Technical Notes Anal. Chem. 1995,67,466-471
Improvement in the LonglTerm Stability of an Amperometric Glucose Sensor System by Introducing a Cellulose Membrane of Bacterial Origin H. P. T. A"on,*st
W. Ege,* M. Oppermann,, W. Gl)pel,* and S. Eiselet
Department of Pharmacology, Institute of Pharmaceutical Sciences and Institute of Physical and Theoretical Chemistry, University of mbingen, Auf der Morgenstelle 8, 72076 mbingen, Germany, and Department of Immunology, University of Gottingen, Kreuzbergnng 57,37075 Gottingen, Germany
Classicalamperometricglucose sensors that use cellulose membranes of wood origin (Cuprophan) suffer from the fact that their long-term stability in blood is short; therefore, their clinical use is limited. In the present study, a classical amperometric glucose sensor was covered with a bacterial cellulose (BC) membrane. Its surface in comparison to that of the classical glucose sensor (Cuprophan) and its long-term stability were tested in vitro and in vivo. The surface element composition was -44% oxygen and -56% carbon in both membranes and thus typical for cellulose. BC membranes exhibited fiber structure, whereas cup membranes did not. There was also a qualitative difference in protein adsorption between both membranes on exposure to bovine serum albumin. Treatmentwith Trogamid of one site of the BC membranes allowed linear glucose detection between 0 and 40 mM. Hemocompatibility of BC membranes was improved in comparison to cup membranes on the basis of complement activation (C3a and C5a). In diluted blood ( l : l O ) , the BC-covered sensor exhibited a long-term stability of more than 200 h; in undiluted blood it was stable for about 24 h, which is about 6-7 times longer than the stability of the classical Cup membrane-covered sensor. In in vivo studies, where the BC membrane-covered sensors were connected to the jugular vein of rats, blood glucose levels could be monitored for at least 24 h. In summary, the use of a modified bacterial cellulose membrane to cover the classical amperometric glucose sensor significantlyimproves the sensor's long-term stabilityboth in vitro and in vivo.
were developed as early as the early 1 9 6 0 ~ ,the ' ~ ~main problems, i.e., platelet deposition, failure of long-term stability, and immunological interactions with the sensor surface? still exist. Scheller and his group4 were among the firsts to develop an amperometric glucose sensor for glucose determinations in diluted blood, thus preventing fouling, with a sample turnover of 120 samples/h. Pfeiffer et alS5used this method to develop a double lumen catheter, which allows venous blood sampling and mixing of the blood sample with heparin solution. This mixture flows directly into a connected glucose sensor chamber. Glucose determination in subcutaneous tissue, often combined with mi~rodialysis,6~~ is also a method to slow down surface fouling. Other groups investigated different kinds of membrane materials (silanes, cellulose acetate) as outer membranes of glucose sensor^^^^ and examined the hemocompatibility of these membranes in contact with blood.1° In contrast to the usual cellulose originating from wood (Cuprophan (Cup)), bacterial cellulose (BC) possesses a hydrogel character and it is possible that its membrane properties may be different compared to those of cellulose originating from wood. Therefore, we studied whether or not covering the classical amperometricglucose sensor with a modified bacterial cellulose membrane would improve its long-term stability in vitro and in vivo. EXPERIMENTAL SECTION Reagents. Glucose oxidase from Aspergillus niger (EC1.1.3.4, 236 units/mg) and D(+)-glucose were purchased from Fluka
Department of Pharmacology, Institute of Pharmaceutical Sciences, University of Tubingen. t Institute of Physical and Theoretical Chemistry, University of Tiibingen. Department of Immunology, University of Gottingen.
Updike, S. J.; Hicks, G. J. Nature 1967,21, 986. Clark, L. L;Lyons C. Ann. N. k: Acad. Sci. 1962, 102, 29-45. Andrade, J. D.; Hlady, V. Adv. Polym. Sci. 1986, 79, 3-59. Scheller, F. W.; Pfeiffer, D.; Hmtzsche, R; Dransfeld, I.; Nentwig, J. Biomed. Biochim. Acta 1989,48, 891-896. Pfeiffer, E. F. Verh. Dtsch. Ges. Inn. Med. 1989, 95, 725-736. Meyerhoff, C.; Bischof, F.; Zier, H.; Pfeiffer, E. F. Diabetologia 1992,35, 1087-1092. Aalders; et al. Int. 1.Artg Organs 1992, 14(2), 102-108. Fortier, G.; Brassard, E.; Belanger, D. Biotechnol. Tech. 1988,2(3), 177182. Umana, M.; Waller, J. Anal. Chem. 1986,58, 2979-2983. Baurmeister, U.; Vienken, J.; Ansorge, W.; Lutrell, k Altif: Organs 1989, 13, 52-57.
466 Analytical Chemistry, Vol. 67,No. 2,January 15, 1995
0 1995 American Chemical Society 0003-2700/95/0367-0466$9.00/0
The techniques available for continuous blood glucose monitoring are still far from satisfactory. Although glucose sensors +
Feinchemikalien GmbH, Neu-Ulm, Germany. Cuprophan was obtained from AKZO,Wuppertal, Germany. Sterile normal saline solution was from Delta-Pharma, Pfullingen, Germany, and Liquemin N 25000 was from Hoffmann-La Roche AG, Grenzach Wyhlen, Germany. The bacterial cellulose (BC) was obtained from Fraunhofer Gesellschaft, Institut f i r G r e d a c h e n und Bioverfahrenstechnik, Stuttgart,Germany. Trogamid was purchased from the Hlils AG, Stuttgart, Germany. Ketanest was obtained from Parke Davis & Co., Berlin, Germany, and Rompun was purchased from Bayer, Leverkusen, Germany. The ampercmetric flow-through cell was obtained from Prufgertitewerk Medingen, Medingen, Germany. Human blood samples were from the Department of Intemal Medicine, University of Ttibingen, Germany. Bacterial Cellulose Membrane Preparation. Cellulose can be produced by microbiological means, e.g., with Acetobacter xylinum in appropriate nutrient media, as recently described by Grobe et al." Here, the bacterial cellulose grows in an emerging fashion in standing cultures, in which a fibrous, webtype structure is formed at the surface of the cell culture. This structure can be harvested as a thin film. Since the first f ilm (Le., membrane) is not homogeneous enough, it is discarded. The following membranes, which also grow as a thin film, however, are completely homogeneous, providing a highly porous structure,'2 and are used as a glucose sensor. Bacterial cellulose exhibits a higher degree of polymerization, in the range of 2000 unit structures/molecule, and thus possesses a more tensile strength.12 The membrane exhibits good storage stability and mechanical properties. Furthermore, it allows steam sterilization. Membrane Properties. In order to evaluate whether or not there are d ~ e r e n c e sin the properties between the classical Cup membrane and the BC membrane, we studied the surface structures and hemocompatibilities of both. (1) Surface Structure of Cellulose Membranes. The surface structures of Cup and BC membranes were studied using scanning electron microscopy (SEW (Rasterelektronenmikroskop DSM 962, Carl Zeiss, Oberkochen, Germany). Membranes were first kept for 12 h in bidistilled water and were then dried for 6 h at 80 "C. After this preparation procedure, the membranes could be examined with SEM. To allow investigation of protein adsorption at the surface of the membranes, Cup and BC membranes were incubated in a 7%bovine serum albumin solution for 12 h and then rinsed with 5 mL of bidistilled water. After being dried for 6 h at 80 "C, the samples were examined with SEM. (2) Hemocompatibility. The activation of the complement system as a parameter for the hemocompatibility of the membrane material13 was determined by measuring the complement activation products C3a and C5a by enzymelinked immunosorbent assay @USA) using monoclonal antibodies. To obtain the samples for the C5a and C3a determination, the membranes were incubated for 2 h in fresh human blood serum (25 mg of membrane/mL of serum). After an end-over-end rotation at 37 "C, the incubation was stopped by addition of sodium ethylenediaminetetraacetic acid (EDTA, Serva, Germany) (final concentration,20 mM). For quantification of C5a/C5a(desArg), (11) GrBbe, A; Chmiel, H. A; Strathmann, H. Eur. Pat. Appl. 0 416 470 A2, 1990. (12) Grbbe, A; Chmiel, H. A Strathmann, H. Proc. Int. C o w . Membr. 1990, 2, 1430-1434. (13) Baumann, H.; Linssen, M.; Keller, R Das Papier 1989,43,674-680.
wells of microtiter plates (Nunc-Immunoplate I1 F, Nunc, Wiesbaden, Germany) were coated with 100pL of antiC5a mAb C17/5 (40 pg/mL) in coating buffer (50 pM carbonate, pH 10.0) for 16 h at 4 "C. After nonspecific binding sites were blocked with 200 pL of blocking buffer (PBSl%gelatine) for 30 min, the wells were rinsed with 200 pL of PBSTween, and 100 pL samples were applied per well. The samples were allowed to bind to the solid phase for 2 h and were detected by adding, in sequence, the second, biotinylated anti-C5a mAb G25/2 (1 pg/pL) and a 1OOG fold dilution of streptavidin peroxidase (Amersham, Braunschweig, Germany) in PBSTween, both incubation steps lasting 1 h. C5a assays were carried out by the colorimetric analysis of peroxidasemediatedhydrolysis of 2 mM 2,Y-azinobis(3ethylbenzothiazolin&sulfonic acid) (AB%, Boehringer Mannheim, Germany) in the presence of 2.5 mM H2Oz using a microplate photometer (Dynatech MR 600, Dynatech, Denkendorf, Germany). Calculations were performed by evaluating the increase of absorbance at 410 nm/min during a 5min reaction period.14 The quantification of C3a was performed in a similar manner, using two different mAbs specifiable for epitopes on C3a.15 Membrane Modifications (Trogamid Coating). Linearity of glucose sensoring is a prerequisite for the adequate use of a glucose sensor. Linearity depends, among other things, on the diffusion resistance of the sensor membrane to glucose.16 To achieve this, several materials, e.g., ~ilanes,'~Ja can be used. In this study we treated the BC membrane with an aromatic polyamide, which is normally used in the production of dialyzers for kidney dialysis. Immediately after harvesting, the BC membranes were dried under tension and then put into a matrix, with several circular gaps (4 2 cm). To cover the bacterial cellulose membrane with Trogamid, a lO%Trogamid solution (1.0 g of Trogamid/lO mL of dimethylformamide) was used. To deposit one layer on the membrane, 200 pL of the Trogamid solution was spread consistently over the membrane area of 2cm diameter. Afterwards the membrane was dried for 30 min at 50 "C. To obtain a Trogamid layer consisting of 400 and 600 pL of Trogamid on the membrane, this procedure was repeated twice. This treatment did not signiscantly affect the sensitivity and response time of the sensor. Glucose Sensor Buildup. For the determination of glucose concentrations, we used an amperometric system containing a platinum working electrode, set at +600 mV, and a silver/silver chloride reference electrode. These two components are parts of a flow-through cell developed at Priifgeratewerk Medingen.lg On top of the working electrode, we attached a membrane sandwich by means of an O-ring. There the enzyme glucose oxidase (GOX) is embedded in a polyurethane (PU)layer between two Cup membranes.20After diffusion of glucose into the enzyme layer compartment, glucose is catalytically transformed to gluconic acid ,andHzOz. At a potential of +600 mV, HzOz is then reoxidized to 0 2 . The resulting current is directly proportional to the glucose (14) Oppermann, M.; Schulze, M.; Gotze, 0. Complement Infimmation 1991, 8, 13-24. (15) Oppermann, M.;Haubitz, M.; Quentin, E.; GBtze, 0. Min. Wochenschr. 1988,66, 857-864. (16) Scheller, F.; Schubert, F. Biosensoren; Brkhiiuser-Verlag: Basel, Switzerland, 1988. (17) Amine, A; et al. Anal. Chim. Acta 1991,242, 91-98. (18) Mullen, W. H.; Churchouse, S. J.; Keedy, F. H.; Vagdama, P. M. Anal. PYOC. 1986,23, 145-146. (19) Bertermann, IC; Scheller, F.; Pfeiffer, D.; Jhchen, M.; Lutter, J. 2.Med. Laboratoriumsdiagn. 1981,22, 83-88. 1046-1052. (20) Olsson, B.; et al. Anal. Chem. 1986,58,
Analytical Chemistry, Vol. 67, No. 2, January 15, 1995
467
Figure 1. Scanning electron microscope (SEM) picture of a bacterial cellulose (BC) membrane (left) and a Cuprophan (Cup) membrane (right) without incubation (top) and after 2 h of incubation in 7% bovine serum albumine solution (bottom).
concentration. The classical glucose sensor possesses a dialysis membrane (Cup) as the outer membrane. For our investigations in diluted and undiluted human blood, we compared this classical sensor with a sensor in which the outer Cup membrane was replaced by the BC membrane. The nonTrogamid covered site was the blooddirected site. Glucose Measurementsin Diluted and Undiluted Human Blood. For experiments in undiluted blood, the tubing system of the experimental setup was rinsed first with heparin solution (5 units/mL) and then with sterile normal saline solution. For measurements in 1:lO diluted blood, we used sterile normal saline solution for rinsing only. The blood samples used for these experiments were obtained from the Department of Internal Medicine, University of Tubingen, in heparinized syringes. To define the lifetime of the sensor unequivocally,we choose the time of 50%initial activity. The activity of the enzyme electrode after preparation and conditioning was 100%by definition. To determine the sensor activity, we compared the sensor signal for a standard glucose solution (1mM) before and after blood contact. The sensor activity was determined every second hour of contact with diluted or undiluted human blood, respectively. With the bacterial cellulose in undiluted blood, however, the sensor activity was determined after every fifth hour. sensor activity = (dl/dt),/(dI/dt),.,
x
100%
For the quantitative detection of glucose, we monitored the time dependence of the sensor signal, i.e., the current I@), after 468 Analytical Chemisfry, Vol. 67, No. 2, January 15, 1995
exposure to glucose and chose the initial slope dI/dt as the sensor signal, which was found to be proportional to the glucose concentration. Animal Experiments. The animal experiments were performed on male Wistar rats with an average body weight of 500700 g. (1) Studies with Anesthetized Rats. The first set of experiments was performed on rats which were anesthetized during the duration of the experiment (6 h) with a mixture of Rompun and Ketanest. In these rats, one of the jugular veins was connected to the glucose sensor with a special cannula. For these experiments we used a membrane sandwich, consisting of CupPU/GOX-Cup BC (400 pL of Trogamid)-BC. Glucose measurements were performed every 15 min. During the time periods when the blood glucose was not being determined, the sensor system was rinsed with sterile normal saline solution. In this manner we monitored the blood glucose concentrations in three individual rats over a period of 6 h. (2)Studies with Free-MovingRats. For glucose sensoring over a period of 3 days, a second setup of experiments was performed, keeping rats in a free-moving system. Here the rats were anesthetized with a mixture of Rompun and Ketanest only for the insertion of a Teflon catheter near the cardiac atrium. The catheter system was then led through the fatty tissue near the throat to the back of the neck and connected with the swivel of a so-called free-moving system. This swivel is placed at one end of a rod, and the other end of the rod is fixed to the rat’s cage. By
Table 1. Effect of Cuprophan and Bacterial Cellulose Membranes on the Actlvatlon of the Complement System
a.
490
0
activation (ng/mL) C5a Generation negative control zymosan=
H
104 2369 775
Cuprophanb
571
BCc BC-SDSd
335
C3a Generation negative control
Cuprophanb
3.57 19.53 16.78
BCC BC-SDSd
12.40
zymosana
12.45
BC + 200 pI Tro BC 0
2
4
6
0
10
12
14
16
glucose [mmoUI]
Activation of complement system by 5 mg/mL Zymosan (positive control, 100%). The test materials were incubated for 2 h with human blood serum. * Cuprophan membrane 25 mg/mL serum. Bacterial cellulose 25 mg/mL serum. Bacterial cellulose 25 mg/mL serum, treated with sodium dodecyl sulfate (SDS), a
using this freemoving system,it was guaranteed that the rat could reach water and food, but it was impossible for the animal to damage its catheter. After this operation, the animals were allowed to recover overnight. During the recovery period of 12 h, a 0.9% NaCl solution (0.5 mL/h) was infused through the catheter to prevent the obstruction of the catheter system. To monitor the blood glucose levels, glucose determinations were performed every 15 min over a period of 6 h/day. During the night period (12 h), 0.9% NaCl solution was again infused into the rat. RESULTS AND DISCUSSION
Membrane Properties. Before the BC could be used as the outer membrane of the sensor, it was necessary to obtain some information about its membrane properties, especially in comparison to the Cup membrane. (1) Surface Structure of Cellulose Membranes. In Figure 1, the SEM pictures of a BC (left) and a Cup (right) membrane are shown. It is obvious that BC membranes exhibit a typical fiber structure, resulting from the cellulose chains. Using the same magnification, Cup membranes show no fibers. The wrinkle structure seen in Cup membranes is not derived from fibers but is the result of the washing and drying process during sampling preparation for SEM, which was not observed in BC membranes. When cellulose membranes are exposed to blood, it must be expected that protein adsorption takes place. To investigate whether or not there are differences in protein adsorption between the two membranes, we also examined the surface structure of both cellulose membranes by SEM after both were incubated in a protein solution. Figure 1 shows that, whereas the incubated BC membrane exhibits a widespread distribution of proteins, on the incubated Cup surface, proteins are deposited as clusters. (2) Hemocompatibility. It is very well known that if blood comes in contact with artificial surfaces, immunological reactions are induced. One of the parameters for immunological responses which is very sensitive is the generation of complement factors, i.e., C3a and C5a. When Zymosan was used as a complementinducing reference (loo%),Cup membranes exhibited generation of C3a and C5a of only 80%and 30%,respectively (Table 1). As is
b.
0 BC + 400 pl Tro BC + 600 pI Tro 0
50
100
150
200
250
glucose [mmolll]
Figure 2. Linearity range of glucose sensors covered by bacterial cellulose membranes: (a) without (O),and with 200 pL (H) of 10% Trogamid solution per 4 = 2.5 cm; (b) with 400 (0)and 600 pL (0) of 10% Trogamid solution per 4 = 2.5 cm. Sensor output in the presence of various glucose concentrations is given as nNs.
also shown in Table 1, the BC membrane produced 30%less C5a than the Cup membrane. Similar results were obtained using a C3a assay. This suggests that BC membranes possess better hemocompatibility than cup membranes. (3) Sensor Linearity Range after Membrane Modiflation with Trogamid. The measurements of long-term stability in vitro (Figure 3) were performed with only one glucose concentration (1 mM). In order to achieve linearity over a range sufficient to accurately measure variations of glucose concentrations in vivo, we modfied the glucose sensor by introducing an additional BC membrane, covered with Trogamid. In a SEM picture of bacterial cellulose covered with 200, 400, and 600 pL of Trogamid, it is obvious that BC membranes covered with the highest quantity of Trogamid became completely smooth, indicating increased diffusion resistance (data not shown). BC membranes covered with 200 pL of Trogamid showed linearity of glucose measurements within a range of 0-3 mM glucose (Figure 2a). When the amount of Trogamid (Figure 2b) was increased to 400 pL, linearity was achieved in a range of 0-40 mM glucose, which is sufficient for blood glucose measurements in normal and diabetic patients. Moreover, it was possible to further increase the linearity range when the BC was covered with 600 pL of Trogamid. Here linearity was achieved between 5 and 170 mM glucose. Analytical Chemistry, Vol. 67, No. 2,January 15, 1995
469
a.
120,
+ rat I 4 rat 2
16 15
rat 3
12 11
.-b5 .-
c.
u m
1 8 -am Cuprophan@
+ Bacterial cellulose 50
0
100
:I
,
,
,
,
,
,
150
0
200
,
,
,
,
, ,
,
,
,
,
,
,
.
,
,
,
3
4
5
6
time [h]
Figure 4. Monitoring of blood glucose in anesthetized rats over a period of 6 h, employing a glucose sensor with a membrane sandwich consisting of Cup-PUIGOX-Cup-BC(400 p L of Trogamid)-BC. Measurements were performed every 15 min. During the time periods when blood glucose was not being determined, the sensor was kept in 0.9% NaCI. The graph shows the results of three individual experiments, carried out over a period of 6 h.
b.
_I
, 2
1
time [h] 120
,
1 0
u m
P
c
40
50%-level
20
e
0 0
5
10
Cuprophan Bacterial cellulose 15
20
25
30
time [h]
c
Figure 3. Long-term stability of a glucose sensor covered with Cup and BC membranes (a) in diluted (1:lO) blood and (b) in undiluted blood. Sensor activity is 100% immediately after its preparation and conditioning in a standard glucose solution (1 mM). Each value of the curve represents a measurement of the standard glucose solution (1 mM) after a 2-h period of exposure to diluted blood. The sensor was regarded as being stable when its activity was more than 50%. For this experiment, the sensor was permanently kept in 1:lO diluted blood. It was disconnected from the blood stream only for the determination of the standard glucose concentration.
Figure 5. Monitoring of blood glucose in nonanesthetized rats, employing a glucose sensor with a membrane sandwich consisting of Cup-PUIGOX-Cup-BC(400 p L of Trogamid)-BC. Measurements were performed over a period of 6-8 h on three successive days.
From the above results, it is evident that linearity of glucose measurements increases with the amount of Trogamid used for covering BC membranes. For the in vivo studies, BC membranes covered with 400 p L of Trogamid were sufticient. For use as the outer membrane of the glucose sensor, the membrane was added to the sandwich in such a way that the Trogamid-treated site did not come in contact with the blood. Sensor Stability. (1) Studies in Vitro. When the BC membrane was used without Trogamid, the long-term stability of a glucose sensor was tested in the presence of 1mM glucose in comparison to a glucose sensor, covered by a Cup membrane, over a period of 200 h in diluted blood (1:lO). Whereas the stability of a sensor covered by Cup lost 50% of its activity after 30 h, the activity of the sensor covered with the BC was still 60% after 200 h of contact with diluted blood (Figure 3a). In Figure 3b, the long-term stability of both sensors in undiluted blood is shown. The sensor activity of the Cupcovered sensor decreased to 50% after about 3 h. With the sensor covered with the BC, a 50%loss of activity was observed only after 25 h. These data clearly show that the properties of the BC membrane allow
marked increase of the long-term stability of the sensor system, even when permanently exposed to undiluted blood. The stability of the BC membrane was not changed in the presence of ascorbic acid in a concentration l@foldto physiological blood levels when the glucose sensor was used in the presence of 1-4 mM glucose (data not shown), suggesting that, especially at low glucose, there is no interference by ascorbic acid which might lead to false glucose measurements. (2) Studies in Vivo. For in vivo experiments we used the glucose sensor, including a BC membrane covered with 400 pL of Trogamid. Figure 4 shows the blood glucose levels of three individual anesthetized rats over a period of 6 h. In separate experiments, blood glucose determinations at the same time points were performed using the usual Beckman Analyzer 2. It turned out that blood glucose levels tested by this method and blood glucose levels tested by the glucose sensor were similar (data not shown). Variations of blood glucose seen in Figure 4 are therefore individual changes and are not due to instability of the sensor. The same holds for the data shown in Figure 5.
470 Analytical Chemistry, Vol. 67, No. 2, January 15, 7995
:/drl 0
lpr"7p.m.
day2
day3
Ip.m.-7p.m.
Ip.m.-7p.m
Time of experiment
In vivo experiments were also carried out on nonunesthetized rats on three different days. The data are shown in Figure 5. Here the data obtained with the glucose determination using the Beckman Analyzer 2 were similar to the data obtained with the BC membranecovered glucose sensor (data not shown), indicating that the activity of the sensor remained constant over a period of 3 days, including a measuring time of 6-8 h/day and a storage time of 12 h/day in the refrigerator. In conclusion, the use of BC membranes to cover the classical amperometric glucose sensor improves its long-term stability in vitro and in vivo as well as its hemocompatibility. These properties may be due to differences in surface structures between BC and Cup membranes.
ACKNOWLEWMENT
The authors are most grateful to Dr. Bauser from the FHG/ IGB, Stuttgart-Vaihingen, for his readiness to support them with devices and bacterial cellulose. Also, the authors wish to thank Dr. Reinauer from the Department of Internal Medicine, University of Tubingen, for delivery of blood samples and Dr. Flosch from the Institute of Physical and Theoretical Chemistry for helpful discussions. Received for review March 11, 1994. Accepted October
27, 1994.@ AC940246K @Abstractpublished in Advance ACS Abstracts, December 1, 1994.
Analytical chemistry, Vol. 67, No. 2, January 15, 1995
471
