Pharmacological modulation of LMNA SRSF1-dependent splicing abrogates diet-induced obesity in mice

Bakground/Objectives:Intense drug discovery efforts in the metabolic field highlight the need for novel strategies for the treatment of obesity. Alternative splicing (AS) and/or polyadenylation enable the LMNA gene to express distinct protein isoforms that exert opposing effects on energy metabolism and lifespan. Here we aimed to use the splicing factor SRSF1 that contribute to the production of these different isoforms as a target to uncover new anti-obesity drug.Subjects/Methods:Small molecules modulating SR protein activity and splicing were tested for their abilities to interact with SRSF1 and to modulate LMNA (AS). Using an LMNA luciferase reporter we selected molecules that were tested in diet-induced obese (DIO) mice. Transcriptomic analyses were performed in the white adipose tissues from untreated and treated DIO mice and mice fed a chow diet.Results:We identified a small molecule that specifically interacted with the RS domain of SRSF1. ABX300 abolished DIO in mice, leading to restoration of adipose tissue homeostasis. In contrast, ABX300 had no effect on mice fed a standard chow diet. A global transcriptomic analysis revealed similar profiles of white adipose tissue from DIO mice treated with ABX300 and from untreated mice fed a chow diet. Mice treated with ABX300 exhibited an increase in O2 consumption and a switch in fuel preference toward lipids.Conclusions:Targeting SRSF1 with ABX300 compensates for changes in RNA biogenesis induced by fat accumulation and consequently represents a novel unexplored approach for the treatment of obesity.


Introduction
Obesity, defined as an imbalance of income and expenditure of energy, is a rapidly growing epidemic worldwide. Obesity is characterized by increased fat mass due to energy storage in WAT. In addition, obesity is associated with a collection of clinical problems known as the metabolic syndrome, including insulin resistance, diabetes, cardio-vascular disease, dyslipidemia, and fatty liver 1 . Owing to changes in the diet in Western countries and an increasingly "aged" population, metabolic syndrome is an epidemic expected to double in incidence in the near future 2 . Dietary management and exercise are not usually successful as an intervention, underscoring the need for efficient medication to treat metabolic disorders 3 . Integrated metabolic networks, which are governed at the transcriptional level by transcription factors and co-regulators, enable the organism to adapt the metabolic state of different organs based on nutrient availability 4 . The limited success to date in the pharmacotherapy of obesity likely reflects the existence of multiple redundant and compensatory pathways in energy homeostasis 5 .
Intense drug discovery efforts in the metabolic field highlight the need for novel strategies for the treatment of obesity. Compelling data suggest that targeting cellular bioenergetics may provide an exciting new therapeutic approach for the treatment or prevention of this disorder 6 .
RNA processing of the LMNA gene was recently reported to generate different isoforms with opposing effects on the energy expenditure (EE) and lifespan of mice 7 . The alternative RNA processing of LMNA pre-mRNA produces three nuclear protein isoforms: lamin A, progerin, and lamin C.
Mutations in lamin A/C can cause several syndromes collectively described as laminopathies, which include several lipodystrophies and progeroid syndromes 8 . The lipodystrophy mutations strongly suggest that lamins can have cell-type-specific functions 9 . Moreover, a de novo silent mutation in LMNA (c.1824C>T; p.Gly608Gly) results in the aberrant AS of the lamin-A isoform, which gives rise to a truncated isoform of the protein termed progerin 10 . The exacerbated accumulation of progerin is the predominant cause of Hutchinson-Gilford Progeria Syndrome (HGPS), a severe form of early-onset premature aging associated with lipodystrophy 11,12 . In HGPS, a loss of adipose tissue correlates with

Luminescence assay
HEK-293 FRT-LMNA-Luc cells were seeded 48 h prior to luminescence lecture at 10,000 cells per well. The culture medium was then removed, and the cells were washed once with sterile DPBS. The cells were lysed using 1X Passive Lysis Buffer (E1941, Promega, Madison, WI, United States), which was directly added to the cells. Half of the cell lysate was then mixed with Luciferase Assay Reagent (E1501, Promega, Madison, WI, United States), and the luciferase activity in each well was measured within 1 second. The remaining cell lysate was used for protein normalization in the well (Pierce™ 660 nm Protein Assay Reagent).

Immunofluorescence analysis
SRSF1-GFP HeLa cells were maintained in Dulbecco"s modified Eagle"s medium (DMEM, Invitrogen) containing 10% heat-inactivated fetal bovine serum (FBS, PAA), 2 mM L-Glutamine, and 1% penicillin and streptomycin (all from Thermo Fisher) and untreated (DMSO), treated with 5 M of either ABX300 or ABX460. The cells were fixed in 4% (w/v) paraformaldehyde for 20 min. The nuclei were stained with Hoechst 33342 (Sigma). The cells were washed in PBS and water, mounted with DAKO mounting medium, and observed under a fluorescence microscope. Cell imaging was performed with a confocal Leica SP5-SMD. The images were processed using the FiJi software.

Fluorimetric titrations
The fluorimetric titrations were performed at a constant dye concentration (ABX300) (1.2 µM) with increasing concentrations of SRSF1 in 0.6 mM guanidine buffer (up to 120 µM). The fluorescence spectra were recorded on a FluoroMax-3 (Jobin Yvon) at room temperature. The measurements were performed with solutions of OD <0.1 to avoid reabsorption of the emitted light, and the data were corrected with a blank and based on variations in the detector with the emitted wavelength. The binding curves were obtained by plotting the fluorescence enhancement F/F0 (F= integrated fluorescence area of Comment citer ce document : Santo , J., Lopez-Herrera, C., Apolit , C., Bareche , Y., Lapasset, L., Chavey, C., Capozi , S., Mahuteau , F., Najman , R., Fornarelli , P., Lopez-Mejia, I. C., Béranger , G., Casas, F., Amri, E. Z., Pau, B., Scherrer, D., Tazi, J. (2017). Pharmacological modulation of LMNA SRSF1-dependent splicing abrogates diet-induced obesity in mice. International Journal of Obesity, 41 the complex and F0 = integrated fluorescence area of the free dye) versus the concentration of SRSF1.

Animals and ethics statement
All animal procedures were executed according to European Directive 2010/63/UE. The mice were maintained under pathogen-free conditions in our animal facility (E34-172-16), and the experiments were conducted by authorized personnel. The study plan was approved by the Institutional Review Board at the Animal Facility of the Institut de Génétique Moléculaire de Montpellier and the Regional Ethics Committee for Animal Experimentation of Languedoc-Roussillon (agreement n° CEEA-LR-1061).

Animal feeding and treatment
Six-week-old male C57BL/6J mice were purchased from Charles River Laboratory (69592 L'Arbresle) and housed under a normal light cycle (12 h light-dark) and a constant temperature of 22°C, with free access to food and water. The experiments began after one week of acclimatization to our installations under CD conditions. At this point, mice either continued to receive the CD (Safe, 3% energy from fat) or were switched to an HFD (Research Diet 58Y1, 60% energy from fat). For the curative scheme, mice were fed for up to ten weeks to generate the DIO model and were then treated with the text compound.
The mice were weighed weekly, and food intake was measured four times per week. For the preventive scheme, the treatment started along with HFD feeding. For both nourishment conditions, the mice were divided into 2 groups of treatment: half of the mice received the vehicle (Labrafil M1944CS -5% DMSO), and the other half received the active compound ABX300 at 50 mg/kg. For both conditions the selected route of administration was intragastric to accurately manage the delivered dose.

Synchronized pair-feeding test
DIO mice were divided into three groups and housed by experimental class. HFD untreated and ABX300-treated animals were provided ad libitum access to food, whereas in the untreated pair-fed (PF) group, the food was calorie restricted. These animals received the same amount of aliment consumed by © 2016 Macmillan Publishers Limited. All rights reserved.

Organ collection
After one month of treatment, the mice were anesthetized by intraperitoneal injection (xylazine/ketamine combination) and euthanized by cervical dislocation. The organs were collected, rapidly weighed and immediately snap-frozen in liquid nitrogen or fixed in 4% paraformaldehyde.

Histological analysis
Organs fixed in 4% paraformaldehyde were dehydrated, embedded in paraffin and stained with HE for histological analysis. The slides were scanned with the NanoZoomer equipment (HAMAMATSU).
Virtual slides were analyzed with the NDP.view software and Image J, and the MRI Adipocyte Tool was used to measure the size of WAT cells.

Micro-computed tomography
Adipose tissue quantification was performed with a SkyScan-1178 X-ray micro-computed tomography system as previously described (Lopez-Mejia et al, 2014).

Total RNA extraction
To extract total RNA, 30 mg of WAT was ground in liquid nitrogen, and the powder was dissolved in TRI reagent (SIGMA) according to the manufacturer"s protocol, followed by DNAse digestion and phenol-chloroform-isoamyl alcohol purification. The RNA quantity and integrity were controlled using a Nanodrop ND-1000 (Thermo Scientific) and an Agilent Bioanalyzer (Agilent Technologies), respectively.

miRNA array experiments
Small non-coding RNA transcripts were assessed using an Affymetrix assay, in which RNA was hybridized to Affymetrix miRNA v3.0 arrays (GenoSplice-Evry-France).

Mitochondrial target prediction analysis
© 2016 Macmillan Publishers Limited. All rights reserved. Putative miRNA targets were identified using the mirDB online database (http://mirdb.org/miRDB/), which is a database for miRNA target prediction and functional annotations. Targets are predicted by an algorithm (MirTarget) that analyzes all NCBI RefSeq 3'-UTRs in order to predict both conserved and non-conserved target motifs. A target prediction score is also computed to inform the user about the confidence of the prediction. The range is from 50 to 100, but caution is recommended for values below 80. Because the microarray analysis was based on a previous version of mirBase (http://www.mirbase.org/) (v.17), the selected miRNA names were updated based on the current mirBase version (v.21). The miRNAs were then compared against the mirDB database in order to retrieve all predicted targets with a prediction score ≥ 80. Using the MitoCarta database (https://www.broadinstitute.org/pubs/MitoCarta/), the selected target genes were then filtered in order to exclusively retrieve specific mitochondrial target genes. The two heat maps presented in Figures 5 A and B were generated using a custom R script (https://cran.r-project.org/) containing gplots (https://cran.rproject.org/web/packages/gplots/gplots.pdf). The data were clustered using the Pearson correlation coefficient method and the Heatmap2 function (http://www.insider.org/packages/cran/gplots/docs/heatmap.2).  DIO mice were switched to ad libitum CD feeding one week before the metabolic chamber experiments.

Reverse-transcription and PCR
Single-housed animals were allowed to acclimate for 24 h prior to recording. Immediately before the run, the animals were weighed and treated. Oxygen consumption (VO₂) and carbon dioxide production (VCO₂) were monitored every 2 min over a 24-h period using a Comprehensive Lab Animal Monitoring System (Columbus Instruments, Columbus, OH) controlled at 22°C. The RER corresponds to the VCO₂/VO₂ ratio.

Statistical analysis
All results are expressed as the mean +/− standard error mean (S.E.M) or standard deviation (SD). The significance of differences was determined with the appropriate statistical test, including Student"s t test, Mann-Whitney test, analysis of variance (ANOVA) or Kruskal-Wallis test, with significance at * P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001.
© 2016 Macmillan Publishers Limited. All rights reserved.

Identification of ABX300
Based on the IDC16 structure, an indole derivative that specifically targets the SR protein SRSF1 and inhibits HIV multiplication in vitro 18 , a new library was developed by ABIVAX to assess the toxic effect of IDC16 owing to its planar structure. Derivatives with quinoxaline and quinoline backbones were synthesized in order to disrupt the planar structure of IDC16 (Fig. S1A). The interaction of IDC16 with SRSF1 required its carboxyl-terminal serine-arginine-rich (RS) domain, which is involved in the ESE-dependent splicing activity and the localization of SRSF1. To avoid the selection of molecules that would generally affect AS because of SRSF1 mislocalization, CRISPR-edited HeLa cells expressing an endogenous GFP-SRSF1 fusion protein were used to select molecules that did not disturb the GFP distribution ( Fig. 1A). After the elimination of all molecules that induced the mislocalization of SRSF1, a primary screen was conducted with 5 µM of each selected molecule, which was added in triplicate to an HEK-293FRT reporter cell line (HEK-293FRT-LMNA-Luc) stably expressing luciferase fused with a portion of the LMNA gene under the control of the CMV promoter (Fig. 1B). This cell line contained the main mutation responsible for HGPS (c.1824C>T; p.Gly608Gly) and was consequently used as a reporter for the specific use of the LMNA exon 11 5" splice sites (5"SS); the authentic 5" splice site produces Lamin A, and the progerin 5" splice produces progerin, and both sites are under the control of SRSF1, as previously mentioned. In this assay, the inhibition of SRSF1 should lead to the selection of an authentic splice site 14 and a consequent increase in luciferase activity.
LMNA isoforms have been shown to play important opposite roles in the regulation of the metabolic adaptations of adipose tissue in mammals. Therefore, in a secondary screen, selected molecules representative of different structures were tested for their ability to induce weight changes in high-fat diet-induced obese mice (Fig. S1B). Among the 22 molecules tested, one lead compound, ABX300, was identified. The binding of ABX300 to SRSF1 was determined using the intrinsic fluorescence of the drug, following a classical fluorimetric titration protocol. Figure 1C shows that © 2016 Macmillan Publishers Limited. All rights reserved. increasing concentrations of SRSF1 enhanced the ABX300 fluorescence, demonstrating a direct interaction between SRSF1 and ABX300 with a stoichiometric ratio of 1:1. As a negative control, we used hnRNPK protein, which does not bind to ABX300 (Fig. S1C). To examine the potential requirement of the RS domain for ABX300 binding, we used mutant protein in which the entire Cterminal RS domain was deleted. No binding was observed with the mutant protein, indicating that the integrity of the SRSF1 structure is required for efficient binding (Fig. 1C).

ABX300 modulates weight gain under a high-fat diet
To further study the effects of ABX300 on energy metabolism, DIO mice were treated with this compound by gavage, either subsequently or simultaneously to high-fat diet (HFD) condition. The average treatment length was one month, and the selected dose was 50 mg/Kg. This dose was the highest non-toxic dosage according to the maximal tolerated dose test in non-obese mice (Fig. S2A). In a curative context, ABX300 enhanced weight loss ( Fig. 2A,  However, a decrease in food intake was observed in DIO mice after only the five first days of treatment, and the amount of ingested food remained lower than in untreated animals over the period of treatment ( Fig. 2A right panel, Mann-Whitney p<0.001, U=17). Cumulative food intake for the entire treatment period (shown at the bottom of the panels) significantly differed between treated and untreated animals. Nevertheless, the effects on body mass were stable and continuous (Fig. S2B, Kruskal-Wallis followed by Dunn's post test p<0.0001). To explore the contribution of reduced food intake to drug-Comment citer ce document : Santo , J., Lopez-Herrera, C., Apolit , C., Bareche , Y., Lapasset, L., Chavey, C., Capozi , S., Mahuteau , F., Najman , R., Fornarelli , P., Lopez-Mejia, I. C., Béranger , G., Casas, F., Amri, E. Z., Pau, B., Scherrer, D., Tazi induced reductions in body weight, we performed a pair-feeding experiment. The amount of food eaten by the drug treatment group each day was determined, and this amount was provided to a vehicle-treated pair-fed group of animals housed under identical conditions the following day ( Fig. 2D  HFD untreated and Pair-fed q=2.386; F=3.402) shows that ABX300 treatment produced greater weight loss than that observed in the corresponding pair-fed group (left and right panels), implying that ABX300 treatment led to both reduced food intake and increased energy output. These effects are dose dependent because treatment of DIO mice with lower dose of ABX300 (20 mg/kg) did not induce any changes in food intake and weight compared to untreated mice ( Fig.S2 B and C) Importantly, no toxicity was observed in ABX300-treated animals following daily gavage. Up to 50 mg/kg, no significant weight loss was observed in lean mice treated compared to non treated mice ( Fig.S2A) whereas in high fat diet-induced obese mice, treatment with ABX300 at 50 mg/kg decreased body weight (Fig.S2B). Histological examination confirmed the absence of toxicity but the loss of large adipocytes in the metabolic organs (WAT, BAT, liver and pancreas) and skin, between obese and obese mice treated for 65 days with ABX300 ( Fig.S3 A and B). These results make unlikely the possibility that the weight loss is due tissue toxicity. Oral delivery of ABX300 did not change significantly the values of hematocrit (Fig.S2D, Kruskal-Wallis followed by Dunn's post test p>0.05) and lactate blood level was also similar between non treated and treated animals (Fig.S2E, Kruskal-Wallis followed by Dunn's post test p>0.05), revealing no hypoxic regions and no defect in liver clearance. Finally, no mark of inflammation was observed since blood resistin level did not change between treated and not treated animals (Fig.S2F, Mann-Whitney p=0.0553, U=9).

ABX300 decreases adiposity in DIO mice
© 2016 Macmillan Publishers Limited. All rights reserved. The autopsy examination of adipose tissue showed a significant reduction in the WAT size in ABX300 DIO-treated mice compared with untreated mice (Fig. S3A). Specifically, the adipose mass was five times lower in treated mice, whereas the masses of all other organs remained constant (Fig. 3A,   Mann-Whitney p=0.0004, U=0). To obtain a more comprehensive view of the fat distribution in these mice, we performed computed tomography. Based on an evaluation of both the subcutaneous and intraabdominal adipose tissue volumes, all adipose tissue depots appeared to be equally affected by ABX300 treatment (Fig. 3B). The total fat volumes were calculated for each group, and ABX300 treatment was found to induce a 5.2-fold decrease in the adipose tissue volume compared with untreated mice (Fig.   3C, Kruskall-Wallis followed by Dunn's post test p<0.01). Remarkably, ABX300 DIO-treated mice were indistinguishable from untreated animals receiving a CD ( Fig. 3B and 3C).
To verify whether enhanced lipolysis was involved in the observed effect of ABX300, we assessed glycerol, non-esterified fatty acids (NEFA) and triglycerides (TGs) in the serum of ABX300treated and untreated animals. None of these parameters differed between the groups (Fig. S3C). While the treatment had no effect on glycemia (Fig.S3E, Kruskall-Wallis followed by Dunn's post test p>0.05), it significantly enhanced the level of circulating Adiponetin (Fig.S3D, Mann-Whitney p=0.012,   U=4). In addition, studies on differentiated adipocytes derived from the 3T3-L1 cell line (Fig. S4A, S4B and 4C), showed ABX300 treatment does not prevent differentiation or induce lipolysis.

Transcriptomic profiling of changes induced by ABX300
To analyze the profile program of AS changes occurring in the adipose tissue of treated and untreated DIO mice, we designed a large-scale screening strategy using high-throughput PCR on a preexisting panel of 1,328 alternatively spliced events (ASEs) 20 . An analysis of the data allowed further stringent quality control measures; ASEs were only considered if >75% of the products ran at the expected motilities and if the total expected PCR concentration exceeded 20 nM. This filter ensured the use of high-quality PCRs for well-expressed genes and led to the use of 536 high-quality PCRs to conduct our adipose tissue analysis (Table S1). We used these high-quality PCRs to study the ASEs expressed in the WAT of ABX300-treated mice and compare them to untreated obese mice. The splicing profiles of pools of RNA extracted from 4 treated and 4 untreated WAT samples allowed the identification of 40 ASEs that varied by more than 20% between the two conditions (Table S2). Among the 40 selected ASEs, only 12 were reproducibly found in individual samples of each mouse (Fig. 4A and Table S3). Manual end-point PCRs are shown for the most robust changes in Myo9b02 and Pdlim5 (Fig. 4B). To test whether these splicing changes were affected by the HFD and/or ABX300 treatment, they were compared to the splicing profile in the adipose tissue from mice receiving a CD. Figure 4B shows that the splicing profiles of both Myo9b and Pdlim5 from CD mice were similar to the ones from DIO mice treated with ABX300, implying that the treatment reverses the splicing events altered by HFD and obesity.
Because variations in gene expression can also directly or indirectly contribute to the phenotype induced by ABX300, we used a dedicated qPCR array from QIAGEN SABiosciences; the Mouse Adipogenesis RT² Profiler™ PCR Array profiles the expression of 84 key genes involved in the differentiation and maintenance of mature adipocytes (Table S4). Among the 84 genes tested, 21 (25%) were significantly modulated in treated obese mice compared with untreated mice (Fig. 4C transcriptional differences were also observed between DIO mice and CD mice (Fig. 4D, The p values

are calculated based on a Student's t-test of the replicate 2^(-Delta Ct) values for each gene in the
control group and treatment groups, and p values over than 0.05 are highlighted), suggesting that ABX300 treatment restores the transcriptional program modified by an HFD. The comparison between lean mice and treated obese mice confirmed this prediction. With the exception of Shh (Sonic Hedgehog) and Gata3 (Trans-acting T-cell-specific transcription factor GATA-3), which were upregulated in treated DIO mice compared with CD untreated mice, all other transcripts perfectly correlated (Fig. 4E, The p

values are calculated based on a Student's t-test of the replicate 2^(-Delta
Ct) values for each gene in the control group and treatment groups, and p values over than 0.05 are highlighted). Again, this finding reinforced the notion that ABX300 restores the health of adipose tissue.
We next focused on expression of lamin A/C in obese and lean mice and found that both isoforms were downregulated specifically in WAT of obese mice but not in BAT, spleen and kidney ( Fig. S5A and S5B). Interestingly, in ABX300-treated mice, both lamin A and lamin C were upregulated in WAT, BAT but also in kidney and liver (Fig. S5C, S5D, S5E and S5F). Again, this upregulation could be a marker of a healthy tissue.

ABX300 modulates the expression specific microRNAs
MicroRNAs (miRNAs) constitute a growing class of non-coding regulatory RNAs that primarily regulate gene expression by reducing mRNA stability and/or repressing translation 21 . Several miRNAs exhibit tissue-specific expression, and those present in WAT are known to have crucial implications for the regulatory networks underlying adipogenesis and adipose dysfunction in obesity 22 . The biogenesis of microRNA can be altered both at the transcriptional and post-transcriptional levels. SRSF1 recognizes the stem region of specific pri-miRNAs, resulting in enhanced cleavage by Drosha 23 . Furthermore, the microprocessor complex responsible for microRNA biogenesis was shown to competitively interact with the splicing machinery 24 . To assess the potential modulation of non-coding RNAs by ABX300 treatment, the population of miRNAs in the visceral WAT from HFD/CD untreated and ABX300-treated DIO mice was determined using Affymetrix microarrays. Among the 1,111 murine miRNAs that were analyzed, 52 were differentially expressed between untreated and treated animals (Fold-change ≥1.5; Pvalue ≤0.05). The heat map in Figure 5A shows the 26 up-and 26 down-regulated miRNAs (see also   Table S5). Similarly, Fig. 5B and Table S6 show the shifts of the 113 microRNAs (54 up-59 down regulated) varying between HFD and CD Untreated mice. Among the above mentioned variations, 37 of 52 (71%) miRNAs differentially expressed in HFD treated mice were found to have the same profile in CD mice (Table S7), corroborating the notion that ABX300 abolishes most changes in the biogenesis of miRNAs induced by an HFD. For instance, a hyper-caloric diet induced 25 while ABX300 treatment down-regulated the expression of miR-221, miR-222, miR-342-3p, and miR-146b (Fig. 5A).
Conversely, miR-30c, miR-92a, miR-193b and miR-378 were decreased during obesity but up-regulated by ABX300 treatment (Fig. 5A). Furthermore, treating CD mice with ABX300 did not alter the expression of any microRNA except for miR-465c2, whose expression was significantly up-regulated 1.56-fold. This finding further suggests that ABX300 primarily modulates miRNA biogenesis in the context of an HFD (Table S8).
Interestingly, the expression levels of ten miRNAs varied exclusively in response to ABX300 but were not altered by diet: miR5122, miR5130, miR130b, miR199a-5p, miR362-5p, miR150-5p, miR26b-5p, miR714, miR92a, and miR29a. Eight of these ten microRNAs were predicted, based on a miRbase analysis (http://mirdb.org/ miRDB), to target mRNAs whose products are directly or indirectly involved in proper mitochondrial functioning (Fig. 5C and Table S9). This group of identified miRNAs represents a large fraction of all microRNAs mediating mitochondrial regulation, as only 16.47% of all microRNAs are expected to belong to this category (Mouse MitoCarta predicted 315 of 1,912). Among the eight identified microRNAs, miR-29a has been predicted to target Slc16a1 (Table S9), which plays a critical role in the regulation of energy balance when animals are exposed to an obesogenic diet 26 .
Because each miRNA regulates, on average, approximately 200 target genes, ABX300 miR-specific © 2016 Macmillan Publishers Limited. All rights reserved.

ABX300 increases oxygen consumption
If ABX300 modulates the functioning of mitochondria, it is expected to generally impact the respiration of the entire organism. To assess respiration at the level of the whole animal, ABX300treated and untreated mice were placed in metabolic cages. Oxygen consumption (VO₂) and CO₂ production (VCO₂) were monitored during a 12 h dark/ 12 h light cycle using the Oxymax Lab Animal Monitoring System (Fig. 6A). HFD ABX300-treated mice showed no differences in VO₂ and VCO₂ over a period of 24 h (Fig. 6B, Mann-Whitney p=0.103, U=39) and Fig. 6C right panels, t-test p=0.943,   t=0.072, ν=21). However, when only focusing on the light period, i.e., the 8 h following ABX300 administration, treated DIO mice showed a marked tendency for higher O₂ consumption and CO₂ production compared with untreated DIO mice (Fig. 6B Mann-Whitney p=0.001, U=12 and Fig. 6C left   panels t-test p=0.0376, t=2.219, ν=21). These 8 h corresponded to the light period, when mice are normally at rest, and this period is characterized by a decrease in metabolism. These outcomes suggested that ABX300 treatment increased the basal metabolic rate of DIO mice. In addition, the respiratory exchange ratio (RER) was determined by calculating the VCO₂/VO₂ ratio. Likewise, ABX300-treated DIO mice demonstrated a RER close to 0.7, which indicates that fatty acids are the main metabolic source of energy. In contrast, untreated animals presented a RER ratio close to 0.8, showing a mixed consumption of fat and carbohydrates (Fig. 6D right panel, t-test p=0.0130, t=2.714, ν=21). These results indicate that ABX300 increases the metabolic rate, which results in the rapid metabolization of lipids, likely via mitochondrial β-oxidation.

Discussion
In this study, we developed a screening assay based on the AS regulation of the LMNA gene to identify ABX300, a compound that can bind SRSF1 in vitro and abrogates the effect of an HFD in vivo.
Targeting splicing factors in vivo has not been tested mainly because these factors are ubiquitously expressed 27 and may have redundant functions in different tissues 28 . Nevertheless, previous studies have supported the relevance of AS isoforms in the regulation of metabolism 29 . Our data reinforce the idea that modulating the activity of splicing factors in vivo may impact RNA biogenesis in a more tissue-specific manner, providing opportunities for a new unexplored therapeutic approach for obesity.
Although currently approved therapies for obesity focus on reducing caloric intake or blocking food absorption 30 , changes induced by ABX300 favor EE. The effect of ABX300 appears to be mediated, at least in part, by changes in the expression of the LMNA gene and a subset of microRNAs involved in mitochondrial functioning.
The selectivity of ABX300 (weight loss exclusively under HFD conditions) (Fig. 2) pointed to a positive safety profile and further supported the possibility of developing ABX300 for therapeutic purposes. In addition to its influence on EE, ABX300 seemed to control feeding. However, a pairfeeding experiment demonstrated that reduced food intake alone was not sufficient to explain the observed phenotype (Fig. 2D). ABX300 can cross the blood brain barrier (Fig.S2G), presumably by transmembrane diffusion due to its solubility in lipid 31 . In the brain, ABX300 is unlikely to alter the AS of LMNA gene because Lamin A but not Lamin C transcripts are regulated by a specific microRNA miR-9 32,33 leading to preferential expression of Lamin C in the brain. However, ABX300 might alter the biogenesis of microRNA involved in mitochondrial functioning in the brain like miR-29a which may target Slc16a1 to control the energy balance when animals are exposed to an obesogenic diet 26 . Another aspect of the safety profile of the drug is that ABX300 did not globally affect the AS of endogenous genes. Indeed, only a few of the AS events in the visceral WAT of treated animals were found to be altered by this molecule. Moreover, these changes suggested a return to normality because AS in © 2016 Macmillan Publishers Limited. All rights reserved. ABX300-treated mice returned to the profile of lean animals (Fig. 4D). Perhaps the most intriguing concept resulting from the findings presented herein is that small molecules targeting the core splicing machinery exert very small effects on different genes and alternative SS (Fig. 4E). Understanding the molecular basis of these differential effects may pave the way towards the rational design of compounds that can modulate the versatile effects of SS selection on gene expression 34,35 . SRSF1 can "travel" from the nucleus to the cytoplasm and is known to participate not only in the control of splicing and miRNA processing in the nucleus but also in the regulation of translation in the cytoplasm 27 . CLIP (crosslinking immunoprecipitation) experiments using anti-SRSF1-specific antibodies 36,37 , splicing-sensitive RNA sequencing, microarray experiments 36 and polysomal fraction mRNAs analysis 38 have identified numerous potential "metabolic" targets for SRSF1, both at the splicing and the translational levels.
Further studies focusing on the interaction of SRSF1 with other partners in the adipose tissue and how these factors are affected by HFD and ABX300 treatment would be of great utility for exploring the association of AS with metabolic alterations in obesity. AS has also been described to affect microRNA biogenesis and regulation 39 . Splicing regulation in vivo may represent a system-wide response to the composition of miRNAs 40 . Nevertheless, how the impact on gene regulation by miRNAs will affect drug metabolism remains unknown 41 . In recent years, interest in the role of miRNAs in fat cell development and obesity has been rapidly growing.
Understanding this role in the proliferation and differentiation of adipocytes during fat cell development may provide new therapeutic targets for anti-obesity drugs and early biomarkers for clinical diagnosis 22,42 . Moreover, miRNAs have been shown to play important roles in both brown and white fat differentiation 43 and in maintaining cell identity by fine-tuning cell-specific transcriptional networks 44 .
In response to obesity, insulin resistance, or cold exposure, changes in miRNAs have been reported in adipose tissue 45 . These changes are consistent with our results, which indicate that ABX300 modulates mitochondrial microRNAs (Fig. 5), and seem to correlate with the observed phenotype of treated animals. Our data support the notion that targeting SRSF1 activity is a valid strategy to develop safe