FGF21 mimetic shows therapeutic promise

Abstract

FGF21 is a protein with multiple metabolic actions, currently a bright star beckoning on the metabolicdisease therapeutic horizon. Gaich et al. present the first clinical efficacy trial for an FGF21mimetic, finding robust lipid lowering effects, some reduction in body weight, but surprisingly modest effects on glucose homeostasis.

Fibroblast growth factor-21 (FGF21) is a 181-amino acid secreted protein that is produced in the liver, white and brown adipose tissue, and endocrine and exocrine pancreas. Since the unexpected discovery of a metabolic role forFGF21 in 2005_[IC1], FGF21 administration to rodents and non-human primates has been shown to reduce liver and circulating triglyceride levels and to increaseinsulin sensitivity, reducing hepatic glucose production and plasma glucose and increasing adipose glucose uptake. FGF21 also causes brown fatactivation, reduction in adipose tissue and body weight, and improvesβ-cell mass and function. In mice, overexpression of FGF21 producesadditional phenotypesconsistent with energy-conserving effects, including a smaller body size attributed to growth hormone resistance, easier entry into torpor, and decreased bone mass. Althoughit is difficult to weave a single unifying theme, FGF21 appears to be crucial for the regulation of fuel use and in adaptation to starvation (reviewed in (Potthoff et al., 2012)). These effects point to FGF21mimetics aspotential therapeutic agents for the treatment ofdiabetes, obesity, and dyslipidemia. Now Gaich et al. report the first clinical efficacy trial of LY2405319, a modified FGF21. They find that LY2405319 improves the lipid profile of obese, diabetic patients, but does not significantly affect weight or glucose homeostasis.

The molecular mechanisms by which FGF21 producesits physiologic effects are still being elucidated. FGF21 transcription is induced by peroxisome proliferator-activated receptor-γ(PPARγ) and PPARα activators, including drugs that are used to treat diabetes (thiazolidinediones) and hyperlipidemia (fibrates), respectively. FGF21 signals by binding toand activating the ‘c’ splicing isoforms of FGFR1, FGFR2, and FGFR3, with FGFR1c likely beingthe principal receptor. Signaling requires the β-klothocoreceptor, with tissue-specific responsiveness chiefly determined by the tissue distribution of β-klotho. Adipose tissue is required for the antidiabetic actions of FGF21 (Wu et al., 2011). FGF21 increasesadiponectinproduction by adipose tissue, suggesting that adiponectintransmits the effects of FGF21 tomuscle and liver(Holland et al., 2013; Lin et al., 2013). However, recentobservations make clear that the relevant mechanisms are even more complex: in adipose tissue FGF21 is not simply a downstream target of PPARγ, but also contributes to PPARγ activity, indicating a bidirectional interaction (Dutchak et al., 2012), and FGF21 is not required for thiazolidinedione action (Adams et al., 2013).

The excitement about FGF21 as a potential therapeutic agent is reflected in the creative efforts that have been expended to produce FGF21 mimetics, including LY2405319. Many factorsneed to beconsidered in these efforts, includingpotency, selectivity, pharmacokinetics (half-life, brain penetration), intellectual property, and manufacturing (feasibility, cost, stability). LY2405319, for example, contains a synthetic disulfide bond that increases the molecule's physical stability, as well as mutations that facilitate its production in yeast. Despite these modifications, LY2405319 has a specificity and potency comparable to FGF21 (Kharitonenkov et al., 2013). Another important consideration is the protein's half life—a longer half-life allowsless frequent dosing, obviating the need for daily injections. Whereas LY2405319's half-life appears to be similar to that of native FGF21, other mimetics have prolonged half-lives, due to innovations such as coupling to polyethylene glycol, fusion with an immunoglobulin constant domain, and coupling of two FGF21 molecules to a single antibody (Table 1). LY2405319 was the first FGF21 mimetic to reach a Phase 1clinical trial.

Table 1.

FGF21 Mimetic Molecules^*

Name, PMID Reference	Company	FGF21 Modification	Modification Result	Half-life**	Binding Specificity**
LY2405319	Lilly	L118C, A134C	new disulfide to increase stability	unchanged	FGFR1c / β-klotho
Kharitonenkov et al., 2013		S167A	prevent glycosylation in yeast		FGFR2c / β-klotho
PMID 23536797		ΔHPIP	monodisperse product		FGFR3c / β-klotho
PEG FGF21	Ambrx/Merck	R131pAcF	single reactive site for coupling	increased	FGFR1c / β-klotho
Mu et al., 2012		couple to PEG	increase t1/2		FGFR2c / β-klotho
PMID 22210323					FGFR3c / β-klotho
Fc-FGF21(RG)	Amgen	L98R	reduce aggregation	increased	FGFR1c / β-klotho
Veniant et al., 2012		P171G	reduce in vivo degradation		FGFR2c / β-klotho
PMID 22798348		fusion to Fc	increase t1/2		FGFR3c / β-klotho
CVX-343	Pfizer	A129C	single reactive site for coupling	increased	FGFR1c / β-klotho
Huang et al., 2013		couple to Ab	increase t1/2		FGFR2c / β-klotho
PMID 23720456					FGFR3c / β-klotho
FGFR1b/1c agonist antibody	Genentech	not based on FGF21	increase t1/2	increased	FGFR1b, FGFR1c
Wu et al., 2011
PMID 22174314
mimAb1 agonist antibody	Amgen	not based on FGF21	increase t1/2	increased	FGFR1c / β-klotho
Foltz et al., 2012
PMID 23197570
FGFR1c/β-klothobispecific	Amgen	not based on FGF21	increase t1/2	increased	FGFR1c / β-klotho
Smith et al., 2013
PMID 23630589

^*Abbreviations: Standard one-letter amino acid abbreviations are used; Δ, deletion; pAcF, p-acetylphenylalanine; PEG, polyethylene glycol; t1/2, half life; Fc, antibody constant domain; Ab, antibody

^**Some half-lives and binding specificities are inferred.

The table lists published FGF21 mimetics, along with their incorporated modifications, half-life, and binding specificity.

As Gaich et al. report, LY2405319, at 0, 3, 10, or 20 mg, was injected into obese subjectswith type 2 diabetes once daily for 28 days, with approximately10 patients completing treatment per dose. The drug improved the patients’ lipid profiles, lowering LDL levels 20-30%, total cholesterol 15-19%, and triglycerides 26-46%, and raising HDL 15-20%. Subjects receiving the drug also lost 1.3-1.5 kg more than the placebo-treated patients, although this amount did not reach statistical significance. There was also a hint of improved glucose homeostasis (lower glucose and insulin levels), but the changes were surprisingly small and not statistically significant.

One challenge with modified proteins is that they may be perceived as foreign by the immune system. The immune response can range from injection site irritation, to development of anti-drug antibodies without effect, to antibodies that alter pharmacokinetics, neutralize drug, or neutralize endogenous FGF21, to hypersensitivity reactions. InGaich et al., some patients developed injection site reactions, about half developed antibodies, and there was one serious hypersensitivity reaction.However, the LY2405319 antibodies were not neutralizing and did not alter the pharmacokineticsof the drug, suggesting thatthe clinical efficacy of LY2405319 in this trial was likely not compromised by the immune response.

How do these results affect the prospectsfor clinical use of FGF21 mimetics? The results of the LY2405319 trial constitute clinical proof-of-concept that FGF21 mimetics improve lipid profiles.Future studies for lipid efficacy could focus on larger trials, optimizing treatment regimens and drug combinations, and identification of the optimum patient population. For weight loss, the 28 days in the current trial is shorter than the usual 12-week duration for proof-of-concept weight loss trials, precluding adequate assessment of weight loss efficacy. One issue not addressed by the LY2405319 trial is bone health. FGF21 was recently found to decrease bone mass, with inhibition of osteoblastogenesis(Wei et al., 2012). As Gaich et al. did not monitor bone biomarkers,this aspect will need to be examined in future clinical studies.

Understanding the lack of efficacy in reducing fasting glucose levels will also require further analysis. Is FGF21 intrinsically a poorchoicefor treating diabetes, do LY2405319 and FGF21 act differently, or was the trial design responsible? The adiponectin levels in this trial may not have plateaued, suggesting that increased exposure to drug could further increaseadiponectin levels toproduce a greater effect on diabetes. Would a longer-acting mimetic perform better? Three FGF21 mimeticsnot based on FGF21 itself (Table 1) may make it possible to target a different spectrum of physiologic effects. One mimetic is a monoclonal antibody that activates FGFR1c but not FGFR2c or FGFR3c, and does not require β-klotho. Another activating monoclonal antibody is specific for FGFR1c/β-klotho, and yet anothermimetic fuses two synthetic proteins that bind FGFR1c and β-klotho. The variety of engineered FGF21 mimetics should facilitate elucidation of the roles of FGFR1c vs FGFR2c/3c signaling and whether the lipid, glucose, growth, and bone phenotypes can be differentially impacted. Althoughnot the hoped for result, the glucose and insulin data are a minorsetback. Many of the questions can be addressed with further clinical trials and other mimetics.

Dyslipidemia, diabetes, and obesity are associated major health problems in need of improved therapeutic agents, andit is exciting to see the first clinical efficacy data for a new therapeutic approach. We look forward to further studies with FGF21 mimetics to address this need.