2.1.

Design and Manufacturing of CerebraLux

2.1.1.

Overview of the system

CerebraLux consists of an optic and an electronic component (Fig. 1). The optic component includes a baseplate that holds the fiber-optic and its ferrule. The electronic component includes an MCU, an IR receiver, and an SMD LED soldered on a double-sided PCB with a removable LiPo battery. Only the optic component is permanently housed on the mouse’s head. The two components are aligned and held together using small magnets. This allows the electronic component to be removed when not in use.

Fig. 1

Overview of the CerebraLux optogenetic probe. This schematic shows the two components of the probe: (1) the electronic components consist of the battery, female header, infrared receiver, microcontroller, PCB, and LED. (2) The optic components consist of a baseplate, ferrule, and fiber-optic. The optic component is the only part to be permanently implanted and weighs only 0.5 g, whereas the electronic component weighs 2.3 g and is attached when experiments are run. The electronic and optic components both have magnets that allow for easy attachment and correct alignment between the fiber-optic and LED.

2.1.2.

Optic component

An overview of the optic component is shown in Fig. 2. This figure shows the baseplate, made of high-density polyethylene (HDPE), containing the fiber-optic at specific $X$- and $Y$-coordinates and two magnets for alignment with the electronic component. The baseplate was designed in Autodesk Fusion 360 https://www.autodesk.com/products/fusion-360/students-teachers-educators, an open-source three-dimensional (3-D) design software.¹⁴ We used this software to generate the CAM toolpaths, and exported the resulting G-code to Otherplan to be milled by an Othermill computer numerical control (CNC) mill (Other Machine Company https://www.bantamtools.com/products/hdpe, Berkeley, California). Step-by-step instructions for the baseplate design, milling, and files for download are available in the CerebraLux manual (Appendix: Secs. A1.1–A1.3) and lab website (Walwyn Lab http://www.walwynlab.semel.ucla.edu/new-page), respectively. We inserted a $500\text{-}\mu \mathrm{m}$ fiber-optic of specified length based on the coordinates for surgery (Doric Lenses, $480/500\text{-}\mu \mathrm{m}$ diameter, NA 0.63, plastic optical fiber) into a ferrule (Doric Lenses, Zirconia Ferrule OD 1.25 mm). Finally, we inserted the ferrule with the fiber-optic into its channel in the baseplate and glued the magnets (Magnet4us Inc., D0201N40) into their respective slots in the baseplate (Appendix: CerebraLux manual Sec. A1.4).

Fig. 2

Overview of the optics component on the CerebraLux probe. This schematic shows greater detail of the optic component. The milled baseplate has slots for the magnets, LED, and fiber-optic. The alignment magnets are glued into their respective slots, and the fiber-optic and ferrule are inserted in the central channel of the baseplate. The LED is on the underneath of the PCB and aligns magnetically into the upper slot in the baseplate. This component is inserted into the correct region of the skull using a milled stereotaxic adapter that also has two aligning magnets (Appendix, CerebraLux manual; Sec. A1.5).

2.1.3.

Electronic component

i. Software. We used an open-source IR remote library developed by Ken Shirriff for the Arduino platform to send and receive signals, which control the LED. A simple graphical user interface (GUI) was designed to control the stimulator. Figure 3 shows the GUI, which receives the input values for on time, off time, and light intensity, and forwards the information to the device. The GUI communicates serially through a universal serial bus (USB) port to an Arduino Uno (ArduinoBoardUno) to send the corresponding IR LED pulses to the device. We designed the GUI in the Python 2.7 TkInter library, the most widely accessed GUI library. Constructing the GUI in Python instead of proprietary software such as LabVIEW and MATLAB ensures free access and programmability of the interface. By providing our interface as a standalone executable file, researchers can use the interface as is or modify it further to fit their experimental needs. The TkInter library is also well supported, with thorough documentation online in forums and tutorials. The program operates by sending three signals for on time, off time, and light intensity, through the infrared 950-nm LED (SparkFun, Boulder, Colorado) in the form of three IR LED flashes. The IR receiver on the stimulator receives these IR pulses, and the microcontroller processes them to assign the corresponding on–off times and light intensity for the stimulating LED. The order of activation is outlined in Fig. 4. The codes for the sending and receiving protocols in the Arduino interactive development environment (IDE) can be found in the Appendix; CerebraLux manual (Secs. A2.2 and A2.3) while the GUI is downloadable (Walwyn Lab http://www.walwynlab.semel.ucla.edu/new-page).

Fig. 3

The Python-based GUI. The LED on the PCB is controlled through this Python-based GUI, which can be run on any computer interface. Once installed and the IR controller connected to the computer, the LED is turned on by clicking “activate LED” and turned off by clicking “STOP.” The ON time, OFF time, and intensity can be altered to implement pulse width modulation and light intensity. The GUI also calculates and outputs the period and frequency of the on–off times entered into the GUI.

Fig. 4

The flowchart demonstrating CerebraLux activation. The computer GUI is connected to an Arduino Uno-controlled IR LED. After pressing “activate LED” on the GUI, the transmitter sends IR pulses to the head-mounted module, where it is received by the photodiode and sent to the ATMega328p microcontroller for processing. The microcontroller then outputs the desired on-time, off-time, and light intensity to the LEDs for stimulating the region of interest in the mouse brain.

ii. Hardware. We ported the code onto an ATMega328p MCU using the protocol outlined by Arduino. We then soldered the ATMega328p microcontroller (Atmel, San Jose, California) surface mount IR receiver (Vishay Electronics, Malvern, Pennsylvania), and surface-mount the LED (Kingbright USA, City of Industry, California) onto a custom PCB. We designed the PCB in Cadsoft EAGLE and ordered the PCB to be manufactured by OSHpark. The fully assembled PCB is shown in Fig. 5. We connected a 3.7-V, 20-mAh capacity rechargeable LiPo battery (Tenergy, Fremont, California) to the PCB using female headers (SparkFun, Boulder, Colorado). The LiPo battery chosen weighs under 1 g and fits within a $15\times 15\mathrm{mm}$ footprint on the PCB. Finally, we placed two magnets on the bottom side of the PCB that are aligned with those on the baseplate. The final assembly of the electronic component is shown in Fig. 5. Details of all materials used are in the Appendix: CerebraLux manual (Sec. A2.1) and step-by-step instructions to manufacture the PCB are described in the Appendix: CerebraLux manual (Sec. A2.4).

Fig. 5

The electronic component of CerebraLux. A fully assembled PCB is shown here. The battery, female headers, MCU, and IR receiver are on the upper side of the PCB and are shown in the left panels with the schematic in the top panel and the device in the lower panel. The magnets and LED are on bottom portion of the PCB. This view is shown in the right panels with the schematic view in the top panel and the device in the lower panel. The module has a footprint of $15\times 15\mathrm{mm}$ and, along with female headers and battery, has a weight of 2.3 g. This includes magnets that align with and attach to those on the baseplate of the optic component.

2.2.

In Vivo Testing of CerebraLux

3.1.

Transmission Range

We first measured the distance that our device was able to consistently send and receive signals and found this to be 1.8 m. This distance was not affected by glass or plastic in the light path. However, we did find that fluorescent lighting interfered with the IR transmission so it is advisable to use low or alternative sources of light.

3.2.

Stimulator Light Output

The ability to adjust the light output from the stimulator is critical. Excessive light power may compromise neuron function, whereas light power below the activation threshold will not activate ChR2. The minimum light power needed to activate ChR2 depends on a number of factors including the light attenuation properties of the target brain region, ChR2 expression levels, and the age of the tissue. Typically, 2 to $10\mathrm{mW}/{\mathrm{mm}}^2$ is considered an acceptable irradiance range to successfully stimulate opsins.¹⁵ Power can also be manipulated by incorporating pulse-width modulation (PWM) to apply the light at a specific frequency (Hz). A range of frequencies may be used in vivo and in vitro^16–19 but for the tests in this study we used 10 Hz, or 10 ms on and 90 ms off. This is done by changing the “ON” and “OFF” time variables in the GUI (Fig. 3). Power can also be manipulated by changing the forward voltage, shown as the intensity variable in the GUI (Fig. 3). This was assessed in vitro by placing a fully assembled device in an integrating sphere (PMD100, Thorlabs, New Jersey) to assess light power output from the tip of the fiber-optic. Furthermore, the performance of rechargeable batteries may be affected by the number of recharge cycles. We, therefore, tested power output using three different batteries that had been recharged once (battery 1), three times (battery 2), or multiple times (battery 3) beforehand. The data are shown as total power output ($\mu \mathrm{W}$) or irradiance [$\mathrm{mW}/{\mathrm{mm}}^2$, Figs. 6(a) and 6(b)]. We found a linear relationship ($r=0.98$) between the forward voltage and light power output with a peak average output of $825\pm 17\mu \mathrm{W}$ or $4.2\pm 0.09\mathrm{m}\mathrm{W}/{\mathrm{mm}}^2$. Interestingly, the power output declined by only 3% when the forward voltage was reduced by 40%. Across all voltages, the newer battery (1) had a higher power output compared to the older batteries (2,3), but all batteries showed the same % reduction with decreasing forward voltage. Details of the assembly and use of these batteries can be found in the Appendix: CerebraLux manual (Sec. A2.5).

Fig. 6

Power output as a function of forward voltage. The effect of varying the forward voltage or intensity (%) on light power output was assessed in the same device connected to three fully charged batteries of different recharge cycles. The data are shown as the total power output (a) and irradiance (b) and show that power output was reduced by 3% if the forward voltage is reduced by 40%. Across all voltages, the battery with fewer recharge cycles (1) had a higher power output compared to batteries with more recharge cycles (2, 3) but all batteries showed the same % reduction with a decrease in forward voltage.

3.3.

Stimulator Light Output over a Single Discharge Cycle (Battery Runtime)

Well-characterized battery life and power output over a discharge cycle provide critical information for a proposed experimental design. We measured the runtime of the 3.7-V 20 mAh battery tested at a 10-Hz frequency and 100% forward voltage while recording light power output from the fiber-optic tip every 5 min until the battery was not able to maintain the threshold light output of $200\mu \mathrm{W}$. We used the same three batteries as used in the previous test that been recharged once (battery 1), three times (battery 2), or multiple times (battery 3) beforehand. The data show a decline in power output over time (Fig. 7). During the first 30 min, the power output of all three batteries show a steady decrease of $\sim 10\%$ for each 10 min of stimulation that is independent of the initial power output. Thereafter, power output declined but remained above 50% of the initial output for 35 min (battery 3) or 50 min (batteries 1 and 2). Together, these data (Figs. 6 and 7) show that power output varies relatively little if the forward voltage is reduced, but it is influenced by the number of prior recharge cycles of the battery and the time in use. These data show that it is important to monitor these two parameters during and between experiments.

Fig. 7

Battery runtime. The lifetime and performance of three batteries was assessed by recording light power output at experimental conditions (10 Hz, 100% forward voltage or intensity) until a threshold of $200\mu \mathrm{W}$ was passed. The battery with fewer recharge cycles (1) had a highest initial power output compared to a battery that had recharged 3 (2) or 10+ times (3). During the first 30 min, the power output of all three batteries showed a steady decrease of $\sim 10\%$ for each 10’ of stimulation that was independent of the initial power output. Thereafter, power output declined but remained above 50% of the initial output for 35 min (3) or 50 min (1, 2).

3.4.

Testing the Stimulator In Vivo

Activating dopamine 1 (D1) receptors in direct pathway striatal projection neurons increases turning behavior.²⁰ Optogenetic activation of these neurons also increases locomotion and velocity.^21,22 We proposed that CerebraLux activation of ChR2 in neurons expressing D1 receptors in the right side of the striatum would increase both counter-clockwise rotations and total mobility. This was tested by connecting the electronic component to the in situ baseplate, made of white HDPE [Fig. 8(A)], and placing the mouse, from the genetically engineered Ai27 × D1 cre line, in the recording chamber. Details of the operating instructions of the probe and software can be found in the Appendix: CerebraLux manual (Sec. A3). Three mice were used for these tests. After 5 min of basal recording, the optoprobe was turned on (10 Hz, 100% intensity) for 5 min and then off for the final 5 min of the test. Figure 8(Aa) shows the mouse with the optoprobe turned on in the arena used. During this period, counter-clockwise rotations increased [$F_{(\mathrm{2,4})}=14.19$, $p<0.05$] when compared with the basal off period [Fig. 8(Ab)] without changing clockwise rotations [$F_{(\mathrm{2,4})}=4.19$, Fig. 8(Ac)]. Activation of the optoprobe also increased distance traveled [$F_{(\mathrm{2,4})}=7.367$, $p<0.05$, Fig. 8(Ad)], velocity [$F_{(\mathrm{2,4})}=13.13$, $p<0.05$, Fig. 8(Ae)], and mobility [$F_{(\mathrm{2,4})}=23.45$, $p<0.005$, Fig. 8(Af)] versus % basal levels. An example of this behavior is seen in the video (Fig. 9). Together these data show that CerebraLux produces the predicted behavioral outcome, an increase in turning behavior, in mice expressing Ch2R in direct pathway striatonigral neurons.

Fig. 8

Validation of CerebraLux in vivo. (A) White HDPE. Mice (Ai27 x D1-cre, $n=3$) were implanted with the baseplate, made from white HDPE, containing the fiber-optic (the optic component) and allowed to recover. Shortly before the experiment, the PCB (the electronic component) was attached and mouse placed in the open field for 5 min, the probe was turned on for 5 min (a) and then off for the last 5 min. The behavior was video-tracked and the data exported and analyzed. We found that ChR2 activation in the right striatum (b) increased the number of counter-clockwise rotations, (c) did not alter the number of clockwise rotations, and increased distance traveled (% baseline, d), velocity (% baseline, e), and time spent mobile or % mobility (% baseline, f). *$p<0.05$ versus the first “OFF” period. (B) Black HDPE. Mice (Ai27 x D1-cre, $n=2$) were implanted with the baseplate, made from black HDPE, and the same experiment was conducted as in A (experiment). There was a marked reduction in light seen when the probe was turned on (g) but the same relative increase in counter-clockwise, but not clockwise, rotations was observed as in mice implanted with a white HDPE baseplate. A sham mouse (Ai27 × D1 cre), implanted with an optoprobe without the intracerebral fiber-optic, showed no counter-clockwise rotations and a decrease in clockwise rotations when the probe was turned on (h) and (f).

Fig. 9

Increased turning when Cerebralux is turned on. This video shows turning behavior in a mouse implanted with a CerebraLux probe. The mouse, with a white HDPE-CerebraLux probe, placed in the right striatum, was placed in an open-field chamber on an elevated, semitransparent, and glass plate. The video, was taken by a camera mounted below the subject and shows turning behavior for 10 s before the optoprobe was turned on, for 5 min when the probe was on and for $\sim 10\mathrm{s}$ after stimulation has finished. Note that the view of subject is from below this reversing the direction of rotation and that the speed of the video is increased to a $2\times$ speed throughout (Video 1, MP4, 10.3 MB [URL: http://dx.doi.org/10.1117/1.NPh.4.4.045001.1]).

So as to reduce light scatter, we repeated these tests using a baseplate made of black HDPE in two additional mice of the same genotype [Fig. 8(B)]. This reduced light scatter [shown by the frame taken when the oproptobe was turned on in Figs. 8(Bg) versus 8(Aa)] but did not change the % increase in counter-clockwise rotations during the on versus basal off period [350% versus 348% in mice with black versus white HDPE, respectively, Fig. 8(Bh)]. Turning the optoprobe on also induced a similar % change in clockwise rotations compared with the basal off period; 138% versus 157% in mice with black versus white HDPE, respectively [Fig. 8(Bi)]. In addition, the sham mouse, from the Ai27 × D1 cre line, in which an optoprobe lacking the fiber-optic was implanted, showed no change in counter-clockwise rotations [Fig. 8(Bh)] and a decrease in clockwise rotations [Fig. 8(Bi)] when the probe was turned on. These data show that the use of the black HDPE baseplate did reduce light scatter but did not affect turning behavior resulting from activation of striatal D1 neurons.

4.1.

Ease of Use

There are several features of the CerebraLux design that make it easy for a researcher to use. The low-profile lightweight baseplate is stable, so it is unlikely to come loose once secured to the skull. The magnets on both the baseplate and PCB allow easy application of the PCB to the mouse. This plug-and-play feature contrasts with wired system optoprobes where anesthesia is often required to connect the mouse. Once in place, the IR-based controller with a transmission range of 1.8 m facilitates complex experimental designs that are not possible using wired or some wireless devices. Once the device has been activated, the mouse can move through mazes or doors allowing further flexibility in experimental designs. Not only is the total light power output adjustable using PWM, the rechargeable battery is also able to sustain 30 min of activation, considerably longer than the typical activation period. Finally, the ability to control CerebraLux through downloadable software on any computer is both easy and inexpensive. The GUI is written in open-source Python to ensure access and programmability by researchers, whereas the IR communication protocol is written in the widely documented Arduino IDE. With all of the implemented features, the total weight of CerebraLux is 2.8 g, of which the mouse permanently carries 0.5 g of the optic component, with the remaining 2.3 g electronic component being detachable. This allows the mouse to move unencumbered during and after experiments. A summary of the features of our device in comparison with similar devices is shown in Table 1.

Table 1

Comparison of CerebraLux with other available devices.

4.2.

Ease of Assembly

The CerebraLux device is designed to be assembled without prior electronic and manufacturing experience. Step-by-step instructions detailing how to manufacture every component of the device are included in the CerebraLux manual. It is also possible to outsource the manufacturing of certain components (e.g., the baseplate) to high-resolution 3-D printing or laser sintering services such as Protolabs. Most of the components, except those that are milled, are commercially available. The baseplate can be machined in 1 h using a low-cost, consumer-grade CNC mill, and the stereotaxic adapter printed in 15 min using a 3-D printer (uPrint SE, Stratasys, Eden Prairie, Minnesota). The PCB board can also be manufactured on the Othermill on a copper board, which is less expensive than OSH Park. Once the PCB is manufactured, researchers can follow the instructions in the CerebraLux manual on soldering all the surface mount components (Sec. A2.4). Manufacturing and assembly of CerebraLux requires a total of 3 h once all the components are attained. In contrast, the cellular-scale stimulator created by the Bruchas research group requires specialized microfabrication facilities and a lengthy production time of two weeks to complete.⁷ CerebraLux’s rapid turnaround time of a day in manufacturing means that design iterations can be developed and tested in a shorter span of time. Additionally, multiple devices can be assembled and implanted at the same time, thus increasing the number of optogenetic studies that a laboratory can conduct within a particular time frame.

4.3.

Cost Analysis

CerebraLux alleviates the financial barriers to optogenetics technology with its open-source design and low cost. Table 2 shows a price breakdown of all components used in our device and a cost analysis comparing our device with the current standards in the field of optogenetics. Our starting price of $108.10 is considerably less expensive than the Cambridge Neurotech device⁶ or the RF resonant cavity proposed by the Poon group⁹ and other suppliers. It is important to note that this cost analysis does not include the cost of purchasing a CNC mill. However, these are often available on university campuses and, even if outsourced, CerebraLux still remains less expensive than similar commercially available devices.

Table 2

Price breakdown of all components used in CerebraLux.

4.4.

In Vivo Testing

Our behavioral experiments demonstrate proof-of-concept that we are able to target and stimulate opsin-expressing cells in the brain with our low-cost device. As previously mentioned, the appropriate light power needed for an experiment often falls within a narrow range. Our device is able to produce $824\pm 17\mu \mathrm{W}$ of power, which, for a $500\text{-}\mu \mathrm{m}$ diameter fiber, corresponds to irradiance values of $4.2\pm 0.1\mathrm{mW}/{\mathrm{mm}}^2$. Our behavioral testing shows that this is adequate to stimulate ChR2. It must be noted that this is affected by the battery, in particular, the number of recharge cycles. Our data show that a greater number of recharge cycles reduces the maximum light power output and runtime. This reduction can be partially offset by careful soldering of the connections and observing the manufacturer guidelines for care and recharging of the battery. In addition, as these batteries are relatively inexpensive and available, a supply of backup batteries would be advisable.

4.5.

Future Directions

In providing the complete specifications of CerebraLux, we are providing the foundation, in which this and future iterations of CerebraLux may be built. We have shown that, as is, CerebraLux works in vivo. As with any engineering project, there are also future directions and improvements that would enhance and improve this probe. One of these is the modality used to activate the probe. We successfully utilized IR signaling as a medium of communication between the researcher and the stimulator. IR signaling is easy to implement using commercial off-the-shelf components such as IR LEDs and Arduino boards. The ease of IR programmability was also demonstrated by sending signals with an Arduino Uno utilizing popular open-source IR libraries. However, IR requires line-of-sight communication and is affected by ambient light sources, which may restrict the type of behavioral studies that can be done. In future iterations, an approach similar to Hirase et al.¹³ can be adopted using an array of LEDs to transmit the signals from multiple angles, ultimately reducing interference and signal loss. Alternatively, the fragility of IR transmission can be bypassed through use of Bluetooth Low Energy 5.0 technology embedded in microcontrollers; however, this method also presents challenges, such as implementing Bluetooth communication protocols.

Future iterations of our device should also aim to maximize efficiency of the optic assembly. Currently, the LED we used produces 8.1 mW of light power at maximum forward voltage, but 0.824 mW of total power or $4.2\mathrm{mW}/{\mathrm{mm}}^2$ is transmitted through the fiber-optic at a 10-Hz frequency, a transmission efficiency of 10.6%. It must be noted that our device utilizes direct butt-coupling between the diode and the waveguide with no optical focusing mechanism, explaining the low transmission efficiency. An optical focusing mechanism could be established by lenses and condensers that would collimate the light from the LED before transmission through the fiber-optic and allow the device to be operated in experimental setups with greater light power needs. However, incorporating such focusing elements would also increase the weight and height profile of the device. The baseplate could also be manufactured with higher precision and tolerance to improve alignment between the diode and fiber-optic and to further improve light power transmission. These iterations would increase power output and would also allow a lower voltage to be used, prolonging overall battery life. Another iteration would be to use black HDPE [Fig. 8(B)] for the baseplate so as to minimize light leakage and prevent the mouse from visually reacting to the light.

Our baseplate design can also be altered to target different classes of opsins. By simply changing the location of the ferrule hole in the baseplate during milling, we can align the fiber-optic with a different color diode on the LED. The LED used here has three color settings (i.e., red, green, blue), which can be programmed to allow for stimulation of a wider range of opsin subtypes. A new baseplate design would also be needed to include bilateral probes. This would require multiple LEDs or a single LED with a light splitter being used to stimulate different targets in the brain. A change in the baseplate design may also be needed for different $XY$-coordinates for different brain regions or for use in other animal models.

One more consideration for a future design is the addition of electrophysiological recording capabilities. To acquire measurements such as temperature and electric signals, our device will require extensive modifications to the device programming, microcontroller selection, and chassis design. However, the addition of this function would require specialized fabrication techniques, which may make the device harder to manufacture and less accessible for all to use.