2023 年 20 巻 1 号 論文ID: e200015
Plasticity is the key feature of our brain function. Specifically, plasticity of hippocampal synapses is critical for learning and memory. The functional properties of the neuronal circuit change as a result of synaptic plasticity. This review summarizes the use of voltage-sensitive dyes (VSDs) to examine neuronal circuit plasticity. We will discuss the significance of plastic changes in circuit function as well as the technical issue of using VSDs. Further, we will discuss the neural circuit level plasticity of the hippocampus caused by long-term potentiation and the entorhinal-perirhinal connection. This review article is an extended version of the Japanese article, Membrane Potential Imaging with Voltage-sensitive Dye (VSD) for Long-term Recording, published in SEIBUTSU BUTSURI Vol. 61, p. 404–408 (2021).
This review discusses the application of voltage-sensitive dyes for imaging plastic changes of neural activity in hippocampal brain slices. The history and the technical challenges of this method will be highlighted.
The core feature of our brain is its plasticity. To investigate its mechanism, we must observe the functional reorganization of the neural circuit. Optical recording of membrane potential changes in circuit neurons with voltage-sensitive dyes (VSDs) [1–5] is a critical method to observe plasticity in the neural network.
Since the discovery of synaptic plasticity [6], the hippocampus has been recognized as a key component for learning and memory [7,8]. Its significance in learning and memory was established by William B. Scoville’s neurosurgery on Henry G Molaison (patient HM), a patient with intractable epilepsy, and subsequent detailed studies by Brenda Milner [9], which have left no doubt that the hippocampus is an essential neural circuit for learning and memory [10,11].
The hippocampus is divided into three sections: The dentate gyrus (DG), cornu ammonis 1 and 3 (CA1, CA3). The hippocampus receives input from entorhinal cortex to the DG. The DG projects to CA3 via mossy fibers. And CA3 projects to CA1 via Schaffer collaterals. Together these projections form a trisynaptic circuit. Plasticity exists at all three synapses. Long-term potentiation (LTP) at the synapse between Schaffer collaterals and CA1 pyramidal cells is arguably the best-studied kind of plasticity to date. However, most published studies of LTP describe LTP as a plastic change on a single synapse level and rarely describe how the input-output relationship of the CA1 circuit is altered. For neuroplasticity, it is important to know how the input-output (I/O) relationship of a circuit is changing. Synaptic plasticity can alter the function of the neuronal circuit and alter how the neuronal signal travels through the circuit. For example, with voltage imaging, we recently found that the spread from the perirhinal cortex to the entorhinal cortex changed and that this change persisted for a long time (6 hrs to 12 hrs) [12,13].
Brain function depends on the complex interactions of many neurons. The mechanism of learning and memory is based on the plasticity of neural circuits. In the actual nervous system, Bliss and Lømo showed its presence in the rabbit hippocampus in 1973. This is the most famous example of LTP in the hippocampus. Particularly in the CA1 area of the hippocampus, LTP occurs after high-frequency stimulation (HFS; typically, 100 Hz for 1 second) or theta burst stimulation (TBS; pattern stimulation that mimics theta waves, which are brain waves seen during animal behavior; ten bursts of high-frequency (100 Hz) stimulation, each consisting of four pulses, with an inter-burst interval of 200 ms) is applied to the Schaffer collaterals. In many situations, LTP produced by TBS is measured as an indicator of learning and memory [14–16]. It is a cellular physiological model that has been validated in numerous physiology papers as well as in pathology research, such as in dementia research, and research on various neuropsychiatric disorders.
LTP in the hippocampal CA1 area is often measured with an extracellular recording electrode. In most cases, the electrical signal is recorded from a single location. The hippocampal CA1 area is thus described as a one-synapse circuit with one input fiber connecting to one postsynaptic pyramidal cell. However, changes not only in the temporal but also spatial pattern during LTP induction [17,18] are important features for a mechanistic understanding of the circuit. With VSDs, we successfully imaged the changes induced by LTP in the neural circuit of the hippocampus [19].
Neurons, the building blocks of the neural circuit, share all basic features with all other cells. For example, their cell membrane consists of a 5-nm-thick lipid bilayer. The lipid is insulating and therefore acts as an electrical capacitance. Due to an asymmetric distribution of ions inside and outside the cell and leak currents of these ions through the membrane, a potential difference (membrane potential) over the membrane is generated. This is true for all cells. Neurons have the additional feature of voltage gated ion channels which can on a time scale of milliseconds change the current of specific ions through the membrane. Thereby, neurons can generate action potentials and graded voltage changes. These membrane voltage changes are used to propagate and process information within a neuron, and, importantly, to integrate synaptic inputs. It is crucial to record membrane voltage changes to understand brain activity on a cellular, circuit, or systems level.
Typically, electrophysiological instruments are used to detect electrical signals from cells. At first, microelectrodes were used to measure the electrical potential of large plant cells the 1920s. Then, also animal cells were examined in the 1930s–1940s), at first larger, then also smaller cells [20]. The patch clamp technique [21] increased the applicability of electrophysiological methods to much smaller cells, including mammalian neurons. The method is a direct measurement of the voltage across the membrane.
Another path to measure membrane potential changes was developed already in the 1940s: This method is based on tiny changes of the optical properties, such as birefringence and light scattering of membranes (ΔF/F typically in the range of 10–5 to 10–6), associated with membrane voltage changes due to ion currents over the membrane during neural excitation [22–26]. It is surprising how early changes in optical properties were associated with changes in membrane potential, even before Alan L Hodgkin and Andrew F Huxley’s Nobel Prize-winning paper on the mechanism of neural excitation was published in 1952. After the discovery of intrinsic optical property changes, optical measurement methods were further developed to include the use of externally applied dye molecules, which could amplify the signal and improve the sensitivity of the measurements [27]. In the early 1970s, Larry Cohen, Brian Salzberg, and their colleagues screened hundreds of dyes from different companies at the Marine Biological Laboratory in Woods Hole, searching for a molecular probe that effectively converts an electrical signal into a change in light intensity [28]. The goal was to “see” the spread of voltage change along neuronal processes. Indeed, they found Merocyanine-540 as the first voltage-sensitive dye for neuronal membranes [29–32]. For a mechanistic description of voltage-sensitivity, see Kuhn & Roome [33]. Although the measurements were made with a photodiode, it was clear to the authors that “such a probe could provide a powerful technique for measuring membrane potential in systems where, for reasons of scale, topology, or complexity, the use of electrodes would be inconvenient or impossible” [30]. Several groups quickly followed up to add a spatial dimension by designing linear imaging devices based on photodiode arrays [34–36]. They also made these devices commercially available. This imaging devices are, of course, no longer available, however, some laboratories continue to use them for more than 40 years, and, if necessary, repairing them with soldering irons. Until the year 2000, photodiode arrays were the best imaging device due to their huge photon-well depth. This means that photodiodes, which convert photons to electrons, can store large quantities of electrons (millions). Thereby it is possible to raise the signal level above the photon shot noise [3,37].
Since voltage-sensitive dyes exhibit small fractional changes upon a physiological membrane voltage change, the light intensity of the recording device is critical. When we first started recording VSD imaging from mammalian brain slices, we used RH-155, a light-absorbing dye [38]. Because the light transmitted through the tissue was much brighter than the fluorescence signal, we imaged the transmitted light to overcome the photon shot noise.
Special techniques were used to address the problem of dye washout with continuous perfusion of low concentrations of VSD [39]. Even though the response became smaller, it was difficult to determine whether it was due to dye washout or other physiological changes. Therefore, we chose Di-4-ANNEPS for the brain slice imaging. The initial fluorescence intensity can be used to calibrate the amount of dye molecules. We selected and optimized excitation and absorption filters for measuring hippocampal CA1 responses based on actual measurements with a combination of filters. In our selection process, we prioritized maximizing the signal-to-noise ratio (S/N) over fluorescent intensity, recognizing that the maximum fluorescence wavelength listed in catalogs may not always correspond to the best wavelength for S/N. We consider that fluorescence properties change due to the membrane potential. The maximum S/N is found where the rate of change of fluorescence properties is greatest for the input wavelength [40].
Also, the numerical aperture (NA) of the objective lens is critical in epifluorescence measurement. As the light passes through the objective on its way to and from the object, NA acts in the fourth power on the fluorescent intensity. The hippocampal system in mice is approximately 2–3 mm in diameter and in rats approximately 5 mm. To visualize the neuronal activity in the circuit, low magnification optics are necessary. Since there were only a few such optics available which can image such large areas, that is with low magnification, and with bright fluorescent, we build our own setup. Our design was based on a tandem-lens microscope design proposed by Amiram Grinvald et al. proposed, which we replicated and improved [41]. The difference between the original design is that we used the Leica objective lens of a fluorescence stereo microscope [37]. The large aperture of this lens is an advantage. By installing a huge dichroic mirror on it, it was possible to obtain an extensive NA even at low magnifications. Figure 1 compares low-magnification bright fluorescent optics (Figure 1A) to a conventional microscope (Figure 1B) under simplified conditions. In conventional microscopes, the focus length L2 (tube length) is fixed by the manufacturer (typically 160, 170, or 180 mm), and the pupil size (φ2) is usually less than 25 mm. To achieve low magnification, such as 1x, the objective lens must have a small numerical aperture (NA) due to the small angle of incidence (θ2) resulting from the fixed L2. In contrast, our system allows for the use of a high NA due to the large pupil (φ1, up to 50 mm) and small focus length (L1, approximately 55 mm), resulting in a large angle of incidence (θ1). To achieve low magnification, a large imager is also preferable, but this requires a large pupil.
A conceptual illustration comparing a wide-field macroscope to a conventional microscope at a magnification of 1x.
(A) Wide-field macroscope: The same large lens with a focus length (F)=L1 is used for both the objective and projection lenses. The optics have a large pupil of φ1 and a large angle of incidence θ1, resulting in a large numerical aperture (NA).
(B) Conventional microscope: The microscope uses a fixed tube length of L2, and to achieve a magnification of 1x, the objective lens must have F=L2, which results in a small pupil φ2 and a small angle of incidence θ2, leading to a small NA.
The amount of detected light is crucial for imaging voltage signals at high speed (10–3) to overcome the photon shot noise [3,37,42]. The small VSD signal also requires minimal mechanical disturbances. The size of the VSD signal is easily exceeded by only subtle light fluctuations introduced into the field of view. To avoid such complication, we also introduced a new chamber system for handling brain slice specimens [37,42]. The usual physiological chamber requires weights or holders to hold the slice in place under a microscope, while our system does not. In our system, hippocampal slices adhere to a 0.45 μm pore a hydrophobic, polytetrafluoroethylene (PTFE) polymer membrane filter membrane (Omnipore, Millipore, USA) when they are in good physiological condition. Our chamber system ensures streamlined perfusion of physiological solution (artificial cerebrospinal fluid, ACSF) from the bottom and over the slice. The membrane filter offers good optical transmittance, enabling the use of transmitted light for slice observation and electrophysiological measurements with upright optics. We also designed the system to minimize the use of costly VSD staining and physical damage to the slices. Since each slice specimen was held in a separate filter, it was easy to distinguish and relate them to the corresponding anatomical location of the brain during serial sectioning. This was particularly useful when dealing with non-hippocampal slices, where the ability to accurately identify the location of each slice relative to a brain atlas was important. In the case of hippocampal slice preparation, this was less critical, but the ability to easily distinguish between slices still offered benefits in terms of organization and record-keeping. Additionally, this approach was useful for creating functional maps of brain optical response. Our membrane holders are designed to accommodate more than one slice, which enables the co-culture of different tissue types and the examination of inter-tissue interactions using voltage-sensitive dye imaging (VSDI). This capability enables the investigation of complex biological processes in the brain.
We used our optimized imaging setup for imaging voltage changes on a circuit level induced by LTP in the hippocampal CA1 area. The initial LTP measurements were a comparison of two time points, capturing the average of 8–16 measurements of the neural response before and after LTP induction. With this method, we can measure hippocampal neuronal activity with light exposure once every 30 seconds for about 500 ms during which we acquire 100 images at a rate of 10,000 frames per second without any attenuation of bioactivity or fluorescence intensity due to bleaching or phototoxicity [43].
As shown in Figure 2, by adding TBS to the Schaffer collaterals while measuring the entire stimulus response once every 30 seconds, we were able to measure late-phase LTP (L-LTP) [44,45] – that occurs over 12 h. L-LTP is a phenomenon that, even for electrophysiological measurements, is technically difficult to measure.
A. Illustration of the photometric technique in the hippocampal CA1 area and continuous LTP measurements over a 12-hour period (B, C). D. The figure shows data from the first 6 hours. Reproduced from from [42]. The citation is permitted under CC-BY.
The results show that the neural response pattern is established at 2, 4, and 6 hours after LTP induction, although the magnitude of LTP varies by location. In other words, LTP is observed throughout the circuit. This may seem obvious, but it is a new finding that has not previously been observed. Once a circuit response is modified by LTP, it remains modified, just like a single synaptic response.
The next step was to determine if the way the circuit response changed differed between HFS and TBS. To examine the data, the same measurements were taken across multiple slice samples in a similarly shaped CA1 region and averaged. So, we calculated the data by averaging the responses from nearly the entire hippocampal CA1 area. The pattern of LTP in the circuits of HFS and TBS differs in that HFS produces more transient response enhancements, known as short-term potentiation, especially in the distal regions, whereas TBS produces less transient LTP in a staircase fashion throughout the circuit.
Such differences would be useful to understand when LTP induction experiments are used to study drug or gene effects. Furthermore, changes in the input-output relationships of the circuit are described. Such measurements were quite unimaginable from the experiments in the 1990s [38,46]. This technique is now automated, allowing for more quantitative tests, and is being used to verify the modification of neural circuits by various drugs, genetic modifications, and environmental factors.
The stability of the VSD recording also allowed us to quantify functional modifications in neuronal circuit activity caused by genetically modified mice and rats [47–51] and drugs [52–54]. It was also used to record intrinsic optical signals [55–58]. In addition, the method has the advantage of recording one-time events such as oscillation of the cortex [56,57].
How do these circuit modifications manifest themselves in plastic changes between large areas? The olfactory entorhinal cortex (EC), which interfaces with the hippocampus, is thought to be a circuit involved in the semantic formation of memory. One of the authors, Kajiwara [59], discovered that stimulation of the amygdala, which is thought to be involved in emotional responses, significantly alters how the perirhinal cortex (PC), which is located adjacent to the EC, propagates. The amygdala, which is involved in emotional responses, is stimulated in the PC. Once observed, this change lasts for a long time and the response of the EC-PC circuit to the same stimulus remains unchanged for a long time (Figure 3). This is a true plastic change in the circuit [12,13]. The time course was then followed by an optical measurement over time (1000 frames per second, about 1 second exposure) with one stimulus every 30 seconds. Although the stimulus was applied to a single-point on the surface layer of the PC, the response spread throughout the PC. The addition of 4-aminopylidine (4-AP; 40 μM), an inhibitor of D-type potassium channels, caused the neural activity that had previously only propagated to the PC to also propagate to the EC side, resulting in strong and long-lasting excitation. Furthermore, this PC-EC neural transmission continued for at least an hour after the 4-AP was washed away. We also confirmed that the combination of superficial and deep stimulation (Figure 3) caused neural propagation between PCs and ECs and that once this occurred, neural propagation between PCs and ECs persisted throughout the experiment. This is a significant finding because it shows that the response of neural circuits across large areas can undergo plastic changes, which cause a change in the response relationship over time if the response is changed in some way. Further, it demonstrates the plasticity of circuits. Although the strength of the response in the EC and PC is shown in the temporal recordings, it should be noted that it is difficult to measure using electrophysiological techniques and conclude that the propagation of excitation between the EC and PC has been altered. Only stable recording with a wide-field of view allows for changes in the function of the neural circuit. This type of wide field recording is also useful for in vivo recordings [60]. Similar studies are currently being conducted in cortical areas such as the anterior cingulate cortex. The ability to make such one-by-one measurements has enabled the recording of oscillatory responses that are important for information processing in the brain.
The combination of PC surface and deep stimulation (A) continuously measured the plastic change (B) in the pattern of excitation propagation between PC and EC (C) and the emergence of responses in the EC (D). From [13]. The citation is permitted under CC-BY.
Recently, Genetically Encoded Voltage Indicators (GEVIs) have received a lot of attention [51–68]. Cell-specific measurements are now possible. For genetically encoded indicators, it is difficult to introduce their gene equally across a wide range of tissue through viral infection, so knock-in mice are used [69].
Summarizing, voltage imaging, which has a long history but faced many technical challenges, is reaching a point where it can be used by researchers to address a wide range of questions on very different levels of brain organization. We hope that voltage imaging will contribute to solving urgent health-related problems, such as dementia, in the near future.
N.A.
TT, RK, YT wrote the manuscript.
The evidence data generated and/or analyzed during the current study are available from the corresponding author on reasonable request.
This work was supported by JSPS KAKENHI Grant Numbers JP22H05698, JP21H03606, JP21K06702, JP21H03532, JP21H00447, JP20H04341, JP19H01142, JP16H06532 to TT, 19K12190 to RK and JP21K15247 to YT. The authors would like to thank Enago (www.enago.jp) for English language review.
