A hybrid P3HT-Graphene interface for efficient photostimulation of neurons
Mattia L. DiFrancesco, Elisabetta Colombo, Ermanno D. Papaleo, José Fernando Maya-Vetencourt, Giovanni Manfredi, Guglielmo Lanzani, Fabio Benfenati
AA hybrid P3HT-Graphene interface for ef fi cient photostimulation ofneurons Mattia L. DiFrancesco, PhD a , b , , Elisabetta Colombo, PhD a , b , ,Ermanno D. Papaleo, MD a , b , Jos (cid:1) e Fernando Maya-Vetencourt, PhD a ,Giovanni Manfredi, PhD c , Guglielmo Lanzani, PhD c , Fabio Benfenati, MD a , b , * a Center for Synaptic Neuroscience and Technology, Istituto Italiano di Tecnologia, Largo Rosanna Benzi 10, 16132, Genova, Italy b IRCCS Ospedale Policlinico San Martino, Genova, Italy c Center for Nano Science and Technology, Istituto Italiano di Tecnologia, Milan, Italy a r t i c l e i n f o
Article history:
Received 7 December 2019Received in revised form12 February 2020Accepted 16 February 2020Available online 17 February 2020 a b s t r a c t
Graphene conductive properties have been long exploited in the fi eld of organic photovoltaics and op-toelectronics by the scienti fi c community worldwide. We engineered and characterized a hybrid bio-interface in which graphene is coupled with photosensitive polymers, and tested its ability to elicit light-triggered neural activity modulation in primary neurons and blind retina explants. We designed such agraphene-based device by modifying a photoactive P3HT-based retinal interface, previously reported torescue light sensitivity in blind rodents, with a CVD graphene layer replacing the conductive PEDOT:PSSlayer to enhance charge separation. The new graphene-based device was characterized for its electro-chemical features and for the ability to photostimulate primary neurons and blind retina explants, whilepreserving biocompatibility. Light-triggered responses, recorded by patch-clamp in vitro or MEA ex vivo ,show a stronger light-transduction ef fi ciency when the neurons are interfaced with the graphene-baseddevice with respect to the PEDOT:PSS-based one. The possibility to ameliorate fl exible photo-stimulatingdevices via the insertion of graphene, paves the way for potential biomedical applications of graphene-based neuronal interfaces in the context of retinal implants. ©
1. Introduction
The development of new technologies for the stimulation ofneuronal cells and tissues paces with the performance tuning ofstate-of-art strategies, which fl ourish thanks to the constantinnovation in the fi elds of nano- or opto-electronics and materialscience.Graphene is a 2D material, fi rstly isolated and identi fi ed in 2004[1], showing outstanding properties, such as one-atom-thicknessand consequent low weight/surface ratio, extremely high elec-trical conductivity, high resistance to stretch and lateral deforma-tion despite a high degree of fl exibility, transparency, andcompatibility with live cells and tissues. Thanks to these features,graphene has long been studied for applications in the fi eld of biomedical technologies. In the last decades, an increasing interestrose in the scienti fi c community for the development of novelgraphene-based scaffolds characterized by biocompatibility [2,3],long-term durability [4], and fl exibility [5]. Prototypes under studyare wearable devices [6], implantable microelectrodes and micro-transistors [7], or in vivo cortical, retinal and peripheral nerveprobes [8].On the other hand, graphene has also been extensively exploitedto enhance photoconversion processes in organic photovoltaic andphotocatalytic devices [9]. In this framework, graphene has beenemployed as transparent electrode [10], catalytic counter electrode[11], and charge transport layer [12], just to mention a few.We previously characterized light-sensitive interfaces based onphotovoltaic polymers able to modulate the activity of HEK-293 cells [13] and neurons [14,15], which proved to be successful invisual restoration in the Royal College of Surgeons (RCS) rat modelof retinal dystrophy [16,17], characterized by photoreceptordegeneration associated with mutation of the mertk geneexpressed by the retinal pigmented epithelium. This triggers a * Corresponding author. Center for Synaptic Neuroscience and Technology, Isti-tuto Italiano di Tecnologia, Italy.
E-mail address: [email protected] (F. Benfenati). Equal contribution.
Contents lists available at ScienceDirect
Carbon https://doi.org/10.1016/j.carbon.2020.02.0430008-6223/ © e hagocytosis defect progressive loss of photoreceptors with accu-mulation of cytotoxic debris in the subretinal space during the fi rstmonths of postnatal life [18]. For this in vivo application, the full-organic implant was composed of a silk fi broin substrate, a poly(3,4-ethylenedioxythiophene)-poly (styrenesulfonate)(PEDOT:PSS) and a photoactive poly-3-hexylthiophene (P3HT)layers, inspired by the core architecture of a typical organic solarcell. In this con fi guration, P3HT is in direct contact with the elec-trolytic extracellular fl uid, which acts as a cathode together withthe tissue itself. Following light-evoked exciton dissociation, theP3HT surface at the tissue interface negatively polarizes and in turnmodulates the membrane voltage of neurons grown in contact withthe photoactive layer [19].We therefore realized a novel neural interface prototype on a fl exible PolyEthylene Terephthalate (PET) substrate by substitutingthe PEDOT:PSS layer with CVD graphene (G-CVD) in order toexploit the more favourable work function of G-CVD for chargeextraction in this architecture, and therefore enhance neuronalphoto-modulation. We ran parallel characterizations of the newand the previous prototypes containing PEDOT:PSS from a bio-physical point of view, in contact with primary hippocampal neu-rons and blind retinal explants.The results exploit the unique performance of graphene forincreasing the photoconversion performances of conjugated poly-mers in a hybrid neural interface that may result in the possibility ofreducing the light power threshold needed to trigger neuronalmodulation.
2. Results
Graphene was produced by Chemical Vapour Deposition (CVD)(Graphenea, Spain), transferred to PET substrates (30 m m thickness)and characterized by spectroscopy. Raman spectra show the G and2D bands typical of a few layers CVD graphene, as highlighted bythe two peaks deconvolutions (Fig. 1a). Successively, P3HT (30 mg/ml; Sigma Aldrich) was spin-coated onto the PET/G-CVD structureto a fi nal fi lm thickness of about 100 nm.The comparison between the optical transmittance of the three-layered devices with G-CVD and PEDOT:PSS as interlayers resultedin very similar spectra, denoting that graphene hydrophobicity didnot interfere with the wettability of the substrate before P3HTcoating, nor with the P3HT optical properties and absorption range(400 e
650 nm) (Fig. 1b). Next, we measured the ability of the gra-phene interlayer to enhance the surface photo-potential by meansof a patch-clamp pipette micromanipulated in the close proximityof the device surface in electrophysiological solution. Illuminationwas provided at 540 nm in order to activate the photosensitivelayer P3HT with light close to its absorbance peak [14]. Surfacepotential recordings in response to green light pulses were realizedwith the G-CVD and PEDOT:PSS interlayers, both terminated with aP3HT-based heterojunction fi lm in order to emphasize the photo-potential and improve the signal to noise ratio. The recordingsshowed a signi fi cant enhancement of the photogenerated chargesin the devices containing graphene with respect to the one withPEDOT:PSS (Fig. 1c), setting the basis for an improved engineeringof the previous prototype.We next performed cyclic-voltammetry measurements in thedark and under illumination at 20 mW/mm . The neural interfacearchitecture containing G-CVD shows the typical cathodic evolu-tion of a P3HT thin fi lm on a conductive layer (such as on ITO), bothin the dark and upon illumination (Fig. 1d). The current densityquanti fi cation for the two prototypes depicted a strong improvement of the photogenerated current whenever graphenewas present as interlayer (Fig. 1e).We also compared the thermal properties of the two neuralinterfaces in electrophysiological environment with a patch-clamppipette and ampli fi er, exploiting the intrinsic resistance depen-dence upon temperature (calibrated pipette measurements; insetFig. 1f) [13]. For both the PEDOT:PSS-based and the G-CVD-baseddevices, the temperature increase in proximity of the P3HT surfacewas recorded at increasing light power densities for 50 and 500 msillumination durations, showing that heating starts to play a rele-vant role only with long stimuli (500 ms) at high power ( >
10 mW/mm ). Fully overlapping temperature curves were obtained withthe two devices, indicating that light absorption by P3HT isresponsible for the light-induced heat generation.Altogether, these results depict a suitable new architecture ofthe previous P3HT-based neural interface, with the substitution ofthe PEDOT:PSS layer with G-CVD able to improve the electro-chemical properties of the device upon illumination, withoutaffecting materials integrity or producing excessive overheating. Neuronal adhesion and viability of primary neurons grown incontact with ITO/P3HT:PCBM device was previously reported [14].We address here whether G-CVD interferes with the properties ofthe device, by comparing growth and viability of primary hippo-campal neurons on PET/G-CVD/P3HT ( þ G-CVD) devices with thoseon PET/PEDOT:PSS/P3HT ( þ PEDOT/PSS) or on a pure PET surfaceused as control. Hippocampal neurons were cultured for 14 days in vitro (DIV) onto the respective substrates coated with poly- L -lysine (PLL), and their viability was estimated as the ratio between fl uorescein diacetate-positive and nuclear DAPI labelling-positivecells. As shown in Fig. 2a,b, a high percentage viability resultedfrom all tested experimental groups (mean ± SEM:PET ¼ ± þ PEDOT:PSS ¼ ± þ G-CVD91.51 ± fi cant changes within the three testeddevices (Fig. 2a and b). This indicated that G-CVD did not affect thefeatures of the polymeric device, allowing normal adhesion, growthand survival of neuronal networks. In view of the increased photo-potential due to the enhancedcharge separation in the presence of G-CVD revealed by surfacepotentials and cyclic voltammetry measurements, we investigatedwhether the larger light transduction ef fi ciency could be translatedinto an improved modulation of neuronal activity. Primary hippo-campal neurons were cultured in vitro on top of either þ G-CVDor þ PEDOT:PSS devices (in addition to a pure PET control), asdescribed above. Fourteen days after plating (14 DIV), neurons weresubjected to patch-clamp measurements in the current-clamp (CC)con fi guration with two distinct protocols: (i) without current in-jection (I ¼ (cid:2) ± (cid:2) ± þ PEDOT:PSS and (cid:2) ± þ G-CVD; see Suppl. Figure 1), and (ii) by injecting an outward currentto maintain the membrane voltage at (cid:2)
70 mV (I
Holding (cid:2)
70 mV) toreproducibly mimic the physiological conditions. Illuminationprotocols were applied at 1 or 15 mW/mm light-power density for50 e
500 ms, using a 550/15 nm LED source.During light stimulation, neurons showed a hyperpolarizingresponse when measured in the I ¼ Holding (cid:2)
70 mV modality, with no-
M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e317
M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e317 embrane voltage modulation of neurons in contact with the PETcontrol device (Fig. 2c and d). A minority of cells showed anopposite behaviour with depolarization at I ¼ , 500 ms) and hyperpolarization at I Holding (cid:2)
70 mV(14.82% of neurons at 15 mW/mm , 500 ms). Light-induced hy-perpolarization and depolarization were measured between theonset and the offset of the light stimulus. To measure net effects oflight stimulation, the voltage change measured with neurons on the PET alone control was subtracted from the hyperpolarizationand depolarization amplitudes measured with the þ G-CVD andthe þ PEDOT:PSS devices for each stimulation protocol ( D Device e D Ctrl; Fig. 2c and d). The light-dependent membrane voltagemodulation resulted signi fi cantly enhanced in the presence of G-CVD in most stimulation protocols, both in the I ¼ Holding (cid:2)
70 mV modalities with stronger effects with the lattermodality. We concluded that the G-based device modulates
Fig. 1.
Graphene-based neural interface characterization. a ) G-CVD layer transferred onto a PET substrate shows a typical Raman spectrum, where the G and 2D band of graphene clearly emerge from the superimposed PET signal. Bothpeaks were deconvoluted from the acquired spectrum and their respective Gaussian fi t reported. b ) Optical transmittance of the full device in the visible range shows no inter-ference of the graphene layer with the P3HT main absorption range (400 e
650 nm). c ) Comparison of the surface potentials of P3HT:PCBM fi lms on either PEDOT:PSS/PET or G-CVD/PET shows a signi fi cantly more pronounced photovoltaic effect on graphene-based devices. The recording is performed with an open patch-clamp pipette (~4 M U ) in current-clampat I ¼
0, micromanipulated in the proximity of the device surface and under illumination at 1 mW/mm d ) Cyclic-voltammetry of a P3HT/G-CVD/PET device measured in phosphate-buffered saline (PBS) in the dark and under illumination at 20 mW/mm and 100 mV/s e ) The comparison between the cathodic current densities of the two prototypes under studyshows a stronger effect on G-CVD-based devices. f ) Temperature characterization of the P3HT/G-CVD/PET device. The temperature variation in proximity of the device surface underlight stimulation of different intensities (540 nm) is monitored by exploiting the intrinsic temperature dependence of a patch-clamp open pipette resistance (calibrated as shown bythe Arrhenius plot in the inset). g ) I/I values and the extrapolated temperature variation curve are shown for the experimental conditions used in the following in vitro experiments(50 and 500 ms durations). (A colour version of this fi gure can be viewed online.) M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e euronal activity more ef fi ciently, particularly when neurons areuniformly polarized at (cid:2)
70 mV.
Taken into account the bimodal response of membrane voltagemodulation to light stimuli, we asked whether the same featurewas preserved over a wide range of injected currents and mem-brane voltages. To this aim, primary hippocampal neurons weresubjected to patch-clamp measurements in the CC con fi gurationranging from (cid:2)
300 pA to þ
300 pA, and in the voltage-clampcon fi guration (VC) ranging from (cid:2)
110 mV to þ
10 mV. The currentinjection or the membrane voltage was maintained all along therecorded sweeps, before, during and after light-stimuli that wereapplied using the same protocols described above (15 mW/mm for500 ms; Fig. 3).In the CC con fi guration, we observed an increasing depolarization towards negative clamped currents and, on thecontrary, increasing hyperpolarization towards positive clampedcurrents (Fig. 3a; blue line and red line for þ PEDOT:PSS and þ G-CVD respectively). Such a current-dependency of the light-evokedmembrane voltage response could be described for neurons incontact with both the þ PEDOT:PSS and the þ G-CVD devices(Fig. 3c, left panel), and was signi fi cantly higher with G-CVDat (cid:2)
300 pA. On the contrary, no current injection-dependency wasdescribed for neurons cultured in contact with PET control devices(Fig. 3a,c; black). With less intense and shorter light stimuli,the þ G-CVD interface outperformed the þ PEDOT:PSS one: whilesmall and non-signi fi cant responses were observed to 1 mW/mm for 50 ms for both interfaces, 15 mW/mm for 50 ms were suf fi cientfor achieving a signi fi cant voltage modulation for the þ G-CVDinterface (Suppl. Figure 2).The same dependency was maintained when switching themeasurement from CC to VC con fi guration. As expected, lightstimuli induced an inward current at hyperpolarizing voltages, that Fig. 2.
Interaction of primary hippocampal neurons with G-based polymeric devices. a ) Viability of primary hippocampal neurons grown onto the various devices. Neurons grown for 14 DIV in contact with PET control, a P3HT/PEDOT:PSS/PET device ( þ PEDOT:PSS) ora P3HT/G-CVD/PET ( þ G-CVD) were stained with the DAPI nuclear stain and the live stain Fluorescein Diacetate. b ) The resulting percent viability was measured as the ratio betweenthe number of Fluorescein Diacetate-positive cells and the total number of cells stained with DAPI. One-way ANOVA/Kruskal-Wallis tests (n ¼
6, 6, 5 for PET, þ PEDOT:PSS and þ G-CVD respectively). c , d ) Light-induced membrane voltage modulation measured in current-clamp con fi guration with no-current injection ( c , I ¼
0) or by setting the holding potentialat (cid:2)
70 mV ( d , I Holding (cid:2)
70 mV).
Upper panels : Representative traces of primary hippocampal neurons in contact with control PET (black traces), þ PEDOT:PSS (blue traces) or þ G-CVD (red traces) devices in response to light stimulation at 15 mW/mm for 500 ms (green shaded area). Lower panels: membrane voltage hyperpolarization ( c ) or depolarization ( d )in response to various light-stimulation protocols as indicated. Unpaired Student ’ s t -test (D ’ Agostino/Pearson ’ s normal distribution) or Mann-Whitney ’ s U test (D ’ Agostino/Pearson ’ snon-Gaussian distribution) between þ PEDOT:PSS and þ G-CVD within the same experimental condition ( c ) N ¼
35, 27, 41, 30, 38, 27, 41, 33; d ) N ¼
30, 24, 30, 25, 29, 24, 28, 26following the order in the plots from left to right). (A colour version of this fi gure can be viewed online). M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e ecame outward during depolarization (Fig. 3b). When the I/Vcurves of the light-triggered current were plotted as a function ofthe clamped voltage (Fig. 3c; right panel), no signi fi cant differenceswere found between þ PEDOT:PSS and þ G-CVD interfaces, despitea tendency towards a higher light-sensitivity of þ G-CVD interfacesat very negative voltages. On the contrary, when we calculated theslope of I/V curves per each cell, we found a signi fi cantly strongerimprovement of current modulation from neurons in contact withthe þ G-CVD devices with respect to the þ PEDOT:PSS benchmark(Fig. 3d). As described for CC measurements, we repeated experi-ments at various intensities and/or durations of the light stimulialso in the case of VC con fi guration (Suppl. Figure 1). Under thiscon fi guration, we found that G-CVD improved the light-triggeredcurrents with respect to the þ PEDOT:PSS device. The enhance-ment was signi fi cant with stimuli at 1 mW/mm for 500 ms or at15 mW/mm for 50 ms, while only a trend was present at 1 mW/mm for 50 ms.To better elucidate whether membrane ionic conductances wereinvolved in phototransduction, we calculated, per each cell, thereversal voltage (E rev ) of the light-triggered current (Fig. 3e). Wefound that the reversal potentials of the light-evoked effects of thetwo devices were closely similar. This suggests that the samecellular mechanism underlies the generation of voltage-dependentcurrents under light-stimulation, and that the enhanced effect of G-CVD is attributable to its improved capability to separate the charges generated in the P3HT photosensitive layer. fi ciency of light-induced fi ringmodulation in primary hippocampal neurons We previously reported the ability of PEDOT:PSS/P3HT-basedneuronal interfaces to decrease spontaneous neuronal fi ring [15]as a consequence of the hyperpolarization induced by light-stimuli.In the context of the enhanced membrane voltage modulation ofthe þ G-CVD device, a more powerful inhibition of neuronal fi ringcould be expected. To assess this aspect, we cultured primaryhippocampal neurons on either þ PEDOT:PSS or þ G-CVD devicesat high density, in order to increase network activity and basalneuronal fi ring, and recorded action potentials (APs) in the CCcon fi guration at I ¼
0, using the previously described illuminationprotocols. Peristimulus time histograms (PSTHs, Bin size 25 ms)were then computed to describe the AP fi ring under illumination.The analysis revealed a pronounced inhibition of neuronal ac-tivity during light stimulation (Fig. 4a and b). Such a modulation ofAP fi ring in response to the light-evoked hyperpolarization was nothomogeneously distributed in the neuronal populations, with someneurons that were insensitive to illumination. Recorded neuronswere therefore sorted into responding cells, showing light-induceddecrease in fi ring, and non-responding cells, whose fi ring wasunaffected by illumination. While the ratio between responding Fig. 3.
Current- and voltage-dependence of light-induced modulation of membrane potential in G-based polymeric devices. a,b)
Representative traces of light-induced voltage ( a ) and current ( b ) modulation in primary hippocampal neurons grown on PET control (black), þ PEDOT:PSS (blue) and þ G-CVD(red) devices and subjected to light stimulation at 15 mW/mm for 500 ms (green shaded area). c ) I/V curves showing the light-induced voltage modulation (15 mW/mm for500 ms) measured in the current-clamp (CC) con fi guration from (cid:2)
300 to þ
300 pA (left ), and the current modulation measured in the voltage-clamp (VC) con fi guration from (cid:2) þ
10 mV ( right ). Two-way ANOVA/Tukey ’ s tests (CC e I/V, N ¼
7, 15, 13 for PET, þ PEDOT:PSS and þ G-CVD respectively; VC-IV N ¼
6, 9, 8 for PET, þ PEDOT:PSS and þ G-CVDrespectively). d ) Slope of single-cell current modulation extrapolated from linear fi tting of V/C e I/V curves. Mann-Whitney ’ s U test (Rout test for outliers, Q ¼ ¼
7, 8for þ PEDOT:PSS and þ G-CVD respectively). e ) Reversal membrane voltage of the light-triggered current extrapolated from V/C e I/V curves. Unpaired Student ’ s t -test (N ¼
9, 8for þ PEDOT:PSS and þ G-CVD respectively). (A colour version of this fi gure can be viewed online). M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e nd non-responding neurons was similar between the two hybriddevices (Fig. 4c), PSTH analysis performed during the illuminationperiod with a bin size of 500 ms (0 e fi cant more intense fi ring inhibition with the G-CVD-baseddevice with respect to the þ PEDOT:PSS device (Fig. 4d). fi ciently recover light-sensitivity in blindretina explants In order to measure the degree of neural activity modulationexerted by devices, we recorded extracellular action potentialselicited by a fl ash of light onto explanted dystrophic retina from 12to 14 months-old dystrophic RCS rats, an experimental model of Retinitis Pigmentosa in which photoreceptor degeneration is due toa mutation in the
Mertk gene [17,20]. For the recordings, theexplanted retina was placed with the retinal ganglion cells (RGCs)in contact with multielectrode array (MEA) electrodes (epiretinalrecording) with the devices layered on the external retina in placeof the degenerated photoreceptors. In this subretinal implantcon fi guration, the P3HT surface was in contact with the remnant ofthe outer plexiform layer and the inner nuclear layer (INL) repre-sented by bipolar cells that are not directly affected by the degen-erative process. The subretinal con fi guration of the device, withlight passing through the inverted microscope and reaching theRGC layer fi rst, recapitulates the physiological pathway of visualstimulation that reaches the light sensitive outer retina after crossing the inner retina layers. While no signi fi cant effect waspresent in the absence of any device or with PET alone substrates, asigni fi cant fi ring modulation was observed in the presence of P3HT-based prototypes in response to light stimulation, as shown byPSTH analysis (Fig. 5a). Interestingly, þ G-CVD triggered a robustlight-dependent modulation of the overall fi ring rate of RGCs that,at the highest power (500 ms, 540 nm at 41 mW/mm ), wassigni fi cantly more intense than that observed in the presenceof þ PEDOT:PSS devices (Fig. 5b).A better understanding of the dynamics of the effect mediatedby the P3HT-based prototypes was obtained by separately consid-ering light-dependent decreases and increases in the fi ring rate ofRGCs, reminiscent of the physiological OFF and ON responses in theretina. Thus, we de fi ned the recorded patterns as OFF-like re-sponses when RGCs were silenced by light stimulation and ON-likeresponses when RGC fi ring was stimulated during the light pulse.The two types of light-modulation are represented by the rasterplots of Fig. 5c and d for OFF-like and ON-like type of light re-sponses, respectively. The fi ring modulation was quanti fi ed forthe þ PEDOT:PSS and þ G-CVD devices at both 16 and 41 mW/mm light power densities during and after the illumination pulse andcompared with the basal dark fi ring rate. OFF-like responses wereindependent of the type of device employed at both light in-tensities, and their rebound fi ring occurring after light offset wassigni fi cantly reduced with þ G-CVD devices at both power densities(Fig. 5c). On the other hand, ON-like responses were strongly and
Fig. 4.
Enhanced light-dependent fi ring modulation in G-based polymeric devices. a) Representative traces of action potentials (black line) from hippocampal neurons grown on PET (top), þ PEDOT:PSS (middle) and þ G-CVD (bottom) devices under light stim-ulation at 15 mW/mm for 500 ms (green box). Red lines represent average traces, showing the light-induced hyperpolarization. b) Peristimulus Time Histogram (PSTH) analysis(bin size ¼
25 ms) of action potential fi ring normalized to the pre-pulse fi ring activity (from (cid:2) for 500 ms;dashed line corresponds to 100%. N ¼
13 and 16 light-responding cells for þ PEDOT:PSS and þ G-CVD, respectively. c) Pie-charts of neurons with no-response to light stimulation(grey) versus neurons responding to light-stimuli with a decrease in fi ring rate (colour) on þ PEDOT:PSS devices (blue; 13 responding cells versus þ G-CVD devices (red; 16 responding cells versus d) PSTH analysis (bin size ¼
500 ms; means ± SEM) of action potential fi ring during the 500 ms light stimulus.Two-way ANOVA/Tukey ’ s tests (n ¼
5, 13, 16 for PET, þ PEDOT:PSS and þ G-CVD, respectively). (A colour version of this fi gure can be viewed online). M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e igni fi cantly potentiated with þ G-CVD devices, as compared withthe þ PEDOT:PSS devices. Indeed, ON-like RCSs onto þ G-CVD de-vices exhibited a more than two-fold increase in fi ring duringillumination, in the absence of signi fi cant effects in the fi ring ac-tivity after light offset (Fig. 5d).When the frequency of occurrence of ON-like and OFF-like re-sponses of the recorded RGC population was analysedwith þ PEDOT:PSS and G-CVD devices, we found a signi fi cant in-crease of the frequency of ON-like fi ring, that was accompanied by a corresponding decrease in the OFF-like responses to light withthe þ G-CVD device. This feature can be particularly interesting inre-establishing a correct balance of ON and OFF responses in animalmodels of
Retinitis Pigmentosa, in which the progression of retinaldegeneration is associated with inner retina rewiring anddecreased ON/OFF RGC ratio, as it is the case of the RCS rat used inthis study [21].
Fig. 5.
Light triggered ON/OFF-like retinal ganglion cell responses from blind RCS retinas. a ) Peristimulus time histograms (PSTHs) (bin size ¼
20 ms; means ± SEM) of the average RGCs activity epiretinally recorded on MEAs in the presence of PET, P3HT/PEDOT:PSS/PET( þ PEDOT:PSS) or P3HT/G-CVD/PET ( þ G-CVD) positioned subretinally. RGC fi ring changes in response to light stimuli (green shadowed area; 500 ms at 41 mW/mm ) werenormalized to the pre-pulse fi ring activity (from 0 to 1.5 s). b ) The average fi ring rate during the 500 ms of green light pulse is reported as a function of the light power density(more than 20 RGCs for each group from 5 animals, ***p < ’ s tests). c ) Representative raster plot of an OFF-LIKE light response recorded from a blind RCSretinal explant in subretinal con fi guration and stimulated by 500 ms (green light at 41 mW/mm ) ( left ). The quanti fi cation of the mean fi ring modulation during and after lightstimulation is shown for illumination intensities of 16 mW/mm ( middle ) and 41 mW/mm ( right ). d ) The same representation of the fi ring rate of panel c) for ON-LIKE lightresponses (n ¼ < < ’ s U test). e) Pie-charts of ON- and OFF-like responses from RGC neurons evoked by lightstimulation with subretinal þ PEDOT:PSS or þ G-CVD devices. (A colour version of this fi gure can be viewed online.) M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e . Discussion and conclusions An increasing interest is growing in the scienti fi c community fornovel technologies able to modulate cellular activity using light as atrigger. A major issue is to provide more ef fi cient and spatiallyresolved technologies, without the drawback of genetic manipu-lation. In this scenario, organic neuronal interfaces based on con-jugated polymers have already proved to be a viable tool for themodulation of neuronal activity in vitro [14,15], and for the recoveryof visual properties in rodent models of retinal dystrophy [16].We report here a novel strategy for the improvement of con-jugated polymer-based light-sensitive devices by the substitutionof the PEDOT:PSS conductive layer with highly conductive G-CVD.Thanks to the more favourable work function of G-CVD for chargeextraction with respect to PEDOT:PSS, this architecture allows amore ef fi cient separation of charges generated under illuminationin the neighbouring semiconductive P3HT, thereby enhancingneuronal photostimulation. Such a gain of function exhibited by theG-based device allows a more effective modulation of neuronalactivity in vitro , as described by the analysis of membrane voltagechanges and AP fi ring modulation, as well as ex vivo in explants oflight-insensitive, degenerate retinas.Graphene revealed to be effective in improving the ef fi ciency ofseveral devices such as neural interfaces [22] or photovoltaic solarcells [23], and the enhanced work function and consequentneuronal photostimulation reported here stands out anotherfruitful application of Graphene in the fi eld of neuronal stimulatingdevices.Interestingly, the light-induced modulation of neuronal activityappeared to be related to the initial state of membrane voltage: themore hyperpolarized is a neuron, the more pronounced will be thedepolarization triggered by illumination and, conversely, the moredepolarized is a neuron, the more it will hyperpolarize under illu-mination. An analogous phenomenon was also found varying themembrane voltage of HEK-293 cells during light-stimulation pro-tocols [13], that was described as a concerted modulation of cellularcapacitance, membrane resistance and reversal potential. Thisfeature could bring about an effect of conjugated polymer-baseddevices, depending on the speci fi c excitation/inhibition balance ofthe neuronal network subjected to illumination, which has to betaken into account for possible in vivo applications, such as retinalimplantable devices, where a different modulatory effect could beachieved as a function of the targeted retinal pathway. Such aprosthesis is meant to be implanted in vivo in the subretinalcon fi guration, with the photoactive P3HT layer in contact with bi-polar cells. In the retina, both photoreceptors and bipolar cells workexclusively by oscillations of the membrane potential, not beingable to generate action potentials. Depolarizations and hyperpo-larizations occur in response to light in on- and off-bipolar cells,respectively, and mediate activation or silencing of the respectiveon- and off-RGCs. In this respect, the under-threshold modulationsof the membrane potential of bipolar cells by the device can bephysiologically relevant for generating light-mediated signals inthe degenerate retina devoid of photoreceptors. Ex vivo measurements with explanted samples of blind retinademonstrated that graphene-based stimulation, with respect toPEDOT:PSS, was overall more effective in eliciting light-dependentresponses. Furthermore, the results obtained by the new prostheticprototype were also better resembling an ideal physiologicalpattern of light response: the ON/OFF-like responses balance,altered by degeneration-induced inner retina rewiring, was re-established by the graphene-based device, differently from theprevious device with PEDOT:PSS that was promoting an OFF-likeresponse predominance. This feature is very promising, suggest-ing that the introduction of graphene could be useful to elicit speci fi c ON or OFF responses, thereby paving the way to furtherimprovements of the device architecture for the design of moreef fi cient retinal prosthetics.
4. Methods
Commercial polyethylene terephthalate (PET) was used as pas-sive substrate for all devices. On top of PET, a conductive layercomposed by PEDOT:PSS or G-CVD (Graphenea) was deposited. Awater dispersion of PEDOT:PSS (Clevios PH1000; Heraeus) wasprepared by adding the following additives: the cosolvent dime-thylsulfoxide (9% in volume, purchased from Sigma-Aldrich) toincrease the overall electrical conductivity; the crosslinker 3-glycidoxypropyltrimethoxysilane (0.9% in volume; Sigma-Aldrich)to enhance the adhesion of the PEDOT:PSS layer to the substrateand avoid delamination; the surfactant Zonyl FS-300 (0.18% involume, Sigma-Aldrich) to promote dispersion wettability.PEDOT:PSS dispersion was then sonicated in an ultrasonic bath for20 min, cooled at room temperature and deposited by spin-coatingin two identical steps (rotation speed 2000 rpm, duration 60 s).After deposition, the substrates underwent a thermal annealingprocess in air (120 (cid:3)
C, 10 min). Monolayers of graphene were syn-thetized by Chemical Vapour Deposition on a silicon substrate andthen transferred to PET (Graphenea) for the preparation of G-baseddevices. A chlorobenzene solution of P3HT with a regio-regularityof 99.5% (15 000 e
45 000 molecular weight, Sigma-Aldrich, 30 g/L) was stirred overnight at 70 (cid:3)
C and then deposited on top of theconductive layer by a two-steps spin coating process (800 rpm, 5 s;1600 rpm, 120 s). A thermal annealing (120 (cid:3)
C, 20 min) completedthe fabrication of the devices that were successively cut with asurface of 1.2 cm for cyclic voltammetry, patch-clamp measure-ments or 0.8 cm for surface potential analysis, and approximately0.5 (cid:4) for MEA measurement. All samples dedicated for cellculture preparation were stuck on glass support with Sylgard 184silicon elastomer, and sterilized at 180 (cid:3) C for 2 h.
Cyclic voltammograms were recorded using Patchmaster V2.73(HEKA Elektroniks) with a two-electrode setup with copper-coatedglass slides on which the prototypes were glued and contacted withsilver paste to the P3HT layer. A PTFE tape with a hole of approxi-mately 1 mm was employed to expose to the electrolyte the P3HT,acting as the working electrode. All measurements were referencedto an Ag/AgCl electrode. The measurements were carried in phos-phate buffer solution at a scan rate of 100 mV/s. Analysis wasperformed with FitMaster v2x90.1 software, Prism 6.07 (GraphPad)and OriginPro 9 (OriginLab). Primary cultures of hippocampal neurons were prepared fromembryonic day 18 rat embryos (Charles River). Rats were sacri fi cedby CO inhalation, and embryos removed immediately by cesareansection. Photoactive planar devices ( þ PEDOT:PSS, þ G-CVD, and PETas sham device) were pre-treated with Poly- L -lysine (0.1 mg/ml inborate buffer) prior to cell-plating. Brie fl y, hippocampi weredissociated by a 30-min incubation with 0.25% trypsin at 37 (cid:3) C andcells were plated on polymeric devices in Neurobasal supple-mented with 2 mM L -glutamine, 2% B27, 100 m g/ml penicillin and100 mg/ml streptomycin and with 10% horse serum (Life Technol-ogies) in the fi rst 4 h of plating. Low-density cultures were preparedplating 40 _
000 to 80 _
000 neurons per device, while high-density
M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e317
M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e317 ultures were prepared plating 80 _
000 to 160 _
000 neurons per de-vice. All animal manipulations and procedures were performed inaccordance with the guidelines established by the EuropeanCommunity Council (Directive 2010/63/EU of March 4th, 2014) andwere approved by the Institutional Ethics Committee and by theItalian Ministry of Health.
Photocurrent measurements were performed in voltage-clampmode at room temperature in recording extracellular solutionwith patch pipettes (4 e fi lled with the same solution. Re-sponses were ampli fi ed, digitized at 10 or 20 kHz and stored withPatchmaster V2.73 (HEKA Elektroniks). FitMaster v2x90.1 wereemployed for data analysis, together with Prism 6.07 (GraphPad)and OriginPro 9 (OriginLab). Whole-cell patch-clamp recordings of hippocampal neurons(between 14 and 18 day in vitro , DIV) were performed at roomtemperature using borosilicate patch pipettes (3.5 e U ) andunder G U patch seal. The extracellular solution contained (in mM):135 NaCl, 5.4 KCl, 1 MgCl , 1.8 CaCl , 5 HEPES, 10 glucose adjusted topH 7.4 with NaOH. The intracellular solution contained (in mM):126 K-Gluconate, 4 NaCl, 1 MgSO , 0.02 CaCl , 0.1 EGTA, 10 Glucose,5 Hepes, 3 ATP-Na , and 0.1 GTP-Na. Recordings were carried outusing an EPC10 (HEKA Elektroniks) ampli fi er. Measurement ofmembrane voltage modulation under light stimulation were per-formed in current-clamp con fi guration with i) no current injection(I ¼ (cid:2)
70 mV (I
H -70 mV ), iii) clamping currents ranging from (cid:2) þ
300 pA, and iv) in voltage-clamp con fi guration at voltagesfrom (cid:2)
110 to þ
10 mV. Responses were ampli fi ed, digitized at 10 or20 kHz and stored with Patchmaster V2.73 (HEKA Elektroniks).FitMaster v2x90.1 were employed for data analysis, together withPrism 6.07 (GraphPad) and OriginPro 9 (OriginLab). Photo-stimulation of neurons was carried out on a Nikon FN1 uprightmicroscope (Nikon Instruments) by using a combination of 510 nmand 550 nm wavelengths of Spectra X LED system (Lumencor) tomatch the P3HT absorption spectrum. The light source had a fi nalpower of 1 or 15 mW/mm , with stimulations of 50 or 500 ms induration.The slope of current modulation was extracted per each cellthrough linear fi tting of VC-IV curves, as calculated with OriginPro9. Reversal voltage (E rev ) was calculated per each cell as the inter-cept of the VC-IV linear fi tted curve on the Y-axis at 0 pA. Following 30 min of dark adaptation, animals were euthanizedby means of high-dose carbon dioxide (CO ) inhalation. All theprocedures were carried out in dim red light since rats are known tolack photoreceptors responding to this wavelength. Both eyeballswere surgically enucleated and quickly placed in oxygenated (95%O , 5% CO ) AMES Medium solution (Sigma-Aldrich), minimizingthe time in which retinal tissue was left deoxygenated. Dissectionof the eyeball was performed fi rst on a dry surface and then inliquid. The procedure is performed as follows: i) a fi rst hole throughthe eyeball at the level of the ora serrata is obtained by means of thetip of a scalpel blade, ii) dissecting scissors are passed through thehole and used to cut and dissect the cornea from the sclera, choroidand retina, iii) lens and vitreous body are removed, iv) residualtissues are cut in half in order to facilitate the following steps andthen moved into a separate AMES solution Petri dish in order to continue the dissection under the microscope. In liquid, the twohemiretinas are gently dissected from the sclera by means of dis-secting forceps and from the choroid, avoiding pointing or grasping,in order to minimize the stress on the tissue. Because the manip-ulation may initially alter electrophysiological properties of thetissue, recordings are initiated after about 10 min of rest inoxygenated AMES solution. Each hemiretina is divided into smallerpieces prior to the settling onto the MEAs. Recording of extracellular activity and action potentials wascarried out on 60-electrodes planar MEA devices. Electrodes of30 m m in diameter are disposed in a 8x8 matrix, with an inter-electrodes distance of 200 m m; one of the electrodes works as aninternal reference (iR) (MEA 200/30iR-ITO-gr, Multi Channel Sys-tems GmbH). The devices were also provided with a 1.5 ml capacityglass ring/culture chamber that allowed perfusing the oxygenatedsolution. Recording electrodes are made of nanostructured tita-nium nitride (TiN) featuring impedance less than 100 k U (speci fi cfor 30 m m Ø electrode). Traces and contacts are made of transparentIndium tin oxide (ITO), allowing for a clear vision of the sampleunder the microscope. Electrical passivation is provided by siliconnitride. Pieces of retina to be tested were cut out from the previ-ously isolated hemiretinas in AMES ’ Medium, with arbitrary di-mensions matching as closely as possible the electrodes surface onthe MEA (~1 mm ).Experimental samples were placed onto the MEA with the RGCslayer facing the electrodes surface (face down on the plate). Thedifferent devices were place directly on top the tissue, with theP3HT layer facing it. Eventually, a plastic net and a metal anchorwere further placed on top of the tissue þ device to favour a moreuniform adhesion. The MEA chamber was fi nally fi lled with AMESMedium and transferred to the recording set-up. To keep tissueoxygenated, the MEA chamber was continuously perfused by aperistaltic pump (ISMATEC) with oxygenated (95% O
5% CO )AMES Medium. Electrical activity was recorded using theMEA1060-INV BC by Multichannel Systems (MCS GmbH, Reutlin-gen, Germany), with a total ampli fi cation of 1200 (two stageampli fi cation). MC Rack software (MCS GmbH) was used for therecording, detection and sorting of the data. In order to distinguishaction potentials from background noise (~15 m V, peak-to-peak), wearbitrarily chose a threshold of detection equal to 4.5 times the SDof the signal (automatically calculated by the acquisition system foreach electrode). The fi ring activity of retinal ganglion cells has beenrecorded for 4 s across each light pulse and the intervals of interestwere the 500 ms during illumination and the following 500 msstarting at light offset. After each light pulse, the fi ring increasedthanks to P3HT-based prototypes, and returned back to a baselinerate in less than a second. The illumination pulse was set at 0.25 Hzand lasted around 1.5 min (25 sweeps). An integrated deviceallowed also to keep the temperature of the perfused AMES Me-dium and of the MEA plate at a desired value of 37 (cid:3)
C. The MCSsystem was optically coupled with light stimulation by means ofthe inverted microscope Nikon Eclipse Ti, using a VSD IR fi lter and a20x objective, producing an illumination spot on the sample of0.93 mm . The microscope was coupled with a CCD HamamatsuOrca D2 camera for image capturing. The entire setup was placedon a vibration isolation table and shielded from electromagneticradiation by a Faraday cage. Light stimulation was provided by theLumencor Spectra X Light engine, composed of 6 independentsolid-state LED that can be individually activated. The excitationbeam is coupled to one of the microscope ’ s ports. Light intensitywas measured using a power meter ThorLabs PM100D and con-verted into power density given the area of the illumination spot. M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e C-Rack recordings and Spectra X stimulations were synchronizedby the Stimulus Generator STG4008 (MCS GmbH). By means of theMC-Stimulus II software it is possible to drive TTL signals (Tran-sistor-Transistor Logic) that synchronize MEA recordings with anexternal device, in this case the Spectra X light source.
Declaration of competing interest
All co-authors have agreed to the submission of the fi nalmanuscript and have no fi nancial con fl ict of interest that might beconstrued to in fl uence the results or interpretation of theirmanuscript. CRediT authorship contribution statementMattia L. DiFrancesco:
Formal analysis, Writing - original draft.
Elisabetta Colombo:
Formal analysis, Writing - original draft.
Ermanno D. Papaleo:
Formal analysis.
Jos (cid:1) e Fernando Maya-Vetencourt:
Validation.
Giovanni Manfredi:
Formal analysis.
Guglielmo Lanzani:
Conceptualization, Writing - review & editing. Fabio Benfenati:
Conceptualization, Writing - review & editing,Funding acquisition. Acknowledgments
We thank D. Mattia Bramini (Istituto Italiano di Tecnologia,Genova, Italy) for help and advice in the neuron viability experi-ments and R. Ciancio, I. Dall ’ Orto, A. Mehilli, R. Navone and D.Moruzzo (Istituto Italiano di Tecnologia, Genova, Italy) for technicalassistance. This project has received funding from the EuropeanUnion ’ s Horizon 2020 research and innovation programme undergrant agreements No. 696656 (Graphene Flaghship - Core 1) andNo. 785219 (Graphene Flaghship - Core 2). Appendix A. Supplementary data
Supplementary data to this article can be found online athttps://doi.org/10.1016/j.carbon.2020.02.043.
References [1] I.V. G, A.A.F.K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang,S.V. Dubonos, Electric Field Effect in Atomically Thin Carbon Films, 306, 2004,pp. 666 e (cid:1) alez-Domínguez, R. Rauti, L. Ballerini, M. Prato,E. V (cid:1) azquez, Graphene improves the biocompatibility of polyacrylamidehydrogels: 3D polymeric scaffolds for neuronal growth, Sci. Rep. (2017),https://doi.org/10.1038/s41598-017-11359-x.[3] S. Mateti, C.S. Wong, Z. Liu, W. Yang, Y. Li, L.H. Li, Y. Chen, Biocompatibility ofboron nitride nanosheets, Nano Res (2018), https://doi.org/10.1007/s12274-017-1635-y.[4] P. Yang, X. Wu, Z. Xie, P. Wang, C. Liu, Q. Huang, Durability improving andcorrosion-resistance mechanism of graphene oxide modi fi ed ultra-thin car-bon paper used in PEM fuel cell, Corrosion Sci. (2018), https://doi.org/10.1016/j.corsci.2017.10.025.[5] S. Patra, R. Choudhary, R. Madhuri, P.K. Sharma, Graphene-based portable, fl exible, and wearable sensing platforms: an emerging trend for Health careand biomedical surveillance, in: Graphene Bioelectron, 2018, https://doi.org/ 10.1016/B978-0-12-813349-1.00013-5.[6] E.O. Polat, G. Mercier, I. Nikitskiy, E. Puma, T. Galan, S. Gupta, M. Montagut,J.J. Piqueras, M. Bouwens, T. Durduran, G. Konstantatos, S. Goossens,F. Koppens, Flexible graphene photodetectors for wearable fi tness monitoring,Sci. Adv. (2019), https://doi.org/10.1126/sciadv.aaw7846.[7] E. Masvidal-Codina, X. Illa, M. Dasilva, A.B. Calia, T. Dragojevi (cid:1) c, E.E. Vidal-Rosas, E. Prats-Alfonso, J. Martínez-Aguilar, J.M. De la Cruz, R. Garcia-Corta-della, P. Godignon, G. Rius, A. Camassa, E. Del Corro, J. Bousquet, C. H (cid:1) ebert,T. Durduran, R. Villa, M.V. Sanchez-Vives, J.A. Garrido, A. Guimer (cid:3) a-Brunet,High-resolution mapping of infraslow cortical brain activity enabled by gra-phene microtransistors, Nat. Mater. (2019), https://doi.org/10.1038/s41563-018-0249-4.[8] K. Kostarelos, M. Vincent, C. Hebert, J.A. Garrido, Graphene in the design andengineering of next-generation neural interfaces, Adv. Mater. (2017), https://doi.org/10.1002/adma.201700909.[9] H. Tang, C.M. Hessel, J. Wang, N. Yang, R. Yu, H. Zhao, D. Wang, Two-dimen-sional carbon leading to new photoconversion processes, Chem. Soc. Rev. 43(2014) 4281 e fi cient stableelectron buffer layer for high-performance solar cells, Carbon N. Y. (2018),https://doi.org/10.1016/j.carbon.2018.01.090.[13] N. Martino, P. Feyen, M. Porro, C. Bossio, E. Zucchetti, D. Ghezzi, F. Benfenati,G. Lanzani, M.R. Antognazza, Photothermal cellular stimulation in functionalbio-polymer interfaces, Sci. Rep. 5 (2015), 8911, https://doi.org/10.1038/srep08911.[14] D. Ghezzi, M.R. Antognazza, M. Dal Maschio, E. Lanzarini, F. Benfenati,G. Lanzani, A hybrid bioorganic interface for neuronal photoactivation, Nat.Commun. 2 (2011) 166, https://doi.org/10.1038/ncomms1164.[15] P. Feyen, E. Colombo, D. Endeman, M. Nova, L. Laudato, N. Martino,M.R. Antognazza, G. Lanzani, F. Benfenati, D. Ghezzi, Light-evoked hyperpo-larization and silencing of neurons by conjugated polymers, Sci. Rep. 6 (2016),22718, https://doi.org/10.1038/srep22718.[16] J.F. Maya-Vetencourt, D. Ghezzi, M.R. Antognazza, E. Colombo, M. Mete,P. Feyen, A. Desii, A. Buschiazzo, M. Di Paolo, S. Di Marco, F. Ticconi,L. Emionite, D. Shmal, C. Marini, I. Donelli, G. Freddi, R. MacCarone, S. Bisti,G. Sambuceti, G. Pertile, G. Lanzani, F. Benfenati, A fully organic retinalprosthesis restores vision in a rat model of degenerative blindness, Nat. Mater.(2017), https://doi.org/10.1038/nmat4874.[17] A. Gal, Y. Li, D.A. Thompson, J. Weir, U. Orth, S.G. Jacobson, E. Apfelstedt-Sylla,D. Vollrath, Mutations in MERTK, the human orthologue of the RCS rat retinaldystrophy gene, cause retinitis pigmentosa, Nat. Genet. 26 (2000) 270 e fl uence of eye pigmentation and light deprivationon inherited retinal dystrophy in the rat, Exp. Eye Res. 21 (1975) 167 e fi cant improvement in the photovoltaic stability of bulk heterojunctionorganic solar cells by the molecular level interaction of graphene oxide with aPEDOT: PSS composite hole transport layer. https://doi.org/10.1016/j.solener.2018.03.083, 2018. M.L. DiFrancesco et al. / Carbon 162 (2020) 308 e317