Visible light biophotosensors using biliverdin from Antheraea yamamai

We report an endogenous photoelectric biomolecule and demonstrate that such a biomolecule can be used to detect visible light. We identify the green pigment abundantly present in natural silk cocoons of Antheraea yamamai (Japanese oak silkmoth) as biliverdin, using mass spectroscopy and optical spectroscopy. Biliverdin extracted from the green silk cocoons generates photocurrent upon light illumination with distinct colors. We further characterize the basic performance, responsiveness, and stability of the biliverdin-based biophotosensors at a photovoltaic device level using blue, green, orange, and red light illumination. Biliverdin could potentially serve as an optoelectric biomolecule toward the development of next-generation implantable photosensors and artificial photoreceptors. © 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

One of the sources to obtain a decent amount of biliverdin is silkworms and silk cocoons, in particular Antheraea yamamai (Japanese oak silkmoth). Specifically, Antheraea yamamai has been cultivated for centuries as a native in East Asia [20,21]. Mimicking the natural sunlight exposure, when light with high intensity (5000 lux) is irradiated onto Japanese oak silkmoths, the silk cocoon is colorized to be green due to the photochemical transformation to biliverdin [8,9]. In other words, biliverdin is also present in the silk cocoons of Antheraea yamamai [8,9]. As biliverdin absorbs the visible light in the broad wavelength range of mainly blue and red [22], the green color of the silk cocoon may provide a camouflage effect in nature. Given that silkworms have been considered as bioreactors for producing recombinant proteins, the silk cocoons of Antheraea yamamai can be a source for massproducing biliverdin. It should also be noted that silk cocoons and processed silk polymers are actively used for biomedical and optical applications, owing to the biocompatibility/biodegradation [23][24][25] and the superior mechanical and optical properties [26][27][28].
Although there are recent advances in implantable photosensors, these still rely on biocompatible, yet exogenous, materials (e.g., MoS 2 , graphene, Al 2 O 3 , Si 3 N 4 , and polyimide) [29]. In our study, we investigate whether there is any endogenous biomaterial that can be used for biological photosensors. Interestingly, animals, including of Antheraea yamamai and humans, can endogenously generate and metabolize biliverdin [4,30]. This means that biliverdin-based photosensors can be more biologically friendly to the human body and be useful for potentially developing next-generation implantable photosensors and artificial photoreceptors.
In this study, we introduce the photoelectric conversion property of biliverdin at a device level as a biophotosensor for detecting visible light. First, after harvesting Antheraea yamamai silk cocoons, we describe a processing method to extract biliverdin from the silk cocoons. Second, we fabricate prototype devices of the biophotosensors. Third, we conduct comprehensive physical and chemical analyses, including mass spectrometry, optical measurements, and photoluminescence (PL) measurements. Finally, we demonstrate the basic current-voltage characteristics and stability of the biophotosensors under visible light illumination with distinct colors. We envision that the reported photovoltaic characteristics of biliverdin could potentially be used for biological and medical applications.

Production, materials, and chemicals
We harvested silk cocoons produced by Antheraea yamamai in the experimental farm of the National Institute of Agricultural Sciences in the Republic of Korea. Antheraea yamamai silkworms were reared in an outdoor-breeding farm planted with oak trees, as shown in Fig.  1(a). To protect the hatched larvae from damage caused by birds and rats, the farm was covered with a net. The eggs of Antheraea yamamai were released to oak trees. After completion of cocoon spinning, the green cocoons were collected and then the silkworm pupae were removed. The silk cocoon of Antheraea yamamai, which was reared in the outdoor sunlight environment, exhibits a green color due to its strong reflectance in the wavelength range of 490 -590 nm, as shown in Fig. 1(b). The rearing time for a growth period of Antheraea yamamai is summarized in Table 1.

Extraction of biliverdin from Antheraea yamamai silk cocoons
To facilitate the extraction of biliverdin, we first removed sericin in the silk cocoons using a chemical and physical degumming method via a low-temperature process for minimizing the heat-induced denaturation of biliverdin. The silk cocoons were soaked into a degumming solution (Na 2 CO 3 , 2.5 g 1 -l ) with stirring of 400 rpm at 60 °C for three hours and then were washed with warmed deionized water. The degumming process was repeated three times. The resultant degummed silk cocoons were dried at room temperature under a dark condition. We developed a simple protocol to extract biliverdin from the degummed silk cocoons using dimethyl sulfoxide (DMSO) as a solvent (Fig. 2). DMSO has relatively low toxicity and is an environmentally benign organic solvent [31]. After the degumming process, the silk cocoons were cut into small pieces with sizes of < 10 mm and then were immersed in the DMSO solvent with a concentration of 0.1 g m1 -1 with stirring of 400 rpm at 45 °C for 24 hours. Then, the DMSO solvent effectively released biliverdin from the silk cocoons. After removing the cocoon pieces, the green residual solution was centrifuged with 9000 rpm for 20 minutes at room temperature. The final solution was stored at room temperature in dark before use.

Device fabrication of biliverdin-based biophotosensor prototypes
As the first testing device using a biliverdin solution, we constructed working prototypes of the biliverdin-based biophotosensors. Specifically, we mimicked the simple device structure of typical dye-sensitized solar cells, as commonly used elsewhere [32,33]. First, a photoanode electrode was prepared on the conducting side of FTO glass substrate by coating a commercial transparent TiO 2 paste. The uniform coating thickness of 10 μm was achieved using a doctor-blade (tape casting) method. Before coating, the substrate was cleaned in an ultrasonic bath using acetone, methanol, and deionized water, followed by blowing with nitrogen gas. To remove organic components in the coated TiO 2 paste, the substrate was subsequently annealed at 120 °C for 15 minutes, at 350 °C for 15 minutes, and at 475 °C for 30 minutes on a hot plate, followed by naturally cooling down to 100 °C. To ensure the physical contact among TiO 2 nanoparticles, the TiO 2 photoanode was additionally immersed into an aqueous solution of 15 mM TiCl 4 at 70 °C for 30 minutes in an oven. Second, a biliverdin solution (10 μl) extracted from the silk cocoons was dropped on the TiO 2 -coated photoanode. The volume of the biliverdin solution was controlled by a punched polydimethylsiloxane (PDMS) template with a diameter of 6 mm, which prevented the solution from diffusion. The sample was dried at 40 °C for above six hours in an oven. Third, an anode electrode was assembled with a platinum (Pt)-coated FTO glass substrate as a counter electrode in a sandwich geometry by a hot-melt Parafilm. For good electrical contact, the external electrodes were coated with a silver paste. Finally, an electrolyte solution was injected into a space created by the Parafilm through pre-drilled channel holes on the counter electrode, followed by sealing with the Parafilm and a thin glass cover. For devices without using an electrolyte, the original biliverdin solution was also directly tested by injecting it into the assembled structure.

Mass spectrometry analysis
We obtained positive ion spectra using a mass spectrometer (6550 iFunnel Q-TOF, Agilent Technologies, Santa Clara, CA, USA). A biliverdin solution (3 μl) extracted from the silk cocoons was injected into a Zorbax Extend C 18 column. The two components of the mobile phase (A and B) were used such that A was 0.1% formic acid in water and B was methanol. The gradient was 10% B for one minute, 55% B for two minutes, and 90% B for four minutes, respectively. The flow was 0.2 ml min −1 . Selected reaction monitoring (SRM) spectra were obtained with 583[M + H] + transitioning to m/z 297.

Optical characterizations
We characterized the optical properties including reflectance, transmittance, absorption, and PL spectra/images using a fiber bundle-coupled spectrometer with an integrating sphere and a customized mesoscopic spectrometer setup, as reported in our previous studies [24,34]. In particular, PL analysis was highly useful to better understand the charge carrier trapping, migration, and recombination processes between photosensitizing molecules and metal oxide materials, because PL emission originates from the recombination of free carriers in the molecules [35][36][37]. A customized template for containing the biliverdin solution was prepared as shown in Fig. 3(a). A TiO 2 film with a diameter of 4 mm was coated on the cleaned glass microscope slide using the doctor-blade method and was subsequently cured at different temperatures of 120 °C (15 minutes), 350 °C (15 minutes), and 475 °C (30 minutes), respectively. A 1 mm-thick PDMS template was made from Sylgard 184 silicone elastomers kit with a ratio of 10:1 (base:agent). After curing, the PDMS template was punched with a diameter of 10 mm and then was laminated on the TiO 2 -coated glass. A biliverdin solution (50 μl) in DMSO was dropped on the TiO 2 -coated glass area in the PDMS template. To detect PL emission of biliverdin, the fabricated template was placed within the field of view of the mesoscopic imaging setup, in which an excitation source at λ ex = 470 nm and an emission filter with λ em = 600 -800 nm were used, as shown in Fig. 3(b).

Current density-voltage response analyses
We measured the current density-voltage (J-V) characteristics of the biophotosensors using a source meter (2400, Keithley Instruments Inc., Cleveland, OH, USA) controlled by a customized LabView interface. For time-resolved photocurrent measurements, a low-noise current preamplifier (SR570, Stanford Research Systems) and a real-time digital oscilloscope (DPO7104, 10 GS/s, 1 GHz, Tektronix) were employed. As visible light sources, blue, green, orange, and red LEDs (Thorlabs Inc., Newton, NJ, USA) were used with different center wavelengths of 470 nm (FWHM = 25 nm), 530 nm (FWHM = 33 nm), 565 nm (FWHM = 104 nm), and 625 nm (FWHM = 18 nm), respectively. The LED power was tuned by controlling the LED drive current (LEDD1B, Thorlabs Inc.) and was calibrated using a power meter (PM100D, Thorlabs Inc.). Importantly, a black mask with a diameter of 6 mm was utilized to ensure accurate measurements of the incident light. All of the measurements were carried out in a dark and shielded enclosure at room temperature.

Results and discussion
First, we confirmed that biliverdin is the dominant molecule in the green solution extracted from the silk cocoons of Antheraea yamamai, using the mass spectral analysis. As shown in the mass spectrum ( Fig. 4(a)), the main peak at m/z 583.2877 corresponds to the molecular ion of biliverdin (exact mass 583.2557 Da, protonated form) [38,39]. The light absorption spectrum also shows the green color; the strong light absorption is observed at around blue (maximum λ abs = 445 nm) and red (maximum λ abs = 603 nm) wavelength ranges, while its weak absorption is shown in the green wavelength (minimum λ abs = 515 nm) in Fig. 4(b). Importantly, biliverdin is a fluorescent biomolecule; it emits red fluorescent light with a maximum emission wavelength (λ em ) of 674 nm when excited at 450 nm, as shown in Fig.  4(c). Indeed, all of the optical properties of the extracted green solution are in good agreement with the previous studies about biliverdin [22,40]. Second, we studied photoinduced electron transfer between biliverdin (donor) to TiO 2 (acceptor) on the basis of the PL emission of biliverdin. Figure 5 depicts the PL image of biliverdin on the TiO 2 -coated glass surrounded by the PDMS template and the cross-sectional spatial profile along A to B (white dashed line). The PL intensity of biliverdin on the TiO 2coated area is lower than that of the glass area without TiO 2 . The percentage decrease in the PL intensity is approximately 15%. This reduction was attributed to the quenching of biliverdin molecules, which were adsorbed to TiO 2 nanoparticles (i.e. physical contact between biliv transfer from a photoelectri Fig vice performan e as shown in F ectly injected i e-drilled chann and 15 mW cm ro bias was obt the absence of s attributed to Ca), sodium (N natural silk coc k fibers to be us, we checke tribute to the p coons of Bomb s of the biliverd injected into t tion (λ = 470 ds, other chem t generation un he photoinduc yte. Figure 6  signals was evaluated as high as ~2.53 × 10 3 (i.e. J dark = 0.003 μA cm −2 and J light = 7.579 μA cm −2 ). The photoresponsivity R = J photo /P light was also calculated to be ~0.505 mA W −1 , where the photocurrent density J photo is determined by J light -J dark and P light is the incident light intensity (i.e. 15 mW cm −2 ). As another key property for photosensors, the relationship between the photogenerated current and voltage and the intensity of the incident light was further studied under the blue light illumination (λ = 470 nm) in Fig. 6(c). J sc increases linearly from 2.217 to 23.868 μA cm −2 as a function of the light intensity in a variation range from 5 to 45 mW cm −2 . On the other hand, the increase in the open-circuit voltage (V oc ) follows nonlinear dependence on the light intensity due to the proportional relationship of V oc and ln(J light /J dark ). Next, we tested the responsiveness of the biliverdin-based biophotosensors. Specifically, we quantified switching behavior of J sc and V oc in the light-on and -off states upon visible light illumination in different central wavelengths of 470 nm, 530 nm, 565 nm, and 625 nm at 15 mW cm −2 , as shown in Fig. 7(a) and 7(b). Both J sc and V oc were reproducibly switched from the 'high' state to the 'low' state by periodically turning each light source on and off. As expected, the dependence of J sc and V oc on the incident wavelength reflected the absorption spectrum of biliverdin in Fig. 4(b). The light on/off ratios of the biophotosensors for J sc were estimated to be ~310, ~595, and ~39 at 530 nm, 565 nm, and 625 nm, respectively.
We also characterized the response speed of the biophotosensors by calculating the rise time R t and the decay time D t . R t was defined as a time interval to rise from 10% to 90% of its peak value, while D t was defined as a time interval to decay from 90% to 10% of its peak value. As shown in the current and voltage responses in Fig. 7(c) and 7(d), R t and D t values of the biophotosensors are ~0.09 and ~0.08 second for J sc and ~0.305 and ~1.134 second for V oc , respectively. For J sc , the biliverdin biophotosensors show the relatively rapid photoresponse characteristics. On the contrary, for V oc , the biophotosensors exhibit the slow photoresponse with much longer R t and D t . This behavior can be understood by a slow recombination rate of photogenerated electrons by electrolyte ions on the TiO 2 /electrolyte interface, because there was no current passing through the external circuit under the V oc condition [44]. Finally, given that all fluorescent molecules are subject to photobleaching [45,46], we determined the stability of the biophotosensors. We characterized real-time photocurrent (J sc ) transient responses by periodically switching an optical signal on and off at 0.5 Hz under the blue light illumination (λ = 470 nm, 15 mW cm −2 ). Figure 8(a) shows that the biophotosensor is switched between the light on-and off-states with relatively good reproducibility and stability. After photoswitching cycles of 2000, the photocurrent was reached to be 6.828 μA cm −2 , but the percentage decrease was only 10% compared to its initial value (i.e. J sc = 7.579 μA cm −2 ). In addition, to check the practicality of the biliverdin-based biophotosensors, we determined the long-term stability for an extended period of 1440 hours (60 days) in Fig.  8(b). The biophotosensors were stored in the lab under the ambient conditions (22 ± 2 °C and 40 -50% relative humidity). After 1440 hours, the photocurrent was highly retained with a percent decrease of ~8% (J sc = 6.981 μA cm −2 ). The decrease in J sc is attributable mainly to the photobleaching of biliverdin by light excitation. In addition, other deteriorating factors at the device level over time can include reduction in the triiodide (I 3 − ) concentration in the acetonitrile-based iodide/triiodide (I − /I 3 − ) electrolyte and degradation of the anode and counter electrodes [45][46][47]. However, this result supports that idea that biliverdin can be stable even when contacted with TiO 2 and the electrolyte.

Conclusion
We have demonstrated the photovoltaics of biliverdin extracted from the silk cocoons of Antheraea yamamai in the photosensor prototypes. Although biliverdin has been used in bioimaging, the photoelectric conversion of biliverdin has not yet been used at a device level to the best of our knowledge. Mass spectroscopy and optical spectroscopy confirm that the silk cocoons of Antheraea yamamai contains biliverdin. Fluorescent quenching of biliverdin in the presence of TiO 2 in the PL analysis supports possible photoinduced electron transfer. In the self-power mode (i.e. zero bias), the biliverdin-based biophotosensors are successfully operated under visible light illumination at 470 nm, 530 nm, 565 nm, and 625 nm. The reliable responsiveness is tested from repeated on/off switching. In addition, this environmental-friendly production and biocompatible nature of biliverdin could potentially provide with a simple and cost-effective manufacturing strategy.