- Email: [email protected]
A molecular and genetic arsenal for systems neuroscience
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
A molecular and genetic arsenal for systems neuroscience Edward M. Callaway Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA
Neural circuits are composed of a meshwork of numerous neuron types, each with their own distinctive morphological and intrinsic physiological properties, connectivity and biochemistry. How do distinct neural subcircuits, composed of different classes of neuron, contribute to brain function? Approaching this question requires methods that can target specific neurons types. This can be achieved by harnessing the same machinery that builds sophistication into the brain and using it to make novel tools for investigating and manipulating the brain: molecular and genetic technology. These tools can be used to target gene expression to specific neuron types within complicated neuronal circuits, and the transgenes that are expressed can be used to elucidate and manipulate these circuits with unprecedented precision and control. These methods are likely to become the archetype for future studies linking perception, cognition and behavior to specific components of the brain. Introduction Systems neuroscience aims to understand how complex interactions between networks of neurons give rise to perception and behavior. This is a daunting task. The primary difficulty is that neural circuits are composed of a complex network of numerous neuron types whose dendrites and axons are intricately intertwined. This creates unique challenges for separately identifying the contributions of each cell type to the intact functioning network. Traditional methods have contributed a great deal to our understanding of brain function, but they fail in their ability to approach questions at this level of specificity. This review will focus on methods being developed that use our understanding of cell and molecular biology to attack questions in systems neuroscience. The primary focus will be on methods that will enable the organization and function of complex neural circuits to be investigated at the level of specific cell types. There has been considerable progress in the development and use of molecular and genetic methods for studying the nervous system in mice. The focus of this review is on how these and related methods can be used in higher mammals, particularly monkeys, in which transgenic technology is not practical owing to the long gestation period (w165 days) and time to reach sexual maturity (several years). The ability to take advantage of Corresponding author: Callaway, E.M. ([email protected]).
molecular and genetic methods in primates should enable studies of more complex behaviors and brain functions than are found or easily studied in mice. Many practical considerations influence decisions about what methods are best to achieve any given experimental goal, and many of the genetic methods that have been developed for studies of the mouse nervous system have been based on decisions that apply more to mice than to monkeys. But, despite these differences, there is also a good deal of common ground. There are two primary advantages to genetic approaches, regardless of species. These are the ability to target specific cell types within complex tissues and the ability to perform complex manipulations with great repeatability. These two goals are interrelated. With transgenic mice, transgene expression is often targeted to specific cell types in the brain and expression is consistent across animals in the same genetic line. Thus, large numbers of precisely manipulated animals can be generated and studied. By contrast, the use of large numbers of primates is neither practical nor desirable. Primates are a precious resource and considerable investment is required for their breeding, care and training. Primate-based studies are therefore designed to minimize the numbers of animals used and it is typical to train a small number of animals on a specific task and then study those animals extensively – often for years. Thus, genetic methods for studying primates must address this reality by developing and using manipulations that are stable over long periods of time. This strategy allows for repeatability in primate models, by enabling long-term investigation of a single animal in which the particular genetic manipulation has been very carefully characterized. Because the use of transgenic primates is impractical, the most useful methods presently available for delivery of genetic material in these animals are based primarily on viral vectors [1,2]. Many of these vectors were developed for gene therapy in humans and therefore incorporate features that are ideal for long-term studies in primates; most notably, they can efficiently transduce nearly every neuron in the vicinity of a brain injection and can yield stable gene expression for months or years without toxicity [3–7]. The main part of this review will therefore begin with a brief overview of properties of some of the most promising viral vectors and considerations involved in the choice of vector.
www.sciencedirect.com 0166-2236/$ - see front matter Q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.tins.2005.01.007
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
Cell-type-specific targeting is one of the chief advantages of genetic methods. The need for cell-type specificity has become apparent from our increasingly sophisticated understanding of the nervous system. For example, it is now clear that when a brain structure contains multiple cell types with interwoven dendritic arbors, afferent input to that structure can selectively connect to some cell types while avoiding others [8]. And it should be self-evident that when there are multiple cell types, each type is likely to have a unique role in the neural network. Conventional methods have allowed tracing of connections and monitoring or manipulation of neuronal activity, but these methods typically do not have the resolution required to address questions at the level of specific cell types. Methods by which gene expression can be targeted to particular cell types of interest will therefore also be discussed in this review. Finally, once gene expression is achieved in the targeted neuronal population, it is useful only if the gene that is expressed provides insight into the organization or function of the nervous system. Thus, the remainder of the review will summarize recent progress in the development and use of genetic methods for achieving three goals: elucidation of neural circuits, monitoring of neural activity, and manipulation of activity. Traditional methods have achieved these, but not at the level of cell-type specificity or with the repeatability that can be achieved using genetics. Viral vectors for gene delivery The ideal vector for transgene delivery depends on the experimental goals. It should be clear that a first consideration is that the chosen vector must be able to deliver genetic material efficiently to the cell type of interest (i.e. the cells must be susceptible to the vector). Furthermore, once the genetic material has entered the cell, the transgene must be expressed (i.e. the cell must be permissive for expression). For the typical systems neuroscientist, it might be surprising to learn that there is not one best, universal vector that can deliver DNA (or RNA) to all cell types. It is beyond the scope of this review to discuss the known tropisms of each viral vector, so the interested reader is referred to selected references describing adenoassociated viruses (AAV) [9–12], lentiviral vectors [6,7] and herpes simplex virus (HSV) amplicon vectors [5]. These are all replication-incompetent vectors that can deliver DNA (or RNA) efficiently to non-dividing adult neurons, potentially yielding transgene expression in nearly every neuron within w0.5–1.0 mm of the injection site (depending on viral titers and volumes). Because the vectors are replication-incompetent, there is no further spread and they enable stable transgene expression without apparent toxicity. A first step in any course of study would be to test empirically which viruses can transduce the cell type of interest. Viral vectors have many advantages but they also have limitations, and the tropism of viruses (not all cell types are both susceptible and permissive for a given vector) could be considered as both. When a particular virus does not infect or generate transgene expression in the cell type of interest, it is a limitation. But tropism can also be used www.sciencedirect.com
197
to target particular cell types of interest [9–11,13–15]. Designer viruses have also been manufactured specifically to create tropism for a particular cell type or to expand the range of cell types that can be transduced [12–14,16]. The most prominent limitation of viral vectors is their capacity. This is most relevant for cell-type-specific promoters, which are often too large to fit into viruses (as will be discussed later). AAV and lentivirus can package genetic sequences up to w5 Kb and w8 Kb, respectively [17,18]. Wild-type HSV has a much larger genome of 152 Kb, and HSV amplicon vectors can therefore, in theory, package amplicons up to this size. In practice, viral titers can be reduced when amplicons Ow10 Kb are packaged. Nevertheless, there have been recent successes in making high-titer HSV amplicons incorporating bacterial artificial chromosomes (BACs) that are 128–135 Kb [19,20]. Continued refinements are therefore likely to increase greatly the utility of HSV amplicon vectors, particularly for cell-type-specific gene expression. Targeting specific cell types Cell-type-specific targeting is one of the chief advantages of genetic methods, and the ways in which it can be achieved have practical considerations for development of the tools that will be discussed here. The most common way to achieve cell-type-specific gene expression is by utilizing a cell-type-specific promoter. For example, the gene encoding Purkinje cell protein 2 (Pcp2/L7) is expressed exclusively in cerebellar Purkinje cells and retinal bipolar cells [21–23]. The 2.9 Kb DNA sequence upstream from this gene has been used successfully to drive transgene expression preferentially in these cell types [24] (although small amounts of aberrant expression were also observed). There are innumerable other examples of success in generating mouse lines using cell-type-specific gene expression, but there are just as many (or probably many more) attempts that are unsuccessful. As a general rule, however, the probability of success and degree of specificity are greatly increased when larger stretches of potential regulatory sequences are used. BAC transgenics, which can introduce O200 Kb of DNA into the germ lines of transgenic mice, have been most successful [25]. For example, improved specificity of transgene expression in Pcp2-expressing cells was achieved using a 173 Kb BAC [26]. The difficulty arises primarily because the DNA sequences and mechanisms that regulate gene expression are not fully understood. To target gene expression to specific cell types using cell-type-specific promoters in viral vectors, it is of course first necessary to identify a vector that can transduce the cell type of interest. But because other cell types in the vicinity will also be susceptible and permissive to the vector, the cell-type-specific promoter is necessary to restrict transcription of RNA from DNA to the cell type of interest (Figure 1). It should therefore be clear that among the most important differences between work with transgenic mice and gene delivery by viruses is the limited capacity of viruses. Because there is no definitive method for identifying the minimal DNA sequence necessary for efficient and selective gene expression in a particular cell type, it is often very difficult to find short sequences that
Review
198
TRENDS in Neurosciences Vol.28 No.4 April 2005
(a)
(c) Cell-type-specific promoter
Molecular switch
(b) (i)
(ii)
• K+ channels open when AL applied → excitability decreases • Normal excitability when AL not present
(iv)
(iii)
TRENDS in Neurosciences
Figure 1. Example of experimental and conceptual steps involved in targeting a specific neuron type for reversible inactivation using a viral vector. (a) A viral vector is constructed to carry the genetic coding sequence for a transgene, in this case the gene encoding a molecular switch. This coding sequence is downstream of a genetic regulatory sequence, or promoter, that will permit and/or drive expression of the transgene only in a specific cell type. (b) The steps involved in generating specific expression of the genetic switch and reversible neuronal inactivation begin with the injection of virus into the brain region of interest. (i) Billions of viral particles (e.g. 1 ml of particles at 1012 particles mlK1) are injected into the brain region of interest, which here contains three neuron types (purple, blue and green), including the cell type that is being targeted (green). (ii) The viral vector efficiently delivers its genetic material to nearly every neuron within w0.5–1.0 mm of the injection site, including all three neuron types (neurons now shown in brown). Larger regions can be influenced by viral injection at more locations. (iii) Expression of the transgene (genetic switch) is restricted to the targeted cell type (now shown in yellow) because only this cell type drives the promoter that was included in the viral construct. (iv) When the switch is flipped (c), the targeted cell type is selectively inactivated. Because this is a reversible inactivation method, the neurons can return to the active state. (c) One molecular switch that has been developed is the Drosophila allatostatin receptor (AlstR). When AlstR is expressed in mammalian neurons, it initially has no effect on their activity, presumably because the mammalian brain lacks an endogenous ligand for the insect receptor. However, when the Drosophila neuropeptide allatostatin (AL) is injected in the vicinity of the neurons expressing the AlstR, the AL flips the genetic switch (inset). AL binds to AlstR, triggering a conformational change and G-protein-mediated opening of inwardly rectifying KC (GIRK) channels. The increase in KC conductance causes neuronal inactivation. The effect of AL is specific for the AlstR-expressing cells and therefore confers selective neuronal inactivation. When AL is removed by diffusion or brain peptidases, normal neuronal function returns.
work. Trial and error is most common and often simply boils down to trying the largest sequence upstream from the coding sequence that can still fit in the vector. This typically leads to exercises in success or failure rather than trial and error. Hopefully, future studies will lead to better methods for predicting the relevant regulatory sequences. For example, crucial regulatory elements might be identified by comparing sequences across species to identify conserved motifs, or by identifying common sequences across multiple genes that are all expressed selectively in a specific cell type [27,28]. Cell-type-specific promoters are not the only way to target gene expression to specific neuron types within complex neuropil. Neuron types can differ not only in the genes they express but also in the distant structures to which they project their efferent axons. In recent experiments, a lentivirus has been pseudotyped with the envelope protein from rabies virus, enabling it to infect neurons efficiently through their axon terminals [29,30]. HSV amplicons can also infect neurons via their axon terminals [5]. Thus, specific types of projection neurons can be infected by injection of an appropriate viral vector into distant axonal arbors. Genetic methods for tracing connections of specific cell types The most advanced methods for tracing neuronal connections, in terms of cell-type specificity and ease of use, include those based on expression of gene products that www.sciencedirect.com
are transported either anterogradely or retrogradely and cross synapses. These include wheat germ agglutinin (WGA), which is transported primarily anterogradely [31–33], and tetanus toxin C fragment (TTC), which is transported retrogradely [34,35]. Both of these methods have been used successfully to trace connections from specific cell types in transgenic mice, either by placing tracer gene expression under the control of a specific promoter or by making expression dependent on Cre recombination that is restricted to cell types of interest. This approach has also been successfully implemented using replication-incompetent adenovirus delivery of genetic constructs that carry coding sequence for the tracer [32]. Transported tracer can be detected either using antibodies specific for the tracer or using expression of green fluorescent protein (GFP)–tracer fusion proteins. It should also be possible to use these methods in primates, with gene delivery mediated by viral vectors. Despite the usefulness of these methods, they have two major, related limitations. Weak connections can be difficult or impossible to detect, and label resulting from direct or monosynaptic connections cannot usually be distinguished from label arising from multisynaptic pathways. In particular, strong multisynaptic connections could result in detection of more transynaptic labeling than weak monosynaptic connections. An alternative to genetically expressed transynaptic tracers is replication-competent neurotropic viruses, which progress transynaptically through the nervous
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
system as part of their natural life cycle [36]. Three neurotropic viruses are used most commonly for transynaptic labeling studies – pseudorabies virus (PRV; despite the name, this is an a-herpes virus), HSV and rabies virus (RV). These viruses have been used extensively as transynaptic tracers and, depending on the type of virus and strain, they can be transported selectively in either the anterograde or retrograde direction [21,36–39]. Used in their replication-competent form, they lack celltype specificity but have the advantage of amplifying signals by replicating at each synaptic stage. Thus, weak connections are more likely to be detected than by using WGA or TTC. But it is still not always possible to distinguish weak, direct connections from strong disynaptic or polysynaptic connections [37]. De Falco et al. [40] greatly increased the utility of PRV by engineering the virus for cell-type specificity. They deleted from the viral genome the thymidine kinase (TK) gene, which is required for viral replication in neurons. They inserted into the viral genome a conditionally expressed TK gene and also the GFP coding sequence. In cells that express Cre recombinase, the viral genome is recombined so that TK and GFP are both expressed. In this way, the PRV replicates and expresses GFP only in parent cells that are targeted for Cre expression. This newly competent, recombined and GFP-expressing PRV can then spread retrogradely to connected cells, where it can further replicate and continue to spread. GFP is therefore expressed only in the multisynaptic chain of neurons connected to the genetically targeted parent cells. Although this method again suffers from difficulty in distinguishing order of connectivity, it might prove possible to restrict viral spread to monosynaptic connections by selectively expressing TK along with Cre in parent cells, and completely eliminating TK from the viral genome. One of the limitations (but also a safety feature) of PRV is that it does not infect primates [36], but it is possible to make similar genetic modifications of HSV [41], which does infect primates. Genetically expressed sensors of activity Among the most exciting potential applications of genetic technology to systems neuroscience is measurement of activity in specific neuron types using genetically encoded sensors [42–44]. Such sensors have been engineered for detection of membrane voltage in addition to Ca2C and other intracellular messengers, and they can also be used to measure vesicular release of neurotransmitter. Each of these signals can be directly or indirectly correlated with neuronal activity. At present, most physiological studies use metal electrodes to record from single neurons in complex structures, without knowledge of the cell types being studied. Genetic methods should allow optical recordings of activity from neurons of known type, perhaps even in awake, behaving primates. However, despite success in the development of probes that can yield signals detectable under in vitro conditions or in invertebrates, success with in vivo recordings in mammals has been extremely limited [45]. The most notable exception is a study in which synaptopHluorin [46] was expressed in mouse olfactory sensory www.sciencedirect.com
199
neurons [45]. Synapto-pHluorin is a genetically encoded pH sensor that indirectly reports activity at synaptic terminals when synaptic vesicles fuse and change their pH during transmitter release [46]. When expressed in transgenic mice under the control of an olfactory-neuronspecific promoter, synapto-pHluorin was detected optically in the terminal arbors in the olfactory bulb. Large fluorescence changes in the olfactory bulb could then be detected following olfactory stimulation. Although these studies represent a tremendous step forward, there are still considerable difficulties to be overcome. An ongoing challenge for these methods is the small size of the signals that are typically generated. Thus, with present technology, in vivo measurements are likely to require averaging over both time and neuronal populations. Future improvements are likely to enable recordings from identified single neurons to be made in vivo using two-photon microscopy. Ideally, temporal resolution will be sufficient to identify individual action potentials. Optical scanning across populations of neurons expressing genetic sensors might also enable measurement of the dynamic activities of identified neuronal assemblies [43]. Genetic methods for selective neuronal inactivation As the spatial resolution of methods for elucidating connectivity and measuring activity of neurons has improved, so too have methods for perturbing neural activity. To test the role of specific cell types within complex circuits it is necessary to target specific cell types for inactivation. Because genetic methods allow cell-type-specific targeting, inactivation methods based on expression of transgenes can be used for cell-type-specific inactivation. Although several in vitro studies describe potential genetic methods for neuronal inactivation (as will be discussed later in this article), the progression to in vivo studies is slow. Therefore, the most successful in vivo studies have used older methods that lack reversibility because the cells are killed. In 1995, Kobayashi et al. [47] described an immunotoxin-based method for selective neuronal inactivation; following selective expression of the interleukin-2 receptor a-subunit, those cells are killed by exposure to immunotoxin. Yoshida et al. [48] used this method, along with a promoter for the metabotropic glutamate receptor mGluR2, to kill selectivity retinal starburst amacrine cells. They were then able to infer the role of this cell type in generating the direction-selectivity of retinal ganglion cells. This method has also been used to kill cerebellar Golgi cells selectively [49]. Despite success with selective cell killing, reversible methods will be crucial for future studies. Such methods might avoid the possibility that functions measured after cell killing reflect compensatory mechanisms, and they are essential for experiments that require an intact system both before and after manipulations. As will be discussed, in vitro studies have been used to characterize reversible methods that are likely to work in vivo in the near future. Several genetic methods for neuronal inactivation can be made reversible on a slow timescale by regulating transcription of the transgene. These methods include overexpression of KC channels [50,51] and expression of
200
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
tethered toxins that can block ion channel function [52]. Such methods could be ideal for studies requiring longterm, reversible inactivation, such as investigations of plasticity or the roles of particular neuron types in learning paradigms that might develop over several days. But these methods are too slow for some other applications. For example, it would be useful to measure the visual receptive field of a neuron in the monkey visual cortex, then inactivate a particular neuron type, reevaluate the receptive field and recover normal function, all within the 1–2 h period in which activity of a single neuron can be recorded. Methods for quickly reversible inactivation are generally based on the expression of a gene that, by itself, does not affect function (Figure 1c). But when the gene product is activated by an exogenous compound, there is selective inactivation that reverses when the compound is washed away. A prototype of this class of genetic method was developed by Conklin and colleagues [53,54]. They modified a k opiate receptor gene by deleting the sequence encoding the part of the receptor that normally binds to natural opiates, but leaving intact a site activated by synthetic opiates. Conklin and colleagues called the engineered receptor a receptor activated solely by synthetic ligand (RASSL). When this RASSL was activated by synthetic opiates in vitro, it caused G-protein-mediated opening of KC channels [54], and when expressed in heart muscle in vivo, synthetic opiates reduced heart rate [53]. Although this system also causes KC channel opening in neurons in vitro (E.M. Callaway, unpublished) and could therefore potentially be used for neuronal inactivation in vivo, it might not be ideal for neuronal inactivation: synthetic opiates activate naturally occurring opiate receptors in the mammalian brain and thus would lack specificity for the cells genetically targeted for RASSL expression. This approach inspired the search for other receptors that can mediate increases in KC conductance via G proteins but that are not activated by ligands for which the mammalian brain has endogenous receptors. The ideal receptor would be activated by a ligand that does not activate any receptors endogenous to the mammalian brain. The insect allatostatin receptor (AlstR) appears to meet these criteria [55] (Figure 1c). When tested in mammalian neurons expressing this receptor, the peptide ligand allatostatin quickly and reversibly increased KC conductance, effectively silencing the neurons [56]. Control neurons were unaffected by allatostatin. A similar strategy for selective and reversible inactivation uses genetic expression of an invertebrate glutamategated ClK channel that has been mutated to make it insensitive to glutamate [57]. However, the channel remains sensitive to the medication Ivermectin, which can silence cultured neurons expressing the receptor. Reversibility occurs slowly, over hours, apparently due to tight binding of the drug to the receptor. Concluding remarks In summary, it is clear that molecular and genetic methods will have an increasingly important role and eventually will have an enormous impact in systems neuroscience. www.sciencedirect.com
However, many of the methods are still in their infancy. For example, there are several vectors for gene delivery that can efficiently deliver genetic material to neurons, but they are limited in their capacity. There are also many short promoter sequences that enable cell-type-specific gene expression, but more systematic and reliable approaches are needed for identifying short promoters to target gene expression to particular cell types. Cell-typespecific gene expression can be reliably used to kill neurons, but reversible inactivation in vivo, particularly on a fast timescale, is still in development. Genetically encoded transynaptic tracers are providing excellent new insights, but they could be improved for some applications by increasing efficiency and restricting spread to monosynaptic connections. And although genetically expressed sensors of neural activity work, they provide small and slow signals; such sensors are likely to be greatly improved in speed and sensitivity by further engineering. The road will be hard, but the future is nearly unlimited and the pioneers who choose this road will be rewarded.
References 1 Kootstra, N.A. and Verma, I.M. (2003) Gene therapy with viral vectors. Annu. Rev. Pharmacol. Toxicol. 43, 413–439 2 Davidson, B.L. and Breakefield, X.O. (2003) Viral vectors for gene delivery to the nervous system. Nat. Rev. Neurosci. 4, 353–364 3 Xiao, X. et al. (1997) Gene transfer by adeno-associated virus vectors into the central nervous system. Exp. Neurol. 144, 113–124 4 Rabinowitz, J.E. and Samulski, J. (1998) Adeno-associated virus expression systems for gene transfer. Curr. Opin. Biotechnol. 9, 470–475 5 Sandler, V.M. et al. (2002) Modified herpes simplex virus delivery of enhanced GFP into the central nervous system. J. Neurosci. Methods 121, 211–219 6 Naldini, L. et al. (1996) Efficient transfer, integration, and sustained long-term expression of the transgene in adult rat brains injected with a lentiviral vector. Proc. Natl. Acad. Sci. U. S. A. 93, 11382–11388 7 Blomer, U. et al. (1997) Highly efficient and sustained gene transfer in adult neurons with a lentivirus vector. J. Virol. 71, 6641–6649 8 Callaway, E.M. (2002) Cell type specificity of local cortical connections. J. Neurocytol. 31, 231–237 9 Davidson, B.L. et al. (2000) From the cover: recombinant adenoassociated virus type 2, 4, and 5 vectors: transduction of variant cell types and regions in the mammalian central nervous system. Proc. Natl. Acad. Sci. U. S. A. 97, 3428–3432 10 Rabinowitz, J.E. et al. (2002) Cross-packaging of a single adenoassociated virus (AAV) type 2 vector genome into multiple AAV serotypes enables transduction with broad specificity. J. Virol. 76, 791–801 11 Rabinowitz, J.E. et al. (2004) Cross-dressing the virion: the transcapsidation of adeno-associated virus serotypes functionally defines subgroups. J. Virol. 78, 4421–4432 12 Shi, W. and Bartlett, J.S. (2003) RGD inclusion in VP3 provides adenoassociated virus type 2 (AAV2)-based vectors with a heparan sulfateindependent cell entry mechanism. Mol. Ther. 7, 515–525 13 Rabinowitz, J.E. and Samulski, R.J. (2000) Building a better vector: the manipulation of AAV virions. Virology 278, 301–308 14 Grandi, P. et al. (2004) HSV-1 virions engineered for specific binding to cell surface receptors. Mol. Ther. 9, 419–427 15 Watson, D.J. et al. (2002) Targeted transduction patterns in the mouse brain by lentivirus vectors pseudotyped with VSV, Ebola, Mokola, LCMV, or MuLV envelope proteins. Mol. Ther. 5, 528–537 16 Bartlett, J.S. et al. (1999) Targeted adeno-associated virus vector transduction of nonpermissive cells mediated by a bispecific Fabg2 antibody. Nat. Biotechnol. 17, 181–186 17 Samulski, R.J. (2000) Expanding the AAV package. Nat. Biotechnol. 18, 497–498 18 Kafri, T. (2001) Lentivirus vectors: difficulties and hopes before clinical trials. Curr. Opin. Mol. Ther. 3, 316–326
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
19 Wade-Martins, R. et al. (2003) Infectious delivery of a 135-kb LDLR genomic locus leads to regulated complementation of low-density lipoprotein receptor deficiency in human cells. Mol. Ther. 7, 604–612 20 Xing, W. et al. (2004) HSV-1 amplicon-mediated transfer of 128-kb BMP-2 genomic locus stimulates osteoblast differentiation in vitro. Biochem. Biophys. Res. Commun. 319, 781–786 21 Nordquist, D.T. et al. (1988) cDNA cloning and characterization of three genes uniquely expressed in cerebellum by Purkinje neurons. J. Neurosci. 8, 4780–4789 22 Oberdick, J. et al. (1990) A promoter that drives transgene expression in cerebellar Purkinje and retinal bipolar neurons. Science 248, 223–226 23 Berrebi, A.S. et al. (1991) Cerebellar Purkinje cell markers are expressed in retinal bipolar neurons. J. Comp. Neurol. 308, 630–649 24 Barski, J.J. et al. (2000) Cre recombinase expression in cerebellar Purkinje cells. Genesis 28, 93–98 25 Heintz, N. (2001) BAC to the future: the use of BAC transgenic mice for neuroscience research. Nat. Rev. Neurosci. 2, 861–870 26 Zhang, X.M. et al. (2004) Highly restricted expression of Cre recombinase in cerebellar Purkinje cells. Genesis 40, 45–51 27 Wasserman, W.W. and Sandelin, A. (2004) Applied bioinformatics for the identification of regulatory elements. Nat. Rev. Genet. 5, 276–287 28 Lenhard, B. et al. (2003) Identification of conserved regulatory elements by comparative genome analysis. J. Biol. 2, 13 29 Wong, L.F. et al. (2004) Transduction patterns of pseudotyped lentiviral vectors in the nervous system. Mol. Ther. 9, 101–111 30 Mazarakis, N.D. et al. (2001) Rabies virus glycoprotein pseudotyping of lentiviral vectors enables retrograde axonal transport and access to the nervous system after peripheral delivery. Hum. Mol. Genet. 10, 2109–2121 31 Braz, J.M. et al. (2002) Transneuronal tracing of diverse CNS circuits by Cre-mediated induction of wheat germ agglutinin in transgenic mice. Proc. Natl. Acad. Sci. U. S. A. 99, 15148–15153 32 Kinoshita, N. et al. (2002) Adenovirus-mediated WGA gene delivery for transsynaptic labeling of mouse olfactory pathways. Chem. Senses 27, 215–223 33 Yoshihara, Y. et al. (1999) A genetic approach to visualization of multisynaptic neural pathways using plant lectin transgene. Neuron 22, 33–41 34 Kissa, K. et al. (2002) In vivo neuronal tracing with GFP-TTC gene delivery. Mol. Cell. Neurosci. 20, 627–637 35 Maskos, U. et al. (2002) Retrograde trans-synaptic transfer of green fluorescent protein allows the genetic mapping of neuronal circuits in transgenic mice. Proc. Natl. Acad. Sci. U. S. A. 99, 10120–10125 36 Enquist, L.W. (2002) Exploiting circuit-specific spread of pseudorabies virus in the central nervous system: insights to pathogenesis and circuit tracers. J. Infect. Dis. 186 (Suppl. 2), S209–S214 37 Ugolini, G. (1995) Specificity of rabies virus as a transneuronal tracer of motor networks: transfer from hypoglossal motoneurons to connected second-order and higher order central nervous system cell groups. J. Comp. Neurol. 356, 457–480 38 Sun, N. et al. (1996) Anterograde, transneuronal transport of herpes simplex virus type 1 strain H129 in the murine visual system. J. Virol. 70, 5405–5413
www.sciencedirect.com
201
39 Kelly, R.M. and Strick, P.L. (2000) Rabies as a transneuronal tracer of circuits in the central nervous system. J. Neurosci. Methods 103, 63–71 40 DeFalco, J. et al. (2001) Virus-assisted mapping of neural inputs to a feeding center in the hypothalamus. Science 291, 2608–2613 41 Oyama, M. et al. (2000) Application of conditionally replicating herpes vector for gene therapy treatment of urologic neoplasms. Mol. Urol. 4, 83–87 42 Guerrero, G. and Isacoff, E.Y. (2001) Genetically encoded optical sensors of neuronal activity and cellular function. Curr. Opin. Neurobiol. 11, 601–607 43 Miesenbock, G. (2004) Genetic methods for illuminating the function of neural circuits. Curr. Opin. Neurobiol. 14, 395–402 44 Tsien, R.Y. (2003) Imagining imaging’s future. Nat. Rev. Mol. Cell Biol. (Suppl), SS16–SS21 45 Bozza, T. et al. (2004) In vivo imaging of neuronal activity by targeted expression of a genetically encoded probe in the mouse. Neuron 42, 9–21 46 Miesenbock, G. et al. (1998) Visualizing secretion and synaptic transmission with pH-sensitive green fluorescent proteins. Nature 394, 192–195 47 Kobayashi, K. et al. (1995) Immunotoxin-mediated conditional disruption of specific neurons in transgenic mice. Proc. Natl. Acad. Sci. U. S. A. 92, 1132–1136 48 Yoshida, K. et al. (2001) A key role of starburst amacrine cells in originating retinal directional selectivity and optokinetic eye movement. Neuron 30, 771–780 49 Watanabe, D. et al. (1998) Ablation of cerebellar Golgi cells disrupts synaptic integration involving GABA inhibition and NMDA receptor activation in motor coordination. Cell 95, 17–27 50 Johns, D.C. et al. (1999) Inducible genetic suppression of neuronal excitability. J. Neurosci. 19, 1691–1697 51 Nadeau, H. et al. (2000) ROMK1 (Kir1.1) causes apoptosis and chronic silencing of hippocampal neurons. J. Neurophysiol. 84, 1062–1075 52 Ibanez-Tallon, I. et al. (2004) Tethering naturally occurring peptide toxins for cell-autonomous modulation of ion channels and receptors in vivo. Neuron 43, 305–311 53 Redfern, C.H. et al. (1999) Conditional expression and signaling of a specifically designed Gi-coupled receptor in transgenic mice. Nat. Biotechnol. 17, 165–169 54 Coward, P. et al. (1998) Controlling signaling with a specifically designed Gi-coupled receptor. Proc. Natl. Acad. Sci. U. S. A. 95, 352–357 55 Birgul, N. et al. (1999) Reverse physiology in Drosophila: identification of a novel allatostatin-like neuropeptide and its cognate receptor structurally related to the mammalian somatostatin/galanin/opioid receptor family. EMBO J. 18, 5892–5900 56 Lechner, H.A. et al. (2002) A genetic method for selective and quickly reversible silencing of mammalian neurons. J. Neurosci. 22, 5287–5290 57 Slimko, E.M. et al. (2002) Selective electrical silencing of mammalian neurons in vitro by the use of invertebrate ligand-gated chloride channels. J. Neurosci. 22, 7373–7379
TRENDS in Neurosciences Vol.28 No.4 April 2005
A molecular and genetic arsenal for systems neuroscience Edward M. Callaway Systems Neurobiology Laboratories, The Salk Institute for Biological Studies, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA
Neural circuits are composed of a meshwork of numerous neuron types, each with their own distinctive morphological and intrinsic physiological properties, connectivity and biochemistry. How do distinct neural subcircuits, composed of different classes of neuron, contribute to brain function? Approaching this question requires methods that can target specific neurons types. This can be achieved by harnessing the same machinery that builds sophistication into the brain and using it to make novel tools for investigating and manipulating the brain: molecular and genetic technology. These tools can be used to target gene expression to specific neuron types within complicated neuronal circuits, and the transgenes that are expressed can be used to elucidate and manipulate these circuits with unprecedented precision and control. These methods are likely to become the archetype for future studies linking perception, cognition and behavior to specific components of the brain. Introduction Systems neuroscience aims to understand how complex interactions between networks of neurons give rise to perception and behavior. This is a daunting task. The primary difficulty is that neural circuits are composed of a complex network of numerous neuron types whose dendrites and axons are intricately intertwined. This creates unique challenges for separately identifying the contributions of each cell type to the intact functioning network. Traditional methods have contributed a great deal to our understanding of brain function, but they fail in their ability to approach questions at this level of specificity. This review will focus on methods being developed that use our understanding of cell and molecular biology to attack questions in systems neuroscience. The primary focus will be on methods that will enable the organization and function of complex neural circuits to be investigated at the level of specific cell types. There has been considerable progress in the development and use of molecular and genetic methods for studying the nervous system in mice. The focus of this review is on how these and related methods can be used in higher mammals, particularly monkeys, in which transgenic technology is not practical owing to the long gestation period (w165 days) and time to reach sexual maturity (several years). The ability to take advantage of Corresponding author: Callaway, E.M. ([email protected]).
molecular and genetic methods in primates should enable studies of more complex behaviors and brain functions than are found or easily studied in mice. Many practical considerations influence decisions about what methods are best to achieve any given experimental goal, and many of the genetic methods that have been developed for studies of the mouse nervous system have been based on decisions that apply more to mice than to monkeys. But, despite these differences, there is also a good deal of common ground. There are two primary advantages to genetic approaches, regardless of species. These are the ability to target specific cell types within complex tissues and the ability to perform complex manipulations with great repeatability. These two goals are interrelated. With transgenic mice, transgene expression is often targeted to specific cell types in the brain and expression is consistent across animals in the same genetic line. Thus, large numbers of precisely manipulated animals can be generated and studied. By contrast, the use of large numbers of primates is neither practical nor desirable. Primates are a precious resource and considerable investment is required for their breeding, care and training. Primate-based studies are therefore designed to minimize the numbers of animals used and it is typical to train a small number of animals on a specific task and then study those animals extensively – often for years. Thus, genetic methods for studying primates must address this reality by developing and using manipulations that are stable over long periods of time. This strategy allows for repeatability in primate models, by enabling long-term investigation of a single animal in which the particular genetic manipulation has been very carefully characterized. Because the use of transgenic primates is impractical, the most useful methods presently available for delivery of genetic material in these animals are based primarily on viral vectors [1,2]. Many of these vectors were developed for gene therapy in humans and therefore incorporate features that are ideal for long-term studies in primates; most notably, they can efficiently transduce nearly every neuron in the vicinity of a brain injection and can yield stable gene expression for months or years without toxicity [3–7]. The main part of this review will therefore begin with a brief overview of properties of some of the most promising viral vectors and considerations involved in the choice of vector.
www.sciencedirect.com 0166-2236/$ - see front matter Q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.tins.2005.01.007
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
Cell-type-specific targeting is one of the chief advantages of genetic methods. The need for cell-type specificity has become apparent from our increasingly sophisticated understanding of the nervous system. For example, it is now clear that when a brain structure contains multiple cell types with interwoven dendritic arbors, afferent input to that structure can selectively connect to some cell types while avoiding others [8]. And it should be self-evident that when there are multiple cell types, each type is likely to have a unique role in the neural network. Conventional methods have allowed tracing of connections and monitoring or manipulation of neuronal activity, but these methods typically do not have the resolution required to address questions at the level of specific cell types. Methods by which gene expression can be targeted to particular cell types of interest will therefore also be discussed in this review. Finally, once gene expression is achieved in the targeted neuronal population, it is useful only if the gene that is expressed provides insight into the organization or function of the nervous system. Thus, the remainder of the review will summarize recent progress in the development and use of genetic methods for achieving three goals: elucidation of neural circuits, monitoring of neural activity, and manipulation of activity. Traditional methods have achieved these, but not at the level of cell-type specificity or with the repeatability that can be achieved using genetics. Viral vectors for gene delivery The ideal vector for transgene delivery depends on the experimental goals. It should be clear that a first consideration is that the chosen vector must be able to deliver genetic material efficiently to the cell type of interest (i.e. the cells must be susceptible to the vector). Furthermore, once the genetic material has entered the cell, the transgene must be expressed (i.e. the cell must be permissive for expression). For the typical systems neuroscientist, it might be surprising to learn that there is not one best, universal vector that can deliver DNA (or RNA) to all cell types. It is beyond the scope of this review to discuss the known tropisms of each viral vector, so the interested reader is referred to selected references describing adenoassociated viruses (AAV) [9–12], lentiviral vectors [6,7] and herpes simplex virus (HSV) amplicon vectors [5]. These are all replication-incompetent vectors that can deliver DNA (or RNA) efficiently to non-dividing adult neurons, potentially yielding transgene expression in nearly every neuron within w0.5–1.0 mm of the injection site (depending on viral titers and volumes). Because the vectors are replication-incompetent, there is no further spread and they enable stable transgene expression without apparent toxicity. A first step in any course of study would be to test empirically which viruses can transduce the cell type of interest. Viral vectors have many advantages but they also have limitations, and the tropism of viruses (not all cell types are both susceptible and permissive for a given vector) could be considered as both. When a particular virus does not infect or generate transgene expression in the cell type of interest, it is a limitation. But tropism can also be used www.sciencedirect.com
197
to target particular cell types of interest [9–11,13–15]. Designer viruses have also been manufactured specifically to create tropism for a particular cell type or to expand the range of cell types that can be transduced [12–14,16]. The most prominent limitation of viral vectors is their capacity. This is most relevant for cell-type-specific promoters, which are often too large to fit into viruses (as will be discussed later). AAV and lentivirus can package genetic sequences up to w5 Kb and w8 Kb, respectively [17,18]. Wild-type HSV has a much larger genome of 152 Kb, and HSV amplicon vectors can therefore, in theory, package amplicons up to this size. In practice, viral titers can be reduced when amplicons Ow10 Kb are packaged. Nevertheless, there have been recent successes in making high-titer HSV amplicons incorporating bacterial artificial chromosomes (BACs) that are 128–135 Kb [19,20]. Continued refinements are therefore likely to increase greatly the utility of HSV amplicon vectors, particularly for cell-type-specific gene expression. Targeting specific cell types Cell-type-specific targeting is one of the chief advantages of genetic methods, and the ways in which it can be achieved have practical considerations for development of the tools that will be discussed here. The most common way to achieve cell-type-specific gene expression is by utilizing a cell-type-specific promoter. For example, the gene encoding Purkinje cell protein 2 (Pcp2/L7) is expressed exclusively in cerebellar Purkinje cells and retinal bipolar cells [21–23]. The 2.9 Kb DNA sequence upstream from this gene has been used successfully to drive transgene expression preferentially in these cell types [24] (although small amounts of aberrant expression were also observed). There are innumerable other examples of success in generating mouse lines using cell-type-specific gene expression, but there are just as many (or probably many more) attempts that are unsuccessful. As a general rule, however, the probability of success and degree of specificity are greatly increased when larger stretches of potential regulatory sequences are used. BAC transgenics, which can introduce O200 Kb of DNA into the germ lines of transgenic mice, have been most successful [25]. For example, improved specificity of transgene expression in Pcp2-expressing cells was achieved using a 173 Kb BAC [26]. The difficulty arises primarily because the DNA sequences and mechanisms that regulate gene expression are not fully understood. To target gene expression to specific cell types using cell-type-specific promoters in viral vectors, it is of course first necessary to identify a vector that can transduce the cell type of interest. But because other cell types in the vicinity will also be susceptible and permissive to the vector, the cell-type-specific promoter is necessary to restrict transcription of RNA from DNA to the cell type of interest (Figure 1). It should therefore be clear that among the most important differences between work with transgenic mice and gene delivery by viruses is the limited capacity of viruses. Because there is no definitive method for identifying the minimal DNA sequence necessary for efficient and selective gene expression in a particular cell type, it is often very difficult to find short sequences that
Review
198
TRENDS in Neurosciences Vol.28 No.4 April 2005
(a)
(c) Cell-type-specific promoter
Molecular switch
(b) (i)
(ii)
• K+ channels open when AL applied → excitability decreases • Normal excitability when AL not present
(iv)
(iii)
TRENDS in Neurosciences
Figure 1. Example of experimental and conceptual steps involved in targeting a specific neuron type for reversible inactivation using a viral vector. (a) A viral vector is constructed to carry the genetic coding sequence for a transgene, in this case the gene encoding a molecular switch. This coding sequence is downstream of a genetic regulatory sequence, or promoter, that will permit and/or drive expression of the transgene only in a specific cell type. (b) The steps involved in generating specific expression of the genetic switch and reversible neuronal inactivation begin with the injection of virus into the brain region of interest. (i) Billions of viral particles (e.g. 1 ml of particles at 1012 particles mlK1) are injected into the brain region of interest, which here contains three neuron types (purple, blue and green), including the cell type that is being targeted (green). (ii) The viral vector efficiently delivers its genetic material to nearly every neuron within w0.5–1.0 mm of the injection site, including all three neuron types (neurons now shown in brown). Larger regions can be influenced by viral injection at more locations. (iii) Expression of the transgene (genetic switch) is restricted to the targeted cell type (now shown in yellow) because only this cell type drives the promoter that was included in the viral construct. (iv) When the switch is flipped (c), the targeted cell type is selectively inactivated. Because this is a reversible inactivation method, the neurons can return to the active state. (c) One molecular switch that has been developed is the Drosophila allatostatin receptor (AlstR). When AlstR is expressed in mammalian neurons, it initially has no effect on their activity, presumably because the mammalian brain lacks an endogenous ligand for the insect receptor. However, when the Drosophila neuropeptide allatostatin (AL) is injected in the vicinity of the neurons expressing the AlstR, the AL flips the genetic switch (inset). AL binds to AlstR, triggering a conformational change and G-protein-mediated opening of inwardly rectifying KC (GIRK) channels. The increase in KC conductance causes neuronal inactivation. The effect of AL is specific for the AlstR-expressing cells and therefore confers selective neuronal inactivation. When AL is removed by diffusion or brain peptidases, normal neuronal function returns.
work. Trial and error is most common and often simply boils down to trying the largest sequence upstream from the coding sequence that can still fit in the vector. This typically leads to exercises in success or failure rather than trial and error. Hopefully, future studies will lead to better methods for predicting the relevant regulatory sequences. For example, crucial regulatory elements might be identified by comparing sequences across species to identify conserved motifs, or by identifying common sequences across multiple genes that are all expressed selectively in a specific cell type [27,28]. Cell-type-specific promoters are not the only way to target gene expression to specific neuron types within complex neuropil. Neuron types can differ not only in the genes they express but also in the distant structures to which they project their efferent axons. In recent experiments, a lentivirus has been pseudotyped with the envelope protein from rabies virus, enabling it to infect neurons efficiently through their axon terminals [29,30]. HSV amplicons can also infect neurons via their axon terminals [5]. Thus, specific types of projection neurons can be infected by injection of an appropriate viral vector into distant axonal arbors. Genetic methods for tracing connections of specific cell types The most advanced methods for tracing neuronal connections, in terms of cell-type specificity and ease of use, include those based on expression of gene products that www.sciencedirect.com
are transported either anterogradely or retrogradely and cross synapses. These include wheat germ agglutinin (WGA), which is transported primarily anterogradely [31–33], and tetanus toxin C fragment (TTC), which is transported retrogradely [34,35]. Both of these methods have been used successfully to trace connections from specific cell types in transgenic mice, either by placing tracer gene expression under the control of a specific promoter or by making expression dependent on Cre recombination that is restricted to cell types of interest. This approach has also been successfully implemented using replication-incompetent adenovirus delivery of genetic constructs that carry coding sequence for the tracer [32]. Transported tracer can be detected either using antibodies specific for the tracer or using expression of green fluorescent protein (GFP)–tracer fusion proteins. It should also be possible to use these methods in primates, with gene delivery mediated by viral vectors. Despite the usefulness of these methods, they have two major, related limitations. Weak connections can be difficult or impossible to detect, and label resulting from direct or monosynaptic connections cannot usually be distinguished from label arising from multisynaptic pathways. In particular, strong multisynaptic connections could result in detection of more transynaptic labeling than weak monosynaptic connections. An alternative to genetically expressed transynaptic tracers is replication-competent neurotropic viruses, which progress transynaptically through the nervous
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
system as part of their natural life cycle [36]. Three neurotropic viruses are used most commonly for transynaptic labeling studies – pseudorabies virus (PRV; despite the name, this is an a-herpes virus), HSV and rabies virus (RV). These viruses have been used extensively as transynaptic tracers and, depending on the type of virus and strain, they can be transported selectively in either the anterograde or retrograde direction [21,36–39]. Used in their replication-competent form, they lack celltype specificity but have the advantage of amplifying signals by replicating at each synaptic stage. Thus, weak connections are more likely to be detected than by using WGA or TTC. But it is still not always possible to distinguish weak, direct connections from strong disynaptic or polysynaptic connections [37]. De Falco et al. [40] greatly increased the utility of PRV by engineering the virus for cell-type specificity. They deleted from the viral genome the thymidine kinase (TK) gene, which is required for viral replication in neurons. They inserted into the viral genome a conditionally expressed TK gene and also the GFP coding sequence. In cells that express Cre recombinase, the viral genome is recombined so that TK and GFP are both expressed. In this way, the PRV replicates and expresses GFP only in parent cells that are targeted for Cre expression. This newly competent, recombined and GFP-expressing PRV can then spread retrogradely to connected cells, where it can further replicate and continue to spread. GFP is therefore expressed only in the multisynaptic chain of neurons connected to the genetically targeted parent cells. Although this method again suffers from difficulty in distinguishing order of connectivity, it might prove possible to restrict viral spread to monosynaptic connections by selectively expressing TK along with Cre in parent cells, and completely eliminating TK from the viral genome. One of the limitations (but also a safety feature) of PRV is that it does not infect primates [36], but it is possible to make similar genetic modifications of HSV [41], which does infect primates. Genetically expressed sensors of activity Among the most exciting potential applications of genetic technology to systems neuroscience is measurement of activity in specific neuron types using genetically encoded sensors [42–44]. Such sensors have been engineered for detection of membrane voltage in addition to Ca2C and other intracellular messengers, and they can also be used to measure vesicular release of neurotransmitter. Each of these signals can be directly or indirectly correlated with neuronal activity. At present, most physiological studies use metal electrodes to record from single neurons in complex structures, without knowledge of the cell types being studied. Genetic methods should allow optical recordings of activity from neurons of known type, perhaps even in awake, behaving primates. However, despite success in the development of probes that can yield signals detectable under in vitro conditions or in invertebrates, success with in vivo recordings in mammals has been extremely limited [45]. The most notable exception is a study in which synaptopHluorin [46] was expressed in mouse olfactory sensory www.sciencedirect.com
199
neurons [45]. Synapto-pHluorin is a genetically encoded pH sensor that indirectly reports activity at synaptic terminals when synaptic vesicles fuse and change their pH during transmitter release [46]. When expressed in transgenic mice under the control of an olfactory-neuronspecific promoter, synapto-pHluorin was detected optically in the terminal arbors in the olfactory bulb. Large fluorescence changes in the olfactory bulb could then be detected following olfactory stimulation. Although these studies represent a tremendous step forward, there are still considerable difficulties to be overcome. An ongoing challenge for these methods is the small size of the signals that are typically generated. Thus, with present technology, in vivo measurements are likely to require averaging over both time and neuronal populations. Future improvements are likely to enable recordings from identified single neurons to be made in vivo using two-photon microscopy. Ideally, temporal resolution will be sufficient to identify individual action potentials. Optical scanning across populations of neurons expressing genetic sensors might also enable measurement of the dynamic activities of identified neuronal assemblies [43]. Genetic methods for selective neuronal inactivation As the spatial resolution of methods for elucidating connectivity and measuring activity of neurons has improved, so too have methods for perturbing neural activity. To test the role of specific cell types within complex circuits it is necessary to target specific cell types for inactivation. Because genetic methods allow cell-type-specific targeting, inactivation methods based on expression of transgenes can be used for cell-type-specific inactivation. Although several in vitro studies describe potential genetic methods for neuronal inactivation (as will be discussed later in this article), the progression to in vivo studies is slow. Therefore, the most successful in vivo studies have used older methods that lack reversibility because the cells are killed. In 1995, Kobayashi et al. [47] described an immunotoxin-based method for selective neuronal inactivation; following selective expression of the interleukin-2 receptor a-subunit, those cells are killed by exposure to immunotoxin. Yoshida et al. [48] used this method, along with a promoter for the metabotropic glutamate receptor mGluR2, to kill selectivity retinal starburst amacrine cells. They were then able to infer the role of this cell type in generating the direction-selectivity of retinal ganglion cells. This method has also been used to kill cerebellar Golgi cells selectively [49]. Despite success with selective cell killing, reversible methods will be crucial for future studies. Such methods might avoid the possibility that functions measured after cell killing reflect compensatory mechanisms, and they are essential for experiments that require an intact system both before and after manipulations. As will be discussed, in vitro studies have been used to characterize reversible methods that are likely to work in vivo in the near future. Several genetic methods for neuronal inactivation can be made reversible on a slow timescale by regulating transcription of the transgene. These methods include overexpression of KC channels [50,51] and expression of
200
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
tethered toxins that can block ion channel function [52]. Such methods could be ideal for studies requiring longterm, reversible inactivation, such as investigations of plasticity or the roles of particular neuron types in learning paradigms that might develop over several days. But these methods are too slow for some other applications. For example, it would be useful to measure the visual receptive field of a neuron in the monkey visual cortex, then inactivate a particular neuron type, reevaluate the receptive field and recover normal function, all within the 1–2 h period in which activity of a single neuron can be recorded. Methods for quickly reversible inactivation are generally based on the expression of a gene that, by itself, does not affect function (Figure 1c). But when the gene product is activated by an exogenous compound, there is selective inactivation that reverses when the compound is washed away. A prototype of this class of genetic method was developed by Conklin and colleagues [53,54]. They modified a k opiate receptor gene by deleting the sequence encoding the part of the receptor that normally binds to natural opiates, but leaving intact a site activated by synthetic opiates. Conklin and colleagues called the engineered receptor a receptor activated solely by synthetic ligand (RASSL). When this RASSL was activated by synthetic opiates in vitro, it caused G-protein-mediated opening of KC channels [54], and when expressed in heart muscle in vivo, synthetic opiates reduced heart rate [53]. Although this system also causes KC channel opening in neurons in vitro (E.M. Callaway, unpublished) and could therefore potentially be used for neuronal inactivation in vivo, it might not be ideal for neuronal inactivation: synthetic opiates activate naturally occurring opiate receptors in the mammalian brain and thus would lack specificity for the cells genetically targeted for RASSL expression. This approach inspired the search for other receptors that can mediate increases in KC conductance via G proteins but that are not activated by ligands for which the mammalian brain has endogenous receptors. The ideal receptor would be activated by a ligand that does not activate any receptors endogenous to the mammalian brain. The insect allatostatin receptor (AlstR) appears to meet these criteria [55] (Figure 1c). When tested in mammalian neurons expressing this receptor, the peptide ligand allatostatin quickly and reversibly increased KC conductance, effectively silencing the neurons [56]. Control neurons were unaffected by allatostatin. A similar strategy for selective and reversible inactivation uses genetic expression of an invertebrate glutamategated ClK channel that has been mutated to make it insensitive to glutamate [57]. However, the channel remains sensitive to the medication Ivermectin, which can silence cultured neurons expressing the receptor. Reversibility occurs slowly, over hours, apparently due to tight binding of the drug to the receptor. Concluding remarks In summary, it is clear that molecular and genetic methods will have an increasingly important role and eventually will have an enormous impact in systems neuroscience. www.sciencedirect.com
However, many of the methods are still in their infancy. For example, there are several vectors for gene delivery that can efficiently deliver genetic material to neurons, but they are limited in their capacity. There are also many short promoter sequences that enable cell-type-specific gene expression, but more systematic and reliable approaches are needed for identifying short promoters to target gene expression to particular cell types. Cell-typespecific gene expression can be reliably used to kill neurons, but reversible inactivation in vivo, particularly on a fast timescale, is still in development. Genetically encoded transynaptic tracers are providing excellent new insights, but they could be improved for some applications by increasing efficiency and restricting spread to monosynaptic connections. And although genetically expressed sensors of neural activity work, they provide small and slow signals; such sensors are likely to be greatly improved in speed and sensitivity by further engineering. The road will be hard, but the future is nearly unlimited and the pioneers who choose this road will be rewarded.
References 1 Kootstra, N.A. and Verma, I.M. (2003) Gene therapy with viral vectors. Annu. Rev. Pharmacol. Toxicol. 43, 413–439 2 Davidson, B.L. and Breakefield, X.O. (2003) Viral vectors for gene delivery to the nervous system. Nat. Rev. Neurosci. 4, 353–364 3 Xiao, X. et al. (1997) Gene transfer by adeno-associated virus vectors into the central nervous system. Exp. Neurol. 144, 113–124 4 Rabinowitz, J.E. and Samulski, J. (1998) Adeno-associated virus expression systems for gene transfer. Curr. Opin. Biotechnol. 9, 470–475 5 Sandler, V.M. et al. (2002) Modified herpes simplex virus delivery of enhanced GFP into the central nervous system. J. Neurosci. Methods 121, 211–219 6 Naldini, L. et al. (1996) Efficient transfer, integration, and sustained long-term expression of the transgene in adult rat brains injected with a lentiviral vector. Proc. Natl. Acad. Sci. U. S. A. 93, 11382–11388 7 Blomer, U. et al. (1997) Highly efficient and sustained gene transfer in adult neurons with a lentivirus vector. J. Virol. 71, 6641–6649 8 Callaway, E.M. (2002) Cell type specificity of local cortical connections. J. Neurocytol. 31, 231–237 9 Davidson, B.L. et al. (2000) From the cover: recombinant adenoassociated virus type 2, 4, and 5 vectors: transduction of variant cell types and regions in the mammalian central nervous system. Proc. Natl. Acad. Sci. U. S. A. 97, 3428–3432 10 Rabinowitz, J.E. et al. (2002) Cross-packaging of a single adenoassociated virus (AAV) type 2 vector genome into multiple AAV serotypes enables transduction with broad specificity. J. Virol. 76, 791–801 11 Rabinowitz, J.E. et al. (2004) Cross-dressing the virion: the transcapsidation of adeno-associated virus serotypes functionally defines subgroups. J. Virol. 78, 4421–4432 12 Shi, W. and Bartlett, J.S. (2003) RGD inclusion in VP3 provides adenoassociated virus type 2 (AAV2)-based vectors with a heparan sulfateindependent cell entry mechanism. Mol. Ther. 7, 515–525 13 Rabinowitz, J.E. and Samulski, R.J. (2000) Building a better vector: the manipulation of AAV virions. Virology 278, 301–308 14 Grandi, P. et al. (2004) HSV-1 virions engineered for specific binding to cell surface receptors. Mol. Ther. 9, 419–427 15 Watson, D.J. et al. (2002) Targeted transduction patterns in the mouse brain by lentivirus vectors pseudotyped with VSV, Ebola, Mokola, LCMV, or MuLV envelope proteins. Mol. Ther. 5, 528–537 16 Bartlett, J.S. et al. (1999) Targeted adeno-associated virus vector transduction of nonpermissive cells mediated by a bispecific Fabg2 antibody. Nat. Biotechnol. 17, 181–186 17 Samulski, R.J. (2000) Expanding the AAV package. Nat. Biotechnol. 18, 497–498 18 Kafri, T. (2001) Lentivirus vectors: difficulties and hopes before clinical trials. Curr. Opin. Mol. Ther. 3, 316–326
Review
TRENDS in Neurosciences Vol.28 No.4 April 2005
19 Wade-Martins, R. et al. (2003) Infectious delivery of a 135-kb LDLR genomic locus leads to regulated complementation of low-density lipoprotein receptor deficiency in human cells. Mol. Ther. 7, 604–612 20 Xing, W. et al. (2004) HSV-1 amplicon-mediated transfer of 128-kb BMP-2 genomic locus stimulates osteoblast differentiation in vitro. Biochem. Biophys. Res. Commun. 319, 781–786 21 Nordquist, D.T. et al. (1988) cDNA cloning and characterization of three genes uniquely expressed in cerebellum by Purkinje neurons. J. Neurosci. 8, 4780–4789 22 Oberdick, J. et al. (1990) A promoter that drives transgene expression in cerebellar Purkinje and retinal bipolar neurons. Science 248, 223–226 23 Berrebi, A.S. et al. (1991) Cerebellar Purkinje cell markers are expressed in retinal bipolar neurons. J. Comp. Neurol. 308, 630–649 24 Barski, J.J. et al. (2000) Cre recombinase expression in cerebellar Purkinje cells. Genesis 28, 93–98 25 Heintz, N. (2001) BAC to the future: the use of BAC transgenic mice for neuroscience research. Nat. Rev. Neurosci. 2, 861–870 26 Zhang, X.M. et al. (2004) Highly restricted expression of Cre recombinase in cerebellar Purkinje cells. Genesis 40, 45–51 27 Wasserman, W.W. and Sandelin, A. (2004) Applied bioinformatics for the identification of regulatory elements. Nat. Rev. Genet. 5, 276–287 28 Lenhard, B. et al. (2003) Identification of conserved regulatory elements by comparative genome analysis. J. Biol. 2, 13 29 Wong, L.F. et al. (2004) Transduction patterns of pseudotyped lentiviral vectors in the nervous system. Mol. Ther. 9, 101–111 30 Mazarakis, N.D. et al. (2001) Rabies virus glycoprotein pseudotyping of lentiviral vectors enables retrograde axonal transport and access to the nervous system after peripheral delivery. Hum. Mol. Genet. 10, 2109–2121 31 Braz, J.M. et al. (2002) Transneuronal tracing of diverse CNS circuits by Cre-mediated induction of wheat germ agglutinin in transgenic mice. Proc. Natl. Acad. Sci. U. S. A. 99, 15148–15153 32 Kinoshita, N. et al. (2002) Adenovirus-mediated WGA gene delivery for transsynaptic labeling of mouse olfactory pathways. Chem. Senses 27, 215–223 33 Yoshihara, Y. et al. (1999) A genetic approach to visualization of multisynaptic neural pathways using plant lectin transgene. Neuron 22, 33–41 34 Kissa, K. et al. (2002) In vivo neuronal tracing with GFP-TTC gene delivery. Mol. Cell. Neurosci. 20, 627–637 35 Maskos, U. et al. (2002) Retrograde trans-synaptic transfer of green fluorescent protein allows the genetic mapping of neuronal circuits in transgenic mice. Proc. Natl. Acad. Sci. U. S. A. 99, 10120–10125 36 Enquist, L.W. (2002) Exploiting circuit-specific spread of pseudorabies virus in the central nervous system: insights to pathogenesis and circuit tracers. J. Infect. Dis. 186 (Suppl. 2), S209–S214 37 Ugolini, G. (1995) Specificity of rabies virus as a transneuronal tracer of motor networks: transfer from hypoglossal motoneurons to connected second-order and higher order central nervous system cell groups. J. Comp. Neurol. 356, 457–480 38 Sun, N. et al. (1996) Anterograde, transneuronal transport of herpes simplex virus type 1 strain H129 in the murine visual system. J. Virol. 70, 5405–5413
www.sciencedirect.com
201
39 Kelly, R.M. and Strick, P.L. (2000) Rabies as a transneuronal tracer of circuits in the central nervous system. J. Neurosci. Methods 103, 63–71 40 DeFalco, J. et al. (2001) Virus-assisted mapping of neural inputs to a feeding center in the hypothalamus. Science 291, 2608–2613 41 Oyama, M. et al. (2000) Application of conditionally replicating herpes vector for gene therapy treatment of urologic neoplasms. Mol. Urol. 4, 83–87 42 Guerrero, G. and Isacoff, E.Y. (2001) Genetically encoded optical sensors of neuronal activity and cellular function. Curr. Opin. Neurobiol. 11, 601–607 43 Miesenbock, G. (2004) Genetic methods for illuminating the function of neural circuits. Curr. Opin. Neurobiol. 14, 395–402 44 Tsien, R.Y. (2003) Imagining imaging’s future. Nat. Rev. Mol. Cell Biol. (Suppl), SS16–SS21 45 Bozza, T. et al. (2004) In vivo imaging of neuronal activity by targeted expression of a genetically encoded probe in the mouse. Neuron 42, 9–21 46 Miesenbock, G. et al. (1998) Visualizing secretion and synaptic transmission with pH-sensitive green fluorescent proteins. Nature 394, 192–195 47 Kobayashi, K. et al. (1995) Immunotoxin-mediated conditional disruption of specific neurons in transgenic mice. Proc. Natl. Acad. Sci. U. S. A. 92, 1132–1136 48 Yoshida, K. et al. (2001) A key role of starburst amacrine cells in originating retinal directional selectivity and optokinetic eye movement. Neuron 30, 771–780 49 Watanabe, D. et al. (1998) Ablation of cerebellar Golgi cells disrupts synaptic integration involving GABA inhibition and NMDA receptor activation in motor coordination. Cell 95, 17–27 50 Johns, D.C. et al. (1999) Inducible genetic suppression of neuronal excitability. J. Neurosci. 19, 1691–1697 51 Nadeau, H. et al. (2000) ROMK1 (Kir1.1) causes apoptosis and chronic silencing of hippocampal neurons. J. Neurophysiol. 84, 1062–1075 52 Ibanez-Tallon, I. et al. (2004) Tethering naturally occurring peptide toxins for cell-autonomous modulation of ion channels and receptors in vivo. Neuron 43, 305–311 53 Redfern, C.H. et al. (1999) Conditional expression and signaling of a specifically designed Gi-coupled receptor in transgenic mice. Nat. Biotechnol. 17, 165–169 54 Coward, P. et al. (1998) Controlling signaling with a specifically designed Gi-coupled receptor. Proc. Natl. Acad. Sci. U. S. A. 95, 352–357 55 Birgul, N. et al. (1999) Reverse physiology in Drosophila: identification of a novel allatostatin-like neuropeptide and its cognate receptor structurally related to the mammalian somatostatin/galanin/opioid receptor family. EMBO J. 18, 5892–5900 56 Lechner, H.A. et al. (2002) A genetic method for selective and quickly reversible silencing of mammalian neurons. J. Neurosci. 22, 5287–5290 57 Slimko, E.M. et al. (2002) Selective electrical silencing of mammalian neurons in vitro by the use of invertebrate ligand-gated chloride channels. J. Neurosci. 22, 7373–7379