Promoters and serotypes: targeting of adeno-associated virus vectors for gene transfer in the rat central nervous system in vitro and in vivo


Corresponding author S. Kügler: Department of Neurology, Vector Laboratory, Waldweg 33, 37073 Göttingen, Germany. Email:


The brain parenchyma consists of several different cell types, such as neurones, astrocytes, microglia, oligodendroglia and epithelial cells, which are morphologically and functionally intermingled in highly complex three-dimensional structures. These different cell types are also present in cultures of brain cells prepared to serve as model systems of CNS physiology. Gene transfer, either in a therapeutic attempt or in basic research, is a fascinating and promising tool to manipulate both the complex physiology of the brain and that of isolated neuronal cells. Viral vectors based on the parvovirus, adeno-associated virus (AAV), have emerged as powerful transgene delivery vehicles. Here we describe highly efficient targeting of AAV vectors to either neurones or astrocytes in cultured primary brain cell cultures. We also show that transcriptional targeting can be achieved by the use of small promoters, significantly boosting the transgene capacity of the recombinant viral genome. However, we also demonstrate that successful targeting of a vector in vitro does not necessarily imply that the same targeting works in the adult brain. Cross-packaging the AAV-2 genome in capsids of other serotypes adds additional benefits to this vector system. In the brain, the serotype-5 capsid allows for drastically increased spread of the recombinant vector as compared to the serotype-2 capsid. Finally, we emphasize the optimal targeting approach, in which the natural tropism of a vector for a specific cell type is employed. Taken together, these data demonstrate the flexibility which AAV-based vector systems offer in physiological research.


Several recent advances have significantly enhanced the usability of adeno-associated virus (AAV)-based vectors as tool in physiological research (Flotte, 2004): (i) modern production methods easily allow for the generation of highly pure and concentrated vector stocks (Zolotukhin et al. 1999; Potter et al. 2002; Grimm et al. 2003); (ii) eight different serotypes have been isolated so far, which considerably broadens the tropism range of recombinant viruses (Davidson et al. 2000; Rabinowitz et al. 2002; Gao et al. 2002; Grimm et al. 2003); (iii) transcriptional control elements have been functionally tested and found to enhance transgene capacity and expression rates (Paterna et al. 2000; Glover et al. 2002; Hermening et al. 2004); and (iv) size restrictions caused by the small genome size of 5 kb have been addressed by efficient double transduction and in vivo vector-genome ligation techniques (Reich et al. 2003).

Vector genomes

Limitations in transgene capacity due to the relatively small genome size of 5 kb have been a persistent blemish on AAV-based vectors. Cell-type specific promoters tend to be bulky, and thus occupy considerable amounts of the viral genome. However, transcriptional targeting of viral vectors to neurones by using a small fragment of the human synapsin-1 gene promoter has recently been shown to work very efficiently both in adenoviral and AAV vectors (Kügler et al. 2003a,b). In combination with short elements to enhance transcriptional levels (Hermening et al. 2004) a total transgene capacity of 3.9 kb arose, which is sufficient for the expression of a roughly 140 kDa-sized protein (without post-translational modifications). Figure 1A compiles some of the most frequently used transcriptional control elements in AAV vectors (Addison et al. 1997; Paterna et al. 2000; Fitzsimons et al. 2002; Klein et al. 2002). In Fig. 1B, vector genomes as used in our research are schematically depicted: here, small trancriptional control elements allowed for the construction of bi-cistronic vectors, which express not only a functional transgene but also a fluorescent reporter gene and therefore demarkate successfully transduced cells.

Figure 1.

Different combinations of promoters and 3′-transcriptional control elements used for CNS gene transfer in recombinant AAV vector genomes
A, short promoters significantly enhance transgene capacity of recombinant AAV vector genomes. The dotted lines represent the transgene capacity of the respective vector genomes and denote the number of base pairs which can be inserted in between the transcriptional control elements. Arrows, promoters; elipsoid, intron sequences; rectangles, polyadenylation sites; pentagon, woodchuck hepatitis virus post-transcriptional control element (WPRE); hSYN1, human synapsin-1 gene promoter; mCMV, murine cytomegalovirus immediate early promoter; NSE, neurone specific enolase gene promoter; CBA, chicken beta-actin/CMV hybrid promoter; PDGF, platelet derived growth factor gene promoter; SV40, simian virus 40 derived polyadenylation site; bGH, bovine growth hormone derived polyadenylation site. B, bi-cistronic vector genomes used in current work in our laboratory. Vectors used for the expression of a functional transgene also carry a second, independent expression cassette for green fluorescent protein (EGFP) or red fluorescent protein (DsRed-2). Functional transgenes expressed from these bi-cistronic vectors may be up to 1500 bp in size. ITR, inverted terminal repeats of AAV-2; Int, 146 bp chimeric intron.

In vitro applications

Cultured brain cells may serve as excellent model systems for the investigation of basic neuronal physiology, for example by electrophysiological recordings from ion channels, investigations of Ca2+ homeostasis, cytoskeletal de- and re-polymerizations during differentiation, neurone–glia interactions, etc. In order to dissect the functional roles of proteins in either neuronal or glial cells, it would be of value to be able to overexpress the protein under investigation specifically in the respective cell type. As shown in Fig. 2, different combinations of AAV serotypes and promoters allow for targeting transgene expression either exclusively to neurones or to astrocytes (as identified by co-staining with either the neurone-specific marker NeuN or the astrocyte-specific marker GFAP (Kügler et al. 2003b)). The onset of transgene expression after AAV vector transduction is often claimed to be ‘slow’ due to the necessity of conversion of the single stranded genome into double stranded DNA, especially in non-dividing cells. However, when primary hippocampal cultures were infected by AAV-2-hSYN vectors at the time of seeding, readily detectable transgene expression emerged in neurones at 2–3 days after transduction, reaching almost 100% transduction efficacy at 5–7 days after transduction (Fig. 2A). No proteasomal inhibitors were used in these experiments. Comparison of the expression kinetics of the short hSYN promoter (0.48 kbp) with the widely used 1.68 kbp CBA hybrid promoter, which is claimed to be a ‘powerful’ promoter (Fitzsimons et al. 2002), revealed that the hSYN promoter expressed at least equal, or even higher, amounts of the transgene (Fig. 2C). Both expression cassettes contained identical 3′-control regions (WPRE and bGH, see Fig. 1A).

Figure 2.

Neurone- or astrocyte-specific transduction of primary neuronal cultures
AAV vectors expressing EGFP from different promoters, packaged into either AAV-2 or AAV-5 capsids, were used to transduce primary hippocampal neurones at times of preparation (DIV 0). Transduction of astrocytes was performed at days in vitro (DIV) 7 of the same culture. EGFP fluorescence of living cultures was recorded with an inverse fluorescent microscope at either day 3 or day 8 after transduction for neurones, or at day 1 or day 3 after transduction for astrocytes. AD, show high power magnifications of cultures transduced with an AAV-2 vector expressing EGFP from the hSYN promoter (A), (an AAV-5 vector expressing EGFP from the hSYN promoter (B), an AAV-2 vector expressing EGFP from the chicken β-actin/CMV hybrid promoter (C) and an AAV-5 vector expressing EGFP from the mCMV promoter (D). Low power micrographs showing quantitative transduction of either neurones or astrocytes derived from the same culture are shown in (E) and (F).

Transgene expression was also completely restricted to neurones if the AAV-2 genome containing the expression cassettes was cross-packaged into the AAV-5 capsid (Grimm et al. 2003). However, expression levels were substantially reduced (Fig. 2B), indicating that receptors for AAV-5 capsids platelet-derived growth factor receptor (PDGF-R) and sialic acids (Walters et al. 2001; Di Pasquale et al. 2003)) are less abundant on primary hippocampal neurones as compared to the receptors necessary for AAV-2 entry (heparan sulphate proteoglycan moieties (Opie et al. 2003)). It is intriguing that replacing the hSYN promoter by the mCMV promoter drastically changed the transgene expression pattern: rapid onset of transgene expression is detected in fibrous astrocytes (Fig. 2D) and remained confined to this cell type throughout the lifetime of the culture. Thus, AAV-5 vectors can transduce both neurones and astrocytes in primary cultures, and can be transcriptionally targeted to either one of these cell types very efficiently.

In order to achive quantitative transduction of cultured neurones (90–95% transduction efficacy), cells are grown on coverslips in 24-well plates. Transduction is performed at 12 h after seeding, in a culture volume of 250 µl. In each well, 1.5–3 × 109 vector genomes are used as determined by quantitative PCR. This corresponds to 0.5–1 × 108 transducing vector units, as the ratio of infective versus total viral particles routinely obtained is 1 : 30. After overnight incubation, culture medium is filled up to 750 µl. This procedure allows for the generation of ‘transgenic’ cultures of primary brain cells (Fig. 2E and F), which are routinely maintained for about 4 weeks without any signs of cytotoxicity (Kügler et al. 2003b).

In vivo applications

As shown above, AAV vectors packaged into the serotype-5 capsid are relatively inefficient in transducing neurones in primary hippocampal cultures but are excellently suited for astrocyte-specific transgene expression. As the production of neurotrophins may be more effective if secreted from astrocytes in the brain (e.g. in model systems of neuronal degeneration and regeneration) we tested the AAV-5-mCMV vector for its transduction properties in the adult rat brain. In pronounced contrast to the results obtained in cultured cells prepared from the embryonic rat brain, we detected striatal transgene expression mainly in cells which neither co-localized with the astrocyte-specific GFAP marker, nor with the neurone-specific NeuN marker (Fig. 3A and B). The predominant location of the transduced cells inside striosomes (fibre bundles crossing the basal ganglia) suggests that they may be oligodendrocytes, an interesting property still under further investigation.

Figure 3.

AAV-5 vector targeting in the brain
The attempt to target transgene expression from an AAV-5 vector to astrocytes in the rat brain is shown in (A) and (B). EGFP expression driven by the mCMV promoter is shown as an overlay with the astrocyte specific anti-GFAP stain (red; A) or the neurone-specific anti-NeuN stain (red; B). Note that EGFP expression (green) only very rarely co-localized (yellow) to these markers. However, AAV-5 vectors expressing EGFP from the hSYN promoter very efficiently expressed EGFP in neurones distributed all over the striatum (C). This expression was entirely neurone-specific as demonstrated by the co-localization of EGFP expression with the neuronal marker NeuN (EG). An AAV-2 vector used at the same titre showed transgene expression closely confined to the injection tract (arrows). V, ventricle; CC, corpus callosum; Cx, cortex; Cpu, caudate putamen (striatum).

Although the combination of mCMV promoter and serotype-5 capsid failed to do the intended task in the brain, another promoter–serotype combination offered very promising features: while AAV vectors packaged into the most widely used serotype-2 capsid show transduction of brain tissue fairly restricted to areas near the injection site (Fig. 3D), we obtained a much larger transduction area if the serotype-5 capsid is used (Fig. 3C). In combination with the neurone-specific hSYN promoter, transgene expression was found exclusively in neurones (Fig. 3EG) spread all over the striatum. Quantification of transduced neurones versus total neurone number (as revealed by NeuN staining) showed that in a volume of roughly 15 mm3 about 40% of the striatal neurones expressed the transgene EGFP after a single injection of 2 µl containing 1 × 108 transducing units (3 × 109 genomes). While the AAV-5-mCMV vector demonstrated that AAV-5 is not a neurone-specific vector per se, targeting the vector to neurones by the hSYN promoter allowed for efficient and widespread neuronal transduction.

One major topic of the symposium this paper is about was the problem of ‘finding the right vector for the job’. As shown above, targeting a vector by transcriptional control or by utilization of a different capsid (which will bind to different cellular receptors) may be an efficient means to tell the recombinant virus where to go. However, the optimal ‘targeting’ should make use of ‘matching pairs’, consisting of a specific cell type in a complex tissue and a viral vector, the tropism of which is specific for this particular cell type. A good example in which this condition is fulfilled is the specific transduction of retinal ganglion cells (RGCs) by AAV-2 vectors. RGCs project through the optic nerve towards the tectum. The transection of the optic nerve is an easy to perform and highly reproducible model system of neurodegeneration and regeneration (Kügler et al. 1999; Heiduschka & Thanos, 2000; Weishaupt & Bähr, 2001). RGCs thus have been the target in many gene transfer studies performed in our laboratory in order to define proteins efficiently counteracting the neurodegenerative process. Initial studies have been accomplished by the use of adenoviral vectors, which, however, can transduce RGCs only in a retrograde application approach (Kügler et al. 1999, 2000). Although this resulted in specific transduction of RGCs (which may be considered as a case of ‘morphological’ targeting), transduction was relatively inefficient, making it tedious to quantify protective effects. A much more elegant approach emerged by the use of AAV-2-based vectors. Fortunately, the primary receptor of AAV-2 (heparan sulphate proteoglycans) appeared to be highly expressed on RGCs. Thus, the intravitreal injection of AAV-2 vectors into the eye resulted in a highly selective transduction of RGCs; less than 10% of transduced cells were not co-labelled with the RGC-specific retrograde tracer FluoroGold (Fig. 4A–C). These preferential transduction properties of AAV-2 are now being used in our laboratory and by others to evaluate the potency of neuroprotective factors and neuroregenerative strategies (Cheng et al. 2002; Fischer et al. 2004; Malik et al. 2004). However, even the best system may be destroyed by use of inappropriate subcomponents: if the hSYN or CBA promoter is replaced by the mCMV promoter in such an AAV-2 vector, then transgene expression in RGCs is virtually completely abolished, and only a few glial cells are found to express the reporter gene EGFP (Fig. 4E and F).

Figure 4.

Targeting transgene expression to a specific neuronal population in the retina
In cross-sections of retinal tissue specific transduction of retinal ganglion cells (RGCs) by an AAV-2 vector with transgene expression driven by the hSYN promoter is shown (AF). Retrograde labelling of RGCs by the fluorescent tracer FluoroGold (FG) is shown in (A). RGCs make up only 1% of all cells present in the retina, which are labelled by the nuclear DAPI stain in (D). After intravitreal vector injection, transgene expression (B) is highly confined to RGCs as demonstrated by EGFP–FG co-localization (C). This cell-type specific transgene expression can be completely annihilated if in the AAV-2 vector the mCMV promoter is used: only a low number of transduced Müller glial cells is detected in this case (E and F). Double transduced RGCs are shown in flat mounted retinae (the observer looks onto the ganglion cell layer, the other layers of the retina are located beneath the surface of the specimen (GM)). Co-injection of the two vectors expressing either EGFP or DsRed resulted in expression of both, EGFP and DsRed, in the majority of RGCs (low power micrographs in GJ, high power micrographs in KM). GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; PRL, photoreceptor layer; PE, pigment epithelium.

As co-expression of several interacting proteins may be desirable (e.g. in inducible vector systems, or in FRET/FLIM studies on protein–protein interactions, etc.), AAV genome size may constitute restrictions even if small transcriptional control units are used. Therefore, we demonstrate that this problem may readily be overcome by the use of a double transduction approach: after intravitreal injection of two different AAV-2 vectors, one expressing EGFP, the other DsRed (1 × 107 transducing units/3 × 108 genomes each) the majority of RGCs was found to express both transgenes (Fig. 4GM).

In conclusion, this short review demonstrates the high flexibility that AAV-based vector systems are offering, making them very useful tools in physiological research. It should be stressed that selecting the right vector for the job (i.e. an appropriate combination of promoters and serotypes) is still an important issue to take into account.



The work presented here was funded by Deutsche For schungsgemeinschaft through the DFG-Research Center ‘Molecular Physiology of the Brain’. We thank J. Kleinschmidt, Deutsches Krebsforschungs-zentrum, Heidelberg, for the serotype-specific packaging plasmids.