Neuron Activity Dependent Redox Compartmentation Revealed with a Second Generation Red-Shifted Ratiometric Sensor

Abstract

Oxidative stress is a hallmark of several aging and trauma related neurological disorders, but the precise details of how altered neuronal activity elicits subcellular redox changes have remained difficult to resolve. Current redox sensitive dyes and fluorescent proteins can quantify spatially distinct changes in reactive oxygen species levels, but multicolor probes are needed to accurately analyze compartment-specific redox dynamics in single cells that can be masked by population averaging. We previously engineered genetically-encoded red-shifted redox-sensitive fluorescent protein sensors using a Förster resonance energy transfer relay strategy. Here, we developed a second-generation excitation ratiometric sensor called rogRFP2 with improved red emission for quantitative live-cell imaging. Using this sensor to measure activity-dependent redox changes in individual cultured neurons, we observed an anticorrelation in which mitochondrial oxidation was accompanied by a concurrent reduction in the cytosol. This behavior was dependent on the activity of Complex I of the mitochondrial electron transport chain and could be modulated by the presence of co-cultured astrocytes. We also demonstrated that the red fluorescent rogRFP2 facilitates ratiometric one and two-photon redox imaging in rat brain slices and Drosophila retinas. Overall, the proof-of-concept studies reported here demonstrate that this new rogRFP2 redox sensor can be a powerful tool for understanding redox biology both in vitro and in vivo across model organisms.

Graphical Abstract

INTRODUCTION

Reactive oxygen species (ROS) are central mediators of redox signaling within and across physiological compartments^1,2. However, the excessive presence of ROS can lead to oxidative stress that damages cellular components and causes dysfunction and cell death^3,4. To provide a healthy redox balance, cells have evolved antioxidant mechanisms that regulate intracellular ROS levels, including antioxidant proteins (e.g. superoxide dismutases, catalases, peroxidases) and small molecules (e.g. glutathione, coenzyme Q10)⁵. Thus, oxidative stress can be caused by an unbalanced increase in ROS production or an impairment of antioxidant machinery⁶.

Oxidative stress has been linked to several neurodegenerative processes such as in Alzheimer’s and Parkinson’s disease^7–9. Even though increased ROS levels might not directly trigger neurodegeneration, they are ancillary to other critical disturbances like mitochondrial dysfunction and glutamate excitotoxicity that propagate neuronal damage^9–12. Evidence of oxidative damage induced by excessive glutamate is also observed in epilepsy and ischemic stroke, and oxidative stress induced neuronal death has been rescued with the use of antioxidants¹³. In particular, the high metabolic demand and abundance of mitochondria make the brain especially vulnerable to redox imbalance and oxidative stress^14–16. Mitochondria play key and sometimes contradictory roles, helping to regulate metabolic burden by the detoxification of neurotransmitters and utilization of metabolic substrates¹⁷ while also acting as a major source of ROS via the electron transport chain¹⁸. Therefore, much remains to be understood about redox relationships between neuronal activity and mitochondria.

For example, within an individual neuron we have yet to understand how redox dynamics between the mitochondria and different compartments interact because it has remained technically challenging to quantitatively correlate spatially distinct redox dynamics with subcellular resolution. To this end, genetically encoded fluorescent sensors can be targeted to specific subcellular locations to study the redox changes in different compartments during live-cell imaging^19–22. Furthermore, recent developments of red fluorescent redox sensors have also made multicolor imaging possible. However, the majority of these sensors report redox dynamics with changes in a single fluorescence intensity channel, and there remains a need to develop multicolor ratiometric redox sensors^23–29. Ratiometric sensors are valuable for live-cell imaging because they are self-normalizing and therefore independent of expression level, facilitating quantitative comparisons between individual compartments, cells, and experiments³⁰. The roGFP sensors developed by Hanson et al. are prime examples of excitation ratiometric redox sensors that work robustly in live cells²¹. Previously, we developed a novel sensor design that uses a Förster resonance energy transfer (FRET) relay to extend the emission profile of roGFP out to red wavelengths by attaching a red fluorescent protein (RFP) FRET acceptor without altering the desirable redox sensing properties of the roGFP itself³¹.

Here, we now report the development of a second-generation sensor, called rogRFP2, in which the FRET relay has been optimized through linker and RFP donor screening to further enhance the red fluorescent signal. With this improved sensor we were able to simultaneously measure activity-dependent redox dynamics in the cytosol and mitochondria of individual cultured neurons upon glutamate stimulation. We were also able to demonstrate the sensor’s practical use for ratiometric one- and two-photon redox imaging of thick samples, including rodent brain slices and fly eyes.

RESULTS

rogRFP2 Protein Engineering

Our first-generation FRET-relay redox sensors were designed with an roGFP FRET donor fused to an RFP FRET acceptor using a seven amino acid linker (Figure 1). Unlike most FRET sensors that exhibit dynamic changes in FRET efficiency, in our design the FRET from roGFP to RFP is constant, and the ratiometric response originates from the roGFP donor. With our first-generation seven amino acid roGFP-RFP linker we were able to achieve a 20% FRET efficiency without compromising the redox properties or the ratiometric response in the excitation spectrum, and this sensor architecture provided sufficient red fluorescent redox-dependent signal for dual-compartment imaging³¹. In this study, we attempted to improve the FRET efficiency further by optimizing the roGFP-RFP linker to reduce the distance between the fluorescent proteins (FPs) without adversely causing suboptimal orientations or steric clashes.

Figure 1.

FRET relay design. roGFP2 is the FRET donor, and a library of linkers and RFP acceptors were screened for improved red emission.

We generated and screened a combinatorial library of different linker lengths and RFP FRET acceptors. Our initial proof-of-concept studies indicated that using roGFP2 as the redox-sensitive FRET donor achieved larger signal dynamic ranges compared to roGFP1 fusions³¹, so we designed our second-generation library with roGFP2 as the donor. We combined the roGFP2 donor with three different RFPs, mRuby2, mApple, or mCherry, because they each exhibit different peak emission bands and in theory can still achieve high FRET efficiencies. We also varied the order of fusion, with the RFP fused to either the N- or C-terminus of roGFP2 via the linker. Finally, we varied the roGFP2-RFP linker by shortening the artificial peptide linker sequence as well as altering the native fluorescent protein termini. To alter the native termini, we tested deletions of up to eleven amino acids from flexible regions identified from X-ray crystal structures.

In total, we were able to generate, express, and purify 28 novel constructs (Supporting Table S1), which we screened by solution studies using time-resolved and steady-state spectroscopy (Figure 2). We used time-resolved spectroscopy because the decrease in donor fluorescence lifetime is a robust metric for measuring FRET efficiency. We paired this with steady-state spectroscopy to measure the signal dynamic range, which we define as the magnitude of the ratiometric change in the excitation spectrum of the red emission channel upon going from the fully reduced to the fully oxidized state. All of the library constructs exhibited a decrease in donor lifetime relative to the parent roGFP2 control as well as a corresponding sensitized red emission peak upon donor excitation. Most constructs also showed an improvement in FRET efficiency, but interestingly there were no clear patterns of dependency on linker length, choice of RFP acceptor, or the order in which the two proteins were fused. Likewise, all constructs exhibited good dynamic ranges for use, but as expected dynamic ranges were diminished relative to the parent roGFP2 control because of less than perfect FRET and spectral crosstalk³¹.

Figure 2.

Linker and RFP acceptor library screen. FRET efficiency was calculated from the change in donor fluorescence lifetime relative to the roGFP2 control. Dynamic range was calculated as the fold difference of the F₄₀₅/F₄₈₅ ratio measured in the presence of 10 mM DTT or 10 mM Oxidized DTT H₂O₂. N.D., not determined.

The top performing library construct (“g0GG3a”) exhibited an improved FRET efficiency of 34% and dynamic range with a maximal 5-fold change in excitation ratio (Supporting Figure S1), and we renamed this variant “rogRFP2” as our second-generation sensor. This new rogRFP2 sensor contains roGFP2 with an 11 amino acid truncation of its C-terminus fused to a diglycine linker that bridges to mApple with a 9 amino acid truncation of its N-terminus. Solution characterization clearly shows a large 1 ns decrease in donor lifetime, which manifests as a significantly increased sensitized mApple red emission and suppressed roGFP2 green emission in the steady-state spectra (Figure 3A,B). Furthermore, the midpoint potential remains unperturbed relative to the roGFP2 parent (Figure 3C)^21,31,32, and fusion to the mApple FRET acceptor does not significantly inhibit responsiveness to redox changes when expressed in live HEK293 cells when compared side-by-side with roGFP2 (Supporting Figure S2)

Figure 3.

Characterization of the second generation rogRFP2 sensor. (A) Donor fluorescence lifetime of rogRFP2 (red) is significantly decreased compared to roGFP2 donor control (green). Fluorescein is shown for comparison (grey). (B) Fluorescence emission spectra of rogRFP2 (red) exhibits significantly improved red emission and suppressed green emission relative to the roGFP2 control (green) and first-generation sensor (grey). (C) The midpoint potential of rogRFP2 is −287.8 ± 0.4 mV (mean ± s.e.m., n=3). (D) Mitochondrial targeting of rogRFP2 (red) is efficient when co-expressed with cytosolic roGFP2 (green) in primary mouse hippocampal neurons. Scale bar = 25 μm.

Despite incomplete suppression of the donor green fluorescence, we also validated that rogRFP2 could be efficiently targeted to the mitochondria of cultured primary mouse hippocampal neurons. Previously, we found that our first-generation sensors could be targeted well to the mitochondria of HEK293 and Neuro2A cultured cell lines, and the red emission was sufficient to spatially separate signals by subcellular localization³¹. Likewise, confocal imaging demonstrated that the second-generation rogRFP2 sensor targeted well to the mitochondria of primary neurons (Figure 3D) (Supporting Figure S3). Furthermore, in spite of less than perfect FRET efficiency, the improved red fluorescence from mitochondrially-targeted rogRFP2 was easily distinguished from the green fluorescence of co-transfected cytosolic roGFP2. Although the residual donor roGFP2 green fluorescence from mito-rogRFP2 is apparent at a higher intensity than the diffuse cytosolic signal from cyto-roGFP2 in the green emission channel (Figure 3D, left panel), the quantitative difference is not easily separated by simple thresholding, illustrating the need for the spectral separation achieved in the red emission channel with our sensor. Additionally, one concern could be that the rogRFP2 linker might be susceptible to cleavage in cells. Linker cleavage could potentially lead to a diminished dynamic range caused by increased background from spectral crosstalk, but notably it would not lead to a change in the intrinsic redox response because the ratiometric measurement self-normalizes for changes in sensor concentration. Linker cleavage would manifest as a reduction in the average FRET and therefore cause a decrease over time in the acceptor-to-donor red-to-green emission ratio with donor excitation. However, on the timescale of live-neuron imaging experiments in this study, the red-to-green emission ratio was stable, indicating that linker cleavage is not a significant problem (Supporting Figure S4). Therefore, as a proof-of-concept for multicolor imaging applications, we next used this system to study neuron activity-dependent redox changes at the single-cell and compartment-specific levels.

Neuron Activity-Dependent Redox Dynamics

Extracellular levels of glutamate in the brain are regulated and maintained at low levels to avoid excitotoxic effects of glutamate^33,34. In primary neuron cultures from rodent models, acute treatment with excitotoxic levels of glutamate (100 μM) causes changes in ROS levels in different localized regions of neurons³⁵. However, it is unclear if there are differences in redox responses between compartments of an individual cell, and to what extent acute redox changes occur in response to less extreme, sublethal glutamate stimulation that approaches more physiological ranges. We therefore used cultured primary mouse hippocampal neurons co-transfected with cytosolic roGFP2 (cyto-roGFP2) and mitochondrially-targeted rogRFP2 (mito-rogRFP2) to resolve these questions more clearly.

Interestingly, glutamate stimulation caused opposite redox responses in the cytosol and mitochondria as reported by the targeted sensors (Figure 4). In our preliminary studies, stimulation with varying levels of glutamate (5–25 μM)^36,37 or potassium chloride (30–60 mM) elicited similar responses, and therefore all subsequent experiments were carried out using 10 μM glutamate stimulation. At this level of stimulation, neurons maintained healthy morphology throughout the experiment, which included a sensor calibration phase. Neurons were imaged in an artificial cerebrospinal fluid imaging solution at room temperature, and after a baseline period glutamate was added to the bath³⁸. Following glutamate stimulation, the sensors were calibrated by 1mM diamide addition to fully oxidize the sensors followed by 5mM dithiothreitol (DTT) addition to fully reduce the sensors. Using the calibration on a cell-by-cell basis, ratio signals could be converted to the percent oxidation level of cyto-roGFP2 and mito-rogRFP2^30,31. At the population level, glutamate stimulation consistently caused an oxidation of the mitochondria (Figure 4D) accompanied by a concurrent reduction of the cytosol (Figure 4A). Specifically, glutamate stimulation caused significant 25% ± 4.6 (Mean + 95% CI) increase in mito-rogRFP2 oxidation relative to the baseline (n=4 experiments, 41 neurons; p=6×10⁻¹², paired t-test). In contrast, there was a significant −11% ± 1.8 (Mean + 95% CI) reduction of cyto-roGFP2 relative to its baseline (n= 4 experiments, 41 neurons; p=8×10⁻³, paired t-test).

Figure 4.

Neuron activity-dependent redox compartmentation. Cultured mouse primary hippocampal neurons were stimulated with 10 μM glutamate followed by treatments with diamide and DTT to calibrate the (A-C) green fluorescent cytosolic roGFP2 and (D-F) red fluorescent mitochondrial rogRFP2. Glutamate stimulation caused (A) cytosolic reduction concurrent with (C) mitochondrial oxidation (mean ± 95% C.I.). Antioxidant treatment (“AOx”) with N-acetylcysteine and Mito-TEMPO blocked the (B,C) cytosolic and (E,F) mitochondrial redox responses.

We also validated that the sensor responses were specific to cellular redox state. When neurons were treated with a cocktail of antioxidants (“AOx”) containing 0.5 mM N-acetylcysteine and 5 μM Mito-TEMPO^39–43, glutamate stimulation caused a much smaller 3.9% ± 1.5 (Mean + 95% CI) oxidation of mito-rogRFP2 relative to baseline (n=3 experiments, 65 neurons; p= 0.05, paired t-test) (Figure 4E,F), and the −1.8% ± 1.2 (Mean + 95% CI) change in cyto-roGFP2 was not significant (n=3 experiments, 65 neurons; p= 0.4, paired t-test) (Figure 4B,C). It is also possible that an activity-dependent pH change could cause the response because the redox potential is intrinsically dependent on pH^20–21. However, glutamate stimulation did not cause pH changes that could account for sensor responses (Supporting Figure S5). Thus, glutamate stimulation primarily causes changes in ROS levels that likely alter the glutathione redox balance detected by the roGFP2 sensing module.

Single-Cell Analysis of Redox Compartmentation

Our population-level analysis revealed a novel phenomenon in which neuronal activity induces mitochondrial oxidation and cytosolic reduction on average. However, population averaging masks the heterogeneity of individual cell behaviors that may be physiologically important. For example, the mean population responses measured in Figure 4 cannot distinguish whether mitochondrial oxidation and cytosolic reduction occurs within individual neurons or between different populations of neurons.

To better understand the origins of the population behavior, we further scrutinized single-cell responses and found that the anticorrelated cytosolic and mitochondrial redox responses observed at the population level are a reflection of anticorrelated responses within the individual neurons. For example, Figure 5 illustrates the single neuron responses from one of the four independent experiments summarized in Figure 4. The exact time course trajectories of individual neurons were heterogeneous and in some cases were multiphasic in nature (Figure 5C), but in general the response of individual neurons recapitulated the trend seen in the population average (Figure 5D,E) (Supporting Figure S6). Glutamate stimulation caused significant 28% ± 10 (Mean + 95% CI) increase in mito-rogRFP2 oxidation relative to the baseline (n=10 neurons; p=1×10⁻⁵, paired t-test) for this representative experiment. In contrast, there was a significant −10% ± 4.8 (Mean + 95% CI) reduction of cyto-roGFP2 relative to its baseline (n=10 neurons; p=1×10⁻⁴, paired t-test) (Figure 5D,E).We can visualize this trend with a two- dimensional plot in which the fold change in cyto-roGFP2 and mito-rogRFP2 serve as coordinates for an individual neuron (Figure 5F). With glutamate stimulation alone, all cells fall in the upper left quadrant, exhibiting mitochondrial oxidation paired with cytosolic reduction relative to baseline.

Figure 5.

Single-cell analysis of activity-dependent redox compartmentation for a representative experiment. The population averaged redox responses (dark shaded lines) in the (A) cytosol (green) and (B) mitochondria (red) reflect the behavior of individual neurons (light shaded lines). (C) Individual neurons exhibit anticorrelated compartment-specific responses despite single-cell heterogeneity. Glutamate stimulation caused a significant (D) reduction in the cytosol paired with (E) oxidation in the mitochondria. (F) Maximal fold change in mito-rogRFP2 oxidation versus cytoroGFP2 oxidation for all neurons (positive, oxidation; negative, reduction).

Thus, the new red fluorescent rogRFP2 sensor has enabled us to observe previously unappreciated neuron activity-dependent redox compartmentation that arises as a stereotyped trend amongst the heterogeneity of single cell behavior. Although a detailed mechanistic study is beyond the scope of this work, we next sought to demonstrate that this system provides a powerful assay platform to further dissect molecular and cellular aspects of redox homeostasis.

Rotenone Treatment

Rotenone is a selective inhibitor of complex I of the mitochondrial respiratory chain^44,45, and it is also a pesticide that is used as a model for environmental toxins associated with Parkinsonism^46,47. One of the main sources of cellular ROS is from the electron transport chain, and therefore we hypothesized that rotenone inhibition of Complex I would modulate the compartment-specific neuron activity-dependent redox responses we observed¹². In fact, rotenone treatment caused a drastic change in phenotype such that glutamate stimulation no longer caused a significant redox response in either the cytosol or mitochondria (Figure 6).

Figure 6.

Rotenone attenuates activity-dependent redox responses in both compartments. Neurons treated with 100 nM rotenone did not exhibit significant (A, C) cytosolic or (B,D) mitochondrial redox responses (mean ± 95% C.I.) with glutamate stimulation; however, the cytosolic responses of individual neurons exhibited a trend towards oxidation (A,C,E).

In neurons treated with 100 nM Rotenone (n=4 experiments, 51 neurons), glutamate stimulation caused a 2.0% ± 1.9 (Mean + 95% CI; p=0.44, paired t-test) increase in mito-rogRFP2 oxidation relative to the baseline and a 3.8% ± 1.4 (Mean + 95% CI; p=0.25, paired t-test) reduction of cytoroGFP2 relative to its baseline (Figure 6), but neither of these were statistically significant. However, the cytosolic response of rotenone-treated neurons exhibited a trend towards glutamate-stimulated oxidation (Figure 6A), which contrasts with the reduction observed in untreated neurons. This trend is reflected in the mito-rogRFP2 versus cyto-roGFP2 plot (Figure 6E) in which the distribution single-cell responses of rotenone-treated neurons is shifted to the right compared to untreated neurons. Notably, the neurons in these imaging experiments maintained healthy morphology, and both the cyto-roGFP2 and mito-rogRFP2 sensors responded appropriately to the calibration protocol.

The effect of rotenone suggests the activity of mitochondrial Complex I is crucial in shaping the activity-dependent redox response of neurons. Acute rotenone treatment was used to inhibit mitochondrial electron transport, and the reversal of phenotype in the face of inhibition may suggest that the cytosolic response is dependent on the mitochondrial response. In the future, determining how these phenotypes evolve at different timescales will provide greater insight into the mechanisms underlying the intracellular redox coupling we have observed. In the future, it may also be interesting to distinguish the acute effects of rotenone on cellular bioenergetics from chronic effects of extended rotenone treatment given the use of rotenone and other mitochondrial neurotoxins in models of Parkinsonism and other neurodegenerative diseases^46,47. To these ends, in this work we have established that the rogRFP2 sensor provides a tractable system for pharmacological studies in the future.

Neuron-Astrocyte Co-Culture

Redox homeostasis in the brain involves both intracellular and intercellular mechanisms, and in particular astrocytes provide redox support to neurons in many ways. For example, astrocytes play a crucial role in buffering neuronal oxidative stress under excitotoxic conditions by sequestering excess glutamate^48,49, astrocytes activate antioxidant pathways in neurons⁵⁰, and astrocytes exhibit a neuroprotective effect in the face of oxidative challenges^51,52. We therefore asked how the presence of astrocytes would affect the acute redox response of neurons in co-culture.

Interestingly, the presence of astrocytes significantly attenuated the magnitude of mitochondrial oxidation upon glutamate stimulation compared to neuron monocultures, but the cytosol still experienced a reduction compared to baseline (Figure 7). In neurons cultured on an astrocyte feeder layer, glutamate stimulation caused a 5.6% ± 6 (Mean + 95% CI) increase in mito-rogRFP2 oxidation relative to the baseline (n=3 experiments, 37 neurons; p=0.10, paired t-test) which was not statistically significant. In contrast, there was a significant −9.26% ± 1.4 (Mean + 95% CI) reduction of cyto-roGFP2 relative to its baseline (n= 3 experiments, 37 neurons; p=5×10⁻⁵, paired t-test) (Figure 7C, D), which is comparable in magnitude to the cytosolic response of neurons in monoculture. Notably, these neurons exhibited similar morphology to neuron monocultures, and the cyto-roGFP2 and mito-rogRFP2 sensors responded equivalently to calibration. These results are consistent with previous literature on antioxidant roles of glia, and our system reveals an important new facet to the subcellular nature of how astrocytes modulate neuronal redox biology in a way that is now experimentally accessible in future studies.

Figure 7.

Astrocytes attenuate mitochondrial oxidative stress in neurons. The presence of co-cultured astrocytes does not affect the (A,C) cytosolic reduction observed after glutamate stimulation, but (B,D,E) mitochondrial oxidation was significantly attenuated.

Thus far, we have demonstrated that the rogRFP2 redox sensor is an effective new tool for multiplexed ratiometric imaging using cultured primary neurons. Separately, we also wanted to demonstrate proofs-of-concept that the rogRFP2 sensor could be used with thick samples from both vertebrate and invertebrate model organisms.

Two-Photon Brain Slice Imaging

The green fluorescent roGFP-based sensors have been used successfully with two-photon microscopy, and therefore we tested whether the two-photon excitation profile is preserved in rogRFP2^39,53–55. As predicted by the one-photon spectra, purified protein samples exhibited a ratiometric response upon oxidation, and for all tested two-photon excitation wavelengths, a strong red fluorescence emission peak was observed at 600 nm despite incomplete suppression of the donor green emission peak (Supporting Figure S7).

Given the promising in vitro results, we sought to test whether rogRFP2 is suitable for ratiometric redox imaging in situ using acute brain slices of rogRFP2 expressing rats. Neuronal expression was induced by adeno-associated virus (AAV) delivery with viruses containing the rogRFP2 construct under the hSyn promoter targeted to the cytosol. Neonatal rats were sacrificed 2–3 weeks after AAV injections and acute brain slices were prepared for live two-photon microscopy of neurons in cortical layers II-IV. Ratiometric signals were recorded using two alternating channels with excitation wavelengths of 800 and 900 nm to quantify how neurons respond to bath applications of diamide and DTT.

Transient bath application of diamide induced a robust increase in rogRFP2 ratios in the red fluorescence emission channel (+76.7 ± 15%, ** p<0.01, n=4 slices) (Figure 8A, B). As a control we also observed the expected increase in the residual green fluorescence ratio (+290 ± 10%, ****p < 0.0001, n=4 slices) to validate the response originates from the donor roGFP2 (Supporting Figure S8). A subsequent application of DTT reduced the neuronal cytosols back to baseline. However, the reduction induced by DTT did not lower the ratio below the baseline, indicating that the cytosol of neurons in situ are almost fully reduced at rest. Occasionally individual neurons displayed extremely high neuronal overexpression of rogRFP2 that caused redox insensitive bright artifacts in the red emission spectrum (Figure 8A, arrow). These scattered cells were easily recognized and could be excluded for data analysis. Other promoters instead of the strong hSyn promoter may reduce this occurrence.

Figure 8.

Two-photon imaging in acute rat brain slices. (A) Ratio images of cortical neurons in Layers II-IV expressing rogRFP2 were collected with alternating 800 and 900 nm excitation, and responses to diamide and DTT were recorded. Excessive overexpression of rogRFP2 sometimes caused puncta to form (arrow), and these neurons were excluded from analysis. Scale bar = 10 μm. (B) Diamide caused a significant increase in oxidation from baseline that could be reversed with DTT in situ (BL, baseline). (C) DTT did not cause a significant reduction from baseline in resting neurons.

We further repeated this experiment but applied DTT before diamide to exclude the possibility that diamide remaining in the bath was decreasing the efficiency of DTT in reducing rogRFP2. The application of DTT first in order did not cause a significant change in the red fluorescence ratio (+0.5± 1.6%, p=0.93, n=4 slices) (Figure 8B), and only a small, though statistically significant, reduction was observed in the residual green fluorescence ratio (−3.6 ± 1%, * p=0.028, n=4 slices) (Supporting Figure S8).

Taken together, these results suggest that the cytosolic redox potential of neurons at rest in situ are more reduced than when cultured in vitro. This may reflect the increased number and close physical proximity of astrocytes around neurons in brain slices compared to cell culture. We conclude that rogRFP2 is suitable for ratiometric redox imaging in situ using two-photon excitation, and this will enable the mechanisms of redox homeostasis to be studied in greater detail in the future.

Fly Retina Imaging

We also sought to determine if the new rogRFP2 sensor could be used for redox imaging in Drosophila melanogaster. Although the roGFP-based sensors have been used for in vivo imaging of fly larvae, they have limited use for imaging the eye because autofluorescence in the green emission band strongly interferes with sensor signals, particularly with 400 nm one-photon excitation (Supporting Figure S9)^56–59. To test whether the red emission from rogRFP2 circumvents this problem, we generated transgenic flies expressing rogRFP2 in the cytosol of photoreceptor cells. These flies express rogRFP2 in the outer photoreceptor cells (R1 – R6) under control of the Rhodopsin-1 (ninaE) promoter, and are in a white-eyed background. We then performed one-photon confocal microscopy for ratiometric imaging of freshly dissected adult eyes from young flies treated with diamide or DTT. Red fluorescence from the rogRFP2 sensor was easily detectable, and as expected the ratio signal was much larger in magnitude for diamide exposure compared to full reduction with DTT (p=9×10⁻⁴, n=6) (Figure 9). Thus, the rogRFP2 sensor is a versatile tool that can be used in both rodent and fly samples of various formats.

Figure 9.

Ratiometric imaging in Drosophila retina. Eyes from adult transgenic flies expressing rogRFP2 in photoreceptor cells were dissected and imaged using one-photon confocal microscopy. Ratio images overlaid on DIC images show responses to (A) diamide and (B) DTT. (C) Eyes exposed to diamide are significantly more oxidized than those exposed to DTT (n=6). Scale bar, 100 μm.

DISCUSSION

Ratiometric live-cell imaging provides a powerful approach for quantitative redox biology, and we have engineered the second-generation red fluorescent sensor rogRFP2 for this purpose. We demonstrated that rogRFP2 can be used for one- and two-photon ratiometric imaging of a variety of model species and sample formats, including cultured primary mouse neurons, acute rat brain slices, and adult drosophila eyes, making it a versatile tool. Our FRET relay design has the advantage that it preserves the desirable redox sensing properties of the donor roGFP2, one of the mostly widely used genetically encoded fluorescent sensors. Furthermore, our design also preserves the ratiometric response imparted by the roGFP2 donor, which facilitates improved quantitation across compartments, cells, and experiments. For comparison, the other currently available red fluorescent redox sensors, including the rxRFPs, roCherry, rxmRuby2, roKate, and HyPerRed, report single-channel intensity changes that vary in dynamic range from ~1.5 to 5.5-fold maximal changes^{23–25,27–29}.With an increase in FRET efficiency and red emission of ~70% compared to our first-generation sensors, the rogRFP2 sensor exhibits a maximal 5-fold ratio change and has a good dynamic range for in situ imaging of challenging thick specimens. More broadly our work also demonstrates that ratiometric sensors can exhibit large dynamic ranges that can match those of “turn on” sensors, and this is promising for future engineering efforts that may be able to achieve even higher FRET efficiencies through modeling and rational design. Taking advantage of the improved red fluorescence, we then used the rogRFP2 sensor to observe novel redox biology in primary neurons.

To this end, the red fluorescent rogRFP2 sensor facilitated multiplexed ratio imaging of cultured primary mouse neurons and revealed unique activity-dependent redox responses that were compartmentalized between mitochondria and cytosol. We observed that glutamate stimulation caused a cytosolic reduction concurrent with mitochondrial oxidation at the single-cell level, which has not been previously observed to the best of our knowledge. Both rotenone treatment and co-culture with astrocytes modulated the redox response as expected, and inhibition by the mitochondrial toxin rotenone suggests that the cytosolic reduction is dependent on the mitochondrial oxidation. In particular, blockade by rotenone indicates the activity of mitochondrial Complex I is a major contributor to the activity-dependent redox compartmentation, which agrees with the plethora of evidence suggesting that the electron transport chain is a major source of ROS production, likely caused by calcium overload and increased mitochondrial respiration in the face of increased neuronal activity^45,60. However, our observation of a cytosolic reduction indicates that ROS generated in mitochondria are not simply propagated to the cytosolic compartment even though glutamate excitotoxity is broadly associated with neuronal oxidative stress and damage. The cytosolic reduction we observed for the first time here may contribute to a protective response, and our new sensor will enable us to study this level of subcellular redox homeostasis in greater mechanistic detail in the future.

MATERIALS AND METHODS

Reagents.

Chemicals and cell culture media and supplements were purchased from Sigma, Formedium, and ThermoFisher Scientific. Molecular biology enzymes were purchased from New England Biolabs (NEB).

Molecular Biology.

Plasmids with altered linker length were generated using Q5 mutagenesis of the first-generation constructs. The pRSETB vector was used for bacterial expression, and the GW1 vector was used for mammalian expression. Four tandem repeats of the COX8 signal sequence for used for mitochondrial targeting.

Protein Expression and Purification.

The sensor library was expressed in E. Coli strain BL-21 and cultured with shaking in Auto Induction Media (AIM) for 12 hours at 37°C followed by 24–48 hours at room temperature. Cells were then pelleted by centrifugation, resuspended in lysis buffer (25 mM Sodium Phosphate Buffer, pH 7.8, 500 mM NaCl, 10 mM Imidazole, 5% Glycerol, 0.25 mg/ml lysozyme, 0.1% Triton-X, 1 mM PMSF, 5 mM DTT) and lysed by sonication. Crude lysate was cleared by centrifugation at 30,000xg for 30 minutes at 4˚C, and passed through a 0.45 μm syringe filter. Histidine-tagged proteins were purified by nickel-affinity chromatography (GE Chelating Sepharose) using gradient elution from 10–500 mM Imidazole in elution buffer (25 mM Sodium Phosphate Buffer, pH 7.8, 500 mM NaCl, 5% glycerol). Purified protein was dialyzed in storage buffer (50 mM MOPS, 750 mM NaCl, 10% glycerol) at 4°C to remove excess imidazole and stored in the −80°C until use.

Steady-State Fluorescence Spectroscopy.

To determine concentration, the protein was denatured using 1M NaOH and the chromophore absorbance was measured at 447 nm using and extinction coefficient of 44,000 M⁻¹cm⁻¹ while accounting for the number of chromophores. The protein was diluted to a final concentration of 1 μM in 75 mM HEPES, 125 mM KCl, 1 mM EDTA, pH 7 for the spectroscopy and redox measurements. In redox titration experiments, the solutions were degassed under vacuum and purged with argon gas, and protein solutions were equilibrated with 10 mM reduced DTT (1,4-dithiothreitol) or 10 mM oxidized DTT (trans-4,5-dihydroxy-1,2-dithiane) for 1 h before measurements. Redox titrations carried out in solutions in which the total DTT concentration was held constant at 10 mM while the reduced DTT to oxidized DTT ratio was varied. Midpoint potentials were determined as previously described and by fitting titration data to a Boltzmann equation. UV-vis and steady-state fluorescence spectroscopy was performed using a BioTek Synergy H4 microplate reader at room temperature.

Two-photon photoluminescence measurements were performed by employing a home-built confocal PL microscope. The output of a high-repetition rate amplifier (Pharos Light Conversion, 400 kHz) was used to pump an optical parametric amplifier (OPA), which yielded the various excitation wavelengths used for the measurements. The excitation wavelengths ranged from 800 nm to 960 nm, in 20 nm increments. The laser beam was focused onto the sample using a 50X [(numerical aperture) NA=0.95] objective. The PL emission was collected with the same objective, optically filtered to remove residual pump light, dispersed with a monochromator (Andor Technology) and detected using a charge coupled device (CCD) (Andor Technology).

Time-Resolved Fluorescence Spectroscopy.

Fluorescence lifetimes were measured using an Edinburgh Instruments FS5-TCSPC+ with a Fianium WhiteLaseMicro pulsed laser, and the instrument response function was measured with Ludox. Re-convolution fitting was carried out using Fluoracle software to calculate mean fitted lifetimes, and FRET efficiency was calculated as E = 1 – τ_DA/τ_D. Two-photon spectra were measured on a custom-built microscope.

Cultured Primary Mouse Neurons.

Primary hippocampal neurons were isolated from P1 FVB mice and plated on nitric acid washed 18mm #1.5 glass coverslips or glass bottomed 35 mm dishes (Cellvis) that were coated with poly-D-lysine (PDL). Neuronal cultures were maintained in Neurobasal-A medium supplemented with B-27, 5 mM glucose, 0.25 mM pyruvate, 0.5 mM GlutaMax, 1X Pen-Strep at 37°C and 5% CO₂ in a humidified incubator 30% feedings every 1–2 days. The cultured neurons were transfected on DIV 4–5 and imaged on DIV 6–8. Astrocytes were isolated from cortices of P1-P2 FVB mice and maintained in DMEM/F12 supplemented with 10% FBS until passage onto PDL coated coverslips. For neuron-astrocyte co-cultures, astrocytes were grown for 2–7 days on coverslips, and then neurons were plated on top of the feeder layer. All animal procedures were approved by and performed in accordance with guidelines by the Purdue Animal Care and Use Committee.

Neurons were equilibrated for 1 hour prior to imaging in imaging solution consisting of 15 mM HEPES, 120 mM NaCl, 3 mM KCl, 3 mM NaHCO₃, 1.25 mM NaH₂PO₄, 5 mM Glucose, 2 mM CaCl₂, 1 mM MgCl₂, 0.25 mM Pyruvate, 0.5 mM Glutamax, pH 7.3 at 37°C and 5% CO₂ in a humidified incubator. For widefield microscopy, neurons were imaged at room temperature in 1-minute time intervals and treatments were added to a static bath with imaging solution. The neurons were imaged over 120 minutes with a 10 μM glutamate addition at 20 minutes, 1 mM Diamide at 40 minutes and 5mM DTT treatment at 85 minutes. DIC images were taken at the beginning and after the imaging paradigm to monitor neuron health. For the drug treatments, the neurons were pre-incubated as following: 100 nM rotenone for 1 hour, 1 mM N-acetylcysteine for 24 hours, 500 μM N-acetylcysteine for 5 hours and 5 μM Mito-TEMPO for 1 hour, prior to imaging, and both rotenone and antioxidants were also included in respective imaging solutions during the stimulation experiment. For imaging pH changes, neurons were incubated for 20 minutes with pH sensitive dye 0.5 μM BCECF-AM⁶¹ and washed with media and imaging solution. The neurons were then imaged at baseline for 15 minutes, followed by a 10 μM glutamate treatment for 20 minutes. Subsequently, a pH calibration was performed with buffers ranging from pH 6.5–7.8 in a high-potassium solution (15 mM HEPES, 10 mM glucose, 123 mM KCl, 20 mM NaCl, 2 mM CaCl₂, 1 mM MgCl₂) in the presence of 5 μM nigericin.

They were imaged on an Olympus IX83 fluorescence microscope with a Plan Apo VC 20X objective (0.75 NA), equipped with a Prior motorized stage and an Andor Zyla 4.2 sCMOS camera (6.5 μm pixel). The images were taken with a 2×2 pixel binning and exposure times ranging from 50–400 milliseconds. The LED light source was a Lumencor SpectraX light engine and the following filter pairs (Semrock or Chroma) were used: ex. 475/34nm, em. 525/50nm; ex. 395/25nm, em. 525/50nm, ex. 475/34nm, em. 632/60nm, ex. 395/25nm, em. 632/60nm, ex. 575/25nm, em. 632/60nm. Confocal microscopy was performed on a confocal laser scanning microscope (Nikon A1R-MP, Purdue Life Sciences Imaging Facility) with a Plan Apo VC 60X oil DIC N2 objective (1.4 NA, pinhole 26.82 μm). Laser excitation at 488nm or 561nm was used with a 405nm/488nm/561nm/640nm multiband dichroic mirror and 525/50nm or 595/50nm emission filters. 512-by-512 pixel images were acquired using a Galvano scanner at 30 frames per second and collected on an A1-DU4 four detector unit with four photomultiplier tubes at a 0.41 μm, 0.41 μm, and 0.5 μm (x,y,z) pixel size dimensions.

Image stacks were analyzed with ImageJ as previously described³¹. The fluorescent images were background subtracted above the mean background signal and then a threshold was set to three standard deviations over mean to eliminate background pixels. Ratio images were generated by dividing the 400 Ex image by the 490 Ex image for both the red and green emissions and the ratio signals were measured around the cell body and neurites. The fraction of oxidized sensor was calculated according to the equation Y_oxidized = (Ratio - Ratio_reduced) / [(F_475,oxidized / F_395,oxidized)·(Ratio_oxidized - Ratio) + (Ratio - Ratio_reduced) as described in previous works^31,62. The change in oxidation over baseline was determined by comparing fold change of min/max change %oxidized after glutamate addition (T21-T39) over average %oxidized (T15-T19) before glutamate stimulation.

Acute Rat Brain Slices.

Animal care protocols were approved by the University of British Columbia’s Committee on Animal care in compliance with the Canadian Council on Animal Care guidelines. The rogRFP2 construct was packaged into an AAV2/9 under the hSyn promoter by the Neurophotonics Centre (Quebec, Canada). Neonatal Sprague Dawley rats (P1–3) were anaesthetized with isoflurane before 2 μL of virus (titer: 10¹² pfu/mL) was intracranially injected into the lateral ventricles as previously described⁶⁴ to induce neuron specific rogRFP2 expression under the hSyn promoter. Injected rat pups were allowed to recover for at least 3 weeks before they were used for experiments.

The AAV injected rats were sacrificed at an age of P17–30. Animals were anaesthetized by isoflurane before decapitation. After dissection the brain was rapidly submerged in an ice-cold slicing solution consisting of (in mM): 120 NMDG, 2.5 KCl, 25 NaHCO₃, 1 CaCl₂, 7 MgCl₂, 1.2 NaH₂PO₄, 20 glucose, 2.4 Na-pyruvate, 1.3 Na-ascorbate, and saturated with 95% O₂/5% CO₂. Coronal brain slices (350–370 μm thickness) were prepared using a conventional vibratome and stored in standard artificial cerebral spinal fluid (ACSF) consisting of (in mM): 120 NaCl, 25 NaHCO₃, 3 KCl, 1.25 NaH₂PO₄, 2 MgCl₂, 2 CaCl₂, and 10 glucose saturated with 95% O₂/5% CO₂ with a pH of 7.4. Brain slices were kept for storage and experiments at 32°C for no longer than 4 hours after slice preparation.

Acute brain slices were imaged with a 2-photon laser-scanning microscope (Zeiss LSM510-Axioskop-2, Zeiss) using a 40x objective. Cortical neurons expressing rogRFP2 were imaged at depths of 50–80 μm and alternately excited with 800 and 900 nm while green emission was captured between 500–550 (green emission) and 575–640 nm (red emission). Time series recordings were acquired while bath applying DTT (5 mM) or diamide (2 mM) after a baseline of at least 5 min. For slice experiments, every data set consisted of 8 selected cells per slice recording of a total of 3 animals each. All individual cells per recordings were averaged to one single trace. Redox ratio amplitudes were averaged at baseline and plateau phases after DTT and diamide applications and compared by using a paired t-test. Statistical error is given as the standard error of the mean based on 4 individual slice recordings each. Statistical significance was defined as * p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001.

Drosophila Retinas.

The rogRFP2 cassette from GW1-rogRFP2 was cloned as a XbaI/EcoRI fragment into pGaSpeR-ninaE-forward (gift from J. O’Tousa), and a KpnI fragment containing the ninaE 5’ and 3’ regulatory regions flanking rogRFP2 was then subcloned into pattB (DGRC #1420). w*, y¹, P{w^+mC,3xP3-DsRed=[ninaEp-rogRFP2]}attP:ZH-2A flies were generated by injection of the pattB-ninaEp-rogRFP2 plasmid into the X chromosome attP site 2A3 iny¹ M{3xP3-RFP.attP}ZH-2A w*; M{vas-int.Dm}ZH-102D (BDSC #24480) followed by Cre excision of the mini-white and 3xP3-RFP markers. Cre excision of mini-white and 3xP3-RFP was confirmed by examination of visible markers and PCR screening. Red fluorescence is visible in eyes due to mApple expression in photoreceptors.

Drosophila eyes were imaged as described in previous studies⁶³ using the Zeiss LSM 880 with Plan-Apochromatin 20x/0.8 objective. Briefly, eyes were removed from the heads of anesthetized adult flies at 6 days post-eclosion, immersed in PBS, 10 mM DTT, or 1 mM diamide for 5 min on a thin microscope slide under a bridged coverslip, and imaged as ~30×1 μm 512×512 z-stacks. Samples were excited line-by-line sequentially at 405nm and 488nm with emission at 570 – 633 nm, 4x line averaging, 16-bit depth and 2.05 μs pixel dwell.

Software.

Table of Contents graphic was created with BioRender.com.