`
`pubs.acs.org/jmc
`
`Discovery of the Once-Weekly Glucagon-Like Peptide‑1 (GLP-1)
`Analogue Semaglutide
`Jesper Lau,* Paw Bloch, Lauge Schäffer, Ingrid Pettersson, Jane Spetzler, Jacob Kofoed, Kjeld Madsen,
`Lotte Bjerre Knudsen, James McGuire, Dorte Bjerre Steensgaard, Holger Martin Strauss, Dorte X. Gram,
`Sanne Møller Knudsen, Flemming Seier Nielsen, Peter Thygesen, Steffen Reedtz-Runge,
`and Thomas Kruse
`Global Research, Novo Nordisk A/S, Novo Nordisk Park, DK-2760 Måløv, Denmark
`*S Supporting Information
`
`ABSTRACT: Liraglutide is an acylated glucagon-like peptide-1 (GLP-1)
`analogue that binds to serum albumin in vivo and is approved for once-daily
`treatment of diabetes as well as obesity. The aim of the present studies was
`to design a once weekly GLP-1 analogue by increasing albumin affinity and
`secure full stability against metabolic degradation. The fatty acid moiety
`and the linking chemistry to GLP-1 were the key features to secure high
`albumin affinity and GLP-1 receptor (GLP-1R) potency and in obtaining
`a prolonged exposure and action of the GLP-1 analogue. Semaglutide was
`selected as the optimal once weekly candidate. Semaglutide has two amino
`acid substitutions compared to human GLP-1 (Aib8, Arg34) and is deri-
`vatized at lysine 26. The GLP-1R affinity of semaglutide (0.38 ± 0.06 nM)
`was three-fold decreased compared to liraglutide, whereas the albumin
`affinity was increased. The plasma half-life was 46.1 h in mini-pigs following
`i.v. administration, and semaglutide has an MRT of 63.6 h after s.c. dosing to mini-pigs. Semaglutide is currently in phase 3
`clinical testing.
`
`■ INTRODUCTION
`
`GLP-1 receptor agonists (GLP-1 RAs) have become a suc-
`cessful treatment for type 2 diabetes providing effective glucose
`control,
`improved beta-cell
`function, body weight loss, and
`lowering of systolic blood pressure.1 Exenatide for twice daily
`administration was the first approved compound in 2005, fol-
`lowed by once-daily liraglutide in 2009.2,3 Exenatide is a non-
`human peptide analogue originally isolated from the saliva of
`the Gila monster and has an i.v. half-life of 30 min and a half-life
`of 2−3 h after s.c. administration.4−6 Liraglutide is a close
`analogue of human GLP-1 designed to bind to human albumin
`via a fatty acid and a spacer covalently attached to the peptide
`backbone.7,8 Liraglutide has an i.v. half-life of 8−10 h and
`13−15 h after s.c. administration, making it suitable for once-
`daily administration.7,9,10 There is a strong PK/PD relationship
`in the GLP-1RA class where short-acting GLP-1RAs display a
`marked ability to reduce gastric emptying, whereas long-acting
`GLP-1RAs with pharmacologically relevant exposure more
`than 24 h have better glucose lowering effect and less effect on
`gastric emptying.1,11 The next generation of GLP-1RAs are
`mainly aimed for once weekly administration and albiglutide
`was approved as the first once-weekly GLP-1 analogue. Albiglu-
`tide has two Gly8 GLP-1 molecules fused in tandem to human
`serum albumin. The time to maximum plasma concentra-
`tion, Tmax, is 2−4 days and the half-life is around 6−8 days.12
`Recently, a second once weekly analogue dulaglutide, was approved.
`
`Dulaglutide is a GLP-1 analogue fused to a Fc-fragment,
`Gly8Glu22Gly36-GLP-1(7−37)-(Gly4Ser)3Ala-Ala234,235Pro228-
`13 There is still a need for optimization in the once-
`IgG4-Fc.
`weekly GLP-1 analogue field because both albiglutide and
`dulaglutide have been shown to be less efficacious than
`liraglutide with respect to weight loss. For albiglutide, weight
`loss in a 26-week trial was 0.6 kg, whereas it was 2.2 kg for
`liraglutide.14 For dulaglutide, weight loss was 2.9 kg, whereas it
`was 3.6 kg for liraglutide (p < 0.01).15
`We report a series of acylated GLP-1 analogues where in-
`creased albumin affinity relative to liraglutide has led to a profile
`suitable for once-weekly administration. The attachment of long
`fatty acids to peptides has successfully extended the half-life of
`native short-acting GLP-1 to the once daily profile of liraglutide.
`As albumin has a half-life of several weeks, it was hypothesized
`that an increased albumin affinity could extend the circulating
`half-life beyond the 13−15 h seen with liraglutide.16,17 The main
`focus of the studies reported here was to generate GLP-1RA
`analogues with an extended pharmacokinetic profile and efficacy
`suitable for once-weekly s.c. administration without compromising
`the potency. We selected semaglutide for clinical development,
`and the molecule is currently in phase 3 clinical development.
`The half-life in humans is reported to be 165 h.18
`
`Received: May 11, 2015
`Published: August 26, 2015
`
`© 2015 American Chemical Society
`
`7370
`
`DOI: 10.1021/acs.jmedchem.5b00726
`J. Med. Chem. 2015, 58, 7370−7380
`
`Downloaded via 131.226.37.235 on July 6, 2023 at 15:49:58 (UTC).
`
`See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
`
`Novo Nordisk Exhibit 2002
`Mylan Pharms. Inc.v. Novo Nordisk A/S
`IPR2023-00722
`Page 00001
`
`
`
`Journal of Medicinal Chemistry
`
`Article
`
`Table 1. Derivatives of Arg34 GLP-1 (7-37) with C16 to C20 Acids and C16 to C18 Diacids and Linkers Attached to Lys26 and
`Compared to GLP-1 (7−37) and Arg34 GLP-1 (7−37)a
`
`GLP-1R binding
`(IC50, nM) (SEM)
`
`sequence
`modification
`
`acylation
`position
`
`GLP-1R potency (EC50, pM)
`ratio 2%/0%
`(SEM) 0% HSA
`HSA
`2% HSA
`0% HSA
`linkage
`protractor
`analogue
`(GLP-1)7−37
`16.2 (0.9)
`0.5
`0.10 (0.02)
`0.19 (0.03)
`none
`none
`Arg34
`1
`7.6 (1.1)
`0.6
`0.14 (0.01)
`0.25 (0.01)
`none
`none
`γGlu
`Lys26
`Arg34
`8.5 (0.7)
`43
`4.78 (1.01)
`0.11 (0.02)
`C16
`liraglutide
`γGlu
`Lys26
`Arg34
`2
`17.1 (1.3)
`3
`1.95 (0.04)
`0.60 (0.04)
`C18
`γGlu-2xOEG
`Lys26
`Arg34
`3
`3.7 (0.8)
`6
`0.97 (0.05)
`0.16 (0.03)
`C20
`Lys26
`Arg34
`4
`70.9 (10.2)
`85
`74.0 (5.8)
`0.87 (0.06)
`none
`C16 diacid
`Lys26
`Arg34
`5
`238 (42)
`104
`295 (52)
`2.83 (0.37)
`none
`C18 diacid
`γGlu-2xOEG
`Lys26
`Arg34
`6
`3.8 (0.4)
`527
`148 (15)
`0.28 (0.04)
`C18 diacid
`aGLP-1 receptor binding at 0% and 2% albumin (HSA) using BHK cells expressing the human GLP-1R and in vitro potency measured in BHK cells
`that express both the human GLP-1R and a luciferase reporter system.
`
`■ RESULTS AND DISCUSSION
`Several series of derivatized GLP-1 analogues were prepared
`(Tables 1−4). The design of these peptides was mainly inspired
`by liraglutide, which was developed as the first GLP-1 appli-
`cable for once daily dosing with full daily efficacy. One of the
`important design criteria was to keep the analogues structurally
`similar to native GLP-1 and not introducing unnecessary amino
`acid changes in order to avoid immunogenicity responses
`similar to those shown for exenatide and taspoglutide. In the
`latter case the program was terminated following phase
`3 clinical development, whereas liraglutide has shown a very
`low immunogenic profile.19,20 For exenatide, a significant amount
`of patients developed antibodies, some were neutralizing, and
`the antibody numbers and titers were higher with the once
`weekly formulation than the twice daily simple formulation.21
`The design space included investigation of the structural impact
`on in vitro human GLP-1 receptor binding affinity and potency
`by modifying the “side-chain” of liraglutide (Figures 1 and 2).
`The results showed that GLP-1 receptor affinity measured in
`the absence of albumin was in general only marginally affected
`by attachment of an albumin binding moiety (protractor) to a
`lysine residue in positions 16, 22, 26, 37, or 38, and most
`derivatives had IC50 values below 0.5 nM. In some cases, the
`receptor affinity was significantly increased, as seen for analogues
`such as 13 (0.04 nM), 38 (0.06 nM), and 35 (0.06 nM) com-
`pared to native GLP-1 and to the nonderivatized analogues
`Arg34GLP-1 (7−37) and Aib8, Arg34 GLP-1 (7−37). The
`affinities of these analogues were significantly increased com-
`pared to the binding affinity of 0.11 nM for liraglutide. How-
`ever,
`it was also possible to reduce the receptor affinity as
`exemplified by 31, which had a poor binding affinity in the
`absence of albumin. In addition to receptor binding, an in vitro
`functional assay was used to estimate the agonistic potencies of
`the various GLP-1 analogues (Tables 1−4). Native GLP-1 had
`an EC50 value of 16.2 pM, whereas Arg34GLP-1 (7−37) and Aib8,
`Arg34GLP-1 (7−37) are slightly more potent (7.6 and 6.2 pM,
`respectively).
`In order to further improve the pharmacokinetic profile and
`obtain a product that is suitable for once weekly administration,
`the N-terminus was substituted to protect against DPP-4 degra-
`dation. As albumin has a serum half-life of approximately 3 weeks
`in humans,22 an increased affinity to albumin could increase the
`systemic half-life beyond the 13 h seen for liraglutide after s.c.
`administration.
`
`One major risk of increasing the albumin binding affinity is
`that the free active fraction would significantly decrease leading
`to a diminished in vivo potency and an increased dose needed
`to achieve acceptable efficacy. Based on the experience from
`various analogues,8,10 as well as for fusion proteins,23 it was
`a realistic concern whether it was at all possible to obtain
`sufficient albumin binding combined with high receptor affinity.
`The negative effect on receptor binding affinity and potency
`with increasing length and lipophilicity of the fatty acids has
`earlier been described8,24 and is also observed here. Com-
`parison of the data for liraglutide and 2 where the fatty acid was
`increased from C16 to C18 and 4 with 5 where the fatty acid
`was changed from C16 diacid to C18 diacid showed in both
`cases a significant reduction in potency (Table 1).
`It therefore became an ambitious design challenge to dem-
`onstrate that it was possible to design an analogue that was
`efficacious at a low dose while still being reversibly bound to
`albumin with an affinity sufficient to protract the systemic
`clearance.
`A series of Aib8, Arg34 GLP-1 (7−37) analogues derivatized
`at Lys26 were prepared, and the in vitro potency and receptor
`binding was compared with those of
`liraglutide. The first
`campaign was aimed at investigating how much the potency
`and receptor binding could be affected by modification of the
`γGlu linker motif between the peptide and the C16 fatty acid in
`the side-chain of liraglutide (Table 2). The results showed that
`the linker can have significant impact on potency and receptor
`binding. It is interesting to observe that the largest difference
`in binding affinity was in the presence of 2% albumin where
`there was 23-fold difference between 8 with a γGlu linker and
`12 with γGlu-3xOEG linker. The difference in functional
`potency was only four- to five-fold for this pair. There was a
`general tendency for improved binding affinity in the presence
`of albumin with increasing length of linker (8 < 10 = 11 < 12).
`The ratio between binding at low and high albumin affinity
`varied from two- to 42-fold, and the effect of adding albumin
`to the assay was highest for 8, which had the shortest linker.
`The effect of
`the Ala to Aib modification in position 8
`(liraglutide and 8) had no significant effect on binding affinity,
`whereas it seemed to impact the potency slightly in a negative
`direction. This observation is opposite to what has been
`described for nonderivatized GLP-1 where the Aib modification
`has a slightly positive effect on receptor binding.25
`In order to increase albumin binding beyond that of
`liraglutide, the C16 fatty acid was exchanged with other
`
`7371
`
`DOI: 10.1021/acs.jmedchem.5b00726
`J. Med. Chem. 2015, 58, 7370−7380
`
`Novo Nordisk Exhibit 2002
`Mylan Pharms. Inc.v. Novo Nordisk A/S
`IPR2023-00722
`Page 00002
`
`
`
`Journal of Medicinal Chemistry
`
`Article
`
`Figure 1. Abbreviations of linkers and protractors used to derivatize GLP-1 analogues in Tables 1−4.
`
`lipophilic moieties in a series of derivatives with potential for
`increased albumin affinity. Based on the knowledge about the
`interaction of fatty acids with albumin, it was expected that
`increased lipophilicity, as well as the location of the acidic
`group, could have an impact on albumin affinity.26−28 A series of
`derivatives were prepared where the properties of the albumin
`binding moiety were investigated (Table 2). Increasing the fatty
`acid length from C16 to C18 had a slightly negative impact on
`both potency and binding in the absence of albumin using the
`Arg34 GLP-1 peptide (compared 2 with liraglutide). The results
`are in accordance with earlier reported data,8 which concluded
`that lengthening the fatty acid chain had a negative impact on
`receptor affinity. However, it was observed that the binding in
`
`the presence of 2% albumin was improved by going from C16
`to C18. An important observation was that in the Aib8, Arg34
`GLP-1 series, the binding affinity with and without albumin was
`significantly improved by going from C16 to C18 fatty acid
`combined with the γGlu-2xOEG linker (compare 11 and 13).
`This result gave the first insight toward the importance of the
`linker. A series of analogues with increasing fatty acid chain
`length (C18 and C20) using γGlu-2xOEG and γGlu-3xOEG
`was prepared (Table 2). Interestingly, the potency that was
`lost in the C18 analogue 2 was regained by introducing the
`γGlu-2xOEG linker in 13 which is even more potent than
`liraglutide. Even the C20 analogue 14 had improved potency
`compared to liraglutide.
`
`7372
`
`DOI: 10.1021/acs.jmedchem.5b00726
`J. Med. Chem. 2015, 58, 7370−7380
`
`Novo Nordisk Exhibit 2002
`Mylan Pharms. Inc.v. Novo Nordisk A/S
`IPR2023-00722
`Page 00003
`
`
`
`Journal of Medicinal Chemistry
`
`Article
`
`Figure 2. Structures of liraglutide and semaglutide.
`
`Table 2. Derivatives of Aib8, Arg34 GLP-1 (7−37) with C16 to C20 Acids and C12 to C16 Diacids and Attached to Lys26 Using
`γGlu and OEG Based Linkersa
`
`GLP-1R binding
`(IC50, nM) (SEM)
`
`acylation
`position
`
`GLP-1R potency (EC50, pM)
`ratio 2%/0%
`sequence
`(SEM) 0% HSA
`HSA
`2% HSA
`0% HSA
`linkage
`protractor
`modification
`analogue
`Aib8, Arg34
`7
`6.2 (0.2)
`0.5
`0.08 (0.01)
`0.17 (0.01)
`none
`none
`γGlu
`Lys26
`Aib8, Arg34
`8
`19.2 (2.3)
`42
`5.05 (1.52)
`0.12 (0.05)
`C16
`Lys26
`Aib8, Arg34
`9
`11.1 (0.5)
`2.3
`0.75 (0.10)
`0.32 (0.01)
`2xOEG
`C16
`γGlu-OEG
`Lys26
`Aib8, Arg34
`10
`14.7 (1.6)
`21
`3.40 (0.86)
`0.16 (0.03)
`C16
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`11
`2.7 (0.3)
`20
`3.78 (1.05)
`0.19 (0.07)
`C16
`γGlu-3xOEG
`Lys26
`Aib8, Arg34
`12
`4.3 (0.2)
`8.0
`0.22 (0.14)
`0.03 (0.01)
`C16
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`13
`3.2 (0.7)
`50
`1.99 (0.59)
`0.04 (0.01)
`C18
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`14
`4.8 (0.6)
`4.9
`0.88 (0.08)
`0.18 (0.02)
`C20
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`15
`42.6 (7.0)
`0.9
`4.75 (0.80)
`5.16 (1.16)
`C12 diacid
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`16
`15.7 (0.8)
`3.3
`8.87 (1.90)
`2.65 (0.45)
`C14 diacid
`Lys26
`Aib8, Arg34
`17
`166 (61)
`3.9
`12.8 (1.4)
`3.25 (0.11)
`2xOEG
`C16 diacid
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`18
`8.6 (0.9)
`22
`20.5 (2.3)
`0.94 (0.14)
`C16 diacid
`aGLP-1 receptor binding at 0% and 2% albumin (HSA) using BHK cells expressing the human GLP-1R and in vitro potency measured in BHK cells
`that express both the human GLP-1R and a luciferase reporter system.
`
`In the original design of liraglutide, the γGlu linker was
`introduced to compensate for the loss of the acidic group of
`palmitate used for amide linkage. In order to investigate the
`impact of the position of the acidic group and its importance
`for the binding to basic residues of albumin, a series of deriv-
`atives comprising a terminal acid on the albumin binding
`side chain was next prepared and tested.29−31 The first two
`analogues were made using C16 diacids in combination with
`different linkers to the peptide backbone. As seen from Table 2,
`the addition of a terminal acid had a large negative impact on
`potency and binding (compare 17 with 9 and 18 with 11). The
`effect was quite dramatic with about 10-fold decreased affinity
`for 18 compared to 11. The small series of C16 diacid ana-
`logues in Tables 1 and 2 also confirmed that the linker had sig-
`nificant impact on potency. The impact of the Aib8 substitution
`was not considered to be responsible for this huge difference, as
`there was no difference between the two nonderivatized
`analogues 7 and 11 (Tables 1 and 2).
`The GLP-1 receptor affinity in the absence of albumin was
`used to get a measure of the true receptor affinity without
`competition with albumin affinity in the assay. Several attempts
`were made to measure the direct binding of the analogues to
`albumin in a screening mode, but due to the high unspecific
`adhesion of several of the analogues to various surfaces it has
`so far not been possible to obtain a reliable assay for albumin
`
`binding affinities. An alternative but indirect method to identify
`derivatives with high albumin affinity was to pick those ana-
`logues that had a right shift of the binding dose−response curve
`when high concentrations of albumin were in the assay
`compared to binding at low albumin concentration. Thus, we
`introduced the IC50 (high albumin)/IC50 (low albumin) ratio
`(BR ratio) as a surrogate of the albumin affinity with awareness
`of the pitfalls of this approach. This kind of data filter would of
`course not pick those derivatives that may have high affinity to
`albumin and the receptor simultaneously, but as we did not
`have a robust direct albumin affinity assay, this was a pragmatic
`screening plan to select analogues for animal pharmacokinetics
`studies.
`It was clear that there was a remarkable effect on the BR ratio
`by modifying the sequence, the acylation site, and the linker.
`The BR ratio varied from around 0.9 for 15 to 940 for
`semaglutide among the peptides in this study. Also the in vitro
`receptor binding affinities and the potencies were remarkably
`different with a potency span of more than 2000-fold including
`the outlier 31 and greater than 100-fold for the main popula-
`tion of the series.
`Based on the experience of modifying the fatty acids and the
`linker,
`it was concluded that
`the linker could have a
`pronounced impact on the in vitro properties. It was decided
`to focus more on this topic, and we hoped that extension of the
`
`7373
`
`DOI: 10.1021/acs.jmedchem.5b00726
`J. Med. Chem. 2015, 58, 7370−7380
`
`Novo Nordisk Exhibit 2002
`Mylan Pharms. Inc.v. Novo Nordisk A/S
`IPR2023-00722
`Page 00004
`
`
`
`Journal of Medicinal Chemistry
`Article
`Table 3. Derivatives of Aib8, Arg34 GLP-1 (7−37) with C18 to C22 Diacid Protractors Attached to Lys26 Using γGlu, OEG, Abu,
`and Benzyl Containing Linkersa
`
`GLP-1R binding
`(IC50, nM) (SEM)
`
`linkage
`
`GLP-1R potency (EC50, pM)
`ratio 2%/0%
`acylation
`sequence
`(SEM) 0% HSA
`HSA
`2% HSA
`0% HSA
`position
`protractor
`modification
`analogue
`Lys26
`Aib8, Arg34
`19
`269 (19)
`14
`27.0 (7.5)
`1.86 (0.27)
`none
`C18 diacid
`γGlu
`Lys26
`Aib8, Arg34
`20
`9.9 (0.7)
`541
`112 (8)
`0.21 (0.01)
`C18 diacid
`γGlu-OEG
`Lys26
`Aib8, Arg34
`21
`4.8 (0.2)
`477
`79.7 (10.7)
`0.17 (0.01)
`C18 diacid
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`6.2 (0.6)
`940
`357 (98)
`0.38 (0.06)
`C18 diacid
`semaglutide
`C18 diacid DγGlu-2xOEG
`Lys26
`Aib8, Arg34
`22
`7.1 (0.4)
`230
`30.0 (3.0)
`0.13 (0.01)
`γGlu-2xOEG
`23b
`Lys26
`Aib8, Arg34
`5.6 (1.4)
`16
`12.9 (6.8)
`0.80 (0.36)
`C18 diacid
`γGlu-3xOEG
`Lys26
`Aib8, Arg34
`24
`27.7 (2.1)
`8.6
`6.17 (0.13)
`0.71 (0.04)
`C18 diacid
`γGlu-dPEG8
`Lys26
`Aib8, Arg34
`25
`47.3 (8.7)
`2.1
`11.4 (1.6)
`5.31 (0.92)
`C18 diacid
`2xγGlu-2xOEG
`Lys26
`Aib8, Arg34
`26
`27.8 (4.0)
`77
`19.5 (8.1)
`0.25 (0.04)
`C18 diacid
`3xγGlu-2xOEG
`Lys26
`Aib8, Arg34
`27
`67.0 (21.1)
`43
`15.2 (0.8)
`0.36 (0.04)
`C18 diacid
`Abu-2xγGlu-OEG
`Lys26
`Aib8, Arg34
`28
`70.3 (7.1)
`9.5
`16.8 (2.5)
`1.77 (0.29)
`C18 diacid
`Lys26
`Aib8, Arg34
`29
`21.2 (2.1)
`44
`8.80 (4.86)
`0.20 (0.08)
`C18 diacid
`Abu-2xOEG
`Abu−γGlu-OEG
`Lys26
`Aib8, Arg34
`30
`10.7 (1.6)
`36
`20.7 (3.3)
`0.58 (0.11)
`C18 diacid
`Benzyl-βAla-2xOEG
`Lys26
`Aib8, Arg34
`31
`5990 (1040)
`5.6
`51.7 (23.0)
`9.19 (1.73)
`C18 diacid
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`32
`11.5 (1.3)
`85
`8.43 (2.22)
`0.10 (0.02)
`C20 diacid
`γGlu-2xOEG
`Lys26
`Aib8, Arg34
`33
`24.4 (2.9)
`116
`24.8 (2.2)
`0.21 (0.01)
`C22 diacid
`aGLP-1 receptor binding at 0% and 2% albumin (HSA) using BHK cells expressing the human GLP-1R and in vitro potency measured in BHK cells
`that express both the human GLP-1R and a luciferase reporter system. bAnalogue 23 has a C-terminal amide.
`
`linker would impact the albumin affinity and PK in a similar
`manner as earlier reported for monoacid derivatives.8,10,24 After
`the positive results with the C16 diacids, a campaign was
`designed using C18 diacid combined with a larger set of linkers
`and varying the acylation position.
`A series of analogues with the Aib8, Arg34 GLP-1 (7−37)
`backbone was derivatized at Lys26 with C18 diacids via various
`linkers (Table 3). The in vitro potency varied from 5990 pM for
`31 with the bulky benzyl-containing linker (Benzyl-βAla) to
`4−6 pM with the most promising linkers. It was still not clear
`how the structural features of the linker affects the potency, but
`it was observed that simple linkers composed of OEG and/or
`γGlu gave the highest potencies with EC50 values below 10 pM,
`as seen for semaglutide, 20, 21, and 22. The receptor binding
`was also affected by the choice of linker with a span of 70-fold
`from around 9 nM for 31 to 0.1−0.2 nM for 21 and 22 in the
`absence of albumin. When 2% albumin was present in the assay
`the right-shift of the binding dose−response (BR ratio) varied a
`lot, from around two-fold for 25 up to 940-fold for semaglutide,
`due to different binding affinities for albumin.
`There was a dramatic effect on the BR ratio by changing
`from C16 diacid in 18 to C18 diacid in semaglutide, which was
`hypothesized to be due to an increased albumin affinity. An
`attempt was made to increase this BR ratio further by ex-
`panding the fatty acid length to C20 diacid in 32 and C22
`diacid in 33 but the BR ratio actually dropped due to improved
`receptor affinity at 2% HSA for these analogues compared to
`semaglutide. One surprising observation of the series of C12
`diacid to C22 diacid (15, 16, 18, semaglutide, 32, and 33) is
`that the in vitro potency is gradually improved from C12 diacid
`in 15 (42.5 pM) to C18 diacid in semaglutide (6.2 pM) but is
`again attenuated for 32 with C20 diacid (11.5 pM) and for 33
`with C22 diacid (24.4 pM), thus concluding that C18 diacids is
`the optimal choice with respect to in vitro potency.
`An earlier observation was that it seemed possible to acylate
`GLP-1 in several positions without losing receptor affinity,8,10
`and also in this study we decided to investigate how the
`
`the in vitro properties.
`position of acylation could affect
`A relatively complex approach was tried where we permutated
`not only the linker but also the attachment point and a few
`mutations of the peptide backbone. In order to be able to draw
`structural conclusions,
`the C18 diacid was kept constant
`throughout (Table 4). Among this set of GLP-1 derivatives,
`four peptides had EC50 potency values below 10 nM (35, 36,
`38, and 43) combined with high receptor affinity below 0.5 nM.
`However, the BR ratio was relatively low, and only three
`derivatives had a BR ratio above 100 (34, 38 and 43).
`Pharmacokinetic studies are more resource demanding,
`hence, pharmacokinetics were only investigated for a limited
`number of the analogues. The in vivo extension of half-life has
`previously been shown to be greatly impacted by the albumin
`binding moiety for a series of non-DPP-4 stabilized analogues
`(including liraglutide).8 In the present work, we hypothesized
`that for DPP-4 stabilized analogues the increase in albumin
`binder lipophilicity would also increase the exposure (increase
`systemic half-life). This was evaluated using a series of ana-
`logues with the Aib8, Arg34 GLP-1 (7−37) backbone deriv-
`atized at Lys26 linked to various chain length (C12 to C20)
`diacids using a γGlu-2xOEG linker (Figure 3).
`As depicted in Figure 3 the trend was quite clear; there is an
`increased exposure in vivo with increased chain length and
`relative lipophilicity of the albumin binder going from C12 to
`C14, C16, C18, and C20 diacid moiety. However, backbone
`substitution and modifying the linkers gave rise to a much less
`clear picture and unpredictable impact on the pharmacokinetic
`properties. The semaglutide analogue exhibited the highest
`BR ratio (940), but there was no apparent correlation between
`the indirect measure of albumin affinity (BR ratio) and the
`observed protraction (terminal systemic half-life) across the
`analogues tested. This is exemplified by the analogue 32,
`which had a BR ratio of 85 but a protraction quite similar to
`semaglutide.
`Native GLP-1 is rapidly metabolized by DPP-4, and liraglutide
`also has been shown to be metabolized to a minor degree by
`
`7374
`
`DOI: 10.1021/acs.jmedchem.5b00726
`J. Med. Chem. 2015, 58, 7370−7380
`
`Novo Nordisk Exhibit 2002
`Mylan Pharms. Inc.v. Novo Nordisk A/S
`IPR2023-00722
`Page 00005
`
`
`
`Journal of Medicinal Chemistry
`Article
`Table 4. Derivatives of GLP-1 (7−37) Peptides Attached with C18 Diacids via γGlu and OEG Containing Linkers and to 16, 22,
`25, 26, 27, 36, 37, and 38 Positions of the Peptidesa
`
`GLP-1R binding
`(IC50, nM) (SEM)
`
`GLP-1R potency (EC50, pM)
`ratio 2%/0%
`acylation
`(SEM) 0% HSA
`HSA
`2% HSA
`0% HSA
`position
`linkage
`protractor
`sequence modification
`analogue
`Lys36
`26Arg26, 34Arg34,Lys36
`34
`7.0 (0.3)
`127
`14.5 (0.8)
`0.11 (0.01)
`2xOEG
`C18 diacid
`γGlu-2xOEG
`Lys27
`Aib8, Lys27
`35
`4.6 (0.4)
`98
`5.39 (1.39)
`0.06 (0.01)
`C18 diacid
`γGlu-2xOEG
`Lys16
`Aib8, Lys16
`36
`9.9 (1.4)
`51
`25.5 (4.9)
`0.50 (0.13)
`C18 diacid
`Lys37
`Aib8, Aib22, Aib35, Lys37 C18 diacid OEG
`37
`40.9 (0.4)
`25
`19.3 (0.7)
`0.77 (0.06)
`γGlu-2xOEG
`Lys22
`Aib8, Lys22
`38
`2.7 (0.1)
`280
`17.7 (2.8)
`0.06 (0.01)
`C18 diacid
`γGlu-2xOEG
`Lys25
`Aib8, Lys25
`39
`41.8 (4.1)
`37
`41.6 (10.5)
`1.13 (0.29)
`C18 diacid
`Lys36
`Aib8, Arg26, Arg34, Lys36 C18 diacid
`40
`159 (21)
`45
`23.0 (2.4)
`0.52 (0.04)
`none
`Lys38
`Aib8, Arg26, Arg34, Lys38 C18 diacid
`41
`87.3 (15.4)
`24
`7.55 (1.44)
`0.31 (0.13)
`none
`Lys38
`Aib8, Arg26, Arg34, Lys38 C18 diacid
`42
`13.2 (1.4)
`36
`3.99 (0.58)
`0.11 (0.01)
`2xOEG
`γGlu-2xOEG
`Lys38
`Aib8, Arg26, Arg34, Lys38 C18 diacid
`43
`1.5 (0.1)
`334
`30.1 (1.1)
`0.09 (0.01)
`Lys38
`Aib8, Arg34, Lys38
`44
`67.5 (0.9)
`25
`6.73 (3.65)
`0.27 (0.05)
`C18 diacid
`2xOEG
`Lys26
`Gly 8, Arg34
`45
`26.0 (0.5)
`64
`6.5 (1.5)
`0.10 (0.03)
`C18 diacid
`2xOEG
`aGLP-1 receptor binding at 0% and 2% albumin (HSA) using BHK cells expressing the human GLP-1R and in vitro potency measured in BHK cells
`that express both the human GLP-1R and a luciferase reporter system.
`
`(data not shown). In fact, 6 has quite similar in vivo protrac-
`tion as liraglutide indicating the importance of not only the
`stronger albumin binding of semaglutide compared to liraglutide
`but potentially also the extra enzymatic stabilization in order to
`obtain long systemic half-life.
`the risk of
`to limit
`As mentioned earlier,
`in order
`immunogenicity responses in patients, we aimed to select a
`peptide structurally similar to liraglutide and native GLP-1.
`Thus, acylation in position 26 and as few mutations as possible
`were prioritized features. Among the possible candidates, it was
`quite encouraging that semaglutide fulfilled these structural
`priorities and had an attractive pharmacokinetic profile com-
`bined with acceptable receptor potency.
`In Vivo Characterization Studies. The in vivo character-
`ization of the analogues in the screening program included
`evaluation of the pharmacokinetic properties in pigs (Göttingen
`mini-pigs) and in vivo efficacy in db/db mice.
`In Table 5 the pharmacokinetic parameters of liraglutide
`and semaglutide are presented. The volume of distribution of
`
`Table 5. Pharmacokinetic Evaluation in Göttingen Mini-Pigs
`Following Administration of Semaglutide (2 nmol/kg i.v.
`or 2 nmol/kg s.c.) and liraglutide (0.5 nmol/kg i.v. or
`1.0 nmol/kg s.c.)a
`
`i.v. administration
`
`s.c. administration
`
`Cl
`T1/2
`Vz
`MRT
`Tmax
`(L/kg)
`(L/h/kg)
`F (%)
`(h)
`(h)
`(h)
`94%
`63.6
`12
`46.1
`0.1019
`0.0016
`semaglutide
`66%
`23.0
`7
`12.4
`0.0674
`0.0038
`liraglutide
`aThe key pharmacokinetic parameters (Cl, Vz, Tmax, T1/2, MRT and F)
`of semaglutide and liraglutide as reference in Göttingen mini-pigs.
`
`liraglutide after i.v. administration was 0.067 L/kg (67 mL/kg).
`This is very close to the blood volume in mini-pigs (65 mL/kg)
`indicating either a limited distribution outside the circulation
`or a fast equilibrium of the liraglutide concentration between
`the circulation and the peripheral tissue. Body clearance of
`liraglutide after i.v. administration was estimated to 0.0038 L/h/kg
`(0.063 mL/min/kg), and together with the observed volume of
`distribution it leads to a half-life of 12.4 h after i.v. administration.
`
`7375
`
`DOI: 10.1021/acs.jmedchem.5b00726
`J. Med. Chem. 2015, 58, 7370−7380
`
`Figure 3. In vivo protraction in rats following i.v. administration of
`32 (5.5 nmol/kg), semaglutide (4.2 nmol/kg), 19 (3.3 nmol/kg), 16
`(5.5 nmol/kg), and 15 (5.3 nmol/kg).
`
`DPP-4 in the Ala8-Glu9 position of the N terminus.32,33 The
`impact on in vivo protraction of the lack of DPP-4 protection is
`illustrated in Figure 4.
`Figure 4 depicts that semaglutide with Aib8 has a longer
`systemic half-life compared to the analogue 6 with Ala8 and
`
`Figure 4. In vivo protraction in rats following i.v. administration of 6
`(5.8 nmol/kg), liraglutide (5.8 nmol/kg), and semaglutide (4.2 nmol/kg).
`Compound 6 is the Ala8 version of semaglutide and susceptible to
`enzymatic degradation.
`
`linker, and protractor
`otherwise similar backbone structure,
`moiety. The susceptibility to enzymatic degradation in vivo
`of 6 was verified by identification of circulating metabolites
`
`Novo Nordisk Exhibit 2002
`Mylan Pharms. Inc.v. Novo Nordisk A/S
`IPR2023-00722
`Page 00006
`
`
`
`Journal of Medicinal Chemistry
`
`The volume of distribution of semaglutide was 0.102 L/kg
`(102 mL/kg). This is 1.5-fold higher than for liraglutide, indi-
`cating lower concentrations of semaglutide measured in blood,
`which is consistent with a higher degree of albumin binding.
`The body clearance after i.v. administration was also two-fold
`lower for semaglutide compared to liraglutide, suggesting that
`the albumin binding might reduce the clearance. This leads to a
`half-life of semaglutide after i.v. administration of 46.1 h, which is
`more than three-fold longer than observed for liraglutide. The
`observed mean residence time (MRT) following s.c. admin-
`istration was approximately three-fold longer for semaglutide
`compared to liraglutide. An almost complete subcutaneous
`bioavailability was observed for semaglutide (94%), whereas the
`subcutaneous bioavailability for liraglutide was 66%. It was
`speculated that an increased albumin binding and strong protec-
`tion toward DPP-4 degradation could facilitate increased bio-
`availability, but further studies have to be made to conclude.
`The db/db mouse, a hyperglycaemic, hyperinsulinaemic, and
`obese model of type 2 diabetes, was used during the initial
`in vivo screening studies of GLP-1 analogues to investigate the
`antihyperglycaemic and body weight lowering efficacy. A dose−
`response study (0.3 to 100 nmol/kg) showed that semaglutide
`had a dose-dependent efficacy and duration of action of up to
`48 h (Figure 5A). A similar study with liraglutide (Figure 5B)
`
`Figure 5. Dose response for blood glucose lowering of semaglutide
`(A) and liraglutide (B) in diabetic db/db mice after s.c. dosing of
`0.3 to 100 nmol/kg.
`
`clearly showed a shorter duration of action and less potency for
`liraglutide compared to semaglutide. Among the tested GLP-1
`analogues, semaglutide was the most potent with an EC50 of
`<2 nmol/kg (calculated as the delta AUC blood glucose during
`0−48 h after dosing).
`HSA Affinity Assessed by Analytical Ultracentrifuga-
`tion. We did not obtain measurements of the absolute binding
`constant between the ligands and albumin. Instead, the albumin
`binding of a subset of analogues was evaluated in vitro by
`analytical ultracentrifugation34 to obtain measures of relative
`albumin affinities. Figure 6 depicts the fraction of bound
`analogue relative to that of liraglutide. As expected the native
`
`Article
`
`Figure 6. Relative binding affinities of GLP-1 analogues to HSA
`measured by analytical ultracentrifugation. Fraction of analogue bound
`to HSA relative to liraglutide assessed by analytical ultracentrifugation.
`The relative fractions of bound vs free analogue were determined in
`solutions of 100 μM analogue and 10 μM HSA in PBS-buffer, pH 7.4.
`
`peptide displayed negligible binding to albumin. Among those
`analogues that are acylated in position 26 there is a clear trend
`that with increasing length of the carbon chain (18, C16 diacid;
`semaglutide, C18 diacid; and 32, C20 diacid), the analogues
`display a successively increased affinity to albumin binding rela-
`tive to liraglutide (4.3, 5.6, and 5.8 times, respectively). This
`trend relates to the longer plasma half-life observed with
`increasing length of the fatty diacids and correlates much better
`than the BR ratio. However, this correlation does not hold for
`other modifications. As an example, the absence of the Aib8 in
`analogue 6 appeared favorable for HSA binding to the same
`extent as for semaglutide, but unfortunately, this analogue has a
`much shorter half-life in rats compared to semaglutide (Figure 6).
`It could be speculated that the