10A recent article in Cell reported a new form of modified rabies virus that was apparently capable 11 of labeling neurons "without adverse effects on neuronal physiology and circuit function". These 12 "self-inactivating" rabies ("SiR") viruses differed from the widely-used first-generation deletion-13 mutant (∆G) rabies viruses only by the addition of a destabilization domain to the viral 14 nucleoprotein. However, we observed that the transsynaptic tracing results from that article were 15 inconsistent with the logic described in it, and we hypothesized that the viruses used were actually 16 mutants that had lost the intended modification to the nucleoprotein. We obtained samples of two 17 SiR viruses from the authors and show here that, in both "SiR-CRE" and "SiR-FLPo", the great 18 majority of viral particles were indeed mutants that had lost the intended modification and were 19 therefore just first-generation, ∆G rabies viruses. We also found that SiR-CRE killed 70% of 20 infected neurons in vivo within two weeks. We have shown elsewhere that a ∆G rabies virus 21 encoding Cre can leave a large percentage of labeled neurons alive; we presume that Ciabatti et 22 al. found such remaining neurons at long survival times and mistakenly concluded that they had 23 developed a nontoxic version of rabies virus. Here we have analyzed only the two samples that 24 were sent to MIT by Ciabatti et al., and these may not be from the same batches that were used 25 for their paper. However, 1) both of the two viruses that we analyzed had independently lost the 26 intended modification, 2) the mutations in the two samples were genetically quite distinct from 27 each other yet in both cases caused the same result: total or near-total loss of the C-terminal 28 modification, and 3) the mutations that we found in these two virus samples perfectly explain the 29 otherwise-paradoxical transsynaptic tracing results from Ciabatti et al.'s paper. We suggest that 30 the SiR strategy, or any other such attempt to attenuate a virus by addition rather than deletion, 31 is an inherently unstable approach that can easily be evaded by mutation, as it was in this case. 33 39Its core principles are 1) the selective infection of the targeted neuron group with a recombinant 40 rabies virus with one deleted gene (which in all work published to date is the "G" gene encoding 41 its envelope glycoprotein), and 2) the in vivo complementation of the deletion by expression of 42 the deleted gene in trans in the targeted starting neurons. With all its gene products present in 43 the starting cells, the virus can fully replicate within them and spreads, as wild-type rabies virus 44 does, to the cells directly presynaptic to the initially infected neurons. Unlike wild-type rabies virus, 45 however, once inside the presynaptic cells, the deletion-mutant ("∆G", denoting the deletion of G) 46 is unable to produce its glycoprotein and is therefore unable to spread beyond these secondarily 47 infected cells, resulting in labeling of just the neurons in the in...