Journal of Controlled Release 81 (2002) 1–6
www.elsevier.com / locate / jconrel
Rapid communication
New mode of drug delivery: long term autonomous rhythmic
hormone release across a hydrogel membrane
a
Gauri P. Misra , Ronald A. Siegel
a
a,b ,
*
Department of Pharmaceutics, 9 -177 Weaver-Densford Hall, University of Minnesota, 308 Harvard St. S.E., Minneapolis,
MN 55455, USA
b
Department of Biomedical Engineering, University of Minnesota, Minneapolis, MN 55455, USA
Received 4 December 2001; accepted 13 February 2002
Abstract
This note demonstrates a novel mode of rhythmic drug delivery, which is independent of external modulation or
physiological stimulation. Rhythmic behavior is attributed to negative, nonlinear feedback between the swelling state of a
hydrogel membrane and the enzymatic conversion of glucose to hydrogen ion. The system pulsates in the presence of a
constant level of glucose, thus distinguishing it from insulin delivery devices that respond to changes in glucose
concentration. As an example, gonadotropin-releasing hormone (GnRH) was released in short, repetitive pulses over 1 week.
 2002 Elsevier Science B.V. All rights reserved.
Keywords: Rhythmic delivery; Hormone; GnRH; LHRH; Ultradian; Feedback; Hydrogel
1. Introduction
Numerous hormones are endogenously secreted in
an episodic, pulsatile manner [1–3]. For example,
gonadotropin releasing hormone (GnRH) is secreted
by the hypothalamus in nearly regular bursts every
1–2 h [4,5]. Treatment of hormonal disorders associated with GnRH deficiency in both men and women
warrants hormone delivery in a temporal pattern
mimicking the endogenous ultradian rhythm [4,5],
and several approaches to ultradian delivery have
been proposed [6–12].
In this communication we demonstrate rhythmic
delivery of GnRH across a polyelectrolyte hydrogel
*Corresponding author.
E-mail address: siege017@tc.umn.edu (R.A. Siegel).
membrane in the presence of a constant level of
glucose, approximating conditions seen in nondiabetic individuals. Rhythmicity is attributed to
nonlinear feedback between the membrane and the
enzyme glucose oxidase. The results presented here
provide the first positive evidence supporting a
strategy, described previously [13–17], for the ultradian delivery of hormones.
2. Materials and methods
2.1. Membrane
N-Isopropylacrylamide
(NIPA,
Kodak),
methacrylic acid (MAA, Polysciences), and ethyleneglycol dimethacylate (EGDMA, Polysciences)
0168-3659 / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved.
PII: S0168-3659( 02 )00043-3
G.P. Misra, R. A. Siegel / Journal of Controlled Release 81 (2002) 1 – 6
2
with molar composition 90:10:0.7 NIPA–MMA–
EGDMA, were dissolved in 1 ml water–ethanol
(50:50, v / v), with cosolvent mass twice that of the
monomer mixture, along with 5 mg ammonium
persulfate (APS, Polysciences) and 20 ml tetraethylmethylenediamine (TEMED, Aldrich). The solution was polymerized at 10 8C between glass plates
separated by a 250-mm spacer. Resulting hydrogels
were extensively washed and conditioned in saline
media.
2.2. Enzymes
Glucose oxidase (GluOx, Sigma: 234 I.U. / mg)
and catalase (Cat, Sigma: 10 700 I.U. / mg) were
immobilized in small polyacrylamide gel particles
produced by polymerization at 4 8C with the following ingredients: 1 ml water, 0.2 g acrylamide
(Sigma), 5 ml EGDMA, 4 mg APS, 10 ml TEMED,
10 mg albumin (Sigma), 20 mg GluOx and 2 mg Cat
(10,700 I.U. / mg).
2.3. Experimental procedure
Membranes were mounted in a water-jacketed,
side-by-side diffusion cell with 80 ml cell volumes
and transport aperture radius 1 cm, illustrated in Fig.
1. A steady pH gradient was established across the
membrane by pH stat-ing Cells I and II overnight at
37 8C in 75 ml, 50 mM NaCl solutions at pH 7.0 and
6.0, respectively. Just before an experimental run,
enzyme particles were added to Cell II, and Cell II
was stat-ed at pH 4.5 for 45 min.
At the start of a run, the contents of Cell I were
replaced with 75 ml of a stock solution containing 50
mM glucose, 50 mM NaCl, and 0.01 wt.% 2-bromo2-nitro-1,3-propanediol (bronopol, Aldrich, antimicrobial). Steady feed and drainage of stock solution through Cell I were then initiated, controlled by
peristaltic pump (SciLog) with a flow-rate 1.36–1.38
ml / min, and pH in Cell I was clamped at 7.0 by
pH-stat. At the same time, the pH-stat was disconnected from Cell II, and a 12.5-g piece of marble
was introduced there. Finally, tetramethylrhodaminelabeled GnRH (f-GnRH, Mw 51614, Genemed Synthesis), was spiked into Cell II to a concentration of
1 mM.
During a run, the pH in Cell II was followed with
a pH electrode, and the appearance of f-GnRH in
Cell I was monitored spectrofluorimetrically ( lex 5
Fig. 1. Diagram of test cell for rhythmic pulsatile delivery of GnRH. Both Cells I and II are 80 ml and are charged with 75 ml solutions,
which are stirred vigorously using magnetic stir bars (SB) and maintained at 37 8C by circulation from a waterbath through water jackets.
Cell I is fed at 1.36–1.38 ml / min by 50 mM saline containing and 50 mM glucose and 0.01 wt.% bronopol (antibacterial), is drained at
equal rate, and is pH stat-ed at 7.0. Cell II initially contains glucose-free saline solution, a 12.5-g piece of marble, and small particles of
polyacrylamide gel containing enzymes. Small dots refer to f-GnRH molecules, initially introduced at 1 mg / ml into Cell II. pH in Cell II is
monitored by a pH meter and appearance of f-GnRH in Cell I is recorded by a spectrofluorimeter, with both instruments sending data to a
personal computer.
G.P. Misra, R. A. Siegel / Journal of Controlled Release 81 (2002) 1 – 6
520 nm, lem 5572 nm). Vigorous stirring (600 rpm)
was provided in both cells, and temperature was
maintained at 37 8C.
3. Results
Results of a typical run are shown in Fig. 2. The
pH record in Cell II (Fig. 2a) is characterized as an
oscillation, with peaks that converge rapidly to a
near-constant value between pH 5.05 and 5.1, and
troughs that become progressively shallower, approaching |pH 4.9–4.95. The period of oscillation
steadily increases from |4 h initially to |8–12 h at
about 1 week, after which pH oscillations cease.
Fig. 2b shows a series of pulses of f-GnRH
concentration in Cell I that correlate with pH drops
in Cell II. When pH rises in Cell II, f-GnRH
transport ceases, and the subsequent decay in fGnRH concentration in Cell I reflects washout
through the drain line.
The rate of release of f-GnRH across the membrane was estimated from the record in Fig. 2b by
deconvoluting the residence time distribution, r(t), of
hormone in Cell I. Since Cell I is well stirred with
constant volume V 5 75 ml and constant outflow-rate
Fig. 2. (a) Time course of oscillations of pH in Cell II. (b)
Corresponding time course of concentration of f-GnRH in Cell I.
3
F 51.37 ml / min, r(t) 5 (F /V ) exp (2Ft /V ). Results
of deconvolution, shown in Fig. 3, feature bursts of
f-GnRH release separated by quiescent periods.
Pulse width and interpulse interval increase with
time, in register with the pH swings in Fig. 2a.
The pattern of pH oscillations is almost exactly
reproduced in the middle of a run when the contents
of Cell II are replaced with fresh stock solution and
enzyme (data not shown). Some variability between
runs is seen, however, possibly due to nonuniformity
of marble samples or subtle differences in membrane
preparation and mounting. Variations in peak and
trough levels seen in Figs. 2b and 3 are most likely
due to baseline drift in the fluorescence detection
system, initial buildup of an f-GnRH load in the
membrane, and reduction of transmembrane gradient
of f-GNRH with time.
4. Discussion
It is significant that the present system behaves
rhythmically even though it is fed by a constant
glucose concentration. The rhythms are therefore
termed autonomous [18]. In Fig. 4 the present
system’s autonomous behavior is contrasted with the
responsive behavior of systems that have been
previously considered for glucose-sensitive insulin
delivery [19–23].
Autonomous rhythmic behavior is attributed to
nonlinear negative feedback between the NIPA–
MAA membrane and glucose oxidase [13–17]. When
pH in Cell II is sufficiently high (.5.02), the
membrane is charged, swollen, and permeable to
glucose. Glucose enters Cell II and is converted by
glucose oxidase to gluconate 2 and H 1 , resulting in a
pH drop in Cell II. Hydrogen ions then bind to and
neutralize the MAA groups on the membrane, and
the membrane shrinks under the influence of the
hydrophobic NIPA component, diminishing its permeability to glucose. Enzymatic production of H 1 is
then attenuated, leading to a rise in pH in Cell II.
Eventually the hydrogen ions bound to the membrane are released into Cell I, restoring the membrane’s charge, and the membrane reverts to its
swollen, high permeability state. The cycle is now
poised to repeat itself. Cycles of swelling and
4
G.P. Misra, R. A. Siegel / Journal of Controlled Release 81 (2002) 1 – 6
Fig. 3. Estimated rate of f-GnRH delivery through the membrane, obtained by deconvoluting data in Fig. 2a. Deconvolution is based on the
residence time distribution r(t) 5 (F /V ) exp (2Ft /V ) of f-GnRH in Cell I, which is assumed to be well stirred, of constant volume V 5 75
ml, and fed and drained with stock solution at a constant rate F 51.37 ml / min. A symmetric 17-point, third-order Savitzky-Golay smoothing
filter ( ORIGIN 6.0 software, Microcal) was used to suppress noise arising from deconvolution. Inset: detailed view of fifth and sixth peaks,
with no smoothing.
deswelling correspond, respectively, to permeation
and blockade of f-GnRH.
Because of the large volume of Cell II, pH swings
would normally require days instead of hours. In this
work, pH swings were accelerated but kept in range
by increasing glucose concentration and introducing
marble (CaCO 3 ) into Cell II. Marble acts as a shunt
for eliminating H 1 via the reaction 2H 1 1
CaCO 3 →Ca 21 1CO 2 1H 2 O [24].
The characteristic relating glucose permeability of
NIPA–MAA membranes to pH in Cell II exhibits a
sharp, first order transition, with hysteresis [16,25],
and this is an important enabling property for the
mechanism proposed above. From Fig. 2 we estimate
that the hysteresis band has its lower and upper
limits in the pH ranges 4.9–4.95 and 5.02–5.10,
respectively, in rough agreement with observations
in Ref. [25]. As pH increases or decreases within the
hysteresis band, the membrane remains in its col-
lapsed or swollen state until it crosses the band’s
upper or lower limit, respectively [14,16]. Because
of sluggishness of membrane response, measured pH
may overshoot these limits before a transition takes
place, particularly on the low end, as seen at early
times in Fig. 2.
With increasing time, however, pH oscillations
converge to a nearly constant amplitude, and periodicity steadily increases. We attribute these phenomena to a steady accumulation of gluconate 2 , which
acts as a pH buffer, in Cell II [17]. This accumulation slows the upswings and downswings of pH, and
reduces the overshoots. When a sufficiently high
concentration of gluconate 2 is reached, H 1 production may be buffered to the extent that pH in Cell
II cannot drop below the lower limit of the hysteresis
band, and the system is trapped in its high permeability state. At this point pH oscillations cease, as
does rhythmic delivery of f-GnRH.
G.P. Misra, R. A. Siegel / Journal of Controlled Release 81 (2002) 1 – 6
5
Fig. 4. Illustration of difference between glucose-responsive delivery of insulin (top panel) and autonomous, glucose-driven rhythmic
delivery of ultradian hormones such as GnRH (bottom panel). In the top panel, insulin release is minimal until glucose level rises, usually in
response to a meal, at which point a pulse of insulin is released. In the bottom panel, the glucose level does not change with time, but
hormone release occurs in regular pulses due to nonlinear, negative feedback between hydrogel membrane swelling and enzymatic
conversion of glucose to hydrogen ions.
The present system is an early prototype of a
device for rhythmic delivery of hormones. Cells I
and II represent external body fluids and the device
interior, respectively. Methods to maintain rhythmic
behavior indefinitely, in the absence of marble and
the presence of physiologically more realistic glucose concentrations (3–7 mM) and pH-buffering
environment (pH 7.4, 27 mM bicarbonate), are under
investigation.
Acknowledgements
This work was funded by NSF grant CHE-
6
G.P. Misra, R. A. Siegel / Journal of Controlled Release 81 (2002) 1 – 6
9996223. We thank Anish P. Dhanarajan and Yuandong Gu for helpful comments.
[14]
References
[15]
¨
[1] G. Brabant, K. Prank, C. Schofl,
Pulsatile patterns in
hormone secretion, Trends Endocrinol. Metab. 3 (1992)
183–190.
[2] W.F. Crowley, J.G. Hofler, The Episodic Secretion of
Hormones, Wiley, New York, 1987.
[3] A. Goldbeter, Biochemical Oscillators and Cellular Rhythms,
Cambridge University Press, Cambridge, 1996.
[4] N. Santoro, M. Filicori, W.F.J. Crowley, Hypogonadotropic
disorders in men and women: diagnosis and therapy with
pulsatile gonadotropin-releasing hormone, Endocr. Rev. 7
(1985) 11–23.
[5] M.B. Southworth, A.M. Matsumoto, K.M. Gross, M.R.
Soules, W.J. Bremner, The importance of signal pattern in the
transmission of endocrine information: pituitary gonadotropin responses to continuous and pulsatile gonadotropin-releasing hormone, J. Clin. Endocrinol. Metab. 72 (1991) 1286–
1289.
[6] N.J. Medlicott, I.G. Tucker, Pulsatile release from subcutaneous implants, Adv. Drug Del. Rev. 38 (1999) 139–149.
[7] A. Gopferich, Bioerodible implants with programmable drug
release, J. Controlled Release 44 (1997) 271–281.
[8] H.L. Jiang, K.L. Zhu, Pulsatile protein release from a
laminated device comprising of polyanhydrides and pHsensitive complexes, Int. J. Pharm. 194 (2000) 51–60.
[9] E.R. Edelman, J. Kost, H. Bobeck, R. Langer, Regulation of
drug release from polymer matrices by oscillating magnetic
fields, J. Biomed. Mater. Res. 19 (1985) 67–83.
[10] H.J. Lynch, R.W. Rivest, R.J. Wurtman, Artificial induction
of melatonin rhythms by programmed microinfusion, Neuroendocrinology 31 (1980) 106–111.
[11] Y. Ly, Y. Cheng, Electrically-modulated variable permeability liquid crystalline polymeric membrane, J. Membr. Sci. 77
(1993) 99–112.
[12] P. Arunothayanun, T. Sooksawate, A.T. Florence, Extrusion
of niosomes from capillaries: approaches to a pulsed delivery
device, J. Controlled Release 60 (1999) 391–397.
[13] R.A. Siegel, C.G. Pitt, A strategy for oscillatory drug release.
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
General scheme and simplified theory, J. Controlled Release
33 (1995) 173–186.
X. Zou, R.A. Siegel, Modeling of oscillatory dynamics of a
simple enzyme–gel diffusion systems with hysteresis. The
case of lumped permeabilities, J. Chem. Phys. 110 (1999)
2267–2279.
J.-C. Leroux, R.A. Siegel, Autonomous gel–enzyme oscillator fueled by glucose—preliminary evidence for oscillations, Chaos 9 (1999) 267–275.
J.-C. Leroux, R.A. Siegel, pH-Hysteresis of glucose permeability in acid-doped LCST hydrogels—a basis for pulsatile, oscillatory drug release, in: J. DeNuzzio (Ed.), Intelligent Materials and Novel Concepts For Controlled
Release Technologies, American Chemical Society,
Washington, DC, 1999, pp. 98–112.
R.A. Siegel, G.P. Misra, A.P. Dhanarajan, Rhythmically
pulsing gels based on chemomechanical feedback, in: Y.
Osada, A.R. Khokhlov (Eds.), Polymer Gels and Networks,
Marcel Dekker, New York, 2002, pp. 357–372.
L. Edelshtein-Keshet, Mathematical Models in Biology,
Springer-Verlag, Berlin, 1988.
S.Y. Jeong, S.W. Kim, M.J.D. Eenink, J. Feijen, Self-regulating insulin delivery systems. I. Synthesis and characterization of glycosylated insulin, J. Controlled Release 1 (1984)
57–66.
F. Fischel-Ghodsian, L. Brown, E. Mathiowitz, D. Brandenburg, R. Langer, Enzymatically controlled drug delivery,
Proc. Natl. Acad. Sci. 85 (1988) 2403–2406.
K. Ishihara, K. Matsui, Glucose-responsive insulin release
from polymer capsule, J. Poly. Sci. Poly. Lett. Ed. 24 (1986)
413–417.
J. Kost, T.A. Horbett, B.D. Ratner, M. Singh, Glucosesensitive membranes containing glucose oxidase: swelling,
activity, and permeability studies, J. Biomed. Mater. Res. 19
(1985) 1117–1122.
C.M. Hassan, F.J. Doyle III, N.A. Peppas, Dynamic behavior
of glucose-responsive poly(methacrylic acid–g-ethylene glycol) hydrogels, Macromolecules 30 (1997) 6166–6173.
´
G. Rabai,
I. Hanazaki, pH oscillations in the bromate–
sulfite–marble semibatch and flow systems, J. Phys. Chem.
100 (1996) 10615–10619.
J.P. Baker, R.A. Siegel, Hysteresis in the glucose permeability versus pH characteristic for a responsive hydrogel membrane, Macromol. Rapid. Commun. 17 (1996) 409–415.