Main

If you want to know the function of a gene in development, you aim to do two different experiments. You knock the gene out ('loss of function'), and you express it ectopically ('gain of function'). You then hope that the two outcomes give you some idea of what the gene does. These two experiments can be easy or difficult depending on the organism and the system you are studying. The gain-of-function approach in the nervous system—like much else in developmental biology—took an enormous step forward with the advent of transgenic mouse technology. Mice can now be engineered to carry genes driven by tissue-specific promoters to ensure specific neural expression. Yet this method, for all its advantages, is not ideal. It is constrained by the promoters available, and it is both slow and resource intensive, requiring a new strain of mice for each new gene. Certainly, a method for introducing genes directly into the developing brain would be very useful. In this issue of Nature Neuroscience, Gaiano et al. describe what may be the technology we have been waiting for1.

The authors have elegantly combined two existing technologies—retroviral vectors and ultrasound image-guided microscopy—to introduce genes into the developing mouse brain as early as the neural fold stages. They demonstrate that the system works by introducing the alkaline phosphatase gene into the mouse forebrain in utero at embryonic day (E) 8.5 and 9.5 and showing that it is widely expressed. Then, in a proof-of-principle experiment, they use a vector encoding the sonic hedgehog (shh) gene to demonstrate that a specific phenotype can be induced in the infected neural tissue. Sonic hedgehog is a soluble factor that 'ventralizes' developing neural tissue; after exposure to shh, tissue that would otherwise become dorsal (in this case, cerebral cortex) is induced to become ventral (corpus striatum). The authors demonstrated this transformation by showing that the target tissue turned on a number of characteristically ventral genes.

Retroviral vectors have been used for years in applications as diverse as cell lineage and gene therapy2. Retroviruses are naturally evolved vehicles for gene transfer, which work efficiently and accurately under physiological conditions. Unlike adenoviruses (which are also widely used as vectors), retroviruses integrate randomly into the chromosomal DNA of the infected cell and are thereby inherited by its descendants. Wild-type retroviruses may also spread horizontally by infecting other cells, but retroviral vectors are engineered to be replication-defective, so transgene expression is restricted to the infected cells and their progeny. Moreover, retroviruses can only infect dividing cells, so neural precursor cells can be infected but postmitotic neurons cannot.

For gain-of-function experiments, it is usually important to obtain expression of the trangene in a large number of cells. The use of retroviruses in developmental biology has been limited by the low titers that can be achieved with conventional vectors, which are normally derived from Moloney leukemia virus (MLV). These viruses cannot be sufficiently concentrated to allow the injection of more than a few hundred infective particles per embryo, so only a handful of cells can be infected in these experiments. To circumvent this problem, Gaiano et al.1 have taken advantage of the recent success in pseudotyping viruses3. The coat proteins of MLV are easily lost during centrifugation, but by replacing them with coat proteins from another virus (vesicular stomatitis virus, VSV), it is possible to make viral particles that are more resistant to mechanical damage and can therefore be centrifuged to concentrations several orders of magnitude higher than for native MLV. Using this innovation, Gaiano et al. appear to have infected many thousands of neural cells per embryo ( Fig. 1 ).

Figure 1: Cross-section of spinal cord from an E12.5 mouse embryo that was infected at E8.5 with retroviral vector expressing an alkaline phosphatase reporter gene.
figure 1

The dorsal neural tube, dorsal root ganglion and floor plate show strong expression.

Full size image

Achieving high titers is not the only problem to be overcome. When we and others began using retroviral vectors for lineage studies in the mid-eighties, we injected virus into the cerebral vesicles of E14 to E16 rat embryos4. This was not difficult, because at those stages the embryo could be seen clearly through the uterine wall. However, embryos younger than E13 were almost impossible to inject; not only are they tiny, they are packed into an opaque deciduum and therefore impossible to see. We were forced to give up.

Gaiano et al. have overcome this problem by using a recently described technique known as ultrasound backscatter microscopy (UBM). In this system, ultrasound is used to visualize rodent embryos through the uterine wall and to guide an injection set-up under the control of a micromanipulator5. Gaiano et al. have used this apparatus to visualize mouse embryos as young as E8.5 and to inject concentrated virus into the neural folds ( Fig. 2 ). At this early stage, the neural tube has yet to close, so they needed only to get the innoculum within the amniotic cavity to infect the precursor cells of the neural plate. As one might expect, the procedure frequently causes exencephaly, but enough embryos survive undamaged to make the approach viable.

Figure 2: Sagittal ultrasound image of an E9.5 mouse embryo about to be injected with a high-titer retroviral stock.
figure 2

The ventricular system of the brain appears as a curved black space in the center of the image.

Full size image

Why is it important to be able to introduce genes into the early embryonic brain? Many of the important neural patterning events occur very early in development: the division of the brain into fore-, mid- and hindbrain, and the subdivision of those regions into the precursors of the major brain nuclei, all begin long before most neurons and glia are generated (see ref. 6 for review). We now have an enormous list of genes that are expressed in patterns suggesting involvement in these processes. These data, plus the morphological development of the embryonic forebrain, suggest that much of the patterning occurs between roughly E9 and E13. This new method may therefore be enormously valuable for manipulating the brain at a critical time during its development.

Although this is primarily a techniques paper, it also provides some interesting insights into the phenomenon of 'retroviral silencing'7. If undifferentiated embryonic carcinoma (EC) cells are infected with retrovirus, the provirus integrates into the host cell DNA as expected, but its expression is somehow blocked. However, if the cells are made to differentiate before infection, the virus is expressed. Somehow, the virus must be 'silenced' when it infects an undifferentiated cell, such that it cannot be reactivated even by differentiating the cell into a permissive state. There are a few exceptions, but most can be attributed to viral mutations or the effects derived from the site of insertion of the retrovirus7. How silencing works is not entirely understood, but it seems to be a two-stage process involving the viral long terminal repeat (LTR), the part of the viral genome that contains the endogenous retroviral promoter8. First the retroviral enhancer (the part of the promoter that drives tissue-specific expression) is turned off, probably because a 'stem cell factor' binds to the LTR9. Subsequently, the site is methylated, this being a well-known mechanism of inactivating genetic loci. Whatever the molecular details, once the construct is turned off, it stays off even if the cell subsequently differentiates.

We have long suspected that neural precursor cells, like EC cells, silence retroviral promoters. For instance, E9 mouse embryos fail to express retrovirally transduced genes in the neural plate, even though the same construct can be expressed in mesodermal tissues10. The answer, as Gaiano et al. demonstrate clearly, is to use a vector in which expression is not driven by the LTR, but by a separate internal promoter.

If retroviral silencing as a practical problem can be overcome so easily, then why is it so interesting? Silencing probably tells us something about how tissue-specific enhancers are controlled during the process of differentiation. During neurogenesis, the precursor cells from which neurons are derived are found in the ventricular zone (VZ), the germinal layer of the forebrain. The VZ cells (unlike postmitotic neurons) are infected by an injection of virus into the cerebral vesicle, because they are mitotic and because virus penetrates poorly into tissue, so only the VZ cells will be exposed to the virus. Yet when Gaiano et al. infected at E9.5 and analyzed five days later with a virus driven from the LTR, they found expression in postmitotic neurons but not in VZ cells. Presumably the virus remains silenced in the undifferentiated precursor cells of the VZ but is activated in neurons derived from the infected VZ cells. How do differentiating neurons escape the block? Perhaps VZ cells express the stem cell factor but do not perform the methylation step. Perhaps, methylation occurs but can be reversed in neurons. Either way, it would be surprising if this unknown control mechanism was not telling us something important about the control of differentiation during neurogenesis. It would be very interesting to know whether VZ cells express a stem cell factor similar to that expressed by EC cells.

In the meantime, though, neuroembryologists will want to evaluate this novel approach to gain-of-function studies. The most important question is whether the technique is cheap and easy enough to encourage its widespread adoption. For that we will have to wait and see.