Corresponding author S. Kasparov: School of Physiology and Pharmacology, School of Medical Sciences, University of Bristol, Bristol BS8 1TD, UK. Email: email@example.com
We briefly review the current literature where optogenetics has been used to study various aspects of astrocyte physiology in vitro and in vivo. This includes both genetically engineered Ca2+ sensors and effector proteins, such as channelrhodopsin. We demonstrate how the ability to target astrocytes with cell-specific viral vectors to express optogenetic constructs helped to unravel some previously unsuspected roles of these inconspicuous cells.
Optogenetics is an emerging technology combining optical methods and molecular biology, which can be used either to monitor optically various processes in the cells of interest or to control their activity by light. This brief review summarizes recent advances achieved in both of these areas of optogenetics and discusses how their implementation may help our understanding of how astrocytes communicate with each other and with adjacent neuronal networks.
For a long time, glial cells were considered passive elements in the brain, providing structural and metabolic support to neurones. Novel experimental evidence, however, suggests that glial cells are anything but passive; they are now regarded as critical elements that sense and respond to neuronal activity and actively participate in information processing (Araque et al. 1999). Astrocytes are the most abundant type of glial cell in the brain and are closely associated with both blood vessels and neurones. A single astrocyte may contact numerous synapses (Bushong et al. 2002), potentially regulating neuronal activity and synaptic transmission (Pascual et al. 2005; Haydon & Carmignoto, 2006).
Astrocytes lack the ability to generate action potentials and thus do not communicate via propagating electrical signals. Instead they respond to stimulation by intracellular calcium [Ca2+]i transients or waves, which are frequently referred to as ‘Ca2+ excitability’ and which have been recorded in vivo (Hirase et al. 2004; Wang et al. 2006; Dombeck et al. 2007; Bekar et al. 2008; Schummers et al. 2008; Shigetomi et al. 2010), in vitro (Cornell-Bell et al. 1990) and also in human brain slices (Oberheim et al. 2009) using conventional chemical Ca2+ indicators. Therefore, Ca2+ signalling in astrocytes is seen as an important clue to their physiological roles. Due to the limitation in current methodology, the measurement of [Ca2+]i in small-volume compartments, such as astrocytic processes and perimembrane microdomain transients, has been somewhat hampered; instead most studies have concentrated on the total cytoplasmatic [Ca2+]i in somatic regions of astrocytes (Grosche et al. 1999; Nett et al. 2002). To resolve perimembrane Ca2+ microdomains with the fluorescent indicators fluo-4 or x-Rhod-1, a special technique, such as total internal reflection fluorescence, was necessary (Marchaland et al. 2008). In contrast to the conventional chemical Ca2+ indicators, genetically encoded Ca2+ sensors can be expressed in astrocytes selectively (Gourine et al. 2010) and targeted to specified cellular compartments, for example plasma membranes, by using well-characterized targeting motifs (Shigetomi et al. 2010). Recently published data based on these approaches are presented in the first part of this review.
Optogenetics also offers numerous light-sensitive proteins as tools for selective activation (and possibly deactivation) of astrocytes (Nagel et al. 2005). Such tools can be used to manipulate these cells in the context of heterogeneous brain tissue and shed light on what physiological responses they trigger in vitro and in vivo. Astrocytes can affect the functions of neuronal networks in a variety of ways. A full account is beyond the scope of this review, but briefly, the currently discussed hypotheses include the following: release of ‘glio-transmitters’, such as ATP, glutamate and d-serine; increased or decreased production of lactate, which may be an important metabolic substrate and messenger in astrocyte–neurone communication (Magistretti, 2006); changes in the activity of uptake mechanisms; and changes in prostaglandin signalling (Gordon et al. 2007, 2008). Optogenetic control of astrocytic activity and function is discussed in the second part of this brief review.
Finally, as all successful optogenetic experiments on astrocytes in vivo published to date have used cell-specific viral vectors for gene delivery, we will briefly describe the currently available options.
Optical analysis of astrocytic [Ca2+]i signalling
Genetically encoded calcium indicators (GECIs) consist of one or two fluorescent proteins (FPs) and a Ca2+-sensitive domain. In the presence of Ca2+, GECIs respond by altering their fluorescence intensity or by a wavelength shift.
Single green fluorescent protein (GFP)-based biosensors. The chromophore of GFP is enclosed in a tight β-barrel structure, which is essential for GFP to be fluorescent and makes access to its internal space particularly difficult (Yang et al. 1996). Yet, interruptions at certain positions, circular permutation or even insertion of foreign proteins may still result, with appropriate optimization, in functional constructs. The most common single GFP-based biosensor platform relies on circularly permuted GFP (cpGFP; Baird et al. 1999), in which GFP is rearranged so that the original N- and C-termini are connected with a flexible peptide linker and another peptide bond is interrupted to form new N- and C-termini that contain ‘sticky ends’: the Ca2+-binding motif from calmodulin (CaM) and its target binding protein, M13 (derived from myosin light chain kinase; Fig. 1A). The linker sequences used to fuse M13 and CaM to the cpGFP moiety and the linker in the middle of the molecule are critical for tuning the response properties of cpGFP-based Ca2+ indicators (Leder et al. 2010). An ideal indicator from this family should increase its fluorescence proportionally to Ca2+ concentration (Fig. 1B, left panel).
In 2001, Nakai et al. generated a construct which they named GCaMP (Fig. 1A). The first version of GCaMP displayed dim fluorescence at resting states, poor folding, and slow maturation at 37°C, as well as non-linear bleaching (Nakai et al. 2001; Lin et al. 2004). Furthermore, in the presence of Ca2+, the GCaMP absorption spectrum changed (Nakai et al. 2001). Subsequent modifications led to generation of GCaMP1.6 (Ohkura et al. 2005) and GCaMP2 (Tallini et al. 2006), demonstrating a five- and sixfold fractional fluorescence change, respectively, of its predecessor (please note that these values reflect the results of in vitro tests in cell-free systems). The most recently published addition to that family is GCaMP3, which has a further improved dynamic range (10-fold increase in fluorescence between resting and stimulated state; Tian et al. 2009).
In 2007, another cpGFP-based indicator was introduced (Souslova et al. 2007), named Case12 because of its exceptionally high contrast ratio (12 times that of resting to fully Ca2+-bound state). This sensor so far is the only one successfully used for astrocytic [Ca2+]i imaging not only in vitro (Guo et al. 2010) but also in vivo (Gourine et al. 2010). Given that in cultured astrocytes untargeted GCaMP2 was not very effective (Shigetomi & Khakh, 2009), it could be that the lower Ca2+ affinity of Case12 may be advantageous in vivo.
Current cpGFP-based Ca2+ sensors have high dynamic ranges but they also have limitations. Firstly, opening of the fluorescent barrel allows entry of protons into the internal cavity and inevitably makes them pH sensitive. For Case12 and GCaMP1.6, the pKa is ∼7.2, which is quite close to the physiological range (Souslova et al. 2007). For GCaMP2 and GCaMP3, we were unable to find this information, but it could be postulated that they are also pH sensitive. Such pH sensitivity can complicate interpretation of some experiments, as illustrated below (see section ‘Monitoring[Ca2+]isignalling in astrocytes using GECI’). Secondly, photostability of cpGFP may be insufficient for some applications. Case12 undergoes a fully reversible quenching when exposed to high-intensity blue light (Souslova et al. 2007), but its fluorescence recovers fully after a few seconds of darkness. Tian et al. (2009) imaged GCaMP3 using a two-photon excitation microscope with 10 mW of laser power at the specimen, and in this test the photostability of GCaMP3 was superior to some other FPs. However, because illumination was not continuous but interspersed by 30 s episodes of darkness, it is difficult to correctly assess the dynamics of its bleaching and/or reversible quenching.
Förster resonance energy transfer (FRET)-based sensors. Förster resonance energy transfer (FRET) was, in fact, the first successful approach to generation of Ca2+ biosensors. In 1997, two groups reported constructs where CaM and M13 motifs were used to alter the conformation of a molecule which incorporates a FRET pair of FPs with different excitation and emission characteristics [Miyawaki et al. 1997; Persechini et al. 1997; Romoser et al. 1997; Fig. 1B (right panel) and C]. The disadvantage of CaM-based Ca2+ sensors is that neurones in particular may express other potential substrates with which Ca2+-bound CaM may interact, and this applies equally to the cpGFP-based sensors discussed in the previous section. Such interaction will prevent the conformational change required to induce FRET upon [Ca2+]i increase (Palmer & Tsien, 2006). The most recent constructs from this subfamily were reported to detect Ca2+ transients triggered by single action potentials (Wallace et al. 2008; Horikawa et al. 2010). One elegant way around the problem of unsolicited interactions of CaM with endogenous proteins is to use troponin C as a Ca2+-sensing moiety, because outside of the skeletal muscle it seems to have no protein targets with which to interact. This approach has been realised in a construct named TN-XXL (Mank et al. 2008; Fig. 1D). The FRET-based sensors also work in membrane-targeted fusions (Nguyen et al. 2010). To the best of our knowledge, these indicators have not yet been used to monitor astrocytic [Ca2+]i, although they may offer some advantages, for example much better pH stability.
Monitoring [Ca2+]i signalling in astrocytes using GECI. The biosensor GCaMP2 (Tallini et al. 2006) was used to image [Ca2+]i signals in microdomains at the plasma membrane of cultured astrocytes by fusing it to the subunits of the Na+ pump (Lee et al. 2006). This fusion helped to reveal highly focal Ca2+ signalling in the areas of the membrane adjacent to the ‘junctional’ endoplasmatic reticulum (Lee et al. 2006). In a recent study, GCaMP2 and GCaMP3 were fused to the membrane-targeting domain of the tyrosine kinase Lck, which contains palmitoylation and myristoylation domains acting as membrane anchors (Shigetomi et al. 2010). In cultured astrocytes, only Lck-targeted sensors revealed localized [Ca2+]i transients in domains as large as 5–10 μm near the plasma membrane (consistent with Marchaland et al. 2008). These transients either occurred apparently spontaneously, in which case they were insensitive to a blocker of neuronal action potentials, tetrodotoxin (TTX), or could be evoked by activation of co-cultured neurones by current pulses, in which case they were sensitive to TTX and likely to be mediated by ATP (Shigetomi et al. 2010).
We have generated an adenoviral vector (AVV) that selectively expresses Case12 in astrocytes (AVV–sGFAP–Case12), using an amplified shortened promoter of the glial fibrillary acidic protein (GFAP; see section ‘The delivery systems’). The affinity of Case12 for Ca2+ is lower (Kd≈ 1 μm) than that of GCaMP2 and GCaMP3 (∼50 nm;Shigetomi & Khakh, 2009). This might be advantageous for Ca2+ imaging in astrocytes because we easily detect spontaneous Ca2+ activity in cultured astrocytes with Case12, while Shigetomi et al. 2009 failed to reveal any. If this supposition is correct, then the very recently developed nano-Cameleons (Horikawa et al. 2010) could be best suited for imaging neurones but not astrocytes. When expressed in astrocytes (Fig. 2), Case12 sensitivity compares to that of the conventional chemical Ca2+ indicator, Rhod-2 AM (Gourine et al. 2010; Guo et al. 2010). Case12 fluorescence faithfully follows the signal derived from a conventional indicator, fura-2 (Fig. 2C). At the same time, Case12 is pH sensitive (note that the intensity of the signal falls below baseline at the end of the pH stimulus in Fig. 2Ba and C). Case12 visualizes changes in [Ca2+]i not only in somata but also in the finest astrocytic processes (Gourine et al. 2010). Guo et al. (2010) used AVV–sGFAP–Case12 to study the effects of one of the ‘neurohormones’ of the angiotensin family, angiotensin(1–7) [Ang(1–7)]. The application of this viral vector to brainstem slice cultures resulted in selective expression of Case12 in astrocytes, thus defining the source of the optical signal in the imaging experiments. In addition, neurones were labelled with other viral vectors incorporating different promoters. We were able to show that Ang(1–7) activates a fraction of astrocytes without affecting local neurones. One great strength of the viral vector approach is its applicability to different animal models, and we were able to compare the effects of Ang(1–7) between normotensive and hypertensive rat strains in vitro (Guo et al. 2010).
The AVV–sGFAP–Case12 was also used in vivo to reveal the chemosensitivity of a specific subset of astrocytes located on the ventral surface (VS) of the medulla oblongata (Gourine et al. 2010). We found that a 0.2 pH unit decrease on the VS of anaesthetized and artificially ventilated rats evoked an immediate increase in [Ca2+]i across the field of VS astrocytes. In contrast, in cultured slices, we observed that acidification-induced Ca2+ excitation of VS astrocytes was insensitive to TTX and muscimol (Fig. 2B), two agents which silence retrotrapezoid nucleus neurones, the only known type of pH-responsive neurones in this area. Hence, we were able to pinpoint local astrocytes as the primary chemoreceptors of the VS. Moreover, using [Ca2+]i responses in astrocytes as a means of detecting ATP-mediated signalling, we found that acidification triggered release of ATP. Indeed the ATP-degrading enzyme apyrase almost completely abolished these [Ca2+]i waves (Gourine et al. 2010). Therefore, propagation of pH-evoked Ca2+ excitation among ventral medullary astrocytes is largely mediated by ATP acting on a subset of P2Y receptors.
In summary, GECIs have proved to be highly valuable tools for in vitro and in vivo imaging of astrocytic [Ca2+]i, and any new members of this family developed to overcome the current limitations hold high promise for future research.
Optical control of astrocytic [Ca2+]i and transmitter release
Channelrhodopsin-2 (ChR2). To control their functional activity optogenetically, light-activated constructs can be targeted to selected cell types. A number of recent reviews summarize origins, features of currently available variants of the channelrhodopsin family and their application to experiments on neurones (see other reviews in this issue of Experimental Physiology). Channelrhodopsin-2 (ChR2) is an algal, light-gated, cation-selective membrane channel. Its molecule consists of a seven-transmembrane-spanning apoprotein, channelopsin, and retinal, which binds covalently to it (Bamann et al. 2008; Wang et al. 2009). After absorption of a photon, ChR2 opens rapidly to form a pore permeable to monovalent (Na+, K+ and H+) and divalent cations (Ca2+). The high-light-induced cation conductance leads to strong and rapid depolarization of the membrane (Nagel et al. 2003). It is thought that the photoisomerization of all-trans-retinal to the 13-cis configuration induces the protein to change its conformation and open the channel (Hegemann, 2008). Various mutants of ChR2 are currently available (see Lin, 2011), including ChR2(H134R) (Nagel et al. 2005; Gradinaru et al. 2007). We have selectively expressed ChR2(H134R) in astrocytes under the control of an enhanced GfaABC1D promoter (see section ‘The delivery systems’). In order to visualize transduced astrocytes without concomitant excitation, we fused ChR2(H134R) to a red-shifted fluorescent protein, Katushka1.3 (later re-named as Katushka 2). Katushka1.3 has its emission peak at ∼630 nm (Shcherbo et al. 2009) and can be excited by green or yellow laser light, thus avoiding activation of ChR2(H134R) by blue light. Moreover, the spectral properties of Katushka1.3 allow us to use Rhod-2 AM, a red-shifted Ca2+ indicator, in combined experiments with ChR2(H134R) (Fig. 3A).
In vitro activation of astrocytes was achieved with moderate-intensity blue light (470 nm laser diode microscope illumination system from Rapp OptoElectronic, Wedel, Germany). In order to illuminate the cells and at the same time continue using green (543 nm) or yellow (568 nm) laser, we modified the light path of our confocal imaging system (SP2 Leica confocal upright microscope) and added an additional dichroic mirror (Fig. 3B).
Validation of AVV–sGFAP–ChR2(H134R)–Katushka1.3. To verify the ability of our vector, AVV–sGFAP–ChR2(H134R)–Katushka1.3, to excite astrocytes, several in vitro tests were performed. Firstly, we loaded dissociated cultured astrocytes with Rhod-2 AM and monitored fluorescence using an SP2 Leica confocal microscope (Fig. 3B). Flashing blue light (470 nm, 20 ms–20 ms duty cycle) was used to activate ChR2(H134R). This caused increases in [Ca2+]i in transduced astrocytes (Fig. 3C). It is important to note that in different experiments the speed of [Ca2+]i rises and their dynamics (fast, slow recovery or no recovery during the observation period) were different, and this depended on the expression level of ChR2(H134R) as well as the light intensity used for stimulation. In order to determine the source of the [Ca2+]i, astrocytes were superfused with Ca2+-free buffer with added EDTA. This resulted in an almost complete elimination of light-induced [Ca2+]i increases, suggesting their dependence on the influx of Ca2+ from the extracellular space (Fig. 3D).
The next step was to determine whether the new vector could be used to trigger release of ATP in a more integrated preparation. Organotypic brain slices were prepared, as previously described (Teschemacher et al. 2005b), and transduced with AVV–sGFAP–ChR2(H134R)–Katushka1.3. After 7 days, slices were placed into a tissue chamber, continuously superfused at 34°C, and ATP levels in the outflow of the chamber were measured using a bioluminescent assay. Illumination with 445 nm light at ∼7 mW mm−2 evoked a 95% increase in ATP release (Fig. 3E; Gourine et al. 2010). This demonstrated directly that this vector is able to control astrocytic ATP release.
In vitro and in vivo application of AVV–sGFAP–ChR2(H134R)–Katushka1.3.In vitro AVV–sGFAP–ChR2(H134R)–Katushka1.3 has been used to study the role of astrocyte to neurone signalling in the retrotrapezoid nucleus (RTN) region. This nucleus is located in the lower brainstem and implicated in the process of central chemosensitivity that enables the brain to adjust respiratory activity to maintain stable levels of CO2 and pH (Loeschcke, 1982; Mulkey et al. 2004). We hypothesized that chemosensitive astrocytes in that area may be able to excite local RTN neurones via release of ATP (Fig. 4A). We took advantage of the fact that RTN neurones can be selectively targeted with the PRSx8 promoter (Abbott et al. 2009). Two vectors were applied simultaneously to organotypic slices containing the RTN area: AVV–sGFAP–ChR2(H134R)–Katushka1.3 for optogenetic control of astrocytes and AVV–PRSx8–DsRed2 to label RTN neurones with DsRed2 for identification and patch-clamp recordings.
Stimulation of local astrocytes transduced with AVV–sGFAP–ChR2(H134R)–Katushka1.3 by flashing light evoked lasting depolarizations of patched DsRed2-labelled RTN neurones (Fig. 4B). In the presence of the ATP receptor blocker, MRS2179 (10 μm), these depolarizations were reversibly prevented (Gourine et al. 2010), indicating a pivotal role for ATP in this process.
In vivo, AVV–sGFAP–ChR2(H134R)–Katushka1.3 was used to test whether optogenetic excitation of local astrocytes in the RTN area can trigger a chemoreceptor-like response. In anaesthetized, vagotomized and artificially ventilated rats pre-injected with vector 7–10 days prior to the experiment, the VS was exposed and phrenic nerve activity recorded to monitor central respiratory drive. Unilateral illumination (pulsing 445 nm light from a laser source) of the transduced side of the VS evoked robust respiratory activity from hypocapnic apnoea (Fig. 4C). Furthermore, consistent with the outcome of in vitro experiments, application of MRS2179 (100 μm) reversibly inhibited the respiratory effects of optogenetic stimulation of astrocytes, confirming the involvement of ATP (Gourine et al. 2010). The intensity of light stimulation and the titres of viral vectors used in these experiments had to be carefully adjusted in order to achieve good reversibility and reproducibility of the physiological responses. Use of very high titres of vectors (>1011 transducing units ml−1) resulted in overexpression and cell death.
The delivery systems
The studies outlined above illustrate the power of optogenetics for studies of astrocyte-to-neurone communication. The ability to target astrocytes selectively and, moreover, at the same time target neurones with a different construct opens a tremendous window of opportunity for future research in this direction. These experiments also demonstrate the potential of cell-specific viral targeting as an experimental strategy for neuroscience. For many of the cell types present in mammalian CNS, there are fairly short and specific promoters which can be used in viral backbones, and it may be predicted that in the future their arsenal will expand. The ability to selectively target two phenotypes (e.g. astrocytes and neurones) at the same time by simply mixing two viral vectors compares favourably in terms of speed, costs and flexibility to a protocol where a double transgenic animal has to be generated.
Many cell-specific mammalian promoters are fairly weak compared with the non-selective viral promoters that are highly successful in cell lines. The promoter of the GFAP gene has been used for many years to target astrocytes, and there are several transgenic mice with transgenes under control of this promoter (Pascual et al. 2005; Potokar et al. 2009). However, as mentioned previously, GFAP promoter sequences used for gene targeting have low transcriptional activity. Adeno-associated vectors partly offset this problem by invading the target cells in very high copy numbers (Ortinski et al. 2010), but we choose to use AVVs because of their important advantages for experimentation in slice cultures (high transducing efficiency in vitro, speed of expression, and ease of preparation of large quantities), in addition to being very efficient in vivo (Teschemacher et al. 2005a), especially for targeting astrocytes (Duale et al. 2005). Transcriptional amplification strategy (TAS) can be used to enhance the activity of cell-specific promoters without loss of cell-type specificity (Liu et al. 2008). For the development of the new viral vector, we used the shortened GfaABC1D (694 bp), which has the same level of specificity as the longer versions of GFAP promoter (Su et al. 2004; Lee et al. 2008). A two-step TAS was used to enhance transgene expression of GfaABC1D (Liu et al. 2008; Fig. 4D). Transcriptional amplification strategy employs a minimal core promoter (65 bp) derived from the human cytomegalovirus (CMV), which is joined upstream of the GfaABC1D promoter in antisense orientation. Minimal CMV (mCMV) drives the expression of an artificial transcriptional enhancer (GAL4BDp65; Liu et al. 2006), which then interacts with unique artificial binding sites (GAL4BS) upstream of the cell-specific promoter, leading to an enhancement of the transcription of the ChR2(H134R)–Katushka1.3 cassette (Fig. 4D). Importantly, the mCMV-driven expression is still strongly influenced by the GfaABC1D promoter, which acts bi-directionally in such constructs (Amendola et al. 2005). Therefore, the enhancer expression is also selective to astrocytes. The GAL4 binding sites do not exist in the mammalian genome, and this prevents off-target amplification and potential bias in gene expression. The TAS strategy is not restricted to the GFAP promoter and has been successfully used to amplify other cell-specific promoters in our laboratory, for example synapsin-1 (Liu et al. 2008) and tryptophan hydroxylase II (Benzekhroufa et al. 2009). We hope that its wider implementation will aid further design of highly efficient viral vectors for optogenetic experimentation.
The development of optogenetic tools has made considerable progress in the last few years. This review has briefly summarized current data on the application of genetically encoded Ca2+ sensors and ChR2-derived light-sensitive effectors to studies of astrocytes, thus illustrating how they could potentially aid in our understanding of astrocyte-to-neurone communication. Although biosensors have to yet outperform the best synthetic dyes, many applications in cell biology and physiology are only possible with GECIs. One invaluable advantage of these indicators is the possibility to express them selectively in precise cell types, which permits identity of the cellular source of the fluorescent signal in vitro and in vivo with confidence. The theoretical limits for GECI performance have not yet been reached, and improvements certainly come with the advent of even brighter and possibly red-shifted FPs. This last point is particularly important because it will allow much deeper penetration into the brain tissue without the need for infrared lasers and is expected to improve signal-to-noise levels. We have also reviewed the optogenetic approach to control astrocytic Ca2+ signalling and transmitter release in vitro and in vivo. Expressing ChR2(H1234R) selectively in astrocytes enabled us to trigger in these cells [Ca2+]i elevations, release of ATP, downstream events in RTN neurones and an increase in respiratory activity (Gourine et al. 2010).
As the field of optogenetics matures, more tools transpire (Airan et al. 2009; Looser et al. 2009). These constructs are similar to ChR2 in that they are activated by light but, unlike ChR2, they impinge on the intracellular signalling rather than acting via opening of ion channels in the plasma membrane. With the aid of these new tools, we shall have more ways to improve our understanding of how astrocytes signal to each other and to the surrounding neuronal networks.
S.K., A.G.T., M.F., B.H.L., J.H., A.V.G., N.M. and V.K. are supported by Wellcome Trust and British Heart Foundation. S.L. is supported by Biotechnology and Biological Sciences Research Council. D.M.C. and E.A.S. were supported by Molecular and Cell Biology program Russian Academy of Sciences.