The in utero electroporation has become a highly favored technique to ex-press recombinant genes in somatic cells in the cortex, or even other parts
of the brain. The easiest way to perform this technique (as we do) is by in-jecting the plasmids into the embryo's ventricle and applying electric pulses through electrode tweezers (Tabata, Nakajima, 2001; Saito, Nakatsuji, 2001).
There are clear advantages associated with this technique. First, it has the potential to generate large elds of labeled neurons (see Fig. 2.1 C), which facilitates in vivo imaging experiments if high background and loss of cell identity is not a concern. Second, it has a high throughput potential pro-vided that transfection eciencies and newborn survival rates are high. It should be noted though that it nevertheless takes a considerable amount of time to generate an adult mouse with suitable expression for imaging: rst, timed pregnancies should be produced, which on average takes 1 week; sec-ond, one needs to wait 3 more weeks for the delivery of the electroporated pups; third, one needs to wait until mice have reached the desired age (in our case 6 weeks) to perform the cranial window implantation; nally, one needs to wait another 10 days before the beginning of the imaging session. This ac-cumulates to a total waiting time of 11-12 weeks. This limitation is negligible if the procedure yields large numbers of mice with suitable expression.
However, the technique has also quite strong limitations and drawbacks for high-resolution long-term imaging of neuronal structures and synapses in vivo. One of the major and more general concerns is the expression of re-combinant proteins throughout development. The constructs are usually based on the chicken β-actin promoter to warrant strong and long-lasting expression. However, such an unconditional expression cassette likely starts to drive high levels of expression shortly after the cells get transfected. For
more physiologically relevant proteins it might. In our PSD-95-eGFP ex-periments we could conrm that the auxiliary expression had not heavily impacted synaptic development (Gray et al., 2006). However, when we ex-perimented using Homer1C we could observe dendritic malformations (blobs;
data not shown). The other drawbacks of the in utero electorporation tech-nique relate to the variable co-expression of uorescent proteins, the level and stability of expression, and the poor spatial targeting. The latter three drawbacks were illustrated by the pilot experiments described in the previous paragraphs: 1. In our hands, the electroporation of two plasmids yielded only small populations of co-expressing cells. This strongly reduced the through-put, as the probability of nding suitable cells under the imaging window was low. 2. Expression levels and stability was unpredictable. Since the electrical pulses were applied through relatively large electrodes, which were manually positioned over the embryos' heads, the transfection eciency var-ied between embryos. In addition, we had very little control over the number of vectors that were transfected per cell. The slight instability of expression over time may have been due to fact that the expression vectors remained episomal. Therefore, the initiation of transcription is less well regulated and probably depends on the accidental encounter of transcription factors with the DNA. The inconstancy of expression levels complicated the ratiometric analysis of two imaging channels as well as comparisons between cells and animals. 3. Since the plasmids were injected into the ventricles without de-tailed visual guidance we had very limited spatial selectivity. Depending on the embryonic stage at which the electroporation was performed we could target expression to a particular cortical layer (in our case E16 and L2/3
respectively). Other than that we had a very low control over the number of cells that were transfected as well as their location.
To change the density of transfection one could adjust plasmid concentrations and the delay between the ventricular injection and the actual electropora-tion. However, in practice the method results in either a very high or a very low density of labeling. When the density is high it is very dicult to distin-guish dierent dendrites and the cell bodies of origin. In addition, in densely labeled cortices the background (out of focus) uorescence is high. When the density is low, the probability of nding transfected cells that co-expresses the two plasmids at levels allowing 2-photon imaging under the craniotomy region is dramatically low. In order to achieve a satisfactory density of trans-fected cells, plasmids expressing the uorescent protein of interest under the control of cre-recombinase can be used. To achieve this the injection solution should contain the two oxed plasmids encoding the uorescent proteins and a third plasmid that encodes cre-recombinase. By changing the concentration of the cre expressing plasmid one can titrate the density of labeling, without compromising the degree of co-expression (Chen et al., 2012a). However, it should be noted that co-expression would be as incomplete as for regular plasmids.
To achieve a higher spatial selectivity one could use viral vectors. This allows the targeting of layers and even specic barrel columns (Marik et al., 2010).
Viral vectors would also circumvent the problem that the developmental ex-pression of synaptic proteins may present. However, in utero electroporation alike, viral vectors have limitations regarding the temporal expression prole
and co-expression prole.