Investigations on artificially extending the spectral range of natural vision

Organic semiconductors are being explored as retinal prosthetics with the prime attributes of bio-compatibility and conformability for seamless integration with the retina. These polymer-based artificial photoreceptor films are self-powered with light-induced signal strength sufficient to elicit neuronal firing events. The molecular aspect of these semiconductors provides wide spectral tunability. Here, we present results from a bulk heterostructure semiconductor blend with a wide spectral response range. This combination elicits clear spiking activity from a developing blind-chick embryonic retina in the subretinal configuration in response to white light. The response is largely triggered by the blue–green spectral regime rather than the red-NIR regime for the present polymer semiconductor layer attributes.


I. INTRODUCTION
The possibility of eliciting firing activities from retinal ganglion neurons (RGCs) using artificial photoreceptors has opened up a range of options to restore and augment vision. 1,2In a healthy retina, the photoreceptors absorb light and transform it into electrical signals.These signals are transferred downstream to bipolar and amacrine cells, which synapse onto ganglion cells.The retinal ganglion cells (RGCs) send the information about the visual scene as a series of action potentials (spikes) to the brain via the optic nerve.The brain decodes the spiking of the population of these RGCs and forms a representation of the external world.This encoding-decoding system is the basis of vision in all vertebrates.In diseases such as retinitis pigmentosa (RP) and macular degeneration (MD), the photoreceptors (rods and cones) degenerate with time, leading to vision loss, even though the rest of the network is still present. 3The artificial stimulation of RGCs in patients with such diseases can lead to vision restoration.
Techniques such as optogenetics have been used where bipolar or RGCs are delivered a light-sensitive protein such as channelrhodopsin (ChR) that causes these cells to fire in response to light.Implanting of electrodes like platinum 4 and carbon nanotubes have been used for direct electrical stimulation of RGCs. 5 Using silicon photodiode implants, the incoming light is converted to NIR light with the help of a camera, and this NIR light is absorbed by the photodiodes, resulting in RGC firing due to electrical stimulation. 6These systems either require an external power source or are based on materials that do not work well in environments with liquids and cause neural tissue scarring due to being rigid and nonconformable.Organic semiconductors work without the need for an external power source, can be coated on flexible substrates, and are light-weight, patternable, and bio-compatible. 7,8he initial choice of the polymer semiconductor for the research groups active in this area was the well-studied poly(3-hexylthiophene-2,5-diyl) (P3HT). 9The material has good biocompatibility, and its absorption spectra span the visible range, making it a suitable candidate for a retinal prosthetic.Bulk heterostructure junction (BHJ) blends that exhibit higher quantum yield for light conversion to charges with P3HT as the acceptor and poly[N,N-bis(2-octyldodecyl)-naphthalene-1,4,5,8-bis(dicarboximide)-2,6-diyl]-alt-5,5-(2,2-bithiophene) (N2200) as the donor (4:1) have been used in both the epiretinal and subretinal configurations. 10,11Different systems have been shown to function as retinal prosthetics like p-n junctions of organic materials like phthalocyanine (H2Pc) and N,N-dimethyl perylene-3,4:9,10-tetracarboxylic diimide (PTCDI) on a metallic back contact 12 13 Aside from being used as retinal prosthetics, organic semiconductors have been used in wound healing 14 and optical regulation of neuronal growth. 15wo configurations are commonly used for interfacing the artificial photoreceptor layer with the retina.In the epiretinal configuration, the prosthetic is introduced adjacent to the retinal ganglion cell (RGC) layer.Such a configuration aims to stimulate RGCs directly without using the multilayered network architecture of the retina. 10It requires minimal surgery and can be used for any disease as long as the RGC layer is unaffected.In the subretinal configuration, the prosthetic is introduced near the degenerated photoreceptor to activate lightinduced changes in the bipolar layer.This method makes use of the layered retinal network to elicit activity from RGCs.This network activation of the upper retinal layers will help in eliciting physiologically relevant firing from RGCs unless there has been significant rewiring of the retinal circuitry, which is known to occur after vision loss in the case of diseases like RP.The main disadvantage of the subretinal implants is the requirement for a more invasive surgical technique compared to epiretinal implants.
The developing chick embryo provides an elegant, accessible model for demonstrating the use of organic BHJs as an artificial polymer photoreceptor (APPR) layer. 16The prominent spontaneous activity in the developing retina is well studied, 13 and a range of markers have been identified.The developing chick retina shows spontaneous bursts of correlated activity called retinal waves, 17 which occur every 1-2 min during this evolving developmental stage.There are no specific light-induced activities in the embryonic stage during the E12-E15 period, and this feature provides a suitable platform to observe functional changes in the APPR-interfaced retina.
Experiments with plain P3HT and P3HT: N2200 in the epiretinal and subretinal positions have been done before in the field.These materials have been shown to stimulate activity from the developing blind chick retina for a broad set of visible wavelengths (450-580 nm).However, they were unable to elicit activity from the retina when illuminated with Red/NIR light.This was attributed to the absorption peak of P3HT being at a wavelength of 520 nm.Thus, stimulation of the retina at red (600-700 nm) wavelength is necessary for complete vision restoration, and stimulation with NIR wavelengths of light can be used also for extending the visual range or alternatively can be used as a complementary part of the spectrum for RP or AMD patients with only partial vision loss.The advancement of materials for developing bulk heterostructure-based solar cells offers a wide set of choices for this purpose.However, it is not obvious that the high quantum yield and the wider spectral range aspects within the BHJ layer should translate to enhanced artificial photoreceptor features.
In this pursuit, we explore the viability of the donor-acceptor of DPP-IT4F (1:1) for this purpose.Poly [2,5- (DPP-DTT) is a high-mobility p-type polymer.It is a low bandgap polymer with a peak absorption in the 700-750 nm range.3,9-bis(2methylene- IT4F) is a non-fullerene-based acceptor (NFA) with energy levels appropriately aligned with the donor low-energy (bandgap) polymers.This combination in a device results in good charge separation and efficiency.The spectral range for photocurrent in DPP-DTT: IT4F BHJ devices extends into the infrared (IR) range of the spectrum.
We present results from our experiments using the DPP-DTT: IT4F polymer BHJ films interfaced to a developing blind chick retina (E12-E15) in the subretinal configuration.The response of RGCs to flashes of blue, green, red, infrared (IR), and white light stimuli was recorded using a multi-electrode array (MEA).

II. RESULTS
The BHJ film of DPP-DTT: IT4F in contact with the electrolyte media was studied for its spectral and transient response.The device configuration consisted of a 150 nm coated film on an indium tin oxide (ITO) substrate and a counter electrode of platinum wire in the electrolyte/media solution.The steady-state photocurrent response shown in Fig. 1(c) was recorded over a wide spectral range to measure the external quantum efficiency using a calibrated Si standard.The responsivity matched the absorption profile of DPP-IT4F ITO/BHJ/electrolyte structure as shown in Fig. 1(c).The response peaked around 700 nm, making it an ideal candidate as a photodetector in the near-infrared range.The typical responsivity of this device is in the range of 2-11 mA=W.The transient voltage response [V ph ðtÞ] signal as shown in Fig. 1(d).was also studied using a pulsed light source at different wavelengths with the light incident from the electrolyte side of the ITO/BHJ/electrolyte structure.The measurement shows a developed photovoltage in the ITO/BHJ/electrolyte structure for a wide range of wavelengths (400-750 nm).These evaluated devices were interfaced with the retina in a subretinal configuration.
All MEA recordings were done on developing chick retina (E12-E15) interfaced with the BHJ in the subretinal configuration.RGC firing was recorded using an 8 Â 8 grid multi-electrode array (Multi Channel Systems).The signals from 59 independent electrodes were amplified, digitized, and acquired by the MEA-2100 system and viewed using the software Multi Channel Experimenter.
The BHJ-coupled (APPR) retina showed clear light-elicited firing of RGCs on multiple electrodes in response to the white-light stimulus in the background of spontaneous activity and retinal waves.The white light LED source is essentially a convolution of a narrow-blue emitter (460-500 nm and a wide emission covering 500-630 nm).Bandpass filtered (300-3000 Hz) data from two representative electrodes showing stimulated spiking activity to white and blue LED sources can be seen in Fig. 2(a).To unearth cell-specific responses from the firing on individual electrodes, a technique called spike sorting was used to cluster spikes on a single electrode (multi-unit activity) into single units or templates, from here on called putative RGCs.
Light of different wavelengths was used as stimuli with varying widths of pulse and intensities to probe the nature of the stimulated activity.The pulse duration varied from 20 ms-2 s in the experiments.The stimuli were repeated a minimum of five times in 5 s intervals.
The firing rates were plotted after binning the response in 50 ms bins and then averaged over the number of trials.
The putative RGCs showed various firing patterns, with most cells being ON, some being ON-OFF, and some being OFF.The relative fraction of these characteristic responses depends on the chickembryo developmental stage.The firing rate and raster for one of the putative RGCs are shown in Figs.2(b) and 2(c).The spikes' latency relative to the stimulus's onset shows that the activity is network mediated by stimulation of the upper layers of the network, namely, the underdeveloped photoreceptors or bipolar cells. 18A set of RGCs in the chick and other vertebrates, called intrinsically photosensitive RGCs (ipRGCs), express melanopsin. 19Thus, these RGCs, in principle, can fire spikes in response to light without needing network activation from photoreceptors.These ipRGCs are known to be slower and need light input in the range of a few seconds to change their firing patterns.We observe spiking from RGCs coupled to the BHJ layer even for short pulses (200 ms).This, coupled with a latency of typical RGC responses, shows that the BHJ activates the outer regions of the developing chick retina and causes the firing of RGCs using the existing network.Neurons, in general, with the injection of charge, behave in a non-linear fashion. 20During small current injections, the firing rate of the neuron increases in a linear fashion, but as the current value increases, the firing rate plateaus off.The firing rate of RGCs from the BHJ-coupled retina also showed similar characteristics with respect to pulse width.The firing rate increased as the pulse width was increased from 200-1000 ms.Increasing the pulse width to 2000 ms reduced the firing rate of RGCs.
Local heating of the polymer surface due to non-radiative dissipation of the absorbed light can be a possible mechanism of coupling between the retina and BHJ.In this case, heating is unlikely to be a dominant factor for the stimulation.Previous results from P3HT have shown that a marginal local increase < 1 C for light intensities % 1 mW=cm 2 . 1 Light intensities used in the present case are one order of magnitude lower (0:1 mW=cm 2 ).Furthermore, in the experiments where the polymer was interfaced with the retina, the time between light flashes was typically 5 s.These slow-frequency stimuli have been shown to reduce heating effects in retinal prostheses.Two other main mechanisms, which have been proposed for coupling between the polymer and retina, are capacitive and Faradaic.In the capacitive case, there is a charge redistribution in the artificial cerebrospinal fluid (aCSF) due to the surface of the polymer BHJ getting charged due to light falling on it. 12,21This rearrangement of ions can change the extracellular ionic concentrations and trigger the neuron.In Faradaic processes, i.e., redox reactions, oxygen reduction is one of the primary reactions thought to occur near the polymer surface, 22 causing current to flow through the aCSF, resulting in neuronal firing.In our experimental arrangement, the distance between the retina and BHJ is % 300 lm, as opposed to the epiretinal case when the retina is in contact with the APPR.This implies that the interaction is long-range and of a sufficient magnitude to stimulate the retina.
The retina interfaced with the BHJ in the subretinal configuration exhibits clear RGC firing patterns in response to the periodic-pulse white light source.The response was markedly different from the retinal waves when stimulated with blue, green, and white light but showed no clear stimulated activity when red and IR light was incident.In spite of the higher magnitude of the transient-marker corresponding to the light pulse ON and OFF instant that appears in the MEA recording, the elicited spiking activity due to red (k > 630 nm) light was minimal.In comparison, the response to white or blue pulsed light sources was substantial.This color-dependent feature in the response of the APPR-coupled blind retina is interesting and requires a systematic study.It should be noted that transient photovoltage (V ph ðtÞ) studies of the bare device in contact with the electrolyte indicate a wavelength-dependent temporal response profile.The response to the pulsed light at a lower wavelength regime (400-600 nm) has a capacitive component (charging-discharging).The time constant of the V ph ðtÞ associated with the response to the longer wavelengths (k > 630 nm) appears to be much higher than the response to the shorter wavelengths (k < 630 nm).It was observed that the transient photocurrent [I ph ðtÞ] also decayed faster when red and infrared LEDs (time constant % 0:1 ms) were used compared to blue and white LEDs (time constant % 0:2 ms).This difference in the decay profile likely plays a role in coupling the polymer film to the retina.In this framework, the absence of an equivalent stimulated activity from the retina when the BHJ is illuminated with red/NIR light indicates the role of the time constant and carrier diffusion length vis-a-vis the film thickness.Increasing the intensity of the red source was not found to stimulate activity from the retina.This points to a different mechanism, possibly Faradaic (redox) reactions occurring at the polymer electrolyte interface, which may contribute to polymer film and retina coupling.To sort this out, systematic studies involving the polymer blends and interface electrochemical studies for redox reactions such as ROS production 23 and impedance studies have to be carried out.

III. CONCLUSION
A novel polymer BHJ (DPP: IT4F) has been shown in this work to elicit light-induced spiking activity when interfaced in the subretinal configuration with a developing blind chick retina.The spiking activity is markedly different from the spontaneous activity at these developmental stages.The latency of the response and the proximity of the polymer to the developing photoreceptor/bipolar side of the retina points to a network-mediated activation of the RGCs instead of direct activation.This activation mechanism will help elicit physiologically relevant spiking from the population of RGCs and would be effective as a visual prosthetic compared to direct activation.Even though the polymer BHJ responds to the incident light in the 650-750 nm range, it does not get translated to RGC spiking activity as opposed to light sources of similar power with k < 630 nm.This is most likely due to the differences in the kinetics of the charging and discharging cycle for blue/white light compared to red/NIR.These transient photovoltage properties play a bigger role in successfully coupling the retina with the BHJ.Changing film thickness has been shown to affect the temporal characteristics of photovoltage in other organic semiconductors. 24etailed studies with respect to the thickness and composition ratio of DPP to IT4F in the BHJ film for eliciting activity from the retina coupled with this BHJ in the longer wavelength regions need to be explored.

IV. METHODS A. Retina dissection and MEA recordings
Methods of retina dissection and MEA recordings have been mentioned in an earlier paper. 11Briefly, retinas from developing chick (Gallus gallus domesticus) embryos (E12-E15) were obtained after cracking the egg, sacrificing the animal, and enucleation of the eye.The retina was then placed in ice-cold aCSF (124 mM NaCl, 25 mM NaHCO 3 , 1:2 mM HEPES, 5 mM KCl, :2 mM MgSO4, 2:5 mM CaCl2, and 10 mM glucose) bubbled with carbogen gas (95%O 2 þ5%CO 2 ) to maintain a pH of 7.4.The retina was then cut into smaller pieces to prevent folding and a suitable piece of retina was collected with a nylon mesh and placed RGC side down on the MEA maintained at 37 C. Another two pieces of nylon mesh are folded in half and kept on top of the retina attached to the nylon mesh to act as spacers when interfacing the polymer film with the retina.The absence of the spacers was found to cause fast death (absence of RGC spikes) from the isolated retina piece likely due to there being no recycling of media close to the retina.The BHJ was introduced on top of the retina in a subretinal configuration.The retina was left on the MEA for a 30-min period for the recorded action potentials to become stable.All stimuli were presented only after visual confirmation of retinal waves, indicating a healthy developing retina.The retina is perfused with fresh aCSF at 1 ml=min during the experiments.The data were recorded at 25 kHz.

B. Device fabrication and characterization
The polymer BHJs were spin-coated on ITO-coated glass substrates after sonication with acetone and IPA and RCA treatment.DPP-DTT and IT4F were weighed in a 1 : 1 ratio and dissolved in dichloro benzene to make the final solution concentration of 10 mg=ml.The solution was dissolved overnight at room temperature before spin-coating at 1000 rpm.The freshly coated BHJ films were left to anneal at 90 C for 1 h before interfacing with the retina.The lock-in measurements were done with a halogen vapor lamp as the source.The scan was done from k ¼ 400 to 800 nm at 1 nm steps, with a chopper frequency of 75 Hz.The response of the ITO/BHJ/ electrolyte system was normalized with a Si photodiode as a reference.
The V ph ðtÞ measurements were done on an oscilloscope for different wavelength LEDs with a frequency of 1 kHz.The recorded data were smoothened and then replotted.

C. Light stimuli
Light stimuli were provided as full-field flashes of LED ON and OFF with different intensities and pulse widths.All stimuli were repeated at least five times over a period of five seconds for averaging purpose.The LEDs' wavelength was blue (k ¼ 450 nm), green (k ¼ 525 nm), red (k ¼ 630 nm), IR (k ¼ 750 nm), and white.The power of the LEDs was in the range of 50-100 lW/cm 2 .

D. Spike sorting
The recordings from an MEA consist of data from 59 independent electrodes.Each electrode can possibly have spikes from more than one RGC.The process of grouping the spikes on an electrode, i.e., the multi-unit activity, into putative single units is called spike sorting.Here, we use an open-source spike sorting toolbox largely written in Python called Spyking-Circus. 25Briefly, the data are first bandpass filtered (300-3000 Hz), then spikes are detected using a threshold method (6 S.D).Principal component analysis (PCA) is done after the detection to reduce the problem's dimensionality before clustering.A density-based clustering method is done on a subgroup of the detected spikes.After identifying clusters, the remaining spikes are put into different clusters based on a template matching step.The final templates are checked manually to verify the results of the spike sorting step, and templates which pass certain guidelines, namely, < 3% refractory period violations, firing rate > 10 Hz, are called putative RGCs, which are then used for further analysis.A total of 42 putative RGCs from three recordings were identified using the above metrics and chosen for further analysis.

E. Rasters and firing rates
After spike sorting, putative RGCs are identified, as mentioned earlier.Spikes for repeats of the same stimuli (either wavelength, intensity, or pulse width) are plotted as tick marks after alignment of the spike times with the stimulus time.Spikes until the next stimulus trial (0-5 s) are taken as part of the response.Firing rates are plotted after binning the spikes from the rasters in 50 ms bins and dividing by the total number of trials to obtain the average firing rate.The inter-spike interval (ISI) shown in the inset of the raster and firing rate plots is calculated by binning the time between successive spikes in the response of a putative RGC into 5 ms bins and plotting the counts in the first 20 bins.

FIG. 1 .
FIG. 1.(a) Chemical structure of DPP and (b) IT4F.(c) The normalized absorbance and responsivity (peak rms photocurrent ¼ 11 mA=W) profile of the ITO/BHJ/electrolyte device.The responsivity was measured using a lock-in amplifier and a halogen lamp as the source.The light intensity was normalized using a Si photodiode.(d) V ph ðtÞ of the ITO/BHJ/electrolyte structure for blue (k ¼ 450 nm), red (k ¼ 630 nm), and white LEDs with power in the range of 50-100 lW=cm À2 ; square pulse indicates LED ON and OFF.

FIG. 2 .
FIG. 2. (a) Band-pass filtered data (300-3000 Hz) showing elicited spiking activity from RGCs on two representative electrodes, white LED top electrode and blue LED (k ¼ 425 nm) bottom for a BHJ coupled E12 retina, the square pulse shows the ON and OFF of the LED.(b) Firing rate of two putative RGCs to a 1s flash of white light (bin size ¼50 ms).Inset shows the inter-spike interval (ISI).(c) Raster plot of the response of the first RGC in (b) to 12 trials of the 1 s white light stimuli.(d) A schematic showing the developing chick retina coupled to DPP-IT4F BHJ placed on the MEA with light shining from the bottom.