An Aldehyde Responsive, Cleavable Linker for Glucose Responsive Insulins

Article abstract of DOI:10.1002/chem.202004878

A glucose responsive insulin (GRI) that responds to changes in blood glucose concentrations has remained an elusive goal. Here we describe the development of glucose cleavable linkers based on hydrazone and thiazolidine structures. We developed linkers with low levels of spontaneous hydrolysis but increased level of hydrolysis with rising concentrations of glucose, which demonstrated their glucose responsiveness in vitro. Lipidated hydrazones and thiazolidines were conjugated to the LysB29 side-chain of HI by pH-controlled acylations providing GRIs with glucose responsiveness confirmed in vitro for thiazolidines. Clamp studies showed increased glucose infusion at hyperglycemic conditions for one GRI indicative of a true glucose response. The glucose responsive cleavable linker in these GRIs allow changes in glucose levels to drive the release of active insulin from a circulating depot. We have demonstrated an unprecedented, chemically responsive linker concept for biopharmaceuticals.

Full text of DOI:10.1002/chem.202004878

Accepted Article
Accepted Article
Title: An Aldehyde Responsive, Cleavable Linker for Glucose
Responsive Insulins
Authors: Knud J. Jensen, Karin Mannerstedt, Narendra Kumar MIshra,
Ebbe Engholm, Morten Lundh, Charlotte Stahl Madsen, Philip
J Pedersen, Priska Le-Huu, Søren L Pedersen, Nina Buch-
Månson, Björn Borgström, Thomas Brimert, Lisbeth N Fink,
Keld Fosgerau, and Niels Vrang
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.202004878
Link to VoR: https://doi.org/10.1002/chem.202004878
01/2020

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
An Aldehyde Responsive, Cleavable Linker for Glucose Respon-
sive Insulins
Karin Mannerstedt[a], Narendra Kumar Mishra[a],[c], Ebbe Engholm[a],[c], Morten Lundh[a], Charlotte S.
Madsen[a], Philip J. Pedersen[a], Priska Le-Huu[a], Søren L. Pedersen[a], Nina Buch-Månson[a], Björn
Borgström[b], Thomas Brimert[b], Lisbeth N. Fink[a], Keld Fosgerau[a], Niels Vrang*[a],, Knud J. Jensen*[a],[c]
[a]
K. Mannerstedt, E. Engholm, N. K. Mishra, M. Lundh, C. S. Madsen, P. J. Pedersen, P. Le-Huu, S. L. Pedersen, N. Buch-Mänson, L. N. Fink, K. Fosgerau,
N. Vrang
Gubra ApS,
Hørsholm Kongevej, 11B DK-2970 Denmark
E-mail: niels@gubra.dk
[b]
[c]
B. Borgström, T. Brimert,
Red Glead Discovery AB, Scheelevägen 17, S-223 63 Lund, Sweden
E. Engholm, N. K. Mishra, K. J. Jensen
Department of Chemistry
University of Copenhagen
Thorvaldsensvej 40, DK-1871 Frederiksberg, Denmark
E-mail: kjj@chem.ku.dk
Supporting information for this article is given via a link at the end of the document.
Abstract: A glucose responsive insulin (GRI) that responds to
changes in blood glucose concentrations has remained an elusive
goal. Here we describe the development of glucose cleavable linkers
based on hydrazone and thiazolidine structures. We developed link-
ers with low levels of spontaneous hydrolysis but increased level of
hydrolysis with rising concentrations of glucose, which demonstrated
their glucose responsiveness in vitro. Lipidated hydrazones and thia-
zolidines were conjugated to the LysB29 side-chain of HI by pH-con-
trolled acylations providing GRIs with glucose responsiveness con-
firmed in vitro for thiazolidines. Clamp studies showed increased glu-
cose infusion at hyperglycemic conditions for one GRI indicative of a
true glucose response. The glucose responsive cleavable linker in
these GRIs allow changes in glucose levels to drive the release of
active insulin from a circulating depot. We have demonstrated an un-
precedented, chemically responsive linker concept for biopharmaceu-
ticals.
of the main obstacles in the development of GRIs is the very nar-
row range of glucose control in the order of mM and very narrow
range of circulating insulin, in the order of pM, and hence a ratio
between them in the order of millions.7
Current GRI approaches8,9,10,11,12 can be divided into unimolecular
GRIs and controlled release matrix GRIs. Unimolecular GRIs are
insulin variants with a built-in functionality that in response to in-
creasing glucose concentration, releases an active insulin mole-
cule from an inactive form. Once the glucose concentration de-
creases, release of active insulin from the inactivated form de-
creases or stops. Hoeg-Jensen and coworkers reported an insulin
modified with a polyol moiety and a phenylboronic acid conju-
gated to LysB29 that cross-linked insulin hexamers by formation
of boronate groups from boronic acid and polyol moieties on dif-
ferent hexamers.8 The high-molecular weight complex with boro-
nate linkages disassembled by addition of the polyol sorbitol, thus
releasing the insulin. However, glucose was less efficient in dis-
assembling this complex. Chou et al reported an insulin variant
with a phenylboronic acid at the end of an aliphatic chain attached
to LysB29, the latter to facilitate binding to albumin.9 This insulin
conjugate provides glycemic control in a diabetic mouse model. A
third approach used multivalent mannose conjugated to insulin at
Lys B29 or at Gly A1 by acylation.10 Controlled release matrices
on the other hand are complex biomimetic GRI systems that con-
tain both insulin and proteins directly interacting with glucose, by
binding to it or catalyzing its oxidation. Proteins evaluated so far
include lectins (e.g. concanavalin A) and glucose oxidase.11,13,14
Here we report a fundamentally new glucose responsive cleava-
ble linker that exploits the fact that soluble glucose exists in an
equilibrium between ring-closed forms (hemiacetal), primarily the
α- and β-pyranose, and the open, linear aldehyde form (aldehydo-
D-glucose). We demonstrate that glucose can accelerate the hy-
drolysis of a cleavable bond. The aldehyde group of carbohy-
drates readily reacts with α-nucleophiles such as hydrazides,
Introduction
An estimated 463 M people currently live with diabetes of which
approximately 10% have type 1 diabetes (T1D).1,2 In T1D patients,
little or no insulin is produced by the pancreas, and patients re-
quire daily insulin injections to maintain normoglycaemia. Several
short- and long-acting insulins are marketed and management of
T1D requires regular injections of basal (long-acting) as well as
meal time (short-acting) insulin.3,4 Some basal insulins, such as
insulin glargine5, precipitate after administration and form a sub-
cutaneous depot from which they are gradually released. Others,
like the lipidated insulin degludec, form polymers of hexamers that
leads to slow release from the subcutaneous depot.6 Upon re-
lease lipidated insulins are non-covalently bound to albumin as a
circulating depot that provides further half-life extension.6 While
the basal insulins provide decent 12-24 h glucose control, none
of them are responsive to circulating glucose fluctuations, which
can lead to episodes of hypoglycemia if too much insulin is ad-
ministered. Even with optimal insulin treatment, tight control of
blood glucose is difficult. For decades the development of a glu-
cose responsive insulin (GRI) has remained an elusive goal. One
1
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
Figure 1: (A) Subcutaneous and circulating depots; (B) GRI concept based on a circulating depot, where the GRIs are non-covalently bound to one of the fatty-acid
binding sites in albumin and released by hydrolysis of a cleavable linker; (C) Hydrazone: R1 and R3 are moieties that can be chosen to adjust the reactivity and
glucose responsiveness of the hydrazone (Table 1). R3 is derived from an aldehyde. In addition, R1 and R3 can provide anchor points for insulin and the fatty acid
to form full GRIs (Fig. 2). Thiazolidine: R3 and R5 are moieties that can be chosen to adjust the glucose responsiveness of the thiazolidine (Table 2). In addition, R3
and R5 can provide anchor points for insulin and the fatty acid to form full GRIs (Fig. 2).
which are formed in the cleavage of this bond, to form carbohy-
drate hydrazones. In a series of experiments, we explored both
hydrazones and thiazolidines as the cleavable groups and relied
on lipidation as half-life extension method, as the lipidated insulin
2
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
forms both sub-cutaneous and albumin bound depots that can be
adjusted according to the length of the lipid (Fig. 1a and 1b).
The hydrolyzable, cleavable hydrazone and thiazolidine groups
will in a first equilibrium release a nucleophile, which in a second
equilibrium can react with glucose (Fig. 1b). Glucose can react as
the aldehyde driving the formation of a new hydrazone or a thia-
zolidine, respectively. The reactivity of the linker was adjusted us-
ing electron-withdrawing and -donating substituents, so the linker
is hydrolyzed by glucose in a concentration dependent manner.
Human blood can contain aldehydes other than glucose,
most notably glyceraldehyde and acetaldehyde. However, they
generally occur in low or insignificant amounts, as low molecular
weight aldehydes such as formaldehyde, acetaldehyde, acrolein,
and methylglyoxal are generally toxic compounds and carcino-
A key challenge for a glucose responsive linker is the large
difference in the concentrations of glucose and insulin under
physiological conditions, which are in the range of mM and pM,
respectively.7 This difference is difficult to mimic in an in vitro set-
ting due to limitations in detecting pM concentrations of small mol-
ecules, such as the core structures. We chose to use an UV-
HPLC based analysis and sub-mM concentrations of the core
structures (For an overview of methods for testing glucose re-
sponsiveness see Fig 2S). The glucose reactivity of the core
structures was evaluated with a glucose test, where they wer in-
cutaed in varying concentrations (0.21 mM and 0.42 mM) with or
without glucose (4.2 M) and analyzed by UV-HPLC. The glucose
was added in two different molar ratios, a low ratio (10,000 equiv.
Glc) and a high ratio (20,000 equiv. Glc). Spontaneous hydrolysis
genic15 (see the Section 1.8 in SI for a more extensive discussion). in the absence of glucose was also determined. For the hydra-
zone core structures, the disappearance of the starting material
as well as the formation of hydrazide-glucose adduct were meas-
ured, and the half-life of the compounds calculated. To determine
the differences for slow-reacting hydrazones, the percentage of
Results
The initial step in the cleavage of the hydrazone16 and thiazolidine
groups is the protonation of a hydrazone or amine nitrogen, re-
spectively (Fig. 1c). The pKa of this amine is influenced by nearby
substituents and the reactivity can be adjusted through these sub-
stituents. Furthermore, the hydrolysis of the five-membered thia-
zolidine ring could potentially also release ring-strain, depending
on its substitution pattern. Based on these considerations, we de-
signed an ensemble of hydrazone and thiazolidine core structures
to survey the structural requirements for possible glucose respon-
siveness (i.e. systematic variation of R1, R2, R3 and R4 substitu-
ents (Fig. 1c, Tables 1 and 2, Fig. 2). We chose hydrazides as
precursors for the hydrazones due to their stability. The carbonyl
precursors were in most cases aliphatic or aromatic aldehydes,
with only a few being ketones. The reactivity was further adjusted
by adding electron-withdrawing or electron-donating substituents
to the aromatic side-chains. N-Alkylation of some hydrazones and
a thiazolidine provided additional possibility to adjust the proper-
ties. Finally, the core structures were designed to allow further
extension to lipidated core structures (here called linkers) that
could be coupled to Lys B29 in insulin.
Simple hydrazone core structures were synthesized from
aromatic and aliphatic hydrazide and aldehyde starting materials.
The R3 groups were varied, with 1-3 electron-donating substitu-
ents (-OMe, -NMe2 and -OH) on the benzaldehyde precursors,
while the R1 and R2 groups on the hydrazides were varied less
(Table 1, 1-14). Hydrazone core structures with electron-with-
drawing substituents on the aromatic group were also synthe-
sized but showed no reactivity towards glucose in a HPLC based
assay (data not shown). Acyl hydrazones can be in trans- or cis-
configurations, with the former generally being more stable.17 In
some cases we observed two isomers of the hydrazones, where
the major form likely has the trans-configuration.
formed hydrazide-glucose adduct was also measured at 72 h (Ta-
ble 1).
For the glucose tests, the thiazolidine core structures were
tested with two different glucose ratios, a low ratio (10,000 equiv.)
and a high ratio (20,000 equiv.), as well as incubation without glu-
cose. The thiazolidine-glucose adduct could not be detected, due
to lack of UV activity or slow reaction rate, so the disappearance
of the starting material was measured, and the half-life of the com-
pounds was calculated (Table 2). In summary, we synthesized
hydrazone and thiazolidine core structures of which a majority
showed a glucose responsive cleavage (Table 1 and 2).
The vast majority of glucose is in the ring-closed hemiacetal
form, with only a small fraction in the open-chain aldehydo-form.
Reaching the equilibrium between these forms is relatively slow.
We investigated an additional aldehyde, glyceraldehyde, which
cannot ring-close (4% in aldehyde form, 96% as hydrate; see SI).
The ratio between the aldehyde and the core structures and GRIs,
could therefore be decreased. A glyceraldehyde test was estab-
lished and applied for selected compounds. The core structures
or GRIs were incubated with glyceraldehyde in two different ratios,
20 and 100 molar equivalence (Table 2S). Rewardingly, we ob-
served a glyceraldehyde responsive cleavage.
Full GRIs were evaluated in in vitro experiments to enable a
larger difference in concentration between insulin variants and
glucose. In control experiments to bridge between HPLC-UV
analyses and in vitro measurements of glucose responsiveness,
we synthesized two intermediate insulin variants with a hydrazone
core structure attached, but without a lipid. These insulin conju-
gates (SI IC-1, IC-2) were suitable for HPLC-UV studies with glu-
cose equivalents similar as for the above studied core structures
(Fig. 10S, 11S). We studied the cleavage of the hydrazone link-
age in presence or absence of glucose or glyceraldehyde. Re-
wardingly, both IC-1 and IC-2 showed an increased cleavage in
the presence of Glc and glyceraldehyde after 72 h (Table 3S).
With increasing concentrations of glyceraldehyde (0, 20 and 200
equiv) and glucose (0 and 2000 equiv) an increased cleavage was
observed after 72 h (Table 3S). This showed the hydrazone core
structure retained the glucose responsiveness after coupling to
insulin. IC-1 and IC-2 are regioisomers and their differences in
activity also indicated that the glucose responsiveness could be
adjusted by changing the substituents on the hydrazones.
A smaller set of thiazolidine core structures were synthe-
sized starting from aldehydes and 1,2-amino-thiols derived from
cysteine. Both aliphatic and aromatic aldehydes were used, with
electron-donating substituents such as -OMe on the aromatic
side-chains (Table 2, 15-19). The R- and S-form of the thiazoli-
dines likely have similar reactivity. As controls, N-substituted ox-
azolidines and dioxolane were also synthesized (Table 2, 20 and
21) but were either too labile (half-life less than 20h, dioxolanes)
or too stable (no conversion, N-substituted oxazolidines), for fur-
ther exploration.
3
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
Table 1: Glucose test of hydrazone core structure. Compound (0.21 or 0.42 mM) in 1x PBS buffer containing 10% DMSO, pH 7.4, 37°C; a Glclow 10,000 equiv.
glucose (4.2 M); b Glchigh 20,000 equiv. glucose (4.2 M); c Hydrazone-Glc was analyzed; d The hydrazide formed in the hydrolysis reaction was analyzed by UPLC.
Com-
pound
Core structure
Glucose test
Conversion %
(72h)c
Hydroly-
sis
T (h)
½
(72h, %)d
a
b
a
b
Glclow
Glchigh
Glclow
Glchigh
22
14
79
87
7
17
18
23
-
1
2
3
4
5
6
13
17
4
94
88
42
68
33
99
95
54
78
36
12
18
1
45
21
>72
>72
7
>72
>72
4.4
>72
>72
2.6
3
28
47
89
23
4
8
7
8
40
77
15
22
-
9
>72
>72
10
>72
>72
>72
>72
3
4
10
4
2
1
11
12
>72
>72
>72
>72
5
1
2
1
1
1
13
14
Subsequent studies were performed in vitro to allow higher equiv-
alents of glucose, thus gradually approaching more physiological
conditions.
hydrazone attached to insulin and the cysteine part of the thiazol-
idines attached to insulin. Aldehyde and hydrazide starting mate-
rials with a C18 fatty acid were prepared first, where after the hy-
drazone and thiazolidine linkages were synthesized (see SI). C18
was chosen as the preferred lipidation strategy as Insulin-C18, in
contrast to Insulin- C14, has minimal blood glucose lowering ef-
fects and would thus serve as an ideal GRI depot upon normogly-
cemia, see Fig. 3S.
Acylation of insulin at LysB29 is generally well tolerated; insulin
detemir and insulin degludec are both acylated at this position.3
The Lys side-chain amine can be acylated with high selectivity in
aqueous buffer at pH 10.5.3a,17,18,19 Solubility problems and low
yields for the acylation of LysB29, required the use of organic sol-
vents, which on the other hand lowered the selectivity. Neverthe-
less, pure GRIs were obtained after chromatography (SI: Section
9).
We next set out to select core structures for lipidation (Fig.
1S). Core structures were chosen based on three parameters (i)
low background hydrolysis, (ii) a shorter half-life at high than at
low glucose concentration, and (iii) for some hydrazones, a higher
rate of formation of new glucose hydrazones at high than at low
glucose concentration. The corresponding linkers were synthe-
sized, i.e. core structures with handles for introduction of a fatty
acid and conjugation to HI. The core structures were appended
with additional functional groups to enable attachment of the fatty
acid and eventually insulin. The ease of chemical synthesis was
also considered in the design. The core structures could have two
opposite orientations in the linkers, as the aldehyde part could be
either attached to the insulin side or to the fatty acid (Fig. 2a).
Most of the linkers were synthesized with the hydrazide part of the
4
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
Table 2: Glucose test of thiazolidine core structures. Compound (0.21 or 0.42 mM) in 1x PBS buffer containing 10% DMSO, pH 7.4, 37°C; aGlclow 10,000 equiv.
glucose (4.2 M); bGlchigh 20,000 equiv. glucose (4.2 M). Disappearance of starting material was analysed by UPLC
.
Com-
pound
Core structure
Glucose test
Conversion %
(24h)
Hydroly-
sis (24h,
%)
T (h)
½
a
b
a
b
Glclow Glchigh Glclow Glchigh
15
16
17
11
13
77
71
42
70
5
7
98
94
8
8
9
66
56
75
69
10
9
18
12
19
20
21
60
30
-
>72
>72
0
0
1
100
100
100
In contrast to HI-C18 (HI acylated with C18 lipid at Lys B29; HI-
C18; Fig. 3S), which had no impact on blood glucose (BG) levels,
GRI-2 and GRI-1 both caused a sustained reduction with signifi-
cant effects after 1 and 2 h, respectively. Normoglycemic levels
were ultimately reached after 8 h (Fig. 4a-b). Thus, the data
shows that the two GRIs are potent in T1D model in contrast to
acylated insulin (HI-C18), despite sharing same lipidation strategy
(Fig. 4b), most likely explained by linker hydrolysis. This prompted
us to conduct more sophisticated experiments to evaluate in vivo
glucose responsiveness.
In order to assess the glucose responsiveness in a physio-
logical setting, where blood glucose levels were controllable, we
performed normo- (5 mM) and hyperglycemic (20 mM) clamp ex-
periments in healthy rats (Fig. 9S, graphs showing similar BG for
all groups throughout the study). For additional explanation of the
clamp study see Fig 4S. Somatostatin was infused 30 min prior to
GRI administration to suppress endogenous insulin secretion
(see SI). When comparing GRI-1 and GRI-2 at equimolar concen-
trations (113 nmol/kg) we observed that a lower glucose infusion
rate (GIR) was required in order to achieve normoglycemia (Fig.
4c-d), while an identical GIR was required to achieve hyperglyce-
mia. This prompted us to double the GRI-2 dose aiming for a
matched AUC GIR at normoglycemia. Interestingly, at the double
dose of GRI-2 as compared to GRI-1 we observed that the GIR
required to maintain normoglycemia remained unchanged
(Fig.4c-d), while the GIR required to maintain hyperglycemia was
significantly increased (15% in GIR AUC, Fig. 4d) indicative of a
true glucose response.
For the in vitro studies of the GRI’s we used CHO cells overex-
pressing the human insulin receptor b (hIRb) and a phospho-AK-
Tassay to demonstrate an aldehyde-driven increase in insulin po-
tency. GRIs were pre-incubated with either glyceraldehyde or glu-
cose and added to the cells for the insulin receptor signaling as-
say. Basal hydrolysis was observed for all tested GRIs, causing
slightly lower EC50 values post- than pre-incubation with glucose
for 24 h, however at different rates (Fig. 5S-h). Pre-incubation with
low concentrations of glucose relative to GRI had no effect on the
potency of the GRI’s (data not shown), however, increasing the
equivalents of glucose to GRI, by either increasing the glucose
concentrations (Fig. 3a, GRI-2) or decreasing the GRI concentra-
tion (Fig. 3b, GRI-2), or substituting glucose with glyceraldehyde
(Fig. 3S c-d, GRI-2) led to increased potency of thiazolidine-based
GRIs, confirming that our concept extended to the full GRI mole-
cule. Notably, all tested thiazolidine-based GRIs with the alde-
hyde attached to insulin reacted with glyceraldehyde (SI-Fig. 5S
a-h) and glucose (SI-Fig 6S a-d), whereas none of the tested thi-
azolidine GRIs with reversed orientation (SI-Fig. 7S a-d) or hydra-
zone-based GRIs reacted (Fig. 3e-f), with either with glucose nor
glyceraldehyde. The ratio between the glucose-induced hydroly-
sis and the basal hydrolysis (termed “glucose potency”) was used
to rank the glucose responsiveness of GRIs with GRI-2 showing
the highest score (Fig. 3g-h). Notably, substituting glucose with
glyceraldehyde, resulted in the same overall ranking of thiazoli-
dine-based GRIs (Fig. 3S g-h; Fig. 8S).
We next set out to test the most promising GRIs, GRI-2, (Fig.
3h) and a non-glucose responsive insulin, GRI-1, (Fig. 3e-f) in a
T1D animal using 90% pancreatectomized rats displaying overt
hyperglycemia.20 HI was included as a positive control that had
immediate blood glucose (BG) lowering effect (Fig. 4a-b).
5
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
Figure 2: (A) General GRI structures; the hydrazone and thiazolidine moieties can have two different orientations between insulin and the lipid, with the aldehyde
derivative being attached either to insulin or to the lipid; (B) Full structure of two GRIs.
Discussion
aqueous solutions at pH 7 and 37°C and has been used for bio-
conjugation reactions. The reaction between an aldehyde and a
The first step in realizing this new GRI concept was the ability to hydrazide is reversible, a feature which has been used in dynamic
adjust the reactivity of the cleavable linker. Hydrazones and thia- combinatorial chemistry for exchange reactions between hydra-
zolidines were chosen as core structures for their well-documented zones.24 The exchange with another aldehyde means transfer of
chemistries and their different structures (the former has a central the hydrazide to this aldehyde, forming a new hydrazone. If the
sp2 carbon, the latter a sp3 carbon). The desired reactivity was an new aldehyde is present in excess, the formation of the new hy-
initial hydrolysis that could be followed by reaction of the released drazone can be driven to completion.16 Aldehydes can also react
nucleophile (hydrazide or aminothiol) with glucose. The glucose with divalent nucleophiles such as 1,2-amino-thiols to form thia-
responsiveness would be a higher rate of cleavage of these hydra- zolidines, a five-membered ring.25
zone and thiazolidine core structures at higher glucose concentra-
Hydrazones were derived from hydrazides (i.e. hydrazine
tions. Additionally, an increased rate of hydrolysis at higher glu- with a carbonyl substituent), while thiazolidines were conveniently
cose concentrations, without formation of a new glucose hydra- derived from L-cysteine. Both were reacted with substituted ben-
zone would also support the GRI concept. We prepared core struc- zaldehydes or aliphatic aldehydes. A challenge in designing the
tures that reacted at fast, intermediate, or slow rates. Next, linkers glucose response assays was the required concentration differ-
were assembled that in addition to hydrazones or thiazolidines had ence between the GRI/core structures and glucose, necessitating
a lipid moiety and a carboxylic acid enabling subsequent anchoring detection of the core structures and their cleavage products at
to insulin (see SI).
low concentration. Blood glucose concentrations are 3.9 to 7.1
Glucose in solution exists in an equilibrium between ring- mmol/L in fasting, healthy persons,7 with 0.0040% present as the
closed forms (hemiacetal), primarily the α- and β-pyranose, and linear aldehyde, while insulin concentrations are 45 to 280 pmol/L.
the open, linear aldehyde form (aldehydo-D-glucose). The muta- An additional problem was that at very high glucose concentra-
rotation between the α- and β-pyranose proceeds via the open al- tions the water concentration would decrease markedly. This
dehyde form (ring-chain tautomerism). The mutarotational half-life would be problem as the water is required for hydrolysis step.As
of glucose in pure water is 9.8 min at 37 °C, while in whole blood it an intermediate step we used glyceraldehyde as aldehyde since
is 2.3 min at 37 °C. This increased interconversion could be due to it cannot form a cyclic hemiacetal and thus has a significantly
the presence of mutarotase, phosphate ions, and bicarbonate higher proportion of the aldehydo-form together with its hydrate
ions.21 There is 0.0040% of the linear aldehyde present in this equi- (see SI). This allowed the use of fewer equivalents of the alde-
librium together with 0.0059% of the corresponding hydrate.22 In hyde in in the initial evaluation of hydrazone and thiazolidine. The
chemical reactions, glucose and other carbohydrates can be made first parameter was a relatively low basal hydrolysis in the
to react quantitatively via the open-chain aldehyde form by a con-
tinuous sdisplacement of the equilibrium.23
The aldehyde moiety of carbohydrates readily reacts with α-nu-
cleophiles such as hydrazides to form carbohydrate hydrazones.
This reaction can be carried out in physiological conditions i.e.
6
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
Figure 3: In vitro assay for evaluation of the aldehyde responsiveness of hydrazone and thiazolidine GRI’s towards glyceraldehyde (GAH) or glucose using an
insulin receptor (IR) assay. Human insulin (HI) was included as a positive control. (A) Constant GRI-2 concentration (1µM) incubated with varying glucose concen-
trations for 24 h. (B) Variable GRI-2 concentrations (0.5-10µM) incubated with or without 1M glucose for 24 h. (C) GRI-2 (50µM) incubated for 6 h with or without
1mM GAH. (D) GRI-2 (50µM) incubated for 24 h with or without 1mM GAH. (E) Variable GRI-1 concentrations incubated with or without 1M glucose for 24 h. (F)
GRI-1 (50µM) incubated for 24 h with or without 1mM GAH. (G) Description of glyceraldehyde/glucose potency calculations. (H). GRIs ranked according to glucose
potencies. Data are mean + SEM, n=2.
7
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
Figure 4: In vivo GRI testing. (A) Blood glucose during insulin tolerance test (ITT) in pancreatectomized rats. Rats were re-fed after 480 min. Values expressed as
mean of n = 5-7 + SEM. Dunnett’s test on two-factor linear model with interaction. *: P < 0.05, **: P < 0.01, ***: P < 0.001 compared to Insulin C-18; (B) Blood
glucose during a second ITT. Rats were re-fed after 480 min. Values expressed as mean of n = 7-8 + SEM. Dunnett’s test on two-factor linear model with interaction.
*: P < 0.05, **: P < 0.01, ***: P < 0.001 compared to Insulin C-18; (C) Glucose infusion rate (GIR) from 0 min to 90 min. Values expressed as mean of n = 8-9 +
SEM. Dunnett’s test on two-factor linear model with interaction. *: P < 0.05, **: P < 0.01 compared to GRI-1 113 nmol/kg (low BG), #: P < 0.05, ##: P < 0.01 compared
to GRI-1 113 nmol/kg (high BG); (D) Glucose infusion rate (GIR) AUC from 0-90 minutes. Values expressed as mean of n = 8-9 + SEM. Dunnett’s test one-factor
linear model. #: P < 0.05 compared to GRI-1 113 nmol/kg (high BG). Glucose responsiveness indicated.
8
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
absence of aldehyde. The second parameter was glucose re-
sponsiveness as determined by whether the core structures and
linkers hydrolyzed at higher concentration of glucose and glycer-
aldehyde than at lower concentrations.
The glucose responsiveness of core structures was judged
by three criteria (Tables 1 and 2): (i) the half-lives (T1/2) of the core
structures at low and high glucose ratios (a significant ratio or T1/2
> 72 h); (ii) the conversion to the new glucose hydrazone at low
vs. at high glucose concentration ratios; (iii) for some core struc-
tures an additional third criterion was based on the conversion to
the new hydrazones in the presence of low or high ratios of glycer-
aldehyde.
In the design and evaluation of the full GRI the orientation
of the core structure hydrazones and thiazolidines played a role
(Fig. 2a). The core structure could be oriented in the linker so that
either the aldehyde or the nucleophile (hydrazide, aminothiol, or
glucose adduct) would remain attached to insulin LysB29 after
cleavage. Cleavage of the linker thiazolidine forms an amino-thiol
that could react with glucose to form a new thiazolidine. But a free
thiol attached to insulin could potentially scramble the disulfide
linkages in insulin. However, as active insulin has a short half-life
after release (4-6 min) of HI28, this is less of a concern in vivo.
The in vitro assay was based on measuring phosphorylation
of AKT downstream of activation of the insulin receptor and pre-
incubation of GRIs (5-50 μM) with different concentrations of glu-
cose (0, 0.1, 1 M) or glyceraldehyde (0 or 1 mM) (Fig. 3), for 6 h
or 24 h. Several key learnings were extracted from these con-
trolled experiments. First, basal hydrolysis was seen for all GRIs,
which we later recapitulated in our in vivo setting with pancreatec-
tomized rats. Secondly, the ratio of GRI to glucose (equivalence)
is essential for the concept; increasing the equivalences to levels
that resembles physiological levels drives the hydrolysis, shown
by both glucose and glyceraldehyde. Third, we could evaluate our
GRIs on whether they responded to glucose/glyceraldehye or not
and also their reactivity, which we termed “glucose potency”. We
thus concluded that thiazolidine-, but not hydrazone-based GRIs
indeed were glucose responsive insulins. The ratio between insu-
lin concentration and glucose concentration was approaching
physiological ratios. Thiazolidine-, but not hydrazone-based GRIs,
responded to increased glucose concentrations in vitro. Since
both classes retained potency in pancreatectomized rats, a clamp
study comparing the most glucose-responsive thiazolidine, GRI-
2, to the hydrazone GRI-1. When comparing GRI-2 and GRI-1 at
equimolar concentrations (113 nmol/kg) during hyperglycemia, an
identical GIR AUC was observed. At normoglycemia, a reduced
GIR of GRI-2 compared to GRI-1 was observed (Fig. 4c). When
doubling the dose of GRI-2 the GIR AUC at normoglycemia re-
mained the same. Interestingly, the GIR AUC at hyperglycemic
conditions at double dose was then increased by 15% compared
to both GRI-1 and GRI-2 at low dose. This proved a true glucose
response.
In general, hydrazones derived from aliphatic aldehydes re-
acted very fast with glucose and showed a rate of basal hydrolysis
that was too high to practically make them into full GRIs (Table 1,
1-5). However, N-methylation of these gave slightly more stable
core structures (Table 1, 1 vs 4). In contrast, hydrazones derived
from benzaldehydes were more stable (Table 1, 6-14). The reac-
tivity of the hydrazone was increased by having a para- rather
than a meta-methoxy substituent on the benzaldehyde side (Ta-
ble 1, 7 vs 6 and 10 vs 11). That electron-donation from the sub-
stituents near the hydrazone moiety increased reactivity was fur-
ther shown by increasing the number of ortho/para-substituents
(Table 1, 6, 8, 9). An ortho hydroxy group on the hydrazide gave
a less reactive hydrazone (Table 1, 6 vs 11). Also, substituting a
para-methoxy group on the phenyl with a dimethylamine (result-
ing in an aniline) reduced glucose reactivity (Table 1, 6 vs 12), as
did hydrazones derived from aliphatic hydrazides (Table 1, 13 and
14 vs 6 and 7). Core structures 4, 6, 7, 15, and 16 were also tested
with glyceraldehyde, where they reacted faster than with glucose
(Table 2S).
Thiazolidines derived from benzaldehydes showed a prom-
ising reactivity (Table 2, 15, 16, 19), while thiazolidines derived
from aliphatic aldehydes showed slower reactivity and a lower
rate of hydrolysis (Table 2, 17-18). Electron-withdrawing substitu-
ents were also synthesized; a para-nitro substituted benzalde-
hyde thiazolidine was tested in the glucose test resulting in very
slow reactivity (data not shown). The number of electron-donating
substituents in para- and ortho positions were increased and gave
very reactive thiazolidine core structures (data not shown or SI).
Ortho-hydroxy groups gave side reactions for both hydrazones
and thiazolidines. Acetals reacted too fast and were to hydrolyti-
cally labile (Table 2, 21). Boc-protection of the oxazolidine amine
gave an essentially unreactive structure (Table 2, 20).
We selected hydrazone 7, which had a low hydrolysis of 4%
over 72 h and a significant difference in conversion in response
to low and high glucose, 3% and 28%, respectively. Thiazolidine
15 had a hydrolysis of 20% over 24 h. Hydrazone 7 was derived
from a benzoic acid hydrazide and an alkoxy-benzaldehyde, while
thiazolidinie 15 was derived from cysteine and benzaldehyde.
Full GRI conjugates were obtained by acylation of HI at Lys
B29 with linkers. The coupling reactions were performed in or-
ganic solvents as a precaution to prevent hydrolysis of the linker
during preparation. Subsequent purification of the full GRI conju-
gates was achieved using C-4 reversed-phase columns under
non-acidic conditions, using either medium-pressure high-perfor-
mance flash chromatography or preparative HPLC. That the puri-
fied GRI was acylated at LysB29, rather GlyA1, was confirmed by
reduction of the disulfides26 or enzymatic cleavage with V827 fol-
lowed by MS analysis.
Initially, a GRI with a glucose responsive cleavable linker,
as described here, will be administered by subcutaneous injec-
tions. Ultimately oral administration could be considered. It would
form a subcutaneous depot, from where it would be gradually re-
leased and form a second, circulating, albumin-bound depot, GRI
administration is well adjusted, we anticipate that fewer insulin in-
jections will be required.
This GRI approach directly uses a reactive part of the glu-
cose molecule, which makes it very specific. Thus, this GRI has a
higher selectivity compared to the existing approaches that react
with polyols in general.
Conclusion
We developed an unprecedented glucose responsive chemical
linker that enabled a glucose responsive insulin (GRI). First, we
developed hydrazone and thiazolidine core structures that could
undergo an aldehydo-D-glucose promoted hydrolysis. This was
achieved by tuning the stability and reactivity of the hydrazone
and thiazolidine by adjusting the electron-donating and -withdraw-
ing groups of nearby substituents. The resulting hydrazone and
9
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
[6]
[7]
[8]
[9]
I. Jonassen, S. Havelund, T. Hoeg-Jensen, D. B. Steensgaard, P. O.
Wahlund, U. Ribel, Pharm. Res. 2012, 29, 2104–2114.
M. E. Daly, C. Vale, M. Walker, A. Littlefield, K. G. M. M. Alberiti, J. C.
Mathers, Am. J. Clin. Nutr. 1998, 67, 1186-1196.
thiazolidine core structures hydrolyzed faster when the glucose
concentration increased. Since this was achieved for two different
functional groups (i.e. cleavable groups), this points to the gener-
ality of the concept. The glucose sensitive core structures were
incorporated in lipidated linkers coupled to the LysB29 amine in
HI to form functional GRIs. The thiazolidine-based GRIs showed
glucose responsiveness in vitro. In vivo studies of two GRIs in
90% pancreatomized rats showed a similar profile by lowering the
blood glucose for a sustained period. Lastly, a physiological glu-
cose response was demonstrated for a thiazolidine-based GRI.
While clear differences and trends were identified in the core
structures, the properties of the GRIs fell into three broad groups:
aromatic and aliphatic, as well as N-methylated aliphatic for hy-
drazone. This new GRI concept is based on general organic-
chemical principles and the core structures can likely be further
adjusted for additional improvement of the GRI functionality. We
envision that further fine-tuning of the GRI functionality could be
achieved within the narrow physiological glucose range, based on
the chemical principles disclosed here.
T. Hoeg-Jensen, S. Havelund, P. K. Nielsen, J. Markussen J. Am. Chem.
Soc. 2005, 127, 6158–6159.
D. H-C Chou, M. J. Weeber, B. C. Tang, a. b. Lin, L. S. Thapa, D. Deng,
J. V. Truong, A.B. Cortinas, R. Langer, D. G. Anderson, Proc. Nat. Acad.
Sci. U.S.A, 2005, 112, 2401-2406.
[10] R. Yang, M. Wu, S. Lin, R. P. Nargund, X. Li, T. Kelly, L. Yan, G. Dai, Y.
Qian, Q. Dallas-Yang, P. A. Fischer, Y. Cui, X. Shen, P. Huo, D. Feng,
M. D. Erion, D. E. Kelley, J. Mu, JCI insight 2018, 3, 1–14.
[11] N. A. Bakh, A. B. Cortinas, M. A. Weiss, R. Langer, D. G. Anderson, Z.
Gu, Z., S. Dutta, M. S. Strano, Nat. Chem. 2017, 9, 937–943.
[12] Q. Wu, L. Wang, H. Yu, H., J. Wang, Z. Chen, Chem. Rev. 2011, 111,
7855–7875.
[13] O. Veiseh, B. C. Tang, K. A. Whitehead, D. G. Anderson, R. Langer, Nat.
Rev. Drug Discov. 2014, 14, 45–57.
[14] T. Hoeg-Jensen, S. Ridderberg, S. Havelund, L. Schäffer, P. Balschmidt,
I. B. Jonassen, P. Vedsø, P. H. Olesen, J. Markussen, J. Pept. Sci. 2005,
11, 339–346.
[15] R. P. Dator, M. J. Solivio, P. W. Villalta, S. Balbo, Toxics 2019, 7 (2), 32.
[16] J. Kalia, R. T. Raines, Angew. Chem. Int. Ed. 2008, 47, 7523–7526.
[17] G. Palla, G. Predieri, P. Domiano, C. Vignali, W. Turner, Tetrahedron
1986, 42, 3649–3654.
Experimental Section
[18] N. G. Hentz, Anal. Chem. 1997, 69, 4994-5000.
[19] H. L. Friesen, H. L. The relative reactivity of insulin amino groups as an
indicator of structural accessibility and its use for synthetic approaches
for structure-function studies. In Insulin: Chemistry, Structure and Func-
tion of Insulin and Related Hormones. in In Insulin: Chemistry, Structure
and Function of Insulin and Related Hormones (Brandenburg, D., and
Wollmer, A., Eds.) 125–133 (1980).
Full experimental details for the synthesis of core structure/Insulin
conjugates, characterization for all compound, in vitro assay for
aldehyde responsiveness and in vivo studies for glucose lowering
effects, clamp design/studies included in the Supporting Infor-
mation. All animal experiments were performed according to the
bioethical guidelines of Gubra (Hørsholm, Denmark) and in com-
pliance with internationally accepted laboratory animal care and
use principles under the personal licenses 2017-15-0201-01378
and 2017-15-0201-01383, which was approved by the Danish An-
imal Experimentation Council and issued by The Danish Animal
Experiments Inspectorate.
[20] L. Rossetti, G. I. Shulman, W. Zawalich, R. A. DeFronzo, J. Clin. Invest.
1987, 80, 1037–1044.
[21] J. Okuda, T. Taguchi, A. Tomimur, Chem. Pharm. Bull 1987, 35, 4332–
4337.
[22] Y. Zhu, J. Zajicek, A. S. Serianni, J. Org. Chem. 2001, 66, 6244–6251.
[23] K. Villadsen, M. C. Martos-Maldonado, K. J. Jensen, M. B. Thygesen,
ChemBioChem 2017, 18, 574–612.
[24] D. K. Kölmel, E. T. Kool, Chem. Rev. 2017, 117, 10358–10376.
[25] F. Guillaumie, O. R. T. Thomas, K. J. Jensen, Bioconjug. Chem. 2002,
13, 285–294.
Acknowledgements
[26] J. Gao, M. Mrksich, G. M. Whitesides, F. A. Gomez, Anal. Chem. 1995,
67, 3093–3100.
The Juvenile Diabetes Research Foundation (JDRF) is acknowl-
edged for partly funding this project (Grant number 1-IND-2018-
673-I-X). We appreciate very valuable discussions with Dr.
Sanjoy Dutta and Dr. Jonathan Rosen, JDRF. We acknowledge
the assistance from colleagues Annika Kers, Carsten Lüdtke,
Edvard Torkelsson, Inger Kers, Jakov Ivkovic, Klaus Dreisch, La-
risa Wahlström, Maitham Majeed, Mattias Jönsson and Sanjay
Borhade at Red Glead Discovery AB, Sweden
[27] J. Houmard, Eur. J. Biochem. 1976, 68, 621–627.
[28] W. C. Duckworth, R. G. Bennett, F.G. Hamel, Endocr. Rev. 1998, 19,
608–624.
Keywords: Insulin 1 • Glucose Responsive Insulin 2 • Cleavable
Linker 3 • Aldehyde Responsive 4 • Hydrazine and Thiazoli-
dine 5
[1]
[2]
A. D. Association, Diabetes Care 2004, 27, S5–S10.
I. D. Federation, IDF Diabetes Atlas – 2019. International Diabetes Fed-
eration (2019). doi:10.1289/image.ehp.v119.i03.
[3]
(a) M. Østergaard, N. K. Mishra, K. J. Jensen, Chem. Eur. J., 2020, 26,
8341-8357. (b) L. Moroder, H. J. Musiol, Angew. Chem. Int. Ed. 2017, 56,
10656-10669.
[4]
[5]
A. N. Zykov, J. P. Mayer, R. D. Dimarchi, Nat. Rev. Drug Discov. 2016,
15, 425-439.
R. Hilgenfeld, T. Sicker, M. Dörschug, R. Obermeier, K. Geisen, G.
Seipke, H. Berchtold, Diabetologia 1992, 35.
10
This article is protected by copyright. All rights reserved.

10.1002/chem.202004878
Chemistry - A European Journal
RESEARCH ARTICLE
An Aldehyde Responsive, Cleavable Linker for Glucose Responsive Insulins
463 M peo ple live with diabetes of which approximately 10% have type 1
diabetes. A glucose responsive insulin that responds to fluctuations in blood
glucose concentrations has remained an elusive goal for decades. Here we
describe the chemical development of hydrazones and thiazolidines that are
cleavable in glucose concentration dependent manner. These were used to
construct unprecedented glucose responsive insulins.
Institute and/or researcher Twitter usernames: ((optional))
11
This article is protected by copyright. All rights reserved.