Redox Flow Iontophoresis for Continuous Drug Delivery

Drug delivery into the brain and spinal cord is fundamentally limited by the blood‐brain barrier which impedes the use of the vast majority of drugs. Implants based on iontophoresis use an applied voltage to deliver charged drug molecules, allowing solvent‐free delivery directly into the site of interest and overcoming issues associated with systemic exposure to the drug. However, during continuous delivery over long periods, electrochemical reactions occur at the electrodes leading to corrosive gas formation. Here, the concept of redox flow iontophoresis is presented, where a redox mediator solution is used to control electrode reactions and sustain continuous delivery for theoretically unlimited duration. As a proof‐of‐concept, a redox flow iontophoresis‐based brain implant that can continuously deliver the cancer drug doxorubicin at stable rates exceeding 2 nmol min−1 is demonstrated. This new concept enables the continuous delivery of various potent drugs into the brain and spinal cord and therefore has the potential to improve treatment options for various diseases.


Introduction
Iontophoresis describes the process of transporting charged molecules into the human body using voltage gradients. [1]Iontophoresis is typically performed by placing two electrodes onto a tissue and applying an appropriate voltage to sustain a DC current, driving charged drug molecules away from the electrode of the same charge polarization as the drug (working electrode) toward the electrode of opposite polarization (counter electrode).The technique has been widely investigated since the early 19th century, and a large volume of articles since then have been DOI: 10.1002/admt.202301641published about its use with a variety of drugs in applications such as transdermal, [2][3][4][5][6][7] transscleral [8,9] or intravesical [10,11] drug delivery.17] In contrast to pressure-driven bulk injection of drug formulations, also known as convection enhanced delivery, iontophoresis allows the delivery of charged drug molecules alone without solvent.Therefore, iontophoretic delivery is not associated with localized pressure increases in the brain tissue avoiding drug reflux, a major limitation of convection enhanced delivery, and promising delivery of larger drug quantities. [18]o sustain a DC electrophoretic current and allow continuous drug delivery, a mechanism for charge transfer at the electrodes which converts from electric to ionic charge carriers is necessary.The mechanism is usually electrolysis of water, however, this gives rise to corrosive gas evolution and pH drifts. [19]Hydrolysis is especially problematic for implantable devices as gas (Cl 2 and H 2 ) buildup within the implant can significantly reduce the drug delivery rate, the device can be severely corroded, and other Figure 1.Illustration of the redox flow iontophoresis concept for the delivery of cationic drug molecules.A redox mediator (yellow and orange spheres) flows past the working electrode and undergoes oxidation.This attracts drug counter ions (small blue spheres) from the adjacent drug flow channel through the ion-exchange membrane into the redox mediator channel.As a result, the cationic drug molecules (red spheres) are transported through the porous membrane into the patient tissue.The redox mediator counter ions are not depicted.Inset: Oxidation of the redox mediator ferrocyanide to ferricyanide.Note that also the inverse reaction is possible permitting delivery of anionic drug molecules when a negative voltage is applied at the working electrode.
potentially toxic reaction side products can form which need to be continuously removed.Alternative approaches have been proposed using redox active electrode materials such as Ag/AgCl, but while solving the gas evolution problem, these electrodes create biocompatibility concerns due to the possible release of Ag + ions. [19,20]An additional issue with conventional iontophoretic approaches is that the drug formulations are in physical contact with the electrode surface, where deleterious electrochemical reactions might reduce the efficacy or change the nature of the drug molecules.[23] Using such electrodes, however, imposes significant limitations, as the drug delivery ceases when the electrodes are fully oxidized or reduced.Here, we present redox flow iontophoresis, a concept that uses a redox mediator solution to facilitate charge transfer at the electrode/electrolyte interface of the working electrode of iontophoretic devices.Redox flow iontophoresis addresses the shortcomings of existing iontophoresis approaches allowing the safe and continuous delivery of drugs.

Redox Flow Concept
Our concept consists of a two-channel microfluidic setup visualized in Figure 1: An aqueous solution of a redox mediator molecule, which can be easily oxidized without the formation of side products, flows through a channel along the working electrode.This redox mediator channel is separated from a second channel, the drug channel, by an ion exchange membrane.The drug channel, through which a solution of the charged drug molecule flows, is in turn separated from the tissue by a porous membrane.When applying a positive voltage at the working electrode inside the device with respect to an external counter electrode, the redox mediator undergoes oxidation, creating a positive charge imbalance in the redox mediator channel.The ion exchange membrane in between the redox mediator and drug channels prevents bulk mixing of the two solutions.It is chosen such that molecules of the same charge as the drug cannot permeate (i.e., an anion exchange membrane for cationic drug molecules) preventing ions of the same charge polarization as the drug from being transported from the redox mediator channel into the drug channel which would compete in delivery with the drug ions.When a voltage is applied, anions are attracted from the drug channel through the ion exchange membrane into the redox channel creating a positive charge imbalance in the drug channel.Consequently, the cationic drug molecules migrate through an outer porous membrane toward the tissue.The continuous flow of both the redox mediator and drug solution ensures enough of both species remain available.Applying a negative voltage allows the delivery of anionic drug molecules.
We have identified the ferrocyanide-ferricyanide redox couple (see inset in Figures 1 and S1, Supporting Information) as a suitable redox mediator as both ions can be readily oxidized and reduced into each other, both are highly soluble in water, the charge transfer does not create any side products such as gases or products that may induce pH drifts, and the couple is nontoxic.6][27]

Implant Design and Fabrication
To demonstrate practical application of the redox flow iontophoresis concept we have fabricated workable implantable devices aimed at delivery of cationic drugs.The devices are specifically designed for drug delivery into the spinal cord and brain featuring a thin elongated body for easy implantation and large thin membranes allowing high drug delivery currents.The implants have one electric and four fluidic connectors (two fluid inlets and two outlets).We operate the devices with Ag/AgCl counter electrodes which can be placed on the skin allowing easy replacement when depleted (Figure 2A).As the counter electrodes are not implanted and operated as cathodes, their biocompatibility is not a concern.It should be noted that the devices can also operate with other types of counter electrodes.The implants consist of an additively manufactured polymer body which is 4 mm wide and 2 mm thick incorporating microfluidic flow channels (Figure 2B,C).We encapsulate the polymer body in a parylene C layer (2 μm) to improve biocompatibility.Then, we laminate the ion exchange and porous membrane to the coated device body to allow continuous flow of the redox mediator and drug solution through the microfluidic channels.The devices are driven by a graphite electrode around which the redox mediator solution flows in a U-shaped fashion to maximize contact area.Graphite is a common choice for redox chemistry applications due to its wide water window and chemical inertness.

Device Characterization
To study the drug delivery performance of our devices we conducted in vitro delivery experiments (see methods, below) using a doxorubicin formulation as a drug solution with a sodium ferrocyanide redox mediator solution.Doxorubicin, an anthracyclineclass cancer drug, is a fluorescent, cationic molecule and therefore a suitable drug candidate for iontophoretic delivery.We delivered the drug into NaCl electrolyte solutions recording the electrical current I(t) as a function of time t and quantified the drug concentration in the reservoir using fluorescence spectroscopy.

Dependence of Drug Delivery on Injected Charge
Figure 3A shows the delivered amount of doxorubicin versus the total injected charge Q = ∫ I(t) dt measured in multiple devices.The data suggests that the delivered amount and total injected charge are linearly correlated which agrees well with previous studies on hydrolysis-based transdermal iontophoretic delivery and organic electronic ion pumps. [23,28]Quantifying this correlation for individual devices is crucial as being able to calculate the amount of delivered drug from the recorded electrical current is a valuable tool for clinicians to continuously assess the progress of the drug delivery.We observed very large drug delivery rates of up to (2.3 ± 0.2) nmol min −1 for a medical doxorubicin formulation, a rate which is 1200 ± 300 times higher than for previously reported iontophoretic brain implants delivering the smaller sized cancer drug gemcitabine. [14]This significant enhancement in delivery rate compared to previous work was possible due to the large and thin membranes in our implant design creating a low ionic resistance and the redox flow iontophoresis approach being able to sustain high electrical currents.A high delivery rate is important to keep drug delivery courses in patients of short duration.We observed up to a 13 ± 1-fold increase in drug delivered with redox flow iontophoresis compared to passive diffusion.The error bars in Figure 3A represent the error of the individual drug concentration measurements and are an underestimate of the overall variability which is likely determined by factors such as membrane conditioning at the beginning of each experiment.
The delivery efficiency  is defined as the ratio of the number of drug molecules delivered per injected charge Q. Figure 3B shows the delivery efficiency versus the total injected charge.Despite the observed high drug delivery rates, we measured an average delivery efficiency of  = (5 ± 1.5) nmol C −1 (slope of Figure 3A) which corresponds to only one drug molecule delivered per 2100 ± 600 injected electrons.This is likely a consequence of using a non-ion selective porous membrane to separate the drug channel from the outside of the device which allows transport of small mobile ions (e.g., Na + , Cl − ) into and out of the drug channel competing with the less mobile doxorubicin cations.Choosing an ion selective membrane to separate the drug channel from the patient tissue in a future refinement of the implant will likely improve the delivery efficiency and increase the drug delivery rate.As shown in Figure S2 (Supporting Information), the drug delivery efficiency  appears to be uncorrelated with the applied voltage.

Dependence of Drug Delivery on Drug Formulation and Flow Rates
We further analyzed the impact of the salinity of the drug formulation on the delivery efficiency.An increase in concentration of NaCl in the drug solution makes the solution more conducting, hence increasing the voltage drop at the working electrode/redox mediator interface causing an increase in overall electric current.As shown in Figure 3C, we observed that the delivery efficiency  however decreases with increasing NaCl concentration.We interpret the decrease of delivery efficiency  with increasing NaCl concentration in the doxorubicin formulation as a consequence of the very mobile Na + ions being preferably delivered over the less mobile doxorubicin cations.A similar behavior has been described for traditional iontophoretic devices. [28]As an increase of NaCl concentration causes the total electrical current and therefore initially also the drug delivery rate to increase, adjusting the drug salinity is a trade-off between delivery efficiency and absolute delivery rate.
Finally, we investigated the dependence of the electrical current, and hence the drug delivery rate, on the ferrocyanide solution concentration, and on both the doxorubicin and ferrocyanide solution flow rates.As shown in Figure 3D, we observed that at constant flow rate the electrical current increases initially with rising ferrocyanide concentration and saturates when larger concentrations are reached.We interpret this behavior as a transition from a regime at low ferrocyanide concentration where the current is limited by the availability of unoxidized ferrocyanide to a steady state regime where the current is limited by the ferrocyanide reaction rate at the graphite electrode.This is supported by our observation depicted in Figure 4A that the electrical current saturates at high ferrocyanide flow rates when keeping the ferrocyanide concentration constant.We also observed a similar saturation behavior for the dependence of the electric current on the doxorubicin drug solution flow rate, Figure 4B.We interpret this current saturation with increasing drug flow rate as a consequence of the device continuously removing ions from the drug channel.Increasing the drug flow rate replenishes the ionic strength of the drug solution and therefore increases the device current.During our experiments, no anodic gas evolution was observed.

Optical Monitoring of Drug Distribution
The experiments described above quantify the total delivered drug amount into a liquid electrolyte.While this gives important insights into the impact of device operation parameters on drug delivery it does not allow to quantify the spatial drug distribution versus time.In order to investigate the spatial drug concentration field, we visually observed the delivery of the cationic dye molecule methylene blue into an agarose gel brain phantom placed in between the drug delivery device and an Ag/AgCl counter electrode.By applying an inverse Abel transform on the 2D absorbance data under the assumption of axial symmetry of the dye plume along the device axis, we recovered the methylene blue concentration field.We previously reported this setup and method. [29]Methylene blue has a very high molar absorption coefficient at wavelengths ≈660 nm and is therefore ideally suited for this optical quantification technique.Figure 5A shows the extracted concentration distributions at different timesteps with and without applied voltage (passive diffusion control).Integrating the absorbance gives the total amount of methylene blue delivered as a function of time, Figure 5B.This experiment visualizes the strong enhancement in drug delivery using redox flow iontophoresis compared to passive diffusion.It further shows that the concentration distribution extends significantly farther from the device when actively delivering the molecules.This increase in drug reach is crucial when targeting extended areas in the brain with a single device.Our experiment further confirms that, as expected, we observe the highest drug concentration in the immediate vicinity of the device surface and a decay of concentration with increasing distance.

In Situ CT Analysis
We further employed computed tomography (CT) analysis to investigate the internal structure of the implant during operation.To this end, we used a similar setup as before with the device being inserted into a brain phantom comprising a cuvette filled with an agarose gel.We delivered Ba 2+ cations from a BaCl 2 solution injected into the drug channel which showed excellent X-ray contrast.Figure S3 (Supporting Information) shows the 3D segmentation of sodium ferrocyanide in the redox mediator channel and Ba 2+ in the drug channel and gel matrix confirming structural integrity during operation.We were further able to image the Ba 2+ ions in the gel brain phantom over time showing iontophoretic delivery of the ion, Figure S4 (Supporting Information).

Discussion
The redox flow concept demonstrated here addresses longstanding problems in the field of iontophoresis arising from the necessity of electronic-ionic charge transfer at the electrodes.With our approach, no corrosive or gaseous side products are formed, and the working electrode material is not depleted.This allows a device based on redox flow iontophoresis to operate continuously, as long as the redox mediator solution is supplied.Moreover, as the drug solution does not come in direct contact with This work enables continuous localized drug delivery directly into the central nervous system at high delivery rates.It allows the delivery of drugs that are currently not usable due to low bloodbrain barrier permeation.[32] However, due to the limited permeability of the blood-brain barrier to doxorubicin, the drug is currently not commonly used for brain cancer therapy. [33]Other promising drug candidates are for instance cisplatin for cancer therapy, the opioid peptide DAMGO to complement opioid-based analgesia, [34] -aminobutyric acid for epileptic seizure control, [17] or macromolecules such as monoclonal antibodies.Redox flow iontophoresis is a platform technology enabling the delivery of a multitude of charged drug molecules.
The presented drug delivery device has four fluidic ports (two inlets and two outlets) and one electric connection which need to be connected to syringe drivers and a voltage source for operation.To this end, the device is compatible with commercial skin port systems allowing chronic implantation in patients.Further substantial miniaturization of the device is possible using our additive manufacturing approach.While we do not anticipate that the positioning of the counter electrode will significantly impact drug delivery under physiological conditions, we plan to explore this aspect in our future research.We will also investigate wearable pump systems and in vivo models to study the device's biocompatibility and efficacy in a chronic drug delivery setting.
In summary, in this paper we have presented redox flow iontophoresis, a novel concept using redox mediators to operate iontophoresis, overcoming the significant limitations that have previously prevented iontophoretic devices from being successfully utilized for chronic drug delivery.Based on this concept we have demonstrated a working implantable device and shown that it can achieve very high delivery rates of the cancer drug doxoru-bicin.We have analyzed the influence of operating parameters such as injected charge, drug formulation and flow rates and we have assessed the spatial drug distribution in agarose brain phantoms using optical methods and computed tomography.We believe that redox flow iontophoresis is a promising new approach to enable chronic delivery of drugs into the central nervous system circumventing the blood-brain barrier.

Experimental Section
Device Fabrication: Device polymer bodies were fabricated with an Asiga Max UV SLA 3D printing system using pro3Dure GR-10 dental grade acrylate resin.The polymer bodies were sonicated twice for 10 min in isopropanol and UV cured for 30 min at 365 nm (2.6 mW cm −2 ) in normal atmosphere.The devices were subsequently coated in a 2 μm parylene C layer with an SCS PDS 2010 lab coater deposition system using a dichloro-p-cyclophane precursor (Galentis, Italy).Neosepta AMX ion exchange membranes (Astom, Japan) and Spectra/Por3 regenerated cellulose dialysis membranes (Repligen, USA) were cut to size and attached to the coated polymer body using Loctite 4305 (Henkel, Germany) UV curing cyanoacrylate adhesive.To form the working electrode, a 0.2 mm thick graphite foil (ProGraphite, Germany) was cut to size and attached to cables using Epo-Tek H27D silver epoxy (Epoxy Technology, USA).The silver epoxy bonding site was thoroughly encapsulated in cyanoacrylate adhesive to prevent contact with any liquid.The electrodes were inserted into the polymer bodies, polyurethane tubing (Instech, USA) was attached, and the assembled devices were sealed using Born2Bond Light Lock gel UV curing cyanoacrylate adhesive (Bostik, France).All devices were thoroughly pressure tested to prevent leakage or fluid flow between the microfluidic channels.
Drug Delivery Experiments: For measuring the amount of delivered cancer drug versus injected charge, medial grade doxorubicin hydrochloride (doxorubicin) solution for infusion was purchased from Medac GmbH (Germany) containing 3.45 mm doxorubicin hydrochloride and 154 mm NaCl at pH 3 and utilized as supplied.Doxorubicin is cationic in its hydrochloride form.For measurements of the delivery efficiency versus drug NaCl concentration doxorubicin hydrochloride powder and sodium chloride were purchased from Merck (UK).Aqueous drug solutions were created containing 3.45 mm doxorubicin hydrochloride and the desired con-centration of NaCl.The solutions were adjusted to pH 4 with HCl (Fisher Scientific, UK).Fresh 200 mm sodium ferrocyanide solutions (Merck, UK) were prepared and used as redox mediator solutions.
Ag/AgCl counter electrodes were created by submerging 1 cm x 2 cm large silver foils (Merck, UK) in 1 m HCl and applying ca.1.1 V versus a Pt counter electrode for 15 min.
To conduct the drug delivery experiments the devices were submerged in a 100 mm NaCl solution with the outer membrane parallel to the Ag/AgCl counter electrode both ca. 5 mm apart from each other.The redox mediator and drug solutions were flushed through the devices at a flow rate of 400 and 17 μl min −1 , respectively.If not stated otherwise, a voltage between 1-7 V was applied between the device (anode) and counter electrode (cathode) using a Keysight U2700A source measurement unit in either potentiostatic or galvanostatic operation.For each device, an open circuit passive diffusion control was taken.The doxorubicin concentrations were quantified by fluorescence using a Tecan Spark plate reader at 480 nm excitation and 590 nm readout wavelength.The electrical currents depicted in Figures 3D, 4A,B were normalized with respect to the initial value of each data series.The delivery experiments were conducted under ambient laboratory conditions at room temperature.We calculated a maximum average drug delivery rate of (2.3 ± 0.2) nmol min −1 from the highest data point in Figure 3A which was obtained after continuously delivering doxorubicin for 15 min.The slope of the linear fit in Figure 3A is (5 ± 1.5) nmol C −1 with a coefficient of determination R 2 of 0.82.
Delivery Into Gel Brain Phantoms: Optical absorbance images were used to measure the mass and concentration maps of methylene blue delivered into 0.6% w/w agarose gels.The gels were prepared by dissolving agarose powder in a 100 mm NaCl solution at 100 °C and the solution poured the solution into a transparent cuvette (10 mm optical path length).Once set the delivery devices under investigation were inserted manually into the gel along one side of the cuvette, with the membrane facing the gel, and an Ag/AgCl electrode was inserted along the other side.In the case of active release, a fixed electric potential (5 V) was applied between to device electrode and the counter electrode.A 0.63 mm methylene blue solution in 100 mm NaCl was pumped into the drug channel at 17 μl min −1 , and a 200 mm ferrocyanide solution was pumped into the redox mediator channel at 8 μl min −1 in the case of passive and at 30 μl min −1 in the case of active delivery.The optical system, calibration and data processing workflow are described in. [29]Absorbance fields were directly computed from pixel intensity values and the mass of methylene blue in the gel was derived for each frame and, assuming axial symmetry, the concentration field was reconstructed by applying a stabilized inverse Abel transform to the absorbance field.Axial symmetry is not valid in the immediate vicinity of the planar membrane; however, this method gives a reasonable estimate of the concentration field without the need of more complex inversion algorithms.
Cyclic Voltammetry: Cyclic voltammograms of a 20 mm sodium ferrocyanide solution in 1.5 m KCl supporting electrolyte were recorded using a PalmSens 5 potentiostat system.The working electrode was a 0.7 mm 2 glassy carbon electrode (Metrohm) which was carefully polished before the measurement.A 1 cm 2 surface area Pt foil (Metrohm) was utilized as a counter electrode and an Ag/AgCl electrode (ALS Japan) as a reference electrode.
CT Imaging: In situ CT scans of the drug delivery device were taken using a Zeiss Xradia Versa 510 3D X-ray microscope with the X-ray source operated at 80 kV/6 W, and with a pixel size of 14.72 μm.The devices and Ag/AgCl counter electrodes were inserted into cuvettes filled with agarose gel phantoms as described above.A 200 mm sodium ferrocyanide solution was flushed through the redox mediator channel at a flow rate of 300 μl min −1 and 1.5 m BaCl 2 was flushed through the drug channel at 17 μl min −1 .Projections were imaged from multiple angles with an exposure time of 2 s.The data was reconstructed using proprietary Zeiss software and image visualization and threshold segmentation was performed using the medical imaging software 3DSlicer.The 2D images in Figure S4 (Supporting Information) were calibrated relatively to each other by adjusting the pixel value histograms such that the 0.1% and 70% quantiles matched.

Figure 2 .
Figure 2. Architecture of the implantable device.A) Application scenario for drug delivery into the brain.The device is implanted into the brain region of interest through a cranial window.It is then connected to two syringe drivers to continuously pump the redox mediator and drug solutions and to waste containers to collect the spent solutions.An electric current is applied between the device and a counter electrode on the skin of the patient.Created with BioRender.com.B) Photograph of the device with fluidic tubing and electric connection cable attached.Scale bar: 1 cm.C) Schematic of the device.Scale bar: 2 mm.

Figure 3 .
Figure 3. Drug delivery analysis.A) Delivered doxorubicin amount versus total injected charge Q.The colors indicate data points for different devices (N = 7).The datapoints at 0 C injected charge represent the passive diffusion control.The dashed line is a linear fit of the data (R 2 = 0.82), while the shaded area indicates the 95% confidence band.B) Corresponding drug delivery efficiency  versus the injected charge.The dashed line represents the mean value.C) Drug delivery efficiency  versus the drug solution NaCl concentration measured for N = 3 devices.The dashed line is a linear fit of the data as guide to the eye (R 2 = 0.4), while the shaded area indicates the 95% confidence band.We conducted the measurement with a custom 3.45 mm doxorubicin formulation adjusted to pH 4. D) Normalized electric current versus ferrocyanide redox mediator concentration.The dashed line is a logarithmic fit as a guide to the eye.

Figure 4 .
Figure 4. Flow rate dependence of drug delivery.A) Normalized electric current versus ferrocyanide redox mediator flow rate.B) Normalized electric current versus doxorubicin drug solution flow rate.

Figure 5 .
Figure 5. Optical monitoring of methylene blue delivery into an agarose brain phantom.A) Concentration field of delivered methylene blue at different points of time.Top: passive diffusion control, bottom: active delivery.Scale bar: 5 mm.B) Total delivered methylene blue as function of time.