The nervous system is an extremely complex and intricately specific network of neuronal circuitries, in which billions of cells work together as an exquisitely organized, interconnected circuit that provides the cellular basis for the vast diversity of animal behavior, including human consciousness, cognition, and emotion. Also, studies indicate that abnormal neuronal wiring may be the cause of several devastating diseases such as autism and schizophrenia, further emphasizing the importance of defining brain communication networks and neural circuit organization. Unraveling the structure and function of brain communication networks or specific neural pathways is essential to understand the complex wiring of the central nervous system and to visualize the architectures that underlie a variety of brain functions.

The technique retrograde tracing explores the neuronal retrograde axonal transport and is commonly used to elucidate the relationship between neuronal structure and function across the nervous system. Table 1 lists human genes involved in retrograde axonal transport. Retrograde transport requires an uptake of a tracer along the axons projecting from the infusion site and transport to a distal area where it subsequently transduces the host cell nucleus. An impressive array of tools ranging from traditional dyes, fluorescent proteins to viral retrograde tracers are utilized by neuroanatomists to visualize neural connections (Figure 1). These tracing techniques have provided valuable information on neural circuit organization and draw links between the functions of neurons located in widely separated brain areas; however, much remains to be discovered. This review focuses on the evolution of different neuronal tracing techniques from conventional tracing to viral vectors with emphasis on their advantages and limitations.

[enlarge]

Figure 1. Viral and classical retrograde tracing.

Traditional methods of neuronal tracing are based on the principle of axonal transport. Axonal transport manages the intracellular movement of small or big molecules or organelles in a neuron. It is a powerful way of transporting a variety of substance ranging from important physiological molecules to pharmaceutical drugs. Based on the direction of transport, axonal transport can be either retrograde or anterograde. In anterograde transport macromolecules such as actin, myosin, and clathrin and membranous organelles (e.g., mitochondria) are taken up by the perikaryon and dendrites and transported to the axon terminal. While in retrograde transport the direction is from the axon terminal to the cell body. Neuronal tracer methods that employ this class of transport are called retrograde tracer. These ‘classical’, or ‘conventional’ retrograde, tracers include plant enzyme horseradish peroxidase (HRP) [4-6] albumin protein labeled with HRP [7, 8], plant lectins (wheat germ agglutinin WGA) [9, 10], Evans blue dye (EB) [11] or other fluorescent molecules such as fluoro-gold (FG) [9, 12-14] and fluoro-ruby (FR) [15, 16]. Table 2 provides an overview of the main categories of traditional retrograde tracers.

A novel method to detect axonal projection patterns from a source to a set of target regions and the count of neurons has recently been described [25]. An initial region projecting to n targets could have 2n-1 potential projection types. To count the cells with different color combinations, the injection of uniquely labeled retrograde tracers in k target regions (k < n) was proposed. Thus, the cell counts for color combinations from n-choose-k experiments can serve as a foundation of a model, which can be implemented using evolutionary algorithms. The study applied this method to evaluate the projections of mouse primary motor cortex to the somatosensory and motor cortices.

Kristensson and Olsson first demonstrated the retrograde axonal transport of HRP in 1971 in motor neurons [7, 8]. HRP was deposited in the gastrocnemius muscle of rats; HRP-activity was detected in the motor neurons in the spinal cord after a few days using the Karnovsky‘s diaminobenzidine (DAB) substrate staining method (sections were incubated in a mixture of hydrogen peroxide and diaminobenzidine) [26]. They further extended their study by injecting HRP tracer into the tongue muscle of suckling mice. Sixteen to 24 h after the injection, labeled motoneurons in the hypoglossal nucleus [7, 8] was documented which demonstrated retrograde axonal transport of HRP. Following this, LaVail and LaVail (1972) demonstrated uptake and retrograde transport of horseradish peroxidase by axons of two different neuronal populations in the central nervous system of the chick: from terminals in the retina to their cell bodies in the isthmo-optic nucleus (ION), and from terminals in the optic tectum to retinal ganglion cell bodies [27]. The technique was further improvised by replacement of hydrogen peroxide and diaminobenzidine substrate by tetramethylbenzidine [6], (TMB), o-dianisidine, benzidine dihydrochloride (BDHC) [28], p-phenylenediamine (PPD-PC), 1-naphthol/azur B [29], alkaline phosphatase (purple reaction product), [30], the magenta chromogen, Vector-VIP [31], and HistoGreen [32, 33]. Discovery of HRP as a neuronal tracer was a breakthrough in retrograde tracing; however, its use was limited as HRP exhibits no specific affinity for the cell surface as its uptake into neurons occurs via a passive process of endocytosis resulting in inefficient uptake at the injection site. This limitation was improved by using wheat germ agglutinin-conjugated (WGA) HRP as tracer instead of native HRP [17]. Active, receptor-mediated mechanisms is one advantage of the lectin. When WGA was conjugated to HRP, its uptake and transport of HRP were greatly enhanced resulting in much higher sensitivity than HRP alone.

Wheat germ agglutinin‐apo HRP gold was further described by Basbaum et al, as a sensitive marker for the retrograde tracing of the projections of central nervous system neurons at the light-microscopic (LM) level. A silver-enhancement procedure was used to detect the gold in the tracer [18]. Wheat germ agglutinin‐apo HRP gold is a colloidal‐gold‐labeled retrograde tracer which is made by conjugating colloidal gold to wheat germ agglutinin (WGA) which is then coupled to enzymatically inactive horseradish peroxidase (apo HRP). This tracer offers several advantages which make it unique: longer persistent time in cells, limited diffusion, use in multiple labeling studies, enhanced sensitivity, compatibility with tissues fixed in a variety of aldehyde fixatives, easy detection. Besides, as these tracers contain an inactive form of HRP molecule, it can be concurrently used with fluorescent and HRP-based immunocytochemical and tracing techniques [18]. However, its use in multiple labeling studies is complicated by several limitations present in autoradiographic and enzyme or lectin histochemical tracing methods including longer development time, low sensitivity and labor. Wheat germ agglutinin can also be conjugated with a fluorophore such as Alexa Fluor 555 or 647 and used [9].

Bacterial toxin fragments (cholera toxin subunit B (CTB) [19, 34] and tetanus toxin [20] were also explored as conjugates for HRP or fluorescent tags such as Alexa Fluor 555 or 488 [35]. CTB is the nontoxic subunit B of cholera toxin protein complex secreted by the bacterium, Vibrio cholerae, which binds to cell surfaces through the pentasaccharide chain of monosialotetrahexosyl ganglioside. A conjugate of non-toxic B fragment of cholera toxin and HRP (choleragenoid) (CTBHRP) as a neuronal tracer offers several advantages. It is highly sensitive due to receptor-mediated uptake; it binds to the monosialoanglioside (GM1) ganglioside of the nerve cell membrane, labels a large number of neurons, and allows immunohistochemical detection (rather than histochemical assay) and diffuses less from labeled neurons than the native HRP. A very recent study by Yao and colleagues, 2018 demonstrates the usefulness of CTB for retrograde labeling of small, non-image forming nuclei in the brain to which specific retinal ganglion cells subtypes project their axons [36]. Szőnyi A et al injected 0.5% Cholera toxin B subunit from List Biologicals into mouse brain regions for studies on memory formation [13]. Beltramo R et al injected Alexa Fluor 647-conjugated cholera-toxin subunit B into mouse cortical area postrhinal cortex to study the collicular primary visual cortex [37]. N Kataoka et al injected CTB conjugated with Alexa Fluor 488 or Alexa Fluor 594 into rat brains [38].

Limitations associated with the use of HRP-based tracers paved the way for the discovery of fluorescent retrograde tracer facilitating direct visualization of labeled neurons. Kristensson in 1970 described the transport of a conjugate of cattle albumin and the fluorescent dye, Evans blue in peripheral nerves; whereas the conjugate was injected into gastrocnemius muscle of rats and a fluorescence signal was detected in the spinal motoneurons [39]. This discovery ushered in various fluorescent dyes: bisbenzimide, diamidinophenylindol (DAPI), propidium iodide [21], True Blue, Fast Blue, Granular Blue [21, 23, 40, 41], Nuclear Yellow, Lucifer Yellow, and Diamidino Yellow [42] in the late 1970s and early 1980s [43], and later FG, CTB, fluorescently tagged beads, Mini Ruby (MR) and fluorescently tagged dextran amines. These tracers offer several advantages: increased fluorescence, high efficacy, minimal fading, and improved stability under ultraviolet light excitation. Several fluorescent retrograde tracers are commercially available now and have been used with variable success [36]. True Blue (TB), FG, FR, and 1,1′-Dioctadecyl-3,3,3′,3′-tetramethyl indocarbocyanine perchlorate (DiI), Fast blue (FB) are commonly used fluorescent dyes, for retrograde tracing alone or in combination. All these fluorescent retrograde tracers have their own set of attributes, advantages, and disadvantages; however, none of them is ideal and can be used in all circumstances.

Fluoro-Gold (FG) is a potent water-soluble crystalline retrograde neuronal tracer with following properties: strong fluorescence, extensive filling of neurons, labeling neuronal soma and dendrites, high resistance to bleaching, no uptake by intact undamaged fibers of passage, no diffusion from labeled cells; and not transported transsynaptically. Furthermore, it has a consistent and pure commercial source, a wide latitude of survival times and compatibility with all other tested neuro-histochemical techniques, widely compatible among different species and ideal for multiple labeling studies, that can be used in combination with other neuronal tracers [9, 14, 44]. The active constituent of FG is hydroxystilbamidine, a weak base which after excitation under UV gives bright gold or blue emission depending on pH. FR can be used at a wide range of concentrations, usually at 2–10%. For example, Szőnyi A et al injected 2% FluoroGold from Fluorochrome into mouse brain regions for retrograde tracing to study neuronal basis for negative experience in mouse [12] and, in a different article, the role of brainstem nucleus incertus GABAergic cells in contextual memory formation [13]. However, FG is neurotoxic resulting in neuronal cell death and may cause tissue damage at the injection site, making it non-suitable for long-term labeling. On the other hand, True Blue is a fluorescent retrograde tracer which is effectively transported over long distances and produces blue fluorescent labeling of neuronal cell body cytoplasm, nucleolus, proximal dendrites and axons with a wideband ultraviolet excitation.

Fluoro-Ruby (FR) (dextran tetramethylrhodamine; MW 10,000) was introduced as a new class of tracer by Schmued et al in 1990 [22]. It is a highly sensitive tracer that can be used for both retrograde and anterograde tracing. FR produces a deep red fluorescence and labels cytoplasm and proximal dendrites [45]. FR is the most appropriate tracer for multiple labeling in combination with other tracers as its fluorescent excitation, and emission profiles are distinct from those of FG and TB [46, 47]. In contrast, FG due to its broad emission spectrum is less suitable in combination with other fluorescent tracers in retrograde double-labeling procedures intended to demonstrate collateralization, because it tends to mask the presence of the second fluorescent tracer in the same cell [14, 33]. Han and colleagues demonstrated that the Fluoro-ruby retrograde fluorescent tracing technique could be accurately used to display the anatomical location of the corticospinal tract in the guinea pig [48]. Recently, Yu and collogues, 2015 compared the retrograde tracing of rat spinal motor neurons using four fluorescent dyes, TB, FG, FR, and DiI in single- and double-labeling experiments [49]. The result of the study indicated that TB, FG, and DiI have similar labeling efficacies in the retrograde labeling of spinal motor neurons in the rat femoral nerve when used alone. FR labeled the fewer number of neurons at day 3, however, at day seven after application its efficacy was not significantly less than other tracers used in the experiments suggesting that FR undergoes slow retrograde axonal transport [49]. Zele and colleagues, also reported that after ten days of application FR labeling efficacy was less in dorsal root ganglions than that of TB, FG, DiI or Diamidino Yellow [50]. In double-labeling experiments, combinations of DiI and TB or FG have similar labeling efficacies and do not interfere with each other's labeling abilities [49].

In addition, FR was also used for retrograde tracing and three-dimensional visualization of the corticospinal tract in guinea pigs [48]. The analysis of the frozen sections showed that the labeling traversed the cervical, thoracic and lumbar segments, and was observed in the ventral region of the posterior funiculus. The 3D-DOCTOR 4.0 software-based reconstruction of spinal cord gray matter and corticospinal tract generated a robust three-dimensional profile. The obtained results demonstrated that the FR retrograde fluorescent tracing can effectively unveil the anatomical location of corticospinal tract in guinea pigs.

Rhodamine-labeled fluorescent latex microspheres (0.02-0.2 micron diameter), introduced by Katz and colleagues in 1984 was the first example of particle-based retrograde tracing. This class of tracers offers distinct advantages: no apparent cytotoxicity or phototoxicity, minimal diffusion, long persistence [24]. Green fluorescent latex microspheres ("beads"), described by Katz and colleagues, are readily transported by neurons in the mammalian central nervous system and diffuse minimally from the site of injection. H Qian et al, for example, injected green Retrobeads IX from Lumafluor to trace the nigrostriatal pathway [51].

Viral vectors constitute a precious class of tools for investigating the neuronal circuit organization. Use of virus as neuronal tracer dates back to the early 1900s when one of the first study exploring the use of Herpes simplex as a retrograde tracer was published [52]. The utility of this technique was further supported by studies obtained from the pseudorabies virus (PRV) and rabies virus [53]. Since then the ever-increasing body of work has led to the development of various neurotracing systems based on neurotropic viruses. Neuronal tracing has been improved dramatically by the ability of these viruses to infect, replicate and migrate within neurons, and across synaptically linked neurons thus overcoming the "dilution" problem of conventional tracers and producing intense transneuronal labeling, as detected immunohistochemically.

Retrograde axonal transport is used naturally by certain viruses to spread from one neuron cell to the next in a chain of neurons (transneuronal transfer). Rabies virus from Rhabdoviridae and PRV (pseudorabies virus) and HSV-1 (herpes simplex virus type-1) from alpha-herpesviruses are the two main classes of viral transneuronal tracers which exhibit retrograde as a part of their natural life cycle [54-57]. In addition to naturally neurotropic viruses, in last two decades there has been a methodological evolution in neuronal tracing with the development of genetically modified viruses and use of a variety of other viral species including pseudorabies virus type 1, vesicular stomatitis virus (VSV), lentivirus, and adeno-associated virus (AAV) [58-60], as listed in Table 3. Other viral vectors have been explored. For example, Siciliano CA et al retrogradely transported a canine adenovirus vector with cre-recombinase from dorsal periaqueductal gray to medial prefrontal cortex in mice [61].

Rabies virus (RBV) is a valuable and powerful tool for explicating neuronal circuit due to its ability to selectively infect neurons and to propagates exclusively between synaptically connected neurons by stringently retrograde transneuronal transfer. ‘Street virus’ and fixed virus, are two well-described classes of RBV. ‘Street’ virus is a wild type isolate from infected animals, whereas the ‘fixed’ virus is a passaged tissue culture-adapted virus. Different strains within these classes may vary considerably in their neurotropism. The early reports of the use of RBV as a transneuronal tracer date back to late 1900 [62, 63]. Ugolini et al, in 1995 demonstrated for the first-time time-dependent transneuronal spread of RBV from post-synaptic to pre-synaptic neurons [64]. Following this study, Kelly and Strick examined transneuronal transport of rabies in the central nervous system of primates after intracortical and intramuscular injections and confirmed that RBV infects synaptically-connected chains of neurons in a time-dependent manner exclusively in the retrograde direction [65]. Subsequent similar studies using different systems for example, mice [66-68], rats [69-71], guinea pig [72], and primates [73, 74] have further demonstrated the merits of RBV as tracers in experimental neuroanatomy [75]. Additionally, significantly reduced cytotoxicity compared to other transsynaptic vectors such as VSV and HSV-1, efficient long-range transport in neurons, high-level gene expression, broad host species range and the ability to be used in primates makes RBV a useful tool for tracing neural circuits [76, 77]. Furthermore, in contrast to the herpes virus, RBV is an RNA virus with a simple life cycle, which is completed in cytoplasm resulting in rapid and effective amplification of the viral genome. The simplicity of RBV genome with 12Kb size that codes for only 5 genes, nucleoprotein (N), phosphoprotein (P), matrix protein (M), glycoprotein (G) and the RNA polymerase (L) allow insertion of any gene of interest without the issues of instability associated with other RNA viruses [77]. All these characteristics make RBV an invaluable viral tracer; however, RBV has its limitations. Extreme caution is required in using a replication-competent RBV since RBV is a deadly virus with broad host range, which limits its use only to the laboratories which are well equipped in the handling of RBV. Also, RBV being a polysynaptic tracer creates ambiguity in determining the direct or indirect projection targets in neural pathways. To address this issue, RBV was genetically modified to create a pseudotyped, recombinant RBV. The resulting virus target infection to specific starter cell types and restrict spread only to monosynaptically connected inputs thus allowing neuronal circuit tracing controllably [76, 77]. RAB envelope glycoprotein (RG), is essential for the trans-synaptic spread and virulence of RBV as it facilitates the budding of viral particles from the host cell. RG-deleted rabies virus (RGdG) can initiate an infectious cycle but cannot spread across synapses due to lack of RG expression. Complementation of G in trans enables trans-synaptic spreading of RG-deleted rabies viruses to directly connected, presynaptic neurons. For example, IC Clark et al developed rabies barcode interaction detection followed by sequencing (RABID-seq) to study microglia-astrocyte interactions that promote CNS pathology in experimental autoimmune encephalomyelitis and multiple sclerosis [78]. Szőnyi A et al injected into vGAT-Cre mouse nucleus incertus AAV2/8-hSynFLEX-TVA-p2A-eGFP-p2A-oG and AAV2/5-CAG-FLEX-oG with an upgraded version of the rabies glycoprotein (oG) increasing trans-synaptic labeling potential and injected Rabies(ΔG)-EnvA-mCherry, and identified the neurons with both eGFP and mCherry colors as starter cells [13], and in a similar fashion for another study [12]. Further, improving the design to target the initial viral infection to a specific cell population, the EnvA-pseudotyped RBV virus is used. In this modification, the RBV virus is pseudotyped with an envelope protein from a separate avian virus, the avian sarcoma, and leucosis virus (ASLV-A) envelope protein EnvA. By doing this, RBV expressing the viral protein EnvA on its surface only infects the cells with TVA, an avian receptor for the envelope protein EnvA. As mammalian cells lack an endogenous receptor for EnvA, the only cell population that expresses TVA will be susceptible to RV transduction. Insertion of a gene of interest like fluorescent protein such as GFP in the place of the deleted RG allows the visualization of both the presynaptic input population and the starting. In this process, RGdG is introduced into a neuronal cell type of interest with G protein supplied in trans, which enables the virus to spread transsynaptically to cells in direct synaptic contact with starter cells. Once in these presynaptic neurons, the G-deleted RABV can divide and replicate but cannot spread further due to the deficiency of RG hence restraining the approach to a monosynaptic spread [77]. For example, Zhang J et al injected RABV-deltaG-GFP into the rostral nucleus of the solitary tract to study sour sensing [79]. Sando R et al injected RbV-CVS-N2c-deltaG-GFP (EnvA) into mouse hippocampi to study the role of latrophilin in synaptogenesis [80]. Loureiro M et al injected EnvA36 G-deleted Rabies-GFP-RbE into mouse medial prefrontal cortex to investigate the monosynaptic pathway for social transmission of food preference [81]. Pseudotyped rabies virus is a safe and excellent tool to decipher brain circuitry in both the central and peripheral nervous system [80, 82]. RG-deleted rabies virus (RGdG) offers several advantages over non-viral tracers such as HRP, which cannot mark and trace the input populations to specific cell types. Tetanus Toxin C fragment and Wheat Germ Agglutinin allows cell type-specific targeting, in a Cre-dependent manner; however, these vectors come with the major downside of low sensitivity and weak signals making it challenging to visualize connectivity.

Mono-trans-synaptic labeling with RABV-G has been used to systematically map long-distance connections between mitral cells in the olfactory bulb and their postsynaptic targets in the olfactory cortex [83]. Deletion-mutant rabies viral (RV) vectors have been proven to be a powerful tool for studying the organization of neuronal circuit, via monosynaptic tracing and rapid, high-level transgene expression of fluorophores, opsins or activity indicators in neurons targeted by multiple routes. However, due to their cytotoxicity leading to the death of infected cells, their use has been confined only to short-term experiments. Studies have also shown evidence of morphological changes related to cytotoxicity in RVdG transduced cells after 16 days of infection, thus limiting its use to short term experiments up to two weeks [84].

Recently a monosynaptic tracing system based on double-deletion-mutant rabies viral vectors was described by Chatterjee et al [85]. In addition to deletion of RAB envelope glycoprotein gene, viral polymerase, which is required both for transcription of viral genes and for replication of the viral genome has also been deleted in the new monosynaptic tracing system. The authors demonstrated that deletion of the viral polymerase gene abolished the cytotoxicity and reduced transgene expression to trace levels but left vectors still able to retrogradely infect projection neurons and express recombinases, allowing downstream expression of other transgene products such as fluorophores and calcium indicators [85]. Neurons transduced with double-deletion-mutant rabies viral vector survived at least for four months and showed no detectable morphological changes at 1-year post-injection [85]. Chatterjee et al, also demonstrated that in comparison to other retrograde viruses such as retro-AAV, double-deletion-mutant rabies viral vector has a wider tropism. However, as the double-deletion-mutant rabies viral vector is not yet compatible with the transsynaptic tracing system and can only be used for delivering genes that do not need much expression levels to be functional, such as cre and flpase, future studies are warranted to gain more in-depth insight for overcoming these limitations.

An optimized system for preparation of a high-titer CVS-N2c-deltaG virus has been developed to evaluate the application of rabies viral vectors in studies of functional networks [86]. N2cG-coated CVS-N2c-deltaG was efficiently used for the retrograde access to projection neurons. Furthermore, CVS-N2c-ΔG was shown to express sufficient recombinases for transgene recombination. The study indicated that the CVS-N2c-ΔG-based toolkit may be used as an effective tool for research on neural circuits.

Another study applied the rabies virus-based, monosynaptic retrograde tracing assay to detect de novo synaptic connections between early retinal cell types in retinal organoids [87]. Among the observed presynaptic cells, the authors analyzed photoreceptors and retinal ganglion cells, which are crucial for retinal cell replacement. The described method is applicable to evaluate synaptic connections in retinal neurons in culture and characterize different stages of synaptogenesis.

Few of the early reports demonstrating the neurotropic properties of herpesvirus were made by Goodpasture and Teague, in 1923 [88] and by others [57, 89]. Ugolini in 1987 was first to explore the tracing properties of the herpes virus as a retrograde tracer [57]. Herpes simplex virus (HSV-1) and PRV (suid herpesvirus 1 or pseudorabies virus) belonging to the subfamily, alphaherpesvirinae of Herpesviridae are most extensively studied neuronal tracers. HSV-1 belongs to genera simplex virus, and is a pervasive opportunistic human pathogen, usually causing cold sores while PRV from genera varicella virus is a veterinary pathogen producing Aujeszky's disease in swine and does not infect humans or non-human primates. The alpha herpes viruses are DNA viruses with large and complex genome surrounded by a layer of proteins known as the tegument, which is contained within a membrane envelope. They can establish lifelong infections through latency in ganglion post-primary infection. The key features that make both HSV-1 and PRV excellent for tracing neural circuits are natural neuronal tropism, transneuronal spread, and broad host range. Zhang X et al injected PRV-CMV-mRFP and PRV-CMV-GFP into mouse spleens to study the neuronal control of humoral immune responses in spleen [90].

The infectious process that leads to the labeling of synaptically connected neurons over long distances is initiated by attachment and fusion of viral particles to the cell surface of target neurons. Once inside the peripheral nervous systems (PNS) of their hosts, viral particles are transported to the cell body of neurons by retrograde transport, where they establish a latent infection. The processes of retrograde axon transport for HSV-1 and PRV are strikingly conserved and host species-independent [91]. Upon reactivation from latency, anterograde transfer of progeny viral particles travels from the ganglia toward the nerve terminals occurs, which can manifest in different forms. Occasionally, the infection can spread into the central nervous system causing encephalitis or herpes keratitis [91, 92]. Carta I et al injected GFP–tagged H129 strain of herpes simplex virus type 1 into cerebellar nuclei and allowed anterograde one synaptic transmission in 50 hours to dopaminergic and nondopaminergic neurons in the ventral tegmental area [93].

Specific strains of the alphaherpesvirus have a preference in transneuronal transmission and can spread both anterogradely and retrogradely, strain H129 of HSV-1 displays transneuronal transmission only in the anterograde direction, whereas PRV Bartha strain transmits only in the retrograde direction [94-96]. PRV is often preferred as a transneuronal tracer for experimental studies for many reasons; it does not infect humans, and produces high titer infection in a wide variety of cell types and non-primate species, it can be easily manipulated and is safe to use in laboratory settings. Furthermore, PRV spreads directionally within chains of synaptically connected neurons by direct cell-cell contact. Studies in a variety of non-primate species have successfully used PRV as a transneuronal tracer [97, 98]. Virulent wild type strains of PRV, such as PRV-Becker, PRV-Kaplan, and NIA3, often induce a robust inflammatory response that kills animals within a few days of infection before much of the circuitry is infected. This limitation has led to the development of various attenuated strains of PRV which has proven to be particularly successful in neuronal tracing. The Bartha strain of PRV is one of the best characterized attenuated strain which induces a significantly controlled inflammatory response, with increased host survival time and spread trans neuronally specifically in the retrograde direction when injected at peripheral sites, unlike wild-type PRV strains which spread in both the anterograde and retrograde directions and are highly virulent. Reduced virulence and monodirectional retrograde spread of PRV-Bartha strain are attributed to mutations in gE, gI and US9 genes from the wild type genome. Subsequently, effective genetic modifications in PRV-Bartha strain leading to development of isogenic recombinants of PRV- Bartha that express unique reporters of infection such as enhanced green fluorescence protein (EGFP) or-galactosidase or red (monomeric red fluorescence protein) fluorescence protein has made it a valuable neuronal tracer for identification and characterization of neurons embedded within complex neural networks. Table 3 shows a list of recombinants made from PRV-Bartha for retrograde tracing. PRV-Ba2001 a Cre-recombinase dependent retrograde only tracer with conditional expression in cre expressing cells has been used widely to trace connections to specific neural cell types [99, 100]. An improvised version of PRV-Ba2001 called as “PRV-Introvert” in combination with Retro-TRAP a molecular profiling method was successfully used for anatomical mapping and molecular profiling of polysynaptic inputs to molecularly defined neural populations to study projections to the mesolimbic dopamine system [101]. Recently, Hogue and colleagues have developed a new PRV-Bartha recombinant (PRV 290) virus that constitutively expresses mTurquoise2, an intense and stable cyan fluorescent reporter protein [102]. Their results demonstrated that PRV-290 is selectively transported retrogradely through polysynaptic circuits with temporal kinetics similar to those of its isogenic recombinants, replicates productively in neurons also co-infected with PRV152 and/or PRV-614 and produces a bright cyan fluorescent reporter that is resistant to fading, fills all compartments of infected cells, and is stable at survival intervals extending to 96 hours [102]. Jansen et al in (1995) were first to introduce dual viral tracing by recombinant strains of PRV-Bartha [103].

Although targeted retrograde gene delivery strategies using rabies virus, HSV, VSV, and pseudo-rabies virus have shown quite encouraging results, toxicity from these vectors typically results in low levels of expression and variable results. Recombinant adeno‐associated virus (AAV) vector‐mediated gene transduction into targeted neuron has recently emerged as a powerful tool to understand specific neuronal morphology in the brain and to track their functions [113-115]. AAV is a small, non-pathogenic single-stranded DNA parvovirus that is not currently known to be host-specific when administered directly into the brain [114]. AAV is particularly attractive due to its low immunogenicity and toxicity and its long-lasting transgene expression. A large number of evolutionarily different AAV serotypes have been identified so far [115]. When compared directly after brain injection, these serotypes demonstrated distinct transduction efficiency, travel and distribution patterns in the nerve cells. In a recent report, Haenraets 2017 analyzed the ability of seven different serotypes (1, 5, 6, 7, 8, 9, rh10) to transduce neurons retrogradely and demonstrated that most of the tested AAV vectors have very similar transduction efficiency in spinal neurons [113].

Interestingly, AAV serotypes 1 and 9, frequently transduce cells at a substantial distance from the injection site; however, serotype 8, rarely shows any distal transductions [116]. Previous studies have also confirmed that some of the AAV serotypes can travel in both anterograde and retrograde directions efficiently. To understand the directional AAV transport, Castle and colleagues, 2014 applied dye-labeled AAV serotypes 1, 8 and 9 to the axon termini or the cell bodies of primary rat embryonic cortical neurons and compared the axonal transport pattern in vitro and in vivo [116]. The authors reported that all three serotypes tested were actively transported along axons in both the anterograde and retrograde directions and shared conserved mechanisms for axonal transport both in vitro and in vivo. AAV serotype 9 has also been shown to efficiently travel in both anterograde and retrograde directions by several other groups [116-118]. Additionally, Castle and colleagues, 2014 also observed that AAV serotype 1 exhibits a higher frequency of axonal transport as compared to serotypes 8 and 9 [116]. Other have also confirmed that the AAV serotype 1 is highly efficient to retrogradely transduce DRG neurons after intraspinal injection [118, 119].

In a recent report, Low and colleagues compared the efficiency of AAV serotypes 6, 8, and 9 to transduce dopaminergic neurons in the substantia nigra either through direct nigral injection or striatal delivery [120]. While they found that all three serotypes transduced nigral dopaminergic neurons with almost equal efficiency after direct nigral injection, AAV serotype 6 was found much superior to serotypes AAV8, and AAV9 for retrograde transduction of nigral neurons post striatal administration [120]. For sequential administration to nigral dopaminergic neurons, the combination of serotypes AAV9 with AAV6 was associated with increased transduction efficiency than serotypes AAV8 with AAV6 or repetitive serotype AAV6 administration. Zheng and colleagues, 2010 further investigated the capacity of the AAV8 vector in transducing PNS, dorsal root ganglion, and CNS by intraperitoneal injection in neonatal mice [121]. Additionally, to find more direct evidence that AAV8 vector can enter into spinal cord and PNS via retrograde transport, they also injected AAV8 vector by direct intramuscular injection in adult mice. These results indicated the mechanism of retrograde transport of AAV8 and their use for potential gene therapy application including, chronic pain, demyelination diseases, and lower motor neuron diseases. Later, Matsuzaki and colleagues explored the marmoset model to better understand the serotype AAV9 transduction profile in the CNS pathways after intracisternal and cerebellar parenchymal injection [122].

Interestingly, the intracisternally injected AAV9 vectors were retrogradely transported to CNS via either intermediary transcytosis or neuronal axons route, resulting in diffuse and global transduction within the CNS. In contrast, cerebellar parenchymal injection of AAV9 serotype led to robust transduction in more confined locations, including the cerebellar cortex and cerebellar afferents, including the vestibular nucleus, neurons of the pontine nuclei and inferior olivary nucleus. Additionally, in the spinal cord, both instillation methods produced retrograde transportation and subsequent transduction of neurons and labeling of the dorsal column-medial lemniscus, spinocerebellar tracts, and spino-olivary fibers. These results suggest that intracisternal route of AAV instillation should be utilized if diffuse and global transduction in the CNS is required, for example against gene therapy diseases affecting broader brain areas, such as Alzheimer and mucopolysaccharidosis diseases. Whereas, the cerebellar parenchymal injection may be of the route of choice for more limited and stronger transduction of cerebellar cortex and cerebellar efferent neurons areas, for example, against diseases that mainly affects the cerebellum, the associated brainstem nuclei, and spinal cord, such as spinocerebellar degeneration.

It is important to recognize that the axonal transport of some of the AAVs is almost exclusively serotype-dependent [118]. For example, AAV serotype 6 is considered highly neuron-specific in both rat and non-human primate brain and transport almost exclusively in a retrograde direction. The ability of AAV serotype 6 to retrogradely transport through axons is of great importance for AAV6-based gene therapy applications including for lysosomal storage diseases, Alzheimer disease, Parkinson’s disease and Huntington’s disease [118]. In contrast, AAV serotype 2 has been typically found to be associated with anterograde axonal transport in rat and non-human primate brain [118].

Since the original report first presented by Kasper and colleagues [123], many other studies have confirmed the retrograde transportation ability of AAVs; however, the overall tendency for retrograde transportation of AAV serotypes has been relatively low, hindering long-term, high-level transgene expression for clinical-grade gene therapy applications. For example, when Hollis and colleagues investigated the retrograde transportation abilities of self-complementary AAV (scAAV) serotypes 1-6 after peripheral injection into either rat extensor carpi muscle or sciatic nerve, the overall retrograde transduction with scAAV Serotype 1 was found approximately 4% of extensor carpi motoneuron pool after intramuscular injection and about 7% after sciatic nerve injection. In contrast, retrograde transduction with scAAV2 was barely visible after intramuscular injection, and less than >1% of MNs following intranerve injection [124]. Whereas, Towne and colleagues found that retrograde delivery of recombinant rAAV2/6 construct in primates could result in a maximum 15% transduction of the motoneuron pool [125]. In a recent report just published, when Yamaguchi and colleagues (2018) tested efficiency of four different AAV vectors (AAV9.hSyn.hChR2(H134R)eYFP.WPRE.hGH, retroAAV mCherry-Cre, AAV2/1.Syn.ChR2(H134R)eYFP.AWP.hGH and AAV-PHPeB:Cag-mNeonGreen) to introduce transgenes into the telencephalon of adult X. laevis, the probability of infecting neurons was found extremely low [126]. Out of 49 adult X. laevis administered with AAVs, only two animals (injected with AAV9.hSyn.hChR2) demonstrated stronger reporter gene expression in neurons three weeks post instillation. Moreover, none of the remaining animals showed any sign of neuronal expression after the same post-infection period (4% success rate) suggesting that AAV may not be a suitable vector to be used in adult X. laevis nervous system. Therefore, the overall efficacy of retrograde transduction using AAV vectors is still considered inadequate and highly serotype dependent.

Although AAV1 vectors are generally considered superior to AAV2 vector for retrograde axonal transport and neuronal gene transfer, only a tiny fraction of intramuscularly injected AAV1 vector was delivered to the spinal cord [127]. To improve the AAV1 vector-based gene delivery, Davis and colleagues engineered the AAV1 vector capsid by Tet1 peptide insertion [128]. The modified AAV1 vector exhibited markedly superior transduction efficiency in cultured motor neurons, as well as enhanced retrograde delivery when compared with the unmodified AAV1 vector. Therefore, adopting such a strategy could be extremely beneficial in other serotypes that are considered more effective than AAV1 for retrograde transport. Recently Tervo and colleagues have developed AAV serotype 2 capsid variant ‘rAAV2retro’ that has been shown to enhance significantly retrograde transduction in neurons with almost equal efficiency to classical synthetic retrograde tracers and also effective cross-functional interrogation of the neural circuit and in vivo genome manipulations in targeted neuronal populations [58]. Haenraets et al 2017 compared transduction efficiency of rAAV2retro with commonly used rAAV serotypes and reported a > 20‐fold increase in retrograde transduction efficiency of descending supraspinal neurons [113]. A recent study just published by Wang and colleagues has successfully demonstrated that injection of AAV2-Retro to the lumbar and cervical spinal cord of adult female mice results in extremely robust and efficient transduction of supraspinal populations throughout the brainstem, midbrain, and cortex (including the reticular, red nucleus, and corticospinal) [129].

Interestingly, supraspinal populations, (including corticospinal and rubrospinal neurons), were transduced with greater than 90% efficiency, with strong transgene expression reached within three days of administration. Furthermore, retrograde transduction was also found highly efficient when vectors were delivered after a spinal injury, highlighting the therapeutic potential of AAV2-Retro. Finally, they found that after retrograde delivery of inhibitory Designer Receptors Exclusively Activated by Designer Drugs (DREADDs), receptor activation in transduced supraspinal populations was adequate to induce complete forelimb paralysis. Therefore, rAAV2-retro is by far the most efficient AAV vector in transducing supraspinal descending projection neurons and holds a grand promise for many clinical gene therapy applications such as Alzheimer’s, lysosomal storage diseases, and Parkinson’s disease. Table 5 summarizes axonal transport properties of some of the well-characterized AAV serotypes in mammalian CNS.

An editable toolkit was generated to produce AAV11, which showed potent retrograde labeling of projection neurons in male wild-type or Cre transgenic mice [130]. When combined with fiber photometry, AAV11 was applied to monitor neuronal activities by retrograde delivering calcium-sensitive indicator regulated by the Cre-lox system. Moreover, GfaABC1D promoter embedding AAV11 was found to be superior to AAV8 and AAV5 in astrocytic tropism demonstrating that AAV11 can be effective to analyze neuron-astrocyte connection. Besides, AAV11 was effective for evaluation of circuit connectivity difference in the brains of the experimental mice with Alzheimer's disease and control mice.

Regarding the research of the bed nucleus of the stria terminalis (BNST)-centered mechanisms, a recent study applied a novel viral-genetic tracing to verify synaptic circuit effects on subregions of BNST in mice [131]. The method was based on monosynaptic canine adenovirus type 2 (CAV2) and rabies virus retrograde tracers. The results showed that while the lateral BNST region had more connections from prefrontal, insular and ectorhinal/perirhinal cortices and anterior thalamus, the medial BNST area had inputs from the medial amygdala, lateral septum, hypothalamus nuclei and ventral subiculum. Overall, the study presented a comprehensive map of the afferent inputs to lateral and medial BNST subregions.

The Retroviral vector systems based on human immunodeficiency virus type 1 (HIV-1) provides stable, long-term gene expression (by virtue of its ability to stably integrate its genome DNA into a host cell) and are widely used for gene transduction applications including neurological and neurodegenerative diseases. The unique ability of stable integration distinguishes retroviruses from other commonly used viral vectors, including VSV, HSV, and AAVs. The two common retroviruses that have been used extensively as viral vectors are the gamma-retroviruses and lentiviruses (LVs). Gamma-retroviruses does not enter to the nucleus intact, whereas the LVs DNA can directly enter the nucleus without nuclear envelope breakdown, and this unique feature makes LVs particularly attractive for gene delivery into neurons. As such, the natural envelope of the LVs is replaced (pseudotyped) with a heterologous envelope. Most LVs are traditionally packaged with vesicular stomatitis viral glycoprotein envelope (VSV-G), that provides LVs a unique ability to transfer genes to a broad range of cell types, however, restrict their ability to retrogradely infect neurons from injected sites [132].

[enlarge]

Figure 2. Schematic drawing of FuG-B/B2 and FuG-C; highly efficient retrograde gene transfer (HiRet) vectors [1, 2].

Furthermore, retrograde infections with VSV-G are not suitable for long-term studies of neural circuitries due to their toxicity (if expressed constitutively) or latency-induced suppression of gene expression [133]. To address this, rabies virus glycoprotein (RV-G) based pseudotyping of LVs has been tested in several studies [2, 132, 133]. Pseudotyping based on particular variants of LVs, for example, RV-G pseudotyped equine infectious anemia virus (a variant LVs) (RV-G-EIAV), is particularly attractive to promote the retrograde gene transportation into the central nervous system [132]. Mazarakis and colleagues compared the transduction abilities of (VSV-G-LVs and RV-G-EIAV) and found that pseudotyping with RV-G-EIAV led to an increased gene transfer to neurons due to retrograde axonal transport and transduction of distal neurons connected to the injection site [132]. Interestingly, this retrograde transport was not restricted to specific neuronal populations and occurred after vector transportation to the striatum, spinal cord, substantia nigra, and muscle. Cetin and colleagues further used this LV system based on EIAV to generate rabies glycoprotein-pseudotyped LVs vector that uses a positive feedback loop composed of a Tet promoter driving both its tetracycline-dependent transcription activator (tTA) (“TLoop”) and channelrhodopsin-2-YFP (ChR2YFP) [134]. They showed that after the initial injection, the virus traveled retrogradely and stably expressed gene products in a drug-controllable fashion in neurons that project to injection sites within the mouse brain. The expression was strong and allowed optogenetic studies of neurons projecting to the location of virus administration, as demonstrated by fluorescence-targeted intracellular recordings [134].

[enlarge]

Figure 3. Localization of GFP signals in spinal motor neurons with HiRet vectors. Figure adopted from Hirano et al [2].

To further improve retrograde transportation, Kato and colleagues developed a novel vector system by pseudotyping LVs with a fusion envelope glycoprotein (termed FuG-B) in which the cytoplasmic domain of RV-G was replaced with the cytoplasmic domain from VSV-G glycoprotein [1]. This FuG-B pseudotyping increased the LVs titer in various cell lines and resulted in robust retrograde transport-mediated gene transfer into different brain regions driving the striatum with greater efficiency than that of the RV-G in the mice model. Furthermore, FuG-B administration into the primate model resulted in strong gene expression into the nigrostriatal dopamine region (a primary target for gene therapy of Parkinson’s disease gene therapy). The gene delivery efficiency of FuG-B vector was further improved by employing a variant of FuG-B, named FuG-B2 (Both FuG-B and B2 consists of the extracellular and transmembrane domains of RV-G fused to the cytoplasmic domain of VSV-G) [2]. Later, the same group developed another novel type of vector FuG-C, in which the 16 amino-acid short C-terminal segment of the extracellular domain and transmembrane/cytoplasmic domains of RV-G were substituted with the corresponding domain of VSV-G [2]. (Figure 2). Both vectors (FuG-B/B2 and FuG-C) were termed highly efficient retrograde gene transfer (HiRet) vectors. All of these vectors showed stable, and highly efficient retrograde gene transfer abilities into various neuronal populations in the brain. HiRet vector transduced both neuronal and glial cells around the inoculation site, whereas neuron-specific retrograde gene transfer (NeuRet vector) exclusively transduced the neuronal cells [2] (Figure 3 and 4). As such, HiRet vectors is by far the most efficient retrograde gene delivery into both hindbrain motor neurons and spinal cord, presenting a robust approach for gene therapy of motor neuron diseases. In a recent report, Sheikh and colleagues, 2018 successfully used constitutively active HiRet-GFP to map supraspinal and propriospinal connections terminating in the cervical or lumbar regions of the rat spinal cord [135]. They also suggested that coupling HiRet LVs with a tetracycline-inducible promoter to drive transgene expression with a second virus, AAV2-TetOn, allows a very tight retrograde labeling of specific neuronal populations. Therefore, such gene delivery tools hold a grand promise for many retrograde labeling applications such as neurological and neurodegenerative diseases by delivering genes required for neuronal survival and protection.

[enlarge]

Figure 4. Localization of GFP signals in hindbrain motor neurons with HiRet vectors. Figure adopted from Hirano et al [2].

Vesicular stomatitis virus (VSV), a negative-strand RNA virus, is recently being explored as an attractive candidate for neural tracing mainly due to its ability to be pseudotyped with other virus glycoproteins. In addition to its wide host range and strong gene expression capabilities, VSV is also not as biohazardous as other neuronal viruses currently being used for retrograde tracing. Previous studies have shown that VSV can spread across synapses in anterograde or retrograde directions depending on the types of glycoprotein that are encoded [83, 136-139]. Beier and colleagues, 2011, 2013 and 2016 have successfully demonstrated that VSV encoding rabies glycoprotein (RABV-G), travels across synapses retrogradely [83, 136-138]. Mudel and colleagues, 2015 have confirmed retrograde transsynaptic transmission of recombinant (rVSV) in the zebrafish, mouse, and chickens [139]. More recently, Yamaguchi and colleagues, 2018 reported the successful use of VSV as a transgene vector in amphibian brains [126]. Although VSV is generally considered toxic to the host cells, the neurons transduced with VSV exhibited normal physiological functions up to 7 days of infection [126].

Additionally, infection with VSV does not prevent the vocal circuits from generating normal fictive vocalizations up to 9 days after infection. Interestingly, the authors did not find any evidence of transsynaptic spread of VSV in frog neurons in retrograde or anterograde directions, as described previously [83, 136-138]. The authors concluded that the properties of VSV spread across synapses may vary among different host species and possibly the efficiency of the spread is significantly lower in amphibians. Nevertheless, the efficacy of retrograde transduction using VSV vectors is still under infancy and highly dependent on the envelope glycoproteins that coat the viral particle. Another major drawback with VSV is its vector-induced cell toxicity [83, 136-138, 140].

[enlarge]

Figure 5. A receptor complementation technique activates neurons for the infection by canine adenovirus type-2 (CAV-2). Adeno-associated viral vectors express the coxsackievirus and adenovirus receptor (CAR) in the target projection neurons (modified from [3] ). The expression of CAR enhances the retrograde delivery of CAV-2 in the neurons.

A recent paper by Li et al has described a receptor complementation technique aiming to enhance viral infection of neurons [3]. The novel receptor complementation method allows infection of the target neuronal cells with canine adenovirus type 2 (CAV-2). The authors generated adenoviral vectors to induce the coxsackievirus and adenovirus receptor (CAR) using a dual virus approach with Cre-recombinase-encoding CAV-2 combined with Cre-dependent adeno-associated viral (AAV) vectors [141] (Figure 5). The AAV vectors were used to introduce CAR and a fluorescent protein expression in the target neurons. After infection with the AAV vectors with CAR construct, the projection neurons begin expressing CAR. Due to its tropism, CAV-2 retrogradely passes from axons to the soma of neurons and expresses Cre recombinase only in the cells projecting to target. The potentiation with CAR significantly elevated the number of GFP-positive pyramidal cells, suggesting that CAR-positive neurons demonstrated a higher degree of CAV-green fluorescent protein (GFP) infectivity. In vivo studies performed in both mice and rats have shown that CAR complementation significantly enhanced the retrograde delivery of CAV-2 in basolateral amygdala neurons. This retrograde labelling approach has been used in a study targeting striatum-projecting orbitofrontal cortical neurons [142]. The authors injected adeno-associated virus into the orbitofrontal cortex and striatum and found that orbitofrontal neurons carry a representation of a single decision variable.

Also, Keefe et al. has recently published a protocol describing a viral vector technology for inserting target genes into neurons [143]. According to the described method, a viral construct, which was designed for highly efficient retrograde transport (HiRet), was delivered to the synapses of propriospinal neurons. The authors suggested that HiRet is mainly applicable to map regenerating and reconnecting areas in the damaged nervous system and to target neurons in ablation experiments.

A recent study combined retrograde and anterograde trans-synaptic viruses to identify regions with direct and indirect effects on the dorsal and ventral prefrontal cortex (PFC) in rats [144]. Using the retrograde tracing using pseudorabies virus (PRV), the authors demonstrated that both dorsal and ventral parts of the PFC are influenced by the dorsal CA3 (dCA3) area of the hippocampus. In addition, the study applied the retrograde monosynaptic tracer Fluoro-Gold, the anterograde monosynaptic tracer Fluoro-Ruby, polysynaptic anterograde viral tracer (HSV-1) and the combination of retrograde AAV with an anterograde AAV. The findings of the study identified parallel disynaptic pathways from the dCA3 to the PFC, which are involved in the hippocampal–prefrontal interactions.

Retrograde tracing is a powerful tool to understand the neuronal circuit and structure/function of brain communication networks in a broad range of animals including rodents and primates. A large number of novel retrograde neuronal tracers based on conventional approaches (such as dyes and fluorescent molecules) are being explored by neuroanatomists to visualize neural connections. Nevertheless, conventional tracing techniques are generally time-consuming, and often associated with low sensitivity and labor intensiveness. On the other hand, retrograde tracing techniques using naturally occurring neurotropic (Rabies virus, PRV, and HSV-1) and genetically modified viruses (VSV, lentiviruses and AAVs) are continually emerging and expanding. Overall these tracing technologies hold a grand promise for many retrograde labeling applications such as neurological and neurodegenerative diseases by transporting genes required for neuronal survival and protection. Although there still are some problems remains, recent developments in genetic engineering and molecular biology tools regarding viral vectors provides the hint of solutions to these problems.

Pankaj Kumar has written this article in his private capacity. No official support or endorsement by Neurocode Labs Inc. is intended or should be inferred. Dr. Konstantin Yakimchuk added the section "Retrograde tracing by receptor complementation" in February 2020.