Transgenic models for molecular and physiological studies in the central nervous system

The near completion of the sequencing of the human and mouse genomes, and those of a number of other eukaryotic organisms, as well as easy access to these sequences via various databases, has focused increased attention on the need for model systems to determine the physiological significance of these sequences. The diverse cellular phenotypes found in the mammalian central nervous system (CNS) and their complex connections and interactions make this effort a daunting task often requiring the development of novel mammalian models in which mutations of specific genes are selectively expressed in the CNS in space and time.

Transgenic and homologous recombination approaches have been important for studies of gene function (Herrup, 1995). An elegant example of this approach is a study that used two transgenic models and viral vectors to identify sites in the the caudate-putamen that use dopamine in the control of feeding behaviour (Szczypka et al. 2001). An embryonic lethal tyrosine hydroxylase (TH) knockout mouse was rescued by selective expression of a TH transgene targeted only to neurons that make noradrenaline (norepinephrine). The resulting dopamine (and not noradrenaline)-deficient mouse models exhibit hyperactivity and hypophagia, and the authors then used these models to examine the behavioural effects of microinjections of mixed adeno-associated viral vectors expressing enzymes that produce dopamine at selected CNS sites.

While such transgenic approaches have been very valuable for the study of gene function, they also have significant limitations, such as position effects when the relatively small transgenes (usually < 10 kb in size) integrate in the genome at positions that have either strong enhancers (causing ectopic expression) or repressors (preventing expression). Another limitation is that the small inserts used for transgenesis often lack distant cis-elements that are critical for directing regional cell-specific gene expression. This can be overcome by using large genomic clones such as cosmids (30-50 kb inserts), bacterial artificial chromosomes (BACs, 75-300 kb), or yeast artificial chromosomes (YACs, up to 1 Mb) as transgenes. Since the average mouse gene is about 30 kb in size, it has been argued that a BAC, which also has high transformation efficiency, would be an ideal vector to use for transgenic studies (Heintz, 2000). Some advantages of BACs are that: (1) expression is usually independent of the insertion site, (2) expression is gene dose dependent, so that level of expression is well correlated with copy numbers, (3) homologous recombination in E. coli permits the precise modification of sequences (from single nucleotide changes to kilobase deletions or insertions of several kilobases of marker genes), and (4) use of internal ribosomal entry sites (IRES) allows the production of polycistronic mRNAs from which several proteins can be expressed under the control of the endogenous transcriptional regulatory sequences. The latter would allow for the simultaneous expression of marker proteins and various perturbant protein mRNAs and proteins in the same cell that expresses the endogenous gene, offering a powerful tool for functional analysis.

The oxytocin (OT) and vasopressin (VP)-synthesizing magnocellular neurons (MCNs) of the mammalian hypothalamo-neurohypophysial system (HNS) have been extensively studied with respect to the regulation of gene expression and secretion of their neuropeptides (Burbach et al. 2001). Because of the absence of homologous cell lines, transgenic models have been used to elucidate the identity of the cis-regulatory elements in the genes that are responsible for their cell-specific expression (reviewed in Gainer & Young, 2001; Murphy & Wells, 2003). Conventional transgenic studies, using relatively short transgenes, have succeeded in targeting reporters selectively to the OT or VP MCNs (Young et al. 1999; Jeong et al. 2001; Davies et al. 2003). However, in none of these studies has expression been observed in the VP-expressing parvocellular neuronal populations in the hypothalamus. In this issue of The Journal of Physiology, Wells et al. (2003) used a VP transgene containing a human growth hormone (hGH) reporter in a cosmid to determine whether the longer inserts would exhibit hGH expression in MCNs and parvocellular neurons. Of the two founder lines that were analysed, one showed modest expression in parvocellular neurons in the suprachiasmatic nucleus (SCN), amygdala and some other areas, whereas the second line did not. The absence of parvocellular neuronal expression has been interpreted as being due to the absence of a relevant enhancer element in the transgene (Murphy & Wells, 2003). However, another possibility is that the efficiencies of the transgenic constructs are very low in comparison to the endogenous gene, which might make it difficult to detect expression of the transgenes in the poorly expressing parvocellular neurons. In this regard, it is important to note that much shorter VP constructs which are present at high copy numbers can produce expression in SCN VP neurons in vitro (Gainer et al. 2001).

In principle, the use of longer genomic clones as transgenic constructs should provide the most reliable and robust targeting of reporters for functional studies. In fact, there is little doubt that use of a BAC construct, or a ‘knock-in’ containing any desired reporter genes or sequences of interest would be the best strategy for transgenic targeting to specific neurons in the CNS.