Application of Recombinant Rabies Virus to Xenopus Tadpole Brain

Abstract The Xenopus laevis experimental system has provided significant insight into the development and plasticity of neural circuits. Xenopus neuroscience research would be enhanced by additional tools to study neural circuit structure and function. Rabies viruses are powerful tools to label and manipulate neural circuits and have been widely used to study mesoscale connectomics. Whether rabies virus can be used to transduce neurons and express transgenes in Xenopus has not been systematically investigated. Glycoprotein-deleted rabies virus transduces neurons at the axon terminal and retrogradely labels their cell bodies. We show that glycoprotein-deleted rabies virus infects local and projection neurons in the Xenopus tadpole when directly injected into brain tissue. Pseudotyping glycoprotein-deleted rabies with EnvA restricts infection to cells with exogenous expression of the EnvA receptor, TVA. EnvA pseudotyped virus specifically infects tadpole neurons with promoter-driven expression of TVA, demonstrating its utility to label targeted neuronal populations. Neuronal cell types are defined by a combination of features including anatomic location, expression of genetic markers, axon projection sites, morphology, and physiological properties. We show that driving TVA expression in one hemisphere and injecting EnvA pseudotyped virus into the contralateral hemisphere, retrogradely labels neurons defined by cell body location and axon projection site. Using this approach, rabies can be used to identify cell types in Xenopus brain and simultaneously to express transgenes which enable monitoring or manipulation of neuronal activity. This makes rabies a valuable tool to study the structure and function of neural circuits in Xenopus.

2 Introduction 83 8 n=28/33 tadpoles, 85% infected), presumably from uptake at local axon terminals. We found that injection of 264 SADΔG-EGFP(B19G) into the ventricle produced widespread infection. Therefore, care was taken to ensure 265 that virus was injected directly into brain tissue without leaking into the ventricle in all experiments. Tadpoles 266 are reared at 22°C and we postulated that this decreased body temperature might lead to decreased efficiency 267 of viral infection compared to warm-blooded vertebrates, like rodents, in which rabies virus has been used 268 extensively. Short term incubations at increased temperature have previously been shown to improve infection 269 efficiency in Xenopus with other viruses (Dutton et al., 2009) and with G-deleted rabies in fish (Dohaku et al.,270 2019). To test whether increasing the rearing temperature increased the proportion of tadpoles infected with 271 rabies virus, we injected animals with SADΔG-EGFP(B19G) virus and incubated them at 26°C or 28°C for 4 272 hours immediately following viral injection and again 24 hours later. Tadpoles were housed at 22°C for the 273 remainder of the experiment. Five to seven days later, we found that 83% of tadpoles (n=10/12 tadpoles) 274 reared continuously at 22°C were infected, while 64% of tadpoles (n=7/11 tadpoles) temporarily incubated at 275 26-28°C were infected, suggesting that increased temperature did not impact infection rates (p=0.37 a ). In the 276 infected animals, the number and brightness of EGFP + neurons did not appear different between the groups. 277 These results demonstrate that at normal rearing temperature, a large majority of tadpoles have infected 278 neurons with robust expression of EGFP following injection of B19G phenotypically complemented rabies 279 virus. 280 281 Next, we tested whether TVA expression could be used to mediate infection of SADΔG-EGFP(EnvA) virus in 282 targeted neuronal populations. We transfected tectal neurons with a dual CMV promoter expression plasmid to 283 drive pan-neuronal TVA and turboRFP expression (CMV::TVA/tRFP) using whole-brain electroporation (Haas 284 et al., 2002). Four days later, when tRFP expression was strong, we injected SADΔG-EGFP(EnvA) virus 285 directly into the transfected optic tectum. Three days after viral injection we observed a subset of tRFP + 286 neurons were infected with pseudotyped rabies virus, as identified by EGFP expression ( Figure 1C; n=41/65 287 tadpoles, 63% infected). Therefore, we achieved infection and robust EGFP expression in a majority of animals 288 of EGFP + cells without detectable expression of TVA/tRFP in these experiments ( Figure 1C). To determine 299 whether TVA expression is required for infection with SADΔG-EGFP(EnvA) virus in Xenopus, we injected 300 SADΔG-EGFP(EnvA) virus into the tectum of untransfected tadpoles. TVA-expressing animals were also 301 injected with the same viral aliquot as a positive control and animals were imaged using identical imaging 302 parameters 6 days later. While the majority of TVA-expressing tadpoles were infected 6 days after viral 303 injection (n=13/18 tadpoles), we did not observe any EGFP + cells in the absence of TVA ( Figure 1E; n=0/21 304 tadpoles). This result indicates that the virus cannot infect Xenopus neurons in the absence of the TVA 305 receptor and suggests that EGFP + cells that we observed in TVA/tRFP transfected animals express a low level 306 of TVA which is sufficient to mediate infection, but the tRFP is below detection threshold. These so-called 307 "invisible TVA" neurons have been noted in other studies, in particular when using Cre-dependent gene 308 expression which can have some leakage, because of the very sensitive interaction between EnvA and TVA 309 have been used to reduce the number of invisible TVA cells (Miyamichi et al., 2013). 312

313
To assess whether increasing tRFP expression would decrease the number of invisible TVA cells, we used the 314 gal4-UAS bipartite transcriptional system to amplify gene expression (Chae et al., 2002;Hirsch et al., 2002). In 315 addition to using the CMV promoter, we drove TVA and tRFP expression using the VGAT promoter, which has 316 previously been shown to increase transfection of inhibitory neurons in the tectum (He et al., 2016). The 317 proportion of excitatory:inhibitory neurons in the optic tectum is 70:30 (Miraucourt et al., 2012). Using post-hoc 318 immunohistochemistry for GABA, it has been demonstrated that the VGAT promoter increases expression in 319 inhibitory neurons so that the transfected population is 50:50 excitatory:inhibitory (He et al., 2016). Co-320 electroporation of VGAT::gal4 or CMV::gal4 with UAS::TVA, and UAS::tRFP into the tectum, followed by 321 injection of SADΔG-EGFP(EnvA) virus 4 days later resulted in infection in the majority of tadpoles ( Figure 1F; 322 n=25 tadpoles, 60% infected). The proportion of tadpoles with infected neurons was similar between the two 323 promoters (CMV::gal4: n=3/4 tadpoles; VGAT::gal4: n=12/21 tadpoles; p=0.63 b ). We observed a significant 324 decrease in the number of invisible TVA neurons which were infected by SADΔG-EGFP(EnvA) virus when 325 TVA and tRFP expression was amplified using the gal4-UAS system ( Figure 1G). When animals were 326 electroporated with CMV::TVA/tRFP and injected with SADΔG-EGFP(EnvA) virus 4 days later, an average of 327 85% of EGFP + neurons per animal lacked detectable tRFP expression. In contrast, electroporating tadpoles 328 with CMV::gal4 or VGAT::gal4 along with UAS::TVA and UAS::tRFP, resulted in only 28% of EGFP + neurons 329 which lacked detectable tRFP (p<0.0001 c ). In addition, these data demonstrate that different promoters can be 330 used to target infection with SADΔG-EGFP(EnvA) showing its utility to target genetically defined neuronal 331

populations. 332 333
To test whether targeting electroporation to a few cells per tectum could reduce the number of invisible TVA 334 neurons, we electroporated tectal cells sparsely using a micropipette (Bestman et al., 2006). For this 335 experiment we used CMV::TVA/tRFP because micropipette-mediated electroporation of multiple plasmids is 336 inefficient. We electroporated 3-4 sites within one tectum. Four days later, we injected animals with successful 337 targeted TVA/turboRFP electroporation with SADΔG-EGFP(EnvA) virus and imaged the animals five days 338 later. While the majority of animals were infected (77%, n=13/17 tadpoles), they still had a number of EGFP-339 only cells present ( Figure 1H). Next, we limited electroporation even further by electroporating only one site per 340 tectum with CMV::TVA/tRFP and screened for animals with a single tRFP + cell. These tadpoles were injected 341 with SADΔG-EGFP(EnvA) virus four days after electroporation. However, 4-7 days after viral injection, no 342 infected cells were detected (n=11 tadpoles). These data demonstrate that using targeted micropipette 343 electroporation did not limit viral infection to cells with detectable tRFP expression, however these data do 344 show that increasing tRFP co-expression in TVA + neurons decreased the proportion of infected EGFP-only 345 cells but did not eliminate them.

Lack of presynaptic infection may be due to insufficient glycoprotein expression 389
The most likely explanations for a lack of presynaptic infection are (1) presynaptic terminals do not contain the 390 receptor(s) necessary for viral uptake, (2) the virus is not packaged and transported appropriately, (3) 391 synapses are too weak to mediate presynaptic infection, (4) electroporated B19G is not expressed sufficiently 392 or in the correct place to coat budding virions, or (5) the virus is not released at the appropriate location. We 393 ruled out the first possibility because B19G phenotypically complemented virus infects tectal neurons, 394 demonstrating that the B19G receptor(s) exist in Xenopus tadpoles at this stage. We tested two of the other 395 possibilities to understand why we did not detect transsynaptic infection and to identify strategies for 396 improvement in the future. 397

398
There has been speculation that the extent of presynaptic spread might depend on the strength or number of 399 synapses (Callaway 2008). It is possible that the synaptic connections in young tadpoles might be too weak to 400 efficiently mediate presynaptic infection. To test this, we exposed tadpoles to a visual stimulus prior to viral 401 injection, which has previously been shown to increase synaptic strength (Aizenman and Cline 2007; Ruthazer 402 et al., 2006). We electroporated tadpoles with VGAT::gal4, UAS::TVA, UAS::tRFP, and UAS::B19G in one 403 tectal lobe and then exposed them to visual experience (VE) for either: (1) Four hours of VE 2 days prior to 404 viral injection, or (2) 12 hours of VE the night before viral injection. Similar to other experiments, 64% of 405 tadpoles (n=9/14 tadpoles) provided with visual stimulus were infected by SADΔG-EGFP(EnvA) virus. 406 However, we did not observe EGFP + cells outside of the injected tectal lobe in any of the groups (data not 12 shown) suggesting that increasing synaptic strength with these protocols is insufficient to produce presynaptic 408

transfer of virus. 409
For presynaptic infection to occur, B19G expression is required on the cell membrane so that it coats the 411 surface of budded viral particles. In addition, surface B19G expression increases the number of virions which 412 bud from infected neurons (Mebatsion et al., 1996a). To test the possibility that B19G is not expressed 413 sufficiently in transcomplemented cells, we first assessed expression of B19G in vitro. We transfected 293T 414 cells with CMV::B19G and α-actin::GFP. One day after transfection, we harvested cells and did western blots 415 on membrane fractions using anti-rabies glycoprotein antibody previously shown to detect B19G (Beier et al., 416 2019). We detected B19G in the membrane fraction of 293T cells, but no signal was found in untransfected 417 cells ( Figure 3A). To test whether B19G expression occurs in frog cells, we transfected XLK-WG Xenopus 418 kidney cells with CMV::B19G and α-actin::GFP, harvested cells one day after transfection, and did western 419 blots on membrane fractions. While there was a band of similar size to rabies glycoprotein found in both 420 transfected and untransfected cells, we observed an additional band specifically in the transfected XLK-WG 421 cells ( Figure 3B). Next, we performed immunohistochemistry for the rabies glycoprotein in CMV::B19G-422 expressing XLK-WG cells. Cells were transfected with CMV::B19G and α-actin::GFP and fixed for 423 immunohistochemistry 48 hours later. We found that B19G was expressed on the membrane of XLK-WG cells 424 using non-permeabilized immunohistochemistry conditions with anti-rabies glycoprotein antibody ( Figure 3C). 425 In contrast, XLK-WG cells transfected with α-actin::GFP alone and imaged using identical parameters had no 426 detectable staining with the anti-rabies glycoprotein antibody. Together, these results indicate that B19G can 427 be expressed on the cell membrane in Xenopus cells in vitro. 428 429 Finally, we assessed the expression of B19G in tadpoles in vivo. We electroporated CMV::B19G/tRFP and 430 examined expression of the glycoprotein using immunohistochemistry. Four days following electroporation, we 431 fixed the animals, dissected and embedded their brains, sectioned them on a vibratome, and performed 432 immunohistochemistry with anti-rabies glycoprotein antibody. Expression of B19G was either observed at low 433 levels ( Figure 3D) or not at all (data not shown). When B19G signal was present, it was observed in the 434 membrane of the apical cell body and in the proximal dendrite. By comparison, B19G expression in XLK-WG 435 cells appeared much stronger and more uniform around the cell membrane. While we cannot rule out 436 differences in antibody penetration in intact tissue compared to culture, these results are consistent with 437 insufficient expression of B19G in vivo contributing to the lack of presynaptic spread of rabies virus. 438 439

Retrograde labeling of neural circuits with recombinant rabies virus 440
In mammals, recombinant rabies virus can be used as a retrograde tracer since it infects at the axon terminal 441 and is retrogradely transported to the cell soma. We tested the utility of using recombinant rabies virus as a 442 retrograde tracer in Xenopus tadpoles. As demonstrated in Figure 1B, unilateral injections of SADΔG-443 13 EGFP(B19G) virus into the optic tectum transduced local axon terminals and yielded robust labeling at the 444 injection site. In 43% of the infected animals (n=15/35 tadpoles), we also observed retrogradely infected 445 EGFP + cells in other brain regions which project to the optic tectum ( Figure 4A). From these data we generated 446 a schematic of neurons which were retrogradely labeled by unilateral tectal injection of SADΔG-EGFP(B19G) 447 virus ( Figure 4B). We found retrograde labeling of neurons in regions known to project to the optic tectum 448 including the contralateral optic tectal lobe, hindbrain, pre-tectum and forebrain, as well as the ipsilateral 449 hindbrain. EGFP + cells were also present in very high numbers in the ipsilateral pre-tectum making them 450 difficult to count and were not included in the schematic. Retinal ganglion cells are known to be a primary 451 source of input to the optic tectum, but no EGFP was observed in the optic chiasm and the eye was not 452 examined in these experiments. It is possible that viral injections were made too deep to label the superficially 453 located retinal ganglion cell axons, or that rabies poorly infects the axon terminals of some cell types in the 454 tadpole. Nonetheless, we found robust infection in several known presynaptic areas demonstrating that rabies 455 can transport between brain regions and act as a retrograde tracer in tadpoles. in non-neuronal tissue (Dutton et al., 2009;Kawakami et al., 2006). However, tadpoles must be maintained at 515 increased temperature immediately following viral injection for infection to occur and it is not currently known 516 whether neurons in the central nervous system can be infected by adenovirus. Vaccinia is a DNA virus with 517 large packaging capacity that widely infects tadpole neurons when injected into the brain ventricle and can be 518 targeted to specific brain regions when directly injected into brain tissue (Wu et al., 1995). Vaccinia produces 519 robust transgene expression at normal rearing temperature, however transgene expression is transient, 520 decreasing over 10 days after infection. The versatility of vaccinia is limited because it cannot be restricted to 521 specific cell types using promoters. Vesicular stomatitis virus (VSV), like rabies, is an enveloped negative-

Lack of presynaptic viral spread in Xenopus 535
We were unable to detect retrograde transneuronal spread of pseudotyped rabies in Xenopus tadpoles. Rabies glycoprotein has multiple glycosylation sites and at least one of them needs to be glycosylated for 579 expression of glycoprotein on the membrane (Conzelmann et al., 1990;Dietzschold 1977;Shakin-Eshleman et 580 al., 1992). Previously, injection of rabies glycoprotein RNA into Xenopus oocytes was found to produce an 581 unglycosylated protein product (Wunner et al., 1980). Whether rabies glycoprotein is glycosylated in Xenopus 582 in vivo remains to be investigated. Expressing the glycoprotein from a different strain of rabies could be a way 583 to achieve transneuronal tracing in Xenopus. The pathogenicity of different rabies strains is determined, in suggests that application of rabies virus to study mesoscale connectomics in Xenopus will generate new 615 insights into circuit components and circuit function. For instance, we recently investigated the development 616 and function of a direct intertectal projection in tadpoles. Using rabies virus injection followed by posthoc 617 immunohistochemistry, we found that both excitatory and inhibitory tectal neurons contribute to intertectal 618 communication, which has major implications for how this neural circuit contributes to tectal function (Gambrill 619 et al., 2016). We found that injection of SADΔG-EGFP(B19G) directly into the brain ventricle resulted in 620 widespread infection near the injection site. This strategy could be exploited to infect neurons and express 621 genes of interest when retrograde tracing from a specific target is not needed. Rabies virus variants are 622 available which drive expression of many fluorescent proteins including GFP, DsRed, mCherry, and BFP 623 (Osakada et al., 2011). In principle, intersectional analysis using simultaneous injection of rabies variants 624 expressing different color fluorescent proteins into different target areas could be used to assess the 625 distribution of neurons projecting to those different targets. In addition, by identifying doubly-or triply-labeled 626 neurons, the degree to which single neurons send axon collaterals to multiple targets can be evaluated.  For each statistical test run in the study, the data structure, statistical test, p-value, and sample size are listed.