Quantitative comparison of expression with adeno-associated virus (AAV-2) brain-specific gene cassettes

Article metrics


This study compared a range of mammalian CNS expression cassettes in recombinant adeno-associated virus (AAV-2) vectors using strong endogenous promoter sequences, with or without a strong post-regulatory element and polyadenylation signal. Changes in these elements led to transgene expression varying by over three orders of magnitude. In experiments conducted in primary cell culture and in >100 stereotactically injected rats, we observed highly efficient and stable (>15 months) gene expression in neurons and limited expression in glia; the highest expression occurred with endogenous, nonviral promoters such as neuron-specific enolase and β-actin. The packaging size of AAV-2 was maximized at 5.7 kb without impairing gene expression, as judged by direct comparison with a number of smaller AAV-2 constructs. The genomic insert size and titer were confirmed by Southern blot and quantitative PCR, and infectivity was tested by particle titer using ELISA with a conformation-dependent epitope that requires the full intact capsid. A packaging and purification protocol we describe allows for high-titer, high-capacity AAV-2 vectors that can transduce over 2 × 105 neurons in vivo per microliter of vector, using the strongest expression cassette.


One of the challenges facing gene transfer to the central nervous system (CNS) has been the development of vectors with appropriate cell tropism, efficient transduction, and stable levels of expression. The main approach for optimizing vectors such as adeno-associated virus (AAV-2) has traditionally been at the nucleic acid level, involving promoter and other gene regulatory elements, although recent work on viral subtypes or pseudotypes suggests that capsid modifications may be equally important in increasing expression or modifying the tropism of recombinant virus.1,2,3,4 To date, most studies on brain gene transfer have focused on vectors carrying viral promoters,5,6,7,8,9,10 although some have looked at mammalian cell type-specific promoters.11,12,13,14,15,16,17 In this study, we compared the strength and stability of expression, as well as cellular specificity, of AAV-2 vectors expressing reporter genes under the control of constitutive, viral, or other cellular promoters.

Despite its importance for gene transfer in the brain, a quantitative comparison of constitutive and neural promoters has not been published, and there is still some controversy regarding the ideal promoters to direct strong and stable gene expression in the mammalian CNS. For example, two semi-quantitative studies14,15 found that AAV-mediated transduction and expression of a reporter gene in the brain with a neuron-specific enolase (NSE) promoter was more efficient and persistent than a construct containing either a viral (CMV) or another cellular promoter (PDGF). The use of an identical NSE promoter in a herpes simplex virus (HSV) vector resulted in much lower expression than that of a HSV immediate–early promoter,12,18 and one study looking at AAV-mediated transduction of neurons in culture had reported that a CMV promoter gave significantly greater expression than a strong cellular promoter, β-actin.13 Moreover an in vivo study involving plasmid DNA under control of different cellular (NSE, GFAP) or viral (CMV) promoters had suggested that GFAP and CMV were both superior to NSE in promoting the expression of a reporter gene in the brain.19 In light of these conflicting results, we sought to test a range of promoters and regulatory elements in AAV-2.

Previous studies have shown that AAV-2 transduces post-mitotic cells, and that neurons are particularly susceptible.6,20 Yet the numbers of transduced cells and the level and stability of expression have been limiting factors for gene transfer.21,22,23 Here, we tested the hypothesis that high levels of transgene expression can be obtained in a variety of brain regions using AAV-2 and that regardless of the regions targeted, strong mammalian promoters are superior to viral promoters such as CMV. Given that expression is related in part to viral capsid proteins and host cell-surface factors24,25,26,27,28,29 that promote internalization, the role for ‘promoter targeting’ with AAV-2 has been de-emphasized, because AAV-2 preferentially transduces neurons independently of promoter effects.30 However, other work supports a role for ‘promoter targeting’ with AAV-2; for example, the myelin basic protein (MBP) promoter has been reported to drive expression in oligodendrocytes using AAV-2.16 Therefore, in addition to testing promoter effects on gene expression in neurons, we wished to determine if expression would occur in glia using a strong glial promoter, despite the clear tendency of AAV-2 to bind to neurons. Once promoter effects were tested, we reasoned that it would be possible to modify the AAV-2 capsid to increase expression in glia, for instance using the pseudotyped AAV-2/5 capsid which shows increased glial uptake (unpublished data). The promoter elements chosen for this study were representative of the strongest promoters described for gene expression in the CNS. We were especially interested in studying promoters in AAV-2 that had been previously looked at in other viral vectors such as HSV, and applying quantitative methods to previous work that had looked at a more limited range of promoters in AAV-2.

The main goal of this study was to look at total reporter gene expression within four brain regions (cortex, hippocampus, striatum, nigra) using promoters that are known to have substantial activity in vivo. We did not describe the phenotype and spatial distribution of all the cells infected. Every subtype of neuronal cell within a brain region may not naturally drive expression from all of these promoter elements and thus differences in expression in vivo are due in part to the total numbers of distinct neuronal populations within the region targeted. In particular, promoters such as hENK or nAChR may not ordinarily drive expression in the majority of cells within these brain regions, especially compared with a pan-neuronal promoter such as NSE. Nevertheless a number of non-mammalian promoters (eg avian β-actin) are known to drive high levels of gene expression in human cells in which they do not ordinarily exist and thus we considered that neurons might express transgenes from a promoter element that is not ordinarily active in a given transduced neuronal subtype, perhaps because other negative regulatory elements are present in the endogenous promoter.

For this study we generated AAV-2 vectors capable of accepting a variety of transgenes in a versatile expression cassette. We introduced reporter (rather than physiologic) genes, although the high levels of expression we report should apply to any gene in the proper size range. The AAV-2 vector is devoid of viral DNA except for two 145 base pair inverted terminal repeats (ITRs). One advantage of AAV-2 virus for in vivo functional genomic studies is its lack of toxicity,31 which is related to this paucity of viral elements; infection with AAV-2 in the brain is not known to elicit a systemic immune response, nor is there inflammation or injury in brain regions where the vector has been injected.5,7,9,16,31 Despite the safety advantage related to minimal native AAV elements, the small packaging capacity of AAV vectors has limited gene transfer applications. Previous studies suggested that up to 6.0 kb can be packaged into virus, but these ‘over-sized’ viruses were not infectious,32 hence the common assumption is that the packaged genome must be limited to just 110% of wild-type AAV or no greater than 5.1 kb.33 This limited DNA insert size restricts the range of open reading frames which can be inserted for any given promoter and post-regulatory element and polyadenylation signal.34

A secondary goal was to test the outer limits of AAV-2 genome size, while maintaining packaging and infectious efficiency. In contrast to earlier studies which inserted non-essential ‘stuffer’ sequences into the vector, we used regulatory elements which increased the size of the expression cassette but still provided for high levels of expression. For the largest construct, a luciferase reporter gene was combined with a preproenkephalin (PPE) promoter and the woodchuck hepatitis B virus post-transcriptional regulatory element, a hepadnavirus post-regulatory element (PRE) known to be important for high-level expression of native mRNA transcripts, acting to enhance mRNA processing and transport of intronless genes.35,36 The bovine growth hormone polyadenylation (BGH-polyA) sequence, reported to drive high levels of expression,37,38 was also incorporated to test its effects on expression. Although it is unclear why this virus was able to accomodate a 5.7 kb insert, some transgenes are thought to adopt unique conformations and may package differently. It is interesting that other sequences we attempted to package with the NSE promoter would not package even at considerably smaller insert sizes (<4.8 kb), suggesting that the capsid may interact in a sequence-specific manner with the packaged DNA, and that the PPE promoter may be unique in some way. This interpretation would be consistent with reports that 5.3 kb is the upper limit of AAV-2 packaging size for most genes.


Counting the number of cells expressing a reporter gene has often been used to estimate transduction efficiency,5,15,39 yet transduced cells can show a variability in expression levels and immunocytochemical protein detection methods can also be variable, making it difficult to distinguish cells with low expression levels from cells not expressing the transgene. To more precisely measure expression levels and eliminate observer bias, we assayed gene expression by quantitative enzyme activity assay. To confirm the biochemical analysis, we observed expression of luciferase and green fluorescent protein reporter genes with immunocytochemistry. Using the luciferase reporter gene in primary cell culture or intact rat brain, we found that different promoter fragments led to gene expression ranging over two orders of magnitude, in the setting of an otherwise identical AAV expression cassette (Figure 1a). In selected constructs (Figure 1b), the post-regulatory element (WPRE) boosted expression by a further order of magnitude. For instance, the WPRE in the 5.7 kb preproenkephalin (PPE) construct increased expression six-fold in both cortical and striatal cultures and three-fold in nigral cultures compared with the construct lacking WPRE. Higher gene expression was also found in vivo, with a 13-fold difference in striatum and 35-fold difference in hippocampus 2 weeks after injection. Based on previous data showing that AAV expression increased most rapidly in vivo by 2 weeks following injection of vector, while in vitro expression generally peaked after 7 days, all primary cultures were analyzed at 10 days after infection and rat brains were collected for analysis at time intervals ranging from 2 weeks to 15 months after infection.

Figure 1

(a) Schematic diagram of rAAV luciferase constructs with alternative choices of promotor. These constructs contain ITRs and polyA sequences but do not contain the WPRE element, ranging in size from 2.65 to 5.05 kb. (b) Schematic diagram of maximally sized rAAV luciferase constructs with alternative choices of promoter. These constructs contain the ITRs, polyA sequences, and WPRE element, ranging in size from 4.8 to 5.7 kb. (c) Western blot of viral capsid proteins, illustrating that all viral structural proteins (VP-1,2,3) are present in equal ratios in the different viruses assayed. Lane 1, rAAV-luc with GFAP promoter; lane 2, rAAV-luc with GFAP promoter and WPRE; lane 3, rAAV-luc with PPE promoter; lane 4, rAAV-luc with PPE promoter and WPRE. (d) Southern blot of vector DNA isolated from rAAV virions. Lane 1, size standard (kb units); lane 2, rAAV-luc with PPE promoter and WPRE element (5.7 kb); lane 3, rAAV-luc with PPE promoter (5.0 kb); lane 4, rAAV-luc with GFAP promoter and WPRE (5.25 kb); lane 5, rAAV-luc with GFAP promoter (4.6 kb); lane 6, negative control. This blot suggests that full-length DNA was packaged into the respective virions.

Western blots indicate that all three AAV structural proteins (VP-1,2,3) which form the viral capsids were present (Figure 1c). We also confirmed the packaging size of our constructs (Figure 1d) with Southern blots, which indicate that the largest viral genome packaged was 5.7 kb. We cannot completely rule out the possibility that in addition to fully packaged vector, some partially deleted genomes were also packaged – however, there were no minor bands visible on Southern blots, which might indicate trace contaminants, and moreover the vector stock with the largest gene cassette provided high levels of expression, which would not be the case with non-packaged virus.

The 12 constructs tested in primary culture revealed a wide range of luciferase activity (Table 1). Levels of luciferase expression were uniformly high in all brain regions using the NSE promoter. In descending order, the promoter activities in vitro were: NSE 1.8 kb > EF > NSE 0.3 kb ≈ GFAP > CMV > hENK > PPE ≈ NFL ≈ NFH > nAchR. Consistent with these results, immunohistochemical analysis of striatal primary culture cells demonstrated that luciferase-immunoreactive (luc-IR) cells transduced using NSE, EF or GFAP promoters had more intense staining than cells transduced using other promoter elements that we tested (Figure 2b). The number of luc-IR cells counted for a subset of promoters indicated that the relative promoter strengths in vitro were: NSE 1.8 kb > EF > GFAP > PPE > hENK. The most intense staining was observed with full-length NSE-luc-WPRE, followed by full-length NSE promoter without WPRE, and truncated NSE 0.3 kb (Figure 2A). Nearly all the striatal cells transduced using the NSE 1.8 kb promoter had multipolar neuronal morphology. The long fibers were either radial processes near the cell somata or longitudinal processes which passed parallel along the long axis of the nerve fibers. No luc-IR cells were detected in mock-infected controls (cf. Table 1; Figure 2a). Of note, we subsequently tested another pan-neuronal promoter (avian β-actin 0.6 kb promoter/enhancer) in the same expression cassette as 1.8 kb NSE, and obtained expression levels that were comparable or slightly greater than those obtained with the 1.8 kb NSE element (unpublished results).

Table 1 Luciferase activity in rAAV-transduced primary culture (units = pg/106 cells)
Figure 2

(A) In vitro immunostaining to luciferase in striatal primary neurons, showing the effect of the WPRE element on the intensity of staining. The boxes are (a) control/mock infected (b) NSE 0.3 kb (c) NSE 1.8 kb (d) NSE 1.8 kb + WPRE element. (B) In vitro immunostaining to luciferase in striatal primary neurons, showing the differences in expression in identical cells using alternate promoters, with the same viral genomic titer for all samples.

Luciferase expression was also tested by injection to the rat brain. At 2 weeks to 3 months after injection, 1 μl of each virus produced strong transgene expression in all injection sites (Table 2). The NSE 1.8 kb promoter showed higher luciferase activity than any other promoter, with the rank of promoter activities being NSE 1.8 kb + WPRE > NSE 1.8 kb > EF > GFAP > NSE 0.3 kb > PPE > CMV > hENK > nAchR > NFL > NFH in the striatum and NSE 1.8 kb + WPRE > NSE 1.8 kb > EF > GFAP > NSE 0.3 kb > PPE > hENK > CMV > NFL ≈ nAchR > NFH in the hippocampus. No luciferase activity was found in control injected animals (Figure 3f). For the PPE-luc-WPRE and NSE-luc-WPRE vectors, gene expression reached peak levels at 1 month after injection and remained high for the 3 months examined, while for NSE without WPRE, expression decreased at 3 months to only 40% of the peak level at 1 month. A subset of animals (n = 4) injected with the NSE 1.8 kb luc-WPRE cassette was maintained for 15 months, at which time luciferase activity was greater than or equal to levels at 1 month (Table 2). The striatal and hippocampal cells transduced using the NSE promoter had multipolar neuron morphology, and double-staining with NeuN and MAP-2 supported their identity as neurons. The luciferase completely filled the cytoplasm and neuronal processes down to the distal axon terminals, suggesting that the strength of expression on a cell-by-cell basis was high. Of interest, the GFAP-luc-WPRE constructs injected into the hippocampus were expressed almost exclusively in neurons, as seen with double-labeling to NeuN/luciferase and GFAP/luciferase (Figure 3g–i). However, in the striatum we found that some astrocytes were also transduced (<5% of cells by volume) in addition to neurons, suggesting a regional tropism of AAV-2 across different brain areas, where AAV co-receptors may be more or less expressed and/or where GFAP gene expression may be differently regulated.

Table 2 Luciferase activity in rAAV-transduced rodent brain (units = pg/injection site)
Figure 3

AAV-2 injections to rat striatum with three different promoters (EF, NSE, GFAP) and WPRE, showing the effect of changing the promoter on neuronal expression of green fluorescent protein (GFP) or luciferase reporters. (a) EF promoter with GFP signal; (b) EF promoter with GFP/NeuN signal; (c) EF promoter with GFP/NeuN signal (close-up); (d) NSE promoter with GFP signal; (e) NSE promoter with GFP/NeuN signal; (f) control – no reporter; (g) GFAP promoter with luciferase signal, NeuN (blue)/luc (green) double-staining; (h) GFAP promoter with luciferase signal, GFAP (red)/luc (green) double-staining; (i) GFAP promoter with luciferase signal, close-up section from striatum showing luc(+) astrocytes in this region.

We injected three additional constructs containing the green fluorescent protein (GFP) gene under control of the three strongest promoter elements (NSE, EF, GFAP) into the rat hippocampus or striatum. Four weeks following stereotactic injection of 2 μl of rAAV-NSE-GFP-WPRE virus into the striatum, using the modified Abercrombie method,40 we counted approximately 400 000 neurons locally expressing GFP, a transduction efficiency of >105 cells per microliter. We obtained dramatic gene expression in vivo using all three promoters (NSE, EF, GFAP), which followed the same pattern of promoter activity seen with luciferase quantification. We determined, based on cell morphology and double-labeling to NeuN (Figure 3d, e; Figure 4e), that >99% of cells transduced with the NSE promoter cassette were neurons. Using the 1.8 kb NSE promoter, we also observed striking anterograde transport, previously reported by others,10 when the vector was injected into the hippocampus; the GFP protein was expressed at high levels in well-defined neuronal layers of the contralateral hippocampus (Figure 4f). With the EF promoter, we also found that the vast majority of cells transduced were neuronal (Figure 3b, c; Figure 4b, c) as defined by NeuN or MAP-2 expression. Using the GFAP promoter, the vast majority of cells transduced were neuronal in the hippocampus (Figure 4h, i), however in the striatum some transduced cells were astrocytes (<5%) using the GFAP construct with WPRE and BGH-polyA (Figure 5).

Figure 4

AAV-2 injections to rat hippocampus with three different promoters (EF, NSE, GFAP) and WPRE, showing the effect of changing the promoter on neuronal expression of green fluorescent protein (GFP) or luciferase reporters. (a) EF promoter with GFP; (b) EF promoter with GFP/NeuN; (c) EF promoter with GFP/NeuN (close-up); (d) NSE promoter with GFP; (e) NSE promoter with GFP/NeuN; (f) NSE promoter with GFP (contralateral hippocampus); (g) NSE promoter with GFP showing CA1 transduction; (h) GFAP promoter with luciferase; (i) GFAP promoter with luciferase (close-up).

Figure 5

Luciferase expression in striatal astroyctes using AAV-2 with GFAP promoter. The GFAP promoter does not provide glial-specific expression with AAV-2, although up to 5% of glia showed reporter gene expression using this promoter. Comparison of confocal images shows identical cells positive for luciferase and the glial marker GFAP. It is possible that some of the GFAP+ and luc+ cells are spiny neurons with atypical expression of GFAP and other proteins. (a) Striatal section immunostained for GFAP. (b) Identical section immunostained for luciferase.


There are two separate issues affecting the targeting and expression of AAV-2 in the brain: vector uptake through the cellular membrane, with escape from the endocytic/ lysosomal compartment and transport into the nucleus, which is regulated by capsid and DNA interactions with the host cell; and effective expression of transgenes by the promoter and regulatory elements, which includes production of the transcript and its interaction with cellular factors. With respect to the first issue, AAV has a neurotropism, largely attributable to its protein capsid, which makes it ideal for the study or manipulation of neurons. This tendency to infect neurons can be accentuated through the use of a neuron-specific promoter such as NSE, which we found confers >99% neuronal specificity. In a previous kinetic study by Bartlett et al,30 wild-type AAV-2 capsids labeled with a fluorophore were taken up into the nucleii of neuronal cells in vivo within minutes, and by 24 h AAV-2 particles were also found in scattered microglia. These experiments suggest that viral binding and uptake primarily determines the neurotropism of AAV-2. However, we also observed limited glial expression of AAV-2 in striatum using a strong glial promoter (GFAP), as well as variable expression in neurons depending on the promoter chosen, suggesting that factors other than cellular binding contribute significantly to AAV-mediated expression in neurons and glia.

The first cellular receptor to be identified for AAV was heparan sulfate proteoglycan (HSPG), which led to the use of heparan columns for AAV purification.41 Recently other co-receptors have been proposed, such as the alphaVbeta5 integrin receptor28 and the FGF-1 receptor.29 Although this interpretation is not universally accepted,42 it appears that the presence of heparan sulfate and the above proteins renders cells more permissive to AAV uptake. The expression of HSPGs, FGF-1 receptor, and integrin alphaVbeta5 receptor are high in neurons,43,44 consistent with high levels of AAV neuronal expression. Yet all neurons are not equally permissive to AAV infection, for example, we found that irrespective of the promoter, hilar neurons and CA1 hippocampal neurons are readily transduced (cf. Figure 4) whereas other neurons such as dentate granule cells are transduced poorly if at all, suggesting that both extra- and intra-cellular factors are responsible for differences in expression among neurons with AAV-2. It is interesting that glia can also express a wide range of heparan sulfate proteoglycans as well as FGF-1 and integrin receptors,45,46,47 but are not permissive to AAV-2. Different subsets of neurons or glia may express varying levels of viral binding factors, and future studies will help to define better the receptors across different brain regions. It is possible, for example, that oligodendrocytes express higher levels of AAV-binding receptors than other macroglia, which would help to explain the reported ability of AAV-2 to transduce oligodendrocytes.16 In addition, pathological responses may alter the expression profile of neurons or glia, for instance, reactive astrocytes often exhibit an up-regulation of GFAP and other cellular factors, which may affect AAV expression using a glial promoter.

With respect to the second issue affecting expression, several steps following transduction and before transcription may contribute to variability in AAV gene expression among neuronal cells. Recently it was reported that one mechanism of AAV entry into cells is dependent on dynamin, a GTPase protein involved in the clathrin-coated endocytotic pathway.48 In addition the role of ‘nuclear shuttling’ has been reported with nucleolin, a nuclear protein that specifically binds to the AAV capsid.49 Gene expression also depends on appropriate levels of host-encoded replication factors, and it is possible that AAV enters certain cells but may not be transcriptionally active. In this context, it has been suggested that single-strand binding proteins are important for second-strand DNA synthesis, a rate-limiting step in AAV gene expression.50 Also, the AAV vector genome can stably integrate in neurons following delivery, although the frequency of genomic insertion is believed to be relatively rare and not required for initial or stable gene expression,51 and variability in expression may be partly attributable to facilitated integration into open, transcriptionally active chromosomal sites in the vectors containing the strongest promoters. Therefore, in addition to differences in host cell receptors for AAV, regional variability in expression may be related to differences in nuclear localization, ssDNA synthesis, or vector integration.

The promoter and regulatory elements that interact with the host cell transcriptional apparatus are important for efficient transcription. Regulation of heterologous promoters usually matches the intrinsic host cell transcriptional state of the relevant host gene. For example, with the NSE promoter, the subset of neurons which naturally expresses the highest level of NSE would be expected to drive the highest level of the transgene. Yet, heterologous promoter effects cannot always be easily explained in terms of a simple correlation with the amount of promoter-binding factors. One report described CFTR expression from an AAV vector devoid of any promoter elements,51 suggesting that AAV ITRs have intrinsic promoter-enhancer activity and may lead to heterologous promoter ‘leakiness’, or expression outside of an attached promoter, which we observed in vivo in the extent of neuronal expression that was possible with a glial promoter. This phenomenon has been reported by others, for example, transfection of cells using the AAV vector with myelin basic protein (MBP) promoter resulted in expression in cell lines such as HeLa and h293, which ordinarily do not express MBP.16 Using an adenoviral vector, another study reported gene expression in both neuronal and non-neuronal cell lines using a NSE promoter. Paradoxically, the use of a GFAP promoter in GFAP-negative neural cells (Neuro-2a) gave good expression and a NSE promoter in glial cells gave equivalent expression as when a GFAP promoter was used.17 In our study, there was negligable glial expression in vivo using the full-length NSE promoter, but high levels of neuronal expression and low levels of glial expression using the GFAP promoter. The ‘leakiness’ of the GFAP promoter and the ‘tightness’ of the NSE promoter, with the majority of expression in neurons, probably relates to the preferential neuronal binding of AAV-2 as well as promoter-specific effects on enhancer elements within the ITRs. Both the 1.8 kb and 0.3 kb NSE promoters were able to confer cell-type specificity on the luciferase gene in vivo, supporting an earlier report that the proximal 5′ flanking region of the NSE gene is capable of conferring specificity on a heterologous gene in transfected cells.52

The immunostaining pattern of hippocampus we observed in vivo suggests that certain brain micro-environments may have very different propensities to take up virus particles, and local differences in intracellular factors also may influence gene expression. Because this study did not include quantitative PCR to viral sequences from microdissections of injected brain tissue, it was not possible to determine whether there were differences in the AAV infection rate of neurons from the striatum, hippocampus, cortex, and nigra. Based on the relative uniformity of reporter gene expression in previous studies in which transgenic animals were created using the NSE promoter,53 it is likely that cellular membrane interactions with AAV are the predominant factor in variable levels of expression in neurons for any given promoter. Nevertheless, when (otherwise identical) AAV-2 constructs containing different promoters are compared across the four brain regions, there did not appear to be a consistent variation of expression by region, suggesting that promoter effects may be as important as cellular transduction events for driving expression in different subsets of neurons.

The CMV promoter has been regarded as one of the strongest constitutively active virus-derived transcription elements and has been used for expression of a multitude of genes in many cell types and vector systems including AAV-2. The CMV early promoter was used in a AAV-based plasmid construct that was introduced to the human brain as part of a clinical trial54. However, in the present study the CMV promoter showed a relatively low luciferase activity in vivo in all the brain regions tested, suggesting serious limitations to a CMV-based gene transfer strategy for functional genomics or clinical applications using AAV-2. In fact, we found that it is among the weakest of promoters tested both in vitro and in vivo. In contrast, the rat NSE 1.8 kb promoter gave the highest level of luciferase activity in vitro and in vivo, suggesting that (along with chicken β-actin) it is the best promoter among those we tested to drive neuronal expression in the brain with AAV-2. One recent study using an AAV-2 construct with CMV promoter and SV40 polyA to drive a GFP reporter gene10 found that 50 nl of vector transduced ‘hundreds of neurons’ at the injection site. At best this corresponds to <10 000 neurons/μl, which is over an order of magnitude less than observed with this optimized vector system. Others have reported 3000–15 000 GFP-positive cells per injection using CMV immediate–early and NSE promoters,15 also at least an order of magnitude less than we obtained with the strongest expression vector containing NSE elements.

While the promoter data we present are valid for studies using AAV, they may not be applicable to vectors that contain viral elements which may interfere with transgene expression. Recombinant AAV-2 is unique among viral vectors because it does not contain viral nucleic acid aside from the ITRs, unlike most deleted vectors such as herpes simplex or adenovirus which frequently retain viral structural or replicative genes. HSV is difficult to compare directly with AAV for promoter effects, due to the effects of heterologous DNA elements. It was reported12 that HSV-NSE-lacZ drives short-term expression in vitro and in vivo, yet lytic promoters as well as heterologous promoters are effectively silenced or down-regulated during viral latency,55 and latency-associated promoters show variable long-term expression. Thus, other vectors may have idiosyncratic expression that relates to the viral life cycle rather than being intrinsic to the promoters or cis-regulatory elements. Another important point is that under some physiologic conditions that we did not address in this study, any of these promoters may be differentially regulated, and may be more or less suitable for a specific application. For example, the hENK promoter may be up-regulated in response to dopamine56 and thus might provide a self-regulatable promoter for functional studies of genes involved in movement or psychiatric disorders.

In summary, we found that very high levels of expression (over 2 × 105 cells/μl of vector) were possible in neurons with an optimized AAV-2 gene cassette. The promoters we tested varied in their ability to drive neuronal expression, with NSE and chicken β-actin being superior to the others. Although we carefully matched equivalent vector stocks, some changes in expression levels might not only reflect the relative strength of promoters, and the presence of the post-regulatory element, but might also be influenced by non-equivalence of viral infectivity for a given genomic titer. We found that the packaging efficency of AAV is dependent upon the length of the genome, in support of results by others,34,35 although the typical 5.1–5.3 kb packaging limit may not be absolute. The largest construct we made, which still showed acceptable packaging and transduction efficiency, was 5.7 kb. This is likely to be the upper limit of AAV-2 packaging size, although some genes appear to package more easily than others. Taken together, the ITR, WPRE, and BGH-polyA sequences we incorporated are approximately 1.2 kb. A strong promoter adds 0.6–2.2 kb to the vector genome, making the total size 1.8–3.4 kb without a gene insert. Therefore, an insert of >3.0 kb is feasible, and many genes of this size may be introduced for functional genomics studies. For example, genes directly relevant to Parkinson's or Alzheimer's disease such as parkin, alpha-synuclein, dopaminergic receptors or biosynthetic enzymes, amyloid precursor protein, presenilin, and many others are likely to be expressed efficiently in neurons with these vectors. Careful deletion analysis of test genes will allow the removal of non-essential sequence before packaging, adding to the possible repertoire of constructs that can be packaged. Optimization of AAV-2 has led to a powerful and versatile tool for the study of nervous system genes, and additional refinements in capsid design will facilitate targeting of a wide range of neuronal and non-neuronal cell types in the vertebrate brain.

Materials and methods

Plasmid construction and vector packaging

pSub201 was digested with XbaI in order to remove the entire AAV genome except for flanking ITRs. The reporter gene, WPRE and BGH-poly-A sequences, and a variety of promoters were inserted between the ITRs using appropriate restriction enzymes. Luciferase-containing vector stocks were generated using a three plasmid, helper-virus free packaging and purification method.8 Plasmids were prepared using alkaline lysis, purified with HPLC (BioCAD, Sprint, PerSeptive Biosystems, Foster City, CA, USA), and concentrated with two volumes of 100% ethanol. HPLC eluate buffers used for purification were pre-autoclaved and filtered. Packaged virus was subjected to ultracentrifugation twice to remove empty viral particles, and ELISA was used for quantitating AAV particles using commercial analysis kits (Progen, Germany). AAV-2 vector stocks were generated under identical conditions and matching titers were used with identical volumes administered. The GFP-containing constructs were generated in 293 cells by calcium phosphate-mediated co-transfection of vector plasmid with a helper virus gene, pDG plasmid (gift of Dr J Kleinschmidt, Deutsches Krebsforschungszentrum, Heidelberg, Germany) and purified by the method of Clark et al57 with modification. The clarified supernatants were applied at a flow rate of 3 ml/min on a BioCAD Sprint HPLC system with a POROS HE heparin column at 3 ml/min at ambient temperature. Aliquots of peak fractions were analyzed by SDS-PAGE and ELISA. Peak virus fraction was dialyzed against 100 mM NaCl, 1 mM MgCl and 20 mM sodium mono- and di-basic phosphate, pH 7.4, and subjected to ultrafiltration. Aliquots of viral vectors were stored at −80°C before injection. A portion of the samples was subjected to quantitative PCR analysis using the Perkin Elmer 7700 system, to quantify genomic titer. The PCR Taqman assay was a modified dot-blot protocol whereby AAV was serially diluted and sequentially digested with DNase I and proteinase K.58 Viral DNA was extracted twice with phenol–chloroform to remove proteins, and then precipitated with 2.5 equivalent volumes of ethanol. A standard amplification curve was set up at range from 102 to 106 copies and the amplification curve corresponding to each initial template copy number was obtained.

Primary neuronal culture

Primary cortical, striatal, hippocampal and nigral cell cultures were prepared from E15 pregnant Wistar rats. The embryos from two litters (approximately 24 embryos) were put in a dish of warm dissecting medium (Ca2+- and Mg2+-free Hank's balanced salt solution containing 0.6% glucose, 100 U/ml penicillin, 100 μg/ml streptomycin, 15 mM Hepes) and brain regions were dissected and collected in tubes containing medium. Tissues were then digested in a trypsin solution (0.25% trypsin, 200 μg/ml DNase in dissecting medium) for 15 min at 37°C in a shaking waterbath. The reaction was terminated by the addition of trypsin inhibition medium (100 μg/ml soybean trypsin inhibitor, 20% fetal bovine serum, 200 μg/ml DNase in dissecting medium), and tissues were washed twice. A cell suspension was obtained by triturating the tissue using a Pasteur pipette until tissue clumps were no longer visible, and then filtering the suspension through a 100 μm nylon filter to eliminate remaining cell clumps. The cells were counted and plated on to poly-L-lysine-coated dishes at a density of 250 000 cells per well in 24-well plates and at a density of 105 cells per well in 96-well plates in Neurobasal medium (Invitrogen/Life Technologies, USA) containing B27 supplement (Invitrogen/Life Technologies, USA) and 0.5 mM L-glutamine, and replenished every 48 h.

Viral infection of primary cultures

After 1 week, primary cells were infected with 108 particles of each recombinant virus per well for 24-well plates and incubated at 37°C, 5% CO2 for 1 week before assay. The particle titer of the recombinant viruses was standardized by ELISA and genomic quantification. For 96-well plates, virus was added at the same multiplicity of infection (MOI = 1000). To keep culture medium fresh, 200 μl of culture medium was added into each well 4 days after virus transdution. For immunohistochemical study, cells in 96-well plates were fixed in 4% paraformaldehyde for 30 min and washed once with PBS before treatment with rabbit polyclonal anti-luciferase antibody (Promega, Madison, WI, USA), 1:1000 dilution, and incubation overnight at 4°C. Following removal of primary antibody, cells were washed with PBS and anti-rabbit biotinylated secondary antibody (Sigma, St Louis, MO, USA) was added, 1:250 dilution. Following a 3-h incubation, cells were washed and treated for 2 h with ExtrAvidin peroxidase (Sigma), 1:250 dilution. Staining was performed using diaminobenzidine. The percentage of immunopositive cells was determined from five separate fields per slide (n = 6 wells per treatment).

Vector delivery in vivo

For in vivo analysis of expression, AAV vectors were stereotactically injected into four different brain regions with 2.5 × 109 particles in 1 μl per site, using a Kopf stereotactic frame. The particle titer of the recombinant viruses was standardized by ELISA and genomic quantification. The animals were anesthetized with ketamine/xylazine and burrholes were drilled at the injection sites. Viral stocks were injected through a 27-gauge cannula connected via polyethylene tubing to a 10 μl Hamilton syringe mounted in a CMA/100 microinjection pump, set to deliver 1 μl over 40 min. The needle was left in place for an additional 5 min before retraction. The skin incision was sutured with synthetic absorbable sutures and the animals were given post-operative analgesia and subcutaneous fluids in a heated recovery unit.

Tissue analysis

Animals were killed at 2 weeks, 1 month, or 3 months after injection. The brains were removed, dissected, and stored at −80°C before enzyme analysis. Luciferase activities were determined by luminometric assay using the Promega Luciferase Assay System. For immunohistochemistry, animals were perfused with 150 ml ice-cold PBS followed by 150 ml ice-cold 4% paraformaldehyde. The brains were removed and equilibrated in a cryoprotectant solution of 30% sucrose/PBS at 4°C. Coronal sections (30 μm) were cut using a cryostat, and treated with antibody as described above. For double-labeling studies, the luciferase or GFP-containing brain slices were incubated with NeuN or GFAP antibody as described previously.

Western blotting

Western blots of viral capsid proteins were performed in order to demonstrate the purity of virus and equivalency of vector stocks. Virus samples were aliquoted and stored at −80°C before electrophoresis. Total protein was quantified using the Bradford assay kit (Bio-Rad). For each sample, 20 μg protein was mixed with gel loading buffer and heated at 95°C for 5 min. Following standard electrophoresis on a SDS-PAGE mini-gel (10%) and equilibration in transfer buffer for 15 min, proteins were transferred to a nitrocellulose membrane. Following blocking and washing, the membranes were incubated with anti-VP1, VP2 and VP3 primary antibodies (1:200; Progen) overnight at 4°C. Membranes were washed and incubated with biotinylated secondary antibodies (1:500; Sigma), then incubated with ExtrAvidin peroxidase (1:500; Sigma) for 3 h at room temperature and bands were visualized using the ECL system (Amersham).

Southern blotting

Southern analysis of vector DNA showed full-length GFAP-luc, GFAP-luc-WPRE, PPE-luc and PPE-Luc-WPRE (Figure 1d). Viral genomes were extracted by a modified Hirt's extraction procedure, and transferred to nylon membrane. The probes were hybridized overnight at 65°C in hybridization buffer and washed twice at 60°C in 2 × SSC with 1% SDS for 15 min, following by washing in 0.1 × SSC with 0.1% SDS. The 500 b.p. probe was synthesized from the luciferase gene and labeled with Digoxigenin-11-dUTP (Boehringer-Mannheim) through PCR reaction. Signal was detected using the Dig Luminescent Detection kit (Roche, USA).


  1. 1

    Douglas JT et al. A system for the propagation of adenoviral vectors with genetically modified receptor specificities Nat Biotechnol 1999 17: 470–475

  2. 2

    Romanczuk H et al. Modification of an adenoviral vector with biologically selected peptides: a novel strategy for gene delivery to cells of choice Hum Gene Ther 1999 10: 2615–2626

  3. 3

    Girod A et al. Genetic capsid modifications allow efficent re-targeting of AAV-2 Nat Med 1999 9: 1052–1056

  4. 4

    Davidson B et al. Recombinant adeno-associated virus type 2, 4 and 5 vectors: transduction of variant cell types and regions in the mammalian cental nervous system Proc Natl Acad Sci USA 2000 97: 3428–3432

  5. 5

    Kaplitt MG et al. Long-term gene expression and phenotypic correction using AAV vectors in the mammalian brain Nat Genet 1994 8: 148–154

  6. 6

    Du B, Wu P, Boldt-Houle DM, Terwilliger EF . Efficient transduction of human neurons with an AAV vector Gene Therapy 1996 3: 254–261

  7. 7

    McCown TJ et al. Differential and persistent expression patterns of CNS gene transfer by an AAV vector Brain Res 1996 713: 99–107

  8. 8

    During MJ et al. Peroral gene therapy of lactose intolerance using an AAV vector Nat Med 1998 4: 1131–1135

  9. 9

    During MJ et al. In vivo expression of therapeutic human genes for dopamine production in the caudates of MPTP-treated monkeys using an AAV vector Gene Therapy 1998 5: 820–827

  10. 10

    Chamberlin NL, Du B, Lacalle SD, Saper CB . Recombinant AAV vector: use for transgene expression and anterograde tract tracing in the CNS Brain Res 1998 793: 169–175

  11. 11

    Andersen JK, Garber DA, Meaney CA, Breakefield XO . Gene transfer into mammalian central nervous system using herpes simplex virus vectors: extended expression of bacterial lacZ in neurons using the neuron-specific enolase promoter Hum Gene Ther 1992 3: 487–499

  12. 12

    Andersen JK, Frim DM, Isacson O, Breakefield XO . Herpesvirus-mediated gene delivery into the brain: specificity and efficiency of the neuron-specific enolase promoter Cell Mol Neurobiol 1993 13: 503–515

  13. 13

    Doll RF et al. Comparison of promoter strengths on gene delivery into mammalian brain cells using AAV vectors Gene Therapy 1996 3: 437–447

  14. 14

    Peel AL, Zolotukhin S, Schrimsher GW, Muzyczka N . Efficient transduction of green fluorescent protein in spinal cord neurons using AAV vectors containing cell type-specific promoter Gene Ther 1997 4: 16–24

  15. 15

    Klein RL et al. Neuron-specific transduction in the rat septohippocampal or nigrostriatal pathway by recombinant AAV vectors Exp Neurol 1998 150: 183–194

  16. 16

    Chen H, McCarty DM, Bruce AT, Suzuki K . Gene transfer and expression in oligodendrocytes under the control of myelin basic protein transcriptional control region mediated by AAV Gene Therapy 1998 5: 50–58

  17. 17

    Morelli AE et al. Neuronal and glial cell type-specific promoters within adenovirus recombinants restrict the expression of the apoptosis-inducing molecule Fas ligand to predetermined brain cell types, and abolish peripheral liver toxicity J Gen Virol 1999 80: 571–583

  18. 18

    Wang Y, Yu L, Geller AI . Diverse stabilities of expression in the rat brain from different cellular promoters in a helper-virus-free herpes simplex virus type 1 vector system Hum Gene Ther 1999 10: 1761–1771

  19. 19

    Hannas-Djebbara Z et al. Transgene expression of plasmid DNAs directed by viral or neural promoters in the rat brain Mol Brain Res 1997 46: 91–99

  20. 20

    Podsakoff G, Wong KK, Chatterjee S . Efficient gene transfer into non-dividing cells by AAV-based vectors J Virol 1994 68: 5656–5666

  21. 21

    Xiao X, Li J, McCown TJ, Samulski RJ . Gene transfer by AAV into the central nervous system Exp Neurol 1997 144: 113–124

  22. 22

    Wu P, Phillips MI, Bui J, Terwilliger EF . AAV vector mediated transgene integration into neurons and other nondividing cell targets J Virol 1998 72: 5919–5926

  23. 23

    Lo WD et al. AAV-mediated gene transfer to the brain: duration and modulation of expression Hum Gene Ther 1999 20: 201–213

  24. 24

    Mizukami H, Young NS, Brown KE . AAV-2 binds to a 150kD cell membrane glycoprotein Virology 1996 217: 124–130

  25. 25

    Duan D et al. Polarity influences the efficiency of recombinant AAV infection in differentiated airway epithelia Hum Gene Ther 1998 9: 2761–2776

  26. 26

    Summerford C, Samulski RJ . Membrane-associated heparan sulfate proteoglycan is a receptor for AAV-2 virions J Virol 1998 72: 1438–1445

  27. 27

    Ruffing M, Heid H, Kleinschmidt JA . Mutations in the carboxy terminus of AAV capsid proteins affect viral infectivity: lack of an RGD integrin-binding motif J Gen Virol 1994 75: 3385–3392

  28. 28

    Summerford C, Barlett JS, Samulski RJ . aVb5 integrin: a co-receptor for AAV-2 infection Nat Med 1999 5: 78–81

  29. 29

    Qing K et al. Human fibroblast growth factor receptor is a co-receptor for infection by AAV-2 Nat Med 1999 5: 71–77

  30. 30

    Bartlett JS, Samulski RJ, McCown TJ . Selective and rapid uptake of AAV-2 in brain Hum Gene Ther 1998 9: 1181–1186

  31. 31

    Linden RM, Woo SLC . AAVant-garde gene therapy Nat Med 1999 5: 21–22

  32. 32

    Dong JY, Fan PD, Frizzell RA . Quantitative analysis of the packaging capacity of recombinant AAV Hum Gene Ther 1996 7: 2101–2112

  33. 33

    Hermonat PL, Quirk JG, Bishop BM, Han L . The packaging capacity of AAV and the potential for wild-type-plus AAV gene therapy vectors FEBS Lett 1997 407: 78–84

  34. 34

    Wang D et al. Efficient CFTR expression from AAV vectors packaged with promoters – the 2nd generation Gene Therapy 1999 6: 667–675

  35. 35

    Huang ZM, Yen TS . Role of the hepatitis B virus posttranscritional regulatory element in export of intronless transcripts Mol Cell Biol 1995 15: 3864–3869

  36. 36

    Donello J, Loeb JE, Hope TJ . Woodchuck hepatitis virus contains a tripartite posttranscriptional regulatory element J Virol 1998 72: 5085–5092

  37. 37

    Pfarr DS et al. Differential effects of polyadenylation regions on gene expression in mammalian cells DNA 1986 5: 115–122

  38. 38

    Yew NS et al. Optimization of plasmid vectors for high-level expression in lung epithelial cells Hum Gene Ther 1997 8: 575–584

  39. 39

    Blomer U et al. Highly efficient and sustained gene transfer in adult neurons with a lentivirus vector J Virol 1997 71: 6641–6649

  40. 40

    McLachlan EM, Janig W . The cell bodies of origin of sympathetic and sensory axons in some skin and muscle nerves of the cat hindlimb J Comp Neurol 1983 214: 115–130

  41. 41

    Zolotukhin S et al. Recombinant AAV virus purification using novel methods improves infectious titer and yield Gene Therapy 1999 6: 973–985

  42. 42

    Qiu J, Mizukami H, Brown KE . AAV-2 co-receptors? Nat Med 1999 5: 467–468

  43. 43

    Nurcombe V et al. Developmental regulation of neural response to FGF-1 and FGF-2 by heparan sulfate proteoglycan Science 1993 260: 103–106

  44. 44

    Pinkstaff JK et al. Integrin subunit gene expression is regionally differentiated in adult brain J Neurosci 1999 19: 1541–1556

  45. 45

    Bansal R et al. Development and FGF-2-mediated regulation of syndecans and glypican in oligodendrocytes Mol Cell Neurosci 1997 7: 276–288

  46. 46

    Takami K et al. FGF-1 expression in the cortex and hippocampus in Alzheimer's Disease Brain Res 1998 802: 89–97

  47. 47

    Tawil NJ et al. Expression and distribution of functional integrins in rat CNS glia J Neurosci Res 1994 39: 436–447

  48. 48

    Duan D et al. Dynamin is required for recombinant AAV-2 infection J Virol 1999 73: 10371–10376

  49. 49

    Qiu J, Brown KE . A 110kDa nuclear shuttle protein, nucleolin, specifically binds to AAV-2 capsid Virology 1999 257: 373–382

  50. 50

    Qing K et al. AAV-2-mediated gene transfer: correlation of tyrosine phosphorylation of the cellular single-stranded D sequence-binding protein with transgene expression in human cells in vitro and murine tissues in vivo J Virol 1998 72: 1593–1599

  51. 51

    Flotte TR et al. Stable in vivo expression of the cystic fibrosis transmembrane conductance regulator with an AAV vector Proc Natl Acad Sci USA 1993 90: 10613–10617

  52. 52

    Twyman RM, Jones EA . Sequences in the proximal 5′ flanking region of the rat neuron-specific enolase (NSE) gene are sufficient for cell type-specific reporter gene expression J Mol Neurosci 1997 8: 63–73

  53. 53

    Forss-Petter S et al. Transgenic mice expressing beta-galactosidase in mature neurons under neuron-specific enolase promoter control Neuron 1990 5: 187–197

  54. 54

    Leone P et al. Global CNS gene transfer for a childhood neurogenetic enzyme deficiency: Canavan disease Curr Opin Mol Ther 1999 1: 487–492

  55. 55

    Lachmann RH, Efstathiou S . Use of herpes simplex virus type 1 for transgene expression within the nervous system Clin Sci (Colch) 1999 96: 533–541

  56. 56

    La Gamma EF et al. Dopaminergic regulation of a transfected preproenkephalin promoter in primary rat astrocytes in vitro and in vivo Exp Neurol 1994 130: 304–310

  57. 57

    Clark KR et al. Highly purified recombinant AAV vectors are biologically active and free of detectable helper and wild type virus Hum Gene Ther 1999 10: 1031–1039

  58. 58

    Wistuba A et al. Subcellular compartmentalization of AAV-2 assembly J Virol 1997 71: 1341–1352

Download references


We would like to thank the following individuals for kindly providing gene constructs and plasmids: the human enkephalin promoter was provided by Dr M Comb; the NF promoters were donated by Dr J-P Julien; the human nAChR promoter was donated by Drs A Bessis and J-P Changeux at Institut Pasteur; and the human EF promoter was donated by Dr A Shibui. We would also like to thank S McPhee for his help with the digital images.

Author information

Correspondence to MJ During.

Rights and permissions

Reprints and Permissions

About this article


  • gene cassette
  • adeno-associated virus
  • neuron
  • promoter
  • post-regulatory element
  • functional genomics

Further reading