High-brightness organic light-emitting diodes for optogenetic control of Drosophila locomotor behaviour

Organic light emitting diodes (OLEDs) are in widespread use in today’s mobile phones and are likely to drive the next generation of large area displays and solid-state lighting. Here we show steps towards their utility as a platform technology for biophotonics, by demonstrating devices capable of optically controlling behaviour in live animals. Using devices with a pin OLED architecture, sufficient illumination intensity (0.3 mW.mm−2) to activate channelrhodopsins (ChRs) in vivo was reliably achieved at low operating voltages (5 V). In Drosophila melanogaster third instar larvae expressing ChR2(H134R) in motor neurons, we found that pulsed illumination from blue and green OLEDs triggered robust and reversible contractions in animals. This response was temporally coupled to the timing of OLED illumination. With blue OLED illumination, the initial rate and overall size of the behavioural response was strongest. Green OLEDs achieved roughly 70% of the response observed with blue OLEDs. Orange OLEDs did not produce contractions in larvae, in agreement with the spectral response of ChR2(H134R). The device configuration presented here could be modified to accommodate other small model organisms, cell cultures or tissue slices and the ability of OLEDs to provide patterned illumination and spectral tuning can further broaden their utility in optogenetics experiments.

Scientific RepoRts | 6:31117 | DOI: 10.1038/srep31117 have been considerable efforts to customise light delivery methods for optogenetics, in particular with respect to bio-implantable light sources [14][15][16][17] . There is also strong interest in spatial patterning of optical stimulation to restrict optical excitation to defined cells or to excite multiple sites independently. Examples from in vitro studies include modulation of laser illumination with digital micromirror devices 18 , or the projection of illumination from arrays of microscopic LEDs 19 . Significant technological advances have been motivated by the desire to achieve spatial patterning in vivo 16,20 . These extend from bio-implantable devices based on microscale inorganic LEDs coupled with recording electrodes 21,22 , to holographic techniques for distributing laser illumination for use alongside fluorescent imaging of neuronal activity 23,24 .
Organic light-emitting diodes (OLEDs) possess a number of properties that render them potentially very useful for optogenetics. Like conventional inorganic LEDs, they have μ s or sub-μ s response times, enabling frequency-modulated optical stimulation to be delivered with sub-microsecond temporal resolution. However, compared to their inorganic counterparts, they can be more readily fabricated on thin, flexible or even elastic plastic films [25][26][27] which is of potential benefit for implantable optical-neural interfaces. OLEDs can also be highly efficient, even when emitting in the blue part of the visible spectrum [28][29][30] and thus generate little heat during high brightness operation, ameliorating tissue heating that can be a concern with existing light delivery systems. OLEDs can also offer considerable versatility in their individual spectral tuning, which is determined by the choice of luminophore 31,32 and can be tuned further by introducing optical micro-cavities 33,34 . Furthermore, emitters with different colours can be layered or multiplexed to form white OLEDs, giving complete coverage of the visible spectrum from a single device 35 .
Previous reports proposed the possibility of using OLEDs for optogentics 36 and we have very recently demonstrated switching of ChRs in non-neuronal systems using OLED micro-arrays 37,38 . However, to the best of our knowledge, OLEDs have not yet been used to modulate the activity of neuronal networks. A main constraint before now has been that their maximum optical power densities fell below those required for efficient ChR activation; most optogenetics experiments use illumination power densities in the range of 0.1-10 mW.mm −2 to elicit action potential firing optically. Here, we report a working implementation of OLEDs providing sufficient illumination intensity to achieve optical activation of ChRs in vivo for optogenetics. Due to the excellent electrical performance of the pin OLEDs used here, driving voltages of 5 V were sufficient to evoke robust optogenetic responses, making these light sources highly attractive for integration with standard electronics.

Results and Discussion
In this study we used a transgenic line of Drosophila (OK371-GAL4/UAS-H134R-ChR2) in which larvae in their third instar stage of development (about 4-7 days after fertilisation) express ChR2(H134R) in motor neurons 39 . In these animals, upon ChR2(H134R) activation with blue light, robust firing of action potentials is evoked in motor neurons, which generates muscular contractions that temporarily immobilise larvae 39 . With this behavioural response as our readout, we used the Drosophila larvae to test if OLEDs provide sufficient luminance intensities for optical activation of neuronal circuits expressing ChRs and to identify OLEDs with suitable spectral profiles.
In our experimental setup (Fig. 1a), an individual ChR2(H134R)-expressing larva was constrained in a silicone chamber containing sucrose solution and located above an OLED device. The animals were imaged with long-pass filtered light (> 600 nm) to assess larval behaviour without background ChR activation. The OLEDs used here employ a pin structure where the active emissive layer (i) is sandwiched between doped hole (p) and electron (n) transport layers to reduce charge carrier injection barriers and voltage drop across the charge transport layers and thus achieve high brightness at low operation voltage 40,41 . For the blue-emitting OLEDs a fluorescent host-guest emitter system was employed, whereas phosphorescent emitters were used for the green-and the orange-emitting devices (see Methods for details of OLED fabrication and Fig. 1a for a schematic of the OLED stack).
The spectral profile of emission from all three OLEDs is shown in Fig. 1b. Our blue OLEDs peaked at 464 nm, matching well with the activation spectrum for ChR2(H134R), which is reported to peak around 450 nm 12,42 (Fig. 1b, lower panel). Emission from green OLEDs (maximum: 515 nm) also overlapped the ChR2(H134R) action spectrum substantially (responses to 515 nm excitation are approximately 50% of the maximum 12 ). Peak emission of the orange OLEDs was recorded at 606 nm and these devices showed negligible emission below 550 nm so were not expected to activate ChR2(H134R). Compared to conventional LED light sources, the OLED emission has a larger spectral bandwidth. However, as the activation spectra of ChR2(H134R) and most other ChR2 are even broader than the spectra provided by our OLEDs, we do not expect that this leads to a reduction in the efficiency of stimulation. If required, the bandwidth of the OLED emission can be reduced by employing optical microcavities 33,34 or emitter systems with intrinsic narrow band emission 43 . Figure 1c shows the voltage-power density characteristics for our blue, green and orange emitting OLEDs. For the following experiments, all OLEDs were driven with 5 V square wave pulses and at this voltage the different OLEDs provided comparable power densities (ranging between 0.25 and 0.4 mW.mm −2 , Fig. 1c). We expect that a further substantial increase in power density can be achieved by further optimisation of device architecture, e.g. by reducing the voltage drop along the resistive indium tin oxide (ITO) connection lead and by fine-tuning of doping levels in the charge transport layers.
In resting conditions, larvae predominantly remained outstretched along their anterior-posterior axis, while exhibiting some feeding behaviour. Rapidly, upon blue OLED illumination, we observed a robust behavioural response in the form of a strong muscular contraction throughout the body plan of the larva, consistent with previous descriptions of responses to ChR2(H134R) activation in this model organism 39 . This manifested as a cessation of feeding behaviour, plus a rapid contraction that drew the head and tail of the larva closer together, followed by relaxation to the original elongated state after switching off the OLED ( Fig. 2a; see also Supplementary Movie S1).
Scientific RepoRts | 6:31117 | DOI: 10.1038/srep31117 To measure the magnitude and kinetics of this behavioural response, we tracked head and tail positions of larvae exposed to alternating 5 s periods of OLED illumination and darkness. From these coordinates we then calculated a head-tail distance (Fig. 2b). During an imaging period of 60 s, sequential 5 s pulses of illumination from a blue OLED reliably triggered contractions in ChR2(H134R) larvae, appearing as shortenings in the measured head-tail distance (Fig. 3a). We investigated the temporal correlation between behaviour and timing of OLED illumination by performing a Fourier transform on the head-tail distance measurements. A dominant peak in amplitude in the transformed data was observed at 0.1 Hz (Fig. 3b), matching the frequency with which light pulses were delivered from the OLED (5 s illumination, followed by 5 s darkness). Thus, robust and repeatable optical activation of ChR2(H134R) in vivo was achieved with these blue OLEDs.
To examine the spectral selectivity of this behaviour and control for possible confounding factors such as changes in temperature or overall brightness, we replicated the experiment with green-emitting and orange-emitting OLEDs. Notably, when substituting with green-emitting OLEDs, we also observed a significant response of ChR2(H134R) larvae that was strongly temporally coupled to the OLED signal (Fig. 3c,d). In contrast, with an orange OLED, no discernible behavioural response could be attributed to optical stimulation (Fig. 3e,f).
To compare contractile responses across larvae and between experiments performed with differently-coloured OLEDs, head-tail distances for each 5 s period of OLED illumination were plotted after normalising to the initial length of each larva immediately before illumination (Fig. 4a). With blue and green OLEDs, head-tail movements started rapidly upon onset of illumination, detectable in the first frame of illumination (frame duration: 100 ms). We determined the initial rate of contraction during the first 0.5 s of illumination from the gradient of a linear fit to this initial period and scaled to the average starting length of larvae measured in this study (4.2 mm). We found that the rate of contraction was faster for blue OLEDs (1.3 mm.s −1 ) than green OLEDs (0.9 mm.s −1 ). Head-tail movements then gradually slowed, reaching minima by 2.4 s under green OLEDs and 3.3 s under blue OLEDs. Therefore, the initial rate and overall duration of head-tail movements initiated by green OLEDs were both roughly 70% of those induced by blue OLEDs.
The minima reached in the fractional head-tail distance traces were also compared for the different OLEDs (Fig. 4b). The largest overall behavioural responses in any conditions were observed in response to blue OLED illumination (mean fraction of initial head-tail distance ± standard error: 0.73 ± 0.04). Substantial responses, albeit slightly smaller overall, were also recorded from periods of green OLED illumination (0.83 ± 0.01). As expected, orange OLED illumination did not trigger detectable behavioural responses, with larvae remaining predominantly in their resting outstretched orientation. The minima reached for each OLED colour differed from each other significantly (ANOVA followed by Bonferroni test, α = 0.05).
Finally, for the purposes of comparison with an established light source for optogenetic stimulation, we measured the effectiveness of a blue inorganic LED at triggering behavioural responses in the same assay used to assess the OLEDs (Fig. 4c). Illumination from the blue LED (emission peak: 477 nm; 5 s on/5 s off) was relayed via a condenser lens from beneath the sample (power density on sample: 0.3 mW.mm −2 ) while imaging larvae as in the OLED experiments. The larvae showed behavioural responses much like those elicited by OLEDs, with contractions that were tightly coupled to the onset of optical stimulation from the blue LED and that were sustained for its entire duration (mean fraction of head-tail distance ± standard error: 0.78 ± 0.02). As an additional negative control, we also tested ChR2(H134R)-expressing larvae that had been cultured on growth medium not supplemented with all-trans-retinal (the essential cofactor for ChR activity). As expected, these animals did not exhibit contractions in response to the blue light stimulus (Fig. 4c).
In conclusion, we have demonstrated that OLEDs are capable of robust in vivo ChR activation in Drosophila larvae and developed a simple setup that could be readily adapted for other small model organisms, or for neuronal cultures or brain slices. Such a system will enable rapid prototyping of new devices with modified layouts and emitters, with a view to optimising further OLEDs for optogenetics.
In the future, by using thin-layer encapsulation, OLEDs may enable lens-free devices in which cells or tissue can be brought to within micron proximity of the emissive surface. Physical patterning of emissive pixels at microscopic resolutions can thereby allow single cells, and even subcellular compartments, to be addressed optically while mitigating the effects of divergence of light due to the close proximity between the cells and the light source. This has been recently implemented to control phototactic behaviour of the highly light sensitive single-cell green alga Chlamydomonas reinhardtii 37 and the membrane potential of non-neuronal cells in culture 38 but further improvements of the emission intensity are required to robustly trigger responses in neuronal systems. By combining high brightness OLEDs such as those used in this study, with further improvements in microscopic patterning, driver electronics, and thin film encapsulation (e.g. via combined organic/inorganic laminates), OLEDs may become a new method of choice for cellular-level optogenetic control of neuronal networks. OLEDs are also a candidate technology for bio-implantable devices for light delivery in vivo. A number of interesting properties of OLEDs could be harnessed for this purpose: devices can be (semi)transparent 44,45 , which may aid high-resolution in vivo imaging of neural activity; using stacked device architectures, OLEDs can emit several distinct spectral bands (e.g. blue and yellow light) that can be controlled separately 46 , which could be harnessed to address different optogenetic constructs at the same site; finally, devices can be mechanically flexible and could conceivably be integrated with analogous organic electronic devices for recording the activity of neurons 47 to form highly conformable bi-directional bioimplants.
Sample illumination was from a fibre-coupled mercury light source (Nikon intensilight C-HGFI) attenuated (to < 1 mW.mm −2 ) with neutral density (ND) filters and long pass filtered (> 600 nm) to avoid non-specific activation of ChRs through imaging alone. Reflected light was collected with a 4X/0.13 NA air objective and slightly demagnified with an achromatic doublet lens (f = 50 mm) to collect the whole field of view on the sensor of the attached sCMOS camera (Andor Neo). Additional ND filters were placed in front of the camera, to prevent OLED emission saturating the detector. 60 s image sequences were acquired at 10 s −1 , while externally modulating OLED emission with a pulse generator by applying square wave 5 V pulses of 5 s duration at 0.1 s −1 . The optical power density provided by each OLED was measured by a calibrated power meter (Gentec-EO) with the photodiode (detector area: 10 mm Ø) placed in direct contact with the OLED. Emission spectra were recorded with a grating spectrograph coupled to a CCD detector (Andor).
For experiments with the blue inorganic LED, blue light pulses were projected onto the underside of the samples with a collimating lens. Spot size and driving current were adjusted to obtain an optical power density of 0.3 mW.mm −2 . Responses to blue LED stimulation were compared between ChR2(H134R)-expressing larvae grown on media either supplemented or not supplemented with 1 mM all-trans-retinal.
Analysis. Movies of Drosophila behaviour were imported as image sequences into the FIJI distribution of ImageJ. Using the plugin MtrackJ 48 the position of the head and tail of each larva was manually recorded in each frame, then from these coordinates, a head-tail distance was calculated geometrically. Periods of OLED illumination were clearly apparent in the movies, so OLED illumination timings were reconstructed directly from the image sequences by placing a region of interest over the OLED and extracting an intensity value for each frame. Fourier transformation of head-tail distance traces and all numerical and statistical analyses were performed in Origin.