Third-generation rabies viral vectors have low toxicity and improved efficiency as retrograde labeling tools

Abstract

In this issue of Cell Reports Methods, Jin et al. report the generation and validation of a rabies variant, RVΔL, for projection-based neuronal labeling. RVΔL shows little toxicity in vivo and has an improved growth advantage over another variant, RVΔGL, making it a useful tool for a wide variety of systems neuroscience-based studies.

In this issue of Cell Reports Methods, Jin et al. report the generation and validation of a rabies variant, RVΔL, for projection-based neuronal labeling. RVΔL shows little toxicity in vivo and has an improved growth advantage over another variant, RVΔGL, making it a useful tool for a wide variety of systems neuroscience-based studies.

Main text

A central goal of neuroscience research is to decipher how neurons in the brain organize their projections, as a step toward understanding how these connections facilitate functional actions. While neuroscientists have many tools at their disposal to study brain connectivity, one limitation is that they are often optimized for a specific scale. For instance, functional imaging can correlate connectivity across many brain areas in awake-behaving subjects (from small animals to humans) but cannot resolve the contribution of individual cells. Meanwhile, methods like electron microscopy or electrophysiology can reveal details of connectivity at the level of individual synapses but are not easily implemented across large expanses of the mammalian brain.

Over a decade ago, modified rabies virus emerged as a molecular tool with the potential to bridge this gap. While rabies virus had existed as a neuroanatomical tool for much longer than that, its uncontrolled spread can quickly obscure neuronal circuit order.¹ To address this, researchers made two clever changes to the virus that significantly altered its functionality. The rabies genome encodes five functional genes: nucleoprotein (N), phosphoprotein (P), matrix protein (M), glycoprotein (G), and RNA polymerase (L). G isn’t required for initial infection or transcription of the viral genome within infected cells but instead facilitates budding of the virus from host cells. To alter the functionality of rabies for improved circuit tracing, scientists first substituted the rabies G gene with GFP.² This reagent, referred to as RVΔG-GFP, is a useful retrograde tracer for neuroanatomical studies, due to the efficiency of rabies to infect axons and be retrogradely trafficked back to the cell body (Figure 1). Meanwhile, robust expression of GFP provides easy identification of infected cells in the brain, including fine structures such as axons and dendrites. The second modification researchers made was to pseudotype the modified rabies with an avian coat protein (EnvA) so that its initial infection is restricted to cells that express its cognate receptor, TVA.³ The trick here is that TVA is not endogenously present in mammals. However, one can express it within defined cell types in the brain to direct infection of the pseudotyped rabies (referred to as RVΔG-GFP(EnvA)) selectively into those cells. Finally, if glycoprotein is also supplied in trans within the TVA-expressing cells, this facilitates spread of modified rabies from infected cells (often referred to as “starter cells”) to their pre-synaptic partners. At this point, viral spread is halted because the pre-synaptic cells lack glycoprotein.

Figure 1.

Comparison of modified B19-based rabies strains RVΔG, RVΔGL, and RVΔL

Deletion of specific genes from the B19 genome modifies tool functionality in different ways.

With these modifications, rabies has become widely adopted in the field of systems neuroscience, elucidating previously unknown connectivity at single-cell resolution throughout the nervous systems of many mammalian model systems. Along the way, important improvements have also been made, including replacement of EGFP for other genes⁴ and production of additional modified strains^5,6 and reagents⁷ with improved efficiency and specificity of trans-synaptic spread. Still, a limitation that remains for many rabies-based tools is their inherent cytotoxicity, precluding their usefulness for long-term studies. To address this, groups are pursuing several unique strategies. In self-inactivating rabies, or SiR, components have been added into the rabies genome to facilitate its degradation shortly after infection, before it can kill the host cell.⁸ SiR-infected neurons have been shown to live for many months without obvious functional deficit, though care should be taken to ensure that critical domains within the SiR are not altered by viral mutation.⁹

Independently, removal of an additional gene (L) from the RVΔG rabies variant dramatically improves the long-term health of infected cells¹⁰ (Figure 1). RVΔGL is well tolerated by host cells because loss of G and L reduce viral gene expression to a minimal level, a feature that will also preclude sufficient expression of most transgenes for experimental use. However, recombinases such as Cre and Flp are effective even in miniscule amounts. Thus, expression of these transgenes by RVΔGL, even in a diminished capacity, is sufficient to turn on recombinase-dependent molecular tools for longitudinal structural and functional experiments.

In this issue of Cell Reports Methods, Jin and colleagues describe the synthesis and validation of a “third generation” B19-based rabies viral vector called RVΔL (Figure 1).¹¹ RVΔL builds on the best features of its predecessors, specifically low toxicity (like RVΔGL) and high viral titer (like RVΔG). Specifically, the authors show that when only L is deleted from the rabies genome, it gains a major growth advantage over RVΔGL when propagated on complementing cells in vitro, reaching a viral titer similar to first generation RVΔG rabies. The practical benefits of this growth advantage for RVΔL are demonstrated with in vivo retrograde tracing studies, where Cre- or Flp-expressing RVΔGL and RVΔL variants were delivered to the somatosensory thalamus of reporter mice (Ai14 or Ai65f) and labeled cell bodies were quantified in the cortex at various time points post injection. RVΔL-Flp was especially impressive compared to its predecessor, labeling approximately six times as many cells. The improved infectivity of RVΔL-Cre was more modest, an outcome the authors speculate may be due to a “ceiling effect” where both variants have labeled a majority of the available projections in this pathway. Still, an improved efficiency for Flp-dependent tools is an important advance as it facilitates intersectional use of RVΔL-based retrograde labeling with other Cre-dependent tools or transgenic lines.

Another consideration when applying any viral tool for experimental use is its tropism, or the ability of a given virus to infect different cell types. RVΔL’s main utility is as a retrograde tracer, or a reagent that can be used to label neuron populations via their axonal projections. Several other viral retrograde tracers are widely used by neuroscientists, such as CAV-based vectors¹² and rAAV2-retro,¹³ which might prompt end users to ask why RVΔL-based tools would be preferred. Toward addressing this, the authors perform a comparison, targeting anterior cingulate cortex (ACA) or anteromedial cortex (AMA) in Ai14 reporter mice with CAV-Cre, rAAV-retro-hSyn-Cre, and RVΔL-Cre and quantifying the density of labeled inputs from various brain areas. The results were mixed, which is perhaps not surprising considering the inherent differences of these tools and experimental variability. Still, RVΔL-Cre was either the second or top performer for retrograde labeling in each of the pathways that were analyzed, suggesting that it will have utility for targeting a range of neural circuits. The authors also compared electrophysiological properties of nucleus accumbens-projecting basolateral amygdala neurons at either 4 or 12 weeks post-infection with RVΔL-Cre or rAAV-retro-hSyn-Cre. No differences were observed between cells, emphasizing that, at least through this time window, RVΔL minimally perturbs endogenous cell function. Finally, while not addressed directly, the authors speculate that RVΔL-based tools should function in animals beyond mice due to shared structural features between current variants and previous generations that have been successful in a wide range of model systems. Overall, these experiments indicate that RVΔL-based rabies variants are a useful addition to the molecular toolkit for neural circuit-based research, with added appeal for users who have struggled to achieve robust projection-based neuronal labeling with other viruses.

With this development of third-generation rabies viral vectors, fans of neural circuit tracing will likely wonder if RVΔL can be further modified to undergo monosynaptic spread, especially since this was recently achieved with second-generation vectors.¹⁰ The authors caution that generating an RVΔL variant capable of synaptic jumping is not a straightforward process and will require overcoming two technical hurdles. First of all, RVΔL needs to be complemented with L in trans for the virus to spread beyond initially infected cells. There is indirect evidence to suggest that this is feasible, as the authors previously demonstrated that complementation of RVΔGL with L was sufficient for neuron-to-neuron viral spread in vivo.¹⁰ The size of L (∼6.4 kb) precludes its expression via an AAV helper viruses, as has been done with G and the first-generation RVΔG. Instead, a more complex transgenic strategy is required, which could limit the model systems within which RVΔGL and RVΔL monosynaptic tracing could be performed. A second challenge with implementing RVΔL for trans-synaptic use is that there is not a simple way to direct its initial infection to specific cell types, as was done with previous generation vectors using the EnvA/TVA approach. This is because G must be absent from the rabies viral genome for pseudotyping to be successful; otherwise, a mixture of EnvA and glycoprotein envelope proteins will be present on the surface of the produced virions, allowing them to infect cells independent of TVA.

Despite these future challenges, the authors convincingly demonstrate that RVΔL is a useful addition to the systems neuroscience toolkit for retrograde labeling of neurons with improved efficiency and minimal disruption to cellular function. Furthermore, the experiments described here lay important groundwork for potential development of RVΔL as a monosynaptic tracing tool in the future.