Pharmacological glycerol-3-phosphate acyltransferase inhibition decreases food intake and adiposity and increases insulin sensitivity in diet-induced obesity

Francis P. Kuhajda, Susan Aja, Yajun Tu, Wan Fang Han, Susan M. Medghalchi, Rajaa El Meskini, Leslie E. Landree, Jonathan M. Peterson, Khadija Daniels, Kody Wong, Edward A. Wydysh, Craig A. Townsend, Gabriele V. Ronnett


Storage of excess calories as triglycerides is central to obesity and its associated disorders. Glycerol-3-phosphate acyltransferases (GPATs) catalyze the initial step in acylglyceride syntheses, including triglyceride synthesis. We utilized a novel small-molecule GPAT inhibitor, FSG67, to investigate metabolic consequences of systemic pharmacological GPAT inhibition in lean and diet-induced obese (DIO) mice. FSG67 administered intraperitoneally decreased body weight and energy intake, without producing conditioned taste aversion. Daily FSG67 (5 mg/kg, 15.3 μmol/kg) produced gradual 12% weight loss in DIO mice beyond that due to transient 9- to 10-day hypophagia (6% weight loss in pair-fed controls). Continued FSG67 maintained the weight loss despite return to baseline energy intake. Weight was lost specifically from fat mass. Indirect calorimetry showed partial protection by FSG67 against decreased rates of oxygen consumption seen with hypophagia. Despite low respiratory exchange ratio due to a high-fat diet, FSG67-treated mice showed further decreased respiratory exchange ratio, beyond pair-fed controls, indicating enhanced fat oxidation. Chronic FSG67 increased glucose tolerance and insulin sensitivity in DIO mice. Chronic FSG67 decreased gene expression for lipogenic enzymes in white adipose tissue and liver and decreased lipid accumulation in white adipose, brown adipose, and liver tissues without signs of damage. RT-PCR showed decreased gene expression for orexigenic hypothalamic neuropeptides AgRP or NPY after acute and chronic systemic FSG67. FSG67 given intracerebroventricularly (100 and 320 nmol icv) produced 24-h weight loss and feeding suppression, indicating contributions from direct central nervous system sites of action. Together, these data point to GPAT as a new potential therapeutic target for the management of obesity and its comorbidities.

  • fatty acid oxidation
  • acylglyceride synthesis
  • hepatic steatosis
  • hypothalamus
  • diabetes

glycerol-3-phosphate acyltransferases (GPATs) esterify long-chain fatty-acyl-CoAs to the sn-1 position on glycerol-3-phosphate to produce lysophosphatidic acid. As catalysts for this first committed step in the synthesis of triglycerides and phospholipids (reviewed in Ref. 4), GPATs have attracted interest as potential targets for pharmacological treatment of the complications of obesity (44). Initially, GPAT was thought to exist as two isoforms, a mitochondrial isoform responsible predominantly for triglyceride synthesis and a microsomal isoform for synthesis of other acylglycerides (reviewed in Ref. 4). There are now four identified GPAT isoforms, two mitochondrial (GPAT1, GPAT2), and two microsomal (GPAT3, GPAT4), each encoded by separate genes (67, 18, 33, 50, 57).

GPAT1, the first mammalian isoform cloned (57), resides in the outer mitochondrial membrane and shows preference for saturated long-chain acyl-CoA, particularly palmitoyl-CoA (30). GPAT1 can be distinguished from other isoforms by its resistance to inactivation by sulfhydryl-reactive compounds such as N-ethylmaleimide (NEM) (29). GPAT1 is most highly expressed in liver and adipose, tissues with high capacity for lipogenesis (24). GPAT1 has been reported to account for 30–50% of total hepatic GPAT activity and catalyzes the bulk of hepatic triglyceride (16, 34, 57). Like other key lipogenic enzymes acetyl-CoA carboxylase (ACC) and fatty acid synthase (FAS), GPAT1 is under transcriptional control by sterol regulatory element binding protein-1, and is thusly upregulated in response to feeding and nutritional states of positive energy balance, such as refeeding and maintenance on high-sucrose or high-fat obesogenic diets (reviewed in Ref. 10). GPAT2, the other mitochondrial isoform, was first recognized in liver mitochondria from GPAT1−/− mice (24) and also resides in the outer mitochondrial membrane. Unlike GPAT1, GPAT2 is sensitive to inactivation by NEM and does not hold preference for palmitoyl-CoA over oleoyl-CoA (24). Although GPAT2 has been reported to account for up to 60% of hepatic mitochondrial GPAT activity in lean wild-type mice and synthesizes triglycerides, it is less responsive to dietary control. GPAT2 expression does not increase in liver of rats fed obesogenic diet (50), and although it is active in GPAT1−/− mice, it does not substitute for GPAT1 functionally; GPAT1−/− mice still have lower liver triacylglyceride (16). GPAT2 expression is 50-fold higher in testis than in other tissues, including liver (50).

GPAT3 and GPAT4 are microsomal isoforms, named for that subcellular fraction. They are localized to the endoplasmic reticulum, sensitive to inactivation by NEM, and can use a range of saturated and unsaturated fatty acyl-CoAs 12–20 carbons in length, although preference is for 16- and 18-carbon species (67, 33). GPAT3 expression is highest in visceral (particularly gonadal) white adipose tissue (WAT), and it is also present in small intestine, kidney, and heart, but relatively low in liver (6). Transcription of GPAT3 is significantly upregulated in 3T3-L1 cells (9) and adipocytes (6) undergoing differentiation. GPAT3 was originally classified as an AGPAT (1-acylglycerol-3-phosphate acyltransferase, second enzyme in acylglyceride synthesis), and it may, in fact, have both GPAT and AGPAT activities (42). GPAT4 was originally classified as AGPAT6, but it was found instead to have NEM-sensitive GPAT activity (7, 33). GPAT4 is highly expressed in brown adipose tissue (BAT), liver, and lactating mammary gland, as well as testis (3, 7, 33). Liver and BAT from GPAT4−/− mice have 50% lower total GPAT activity than those in wild-type controls, and 65% lower total NEM-sensitive GPAT activity (33). Pups from GPAT4−/− dams die within 48 h unless cross-fostered, due to extremely low triglyceride in milk from the dam (3) GPAT4 may play a role in spermatogenesis, as it is abundantly expressed in spermatocytes and spermatids during meiosis (35). Subdermal adipose tissue is greatly decreased in GPAT4−/− mice, although it is yet to be determined whether this is due to attenuated triglyceride production or to inhibited adipocyte differentiation, as occurs in congenital generalized lipodystophy due to mutations in AGPAT2 (12). In summary, triglyceride synthesis is not restricted to a single GPAT isoform. Thus, multiple isoforms may serve as potential targets for pharmacological intervention.

The ability to modulate GPAT activity in vivo has thus far been restricted to knockout mouse models. These models have shown some select reversals or prevention of obesity or other related metabolic disorders. There was an initial report of moderate reduction of adipose tissue mass in GPAT1−/− mice (16). When fed the Surwit diet comprising high levels of sucrose and short- and medium-chain fatty acids, GPAT1−/− animals became obese (34), although they displayed reduced hepatic steatosis with improvement in insulin resistance compared with controls. Similarly, selective hepatic GPAT1 knockout in ob/ob mice exhibited reduced hepatic steatosis and improved plasma glucose and cholesterol levels (55), whereas hepatic overexpression of GPAT1 caused insulin resistance (31). While the precise mechanism is unclear, GPAT4-null mice have reduced liver and adipose triacylglycerol (33) and are resistant to diet-induced obesity (DIO) (48). Mouse models that are null for GPAT2 or GPAT3 have not yet been made or reported. However, in mouse 3T3-L1 adipocytes, it appears that the GPAT3 isoform may contribute significantly to triglyceride accumulation (9) and GPAT3 is highly expressed in adipocytes (6). Furthermore, GPAT3 and GPAT4 are phosphorylated and activated by insulin, thus establishing a connection between glucose metabolism and acylglyceride synthesis (40). Thus, a reduction in GPAT activity may have a beneficial effect on the obese phenotype, irrespective of the particular GPAT isoforms that are targeted.

We recently developed small-molecule inhibitors of GPAT whose design was based on the crystalline structure of squash GPAT (53). The compounds had variable inhibitory activity against GPAT in an assay with mouse mitochondrial GPAT, with no NEM. We have selected a compound with low IC50 for mitochondrial GPAT activity from those studies (53), herein called FSG67, to investigate the biological consequences of pharmacological GPAT inhibition in mice. The data show that pharmacological GPAT inhibition with FSG67 reduces body weight, adiposity, and food intake in mice, while enhancing fatty acid oxidation and protecting against hypophagia-induced decreases in energy expenditure. FSG67 resolved hepatic steatosis and increased glucose tolerance and insulin sensitivity in DIO mice. These results suggest that pharmacological GPAT inhibition holds promise as potential therapy for obesity and attendant metabolic disorders.


FSG67 preparation and administration.

FASgen provided FSG67, a synthetic, small-molecule inhibitor of glycerol-3-phosphate-acyl-transferase (GPAT). Specifically, the GPAT assay had used mouse mitochondrial GPAT, without addition of NEM. Therefore, FSG67 had the potential to act on both the GPAT1 and GPAT2 isoforms.

For the current in vivo studies, FSG67 was dissolved and neutralized as needed with NaOH in glucose-free RPMI 1640 (Invitrogen, Carlsbad, CA), or PBS for all studies, at doses indicated per experiment. Mice were administered FSG67 or vehicle by intraperitoneal injection in total injection volumes of 50 μl. In acute-effect studies (Figs. 2 and 3), lean or DIO mice were treated with a single intraperitoneal dose of 5 or 20 mg/kg (15.3 or 61.1 μmol) just prior to dark-onset. Chronic FSG67 treatments in DIO mice were accomplished with intraperitoneal injections of FSG67 at 5 mg/kg (15.3 μmol) just prior to dark-onset daily. Studies in vitro used DMSO as vehicle, with final concentration of 0.25%, which has not affected viability in cell-based viability assays (data not shown).

GPAT activity assay in isolated liver mitochondria.

The mitochondrial GPAT assay has been described (39). As we have done previously (5354), isolated mitochondria were added to an incubation mixture with [14C]glycerol-3-phosphate, palmitoyl-CoA, and varying concentrations of FSG67. For the current determinations, the assay was run with or without the sulfhydryl-reactive compound NEM (100 μM) to distinguish the NEM-insensitive GPAT1 activity from that of NEM-inactivated GPAT2 (29). The reaction was stopped and processed after 10 min with chloroform, methanol and 1% perchloric acid to retain the organic layer, which was evaporated under nitrogen and counted for 14C to determine the extent of reaction inhibition. Data points were recorded in triplicate, and IC50 values were calculated by linear regression.

3T3-L1 adipocytes.

3T3-L1 cells were differentiated into adipocytes, as described previously (36). Seven days postdifferentiation, cells were treated with FSG67 at indicated concentrations for 18 h, then labeled with [14C]palmitate for 2 h. Following Folch extraction, lipids were subjected to polar and nonpolar thin-layer chromatography (21). Triglyceride and phosphatidylcholine fractions were quantified by phosphorimaging (Storm 840, Molecular Dynamics, Piscataway, NJ) and image analysis (ImageQuant, Molecular Dynamics, ver. 3.3).

DIO and lean mouse models.

All animal experimentation was done in accordance with protocols submitted to and approved by the Johns Hopkins University School of Medicine Institutional Animal Care and Use Committee for these studies. Male C5BL/6J mice obtained from Jackson Laboratories (Bar Harbor, ME) were maintained on a 12:12-h light-dark cycle at 25°C and acclimatized for 1 wk prior to treatment. For lean animal studies, 10- to 12-wk-old mice (8 mice/group) were fed rodent chow with 13% calories from fat, 58% from carbohydrate, and 29% from protein (4.1 kcal/g) (Prolab RMH 2500, PMI Nutrition International, Brentwood, MO). For studies of diet-induced obesity (DIO), mice at 4 or 5 wk of age were fed a high-fat synthetic diet (HF) composed of 60% calories from fat, 20% from carbohydrate, and 20% from protein (5.2 kcal/g) (D12492; Research Diets, New Brunswick, NJ) for 12–16 wk until studies commenced. Some mice fed HF are DIO-resistant; thus, our accepted weight range for studies was 40–45 g (4–10 DIO/group, described per study).

Food intake measurement and food-restricted controls.

Regardless of cage-type, food intake was measured by hand, as preweighed food in minus food out, minus spilled food captured on cage paper beneath a perforated plastic or wire-mesh flooring insert. Data were expressed in kilocalories. Acute-effects studies included a control group that was fasted for 18 h. Chronic studies included groups of mice that were pair-fed (PF) to the level of food intake exhibited by FSG67-treated index group. The PF data were collected 2 days behind the FSG67 group. Paired-feeding was accomplished by supplying all PF mice with the average amount of food eaten by the index group, plus the average spillage by the index group.

Conditioned taste aversion test.

We tested for the ability of FSG67 to produce conditioned taste aversion, to determine whether the compound induced malaise. Adult, lean mice maintained on chow were trained for 10 days to consume their daily water in a scheduled 2-h access period during the light period. Chow was withheld during the 2-h water access. Body weight, food intake, and water intake (30-min, 30–120-min) were measured daily. Mice had stable water intakes (30 min, ∼1.5 ml, 30–120 min, ∼0.5 ml) for at least the last 3 days. Mice were divided into three groups of n = 8, and given access to 0.15% sodium saccharin rather than water for 30 min. At the end of saccharin access, mice were injected intraperitoneally with vehicle or FSG67 (5 or 20 mg/kg, 15.3 or 61.1 μmol/kg) and permitted water access for the remaining 90 min. The next day, mice were given 30-min access to two bottles, one with 0.15% saccharin, and the other with water. Intakes of both solutions were recorded. Data are expressed as % saccharin preference [100 × saccharin intake/(saccharin + water intakes)].

Indirect calorimetry.

Some DIO mice were used to assess whole-body metabolic profile by indirect calorimetry, with simultaneous daily body weight and HF diet intake monitoring. Mice retested in an open-flow indirect calorimeter (Oxymax, Columbus Instruments). Mice were adapted to individual housing in home cages for 3 days, prior to acclimation for 2 days in the calorimetry chambers, before treatments began. DIO mice (three groups, 4 or 5 mice/group) were injected intraperitoneally with vehicle or FSG67 (5 mg·kg−1·day−1) for 9 days of ad libitum access to HF in the chambers or injected with vehicle and PF to the FSG67 index group. After 9 days of intraperitoneal treatments, mice were monitored in the calorimeter for another 6 days without injections, but with continued PF control. Rates of oxygen consumption (V̇o2, ml·kg−1·h−1) and carbon dioxide production (V̇co2) were measured for each chamber every 15 min throughout the study. Respiratory exchange ratio (RER = V̇co2/V̇o2) was calculated by Oxymax software (v. 4.02) to estimate relative oxidation of carbohydrate (RER = 1.0) vs. fat (RER approaching 0.7), not accounting for protein oxidation. Energy expenditure was calculated as EE = V̇o2 × [3.815 + (1.232 × RER)] (27), and normalized for subject body mass (kcal·kg−1·h−1). Average metabolic values were calculated per subject, for each day, and averaged across subjects for each day.

Quantitative-NMR assessment of adiposity.

An EchoMRI-100 (Echo Medical Systems, Houston, TX) in the Molecular and Comparative Pathobiology Phenotyping Core at Johns Hopkins University School of Medicine was used to measure fat, lean, and water masses by quantitative-NMR in DIO mice treated chronically (at 13th day of a 28-day study) with FSG67 or vehicle, or vehicle plus PF (n = 6 or 7/group).

Glucose and insulin tolerance testing, and other blood chemistry measurements.

A glucose tolerance test (GTT) was performed on day 18 of the 28-day chronic study in DIO mice. Mice were subjected to a 6-h food deprivation (2) after which glucose (1 mg/kg) was injected intraperitoneally. Clean tail blood droplet was obtained at time points (0, 15, 30, 60, 120, 180 min) for measurement of blood glucose with a glucometer (BD Logic, NovaMax strips), and 20-μl volumes were collected (0, 30, 60, 180 min) for serum insulin measurement by ELISA (Millipore, Beverly, MA, USA). The ITT was performed on day 26 of the chronic treatment study. After a 2.5-h food deprivation, mice were injected with insulin (1 U/kg ip) and glucometer readings were taken from tail blood (0, 15, 30, 45, 60, 90 min). On day 28, mice were euthanized for tissue harvest, and blood samples were taken for serum measurements of leptin by ELISA (Millipore). DIO mice from a separate 16-day chronic treatment study were euthanized 18 h after final injection for tissue harvests for RT-PCR, and blood was collected for serum measurements of triglycerides, cholesterol, and glucose (Bioanalytics, Gaithersburg, MD).

Chronic lateral cerebroventricle cannulas.

For experiments requiring intracerebroventricular administration of compounds, adult mice on chow were outfitted unilaterally with chronic indwelling cannulas aimed at the lateral cerebral ventricle. After 1 wk of recovery, cannula placements were assessed by measuring food intake in response to intracerebroventricular neuropeptide Y (NPY; American Peptide, Sunnyvale, CA). Mice were given NPY (0.25 nmol/2 μl injection) or sterile 0.9% saline vehicle via the intracerebroventricular cannula, and allowed 1-h access to chow during the light phase. Mice that ate at least 0.5 g of food after NPY were used in the experiments. Two experiments were performed. In each, 5 or 6 mice were injected intracerebroventricularly with vehicle and 24-h body weight changes, and food intakes were measured. Three days later, a dose of FSG67 (100 nmol or 320 nmol) was injected intracerebroventricularly, with weight and food intake measurements.

Real-time quantitative RT-PCR.

Hypothalamus, liver, and WAT of DIO and lean mice were harvested RNAse-free and immediately frozen in liquid nitrogen. Total RNA was isolated, and RT-PCR was performed as described in previous work (46). Gene-specific primer pairs were designed using Primer3 software ( Sequences of the primer pairs are listed in Table 1.

View this table:
Table 1.

Primer pairs used in real-time quantitative RT-PCR


Fresh tissues were fixed in buffered formalin or snap-frozen in liquid N2 for histological processing. Epididymal WAT, BAT, liver, and skin were embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Frozen samples were cryosectioned and stained with Oil-Red-O to identify triglyceride droplets, with hematoxylin counterstain. Hepatic glycogen was assessed on sections using the periodic acid-Schiff stain (PAS) with and without diastase treatment.

Statistical analyses.

All data are presented as means ± SE. Statistical analyses were performed with Prism 5.0 (Graph Pad Software, San Diego, CA). The IC50 determinations were performed with linear regression. Food intake data from the acute-effect studies (Fig. 2B), RT-PCR data from the lean-acute study (Fig. 12A), and data from intracerebroventricular administration studies (Fig. 11) were analyzed by unpaired, two-tailed t-test. Other data sets with one time point or cumulative measure were analyzed by one-way ANOVA followed either with Bonferroni tests for all group comparisons (Figs. 2, A and C, 3B, 6, 7, 8, and 9, B, D, F), or with Dunnett's test for comparisons with vehicle control (Fig. 3A). Data with repeated time points (Figs. 4, 5, 9, A, C, E) were analyzed by two-way ANOVA, with time and treatment as variables, followed by Bonferroni comparisons of all groups. Test statistics and group sizes are reported in the text. Statistically significant differences were determined at P < 0.05.


FSG67 inhibits both GPAT1 and GPAT2 isoforms of liver mitochondrial GPAT.

An earlier evaluation showed that FSG67 had GPAT inhibitory activity in mouse mitochondria (53). We included NEM (100 μM) to distinguish the activities of GPAT1 and GPAT2. The final concentrations of FSG67 in the assays were 0, and 0.2–122.2 μM (0.625–40 μg/ml). Counts of 14C after the reaction were higher in the non-NEM baseline condition (11,330 ± 1,203) than in the baseline with NEM to inactivate GPAT2 (7,847 ± 558). FSG67 decreased GPAT activity in both conditions as concentration was increased, resulting in IC50 of 30.2 μM for total mitochondrial GPAT (P < 0.0001, r2 = 0.83, n = 3) and IC50 of 42.1 μM for GPAT1 (P < 0.0001, r2 = 0.86, n = 3) in this experiment.

FSG67 reduces acylglyceride synthesis in 3T3-L1 adipocytes.

Mouse 3T3-L1 cells were used to measure the effect of FSG67 on acylglyceride synthesis in vitro. The 3T3-L1 cells were treated at 7 days postdifferentiation with FSG67 at concentrations of 7.6 μM to 150 μM (2.5–50 μg/ml), and the IC50 values for inhibition of triglyceride and phosphatidylcholine synthesis were determined from thin-layer chromatography using linear regression. The IC50 values were 33.9 μM for cellular triglyceride synthesis (P = 0.023, r2 = 0.86, n = 3) and 36.3 μM for phosphatidylcholine synthesis (P = 0.015, r2 = 0.89, n = 3). Phosphatidylcholine, as the predominant phospholipid synthesized in the 3T3-L1 adipocytes, was representative of overall cellular phospholipid synthesis. Consistent with this inhibition of acylglyceride synthesis, there was a reduction of the apparent number and average size of lipid droplets in 3T3-L1 adipocytes after 48 h of FSG67 treatment (Fig. 1). Control 3T3-L1 cells had plentiful lipid droplets, many of large size (Fig. 1A). An FSG67 concentration of 30 μM, similar to IC50 values for triglyceride and phospholipid syntheses (above), and to IC50 for mouse mitochondrial GPAT activity (see Ref. 53 and results), greatly decreased the number and typical size of droplets (Fig. 1B). Further increases of FSG67 to 75 μM and 150 μM decreased the number of lipid droplets (Fig. 1, C and D). Cultured cells appeared healthy and intact. These results demonstrate that FSG67 reduces acylglycerol synthesis and lipid accumulation in vitro, and support the hypothesis that FSG67 inhibits GPAT activity in cells.

Fig. 1.

FSG67 decreases lipid droplet number and size in 3T3-L1 adipocytes. 3T3-L1 cells at 7 days postdifferentiation were treated with vehicle (A) or the glycerol-3-phosphate acyltransferase (GPAT) inhibitor FSG67 at indicated concentrations (B, C, D) for 18 h. Phase-contrast photomicrographs (×400) show reduced number and average size of lipid droplets in the cells with increasing final well concentration of FSG67. Arrows indicate typical droplet appearance, with bright centers and surrounding dark halo.

Acute FSG67 treatment produces weight loss and decreases food consumption in lean and DIO mice.

We administered a single dose of FSG67 (20 mg/kg ip) to both lean and DIO mice to determine its effects on body weight. We also measured food intake, as this had not been examined in studies of GPAT−/− mice. In separate studies, chow-fed lean mice (n = 8/group) and DIO mice on high-fat diet (HF) (n = 6 or 8/group) were injected with vehicle or FSG67 at the beginning of dark cycle, or were fasted for 18 h. Outcomes for body weight showed significant differences [Lean: F(2,21) = 163.1, P < 0.0001; DIO: F(2,17) = 13.04, P = 0.0004].

FSG67-treated lean mice lost 3.7 ± 0.9% of preinjection body weight (−1.0 ± 0.2 g), whereas vehicle-injected controls gained 2.5 ± 0.5% (0.6 ± 0.1g) (Fig. 2A; P < 0.0001). Fasted mice lost 15.5 ± 0.7% (−3.9 ± 0.2 g), significantly more than FSG67-treated mice (P < 0.0001). FSG67 decreased chow intake to 35% of vehicle control, 6.0 ± 1.0 kcal vs. 17.2 ± 0.7 kcal [Fig. 2B; t(14) = 9.338, P < 0.0001] but did not completely prevent energy intake.

Fig. 2.

Intraperitoneal administration of FSG67 can produce acute decreases in body weight and energy intake in lean mice and mice made diet-induced obese (DIO) with a high-fat diet (HF). Body weight changes and food intakes were measured following a single dose of vehicle (white bars) or FSG67 at 20 mg/kg ip (gray bars) in lean mice (n = 8/group) or DIO mice (n = 6–8/group) on ad libitum diets. Additional groups of mice were fasted for comparisons (black bars). A: FSG67 produced weight loss in lean mice. B: FSG67 decreased chow intake by lean mice but did not completely prevent energy intake. C: FSG67 produced weight loss in DIO mice. D: FSG67 reduced HF intake by DIO mice. All panels had significant effects by one-way ANOVA (A, C) or unpaired, two-tailed t-test (B, D). Bonferroni tests: *P < 0.05 vs. vehicle control, #P < 0.05 for FSG67 vs. fasted.

FSG67 also decreased body weight in DIO mice consuming a calorically dense diet rich in long-chain fatty acids (Fig. 2C). FSG67-treated DIO mice lost 4.3% ± 0.5% of preinjection body weight (−1.7 ± 0.2 g), significantly more than vehicle-injected controls (0.6%, P < 0.001), a similar response to that seen in lean mice. As expected, DIO mice were less responsive than lean mice to food deprivation. Fasted DIO mice lost 4.7 ± 0.8% weight, similar to weight loss seen in the FSG67-treated group. FSG67 significantly reduced energy intake by DIO on HF diet to 33% of vehicle control [Fig. 2D; 5.4 ± 1.6 vs. 16.2 ± 0.7 kcal; t(10) = 5.326, P = 0.0003], much as it had decreased energy intake from chow diet in lean mice.

FSG67 does not produce conditioned taste aversion at doses used for gradual weight loss.

Because FSG67 decreased food intake, we performed conditioned taste aversion (CTA) testing to determine whether FSG67 produced an aversive interoceptive stimulus. CTA might suggest malaise as a contribution to the hypophagia. Lean mice (8/group) trained to consume water on a constrained daily schedule were allowed briefly to drink 0.15% saccharin solution, then injected intraperitoneally with vehicle or with FSG67 at 5 mg/kg or 20 mg/kg. In the two-bottle test the next day, FSG67 failed to produce CTA [Fig. 3A; F(2,20) = 2.456, P = 0.1113, not significant (ns)] at either 5 mg/kg (79.7 ± 2.5% saccharin preference, q = 0.5646) or 20 mg/kg (64.6 ± 10.7% saccharin preference, q = 2.149) vs. vehicle control (85.3 ± 3.8% saccharin preference). Thus, the reduction in food intake from FSG67 was not likely due to sickness behavior, but rather to other effects on appetite.

Fig. 3.

Conditioned taste aversion testing and selection of FSG67 dose for chronic studies. A: Conditioned taste aversion (CTA) testing using a two-bottle choice paradigm did not reveal significant reduction in saccharin drinking in lean mice (n = 8/group) injected intraperitoneally with FSG67 at 5 mg/kg (gray bar) or 20 mg/kg (black bar). B: daily injections of low doses of FSG67 produced weight loss by the 5th day in mice that received 5 mg·kg−1·day−1 (black bar, 15.3 μmol·kg−1·day−1). B: one-way ANOVA is significant. Bonferroni test: *P < 0.05 vs. vehicle control (white bar).

To begin to model FSG67 as a potential antiobesity treatment, we determined a dose of FSG67 that would be suitable for multiple dosing and produce gradual weight loss in DIO mice (Fig. 3B). DIO mice (4/group) were given intraperitoneal injections daily of vehicle or FSG67 (1, 2, 5 mg/kg). After 5 days, FSG67 at 5 mg/kg produced weight loss of 3.9 ± 0.1% compared with vehicle control [q = 3.143, P = 0.0364; F(3,12) = 3.928]. This dose was chosen for subsequent chronic treatment experiments.

Chronic low-dose FSG67 treatment of DIO mice produces weight loss and decreases food intake, while protecting against hypophagia-induced decreases in metabolic rate, and increasing fat oxidation.

A second chronic treatment study utilized indirect calorimetry to measure changes in whole-body metabolism during GPAT inhibition with FSG67. Body weight and HF diet intake were measured daily. DIO mice (4 or 5/group) were injected daily for 9 days with FSG67 (5 mg/kg ip), vehicle, or pair-fed to the average food intake of the FSG67-treated group. Monitoring continued for another 6 days after cessation of FSG67. Indirect calorimetry was used to measure rates of oxygen consumption (V̇o2) and carbon dioxide production (V̇co2) to calculate the respiratory exchange ratio (RER = V̇co2/V̇o2), and estimate rates of EE (27).

This longer chronic experiment revealed significant effects by treatment [F(df2) = 60.41, P < 0.0001] and time [F(df14) = 2.699, P = 0.0014] on weight change (Fig. 4A). Nine days of FSG67 treatment resulted in weight loss of 12.1 ± 3.6% compared with vehicle control (t = 3.106, P < 0.05), and pair-fed controls lost 5.5 ± 2.0% weight (ns from FSG67 at day 9). After cessation of FSG67, mice maintained weight losses of 15.6–17.6% that resulted in significant differences from the pair-fed control losses of 5.9–6.8% during days 11–15 (Fig. 4A; all P < 0.05).

Fig. 4.

Chronic FSG67 treatment of DIO mice reversibly reduces body weight and food intake while enhancing fatty acid oxidation. DIO mice (4 or 5/group) were tested in an indirect calorimeter. Mice were injected daily for 9 days with vehicle (○) or FSG67 (▴, 5 mg·kg−1·day−1 ip) and fed HF ad libitum, or received vehicle injection while being pair-fed (PF; ▿) to the FSG67 index group HF intake. Subsequently, mice were monitored for 6 more days on ad libitum feeding or PF. A: daily FSG67-produced weight loss beyond that caused by the hypophagia (PF). B: daily FSG67 produced an average daily hypophagia of 30% in DIO mice (PF included in the two-way ANOVA, data not shown in graph). C: FSG67 decreased cumulative HF intake by 30%. D: daily FSG67 provided partial protection from the decrease in rate of oxygen consumption (V̇o2, ml·kg−1·h−1) characteristic of feeding reduction/restriction (PF). E: FSG67 produced a decrease in respiratory exchange ratio (RER = V̇co2/V̇o2) beyond that seen in PF controls, indicating enhanced fat oxidation. F: daily FSG67 provided partial protection from the decrease in rate of energy expenditure (EE; kcal·kg−1·h−1) characteristic of feeding reduction/restriction (PF). All panels have significant effects of treatment by two-way ANOVA. Bonferroni tests: *P < 0.05 FSG67 vs. vehicle control; #P < 0.05 for FSG67 vs. PF.

The experiment also revealed a significant overall effect of treatment on daily energy intake from the HF diet [Fig. 4B, F(df2) = 19.97, P < 0.0001]. Chronic FSG67 at 5 mg·kg−1·day−1 decreased energy intake by 30% on average (−10.1 ± 0.5 kcal/day vs. 14.4 ± 0.3 kcal/day). This hypophagia resulted in significant effects for cumulative energy intake (Fig. 4C) over time [F(df14) = 130.5, P < 0.0001] and by treatment [F(df2) = 71.79, P < 0.0001], with significant interaction [F(df28) = 1.567, P = 0.0447]. Cumulative energy intake by FSG67-treated DIO mice was 87.7 ± 13.8 kcal by 9 days of treatment compared with vehicle controls at 127.1 ± 7.0 kcal (P < 0.05). By day 15, 6 days after FSG67 treatment stopped, FSG67-treated mice had eaten 148.7 ± 27.6 kcal, still 31% less than vehicle controls (P < 0.001).

Indirect calorimetry data from this experiment showed significant treatment effects (all P < 0.0001) for V̇o2 [F(df2) = 33.45], RER [F(df2) = 15.07], and calculated energy expenditure [F(df2) = 23.82]. Despite the hypophagic effect of FSG67, it appeared to protect somewhat against the characteristic decrease in rates of oxidative metabolism seen in the pair-fed group (Fig. 4D; average V̇o2 during days 1–9: FSG67, 2,930.8 ± 33.6 kcal·kg−1·h−1, PF, 2,760.2 ± 36.8 kcal·kg−1·h−1, 5.8% difference). PF mice will run out of food before the end of the day, and they would, therefore, begin to oxidize their own stores of fat. Hence, in addition to the low-average RER caused by intake of HF diet (average RER for vehicle control days 1–9, 0.748 ± 0.002), PF mice had further decreased RER (0.727 ± 0.004). FSG67-treated mice that were hypophagic comparable to the pair-feeding had even lower average RER (0.715 ± 0.002) than PF controls, indicating additional fat oxidation. The V̇o2 and RER were used to calculate rates of energy expenditure. FSG67 treatment appeared to protect slightly against the decrease in energy expenditure that is characteristic of feeding restriction (FSG67, 13.8 ± 0.2 kcal·kg−1·h−1; PF, 13.2 ± 0.2, 4.3% difference), consistent with the V̇o2 data. In summary, DIO that were treated chronically with the GPAT inhibitor FSG67 had lower energy intakes, increased rates of oxidative metabolism, a shift toward enhanced whole-body fat oxidation, and increased energy expenditure despite hypophagia. The result was significant loss of body weight.

Extended chronic treatment of DIO with FSG67 maintains weight loss.

In a third chronic treatment study using DIO mice (6 or 7/group), daily injections of vehicle or FSG67 at 5 mg/kg were continued during ad libitum access to HF for 28 days. Again, a PF group was included to control for consequences of FSG67-induced hypophagia. Weight changes and HF intakes were monitored daily (Fig. 5). This experiment yielded significant effects of treatment [F(df2) = 185.3, P < 0.0001] on cumulative weight change, interacting with day of treatment [F(df24) = 1.62, P = 0.0074]. The first part of the study showed a treatment-dependent pattern of weight loss (Fig. 5A), consistent with the previous study (Fig. 4A). FSG67 treatment resulted in ∼12% weight loss vs. vehicle control (P < 0.05, day 8 and beyond). Pair-fed controls showed maximal weight loss of 5.5% on day 10. Just a few days later, the pair-fed mice weighed the same as vehicle ad libitum controls. In contrast, the FSG67-treated mice maintained their 12% weight loss throughout the remainder of the study (Fig. 5A). This occurred despite the temporary nature of the hypophagia from FSG67 (and therefore, the food restriction of the PF group), which lasted for the first 10 days of treatment only [Fig. 5B, daily energy intake, treatment effect: F(df2) = 26.49, P < 0.0001; day effect: F(df25) = 5.088, P < 0.0001]. FSG67-treated mice did not show rebound hyperphagia (Fig. 5B), and thus cumulative energy intakes remained significantly lower for the FSG67 group vs. vehicle control from day 9 until the end of the study (Fig. 5C, all P < 0.05). Effects of treatment [F(df2) = 184.7 and day F(df25) = 408.3; both P < 0.0001] and interaction [F(df50) = 1.537, P = 0.0139].

Fig. 5.

Long-term daily FSG67 maintains weight loss in DIO mice. DIO mice were injected intraperitoneally with vehicle (○) or FSG67 (▴, 5 mg·kg−1·day−1, 15.3 μmol·kg−1·day−1) on ad libitum HF, or were vehicle-treated and PF (▿) to the FSG67 index group (6 or 7 mice/group), for 28 days. At time points during the study, mice were subjected to other tests (top x-axis, data shown in other figures). A: long-term daily FSG67 decreased body weight in DIO by day 8, and maintained the weight loss. B: daily FSG67 produced hypophagia until days 9 and 10, followed by return to normal intake levels. C: cumulative 29-day differences in HF intake were maintained. A–C: significant effects of treatment by two-way ANOVA. Bonferroni tests: *P < 0.05 FSG67 vs. vehicle control, #P < 0.05 for FSG67 vs. PF.

FSG67 treatment reduces adiposity.

On day 13 during the 28-day study, noninvasive quantitative NMR was used to measure fat mass and fat-free lean mass (Fig. 6). The FSG67-treated mice weighed 35.8 ± 1.3 g, 4 g less than DIO vehicle controls (39.8 ± 0.5 g; q = 3.023, P < 0.05). Pair-fed mice had body mass similar to that of the vehicle group (35.8 ± 1.6 g). Fat mass in FSG67-treated mice was 14.3 ± 0.6 g, 4.2 g less than vehicle controls (10.1 ± 1.1; q = 3.725, P < 0.01) and 3.35 grams less than average fat mass in the pair-fed group (13.5 ± 0.4 g, q = 3.005, P < 0.05). There were no differences among groups in lean mass. Thus, the weight loss after chronic GPAT inhibition with FSG67 was primarily from fat mass. Water content, which is mainly associated with the lean mass, was unchanged among groups (data not shown).

Fig. 6.

Chronic daily treatment of DIO mice with FSG67 for 13 days caused weight loss specifically from the body fat mass. Significance by one-way ANOVA, n = 6 or 7/group, Bonferroni tests: *P < 0.05 FSG67 (checkered bars) vs. vehicle (white bars) and PF controls (black bars).

Systemic administration of FSG67 downregulates gene expression of lipogenic enzymes in DIO mice.

From the in vivo and in vitro studies, it seemed that FSG67 could either prevent fat accumulation, enhance fatty acid oxidation, or both. To investigate mechanisms by which FSG67 treatment could have these effects, we used RT-PCR to measure the expression of mRNAs for several key lipogenic and fat-oxidative enzymes in white adipose tissue (WAT) and liver from vehicle control (7 or 8 samples), FSG67-treated (4 or 5 samples), and pair-fed (5 or 6 samples) DIO mice from a separate chronic trial with FSG67 (5 mg/kg ip) for 16 days (data not shown).

We measured expression of lipogenic enzymes: FAS, responsible for the de novo reductive synthesis of fatty acid (49), acetyl-CoA carboxylase-1 (ACC1), the cytoplasmic isoform of ACC expressed in lipogenic organs that synthesizes malonyl-CoA used as a substrate of FAS for fatty acid synthesis (19), and GPAT. In WAT, we also measured expression of peroxisome proliferator-activated receptor gamma (PPARγ) a key transcription factor for adipogenesis (37), lipid partitioning (28), and postprandial lipid storage (45). WAT from FSG67-treated DIO mice showed substantial downregulation of ACC1, FAS, and GPAT by 50–60% vs. vehicle control and pair-fed conditions, which were the same (Fig. 7A) [ACC1: F(2,16) = 15.62, P = 0.0002. t = 4.503, P < 0.01 vs. vehicle; t = 5.307, P < 0.001 vs. pair-fed. FAS: F(2,16) = 19.7, P < 0.0001, t = 6.287, P < 0.001 vs. vehicle, t = 3.727, P < 0.01 vs. pair-fed. GPAT: F(2,16) = 9.218, P = 0.0022. t = 4.235, P < 0.01 vs. vehicle; t = 3.033, P < 0.05 vs. pair-fed]. Together, these data demonstrated reduced expression of three key enzymes for synthesis of fatty acids and acylglycerides in WAT from FSG67-treated DIO mice. Gene expression for PPARγ followed a similar pattern [data not shown; F(2,13) = 7.849, P = 0.0058], with significant differences between FSG67 and pair-fed conditions (t = 3.955, P < 0.01). To examine the potential for lipid oxidation, we measured expression of the “liver” form of CPT-1, the rate-limiting enzyme for entry of long-chain fatty acids into mitochondria for oxidization (7 or 8 samples/group). CPT-1 expression showed no group differences. We also measured gene expression of UCP-2 in WAT. Interestingly, FSG67-treated DIO WAT seemed protected from a decrease in UCP-2 seen after restricted energy intake (Fig. 7B) [UCP: F(2,19) = 3.952, P = 0.0367; t = 2.66, P < 0.05].

Fig. 7.

Pharmacological GPAT inhibition reduced adiposity and down-regulated lipogenic gene expression in DIO mice. DIO mice were vehicle-injected controls (white bars), FSG67-treated (5 mg·kg−1·day−1; checkered bars), or vehicle-injected pair-fed controls (solid bars) for the hypophagic FSG67-treated group, for 16 days. Data were obtained from real-time quantitative RT-PCR of mRNAs white adipose tissue (WAT) and liver, expressed in arbitrary units with the average of vehicle control = 1 unit. A: FSG67 decreased the expression of mRNA in WAT for lipogenic enzymes acetyl-CoA carboxylase-1 (ACC1), fatty acid synthase (FAS), and glycerol-3-acyltransferase GPAT (”mitochondrial“ primers, GPAT-1). B: RT-PCR for mRNAs in WAT: carnitinepalmitoyl-transerase-1 (CPT-1, liver form) and UCP-2. C: RT-PCR for lipogenic enzyme mRNAs in liver: ACC-1, FAS, and mitochondrial GPAT (GPAT-1). D: RT-PCR for CPT-1 and UCP-2 mRNAs in liver. Each mRNA/tissue significant by one-way ANOVA, Bonferroni comparisons: *P < 0.05 vs. vehicle and PF controls; #P < 0.05 vs. PF; @P < 0.05 vs. vehicle.

FAS gene expression in liver followed a pattern similar to that seen in WAT, suggesting downregulation of hepatic de novo fatty acid synthesis. FAS expression was decreased in FSG67-treated DIO mice vs. vehicle control (Fig. 7C) [F(2,18) = 8.604, P = 0.0024. t = 4.133, P < 0.001], though not relative to pair-fed control. ACC1 showed a similar pattern of expression but did not exhibit overall statistical significance (P = 0.088). Liver GPAT expression was unchanged after either FSG67 or pair-feeding. These RT-PCR data suggest that the capacity for de novo fatty acid synthetic capacity was diminished in liver after systemic treatment with the GPAT inhibitor FSG67. Gene expressions of the liver-form CPT-1 and for UCP-2 were measured; neither showed significant group differences.

Chronic FSG67 treatment of DIO mice improves glucose tolerance and insulin sensitivity.

DIO mice from the 16-day study were euthanized in an ad libitum-fed state (except PF group). We measured serum levels of triglyceride, cholesterol, and glucose (Fig. 8). Cholesterol levels were unchanged by treatment. Although triglyceride levels in FSG67 mice were 104.5 mg/dl, down from 144.0 and 134.9 mg/dl for vehicle and pair-fed controls respectively, the overall test did not reveal statistical differences among groups (P = 0.0892). Fed-state serum glucose in FSG67-treated DIO mice was 153.3 mg/dl, down from 200.6 and 189.0 mg/dl for vehicle and pair-fed controls respectively, but analysis did not reveal significant differences among groups (P = 0.2238).

Fig. 8.

Serum levels of triglyceride, cholesterol, and glucose in DIO mice. Mice were treated with vehicle (white bars) or FSG67 (5 mg·kg−1·day−1 ip, checkered bars) on ad libitum HF diet for 16 days, or pair-fed (PF; black bars) to the FSG67 group HF intake. Mice were killed at 18 h after final intraperitoneal injection, 6 h into the light phase.

Using mice in the 28-day chronic study (Fig. 5), we subjected mice on day 18 to GTTs (Fig. 9, A and B). Analysis of the GTT indicated significant effects of treatment [F(df2) = 16.15, P < 0.0001] over time [F(df5) = 52.24, P < 0.0001]. Although fasted glucose levels were not significantly different among treatments, the FSG67-treated mice showed significant reductions in both peak (Fig. 9A) and total blood glucose elevation [Fig. 9B, area under the curve (AUC): F(2,16) = 6.871, P = 0.007] vs. both vehicle and pair-fed controls (30-min: t = 4.222, P < 0.001 vs. vehicle, t = 2.83, P < 0.05 vs. pair-fed; AUC: t = 3.061 vs. vehicle, t = 3.376 vs. pair-fed, both P < 0.05). Thus, chronic FSG67 enhanced the ability of DIO mice to clear the blood of glucose. We also measured plasma insulin by ELISA at several time points during the GTT (Fig. 9, C and D), to ascertain how much insulin was released to exert these effects on glucose clearance. These data also showed significant effects of treatment [F(df2) = 9.8, P = 0.0002] and over time [F(df3) = 3.507, P = 0.0202], with significant differences in insulin AUC [F(2,16) = 5.039, P = 0.0201]. The FSG67 group had lower serum insulin than the pair-fed control at 0 min and 15 min (0 min: t = 2.778; 30 min: t = 2.631, both P < 0.05). Together, the data indicated that chronic treatment with the GPAT inhibitor FSG67 allowed more rapid clearance of blood glucose, with less insulin output required, and suggested enhanced insulin sensitivity.

Fig. 9.

Chronic daily treatment of DIO mice for 18 days with FSG67 increased glucose tolerance and insulin sensitivity. Mice received daily intraperitoneal injections of FSG67 (5 mg·kg−1·day−1, checkered bars) or vehicle (white bars), or received vehicle treatments plus pair-feeding (PF, black bars) to the level of HF intake by the FSG67 group. A: blood glucose levels during a glucose tolerance test (GTT) show enhanced clearance of a glucose bolus (1 g/kg) after FSG67. B: glucose area under the curve (AUC) during the GTT. C: serum insulin was measured in samples taken during the GTT. D: AUC for serum insulin during the GTT, showing lower insulin levels in FSG67-treated mice during the GTT. E: blood glucose levels during an insulin tolerance test (ITT) show that chronic FSG67 treatment enhanced glucose clearance after exogenous insulin bolus (1 U/kg). F: AUC for glucose during the ITT. All AUC panels had significant effects by one-way ANOVA. Time-course panels had significant effects of treatment and time by two-way ANOVA followed by Bonferroni group comparisons. For all panels: *P < 0.05 vs. vehicle control; #P < 0.05 vs. PF.

We followed up on these data by using the mice on day 26 for insulin tolerance tests (ITT; Fig. 9, E and F). Again, there were significant effects of treatment [F(df2) = 26.94, P < 0.0001] and time [F(df5) = 10.71, P < 0.0001], and significant differences in glucose AUC during the ITT [F(2,15) = 4.754, P = 0.0252]. The FSG67 group had glucose AUC lower than pair-fed controls during the ITT (t = 3.032, P < 0.05), with lower levels during 15–60 min (all t > 2.7, all P < 0.05). FSG67 treatment made mice on HF diet more sensitive to exogenous insulin and agree well with the data from the GTT analysis.

Chronic FSG67 decreases the size of white adipocytes and resolves hepatic steatosis in DIO mice.

On the day 28 of treatment, FSG67 and vehicle control mice were killed for terminal measurements. Consistent with the reductions in body weight and adiposity seen at earlier time points, chronic FSG67 decreased serum leptin levels to 5 ± 2.8 mg/dl vs. 28.1 ± 3.0 mg/dl in controls [t(df11) = 4.258, P = 0.0013]. FSG67 mice were clearly smaller (Fig. 10, A and B) and had smaller adipocytes in samples of epididymal WAT (Fig. 10, C and D). We noted that WAT from both groups, both maintained on a HF diet, had macrophage infiltration (termed “crown lesions,” arrow in Fig. 10, C and D) that had not been resolved with FSG67 treatment. FSG67-treated mice had complete or near-complete resolution of hepatic steatosis, as seen from Oil-Red-O staining of liver tissues (Fig. 10, E and F). In an earlier study, pair-fed mice had some resolution of fatty liver, with smaller lipid depots, but it was not as impressive as seen with FSG67 in that study (data not shown). Histopathological analysis of the liver showed no inflammation or hepatocellular injury. Glycogen, assessed in histologic sections of liver with PAS staining with and without diastase treatment, showed no consistent changes between treatment groups (data, images not shown). Inhibition of the GPAT4 isoform has been associated with dermal depletion of adipose tissue, and GPAT3 is expressed in the epidermis. Skin from representative treated and vehicle control animals are shown in Fig. 10, I and J. Although the FSG67-treated DIO mice were leaner than controls and had somewhat smaller subdermal adipocytes (Fig. 10, I and J), histopathological analysis of the epidermis and dermis confirmed the outward appearance of the mice, which showed no evidence of lipodystrophy or hair loss. Epidermal tissue, including sebaceous glands and dermal adipocytes, were normal in appearance. In addition to the skin, GPAT4 is highly expressed in mouse testis and is thought to function in sperm maturation. Histopathological analysis of testes in treated vs. control animals was unremarkable, with no evidence of dysmaturation of the sperm or reduction of spermatogonia (data not shown).

Fig. 10.

Visual and histological differences between chronic FSG67 and vehicle treatment after 28 days. A: representative mouse treated with FSG67 (5 mg·kg−1·day−1 for 28 days) is leaner than DIO vehicle-treated control (B). C: epididymal white adipose tissue from FSG67-treated mouse has smaller adipocytes than the DIO vehicle-treated control (D). “Crown lesions,” macrophages between the adipocytes were present in both groups. E: FSG67 substantially resolved hepatic steatosis in DIO mice; note intracytoplasmic large and small droplet fat accumulation prominent in the vehicle control (F). FSG67 also decreased HF-induced white adipose accumulation in brown adipose tissue (G) vs. vehicle control DIO (H). FSG67 treatment did not result in outward signs of lipodystrophy, and histopathology of the epidermis and dermis of the FSG67-treated (I) and vehicle control (J) animals showed no remarkable differences besides smaller adipocytes.

Intracerebroventricular FSG67 treatment reduced food consumption and body weight.

Systemic administration of FSG67 decreased food intake without causing malaise (Fig. 3A). Food intake modulation necessarily involves the central nervous system (CNS). We administered FSG67 intracerebroventricularly to determine whether GPAT inhibition could act centrally to reduce food intake. In separate groups, lean mice on chow (5 or 6/group) were treated with FSG67 intracerebroventricularly at doses of 100 or 320 nmol vs. vehicles. Within 24 h, mice treated with 100 nmol lost 3.0 ± 1.5% body weight (0.75 ± 0.4 g) [t(df5) = 2.913, P = 0.0333], while the 320 nmol group lost 7.2 ± 1.2% body weight (1.8 ± 0.3 g) [t(df5) = 5.783, P = 0.0044]; vehicle controls in both studies gained weight (Fig. 11A). Significant reduction in chow intake only occurred in the 320-nmol treatment group [Fig. 11B, 15.7 ± 0.3 kcal control, 10.4 ± 1.3 kcal FSG67; t(df4) = 3.614, P = 0.0225]. These data indicate that the reduction in food consumption accompanying GPAT inhibition may have a significant contribution from sites of direct FSG67 action in the CNS. Moreover, the occurrence of weight loss without a reduction of food intake in the 100-nmol group suggests a central effect on metabolism independent of changes in food intake behavior.

Fig. 11.

Intracerebroventricular administrations of FSG67 decreased mouse body weight and consumption of chow. Vehicle (2 μl sterile saline, white and dotted bars) and FSG67 (100 nmol, striped bar; 320 nmol, checkered bar) were administered intracerebroventricularly to the lateral ventricle in separate groups (5 or 6) in repeated measures design. A: intracerebroventricularly administered FSG67 produced weight loss. B: FSG67 reduction of food intake was dose-dependent. *P < 0.05 vs. matched vehicle control in unpaired two-tailed t-tests.

Acute and chronic FSG67 treatment alters hypothalamic neuropeptide expression.

Relative levels of gene expression for hypothalamic peptides were determined in lean and DIO mice treated with a single dose of FSG67 (20 mg/kg ip; see Fig. 1) and in chronically treated DIO mice from the 16-day study (data not shown) to identify potential neuropeptide mechanisms responsible for acute and chronic reduced food intake. The RT-PCR for lean mice is shown in Fig. 12A; data from the fasted mice are shown for reference, but not included in the analysis because feeding suppression by FSG67 was not akin to total food deprivation. In mice treated with high-dose FSG67, expression of the orexigenic hypothalamic neuropeptide Agouti-related peptide (AgRP) was significantly decreased compared with vehicle controls [t(df=9) = 2.815, P < 0.0202]. The orexigenic peptide NPY showed a similar pattern, although not statistically significant [t(df9) = 2.117, P = 0.0633]. Gene expression for proopiomelanocortin (POMC), precursor for the anorexigenic alpha-melanocyte-stimulating hormone was slightly but not significantly decreased compared with vehicle control [t(df9) = 2.126, P = 0.0625]. On balance, these data indicate that acute FSG67 did not elicit increases in expression of orexigenic neuropeptides or decreases in anorexigenic neuropeptides that could be expected from hypophagia. Indeed, AgRP expression was decreased. Unlike lean mice, single high-dose FSG67 treatment of DIO mice increased AgRP expression over that in vehicle controls and food-deprived animals (data not shown). Furthermore, food deprivation did not result in increased hypothalamic NPY or AgRP message in DIO mice. The pattern of increased orexigenic neuropeptide expression with acute FSG67 treatment is consistent with a hunger response and may indicate a rebound of orexigenic peptide expression in the DIO mice, or could represent a further example of dysregulated neuropeptide signaling in DIO mice (11). After chronic treatment with low-dose FSG67 (Fig. 12B), DIO mouse hypothalamus showed decreased expression of NPY vs. control [F(2,17) = 8.077, P = 0.0034; t = 3.417, P < 0.01], an indication that FSG67 suppressed expression of a “hunger” signal in the long term. NPY was also suppressed, however, in pair-fed DIO mice. This profile was similar to the acutely treated lean mice and may reflect normalization of the appetite response in the chronically treated DIO mice. This would seem consistent with the transient nature of hypophagia during chronic FSG67 treatment (Fig. 5B).

Fig. 12.

Systemic treatments with FSG67 decreased mRNA levels of orexigenic neuropeptides in hypothalamus. A: mice received ip FSG67 at 20 mg/kg (checkered bars) or vehicle (white bars). Another group was fasted (gray-striped bars). Fresh hypothalamus was removed 18 h later during the light period and was processed for RT-PCR. Because FSG67 mice did eat some food (Fig. 2B), the fasted group is included in the figure for illustration, but not in the analysis (unpaired two-tailed t-test). FSG67 decreased mRNA level for Agouti-related peptide (AgRP). B: DIO mice had chronic intraperitoneal treatments of FSG67 or vehicle, or were PF for 16 days. At 18 h after final injection, mRNA for neuropeptide Y (NPY) was decreased in lysates of whole hypothalamus from mice in FSG67 and PF groups (one-way ANOVA, Bonferroni comparisons). *P < 0.05 vs. vehicle control. POMC, proopiomelanocortin; CART, cocaine- and amphetamine-regulated transcript.


Typical obesity is accompanied by a series of complex and incompletely understood biochemical and physiological changes that are remarkably refractory to pharmacological manipulations that target single hormonal or neuropeptide factors, perhaps because of the redundancy of these pathways in regulating energy balance. In this study, we demonstrate that pharmacological inhibition of GPAT can reduce food intake, decrease body weight and adiposity, enhance energy utilization as fatty acid oxidation, reverse hepatic steatosis, and enhance insulin sensitivity. Overall, pharmacological GPAT inhibition ameliorates the dysregulated metabolic phenotype in DIO mice.

GPATs esterify long-chain fatty acyl-CoAs to the sn-1 position of glycerol-3-phosphate to produce lysophosphatidic acid, the first committed step in the synthesis of triglycerides and phospholipids (4). As such, they have garnered attention for their role as potential therapeutic targets for metabolic diseases such as obesity and diabetes (44). There are at least four mammalian GPAT isoforms that have been reviewed recently, encoded by different genes, with different subcellular and tissue distributions and sensitivities to inactivation by sulfhydryl-reactive compounds, such as NEM (13, 43, 51). Despite this diversity, they are all thought to share a common catalytic domain (3, 6, 33, 50) that can be exploited for the design of small-molecule inhibitors.

Rodent knockout and overexpression studies have provided some insight into the roles of GPAT isoforms in health and metabolic disease. Mice deficient in GPAT1, the mitochondrial NEM-resistant isoform, were the first to be developed, and showed moderate reduction of adipose tissue mass (16). Later studies demonstrated that GPAT1-null mice developed an obese phenotype when fed high-fat diets, although liver, plasma, and cardiac triglyceride contents were reduced (22, 34). Conversely, rodents with increased GPAT1 expression had increased hepatic and serum triglycerides (25, 31). GPAT2 was discovered as the residual mitochondrial NEM-sensitive GPAT activity remaining in GPAT1-null mice (24). Although GPAT2-deficient mice have not yet been produced, the activity of the GPAT2 isoform could have contributed to the development of the obese phenotype in the GPAT1-deficient mice. GPAT4-deficient mice were unable to nurse due to a lack of milk production (7). In contrast to GPAT1 knockout mice, these animals were not only resistant to obesity, but developed a marked reduction in subcutaneous adipose tissue, similar to that seen in subdermal lipodystrophy (48).

Given the experience with GPAT isoform knockouts, a key consideration for pharmacological GPAT inhibition has been whether to target a particular isoform, or to target GPAT activity more broadly. In practice, it might not yet be feasible to design compounds to target the specific isoforms, due to their similar catalytic domains. The crystal structure of GPAT from squash chloroplast was reported in 2001 (47) and has served as our template for designing GPAT inhibitors (53, 54) because crystal structures for the mammalian isoforms are not yet available. In the current studies, our small-molecule GPAT inhibitor FSG67 likely inhibited multiple GPAT isoforms, although possibly to different extents. FSG67 is effective against both mitochondrial GPAT1 and GPAT2 isoforms (see results and Ref. 53). In addition, FSG67 inhibits acylglyceride synthesis in 3T3-L1 adipocytes, which preferentially express GPAT3, the major microsomal isoform (6). The effect of FSG67 on GPAT4 has not yet been investigated, though it is worth noting that there was no evidence of subdermal lipodystrophy in the FSG67-treated animals. In 3T3-L1 adipocytes, FSG67 reduced triglyceride synthesis, which was reflected in a concentration-dependent reduction of number and typical size of intracellular lipid droplets. With systemic treatment, FSG67 decreased the size of adipocytes in epididymal WAT, a tissue similar to differentiated 3T3-L1 cells in their high expression of GPAT3.

Although phospholipid synthesis was also decreased in 3T3-L1 cells, FSG67 was not toxic to the cells at concentrations tested up to 150 μM. One possible explanation for the lack of in vitro toxicity in 3T3-L1 adipocytes is that a recently described GPAT-independent peroxisomal acylglyceride synthesis may provide adequate lipid production for cell survival (14). Overall, the lack of in vitro and in vivo toxicity of FSG67, despite its other effects is encouraging. The IC50 values in vitro for FSG67 effects on GPAT activity and lipid synthesis have been in the micromolar range. More typically, pharmacological efficacy is considered strong when a compound has in vitro effects in the nanomolar range. We are mindful that the IC50 values that we measured did not use purified GPAT in the assays, and thus actual IC50 may be significantly lower. However, if overinhibiting GPATs, or any one isoform, may carry serious health risk, it could be desirable instead to inhibit multiple GPATs at a more modest level. Nevertheless, work continues in the effort to design new GPAT inhibitors that target the catalytic domain more effectively (54), and such compounds may offer similar degree of biological effect as FSG67, but at lower concentration and dose ranges.

The phenotype of the DIO mice changed with chronic GPAT inhibition. Chronic daily systemic treatment of DIO mice with a low dose of FSG67 produced a gradual, yet significant, decrease in body weight accompanied by overall reduction in energy intake. Across studies, DIO mice typically lost 12% of their initial body weight, whereas vehicle controls gained slight weight, and pair-fed controls lost no more than 6% body weight. Cessation of FSG67 at day 9 did not result in rebound hyperphagia or rapid regain of body weight during the next 6 days of posttreatment monitoring. When FSG67 treatment was continued, mice remained at the lower plateau of body weight for the remainder of long experiment. Weight loss maintenance occurred, despite a return to normal energy intake between days 10 and 11. Interestingly, there was no rebound hyperphagia over the next 16 days of food intake measurements in that experiment. Together, these data suggest several possibilities. The initial hypophagia probably did contribute to the weight loss with FSG67; pair-feeding produced some weight loss. However, FSG67 likely caused other alterations in metabolic profile that enabled additional weight loss beyond that accounted for by hypophagia, Indeed, it is well known that one of the consequences of reduced energy intake is a decrease in overall metabolic rate, due to numerous interrelated and redundant mechanisms that conserve energy expenditure when the organism is in negative energy balance. Yet, FSG67-treated mice lost significantly more weight than pair-fed controls. These other metabolic changes were likely significant, because even after energy intake returned to normal, FSG67 mice remained lean.

Noninvasive indirect calorimetry of DIO mice during chronic treatment showed that the food-restricted pair-fed controls had characteristic decreases in rates of oxidative metabolism (V̇o2) and energy expenditure. Pair-fed mice also had lower respiratory exchange ratio, indicative of a shift toward fat oxidation. This effect was evident, even against the background of already-low RER due to the high availability of fat from the HF diet. The FSG67 mice, though hypophagic, did not show the same degree of suppression of V̇o2 as pair-fed mice, suggesting relative increase in rates of oxygen usage in oxidative metabolism. In addition, the FSG67 had decreased RER in the FSG67 group, relative to vehicle control, and importantly also compared with pair-fed control. Complete fat oxidation requires greater oxygen input than for carbohydrate oxidation, because the mole-to-mole ratio of carbon to hydrogen for fat is lower than the 1-to-1 ratio in carbohydrate, thus resulting in lower RER values (RER = V̇co2/V̇o2). Although the degree of completion of fat oxidation at the level of individual tissues was not determined in these studies, we interpret the current calorimetry data as an increase in fatty acid oxidation that contributes to relative increase in V̇o2 vs. pair-fed controls. Just as FSG67-treated mice were somewhat protected from the lower V̇o2 from hypophagia, they were also protected from the concomitant decrease in overall energy expenditure. We think that this contributed to the ability of FSG67 treatment to maintain mice at lower body weights.

That a treatment that inhibits lipogenesis could increase fatty acid oxidation is not without precedent. The location of GPATs on the outer mitochondrial membrane allows them to compete for long-chain acyl-CoAs with carnitine palmitoyltransferase-1 (CPT-1) (13). As such, GPAT inhibition would tend to direct long-chain acyl-CoAs away from acylglyceride synthesis, and toward CPT-1-mediated transport into the mitochondria for oxidation. This hypothesis has been tested by other laboratories, with a focus on the liver. Consistent with the hypothesis, increased levels acyl-carnitines and β-hydroxybutyrate were identified in livers from GPAT1-deficient mice (17), and hepatic fat oxidation was increased in ex vivo preparations. Conversely, adenoviral overexpression of hepatic GPAT1 led to steatosis and reduced hepatic fatty acid oxidation ex vivo (25). However, overall indirect calorimetry data from GPAT1 knockout mice had not been consistent with the hepatic and blood-based data (56). In our study with pharmacological GPAT inhibition, DIO mice showed increases in fat oxidative metabolism relative to pair-fed controls.

After nearly 2 wk of chronic FSG67, quantitative NMR analysis of whole animal adiposity showed clearly that a selective loss of fat mass accounted for the decrease in body weight. Lean, fat-free mass was unaffected. Because the FSG67-treated mice were no longer hypophagic, it followed that the pair-fed controls by this point had opportunity to regain their previous weight losses; they weighed the same as vehicle controls and had similar levels of adiposity at the time of quantitative NMR. This experiment likely missed a time point at which pair-fed controls would have shown an intermediate fat loss, with fat mass between that of FSG67- and vehicle-injected control DIO mice. The return to normophagia between days 10 and 11 had not been anticipated, because pair-fed mice in a different trial had shown clear intermediate weights between those of vehicle controls and FSG67-treated mice on day 16 (data not shown), when tissues were taken for RT-PCR analyses, and because mice in the indirect calorimetry study did not show rapid return to normophagia after cessation of FSG67 on day 9. Intertrial differences in rate of weight loss aside, FSG67 at a dose of 5 mg·kg−1·day−1 has reliably produced ∼12% weight loss across trials. The weight loss is from the fat mass. Adipocytes are smaller in FSG67-treated mice than in controls. Consistent with this, serum leptin levels are substantially deceased after chronic FSG67.

The phenotype of metabolic enzyme expression changed with GPAT inhibition in DIO mice. Lipogenic enzyme gene expression was substantially decreased in both WAT and liver, suggesting a reduction of de novo lipogenesis (ACC1, FAS). Moreover, both GPAT and PPARγ expression were downregulated in WAT. Therefore, decreased fat deposition, as well as potential increased fat oxidation, contributed to the decreased adiposity and body weight in FSG67-treated DIO mice. While fatty acid oxidation was increased relative to pair-fed animals by whole-body calorimetry measurement, it was not reflected in upregulation of l-CPT-1 expression in liver or WAT. This is just one of the enzymes important for fatty acid oxidation in mitochondria. The current RT-PCR analysis did not include skeletal muscle, which is a large organ mass that can oxidize fatty acids. It also did not include analysis of BAT, which can oxidize fat as well, and has the ability to uncouple oxidative phosphorylation using UCP-1, which could allow for more energy expenditure in the form of liberated heat. Histological analysis of BAT did show decreased fat content. Further analysis of WAT in regard to lipolytic enzymes could provide further insight into mechanisms of FSG67 effects to decrease fat mass in vivo. In WAT, a decrease in UCP2 expression was seen in pair-fed mice, but not with FSG67 treatment. Increased UCP2 expression has been associated with increased fatty acid oxidation (41). It also mediates proton leakage in the mitochondria, reducing the efficiency of oxidative phosphorylation (5). Thus, preservation of UCP2 expression with FSG67 could help permit oxidation of fat from WAT.

There was remarkable decrease of fat from the livers of FSG67-treated mice. Hepatic steatosis was clearly resolved by 28 days of FSG67, with Oil-Red-O-stained lipid droplets virtually gone in most mice. This is consistent with previous trials, in which pair-fed mice showed intermediate effect (data not shown). The resolution of steatosis occurred without visual evidence of liver histopathology. Livers from FSG67-treated and control mice had similar levels of glycogen as evident from PAS staining, though this was not a quantitative measure. Expression of mRNA for FAS was decreased in livers from FSG67-treated mice compared with both vehicle- and pair-fed controls, indicating reduced capacity for de novo lipogenesis. The hypothesis that GPAT inhibition could result in increased fatty acid oxidation as well has been supported by other laboratories focused on hepatic effects of GPAT1. They have shown clear effects of GPAT1 knockout and overexpression on hepatic fat content (17, 25). FSG67 inhibits GPAT1 (results). The resolution of hepatic steatosis in our DIO mice given FSG67 is thus consistent with reports from these other laboratories.

Hepatic steatosis is well known to contribute to insulin resistance seen in both mice (32) and humans (20). The association of hepatic steatosis and insulin resistance has been dramatically demonstrated in clinical studies in which treatment of diabetic patients with PPARγ agonists, such as pioglitazone and rosiglitazone, reduced hepatic steatosis and improved insulin resistance (52). Although there are a number of proposed mechanisms linking hepatic steatosis to insulin resistance, accumulation of diacylglycerol in the hepatocyte or other insulin-sensitive tissues is thought to activate PKC isoforms that interfere with insulin action leading to insulin resistance (38). Since inhibition of GPAT would reduce the amount of lysophosphatidic acid available for diacylglycerol synthesis, GPAT inhibition would be expected to enhance insulin sensitivity. After chronic treatment of DIO mice with FSG67 for 18 days, glucose tolerance was improved in a GTT. Furthermore, insulin secreted in response to the glucose bolus was also significantly decreased. This suggested that the insulin secreted was more effective at clearing blood glucose after FSG67 treatments. An ITT performed at day 26 of treatment showed that the mice treated with FSG67 had enhanced ability to clear blood glucose in response to exogenous insulin bolus. The mechanisms by which systemic treatment with a GPAT inhibitor improves glycemic control need to be identified. Insight into tissue-specific contributions and effects can certainly come from studies in mice with knockout or knockdown of GPATs in select tissues. Hepatic GPAT1 knockout in ob/ob mice exhibited reduced steatosis and improved plasma glucose and cholesterol levels (55), whereas hepatic overexpression of GPAT1 caused insulin resistance (31). Examination of all tissues that could contribute to the effect, including skeletal muscle, adipose depots, liver, and pancreas, should be undertaken, using both pharmacological and genetic approaches.

Although resolution of hepatic steatosis seems desirable, it may come with certain risks, depending on the approach, with regard to enzyme targeting and degree of inhibition. Modest inhibition of GPATs might provide some advantage over other methods. Studies in animal models of nonalcoholic fatty liver disease demonstrate that inhibition of diacylglycerol acyltransferase, the enzyme that catalyzes the final step in triglyceride synthesis, improves hepatic steatosis and insulin sensitivity [reviewed in (8)]. Hepatic-specific inhibition of diacylglycerol acyltransferase with antisense oligonucleotides improved hepatic steatosis in obese, diabetic mice. However, it also exacerbated injury and fibrosis. Free fatty acids accumulated in the liver, leading to induction of fatty acid oxidizing systems that increased hepatic oxidative stress and liver damage (8). In contrast, total knockout of GPAT1 in mice, though it also resulted in resolution of hepatic steatosis with concomitant increase in reactive oxygen species and apoptosis, did not result in liver damage, because increased liver cell proliferation counterbalanced the apoptosis (15). This was observed in a total knockout of GPAT-1. Our approach with FSG67 in vivo is likely a moderate inhibition, which may make an important difference in risk of potential for liver disease. These possibilities must be assessed in future studies. It is important to point out that our DIO mice were continually maintained on high-fat diet during FSG67 treatment. Continued provision of fatty acids from a HF diet could lead to higher fatty acyl-CoA in liver, increased fatty acid oxidation, and generation of reactive oxygen species, carrying some risk for potential liver disease. Furthermore, histological examination of WAT in our study revealed that FSG67 did not resolve the macrophage infiltration characteristic of consumption of HF. In a clinical setting, FSG67 would more likely be an adjunct treatment to dietary and lifestyle modifications.

One of the features of pharmacological GPAT inhibition was the reduction in food intake in both lean and DIO mice. Analyses of food intake were not included in the GPAT isoform knockout studies. Moreover, the lack of significant CTA with the doses of FSG67 used in these experiments indicated that the decreased food intake was not likely due to sickness behavior, but it was likely a specific effect on appetite. Alterations in food intake necessarily involve CNS involvement, whether as a response to feedback signals from the periphery, or by CNS initiation. We assessed whether FSG67 could exert some of its actions by direct action in the brain. FSG67 administered at nanomole doses intracerebroventricularly achieved dose-dependent, reductions in animal weight within 24 h. Only the high (320 nmol) dose of FSG67 induced a reduction in food intake. This dose is on par on a per-weight basis with the 15.3 μmol/kg (5 mg/kg) systemic dose that did not reduce food intake within the initial 24 h, and 6-fold lower than the 61.1 μmol/kg dose (20 mg/kg) that produced clear acute weight loss. These comparisons suggest that the FSG67 given intracerebroventricularly did not leak to the periphery at levels that would be sufficient to produce hypophagia or weight loss. The reduction in animal weight without a concomitant reduction in food intake in the 100-nmol group, suggests that a centrally mediated effect on whole body animal metabolism could account for some of the weight loss. The ability of intracerebroventricular FSG67 to decrease food intake at higher doses suggests that CNS sites of action may also play a role in the reduction of food intake with GPAT inhibition. Mitochondrial GPAT (23) and the microsomal isoforms (6, 7) have measurable expression by mRNA analysis in the brain, but regional differences are unknown. Although there was little to no mRNA expression for GPAT1 in whole brain (18), samples of specific regions might yield different results. Hypothalamic and hindbrain neurons in brain sites that control food intake may be more readily exposed to systemic circulation, enabling access of small molecules such as FSG67. The ability of FSG67 to access brain sites will have to be assessed in pharmacokinetics studies. A different small-molecule compound that we have made, C75, which decreases food intake and body weight through other modulations of fatty acid metabolism, is effective by intracerebroventricular administration (1, 26).

Hypothalamic neuropeptide expression was measured in lean and DIO mice treated intraperitoneally with higher acute dose of FSG67 and also after 16 days of chronic treatment of the DIO mice. In acutely treated lean mice, FSG67 significantly decreased the expression of AgRP, further supporting the hypothesis of a centrally mediated reduction in food intake. NPY expression was reduced in the chronically treated DIO mice. Similar changes in hypothalamic neuropeptide expression were seen with prior studies of pharmacological fatty acid oxidation stimulation with C75 in DIO mice (26, 46), as well as in rats (1). In the acutely treated DIO mice, AgRP expression was increased (data not shown), suggesting a rebound of orexigenic peptide expression in the DIO mice and a hunger-like response. This paradoxical effect could represent a further example of dysregulated neuropeptide signaling in DIO mice (11). Nonetheless, both intracerebroventricular treatment with FSG67 and analysis of hypothalamic peptide profiles suggest a CNS component contributing to the mechanism of action of FSG67 involving both the regulation of systemic metabolism and food intake. The interplay between GPAT inhibition in peripheral targets, such as liver and adipose tissue and in the brain is no doubt complicated and may involve alteration of triglyceride synthesis, fatty acid oxidation, adipose-derived hormones, and neural afferent and efferent pathways. Sorting out the relative contributions of these mechanisms will further our understanding of how fatty acid metabolism contributes to overall metabolic homeostasis.

Strategies to treat metabolic syndrome will likely require a multiple-target approach, given the complex and redundant regulation of metabolism and feeding behavior. The mouse knockout studies of GPAT isoforms identified this enzyme as a potential target for obesity drug development. Our studies extend these initial molecular observations, demonstrating that pharmacological inhibition of GPATs can decrease adiposity and hepatic steatosis, enhance fatty acid oxidation, and reduce food intake leading to enhanced insulin sensitivity and progression toward a lean phenotype.


This work was supported by National Institutes of Health National Institute of Neurological Disorders and Stroke Grant R01 NS041079 to G. V. Ronnett, Sponsored Research Agreement between FASgen and G. V. Ronnett, and Research Collaboration between S. Aja and the Daegu-Gyeongbuk Institute of Science and Technology Convergence Science Center.


Under a license agreement between FASgen, and the Johns Hopkins University, F. P. Kuhajda, G. V. Ronnett, C. A. Townsend, and L. E. Landree are entitled to a share of royalties received by the University on future sales of products that might be developed related to reagents described in this article. G. V. Ronnett and C. A. Townsend own FASgen stock, which is subject to certain restrictions under university policy. The Johns Hopkins University, in accordance with its conflict of interest policies, is managing the terms of this arrangement.


We thank J. Thupari for technical assistance. K. Daniels and K. Wong worked with S. Aja through participation in Research Practicum, Baltimore Polytechnic Institute.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
View Abstract