`
`Discovery, characterization, and clinical development
`of the glucagon-like peptides
`
`Daniel J. Drucker,1 Joel F. Habener,2 and Jens Juul Holst3
`1Lunenfeld-Tanenbaum Research Institute, Mt. Sinai Hospital, University of Toronto, Toronto, Ontario, Canada. 2Laboratory of Molecular Endocrinology, Massachusetts General Hospital, Harvard University,
`Boston, Massachusetts, USA. 3Novo Nordisk Foundation Center for Basic Metabolic Research, Department of Biomedical Sciences, University of Copenhagen, Copenhagen, Denmark.
`
`The discovery, characterization, and clinical development of glucagon-like-
`peptide-1 (GLP-1) spans more than 30 years and includes contributions from
`multiple investigators, science recognized by the 2017 Harrington Award
`Prize for Innovation in Medicine. Herein, we provide perspectives on the
`historical events and key experimental findings establishing the biology
`of GLP-1 as an insulin-stimulating glucoregulatory hormone. Important
`attributes of GLP-1 action and enteroendocrine science are reviewed, with
`emphasis on mechanistic advances and clinical proof-of-concept studies.
`The discovery that GLP-2 promotes mucosal growth in the intestine is
`described, and key findings from both preclinical studies and the GLP-2
`clinical development program for short bowel syndrome (SBS) are reviewed.
`Finally, we summarize recent progress in GLP biology, highlighting emerging
`concepts and scientific insights with translational relevance.
`
`The endocrine activity of the gastroin-
`
`testinal tract has been studied for more
`than a century, with gut hormones such as
`secretin emerging from the seminal stud-
`ies of Bayliss and Starling (1). The concept
`that the gut also controlled pancreatic
`islet secretions was supported by experi-
`ments demonstrating that administration
`of crude gut extracts lowered blood glu-
`cose in animals. The development of the
`insulin radioimmunoassay enabled the
`description of the incretin effect, namely
`that glucose administered into the gut
`potentiated insulin secretion to a greater
`extent than isoglycemic stimulation of
`insulin secretion achieved through i.v.
`glucose administration. The first incretin
`hormone, glucose-dependent insulinotro-
`pic polypeptide (GIP), was isolated by John
`Brown in the 1970s (2). Here, we describe
`the discovery and characterization of the
`
`second incretin hormone, GLP-1. We high-
`light 3 decades of science from multiple
`laboratories supporting the development
`of GLP-1– and GLP-2–based therapies.
`GLP-1 is now used for the treatment of
`type 2 diabetes (T2D) and obesity, whereas
`GLP-2 is used for the therapy of short bow-
`el syndrome (SBS).
`
`Discovery of GLP-1
`Although GIP was isolated through classical
`peptide purification and protein sequenc-
`ing methodology, the discovery of the
`GLP-1 sequence stemmed from the appli-
`cation of recombinant DNA approaches
`developed in the laboratories of Stanley
`Cohen, Paul Berg, and Herb Boyer in the
`early 1970s. This remarkable new techn-
`ology allowed for a rapid and accurate
`prediction of the amino acid sequences of
`proteins by the decoding of the nucleotide
`
`Conflict of interest: D.J. Drucker is supported by the Canada Research Chairs program and a BBDC-Novo Nordisk
`Chair in incretin biology, and has received speaking or consulting honorarium from Arisaph, Intarcia, Merck, Pfizer,
`and Novo Nordisk Inc. Mt. Sinai Hospital receives research support from GSK, Merck, and Novo Nordisk, for preclin-
`ical studies in the Drucker laboratory. J.J. Holst has served as a consultant or advisor to Novartis Pharmaceuticals,
`Novo Nordisk, Merck Sharp & Dohme, and Roche and has received fees for lectures from Novo Nordisk, Merck Sharp
`& Dohme, and GlaxoSmithKline.
`Reference information: J Clin Invest. 2017;127(12):4217–4227. https://doi.org/10.1172/JCI97233
`
`sequences of cloned recombinant cDNA
`copies of messenger RNAs. The Habener
`lab utilized this technology to elucidate
`proglucagon amino acid sequences from
`cDNAs and genes isolated from angler-
`fish in the early 1980s (3–5) and the rat
`proglucagon cDNA and gene sequences
`followed shortly thereafter (refs. 6, 7, Fig-
`ure 1, and Figure 2). Corresponding pro-
`glucagon sequences from hamster, bovine,
`and human were identified by Graeme
`Bell and others in the early 1980s (8–10).
`These sequences revealed that glucagon
`and re lated GLP sequences were encoded
`by larger protein precursors, termed pre-
`prohormones (Figure 2). The anglerfish
`preproglucagons (Figure 2A), isolated and
`characterized by Lund and Goodman (3–5),
`were interesting as there were two different
`cDNAs encoded by separate (nonallelic)
`genes and they each contained a glucagon-
`related sequence, in addition to glucagon.
`The two anglerfish glucagon–related pep-
`tides resembled GIP, a glucoincretin hor-
`mone released from the gut into the circu-
`lation during meals, subsequently shown
`by Dupre and Brown in 1973 to augment
`glucose-dependent insulin secretion (11).
`Unlike the two anglerfish preproglucagons,
`each of which harbored glucagon and a sin-
`gle glucagon-related peptide, the mammali-
`an preproglucagons all contained glucagon
`and two additional glucagon-related pep-
`tides, designated GLP-1 and GLP-2 (Figure
`2B). Notably, the corresponding amino acid
`sequences of the GLP-1s in the four mam-
`malian species were identical (12), with
`conservation of sequence implying as yet
`unknown but potentially important biolog-
`ical actions of GLP-1.
`Collectively, these findings further
`supported the evolving notion at the
`time that small peptide hormones are
`synthesized in the form of larger prohor-
`mones and that the final bioactive pep-
`tides are formed posttranslationally by
`selective enzymatic cleavages from the
`prohormones (Figure 2B). Earlier studies
`
`jci.org Volume 127 Number 12 December 2017
`
`4 2 1 7
`
`H A R R I N G T O N P R I Z E E S S A Y
`
`MPI EXHIBIT 1018 PAGE 1
`
`
`
`The Journal of Clinical Investigation
`
`Figure 1. Historical timeline depicting seminal events in the discovery and development of gluca-
`gon-like peptide therapeutics.
`
`had demonstrated that insulin and para-
`thyroid hormone are synthesized as pre-
`prohormones. The hydrophobic amino
`terminal sequences, termed the pre- or
`signal sequences, are cleaved from the
`nascent polypeptide chains during their
`translation on ribosomes, leaving the
`prohormones as the precursor of the
`peptide hormones. The A and B chains
`of insulin and parathyroid hormone were
`found to be cleaved from their respective
`prohormones by the actions of specific
`endopeptidases, prohormone conver-
`tases, which cleave proteins at the sites
`of two consecutive basic amino acids,
`arginine and lysine (13). These earlier
`findings of prohormone convertase spec-
`ificity were applied to deduce potential
`cleavage sites in proglucagon.
`
`Structure-activity properties
`of GLP-1
`Examination of the amino acid sequence
`of proglucagon
`initially presented a
`conundrum regarding the processes by
`which potentially bioactive GLP-1 pep-
`tides might be liberated from the pro-
`hormone (Figure 2B). In keeping with
`the rule that bioactive peptides are
`cleaved from prohormones at sites con-
`sisting of two basic amino acids (13),
`several investigators initially predicted
`that bioactive GLP-1 would be a peptide
`of 37 amino acids beginning with histi-
`dine and ending in glycine, GLP-1(1-37).
`However, further inspection of the pro-
`hormone sequences revealed a second
`single basic amino acid followed by his-
`tidine residing 6 amino acids carboxyl-
`proximal to the first histidine, predict-
`ing a GLP-1 peptide of 31 amino acids,
`
`GLP-1(7-37). Further, at the carboxyl-
`terminal region of the putative GLP-1 pep-
`tide resides a sequence RGRR predicting
`a prohormone convertase–directed cleav-
`age site followed by an amidation of the
`penultimate arginine by a peptidylglycine
`α-amidating monooxygenase (14) result-
`ing in peptides of 36 and 30 amino acids,
`GLP-1(1-36)amide and GLP-1(7-36)amide.
`The availability of the amino acid
`sequences of GLP-1 and GLP-2 obtained
`from the decoding of the nucleotide
`sequences based on the genetic code
`allowed for the chemical synthesis of the
`predicted peptides, examination of their
`biological activities, and preparation of
`peptide-specific antisera. Daniel Drucker
`joined the Habener laboratory in the sum-
`mer of 1984, with the original intent of
`studying the molecular control of thyroid
`hormone biosynthesis. However, the thy-
`roid group, led by Bill Chin, was decamp-
`ing for the Brigham and Women’s Hospi-
`tal, and Daniel was assigned to work on the
`proglucagon gene. To understand whether
`proglucagon might be processed to yield
`multiple proglucagon-derived peptides
`(PGDPs), including GLP-1 and GLP-2,
`Drucker transfected a proglucagon cDNA
`expression vector into fibroblasts, pituitary
`cells, and islet cells (15). Although minimal
`processing was observed in fibroblasts,
`immunoreactive GLP-1 and GLP-2 was
`detected by chromatography and radio-
`immunoassays in medium and extracts
`from transfected pituitary and islet cells
`(15). Application of similar chromatogra-
`phy and radioimmunoassay techniques
`in studies of rat (16), pig (17), and human
`(18) tissues revealed distinct profiles of
`PGDPs in pancreas and gut (Figure 3). Of
`
`note, in addition to glucagon, the pancreas
`contained a large peptide with immuno-
`reactive determinants for both GLP-1 and
`GLP-2 (but not glucagon), consistent with
`incompletely cleaved proglucagon (major
`proglucagon fragment [MPGF]) (Figure 3,
`A and C, and ref. 19). In contrast to find-
`ings in pancreas, immunoreactive GLP-1
`peptides detected in gut extracts consisted
`entirely of smaller peptides (refs. 15 and
`16, and Figure 3, B and C). These findings
`indicated that the cleavage of proglucagon
`into small GLP-1–immunoreactive pep-
`tides was more efficient in the gut com-
`pared with pancreas. These observations
`were also consistent with the incretin con-
`cept in which, in response to oral nutrients,
`glucoincretin hormones such as GIP (and
`subsequently GLP-1) originate from the
`gut and not the pancreas.
`Svetlana Mojsov in the Habener lab-
`oratory detected glycine-extended and
`arginine-amidated isoforms of GLP-1 — as
`well as both the amino-terminally extended
`peptides GLP-1(1-37) and GLP-1(1-36)amide
`and the amino-truncated peptides GLP-1
`(7-37) and GLP-1(7-36)amide — in extracts
`of pancreas (Figure 3A). Nevertheless, the
`abundance of these peptides was much
`greater in the gut (20).
`
`Bioactivities of the GLP-1
`peptides
`The earliest studies of the bioactivities of
`GLP-1(1-37) were indecisive. One study
`found that the 37 amino acid peptide
`activated adenylyl cyclase in membranes
`prepared from rat pituitary and hypothal-
`amus (21), whereas another study failed to
`detect any effects of the peptide on glucose
`and insulin in cortisone-treated rabbits
`
`4 2 1 8
`
`jci.org Volume 127 Number 12 December 2017
`
`H A R R I N G T O N P R I Z E E S S A Y
`
`MPI EXHIBIT 1018 PAGE 2
`
`
`
`The Journal of Clinical Investigation
`
`bioactivities for the extended forms, GLP-1
`(1-37) and GLP-1(1-36)amide, have yet been
`determined. Furthermore, no distinctive bio-
`logical activities have been attributable spe-
`cifically to the amidated forms of GLP-1.
`
`The view from Denmark
`In Copenhagen, Jens Holst and colleagues
`were interested in the incretin effect and
`were studying the condition of postpran-
`dial reactive hypoglycemia after gastric
`surgery (27). This type of hypoglycemia
`was clearly hyperinsulinemic, yet the sig-
`nal for insulin secretion was unknown.
`Looking for possible candidates, they were
`inspired by Lise Heding’s work on gluca-
`gon and her identification of the immu-
`nological differences between gut and
`pancreatic glucagon (28). Knowing that
`glucagon would stimulate insulin secre-
`tion, they were interested in the numer-
`ous cells in the gut that produce immu-
`noreactive glucagon (29). Eventually, this
`work led to the identification of glicentin
`and oxyntomodulin (Figure 2B), which
`both contain the full glucagon sequence,
`explaining the immunoreactivity in the
`gut (30–32).
`Having identified all of the molecular
`components of glicentin also in the pan-
`creas (33), they proposed that glicentin
`represents at least part of a common gut
`and pancreatic glucagon precursor, which
`undergoes differential processing in the
`two tissues, a hypothesis subsequently con-
`firmed through identification of the human
`proglucagon gene by Graeme Bell and col-
`leagues (7). However, it was also clear that
`proglucagon was larger than glicentin, and
`the interest focused on peptides contained
`within the MPGF representing the remain-
`der of proglucagon (minus glicentin) (34).
`The early work decoding the anglerfish
`proglucagon cDNA by Lund and Habener
`(3) followed by elucidation of the ham-
`ster proglucagon cDNA by Graeme Bell
`(9) supported a hypothesis that cleavage
`of MPGF might result in liberation of the
`GLPs. The Holst group quickly developed
`radioimmunoassays for GLP-1 and GLP-2
`to test this hypothesis. To their excitement,
`they found that MPGF was indeed differ-
`entially processed in the gut, but not in the
`pancreas, to yield two GLPs (ref. 17 and
`Figure 3C). However, using the perfused
`pancreas preparation, they soon realized
`that neither of the two GLPs used in these
`
`Figure 2. Structure and processing of anglerfish and human proglucagon. Representation of the
`structures of proglucagon cDNAs from Anglerfish (A) and Human (B), with tissue-specific liberation of
`individual proglucagon-derived peptides in pancreas or intestine. PC1, prohormone convertase 1; PC2,
`prohormone convertase 2; MPGF, major proglucagon fragment. Arrows in A represent sites of cleave by
`prohormone convertase enzymes.
`
`(22). To determine whether the shorter
`GLP-1 isoforms were bioactive, Drucker
`incubated truncated forms of GLP-1 with
`multiple cell lines, examining cell growth,
`gene expression, and signal transduc-
`tion. Remarkably, insulin-producing islet
`cells responded to N-terminally truncated
`forms of GLP-1, exhibiting increased lev-
`els of cyclic AMP accumulation within
`minutes of exposure to 50 pM–5 nM GLP-
`1(7-37) (Figure 4A). In contrast, neither
`glucagon, GLP-1(1-37), or GLP-1(1-36)–
`NH2 increased cAMP accumulation at
`these concentrations (23). GLP-1(7-37)
`also increased the levels of insulin mRNA
`transcripts and stimulated insulin secre-
`tion at 25 mM but not 5.5 mM glucose in
`studies with isolated islet cell lines (Figure
`4, B and C, and ref. 23). In contrast, nei-
`ther glucagon nor GLP-2 induced insulin
`gene expression. Furthermore, GLP-1
`(7-37) had no effect on levels of angioten-
`sin gene expression in the insulinoma cell
`line and did not change levels of glucagon,
`prolactin, and corticotropin mRNAs in islet
`or pituitary cell lines (23). Hence, these
`findings (Figure 4, A–C) published in 1987
`(23) established that GLP-1(7-37) directly
`augments glucose-dependent insulin bio-
`synthesis and secretion from β cells.
`
`The demonstration that GLP-1 directly
`increased cAMP levels provided condi-
`tional evidence for the existence of a Gs
`protein-coupled receptor in β cells. In
`studies of insulin secretion using the iso-
`lated perfused pancreas, Mojsov and Weir
`demonstrated that GLP-1(7-37) and not
`GLP-1(1-37) stimulated insulin secretion
`at concentrations as low as 50 pM (ref. 24
`and Figure 5, A–C). Likewise, as outlined
`below, Jens Holst and colleagues showed
`that luminal glucose stimulated GLP-1
`secretion from the perfused intestine (Fig-
`ure 5D), and doses from 500 pM to 5 nM
`GLP-1(7-36)amide stimulated insulin secre-
`tion in the perfused pig pancreas (Figure
`5E and ref. 25). Thus, it turned out that a
`cleavage in proglucagon at the single basic
`amino acid, arginine, and not the double
`basic amino acids, generated the active
`GLP-1 peptides, GLP-1(7-37) and GLP-1
`(7-36)amide (Figure 5C). First in man studies
`reported in December 1987 by Kreymann
`and Bloom (26) rapidly established the
`insulinotropic actions of GLP-1(7-36)amide
`in human subjects. Although numerous
`studies have demonstrated that the amino-
`terminally truncated forms of GLP-1, GLP-
`1(7-37), and GLP-1(7-36)amide are active
`glucoregulatory hormones, no compelling
`
`jci.org Volume 127 Number 12 December 2017
`
`4 2 1 9
`
`H A R R I N G T O N P R I Z E E S S A Y
`
`MPI EXHIBIT 1018 PAGE 3
`
`
`
`The Journal of Clinical Investigation
`
`Mentlein in Kiel, Holst and Deacon showed
`that the GLP-1 molecule was cleaved by the
`enzyme
`dipeptidyl-peptidase-4(DPP-4)
`in vivo and that inhibitors of this enzyme
`could completely protect the molecule
`(43). In fact, the circulating half-life of
`GLP-1 was only 1.5–2 minutes in human
`subjects with diabetes, and they proposed
`that inhibitors of DPP-4 could maintain
`higher levels of intact active endogenous
`GLP-1 for therapeutic purposes (44). Sub-
`sequent studies soon demonstrated that
`DPP-4–resistant GLP-1 analogues were
`longer-acting than native GLP-1 (45). Fur-
`thermore, inhibitors of DPP-4 completely
`prevented the breakdown of GLP-1 in the
`circulation and amplified the insulinotro-
`pic actions of GLP-1 (46). This exciting
`development, presented in a Perspectives
`article in Diabetes in 1998 (47), was soon
`followed by the development of clinically
`useful inhibitors, first vildagliptin and sub-
`sequently sitagliptin.
`It remained to be understood whether
`GLP-1 receptor agonists would actually
`be useful for clinical diabetes therapy or
`whether tachyphylaxis would develop
`upon chronic administration. The group
`in Copenhagen administered synthetic
`GLP-1 by constant s.c. infusion for 6
`weeks to a group of individuals with
`long-standing T2D (48). Fortunately, no
`tachyphylaxis was observed; GLP-1 ther-
`apy reduced fasting and mean plasma
`glucose by 4.3 and 5.5 mmol/l; glycated
`hemoglobin by 1.3 %; and body weight by
`2 kg. Moreover, insulin sensitivity and β
`cell function, assessed by clamp studies,
`greatly improved. Importantly, no limit-
`ing side effects were re corded (48), pro-
`viding proof of concept in 2002 for GLP-1
`therapy in subjects with T2D. It was now
`clear that GLP-1–based therapies had tre-
`mendous potential.
`
`The Toronto perspective:
`the physiology of GLPs and
`discovery of GLP-2
`Building on the availability of cloned pro-
`glucagon gene sequences in the Habener
`lab (6), Daniel Drucker and Jacques
`Philippe pursued the analysis of the molec-
`ular control of islet α cell proglucagon
`gene expression in the mid 1980s (49, 50).
`Upon returning to Toronto in 1987, Druck-
`er extended these studies to examine pro-
`glucagon gene expression in the intestine
`
`Figure 3. Processing of proglucagon and glucagon-like peptides in the pancreas and intestine.
`Detection of immunoreactive forms of GLP-1 in extracts from rat pancreas (A) and intestine (B) as
`adapted from ref. 16. Characterization of proglucagon-derived peptide immunoreactivity secreted
`from perfused pig pancreas and intestine (C) using peptide-specific antisera reveal tissue-specific
`posttranslational processing of the PGDPs, as outlined in ref. 17.
`
`studies had any effect on pancreatic hor-
`mone secretion. They therefore decided
`to isolate the naturally occurring hormone
`from porcine and human and gut extracts,
`and they found that the naturally occur-
`ring peptide was a truncated from of GLP-1
`representing proglucagon
`(aa 78-108)
`(Figure 5C) and subsequently found to be
`amidated, corresponding to proglucagon
`78-107amide (35). Importantly, this hormone
`was potently insulinotropic (Figure 5E and
`ref. 25), so they had also described a new
`incretin hormone in 1987; however, they
`wondered whether GLP-1 was more inter-
`esting than the already known incretin
`GIP, which exhibited a diminished effect
`on insulin secretion in patients with T2D
`(36). They soon found, using the perfused
`pancreas, that GLP-1 — in contrast to GIP —
`also powerfully inhibits glucagon secretion
`(37). Eventually, they demonstrated that
`during infusions of physiological amounts
`of GLP-1 into humans, insulin secretion
`would be stimulated and glucagon secre-
`tion inhibited, resulting in a decrease in
`hepatic glucose production (38). However,
`the effect was self-limiting, with the insulin-
`
`stimulating activity attenuated as plasma
`glucose levels started to fall, limiting the
`fall to 0.5–1 mmol/l.
`At that time, the Copenhagen group
`realized that GLP-1 was extremely interest-
`ing and, in further studies, demonstrated
`that it strongly inhibited gastric motility
`and gastric and pancreatic exocrine secre-
`tion (39), consistent with an important
`role for this hormone as a regulator of
`upper gastrointestinal function. They also
`demonstrated that infusions of GLP-1 in
`humans inhibited appetite and food intake,
`actions subsequently exploited in the clinic
`to treat obesity (40). In studies published
`in 1993 by Michael Nauck and colleagues
`in Göttingen, i.v. infusion of GLP-1 com-
`pletely normalized severely elevated fast-
`ing glucose concentrations in patients with
`long-standing T2D as a consequence of the
`actions of GLP-1 to stimulate insulin and
`inhibit glucagon secretion (41). Although
`GLP-1 clearly had therapeutic potential,
`s.c. injections of GLP-1 were disappoint-
`ingly ineffective (42). The explanation was
`an extremely rapid metabolism and inacti-
`vation of GLP-1. With inspiration from Rolf
`
`4 2 2 0
`
`jci.org Volume 127 Number 12 December 2017
`
`H A R R I N G T O N P R I Z E E S S A Y
`
`MPI EXHIBIT 1018 PAGE 4
`
`
`
`The Journal of Clinical Investigation
`
`Figure 4. GLP-1 directly stimulates signal transduction and glucose-dependent insulin secretion in islet cells. GLP-1(7-37) directly stimulates cyclic AMP
`accumulation (A), insulin gene expression (B), and glucose-dependent insulin secretion (C) in a rat insulinoma cell line, as adapted from ref. 23.
`
`and CNS. He isolated human neonatal
`brainstem cDNAs encoding proglucagon,
`which exhibited an identical sequence to
`that described for islet proglucagon (51).
`Molecular cloning of the rat intestinal pro-
`glucagon cDNA similarly revealed an open
`reading frame identical to that elucidated
`for the rat pancreatic islet proglucagon
`cDNA (52). Moreover, in studies carried
`out in collaboration with Patricia Brubak-
`er, forskolin, cholera toxin, and dibutyryl
`cyclic AMP increased the synthesis and
`secretion of intestinal PGDPs from primary
`cultures of rat intestinal cells (52). At the
`time, there were no differentiated GLP-1–
`secreting enteroendocrine cell lines suit-
`able for studies of intestinal proglucagon
`gene expression. Accordingly, Ying Lee, a
`fellow in the Drucker laboratory, generated
`a transgenic mouse expressing the SV40 T
`antigen cDNA under the control of the pro-
`glucagon gene promoter. This transgenic
`mouse reproducibly developed GLP-1–
`secreting enteroendocrine tumors of the
`colon (53), enabling isolation of the first
`differentiated GLP-1–producing entero-
`endocrine L cell line in 1992, designated
`GLUTag cells. GLUTag cells were easily
`propagated ex vivo; secreted immunoreac-
`tive GLP-1, glicentin, oxyntomodulin, and
`GLP-2 in response to cyclic AMP analogues
`(54, 55); and resembled primary cultures
`of nonimmortalized gut endocrine cells
`in regard to their response to a battery of
`secretagogues (56).
`
`Two unexpected observations were
`made during isolation of GLUTag cells.
`First, mice harboring s.c. GLUTag cell
`tumors exhibited a marked reduction
`of pancreatic islet α cell mass (57). Sec-
`ond, mice with s.c. GLUTag, InR1-G9, or
`RIN1056A glucagon-producing
`tumors
`all exhibited marked enlargement of the
`small bowel. These findings led Drucker
`to reinvestigate the link between glucagon-
`producing tumors and gut growth, first
`reported in a human subject studied at the
`Hammersmith hospital in 1970 by Dowl-
`ing and colleagues in London (58). A series
`of simple experiments from the Drucker
`lab published in 1996 identified GLP-2 as
`the PGDP with the most potent intestino-
`trophic activity in mice (58). Remarkably,
`although immunoreactive GLP-2 had been
`detected in intestinal extracts of various
`species (refs. 16, 59, and Figure 3), no pre-
`vious biological activity had yet been iden-
`tified for GLP-2 in vivo.
`The actions of GLP-2 to stimulate
`small bowel growth were rapid, detectable
`within days, and associated with increased
`crypt cell proliferation (58). Surprising-
`ly, when similar doses of GLP-2 were
`administered to rats, intestinal growth
`was not significantly increased, although
`an increase in crypt plus villus height was
`observed (60). With hindsight, these find-
`ings reflected the importance of DPP-4 for
`the degradation of GLP-2, more evident
`in rats than in mice. Subsequent studies in
`
`the Drucker lab demonstrated that native
`GLP-2 robustly increased small bowel
`growth in Fischer 344 rats with an inac-
`tivating mutation in the Dpp4 gene (60).
`Furthermore, a GLP-2 analogue with a sin-
`gle amino acid substitution [Gly2]-GLP-2
`exhibited substantial resistance to DPP-4
`cleavage and robust intestinotrophic activ-
`ity in normal rats in vivo (60). Hence, the
`importance of DPP4 for cleavage of both
`GLP-1 and GLP-2 became evident very
`early in the study of the GLPs.
`The identification of GLP-2 as an
`intestinal growth factor spurred a series
`of experiments examining the actions of
`GLP-2 in the context of experimental gut
`injury. GLP-2 administration was gener-
`ally associated with preservation of gut
`mucosal structure and function in the
`setting of chemical, radiation, or surgical-
`ly induced intestinal injury in preclinical
`studies (61–63). Notably, GLP-2 also rapid-
`ly increased nutrient absorption in normal
`rodents (64) and in animals with surgical
`gut resection mimicking SBS (65). Excit-
`ingly, the findings in animals were soon
`extended to humans with SBS. In a pilot
`study carried out around 2000–2001,
`Jeppesen and colleagues demonstrated
`that native GLP-2 administered twice dai-
`ly for 35 days increased nutrient absorp-
`tion, energy absorption, and weight gain
`in human subjects with SBS (66). These
`findings, namely expansion of intesti-
`nal mucosal surface area coupled with
`
`jci.org Volume 127 Number 12 December 2017
`
`4 2 2 1
`
`H A R R I N G T O N P R I Z E E S S A Y
`
`MPI EXHIBIT 1018 PAGE 5
`
`
`
`The Journal of Clinical Investigation
`
`generation of Glp1r–/– mice. These ani-
`mals generated by Louise Scrocchi exhib-
`ited impaired oral glucose tolerance and
`reduced insulin levels after glucose stim-
`ulation, demonstrating the critical role of
`endogenous GLP-1 as an incretin hormone
`in 1996 (69). Unexpectedly, Glp1r–/– mice
`also exhibited fasting hyperglycemia, and
`impaired i.p. glucose tolerance, establish-
`ing the importance of the GLP-1 for β cell
`function beyond its original description as
`an incretin. Moreover, GLP-1R–deficient β
`cells exhibit enhanced sensitivity to apop-
`totic injury, whereas pharmacological acti-
`vation of GLP-1R signaling attenuated β
`cell death in mice in vivo, as well as in cul-
`tures of purified rat β cells studied ex vivo
`(70). The importance of GLP-1R signaling
`for the response to cellular stress was high-
`lighted by findings that activation of GLP-
`1R signaling attenuated the development
`of ER stress in β cells by enhancing ATF-4
`induction and accelerating recovery from
`translational repression via augmentation
`of ER stress–stimulated ATF-4 translation
`(71). Collectively, these findings explain
`how GLP-1, despite acting to simultane-
`ously enhance insulin biosynthesis and
`secretion, maintains functional β cell mass
`through attenuation of ER stress.
`The actions of GLP-1 to stimulate insu-
`lin and inhibit glucagon secretion, together
`with its inhibitory effects on food intake
`and gastric emptying, spurred considerable
`effort in development of GLP-1–based ther-
`apeutics. Remarkably, the first GLP-1R ago-
`nist approved for clinical use was exenatide
`(synthetic exendin-4), a peptide originally
`isolated from Heloderma suspectum lizard
`venom by John Eng in 1992 (72). Subse-
`quent molecular cloning studies published
`by the Drucker laboratory in 1997 demon-
`strated that the lizard genome contains
`two distinct proglucagon gene sequences
`encoding GLP-1 and a separate proexendin
`gene, largely restricted in its expression to
`the salivary gland (73). The original clinical
`development program for exendin-4 uti-
`lized twice-daily injections of the unmodi-
`fied peptide. Pivotal studies in subjects with
`T2D ultimately resulted in the approval
`of twice-daily exenatide, the first GLP-1R
`agonist in 2005 (Figure 1). These efforts
`spurred the development of a once-weekly
`form of exenatide in a microsphere suspen-
`sion, ultimately the first once-weekly ther-
`apy approved for diabetes in 2012 (Figure
`
`Figure 5. Truncated forms of GLP-1 stimulate glucose-
`dependent insulin secretion. GLP-1(7-37) (A) but
`not GLP-1(1-37) (B) stimulates insulin secretion from
`the perfused rat pancreas as adapted from ref. 24.
`Sequences of GLP-1(1-37), GLP-1(7-37), and GLP-1
`(7-36amide). (C) Intraluminal glucose stimulates entero-
`glucagon, GLP-1, and GLP-2 secretion from the perfused
`pig ileum (D), as adapted from ref. 17. Demonstration
`that GLP-1(7-36)amide stimulates insulin secretion from
`the perfused pig pancreas (E), as adapted from ref. 25.
`
`enhanced nutrient absorption (67), togeth-
`er with substantial preclinical data, sup-
`ported the initiation of a drug development
`program to test the efficacy of GLP-2 in
`human subjects with parenteral nutrition–
`dependent (PN-dependent) SBS. Clinical
`studies were initiated with h[Gly2]-GLP-2,
`a degradation-resistant GLP-2 analogue
`discovered in the Drucker lab (60) and sub-
`sequently designated teduglutide. Once
`daily teduglutide administration in human
`subjects with PN-dependent SBS resulted
`
`in increased fluid and energy absorption,
`while reducing the requirements for PN, in
`two separate placebo-controlled Phase 3
`studies (67, 68). Teduglutide was approved
`for chronic therapy of subjects with
`PN-dependent SBS in the US in Decem-
`ber 2012 (67).
`
`The physiology of GLP-1 action
`Complementary studies in the Drucker lab
`were focused on defining the physiological
`importance of endogenous GLP-1 through
`
`4 2 2 2
`
`jci.org Volume 127 Number 12 December 2017
`
`H A R R I N G T O N P R I Z E E S S A Y
`
`MPI EXHIBIT 1018 PAGE 6
`
`
`
`The Journal of Clinical Investigation
`
`1). Exenatide once weekly was more effec-
`tive than the twice-daily preparation, with
`greater reduction of glycemia and compa-
`rable control of body weight (74). Today,
`multiple GLP-1R agonists (small peptides
`and high molecular weight proteins) are
`approved for the treatment of T2D, and a
`single drug, liraglutide, was approved for
`the treatment of obesity in 2014 (Figure 1).
`The widespread distribution of GLP-1
`receptors in extrapancreatic tissues stimu-
`lated considerable research into the nong-
`lycemic actions of GLP-1. Indeed, we now
`understand that GLP-1 controls inflam-
`mation (75), reduces experimental kidney
`injury (76), and like GLP-2, acts as a potent
`intestinal growth factor through mech-
`anisms including stimulation of crypt
`fission (ref. 77 and Figure 6). Among the
`actions of GLP-1 that have engendered the
`most interest are its effects in the cardio-
`vascular system. Native GLP-1 increases
`flow-mediated vasodilation, enhances
`heart rate (HR) and cardiac output, and
`is cardioprotective in preclinical studies,
`most notably in animals with ischemic
`cardiac injury (78). Although degradation-
`resistant GLP-1R agonists have minimal
`effects on blood vessels (79), they increase
`HR and reduce cardiac
`inflammation
`and infarct size in animals with experi-
`mental myocardial infarction (75, 80).
`Understanding the mechanisms through
`which GLP-1 exerts cardioprotection is
`challenging, since the GLP-1R is largely
`expressed in atrial and not ventricular car-
`diomyocytes (81). Furthermore, the GLP-
`1R–dependent control of inflammation in
`heart and blood vessels is complex, as the
`predominant site of GLP-1R expression in
`the (murine) immune system is within the
`intestinal intraepithelial lymphocyte (82).
`Several cardiovascular outcome stud-
`ies have documented the safety of GLP-1R
`agonists in human subjects with T2D and
`preexisting cardiovascular disease (83, 84).
`Excitingly, reduction of major adverse car-
`diovascular events was reported in human
`subjects treated with long-acting liraglutide
`and semaglutide (85, 86). These findings,
`reve