Retinal Prosthesis Safety: Alterations in Microglia Morphology due to Thermal Damage and Retinal Implant Contact

Abstract

Purpose.: In order to develop retinal implants with a large number of electrodes, it is necessary to ensure that they do not cause damage to the neural tissue by the heat that the electrical circuits generate. Knowledge about the threshold of the amount of power that induces damage will assist in developing power budgets for retinal implants.

Methods.: Heat-induced retinal damage was evaluated by measuring changes in the morphology of the resident immune cells, the microglia, which are the first cells to respond to retinal trauma. Microglial soma and arbor areas were assessed in rat retinal tissues in vitro to determine the effects of increasing temperatures, implant contact, and heating and implant contact combined.

Results.: In response to increasing incubation temperatures (no implant), microglial somas enlarged and arbor areas retracted, indicative of retinal stress. Thermal damage thresholds, defined as a significant change in microglial morphology from that observed at the upper limit of normal body temperature, occurred at a temperature of 38.7°C. Implant contact, induced when a passive implant was placed on the retina, also caused similar morphological alterations in microglia, indicating retinal damage. Heated-implant contact exacerbated the effects of temperature alone but still resulted in a thermal damage threshold of 38.7°C, the same as with heating alone.

Conclusions.: Our conservative recommendations are that implanted retinal electronics keep power dissipations to less than 19 mW/mm² to stay below the microglial thermal damage threshold (2.1°C) and to comply with international standards for implantable devices (2°C).

Introduction

For over 27 million people worldwide suffering from visual loss and impairment caused by retinitis pigmentosa and age-related macular degeneration, ¹ a retinal implant with the ability to restore their sight at even a rudimentary level is desirable. Current clinical trials, performed with up to 1500 electrodes, have had some success in eliciting visual perceptions through electrical stimulation of the remaining retinal neurons. ^2–6 These studies, as well as simulations performed using normal sighted subjects, suggest that in excess of 1000 electrodes may be required to provide mobility, facial recognition, and the ability to read. ^7,8 As the desire for enhanced visual acuity and resolution increases, so too may the power requirements of the implanted circuitry. The amount of thermal energy released from the implant into the delicate retinal tissue becomes a serious safety concern. Experimental investigation of the detrimental effects of thermal heating on retinal tissue is important to ensure that energy dissipated by implanted visual prostheses does not cause or exacerbate any cellular damage or neuronal death. Determination of temperatures that cause thermal damage can be used to assist in the design of implant circuitry and stimulation protocols to maximize the performance of visual prostheses while ensuring that patient safety is not compromised. While power dissipations above 50 mW have been reported to cause immediate trauma, inducing retinal whitening, edema, and other lesions, ⁹ the thermal increases that can be tolerated by retinal tissues remain largely unknown. 

Damage to the central nervous system (CNS) caused through the implantation of an electrode or foreign body induces a series of events including astrocyte gliosis, activation of microglia, and neuronal cell death. ¹⁰ These responses occur immediately following, and also in response to, the sustained presence of a device. Astrogliosis, perhaps the most well studied change in the CNS in response to an implanted device, involves hypertrophy of glial cells, changes in function and proliferation, and, ultimately, the formation of a glial sheath that interrupts the device tissue interface. ^11,12 However, changes such as neuronal cell death and macrogliosis are slow to occur and require chronic implantation of the device in vivo. 

Microglia, the resident immune cells of the CNS, are highly sensitive to environmental changes and rapidly alter their phenotype with even minor disturbances of CNS and retinal homeostasis. ^13–15 Notably, they are one of the first cell types to respond when the CNS is injured, before the onset of neuronal cell death and before the development of astrocyte gliosis. ^16,17 Indeed, retinal microglia are thought to respond to signals released by stressed neurons in as little as 82 seconds ^16,17 and are thus an excellent way to determine the tissue response to disease or injury. Microglia are uniformly distributed within the CNS and retina and have a small cell soma, little perinuclear cytoplasm, and a number of fine, branched processes that are covered in numerous protrusions. When the neural tissue is stressed, microglial cells become activated, altering their cellular morphology, tissue distribution, and migratory characteristics, changing their process structure and expression of cell surface antigens. ^14,17–19 Morphological changes to normal, ramified microglial cells include the retraction of and decrease in number of primary and terminal processes and increase in soma volume. ¹⁴ This has been observed to occur within the eye in response to a range of induced stressors including cell axotomy, ²⁰ ischemia, ^21–23 photoreceptor degeneration, ²⁴ and laser-induced retinal damage. ^16,17 Similarly, microglia become activated in response to implantation of a foreign body ²⁵ or hyperthermia, ^26,27 releasing proinflammatory cytokines. ²⁷ Our hypothesis was that changes in microglial morphology in response to foreign body contact and heat could be used as an indicator of stress in the retina, and that quantification of this response would allow determination of power thresholds for safe levels of heat produced by retinal prostheses. 

Microglial morphology was assessed in the rat retina in response to temperature increases, implant contact, and a combination of implant contact and localized heating. While investigation of gliosis and cell death can be performed in vivo, assessment of thermal trauma is limited. By utilizing the fast morphological response of retinal microglia, in vitro experimentation allows retinal changes caused by excessive thermal trauma to be observed and isolated from damage induced by foreign body contact. Experiments were designed to investigate the threshold temperatures that induced retinal microglial responses, comparing the induced temperature elevations at the surface of the tissue with power dissipated by the implant to provide power budgets to retinal prosthesis architects. 

Methods

All experiments involving animals were conducted in accordance with the University of Melbourne animal experimentation ethics committee recommendations (U of M Ethics ID No. 0911158) and adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. All efforts were made to minimize animal suffering and the number of animals used. 

Preparation of Rat Retina

Twelve-week-old Sprague-Dawley rats were deeply anesthetized with an intramuscular injection of ketamine (5 mg/kg) and xylazine (60 mg/kg) prior to sacrifice with a lethal overdose of sodium pentobarbital (Lethobarb, 120 mg/kg; Merial Australia, North Ryde, NSW, Australia). Both eyes were enucleated with the aid of a surgical microscope and the anterior segments carefully dissected. The cornea, lens, and vitreous were removed, with care taken not to mechanically damage the retinal tissue within the posterior eye cup. The posterior eye cups, containing retinae, were then used in one of four different experiments as outlined below. 

Experimental Paradigms

Baseline Experiments to Assess Microglial Morphology.

Baseline microglial morphology was assessed in retinae (n = 6) that were immediately fixed following death of the rat (baseline). After dissection, retinae were immediately fixed for 40 minutes in 4% (w/v) paraformaldehyde prepared in 0.1 M phosphate buffer (PB) and fluorescently immunolabeled for microglia, as described below. 

Control Experiments to Assess Microglial Morphology.

The control for all subsequent experiments was to assess microglial characteristics following 1 hour of incubation in a physiological saline solution, Ames' medium, at body-simulating temperatures of 36.7 ± 0.8°C ²⁸ (incubation control). Posterior eye cups (n = 22) were immersed in Ames' medium (A1420-1L; Sigma-Aldrich, Castle Hill, NSW, Australia) and bubbled with carbogen (5% CO₂, 95% O₂; BOC Medical, North Ryde, NSW, Australia) for 1 hour prior to fixation. 

Effect of Temperature on Microglial Morphology.

In order to test the effect of temperature alone on microglial morphology, posterior eye cups (n = 46) were immersed in Ames' medium and bubbled with carbogen at temperatures ranging from 33°C to 45°C for 1 hour (thermal response). Exact temperatures were maintained by placing the eye cups in a vessel in the center of a custom-made heat bath. This heat bath could be used to heat retinae at single constant temperatures between 30°C and 46°C, deviating by less than ±0.1°C over 1 hour and less than ±0.3°C over 3 hours. ²⁹ Tissues were incubated for 1 hour prior to fixation and immunocytochemical labeling for microglia. 

Effect of Implant Contact and Heated-Implant Contact on Microglial Morphology.

Foreign body response, induced by placement of an implant in contact with the retinal surface (ganglion cell layer) while bath temperature was maintained at 36.7°C, was assessed (implant contact response). Heated-implant contact, induced by heating the implant while bath temperature was maintained at 36.7°C, was also assessed (heated-implant contact response). For these experiments, posterior eye cups (implant contact response, n = 17; heated-implant contact response, n = 19) were placed in a specialty holder in a Petri dish containing Ames' medium bubbled with carbogen on a heating plate. An X,Y,Z micromanipulator was used to centrally position implants in contact with the retinal surface. Uniform contact was ensured by lowering the implant to a set distance throughout each experiment, with gentle retinal tissue contact verified through initial pilot trials. For heated-implant contact responses, retinae were allowed 10 minutes to equilibrate prior to implant-driven thermal heating for 40 minutes. For a schematic of the experimental apparatus, see Figure 1. 

Figure 1. 

 

Schematic of experimental apparatus for implant contact eyes. Dissected eye cups were placed in a Petri dish containing Ames' medium bubbled with carbogen on a heated plate such that bath temperature was maintained at 36.7°C. Examples of an incubation control eye and implant contact response eye are presented. The implant contact eye was placed in a specialty holder, and an X,Y,Z micromanipulator was used to centrally position the implant in contact with the ganglion cell side of the retinal surface. The implant was constructed by attaching a 20 Ω surface-mounted resistor to a 1000 Ω platinum resistance temperature detector (PRTD), with the two 1.6 × 0.85 mm components overlaid to form a cross (heating element surface area, 1.36 mm²). Temperature at the surface of the retinal tissue was detected with the PRTD sensor in the implant, while bath temperature was detected by a sensor in the bath. The Petri dish, specialty holder, eye cup, and implant are all to scale (scale bar, 10 mm).

Implant Design

The implants placed on the retinae used in these tests were constructed by attaching a 0603-size, 20 Ω surface-mounted resistor to 1000 Ω platinum resistance temperature detectors (PRTDs; Heraeus Sensor Technologies, Inc., Kleinostheim, Germany). Short circuiting was avoided by a thin layer of conformal coating, with the two 1.6 × 0.85 mm components overlaid to form a cross (heating element surface area, 1.36 mm²). To protect the device from tissues and fluids while minimizing the effect on thermal measurements, a 50 μm coating of silicone sealant (type 65AR flowable; Permatex, Ballarat, VIC, Australia) was applied over the whole device. ²⁹  

Temperature Calibration and Detection

The temperatures of the heat bath and tissue were measured using 1000 Ω PRTDs connected to a custom-made high-resolution temperature measurement system (HRTMS; Gavin Pearce, National Information and Communication Technology Australia Ltd. [NICTA]). The HRTMS used alternating currents to remove thermocouple effects between the dissimilar metals of the PRTDs and the copper connection wires, four-wire Kelvin connections to minimize the effect of lead wire resistances, and a precision current pump to enable control of self-heating effects. ³⁰ Voltage measurements from the HRTMS were acquired via computer through a data acquisition card (NI-USB 6215; National Instruments, Sydney, Australia), allowing multiple temperatures to be simultaneously recorded and displayed using LabView Signal Express 3.5.0 (National Instruments). All measurements were recorded at 1 Hz and converted to temperatures in accordance with IEC Specification DIN EN 60751-1998. ³¹ Calibration of the HRTMS over at least seven independent saline bath trials indicated a highly linear voltage–temperature relationship (correlation coefficient of r ² = 0.9999), 

 

where V is the recorded voltage for temperatures (T) between 30°C and 50°C (±0.002°C). 

Sensors connected to the HRTMS were used to monitor and record temperatures within the baths and implants. For eye cups subjected to heated-implant contact, currents, measured as the voltage drop across a 0.1%, 1.0 Ω resistor, of up to 93.2 mA (173.6 mW, 127.6 mW/mm²) were passed through the heating element coils. 

Immunocytochemical Labeling of Microglia in Retinal Flat Mounts

Posterior eye cups were fixed for 40 minutes in 4% (w/v) paraformaldehyde in 0.1 M phosphate buffer (PB). Retinae were then dissected out of the eye cups, cryoprotected in graded sucroses (10% for 1 hour, 20% for 1 hour, 30% overnight), and frozen and thawed three times to enhance antibody penetration. Microglial cells were detected using immunocytochemical labeling, carried out on retinal whole mounts using the indirect fluorescence method. ^32,33 Retinal sections were blocked for 1 hour in a solution containing 10% normal goat serum (NGS), 1% bovine serum albumin (BSA), and 0.05% Triton X-100 in PB. The primary antibody, rabbit polyclonal anti-ionized calcium binding adaptor molecule 1 (IbA1) (Cat. No. 019-19741; Wako Chemicals, Richmond, VA), generated to synthetic peptide corresponding to the C-terminus of IbA1 (sequence PTGPPAKKAISELP), was used to label microglia. Rabbit anti-IbA1 was diluted (1:1500) in 3% NGS, 1% BSA, 0.05% Triton X-100 in PB and applied for 4 days at room temperature. After washing in PB, goat anti-rabbit IgG conjugated to Alexa TM 594, secondary antibody (Cat. No. A-11037; Invitrogen, Life Technologies, Mulgrave, VIC, Australia; diluted 1:800), was applied overnight. After labeling, retinae were flattened, mounted on a glass slide ganglion cell side up, coated in a Mowiol/glycerol-based mounting medium (Mowiol4-88; Sigma-Aldrich), and coverslipped. These fluorescent immunocytochemical labeling procedures were performed on all tissues from all test protocols. 

Microglial Morphological Analysis

Analysis of microglial morphology was performed to investigate the effect of temperature, implant contact, and the combined effect of implant contact and localized temperature increases. Confocal images were taken in three locations around the optic nerve head (central location) and also in the peripheral retina. This enabled morphological changes from microglia in direct contact with the implant (central location) to be compared with those found distant to the implant. Microglial images were taken on a Carl Zeiss Laser Scanning System LSM 5 PASCAL (Carl Zeiss MicroImaging GmbH, Jena, Germany). A Zeiss EC Plan-Neofluar 20×/0.5 M27 air objective was used to observe the fluorescent microglia and photograph the images. Six sequential Z stacks were imaged (three central and three peripheral), consisting of up to fifteen 2.6 μm slices at each location. Of the 15 Z-stack slices collected in a given location, microglia could be separated into two plexuses within the retina, specifically into microglia at the surface of the retina, the ganglion cell layer (GCL), and microglia deeper in the retina in the outer plexiform layer (OPL). Stacks of images of microglia in the GCL were separated from stacks of images of microglia in the OPL. These Z stacks were separately flattened to produce two single images, one image of microglia in the GCL and one image of microglia within the OPL. 

Quantification of the morphological changes was performed via measurement of soma area and the arbor area, the area covered by the microglial processes. ²² At the time of analyses, the person conducting the analysis was blinded to trial type (incubation or implantation) and temperature. The soma and arbor areas were calculated using Zeiss LSM Image Browser 4.2.0.121 (Carl Zeiss MicroImaging GmbH). Comparison between microglia was made in four regions: (1) central ganglion cell layer (CGCL), the region in direct contact with the implant; (2) central outer plexiform layer (COPL), close to the implant but located in a deeper layer of the retina; (3) peripheral ganglion cell layer (PGCL), a region away from any direct contact with the implant; and (4) peripheral outer plexiform layer (POPL), a deeper layer of the retina well away from the implant. As the implant was located centrally, it was expected that microglia in that region, specifically microglia located in the immediately adjacent ganglion cell layer (CGCL), would be more affected by implant contact and heated-implant contact. 

To determine the effect of temperature on microglial morphology, temperature–response curves were generated by plotting microglial soma and arbor area against temperature, and fitted using GraphPad Prism 5.0.4 (GraphPad Software, San Diego, CA) with the sigmoidal equation 

 

where S (μm²) is the soma (or arbor) area as a function of the baseline response in the absence of additional heat, B (μm²); the upper response at which additional heating would not induce a greater change, U (μm²); the temperature of the tissue, T (°C); the slope of the temperature response curve, H (μm²/°C); and the temperature at which the microglial response was halfway between the basal and upper limit, T₅₀ (°C). Thermal damage threshold was determined as the lowest temperature at which there was a statistically significant change in microglial morphology from that observed at less than and up to the upper limit of normal body temperature, 37.5°C. ²⁸ This threshold was applied to provide a conservative approach to ensuring implant safety; that is, a temperature increase that induces a significant change in microglial morphology from that observed at normal body temperatures might result in neurotoxic effects on the retina. 

All data are plotted as the mean ± SEM. Differences in soma and arbor area were determined by one-way analysis of variance (ANOVA) and Tukey's multiple comparison test for multiple regions under baseline conditions. When multiple regions and experimental paradigms were considered, a two-way ANOVA and a Bonferonni multiple comparison post hoc test were applied as appropriate using GraphPad Prism. Differences were considered statistically significant if P < 0.05. 

Results

Incubation Control Microglia Soma and Arbor Areas Were No Different from Baseline

In order to assess whether incubation in physiological Ames' medium had any effect on microglial morphology, retinae were incubated for 1 hour in Ames' medium at body temperature (incubation control) and compared to retinae from tissue that was immediately fixed after enucleation (baseline). Micrographs of microglia from incubation control retinae are presented in Figure 2 (A, low power; B, close-up of microglia from A) and were found to be qualitatively similar to baseline microglia (not shown). Soma area and arbor area of microglia from incubated-control and baseline retinae were quantified and average responses compared in the different retinal regions (Figs. 3A–D). Average soma areas for microglia from baseline retinae from each of the four regions (central ganglion cell layer, CGCL; peripheral ganglion cell layer, PGCL; central outer plexiform layer, COPL; peripheral outer plexiform layer, POPL) were 21.27 ± 0.80 μm², 21.52 ± 0.56 μm², 21.52 ± 1.11 μm², and 21.90 ± 0.93 μm², respectively. The average dendritic areas were 2041 ± 174 μm², 2053 ± 178 μm², 2073 ± 198 μm², and 2111 ± 184 μm² in the CGCL, PGCL, COPL, and POPL regions, respectively. There were no statistically significant differences between soma areas (one-way ANOVA, across retinal region, P > 0.05) or arbor areas (one-way ANOVA, across retinal region, P > 0.05) in the ganglion cell layer and outer plexiform layer, or between centrally and peripherally located microglia in baseline retinae. In addition, in all cases, these measurements from microglia in baseline retinae were statistically indistinguishable from those quantified from incubation control retinae (two-way ANOVA, retinal region, P > 0.05, and paradigm, baseline versus incubation, P > 0.05, for both soma and arbor area). Thus, incubating microglia in Ames' medium for 1 hour at body temperature did not alter microglial morphology compared to baseline. 

Figure 2. 

 

Microglial morphology in incubation control and implant contact rat retinae. Microglia were fluorescently labeled with an antibody against IbA1 and imaged using confocal microscopy. (A) An example incubation control retina, showing a low-power image of immunofluorescent microglia labeled within the ganglion cell layer of the central rat retina. (B) Magnification of an individual microglia from (A). (C) An example of an implant contact retina, showing microglia labeled within the ganglion cell layer of the central retina. (D) Magnification of an individual microglia from (C). Microglia from implant contact retinae had larger somas and smaller arbors, indicative of an activated phenotype and mechanical trauma. Scale bars are 50 μm.

Figure 3. 

 

Quantitative analysis of changes in microglial morphology in baseline, incubation control, and implant contact retinae. Retinae were either fixed immediately (baseline, n = 6), incubated for an hour at body temperature (incubation control, n = 22), or placed in contact with an implant (implant contact, n = 17). Microglia were fluorescently immunolabeled using IbA1, and soma areas were quantitatively assessed in the (A) central and peripheral ganglion cell layer and (B) central and peripheral outer plexiform layer. Microglial arbor areas in the (C) central and peripheral ganglion cell layer and (D) central and peripheral outer plexiform layer were also assessed. There was no difference in microglial morphology between retinae that were immediately fixed and those incubated for 1 hour at body temperature. In contrast, retinae that were placed in contact with an implant had microglia with significantly larger soma area and smaller arbor area than baseline or control retina in all retinal regions. Data are presented as mean ± SEM; asterisk indicates significant difference in soma or arbor area, P < 0.05.

Implant Contact Induced Morphological Alterations in Retinal Microglia

In order to assess the effect of implant contact on microglial response, retinae were immersed in Ames' medium at body temperature, and a passive element was placed on the surface. Micrographs of microglia from implant contact retinae are presented in Figure 2 (Fig. 2C, low power; Fig. 2D, close-up of microglia from Fig. 2C). Microglia from implant contact retinae were found to have larger soma and smaller arbor areas than microglia from baseline and incubation control retinae (Figs. 3A–D). In the CGCL, a region that was in direct contact with the implant, microglial somas were significantly larger than observed in incubation controls (∼9.3 μm² larger, two-way ANOVA, for paradigm of incubation control versus implant contact, P < 0.05; Fig. 3A). Similarly, microglia in the peripheral regions were significantly affected by the presence of an implant, with somal areas increased by ∼5.6 μm² in the GCL and ∼6.1 μm² in the OPL (two-way ANOVA for paradigm of incubation control versus implant contact, P < 0.05 each area; Figs. 3A, 3B, respectively). Microglial arbor areas, regardless of region, were also significantly reduced by the presence of an implant (two-way ANOVA, for paradigm of incubation control versus implant contact, P < 0.05 each area; Figs. 3C, 3D). These changes in microglial morphology are consistent with activation of microglia from a resting state in response to implant contact and indicate that microglia are sensitive to mechanical trauma even in regions distant from the area touched by the implant. 

Temperature Increases Induced Morphological Alterations in Retinal Microglia

Microglial morphology was assessed in response to incubation at varying temperatures up to 44°C (no implant). Representative images of microglia from incubation control retinae (Figs. 4A, 4B), 39°C heated retinae (Figs. 4C, 4D), and 43°C heated retinae (Figs. 4E, 4F) are presented. Qualitatively, high temperatures induced increases in the soma area, and this was found to correlate with increasing temperature. Similarly, arbor area was observed to decrease as the incubation temperature was increased. These changes in microglial morphology are consistent with activation of microglia from a resting state and are likely occurring in response to thermal damage. 

Figure 4. 

 

Representative images of microglia from retinae incubated at physiological and high temperatures. Retinae were incubated for 1 hour at body temperature (incubation control) and at temperatures up to 45°C (no implant). Whole-mount retinae were fluorescently immunolabeled for microglia using an IbA1 antibody, and images from the central ganglion cell region are presented. (A) Example of microglia from a retina incubated for an hour at control temperatures, 36°C, and (B) magnification of an individual microglia from (A). (C) Example of microglia from a retina incubated for an hour at moderately high temperatures, 39°C, and (D) magnification of an individual microglia from (C). (E) Example of microglia from a retina incubated for an hour at high temperatures, 43°C, and (F) magnification of an individual microglia from (E). Microglia incubated at high temperatures had larger somas and smaller arbors, indicative of an activated phenotype and thermal trauma. Scale bars are 50 μm in each image.

Changes in microglial morphology in response to increasing temperatures were quantified. Increasing temperature induced an increase in soma area from a minimum of ∼21 μm² at body temperature to a maximum of ∼45 μm² at 43°C. Soma area was plotted against temperature; and a sigmoidal temperature–response curve was fitted for each of the four retinal regions, which were found to be almost identical irrespective of region, with strong correlation coefficients between r ² = 0.8988 and r ² = 0.9563. An example of the temperature–response curve for microglial soma in the central retinal ganglion cell region is presented in Figure 5A (r ² = 0.9563). The thermal damage threshold, defined as the lowest temperature at which there was a statistically significant change in soma from that observed at the upper range of normal body temperature, occurred at 38.6°C, represented by the vertical dotted line on Figure 5A. 

Figure 5. 

 

Temperature–response curves for microglial soma and arbor area. Retinae were incubated for 1 hour at temperatures up to 45°C (no implant). Whole-mount retinae were fluorescently immunolabeled for microglia using an IbA1 antibody, and soma and arbor areas of microglia from the central ganglion cell region were assessed. (A) Soma area was plotted as a function of temperature (gray data points) and found to increase with increasing temperature. A sigmoidal, temperature–response curve could be fit to the data (solid gray line; correlation coefficient R = 0.9563), and from this curve the temperature at which soma area significantly deviated from body temperature was assessed as the thermal damage threshold (broken black line), 38.6°C. (B) Arbor area was plotted as a function of temperature (gray data points) and found to decrease with increasing temperature. A sigmoidal, temperature–response curve could be fit to the data (solid gray line; correlation coefficient R = 0.6951), and from this curve the thermal damage threshold was assessed (broken black line) as 38.7°C.

The change in microglial arbor area in response to altered incubation temperature was also assessed in each of the four retinal regions. Figure 5B shows the temperature–response curve for the CGCL microglial arbor area as a function of temperature. The microglial arbor area in the CGCL was approximately 2030 μm² at body temperature, and reduced to an area approximately 1539 μm² at temperatures upward of 40°C. Microglial processes are known to be in constant motion, extending and retracting their processes to survey the synaptic regions of the retina. ^13,15,17 The variation in arbor area observed across the various temperatures and the resulting poor fit of the data (correlation coefficient r ² = 0.6951) are likely to be due to this phenomenon. Indeed, temperature–response curves could not be fitted to microglial arbor area data collected from cells in both the central and POPL (data not shown). Despite the variation, the thermal damage threshold determined from microglial arbor area in the CGCL was similar to that observed from measuring soma area, and occurred at a temperature of 38.7°C (Fig. 5B; vertical dotted line). 

These data suggest that evaluating changes in microglial morphology is a good indicator of alterations in retinal homeostasis, with rapid changes in soma and arbor area observed in response to temperature increases. While both decreases in arbor area and increases in soma area could be used to evaluate potential trauma, a stronger correlation was observed between soma area and temperature. All further correlations were carried out through examination of the soma area of microglia in the CGCL as a function of temperature. 

Heated-Implant Contact Exacerbated Microglial Morphological Changes

To examine the effect that retinal contact with an active, power-dissipating heating element had on microglial morphology, experiments were performed with localized implant-induced temperatures of up to 44°C. As was seen for temperature alone, localized heat induced temperature-dependent changes in microglia, including increases in soma size and decreases in arbor spread. For microglia directly under the implant in the CGCL, a temperature–response curve was fit to soma area as a function of temperature (correlation coefficient r ² = 0.7958), and this is shown in Figure 6A. Microglia had a minimum soma area of ∼30 μm² at body temperature, which increased to a maximum soma area of ∼50 μm² at temperatures above 40°C. From this curve, the thermal damage threshold was determined to be 38.7°C, represented by the vertical dotted line on Figure 6A. 

Figure 6. 

 

Temperature–response curve for microglial soma area from heated-implant contact retinae. Retinae were incubated for 1 hour at body temperature in contact with an implant and heated to temperatures of up to 45°C by the implanted element. Whole-mount retinae were fluorescently immunolabeled for microglia using an IbA1 antibody, and soma areas of microglia from the central ganglion cell region were assessed. (A) Soma area was plotted as a function of temperature (black data points) and found to increase with increasing temperature. A sigmoidal, temperature–response curve could be fit to the data (black line; correlation coefficient R = 0.7958), and from this curve the thermal damage threshold was assessed (broken black line), 38.7°C. (B) The same data from (A) (black line) were plotted against the soma areas obtained from heated retinae (no implant, gray line). The temperature–response curve from heated-implant contact retinae was shifted upward from that obtained in response to heating alone, indicating that mechanical trauma exacerbates the effects of thermal damage. However, a similar thermal damage threshold of ∼38.7°C was determined for each (broken black line).

To compare the combined effect that temperature and implant contact had on microglial morphology, the temperature-response curves for the soma area in the CGCL from retinae with thermal effects were plotted on the same graph as those measured with heat-implant contact (Fig. 6B). Microglia from tissues heated with the implant had soma areas between 5 and 10 μm² larger than heated incubation controls (no implant) at all except the highest temperatures. This indicates that implant contact exacerbates the effects of temperature increases alone. However, despite this upward shift of the heated-implant contact temperature–response curve, the thermal damage thresholds were similar for heated-implant contact effects and thermal effects alone, at approximately 38.7°C in both cases (Fig. 6B). 

Relationship between Power Dissipated and Tissue Surface Temperature Increase

A linear relationship between the power dissipated by the implant, measured as the voltage drop across a 0.1%, 1 Ω resistor, and the temperature increase induced and measured at the surface of the tissue was observed to obey 

 

where the surface tissue temperature, T_O (°C), is a function of the overall power dissipated, P_O (mW), for powers up to 173.6 mW (correlation coefficient r ² = 0.8125). 

We then normalized this to the surface area of the implant to obtain retinal surface temperature as a function of power dissipated per area of implant contacting the tissue (mm²): 

 

where the retinal surface temperature, T_A (°C), is expressed as a function of the power dissipated per surface area of implant, P_A (mW/mm²). In this case, P_A = P_O /1.36 mm². 

To enable retinal prosthesis architects to construct devices that do not cause deleterious thermal effects on neighboring tissue, implanted electronics must be carefully designed. These data indicate that an implant power dissipation of 19.2 mW/mm² will induce a 2.1°C increase in retinal surface temperature in vitro. Based on our thermal damage thresholds of 38.7°C, to reduce potentially deleterious effects on retinal integrity, our conservative recommendations are that implants restrict power to below 19.18 mW/mm² and temperature increases to less than 2.1°C. 

Discussion

The main findings of this study were that microglia morphology was altered in response to increasing temperature and implant contact following as little as 1 hour. Moreover, the type of morphological change, reflected by a reduction in area covered by processes and an increase in soma size, suggests microglial activation indicative of retinal trauma. ¹⁴ Neither the location within the tissue (centrally or peripherally) nor the tissue layer (ganglion cell layer or outer plexiform layer) was observed to alter microglial thermal tolerance, with all regions affected equally. The data presented here suggest that to avoid a temperature rise greater than the observed thermal damage thresholds assessed (2.1°C) and to stay in line with international recommendations (international standard, ISO 14708-1:2000 for implantable devices <2°C ³⁴ ), implanted electronics should keep power dissipations to less than 19 mW/mm². 

It is well known that implants placed into the CNS induce a variety of cellular changes when implanted chronically, depending on the device material, size of electrodes, and stability of the device in situ. ^{10–12,35,36} Macroglia are known to change over weeks, displaying a hypertrophic effect that can lead to scar formation and loss of implant functionality. ^11,12 Also, neuronal effects can occur either as a consequence of bystander glial responses or because of direct effects of the implant itself. However, few studies have looked at the short-term consequences of device implantation and temperature increases. 

Our results show that microglia change morphology in response to implant contact or temperature increases, and that this is a sensitive measure with which to assess retinal response to a device. Microglia are known to respond to early neuronal stress by changing morphology and releasing proinflammatory mediators, rapidly migrating to areas of damage to phagocytose and remove debris. ¹⁴ Microglia have previously been shown to become activated in response to implantation of a foreign body ²⁵ or hyperthermia, ^26,27 reducing their dendritic spread and increasing their soma size. Furthermore, with time, microglia have been shown to migrate to the region of an implant, surrounding the device. ¹⁰ In line with this, our results indicate that microglia are quick to respond to implant contact and temperature with significant changes in morphology consistent with activation observed in less than 1 hour. However, this time frame was too short to induce significant microglial proliferation/recruitment or migration toward the implant. Furthermore, our data indicate that microglial morphologies change in a sigmoidal dose–response like fashion to temperature increases such that they could be used to quantitatively assess tissue response to trauma in vitro. By using this test system, we were able to assess the effects of mechanical and thermal damage on microglial morphology independently. Specifically, placing an implant on the inner surface of the retina induced a microglial activation response, in line with the finding that microglia are responsive to foreign bodies. ^10,25 Moreover, temperature increases exacerbated any mechanical effects, leading to even greater changes in microglia in a dose–response like fashion, in line with the finding that microglia respond to neuronal trauma induced by hyperthermia. ^26,27  

In addition, microglia were uniformly affected across the retina, irrespective of the placement of the device. Although the current work was based on an epiretinal device placement and stimulation strategy, the present findings have implications for all retinal prostheses under development that are designed to have direct contact with the retina, both epiretinal and subretinal devices. ^6,37 Specifically, implants in contact with the retina are likely to have widespread effects on retinal microglia, and this may have implications for ongoing neural damage outside the region in direct contact with the device. Uniform changes to microglia across the retina have been reported in retinal diseases that are known to have a localized pathology. For example, the retinal vascular disease retinopathy of prematurity is known to cause aberrant blood vessel growth in the midperipheral retina in the mouse. ²² Despite this localized vascular pathology, microglia were noted to be activated in this disease across the entire retina. In the current study, implant contact and heated-implant contact induced damage and microglial activation in regions outside of those in direct contact with the implant, in this case likely indicating significant and widespread retinal trauma. These morphological changes are consistent with a neurotoxic microglial response. Indeed, microglia have previously been shown to release proinflammatory cytokines in response to hyperthermia (39°C). ²⁷ However, more research is required to confirm that the microglial response measured at threshold (38.7°C) in the present study is indeed neurotoxic. 

While this work has shown alterations in microglial morphology due to thermal trauma, implant contact, and heated-implant contact, we could not assess if morphological alterations were reversible or if they permanently affect the neurons, due to the short-term nature of the in vitro assay. Thus, it is possible that the mechanical trauma associated with implantation of a device will be reduced over a longer time course. In addition, the effects of temperature increases generated by an implant in vivo may also be reduced over time, modulated by changes in blood flow in the choroid, which plays an important role in regulating retinal temperature. However, this study investigated the effects of heating over only a 1-hour period, and it is possible that chronic thermal increases may instead increase the severity of morphological change and may cause additional damage to retinal tissue. We recommend that future studies be conducted to investigate the effects of prolonged, subthreshold, implant-induced and localized heating to ensure that permanently inserted devices are safe for the lifetime of the patient. 

Previous studies have attempted to assess the effects of retinal temperature increases on ocular viability. One study in rabbits found that a plaque applied to the sclera and heated to 45°C for 45 minutes was the maximum temperature that could be applied to the eye before diffuse neuronal damage was observed, while 47°C induced irreversible damage to retinal function. ³⁸ This heat was applied through the sclera, and it would be expected that the temperature in the neural retina would be less than that applied to the sclera. However, direct temperature in the neural retina was not assessed. More recent studies have assessed the effects of applying heat to the retina directly, finding that 50 mW induced retinal whitening, ⁹ while 40 mW induced a 4.5°C increase in retinal temperature. ³⁹ This second study, by Sailer et al., suggested that a 3.2°C increase induced by 15 mW (4.8 mW/mm²) of heat applied by infrared radiation was a tolerable increase for ocular tissues, based on intact gross morphology of the retina. ³⁹ However, based on our more specific morphological assessment of retinal microglia, a 3.2°C increase in temperature would induce significant microglial activation and likely cause release of proinflammatory cytokines. ^26,27 Furthermore, the heat induced by infrared laser spot of 4.8 mW/mm² may be different from that induced by an electrical device in contact with the retina of 19 mW/mm², such as that used in the present study. 

Our data indicate that if the surface temperature of an implanted device remains below 38.7°C, the retinal effects will be negligible. One caveat is that this estimate is constrained by the number of samples tested in the present study. Indeed, it is possible that lesser temperatures might induce significant changes in microglial morphology if many more experiments were completed. However, at temperature increases greater than 2.1°C above the upper range of body temperature, we expect that irreversible thermal damage will occur. Moreover, our observations imply that power budgets of 19 mW/mm² or less are indicated. Our experiments were completed with rat only, and more extensive work in other species is required to confirm the relevance of the current work to humans. Interestingly, this temperature threshold of 2.1°C determined in the rat retina is along the lines recommended by the international regulatory committees, which enforce an upper thermal increase of 2°C on implant surfaces for humans. International standard ISO 14708-1:2000 requires that no outer surface of an implantable part of the active implantable medical device be greater than 2°C above body temperature. ³⁴ This provides support for the use of the current technique of assessing microglial morphology changes as an assay of retinal response to temperature increases and implant contact. 

In conclusion, this study has shown that differentiation between heating, implant contact, and heated-implant contact is possible in vitro through assessment of alterations in retinal microglial morphology. Morphological changes including a decrease in arbor area and an enlargement of microglial somas were observed when the retinal tissue was heated for 1 hour. Implant contact induced mechanical trauma and caused microglial alterations that were exacerbated when this was compounded with thermal insult. To ensure that an implanted retinal prosthesis will not cause extensive damage to neighboring neurons, careful monitoring of implant temperature is strongly recommended. Thermal increases of 2.1°C (temperature of 38.7°C) were observed to cause significant activation of microglia, and we recommend that retinal implants restrict power dissipation to below 19 mW/mm² to the reduce potentially deleterious effects caused by tissue heating. 

Footnotes

 Supported by the Australian Research Council (ARC) through its Special Research Initiative (SRI) in Bionic Vision Science and Technology grant to Bionic Vision Australia (BVA). The Bionics Institute acknowledges the support it receives from the Victorian Government through its Operational Infrastructure Support Program.

Footnotes

 Disclosure: N.L. Opie, None; U. Greferath, None; K.A. Vessey, None; A.N. Burkitt, None; H. Meffin, None; D.B. Grayden, None; E.L. Fletcher, None