2FA-Platform Generates Dual Fatty Acid-Conjugated GLP-1 Receptor Agonist TE-8105 with Enhanced Diabetes, Obesity, and NASH Efficacy Compared to Semaglutide

Abstract

Conjugating two fatty acids (2FAs) to peptide drugs can improve pharmacokinetics and therapeutic effects. However, optimizing FA spacing, chain combination, and attachment site to simultaneously enhance albumin binding and drug efficacy remains challenging. We introduce a multiarm linker technology enabling precise control of 2FA spacing, composition, and attachment. By applying this technology to a modified glucagon-like peptide-1 (GLP-1) and screening various 2FA-GLP-1 conjugates differing in linkage, linker, and FA properties for improved albumin affinity, pharmacokinetics, and pharmacodynamics, TE-8105 emerged as a promising candidate. TE-8105 outperformed semaglutide, showing improved long-term glycemic control, weight loss, and liver health in diabetic mice, and dose-dependent weight loss and favorable body composition changes in obese mice. A distinct advantage of TE-8105 over semaglutide is its low-dose reduction of liver steatosis and improvement of liver health in nonalcoholic steatohepatitis mice. The multiarm linker technology provides a versatile platform for developing improved 2FA-peptide therapeutics.

Introduction

Glucagon-like peptide-1 receptor agonists (GLP-1RAs) have become frontline therapies for managing type 2 diabetes (T2D), obesity, and associated conditions such as fatty liver disease and cardiovascular complications.¹⁻⁴ These drugs mimic the native incretin hormone GLP-1, which interacts with its receptor to enhance insulin secretion, suppress glucagon release, delay gastric emptying, and reduce appetite in a glucose-dependent manner.⁵⁻⁷ Technological advancements have been instrumental in developing these pharmaceuticals, as the clinical utility of native GLP-1 is limited by its short intravenous half-life of 1.5–2 min,^2,4,8 Strategies to enhance GLP-1 stability include: (i) replacing vulnerable GLP-1 residues prone to enzymatic cleavage with protease-resistant unnatural residues,⁹ (ii) conjugating a protein or polyethylene glycol (PEG) to GLP-1 to protect against reticuloendothelial clearance,¹⁰ and (iii) attaching a fatty acid (FA) to GLP-1,^11,12 enabling reversible binding to human serum albumin (HSA), the most abundant blood protein with a half-life of 19 days¹³ thus extending GLP-1’s half-life.

In 1991, Chang invented a method to extend the pharmacokinetics of therapeutic proteins and peptides¹⁴ through site-specific conjugation of a lipophilic moiety to a therapeutic agent. Early studies showed that site-specific conjugation of a FA to therapeutic proteins/peptides such as human insulin,¹⁵ Bowman-Birk protease inhibitor,¹⁶ interferon-α,¹⁷ adrenocorticotropin,¹⁸ and salmon calcitonin¹⁹ resulted in longer circulating half-lives of the modified protein/peptide.^20,21 Site-specific FA modification was also applied to improve the kinetic properties of GLP-1RAs and fusion peptides of GLP-1 and other incretins, leading to the development of FA-modified GLP-1 drugs such as liraglutide,^12,22,23 semaglutide,^12,24 and tirzepatide.²⁵⁻²⁷ Although these 1FA-GLP-1 drugs are widely accepted for treating T2D, obesity, and related pathological conditions, they show dose-dependent gastrointestinal side effects such as nausea and vomiting and are not effective for all patients.^7,28−30 Notably, clinical data indicate that the efficacy of 1FA-GLP-1 drugs is strongly correlated with their in vivo stability.³¹

Herein, we present a 2FA platform based on a novel multiarm linker for generating more efficacious 2FA-conjugated peptide drugs and apply it to develop TE-8105, a 2FA-conjugated GLP-1RA that demonstrates enhanced efficacy over semaglutide in preclinical models of T2D, obesity, and nonalcoholic steatohepatitis (NASH). Our multiarm linker creates a 2FA-bundle consisting of two specifically paired FAs, which can be attached to a single residue in the therapeutic peptide. The simplest version of a 2FA-bundle comprises a tetrapeptide with two Lys residues, each linked via a glutamate bridge to a methylated or carboxylated C16–C20 FA. At the tetrapeptide’s N-terminus is an alkyne or maleimide group for conjugating to the therapeutic peptide via click chemistry.³² A PEGylated amino acid (aa) residue with 2, 4, or 6 ethylene glycol (EG) units connects the N-terminal reactive group to the first Lys, which is linked to the second Lys via another PEGylated residue. A CH₃ or COOH group at the C-terminus caps the 2FA-bundle.

Importantly, our methodology addresses the “molecular docking challenge” of matching FA characteristics to albumin’s seven FA-binding pockets of varying shape, depth, and electrostatic properties,³³ achieving effective albumin binding, without compromising the biological function of the peptide drug. The multiarm linker enables the creation of diverse 2FA-bundles that can be site-specifically conjugated to a native GLP-1 Lys or to a Cys introduced via a (GGGSG)₃ linker at the peptide’s C-terminus. This generates 2FA-GLP-1 conjugates with varying linkages, FA characteristics, and the multiarm linker’s C-terminal group. With this library of diverse 2FA-GLP-1 conjugates, we can explore which 2FA combination and configuration yields better pharmacokinetic and therapeutic properties than the commercially available 1FA-GLP-1, semaglutide, which comprises a single C18 diacid.

We compared these 2FA-GLP-1 molecules and semaglutide in terms of their HSA affinities and their effectiveness in controlling blood sugar in diabetic db/db mice. By comparing the pharmacokinetic properties of 2FA-GLP-1 molecules with better albumin binding and glucose control efficacy than semaglutide, a specific 2FA-GLP-1 molecule named TE-8105 exhibited (i) prolonged half-life, (ii) improved efficacy in preclinical models of T2D, obesity, and NASH, and (iii) potential for more convenient and effective treatment of T2D and obesity. Our approach to optimize albumin binding, drug stability, and efficacy could be broadly applicable to develop more efficacious peptide drugs across various therapeutic areas.

Results

Multiarm Linker for Generating Diverse 2FA-GLP-1 Conjugates

We developed a versatile multiarm linker to create a library of diverse 2FA bundles that can be used to generate 2FA-GLP-1 conjugates with varying characteristics, enabling the optimization of albumin binding and therapeutic efficacy. The multiarm linker comprises 2 lysine and 2 PEGylated residues with a reactive N-terminal alkyne/maleimide for peptide-drug attachment and an esterified or acidic C-terminus (Figure 1A–C). Two FAs with identical or differing chain lengths and/or terminal groups were attached to the linker’s Lys ε-amino groups via a glutamate moiety. This linker design allows for diverse 2FA-bundle configurations, varying in (i) conjugation linkage (alkyne–azide or maleimide–thiol), (ii) FA chain length (C16, C18, or C20), (iii) FA terminal group (CH₃ or COO^–), (iv) FA spacing (n = 2, 4, or 6 EG units), and (v) the linker’s C-terminal group (CH₃ or H).

Figure 1.

Schematic of 2FA-GLP-1 conjugates generated using a multiarm linker. (A) General structure of 2FA-bundles created by a multiarm linker. (B) Modified GLP-1 conjugated to a 2FA-bundle at native Lys-26 via a Glu-azide-alkyne reaction. (C) Modified GLP-1 conjugated to a 2FA-bundle at Cys-53 linked by (GGGSG)₃ via thiol-maleimide reaction. n is the number of EG units; L₁ and L₂ denote the number of CH₂ groups, whereas T₁ and T₂ denote the terminal groups in FA-1 and FA-2, respectively; C^ter represents −CH₃ or −H.

The synthesized 2FA-bundles were conjugated to GLP-1 (residues 7–37) modified by substituting Ala-8 with α-aminoisobutyric acid (Aib) and Lys-34 with Arg to improve the peptide’s resistance to enzymatic cleavage without compromising GLP-1RA potency.¹² The Lys-34 → Arg substitution also enables site-specific 2FA conjugation to Lys-26. To enable click reaction with the linker’s reactive alkyne or maleimide group, a complementary reactive group (azide or Cys) was introduced to the modified GLP-1. For alkyne–azide coupling, an azide-modified glutamate, Glu-CO-CH₂-N₃, was linked via the γ-COOH group of the Glu side chain to the Lys-26 ε-amino group (Figure 1B). Alternatively, for maleimide–thiol coupling, Cys-53 was engineered at the modified GLP-1 C-terminus via a 15-mer (GGGSG)₃ linker (Figure 1C). The GLP-1 analog was then site-specifically conjugated to a 2FA-bundle via copper-catalyzed alkyne–azide cycloaddition^34,35 or maleimide–thiol coupling (Figures S1 and S2).^32,36 The synthesized 2FA-GLP-1 molecules were purified to >95% purity, as determined by high-performance liquid chromatography (HPLC)-mass spectrometry (MS), with molecular weights ranging from 5092–6331 Da (Table S1).

2FA-GLP-1 Can Enhance Albumin Affinity Compared with 1FA-GLP-1

We measured the association rate constant (K_on), dissociation rate constant (K_off), and equilibrium association constant (K_a = K_on/K_off) of each 2FA-GLP-1 variant and semaglutide. The relative K_a values, K_a′ = K_a(2FA-GLP-1)/K_a(semaglutide), are presented in Table 1. Most 2FA-GLP-1 variants with an esterified C-terminus (C^ter = CH₃) displayed K_a′ > 1, indicating stronger albumin binding than semaglutide.

Table 1. HSA Affinity and Glucose Levels of GLP-1 2FAs Relative to Semaglutide.

2FA-GLP-1a	Linkage	Multiarm linkerb		FA chain-1c		FA chain-2c
Name	Site	n	Cter	L1	T1	L2	T2	Ka′d	AUC′e
TE-8101	Lys-26	4	–CH3	C18	–COOH	C18	–COOH	28	1.43
TE-8102	Lys-26	4	–CH3	C16	–CH3	C18	–COOH	18	0.77
TE-8103a	Lys-26	4	–CH3	C16	–CH3	C16	–CH3	8	1.43
Effect of FA Chain Length
TE-8111	Lys-26	4	–CH3	C16	–COOH	C16	–COOH	0.2	1.43
TE-8112	Lys-26	4	–CH3	C18	–COOH	C20	–COOH	77	1.25
TE-8113	Lys-26	4	–CH3	C20	–COOH	C20	–COOH	139	1.25
Effect of FA spacing
TE-8121	Lys-26	2	–CH3	C18	–COOH	C18	–COOH	30	1.25
TE-8122	Lys-26	6	–CH3	C18	–COOH	C18	–COOH	1.7	1.67
TE-8123	Lys-26	6	–CH3	C16	–CH3	C18	–COOH	19	0.63
Effect of relocating conjugation site to Cys-53
TE-8131	Cys-53	4	–CH3	C16	–CH3	C18	–COOH	15	0.91
TE-8132	Cys-53	4	–CH3	C18	–COOH	C18	–COOH	18	1.00
TE-8133	Cys-53	4	–CH3	C18	–COOH	C20	–COOH	95	1.25
TE-8134	Cys-53	4	–CH3	C20	–COOH	C20	–COOH	160	1.25
Effect of multiarm linker’s C-terminal group
TE-8105	Lys-26	4	–H	C16	–CH3	C18	–COOH	2.2	0.83

^aFor TE-8103, Ala-8 remains unchanged.

^bn is the number of EG units; C^ter is the multiarm linker’s C-terminal group.

^cL₁/L₂ and T₁/T₂ denote the chain length and terminal group of the FA chain-1/2.

^dK_a′ = K_a(2FA-GLP-1)/K_a(semaglutide).

^eAUC′ = AUC(2FA-GLP-1)/AUC(semaglutide). Rows in italics highlight molecules with hybrid −CH₃ and −COOH chains.

Notably, 2FA-GLP-1 molecules containing two C18-COOH, C16-CH₃ + C18-COOH, and two C16-CH₃ chains displayed 28-, 18-, and 8-fold greater HSA affinity compared with semaglutide, respectively. The K_a′ values of the other 2FA-GLP-1 molecules reveal how the FA chain length, spacing, and attachment site affect HSA affinity.

• 1.FA chain length: for carboxylated FAs separated by 4 EG units, increasing chain length enhanced albumin binding: the K_a′ of C16 + C16, C18 + C18, C18 + C20, and C20+C20 increased from 0.2 to 28 to 77 to 139, respectively.
• 2.FA spacing: the effect of FA spacing on albumin binding depended on the second FA chain length and charge. Increasing the FA spacing from 4 to 6 EG units reduced albumin binding for a homogeneous C18-COOH 2FA-bundle (K_a′ decreased from 28 for TE-8101 to 1.7 for TE-8122), but had minimal impact on a hybrid C18-COOH + C16-Me 2FA-bundle (K_a′ = 18 for TE-8102 and 19 for TE-8123).
• 3.Conjugation site: shifting the conjugation site from Lys-26 to Cys-53 marginally decreased the albumin affinity of hybrid C16-CH₃ + C18-COOH chains (K_a′ = 18 for TE-8102 and 15 for TE-8131), but had a more significant effect on C18-COOH chains (e.g., K_a′ dropped from 28 for TE-8101 to 18 for TE-8132). However, the trend observed for homogeneous C18-COOH chains reversed for longer C20-COOH chains (e.g., K_a′ increased from 139 for TE-8113 to 160 for TE-8134).

Effective 2FA-GLP-1 Conjugates for Blood Glucose Control

Having established the improved HSA binding capabilities of most 2FA-GLP-1 molecules, we next evaluated their efficacy in controlling blood glucose levels in diabetic db/db mice compared with semaglutide and a negative control group receiving phosphate-buffered saline (PBS). First, we measured baseline blood glucose levels in mice (ranging from 16 to 29 mmol/L) and set this as 100%. Next, we monitored glucose levels at various time points after a single subcutaneous dose of semaglutide/2FA-GLP-1 and computed the change in glucose levels (Δglucose) from baseline (Figure S3). To quantify the overall blood sugar response over time, we calculated the area under the curve (AUC) values for blood glucose levels up to 96 h for each molecule (Table S2). The relative AUC values, AUC′ = AUC (2FA-GLP-1)/AUC(semaglutide), are presented in Table 1.

Several 2FA-GLP-1 conjugates showed AUC′ < 1, indicating stronger blood sugar-lowering effects than semaglutide. Those with hybrid C16-Me + C18-COOH chains separated by 4 or 6 EG units had lower blood glucose levels than semaglutide (AUC′ = 0.77 for TE-8102 and 0.63 for TE-8123), maintaining lower blood glucose levels for up to 72 h postinjection: at 48 h postadministration, mice treated with TE-8102 or TE-8123 exhibited lower random blood glucose levels (31%) than semaglutide-treated mice (41%). This difference was even greater at 72 h, when glucose levels in semaglutide-treated mice rose to 81%, whereas those in mice treated with TE-8102 (50%) or TE-8123 (38%) remained significantly lower. Relocating the conjugation site from Lys-26 in TE-8102 to Cys-53 (TE-8131) slightly reduced HSA affinity and efficacy (AUC′ increased from 0.77 to 0.91), but the blood sugar reduction at 72 h postinjection for TE-8131 (−8.2 ± 1.4 mmol/L) still exceeded that for semaglutide (−3.8 ± 0.8 mmol/L).

Importance of Balancing Albumin Affinity and Blood Sugar Control

Enhanced albumin binding did not always correlate with improved glycemic control. For instance, excessively strong albumin binding in TE-8113 (K_a′ = 139) and TE-8134 (K_a′ = 160), led to reduced efficacy compared with semaglutide (AUC′ = 1.25). Conversely, reduced albumin affinity seen in TE-8132 where the conjugation site relocated from Lys-26 in TE-8101 to Cys-53 (K_a′ dropped from 28 to 18) improved glycemic control potency to a level comparable to semaglutide (AUC′ decreased from 1.43 to 1). These findings highlight the intricate interplay between HSA binding affinity and glycemic control efficacy: overly strong HSA binding of 2FA-GLP-1 may affect GLP-1R activation. Thus, our approach enables the fine-tuning of these interactions and identification of optimal 2FA-GLP-1 configurations with enhanced therapeutic properties.

The Multiarm Linker’s Carboxylate C-Terminus Enhances In Vivo Stability

Initially, we modified the multiarm linker by replacing its C-terminal COOH with a methyl group (CH₃) to protect it from breakdown by endogenous peptidases. This esterification did not enhance stability as anticipated, and the molecules often reverted to COOH C-terminus upon long-term storage (Figure S4). Hence, we synthesized TE-8105, similar to TE-8102, but with an unmodified COOH C-terminus in the linker, and evaluated its efficacy in controlling blood sugar. TE-8105 showed lower albumin binding (K_a′ = 2.2) but comparable glucose-lowering activity to its esterified counterpart TE-8102 (AUC′ ∼ 0.8).

TE-8105 as a Lead Candidate

To identify a lead candidate among TE-8102, TE-8123, and TE-8105, which showed better glucose control efficacy than semaglutide, we compared their pharmacokinetic properties. A single dose of ¹³¹I-radiolabeled semaglutide, TE-8102, TE-8123, or TE-8105 was injected into Sprague–Dawley rats and half-lives and mean residence times (MRTs) were determined. All three 2FA-GLP-1 molecules showed prolonged half-lives and MRTs compared with semaglutide (Table 2), consistent with their better glucose-lowering efficacy in db/db mice. TE-8105 displayed better in vivo stability, longer half-life, and MRT than TE-8102 or TE-8123, making it the choice candidate for further investigation as a potential T2D treatment.

Table 2. Pharmacokinetics of Semaglutide, TE-8102, TE-8123, and TE-8105 in Rats.

GLP-1 RA	t1/2a (h)	Tmaxb (h)	MRTc (h)
Semaglutide	17.64	16	30.91
TE-8102	23.88	16	41.17
TE-8123	24.55	16	43.61
TE-8105	31.34	24	53.30

^aThe time it takes for the drug concentration in the bloodstream to reduce by half.

^bThe time it takes for a drug to reach its maximum concentration in the bloodstream after administration.

^cThe mean residence time a drug spends in the body after administration.

TE-8105 Exhibits Enhanced Longer-Lasting Glycemic Control

To assess TE-8105’s efficacy in controlling blood glucose compared with semaglutide, db/db mice received subcutaneous doses (1, 3, 10, 30, and 100 nmol/kg) of PBS, semaglutide or TE-8105 with three mice per dose group. Blood glucose levels were monitored before dosing and at multiple time points postdosing. TE-8105 and semaglutide dose-dependently reduced blood glucose levels (Figures 2A and S5A). The half-maximal effective concentration (EC₅₀) values of semaglutide and TE-8105 were derived from their dose–response profiles. TE-8105 (EC₅₀ = 1.19 nM) is ∼7.8 times more potent than semaglutide (EC₅₀ = 9.24 nM) in regulating blood glucose levels. At 3 nmol/kg, TE-8105 exhibited more effective and prolonged glucose-lowering action than semaglutide (Figure 2A). At higher doses (e.g., 30 nmol/kg), TE-8105 maintained effective blood glucose control for 72 h, with partial effectiveness even at 120 h postinjection, whereas semaglutide’s efficacy waned after 48 h (Figure 2B). TE-8105’s improved stability and enhanced pharmacokinetics properties compared to semaglutide, especially at the later time points, likely contribute to its improved glucose-lowering efficacy.

Figure 2.

Dose–response relationships of semaglutide and TE-8105. Comparison of dose–response relationships between TE-8105 and semaglutide in db/db mice at (A) 3 nmol/kg and (B) 30 nmol/kg. Data are mean ± standard error of the mean (SEM). Comparison of dose–response relationships between semaglutide and TE-8105 in triggering insulin release from isolated mouse pancreatic islets at (C) high glucose concentration (17.5 mM) and (D) low glucose concentration (2.5 mM). Data were shown as mean ± SEM and were analyzed by 2-way ANOVA. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

In Vitro Characterization of TE-8105

To elucidate the pharmacodynamic (receptor activation) vs pharmacokinetic (albumin binding and half-life) components contributing to the glucose-lowering activity of TE-8105, we measured cAMP production in GLP-1R-overexpressing CHO-K1 cells with and without HSA. In the absence of albumin, TE-8105, semaglutide, and GLP-1 showed comparable potencies in stimulating cAMP production (Figure S5B) with EC₅₀ values of 0.92 ± 0.13 nM, 0.28 ± 0.05 nM, and 0.71 ± 0.18 nM, respectively. This indicates that all 3 molecules have similar intrinsic receptor activation abilities. However, in the presence of albumin, the dose–response curves for TE-8105 (EC₅₀ = 43.3 ± 8.2 nM) and semaglutide (EC₅₀ = 25.6 ± 6.1 nM) were significantly right-shifted compared to GLP-1 (EC₅₀ = 0.16 ± 0.01 nM). This shows that albumin binding of TE-8105 and semaglutide diminished their receptor potency relative to unmodified GLP-1.

TE-8105 Mimics GLP-1 to Stimulate Glucose-Dependent Insulin Secretion

When blood glucose levels rise, pancreatic β cells naturally release insulin to lower them. By binding to its receptor, GLP-1 amplifies this response and enhances insulin secretion by β cells.³⁷ This response ceases when the blood glucose level reaches 2.8 mM.³⁸ We assessed TE-8105’s ability to mimic GLP-1’s insulin secretion effect by measuring insulin levels from isolated pancreatic islets stimulated with GLP-1, semaglutide, and TE-8105 at varying doses (10 or 100 nM) and glucose concentrations (2.5 or 17.5 mM) in the presence of albumin-containing fetal bovine serum (FBS). At high glucose concentration (17.5 mM), TE-8105 induced greater insulin secretion than GLP-1 or semaglutide (Figure 2C). However, at low glucose concentration (2.5 mM), none of the molecules triggered insulin release (Figure 2D). These results affirm TE-8105’s ability to function similarly to native GLP-1 in vitro, stimulating insulin secretion under high blood sugar conditions.

To rationalize TE-8105’s enhanced insulin secretion at high glucose concentrations compared with GLP-1 or semaglutide, we hypothesized that albumin-bound TE-8105 may delay the internalization and degradation of the GLP-1R complex within endosomes, allowing GLP-1Rs to remain longer on the cell surface, leading to increased insulin secretion. To investigate this hypothesis, we monitored GLP-1R internalization after ligand binding under high-glucose conditions using confocal imaging of HEK293-T cells overexpressing GLP-1R tagged with enhanced green fluorescent protein (EGFP). Initially, GLP-1Rs were abundant on the cell membrane, appearing as intense, continuous plasma membrane fluorescence (Figure 3, leftmost images).

Figure 3.

Internalization of GLP-1R under GLP-1, semaglutide, and TE-8105 treatment. Confocal imaging of HEK-GLP-1R-EGFP cells under GLP-1, semaglutide, and TE-8105 induction without FBS (first 3 rows) and with FBS preincubation (last 3 rows).

TE-8105 Retards GLP-1R Internalization

To infer the rate and extent of GLP-1R internalization, we compared time-lapse images of cells treated with GLP-1, semaglutide, or TE-8105 with and without FBS preincubation. Without FBS preincubation, all 3 molecules showed similar patterns of concentrated intracellular fluorescence, peaking around 30 min (Figure 3, first 3 rows), indicating that they all triggered comparable GLP-1R internalization. With FBS preincubation, semaglutide and TE-8105 showed reduced internalized fluorescence intensity compared to GLP-1 (Figure 3, last 3 rows), suggesting that albumin-bound semaglutide and TE-8105 retarded GLP-1R internalization compared with their free forms, in line with previous findings that a C16 diacid-modified exendin-4 analog reduced GLP-1R endocytosis more than unacylated exendin-4.³⁹ Notably, TE-8105 displayed more sustained surface fluorescence than semaglutide, indicating slower GLP-1R internalization. The differences in receptor internalization triggered by TE-8105, semaglutide, and GLP-1 in the presence of albumin are due to differences in albumin binding rather than differences in their receptor activation potencies, which are similar in the absence of albumin (Figure S5B). These results suggest that the 2FA design of TE-8105 not only enhances albumin binding, but also helps retain the receptor on the cell surface for a longer duration, potentially prolonging GLP-1R signaling and enhancing insulin secretion.

Altogether, the in vitro results suggest that compared to semaglutide, TE-8105’s enhanced albumin binding and in vivo stability, leading to its longer half-life, slower GLP-1R internalization, and sustained GLP-1R activation, likely contribute to its enhanced in vivo glucose-lowering effects.

TE-8105 Shows Long-Term Efficacy in Diabetic Mice

TE-8105 Achieved Lower Blood Glucose Levels Than Semaglutide

To assess the long-term effects of TE-8105 treatment, db/db mice received subcutaneous injections of PBS, semaglutide, or TE-8105 at doses of 10 or 30 nmol/kg every 2 days for 8 weeks. Blood glucose levels were monitored throughout the treatment period. PBS-treated db/db mice maintained high blood glucose levels (17–30 mmol/L), whereas both TE-8105 and semaglutide significantly reduced blood glucose levels, with TE-8105 showing greater reduction at the higher dose (Figure 4A): at 30 nmol/kg, TE-8105 achieved a lower steady-state blood glucose (10.2 ± 1.2 mmol/L) than semaglutide (17.7 ± 2.0 mmol/L).

Figure 4.

Chronic treatment of TE-8105 improved glycemic control in db/db mice. (A) Blood glucose levels in db/db mice treated with PBS (gray), semaglutide (blue), or TE-8105 (red) at 10 or 30 nmol/kg dose every 2 days for 8 weeks. (B) HbA1c changes after long-term treatment. (C) Body weight change on day 58 vs day 0. (D) Heart, liver, iWAT, and BAT weights relative to body weights of db/db mice. (E) H&E staining of iWAT and liver sections after long-term treatment; 20× magnification. Arrows indicate macrovesicular steatosis (dotted line); microvascular steatosis (bold); hypertrophy (blue). (F) iWAT cell area and (G) diameter analyzed using StrataQuest Plus Analysis Software. Data shown as mean ± SEM and analyzed using ANOVA with multiple comparisons; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

TE-8105 Improved Long-Term Glycemic Control

To assess long-term blood glucose control, we measured glycated hemoglobin A1c (HbA1c) levels in db/db mice before the study (day 0) and on day 58. Mice treated with TE-8105 at both 10 and 30 nmol/kg showed a significant reduction in HbA1c levels compared to PBS- or semaglutide-treated mice (Figure 4B), indicating better long-term glycemic control with TE-8105.

TE-8105 Promoted Weight Loss and White Fat Tissue Reduction

Figure 4C shows the body weight change in db/db mice on day 58 vs day 0. PBS-treated mice exhibited a steady increase in body weight, from 40.7 ± 2.5 g (day 0) to 51.0 ± 2.4 g (day 58). TE-8105-treated mice showed significantly less weight gain, with the 30 nmol/kg dose resulting in significantly less weight gain than semaglutide-treated mice. Both semaglutide and TE-8105 treatments significantly reduced the proportion of inguinal white adipose tissue (iWAT) weight to total body weight without significant changes observed in the weight of other organs such as the heart, liver, and brown adipocyte tissue (BAT) (Figure 4D). Additionally, diabetic symptoms such as excessive eating (polyphagia) and drinking (polydipsia) were significantly reduced in treated mice compared to control mice.

TE-8105 Promoted Smaller Fat Cells

Hematoxylin and eosin (H&E) staining of iWAT revealed smaller adipocytes in TE-8105-treated mice compared to control mice (Figure 4E). TE-8105 treatment increased the percentage of smaller iWAT adipocytes with area <4,000 μm² (Figure 4F) and diameter of 15–30 μm (Figure 4G) compared with control. The 30 nmol/kg TE-8105 dose nearly eliminated large iWAT cells with area ≥16,000 μm².

TE-8105 Improved Liver Health

Figure 4E shows H&E staining of liver tissue. PBS-treated mice with dysfunctional leptin receptors showed signs of hypertrophy (blue arrow), macrovesicular and microvesicular steatosis (dotted and solid arrows), characterized by large lipid droplets and swollen liver cells (hepatocyte ballooning). These pathological features were markedly reduced in the semaglutide- and TE-8105-treated groups.

In summary, chronic treatment with TE-8105 significantly improved glycemic control in db/db mice, as evidenced by reduced blood glucose and HbA1c levels. TE-8105 also resulted in less weight gain, reduced relative iWAT weight, and improved histological profiles of iWAT and liver tissues. These effects were dose-dependent, with the 30 nmol/kg dose of TE-8105 showing the most significant improvements across the measured parameters.

TE-8105 Enhanced Weight Loss and Metabolic Improvements in Obese Mice

We investigated TE-8105’s efficacy in reducing body weight and fat mass in C57BL/6 diet-induced obese (DIO) male mice. Mice received subcutaneous injections of PBS, semaglutide, or TE-8105 at doses ranging from 3 to 100 nmol/kg every 4 days for 40 days. Body weight, composition, and blood glucose levels were monitored throughout the study.

TE-8105 Showed Dose-Dependent Weight Loss and Improved Glycemic Control

TE-8105 achieved greater weight loss than semaglutide at equivalent doses (Figure 5A,B). At 10 nmol/kg, TE-8105 reduced body weight by 10.1 ± 1.1%, whereas semaglutide resulted in a nonsignificant weight loss of 2.8 ± 1.8%. At the maximal dose (100 nmol/kg), TE-8105 reduced body weight by 16.3 ± 3.9%, compared to 9.6 ± 1.4% for semaglutide. Additionally, TE-8105 (30 nmol/kg) significantly lowered blood glucose levels, whereas semaglutide did not show a significant effect (Figure 5C).

Figure 5.

Chronic treatment of TE-8105-induced weight loss in DIO mice. (A) Body weight of DIO mice treated with PBS, and 3–100 nmol/kg of semaglutide or TE-8105 every 4 days for 40 days. (B) Body weight change of day 40 vs day 0. (C) Blood glucose levels and (D) % fat and lean mass post-treatment. (E) Percentage of iWAT cells with diameter ≤60 μM (black) and >60 μM (white) after treatment. (F) Total cholesterol and ALT levels post-treatment. Data shown as mean ± SEM and analyzed using ANOVA with multiple comparisons; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

TE-8105 Promoted Favorable Body Composition Changes

As the body weights of mice vary, the weight of fat or lean mass (in grams) was divided by individual body weight to yield the percentage of fat/lean mass. In the PBS control group, the mean fat mass (13.8 ± 0.3 g) and lean mass (23.2 ± 0.6 g) comprised 31.4 ± 0.4% and 52.7 ± 0.4% of the mean total body weight. Compared to PBS, TE-8105 dose-dependently reduced fat percentage: at 100 nmol/kg, TE-8105 reduced fat mass from 31.4 ± 0.4% to 19.8 ± 3.1% and increased the lean mass from 52.7 ± 0.4% to 62.5 ± 2.8% (Figure 5D). It also increased the proportion of small adipocytes (≤60 μm), while decreasing the proportion of larger ones (>60 μm) compared to PBS (Figures 5E and S6). Semaglutide showed no significant changes compared to PBS.

TE-8105 Improved Dyslipidemia and Liver Health

TE-8105 also improved dyslipidemia and liver health markers. TE-8105-treated mice exhibited lower total cholesterol levels at 10, 30, and 100 nmol/kg, whereas semaglutide-treated mice showed significant reduction only at the highest dose (Figure 5F). Although neither treatment significantly lowered alanine aminotransferase (ALT) levels, TE-8105 significantly reduced liver fat droplets and lipid accumulation after 40 days compared to PBS-treated mice (Figure S6).

TE-8105’s Potential to Treat NASH at Lower Doses

We investigated TE-8105’s potential to reduce fatty liver in a 10-week study using Gubra Amylin DIO NASH (GAN-DIO-NASH) male mice on a GAN diet. Mice received daily doses of PBS, semaglutide (5 or 30 nmol/kg), or TE-8105 (5, 10, or 30 nmol/kg). TE-8105 outperformed semaglutide in weight loss and liver function at the lowest dose (5 nmol/kg), achieving significantly greater reductions in weight (∼20% vs 5.7 ± 1.4%), fat mass, ALT levels, liver lipid droplets, and liver triglyceride levels, (Figure 6A–F). However, at the highest dose (30 nmol/kg), both TE-8105 and semaglutide improved liver health compared with control mice.

Figure 6.

Chronic treatment of TE-8105-induced weight loss in GAN-DIO-NASH mice. (A) Body weight of GAN-DIO-NASH mice treated daily with PBS, semaglutide (5 or 30 nmol/kg), or TE-8105 (5, 10, 30 nmol/kg) for 10 weeks. (B) Body weight change from day 0 to day 70. (C) Fat mass and (D) ALT levels of mice on days 0 and 70. (E) H&E staining of liver; 20× magnification. (F) Liver triglyceride levels post-treatment. Data shown as mean ± SEM. ANOVA with multiple comparisons for (B) and (F), Student’s t test for (C) and (D). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Discussion and Conclusions

This study presents a 2FA platform based on a novel multiarm linker for creating 2FA-peptide drugs with improved pharmacokinetic and therapeutic properties. Using this 2FA platform, we generated diverse 2FA-GLP-1 conjugates, and one of them, TE-8105, showed a much longer serum half-life and better long-term efficacy than semaglutide in preclinical models of T2D, obesity, and NASH.

Advantages of the Multiarm Linker Technology

Our technology, allowing precise control over FA spacing, combination, and attachment site, confers several advantages over traditional methods using PEGylation or a single FA to extend peptide half-life:

• 1.It enables precise 2FA-bundle attachment at a single peptide site, overcoming limitations of bulky PEGylation or dual-site FA attachments that can alter receptor interactions and function. For example, GLP-1 with C16 diacids attached to native Lys-26 and Lys-34 decreased receptor potency by 304-fold in the absence of albumin.^12,23 In contrast, TE-8105 showed similar receptor potency to GLP-1 without HSA (Figure S5B).
• 2.It enables systematic screening of different combinations of linkage, FA characteristics, and linker properties to identify optimal 2FA-peptide conjugates with improved albumin affinity, pharmacokinetics, and therapeutic properties. It addresses limitations of fixed FA spacing in previous studies using a Lys linker to attach two FAs of fixed spacing via γGlu linked to its N-terminal and side chain amino groups.^40,41
• 3.By utilizing endogenous peptides/FAs and only 2 exogenous PEGylated aa residues, our approach minimizes the risk of immunogenicity. In contrast, drugs with nonhuman peptide sequences may elicit antibodies that may affect drug safety and efficacy.
• 4.Our technology simplifies synthesis of homogeneous 2FA-peptide candidates, whereas large dual/triple agonist peptide drug candidates may face difficulty in synthesis.⁴²
• 5.It can generate more efficacious 2FA-peptide conjugates, potentially allowing for lower therapeutic doses, reducing side effects and improving tolerability.
• 6.It is versatile, allowing modification of the N-terminal reactive group, the number of lysines and their spacing in the multiarm linker to enable the conjugation of other moieties such as cytotoxic drugs or radionucleotides to antibodies.^43,44

Advantages of TE-8105

The multiarm linker technology enabled precise dual FA control, optimizing TE-8105’s binding to multiple pockets on albumin for simultaneous enhanced albumin affinity, pharmacokinetics, and glycemic control compared with semaglutide (Tables 1 and 2). The C16-Me and a C18-COOH FA spaced 4 EG units apart in TE-8105 improve hydrophobic and electrostatic interactions with albumin’s multiple FA-binding sites, increasing albumin affinity compared to semaglutide’s single C18-COOH FA. TE-8105’s enhanced albumin binding, coupled with the high albumin concentration in the blood (∼0.6 mM¹³) and interstitial space (∼0.3 mM⁴⁵) suggests the majority of TE-8105 is in the albumin-bound state. We hypothesize that the higher albumin affinity of TE-8105 compared to semaglutide would (i) slow release from the injection site to the bloodstream, (ii) enhance the distribution of the drug in high albumin-containing tissues, (iii) reduce renal and metabolic clearance, and (iii) delay GLP-1R internalization and degradation. These factors collectively prolong TE-8105 circulation and GLP-1R residence on the cell surface, leading to sustained GLP-1R activation and improved long-term glycemic control of TE-8105 compared to semaglutide. Importantly, like native GLP-1, TE-8105’s glucose-dependent insulin secretion minimizes the risk of hypoglycemia, making TE-8105 safer than drugs that stimulate insulin secretion through glucose-independent pathways. Furthermore, TE-8105’s acidic C-terminus provides excellent solubility (up to 20 mg/mL in buffered conditions).

TE-8105’s Therapeutic Potential for Treating Diabetes, Obesity, and Fatty Liver Disease

TE-8105 exhibited better long-term efficacy than semaglutide in reducing blood glucose levels, HbA1c, body weight, iWAT cell size, and ameliorating fatty liver disease symptoms in diabetic mice (Figure 4). This dual action on glucose regulation and weight management is particularly beneficial for diabetic patients with obesity. TE-8105 induced dose-dependent weight loss, reduced fat mass, and shifted fat cell distribution toward healthier, smaller adipocytes in obese mice (Figure 5). Unlike semaglutide, TE-8105 at low doses (5 nmol/kg) reduced liver fat content and improved liver health in GAN-DIO-NASH mice (Figure 6). Given the dose-dependent side effects of semaglutide,^31,46 TE-8105’s greater efficacy at lower doses suggests a reduced treatment burden.

Potential Impact

Compared with semaglutide, TE-8105’s longer duration of action and efficacy in reducing blood glucose and body weight imply less frequent administration and lower dosage requirements. This could reduce side effects and treatment costs, and improve patient compliance. TE-8105’s promising results, especially at low doses, in reducing body weight and fat mass, and improving liver health make it a strong candidate for treating NASH, a condition with limited treatment options. Although resmetirom (Rezdiffra), a once-daily oral, liver-directed thyroid hormone receptor-β agonist, has received accelerated approval for NASH patients with moderate to advanced liver fibrosis, it may cause liver toxicity, necessitating the monitoring of liver enzymes. TE-8105, like GLP-1, may reduce liver fat synthesis and accumulation by inducing autophagy and fat degradation,⁴⁷ potentially halting the progression of liver damage in NASH patients.

Limitations and Future Studies

Further studies are needed to elucidate the precise mechanism by which TE-8105’s unique 2FA structure enhances albumin binding and impacts receptor activation, tissue distribution, central nervous system penetration, and adipocyte remodeling or lipolysis, leading to improved efficacy compared with semaglutide. Ongoing clinical studies will evaluate TE-8105’s translation potential and efficacy compared with semaglutide, tirzepatide or other coagonists/triagonist. Future research will explore potential synergistic effects with other T2D/obesity drugs and applications in other metabolic disorders.

In conclusion, our multiarm linker technology provides a versatile platform for developing improved 2FA-peptide drugs for various therapeutic applications. TE-8105, a product of this approach, exhibits a favorable pharmacokinetic profile and sustained therapeutic action, potentially offering significant benefits for patients with T2D, obesity, and NASH.

Experimental Section

Study Design

This study designed a novel multiarm linker technology for creating diverse 2FA-peptide conjugates. The study compared the abilities of the 2FA-GLP-1 conjugates and semaglutide to bind HSA and lower blood glucose levels, as well as their pharmacokinetic properties, to identify the lead candidate, TE-8105. The efficacy of TE-8105 compared with semaglutide was evaluated using preclinical animal models: T2D (db/db mice), obesity (C57BL/6 diet-induced obese mice), and NASH (Gubra Amylin DIO NASH mice). The mice were randomized into groups and administered subcutaneous injections of PBS (as negative control) or varying doses of semaglutide or TE-8105. Blood glucose levels and body weight were assessed in db/db mice. Body weight, fat mass, and adipocyte or liver histology of obese and NASH mice were assessed to determine treatment effects.

Syntheses of GLP-1-2FA Analogues

The designed GLP-1-2FA analogues were synthesized by WuXi AppTec Company (Shanghai, China). Figure S1A–C outlines the 3-step synthesis: (i) synthesis of the modified GLP-1 peptide, (ii) synthesis of the 2FA-bundle, and (iii) conjugation of the 2FA-bundle to GLP-1 via azide–alkyne or thiol-maleimide reaction. The azido-modified GLP-1 peptide in Figure 1b (1, Figure S1A) was synthesized using Fmoc-based solid-phase peptide synthesis (SPPS), starting from Gly-37 to His-7. Following removal of the Dde protecting group from the Lys-26 ε-amino group with 3% hydrazine hydrate in N,N-dimethylformamide (DMF) at 20 °C for 10 min, Glu and 2-azidoacetic acid were sequentially coupled. Similarly, the (GGGSG)₃-Cys-53 coupled GLP-1 peptide in Figure 1c (2, Figure S1A) was also synthesized via Fmoc-based SPPS, but from Cys-53 to His-7. The resin was then washed with methanol and dried under vacuum to obtain the side chain-protected peptide. Peptide cleavage from the resin and deprotection were achieved by stirring with a cleavage buffer (92.5% trifluoroacetic acid (TFA)/2.5% triisopropylsilane/2.5% H₂O/2.5% 3-mercaptopropionic acid) at 20 °C for 2.5 h. Both peptides were purified by preparative HPLC and their molecular weights were confirmed by HPLC-MS (Figure S2): The observed molecular weights for peptide 1 (3610.2 Da) and peptide 2 (4446.9 Da), derived from the m/z values at 1204.4 and 1483.3 ([M + 3H]³⁺), respectively, matched their theoretical molecular weights (3609.9 and 4446.8 Da).

The 2FA-bundles were synthesized by Fmoc-based SPPS, using Dde-Lys(Fmoc)–OH, Fmoc-Lys(Dde)–OH, and Fmoc-Glu-OtBu for orthogonal coupling, then modified with either a maleimide or an alkyne group, yielding the respective 2FA-bundle 3 or 4 (Figure S1B). After total deprotection and cleavage, the products were recovered by precipitating with cold isopropyl ether, yielding a white solid ready for subsequent use. Finally, the GLP-1 peptide was linked to the 2FA-bundle via a copper-catalyzed azide–alkyne cycloaddition (CuAAC) or maleimide–thiol click reaction in solution (Figure S1C). For the CuAAC reaction, a solution of peptide 1 (1.00 eq, TFA) in DMF (30.0 mL) was added to 2FA bundle 4 (1.31 eq, TFA), N,N-diisopropylethylamine (2.00 equiv), and CuI (0.30 equiv), and the mixture was stirred at 25 °C for 30 min. For the maleimide–thiol reaction, a solution of peptide 2 (1.00 eq, TFA) in acetonitrile/H₂O (30:70 (v/v), pH 7.5, 30.0 mL) was added to 2FA bundle 3 (1.1 eq, TFA), and the mixture was stirred at 25 °C for 10 min. The synthesized GLP-1–2FA analogues were purified by HPLC with a C18 column at 220 nm, achieving ≥95% purity by HPLC-MS (Table S1). Their molecular weights were determined using liquid chromatography-electrospray ionization (LC-ESI) and matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF).

Albumin Affinities of GLP-1 Analogues Using Surface Plasmon Resonance

Surface plasmon resonance was employed to assess the HSA binding affinities of the GLP-1 analogues using the Biacore T200 system (GE Healthcare). Initially, 2 μg/mL of HSA (A3782, Sigma-Aldrich) was immobilized onto a Biacore CM5 sensor chip (GE Healthcare) to achieve a surface density of approximately 880 resonance units (RU). Analyte solutions containing seven concentrations (40, 20, 10, 5, 2.5, 1.25, and 0.625 μM) of each GLP-1 analog were prepared. These solutions were then flowed over the sensor chip surface at a rate of 30 μL/min using PBS (pH 7.4) with 0.005% Tween 20 as the running buffer at 25 °C. The GLP-1 analog was allowed to interact with the immobilized HSA for up to 90 s. Subsequently, only the running buffer flowed over the chip surface for 180 s to allow the analyte to dissociate from albumin. After dissociation, 50 mM NaOH was injected for 30 s to regenerate the chip surface for subsequent cycles of association, dissociation, and regeneration. The binding responses recorded during these phases were fitted to a 1:1 Langmuir binding model using Biacore T200 Evaluation Software (GE Healthcare) to compute the association rate constant K_on (M^–1 s^–1), the dissociation rate constant K_off (s^–1), and the equilibrium association constant K_a (M).

GLP-1 RA Activity Assay

GLP-1R-overexpressing CHO-K1 cells (2 × 10⁴ per well) were seeded in a 96-well White/Clear bottom plate and cultured at 37 °C with 5% CO₂ for 24 h. The next day, the culture medium was replaced with serum-free F12-K medium. Varying concentrations of GLP-1, semaglutide, or TE-8105, either without HSA or preincubated with HSA for 30 min, were added to the wells. The cells were then incubated at 37 °C and 5% CO₂ for 30 min. Each condition was tested in triplicate. After incubation, intracellular cAMP levels were quantified using HitHunter cAMP assay for Biologics (Eurofins DiscoverX #90-0075LM10). Luminescence was detected with a Synergy H1 multimode microplate reader, and EC₅₀ values were calculated using GraphPad Prism.

Animals

All animal experiments were conducted following internationally accepted principles for the care and use of laboratory animals and were approved by the Institutional Animal Care and Use Committee (IACUC) of the Institute of Cellular and Organismic Biology, Academia Sinica and National Yang-Ming Chiao Tung University. Male C57BL/6 mice (5–6 weeks old) and BKS.Cg-Dock7^m +/+ Lepr^db/JNarl (db/db) diabetic mice (6 weeks old) purchased from Taiwan’s National Laboratory Animal Center, and Sprague–Dawley rats purchased from BioLASCO Taiwan Co. Ltd., were housed in groups of 3–5 mice or 2 rats in a well-ventilated humidified room with a 10h:14h dark-light cycle. They had access to a standard chow diet and water ad libitum. To minimize the stress on animals due to handling, all animals were accustomed to blood sampling and dosing procedures 1 week before the experiments.

Evaluating GLP1-FA Analogues for Blood Glucose Control in Db/Db Mice

To assess the effect of the GLP1–1FA/2FA analogues in Table 1 on blood glucose levels, db/db mice were divided into 3 groups of four. Mice were subcutaneously administered with PBS, serving as control, or 100 nmol/kg GLP1–1FA/2FA analogues. Blood glucose levels were monitored before dosing (time 0) and at 2, 4, 8, 12, 24, 48, 72, and 96 h postdosing using an Accu-Chek Active blood glucose meter (Roche) on blood samples obtained from the tail-tip capillaries of conscious mice.

Pharmacokinetics Analysis

2FA-GLP-1 analogues (TE-8102, TE-8123, and TE-8105) and semaglutide were labeled with radioactive ¹³¹I. GLP1-FA (0.6 mg) in 50 mM phosphate buffer solution (0.6 mL, pH 7.2) was added to a tube precoated with Iodogen containing 20 μL of carrier-free Na¹³¹I solution (11.0–37.0 MBq). After incubation at room temperature for 15 min, the sample was removed from the reaction tube to stop iodide oxidation. Radiolabeled GLP1-FA was separated from unbound radioactive iodine by spinning filtration (Amicon centrifugal filter, MWCO 1 kDa) at 6,000g three times to remove free 131I-iodide. The purity of each ¹³¹I-labeled GLP1-FA analog was assessed using reverse-phase thin-layer chromatography (RP-TLC). The RP-TLC strips were developed by ascending chromatography (EtOH/H₂O = 1:1) and scanned using a radio-TLC scanner. The peptide-associated radioactivity was measured using a dose calibrator and expressed as a percentage of the total radioactivity.

Three male Sprague–Dawley rats aged 6–8 weeks received subcutaneous injections of ¹³¹I-labeled GLP1-FA analogues (3.8 MBq, ∼200 μg in 200 μL of phosphate-buffered saline). Blood samples were collected at selected time points postinjection via tail vein puncture. The radioactivity of ¹³¹I in the blood samples was measured using a gamma counter. The GLP-1 analog concentration in blood was estimated by dividing specific concentration (MBq/mL) by specific activity (MBq/nmol) after decay correction. Pharmacokinetic parameters were determined using PKsolver with an extravascular, noncompartmental model.

Isolating Pancreatic Islets

Pancreatic islets were obtained from donor mice (C57BL6/J, 8–10 weeks) using mechanical enzyme digestion and density gradient purification methods. After euthanizing mice under general anesthesia, their abdomens were dissected to access the pancreas. To separate the insulin-producing islets from the surrounding pancreatic tissue, collagenase D solution (2.5 mg/mL, Roche) in ice-cold Kreb buffer (3 mL) was injected into the bile duct. The pancreas was incubated with collagenase D at 37 °C for 20 min with gentle shaking. The enzyme digestion was stopped using 10 mL of RPMI1640 medium containing 10% FBS and 10 mM EDTA, keeping the mixture on ice. The mixture was then filtered through a prewetted 70 μm cell strainer and washed with 10 mL of medium. The pancreatic islets, being denser, were retained on the cell strainer’s inner surface and harvested with a 15 mL medium. They were overlaid with Ficoll Paque Plus (1.077) and centrifuged at 900g for 20 min to separate them from other cellular components. Pancreatic islets were collected from the Ficoll/media up-face fraction, passed through the cell strainer, washed with 10 mL of medium, and hand-picked using wide-open tips.

Insulin Secretion Assay

Isolated pancreatic islets were cultured overnight at 5% CO₂ and 37 °C in RPMI medium (100 μL) in a 96-well plate with 10 islets seeded in each well. The next day, the medium was replaced with fresh medium containing different concentrations of glucose (2.5 mM or 17.5 mM) and GLP-1 analogues (10 nM or 100 nM). After 24 h of incubation, the supernatant was collected from each well for insulin enzyme-linked immunosorbent assay (ELISA) assay (Ultrasensitive Mouse Insulin ELISA 10-1249-01, Mercodia) to measure the amount of insulin secreted by the islets under each condition.

Live-Cell Imaging of GLP-1R Internalization

Human embryonic kidney 293 (HEK293-T) cells overexpressing EGFP-tagged GLP-1Rs (HEK-GLP-1R-EGFP) were cultured on coverslips and incubated overnight with CO₂ at 37 °C. The next day, GLP-1 analogues (100 nM), preincubated with or without 10% FBS for 30 min, were added to the cells. To observe GLP-1R internalization, live-cell imaging using a confocal microscope (Zeiss LSM 880) was performed with images captured every 10 min. The cell culture medium used in the confocal study contained 4500 mg/L glucose.

Chronic Studies in db/db Mice

To assess the effects of long-term treatment, db/db mice were subcutaneously administered PBS, semaglutide, or TE-8105 (10 and 30 nmol/kg) every 2 days for 8 weeks. Blood glucose levels were measured every 2 days before dosing. The glucose AUC was calculated by a trapezoidal approximation of the percentage glucose levels measured every 2 days. Blood glucose profiles were plotted as mean ± SEM versus time. Additionally, body weight, food intake, and water intake of db/db mice were recorded every 2 days. On day 59, mice were euthanized and organs such as the heart, liver, iWAT, and BAT were acquired and weighed. Histological analysis was performed on the liver and iWATs fixed in 4% paraformaldehyde.

Diet-Induced Obesity (DIO) Study in DIO Mice

To induce obesity, male C57BL/6 mice, starting at 6 weeks old, were fed a high-fat diet, 45% kcal from fat, (D1245li, Research Diets, Inc.) until they reached 18–20 weeks old. To evaluate the effect of GLP-1-2FA on weight loss, DIO mice maintained on a high-fat diet received subcutaneous injections of PBS (100 μL), semaglutide (5 or 30 nmol/kg), or TE-8105 (3, 10, 30, 100 nmol/kg) every 4 days for 40 days. The blood glucose levels of mice were regularly monitored. Body weight, body composition, blood chemistry, and tissue histology of livers fixed in 4% paraformaldehyde were performed by the Taiwan Mouse Clinic. The weight of fat or lean mass was measured using a Minispec LF50 TD-NMR Body Composition Analyzer (Bruker).

NASH Study in GAN-DIO-NASH Mouse Model

To induce NASH development, 6-week-old male C57BL/6JNarl mice were fed a GAN diet (D09100310, Research Diets, Inc.) for 28–30 weeks. This GAN-DIO-NASH mouse model exhibits clinical translatability in terms of the physiological, metabolic, liver fat accumulation, inflammation, hepatocyte ballooning, and histopathological aspects of fibrosing NASH. Additionally, this model shows strong concordance with the liver transcriptome signature of human NASH.⁴⁸⁻⁵⁰ GAN-DIO-NASH mice were subcutaneously administered with PBS (100 μL), semaglutide (5 or 30 nmol/kg), or TE-8105 (5, 10, or 30 nmol/kg) once daily for 70 days. Body weight, body composition, blood chemistry, and tissue histology of livers fixed in 4% paraformaldehyde were performed by the Taiwan Mouse Clinic.

HbA1c Test

Twenty microliter blood samples were collected from chronic study db/db mice. The HbA1c level in each sample was measured using a Mouse Hemoglobin A1c Assay kit (cat. no. 80310, Crystal Chem).

Tissue Histology

Tissues were fixed in 4% paraformaldehyde and embedded in paraffin before sectioning. For general morphological observations, multiple sections were prepared and stained with H&E. The area of each cell WAT section was analyzed using the StrataQuest Plus Analysis System (TissueGnostics).

Statistical Analysis

Statistical analyses were performed using GraphPad Prism version 8.0 (GraphPad Software, La Jolla, CA). Data were expressed as mean ± SEM. Statistical significance between groups was determined using a two-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for multiple comparisons. For comparisons between two groups, an unpaired two-tailed Student’s t test was used. Differences were considered statistically significant at a p-value <0.05. The EC₅₀ values were calculated using nonlinear regression analysis. All experiments were performed in triplicate to ensure reproducibility and reliability of the results.

Glossary

Abbreviations

Aa

amino acid

ALT

alanine aminotransferase

AUC

area under the curve

BAT

brown adipocyte tissue

db/db

BKS.Cg-Dock7m + /+ Leprdb/JNarl

DIO

diet-induced obesity

DDP-4

dipeptidyl peptidase-4

EC₅₀

half maximal effective concentration

EG

ethylene glycol

EGFP

enhanced green fluorescent protein

ELISA

enzyme-linked immunosorbent assay

ESI

electrospray ionization

FA

fatty acid

FBS

fetal bovine serum

GLP-1

glucagon-like peptide-1

GLP-1R

glucagon-like peptide-1 receptor

HbA1c

hemoglobin A1c

H&E

hematoxylin and eosin

HEK293

human embryonic kidney 293

HSA

human serum albumin

HPLC

high pressure liquid chromatography

iWAT

inguinal white adipose tissue

k_on

association constant

k_off

dissociation constant

K_D

affinity constant

LC

liquid chromatography

MALDI-TOF

matrix-assisted laser desorption/ionization time-of-flight

MS

mass spectrometry

NASH

nonalcoholic steatohepatitis

PBS

phosphate-buffered saline

RP-TLC

reverse-phase thin-layer chromatography

RU

resonance units

SD rats

Sprague–Dawley

SEM

standard error of the mean

T2D

type 2 diabetes

WAT

white adipose tissue

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.4c02153.

• Molecular Formula Strings (CSV)

• Table S1: analytical data of GLP-1 RAs and HLPC spectra; Table S2: area under the curve (AUC) values for blood glucose levels from 0 to 96 h for various GLP-1 RAs; Figure S1: (A) GLP-1 peptide synthesis, (B) 2FA bundle synthesis, (C) Click reaction between GLP-1 peptide and 2FA bundle; Figure S2: (A) HPLC and ESI-MS spectra of azido-modified GLP-1 peptide 1, (B) HPLC and ESI-MS spectra of (GGGSG)3-Cys-53 coupled GLP-1 peptide 2; Figure S3: pharmacodynamic profiles of GLP-1 RAs in db/db mice; Figure S4: stability analysis of TE-8102 by MALDI-TOF MS; Figure S5: blood glucose response curves and cAMP accumulation response curves; Figure S6: (A) H&E staining of iWAT, (B) H&E staining of liver.(PDF)

Author Contributions

T.-W.C. conceived the platform and designed the multiarm linker. M.-T.W., H.-M.C., and T.-W.C. designed the study and interpreted the data. W.-C.L. performed Biacore assays; M.-T.W. performed imaging experiments. M.-T.W. and P.-H.L. performed in vivo experiments; C.-J.P. performed radioactivity experiments, H.-J.L. performed the GLP-1 RA activity assay, and J.D.W. provided modeling support. M.-T.W. and J.D.W. wrote first draft. C.L. analyzed data and edited the paper along with T.-W.C.

The authors declare the following competing financial interest(s): M.-T. Wong, J. D. Wright, P.-H. Lin, and H.-M. Chu are employees of Immunwork, Inc. W.-C. Lin, H.-J. Lee, C.-J. Peng, and H.-M. Chu are employees of T-E Meds, Inc. T-W Chang is the founder of Immunwork, Inc. and T-E Meds, Inc., and C. Lim is scientific advisor to both companies. All authors hold stock or have stock options from Immunwork, Inc. or T-E Meds, Inc.