The insulin secretory action of novel polycyclic guanidines: Discovery through open innovation phenotypic screening, and exploration of structure-activity relationships

Article abstract of DOI:10.1016/j.bmcl.2014.01.021

We report the discovery of the glucose-dependent insulin secretogogue activity of a novel class of polycyclic guanidines through phenotypic screening as part of the Lilly Open Innovation Drug Discovery platform. Three compounds from the University of California, Irvine, 1-3, having the 3- arylhexahydropyrrolo[1,2-c]pyrimidin-1-amine scaffold acted as insulin secretagogues under high, but not low, glucose conditions. Exploration of the structure-activity relationship around the scaffold demonstrated the key role of the guanidine moiety, as well as the importance of two lipophilic regions, and led to the identification of 9h, which stimulated insulin secretion in isolated rat pancreatic islets in a glucose-dependent manner.

Full text of DOI:10.1016/j.bmcl.2014.01.021

Accepted Manuscript
The Insulin Secretory Action of Novel Polycyclic Guanidines: Discovery
through Open Innovation Phenotypic Screening, and Exploration of Structure-
Activity Relationships
Michael B. Shaghafi, David G. Barrett, Francis S. Willard, Larry E. Overman
PII:
S0960-894X(14)00040-7
DOI:
http://dx.doi.org/10.1016/j.bmcl.2014.01.021
BMCL 21256
Reference:
Bioorganic & Medicinal Chemistry Letters
To appear in:
Received Date:
Revised Date:
Accepted Date:
19 December 2013
6 January 2014
8 January 2014
Please cite this article as: Shaghafi, M.B., Barrett, D.G., Willard, F.S., Overman, L.E., The Insulin Secretory Action
of Novel Polycyclic Guanidines: Discovery through Open Innovation Phenotypic Screening, and Exploration of
Structure-Activity Relationships, Bioorganic & Medicinal Chemistry Letters (2014), doi: http://dx.doi.org/10.1016/
j.bmcl.2014.01.021
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

The Insulin Secretory Action of Novel Polycyclic Guanidines: Discovery through Open Innovation
Phenotypic Screening, and Exploration of Structure-Activity Relationships
¶
Michael B. Shaghafia , David G. Barrettb, Francis S. Willardb, and Larry E. Overmana*
aDepartment of Chemistry, 1102 Natural Sciences II, University of California, Irvine, California 92697-
2025
bLilly Research Laboratories, Eli Lilly and Company, Indianapolis, Indiana
*Corresponding author. Tel.: +1 949-824-7156; fax: +1 949-824-3866
E-mail Address: leoverma@uci.edu (L. Overman).
ABSTRACT: We report the discovery of the glucose-dependent insulin secretogogue activity of a novel
class of polycyclic guanidines through phenotypic screening as part of the Lilly Open Innovation Drug
Discovery platform. Three compounds from the University of California, Irvine, 1–3, having the 3-
arylhexahydropyrrolo[1,2-c]pyrimidin-1-amine scaffold acted as insulin secretagogues under high, but
not low, glucose conditions. Exploration of the structure-activity relationship around the scaffold
demonstrated the key role of the guanidine moiety, as well as the importance of two lipophilic regions,
and led to the identification of 9h, which stimulated insulin secretion in isolated rat pancreatic islets in a
glucose-dependent manner.
1

There are greater than 350-million people living with diabetes mellitus worldwide.1 The majority
of diabetes mellitus patients incur the type 2 diabetes, the adult onset variant of the disease,
characterized by elevated blood glucose levels in the settings of elevated peripheral insulin resistance
and impaired insulin secretion from pancreatic β cells. Several classes of medications are used to control
blood glucose in type 2 diabetes patients, including insulinotropic agents such as sulfonylureas, DPP-IV
inhibitors and GLP-1 analogues. The use of currently available therapies is tempered by their undesired
side effects, increased risk of hypoglycemia and short durability. The identification of small-molecule
insulin secretion agents that operate selectively under high-glucose conditions and amplify glucose-
induced insulin secretion by novel mechanisms therefore constitutes an opportunity for improvement in
the current standard of care for type-2 diabetes.
We previously described a medium-throughput insulin secretion assay that utilized INS-1E
insulinoma cells to measure insulin secretion under conditions of both high (5 mM) and low (0.1 mM)
glucose concentrations.2 Through the Lilly Open Innovation Drug Discovery platform,3 University of
California, Irvine (UCI) compounds were submitted in a structurally blinded fashion for phenotypic
screening, including in the insulin secretion assay. Three UCI compounds, 1–3, representing the 3-
arylhexahydropyrrolo[1,2-c]pyrimidin-1-amine scaffold acted as insulin secretagogues under high
glucose conditions (Table 1).4 Compound-stimulated insulin secretion could not be detected at basal
glucose concentration, suggesting a glucose dependency of their insulinotropic activity. In comparison,
the sulfonylurea glibenclamide is a more potent insulin secretogogue than the UCI compounds in this
assay, but also stimulates insulin secretion at low glucose concentrations as shown in Table 1. The
2

observed insulin secretion at low glucose concentrations is in line with the known propensity for
glibenclamide to stimulate insulin secretion in vivo in a glucose-independent fashion.5 ,6
Table 1
Phenotypic screening results identifying 3-arylhexahydropyrrolopyrimidin-1-amine lead compounds 1–3
as glucose dependent secretagogues in comparison with glibenclamide standard.a
aCompounds were tested for insulin secretagogue activity at 5 mM and 0.1 mM glucose in INS1-E cells (as
described in footnote 4). Data were fit to pharmacological models and mean and error values for potency
and efficacy (%stimulation) were calculated.
3

bNo increase in insulin secretion was detected at compound concentrations up to 33 uM.
As compound 3 stimulated insulin secretion at high glucose in isolated rat pancreatic islets (data
not shown), it was decided to explore the SAR of this scaffold in an effort to improve potency and to
confirm glucose dependency. We report herein the synthesis and evaluation of 3-
arylhexahydropyrrolo[1,2-c]pyrimidin-1-amines as potential glucose-dependent insulin secretagogues.
Bicyclic guanidines 1–3 could be prepared in six linear-steps by the route disclosed previously by
Overman and Wolfe (Scheme 1).7 Beginning with hydrocinnamaldehyde, Grignard reaction with 3-
(ethylenedioxy)propylmagnesium chloride and Mitsunobu installation of the azide unit gave 4 in good
yield.8 Reduction of the azide with LiAlH4 provided amine 5 in quantitative yield, which could then be
united with 1,3-di(Boc)thiourea utilizing Mukaiyama’s salt to obtain alkylguanidine acetal 6. Cleavage of
protecting groups and subsequent cyclization upon exposure to TFA at room temperature, followed by
quenching with thiophenol and a salt switch with dilute aqueous HCl, provided pyrrolidine N,S-
thioaminal hydrochloride salt 7 in 40% yield over three steps. Copper-mediated cyclocondensation of 7
with the appropriate alkenes afforded the desired bicyclic guanidines as mixtures of C3-epimers; this
step is exemplified in Scheme 1 by the synthesis of 2. In this and other similar constructions, the C3
substituent of the major epimer is equatorial, as readily determined by 1H NMR analysis.7
4

Scheme 1. Initial route to bicyclic guanidines illustrated by the synthesis of 2.
An abbreviated, more scalable route to bicyclic guanidines such as 2 was also utilized (Scheme 2).
In this sequence, Mitsunobu displacement of secondary alcohol 8 with commercially available di-
(Boc)guanidine provided intermediate 6 directly. 9,10 Incorporation of this transformation in the
synthetic sequence obviates the previously utilized azide chemistry and the necessary oxidation state
change, while reducing the synthetic sequence to only four linear steps.
5

Scheme 2. Abbreviated route to bicyclic guanidines utilizing Mitsunobu incorporation of the guanidine
unit.
Our exploration of the SAR surrounding UCI lead compounds 1–3 in the insulin secretion assay
focused on insulin secretion at 5 mM glucose (high glucose). The ability of each compound to stimulate
basal insulin secretion was also examined, and no compound exhibited significant insulinotropic activity
at concentrations up to 30 µM in the presence of low glucose (data not shown). We first examined the
structure activity relationship of the C3-substituent using the cyclocondensation of N-amidinyliminium
ion precursor 7 with a series of terminal alkenes to prepare 9a–9m (Table 2).11, 12 A substituent is
necessary at this position, as evidenced by the lack of activity of the de-phenyl analog 9a.13 The phenyl
epimer, C3 epi-2, was not significantly less potent than 2 itself. The cyclopentyl analog 9b retained
activity, suggesting that lipophilicity, but not necessarily aromaticity, was required in this portion of the
scaffold, as also indicated by the activity of 1. Introduction of methyl or bromo substituents in the aryl
ring did not significantly impact activity, regardless of their position on the ring (entries 5–7 and 9–11).
However, the nature of the para substituent proved important. Whereas bromo and methyl were
6

tolerated, analog 9f bearing the trifluoromethyl group proved inactive over the concentration range
studied (entry 12).14 Furthermore, incorporation of a para carboxylic acid or an N-acylamino substituent
also resulted in inactive compounds (entries 13–14). 15
Table 2
Insulin secretion SAR of 3-susbstituted hexahydropyrrolo[1,2-c]pyrimidin-1-amines.a
7

aCompounds were tested for insulin secretagogue activity at 5 mM glucose in INS1-E cells (as described
in footnote 4). Data were fit to pharmacological models and mean and error values for potency and
efficacy (%stimulation) were calculated.
bThis sample contained up to 10% of the minor epimer based on analysis of the 1H-NMR spectrum.
cThis sample contained ca 30% of the minor epimer based on analysis of the 1H-NMR spectrum.
The SAR of the C7 substituent (phenethyl in 2) was explored next, using the cyclocondensation of
styrene with the appropriate N-amidinyliminium ion precursors to obtain 10a–10m (Table 3).16 Deletion
of the side chain resulted in a loss of activity over the concentration range tested (entry 2). The length
and nature of the spacer between a phenyl group and the core had a large impact on activity. Removal of
the spacer eliminated activity (entry 3). The analog with a methylene spacer was perhaps slightly less
potent than 2 (entry 4), whereas the analog with a propylene spacer retained activity (entry 5).
Replacement of the ethylene spacer with biphenyl abrogated activity (entry 6), as did replacement of
phenethyl with an oxymethylene spacer (entry 7). Replacement of the phenethyl group with n-propyl
resulted in decreased activity (entry 8). Introduction of a 3-bromo substituent on the phenyl ring
afforded a modest increase in potency (entry 9). Of note, incorporation of a methanesulfonamide at either
the 3- or 4- position of the phenyl substituent resulted in a loss of activity, whereas the
methanesulfonamide appended in the 2-position was tolerated (entries 10–12). Finally, the two
heteroaryls studied, 4-pyridyl and N-methylpyrazole, proved inactive (entries 13, 14).
Table 3
8

Insulin Secretion SAR of the C7 substituent.a
aCompounds were tested for insulin secretagogue activity at 5 mM glucose in INS1-E cells (as described
in footnote 4). Data were fit to pharmacological models and mean and error values for potency and
efficacy (%stimulation) were calculated.
bThis sample contained up to 20% of the minor epimer based on analysis of the 1H-NMR spectrum.
9

We next examined the affect of the nature and presence of the guanidine moiety on insulin
secretion (Table 4). Methylation of either guanidine nitrogen had no detectable impact on activity
(entries 1, 2),17 whereas the tert-butoxycarbonyl derivative 11c showed no measurable activity (entry
3).18 Replacement of the exocyclic guanidine nitrogen with oxygen to afford the corresponding urea19
also resulted in a loss of activity (entry 4). Collectively, these data support the necessity of a guanidinium
cation in order to stimulate insulin secretion.
Table 4
Insulin secretion SAR of analogs bearing modification at the guanidine functionality of 2.
10

aCompounds were tested for insulin secretagogue activity at 5 mM glucose in INS1-E cells (as described
in footnote 4). Data were fit to pharmacological models and mean and error values for potency and
efficacy (%stimulation) were calculated.
bThis sample contained ca 30% of 11a based on analysis of the 1H-NMR spectrum.
Finally we examined compounds 12a-e, containing only one of the two rings of guanidinium lead
structure 2 (Table 5).20 Removing the pyrrolidine ring afforded 12a, which, although less potent than 2,
retained measurable activity (entry 1). Analog 12b, in which the hexahydropyrrolopyrimidine ring has
11

been opened, also retained activity (entry 2). That these simple analogs stimulated insulin secretion led
us to determine the activity of other simple guanidines including the synthetic intermediate 7, which
proved equipotent to 2 (entry 3). Of note in this series, compound 12c bearing a phenyl substituent
adjacent to the pyrrolidine nitrogen provided single-digit micromolar activity (entry 4). This result
contrasts with the lack of activity observed for the corresponding phenyl analogue in the bicylic
guanidinium series (Table 3, entry 2). Of further note, replacement of the thioether of 7 with either a
phenethyl or benzyl group resulted in a modest loss in potency (Table 5, entries 5–6). Overall, these data
demonstrate how critical the presence of the guanidine group is for activity in this series of insulin
secretogogues.
Table 5.
Insulin secretion SAR of ring-opened guanidinium salts.a
12

aCompounds were tested for insulin secretagogue activity at 5 mM glucose in INS1-E cells (as described
in footnote 4). Data were fit to pharmacological models and mean and error values for potency and
efficacy (%stimulation) were calculated.
13

bThis sample contained a 1:1 mixture of diastereomers based upon 1H-NMR analysis.
cThis sample contained a ca. 3:1 mixture of epimers based on analysis of the 1H-NMR spectrum. The
major isomer was assumed to be syn, in line with the published data (ref 20).
dThis sample contained a ca. 9:1 syn/anti mixture based on 1H-NMR NOE experiments.
To determine whether the 3-arylhexahydropyrrolo[1,2-c]pyrimidin-1-amines stimulate insulin
secretion in a glucose-dependent manner in primary cells, isolated rat pancreatic islets were treated with
different concentrations of compound 9h (Figure 1). Secretogogue 9h (3 µM) amplified glucose-induced
insulin secretion at glucose concentrations of 11 mM, but did not increase basal insulin secretion when
the glucose concentration was 3 mM, confirming a glucose-dependent mode of action for this compound.
4
*
3
2
1
0
3 mM Glucose
11 mM Glucose
Figure 1. Stimulation of insulin secretion with compound 9h in isolated rat pancreatic islets at 3 mM and
11 mM glucose. Insulin secretion was measured in the presence of vehicle (white bars) or 3 µM
compound 9h (black bars). *P<0.05 versus insulin secretion in the vehicle group at 11 mM glucose.
In summary, using a phenotypic assay, we were able to identify a class of novel bicyclic guanidines
that act as glucose-dependent insulin secretogogues in both an insulinoma cell line and primary rat
14

pancreatic β cells. Exploration of the structure-activity relationship around the scaffold demonstrated the
key role of the guanidine moiety, as well as the importance of two lipophilic regions. Unfortunately, the
exploration did not result in significant increases in potency versus the initial hits, although it is clear that
certain functional groups are not tolerated. The identification of adequate dynamic range in structure
activity relationships remains a key attribute for the progression of phenotypic screening-derived
chemotypes into lead optimization campaigns.
The presence of clear structure-activity relationships in the series studied, as well as the glucose
dependency of the effects, suggests that the compounds elicit a secretory response from cells by a
discrete interaction with a molecular target or set of related targets, rather than through ill-defined or
cytotoxic cell modifications. The necessity of the guanidine group for activity suggests that this scaffold
might share a mode of action with other guanidines21 or even imidazolines that are known to be
insulinotropic agents.22 Even though the glucose-lowering effects of some of these agents have been
demonstrated clinically,23,24,25 their molecular targets remain undiscovered despite intensive
investigation.
Acknowledgements
The authors acknowledge Jefferson R. McCowan, Charles W. Lugar, and Timothy A. Grese for their
input into the SAR analysis, and Karen Cox and Samreen Syed for the generation of data. We also thank
Alexander M. Efanov and Jefferson McCowan for a careful review of the manuscript and supporting data,
and the Eli Lilly and Company OIDD team for their support. Support for research carried out at UC Irvine
was provided by Eli Lilly. NMR and mass spectra of compounds prepared and characterized at UC Irvine
were determined using instruments purchased with the assistance of NSF and NIH shared
15

instrumentation grants. We thank Dr. John Greaves, Department of Chemistry, UC Irvine, for assistance
with mass spectrometric analyses.
Supporting data
Supplementary data associated with this article can be found in the online version, at
References and Notes
16

¶
Current address: eFFECTOR Therapeutics, Inc., Medicinal Chemistry, 11180 Roselle St. San Diego,
CA 92121.
1. WHO fact sheet Nº312, updated March 2013, http://www.who.int/mediacentre/factsheets/
fs312/en/
2. Lee, J. A.; Chu, S.; Willard, F. S.; Cox, K. L.; Sells Galvin, R. J.; Peery, R. B.; Oliver, J. O.; Meredith, S. H.;
Heidler, S. A.; Gough, W. H.; Husain, S. Palkowitz, A. D.; Moxham, C. M. J. Biomol. Screen 2011, 16, 588.
3. To learn more about the Lilly Open Innovation Drug Discovery platform, please visit the program
website at https://openinnovation.lilly.com
4. Biological potency determination was conducted as described in reference (2) with minor
modifications. Data were normalized relative to the maximal insulin secretion induced by the
combination of 40 mM glucose and 50 µM IBMX (as described in reference [2]). Data were fit to the
four parameter logistic equation using PRISM 5.0 (GraphPAD) and reported ‘% Stimulation’ values
correspond to fitted ‘Top’ values derived from curve fits. 20-point dose response curves with a 2-
fold compound dilution factor were used to increase precision and obviate a biphasic concentration
response relationship observed at high compound concentrations. Data were fit to dose response
curves when maximal % stimulation observed was >15 %.
5. Efendic S.; Enzmann, F.; Nylen, A; Uvnas-Wallensten, K.: Luft, R. Proc. Nat. Acad. Sci. U.S.A. 1979, 76,
5901.
6.
Patane, G.; Piro, S.; Anello, M. Rabuazzo, A. M.; Vigneri, R.; Purrello, F. British Journal of Pharmacology
2000, 129, 887.
7.
8.
The rel-(3R, 4aR, 7S) stereoisomer predominates to the extent of 3–5:1, see: Overman, L. E.; Wolfe, J.
P. J. Org. Chem. 2001, 66, 3167.
It was later discovered that diphenylphosphoryl azide (DPPA) was a suitable surrogate for the HN3:
Shaghafi, M. B., UC Irvine, unpublished Work, 2011.
9. Procedure for the synthesis of 6 utilizing the Mitsunobu sequence depicted in Scheme 2: To a flame-
dried 500-mL round-bottom flask having a 200 mL addition funnel attached, was added alcohol 8
(5.30 g, 22.4 mmol). Anhydrous toluene (170 mL) was then added, followed by Ph3P (8.80 g, 33.6
mmol, 1.50 equiv) and di-(Boc)guanidine (11.6 g, 44.8 mmol, 2.0 equiv). The flask was placed in a
bath at 0 °C. Once the stirred solution was homogeneous, diisopropyl azodicarboxylate (DIAD, 6.80 g,
33.6 mmol, 1.5 equiv) was added to the addition funnel via syringe, followed by toluene (55 mL).
The DIAD solution was then delivered to the stirred solution at 0 °C over ~20 minutes, at which time
a yellow precipitate formed. The solution was stirred at 0 °C for a further 2 h (precipitate dissipates),
then allowed to warm to ambient temperature overnight. The solvent was removed in vacuo to
provide a thick yellow residue, which was diluted into ~250 mL of EtOAc (3 x 250 mL) and washed
with 10% citric acid solution saturated with NaCl. The combined organic layers were then washed
with brine, dried over Na2SO4, decanted and concentrated in vacuo. Purification of the residue by
silica gel chromatography (EtOAc/hexanes; 2% ꢀ 17.5%) gave 6 (6.2 g, 58%) as a colorless powder,
which was identical to the product prepared by the published procedure.7
10. Wolfe, J. P. UC, Irvine, unpublished studies, 1999.
11. For a discussion of the observed stereochemical outcome in cyclocondensation reactions providing
3-arylhexahydropyrrolo[1,2-c]pyrimidin-1-amines, see ref. 7.
12 In general, the epimers obtained from the cyclocondensation were separated and only the major
epimer was tested. Examples in which the epimers were not separated are noted in the Tables.
17

13. The de-phenyl analogue 9a was prepared by cyclocondensation of allyltrimethylsilane and 7 in the
presence of Cu(OTf)2, providing the C4a-allylated cycloadduct.7 This compound was elaborated to
9a by sequential ozonolysis/sodium borohydride reduction.
14. Jagodzinska, M.; Huguenot, F.; Candiani, G.; Zanda, M. ChemMedChem 2009, 4, 49.
15 The N-acetyl analog 9l was prepared from 9e by palladium-catalyzed amidation after protection of
the guanidine with a tert-butyloxycarbonyl (Boc) group. Subsequent removal of the Boc group
afforded 9l.
16 The synthetic sequences leading to compounds 10a–10m are described in the supplementary data.
17 The N-Methyl analog 11a was prepared from 2 by selective methylation of the endocyclic nitrogen
with methyl iodide. N-Methyl analog 11b was obtained from 2 by partially selective endo alkylation
with 4-methoxybenzyl (PMB) chloride, followed by exo N-alkylation with methyl iodide, and
removal of the PMB group.
18 The tert-butoxycarbonyl analog 11c was prepared from 2 by treatment with di-tert-butyl
dicarbonate in dioxane containing aqueous sodium hydroxide.
19 For a description of the synthesis of urea 11d, see ref 7.
20 Synthetic sequences leading to compounds 12a–12e are described in the supplementary data. The
intermediate prepared en route to 12d was obtained as a mixture of epimers according to a
published procedure, see: Katritsky, A. R.; Cui, X.-L.; Yang, B.; Steel, P. J. J. Org. Chem. 1999, 64, 1979.
21. Kosasayama, A.; Konno, T.; Higashi, K.; Ishikawa, F. Chem. Pharm. Bull. 1979, 27, 841.
22. Rustenbeck, I.; Herrmann, C.; Ratzka, P. Naunyn-Schmiedeberg’s Arch. Pharmacol. 1997, 356, 410.
23. Morgan, N. G. Exp. Opin. Invest. Drugs 1999, 8, 575.
24. Byrom, W. D.: Rotherham, N. E.; Bratty, J. R. Br. J. Clin. Pharmac. 1994, 38, 433.
25. Kawazu, S.; Suzuki, M.; Negishi, K.; Watanabe, T.; Ishii, J. Diabetes 1987, 36, 216.
18