Research ArticleMetabolic Disease

Development of a therapeutic monoclonal antibody that targets secreted fatty acid–binding protein aP2 to treat type 2 diabetes

See allHide authors and affiliations

Science Translational Medicine  23 Dec 2015:
Vol. 7, Issue 319, pp. 319ra205
DOI: 10.1126/scitranslmed.aac6336

Kill the messenger

A variety of metabolic messengers—many from adipose tissue itself—controls the energy state of organs and organisms. Recently, researchers showed that the fatty acid binding protein aP2, once thought to live and work only in the cytoplasm, is also secreted by adipose tissue and spurs metabolic changes in other organs. Now, Burak and colleagues test whether secreted aP2 can serve as a therapeutic target for type 2 diabetes.

In mice, the secreted form of aP2 regulates glucose production in liver, systemic glucose homeostasis, and insulin resistance. Serum levels of aP2 were shown to be elevated in obese mice and humans and to correlate with metabolic complications. The authors identified a monoclonal antibody to aP2 that lowered fasting blood glucose, increased insulin sensitivity, and lowered both fat mass and fatty liver (steatosis) in obese mouse models, relative to a control antibody, but not in aP2-deficient mice. The antidiabetic effects of the therapeutic antibody were linked to the regulation of hepatic glucose output and peripheral glucose utilization. Together, these findings suggest that an aP2-targeted antibody that kills the messenger is a viable approach for diabetes treatment.


The lipid chaperone aP2/FABP4 has been implicated in the pathology of many immunometabolic diseases, including diabetes in humans, but aP2 has not yet been targeted for therapeutic applications. aP2 is not only an intracellular protein but also an active adipokine that contributes to hyperglycemia by promoting hepatic gluconeogenesis and interfering with peripheral insulin action. Serum aP2 levels are markedly elevated in mouse and human obesity and strongly correlate with metabolic complications. These observations raise the possibility of a new strategy to treat metabolic disease by targeting serum aP2 with a monoclonal antibody (mAb) to aP2. We evaluated mAbs to aP2 and identified one, CA33, that lowered fasting blood glucose, improved systemic glucose metabolism, increased systemic insulin sensitivity, and reduced fat mass and liver steatosis in obese mouse models. We examined the structure of the aP2-CA33 complex and resolved the target epitope by crystallographic studies in comparison to another mAb that lacked efficacy in vivo. In hyperinsulinemic-euglycemic clamp studies, we found that the antidiabetic effect of CA33 was predominantly linked to the regulation of hepatic glucose output and peripheral glucose utilization. The antibody had no effect in aP2-deficient mice, demonstrating its target specificity. We conclude that an aP2 mAb–mediated therapeutic constitutes a feasible approach for the treatment of diabetes.


Adipose tissue is an endocrine organ that secretes numerous cytokines, adipokines, and lipokines that regulate aspects of systemic metabolic homeostasis, and obesity-related alterations in the endocrine output of adipose tissue are linked to the development of a variety of metabolic disorders (13). Adipose tissue integration with systemic metabolism also requires the trafficking and signaling of lipids in adipocytes with proper coordination of local and systemic metabolic adaptations, a facet of adipose tissue function that is not well understood. Fatty acid–binding proteins (FABPs), in particular FABP4/aP2, play a critical role in this capacity and contribute to metabolic maladaptive responses under pathological conditions such as metabolic stress (46). In mice, genetic FABP deficiency results in significant protection against the dysregulated metabolic outcomes of obesity and diabetes, improves the lipid profile in adipose tissue with increased levels of C16:1n7-palmitoleate, reduces hepatosteatosis, and improves control of hepatic glucose production (HGP) and peripheral glucose disposal (710). In addition, genetic deficiency or pharmacological blockade of FABP4/aP2 potently reduces atherosclerotic lesions and improves dyslipidemia (1114). Hence, aP2 plays a critical role in many aspects of the development of metabolic disease in preclinical models.

In the past two decades, the biological functions of FABPs in general and aP2/FABP4 in particular have primarily been attributed to their actions as intracellular proteins (5, 6). Hence, much of the biological function at distant organs has been considered secondary to changes that occur directly in the adipose tissue, including profiles of lipids produced at this site. However, we and others recently demonstrated that, in addition to its cytoplasmic presence, aP2 is actively secreted from adipose tissue through a nonclassical regulated pathway (1519). aP2 secretion is responsive to nutritional cycles and stimulated by signals that are coupled to fasting and lipolysis (1517). The secreted form of aP2 acts as an adipokine—a peptide that is secreted by adipose tissue and acts on distant organs—and regulates HGP and systemic glucose homeostasis (15), and contributes to insulin resistance in mice (20). Serum aP2 levels are significantly elevated in obese mice, and blocking aP2 in these mice with a polyclonal antibody attenuates type 2 diabetes (15). The same patterns are also observed in human populations in which aP2 levels are significantly increased in obesity and strongly correlate with metabolic and cardiovascular diseases in multiple independent human studies (2132). Last, humans who carry a haploinsufficiency allele that results in reduced aP2 expression are protected against diabetes and cardiovascular disease, as demonstrated in independent populations (33, 34). Together, these findings suggest that the biological and hormonal functions of aP2 are conserved and relevant to human pathophysiology and that secreted aP2 might represent a well-validated and promising target for the development of a new kind of diabetes therapeutic. Furthermore, these potentially paradigm shifting lines of evidence related to aP2 biology also illustrate the possibility of generating a new therapeutic entity based on an anti-aP2 monoclonal antibody (mAb), which may overcome the existing translational hurdles of targeting intracellular aP2 in metabolic diseases.

As proof of principle of this therapeutic direction, we produced and evaluated a number of mAbs raised against aP2 (anti-aP2). One of these, designated CA33, improved glucose metabolism, reduced fat mass, increased systemic insulin sensitivity, and reduced liver steatosis in obese mice. We resolved the crystal structure of this antibody and defined its molecular interactions with the aP2/FABP4 protein. Using hyperinsulinemic-euglycemic clamp studies, we showed that the antidiabetic effect of this antibody was predominantly linked to the reduction of hepatic glucose output and increased peripheral glucose utilization. These studies provide strong preclinical support for the potential of aP2-targeted therapies for diabetes and fatty liver disease and present a candidate molecule that could be used for such a purpose.


Anti-aP2 mAb development and evaluation

Obesity is associated with increased levels of circulating aP2, which contributes to the elevation of HGP and reduced peripheral glucose disposal (15) and insulin resistance (20), characteristics of type 2 diabetes (35). Therefore, neutralizing serum aP2 might represent an efficient approach to treat diabetes and possibly other metabolic diseases (15). To explore this therapeutic possibility, we produced and evaluated mouse (HA3) and rabbit-mouse hybrid mAbs (CA186_02033, CA186_02013, CA186_02015, and CA186_02023; hereinafter referred to as CA33, CA13, CA15, and CA23, respectively) raised against the full-length human and mouse aP2 peptides. Measurement of binding affinity to human and mouse aP2 in the presence of lipid by surface plasmon resonance using a Biacore T200 system demonstrated a wide range of affinities for these antibodies, from the micromolar to the low nanomolar range (Fig. 1A). As an initial test for the potential effects of these antibodies in vivo, we administered them subcutaneously for 4 weeks to mice with high-fat diet (HFD)–induced obesity (Fig. 1B). The HFD feeding resulted in a rise in serum insulin levels during the experiment, an effect that was blunted by treatment with the mouse antibody HA3 and reversed by the hybrid antibody CA33, but unaltered by the other three hybrid antibodies tested (Fig. 1C). CA33 also significantly decreased fasting blood glucose (Fig. 1D), whereas the other antibodies tested did not improve glycemia, suggesting that CA33 reduced the insulin resistance associated with HFD and improved glucose metabolism.

Fig. 1. Anti-aP2 mAb development.

(A) Binding affinity of aP2 mAbs to human and mouse aP2, determined by biomolecular interaction analysis. (B) In vivo study design and antibody regimen. Mice (20-week-old) were on HFD for 15 weeks at week 0 (n = 10 mice per group). (C) Plasma insulin levels before and after 4-week treatment with vehicle or anti-aP2 mAbs, measured after 6 hours of daytime food withdrawal. (D) Blood glucose levels in obese mice on HFD before and after a 4-week treatment of vehicle or anti-aP2 mAbs. Blood glucose levels were measured after 6 hours of daytime food withdrawal. (E) GTT performed after 2 weeks of treatment in obese mice on HFD with vehicle or anti-aP2 mAbs (glucose, 0.75 g/kg). (F) ITT performed after 3 weeks of treatment in obese mice on HFD with vehicle or anti-aP2 mAbs (insulin, 0.75 IU/kg). (G) Body weights in obese mice on HFD before and after treatment with vehicle or anti-aP2 mAbs. Weight was measured in the fed state. For (C), (D), and (G), statistical analysis was performed by Student’s t test. For (E) and (F), comparisons are by repeated-measures two-way analysis of variance (ANOVA). *P < 0.05, **P < 0.01.

We verified the systemic improvement in glucose metabolism using a glucose tolerance test (GTT). CA33 therapy resulted in significantly improved glucose tolerance (Fig. 1E), whereas the other antibodies did not improve glucose tolerance, as demonstrated by similar glucose disposal curves compared to vehicle (fig. S1A). Furthermore, only CA33 treatment markedly improved insulin sensitivity [as demonstrated in the insulin tolerance tests (ITTs)], whereas other antibodies tested were similar to vehicle (Fig. 1F and fig. S1B). There was a moderate reduction in weight gain in all but one of the antibody-treated groups (CA15), although this observation did not correlate with improvement in glucose metabolism (Fig. 1G). Together, these results demonstrated that, of the small panel tested, only CA33 has antidiabetic properties.

Characterization of low affinity anti-aP2 mAb CA33

We next sought to characterize the CA33 antibody to better understand its therapeutic properties. In an Octet-binding assay, all of the antibodies tested demonstrated saturable binding to aP2. There was a measurable but low interaction with the related protein FABP3 (~25% of the aP2/FABP4 interaction) and only minor interaction with FABP5 (also known as mal1) that was similar to control immunoglobulin G (IgG) binding (Fig. 2A). We also found that the improvement in glucose homeostasis in CA33-treated mice was related to an effect of this antibody on circulating aP2 levels. After 4 weeks of treatment, CA33-treated mice maintained circulating aP2 levels of a magnitude similar to or slightly lower than that seen in control-treated animals, whereas all other antibodies including HA3 resulted in a marked 10-fold increase in circulating aP2 levels (Fig. 2B). Indeed, circulating aP2 was undetectable by Western blot in both control and CA33-treated mice, but robustly evident in serum after HA3 treatment (Fig. 2B, inset).

Fig. 2. Characterization of the lead antibody CA33.

(A) Octet analysis of antibody binding to aP2 and closely related lipid-binding proteins. (B) Plasma aP2 levels in HFD-fed mice treated with vehicle, CA33, or HA3 for 3 weeks (n = 10 mice per group). Mice had been on HFD for 12 weeks before the experiment was initiated. Western blots were carried out to detect aP2 in the serum of three mice from each group (inset). **P < 0.01 by Student’s t test. (C) Antibody cross-blocking as determined on Biacore. (D) Hydrogen-deuterium exchange mass spectrometry (HDX) identification of aP2 residues involved in the interaction with CA33 and HA3. Residues involved in the interaction are shown in red. (E) Ribbon diagram depicting the secondary structure elements of the Fabs CA33 and HA3 in complex with aP2. Image was generated by superimposing the two independent Fab-aP2 crystal structures. (F) High-resolution mapping of the CA33 epitope (purple) on aP2 (khaki). Interacting residues in both molecules are shown as stick models. Hydrogen bonds are dashed lines. The side chain of K10 in aP2 forms a hydrophobic interaction with the phenyl side chain of Y92.

To begin characterizing the target sites, we performed cross-blocking experiments. CA33 partially blocked binding of the ineffective mouse antibody HA3 to aP2, whereas HA3 binding was completely blocked by the hybrid antibodies CA13 and CA15 (Fig. 2C). These data suggest that the epitope recognized by CA33 only partially overlaps with that recognized by HA3. In further analysis, epitope identification based on hydrogen-deuterium exchange mass spectrometry indicated that CA33 interacts with the first α helix and the first β strand of aP2 on residues 5 to 16, 20 to 28, and 118 to 132, which partially overlapped with the epitope identified for HA3 (Fig. 2D). To understand these epitopes precisely, we then cocrystallized the Fab fragments of CA33 and HA3 with aP2 (Fig. 2E). Analyses of the crystals showed that CA33 binds an epitope that is spread out over the secondary structure elements β1 and β10 and the random coil regions linking α2 to β2 and β3 to β4, and includes the aP2 amino acids T57, K38, L11, V12, K10, and E130 (Fig. 2F).

Despite the partial blocking of HA3 by CA33, we observed no direct overlap of their epitopes. In addition, the low affinity of the CA33 Fab can be explained by the crystal structure. Only one amino acid in the heavy chain of CA33 makes a contact with aP2, and most of the contacts are through the light chain (Fig. 2, E and F). In contrast, HA3-aP2 contact was found to be more conventional, with both Fab chains interacting with aP2. The structure also showed that CA33 does not bind to the “lid” of the β barrel (S14 to A37), which has been postulated to control the access of lipids to the binding pocket (36), or the “hinge,” which contains E14, N15, and F16 (37). In addition, we found that lipid binding to aP2 was not substantially altered by the presence of CA33 (fig. S2). In contrast, HA3 was shown to bind directly to the lid but has limited activity. Cumulatively, these experiments suggest that CA33 activity may be independent of general aP2 lipid binding.

Given the relatively low affinity of CA33 for aP2, we investigated whether its efficacy is related to off-target effects. We reasoned that the existence of such potential effects could be best explored if CA33 generated similar metabolic improvements in the absence of aP2. Accordingly, we tested the effect of CA33 treatment in aP2−/− mice fed a HFD and found that antibody therapy failed to induce any change in weight or fasting glucose in this model (Fig. 3, A and B). Furthermore, CA33 did not affect glucose tolerance in obese aP2−/− mice (Fig. 3C), clearly demonstrating that the antibody’s effects result from targeting of aP2.

Fig. 3. Metabolic profiles and target specificity of CA33 activity.

Mice deficient in aP2 were placed on HFD to examine potential CA33 effects. (A) Fasting blood glucose in aP2−/− mice before and after antibody or vehicle treatment. (B) Body weight in aP2−/− mice before and after 3 weeks of antibody or vehicle treatment (n = 10 mice per group). (C) GTT performed after 2 weeks of antibody treatment in aP2−/− mice. (D) Body weight in ob/ob mice before and after 3 weeks of antibody or vehicle treatment (n = 10 mice per group). (E) Fasting blood glucose in ob/ob mice before and after antibody or vehicle treatment. (F) Plasma insulin levels in ob/ob mice after 3 weeks of vehicle or antibody treatment. (G) GTT performed after 2 weeks of antibody treatment in ob/ob mice. Statistical analysis was performed by Student’s t test (A, B, and D to F) and repeated-measures two-way ANOVA (C and G). *P < 0.05, **P < 0.01.

Last, we investigated the effect of CA33 in a second model of severe genetic obesity and insulin resistance: leptin-deficient ob/ob mice. Over the course of 3 weeks of treatment, both the CA33 and vehicle-treated groups gained weight but the extent was less in the CA33-treated group (Fig. 3D), and hyperglycemia was normalized in CA33-treated ob/ob mice compared to controls (Fig. 3E). The extent of hyperinsulinemia was also partially reduced in the CA33-treated animals (Fig. 3F). Lower serum glucose and insulin levels suggested improved glucose metabolism upon neutralization of aP2. Indeed, after administration of exogenous glucose, CA33-treated ob/ob mice also exhibited significantly improved glucose tolerance compared to vehicle-treated mice, despite the presence of massive obesity (Fig. 3G). These data underscore the broad potential applicability of aP2 neutralization to metabolic disease in independent preclinical models and are consistent with the results obtained with genetic deficiency in the context of dietary and genetic obesity (810).

Improvement in lipid metabolism and hepatosteatosis in obese mice

Having identified and physically characterized a candidate mAb that could generate metabolic benefits, we embarked on detailed functional studies in the HFD mouse model. To further explore the metabolic effects of CA33 treatment, we treated HFD-induced obese mice with CA33 or vehicle for 5 weeks and examined the effects on the liver. As expected, long-term high-fat feeding induced steatosis and triglyceride accumulation in the liver of vehicle-treated mice; however, these effects were significantly ameliorated by CA33 treatment (Fig. 4, A and B), and the improvement in liver lipid homeostasis was accompanied by reduced hepatic expression of key genes involved in de novo lipogenesis (Fig. 4C). Similar to what has been described in the setting of genetic FABP deficiency (9, 3841), CA33-treated mice had moderately higher plasma free fatty acid levels (Fig. 4D) and lower glycerol levels (Fig. 4E), suggesting that circulating aP2 may play a role in this biology. Total cholesterol levels were also lower in the CA33-treated group (Fig. 4F), although there was no significant difference in plasma triglycerides relative to the untreated group (Fig. 4G). Notably, although whole-body genetic aP2 deficiency is associated with a substantial up-regulation of adipose tissue mal1/FABP5 expression (8, 9, 42, 43), we did not detect significant changes in circulating mal1 protein levels with CA33 treatment, indicating a lack of compensatory changes (Fig. 4H). In addition, CA33 treatment did not alter the circulating levels of FABP3 protein (Fig. 4H), glucagon, or adiponectin (Fig. 4, I and J).

Fig. 4. CA33 reduces adiposity and liver fat accumulation in mice with diet-induced obesity.

(A) Representative images of hematoxylin and eosin (H&E)–stained liver after 5 weeks of treatment with vehicle or CA33. Scale bar, 50 μm. (B) Liver TG (triglyceride) content after antibody treatment. (C) Expression of lipogenic genes encoding stearoyl-CoA desaturase (Scd1), fatty acid synthase (Fasn), and acetyl-CoA carboxylase (Acc1) in liver samples after vehicle or CA33 treatment. (D to J) Plasma nonesterified fatty acid (NEFA) (D), plasma glycerol (E), plasma total cholesterol (F), plasma triglycerides (G), plasma FABP3 and FABP5 (mal1) (H), plasma glucagon (I), and plasma adiponectin levels in mice treated with CA33 or vehicle (J). For (B) to (J), n = 6 mice per group. Statistical analysis was performed by Student’s t test. *P < 0.05, **P < 0.01. n.s., not significant.

Effects of CA33 treatment on adipose tissue and body composition

We next examined the impact of CA33 treatment on adipose tissue and body composition. Dual-energy x-ray absorptiometry (DEXA) scans showed that CA33 treatment significantly reduced fat mass (Fig. 5A) and slightly reduced lean tissue mass, which likely reflects the reduced lipid accumulation in solid organs, especially the liver. Indeed, CA33 treatment significantly decreased liver weight relative to vehicle-treated animals, and this reduction remained significant when expressed as percentage of body weight (Fig. 5B). The decrease in body weight in CA33-treated obese mice is reminiscent of the phenotype of mice with aP2 and combined aP2/FABP5 deficiency on a HFD (9, 12, 44), and raised the possibility that the antibody therapy directly altered metabolic parameters. In metabolic cage analysis, neither physical activity nor food intake differed between vehicle and antibody-treated mice (Fig. 5, C and D). In general, adipose tissue depots appeared similar between antibody-treated and control groups; although there was a significant decrease in the size of the perigonadal fat pad (PG-WAT; Fig. 5E), analysis of H&E-stained sections of PG-WAT revealed a similar degree of inflammatory cell infiltration in CA33 and vehicle-treated mice (Fig. 5F). Indeed, we quantitated, by fluorescence-activated cell sorting (FACS) analysis, the infiltration of F4/80+ and CD11c+ myeloid cells in the stromal vascular fraction of adipose tissue preparations and found similar numbers of these immune cells in each group (Fig. 5, G and H). We found no change in the expression of tumor necrosis factor, interleukin-1β (IL-1β), chemokine (C-C motif) ligand 2 (CCL2), chemokine (C-X-C motif) ligand 1 (CXCL1), F4/80, or CD68 in PG-WAT isolated from CA33-treated mice and a modest increase in IL-6 levels (Fig. 5I). This is perhaps not surprising, because the antibody treatment is not expected to influence intracellular aP2 levels. Indeed, adipose tissue aP2 protein levels were not decreased but instead slightly increased upon antibody treatment, and we did not observe compensatory regulation of mal1 (Fig. 5J), suggesting that the neutralization of circulating aP2 in adult mice does not result in molecular compensation by other FABP isoforms and in this way may differ from genetic aP2 deficiency.

Fig. 5. Effects of CA33 treatment on adipose tissue in dietary obesity.

(A) Body fat and lean mass as determined by DEXA after 5 weeks of vehicle or CA33 treatment. (n = 10 per group). (B) Liver weight of obese mice after CA33 treatment. (C and D) Physical activity (C) and total food intake (D) in vehicle or CA33-treated mice on a HFD (n = 8 per group). (E) Weight of perigonadal white adipose tissue (PG-WAT) of mice treated with vehicle or CA33 for 3 weeks. (F) Representative images of H&E-stained perigonadal adipose tissue (PG-WAT) after treatment with vehicle or CA33. Scale bar, 200 μm. (G and H) Effects of CA33 on F4/80+ (G) and CD11c+ (H) cell numbers in adipose tissue determined by FACS analysis. (I) Expression of cytokines, chemokines, and inflammatory cell markers in PG-WAT after vehicle or CA33 treatment. (J) Adipose tissue aP2 and mal1 protein levels measured by Western blot analysis after 5 weeks of vehicle or antibody treatment. Samples from mice with genetic deficiency of aP2 and/or mal1 were included to ensure antibody specificity. β-Tubulin is shown as a loading control, and quantification is shown in the graph on the right. Statistical analysis was performed by Student’s t test. *P < 0.05, **P < 0.01.

Decreased liver glucose production and increased peripheral insulin sensitivity

HGP and peripheral glucose utilization are both critical in the maintenance of normoglycemia and adaptation to feeding and fasting responses (45, 46). Recent studies demonstrated that HGP and liver gluconeogenic activity were regulated by aP2 (15) and suggested that this response, alone or in combination with peripheral effects, may be critical in mediating the antidiabetic effect of aP2 blockade. To determine whether this response underlies the therapeutic properties of CA33, we collected and characterized livers from HFD-fed mice after 5 weeks of CA33 or vehicle treatment and observed only in samples from the CA33-treated obese mice a marked reduction in the expression of gluconeogenic genes that encode phosphoenolpyruvate carboxykinase 1 (Pck1) and glucose 6-phosphatase (G6Pase) (Fig. 6A). In addition, the enzymatic activities of cytoplasmic Pck1 and microsomal G6Pase were significantly reduced in samples from CA33-treated mice (Fig. 6, B and C). These findings were consistent with earlier studies of aP2 function on liver (15).

Fig. 6. CA33 treatment increases hepatic and peripheral insulin sensitivity.

(A) Expression of gluconeogenic genes encoding Pck1 and G6Pase. Liver samples were collected after 6 hours of daytime food withdrawal from obese mice treated with vehicle or CA33 (n = 10 for each group) for 4 weeks. (B) Enzymatic activity of Pck1 in liver samples. (C) Enzymatic activity of G6Pase in the liver microsomal fraction. (D) Blood glucose during hyperinsulinemic-euglycemic clamp experiments. (E to I) Clamp studies performed in obese mice after 5 weeks of treatment on a HFD, treated with either vehicle or CA33 (n = 7 for each group). (E) Glucose infusion rate (GIR). (F) Clamp hepatic glucose production (C-HGP). (G) Rate of whole-body glucose disappearance (RD). (H) Glucose uptake in triceps surae muscle. (I) Whole-body glycolysis. For (B) to (I), n = 7 mice per group. Statistical analysis was performed by Student’s t test. *P < 0.05, **P < 0.01.

Next, we examined whole-body glucose fluxes with the use of hyperinsulinemic-euglycemic clamp studies (Fig. 6D). For these experiments, mice were kept on HFD for 20 weeks before antibody treatment, and clamp studies were performed after 5 weeks of antibody intervention. During the clamp study, CA33-treated obese mice required significantly higher glucose infusion rates to maintain euglycemia (Fig. 6E) and showed decreased clamp HGP (Fig. 6F) and a nonsignificant trend toward a decrease in basal HGP, relative to vehicle-treated mice (fig. S3A; P = 0.07). CA33 treatment also significantly increased whole-body clamp glucose disposal rates (RD) (Fig. 6G), although plasma insulin levels were lower compared to controls despite equivalent insulin infusion (5 mU kg−1 min−1) (fig. S3B). Together, these data suggest that CA33 increases whole-body systemic insulin sensitivity and increased insulin-stimulated glucose utilization. To further assess glucose utilization, we collected peripheral tissues after the clamp experiments and measured 2-[14C]deoxyglucose uptake. Insulin-stimulated glucose uptake tended to be higher in muscle tissue (Fig. 6H; P = 0.05) and PG-WAT (fig. S3C; P = 0.1) isolated from CA33-treated obese mice relative to controls, although these differences were of borderline statistical significance. Furthermore, whole-body glycolysis, measured as the rate of increase in plasma [3H2O] as a by-product of glycolysis, was also increased by CA33 treatment (Fig. 6I). These data support the conclusion that the glucose-lowering effect of CA33 occurs predominantly through decreasing glucose production in liver and to a lesser degree by increasing glucose utilization in peripheral tissues.


Obesity, which is now recognized by the American Medical Association as a stand-alone disease, is also involved in the pathogenesis of many immunometabolic disorders, most notably type 2 diabetes, fatty liver disease, atherosclerosis, and dyslipidemia (1). We recently demonstrated that aP2/FABP4 is a bona fide adipokine that contributes to elevated HGP in obesity (15). Because dysregulated HGP plays a key role in diabetes pathogenesis, this finding led us to hypothesize that aP2, as a link between adiposity and dysmetabolism, represents a promising therapeutic target. Numerous association studies suggest that the effects of secreted aP2 described in mice are highly conserved in humans. In both species, the level of circulating aP2 increases in response to fasting-related signals and obesity, and increased serum aP2 levels are strongly correlated with metabolic disease risk in humans (2132). Independent genetic analyses have shown that a rare FABP4 mutation that results in reduced aP2 expression confers protection against diabetes and cardiovascular disease (33, 34). Together, these findings offer strong evidence that blocking aP2/FABP4 might have therapeutic efficacy against metabolic disease in humans.

Several strategies to block aP2 function have been shown to be efficacious in preclinical models. For example, small-molecule inhibitors of aP2 have been developed, and some, but not all, showed significant benefit against diabetes, hepatosteatosis, and atherosclerosis (12, 4749). However, the high abundance of the target protein and lack of catalytic activity have limited the progression of these more traditional efforts. A targeted and highly specific interfering RNA delivery into adipose tissue against aP2/FABP4 by cell type–specific carrier molecules has been potently effective in both reducing aP2 and improving systemic metabolism (44). Although this approach has been proven to be very effective and extremely promising, there is as yet no experience in translating such platforms into human use. As a result, therapeutically targeting aP2 has been challenging, and the promising success obtained in preclinical models (12, 44, 4749) has been slow to progress toward clinical translation in humans.

The discovery of biological functions of secreted aP2 introduces a paradigm shift both in our understanding of the systemic metabolic effects of this molecule and also in consideration of strategies to exploit it for therapeutic purposes. This potential is supported by the finding that a polyclonal antibody that reduces circulating aP2 levels improved glucose homeostasis in obese mice (15). Here, we provide critical proof of principle in mice that secreted aP2 can be targeted effectively by a mAb to ameliorate diabetes. We produced and evaluated a number of anti-aP2 mAbs, identifying one candidate, CA33, as a potential preclinical molecule that lowered fasting blood glucose, improved glucose metabolism, increased systemic insulin sensitivity, and reduced liver steatosis. These metabolic outcomes were reproducible in two independent models of obesity (genetic and dietary), were absent in aP2-deficient mice, and were linked to the regulation of hepatic glucose output and peripheral glucose use, as shown in hyperinsulinemic-euglycemic clamp studies.

The use of this anti-aP2 mAb in mouse models did not decrease aP2 protein levels in adipose tissue and did not result in compensatory changes in mal1 production, which can undermine the metabolic effects of blocking aP2 in early development (8, 9, 43). As a result, antibody-mediated targeting of aP2 generated phenotypes highly reminiscent of the data generated in mice deficient in both aP2 and mal1/FABP5 (9). This feature might be an advantage of this platform for use in the treatment of metabolic diseases with robust effects.

The results obtained in this study also raise the possibility that a substantial portion of the metabolic biology of aP2, at least in the context of obesity and diabetes, can be attributed to the circulating form of the protein. Increased cardiovascular risk and weight gain are two common liabilities that interfere with the development of new antidiabetic medicines. In preclinical experimental models, perhaps the strongest protective impact of genetic aP2 deficiency has been observed in experimental models of atherosclerosis (11, 13, 14). Although a role for secreted aP2 has not been directly demonstrated in human dyslipidemia and atherosclerosis, multiple human studies, both association and genetic, strongly support a role for aP2 in human cardiovascular diseases (23, 25, 28, 3034). Indeed, an independent group showed that treatment of obese mice with an anti-aP2 mAb resulted in significant improvements in dyslipidemia and moderate effects on insulin sensitivity (50). Our study also demonstrates potential benefits of targeting aP2 to prevent weight gain, at least in the dietary model of obesity, similar to the earlier results in genetic models. On the basis of the data presented here, we hypothesize that therapeutic targeting of aP2 may have significant additional benefits in limiting obesity and in preventing cardiovascular complications, especially in humans with metabolic syndrome. Hence, an important future research avenue is to explore the impact of secreted aP2 on dyslipidemia and cardiovascular disease in preclinical models.

The molecular mechanisms by which secreted aP2 generates its biological functions are not well understood. X-ray crystallography analysis revealed that neither CA33-bound aP2 nor HA3-bound aP2 showed significant deviations from other aP2 structures in the Protein Data Bank (PDB) database, suggesting that CA33 does not cause any gross distortions in the tertiary structure of the aP2 backbone that might account for its activity. In addition, neither structure showed any electron density for palmitic acid within the aP2 fatty acid binding pocket, which is surprising given that the aP2 was refolded in the presence of palmitic acid. This could result from the displacement, by Fab binding, of palmitic acid from the lipid-binding cavity or from preferential crystallization of “empty” aP2, but the fact that both structures lack lipid suggests that lipid binding is not a sole determinant of therapeutic efficacy. Furthermore, in biochemical analyses, we did not observe antibody-mediated displacement of a fluorescently labeled fatty acid (fig. S2), supporting the conclusion that antibody efficacy is not dependent on its targeting of a form of aP2 that carries specific cargo. Last, the location of the epitope recognized by CA33 does not suggest potential interference with the FABP4 lipid-binding cavity. However, potential alterations in a specific lipid-ligand interaction cannot be ruled out conclusively at this stage.

Future studies may focus on the identification of protein complexes or binding partners of circulating aP2 in generating its biological functions, and the availability of unique antibodies such as CA33 along with control antibodies that lack metabolic efficacy may facilitate such mechanistic avenues of research. Further unraveling of the structural determinants of CA33 action and its successful humanization will be central to translating this approach to treating human disease.


Study design

The objective of the study was to develop mAbs that target circulating aP2, characterize their structural and physical properties, and determine their potential as antidiabetic therapeutics. In vivo results shown for wild-type mice are representative of at least two independent cohorts in the Hotamisligil lab, and the key antidiabetic effects have been independently replicated in a separate facility. In vivo experiments (GTT and ITT blood glucose measurements, clamps) were not performed blinded. The elimination criteria for the outliers were based on the visible health of the individual mice or on findings more than 2 SDs from the mean. Quantitative real-time polymerase chain reaction (qPCR) samples were analyzed in technical duplicates with 6 to 10 biological replicates.


Animal care and experimental procedures were performed with approval from animal care committees of Harvard University. Male mice [leptin-deficient (ob/ob) and diet-induced obese (DIO) mice with C57BL/6J background] were purchased from the Jackson Laboratory and kept on a 12-hour light/dark cycle. DIO mice with C57BL/6J background were maintained on HFD (60% kcal fat; Research Diets Inc., D12492i) starting at 5 weeks of age for 12 to 15 weeks before starting treatment except in clamp studies, for which they were on HFD for 20 weeks. Leptin-deficient (ob/ob) mice were maintained on regular chow diet (RD; PicoLab 5058 LabDiet). Animals used were 17 to 21 weeks of age for dietary models and 9 to 12 weeks of age for the ob/ob model. In all experiments, at least seven mice in each group were used, unless otherwise stated in the text. The mice were treated with 150 μl of phosphate-buffered saline (PBS) (vehicle) or 1.5 mg per mouse (~33 mg/kg) of anti-aP2 mAb in 150 μl of PBS by twice weekly subcutaneous injections for 3 to 5 weeks (Fig. 1B). Before and after the treatment, blood samples were collected from the tail after 6 hours of daytime food withdrawal. Body weights were measured weekly in the fed state. Blood glucose levels were measured weekly after 6 hours of food withdrawal or after 16-hour overnight fast. After 2 weeks of treatment, GTTs were performed in mice after 16-hour overnight fast by intraperitoneal glucose injections (0.75 g/kg for DIO, 0.5 g/kg for ob/ob mice). After 3 weeks of treatment, ITTs were performed in DIO mice after a 6-hour fast by intraperitoneal insulin injections (0.75 IU/kg). After 5 weeks of treatment, hyperinsulinemic-euglycemic clamp experiments were performed in DIO mice as previously described (9, 12, 15). Metabolic cage (Oxymax, Columbus Instruments) and total body fat measurement by DEXA (PIXImus) were performed as previously described (12).

Production and administration of anti-aP2 antibodies

CA186_02033, CA186_02013, CA186_02015, and CA186_02023 (hereinafter referred to as CA33, CA13, CA15, and CA23, respectively) were produced and purified by UCB (Union Chimique Belge). New Zealand White rabbits were immunized with a mixture containing recombinant human and mouse aP2 (generated in-house in Escherichia coli; accession numbers CAG33184.1 and CAJ18597.1, respectively). Splenocytes, peripheral blood mononuclear cells, and bone marrow were harvested from immunized rabbits and subsequently stored at −80°C. B cell cultures from immunized animals were prepared using a method similar to that described by Zubler et al. (51). After 7 days of incubation, antigen-specific antibody-containing wells were identified using a homogeneous fluorescence-linked immunosorbent assay with biotinylated mouse aP2 immobilized on SuperAvidin beads (Bangs Laboratories) and a goat anti-rabbit IgG Fcγ-specific Cy5 conjugate (Jackson ImmunoResearch). To identify, isolate, and recover the antigen-specific B cell from the wells of interest, we used the fluorescent foci method (52). This method involved harvesting B cells from a positive well and mixing with paramagnetic streptavidin beads (New England Biolabs) coated with biotinylated mouse aP2 and goat anti-rabbit Fc fragment–specific fluorescein isothiocyanate conjugate (Jackson ImmunoResearch). After static incubation at 37°C for 1 hour, antigen-specific B cells could be identified due to the presence of a fluorescent halo surrounding that B cell. Individual antigen-specific antibody-secreting B cells were viewed using an Olympus IX70 microscope and were picked with an Eppendorf micromanipulator and deposited into a PCR tube. Variable region genes from these single B cells were recovered by reverse transcription PCR (RT-PCR), using primers that were specific to heavy- and light-chain variable regions. Two rounds of PCR were performed, with the nested 2° PCR incorporating restriction sites at the 3′ and 5′ ends allowing cloning of the variable region into a variety of expression vectors: mouse γ1 IgG, mouse Fab, rabbit γ1 IgG (VH), or mouse κ and rabbit κ (VL). Heavy- and light-chain constructs were transfected into human embryonic kidney 293 cells using Fectin 293 (Invitrogen) and recombinant antibody expressed in six-well plates. After 5 days of expression, supernatants were harvested and the antibody was subjected to further screening by biomolecular interaction analysis using the Biacore system to determine affinity and epitope bin.

Mouse anti-aP2 mAb HA3 was produced by the Dana-Farber Cancer Institute Antibody Core Facility. Female C57BL/6 aP2−/− mice, 4 to 6 weeks old, were immunized by injection of full-length human FABP4-Gst recombinant protein suspended in Dulbecco’s PBS (Gibco) and emulsified with an equal volume of complete Freund’s adjuvant (Sigma Chemical Co.). Spleens were harvested from immunized mice, and cell suspensions were prepared and washed with PBS. The spleen cells were counted and mixed with SP 2/0 myeloma cells [American Type Culture Collection (ATCC) no. CRL8006] that are incapable of secreting either heavy- or light-chain immunoglobulins (53) at a spleen/myeloma ratio of 2:1. Cells were fused with polyethylene glycol 1450 (ATCC) in 12 96-well tissue culture plates in HAT (hypoxanthine-aminopterin-thymidine) selection medium according to standard procedures (54). Between 10 and 21 days after fusion, hybridoma colonies became visible and culture supernatants were harvested and then screened by Western blot on strep-His-human-FABP4. A secondary screen of 17 selected positive wells was done using high-protein binding 96-well EIA (enzyme immunoassay) plates (Costar/Corning Inc.) coated with 50 μl per well of a 2 μg/ml solution (0.1 μg per well) of strep-His-human-FABP4 or an irrelevant Gst-protein and incubated overnight at 4°C. Positive hybridomas were subcloned by limiting dilution and screened by enzyme-linked immunosorbent assay (ELISA). Supernatant fusions were isotyped with IsoStrip kit (Roche Diagnostics Corp.).

Large-scale transient transfections were carried out using UCB’s proprietary CHOSXE cell line and electroporation expression platform. Cells were maintained in logarithmic growth phase in CDCHO media (Life Technologies) supplemented with 2 mM GlutaMAX at 140 rpm in a shaker incubator (Kuhner AG) supplemented with 8% CO2 at 37°C. Before transfection, the cell numbers and viability were determined using Cedex cell counter (Innovatis AG), and 2 × 108 cells/ml were centrifuged at 1400 rpm for 10 min. The pelleted cells were washed in Hyclone MaxCyte buffer (Thermo Scientific), spun for a further 10 min, and then resuspended at 2 × 108 cells/ml in fresh buffer. Plasmid DNA, purified using Qiagen Plasmid Plus Giga Kit, was then added at 400 μg/ml. After electroporation using a MaxCyte STX flow electroporation instrument, the cells were transferred in ProCHO medium (Lonza) containing 2 mM GlutaMAX and antibiotic/antimitotic solution and cultured in a wave bag (Cellbag, GE Healthcare), placed on a bioreactor platform set at 37°C and 5% CO2 with wave motion induced by rocking at 25 rpm.

Twenty-four hours after transfection, a bolus feed was added and the temperature was reduced to 32°C and maintained for the duration of the culture period (12 to 14 days). At day 4, 3 mM sodium butyrate (n-butyric acid sodium salt, Sigma B5887) was added to the culture. At day 14, the cultures were centrifuged for 30 min at 4000 rpm, and the retained supernatants were filtered through 0.22-μm Sartobran P (Millipore) followed by 0.22-μm gamma gold filters. CHOSXE harvest expressing mouse mAb was conditioned by the addition of NaCl (to 4 M). The sample was loaded onto a protein A MabSelect SuRe packed column (GE Healthcare) equilibrated with 0.1 M glycine + 4 M NaCl (pH 8.5) at 15 ml/min. The sample was washed with 0.1 M glycine + 4 M NaCl (pH 8.5), and an additional wash step was performed with 0.15 M Na2HPO4 (pH 9). The ultraviolet (UV) absorbance peak at 280 nm was collected during elution from the column using 0.1 M sodium citrate (pH 3.4) elution buffer and then neutralized to pH 7.4 by the addition of 2 M tris-HCl (pH 8.5). The mAb pool from protein A was then concentrated to suitable volume using a minisette tangential flow filtration device before being purified further on a HiLoad XK 50/60 Superdex 200 prep grade gel filtration column (GE Healthcare). Fractions collected were then analyzed by analytical gel filtration technique for monomer peak, and clean monomer fractions were pooled as final product. The final product sample was then characterized by reducing and nonreducing SDS–polyacrylamide gel electrophoresis and analytical gel filtration for final purity check. The sample was also tested and found to be negative for endotoxin using a LAL assay method for endotoxin measurements. The final buffer for all mAbs tested was PBS. For in vivo analysis, purified antibodies were diluted in saline to 10 mg/ml and injected at a dose of 1.5 mg per mouse (33 mg/kg) into ob/ob and wild-type mice on HFD.

Measurement of antibody affinity

The affinity of anti-aP2 binding to aP2 was determined by biomolecular interaction analysis, using a Biacore T200 system (GE Healthcare). AffiniPure F(ab′)2 fragment goat anti-mouse IgG, Fc fragment–specific (Jackson ImmunoResearch Lab Inc.) in 10 mM NaAc, pH 5 buffer was immobilized on a CM5 Sensor Chip via amine coupling chemistry to a capture level between 4500 and 6000 response units (RU) using HBS-EP+ (GE Healthcare) as the running buffer. Anti-aP2 IgG was diluted to between 1 and 10 μg/ml in running buffer. A 60-s injection of anti-aP2 IgG at 10 μl/min was used for capture by the immobilized anti-mouse IgG, Fc, then aP2 was titrated from 25 to 3.13 nM over the captured anti-aP2 for 180-s at 30 μl/min followed by 300-s dissociation. The surface was regenerated by 2 × 60-s 40 mM HCl and 1 × 30-s 5 mM NaOH at 10 μl/min. The data were analyzed using Biacore T200 evaluation software (version 1.0) using the 1:1 binding model with local Rmax. For CA33, 60-s injection of the antibody at 10 μl/min was used for capture by the immobilized anti-mouse IgG, Fc, then aP2 was titrated from 40 to 0.625 μM over the captured anti-aP2 for 180 s at 30 μl/min followed by 300-s dissociation. The surface was regenerated by 1 × 60-s 40 mM HCl, 1 × 30-s 5 mM NaOH, and 1 × 30-s 40 mM HCl at 10 μl/min. Steady-state fitting was used to determine affinity values.

Antibody cross-blocking

The cross-blocking assay was performed by injecting mouse aP2 in the presence or absence of mouse anti-aP2 IgG over captured rabbit anti-aP2 IgG. Biomolecular interaction analysis was performed using a Biacore T200 (GE Healthcare Bio-Sciences AB). Anti-aP2 rabbit IgG transient supernatants were captured on the immobilized anti-rabbit Fc surfaces (one supernatant per flow cell) using a flow rate of 10 μl/min and a 60-s injection to give response levels above 200 RU. Then, mouse aP2 at 100 nM, or mouse aP2 at 100 nM plus mouse anti-aP2 IgG at 500 nM was passed over for 120-s followed by 120-s dissociation. The surfaces were regenerated with 2 × 60-s 40 mM HCl and 1 × 30-s 5 mM NaOH.

FABP cross-reactivity

The recombinant human proteins aP2 (generated in-house in E. coli), hFABP3 (Sino Biological Inc.), and hFABP5 (Sino Biological Inc.) were biotinylated using a fivefold molar excess of EZ-Link Sulfo-NHS-LC-Biotin (Thermo Fisher Scientific) and purified from unbound biotin using a Zeba desalting column (Thermo Fisher Scientific). All binding studies were performed at 25°C using an Octet RED384 system (Pall FortéBio Corp.). After a 120-s baseline step in PBS containing 0.05% Tween 20 (pH 7.4) (PBS-T), Dip and Read streptavidin (SA) biosensors (Pall FortéBio Corp.) were loaded with biotinylated recombinant haP2, hFABP3, or hFABP5 at 60 nM for 90 s. After a 60-s stabilization step in PBS-T, each FABP-loaded biosensor was transferred to a sample of mAb at a concentration of 800 nM, and association was measured for 5 min. Biosensors were then transferred back to PBS-T for 5 min to measure dissociation. Nonspecific binding of antibodies was monitored using unloaded biosensor tips. Maximal association binding, that is, once signal had plateaued, minus background binding, was plotted for each antibody/FABP combination.

aP2 expression and purification

Mouse aP2 cDNA (complementary DNA) optimized for expression in E. coli was purchased from DNA2.0 and subcloned directly into a modified pET28a vector (Novagen) containing an in-frame N-terminal 10 His-tag followed by a tobacco etch virus (TEV) protease site. Protein was expressed in the E. coli strain BL21DE3 and purified as follows. Cells were lysed with a cooled cell disruptor (Constant Systems Ltd.) in 50 ml of lysis buffer [PBS (pH7.4) containing 20 mM imidazole] per liter of E. coli culture supplemented with a Complete EDTA-free protease inhibitor cocktail tablet (Roche). Lysate was then clarified by high-speed centrifugation (60,000g, 30 min, 4°C). Ni-NTA beads (4 ml) (Qiagen) were added per 100 ml of cleared lysate and tumbled for 1 hour at 4°C. Beads were packed in a Tricorn column (GE Healthcare) attached to an ÄKTA fast protein liquid chromatography (GE Life Sciences), and protein was eluted in a buffer containing 250 mM imidazole. Fractions containing protein of interest as judged by 4 to 12% bis-tris NuPAGE (Life Technologies Ltd.) gel electrophoresis were dialyzed to remove imidazole and treated with TEV protease at a ratio of 1 mg per 100 mg of protein. After overnight incubation at 4°C, the sample was repassed over the Ni-NTA beads in the Tricorn column. Untagged (that is, TEV-cleaved) aP2 protein did not bind to the beads and was collected in the column flow through. The protein was concentrated and loaded onto an S75 26/60 gel filtration column (GE healthcare) pre-equilibrated in PBS, 1 mM dithiothreitol. Peak fractions were pooled and concentrated to 5 mg/ml. Six milliliters of this protein was then extracted and precipitated with acetonitrile at a ratio of 2:1 to remove any lipid. After centrifugation at 16,000g for 15 min, the acetonitrile + buffer was removed for analysis of original lipid content. The pellet of denatured protein was then resuspended in 6 ml of 6 M GuHCl PBS + 2 μmol of palmitic acid (5:1 ratio of palmitic acid to aP2) and then dialyzed twice against 5 liters of PBS for 20 hours at 4°C to allow refolding. After centrifugation to remove precipitate (16,000g, 15 min), protein was gel-filtered using an S75 26/20 column in PBS to remove aggregate. Peak fractions were pooled and concentrated to 13 mg/ml.


Purified aP2 was complexed with CA33 and HA3 Fab (generated at UCB by conventional methods) as follows. Complex was made by mixing 0.5 ml of aP2 at 13 mg/ml with either 0.8 ml of CA33 Fab at 21.8 mg/ml or 1.26 ml of HA3 Fab at 13.6 mg/ml (aP2/Fab molar ratio of 1.2:1). Proteins were incubated at room temperature for 30 min and then run on an S75 16/60 gel filtration column (GE Healthcare) equilibrated with 50 mM tris (pH 7.2), 150 mM NaCl + 5% glycerol as the running buffer. Peak fractions were pooled and concentrated to 10 mg/ml for crystallization.

Sitting-drop crystallization trials were set up using commercially available screening kits (Qiagen). Diffraction-quality crystals were obtained directly from primary crystallization screening without any need to optimize crystallization conditions. For the aP2/CA33 complex, the well solution contained 0.1 M Hepes (pH 7.5), 0.2 M (NH4)2SO4, 16% PEG 4K, and 10% isopropanol. For the aP2/HA3 complex, the well solution contained 0.1 M MES (pH 5.5), 0.15 M (NH4)2SO4, and 24% PEG 4K. Data were collected at the Diamond synchrotron on i02 (λ = 0.97949), giving a 2.9 Å data set for aP2/CA33 and a 2.3 Å data set for aP2/HA3. Structures were determined by molecular replacement using Phaser (55) (CCP4) with aP2 and Fab domain starting models. Two complexes were found to be in the asymmetric unit for aP2/CA33 and one for aP2/HA3. Cycles of refinement and model building were performed using CNS (56) and coot (57) (CCP4) until all the refinement statistics converged for both models. Epitope information was derived by considering atoms within 4 Å distance at the aP2/Fab contact surface. The data collection and refinement statistics are shown in Table 1. The aP2/CA33 and aP2/H3 structures have been deposited in the PDB with the PDB IDs 5C0N and 5D8J, respectively.

Table 1. Structure determination and refinement statistics.

Values in parenthesis refer to the high-resolution shell. Rsym = Σ|(I − <I>)|/Σ(I), where I is the observed integrated intensity, <I> is the average integrated intensity obtained from multiple measurements, and the summation is over all observed reflections. Rwork = Σ||Fobs| − k|Fcalc||/Σ|Fobs|, where Fobs and Fcalc are the observed and calculated structure factors, respectively. Rfree is calculated as Rwork using 5% of the reflection data chosen randomly and omitted from the refinement calculations. Epitope information was derived by considering atoms within 4 Å distance at the aP2/Fab contact surface. RMS, root mean square.

View this table:

Ligand binding to aP2

Ligand binding to aP2 was assessed using a fluorescence-based assay system utilizing the fluorescent fatty acid cis-parinaric acid (Life Technologies) as previously described (58). Cis-parinaric acid was dissolved in ethanol (final ethanol 0.2%, v/v). Cis-parinaric acid (1 μM) in 50 mM phosphate buffer at indicated pH was incubated with recombinant aP2 protein (1 μM) for 1 min, and the resultant fluorescent signal was measured (excitation, 375 nm; emission, 475 nm).

Hyperinsulinemic-euglycemic clamp studies and hepatic biochemical assays

Hyperinsulinemic-euglycemic clamps were performed by a modification of a reported procedure (15). Specifically, mice were clamped after 5 hours of fasting and infused with insulin (5 mU kg−1 min−1). Blood samples were collected at 10-min intervals for the immediate measurement of plasma glucose concentration, and 25% glucose was infused at variable rates to maintain plasma glucose at target euglycemic concentrations. Baseline whole-body glucose disposal was estimated with a continuous infusion of [3-3H]glucose (0.05 μCi/min). This was followed by determination of insulin-stimulated whole-body glucose disposal whereby [3-3H]glucose was infused at 0.1 μCi/min. To determine insulin-stimulated glucose uptake in muscle and WAT, 2-[14C]deoxyglucose (2-[14C]DG; PerkinElmer) was administered as a bolus (10 μCi) immediately before the time 0 blood collection, and tissues were collected after 40 min. Glucose uptake was calculated from the plasma decay profile of 2-[14C]DG, which was fitted with an exponential curve, and tissue 2-[14C]DG-6-phosphate content.

Total lipids in liver were extracted according to the Bligh-Dyer protocol (59), and a colorimetric method was used for triglyceride content measurement by a commercial kit according to the manufacturer’s instructions (Sigma-Aldrich). Gluconeogenic enzyme Pck1 activity was measured by a modification of a reported method (60). G6Pase activity was measured by a modification of Sigma protocol (EC Briefly, the livers were homogenized in lysis buffer containing 250 mM sucrose, tris-HCl, and EDTA. Lysates were centrifuged at full speed for 15 min, and the supernatant (predominantly cytoplasm) was isolated. Then, microsomal fractions were isolated by ultracentrifugation of cytoplasmic samples. Microsomal protein concentrations were measured by commercial BCA kit (Thermo Scientific Pierce). Glucose-6-phosphate (200 mM) (Sigma-Aldrich) was preincubated in bis-tris. Microsomal protein (150 μg) or serial dilution of recombinant G6Pase was added and incubated in that solution for 20 min at 37°C. Then, 20% trichloroacetic acid was added, mixed, and incubated for 5 min at room temperature. Samples were centrifuged at full speed at 4°C for 10 min, and the supernatant was transferred to a separate UV plate. Color reagent was added, and absorbance at 660 nm was measured and normalized to a standard curve prepared with a serial dilution of recombinant G6Pase enzyme.

Plasma aP2, mal1, FABP3, adiponectin, glucagon, and insulin ELISAs

Blood was collected from mice by tail bleeding after 6-hour daytime or 16-hour overnight food withdrawal. Terminal blood samples were collected by cardiac puncture or collected from tail vein. The samples were spun in a microcentrifuge at 3000 rpm for 15 min at 4°C. Plasma aP2 (BioVendor R&D), mal1 (CircuLex Mouse mal1 ELISA, CycLex Co. Ltd.), FABP3 (Hycult Biotech), glucagon, adiponectin (Quantikine ELISA, R&D Systems), and insulin (insulin-mouse ultrasensitive ELISA, Alpco Diagnostics) measurements were performed according to the manufacturer’s instructions.

QPCR analysis

Tissues were collected after 6-hour daytime food withdrawal, immediately frozen, and stored at −80°C. RNA isolation was performed using TRIzol (Invitrogen) according to the manufacturer’s protocol. For first-strand cDNA synthesis, 0.5 to 1 ng of RNA and 5× iScript RT Supermix were used (Bio-Rad Laboratories). qPCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems, Life Technologies), and samples were analyzed using a ViiA7 QPCR machine (Applied Biosystems, Life Technologies). Primers used for qPCR are shown in Table 2.

Table 2. Primers for qPCR.
View this table:

Statistical analysis

Results are presented as means ± SEM. Statistical significance was determined by repeated-measures ANOVA or Student’s t test as indicated in the figure legends. *P < 0.05, **P < 0.01.


Fig. S1. Screening cohort.

Fig. S2. Effect of antibody on lipid binding to aP2.

Fig. S3. Hyperinsulinemic-euglycemic clamp of HFD mice treated with CA33.


  1. Acknowledgments: We are grateful to members of the Hotamisligil laboratory for their contributions to this project, especially A. Bartelt for the metabolic cage analysis, K. Claiborn for critical reading and editing of this manuscript, and M. McGrath for coordinating the academic-industry alliance between the biopharmaceutical company UCB and the Hotamisligil laboratory, Harvard University. Funding: The technology described in this manuscript is licensed to UCB, and this work is supported by a sponsored research grant to G.S.H. Author contributions: M.F.B., K.E.I., A.W., A.L., G.T., E.S.C., M.S., A.T., and K.E. designed and performed in vivo and in vitro experiments, analyzed the data, prepared the figures, and revised the manuscript. C.D. performed the crystallization experiments, and G.B. and C.D. analyzed the crystal structures, prepared the figures, and revised the manuscript. D.L., L.H., G.O., H.H., S.W., R.G., H.N., C.D., and A.M. produced the antibodies and performed structural and binding analyses, and revised the manuscript. G.S.H. designed experiments, interpreted the results, and wrote and revised the manuscript. Competing interests: This work was supported by a sponsored research grant to G.S.H. from UCB. Data materials and availability: Monoclonal antibodies described in this work were provided through a materials transfer agreement (MTA) between Harvard University and UCB. All reasonable requests for said materials will be fulfilled via an MTA with UCB (contact
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article