INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS AND DISUSSION REFERENCES

THERE IS A GROWING INTEREST in the development of sensors that can be implanted in tissues for long-term monitoring of metabolites. One class of sensors potentially useful for this application is based on electrochemical sensing of oxygen, including sensors for oxygen itself and oxygen-based enzyme electrode sensors for glucose, lactate, pyruvate, and other key metabolites. Sensors of this type are being developed for a variety of clinical applications, including glucose monitoring in diabetes; lactate monitoring in cardiac therapy, shock, and extreme exertion; oxygen monitoring for feedback control of variable-rate pacemakers; and other applications. Signals reported from these sensors are intended to relate the local tissue concentrations of the analyte of interest to the respective systemic blood concentrations.

Although the development of the sensors themselves is advanced in certain cases, there have been difficulties in validation of these sensors when implanted in tissues and in establishing reliable relationships between the sensor signals and systemic blood concentrations. Some studies have employed an experimental approach in which fine, needlelike sensors are implanted in animals (e.g., Ref. 14) or directly in human cutaneous tissues for several days (e.g., Refs. 9 and 18). These studies have shown that the concentration-dependent sensitivity and response characteristics of sensors are altered when sensors are used as implants compared with their respective in vitro properties (8, 19). However, this experimental approach provides little information about the physiological factors and phenomena characteristic of local tissues that are responsible for modification of the sensitivity and response. This testing approach has also led to inconsistent results when multiple sensors implanted in a given human subject are tested simultaneously (13). These results suggest that a more effective approach is needed.

Several approaches have been proposed to study the general tissue response to implants that may be applicable to implanted sensors. A cage implant system has been employed for studies of cellular responses to implanted materials (12). This device is a small, stainless steel cage into which the material of interest or sensor can be placed, and the entire system is implanted. Tissue grows into the cage mesh, and a reservoir of fluid transudate develops around the implant that can be serially aspirated and analyzed for cellular and protein content. Although this approach is useful for studies of amoeboid cell responses to materials, it is less useful for validation of sensors where the tissue must make close contact with the implant. An alternative approach involves the study of excised tissue mounted in a standard diffusion cell (15-17). This approach has been used in a histological evaluation of mass transfer properties of fluorescent tracer molecules in tissues, from which steady-state mass transfer properties of glucose have been estimated. The studies were, however, carried out without benefit of actual working sensors, did not involve actively perfused living tissues, and did not allow for serial visualization of tissue to document structural changes with time. This approach therefore precludes the ability to correlate sensor signals with metabolite vascular concentrations, study of the effects of local and regional microvascular perfusion, or monitor possible changes in tissue morphology.

New approaches to sensor testing and validation should take into account the unique characteristics of tissues. One such characteristic is the heterogeneous spatial distribution of metabolites in tissues. It is well documented that oxygen distribution in tissues is heterogeneous, as shown by a variety of experimental methods including rodent window chambers and acute applications of oxygen microsensors (4). There are reasons to suspect that other small metabolites such as glucose and lactate, which also diffuse rapidly and may be locally produced or consumed, may also be heterogeneously distributed in tissues. Sensors and experimental test systems for validation of sensors in tissues should therefore be designed to account for this possibility.

Temporal variation in the signals of implanted sensors is another important consideration for validation of sensor performance. Temporal variations originate as transients in the systemic blood concentration of the metabolite, but as blood carrying the metabolite reaches the tissues in which the sensor is located, a variety of factors can affect the resulting metabolite-dependent signal. These factors include variations in regional or local perfusion, lags due to metabolite diffusion to the sensor from nearby microvessels, and long-term changes in microvascular patterns, tissue architecture, or function. These effects can result in complex transient signals reported by the implanted sensors and complicate the interpretation of the sensor signal. The experimental preparation used for sensor validation must provide a means for continuous recording of these signals, exploring their range of dynamic variability, identifying the mechanisms of underlying transient components, and relating the signals to steady state and transient systemic blood metabolite concentrations.

The experimental preparation should also provide a means to quantify any modifications of the inherent sensitivity of the sensor itself as a result of long-term use and exposure to tissues. Modifications could be the result of electrochemical poisoning of the sensor, imbibition of lipids into the sensor membranes, loss of immobilized enzyme activity, or other processes. The preparation should permit sensor retrieval after experiments for in vitro comparison of preimplantation and postexplantation sensor characteristics. Identification of changes, if any, may suggest approaches to improve the sensor design, materials, or fabrication.

The ideal system for tissue sensor validation would allow nondestructive visualization of living tissues, mobile cells, and microvascular structures in the immediate vicinity of the implanted sensor while simultaneously recording dynamic sensor signals and measuring or controlling systemic metabolite concentrations. It is important that the experimental system not require anesthesia to avoid depressive effects on microcirculatory function or direct inhibition of sensor function. An experimental system in which repeated observations could be carried out over a period of weeks to months would be highly advantageous.

A window chamber system that meets these requirements is described here. This experimental apparatus is implanted in hamsters and used in conjunction with an array of independently connected potentiostatic oxygen sensors of a type documented to have long-term stability in vitro and in vivo. Examples are given of tissue visualization, sensor calibration, and signal recordings obtained with this system. The development of this experimental tool and the associated technologies sets the stage for subsequent extensive studies of biological factors that affect sensor response and validation of various implanted oxygen-based electrochemical sensors.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS AND DISUSSION REFERENCES

Window chamber. The hamster dorsal skin fold chamber (3, 5) was modified to support a 100- to 200-µm-thick layer of exteriorized subcutaneous connective tissue and muscle tissue and accommodate a disc sensor array on one side and a glass window on the other side. The apparatus, shown in Fig. 1, consisted of two titanium alloy frames, each having a 12.0-mm-diameter circular opening fitted with a ring for attachment of the window or sensor array. Each ring projected 0.95 mm toward the tissue and, in conjunction with appropriate spacers to separate the frames, produced sufficient separation to allow vigorous perfusion of the microvasculature in the thin sheet of tissue. One ring had three small pins that penetrate the tissue and mate with holes of the opposing ring to restrict tissue movement. A glass microscope coverslip was secured into the window of one frame, and the sensor array disc was secured into the other frame using slotted retaining rings. The edges of the coverslip and sensor disc were coated with silicone adhesive before insertion into the ring to provide a critically important seal. In some experiments, an additional 100-µm-thick plastic disc was glued to the internal surface of the window to minimize the thickness of the fluid layer between the tissue and sensor array surfaces. The frames were held together with four 10-mm screws and 4-mm hexagonal nuts. Before implantation, the frame surfaces, coverslip, and screws were autoclaved and rinsed extensively with sterile saline. The sensor array was sterilized by soaking in 6% glutaraldehyde solution, followed by rinsing in several changes of sterile saline solution over a 24-h period.

View larger version (166K): [in this window] [in a new window]   Fig. 1.   Window chamber mounted on the back of a hamster. The flange at the top supports a fan connector (not shown). Frame dimensions: 5.0 cm in length and 5.0 cm in height.

Implantation procedure. All experiments were conducted under ethical guidelines suggested by the American Physiological Society and using protocols approved by the University of California-San Diego Animal Subjects Committee. Male Syrian golden hamsters weighing 50-120 g were injected intraperitoneally with 50 mg/kg pentobarbital sodium. After a surgical plane of anesthesia was reached, the dorsum was shaved with electric clippers, coated with a depilatory for 5 min, washed with saline, and swabbed with Betadine solution. The dorsal skin was drawn up into a tall, thin fold that extended caudally along the midline. Three strands of sterile 4-0 silk were drawn through the top edge of the skin fold at both ends and the middle to support the fold. A 14-mm-diameter disc of skin was cut away from the middle of the fold on one side, and the coverslip frame was sutured in place. The periphery of the sealing rings was coated with broad-spectrum antibiotic ointment. A disc of skin and retractor muscle corresponding to the window dimensions was removed, and the sensor array was secured to the frame already in place using screws and nuts. Two 23-gauge stainless steel cannulas were inserted into the chamber, one on each side of the tissue, to fill the chamber with sterile saline and remove air bubbles, after which the cannulas were removed and the nuts on the chamber frames were tightened. Sutures of 4-0 silk were added to help secure the frames and provide support to the skin fold. Tissue blood flow was checked using intravital microscopy, and the spacing of the frame apparatus was adjusted if necessary.

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Schematic cross section of the skin fold. Specific tissue layers are excised to permit contact of the sensor array with subcutaneous tissue (left) or skin muscle (right).

Stability of potentiostatic oxygen sensors. For the extended study periods used here, it is essential that stable oxygen sensors are employed. We have previously described a three-electrode potentiostatic oxygen sensor system that is substantially more stable during continuous operation than the well-known Clark oxygen sensor (2). The electrodes are maintained in ionic contact via a thin layer of electrolyte and isolated from the analyte medium by a thin, pore-free layer of silicone rubber, which is highly permeable to oxygen but impermeable to polar compounds and charged chemical species. The potentiostatic oxygen sensor has been demonstrated to be stable to within a few percent over periods of several months in vitro (11) and in vivo (1). The potentiostatic oxygen sensor principle was used here.

Sensor array fabrication. Sensor arrays were fabricated by thick film deposition of platinum paste in a specific pattern on an alumina disc substrate, followed by baking at 700°C (10). The sensors were composed of disc platinum working electrodes of 125 µm diameter, common platinum disc counterelectrodes ~875 µm in diameter, and a ribbonlike common potential reference electrode created by electrodeposition of silver on the platinum reference electrode base, followed by chloridization to the form the Ag/AgCl junction. There were 18 working electrodes separated from each other by distances of 1-2 mm and 6 counterelectrodes on each array. A 25-µm layer of conductive electrolyte (1.0 N NaCl, pH 7.3, in cross-linked polyhydroxyethylmethacrylate gel) was deposited by spin casting on the alumina substrate, followed by spin casting of a 25-µm layer of polydimethylsiloxane. An individual wire connection was made to each sensor from the back of the array disc via an alumna plate having patterned conductive traces to a multipin fan connector, which was connected to a multichannel potentiostat. Data were archived by computer. The sensor array and connectors are shown in Fig. 3, A and B.

View larger version (123K): [in this window] [in a new window]   Fig. 3.   A: oxygen sensor array and connectors. Bottom right, alumina disc (12.0 mm diameter) with a patterned sensor array attached to a rectangular connector plate. The disc has laser-drilled holes for metal feedthroughs that connect individual sensors to leads on the back of the disc. Patterned sensor components were applied by a thick film batch process. Bottom left, the obverse, which shows an insulated connector plate with wire connections and a snap ring for attachment of the disc to the window chamber. A dime is shown for size comparison. Top, fan connector for access to multichannel instrumentation. B: close-up view of the oxygen sensor array. Left, sensor array with small (125 µm diameter) independent platinum working electrodes, large (875 µm diameter) common platinum counterelectrodes, and a curved common Ag/AgCl reference electrode. The membrane is not present. Right, the obverse. Spotwelded 25-µm gold wires connect electrode feedthroughs at the back of the disc to strip leads patterned in the connector plate.

Sensor characterization in vitro. The present study relies on precise calibration of individual sensors rather than highly reproducible sensor fabrication. The in vitro response of individual sensors was determined before implantation, used as a calibration standard for in vivo studies, and determined again in vitro after explantation to verify sensor stability. Sensors were characterized for sensitivity and linearity to oxygen in the stirred liquid phase by exposure to stirred solutions equilibrated with analyzed mixtures of oxygen and argon. Calculations of oxygen concentrations were based on analyzed mixed gasses, corrected for barometric pressure and humidity. Background signals in the absence of oxygen were determined by exposure of the sensors to buffer solutions saturated with sodium disulfite. All studies were carried out at 37°C. Electrode area and membrane thickness were determined by microscopic examination.

Oxygen recording in vivo. For signal recording and tissue observation, the unanesthetized hamsters were guided into a perforated plastic cylinder having a longitudinal slit from which the chamber frame assembly protruded. Animals typically rested quietly or slept during these periods. Sensor response was demonstrated by step and ramp changes in inspired oxygen concentration from atmospheric (21% oxygen) to 15% oxygen. Sensors were disconnected electrically between recording sessions and hamsters were returned to their cages.

Tissue visualization. Microscopic images of the tissue were obtained with epi-illumination in most cases, although nonuniform transillumination was possible for certain sensors by directing the optical source through the alumina disc from behind. At the termination of certain experiments, animals were placed under deep anesthesia and injected with an osmotically neutral solution of India ink to highlight the microvasculature. All images were captured with a digital camera.

	RESULTS AND DISUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS AND DISUSSION REFERENCES

Tissue visualization. Examples of tissue visualization at 2 wk are given in Fig. 4. Figure 4, left, shows vascularized tissue adjacent to a sensor array as viewed by transillumination from behind the array disc. Figure 4, right, is a magnified view (×4) of the indicated area, which contains a disc working electrode. Several feeding vessels and capillaries can be seen in the vicinity of the sensor. Transillumination has the advantage of high contrast but the disadvantage of nonuniformity due to the presence of features on the back of the array disc. The contrast between the blood vessels and surrounding tissue facilitates determination of the local capillary density. Tissue in the window chamber typically remains clear and transparent for periods of weeks to months, as shown by the image in Fig. 5, which was taken 4 wk after implantation (arrows indicate working electrodes). Figure 6 is a postmortem image in which the tissue has been perfused by the injection of India ink. The ribbonlike reference electrode is visible as a shadow in the bottom left quadrant. This image, although of nonliving tissue, is useful to indicate capillary networks. These imaging methods described above can provide morphological information for construction of models of mass transfer to sensors and of microvascular remodeling.

View larger version (93K): [in this window] [in a new window]   Fig. 4.   Sensor array and tissue in the window chamber. Left: window chamber at 2 wk after implantation seen by transillumination through the sensor array disc. Membrane-covered electrodes are visible through the tissue. Arrows indicate working electrodes. Right: magnification (×4) of the indicated rectangular area, showing a single membrane-covered working electrode (dark central disc) and the nearby microvascular pattern.

View larger version (115K): [in this window] [in a new window]   Fig. 5.   Window chamber at 4 wk. Small working electrode discs (arrows), large counterelectrode discs, and the curved reference electrode are visible through the tissue (×7.5 magnification, epi-illumination).

View larger version (131K): [in this window] [in a new window]   Fig. 6.   Postmortem injection of India ink reveals microvascular patterns. A ribbonlike reference electrode is visible as a shadow in the bottom left quadrant.

In vitro and in vivo sensitivity of oxygen sensors. Recordings of oxygen sensor signals were obtained repetitively after various periods of implantation. A full review of these data will be presented and analyzed in subsequent communications, but examples are presented here. The preliminary results presented in Table 1 indicate the type of useful information obtainable from the system. Table 1 contains steady-state current results of selected sensors from a given implant under three conditions: preimplantation, postexplantation, and implantation at 4 wk, with the means and SDs of each group. All signals were collected by exposure to atmospheric oxygen concentrations either directly or by inspiration. The postexplantation results show a mean and SD comparable with preimplantation results. The means and SD of the differences between preimplantation and postexplantation are relatively small, indicating that these sensors were stable over the period of the study. However, the mean of the in vivo measurements is substantially lower and the SD is larger. The difference in means between in vitro and in vivo measurements is an indication of magnitude of diffusional and oxygen solubility differences between in vitro and in vivo conditions but cannot be ascribed to changes in sensor sensitivity. The relatively large SD of in vivo measurements is an indication of oxygen spatial heterogeneity. The data from this experimental subject are representative of a subset of recordings from various animals obtained at various periods after implantation. These preliminary studies set the stage for extensive further observations and quantitative modeling studies of these effects to be reported in subsequent communications.

Dynamic response of implanted oxygen sensors. An example of a signal recording from an implanted sensor array at 2 wk is given in Fig. 7. Simultaneously recorded signal currents are shown as a function of time in response to step changes in inspired oxygen from 21% to 15% and back to 21% (arrows). The recorded signals display a range of sensor responses to tissue oxygen even though the sensitivity of sensors in vivo was very close, reflecting the heterogeneous distribution of oxygen in tissues. As the intent here is to display an example of a dynamic response, the respective signals were not adjusted based on in vitro calibration. Under the conditions shown here, there appear to be slight differences in the time for signals to rise or fall after the change in inspired oxygen, although the responses may include delays due to pulmonary exchange and transport via the blood to the tissue. Some signals also display characteristic temporal variations. This example is typical of many recordings with the sensor array. Further experimental studies and analysis are underway.

View larger version (31K): [in this window] [in a new window]   Fig. 7.   Examples of a recording from an oxygen sensor array at 4 wk. Inspired oxygen was abruptly changed from atmospheric (21%) to 15% and then back to atmospheric (arrows). Although the sensitivity of individual sensors was similar in vitro, differences seen after implantation are largely due to small local differences in tissues. For purposes of illustration, there was no attempt to adjust sensitivity based on in vitro calibrations.

Criteria for a successful window chamber preparation. Several qualitative criteria are indicative of a successful window chamber preparation. First, there must be vigorous and uniform perfusion of the microvasculature over the duration of the implant period. The level of perfusion depends to a great extent on matching of the spacing between the window chamber frames to the tissue thickness, which is related to animal body weight. For animals of a given body weight, adjustment of the spacing to assure maximal and unimpeded perfusion can be readily determined by experience. Second, the thickness of the liquid layer between the tissue and the sensor surfaces must be minimal to avoid interference due to oxygen mass transfer in the liquid layer and consumption by neighboring sensors. The liquid layer thickness can be reduced to several micrometers while maintaining vigorous microvascular perfusion by inclusion of a transparent plate of specified thickness glued to the internal surface of the window. The plate or similar artifice can provide a uniform support that positions the tissue very close to the surface of the sensor array. Our order-of-magnitude estimates of mass transfer suggest that interaction between sufficiently separated neighboring sensors is negligible and contributions to the diffusion lag of individual sensors are minimal when the liquid layer is thin. A third criterion for success is maintenance of clarity and transparency of the tissue for the duration of the experiment, which may be 1 mo or more. Transparency may be a compromised by collagen deposition, ingression of amoeboid cells, tissue remodeling, the presence of microhemorrhages or localized infections. Maintenance of tissue clarity is enhanced by the use of sterile procedure at implantation, minimization of tissue handling, and appropriate fit of the frame.

Range and response. Although there was no attempt in the present study to systematically adjust signals based on in vitro calibration values, these recordings show that there can be substantial differences in sensitivity to oxygen after implantation. The cause of these differences remains to be determined. These observations are nevertheless generally consistent with previous reports of heterogeneous oxygen distributions in tissue based on acute application of sensors (4) as well as reports based on other technologies (e.g., Ref. 7).

Advantages of the preparation. The window chamber and sensor array system has key advantages for sensor validation. The advantages derive from the ability to nondestructively visualize tissue structure and function in the vicinity of individual sensors without the inhibitory effects of anesthesia while simultaneously recording oxygen flux at multiple sites. This allows detailed study on the scale of diffusional distances of the local physiological phenomena that determine oxygen distributions and dynamics and their effects on sensor signals. The capability of maintaining functional oxygen sensors and the window chamber system for periods of 60 days or more is highly beneficial for studying changes in tissues that occur over the long term and measuring the effects of these changes on response of sensor implants. The potentiostatic oxygen electrode system used here is electrically isolated and does not convey current through the tissues. The oxygen-impermeable glass plate and ceramic sensor array on the respective sides of the tissue layer prevent oxygen access from the atmosphere, assuring that tissue oxygen is derived entirely from the vasculature. The oxygen sensors are not implanted directly in the tissue, as many previous electrochemical sensors have been, thereby avoiding localized tissue injury. With appropriate cell-specific markers, the movement of individual cells in the field can be studied (6). The nondestructive visualization capability of the window chamber (as opposed to observations with fixed tissues) is essential for study of dynamic events. The sensors can also be operated continuously or intermittently to allow study of the effects of sustained tissue oxygen consumption. The use of a sensor array is beneficial for determination of the range of signal differences within a single animal and makes for efficient study design. These features of the window chamber system constitute important advantages for sensor validation.