Although nonlinear microscopy and fast (∼1 ms) membrane potential (Vm) recording have proven valuable for neuroscience applications, their potentially powerful combination has not yet been shown for studies of Vm activity deep in intact tissue. We show that laser illumination of neurons in acute rat brain slices intracellularly filled with FM4-64 dye generates an intense second-harmonic generation (SHG) signal from somatic and dendritic plasma membranes with high contrast >125 μm below the slice surface. The SHG signal provides a linear response to ΔVm of ∼7.5%/100 mV. By averaging repeated line scans (∼50), we show the ability to record action potentials (APs) optically with a signal-to-noise ratio (S/N) of ∼7–8. We also show recording of fast Vm steps from the dendritic arbor at depths inaccessible with previous methods. The high membrane contrast and linear response of SHG to ΔVm provides the advantage that signal changes are not degraded by background and can be directly quantified in terms of ΔVm. Experimental comparison of SHG and two-photon fluorescence Vm recording with the best known probes for each showed that the SHG technique is superior for Vm recording in brain slice applications, with FM4-64 as the best tested SHG Vm probe.
Measuring electrical phenomenon of excitable cells is vital to understanding both their intrinsic and network properties. These recordings are still predominantly accomplished through the use of microelectrodes, which are generally limited to recording from only a few positions in a sample (Stuart and Sakmann 1994). Optical methods of recording Vm allow for multiple site recordings from populations, single cells, and fine processes that are too small or fragile for electrode recordings (Grinvald and Hildesheim 2004; Zochowski et al. 2000).
Current optical techniques to record fast Vm events in neural systems rely on one-photon methods (Antic 2003; Cohen et al. 1968; Gonzalez and Tsien 1997; Knopfel et al. 2003; Siegel and Isacoff 1997; Stepnoski et al. 1991; Zochowski et al. 2000) (fluorescence, absorption, etc.). One-photon fluorescence methods can be used to generate high signal-to-noise ratio (S/N) measurements of action potentials (APs) from subcellular regions in a single trial (Antic 2003; Antic and Zecevic 1995; Milojkovic et al. 2004) and subthreshold events with averaging (Antic and Zecevic 1995; Djurisic et al. 2004). Such recordings have been made from neurons in superficial layers of thick specimens (Antic 2003; Djurisic et al. 2004; Milojkovic et al. 2004); however, these techniques are generally limited by light scattering to depths less than ∼50 μm below the surface. To record Vm activity optically deep in intact systems with high spatial resolution, nonlinear optical methods [second-harmonic generation (SHG) or two-photon fluorescence (TPF)] are needed (Denk and Svoboda 1997; Denk et al. 1990; Dombeck et al. 2003; Kuhn et al. 2004; Mertz 2004; Zipfel et al. 2003a). The challenges of observing fast (∼1 ms) Vm events with nonlinear microscopy are significant, mainly because of the relatively small number of photons collected per time-point. The corresponding shot noise places strict limitations on the level of detectable signals and dye response to ΔVm that are needed. Compounding this problem for TPF is the reduction in the measured dye response to ΔVm by dye molecules that are present within the focal volume but are not embedded in the plasma membrane (background). Variations in this background can have the effect of making the observed ΔF/F signal difficult to quantify in terms of ΔVm unless the background can be kept low (as in culture; Bullen and Saggau 1999) or unless prior knowledge of the cellular Vm properties exists (Djurisic et al. 2004).
SHG has the potential to solve these problems. Because of the molecular alignment requirement for signal generation of this coherent two-photon scattering process, the SHG signal only emanates from properly oriented dye molecules in the membrane (Bouevitch et al. 1993; Campagnola et al. 1999; Mertz and Moreaux 2001); thus its effective Vm response is not significantly attenuated by background. The high membrane contrast also allows for a more direct relation between optical signal change and ΔVm. As with TPF, the nonlinear nature of SHG provides optical sectioning and micron scale resolution at depths up to ∼400 μm into scattering brain tissue (Dombeck et al. 2003).
Fast SHG recordings of ΔVm have been achieved in model membranes (Moreaux et al. 2003; Pons et al. 2003) and Aplysia neurons in culture (Dombeck et al. 2004), but have not yet been applied to intact mammalian neural systems where the full advantages could be realized. Previous SHG studies were also restricted to dyes that could only label cells extracellularly and could not be intracellularly applied with a pipette. Here we report on the latest step toward high-resolution SHG imagingof Vm in thick preparations by labeling a specifically targeted mammalian neuron below superficial layers in brain slices through intracellular application of FM4-64. We show the use of SHG microscopy for recording APs and fast voltage steps deep in slice where other techniques are not effective. For a direct comparison between SHG and TPF Vm recording in optically thick preparations, we optically record fast Vm signals with the two methods in brain slices using the best known Vm probes. We found the FM4-64 SHG S/N and effective Vm response to be greater than that for any other tested SHG or TPF probe. Even though TPF can outperform SHG in the culture dish, SHG is superior in brain slice systems because of its low background and high plasma membrane contrast compared with TPF.
FM4-64 brain slice recordings
The basic design of our Radiance 2000 (Bio-Rad)–based imaging system has been described (Dombeck et al. 2003, 2004). Now, however, an upright microscope (BX50WI, Olympus) is used with a 0.8-NA condenser for infrared video microscopy and collection of the transmitted SHG signal, 530/30 and 580/150 optical filters (Chroma Technology) for SHG and TPF detection, respectively, and GaAsP PMTs (H7422, Hamamatsu) for signal detection. A physiology objective (×40, 0.8 NA, overfilled back aperture) was used for illumination and epi (backward) collection. Excitation was provided by a 1,064 ± 15-nm fiber laser (Fianium, FEMPTOPOWER 1060) with ∼300-fs pulses at 70 MHz, ∼10- to 20-mW average power, and near linear polarization at the sample. This excitation was used instead of shorter wavelengths (∼900 nm) because 1) greater SHG responses to ΔVm were observed at 1,064 (∼7.5%/100 mV) than 900 nm (∼5%/100 mV); 2) absorption by intrinsic flavins will affect the 450-nm SHG signal more than the 532-nm signal; and 3) the quantum efficiency of available photodetectors is higher using GaAsP PMTs to collect 532-nm light (∼40%) than using BiAlkali PMTs to collect 450-nm light (∼25%) (Zipfel et al. 2003b).
Animals were treated in accordance with Cornell University Regulations (IACUC protocol 00-46-03). Transverse brain slices (∼250 μm thick) were made with a vibratome (Integraslice 7550 PSDS, Campden Instruments) from Sprague-Dawley rats (P5–P25) in 4°C oxygenated sucrose slicing solution (Kirov et al. 2004) containing (in mM) 212 sucrose, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2, and 25 glucose (pH 7.4; osmolarity, ∼310 mosmol/l). Slices were transferred to an ∼35°C oxygenated extracellular solution containing (in mM) 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2, and 25 glucose (pH 7.4; osmolarity, ∼310 mosmol/l with glucose). The slices were allowed to rest at 35°C for 30 min before being moved to room temperature until use.
ELECTROPHYSIOLOGY, STAINING, AND LINE SCANNING.
Slices were transferred to a recording chamber continually perfused with warm (35 ± 1°C) oxygenated extracellular solution. Whole cell patch-clamp recordings were obtained under infrared video microscopy using an Axoclamp 2b amplifier (Axon Instruments) and 5–10 MΩ pipettes filled with intracellular solution containing (in mM) 130 K-Gluconate, 5 NaCl, 24 KCl, 10 HEPES, 4 Mg-ATP, and 0.4 Na-GTP (pH 7.3 with KOH; osmolarity, ∼305 mosmol/l). A final concentration of 200 μM FM4-64 was added to this solution, sonicated for ∼1 min, and passed through a 0.2-μm filter to remove any dye precipitate. Once a gigaohm seal was obtained, quick negative pressure pulses opened the seal for whole cell recording and dye filling. Neurons from the dentate gyrus (DG), CA1, and entorhinal cortex were used, but not identified, except by physical location. Advasep 7 (1 mM; Cydex) was added to the extracellular solution. As expected, this absorbs dye discharged into the neural tissue when obtaining a patch and likely any dye that flip-flops to the outer plasma membrane leaflet during the course of the experiments (Kay et al. 1999). SHG population labeling with dye crystals has been shown (Mertz and Moreaux 2001), but here deeper labeling was accomplished by pressure (∼50 mmHg) injecting (Stosiek et al. 2003) ∼500 μM FM4-64 in extracellular solution into the tissue through a patch pipette for ∼1–5 min.
TTL pulse triggering was used to synchronize the electrophysiology stimuli to each line scan (triggering jitter < 3 μs). Data were collected during the scanning fly-back to obtain line scanning rates of 1,200 lines/s for AP recordings, whereas unidirectional scanning (300 and 600 lines/s, no fly-back collection) was used for dendrite recordings; 200–256 lines were collected per individual line scan. Two to 5 s were allowed between the 35–60 individual line scans that were later averaged, with typical AP averaging taking ∼2 min and voltage step averaging on dendrites taking ∼3 min. APs were elicited in current clamp during individual line scans by current injection through the pipette. Stimulation of APs was possible with ∼200-pA and ∼8-ms pulses; however, shorter pulses of more intense current (∼1 nA, ∼1-ms duration) were used to reach threshold quickly. This stimulation protocol led to a greater AP temporal stability of less than ∼0.2-ms drift over the minutes of signal averaging not possible with the longer protocol (more than ∼2-ms drift). This was necessary to synchronize the timing of the short-duration APs with the time during which the SHG was collected per line; this synchronization protocol has been previously described (Dombeck et al. 2004). The electrode recorded APs shown in Fig. 3C are the mean of n = 55. This averaging can have the effect of reducing AP amplitudes by ∼10–15% and increasing AP durations up to ∼10–20% compared with single trace recordings because of the small temporal drift of the APs from trial to trial. Voltage steps were applied in voltage clamp during line scans.
A Vm-independent bleaching effect during each individual line scan (∼1–5%/100 ms) was observed; this distortion to the SHG time-course was normalized from the recordings by fitting with a mono-exponential decay. The time allowed between individual line scans (2–5 s) was sufficient for nonbleached dye to diffuse and replenish the scanned membrane region. Preliminary reports of research using SHG from intracellularly filled FM4-64 to detect ΔVm in brain slices have been published in abstract form (Dombeck et al. 2005a,b; Yuste et al. 2005).
Comparison of various SHG and TPF probes
The most important parameter for comparing various dyes for optical Vm recording is the S/N; we attempted to maximize this value for each tested dye as follows. The S/N is proportional to the Vm sensitivity (ΔS/S) and the square root of the photon flux (see Eq. 1). Therefore the dye concentration and wavelength for each specific dye was selected either to produce the highest photon flux and ΔS/S, respectively, or to reproduce previous studies by other researchers. We varied the illumination intensity to obtain a similar photon flux for cells loaded with each dye (except for retinal, see results). The maximal photon flux, and therefore the maximal S/N, is limited by photodamage. The photodamage limit for each probe is difficult to quantify exactly (see discussion). However, various studies (see results for FM4-64 in brain slices and also Dombeck et al. 2004; L.S., personal communication; D.A.D., personal communication) found that, in several dyes, the approximate damage thresholds occur at a similar TPF photon flux. This value set the approximate illumination intensity for each dye for maximal S/N. Because the SHG photon flux is similar to the TPF photon flux, these considerations allow us to compare the S/N simply by measuring ΔS/S (SHG or TPF) for each probe. The ΔS/S was measured by applying fast (millisecond time scale) voltage steps.
Aplysia neurons were prepared according to Dombeck et al. (2004). Brain slices were prepared as listed above. Staining of cultured neurons was accomplished by extracellular perfusion of the dyes: JPW1114 (Di-2-ANEPEQ; 25 μM), ANNINE-6 (as in Kuhn et al. 2004), DHPESBP (as in Dombeck et al. 2004), FM4-64 (25 μM), Di-4-ANEPPS (as in Millard et al. 2003), and Retinal (as in Nemet et al. 2004). Intracellular staining in brain slices was accomplished as listed above: DHPESBP (100–300 μM in intracellular solution containing cyclo-γ-dextran, to prevent dye precipitation, prepared as in Millard et al. 2003), JPW1114 (200 μM in intracellular solution), and ANNINE-6 (1.7 mM stock in DMSO + 20% pluronic diluted 1:20 in intracellular solution; this high concentration of DMSO and pluronic was needed to adequately stain the cells, but did not noticeably affect cell viability). Line scanning and voltage clamping of Aplysia neurons stained with these dyes was the same as in Dombeck et al. (2004); brain slice line scanning and voltage clamping is described above. The following excitation wavelengths, SHG and TPF emission filters, and average powers at the sample were used to measure SHG and TPF Vm sensitivities: JPW1114 [1,064 nm, 570/150 (TPF), 530/30 (SHG), 60 mW], ANNINE-6 [980 nm, 570/150 (TPF), 30 mW; 1,064 nm, 570/150 (TPF), 530/30 (SHG), 80 mW], DHPESBP [940 nm, 575/150 (TPF), 460/30 (SHG), 15 mW], FM4-64 [1,064 nm, 570/150 (TPF), 530/30 (SHG), 15 mW], Di-4-ANEPPS [850 nm, 575/150 (TPF), 425/20 (SHG), 15 mW], and retinal [900 nm, 575/150 (TPF), 460/30 (SHG), 15 mW]. The shorter wavelengths were obtained with a mode locked Ti:Sapphire laser (Spectra Physics, Tsunami). As a control, the sensitivity of ANNINE-6 was measured in HEK293 cells at 980 nm under the same conditions as Kuhn et al. (2004); our value of −35 ± 2%/100 mV (slope at 0 mV) agrees with their value of −35 ± 5%/100 mV. The measurement in HEK293 cells was also repeated at 1,064 nm (see discussion).
Staining and imaging
Neurons in hippocampal slices patch clamped and filled with intracellular solution containing FM4-64 and illuminated with ∼1,064-nm/∼300-fs laser pulses generate an intense SHG signal from the labeled inner leaflet of the plasma membrane, with little background signal from intracellular or extracellular sources (compare SHG in Fig. 1A to TPF in Fig. 1B). The plasma membrane to intracellular SHG signal ratio is M/I = 20.3 ± 11.8 (n = 16 cells) for soma and proximal larger dendrites. In finer dendrites and spines, the intracellular space could not be resolved, precluding measurements of this ratio. The lack of an SHG signal from intracellular membranes is likely caused by 1) dye flip-flop quickly resulting in a dye concentration equilibrium between leaflets of the intracellular membranes, destroying the asymmetry needed for SHG; 2) subresolution invaginations and foldings of the intracellular membranes (Krstiāc 1979; Ladinsky et al. 1999), leading to destructive SHG interference from oppositely oriented dye molecules; or 3) different partitioning coefficients for FM4-64 in the plasma membrane versus intracellular membranes (Sprong et al. 2001).
The staining time for FM4-64 varies depending on many parameters (whole cell access resistance, neuron volume and architecture, etc.), but typically an SHG signal from the soma and dendrites >100 μm distal from the soma was visible ∼5 and ∼25 min after whole cell break-in, respectively (Fig. 1C). Most cell bodies were ∼50–70 μm below the slice surface, with many of the stained dendrites >125 μm below the surface. Axons were also visible in some images (data not shown). Current-clamp recordings from stained cells reveal physiologically normal parameters: resting potential of approximately −65 ± 10 mV, input resistance of ∼150 MΩ to 2 GΩ, AP amplitudes of ∼65–100 mV, and duration of ∼1.0 ± 0.5 ms. Figure 1F shows a current-clamp recording from a stained neuron ∼45 min after whole cell break-in. Average whole cell recordings lasted ∼45 min, with some cells viable for >1.5 h, similar to cell viability without imaging or dye. Toward the end of recordings, many cells had an increased diffuse intracellular SHG signal. Additionally, a fraction (∼25%) of otherwise normal neurons had a greater intracellular SHG signal (M/I ∼ 5) throughout the recording time.
The high spatial resolution (∼0.6 μm) allows for dendritic spine visualization. A comparison between SHG (Fig. 1D) and TPF (Fig. 1E) spine images reveals that not all of the spines visible in the TPF channel are visible in the SHG channel. The filled arrowheads point to spines visible in both channels, whereas the open arrowheads point to spines visible only in the TPF channel. No spines were present in only the SHG and not the TPF channel. It is possible that the size of individual spines and/or the distance between adjacent spines leads to destructive interference of the SHG signal in some cases. Multi-cell labeling is also possible. Figure 1G shows two neurons patch clamped and filled in the same field of view, whereas Fig. 1H shows a larger population stained by pressure injecting a concentrated solution of FM4-64 into the CA1 cell layer. Although M/I appears to be greater than stated above in many figures (Figs. 1, 3, 4), this is an illusion caused by projecting many z-sections onto one plane.
Response of SHG to fast Vm transients
By applying voltage steps to the voltage-clamped neurons during line scanning, we found that the SHG signal is modulated by ΔVm on the physiologically relevant (∼1 ms) time scale (Fig. 2, A–C). A linear response of the SHG signal with respect to ΔVm [linear best fit: ΔSHG/SHG = (0.075%/mV)ΔVm; slope error = 0.004%/mV] is observed by applying a range of voltage steps (Fig. 2D). The SHG signal increases with positive voltage steps and decreases with negative voltage steps, opposite to the response seen from extracellularly labeled neurons in culture. This is expected from the opposite orientation of the chromophore with respect to the transmembrane electric field.
Spatial resolution and S/N
Assuming a shot noise limited system, the S/N for Vm recording can be written (1) where ΔS/S is the normalized relative signal response to ΔVm (Vm sensitivity), n is the number of averaged line scans, φ is the number of detected photons/time, and t is the time the focal volume scans over the membrane per line; t is defined by t = a/v, where a is the distance scanned over the membrane, and v is the focal volume scanning velocity. Line scanning perpendicular to the membrane (t ∼ 10 μs, a reduces to the focal volume diameter) provides high spatial resolution (∼0.6 μm; Fig. 2, A–C) and results in S/N for a single line scan of ∼1 for a 90-mV step and S/N of ∼5 for temporal averaging of n = 50 line scans. The shot noise corresponds to ∼100 photons/membrane pass (∼10 photons/μs on the membrane) and ∼5,000 photons after averaging. Line scanning parallel to (along) the neuronal membranes is also used (Figs. 3 and 4); this amounts to increasing a. The differences in S/N for fast SHG Vm recordings seen in this research are caused by one or more of the following: number of line-scans averaged (n), integration time on the membrane (t), average illumination intensity (I), degree of staining, polarization of the incident illumination with respect to membrane position, and/or ΔVm in the specific recording.
Detection of APs and fast voltage steps
To prove the ability of SHG to record fast neuronal Vm signals in brain slices, APs were elicited and optically recorded. The line scanning position parallel to the somatic membrane is shown in Fig. 3A. Two ∼1-ms APs were elicited by current injection through the patch pipette during line-scanning. Line scan averaging (n = 55) showed that optical intensity modulations in the SHG emission (Fig. 3B) occur during the AP time-course (Fig. 3C) with S/N ∼7–8, 0.83-ms temporal, and 0.6 μm by ∼7 μm x-y spatial resolution. Trains up to four APs were recorded in other experiments (data not shown). The temporal resolution of the scanning system is approximately the same as the AP duration; this can result in undersampled recordings. It is possible, however, to record the approximate AP peak voltage optically by synchronizing the time of the AP peak and the focal volume recording time on the membrane (t). Asynchrony in this regard results in SHG traces not recording the AP peak voltage.
Figure 4 shows the ability to record fast Vm events in the dendritic arbor using SHG. Scanning positions parallel to dendritic and somatic membranes on two neurons in slice are shown. During line scanning, voltage steps were applied through the patch pipette at the soma. After line scan averaging (n ≥ 30), the S/N varied between ∼3.5 and 5.0. Many of the probed dendrites were <2 μm in diameter and ∼100 μm below the slice surface. The recording plane for the oblique dendrite in Fig. 4C (inset) is actually ∼30 μm above the apical dendrite seen in this z-projection image.
Photodamage was seen during some FM4-64 SHG AP recordings when a dose corresponding to I ∼ 7 MW/cm2 and t ∼ 65 μs was reached. In many cases below this dose, the damage is negligible in comparison with nonimaging controls, as determined by AP amplitude and duration stability, resting potential stability, and absence of gross morphological changes. The dendritic arbor is more sensitive to photodamage than the soma, so the dose in the arbor was reduced compared with that used at the soma. It was also found that increasing the time between each individual line scan during averaging helped reduce photodamage. Many seconds (∼5 s) works well, but many successful recordings were also made with ∼2 s between scans.
Backward-directed SHG component
In addition to the large forward propagating SHG signal used for recordings in this research, the stained neuronal membranes also generate a small backward directed SHG signal in slice. A SHG forward versus backward ratio of 6.6 ± 1.1 was deduced from SHG images of filled neurons, both in forward and epi propagation directions, with fluorescence calibration for absorption and scattering in the tissue, and collection efficiency differences of the optics and detectors in each direction. These measurements were made 50–100 μm deep into 250-μm-thick brain slices from P16–P17 rats. Like the forward propagating signal, the backward signal emanates predominately from the plasma membrane, with little background. The S/N for Vm recording with the SHG backward-directed component is reduced by a factor of ∼ compared with the SHG forward propagating signal.
Staining and Vm sensitivity with various TPF and SHG optical probes
We initially compared the optical Vm recording S/N for many of the best known SHG and TPF Vm probes in cultured Aplysia neurons (stained by extracellular perfusion of the dyes) to select the best candidates for use in intact brain tissue: ANNINE-6, FM4-64, DHPESBP, Di-4-ANEPPS, JPW1114, and retinal. The most important parameter for comparing these dyes is the S/N, which depend on many parameters (see Eq. 1); however, we found that the S/N was most strongly related to ΔS/S (see methods), making it possible to compare the S/N simply by measuring ΔS/S (SHG or TPF) for each probe.
Although retinal (Nemet et al. 2004) showed a high SHG ΔS/S (∼20%/100 mV), the S/N was reduced by an effective SHG cross-section far lower than the other probes that could not be overcome because of poor staining efficiency and severe photobleaching problems. In contrast to the measurements performed by Millard et al. (2003, 2004) on the slow (seconds) time scale, no SHG response to fast (milliseconds) ∼100-mV voltage steps was seen from neurons stained with DI-4-ANEPPS, and the TPF ΔS/S was ∼3%/100 mV. This result implies that there is no correlation between fast (electro-optic or reorientational; Moreaux et al. 2003; Pons et al. 2003) and slow SHG Vm response mechanisms. The largest ΔS/S was measured in culture with ANNINE-6, FM4-64, DHPESBP, and JPW1114. The membrane staining ability and ΔS/S (SHG and TPF) of these four probes in culture are summarized in the first column of Fig. 5 and in Dombeck et al. (2004) and Moreaux et al. (2003).
Given the ΔS/S from our culture dish experiments, we focused our brain slice studies on the four best probes. DHPESBP was eliminated because neurons in slice filled with this dye showed no SHG from intracellular or plasma membranes, indicating poor membrane labeling compared with in culture. The membrane staining ability and ΔS/S (TPF and SHG) of ANNINE-6, FM4-64, and JPW1114 probes in brain slice are summarized in the second column of Fig. 5. It is seen that ANNINE-6 was not able to produce SHG in brain slices, again indicating little dye partitioning into the plasma membrane. Because of the large TPF ΔS/S seen in extracellularly stained cultured cells, we were encouraged to measure the ΔS/S in filled neurons. This measurement showed a drastic reduction of the TPF ΔS/S from 28.3%/100 mV in cultured neurons to 1.2%/100 mV in brain slices. A large reduction of the TPF ΔS/S (from 12.2%/100 mV in cultured neurons to 1.4%/100 mV in brain slices) is also shown for JPW1114; however, the JPW1114 brain slice SHG signal seen in Fig. 5 shows this dye is well loaded into the membrane. This indicates that the high background and low membrane contrast, rather than poor membrane loading, is responsible for the reduced TPF ΔS/S. On the other hand, the SHG ΔS/S of FM4-64 and JPW1114 is only slightly affected by intracellular brain slice loading compared with the culture dish. Additionally, FM4-64 presented a higher SHG ΔS/S and slower flip-flop rate between leaflets of the membrane than JPW1114. These results show that the best ΔS/S, and consequently S/N, in brain slice is obtained with FM4-64 SHG.
S/N and photodamage considerations
Increases in S/N are dependent on 1) increasing φ in Eq. 1 to decrease the shot noise and/or 2) increasing ΔS/S in Eq. 1. Both 1 and 2 can be accomplished though the design of better dye molecules. This is the subject of current experiments in our laboratory and reports of others (Barzoukas et al. 1996; Millard et al. 2004; Moreaux et al. 2003; Nemet et al. 2004; Pons et al. 2003). It is hoped that design and screening of many SHG probes on the physiologically relevant fast time scale will soon lead to similar increases in SHG ΔS/S and low bleaching as have been reported for TPF from ANNINE-6 (Kuhn et al. 2004). Other incremental increases in point 1 above have been made and will continue to be sought through the engineering of higher transmission optics and higher quantum yield photo-detectors (such as the GaAsP detectors used here).
With respect to FM4-64 and the current photon collection/detection technology, the factor most limiting S/N is photodamage, and in fact, the shown S/N in Figs. 2–4 was obtained near the damage thresholds for soma and dendrite recordings. Studies aimed at understanding and decreasing this problem should help to increase S/N by accommodating increases in incident laser intensity that will increase φ. Photodamage is not well understood, and its source remains an open question (Zipfel et al. 2003b), although it is generally agreed that this damage stems from absorption of the incident illumination. Damage during SHG is likely therefore to stem from absorption and nonradiative photoproducts of the chromophore. The damage is a complicated function of I, t, staining, cell morphology, and other unknown factors (Hopt and Neher 2001; Koester et al. 1999), and here was found to vary from cell to cell and even within one cell. A study is underway in our laboratory to characterize the photodamage and possibly to reduce its influence through the use of antioxidants. The results of this study may also increase S/N enough for the detection of fast Vm events in a single trial.
To detect Vm signals of 100 mV optically in a single trial using SHG from FM4-64 with S/N = 5, it would be necessary to collect ∼4,500 photons per membrane recording time t. This could be accomplished given the SHG φ and ΔS/S shown here for FM4-64, with the only change being an increase in t to ∼450 μs. In the line scanning configuration, this integration time could be achieved by increasing a, the distance scanned over the membrane; however, this was found to cause photodamage in brain slices unless I was simultaneously reduced. For this reason, damage studies, better photo-detectors, and/or higher effective cross-section dye molecules are needed to increase the number of collected photons. Improvements in dye molecule performance can also increase ΔS/S. Assuming a dye with ΔS/S ∼ 20%/100 mV, the number of collected photons needed to measure a 100-mV Vm signal optically with S/N = 5 is ∼600 photons per membrane recording time t. If we also assume a fourfold increase in Φ over that shown in this research for FM4-64, an integration time of t ∼ 15 μs would be needed. Assuming the proper scanning system such as random access scanning (Bullen and Saggau 1999; Iyer et al. 2004; Otsu et al. 2005; Roorda et al. 2004), this integration time could allow for ∼65 different locations in the brain slice to be sampled at 1 kHz or ∼20 locations at 3 kHz. This shows the potential of SHG to optically record fast Vm signals at many different positions deep within scatting brain tissue.
Comparison of various SHG and TPF probes for Vm recording
SHG and TPF currently provide the best methods for high spatial resolution imaging deep in scattering tissue (Denk and Svoboda 1997; Denk et al. 1990; Dombeck et al. 2003; Mertz 2004; Zipfel et al. 2003a). We have compared many of the best known SHG and TPF Vm probes to select the best candidate for use in intact brain tissue. Of these probes, JPW1114 and FM4-64 can be filled into neurons in brain slice to generate an intense plasma membrane SHG signal. On the other hand, ANNINE-6 shows little dye partitioning into the membrane and high background in brain slice; consequently, we measured a TPF ΔS/S reduction of ∼23 times compared with the culture dish experiments. It is important to observe that the TPF ΔS/S of JPW1114 in brain slice is also drastically reduced by an order of magnitude. In this case the dye is well partitioned into the plasma membrane producing an SHG signal; however, the effective TPF ΔS/S is still greatly affected by the poor membrane contrast and high background. Furthermore, this problem is not specific to intracellular staining; pressure injection of ANNINE-6 into brain slices also results in a low membrane contrast compared with FM4-64 SHG, likely caused by binding of the dye to extracellular components such as the extracellular matrix. The SHG ΔS/S was reduced only slightly (20–30%) between the culture dish and intracellular filling in intact systems. A similar SHG Vm sensitivity dependence on the slow time scale has also been observed between different cell types (Sacconi et al. 2005). These marginal effects may be explained by the vastly different environments of extracellular staining of Aplysia neurons (with a salt concentration approximately twofold greater than mammalian neurons) and intracellular staining of mammalian neurons in brain slice, small SHG background variations (see M/I measurements above), or possibly by the different physical and chemical properties of the plasma membrane environments.
These results show that 1) the SHG ΔS/S is not degraded by background signal (as is TPF), and consequently, the S/N is greater by almost an order of magnitude with SHG than TPF using currently known probes in brain slices; 2) together with the linear response to ΔVm, the low SHG background makes it possible to quantify directly the SHG Vm response in terms of ΔVm. This is more difficult with TPF because of spatial variations in the background (see Fig. 5 or Antic and Zecevic 1995); 3) the SHG ΔS/S is more consistent in different cell types and staining environments than TPF; and 4) of the tested nonlinear optical Vm probes, the best S/N achieved in brain slice is provided by FM4-64 SHG.
Our results suggest that it will be difficult to increase the TPF membrane contrast, given the JPW1114 results. Therefore the most promising direction to follow to increase the optical Vm recording S/N with nonlinear microscopy in intact systems is to increase ΔS/S and/or the effective cross-section of SHG or TPF probes. Simply increasing the TPF ΔS/S and/or cross-section of probes such as ANNINE-6 or JPW1114 without considering the TPF background will likely not lead to a solution to the problem. For example, to overcome this background problem, a TPF ΔS/S ∼ 20 times greater than ANNINE-6 would be needed. According to Kuhn et al. (2004), increases in ΔS/S for ANNINE-6 should be possible at longer wavelengths. Although our ΔS/S measurements at 980-nm agree with Kuhn et al. (see methods), under the same conditions we measure only an ∼30.8 ± 1%/100-mV sensitivity at 1,064 nm. Together, these arguments and results imply that the most promising direction of engineering to follow for nonlinear microscopy Vm recording is to increase the ΔS/S and effective cross-section of SHG probes.
For in vivo applications a large fraction of the TPF signal can be epi-collected, whereas only a small fraction of the SHG signal can be collected in this direction. Wavelength scale inhomogeneities (Mertz and Moreaux 2001) or secondary scattering of the forward propagating SHG signal are the likely cause of the backward propagating SHG signal. The longer scattering length presented to the forward propagating SHG photons in vivo compared with in slice may increase the SHG backward scattered signal. New technologies to manipulate the polarization and phase of the illumination light within the focal volume (Novotny et al. 2001) may also increase the SHG backward propagating signal by optimizing coherent addition conditions. Even if the backward propagating SHG signal cannot be increased in the future, currently the S/N for intact systems Vm recording with back propagating FM4-64 SHG is still approximately three times better than can be achieved with ANNINE-6 epi-collected TPF and approximately the same as JPW1114 epi-collected TPF. In theory, combining the epi-TPF and back-propagating SHG signals from JPW1114 may provide a slight advantage over FM4-64 in vivo; however, the rapid flip-flop rate of JPW1114 compared with FM4-64 must also be considered.
Possible applications and future directions
Micron scale Vm dynamics deep in intact neural systems, where much of the volume of activity and dendritic integration resides, have been inaccessible until now because of the light scattering limitations of linear optical methods. We have shown the optical recording of fast neuronal Vm transients deep in intact mammalian neural tissue with micron scale resolution using SHG microscopy from targeted neurons patch clamped and filled with FM4-64. This technique should find applications in the near future in studies that directly record and quantify fast ΔVm from dendritic spines (Tsay and Yuste 2004) and from fine dendrites. Deeper single cell patching/SHG recording can be implemented using TPF targeted patch clamping (Margrie et al. 2003). Because the SHG emission is always at one-half the illumination wavelength, it can also be combined easily with functional fluorescent indicators such as Ca2+- and Na+-sensitive dyes. With further improvements in the S/N, labeling methods (Hinner et al. 2004), and a combination with faster and more flexible imaging modalities (Bullen and Saggau 1999; Iyer et al. 2004; Kobayashi et al. 2002; Otsu et al. 2005; Roorda et al. 2004; Sacconi et al. 2003; Tsien and Bacskai 1995), SHG should move the imaging field closer to large scale high-resolution Vm recordings of population activity deep in thick tissue preparations.
M. Blanchard-Desce is grateful to CNRS for financial support (Action Coordonée Optique and Physique et Chemie du Vivant grants), L. Sacconi is grateful to the European Laboratory for Non-Linear Spectroscopy for Grant MTKD-CT-2004-509761 and to Cornell University for his visit, and D. A. Dombeck and W. W. Webb are grateful for funding from National Institutes of Health Grants P41 EB-001976-17, GM-08267, and GM-07469.
We thank T. Mallegol for the synthesis of the DHPESBP dye, S. Kim, R. Molloy, M. Williams, and R. Williams for helpful discussions and reading of this manuscript, W. Zipfel and I. Tullis for instrumentation support, and R. Yuste for helpful discussions on the intracellular use of FM4-64. We also thank R. Harris-Warrick, M. Diaz-Rios, and D. Tsay for slice patch clamping advice and guidance and H. Vishwasrao and K. Kasischke for advice on slice preparation.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 by the American Physiological Society