2.1.

Animals

All procedures were performed following the UK Animal Scientific Procedures Act (1986) under Home Office Personal and Project licenses following appropriate ethical review.

2.2.

Cloning and AAV Production

To generate the pCAG-nirButterfly-IRES-EGFP vector, the CAG promoter fragment amplified from pCAG-Kir2.1-T2A-tdTomato (Addgene #98804), the nirButterfly fragment amplified from pCAG-nirButterfly (Addgene #136590), and the IRES2-EGFP fragment amplified from pIRES2-EGFP (Clontech) were subcloned into pcDNA3.1/Hygro(+) (Invitrogen). The pCAG-Archon1-IRES-EGFP vector was generated by subcloning the Archon1 fragment amplified from the pAAV-CaMKII-Archon1-EGFP vector (Addgene #108417) into the pCAG-nirButterfly-IRES-EGFP vector to replace the nirButterfly coding sequence. AAV vector plasmids, pAAV-CaMK2A-nirButterfly and pAAV-CaMK2A-Archon1, were produced by subcloning nirButterfly and Archon1, respectively, into the pAAV-CW3SL vector in place of EGFP (Addgene #61463) to express the GEVIs under identical virus plasmid backbones. The AAV plasmids were separately packaged into the AAV1 capsid by Penn Vector Core to produce AAV1.CaMK2A.nirButterfly or AAV1.CaMK2A.Archon1 at similar virus titers.

2.3.

Cell Culture, Plasmid Transfection, and AAV Transfection

2.4.

Confocal Imaging

Transduced cultures were expressed for at least 1 week before imaging.

2.5.

Surgery

For performance comparison experiments between nirButterfly and Archon1 in acute brain slices, C57Bl6 mice aged 21 to 22 days were used ($N=6$, either sex). All mice underwent isoflurane-induced surgical anesthesia induced at 5% isoflurane carried in oxygen and maintained at 1% isoflurane in oxygen. To transduce the hippocampus with either nirButterfly or Archon1, the skull over the somatosensory cortex was gently thinned, and the pipette was used to puncture through the thinned skull to deliver 1e10 vg of either AAV1.CaMK2A.nirButterfly or AAV1.CaMK2A.Archon1 (Penn Vector Core USA) into hippocampal CA1 or the cortex over three depths at the rate of $0.2\mathrm{nl}/\mathrm{s}$ (Nanoject II, Drummond). The pipette was left at the final depth for an additional 5 min before being retracted, and skin over the scalp was sutured.

2.6.

Population Imaging in Acute Brain Slices

Acute brain slices were prepared from AAV-injected animals after 3 weeks of expression time. All mice were terminally anesthetized (ketamine/xylazine ip), transcardially perfused with 10 ml of ice-cold sucrose-based cutting solution (in mM: 0.3 KCl, 2.6 ${\mathrm{NaHCO}}_3$, 0.125 ${\mathrm{NaH}}_2{\mathrm{PO}}_4$, 0.3 sodium pyruvate, 190.0 sucrose, 25.0 dextrose, 0.5 ${\mathrm{CaCl}}_2$, 4.0 ${\mathrm{MgCl}}_2$), and then decapitated, and brains were removed under immersion in ice-cold cuttings solution that bubbled with $95\%{\mathrm{O}}_2$ to $5\%{\mathrm{CO}}_2$. Coronal sections ($300\mu \mathrm{m}$ thickness) were cut with a vibratome (Leica VT1200S, Leica Microsystems) and transferred into holding artificial cerebrospinal fluid (in mM: 12.6 NaCl, 3.5 KCl, 26 ${\mathrm{NaHCO}}_3$, 0.125 ${\mathrm{NaH}}_2{\mathrm{PO}}_4$, 1.0 dextrose, 2.0 ${\mathrm{CaCl}}_2$, 2.0 ${\mathrm{MgSO}}_4$) bubbled with 95% ${\mathrm{O}}_2$ TO $5\%{\mathrm{CO}}_2$. Slices were first held at 33°C for 30 min and then transferred to room temperature for at least another 30 min recovery. For recordings, single slices were transferred into the recording chamber mounted on the stage of a dual emission widefield epifluorescence immersion microscope (Scientifica, United Kingdom). Slices in the recording chamber were superfused with recording ACSF bubbled with $95\%{\mathrm{O}}_2$ to $5\%{\mathrm{CO}}_2$ (in mM: 12.6 NaCl, 3.5 KCl, 26 ${\mathrm{NaHCO}}_3$, 0.125 ${\mathrm{NaH}}_2{\mathrm{PO}}_4$, 1.0 dextrose, 1.2 ${\mathrm{CaCl}}_2$, 1.0 ${\mathrm{MgSO}}_4$) at 32°C to 34°C.

For synaptic stimulation, borosilicate glass pipettes were pulled on a two-stage vertical puller (PC-10, Narishige, Japan) to a pipette resistance of 0.5 to $1\mathrm{M}\mathrm{\Omega }$ and filled using the recording ACSF (in mM: 12.6 NaCl, 3.5 KCl, 26 ${\mathrm{NaHCO}}_3$, 0.125 ${\mathrm{NaH}}_2{\mathrm{PO}}_4$, 1.0 Dextrose, 1.2 ${\mathrm{CaCl}}_2$, 1.0 ${\mathrm{MgSO}}_4$). For the Gabazine condition, $5\mu \mathrm{M}$ Gabazine (Tocris) was added to the recording ACSF, and the slice was superfused with Gabazine-ACSF for 10 min before recording. For the Gabazine recovery condition, the slice was superfused with the original recording ACSF for 30 min before recording.

The epifluorescence immersion microscope was equipped with a CMOS camera (acA1920-$155\mu \mathrm{m}$, BaslerAG) or photodiode (TILL Photonics, Gräfelfing, Germany), using LED light sources (CoolLED, United Kingdom; FiberOptoMeter, NPI Electronic, Germany) and with the following optical filters: (Semrock and Chroma): miRFP670 (donor) excitation FF02-615/29, excitation beamsplitter 635LP (FF635-Di01), and miRFP670 emission 676/37. Signals were acquired with an Axon 700B Multiclamp amplifier and digitized at 10 kHz with a Digidata 1440A using pCLAMP software (Molecular Devices, San Jose, California) for data acquisition and the synchronization of stimulus delivery. Only the donor channel was recorded for nirButterfly. Identical optical filters were used for functional imaging experiments with Archon1.

2.7.

Data Analysis

Spectral and intensity data were analyzed post-hoc using Leica Application Suite X.

Manual ROIs for optical traces in functional imaging experiments were generated in ImagePro. Fluorescence intensity quantifications for GEVI brightness comparison were performed using ImagePro.

The quantification of photodiode recordings and the fitting of response time constants were analyzed using custom MATLAB scripts from the Knopfel Lab MATLAB scripts library.

All statistical tests are performed using MATLAB.

3.1.

nirButterfly Exhibits Sevenfold Higher Brightness than Archon1

The brightness of the two GEVIs was first compared in cultured HEK293T cells. We generated bicistronic plasmids encoding either nirButterfly or Archon1 (each GEVI equipped with a cytosolic EGFP tag on the same plasmid backbone) and performed side-by-side transfection into cultured HEK293T cells [Fig. 2(a)]. We used the cytosolic EGFP expression to normalize across other factors that may introduce variability in total detected brightness, including size variability of the expressing cells and variation in GEVI protein expression levels.

Fig. 2

Brightness comparison between nirButterfly and Archon1 in HEK cells and cultured neurons. (a) Representative images of HEK293 cells co-expressing the GEVI (NIR channel, top) and EGFP (below) after being side-by-side transfected with pCAG-nirButterfly-IRES-EGFP (left) or pCAG-Archon1-IRES-EGFP (right). For visualization, image gain is enhanced by 1.5× for the EGFP channel of nirButterfly and 5× for the near-infrared channel of Archon1. $\text{Scale bar}=10\mu \mathrm{m}$ (b) Absolute fluorescence intensities (red-green bars) in the near-infrared and EGFP channels for nirButterfly and Archon1 ($N=26$ FOVs for nirButterfly; 24 for Archon1). EGFP-normalized fluorescence intensities (black bars). (c) Cultured neurons were side-by-side transduced with AAV1.CaMK2A-nirButterfly or AAV1.CaMK2A-Archon1. Images are shown at the same display gain, both without supplemental chromophores. The Archon1 inset shows the same FOV with the image gain 4× enhanced. Arrows indicate expression aggregates. $\text{Scale bar}=20\mu \mathrm{m}$. (d) Absolute fluorescence intensities for cultures transduced with nirButterfly ($N=10$ FOVs for nirButterfly with biliverdin; 5 for nirButterfly without biliverdin) and Archon1 ($N=10$ for Archon1 with all-trans-retinal; $N=5$ for Archon1 without all-trans-retinal), both with and without supplements of their respective chromophores (biliverdin for nirButterfly, and all-trans-retinal for Archon1). Data presented in median ± SEM; *** = $p<0.001$; individual data points shown for $N=<10$.

We observed an approximately 1.5-fold higher fluorescence intensity of cytosolic EGFP expression in HEK293T cell cultures transfected with Archon1-IRES-EGFP than nirButterfly-IRES-EGFP [Fig. 2(b), red-green bars]. This could be due to the larger molecular size of nirButterfly than Archon1, which could potentially lead to decreased expression levels. We observed membrane-targeted expression of nirButterfly in the transfected HEK cells, but cytosolic aggregates were typically observed in HEK293T cells expressing Archon1 [Fig. 2(a)]. Such Archon1 aggregates have also been reported previously.^25–27 After normalization on the EGFP fluorescence level, we observed on average sevenfold higher fluorescence intensity from nirButterfly versus Archon1 [$N=26$ FOVs for nirButterfly, 24 for Archon1; $p<0.001$, two-sample $t$-test; Fig. 2(b), black bars].

Next, we compared GEVI brightness in cultured primary neurons as they more closely model the conditions for neurophysiological applications. The transfection of plasmids into cultured neurons often has very low transfection efficiency; hence here we used AAV transduction to express the GEVIs. We generated virus plasmids encoding either nirButterfly or Archon1 using an identical virus plasmid backbone. Neuronal cultures of similar seeding density were transduced side-by-side with AAV1.CaMK2A-nirButterfly or AAV1.CaMK2A-Archon1 at similar virus titers [Fig. 2(c)]. The size of the nirButterfly gene is close to the maximal AAV packaging size; therefore, we could not introduce an additional EGFP tag into the AAV cargo plasmid to perform the normalization step as described in Figs. 2(a) and 2(b). Instead, here we measured the absolute fluorescence intensity of the transduced cultures, and similar to HEK293T cell cultures [Figs. 2(a) and 2(b)], in neurons [Fig. 2(c)] we observed an approximately sixfold higher fluorescence intensity from nirButterfly than Archon1, both with and without the additional supplements of their respective chromophores [biliverdin for nirButterfly, all-trans-retinal for Archon1; both are present naturally in the mammalian brain; $N=10$ FOVs for with chromophore supplements, $p<0.001$, two-sample $t$-test; and $N=5$ FOVs for without chromophore supplements, $p<0.001$, two-sample $t$-test; Fig. 2(d)].

Then, we transduced pyramidal neurons in the hippocampus of wild-type mice using AAV1.CaMK2A-nirButterfly or AAV1.CaMK2A-Archon1 at similar virus titers. After 3 weeks of expression, we made coronal brain slice preparations and imaged them under epifluorescence imaging conditions [Figs. 3(a) and 3(b), left]. Fluorescence was observed from the cell body in the hippocampal stratum pyramidale and from the intermingled processes in the oriens and radiatum. For slices transduced with AAVs delivering Archon1, approximately five times higher illumination intensity was needed to achieve near comparable fluorescence intensities for functional imaging [Fig. 2(d)].

Fig. 3

GEVI sensitivity and kinetic comparisons in acute brain slices. (a) (left) AAV-transduced nirButterfly expression in hippocampal pyramidal neurons, and (right) representative population response in CA3 following Schaffer collateral stimulation under standard ACSF, Gabazine-ACSF, and post-Gabazine recovery conditions (10-trial average). The dotted circle outlines the region of photon sampling for the photodiodes. The position of the stimulation electrode was outside the imaging field of view. (b) Same as in (a) for Archon1. Traces were acquired with five times higher illumination intensity. (c) GEVI sensitivity was measured as the population peak response amplitude normalized to baseline fluorescence for following 1-pulse ($p=0.024$, two-sample $t$-test with unequal variance) and five-pulse synaptic stimulation ($p=0.002$, two-sample $t$-test with unequal variance). (d) GEVI kinetic properties are measured as the decay time constants following 1-pulse ($p=0.002$, two-sample $t$-test with unequal variance) and 5-pulse synaptic stimulation ($p=0.003$, two-sample $t$-test with unequal variance). nirButterfly: $N=8$ slices from 3 mice; Archon1: $N=11$ slices from 3 mice. Data presented in median ± SEM; * = $p<0.05$. ** = $p<0.01$.

3.2.

Archon1 Displays Higher Sensitivity but Slower Kinetics than nirButterfly

Next, we wanted to compare the functional performance of these two GEVIs in brain slices. Here, instead of single cell-level comparisons (Fig. 2), we used epifluorescence population imaging to effectively average across two cardinal factors that are notorious for affecting indicator performance: cell size variability and indicator expression variability; both factors strongly influencing the signal-to-noise ratio (SNR) in voltage imaging traces.

We stimulated Schaffer collaterals that project from CA3 to CA1 and imaged the population responses in CA1 using a photodiode that samples over the region of interest at 10 kHz. Only the donor channel ($676/37\mathrm{nm}$) was recorded for nirButterfly. Given that the emission detection used in our epifluorescence imaging setup transmits a comparable fraction of the emitted fluorescence of both nirButterfly and Archon1 [Figs. 1(b) and 1(d)], we used the identical optical configuration to measure the functional emission of Archon1.

Both GEVIs reported population optical signals in CA1 following single pulse Schaffer collateral stimulation and five pulses at 50 Hz, with a good SNR [Figs. 3(a) and 3(b), right]. With the addition of the ${\mathrm{GABA}}_{\mathrm{A}}$ receptor antagonist gabazine to the perfusate, these population voltage responses increased in amplitude when using both GEVIs, and the response decay rate slowed down, resulting in fused responses to the five pulses. After recovery from gabazine, the responses became shorter-lasting again and more separated in the case of the 5-pulse stimulation.

We also performed population level imaging experiments of synaptic stimulation with the GEVIs being expressed in the cortex. We observed similar optically reported response amplitude for both indicators and therefore pooled the functional data. Quantifications of these population responses were used to assess the sensitivity and kinetics of the GEVI performances in mouse brain tissue. For sensitivity, the absolute peak amplitude for Archon1 was approximately twofold larger than nirButterfly [Fig. 3(c)] for both responses to single pulse synaptic stimulation (Archon1: $0.59\pm 0.15\%$ $\mathrm{\Delta }F/F$, nirButterfly: $0.34\pm 0.03\%$ $\mathrm{\Delta }F/F$, $N=11$ slices from 3 mice for Archon1 and 8 slices from 3 mice for nirButterfly, median ± SEM) and 5-pulse stimulation (Archon1: $1.56\pm 0.36\%$ $\mathrm{\Delta }F/F$, nirButterfly: $0.58\pm 0.05\%$ $\mathrm{\Delta }F/F$, $N=11$ and 8, respectively, median ± SEM).

We also quantified the decay time constant of these responses as a kinetics comparison between the two indicators. The response decay for Archon1 was approximately four times slower than that of nirButterfly [Fig. 3(d)], for both single pulse (Archon1: $87.00\pm 24.37\mathrm{ms}$, nirButterfly: $17.50\pm 3.15\mathrm{ms}$; Tau of a single exponential fit, $N=11$ and 8, respectively, median ± SEM) and 5-pulse stimulation (Archon1: $110.00\pm 27.56\mathrm{ms}$, nirButterfly: $15.50\pm 1.11\mathrm{ms}$; Tau of a single exponential fit, $N=11$ and 8, respectively, median ± SEM). This slower response decay has been noted in previous work (res) and may also contribute to the larger peak amplitude of Archon1 signals observed at sampling rates $<500\mathrm{Hz}$. A slower decay also can cause larger GEVI signal amplitudes with repetitive stimulation. These results were surprising given the faster kinetics generally declared for opsin-based GEVIs, but the potential mechanisms underlying this feature of Archon1 require further elucidation.