CDDO-Me Attenuates CA1 Neuronal Death by Facilitating RalBP1-Mediated Mitochondrial Fission and 4-HNE Efflux in the Rat Hippocampus Following Status Epilepticus

Ras-related protein Ral-A (RalA)-binding protein 1 (RalBP1, also known as Ral-interacting protein of 76 kDa (RLIP76) or Ral-interacting protein 1 (RLIP1 or RIP1)) is involved in the efflux of 4-hydroxynonenal (4-HNE, an end product of lipid peroxidation), as well as mitochondrial fission. In the present study, we found that 2-cyano-3,12-dioxo-oleana-1,9(11)-dien-28-oic acid methyl ester (CDDO-Me) attenuated CA1 neuronal death and aberrant mitochondrial elongations in these neurons coupled with enhanced RalBP1 expression and reduced 4-HNE levels following status epilepticus (SE). RalBP1 knockdown did not affect mitochondrial dynamics and CA1 neuronal death under physiological and post-SE conditions. Following SE, however, cotreatment of RalBP1 siRNA diminished the effect of CDDO-Me on 4-HNE levels, mitochondrial hyperfusion in CA1 neurons, and CA1 neuronal death. These findings indicate that CDDO-Me may ameliorate CA1 neuronal death by facilitating RalBP1-mediated 4-HNE efflux and mitochondrial fission following SE. Therefore, our findings suggest that increased RalBP1 expression/activity may be one of the considerable targets to protect neurons from SE.


Introduction
The morphology of mitochondria is regulated by a balance of fusion (elongation) and fission (fragmentation), which are referred to as mitochondrial dynamics. The maintenance of mitochondrial dynamics is essential for cell development and mitosis. Furthermore, the defects of this process lead to various types of cell death [1,2]. Large GTPase dynaminrelated protein 1 (DRP1) is one of the regulators of mitochondrial dynamics, which reversely acts by phosphorylation of distinct sites: phosphorylation of serine (S) 616 site on DRP1 facilitates mitochondrial fission, while S637 leads to fusion by detaching DRP1 from the outer mitochondrial membrane [2,3].
Status epilepticus (SE, uncontrolled prolonged seizure activities) shows a high mortality rate and causes secondary neurological deficits such as neuronal injury, alterations of neuronal networks, brain edema, neuroinflammation, and cognitive impairments [4,5]. In the hippocampus, CA1-3 pyramidal cells and dentate hilar neurons are vulnerable, while dentate granule cells are relatively invulnerable to SE [6,7]. In addition, SE induces distinct types of neuronal death in the heterogeneous patterns of mitochondrial dynamics: SE results in programmed necrosis of CA1 neurons accompanied by impaired mitochondrial fission and apoptosis of hilar parvalbumin (PV) cells (one of the GABAergic interneurons) by excessive mitochondrial fission without affecting optic atrophy 1 and mitofusin expression levels in the pilocarpine-induced SE rat model [2,8,9]. However, the underlying mechanisms of the regulation of mitochondrial dynamics leading to distinct neuronal death patterns have been elusive.
Here, we show that CDDO-Me facilitated mitochondrial fissions and attenuated CA1 neuronal death following SE. In addition, CDDO-Me enhanced RalBP1 expression in CA1 neurons concomitant with reduced 4-HNE levels following SE. Although RalBP1 knockdown did not affect mitochondrial dynamics and CA1 neuronal death under physiological and post-SE conditions, cotreatment of RalBP1 siRNA abated reductions in 4-HNE signals, CA1 neuronal degeneration, and mitochondrial elongation induced by CDDO-Me under post-SE conditions. These findings indicate that endogenous RalBP1 may basically participate in 4-HNE efflux but be insufficient to affect mitochondrial dynamics in neurons under physiological and post-SE conditions. At a higher expression level (induced by CDDO-Me), however, RalBP1 may accelerate mitochondrial fission in CA1 neurons following SE. Therefore, our findings suggest that CDDO-Me may protect CA1 neurons from SE insults by facilitating 4-HNE efflux and mitochondrial fission.

Experimental Animals and Chemicals
Male Sprague-Dawley (SD) rats (7 weeks old, n = 175, Daehan Biolink, South Korea) were used in the present study. Rats were housed under controlled conditions (22 ± 2 • C, 55 ± 5%, and a 12:12 light/dark cycle with lights) and accessed food and water ad libitum. Four animals died during experimental procedures, and three rats did not respond to pilocarpine treatment, which were ruled out in the present study. Animal experiments were allowed by the Institutional Animal Care and Use Committee of Hallym University (Chuncheon, Korea; No. Hallym 2021-30, approval date: 17 May 2021).

SE Induction and Electroencephalogram (EEG) Analysis
SE was induced by a single dose (30 mg/kg) of pilocarpine in rats pretreated (24 h before pilocarpine injection) with 127 mg/kg LiCl, as previously described [8,9,17,18]. Rats were pretreated with atropine methyl bromide (5 mg/kg i.p.) prior to pilocarpine injection. Two hours after SE, animals were administered diazepam (10 mg/kg, i.p.) to cease behavioral seizures and repeated as needed. As controls, rats were treated with atropine methyl bromide, followed by saline, but not pilocarpine. EEG signals of electrodeimplanted animals were recorded with a DAM 80 differential amplifier (0.1-3000 Hz bandpass; World Precision Instruments, Sarasota, FL, USA) during the 2 h recording session for each animal. The EEG signals were digitized and analyzed using LabChart Pro v7 (AD Instruments, New South Wales, Australia). Time of seizure onset was determined when rhythmic paroxysmal depolarizations (4-10 Hz) lasting >3 s with 2 times higher amplitudes than the basal EEG were detected. Two hours after SE onset, diazepam (Valium; Roche, France; 10 mg/kg, i.p.) was administered and repeated as needed.

Tissue Processing
Three days after SE induction, animals (n = 7 in each group) were deeply anesthetized with urethane (1.5 g/kg i.p.) and transcardially perfused with phosphate-buffered saline, followed by 4% paraformaldehyde. Brains were excised and then postfixed with the same fixative. Brains were cryoprotected with 30% sucrose and sectioned with a cryostat. For Western blot, the hippocampus was collected after decapitation under urethane anesthesia. After homogenization, protein concentration was determined with a BCA Protein Assay Kit (Pierce Chemical, Rockford, IL, USA).

Immunohistochemistry, Measurements of Mitochondrial Length, and Neuronal Damage
Sections were blocked in 10% normal goat serum (Vector, Burlingame, CA, USA) and subsequently incubated overnight at room temperature in primary antibodies (Table 1). Thereafter, sections were reacted with a fluorescein isothiocyanate-or Cy3-conjugated IgG (Vector, Burlingame, CA, USA) for 1 h. For negative control, preimmune serum was used to substitute for the primary antibody. Fluoro-Jade B (FJB) staining was performed to detect degenerating neurons based on the manufacturer's instructions [18]. In five hippocampal sections in an animal (five animals in each group), the randomly selected CA1 neurons (total of 25 cells in each group) were used for mitochondrial morphometry using AxioVision Rel. 4.8 and ImageJ software. The parameters for mitochondrial morphometry were as follows: area-weighted form factor (mitochondrial elongation index) = squared perimeter/4π; form factor (the transition of a mitochondrion to elongated and complex morphology) = squared perimeter/4π × area; cumulative area/perimeter ratio (the transition of an elongated mitochondrion to interconnected mitochondria) = Σarea/Σperimeter [19,20]. Fluorescent intensity was measured from 15 sections per animal and then normalized by setting the threshold levels. A blind test was applied to the measurement of mitochondrial length, fluorescent intensity, and cell count of FJB-positive neurons.

Western Blot
Western blot was performed by the standard protocol. Following electrophoresis, proteins were electrotransferred onto nitrocellulose membranes, blocked with 2% bovine serum albumin in Tris-buffered saline, and then incubated with primary antibodies (Table 1) overnight at 4 • C. After washing, membranes were reacted with secondary antibodies. Signals were detected by chemiluminescence using the ECL Western blotting kit (GE Healthcare Korea, Seoul, Korea). The bands were detected and quantified using an ImageQuant LAS4000 system (GE Healthcare Korea, Seoul, Korea) and normalized to β-actin.

Quantification of Data and Statistical Analysis
Data were analyzed by Shapiro-Wilk W-test to evaluate the values on normality followed by two-tailed Student's t-test and one-way repeated measure ANOVA. One-way ANOVA, followed by Bonferroni post hoc tests, was applied for multiple comparisons. A value of p < 0.05 was considered to be statistically significant.

CDDO-Me Facilitates Mitochondrial Fission in CA1 Neurons under Physiological and Post-SE Conditions
Since the dysregulation of mitochondrial dynamics leads to programmed necrosis in CA1 neurons following SE [8,9,18], we validated whether CDDO-Me affects mitochondrial dynamics in CA1 neurons under physiological and post-SE conditions.

RalBP1 siRNA Does Not Affect Mitochondrial Dynamics and CA1 Neuronal Death Induced by SE
To elucidate the role of RalBP1 in SE-induced aberrant mitochondrial dynamics in CA1 neurons, we applied RalBP1 knockdown prior to SE induction. As compared to control siRNA, RalBP1 siRNA did not influence the seizure susceptibility to pilocarpine (repeated one-way ANOVA; n = 7; Figure 4). Under physiological conditions, RalBP1 siRNA Consistent with our previous study [9], SE resulted in a 1.73-fold increase in mitochondrial length (F (1,48) = 11.9, p = 0.001 vs. vehicle-treated control animals, one-way ANOVA; Figure 3A,B). SE also increased the cumulative area/perimeter ratio to 2.16-fold of control animal levels (F (1,48) = 28.6, p < 0.001 vs. vehicle-treated control animals, one-way ANOVA), while it reduced the form factor to 0.54-fold of control animal levels (F (1,48) = 17.7, p < 0.001 vs. vehicle-treated control animals, one-way ANOVA; Figure 3A,C). These findings indicate that SE may lead to aberrant mitochondrial fusions and their aggregations in CA1 neurons. As compared to vehicle, CDDO-Me attenuated the SE-induced increases in mitochondrial length to 0.63-fold of vehicle-treated SE animal levels (F (1,48) = 8.5, p = 0.005 vs. vehicle-treated SE animal levels, one-way ANOVA; Figure 3A,B), accompanied by RalBP1 upregulation (Figures 2 and 3A). However, CDDO-Me also reduced the cumulative area/perimeter ratio to 0.65-fold of vehicle-treated animal levels (F (1,48) = 6.98, p = 0.01 vs. vehicle, one-way ANOVA) but increased the form factor to 2.3-fold of vehicle-treated animal levels (F (1,48) = 7.6, p = 0.008 vs. vehicle, one-way ANOVA) following SE ( Figure 3A,C). Taken together, our findings suggest that CDDO-Me-induced RalBP1 upregulation may abrogate the abnormal mitochondrial elongations induced by SE.

RalBP1 siRNA Does Not Affect Mitochondrial Dynamics and CA1 Neuronal Death Induced by SE
To elucidate the role of RalBP1 in SE-induced aberrant mitochondrial dynamics in CA1 neurons, we applied RalBP1 knockdown prior to SE induction. As compared to control siRNA, RalBP1 siRNA did not influence the seizure susceptibility to pilocarpine (repeated one-way ANOVA; n = 7; Figure 4). Under physiological conditions, RalBP1 siRNA reduced RalBP1 protein levels 0.67-fold of control siRNA levels in the whole hippocampus (t (12) = 6.6, p < 0.001 vs. control siRNA-treated control animals, Student t-test, n = 7; Figure 5A,B and Supplementary Figure S1). RalBP1 knockdown also did not affect mitochondrial dynamics under physiological conditions ( Figure 5C-E). Furthermore, RalBP1 siRNA did not influence the aberrant mitochondrial fusion and CA1 neuronal loss induced by SE ( Figure 5C-F), although it decreased RalBP1 protein levels 0.64-fold of control siRNA levels in the whole hippocampus (t (12) = 6.6, p < 0.001 vs. control siRNA-treated control animals, Student t-test, n = 7; Figure 5A,B). These findings indicate that endogenous RalBP1 may not be involved in mitochondrial dynamics under physiological conditions. Considering the effects of CDDO-Me on mitochondrial dynamics and RalBP1 expression following SE, however, it is likely that SE-induced RalBP1 downregulation may lead to aberrant mitochondrial elongations in CA1 neurons. To confirm this, we applied cotreatment of RalBP1 siRNA with CDDO-Me prior to SE induction.

RalBP1 Knockdown Diminishes the Effects of CDDO-Me on Aberrant Mitochondrial Dynamics and CA1 Neuronal Death Following SE
In the present study, cotreatment of RalBP1 siRNA with CDDO-Me did not change the seizure susceptibility to pilocarpine, as compared to cotreatment of control siRNA with CDDO-Me (repeated one-way ANOVA; n = 7; Figure 6A,B). Under physiological conditions, RalBP1 siRNA cotreatment reduced RalBP1 protein levels 0.63-fold of control siRNA levels in the whole hippocampus (t (12) = 9.9, p < 0.001 vs. control siRNA cotreatment, Student t-test, n = 7; Figure 6C,D and Supplementary Figure S1). Following SE, RalBP1 siRNA cotreatment also decreased RalBP1 protein levels 0.68-fold of control siRNA levels in the whole hippocampus (t (12) = 7.4, p < 0.001 vs. control siRNA cotreatment, Student t-test, n = 7; Figure 6C,D and Supplementary Figure S1). These findings indicate that RalBP1 knockdown may effectively diminish CDDO-Me-induced RalBP1 upregulation following SE.
Antioxidants 2022, 11, x FOR PEER REVIEW 9 of 5A,B and Supplementary Figure S1). RalBP1 knockdown also did not affect mitochondri dynamics under physiological conditions ( Figure 5C-E). Furthermore, RalBP1 siRNA di not influence the aberrant mitochondrial fusion and CA1 neuronal loss induced by S ( Figure 5C-F), although it decreased RalBP1 protein levels 0.64-fold of control siRNA le els in the whole hippocampus (t(12) = 6.6, p < 0.001 vs. control siRNA-treated control an mals, Student t-test, n = 7; Figure 5A,B). These findings indicate that endogenous RalBP may not be involved in mitochondrial dynamics under physiological conditions. Consid ering the effects of CDDO-Me on mitochondrial dynamics and RalBP1 expression follow ing SE, however, it is likely that SE-induced RalBP1 downregulation may lead to aberran mitochondrial elongations in CA1 neurons. To confirm this, we applied cotreatment RalBP1 siRNA with CDDO-Me prior to SE induction.

Discussion
The major findings in the present study are that CDDO-ME ameliorated SE-induced aberrant mitochondrial dynamics by RalBP1 upregulation, although RalBP1 alone did not affect mitochondrial dynamics under physiological conditions.
RalBP1 is a multifunctional protein showing transport activity. RalBP1 also fosters DRP1 S616 phosphorylation. Indeed, RalBP1 knockdown reduces the amount of DRP1 S616 phosphorylation and inhibits mitochondrial fragmentation [3]. However, the present data show that RalBP1 knockdown did not alter mitochondrial length in CA1 neurons under physiological conditions and following SE. Considering very low RalBP1 protein levels, these findings indicate that RalBP1 may not participate in mitochondrial fission in CA1 neurons under both conditions. In the present study, SE evoked the defects of mitochondrial fission in CA1 neurons and massive neuronal CA1 neuronal death without altering RalBP1 protein levels. However, CDDO-Me effectively attenuated SE-induced CA1 neuronal death concomitant with reduced mitochondrial length and RalBP1 upregulation, although CDDO-Me did not influence RalBP1 expression levels under physiological conditions. Furthermore, RalBP1 siRNA cotreatment abrogated CDDO-Me-induced mitochondrial fragmentation in CA1 neurons under post-SE conditions. Therefore, these findings suggest that CDDO-Meinduced RalBP1 upregulation may facilitate mitochondrial fission following SE.
RalBP1 also transports the oxidized lipid byproducts, including 4-HNE. Indeed, impaired RalBP1 function cells exert the accumulation of 4-HNE. Therefore, RalBP1 plays an important role in the stress defense mechanism [10,11,21]. In the present study, SE increased RalBP1 expression in dentate granule cells that are invulnerable to SE [22,23]. Furthermore, 4-HNE signals were very low in these cells following SE. However, the present study demonstrates that SE induced 4-HNE signals in CA1 neurons, which was ameliorated by CDDO-Me-induced RalBP1 upregulation. Furthermore, RalBP1 siRNA cotreatment diminished the effect of CDDO-Me on 4-HNE induction and CA1 neuronal death following SE. These findings indicate that RalBP1 expression may be inversely relevant to SE-induced 4-HNE synthesis, and that CDDO-Me may also increase CA1 neuronal viability by enhancing RalBP1-mediated 4-HNE efflux following SE. Therefore, SE-induced RALBP1 upregulation in dentate granule cells may be involved in 4-HNE efflux as an adaptive response to oxidative stress. Considering the antioxidant properties of CDDO-Me [16,17], it is postulated that CDDO-Me would attenuate abnormal mitochondrial elongation by reducing oxidative stress following SE. If it is true, CDDO-Me would abrogate SE-induced RalBP1 upregulation in dentate granule cells that are resistant to SE insults [22,23]. However, the present study shows that CDDO-Me greatly increased RalBP1 protein levels in dentate granule cells under post-SE conditions. Therefore, our findings indicate that acceleration of mitochondrial fission induced by CDDO-Me may not be a result of RalBP1-mediated reductions in oxidative stress but a direct result of RalBP1 upregulation.
RalBP1 knockout mice show lower blood glucose levels than wild-type mice [24], which also triggers mitochondrial fission in CA1 neurons [25]. However, the present data show that RalBP1 siRNA cotreatment alleviated mitochondrial fission induced by CDDO-Me following SE, although RalBP1 siRNA alone could not affect mitochondrial dynamic under physiological conditions and following SE. Therefore, it is plausible that CDDO-Meinduced RalBP1 upregulation may facilitate mitochondrial fragmentations, independent of glucose metabolism.
Although the underlying mechanisms of CDDO-Me-induced RalBP1 upregulation have still been unclear, it is considerable that nuclear factor-κB (NF-κB) and p300 transactivate RalBP1 expression [26,27]. RalBP1 is a target of proinflammatory molecules (such as tumor-necrosis factor-α) via a cluster of evolutionarily conserved NF-κB binding sites [27]. However, it is well known that CDDO-Me inhibits NF-κB-mediated signaling pathways [28][29][30][31]. Therefore, it is unlikely that CDDO-Me would increase RalBP1 expression by activating the NF-κB signaling pathway. On the other hand, p300 knockdown decreases RalBP1 expression, indicating that p300 regulates RalBP1 promoter activity and its expression [26]. Interestingly, p300 promotes Nrf2 transcriptional activity, thereby promoting Nrf2 nuclear localization, since p300 competes with Kelch-like erythroid cellderived protein with CNC homology-associated protein 1 (Keap1), a repressor of Nrf2, for NRF2 binding [32]. Thus, it is likely that p300 may act as an endogenous Nrf2 activator. Indeed, p300 overexpression increases cell viability against oxidative stress [32]. Similarly, CDDO-Me exerts the release of Nrf2 from Keap1, which hinders its ubiquitination and increases Nrf2 activity [33]. Under these conditions, CDDO-Me may provide an opportunity to increase the binding of p300 to RalBP1 promoter sites. Therefore, it is likely that CDDO-Me may increase RalBP1 protein expression by enhancing p300 transcriptional activity under pathophysiological conditions. Further studies concerning the effect of CDDO-Me on p300-mediated RalBP1 transactivation are needed in the near future.
On the other hand, the present data demonstrating the mitochondrial hyperfusionmediated CA1 neuronal death under post-SE conditions are opposite to those under other pathophysiological conditions. Unlike other neurological diseases, including Alzheimer's disease, cerebrovascular disease (stroke), Huntington's disease, and Parkinson's disease [2], seizure activity leads to dysregulation of mitochondrial fission coupled with impairments of their distribution and function [34,35]. Indeed, DRP1 pathogenic variant-mediated mitochondrial elongation is relevant to seizures in human patients [36]. In particular, fibroblasts obtained from refractory epilepsy patients show the defects of DRP1-mediated mitochondrial fission without altering bioenergetics [37]. In previous studies [8,38], we have also reported that the increased mitochondrial fission induced by WY14643 protects CA1 neurons from SE, whereas the mitochondrial elongation by Mdivi-1 exacerbates SEinduced CA1 neuronal death. Therefore, aberrant mitochondrial fusion may be involved in the pathogenesis of SE-induced neuronal death.

Limitation of the Study
Lucchi et al. [39] reported that pilocarpine-induced SE leads to neuronal degenerations, including hilus interneurons and entorhinal cortical neurons concomitant with CA1 neurons. In previous studies [29,38,[40][41][42], we have also reported that pilocarpine-induced SE develops neuronal damage in the hilus of the hippocampus, entorhinal cortex, and piriform cortex. Unlike CA1 neurons, however, CDDO-Me does not protect hilus interneurons, although it decreases 4-HNE levels in these neurons following SE [18]. Therefore, SE-induced hilus neuronal death is irrelevant to RalBP1-mediated 4-HNE efflux. Indeed, we have revealed that SE results in hilus neuronal damage by excessive mitochondrial fragmentation through the p47Phox/cyclin-dependent kinase 5/dynamin-related protein 1 signaling pathway [43] and mitochondrial translocation of caspase-3 [38], which are not involved in SE-induced CA1 neuronal death. Furthermore, SE induces severe vasogenic edema in the entorhinal and piriform cortices, which is limited in the hippocampus [42]. Since SE-induced vasogenic edema formations in these regions are prior to neuronal degeneration and the prevention of vasogenic edema attenuates neuronal death in these regions [44,45], vasogenic edema may be one of the underlying mechanisms in the neurodegeneration in the entorhinal and piriform cortices. CDDO-Me also ameliorates neuronal death in the piriform and entorhinal cortices and increases astroglial viability by inhibiting NF-κBmediated vasogenic edema formation [29]. In addition, 4-HNE upregulation is restricted to astrocytes concomitant with vasogenic edema formation in the piriform cortex [46]. Therefore, it is likely that CDDO-Me may mitigate SE-induced neuronal degeneration in the entorhinal and piriform cortices by inhibiting vasogenic edema formation, but not activating RalBP1-mediated 4-HNE efflux from cortical neurons. For these reasons, we focused on CA1 neurons to evaluate the effects of CDDO-Me and RalBP1 siRNA on 4-HNE levels during SE-induced neuronal death in the present study and suggest that the underlying mechanisms of SE-neuronal degeneration may be complicated and heterogenous.
On the other hand, Lucchi et al. [39] applied diazepam treatment 10 min after the pilocarpine-SE onset to improve survival and to standardize SE duration based on the paper of Gualtieri et al. [47]. The early diazepam administration rapidly reduces motor seizure severity (tonic seizures) without altering subtle convulsive movements, such as clonic jerks, whiskers or eyelids, and tremors of the whole body. In addition, this early diazepam treatment dampens, but does not abolish, electrocorticographic seizure activity [47]. In the present study, we applied a 2 h lasting SE rat model. In this model, diazepam treatment at 2 h after SE onset effectively decreases pilocarpine-induced seizure activities on hippocampal EEG, as well as behavioral seizures [48,49]. Therefore, our model provides a long observation to evaluate the anticonvulsive effects of test chemicals without the interference effect of diazepam, as compared to the model of Lucchi et al. [39]. Most of all, Gualtieri et al. [47] reported that the results obtained from the early diazepam-treated model are consistent with those of our 2 h lasting SE model, citing our papers [40,42,44]. As compared to vehicle and control siRNA, the present study shows that CDDO-Me and RalBP siRNA did not affect the seizure latency and EEG power (seizure severity) induced by pilocarpine during 2 h EEG recording. Considering that EEG recording is more reliable to quantify the seizure onset and its severity than the observation of behavioral seizures, our results indicate that CDDO-Me or RalBP siRNA may not have anticonvulsive properties and may regulate CA1 neuronal death without altering pilocarpine-induced SE activity. Further studies are needed to validate the efficacies of diazepam treatment at different time points after SE onset for evaluating anticonvulsants.

Conclusions
In the present study, we found that CDDO-Me upregulated RalBP1 expression, accompanied by a decrease in 4-HNE signals in CA1 neurons following SE. CDDO-Me reduced mitochondrial length in these neurons under physiological and post-SE conditions. Although RalBP1 knockdown alone did not affect mitochondrial dynamics and CA1 neuronal death induced by SE, RalBP1 siRNA cotreatment abrogated the effects of CDDO-Me on aberrant mitochondrial dynamics and CA1 neuronal death following SE. Therefore, our findings indicate that CDDO-Me may ameliorate CA1 neuronal degeneration by accelerating RalBP1-mediated mitochondrial fission and 4-HNE efflux following SE.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.