Engineering of Orally Available, Ultralong-Acting Insulin Analogues: Discovery of OI338 and OI320

Article abstract of DOI:10.1021/acs.jmedchem.0c01576

Recently, the first basal oral insulin (OI338) was shown to provide similar treatment outcomes to insulin glargine in a phase 2a clinical trial. Here, we report the engineering of a novel class of basal oral insulin analogues of which OI338, 10, in this publication, was successfully tested in the phase 2a clinical trial. We found that the introduction of two insulin substitutions, A14E and B25H, was needed to provide increased stability toward proteolysis. Ultralong pharmacokinetic profiles were obtained by attaching an albumin-binding side chain derived from octadecanedioic (C18) or icosanedioic acid (C20) to the lysine in position B29. Crucial for obtaining the ultralong PK profile was also a significant reduction of insulin receptor affinity. Oral bioavailability in dogs indicated that C18-based analogues were superior to C20-based analogues. These studies led to the identification of the two clinical candidates OI338 and OI320 (10 and 24, respectively).

Full text of DOI:10.1021/acs.jmedchem.0c01576

pubs.acs.org/jmc
Article
Engineering of Orally Available, Ultralong-Acting Insulin Analogues:
Discovery of OI338 and OI320
̌
́
Thomas B. Kjeldsen, Frantisek Hubalek, Tina M. Tagmose, Lone Pridal, Hanne H. F. Refsgaard,
̈
Trine Porsgaard, Sanne Gram-Nielsen, Lars Hovgaard, Henrik Valore, Martin Munzel,
̀
Claudia U. Hjørringgaard, Claus Bekker Jeppesen, Valentina Manfe, Thomas Hoeg-Jensen,
Svend Ludvigsen, Peter Kresten Nielsen, Inger Lautrup-Larsen, Carsten E. Stidsen, Erik M. Wulff,
Patrick W. Garibay, Janos T. Kodra, Erica Nishimura, and Peter Madsen*
́
Cite This: J. Med. Chem. 2021, 64, 616−628
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: Recently, the first basal oral insulin (OI338) was
shown to provide similar treatment outcomes to insulin glargine in
a phase 2a clinical trial. Here, we report the engineering of a novel
class of basal oral insulin analogues of which OI338, 10, in this
publication, was successfully tested in the phase 2a clinical trial. We
found that the introduction of two insulin substitutions, A14E and
B25H, was needed to provide increased stability toward
proteolysis. Ultralong pharmacokinetic profiles were obtained by
attaching an albumin-binding side chain derived from octadeca-
nedioic (C18) or icosanedioic acid (C20) to the lysine in position
B29. Crucial for obtaining the ultralong PK profile was also a
significant reduction of insulin receptor affinity. Oral bioavailability
in dogs indicated that C18-based analogues were superior to C20-based analogues. These studies led to the identification of the two
clinical candidates OI338 and OI320 (10 and 24, respectively).
Second, the environment in the gastrointestinal tract is harsh
and composed of a cocktail of numerous proteolytic enzymes.
If sufficient amounts of insulin can survive proteolysis, it still
needs to cross the gastric or intestinal epithelium in order to be
absorbed.
Numerous attempts of developing orally available insulin
have been undertaken by many research groups and
pharmaceutical companies. This research has been reviewed
extensively.6−13 In the vast majority of clinical studies
performed, the insulin that was tested was human insulin
(HI), that is, fast-acting insulin.9 Because the therapeutic
window of fast-acting insulin is very narrow with the risk of
potential fatal hypoglycemic episodes due to insulin over-
dosing, we considered that variation in bioavailability of orally
dosed fast-acting insulin would be of too high risk for the
patient. Instead we envisioned that oral dosing of an ultralong-
acting insulin (>24 h) would overcome the issue of variability
INTRODUCTION
■
Insulin provides a highly efficacious treatment of patients with
type 1 diabetes and of patients with later stages of type 2
diabetes and is currently administered by subcutaneous
injection of fast-acting and/or long-acting insulin analogues.
Ever since the discovery of insulin in 1921,1,2 many efforts have
been made to find a way to make insulin orally available. The
first unsuccessful attempt was reported as early as in 1923.3
Patients have very strong preference for oral insulin. If such
treatment was available, it is believed to improve the quality of
life by reducing the treatment burden of (multiple) daily
subcutaneous injections. Oral insulin might also improve
patient compliance and adherence to their treatment regimes.
Furthermore, orally delivered insulin, like endogenous insulin,
would enter circulation via the hepatic vein, and it is likely that
rapid insulinization of the liver would provide further
physiological benefits relative to subcutaneously delivered
insulin, such as better glucose control because of efficiently
shutting down of hepatic glucose production, fewer hypo-
glycemic episodes because of less insulin reaching peripheral
tissues, reduced weight gain, and increased insulin sensitiv-
ity.4,5
Received: September 9, 2020
Published: December 28, 2020
There are serious barriers for insulin absorption via the oral
route. First, the low pH in the stomach needs to be considered.
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
© 2020 American Chemical Society
J. Med. Chem. 2021, 64, 616−628
616

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 1. Insulin detemir, insulin degludec, and O346. Insulin is desB30 HI and side chains are: insulin detemir: myristoyl, insulin degludec:
hexadecanedioyl-gGlu, and O346: octadecanedioyl.
of day-to-day variation of bioavailability because at steady state,
a larger fraction of the absorbed insulin dose would exert its
action on the day(s) after the administration.14
compromised glucodynamic potency would be detrimental to
this project as this would increase the dose requirements.
Prior to considering clinical testing of any basal oral insulin,
we decided on a set of criteria that the new insulin analogue
should meet in order to increase the likelihood of success. (1)
The insulin should be stabilized toward proteolytic degrada-
tion, (2) the i.v. PK in dogs should be around 1 day, (3) the
pharmacodynamic potency should not be compromised, and
(4) oral bioavailability of the insulin in a pharmaceutically
relevant tablet formulation should be >2% in dogs.
Many approaches and formulation designs have been tested
trying to achieve oral delivery of insulin but with limited
success.6−13 Examples of common of permeation enhancers are
bile salts, fatty acids, surfactants, salicylates, Ca2+-chelators, and
others. The absorption enhancer sodium caprate elicits
dilatations in human intestinal tight junctions and enhances
drug absorption by the paracellular route.18 The so-called
GIPET I (gastrointestinal permeation enhancement technol-
ogy) tablet formulation technology18 based on sodium caprate
was used in the previously reported clinical trials.19,20 Hence,
to align with the clinical results, here we also focused on
caprate as the permeation enhancer in this study. Insulins
formulated in the GIPET formulation are not expected to
aggregate when released in the gastrointestinal tract because of
high concentrations of caprate.
Here, we report our journey from HI to orally available long-
acting protease-stabilized insulin analogues that meet the
defined design criteria, and when compared with a very similar
nonprotease-stabilized analogue, they display much improved
oral bioavailabilities in GIPET I formulations when tested in
dogs. We selected two analogues for further preclinical and
conceptual pharmacology studies.14 These two analogues have
subsequently progressed into clinical testing. One of these
analogues, insulin 338 or OI338, 10 in this publication, has
successfully completed a phase 2a clinical study,19 an 8-week,
randomized, double-blind, double-dummy, active-controlled,
parallel trial with 50 type 2 diabetic patients comparing oral
OI338 with subcutaneous insulin glargine. This study showed
that glycemic control was not different between the groups
receiving oral OI338 and subcutaneous insulin glargine, and
that no severe adverse events were reported. Notably, the
incidence of hypoglycemia was low in both groups and with no
severe episodes. The reported elimination plasma half-life of
OI338 in man was 70 h.19 This type of analogues has
furthermore been shown20 to have a hepatocentric mode of
action which efficiently protects against hypoglycemia caused
by unintentional overdosing, such as foreseen in day to day
variation of oral bioavailability. It was also unequivocally
In order to obtain such ultralong-acting insulins with
sufficient oral bioavailability, we hypothesized that strong
and reversible binding to albumin by fatty acid acylation
technology (lipidation) in combination with insulin substitu-
tions that mediate increased stability toward proteolysis and
insulin substitutions that mediate significant reduction of
insulin receptor affinity would lead to compounds with the
desired properties.14
As our hypothesis on the strategy of lowering of the insulin
receptor affinity to obtain protraction may seem contra
intuitive to most medicinal chemists, it is explained here in
further detail. Insulin, after engagement with and activation of
its receptor, is cleared via receptor-mediated endocytosis.15
A
lowering of the insulin receptor affinity will, thus, lower the
rate of insulin receptor-mediated clearance, and, according to
our hypothesis, mediate a more protracted profile. It has been
shown16 that insulins with decreased or increased receptor
affinities compared to HI (31 and 207%, resp.) have equal
glucose lowering potencies when tested in pigs by the
hyperinsulinemic, euglycemic clamp technique. The analogue
with low affinity displayed a significantly lower in vivo
clearance rate compared to HI and the high affinity analogue
[6.9 mL/(kg × min) vs 19.9 and 26.4 mL/(kg × min),
respectively].16 Based on these data, we hypothesized that
lowering of insulin receptor affinity of fatty acid acylated
insulins would mediate a longer pharmacokinetic profile
without loss of glucose lowering potency, provided that the
insulin in circulation is not subject to other types of clearance.
Here, albumin binding becomes important in order to prevent
renal clearance. If other types of clearance are prevented, or are
not operational, every administered insulin molecule will,
eventually, reach the insulin receptor, and elicit their actions,
resulting in cellular glucose uptake and insulin clearance
without loss of glucose lowering potency.
One insulin with very long systemic half-life has been
pharmacodynamically characterized by i.v. bolus administra-
tion to dogs.17 The analogue, O346 (Figure 1), is desB30 HI
acylated at the LysB29 position with the α,ω-dicarboxylic acid
(fatty diacid) with 18 carbon atoms, octadecanedioic acid. This
analogue was reported to have a high albumin-binding affinity
and, further, to have an extended time-action profile with time
to reach 95% of total O346-stimulated glucose uptake of 45 h.
O346 had, however, markedly reduced glucodynamic (glucose
lowering) potency in dogs, with only 18% biological efficacy
relative to HI.17 As oral bioavailability is expected to be low, a
617
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
a
Scheme 1. General Solution and Solid Phase Syntheses of Side Chains Used for Acylation of Insulin
a
(a) (1) TSTU, DIEA, THF; (2) H-GluOtBu, THF, DMF water; (b) TSTU, DIEA, THF, MeCN; (c) H-OEG-OEG-OH, DIEA, EtOH; (d)
TSTU, DIEA, MeCN; (e) TFA; (f) Fmoc-OEG-OH, DIEA, DCM; (g) (1) Fmoc-Glu-OtBu, HOBt, DIC, DIEA, NMP/DCM; (2) 20% piperidine,
NMP, repeat 1 and 2 with required Fmoc amino acids; (3) 5% TFA/DCM.
a
Scheme 2
a
t
(i) MsCl, Et3N, DCM; (ii) KCN, DMSO, 40 °C, 16 h; (iii) KOH, THF, MeOH, H2O; (iv) Boc2O, DMAP, BuOH/THF (4:1)
shown that the compounds do not have compromised glucose
lowering potencies despite their low insulin receptor affinities
and long duration of action.14
convenient to prepare stock batches of the tert-butyl-protected
side chains (e.g., I) on the solid support on larger scale and
convert these to the unprotected OSu-activated side chains
(e.g., III) in smaller batches and use these until the quality
deteriorated because of hydrolysis of the OSu ester. It was
found to be very important that this product was crystallized,
washed, dried thoroughly, and stored at −18 °C, otherwise
residual TFA rendered the product hygroscopic and highly
prone to hydrolysis of the OSu ester. Shelf life of this side
chain is limited, but 2−3 months is often obtained.
The 1,17-heptadecanedioic acid was prepared in excellent
yields, as shown in Scheme 2; starting with O-mesylation of
methyl 16-hydroxyhexadecanoic acid, substitution with cyanide
and hydrolysis afforded the desired diacid.
The various mono-tert-butyl protected fatty diacids (C16,
C17, C18, and C20) were in turn prepared from the
corresponding fatty diacids, as shown for the C17 diacid in
Scheme 2, last step. By this procedure, briefly, a statistical
mixture of unchanged diacid and mono- and diester was
obtained. The unchanged diacids were insoluble and could be
removed by filtration, whereas the mono- and di-esters were
CHEMISTRY
■
Preparation of Side Chains. Synthesis of fatty diacid
containing side chains (Scheme 1) was performed by simple
amide coupling reactions, either in solution or on solid
support. Briefly, the solution phase method starts with the
mono-tert-butyl protected fatty diacid, which is coupled with
H-Glu(OH)-OtBu using N,N,N′,N′-tetramethyl-O-(N-
succinimidyl)uronium tetrafluoroborate (TSTU) activation.
Depending on sequence of the desired side chain, this product
is further coupled with other amino acids. Penultimately, the
protected side chain (e.g., I) is activated as an N-
hydroxysuccinimide ester (OSu-ester) (e.g., II) using TSTU
and deprotected with trifluoroacetic acid (TFA) to give the
final unprotected, activated side chain (e.g., III). Alternatively,
the key intermediate protected side chains (e.g., I) can be
synthesized on a solid support (2-Cl-trityl chloride polystyrene
resin), as also shown in Scheme 1. We found it most
618
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Scheme 3. Acylation of Insulin. Protected and Unprotected Activated Side Chains were Dissolved in MeCN or DMF and
a
Added Dropwise to a Stirred Solution of Insulin (100 mg/mL) at pH 11
a
tert-Butyl deprotections were carried out with neat TFA.
Figure 2. Abbreviations of fatty diacids and linkers in Tables 2−7.
separated by chromatography. The isolated di-esters could be
saponified and subjected to re-esterification, thus, the practical
yields of mono-esters were substantially higher than the
reported 27%.
Acylation of Insulin. Acylation of insulin at the lysine side
chain in position B29 has been described previously using fatty
diacid containing side chains leading to the discovery of insulin
degludec.21 In this work, we have used two different
approaches for lysine acylation (Scheme 3). Initially, we used
tert-butyl-protected OSu-activated side chains (like II) for
insulin acylation. Briefly, the side chain was dissolved in
acetonitrile or N-methyl pyrrolidone (NMP) and added slowly
to the insulin dissolved in an aqueous buffer at pH 11, while
keeping pH constant by simultaneous addition of 1 M NaOH.
The side chain was added until adequate conversion was
reached, as judged by liquid chromatography−mass spectrom-
etry (LC−MS). The insulin product mixture was precipitated
by adjusting pH to 5.2, followed by isolation by centrifugation
619
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
and lyophilization. The product mixture was then deprotected
by dissolution in TFA. After concentration in vacuo, the
desired product was purified by preparative reversed-phase
high-performance liquid chromatography (RP-HPLC). Later,
we learned that the acylation could be carried out accordingly
but more conveniently with unprotected OSu-activated side
and expressed as fold relative stability to 4. The strength of the
extract was standardized and appropriately diluted to ensure
that degradation kinetics of 4 were comparable from batch to
batch. This assay was used to evaluate and compare proteolytic
stabilities of the insulin analogues.
The receptor-binding assays22 were performed in two
different setups; one setup is using solubilized insulin receptor
A isoform in the absence of human serum albumin (HSA) and
the other setup is in the presence of either 1.5 or 4.5% HSA.
The assay in the absence of HSA gives an indication of the
native affinity of the analogue to the receptor, whereas the
assay performed in the presence of HSA gives an indication of
the functional insulin receptor affinity as the side chains of the
acylated insulin analogues are conferring affinity for HSA.
Insulin receptor affinities are given in percent relative to that of
HI (in the same assay plate).
t
chains (like III) in one step. The Bu-protected OSu-activated
side chains (e.g., II) could be deprotected with anhydrous
TFA, and the desired activated unprotected side chains (e.g.,
III) could be isolated by concentration in vacuo and
crystallization, as described above (Scheme 3).
In the text and tables throughout this publication, the side
chains attached to insulins (insulin backbones) are abbreviated
and the structures of the components are listed in Figure 2.
RESULTS AND DISCUSSION
■
Insulin receptor-binding data (Table 2) show that the
individual substitutions B25H and A14E reduce IR affinity to
30% (6) and 79% (3), respectively. The analogue 4 with the
substitutions A14E, B25H, desB30 displays IR affinity of 23%,
and the relative stability of the analogues in the rat duodenum
stability assay indicates that combination of the two
substitutions is crucial for obtaining stability toward proteolytic
degradation (Table 2) as HI, and HI with the single
substitutions B25H (2), and A14E (3) displays relative
stabilities of 0.4, and 0.3 fold relative to A14E, B25H, and
desB30 (3). Furthermore, it is noted that removal of B30T
from HI (desB30) neither influence insulin receptor binding
nor enzymatic stability (5 and 1).
The side chain of the once weekly glucagon-like peptide-1
receptor agonist semaglutide23 was chosen as a model albumin-
binding side chain, and the analogues 5, 6, 3, and 4 were
acylated at the lysine residue in position B29 accordingly. The
resulting acylated analogues were assayed for insulin receptor-
binding affinity without HSA and in the presence of 4.5% HSA
in the assay buffer (Table 2). Insulin receptor-binding affinity
data without HSA showed that analogues with the sub-
stitutions desB30 (7) and A14E, desB30 (9) both displayed
relatively high affinity (13.9% and 6.19%, respectively). The
two analogues containing the B25H substitution (8 and 10)
had lower affinities (1.9 and 2.3%, respectively). In the
presence of 4.5% HSA, the affinities for the insulin receptor
were further lowered and the two analogues with the B25H
substitution displayed the lowest affinities (8 and 10, both
0.2%). The analogues were also tested for proteolytic stability
relative to 4 in the rat duodenum degradation assay, and again
the importance of both substitutions A14E and B25H being
present is evident [relative stability 0.9 (10) versus 0.1 (7), 0.3
(8), and 0.1 (9)] (see Table 2).
In order to engineer an insulin with increased stability toward
proteolytic degradation relative to HI, a range of insulin
analogues were screened for stability toward pepsin, chymo-
trypsin, and carboxypeptidase A. Two analogues were found
that were more stable than HI (1): B25H HI (2) and A14E,
desB30 HI (3). By combining the two substitutions, A14E,
B25H, desB30 HI (4) were identified. The effects of the two
substitutions were additive or synergistic, depending on the
enzyme studied, and the analogue 4 was significantly more
stable than HI (see Table 1).
Table 1. Stability of Insulins toward Proteolytic Enzymes
carboxypepti-
pepsin
fold
chymotrypsin
fold
dase A
fold
substitutions
in HI
t1/2
(rel. to
HI)
t1/2
(rel. to
HI)
t1/2
(rel. to
HI)
Cpd #
(min)
(min)
(min)
1
2
3
HI
B25H
A14E,
desB30
1.1
17.8
3.7
1
16.2
3.4
5.0
25.7
33.2
1
5.1
6.6
6.9
25.7
15.7
1
2.1
2.3
4
A14E, B25H, 42.4
desB30
38.5
62.4
12.5
61.9
9.0
In Figure 3, the time course of degradation of insulin
analogues by enzymes from rat duodenum is shown. The
additive effects of the substitutions A14E and B25H are
evident.
We developed a proteolytic assay based on aqueous extracts
from luminal rat duodenum, where the kinetics (t1/2) of
degradation of insulins were compared to the degradation of 4
Thus, we decided to fix the insulin backbone to A14E,
B25H, and desB30 (4) because the stability toward enzymatic
degradation is mediated primarily by these insulin substitu-
tions, and the desired low insulin receptor affinity is obtained
primarily by the B25H substitution.
Next, we investigated the length of the albumin-binding fatty
diacid in order to explore how the length of the fatty diacid
influences the PK and the proteolytic stability (Table 3). A
series of closely related analogues were made with 1,16-
hexadecanedioic acid (C16, 11), 1,17-heptadecanedioic acid
(C17, 12), 1,18-octadecanedioic acid (C18, 10), and 1,20-
icosanedioic acid (C20, 13). All compounds had the same
side-chain linker, gGlu-2xOEG, and the same insulin backbone,
A14E, B25H, and desB30 (4). Insulin receptor affinities in the
Figure 3. Time course of degradation of insulin analogues by rat
duodenum enzymes (n = 3, mean SD). Results for each analogue
were fitted using a single exponential decay function in Graph Pad
Prism.
620
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Table 2. Insulin Receptor Affinities and Stability in Rat Duodenum Extract of Insulin Analogues
a
HI receptor A isoform binding (% rel HI, 95% CI)
analogue
side chain
none
none
none
none
none
none
C18-gGlu-2xOEG
C18-gGlu-2xOEG
C18-gGlu-2xOEG
C18-gGlu-2xOEG
insulin substitutions
HI
desB30
B25H, desB30
B25H
A14E, desB30
A14E, B25H, desB30
desB30
0% HSA
4.5% HSA
stability in rat duodenum extract (rel. to 4)
1
5
6
2
3
4
7
8
9
100%
100%
0.2
0.2
ND
b
29.6% (26.2−33.4%)
b
ND
0.4
0.3
79.4% (71.6−88.1%)
22.5% (19.4−26.1%)
13.9% (10.7−18.1%)
1.93% (1.49−2.50%)
6.19% (5.38−7.13%)
2.31% (2.15−2.49%)
c
1
1.09% (0.96−1.22%)
0.22% (0.16−0.30%)
0.86% (0.65−1.15%)
0.1
0.3
0.1
B25H, desB30
A14E, desB30
A14E, B25H, desB30
10
0.19% (0.17−0.2 2%)
0.9
a
b
c
Solubilized hIR-A SPA assay, affinities expressed relative to HI, 95% confidence intervals. Not determined. 5−10 fold more stable than HI.
Table 3. Insulin Receptor Affinities, Stability toward Rat Duodenum Extract, and Rat i.v. PK of a Series of Analogues with
a
Different Side Chains
b
HI receptor A isoform binding (% rel HI, 95% CI)
stability in rat duodenum extract (rel. to
rat i.v.
t1/2
analogue
side chain
0% HSA
1.5% HSA
4.5% HSA
4)
11
12
10
13
14
15
16
17
18
19
20
21
C16-gGlu-2xOEG 2.30% (2.04−2.59%)
C17-gGlu-2xOEG 1.96% (1.73−2.21%)
C18-gGlu-2xOEG 2.31% (2.15−2.49%)
C20-gGlu-2xOEG 0.87% (0.61−1.24%)
0.20% (0.17−0.23%) 0.21% (0.16−0.28%)
0.24% (0.19−0.30%) ND
0.11% (0.10−0.12%) 0.19% (0.17−0.22%)
0.09% (0.08−0.11%) 0.23% (0.17−0.30%)
0.01% (0.01−0.01%) 0.01% (0.01−0.02%)
1.3
1.1
0.9
0.5
0.6
1.2
0.9
1.5
1.3
0.5
0.6
0.9
1.3 h
3.2 h
5.7 h
11.9 h
c
c
C18
1.25% (1.06−1.48%)
1.99% (1.68−2.35%)
1.71% (1.43−2.06%)
1.07% (0.93−1.22%)
0.77% (0.61−0.93%)
1.53% (1.29−1.81%)
ND
C18-gGlu
C18-gGlu-OEG
C18-2xgGlu
C18-4xgGlu
C20
0.02% (0.02−0.03%) 0.05% (0.03−0.08%)
6.0 h
5.6 h
c
ND
0.12% (0.08−0.18%)
c
0.04% (0.03−0.05%) 0.04% (0.03−0.06%)
ND
c
c
0.05% (0.04−0.06%) ND
ND
0.01% (0.01−0.02%) <0.01%
13.0 h
c
C20-gGlu
C20-2xgGlu
2.01% (1.73%−2.35%) 0.02% (0.01−0.03%) 0.03% (0.03−0.04%)
0.84% (0.73−0.98%) 0.03% (0.03−0.04%) 0.03% (0.02−0.05%)
ND
12.0 h
c
a
b
Insulin backbone is A14E, B25H, desB30 (4). Solubilized hIR-A SPA assay, affinities expressed relative to HI, 95% confidence intervals. Not
determined.
absence of HSA were similar to the analogues with fatty diacids
C16, C17, and C18 (11, 12, and 10, around 2%), whereas the
analogue with C20 (13) displayed slightly lower affinity
(0.9%). In the presence of 4.5% HSA, the analogues displayed
similar insulin receptor-binding affinities (0.2%). The ana-
logues with the shortest chain lengths (C16, 11 and C17, 12),
displayed slightly higher insulin receptor-binding affinities in
the presence of lower concentrations, 1.5%, of HSA, (0.2%)
than the analogues with the longer chain lengths (C18, 10 and
C20, 13, both 0.1%). This could indicate that in the presence
of less HSA (1.5%), a larger fraction of the analogues with the
shorter fatty diacid chain lengths (C16 and C17) are not
bound by HSA. The relative proteolytic stability is declining
with increasing chain lengths, and the C20 analogue only
displays 0.5-fold relative stability.
Following rat i.v. dosing, the resulting t1/2 (Table 3) of the
analogues displayed a clear picture of increasing t1/2 with
increasing chain lengths (C16, 11, 1.3 h; C17, 12, 3.2 h; C18,
10, 5.7 h; and C20, 13, 11.9 h).
It was decided to only progress further with analogues
having C18- or C20-based fatty diacid containing side chains
because only these would be likely to match the criteria of
having a human PK profile longer than 24 h. The C16-based
analogue 11 had a clearly too short PK to be considered and
the C17-based analogue 12 would, in case of long enough
human PK, be much more expensive to produce than the C18
and/or C20 analogues because the 1,17-hepadecanedioic acid
needs to be prepared from methyl 16-hydroxyhexadecanoate,
as described in the experimental section or alternatively from
1,18-octadecanoic acid in a number of synthetic steps.
Furthermore, it is noted that the proteolytic stability of the
C20 analogue 13 is less than that of the C18 analogue 7 (0.5 vs
0.9-fold, respectively) which potentially could lead to lower
oral bioavailability.
Consequently, two series of analogues were made with C18-
and C20-based fatty diacids.
In C18 context, analogues were made with varying linker
lengths and linker compositions. From Table 3, it can be seen
that insulin receptor affinity in the absence of HSA is quite
unaffected of truncations in the linker of 10 as analogues with
the linkers gGlu-OEG (16), gGlu (15) and no linker (14) all
display similar affinities. In the presence of 4.5% HSA, there is
a clear drop in receptor affinity as the linker is truncated.
Receptor affinities decreased as follows: gGlu-2xOEG (10,
0.2%), gGlu-OEG (16, 0.12%), gGlu (15, 0.05%), and no
linker (14, 0.01%). The presence of a hydrophilic linker
improves proteolytic stability. Following rat i.v. PK, the series
with the increasingly truncated linkers displayed similar PK
properties, as t1/2 are quite similar: gGlu-2xOEG (10, 5.7 h),
gGlu-OEG (16, 5.6 h), and gGlu (15, 6.0 h). Analogues with
linkers composed of 1, 2, and 4 gGlu residues (15, 17, and 18,
respectively) display similar in vitro properties, maybe with
621
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Table 4. Insulin Receptor Affinities, Stability toward Rat Duodenum Extract, and Rat i.v. PK of a Series of Analogues with
a
Different Compositions of Side-Chain Linkers
b
HI receptor A isoform binding (% rel HI, 95% CI)
analogue
side chain
None
C18-gGlu
C18-gGlu-2xOEG
C20-gGlu
C20-gGlu-2xOEG
0% HSA
1.5% HSA
4.5% HSA
stability in rat duodenum extract (rel. to 4) rat i.v. t1/2
c
c
c
22
23
24
25
26
19.4% (15.7−24.1%) ND
2.22%(2.03−2.46%)
2.97% (2.66−3.32%) 0.23% (0.21−0.27%) ND
1.29% (0.98−1.69%) ND
ND
3.1
5.8
4.3
3.4
2.3
ND
0.03% (0.02−0.03%) 0.03% (0.03−0.04%)
6.0 h
5.6 h
14.0 h
9.1 h
c
c
0.03% (0.02−0.05%)
1.46% (1.21−1.75%) 0.25% (0.22−0.29%) 0.38% (0.32−0.45%)
a
b
The insulin backbone is A14E, B25H, desB27, desB30 (22). Solubilized hIR-A SPA assay, affinities expressed relative to HI, 95% confidence
c
intervals. Not determined.
slightly increased insulin receptor-binding affinities in the
presence of 1.5% HSA with increasing number of gGlu
residues in the linker (0.02, 0.04, and 0.05%, respectively).
In C20 context, a similar series was made with various
linkers (see Table 3). In the presence of 4.5% HSA, the insulin
receptor affinity dropped significantly by truncation of the
linker: gGlu-2xOEG (13, 0.18%), gGlu (20, 0.03%), and no
linker (19, <0.01%). In the presence of 1.5% HSA, the affinity
of 19 is 0.01%. Proteolytic stabilities of these C20-based
analogues were found to be similar (0.5−0.6-fold relative) and
lower than the stability of the comparable C18-based
analogues (0.9−1.2-fold relative). The rat i.v. PK profiles of
the C20-based analogues were shown to be significantly more
protracted than the C18-based analogues, having t1/2’s of 11.9
h (gGlu-2xOEG, 13), 21 h (gGlu, 20), and 13 h (no linker,
19). Analogues with 1 and 2 gGlu’s in the linker (20 and 21,
respectively) were found to possess similar insulin receptor
affinities.
Based on these studies of acylated insulin analogues with
C18- or C20-based fatty diacids as albumin-binding element, it
was concluded that the shorter the linker, the lower the
functional insulin receptor affinity. For C18-based insulin
analogues, the rat PK was relatively unaffected by the length of
the linker and the t1/2’s were around 6 h. Rat PK of C20-based
analogues was generally approximately 2-fold longer than the
corresponding C18-based analogues. For the C20-based
analogues, the stability in the rat duodenum extract assay
was generally lower than that for the corresponding C18-based
analogues, increasing the risk of lower oral bioavailability of
C20-based analogues.
For further preclinical studies, it was decided to proceed
with analogue 10. No other C18-based analogue was superior
with respect to rat PK, and we chose not to decrease insulin
receptor affinity further by truncating the linker. Also, it was
regarded as a major risk of obtaining too low oral
bioavailability with a C20-based analogue because of lower
proteolytic stabilities, so this class was deselected in spite of
having an appealing PK profile.
In order to further improve the proteolytic stability of
analogue 10, we hypothesized that deleting the ThrB27
(desB27) residue would stabilize the chymotryptic triad (B24-
26: FHY) by placing the proline of B28 next to the tyrosine in
B26 (B-chain C-terminal amino acids: FHYTPK → FHYPK).
The unacylated analogue (22, Table 4) proved to be 3-fold
more stable in the rat duodenum extract assay, as compared to
4. A series of acylated analogues of analogue 22 were made
with both C18- and C20-based side chains containing the
linkers gGlu and gGlu-2xOEG (see Table 4). It was evident
that proteolytic stability was increased relative to the analogues
without the ThrB27 deletion. Thus, for the two analogues with
the side chain C18-gGlu (15 and 23), stability was increased
from 1.2 to 5.8-fold relative; for the two analogues with the
side chain C18-gGlu-2xOEG (10 and 24), stability was
increased from 0.9 to 4.3-fold relative; for the two analogues
with the side chain C20-gGlu-2xOEG (13 and 26), stability
was increased from 0.5 to 2.3 fold relative; and for the two
analogues with the side chain C20-gGlu (20 and 25), stability
was increased from 0.6- to 3.4-fold relative. It was also evident
that omission of the 2xOEG part of the linker caused a
significant loss of insulin receptor affinity when assayed in the
presence of HSA. In the absence of HSA, insulin receptor
affinities were unaffected by the length of the linker. In the case
of analogues with C18-based side chains, rat i.v. PK indicated
that half-life was quite unaffected by the ThrB27 deletion,
whereas in the case of C20-based analogues, PK was shorter
for analogues with the ThrB27 deletion. Thus, for the two
analogues with the side chain C18-gGlu (15 and 23), t1/2 were
5.6 and 6.0 h, respectively, and for the two analogues with the
side chain C18-gGlu-2xOEG (10 and 24), t1/2 were 5.7 and
5.55 h, respectively. For the two analogues with the side chain
C20-gGlu (20 and 25), t1/2 were 21 and 14 h, respectively, and
for the two analogues with the side chain C20-gGlu-2xOEG
(13 and 26), t1/2 were 11.9 and 9.1 h, respectively. These
differences are likely to arise from higher insulin receptor
affinities in the presence of HSA of the analogues with the
ThrB27 deletion, especially in the case of C20-based
analogues, giving rise to faster insulin receptor-mediated
clearance.
Based on these data, it was decided to progress with
analogue 24 as a backup candidate to analog 10 to mitigate the
risk of 10 having too little stability toward proteolytic
degradation. Consequently, the C20-based analogues and 23
were not considered.
The HSA-binding affinities of analogues 10, 24, and of the
nonproteolytically stabilized comparator 7 were measured24
and compared to the HSA-binding affinities of insulin degludec
and of insulin detemir (Figure 1). The latter was used as an
assay standard, and the reported HSA-binding affinities are
expressed as fold relative to the affinity of detemir to HSA. The
data are presented in Table 5. Even though 10 only displays
Table 5. HSA-Binding Affinities Measured24 Relative to
Insulin Detemir
analogue
HSA affinity25 (fold relative to detemir)
10
4.2
24
8.9
7
7.9
2.421
Degludec
622
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
half of the HSA-binding affinity relative to that of 7, the PK is
significantly longer. As 24 has roughly the same HSA affinity as
7, this indicates that the long PK profile is mediated also by
other properties than HSA binding. Our hypothesis is that the
significantly lowered insulin receptor-binding affinity contrib-
utes significantly to this aspect.
10 and 24 were further characterized in various animal
models and compared to the similar nonproteolytically
stabilized analogue 7, as will be discussed in the following.
We have verified that the glucodynamic (i.e., blood glucose
lowering) potencies of analogues 10 and 24 have not been
compromised by the extended PK and/or the lowered insulin
receptor affinities. The analogues were dosed (∼33 nmol/rat)
by subcutaneous injection to Sprague-Dawley rats, and blood
glucose was measured up to 48 h after dosing (see Figure 4).
The two analogues potently lowered blood glucose.
affinities for human, rat, and dog insulin receptors of 10 and 24
are not different (data not shown).
10 and 24 both demonstrated encouraging PK and PD
profiles, and the remaining question was whether the analogues
were orally available to the desired extent. As previously
published,14 this was indeed the case. 10, 24, and the
comparator compound 7 were formulated in film-coated
(Opadry II Yellow) gastro-intestinal permeation enhancement
technology one tablets (GIPET I)17 with the absorption-
enhancer sodium caprate (550 mg) and dosed orally to dogs
(Table 7). Bioavailability (relative to i.v. dosing) was found to
Table 7. Bioavailability Following Oral Administration of
Insulin Analogues Formulated in GIPET I Tablets to Dogs
a
insulin
substitutions
fatty
diacid
bioavailability relative
to i.v. dosing mean
analogue
linker
7
desB30
C18
C18
C18
gGlu-2xOEG 0.31%
(CV 75%, n = 8)
gGlu-2xOEG 3.0%
(CV 100%, n = 40)
gGlu-2xOEG 3.9%
(CV 78%, n = 63)
10
24
A14E, B25H,
desB30
A14E, B25H,
desB27,
desB30
13
A14E, B25H,
desB30
C20
gGlu-2xOEG 0.43%
(CV 63%, n = 8)
b
a
The oral doses were 30 nmol/kg for 7 and 13 and 60 to 120 nmol/
b
kg for 10 and 24. Uncoated tablets.
be 3.0% (CV 100%, n = 40, 10) and 3.9% (CV 78%, n = 63,
24) vs 0.31% (CV 75%, n = 8, 7), respectively. A representative
PK/PD profile of 24, dosed orally, is shown in Figure 5. The
large variability observed in both the PK and PD profiles has
been reported14 to be significantly reduced upon multiple
dosing to steady state. As a prototype C20-based analogue, we
also tested 13 orally in dogs formulated in uncoated GIPET I
tablets. Bioavailability was found to be 0.43% (CV 63%, n = 8,
13), significantly lower than that of 10 and 24. This could
indicate that C20-based analogues display lower bioavail-
abilities than corresponding C18 based analogues, 10 and 24.
These dog data were highly encouraging as pharmaceutical and
clinical meaningful bioavailabilities were obtained on insulin
analogues that displayed the desired PK and PD properties.
Figure 4. Blood glucose changes after a single s.c. dose of 10 (32
nmol/rat), 24 (34 nmol/rat), or vehicle to fed Sprague-Dawley rats, n
= 5 in each group.
Furthermore, these two analogues have been tested in
hyperinsulinemic, euglycemic clamp studies in various animal
models (rats, dogs, and pigs) and have both been shown to be
equipotent with HI, as measured by glucose infusion rates.14
Because both 10 and 24 displayed promising rat PK and
pharmacodynamic (PD) properties, we decided to test these
analogues intravenously in dogs along with the nonproteolyti-
cally stabilized comparator analogue 7. Some of the obtained
PK parameters are shown in Table 6 (i.v. t1/2 data of these
CONCLUSIONS
■
In conclusion, we have engineered insulin analogues that
fulfilled all the design criteria, that we had set up to engineer a
safe, orally available basal insulin (proteolytic stability, strong
and reversible albumin binding, and low insulin receptor
affinity). Stability toward proteolytic degradation was achieved
by introducing two insulin substitutions, A14E and B25H. The
latter substitution also mediated the required reduction of
insulin receptor affinity, an affinity reduction that was further
lowered by the side chain. An extensive SAR exploration of the
albumin-binding side chain, attached to the lysine in position
B29, was conducted. We found that the fatty diacid moiety
should be derived from octadecanedioic or icosanedioic acid
(C18 or C20) in order to provide the adequate long PK
profile. Furthermore, the linker connecting the fatty diacid with
the lysine in position B29 was shown to be needed and that
longer linkers are better than shorter linkers. Oral bioavail-
ability of the analogues was shown to be much higher for the
C18-based analogues, as compared to the C20-based
analogues. The identified lead analogues, 10 and 24, both
Table 6. Intravenous PK Parameters for 10, 24, and 7 in
Dogs, Dosed 2.0 nmol/kg
a
clearance, CL
(mL/kg/min SD)
MRT
(h, SD)
t1/2 (h,
analogue
SD)
10 (n = 4)
24 (n = 3)
7 (n = 4)
0.074 0.014
0.057 0.010
0.188 0.017
28
19
6
3
1
0.6
24 0.4
17
1
5.7 0.3
a
Harmonic mean.
compounds have previously been reported14). Both 10 and 24
displayed mean residence times (MRTs) around 24 h in
alignment with decided criteria for candidates for oral use.
These data also underline the importance of introducing the
A14E and B25H substitutions for obtaining the desired long
PK profiles as the analogue without these substitutions (7) has
a PK profile that is clearly too short. It is also noted that
623
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 5. Dog PK and PD profiles (mean SEM, n = 8) for 24 after oral dosing of 60 nmol/kg in GIPET tablets. (A) Mean PK profiles SEM.
(B) Mean PK individual profiles, first 6 h. (C) PD profiles (blood glucose lowering) SEM. (D) Individual PD profiles, first 6 h.
dried in vacuo to afford 91.5 g (77%) of methyl 16-((methylsulfonyl)-
oxy)hexadecanoate (2) as a white solid.
contain the albumin-binding side chain of semaglutide, and
presence of the two insulin substitutions A14E and B25H was
needed in order to confer oral bioavailability and the required
ultralong PK and PD profiles to reduce variability at steady
state. The two identified lead analogues, 10 and 24, represent
the optimal compromise with regard to insulin backbone
substitutions, side-chain composition, in vivo pharmacokinetics
and in vivo dynamics, and oral bioavailability in a conventional
enhancer system. However, the I338 doses were high in
human, challenging commercial access. Recent technological
developments within the arena of oral delivery of insulin, and
particularly progress within oral delivery devices25 together
with the tailor-made insulin molecules described here, may
facilitate oral insulin treatment in the future.
LC−MS m/z: 365.5 (M + H)+. 1H NMR (300 MHz, CDCl3, δH):
4.22 (t, J = 6.51 Hz, 2H) 3.66 (s, 3H); 3.00 (s, 3H); 2.30 (t, J = 7.52
Hz, 2H); 1.80−1.68 (m, 2H); 1.68−1.54 (m, 2H); 1.25 (s, 22H).
Methyl 16-Cyanohexadecanoate. Methyl 16-((methylsulfonyl)-
oxy)hexadecanoate (91.0 g, 250 mmol) was dissolved in dimethyl
sulfoxide (DMSO) (800 mL) and heated to 40 °C. To the solution
was added KCN (24.4 g, 374 mmol), and the mixture was stirred
overnight. Then, the mixture was heated to 75 °C and cyclohexane
(800 mL) was added under vigorous stirring. After 10 min, the upper
cyclohexane layer was separated in a separation funnel. Separation was
repeated three times. The organic phases were combined, washed
with water (3 × 200 mL), dried over anhydrous Na2SO4, and
evaporated in vacuo to obtain 69.5 g (94%) of methyl 16-
cyanohexadecanoate as a white solid. LC−MS m/z: 296.5 (M +
H)+. 1H NMR (300 MHz, DMSO-d6 δH): 11.96 (br s, 2H); 2.18 (t, J
= 7.52 Hz, 4H);1.55−1.41 (m, 4H); 1.26 (s, 22H).
EXPERIMENTAL SECTION
■
1,17-Heptadecanedioic Acid. Methyl 16-cyanohexadecanoate
(69.0 g, 234 mmol) was dissolved in mixture of tetrahydrofuran
(THF)/MeOH/water (1:1:0.3; 1.6 L), and KOH (78.5 g, 1.40 mol)
was slowly added. The mixture was refluxed for 48 h. The organic
solvent was evaporated and water (500 mL) was added, and the
mixture was refluxed for 24 h. The mixture was cooled to room
temperature and acidified with 10% solution KHSO4 to pH 2. The
white precipitate was filtered off, then triturated with water (2 × 1 L),
filtered off, and dried in vacuo to afford 68.5 g (97.6%) of 1,17-
heptadecanedioic acid. LC−MS m/z: 301.6 (M + H)+.
Heptadecanedioic Acid Mono-tert-butyl Ester. 1,17-Heptadeca-
nedioic acid (68.0 g, 226 mmol) and 4-dimethylaminopyridine
(DMAP) (3.25 g, 26.3 mmol) were dissolved in a mixture of
THF/tBuOH (2:8, 1.6 L) under moderate heating (50 °C) to give
creamy solution (some insoluble residue). Di-tert-butyl dicarbonate
(49.4 g, 226 mol) solution in THF (100 mL) was added dropwise at
room temperature, and the resulting mixture was stirred for 24 h.
All final compounds were analyzed by RP-HPLC (key compound
purities >95% in two systems, acidic and neutral), and identities were
verified by LC−MS (electrospray).
Expression, Cultivation, and Purification of Insulin Ana-
logues. Insulin precursors were expressed in yeast (Saccharomyces
cerevisiae) and subsequently converted into two-chain insulin desB30
analogues, as previously described.26
Synthesis of Side Chains. Building Blocks for Side Chains.
1,17-Heptadecanedioic Acid Mono-tert-butyl Ester. A mixture of
methyl 16-hydroxyhexadecanoate (91.0 g, 318 mmol) and Et3N (44.7
mL, 317 mmol) was dissolved in dry dichloromethane (DCM) (800
mL). The solution was cooled to 0 °C, and methansulfonylchloride
(25.2 mL, 334 mmol) was added dropwise over 30 min. The mixture
was stirred at 0 °C for 2 h. 1 M HCl (500 mL) was added, and the
phases were separated. This was repeated two times. The organic
phase was dried over anhydrous Na2SO4 and evaporated in vacuo. The
crude product was crystallized from hot cyclohexane (200 mL) and
624
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Then, the solvents were evaporated, and the residue was redissolved
in DCM (400 mL). The unreacted starting material, heptadecanedioic
acid, begun to precipitate and then it was filtered off. The solvent was
evaporated in vacuo, and the residue was purified by flash column
chromatography (Silicagel 60, 0.040−0.063 mm; eluent: toluene first
and the next with DCM/MeOH 98:2) to afford 22 g (27%) of
heptadecanedioic acid mono-tert-butyl ester as a white solid. LC−MS
m/z: 357.5 (M + H)+. 1H NMR (300 MHz, CDCl3, δH): 2.35 (t, J =
7.52 Hz, 2H) 2.25 (t, J = 7.52 Hz, 2H); 1.69−1.52 (m, 4H); 1.45 (s,
9H); 1.26 (s, 22H).
The synthesis of C16-, C18-, and C20-diacid mono-tert-butyl esters
was made accordingly, starting from the corresponding diacids.
Solid-Phase Synthesis of Side Chains. The general method for
preparation of tert-butyl protected side chains on the solid support is
illustrated by the following procedure for preparation of I, II, and III.
All the other side chains were prepared accordingly:
(30 mL) was added, and the insoluble oil was isolated by decantation
and evaporated in vacuo. This afforded 1.1 g of tert-butyl-protected
C18-gGlu-2xOEG-OH (I) as an oil. LC−MS: m/z = 846.6 (M + 1).
1H NMR (CDCl3) is given in the Supporting Information.
OSu-Activation. The above tert-butyl-protected C18-gGlu-2xOEG
(I) (0.63 g) was dissolved in THF (35 mL). DIEA (0.255 mL, 2
equiv) was added followed by TSTU (0.45 g, 2 equiv), and the
mixture was stirred at room temperature for 16 h. The mixture was
partitioned between ethyl acetate (250 mL) and aqueous NaHSO4 (3
× 100 mL). The organic phase was dried (MgSO4) and concentrated
in vacuo to afford 0.65 g of tert-butyl-protected C18-gGlu-2xOEG-
OSu (II) as an oil, which was used without further purification. LC−
MS: m/z = 943.4 (M + 1). 1H NMR (CDCl3) is given in the
Supporting Information.
TFA Deprotection. The above tert-butyl-protected C18-gGlu-
2xOEG-OSu (II) (10 g) was dissolved in TFA (50 mL) and stirred
at room temperature for 1 h. The mixture was concentrated in vacuo,
and the residue was stripped twice with toluene and twice with 2-
propanol. The residue was crystallized from 2-propanol (75 mL),
filtered, and dried in vacuo to afford 6.6 g of unprotected C18-gGlu-
2xOEG-OSu (III). LC−MS: m/z = 831.7 (M + 1). 1H NMR
(CDCl3) is given in the Supporting Information.
1.0 g of 2-chlorotrityl polystyrene resin, 1.60 mmol/g, was swelled
for 30 min in DCM (10 ml).
Attachment of Fmoc-OEG-OH. 0.39 g (0.63 equiv, 1.0 mmol) of
Fmoc-8-amino-3,6-dioxaoctanoic acid (Fmoc-OEG-OH) was dis-
solved in DCM (15 mL) and added to the resin. N,N-
Diisopropylethylamine (DIEA) (0.44 mL, 2.5 mmol) was added
dropwise. The reaction mixture was vortexed for 30 min, and then,
MeOH (2 mL) was added and the mixture was vortexed for
additional 15 min. The resin was filtered and washed with NMP (2 ×
8 mL) and DCM (8 × 8 mL). 20% piperidine/NMP (8 mL) was
added, standing for 10 min repeated once [filtered and washed with
NMP (2 × 8 mL), DCM (3 × 8 mL), and NMP (5 × 8 mL)]. A
positive (trinitrobenzesulfonic acid) TNBS test gave red-colored
resins.
Acylation with Fmoc-OEG-OH. 0.78 g (2 equiv, 2.0 mmol) of
(Fmoc-OEG-OH) was dissolved in NMP/DCM 1:1 (10 mL). 0.28 g
(2.2 equiv, 2.4 mmol) of HOSu was added followed by addition of
0.37 mL (2.2 equiv, 2.4 mmol) of DIC. The reaction mixture was
allowed to stand for 1 h and was then added to the resin, and finally
0.407 mL (2.2 equiv) of DIEA was added. The mixture was vortexed
for 16 h, filtered and washed with NMP (2 × 8 mL), DCM (3 × 8
mL), and NMP (5 × 8 mL). A positive TNBS test gave colorless
resins. 20% piperidine/NMP (10 mL) was added, standing for 10 min
repeated once [filtered and washed with NMP (2 × 8 mL), DCM (3
× 8 mL), and NMP (5 × 8 mL)]. A positive TNBS test gave red-
colored resins.
Acylation of Insulins. The acylated analogues were all
synthesized similarly, as described in the following examples, and
purities and identities of the products were established by HPLC and
LC−MS. Purities were determined by two methods at acidic pH and
at neutral pH. Purities in both systems were >95% for all key
compounds.
Acylation Using Protected Side Chains. Preparation of 7: 3.15 g of
desB30 HI was dissolved in 30 mL of 100 mM Na2CO3 and 30 mL of
acetonitrile, pH 10. 650 mg (1.25 mol equiv) side chain (tBu-
protected C18-gGlu-2xOEG-OSu) was dissolved in 10 mL of
acetonitrile and slowly added with stirring to the insulin. After 1 h,
the mixture was diluted with water (170 mL) and pH adjusted to 5.2
with 1 M HCl. After standing at 5 °C for 1 h, the precipitate was
isolated by centrifugation and decantation and dried by lyophilization.
This product was dissolved in 120 mL of TFA and left for 30 min.
The mixture was poured into 600 mL of ice-cold Et2O. The
precipitate was isolated by centrifugation and decantation and further
washed twice with Et2O and dried. This product was purified by
preparative RP-HPLC using 0.1% TFA and a gradient of acetonitrile
in water. Purest fractions were pooled and lyophilized and repurified
by anion exchange chromatography using a gradient of NH4OAc
(0.25 to 1.25%) in 42.5% ethanol containing 0.25% TRIS, pH 7.5.
Column: Ressource Q. Pure fractions were pooled, and the product
was isolated by isoelectric precipitation by adjusting pH to 5.2 with 1
M HCl, followed by centrifugation and decantation. Lyophilization
afforded 7, 81 mg.
Acylation Using Unprotected Side Chains. Preparation of 11: 2 g
of 4 was dissolved in 35 mL of 0.1 M Na2CO3. pH was adjusted to
11.2 using 1 M NaOH. 1.5 mol equiv side chain (C16-gGlu-2xOEG-
OSu), dissolved in 2 mL of DMF, was added dropwise, while keeping
pH constant at 11.2 by simultaneous titration with 1 M NaOH. The
reaction mixture was acidified to pH 1.8 using 1 M HCl and purified
by preparative RP-HPLC using a gradient of 0.1% TFA in acetonitrile
in 0.1% TFA in water. Pure fractions were lyophilized to afford 11
(547 mg, 24%).
In Vitro Binding Assays. In vitro binding assays: Competitive
displacement by insulin analogues of 125I-labeled HI from solubilized
HI receptor isoform A was performed, as previously described.22
Insulin receptor affinities for insulin analogues were reported relative
to HI [EC50(HI)/EC50(analogue) × 100%] within each plate. The
human serum albumin-binding assays using 125I-labeled insulin
analogues were accomplished, as published,24 and affinities were
reported relative to insulin detemir.
Enzyme Inhibition Assays. Pepsin. Proteolytic stability of HI
and insulin analogues (0.6 mM, 10 μL) toward pepsin (13.6 μL of 1
μg/μL) was measured after incubation in approx. 35 mM HCl (pH
1.1−1.5) and 25 °C at a final volume of 400 μL. At various time
points (5, 15, 30, and 60 min), samples were quenched with an equal
Acylation with Fmoc-Glu-OtBu. 0.86 g (2 equiv, 2.0 mmol) of
Fmoc-Glu-OtBu was dissolved in NMP/DCM 1:1 (10 mL). 0.32 g
(2.2 equiv, 2.4 mmol) of 1-hydroxybenzotriazole (HOBT) was added
followed by addition of 0.37 mL (2.2 equiv, 2.4 mmol) of DIC. The
reaction mixture was allowed to stand for 20 min and was then
transferred to the resin, and finally 0.407 mL (2.2 equiv) of DIEA was
added. The mixture was vortexed for 16 h, filtered and washed with
NMP (2 × 8 mL), DCM (3 × 8 mL), and NMP (5 × 8 mL). A
positive TNBS test gave colorless resins. 20% piperidine/NMP (10
mL) was added, standing for 10 min repeated once [filtered and
washed with NMP (2 × 8 mL), DCM (3 × 8 mL), and NMP (5 × 8
mL)]. A positive TNBS test gave red-colored resins.
Acylation with 1,18-Octadecanedioic Acid Mono tert-Butyl
Ester. 0.75 g (2 equiv, 2.0 mmol) of 1,18-Octadecanedioic acid
mono tert-butyl ester was dissolved in NMP/DCM 1:1 (10 mL). 0.32
g (2.2 equiv, 2.4 mmol) HOBT was added followed by addition of
0.37 mL (2.2 equiv, 2.4 mmol) of DIC. The reaction mixture was
allowed to stand for 20 min and was then transferred to the resin, and
finally 0.41 mL (2.2 equiv) of DIEA was added. The mixture was
vortexed for 16 h, filtered and washed with NMP (2 × 8 mL), DCM
(3 × 8 mL), and NMP (5 × 8 mL).
Cleavage with TFA. 8 mL of 5% TFA/DCM was added to the
resin, and the reaction mixture was vortexed for 2 h, filtered, and the
filtrate was collected. More 5% TFA/DCM (8 mL) was added to the
resin, and the mixture was vortexed for 10 min, filtered, and the resin
was washed with DCM (2 × 10 mL). The combined filtrates and
washings were pH adjusted to basic using about 800 μL of DIEA. The
mixture was evaporated in vacuo affording an oil (3.5 g). Diethylether
625
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
volume of 0.1 M Tris−HCl (pH 8.0) and cooled to 5 °C. HI and
insulin analogues were immediately analyzed by RP-HPLC at 214 nm,
and the area under the peak corresponding to intact protein was
determined. Half-lives (t1/2) were obtained from the curves, and the
fold increase/decrease compared to HI was calculated (stability
relative fold).
intact insulin following oral administration divided by the plasma
concentration of intact insulin after i.v. administration.
Quantification of Insulin Concentrations in Plasma Sam-
ples. For pharmacokinetic assessment, insulin plasma concentrations
were analyzed, as previously described.14
Chymotrypsin and Carboxypeptidase A. Proteolytic stability of
HI and insulin analogues (0.6 mM, 10 μL) toward chymotrypsin
(0.34 or 3.4 μg, 3.4 μL of 0.1 or 1 μg/μL) or carboxypeptidase A (2.5
U) was measured after incubation in 5 mM sodium phosphate, 140
mM sodium chloride, 70 ppm Tween 20 (pH 7.4) and 37 °C at a final
volume of 100 μL. At various time points (0, 5, 15, 30, and 60 min),
samples were quenched with an equal volume of 0.2% TFA and
cooled to 5 °C. HI and insulin analogues were immediately analyzed
by RP-HPLC at 214 nm, and the area under the peak corresponding
to intact protein was determined. Half-lives (t1/2) were obtained from
the curves, and the fold increase/decrease compared to HI was
calculated (stability relative fold).
ASSOCIATED CONTENT
■
sı
* Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.jmedchem.0c01576.
Compound purity data, MS data, 1H NMR data for I, II
and III, and experimental procedures for solution-phase
synthesis of unprotected, OSu-activated side chains
(PDF)
Rat Duodenum Stability Assay. Degradation of insulin
analogues using duodenum lumen enzymes (prepared by aqueous
wash and filtration of duodenum lumen content) from SPRD rats.
The assay is performed by a robot in a 96-well plate (2 mL) with 16
wells available for insulin analogues and standards. Insulin analogues
∼15 μM are incubated with duodenum enzymes in 100 mM 4-(2-
hydroxyethyl)-1-piperazineethanesulfonic acid (pH = 7.4) at 37 °C.
Samples are taken after 1, 15, 30, 60, 120, and 240 min, and the
reaction quenched by addition of 10% TFA. Intact insulin analogues
at each point are determined by RP-HPLC. Degradation half time is
determined by exponential fitting of the data and normalized to half
time determined for the reference insulins, A14E, B25H, desB30, HI,
in each assay. The amount of enzymes added for the degradation is
such that the half time for degradation of the reference insulin is
between 60 min and 180 min. The result is given as the degradation
half time for the insulin analogue in rat duodenum divided by the
degradation half time of the reference insulin from the same
experiment (relative degradation rate).
In Vivo Studies. Pharmacokinetic studies were conducted in rats
and dogs, in accordance with the Protection of Animals Act, the Act
on Experiments on Animals, and the Standard Operating Procedures
for Experiments on Animals at Novo Nordisk A/S. The experiments
were performed under the supervision and approval of the Danish
Government Animal Experiments Inspectorate and the Novo Nordisk
Ethical Review Counsel.
Rat i.v. Male, nonfasted awake SPRD rats are given a single i.v.
dose of insulin (2 μM insulin, 5 mM phosphate, 140 mM sodium
chloride, 70 ppm polysorbate 20, pH 7.4) in the tail vein, and plasma-
concentration time profiles are followed for 2 days after dosing with
frequent blood sampling.
Rat s.c. Insulin analogues are administered subcutaneously once at
a dose of 33 nmol/animal. The rats are weighed individually and are
divided into groups of 5 of similar mean and standard deviation.
Blood glucose is measured by the glucose oxidase method using a
Biosen-S analyzer (EKF Diagnostic, Germany), and is assessed before
dosing and at 2, 4, 8, 24, and 48 h after dosing.
Dog i.v. Male, trained Beagle dogs, weighing 9−15 kg, are given a
single i.v. 2 nmol/kg dose of insulin (40 μM insulin, 5 mM phosphate,
140 mM sodium chloride, 70 ppm polysorbate 20, pH 7.4) and
plasma-concentration time profiles are followed for 7 days after dosing
with frequent blood sampling. During periods of frequent sampling,
the blood samples were taken from Venflon catheters in cephalic veins
kept open with heparinized saline, and the other blood samples were
taken from a jugular vein.
Dog GIPET I Tablets Oral. Tablets were prepared, according to
standard pharmaceutical techniques. This ensured content uniformity
and technical properties within pharmacopoeial limits. The tablet core
composition was as follows: active ingredient (dose-dependent),
sodium caprate (550 mg), sorbitol (149 mgAPI dose), and stearic
acid (3.5 mg). The tablets were coated with 4.5% opadry (PVA
polymer). Tablets were dosed to trained dogs, as previously
described.14 Bioavailability is defined as the plasma concentration of
AUTHOR INFORMATION
■
Corresponding Author
Peter Madsen − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark; orcid.org/
0000-0003-0398-7309; Phone: +45 30754894;
Email: pem@novonordisk.com
Authors
Thomas B. Kjeldsen − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark; orcid.org/
0000-0003-0094-9286
̌
́
Frantisek Hubalek − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
Tina M. Tagmose − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
Lone Pridal − Novo Nordisk A/S, Global Drug Discovery,
DK-2760 Maaloev, Denmark
Hanne H. F. Refsgaard − Novo Nordisk A/S, Global Drug
Discovery, DK-2760 Maaloev, Denmark; orcid.org/0000-
0002-4996-4084
Trine Porsgaard − Novo Nordisk A/S, Global Drug
Discovery, DK-2760 Maaloev, Denmark
Sanne Gram-Nielsen − Novo Nordisk A/S, Global Drug
Discovery, DK-2760 Maaloev, Denmark; orcid.org/0000-
0001-9576-9397
Lars Hovgaard − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
Henrik Valore − Novo Nordisk A/S, CMC API Development,
DK-2880 Bagsvaerd, Denmark
Martin Munzel − Novo Nordisk A/S, Global Research
̈
Technologies, DK-2760 Maaloev, Denmark
Claudia U. Hjørringgaard − Novo Nordisk A/S, Global
Research Technologies, DK-2760 Maaloev, Denmark
Claus Bekker Jeppesen − Novo Nordisk A/S, Global Drug
Discovery, DK-2760 Maaloev, Denmark
̀
Valentina Manfe − Novo Nordisk A/S, Global Drug
Discovery, DK-2760 Maaloev, Denmark
Thomas Hoeg-Jensen − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark; orcid.org/
0000-0003-3882-6345
Svend Ludvigsen − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
Peter Kresten Nielsen − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
Inger Lautrup-Larsen − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
626
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Carsten E. Stidsen − Novo Nordisk A/S, Global Drug
Discovery, DK-2760 Maaloev, Denmark; orcid.org/0000-
0002-6636-3600
Erik M. Wulff − Novo Nordisk A/S, Global Drug Discovery,
DK-2760 Maaloev, Denmark
Patrick W. Garibay − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
REFERENCES
■
(1) Banting, F. G.; Best, C. H. The internal secretion of the pancreas.
J. Lab. Clin. Med. 1922, 7, 251−266.
(2) Rosenfeld, L. Insulin: Discovery and controversy. Clin. Chem.
2002, 48, 2270−2288.
(3) Harrison, G. A.; Cantab, B. C. Insulin in alcoholic solution by
the mouth. Br. Med. J. 1923, 2, 1204−1205.
(4) Herring, R.; Jones, R. H.; Russell-Jones, D. L. Hepatoselectivity
and the evolution of insulin. Diabetes, Obes. Metab. 2014, 16, 1−8.
(5) Gregory, J. M.; Cherrington, A. D.; Moore, D. J. The peripheral
peril: Injected insulin induces insulin insensitivity in type 1 diabetes.
Diabetes 2020, 69, 837−847.
(6) Berger, M. Oral insulin 1922-1992: The history of continuous
ambition and failure. Frontiers in Insulin Pharmacology; Thieme Verlag,
1993; pp 144-148.
́
Janos T. Kodra − Novo Nordisk A/S, Global Research
Technologies, DK-2760 Maaloev, Denmark
Erica Nishimura − Novo Nordisk A/S, Global Drug
Discovery, DK-2760 Maaloev, Denmark
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.jmedchem.0c01576
(7) Arbit, E.; Kidron, M. Oral Insulin: The rationale for this
approach and current developments. J. Diabetes Sci. Technol. 2009, 3,
562−567.
Author Contributions
P.M., T.B.K., F.H., P.W.G., J.T.K., T.M.T., M.M., C.U.H., and
T.H.-J. designed and prepared the analogues and the side
chains. F.H. and P.K.N. performed the enzymatic experiments.
T.B.K. and I.L.-L. made the yeast molecular biology. F.H.,
T.B.K., S.L., P.K.N., and I.L.-L. designed the unmodified
proteolytically stabilized insulins. T.B.K., E.N., and C.E.S.
characterized the analogues in vitro. L.P., H.R., T.P., S.G.-N.,
and E.M.W. performed the in vivo experiments. L.P. and
H.H.F.R. performed the pharmacokinetic calculations. C.B.J.
and V.M. made the insulin plasma assays. L.H. made the
formulations. F.H., H.H.F.R., L.P., P.M., and T.B.K. led and
coordinated the studies. P.M., T.B.K., and F.H. discussed the
data. P.M. wrote the manuscript. All authors have given
approval to the final manuscript.
(8) Heinemann, L.; Jacques, Y. Oral insulin and buccal insulin: A
critical reappraisal. J. Diabetes Sci. Technol. 2009, 3, 568−584.
̈
(9) Zijlstra, E.; Heinemann, L.; Plum-Morschel, L. Oral insulin
reloaded: A structured approach. J. Diabetes Sci. Technol. 2014, 8,
458−465.
(10) Arbit, E.; Kidron, M. Oral insulin delivery in a physiologic
context: Review. J. Diabetes Sci. Technol. 2017, 11, 825−832.
(11) Lopes, M.; Simoes, S.; Veiga, F.; Seica̧ , R.; Ribeiro, A. Why
̃
most oral insulin formulations do not reach clinical trials. Ther.
Delivery 2015, 6, 973−987.
(12) Iyer, H.; Khedkar, A.; Verma, M. Oral insulin - a review of
current status. Diabetes, Obes. Metab. 2010, 12, 179−185.
́
(13) Fonte, P.; Araujo, F.; Reis, S.; Sarmento, B. Oral insulin
delivery: How far are we? J. Diabetes Sci. Technol. 2013, 7, 520−531.
́
(14) Hubalek, F.; Refsgaard, H. H. F.; Gram-Nielsen, S.; Madsen, P.;
Notes
Nishimura, E.; Munzel, M.; Brand, C. L.; Stidsen, C. E.; Claussen, C.
̈
H.; Wulf, E. M.; Pridal, L.; Ribel, U.; Kildegaard, J.; Porsgaard, T.;
Johansson, E.; Steensgaard, D. B.; Hovgaard, L.; Glendorf, T.;
Hansen, B. F.; Jensen, M. K.; Nielsen, P. K.; Ludvigsen, S.; Rugh, S.;
Garibay, P. W.; Moore, M. C.; Cherrington, D. A.; Kjeldsen, T.
Molecular engineering of safe and efficacious oral basal insulin. Nat.
Commun. 2020, 11, 3746.
The authors declare the following competing financial
interest(s): All authors are or were employees of Novo
Nordisk A/S and are or were shareholders of Novo Nordisk A/
S.
(15) Sonne, O. Receptor-mediated endocytosis and degradation of
insulin. Physiol. Rev. 1988, 68, 1129−1196.
ACKNOWLEDGMENTS
■
Dr. Palle Jakobsen is gratefully acknowledged for valuable
inputs to the project. The skillful technical assistance of Dorthe
E. Jensen, Jan L. Sørensen, Annette Nielsen, Morten Nielsen,
Marta A. Ziemba, Randi Dyrnesli, Annette F. Bjerre, Lene S.
Wallander, Gitte Norup, Anne-Lise Rørbech Gudmundsson,
Svitlana Tkach, Jane Hegestand Eriksen, Mette Frost, Lenette
S. Jørgensen, Dorthe Dreyer Andersen, Jette Lenstrup,
Marianne Bojsen Jappe, Trine Rørmose Løjmand, Anja
Benfeldt, Nete Ibsen, Sandra Nichola Bondo Jensen,
Thorbjørn Nordahl Richter-Olesen, Lone Honore, Stine
Valentin Johansen, and Louise Christiansen is gratefully
acknowledged. Also Drs. Vojtech Balsanek and Lubos Jelínek
of Apigenex, Prague, Czech Republic are thanked for syntheses
of some of the side chains.
(16) Ribel, U.; Hougaard, P.; Drejer, K.; Sorensen, A. R. Equivalent
in vivo biological activity of insulin analogues and human insulin
despite different in vitro potencies. Diabetes 1990, 39, 1033−1039.
(17) Ellmerer, M.; Hamilton-Wessler, M.; Kim, S. P.; Dea, M. K.;
Kirkman, E.; Perianayagam, A.; Markussen, J.; Bergman, R. N.
Mechanism of action in dogs of slow-acting insulin analog O346. J.
Clin. Endocrinol. Metab. 2003, 88, 2256−2262.
(18) Leonard, T. W.; Lynch, J.; McKenna, M. J.; Brayden, D. J.
Promoting absorption of drugs in humans using medium-chain fatty
acid-based solid dosage forms: GIPET. Expert Opin. Drug Delivery
2006, 3, 685−692.
́
(19) Halberg, I. B.; Lyby, K.; Wassermann, K.; Heise, T.; Zijlstra, E.;
̈
Plum-Morschel, L. Efficacy and safety of oral basal insulin versus
̌
̌
́
̌
subcutaneous insulin glargine in type 2 diabetes: a randomised,
double-blind, phase 2 trial. Lancet Diabetes Endocrinol. 2019, 7, 179−
188.
(20) Halberg, I. B.; Lyby, K.; Wassermann, K.; Heise, T.; Plum-
ABBREVIATION
̈
Morschel, L.; Zijlstra, E. The effect of food intake on the
■
pharmacokinetics of oral basal insulin: A randomised crossover trial
in healthy male subjects. Clin. Pharmacokinet. 2019, 58, 1497−1504.
(21) Jonassen, I.; Havelund, S.; Hoeg-Jensen, T.; Steensgaard, D. B.;
Wahlund, P.-O.; Ribel, U. Design of the novel protraction mechanism
of insulin degludec, an ultra-long-acting basal insulin. Pharm. Res.
2012, 29, 2104−2114.
C16, 1,16-hexadecanedioic acid; C17, 1,17-heptadecanedioic
acid; C18, 1,18-octadecanedioic acid; C20, 1,20-icosanedioic
acid; CL, clearance; gGlu, gamma-L-glutamic acid; HI, human
insulin; MRT, mean residence time; OEG, 8-amino-3,6-
dioxaoctanoic acid; OSu, N-hydroxysuccinimide (ester);
TSTU, N,N,N′,N′-tetramethyl-O-(N-succinimidyl)uronium
tetrafluoroborate.
(22) Glendorf, T.; Sørensen, A. R.; Nishimura, E.; Pettersson, I.;
Kjeldsen, T. Importance of the solvent-exposed residues of the insulin
627
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
B chain alpha-helix for receptor binding. Biochemistry 2008, 47,
4743−4751.
̈
(23) Lau, J.; Bloch, P.; Schaffer, L.; Pettersson, I.; Spetzler, J.;
Kofoed, J.; Madsen, K.; Knudsen, L. B.; McGuire, J.; Steensgaard, D.
B.; Strauss, H. M.; Gram, D. X.; Knudsen, S. M.; Nielsen, F. S.;
Thygesen, P.; Reedtz-Runge, S.; Kruse, T. Discovery of the once-
weekly glucagon-like peptide-1 (GLP-1) analogue semaglutide. J. Med.
Chem. 2015, 58, 7370−7380.
(24) Kurtzhals, P.; Havelund, S.; Jonassen, I.; Kiehr, B.; Larsen, U.
D.; Ribel, U.; Markussen, J. Albumin binding of insulins acylated with
fatty acids: Characterization of the ligand-protein interaction and
correlation between binding affinity and timing of the insulin effect in
vivo. Biochem. J. 1995, 312, 725−731.
(25) Abramson, A.; Caffarel-Salvador, E.; Khang, M.; Dellal, D.;
́
Silverstein, D.; Gao, Y.; Frederiksen, M. R.; Vegge, A.; Hubalek, F.;
Water, J. J.; Friderichsen, A. V.; Fels, J.; Kirk, R. K.; Cleveland, C.;
Collins, J.; Tamang, S.; Hayward, A.; Landh, T.; Buckley, S. T.;
Roxhed, N.; Rahbek, U.; Langer, R.; Traverso, G. An ingestible self-
orienting system for oral delivery of macromolecules. Science 2019,
363, 611−615.
(26) Kjeldsen, T. Yeast secretory expression of insulin precursors.
Appl. Microbiol. Biotechnol. 2000, 54, 277−286.
628
https://dx.doi.org/10.1021/acs.jmedchem.0c01576
J. Med. Chem. 2021, 64, 616−628