Millions of individuals around the world suffer vision loss due to retinal degeneration diseases such as retinitis pigmentosa and age-related macular degeneration. These diseases affect the photoreceptor cells rendering them dysfunctional and eventually causing their demise. Retinal prostheses can be used to restore some useful vision to the affected patients improving their quality of life considerably. These prostheses are based on stimulation of secondary neurons in the retina replacing the functionality of photoreceptors. Several methodologies are applied to convert visual images into a train of pulses that are transmitted to an array of electrodes positioned on the retinal surface. These pulses are then used to stimulate the viable secondary neurons. Many groups (refer to the review (1)) worldwide are working on different devices based on the placement of the implant with respect to the retina.
Neural tissue inhomogeneity is an important parameter affecting neural stimulation (2; 3). The vertebrate retina is a dense neural tissue composed of multiple layers each characterised by different cell types and densities (4) rendering it electrically inhomogeneous. By constructing an electric model based on inhomogeneity, it is feasible to compute the electric field distribution in the retina and consequently predict parameters such as threshold and resolution of stimulation for a safe and efficient retinal prosthesis. In order to construct a realistic, passive electric model of a retina, it is necessary to measure layer resistivity locally and precisely.
The resistivity of the retinal layers has been measured for various applications until now such as local electroretinograms (5) and current source density analysis (6). Researchers mainly used the four-terminal (tetrapolar) method to measure the resistivity profiles in the depth of the retina. Double-barrelled (6) and concentric (5) glass micropipettes have been employed as the pick-up electrodes previously. Tetrapolar measurements require a complicated setup due to additional electronics (such as front-end amplifier, current injection electrodes, etc.) and retina sealing issues (in
In this study, an alternative and direct approach to measure local resistivities in the various layers of an isolated retina is presented. The approach uses a thin, flexible microfabricated probe of two electrodes to record impedance by an easy to setup bipolar impedance spectroscopy technique. Bipolar measurement is more suitable if one wants to measure a change at a specific position in an otherwise homogeneous sample (9). Considering each retinal layer to be homogeneous, the changes in resistivity occurring at various layer ayer interfaces can be well detected by bipolar measurements. Our electrodes with close spacing allow high resolution resistivity profiling in relatively thin isolated rat and chick retina samples.
Wistar (
Rectangular electrodes with rounded corners were used to reduce fringing effects. The dimensions and design of the microprobe used in this study is presented in Fig. 1.
The rectangular electrodes are spaced 10μm apart. The spacing between electrodes is based on a compromise between a localised measurement (high resolution) and maximum current penetration in the retina (sensitivity). Average retinal thickness for rat is 150μm (11) and chicken is 175μm (12. High resolution measurements are required to probe the different layers within the rat and embryonic chick retinas. An electrode spacing of 10 μm is sufficient to obtain an elaborate resistivity profile of the retina accounting for typical retinal cell sizes ranging from 10 μm. Both an analytic analytical (9) and finite element method based computation (Comsol Multiphysics 4.0a) of electric field penetration depth in saline (ρ=1.5Ω⋅m) for a 10 μm spacing between electrodes revealed an approximate depth of 8.3 μm (see supplementary information). The he large enough penetration depth ensures probing retinal cells making an electrode spacing of 10μm appropriate for the application under consideration.
The choice of measurement method depends on the degree of sample homogeneity and the measurement hardware (i.e. whether the measurement hardware is better at detecting absolute or relative changes) (9). . By virtue of the he similarity of cells within a retinal layer, we can consider each layer to be homogeneous. Owing to their high sensitivity to small changes near the electrodes, electrodes, bipolar measurements record the resistivity of the layer. Bipolar impedance measurement method is used in this study as it requires a simpler experimental setup compared to multielectrode schemes.
The polyimide-based flexible microprobe 10-12μm thick consisted of two recessed Platinum electrodes (40μm×25μm), separated by 10μm was fabricated based on an established process (13). A photograph of the complete microprobe assembled on a plastic base for easy manipulation is shown in Fig. 2.
In order to measure a resistivity profile of the retina, it is essential to extract the resistance in each layer. layer For the extraction of the tissue resistance from the measured impedance, one of the approaches is to consider the impedance of an electrode-retina configuration represented by an equivalent passive electrical circuit model as shown in Fig. 3. The model consists of contributions due to the electrodes-electrolyte interface and the complex tissue impedance in series with it. it. The constant phase element (CPE), Z ZCPE_E accounts for the non-ideal capacitive behaviour observed in solid metal electrodes (14). The complex tissue impedance is represented by a Cole model (15; 16) of a resistance (R tissue ) in parallel with a series combination of an intracellular resistance (Rintra) and a CPE (ZCPE_T). The model is suitable for AC analysis alone.
A typical impedance/phase spectrum along with its model fit at a depth in the retinal tissue is presented in Fig. 4. Tissue resistance can be extracted from experimental data involving impedance/phase spectra by using fitting algorithms applied on the equivalent circuit model. Alternatively, tissue resistance has been extracted using the peak resistance frequency (PRF) method in brain tissue impedance measurements (17).
The PRF method involves finding the frequency at which the measured impedance is least capacitive (or closest to resistive behaviour). Below this frequency, electrode CPE increases the measured impedance; above this frequency frequency, the tissue CPE and CPAR each separately or jointly decrease the measured impedance. A single choice of frequency to determine tissue resistance is usually defined by the cut-off frequency calculated from the electrode capacitance and tissue resistance itself. The PRF approach helps to define the best measurement frequency for identifying tissue resistance from a typical tissue impedance spectrum spectrum. Although Although, the tissue resistance can be extracted using fitting methods on the electrical equivalent (Fig. 3), its reliability is limited to uncertainties in the various fitting parameters (ZCPE_E, ZCPE_T, Rtissue, ZPAR).
The tissue resistance is extracted from the impedance magnitude at the PRF. At three different depths in the retina, magnitudes magnitude at PRFs (ZPRF) associated with raw experimental data compared within 10% of their corresponding fitted tissue resistance (Rtissue) values. We do not observe tissue relaxation as a result of the dominating electrode interface impedance (owing to the small electrode size). In view of possible misinterpretation (curbed tissue relaxation) and algorithmic errors in the fitting method using equivalent circuits, it was proposed to apply the PRF method to extract tissue resistance from impedance spectra recorded at different depths in the retina.
The resistivity (ρ) at any depth in the retina can be determined from the measured tissue resistance (R) using a simple direct relationship given by ρ = R/k (18), , where k is the cell constant. The cell constant of the geometry used in this study was analytically calculated as 232 232.75cm -1 according to the method described by Jacobs
composition (in mM) of Ringer's solution (Sigma Aldrich, Switzerland): Sodium Chloride (NaCl) – 115, Potassium Chloride (KCl) – 5, Calcium Chloride (CaCl2) – 2, Sodium Bicarbonate (NaHCO3) – 25, Magnesium Sulphate (MgSO4) – 2, Monopotassium phosphate ( KH2PO4) – 1 and Glucose ( C6H12O6)- 30
– In every trial trial, at least three impedance measurements at different depths (every 10μm) in the bath (Ringer’s solution) before entering the retina were performed. The first considerable change in impedance magnitude at PRF indicate indicated the entry into the retina. Visual control using a pair of binoculars confirm confirmed this first electrode-retina contact. Subsequent impedance measurements at every 10μm depth were recorded until an impedance value similar to the one observed in the bath was encountered. Each measurement was recorded with a wait time of 30 seconds for the signal to stabilise after the micromanipulator made the 10μm vertical movement into the retina. This time was determined based on measurement of time taken for the impedance value to stabilise at a random depth in the retina (see supplementary information) information). Three more recordings at 10μm intervals were made to ensure the electrodes contact with the Agar gel before terminating the experiment and retracting the electrodes to the initial position. The system was under ambient laboratory conditions of 21°C during the impedance measurements.
To o be able to compute the resistivity from the measured tissue resistance at PRF, the cell constant needs to be experimentally determined. The cell constant of the bipolar electrodes used in this study was calculated using the impedance/phase spectrum of Ringer’s solution of predetermined conductivity conductivity. The spectrum is as shown in Fig. 6. Based on an average of such measurements with different batches of electrodes, an average experimental cell constant was found to be 225cm-1 (less than 5% error). This value is within 3.5% of the theoretical and simulated values of cell constant for the electrode configuration used in this study. The close agreement of the cell constant with previously calculated values validates the proper working of the electrodes for an experiment.
Following the functional validation of electrodes, they can be employed for impedance measurements at different depths in the retina of the chosen animal model. During these experiments, a shift in the PRF was observed at each depth in the retina as present presented ed in Fig. 7. Starting from the retinal ganglion cell layer as we go deeper into retina towards the photoreceptor layer, the PRF moved to lower frequencies and the impedance magnitude rose from a low to a high value between 80-85% of retinal depth spanned during the experiment. With reference to Fig. 7, a 10% retinal depth corresponded to an approximate microprobe displacement of 14μm in the rat retina.
Impedance spectroscopic measurements at different depths of isolated retinal slices from three rat (14-16 days postnatal) and five chick (three E18 and two E12 embryos) samples were performed. Resistivity values were calculated from the extracted impedance value at the PRF using the direct relation between both for each depth in the retina sample. A resistivity depth of 100% was denoted as the last measurement in the retina before an identical value of resistivity obtained in Ringer’s solution is reached (electrodes in Agar gel). Point zero represent represented the last measurement in the Ringer’s solution before there is a significant change in the resistivity, i.e., an appreciable shift in the PRF is observed observed. Resistivity versus retinal depth profiles for both rats and embryonic chicks are presented in Fig. 8 and Fig. 10 respectively.
In both rat (Fig. 8) and embryonic chick (Fig. 10) measurements, an increasing resistivity-depth profile is observed rising gradually from the retinal ganglion cell layer towards the photoreceptor layer. At an approximate depth of 65%, the resistivity reaches a maximum value and then gradually decreases to attain a value obtained in Agar gel. There is a close interspecies resemblance in the studied resistivity profile shapes.
The maximum mean resistivity reached in rat retina samples is 4.2 ± 0.9 Ω⋅m and for E12 chick is 4.5 ± 0.2 Ω⋅m occurring between 65-70% retinal depths. On the other hand, the maximum mean resistivity in E18 chick retina samples is 7.9 ± 0.6 Ω⋅m which is approximately double the value measured in rats and E12 chicks at the same retinal depth. In rat and E18 chick resistivity profile measurements, at around 80% depth into the retina, there is a definite dip in the resistivity profile gradually decreasing into a low value similar to a measurement in Ringer’s solution.
The he standard deviation (SD) of resistivity from the mean resistivity value at each depth was examined examined. For the embryonic chick resistivity profiles profiles, it is determined that the SD is low in the Ringer’s solution and the Agar gel. gel In contrast, the rat data demonstrates large SD in the these two regions of the resistivity profile profile.
It is known that there is a PRF shift with a resistivity change in different retinal layers. A representation of the relationship between PRF and resistivity based on the experiments on rat and embryonic chick retinal slices is depicted in Fig. 9 and Fig. 11 respectively. The log resistivity is linearly dependent on the log PRF for both the species. For embryonic chicks, owing to similarity in the data across various trials, it can be observed that there is a unique PRF for each resistivity. On the contrary, the three rat trials suggest multiple PRFs for each resistivity.
To the best of our knowledge, planar, bipolar microelectrodes on a flexible substrate were used for the first time in this study to measure resistivity-depth profiles in rat and embryonic chick retinas retinas. We first demonstrated the functionality of the microfabricated device. The electrode cell constant extracted from the measured solution resistance in Ringer’s solution compares well with the value obtained by equivalent circuit fitting. The resistivity values at different depths in the retina established by the PRF method are within 10% of the fitted values. This is a confirmation of the electrode interface impedance not interfering with the measurements. There was a close agreement between the experimental and the theoretical/simulated bipolar cell constant values. The experimental value of 225 225cm -1 is low compared to the combined average of both theoretical and simulated value of 232 232.5cm-1. This low difference of 3.5% % is within the experimental variations. Hence, the rounded corners of the electrodes instead of sharp perpendicular shapes may have contributed to reduction in fringing effects of electric field originating from the electrode edges. The resistivity-depth profiles, in both rat and embryonic chick experiments, indicate the inhomogeneous nature of the retina and the trend they follow are in accordance with the results obtained for various species in previous studies (5; 6). This confirms that our method is valid for retina resistivity profiling studies.
We found the maximum local resistivity occurred in all experiments between 65-70% retinal depths. This can be explained by greater retinal resistivity in regions like the inner nuclear layer (INL) where neurons are packed more tightly than the inner plexiform layer (IPL) (7). Our observations are in close agreement with the local maxima occurring at retinal depths of ~80% in monkey (5), ~75-80% in rat (21) and ~70% in chicken (7). The shape of the resistivity profile of the chick embryo was similar to that of the chicken (7) and the rat resembled mammalian species (5) to a large extent. This was particularly true in the region between the proximal retina and down to the junction of inner and outer photoreceptor segments. We observed an appreciable dip in the resistivity values from a retinal depth of 80% onwards until electrodes come in contact with the Agar gel. This decrease in resistivity in the photoreceptor layer was also found in previous studies using isolated slice models of avian (7) and rat (21) retinas. The local decrease in resistivity might be caused by the relatively large interstitial spaces among the outer and inner receptor segments (21). Resistivity profiles are affected by the type of retinal preparation used (isolated retinal slices and eyecup preparations). For comparing profiles, the anatomical difference between an isolated slice and an eyecup preparation of retina needs to be considered owing to the absence of the retinal pigment epithelium. In an isolated slice preparation, considering that photoreceptors offer low resistance, the effective resistivity profiling is made between the inner and outer limiting membrane (22).
We observed higher resistivity values in E18 compared to E12 chicks. This can be attributed to the ongoing retinogenesis which terminates only at E18 (23). Cell differentiation between E12 and E18 is accompanied by cell polarization, laminar stratification and changes in cell numbers (23; 24) potentially accounting for resistivity changes within the retina. Further exploration of this subject can be interesting for future studies.
The absolute resistivities found in this study are lower compared to previous findings in rats (21) and chicken (7). The values may be difficult to compare with former investigations as the measurements are influenced by various factors like the interracial difference, age difference, measurement technique, electrodes used, etc. A majority of previous studies used the tetrapolar method with a constant current injection. The local damage caused by the pick-up micropipettes in the retina may cause a local current increase due to inflow of the perfusion solution. This could lead to an increased voltage drop resulting in a false increase of measured resistivity. In addition, most of the studies were conducted in the low frequency region (ranging between 1Hz and 100Hz) with very small electrodes (~2-20μm) (5; 6; 25). Regardless of the method used,
The maximum SDs from the mean resistivity at certain retinal depths in our study was found to be high. A large variability between trials was also observed in previous studies of rat and chicken (7; 21) that were used for comparison to our findings. All measurement techniques until now including ours have the following inherent drawbacks that affect the resistivity-depth profiling of the retina – (i) the movement of the electrode relative to the tissue not being accurate due to chip-tissue slippages, (ii) pressure causing damage to the tissue (iii) damage to the tissue by electrodes causing a high-current shunting between them resulting in an erroneous measurement of resistivity in the retinal layers and (iv) unpredictability of resistivity values at the retinal layer boundaries. Apart from these factors, the location on the retinal slice where the electrodes penetrate is a significant reason for the variability in resistivity measurements. A solution, even though it contributes to the experimental complexity, may be to locally stain the retina as a visual aid for electrodes insertion to produce reproducible resistivity profiles of the retina.
The small electrode spacing of the bipolar electrodes permitted high resolution measurements in rat and embryonic chick retinas. The high resolution profiling consisted of 25 depths in embryonic chicks and 15 depths in rats. Assuming a 10μm microprobe displacement into the retina, the 10μm spacing between the electrodes used in our study is more sensitive to capture the subtle changes in resistivity between the layers. Previous investigations employed larger electrode spacing of ~25μm (5) and ~12-16μm (6). Our electrodes design is an improvement in terms of measurement resolution compared to literature.
An important result of our study is that the resistivity at a certain depth within the retina is identified by a unique PRF in embryonic chick experiments. Conversely, distinct profiles (refer Fig. 9) were obtained in the PRF versus resistivity plots for rats. These profiles can be understood based on the large SDs observed in the Ringer’s solution before entering the retina and in the Agar gel (refer Fig. 8). Although the PRF is the frequency at which the measured impedance is most resistive, representing the tissue resistance, it is influenced by interface and parasitic capacitances. The large differences in resistivities observed for calibrated mediums could be attributed to changes in electrode capacitance. This may be perceived as the electrodes not being sufficiently clean before the experiment. There could be a thin layer of adsorbed proteins from the retinal tissue cells or damaged limiting membrane residues that may add to the overall measured impedance. Electrodes were cleaned with mild soap solution for rat experiments whereas with RCA-1 cleaning procedure for the embryonic chick experiments. Thus, we conclude that quality of an electrode surface is crucial for good resistivity profiling in a retina and RCA-1 cleaning process is more effective compared to soap for electrodes used in this study.
An alternative method for high resolution resistivity profiling along the depth in a retina based on bipolar impedance spectroscopy was established. We validated our device by profiling rat and embryonic chick retinas. The resistivity at each retinal depth was calculated based on tissue resistance extracted by peak resistance frequency methodology. Qualitatively, we found the resistivity-depth profiles to be in accordance with earlier studies and that resistivity at any arbitrary retinal depth is characterised by a unique peak resistance frequency. We have shown the potential of planar bipolar microelectrodes as a new technique to probe absolute local resistivity within a retina and multi-layered tissues, in general.
Determining absolute values of resistivities in retina contributes to improved understanding of retinal stimulation by means of modelling studies. The generated resistivity profiles can form the basis for construction of a realistic electric model of a retina. Finite element modelling may be used for estimating and optimising critical parameters such as stimulation thresholds, heat dissipation, resolution, etc. for a given electrode geometry, that are instrumental for the safety and efficacy of a retinal prosthesis. A future improvement of our two-electrode system would be a linear array of electrodes on a single strip. An array of electrodes is capable of probing different layer resistivities with a single insertion into the retina which is expected to cause less damage and provide more reliable measurements.