Enhanced Hemocompatibility and In Vivo Analytical Accuracy of Intravascular Potentiometric Carbon Dioxide Sensors via Nitric Oxide (NO) Release

Abstract

In this report, the innate ability of nitric oxide (NO) to inhibit platelet activation/adhesion/thrombus formation is employed to improve the hemocompatibility and in vivo accuracy of an intravascular potentiometric PCO₂ (partial pressure of carbon dioxide) sensor. The catheter-type sensor is fabricated by impregnating a segment of dual lumen silicone tubing with a proton ionophore, plasticizer and lipophilic cation-exchanger. Subsequent filling of bicarbonate and strong buffer solutions, and placement of Ag/AgCl reference electrode wires within each lumen, respectively, enables measurement of the membrane potential difference across the inner wall of the tube, with this potential changing as a function of the logarithm of sample PCO₂. The dual lumen device is further encapsulated within a S-nitroso-N-acetyl-DL-penicillamine (SNAP)-doped silicone tube that releases physiological levels of NO. The NO releasing sensor exhibits near-Nernstian sensitivity towards PCO₂ (slope = 59.31 ± 0.78 mV/decade) and low drift rates (< 2 mV/24h after initial equilibration). In vivo evaluation of the NO releasing sensors performed in the arteries and veins of anesthetized pigs for 20 h shows enhanced accuracy (vs. non-NO release sensors) when benchmarked to measurements of discrete blood samples made with a commercial blood gas analyzer. The accurate, continuous monitoring of blood PCO₂ levels achieved with this new IV NO releasing PCO₂ sensor configuration could help better manage hospitalized patients in critical care units.

Graphical Abstract

Blood gas (PCO₂, PO₂ and pH) analyses are essential for assessing a patient’s physiological status, especially in intensive care units and operating rooms [1]. Real-time blood PCO₂ monitoring is of particular interest, as hypercapnia and respiratory acidosis ensue from respiratory failure, which can be associated with severe conditions (e.g., chronic obstructive pulmonary disease, central nervous system depression, neuromuscular disorders, thoracic deformities, etc.) that require immediate therapeutic intervention [2, 3]. However, the clinical standard of care at present employs a benchtop blood gas analyzer to test arterial blood samples intermittently collected from patients, which may cause delayed diagnosis/therapeutics. Non-invasive, continuous end-tidal and transcutaneous approaches to PCO₂ measurements have been reported, but these approaches are error-prone [4, 5]. Hence, the availability of intravascular (IV) PCO₂ sensors that enable continuous real-time monitoring of PCO₂ with good accuracy is highly desirable.

Multi-probe IV blood gas monitoring systems with built-in fiberoptic PCO₂ sensors are used clinically but these are primarily employed only to monitor PCO₂ during extracorporeal procedures (e.g. open heart surgery), and they suffer from accuracy problems due to limitations (e.g., sensitivity to ionic strength, photo-bleaching and leaching of immobilized dyes, etc.) and geometrical obstacles (e.g., wall effect) [6–8]. Electrochemical systems, although offering immunity to many of the specific problems of their optical counterparts, have not attracted much attention to date for IV applications. Indeed, conventional Severinghaus-type glass pH sensor-based PCO₂ electrodes are too rigid/fragile to be incorporated into catheter structures for practical in vivo use [9, 10].

To tackle these issues, our group previously reported a catheter-type potentiometric PCO₂ sensor based on a simple concentric dual-lumen silicone tubing configuration, employing the inner wall doped with a proton ionophore tridodecylamine, a sodium tetrakis[3,5-bis(trifluoromethyl)phenyl]borate (NaTFPB) cation-exchanger and nitrophenyloctylether (NPOE) plasticizer, as the H⁺-sensitive membrane [11, 12]. When one lumen is filled with NaHCO₃ solution and the other with a strong buffer, the voltage across the inner wall between the two lumens changes in proportion to the log PCO₂ due to a pH decrease in the lumen filled with the bicarbonate solution. These IV PCO₂ probes produced accurate results in a heparinized dog model for 7 h. However, long-term hemocompatibility without the need for systemic heparinization to prevent clotting on the sensing catheter’s outer surface remains a major challenge for IV sensors placed within blood vessels. Indeed, the host’s response in the form of thrombus formation on sensor surfaces [13, 14] can alter the local concentration of CO₂ via cellular (e.g. platelet) metabolic activities and also limit analyte diffusion from the blood stream. Both processes can cause CO₂ levels in the region surrounding the sensor to deviate from their bulk blood levels and thereby yielding erroneous analytical results [15, 16]. Indeed, most preclinical/clinical studies for continuous blood gas detection in vivo employ heparinized saline flush and/or heparin modification of the sensors, and typically span only a short time frame (a few hours) and often require recalibration [6, 17, 18].

Nitric oxide (NO), an endothelium-derived molecule with anti-platelet/thrombotic and anti-inflammatory properties, has emerged as a promising solution to the problem of IV device biocompatibility. As demonstrated previously in multiple in vivo studies using animal models, physiological levels of NO emitted from amperometric IV glucose and PO₂ sensors could successfully reduce clot formation and improve sensor accuracy for continuous IV blood monitoring [19–24].

In this study we report, for the first time, the fabrication of a NO releasing catheter-type potentiometric PCO₂ sensor (Figure 1). The NO release functionality is derived from the incorporation of the NO-donating molecule, S-nitroso-N-acetylpenicillamine (SNAP), into a silicone tube that the original dual lumen potentiometric CO₂ sensor [11, 12] is inserted within. After initial benchtop studies, continuous in vivo blood PCO₂ monitoring using a porcine model for 20 h was performed to evaluate the accuracy and hemocompatibility of the new NO releasing PCO₂ sensor in comparison with a non-NO release control sensor (Figure S1(a) and (b)) tested in the same animals. It will be shown that the addition of the NO releasing silicone tube over the inner dual lumen PCO₂ sensor significantly enhances the analytical performance of the IV implanted sensors.

Figure 1.

Image (left) and cross-sectional schematic view (right) of a NO-releasing catheter-type potentiometric PCO₂ sensor.

Unlike a glass pH sensitive membrane in the Severinghaus style PCO₂ sensor, H⁺ sensitivity of the catheter-type sensor is derived from modifying a ca. 3 cm section of dual lumen silicone tube (5 cm) with an H⁺-sensing tridodecylamine ionophore, as well as NaTFPB as a lipophilic ion-exchanger and NPOE as a plasticizer to decrease resistance of the tube wall. The membrane potential difference across the H⁺-sensitive silicone inner wall between the lumens is related to the logarithm of sample PCO₂ (log PCO₂) (Figure S1(c)). In order not to disrupt the sensing chemistry within the wall between the lumens, the modified end of a 5 cm long H⁺-sensitive dual lumen tube was encapsulated in a NO releasing sheath consisting of a 2.2 cm length of a SNAP impregnated silicone tube that is sealed by a SNAP/silicone (35/65, w/w) mixture [25] on the tip (Figure 1). Due to SNAP crystal formation [26], SNAP doped silicone is expected to provide excellent long-term stability, and more importantly, sustainable physiological-levels of NO release [27] to potentially enhance hemocompatibility of the sensing section of the IV device when placed within blood vessels. The NO release rate from the SNAP in the wall of the outer tube via thermal decomposition was determined daily using a chemiluminescence nitric oxide analyzer (NOA) over an 8-day period (Figure S2). During this period, the NO flux was relatively stable and never fell below the lowest endothelial level (> 0.5 × 10⁻¹⁰ mol·cm⁻²·min⁻¹) [16], a flux reported to endow implantable SNAP-modified devices with significant antithrombotic/anti-inflammatory functions [21, 28–29]. NO release from the dome-shaped sensor tip (Figure S3) alone was examined, and was also > 0.5 × 10⁻¹⁰ mol·cm⁻²·min⁻¹ for at least 96 h, which should be adequate to inhibit thrombosis specifically initiated at the sensor’s tip.

Fabricated under optimized conditions (see Supporting Information), both the NO releasing sensors and controls (without SNAP) (Figure S1) were tested for simultaneous potentiometric PCO₂ response in a 25 mM NaHCO₃/120 mM NaCl calibration solution saturated alternately by certified CO₂ gases (±1% error) within a physiologically-relevant range of 5.0% to 15.0% (i.e. 38.0 – 114.0 mmHg) at 37°C. As illustrated in Figure S4, the sensors exhibit good reversibility and near-Nernstian sensitivity to PCO₂, without any significant difference in the slopes between the NO releasing sensors (59.31 ± 0.78 mV/decade, n=3) and controls (59.25 ± 0.71 mV/decade, n=3). The response time of the dual-lumen silicone PCO₂ sensor, defined here as the time to reach 90% of cell potential equilibration (t₉₀) for a PCO₂ step change from 38 to 76 mmHg, is ca. 4.8 min for control sensors, faster than ca. 6.1 min for the NO releasing sensors. This disparity is due to the hindrance of CO₂ diffusion by the added outer tube that contains the SNAP crystals in the outer silicone sheath. Response times on the order of several minutes are still clinically useful for continuous monitoring, especially when compared to ex vivo tests with a blood gas analyzer. Much faster response times could be achieved by reducing the diameter and wall thicknesses of the inner PCO₂ sensing dual lumen tube [11], while optimizing the size of outer SNAP doped tube for balanced rapidity in response time and NO flux.

The new dual lumen catheter PCO₂ sensors exhibit initial potential drifts when equilibrated in NaHCO₃ calibration solution of 5.0% CO₂ at 37°C. While the drifts of control sensor go through a relatively long equilibration phase (about 16 h) until a plateau is reached, similar to our previous report [11], the NO releasing sensors have very minor initial drifts that are completed within 8 h (Figure S5). Only negligible drifts/fluctuates were observed subsequently (< 2 mV/24 h) for both types of sensors. It is possible that osmotic process, caused by the osmotic pressure differences, is less significant in the NO releasing sensors due to the obstruction of water diffusion through silicone rubber by SNAP within the encapsulating tube [30–32]. The sensitivity for both controls and NO releasing sensors typically doe not decrease by any significant degree over the first 48 h, which is adequate for the time-frame (20 h after equilibration) of the in vivo animal studies conducted in this study (Figure S5 (a) and (b)). It is also noteworthy that the endothelial-level of NO release exhibited by the sensors does not systematically cause significant fluctuation of potentiometric signal or affect the long-term sensitivity of the underlying PCO₂ sensor over 96 h (Figure S5 (b)).

To evaluate the in vivo accuracy of the new NO releasing PCO₂ probe for continuous IV monitoring, three pairs of NO releasing and control sensors were tested in the jugular veins and femoral arteries/veins on both sides of each pig (n=3 pigs in total), with the distal section of ca. 2.2 cm in length placed within the blood vessels. The sensors were pre-conditioned and pre-calibrated in CO₂ tonometered NaHCO₃/NaCl solutions before implantation. The raw EMF potentiometric signal (mV) was converted into PCO₂ values (mmHg) using one-point calibration (see Supporting Information for details). Figure 2(a) illustrates continuous PCO₂ readings for a representative pair of NO and control sensors implanted in the femoral arteries of the same animal over a 20-h in vivo experiment, compared to values reported by a reference blood gas analyzer on discrete blood samples periodically drawn from the animals. The signal from the control sensors (blue line) started to deviate positively as early as the first 4 h of the experiment. Such a permanent upward shift is attributed to thrombus formation around the catheter surface (Figure 2(b)). Indeed, excess CO₂ generated by metabolic activity of platelets/cells within the surface clot layer creates a local environment with an elevated CO₂ level, driving monitored PCO₂ values falsely higher than the true values in the bulk blood [15, 16]. The decreased sensitivity and response time towards CO₂ challenges for the control sensors, as manifested by blunt peaks in Figure 2(a), results from hampered gas diffusion due to the isolation of sensor by the blood clots. Such an overall loss of accuracy for control sensors highlights the biocompatibility issue for IV PCO₂ sensors. In contrast, more rapid and accurate tracking of PCO₂ changes is observed for the NO releasing sensors (Figure 2(a), green line). Despite the dramatic changes in PCO₂ upon the manipulation of ventilation for a total of five CO₂ challenges, continuously monitored PCO₂ values always correlate well with discrete in vitro blood gas analyzer values throughout the 20 h, showing the robustness of NO-releasing sensors for PCO₂ measurements over an extended implantation period without the need for recalibration. The images of the NO releasing sensor after being explanted at the end of the run suggests minimal clot formation (Figure 2(b)), which agrees well with the excellent in vivo accuracy. Depending on different physiological status of the pigs (e.g. coagulation time, blood pressure, blood vessel size, etc.), controls could also accumulate minor clots of low surface coverage (Figure S6(b)) that cause a smaller baseline shift without serious sensitivity loss (Figure S6(a)). Indeed, individual sensors with high levels of deviation generally show significant clotting. Throughout all animal studies, no interference was observed by anesthetic gases and medicines (e.g., isoflurane and vecuronium bromide) introduced.

Figure 2.

(a) Example of response curves for a NO releasing sensor (green) and a control sensor (blue) for PCO₂ monitoring in pig femoral arteries compared to corresponding discrete blood gas analyzer values (red dots) over a 20-h animal study and (b) Images of NO releasing and control sensors explanted from pig femoral arteries after 20 h.

The numerical accuracy of 3× pairs of intra-arterial PaCO₂ sensors and 5× pairs of intravenous PvCO₂ sensors in three animal studies (after excluding 1× pair of femoral venous sensors that were accidentally misplaced) was evaluated separately by quantitating the mean absolute relative deviation (MARD) between in vivo values measured and in vitro blood sample values at the same time points. Each MARD value was calculated using data points for each type of sensor at 15, 30, 45, 60, 75 and 240 min for each 4 h test period (each CO₂ challenge) during the overall 20-h study. As illustrated in Figures 3(a) and (b), deviation generally increases over time for IV sensors due to clot formation. While control sensors in both arteries and veins exhibited a notable positive MARD (up to 20–30% at the 20-h mark), the corresponding NO-emitting sensors only exhibited an ca. 10% deviation from the reference values over the entire 20 h test period (p < 0.05 at each time point, n=18 for each type of PaCO₂ sensors and n=30 for PvCO₂ sensors in most cases). In fact, the MARD values for control sensors (especially severely clotted/impaired controls) can be understated here, as a positive PCO₂ deviation due to CO₂ produced by clots might be canceled out by diminished magnitude of the response to the CO₂ challenges because of slowed CO₂ diffusion, and this would reduce the overall MARD for controls. This phenomenon is different for IV PO₂ sensors, for which the consumption of surrounding O₂ and declined sensitivity collectively lead to a larger deviation over time [20, 21]. Average clotted area on explanted PCO₂ sensing catheters after 20 h shows close correlation with the MARD, as NO releasing sensors have considerably less clots on their surfaces than controls (Figure 3(c)). Clot reduction is particularly effective for NO releasing PaCO₂ sensors, probably due to a much higher blood flow in arteries. The remaining yet insignificant MARD and clotting for NO release sensors (especially those for PvCO₂) did not result from significantly decreasing NO release, as the NO flux of SNAP-modified sensors before and after the 20-h animal studies remain relatively constant (Figure S7). After all, only moderate decrease in NO flux should be expected in SNAP-modified devices after initial burst on the first day [24, 27]. However, events such as hypercoagulation in individual pigs and exceedingly fast changing of the blood CO₂ levels during these challenges did cause discrepancy in reported values for controls as well as the NO releasing sensors. The partial obstruction of blood flow expected from the current sensors with an outer diameter of 2 mm, despite NO release, could have also increased the likelihood of thrombogenicity and thus loss in sensor accuracy, as the sensors typically exhibit less clotting in blood vessels of higher flow rates and larger size.

Figure 3.

Mean average relative deviation (MARD) of PCO₂ values monitored by (a) NO releasing and control PCO₂ sensors in pig arteries (n=3), and (b) NO releasing and control PCO₂ sensors in pig veins (n=5) from the values measured by blood gas analyzer reference method at 4 h intervals over the 20-h animal studies, and (c) average clotted area on explanted NO releasing sensors vs control sensors from both arteries (left, n=3) and veins (right, n=5) after 20 h.

Clinical accuracy of the intravascular sensors was evaluated with error grids (Figure 4). On the basis of the periodic in vitro blood gas tests as the reference, 93.3% of the measurements from NO releasing PaCO₂ sensors (n = 89) were within ±20% error (known as the clinically accurate zone), while only 66.3% of the measurements from control PaCO₂ sensors were within ±20% error. Similarly, the results of NO releasing and control PvCO₂ sensors (n = 142) are 89.4% and 62.0%, showing a 27.4% suppression of accuracy for the intravenous controls. Based on linear regression (all p-values < 0.001), NO releasing sensors also exhibit substantially better correlation of R² = 0.864 (PaCO₂) and 0.742 (PvCO₂) with the blood gas analyzer reference method, than controls with R² of 0.177 (PaCO₂) and 0.382 (PvCO₂). Again, the generally superior results for PaCO₂ measurements than PvCO₂ with NO release is due to more effective clot reduction by NO release in arteries which normally exhibit much higher blood flow rates (Figure 3(c)). The overall relative deviations as calculated are −1.09% ± 11.52% (PaCO₂) and 3.94% ± 13.93% (PvCO₂) for NO releasing sensors, in sharp contrast to 10.07% ± 35.95% (PaCO₂) and 14.43% ± 26.41% (PvCO₂) for controls. Similarly, Bland-Altman plots also suggest the same trend as illustrated by much wider 95% limits of agreement for the control sensors (Figure S8). The above data further substantiate our hypothesis that thrombus formation and concomitant entrapped metabolically active cells on the control sensors can lead to considerable loss of their accuracy.

Figure 4.

Comparison of PCO₂ values measured by (a) NO releasing PaCO₂ sensors and PvCO₂ sensors and (b) control PaCO₂ sensors and PvCO₂ sensors against the values by a blood gas analyzer throughout the 20-h animal studies. Dashed lines and the solid line indicate 0% error and ±20% error.

Overall, the demonstrated improvement of PCO₂ analytical accuracy using NO release CO₂ sensing catheters during the proof-of-concept 20-h animal studies reported herein, provides the basis for a pathway toward creating continuous IV PCO₂ monitoring sensors for use in critically ill patients. The PCO₂ sensor delivered a strong performance (e.g., NO flux, in vivo accuracy and hemocompatibility, etc.) that is comparable to that of the best NO-releasing intravascular PO₂/glucose/lactate sensors we have previously reported [19–24].

The sensors are immediately applicable for ex vivo clinical procedures (e.g. extracorporeal circuits during cardiopulmonary bypass), where patients are systemically heparinized. But more importantly, with further miniaturization of the sensor configuration described herein, such PCO₂ sensors could be employed for continuous longer-term intra-arterial applications in critically ill patients. Other optimization strategies such as selecting more effective NO donors, and further combining NO release with surface immobilization of CD47 peptide [33] or anticoagulant [34], are also underway in our group for longer-term animal as well as potential clinical studies. Future designs may include a thermocouple for temperature correction. A pH sensing electrode will also be incorporated into the PCO₂ sensor, similar to our previous design [11, 12], for synergistically assessing acid/base homeostasis and the better management of a wide spectrum of medical conditions [35].