Construction of a photoactivated insulin depot

Article abstract of DOI:10.1002/anie.201207264

Light controlled: A material that allows for insulin to be released in a controlled fashion by using light was prepared. A subcutaneous reservoir of such materials could allow for the non-invasive control of blood sugar. Insulin (blue in picture) was linked to an insoluble resin (green) through a photocleavable linker (red). Native insulin is released following a first-order process in response to pulses of light from an LED. Copyright

Full text of DOI:10.1002/anie.201207264

.
Angewandte
Communications
DOI: 10.1002/anie.201207264
Protein Delivery
Construction of a Photoactivated Insulin Depot**
Piyush K. Jain, Dipu Karunakaran, and Simon H. Friedman*
1404
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2013, 52, 1404 –1409

Angewandte
Chemie
Insulin is a primary tool for the treatment of type I diabetes.[1]
However, multiple challenges accompany its administration.
Because of low oral bioavailability, it is typically injected
multiple times per day, a significant lifetime burden on
patients. Moreover, because of fluctuating blood sugar levels,
the required amounts of insulin continuously vary over the
course of a day. A current solution to the challenge of
multiple daily injections is the use of an insulin pump, which
delivers insulin transdermally through a cannula, an awkward
and limiting solution.[1]
We have conceived of a less invasive approach, in which
light is used to activate the release of insulin from a covalently
linked depot (Figure 1). The elements of the system are
a polymer, a photocleavable linker, and the therapeutic,
insulin. Ideally, the polymer should be insoluble, so that it
remains at the site of injection, and biodegradable, so that
after the majority of the insulin has been released, the
polymer can be cleared from the system.
Insulin has features that make it particularly amenable to
the photoactivated depot (PAD) approach. Because blood
sugar levels vary greatly, the required concentrations of
insulin in the blood also vary greatly from hour to hour.
Moreover, the total volume of insulin required in a day is
small, on the order of 1 mg, the equivalent of 1 mL. Even with
a 10:1 ratio of carriers to insulin in a PAD, a typical 250 mL
injection of insulin would contain the equivalent of 25 days
insulin. The principle challenge is to engineer a material in
which insulin maintains its integrity during the synthetic
process, and is retained effectively until light releases it in
a precisely metered fashion and in its native form.
We have successfully made this material by using a new
azide derivative of the 1-(4,5-dimethoxy-2-nitrophenyl)ethyl
(DMNPE) group[2] that reacts with insulin. The final link to
the insoluble resin is made through a “click”-type reaction
with a strained cyclooctyne (DBCO) that is resin-bound. We
demonstrate that insulin modified with both one and two
DMNPE–azide moieties is photoreleased in a similar fashion,
with the dimodified species being photolyzed in two sequen-
tial reactions, each of which has identical kinetics. This
suggests chemically similar sites of modification. Further-
more, we demonstrate that the insulin modified with
DMNPE–azide efficiently reacts with resin-bound DBCO,
and that this final material shows efficient and metered
photorelease of native insulin when using a 365 nm light-
emitting diode (LED).
Figure 1. Overall approach to the photoactivated depot (PAD). A drug,
insulin in this case, is linked to an insoluble but biodegradable resin,
through a photocleavable linker. The conjugate is injected in a shallow
depot cutaneously or subcutaneously. Irradiation breaks the link of
insulin from the resin, thereby allowing it to diffuse away from the
resin and be absorbed by the systemic circulation. Ultimately the resin
is biodegraded.
the most direct ways of accomplishing this is using a “click”
approach to make the final link between resin and insulin
(Figure 2). We started from the known DMNPE derivative 2,
which contains both carboxy and ketone groups.[3] This was
modified with an amino azide 1 to make the corresponding
amide 3 in good yield. The ketone group was then trans-
formed into the hydrazone 4 by using hydrazine, and finally
converted to the diazo derivative 5 by using MnO2, a process
based on previously developed DMNPE chemistry in our
group.[4] We found that a parallel approach, in which
propargyl amine was condensed with 2, was not effective.
The diazo–DMNPE–azide 5 was reacted with insulin
directly in DMSO, using a 2:1 mole ratio. HPLC analysis of
the reaction mixture shows two main, well-resolved peaks as
well as a minor one (Figure 3a). The compounds in these
peaks were isolated and identified by using ESI–MS as
unreacted insulin (58%), monomodified insulin (“insulin
monoazide” 6, 32%), and dimodified insulin (“insulin di-
azide” 7, 9%). A less highly modified sample was prepared
for the resin modification experiments using a 1:1 ratio of
insulin to 5. This reaction product showed 75% unreacted
insulin, 23% insulin monoazide 6, and 2% insulin diazide 7.
For the resin linking, this mixture was used directly, because
the insulin was expected to be “silent” during the coupling,
thus leaving principally insulin monoazide 6 to react with the
resin.
We are pursuing multiple ways of making the photo-
cleavable link between insulin and an insoluble resin. One of
[*] P. K. Jain, D. Karunakaran, Prof. S. H. Friedman
Division of Pharmaceutical Sciences, School of Pharmacy
University of Missouri, Kansas City
For our studies, we have used a ChemMatrix rink amide
poly(ethylene glycol) (PEG) resin. In the long term, we are
interested in using a biodegradable matrix, so that the depot
can be cleared from the site after a majority of the insulin has
been released. The use of the PEG rink amide resin gave us
the ability to analyze intermediates through TFA-induced
cleavage as well as through final photolysis.
2464 Charlotte Street, Kansas City, MO 64108-2718 (USA)
E-mail: friedmans@umkc.edu
Homepage: http://www.friedmanlaboratory.org
[**] We would like to thank Prof. William Gutheil for guidance on mass
spectrometry issues.
To link insulin monoazide to the resin, we first attempted
a copper(I)-based click approach, in which the resin was
derivatized with propargyl amine. This proved unsuccessful
Supporting information (including experimental details) for this
article is available on the WWW under http://dx.doi.org/10.1002/
anie.201207264.
Angew. Chem. Int. Ed. 2013, 52, 1404 –1409
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
1405

.
Angewandte
Communications
Reaction of insulin mono-
azide with the resin-bound
DBCO was monitored by exam-
ining the loss of insulin–azide
from the supernatant surround-
ing the resin. While this test
sample was being prepared, an
analogous control sample was
prepared in which native insulin
was added to DBCO-modified
resin. The purpose of this con-
trol was to insure during pho-
tolysis experiments that the gen-
eration of native insulin was due
to actual photolysis and not
simply loss of adsorption.
The resulting materials were
analyzed in multiple ways. To
confirm the adduct and link
were as intended, the modified
resin was treated with 95%
TFA to cleave the carboxamide
link. The material that was
released from the resin was
analyzed by ESI–MS and
showed
a
single prominent
mass (6593.0), similar to the
mass expected of the insulin–
DMNPE–triazole–DBCO
adduct 11 (6594.0). A second
sample of the modified resin
was irradiated using a 30 W
black-ray fluorescent fixture.
The supernatant as analyzed
showed a single peak in the
HPLC with retention time con-
sistent with a native insulin
standard. This was confirmed
by ESI–MS analysis (mass
expected 5808.0, mass observed
5808.0; Figure 2).
We examined the kinetics of
photolysis in two ways: in the
isolated and purified insulin
mono and diazides (6,7), and
in the final modified resin. The
former experiment provides us
with limiting values of photoly-
sis, as the system is homoge-
nous. The latter is closer to
“real-world” conditions, but is
complicated by the inherent
variability of light propagation
through inhomogenous resin.
Furthermore, solution photoly-
Figure 2. Synthesis and characterization of the photoactivated depot. The depot is constructed in
a convergent fashion, with the alkyne portion being attached to the resin, and the azide being attached to
the insulin through a photocleavable linker. Copper-free click conditions make the final link between
insulin and resin. TFA-induced cleavage of the total conjugate from the resin confirms the covalent
nature of the linkage, while photolysis confirms the release of unmodified, native insulin. HOBT=
1-hydroxybenzotriazole, EDC=1-ethyl-3-(3-dimethylaminopropyl)carbodiimide.
when using a variety of conditions (data not shown).[5] We
then turned to the strained cyclooctyne DBCO 8, which was
efficiently reacted with the resin amine directly to form 9
(Figure 2).[6]
sis of the mono and diazides gives us insights into the site of
attachment of the DMNPE group. Insulin has multiple
potential sites of modification, but literature precedent
suggests that the site of modification will likely be the
1406 www.angewandte.org
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2013, 52, 1404 –1409

Angewandte
Chemie
insulin monoazide and insulin
diazide, we hoped to reveal
these differences.
A solution (150 mL) of puri-
fied
insulin
monoazide
(260.3 mm insulin monoazide 6,
6.5 mm insulin) was irradiated
for varying durations of time
using a black-ray lamp (Model
XX-15 L, 30 W). Samples were
removed and analyzed by
HPLC to quantify the amount
of insulin and insulin mono-
azide. Both of these kinetic
progress curves were plotted
and fit well to a first-order
process, giving
a first-order
rate constant k of (0.028 Æ
0.001) minÀ1 (Figure 3c; also
see the Supporting Information
for details of fitting proce-
dures).
In a parallel experiment,
a purified solution (150 mL) of
insulin diazide (125.2 mm insu-
lin diazide, 26 mm insulin mono-
azide) was similarly photolyzed
and analyzed. In this case, we
determined the insulin diazide
starting material, the insulin
monoazide intermediate, and
the final insulin as a function
of time. The resulting kinetic
plots were fit to two sequential
first-order processes, corre-
sponding to starting material
being converted to intermedi-
ate, finally being converted to
final product and resulting in
two kinetic constants k1 and k2
as shown in Figure 3b.[7] The
consumption of 7 (Figure 3d)
was well fit with a first-order
rate law, giving a first-order
Figure 3. Formation and kinetic analysis of insulin monoazide and diazide photolysis. a) Reaction of
insulin and compound 5 in a ratio of 1:2, resulting in the formation of insulin, insulin monoazide, and
insulin diazide. HPLC analysis shows product ratio, and identity is confirmed by ESI–MS. b) Scheme
showing the conversion of insulin diazide to insulin monoazide to insulin by photolysis. Rate constants
indicated. c) Photolysis of insulin monoazide to insulin. First-order kinetic fit curves are shown.
d) Photolysis of insulin diazide, first to insulin monoazide, then to insulin. Kinetic fit curves of two
sequential first-order processes are shown.
constant
k1
of
(0.054 Æ
0.004) minÀ1. This is essentially
double the constant for the
photolysis of insulin monoa-
zide; this value is consistent
with two DMNPE groups per
molecule. The formation and
carboxy groups. Because of this, insulin monoazide 6 will
likely be a mixture of species with modifications at these
different sites. Furthermore, insulin diazide 7 will be an even
more complex mixture. It is possible that the micro environ-
ment of the modification site will influence the rate of
photolysis, or alternatively, different sites could have essen-
tially identical rates. By examining the kinetic fits of both the
consumption of 6 was well fit by two first-order rate constants,
k1 being (0.052 Æ 0.004) minÀ1 and k2 being (0.023 Æ
0.001) minÀ1. Again, k1 being approximately two times the
value of k2, and the first-order constant for the photolysis of
insulin monoazide. The final step in the photolytic path, the
formation of insulin, was less well fit by two first-order
processes but gave comparable values of k1 of (0.028 Æ
0.017) minÀ1 and k2 of (0.033 Æ 0.019) minÀ1. Taken together,
Angew. Chem. Int. Ed. 2013, 52, 1404 –1409
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
1407

.
Angewandte
Communications
the similarity of rate constants for the consumption of insulin
monoazide and diazide, when corrected for the 2:1 ratio of
photolyzed groups per mole of compound, suggests that the
many possible sites of photolysis have essentially the same
photoreactivity; thus, insulin monoazide, despite being
a likely mixture, can be reasonably modeled as a monolithic
entity with a single first-order photolysis rate.
In addition to photolysis by using a fluorescent UV lamp,
we also examined rates of photolysis of insulin monoazide by
using an LED (200 mW 365 nm LED, Nichia), since this is
a light source closer to what might be used in a real-world
application. A similar experiment was performed, in which
the sample was placed close to the LED. Again, this was done
to simulate a real-world situation, in which the light source
was placed directly above the depot site. Sampling and
analysis was performed identically to the manner described
above. The resulting release was very rapid, giving a k1 of
(0.893 Æ 0.105) minÀ1, approximately 32 times faster than the
photolysis when using the fluorescent UV lamp (see the
Supporting Information for details).
We then examined the photolysis kinetics of the resin-
linked insulin by using the LED. This experiment was
performed in a similar fashion to the solution-based experi-
ments described above, with some key modifications: After
each time period and before sampling, the resin was gently
vortexed in an attempt to approximate homogeneity. In
addition, the supernatant was sampled both directly after the
LED was turned off, and then again after a dark period. This
was done to insure that processes that were not based on
photolysis were not taking place. The resulting “step-plot” is
shown in Figure 4a. We consistently observed no additional
release during the dark phase. The data in Figure 4a was
replotted, with insulin release being plotted as a function of
irradiation time. Even though the solution is inhomogenous
and there were small variations in the total volume, the
resulting curve was reasonably fit to a first-order rate law
(Figure 4b). The resulting first-order constant (k = (0.035 Æ
0.002) minÀ1) was approximately 25-fold slower than the
result obtained for the homogeneous solution of insulin
monoazide when using the same LED source. This is not that
surprising, considering the possibility of scattering of the
irradiation source by the beads, an effect that is not seen in the
homogenous experiment.
In this work, we have prepared and characterized an
insulin photoactivated depot (i-PAD), and in the process we
have shown that insulin can be efficiently linked to an
insoluble resin, maintain its integrity, and then be released in
a predictable fashion by using light. As such, it forms an ideal
starting point for the development of a therapeutic approach,
in which a biodegradable photoactivated depot of insulin is
injected under the skin and transdermal illumination stim-
ulates the release of insulin at the appropriate time and in the
appropriate amounts.
Figure 4. Stepwise photolysis of the photoactivated insulin depot.
a) Cumulative moles of insulin released from the modified resin when
using an LED point source that was turned on and off repeatedly.
Light and dark bars indicate periods of irradiation and darkness.
b) Replot of data from (a) shows the concentration of released insulin
as a function of irradiation time. First-order fit curve is shown.
glucose monitoring and continuous metered insulin admin-
istration should allow for native levels of blood sugar control
and a return to normality for type I diabetics.
To our knowledge, this is the first example of light-
controlled release of insulin from a covalently linked depot.
There have been other descriptions of light-activated drug
release in general. Many of them rely on degrading a bulk
polymer by using light to allow release of the drug.[9] This
process may be harder to control as there is no direct one-to-
one correspondence of a photolysis event with drug release.
In fact, photolysis events will be associated with random
changes to the sequestering polymer, thereby making it
difficult to predict changes in insulin release. Other examples
from the literature use shorter and more toxic wavelengths of
light and/or irreplaceable depots, for example, using 266 nm
light to release fluorouracil from a difficult to replace
intraocular lens.[10] A previously described approach of
using photocleavable groups to make a drug insoluble (e.g.
ibuprofen) would be difficult to adapt for insulin, because
multiple modifications would be required, and this would
again lead to challenges in predictable control of release.[11]
Our approach of linking a single photolysis event to the
release of a single drug should allow for a more predictable
control of the process.
Insulin is particularly suited to this approach, since only
a small amount is required for each day, and the moment-to-
moment needs are highly variable. An insulin PAD would
also put in place half of the components needed for an
artificial pancreas, the other half being a continuous blood
glucose monitor.[8] The combination of continuous blood
Future optimization steps include the use of a biodegrad-
able resin, as well as the exploration of photocleavable groups
that can be removed by applying longer wavelengths. This will
allow depots to be positioned at deeper levels, including
intramuscularly. However, the currently used photocleavable
chromophore, DMNPE, may still be effective in shallow,
1408 www.angewandte.org
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2013, 52, 1404 –1409

Angewandte
Chemie
[3] C. P. Holmes, J. Org. Chem. 1997, 62, 2370 – 2380.
[4] P. K. Jain, S. Shah, S. H. Friedman, J. Am. Chem. Soc. 2011, 133,
440 – 446.
[5] a) C. W. Tornøe, C. Christensen, M. Meldal, J. Org. Chem. 2002,
67, 3057 – 3064; b) Q. Wang, T. R. Chan, R. Hilgraf, V. V. Fokin,
K. B. Sharpless, M. G. Finn, J. Am. Chem. Soc. 2003, 125, 3192 –
3193.
dermal or hypodermal sites, which are within approximately
1 mm of the skinꢀs surface. The dermal layer is vascularized
and adjacent to cells that synthesize vitamin D by using light
of even shorter wavelengths (270–300 nm), which are able to
penetrate to that depth. As such it may be an effective
location for DMNPE-based depots.
The long-term benefit of such materials would be the
transformation of the delivery of insulin and similar ther-
apeutics: molecules that are needed in small and highly
varying amounts in response to physiological cues.
[6] a) N. J. Agard, J. A. Prescher, C. R. Bertozzi, J. Am. Chem. Soc.
2004, 126, 15046 – 15047; b) M. F. Debets, S. S. van Berkel, J.
Dommerholt, A. T. Dirks, F. P. Rutjes, F. L. van Delft, Acc.
Chem. Res. 2011, 44, 805 – 815; c) M. F. Debets, S. S. van Berkel,
S. Schoffelen, F. P. Rutjes, J. C. van Hest, F. L. van Delft, Chem.
Commun. 2010, 46, 97 – 99.
Received: September 8, 2012
Published online: December 3, 2012
[7] B. W. Benson, The Foundations of Chemical Kinetics, McGraw-
Hill, New York, 1960.
[8] a) D. Elleri, D. B. Dunger, R. Hovorka, BMC Med. 2011, 9, 0120;
b) M. Langendam, Y. M. Luijf, L. Hooft, J. H. Devries, A. H.
Mudde, R. J. Scholten, Cochrane database of systematic reviews
(Online) 2012, 1.
[9] C. Alvarez-Lorenzo, L. Bromberg, A. Concheiro, Photochem.
Photobiol. 2009, 85, 848 – 860.
[10] C. Sinkel, M. C. Schwarzer, G. Frenking, A. Greiner, S. Agarwal,
Magn. Reson. Chem. 2011, 49, 70 – 75.
[11] C. P. McCoy, C. Rooney, C. R. Edwards, D. S. Jones, S. P.
Gorman, J. Am. Chem. Soc. 2007, 129, 9572 – 9573.
Keywords: click chemistry · drug delivery · photochemistry ·
photolysis · protein modifications
.
[1] S. M. Switzer, E. G. Moser, B. E. Rockler, S. K. Garg, Endo-
crinol. Metab. Clin. North Am. 2012, 41, 89 – 104.
[2] a) J. W. Walker, G. P. Reid, J. A. McCray, D. R. Trentham, J. Am.
Chem. Soc. 1988, 110, 7170 – 7177; b) J. W. Walker, G. P. Reid,
D. R. Trentham, Methods Enzymol. 1989, 172, 288 – 301.
Angew. Chem. Int. Ed. 2013, 52, 1404 –1409
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 1409