and hypophagia and enhancement of energy expenditure accompanied by lower NPY and increased POMC mRNA levels. Improvement of insulin action decreases fasting glycemia, glycemia after pyruvate injection, and PEPCK protein expression in the liver, suggesting a reduction in hepatic glucose production. The results of this study suggest that the noncanonical IκB kinase IKKε may be a key inflammatory mediator in the hypothalamus of obese mice, and its hypothalamic inhibition meliorates energy and glucose metabolism.

Acknowledgments. The authors thank L. Janeri and J. Pinheiro (Department of Internal Medicine, UNICAMP, Campinas, São Paulo, Brazil) for technical assistance.

Funding. This work was supported by FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo): 2012/10338-6 and 2010/52068-0, CEPID (Centros de Pesquisa, Inovação e Difusão): 2013/07607-8, São Paulo, Brazil, and CNPq INCT (Instituto Nacional de Ciência e Tecnologia de Obesidade e Diabetes): 573856/2008-7 and UNIVERSAL: 481084/2013-4.

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

Author Contributions. L.W. and P.O.P. contributed to researching data and to the design and performance of the experiments, data analysis, discussion, and writing and review of the manuscript. P.G.F.Q., A.C.S., A.H.B.d.M., V.D.B.P., T.M.Z., G.C., D.G., J.M.d.S., and J.C.B. contributed to the performance of the experiments and data analysis. L.A.V., J.C.B., I.L.-C., and M.J.A.S. contributed to the discussion and writing and review of the manuscript. P.O.P. 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.