Removing crosstalk signals in neuron activity by time multiplexed excitations in a two-photon all-optical physiology system

The two-photon all-optical physiology system has attracted great interest in deciphering neuronal circuits in vivo, benefiting from its advantages in recording and modulating neuronal activities at single neuron resolutions. However, the interference, or crosstalk, between the imaging and photostimulation beams introduces a significant challenge and may impede the future application of voltage indicators in two-photon all-optical physiology system. Here, we propose the time multiplexed excitation method to distinguish signals from neuronal activities and crosstalks from photostimulation. In our system, the laser pulses of the imaging beam and photostimulation beam are synchronized, and a time delay is introduced into these pulses to separate the fluorescence signal generated by these two beams. We demonstrate the efficacy of our system in eliminating crosstalk signals from photostimulation and evaluate its influence on both genetically encoded calcium indicators (GECIs) and genetically encoded voltage indicators (GEVIs) through in vivo experiments.


Introduction
Due to its advantages in superior penetration depth, low phototoxicity, and long-term imaging ability, two-photon microscopy has significantly advanced the development of neuroscience [1].Further, the integration of two-photon imaging and optogenetics has broaden the range of potential applications in neuroscience research [2,3].The deployment of all-optical physiology systems has facilitated various discoveries of neuroscience, such as the functionality of the visual cortex [4], the generation of new synapses [5], the interaction between behavior and cognition [6], and other phenomena [7][8][9].However, there exists crosstalks between commonly used in vivo calcium imaging indicators and opsins, which can be categorized into following types [10]: i.The laser beam for two-photon imaging of the activity indicators activates the opsins, causing unintentional activation of opsin-expressing neurons during point by point scanning within the field-of-view (FoV).
ii.The photostimulation laser for opsin activation or inhibition induces unexpected excitation of the activity indicators, generating fluorescence signals that cannot be distinguished from the fluorescence generated by imaging lasers.
iii.The red indicator proteins frequently used to label opsin-expressing neurons can also be excited by the photostimulation beam.Up: Emission spectra of typical red indicators.Data from [12].Green range indicates the green filter range (500 nm ∼ 550 nm) of our two-photon all-optical physiology system.Down: Two-photon excitation spectrum for GCaMP6m.Data from [13].Red vertical line indicates the wavelength (1040 nm) of two-photon photostimulation beam.(c) Typical images of recording channel before (t = 0 ms), during (t = 17∼51 ms) and after (t = 68, 85 ms) photostimulation.Scale bar: 80 µm.
Among them, the first type of crosstalk can be mitigated by selecting a larger scanning FoV, using the lowest possible imaging beam power, and choosing opsins with a low activation probability in the calcium imaging wavelength [10].However, the second and third types of crosstalk directly degrade the image quality of neuronal activity recording [Fig.1(c)] [4,11], which consequently affect subsequent neuronal activity extraction, neuronal network timing, and even interfere the final biological conclusions.In practical experiments involving two-photon all-optical physiology with voltage indicators, crosstalk can obscure the rapid temporal voltage signals, which can significantly interfere with accurately determining the sequence of neuronal firing.Mardinly et al. addressed this challenge by implementing a fast thresholded RC circuit to gate the photostimulation beam [3].However, this method reduces the imaging duty cycle and prevents continuous stimulation of neurons.In practical experiments, even though lowering the stimulation power is often used to decrease the second and third types of crosstalk, it cannot completely eliminate it.Specifically, when a substantial number of neurons need to be stimulated simultaneously, reducing the average power of photostimulation may significantly reduce the success rate of photostimulation.
Here we propose to eliminate the image crosstalk caused by the second and third type, through the time multiplexed excitation method.In our custom-built two-photon all-optical physiology system, the imaging beam and photostimulation beam, both at a repetition rate of 80 MHz, are synchronized with a time delay.The distinct arrival times at the detector makes the fluorescence signal generated by the two beams separable.With our method, the imaging duty cycle remains unaffected, and target neurons could be stimulated at any time without being constrained by the scanning position of the imaging beam.Thus, stimulus pattern with any scanning trajectory as well as extended patterns generated by spatial light modulator (SLM) can be used to stimulate target neurons with our approach.To validate the efficacy of our approach, we conduct various experiments including single-cell activation experiments using Genetically Encoded Voltage Indicators (GEVIs) and Genetically Encoded Calcium Indicators (GECIs), as well as multi-cell sequential activation and multi-cell holographic activation experiments using GECIs in the CA1 region of the mouse hippocampus.Through these experiments, we demonstrate the differences in activity traces, activity correlations, response intensities, response times, and firing orders of neurons during the same experiment before and after the removal of fluorescence signal crosstalk, verifying the effectiveness of our method.Our system is built on a custom two-photon all-optical physiology system, as shown in Fig. 2. The laser (Chameleon Discovery NX, Coherent Inc.) has two ports: one is tunable within the range of 690 nm to 1300 nm, and the other is fixed at 1040 nm.We use the tunable port for two-photon imaging, setting the wavelength to 920 nm to optimize the imaging performance of the activity indicator.And the fixed port is applied for two-photon photostimulation.The beam from the tunable port, generated through optical parametric amplification (OPA), and the beam from the fixed port are temporal synchronized with same seed source, thereby ensuring time synchronization between each other.The intensity of each laser beam is controlled by a separate Pockels cell (Pockels cell: 350-80-LA, Controller: 302RM, Conoptics, Inc.).
The signal from PMT is then amplified by a high-speed transimpedance amplifier (DHPCA-100, Femto) and split into two channels by a power splitter (ZAPD-30-S+, MiniCircuits) [18].To match the optical delay introduced in the microscope, a relative electric delay is introduced by using cables of different lengths (∼1.3 m difference).Thus, both expected fluorescence signal and crosstalk fluorescence signal are digitalized by a high-speed data acquisition board (vDAQ data acquisition hardware, vidiro), synchronized with an 80 MHz laser clock.We adjust the length of the laser clock cable to enable that the data acquisition system can effectively capture fluorescence signals from both channels at their peak intensities.The custom built two-photon all-optical physiology system is controlled by Scanimage (vidiro).

Crosstalk removal
Leakage between sampling channels, caused by the fluorescence lifetime of the fluorophore and the dynamic response of the circuit, cannot be entirely avoided.Therefore, we need to perform subsequent processing on the signals from both channels.With two channels of fluorescence signals of different time delays and assuming a linear leakage ratio of the system, our method can separate the fluorescence signals excited by the imaging beam and photostimulation beam [21].
Assuming linear leakage ratio between channel 1 and channel 2, we have: Where J i is detected signal on channel i, S i,j is channel leakage ratio from beam j to channel i, and I i is the original signal generated by beam i.Before experiments, we calibrate the system by exciting the sample individually for each beam to obtain the accurate S matrix.And the original signal [I] without crosstalk can be calculated as: In this case, I has two components: I 1 and I 2 .Among them, I 1 is the desired neuronal activity image without crosstalk, while I 2 is the crosstalk fluorescence signal generated by the photostimulation beam.

Stereotactic injection
All experimental procedures are carried out in accordance with animal protocols approved by Tsinghua University.C57BL/6J mice were used in this study and were maintained under standard conditions by the Animal Research Center of Tsinghua University.For virus injections into the mouse CA1, mice are first anesthetized with Avertin (0.3 g/kg body weight).Craniotomies on the skull over the right hippocampus are performed using a 0.5 mm-diameter drill, and 300 nl of the virus is injected into the dorsal CA1 (−2.0 mm A/P, −1.5 mm M/L, and −1.4 mm D/V) using a 10 µl nanofil syringe controlled by UMP3 and Micro4 system (WPI) with a speed of 60 nl/min.After the injection, the needle remains in place for 10 min to ensure that the viruses spread to the targeted area before it is slowly withdrawn.After injection, mice are returned to their home cages and are allowed to recover for at least 5 days before performing further experiments.

Imaging window implantation
Anaesthesia is induced with 5% isoflurane before surgery and is kept with 1.5% isoflurane during the surgery.Prior to surgery, mice are administered an anti-inflammatory drug Meloxicam (8 mg/kg) subcutaneously.A 3 mm-diameter craniotomy is made using a micro-drill at the AAV injection site.The cortex tissue above the hippocampus is aspirated with a 27 gauge needle connected to a pump until the fibers of the corpus callosum become visible, exposing the alveus of the hippocampus.A stainless-steel cannula (3 mm diameter, 1.6 mm height) covered by a cover glass (3 mm diameter, 0.17 mm thickness) is inserted into the opening until the glass is in contact with the fibers.The cannula is secured in place with glue and dental cement, and a stainless-steel head-post is also fixed onto the skull using dental cement.Following surgery, mice are returned to their home cages and are monitored daily.Meloxicam (4 mg/kg/day) is given subcutaneously for 3 consecutive days after surgery to prevent inflammation.Mice are allowed to recover for at least 3 weeks before imaging experiments are performed.

Sparse labeling of neurons
Tamoxifen dissolved in corn oil (5 mg/ml) is administrated to mice via intraperitoneal injection (30 mg/kg body weight) 1 week before imaging.This allows the expression of ChRmine-mScarlet in a sparse subpopulation of CA1 neurons.To validate our method, we perform single neuron photostimulation experiment on mouse hippocampus CA1 pyramidal cells.The image of the selected FoV is illustrated in Fig. 3(a).Spiral excitation patterns, lasting for 1 s, are used to stimulate the targeted neurons.Through the frame triggers of the two-photon imaging system, we ensure consistent time intervals for photostimulation.We collect 256 × 256 pixel images at a frame-rate of 60 Hz.The activity traces of all neurons are extracted using suite2p [25] (python) and then filtered with a lowpass filter (scipy.signal,python).The heatmap in Fig. 3(b) displays the neuronal activity before and after crosstalk removal.The presence of blue vertical lines in the neuronal activity heatmap before crosstalk removal indicates the artifact resulting from the photostimulation beam, which is eliminated after crosstalk removal.Next, we compare the correlation coefficient of neuronal activity before and after crosstalk removal.We calculate the activity correlation coefficients between all non-target neurons and target neuron [Fig.3(c)].The average correlation coefficient decreases significantly from 0.355 to 0.088 (p < 0.001, ANOVA).We also calculate the activity correlation coefficient between all neurons [Fig.3(d)], and observed a decrease from an average of 0.595 to 0.530 (p < 0.001, ANOVA).Figures 3(e) and 3(f) demonstrate the activity traces of the target neuron and a non-target neuron, respectively.The fluctuations observed in the residual signal correspond to the activation of the photostimulation beam.

Crosstalk signal removal in serial multi-neuron photostimulation
To demonstrate the improved accuracy of the neuronal activity firing sequence after crosstalk removal, we conduct serial multi-neuron photostimulation experiments in the CA1 region of the mouse hippocampus.Figure 4(a) displays the image of the selected FoV and the order of stimuli.We select three neurons and activate them sequentially using spiral excitation patterns, with each stimulation lasting for 50 ms.The heatmap in Fig. 4(b) illustrates the neuronal activity before and after crosstalk removal.For comparison, the activity trace of each neuron is normalized.The individual and average responses to each stimulus are demonstrated in Figs.4(c-e).The response curve before crosstalk removal is different from the response curve after crosstalk removal whether in response time [Figs.4(f-h)] or response intensity [Figs.4(i-k)].The starting point for analysis is set when the photostimulation of neuron #1 is triggered.We observe that in the absence of crosstalk removal, neuron #3 exhibits an earlier peak compared to neuron #1, statistically (neuron #1: 218 ± 55 ms, neuron #3: 207 ± 59 ms, N = 9, mean ± s.d.).However, after eliminating crosstalk, neuron #1 reached peak intensity earlier than neuron #3 (neuron #1: 237 ± 37 ms, neuron #3: 261 ± 71 ms, N = 9), which is consistent with our activation sequence.This indicates that the presence of crosstalk can affect the determination of the firing order of neurons during two-photon photostimulation.For voltage indicators with significantly shorter durations, the removal of crosstalk becomes even more critical.Holographic photostimulation offers the capability of simultaneously activating multiple neurons, providing an advantage in neuronal circuit regulating [2,3,11,26,27].In our experiment, we use Computer Generated Holography (CGH) to perform parallel photostimulation in the CA1 region of the mouse hippocampus.Figure 5(a) presents the image of the selected FoV and the corresponding stimuli targets.To generate the holographic pattern, we employ the weighted Gerchberg-Saxton (GSw) algorithm [28], resulting in precise targeting of three specific neurons.The SLM generates a point array, which was then spiral scanned across the three neurons simultaneously for a duration of 300 ms. Figure 5(b) illustrates the heatmap of neuronal activity before and after crosstalk removal.For comparison, the activity traces of all neurons are normalized.The individual and averaged responses to each stimulus are demonstrated in Figs.5(c-e).Notably, the response curve before crosstalk removal exhibits clear distinctions in both response time [Figs.5(f-h)] and response intensity [Figs.5(i-k)], compared to the response curve after crosstalk removal.After the removal of crosstalk, we observe a delay in the rising time and a reduction in response intensity.These findings hold significant importance in the study of neuronal network connectivity.fluorescent signals by binding fluorescent proteins on the neuron membrane.This facilitates a more direct and reliable transmission of signals.Furthermore, the signal of GEVI is less susceptible to environmental factors, contributing to its superior stability.Additionally, GEVI can capture a different set of information within neurons, rendering it a valuable tool for studying neuronal activity.These advantages make GEVI a highly promising tool for neuroscience study, particularly when compared with traditional methods such as electrode insertion.

Crosstalk signal removal in holographic multi-neuron photostimulation
To further validate the competence of our method beyond GECIs, we label pyramidal neurons with JEDI-2P [29] in the CA1 region of the mouse hippocampus [Fig.6(a)].In contrast to GECIs, JEDI-2P exhibits negative signal changes where neuronal fluorescent intensity decreases during action potential firing [Figs.6(b, c)], which exacerbates the issue of crosstalk signal in analysis.To address this, we reduce the image sampling to 128 × 24 pixels, resulting in an imaging frame rate of 396.1 Hz, for voltage imaging.Representative images obtained from recording are shown in Figs.6(d, e).Neurons are selected and stimulated using spiral excitation patterns with each lasting for 1 s.Here, we use the z-score normalization method (z-score ∆F/F = (∆F/F -mean(∆F/F))/std(∆F/F)) to present the voltage signals of neurons [19].

Discussion and conclusion
We propose the time multiplexed excitation method to effectively remove crosstalk fluorescence signals in two-photon all-optical physiology experiments.We introduce time delay in the two-photon imaging beam, ensuring that its pulse train does not overlap with that of the photostimulation laser beam.Subsequently, we perform synchronous acquisition and linear demodulation techniques to separate the fluorescence signals generated by the imaging and photostimulation beams.With this method, we successfully eliminate image stripes in neuronal activity recording that commonly arise when the photostimulation beam is on.Notably, our method does not require any customized components, making it readily applicable to existing two-photon all-optical physiology systems.
We compare the neuronal activity trace, correlation, rising time, and intensity during photostimulation, as well as firing order during multi-neuron photostimulation before and after crosstalk removal, to confirm the efficacy of our method.It is crucial to accurately determine the timing of neuronal activity during photostimulation, considering that neurons within a neuronal network fire in a specific sequence.Therefore, through our method, one can gain valuable insights into the structure and functionality of neuronal circuits.
Our approach ensures the effectiveness of crosstalk elimination as long as the imaging laser and photostimulation laser are synchronized.This synchronization allows for the utilization of a low repetition laser for two-photon photostimulation [19], which provides practical benefits with high-energy laser pulses [10].The approach introduced herein presents a crucial advancement for two-photon all-optical physiology systems, and we anticipate that it will prove valuable for further neuroscience research.
With a broad emission band of the red indicators, a portion of the fluorescence signal can leak into the recording channel of neuronal activity [Figs.1(a, b)].

Fig. 2 .
Fig. 2. System scheme of two-photon all-optical physiology system.A time delay of about 6.25 ns is introduced between two-photon imaging beam and two-photon stimulation beam by an optical delay line of ∼1.8 m.The excited fluorescence signals from imaging beam and stimulation beam are detected by a PMT and amplified, followed by synchronous digitization with the clock of the laser.EOM: Electric optical modulator.ETL: Electric tunable lens.BE: Beam expander.SLM: Spatial light modulator.L: Lens.SL: Scan lens.TL: Tube lens.DIC: Dichroic mirror.PMT: Photomultiplier tube.

Fig. 3 .
Fig. 3. Crosstalk signal removal in single neuron photostimulation.(a) Two-photon imaging results of recording FoV.(b) Up: ∆F/F heatmap of extracted neurons before crosstalk signal removal.Down: ∆F/F heatmap of extracted neurons after crosstalk signal removal.Orange arrowheads indicate the moment of two-photon photostimulation.(c) Correlation coefficient between other neurons and the activated neuron before and after crosstalk signal removal.(***: p < 0.001) (d) Correlation coefficient between all neurons before and after crosstalk signal removal.(***: p < 0.001) (e) Normalized ∆F/F of the activated neuron before and after crosstalk signal removal.(f) Normalized ∆F/F of a non-target neuron before and after crosstalk signal removal.

Fig. 4 .
Fig. 4. Crosstalk signal removal in serial multi-neuron photostimulation.(a) Two-photon imaging results of recording FoV.Arrows and spiral scan patterns indicate activated neurons and activation sequence.(b) Up: ∆F/F heatmap of extracted neurons before crosstalk signal removal.Down: ∆F/F heatmap of extracted neurons after crosstalk signal removal.Orange arrowheads indicate the moment of two-photon stimulation.(c-e) Zoomed-in view of neuronal activity during photostimulation of neuron #1-3.Blue and yellow trace indicate every stimuli trial before and after crosstalk signal removal.Red and green trace indicate averaged trial (N = 9) before and after crosstalk signal removal.(f-h) Time to reach the maximum activity intensity after photostimulation of neuron #1-3 before and after crosstalk signal removal.(i-k) Maximum activity intensity after photostimulation of neuron #1-3 before and after removal of signal crosstalk.

Fig. 5 .
Fig. 5. Crosstalk signal removal in holographic multi-neuron photostimulation.(a) Twophoton imaging results of recording FoV.Stars and spiral scan patterns indicate the position of activated neurons and 0th order beam.(b) Up: ∆F/F heatmap of extracted neurons before crosstalk signal removal.Down: ∆F/F heatmap of extracted neurons after crosstalk signal removal.Orange arrowheads indicate the moment of two-photon stimulation.(c-e) Zoomed-in view of neuronal activity during photostimulation of neuron #1-3.Blue and yellow trace indicate every stimuli trial before and after crosstalk signal removal.Red and green trace indicate averaged trial (N = 9) before and after crosstalk signal removal.(f-h) Time to reach the maximum activity intensity after photostimulation of neuron #1-3 before and after crosstalk signal removal.(i-k) Maximum activity intensity after photostimulation of neuron #1-3 before and after crosstalk signal removal.

3. 4 .
Crosstalk signal removal in imaging with voltage indicator JEDI-2P GEVI and GECI are both genetically engineered neuronal indicators used to detect changes in neuronal activity.While GECI monitors fluctuations in calcium ion concentration, GEVI is specifically designed to track alterations in membrane voltage.In comparison to GECI, GEVI exhibits several notable advantages.It directly converts membrane voltage signals into 6.Crosstalk signal removal with fast voltage indicator JEDI-2P.(a) Typical neuron expressing JEDI-2P.Scale bar: 5 µm.(b) The same neuron, as in (a), expressing jRGECO.(c) Simultaneously recording activity of the neuron in (a) and (b) in Z-score ∆F/F of JEDI-2P and jRGECO.Framerate: 207 Hz.(d) Two-photon imaging results of one FoV.Frame rate: 396.1 Hz.Spiral scan pattern and white circle indicate the stimulated neuron and a nearby neuron.(e) Two-photon imaging results of another FoV.Framerate: 396.1 Hz.Spiral excitation pattern and white circle indicate the stimulated neuron and a nearby neuron.(f) Z-score ∆F/F of the stimulated neuron [Cell 1 in (d)] before and after signal crosstalk removal.Blue and green traces indicate activity traces before and after signal crosstalk removal, respectively.Red vertical line indicates the time of photostimulation.(g) Z-score ∆F/F of the nearby neuron [Cell 2 in (d)] before and after removal of signal crosstalk removal.(h) Z-score ∆F/F of the activated neuron [Cell 1 in (e)] before and after signal crosstalk removal.(i) Z-score ∆F/F of the nearby neuron [Cell 2 in (e)] before and after signal crosstalk removal.
demonstrate the activity trace for target and neighboring neurons within the same field of view before and after removing the crosstalk [Figs.6(f,g)].The negative signal changes that are previously obscured by crosstalk are effectively recovered with our method [Figs.6(h,i)].