Nat Methods. 2007 May;4(5):379; author reply 379.
Viral gene delivery is a powerful tool that is rapidly gaining popularity in experimental neuroscience. I believe that the near future will see rapid expansion of this remarkably versatile and adaptable method. Recent communication by Wickersham et al.1 describes yet another interesting implementation of this technology for retrograde tracing of neurons from their target areas. The key point of the paper is demonstration of a high level of enhanced GFP (EGFP) expression using a rabies virus mutant with glycoprotein gene substituted by EGFP. After injecting the vector into the thalamus, the authors showed many bright EGFP-expressing cortical pyramidal neurons. As a reference, they used a lentivirus with rabies virus envelope glycoprotein in which EGFP expression was driven by human cytomegalovirus (hCMV) promoter and found a very low level of EGFP expression with it.
I would like to draw the attention of Nature Methods’ readers to the issue of suitability of hCMV-driven constructs for targeting brain neurons. According to the experience of our laboratory, hCMV expression in brain is only stable in some neuronal phenotypes, for example, cholinergic neurons such as those found in the nucleus of the hypoglossal nerve or vagal motor neurons. Other neurons, for example, in the adjacent nucleus of the solitary tract (which are glutamate, GABA and noradrenergic neurons) yield extremely poor expression with hCMV-containing adeno- and lentiviruses2. Hence, the assumption that hCMV-driven viral constructs are guaranteed to express in all neurons is incorrect. Such constructs are, however, very active in glia, sometimes creating a confusing impression of high level of expression (see Supplementary Fig. 1 online). A striking example of a lack of hCMV activity in medullary neurons was recently given3. Therefore, for the comparison to be fair, it is critical that stable and high-level EGFP expression in pyramidal neurons be demonstrated using the same lentivirus (presumably this would require using viruses coated with vesicular stomatitis glycoprotein rather than rabies virus glycoprotein). Incidentally, we also had demonstrated bright unstained EGFP-expressing noradrenergic neurons targeted using conventional adenoviral vectors injected into spinal cord2. Such cells are easily identifiable in acute slices prepared from pretransfected animals and can be studied using patch clamp. Curiously, it seems that the phenotype of neurons affects their suitability for retrograde transduction, as we never succeeded in achieving the same with glutamatergic neurons, such as cortical pyramidals, which were the target in the study by Wickersham et al.1. Finally, it is likely that the expression profile of hCMV-containing vectors based on adeno-associated virus may obey different rules owing to the effects of the viral inverted terminal repeats4.