Long-term AAV vector gene and protein expression in mouse brain from a small pan-cellular promoter is similar to neural cell promoters

Abstract

The neurogenetic, lysosomal enzyme (LSE) deficiency diseases are characterized by storage lesions throughout the brain; therefore, gene transfer needs to provide wide-spread distribution of the normal enzyme. Adeno-associated virus (AAV) vectors can be effective in the brain despite limited transduction because LSEs are exported to neighboring cells (cross-correction) to reverse the metabolic deficit. The extent of correction is determined by a combination of the total amount of LSE produced by a vector and the spatial distribution of the vector within the brain. Neuron-specific promoters have been used in the brain because AAV predominantly transduces neurons. However, these promoters are large, using up a substantial amount of the limited cloning capacity of AAV vector genomes. A small promoter that is active in all cells, from the LSE β-glucuronidase (GUSB), has been used for long-term expression in AAV vectors in the brain but the natural promoter is expressed at very low levels. The amount of LSE exported from a cell is proportional to the level of transcription, thus more active promoters would export more LSE for cross-correction, but direct comparisons have not been reported. In this study, we show that in long-term experiments (<6 months) the GUSB minimal promoter (hGBp) expresses the hGUSB enzyme in brain at similar levels as the neuron-specific enolase promoter or the promoter from the latency-associated transcript of herpes simplex virus. The hGBp minimal promoter thus may be useful for long-term expression in the central nervous system of large cDNAs, bicitronic transcription units, self-complimentary or other designs with size constraints in the AAV vector system.

Neurogenetic diseases typically have lesions throughout the brain, thus requiring global correction.¹ Several regions of the mouse brain have been transduced using recombinant adeno-associated virus (AAV) vectors but the number of cells that AAV can transduce is relatively limited.² The lysosomal storage diseases (LSDs) are candidates for AAV treatment in the central nervous system (CNS) because the enzymes involved can be secreted from a subset of transduced brain cells and can be taken up by adjacent diseased cells, a process called cross-correction.^3-5 One strategy to deliver more total enzyme to the brain from a limited number of transduced cells is to increase transcription of the transferred gene.³ There is a direct correlation between the amount of lysosomal enzyme (LSE) gene transcription from a vector and the total amounts of LSE that accumulate in the cell and that are secreted into the extracellular space.⁶ Therefore, producing sustained higher levels of vector gene expression by employing stronger promoters may result in more total therapeutic enzyme delivery within the brain.

Viral promoters have been used in a number of AAV vectors in the CNS but can be downregulated in vivo. The cytomegalovirus (CMV) immediate early promoter is effective in some regions of the mouse brain, notably the striatum, but vector gene expression declines significantly in the hippocampus and cortex.⁴ When the CMV enhancer is paired with the housekeeping promoter from b-actin, more sustained gene expression can occur.⁷ The latency-associated transcript (LAT) promoter from herpes simplex virus (HSV)-1 is a potential candidate for long-term expression because it mediates lifelong expression in latent infections in postmitotic neurons.⁸ LAT expression is significantly higher in neuronal than non-neuronal cell lines.⁹ Studies with HSV vectors in animals have demonstrated that the LAT promoter is capable of expressing an LSE for extended periods of time,⁸ but it has not been evaluated in other types of viral vectors.

A number of eukaryotic promoters have been used in AAV vectors in the CNS. Cell-type-specific promoters have high transcriptional activity and neuron-specific promoters have been used in the AAV vectors for the CNS because AAVs predominantly transduce neurons.^10-12 The neuron-specific enolase (NSE) promoter has produced sustained expression in vivo of a transferred gene in neurons in spinal cord and brain, including LSEs.^10-12 Housekeeping promoters have been used to achieve vector gene expression in multiple cell types, although they are generally weak in transcription studies. However, some housekeeping promoters have the advantage of being small, which allows the inclusion of large coding sequences, introns and other elements within the size constraints of AAV vectors. One candidate is the minimal promoter from the human β-glucuronidase (GUSB) gene (378 bp),¹³ which is expressed in all cells and tissues.¹⁴ The hGBp mediates sustained expression of the GUSB cDNA in several different cell types from retrovirus, lentivirus and AAV vectors,^15-17 but the endogenous GUSB gene is expressed at a very slow rate.¹⁸

Most AAV serotypes transduce neurons, but some AAV serotypes can also transduce oligodendrocytes and astrocytes,^19-21 which is advantageous for LSDs because all cell types are affected. Because the goal of gene transfer in the CNS in LSDs is to deliver as much total enzyme as possible, we tested the amounts of total enzyme expressed over long period of time from vectors that expressed high levels of a gene but in a limited number of cells (neurons) vs a vector with a weaker pan-cellular promoter but which is expressed in multiple cell types (Figure 1). The AAV1 serotype was selected because it transduces many structures in the brain and the GUSB cDNA was used as a test gene because it has been widely studied.^23-25

Figure 1.

β-Glucuronidase (GUSB) expressing adeno-associated virus (AAV) vector constructs. (a)HβH encodes the 378 bp hGUSB promoter driving the hGUSB cDNA with a 100 bp mini-intron splice donor/splice acceptor (SD-SA) and an SV40 polyA addition signal, as described.¹⁵ (b) Latency-associated transcript (LAT)-GUSB vector constructed by cloning a 4037 bp MluI-XcmI fragment from 1716-LAT-hGUSB vector described previously²² into the pZAC 2.1 plasmid between the 5′ITR and the SV40 SD-SA signal. (c) Neuron-specific enolase (NSE)-GUSB vector was constructed from the BglII-XhoI fragment containing the 1.8 kb NSE promoter isolated from the pTR-UF4 plasmid (a kind gift from N Muzyczka, University of Florida, FL, USA). The NSE promoter was inserted downstream of the 170 bp 5′ITR, followed by the 2.2 kb hGUSB cDNA, a 177 bp SV40 SD-SA region, 238 bp SV40 polyA and 167 bp 3′ITR. Vectors were packaged into viruses using the AAV1 serotype capsid, as described,⁵,²³ and titers of 3.1 × 10¹³ viral genome equivalents (GE) per ml were used.

The distribution of gene expression was tested using the pan-cellular vector (AAV1-HbH) to maximize the number of cell types. A single injection into the adult mouse striatum resulted in extensive spread in the rostral-to-caudal direction, with most of the transduction occurring on the ipsilateral side and only small amounts of transport^5,20 to the contralateral side (Supplementary Figure 1). The types of cells transduced were determined by colocalization of cell-type-specific antibodies and fluorescent in situ hybridization (FISH) for hGUSB mRNA.^4,21,23 The predominant cell type transduced was neurons (Figures 2a–c) but substantial transduction of white matter tracks also occurred, which colocalized with an oligodendrocyte-specific marker (proteolipid protein, PLP) (Figures 2d–f). Astrocytes were also transduced but to a lesser extent (Figures 2g–i).

Figure 2.

In vivo transduction of multiple cell types by HβH. Transduction was detected by fluorescent in situ hybridization (FISH) and then with immunohistochemistry (IHC) for cell-type-specific markers listed to left. (a, d, g) Human GUSB mRNA FISH; (b) neuronal nuclei (NeuN) IHC; (c) merged image of hGUSB and NSE; (e) oligodendrocyte marker proteolipid protein (PLP) IHC; (f) merged image of hGUSB and PLP with inset showing the FISH signal colocalizes with DAPI staining; (h) astrocyte-specific marker glial fibrillary acidic protein (GFAP) IHC; (i) merged image of hGUSB and GFAP. Arrows (c, i) point out areas of colocalization of FISH and IHC. Panels a–c are × 10 images; Panels d–i are × 40 images. Animal procedures were approved by the IACUC. Injections were administered into the brains of normal C3H/ HeOuJ adult mice using standard stereotaxic procedures.⁴,²⁰ Slow injections of 1 μl vector to target major structures were: a rostral tract 0.1 mm caudal to bregma, 2.3 mm left of midsagittal, and depths of 2.6 mm for striatum and 0.8 mm for cortex; and a caudal tract 2.1 mm caudal of bregma, 1.5 mm left of midline, at depths of 3.0 mm for thalamus and 1.4 mm for hippocampus. At the time of killing, mice were anesthetized and the brains were perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA). The brains were then postfixed in 4% PFA, dehydrated in 30% sucrose, frozen in OCT compound and cut by cryostat into 20 μM coronal sections. FISH and immunofluorescent antibodies (IFA) for cell type identification were performed as described.⁵,²³

Transcription was evaluated in adult mouse brains by unilaterally injecting four sites in two tracks:^20,26 a rostral track to deliver vector into the cortex and the striatum and a caudal track with injections into the hippocampus and the thalamus. Cryosections were evaluated by in situ hybridization for distribution of gene transfer. At 2 months all three vectors generally showed similar patterns of spatial distribution (Figure 3a). Although the in situ hybridization varied from section to section, the overall amount of vector GUSB mRNA was statistically similar (P>0.05) in the injected area of the brains (Figure 3b). Long-term delivery of GUSB was measured by total amounts of GUSB enzymatic activity 6 months after injection (Table 1). The total activity on the ipsilateral side was similar (P>0.05) for all three promoters. However, on the contralateral side the NSE and LAT vectors produced moderately higher levels compared to the HbH vector (P<0.05), suggesting that some neuronal transport had occurred.^4,5,20

Figure 3.

Distribution and relative expression of human β-glucuronidase (hGUSB) transcription using vectors with neuronal promoters. (a) In situ hybridization for hGUSB mRNA in brain sections 2 months after injection with HβH, latency-associated transcript (LAT)-GUSB or neuron-specific enolase (NSE)-GUSB. Upper row, sections at level of cortex and striatum; lower row, level of hippocampus and thalamus. (b) Northern blot analysis from injected hemispheres (n=3 mice). Probes specific for hGUSB were used for phosphoimage analysis and subsequently reprobed with mouse β-actin to normalize transcript levels. Transcript levels are normalized to the levels in transgenic mice, which express the hGUSB on a null mouse background. In situ hybridization was performed as in Figure 2. Northern blot analysis used 30 μg samples of total RNA. Membranes were hybridized with a cDNA probe specific for hGUSB (544 bp NotI-BglI fragment spanning exons 1–5) and phosphoimager readings were obtained for quantitative analysis. Membranes were stripped then reprobed with murine β-actin, and ratios of GUSB to actin were calculated.

Table 1.

Enzyme levels in brain from AAV1 vectors 6 months after injection

Vector	GUSB enzyme activity (nmol h−1 per mg)
	Injected	Contralateral
HβH	304.0	56.7
	629.9	45.4
	1046.0	38.1
	660.0	30.8
	920.5	54.1
Mean (±s.e.m.)	712.1 (257.3)	45.0 (9.7)
LAT-hGUSB	411.0	44.8
	765.2	127.2
	1552.5	169.7
Mean (±s.e.m.)	909.6 (477.1)	113.9 (51.9)a
NSE-hGUSB	460.8	62.9
	1110.2	108.4
	430.7	71.9
	882.4	53.2
Mean (±s.e.m.)	721.0 (282.4)	74.1 (22.9)a

Abbreviations: hGUSB, human β-glucuronidase; LAT, latency-associated transcript; NSE, neuron-specific enolase.

Controls: C3H untreated (n=3): left 5.8 (6.0); right 8.1 (4.6). Transgenic GUSB (n=3): left 633.8 (144.4); right 779.6 (249.1). Differences between hemispheres were not significant. Brain tissue from the injected and contralateral sides were homogenized separately in lysis buffer (0.2% Triton X-100, 0.9% saline) and frozen at −80 °C. Homogenates were thawed at 37 °C and centrifuged for 2 min at 12 000 g. Supernatants were assayed for protein levels using a Bio-Rad protein assay, then added to 10 mm 4-methyl-umbelliferyl-β-d-glucuronide substrate for 1 h at 37 °C in parallel with 4-methyl-umbelliferone standard. Reactions were terminated with stop buffer (0.32 m glycine, 0.2 m sodium carbonate) and read in a fluorometer, as previously described.⁶ Control mice were uninjected C3H/HeOuJ mice (background control) and a transgenic strain¹⁴ carrying the normal human GUSB gene on a background with a homozygous mutation in the mouse GUSB gene, which expresses high levels of human GUSB on the null mouse background (positive control). Animal procedures and analyses were performed as described in Passini et al.⁴ and Cearley and Wolfe,²⁰ and in Figure 2.

^aGUSB significantly higher than detected with HβH vector (P<0.05).

The transferred enzyme was shown to be functional by testing in GUSB-deficient mice. GUSB is responsible for one step in the sequential degradation of glucuronic-acid-containing glycosaminoglycans (GAGs). A genetic deficiency of GUSB results in lysosomal accumulation of partially degraded GAGs, causing the disorder mucopolysaccharidosis (MPS) type VII. The hallmark lesion is distended lysosomes, which appear as large vacuoles in cell bodies.³ Supplementary Figure 2 shows reversal of the storage lesions in areas of the brain where the enzyme was distributed after AAV-1 vector injection.

AAV vectors are effective for delivering enzymes such as GUSB to the CNS. Injecting AAV into the developing mouse brain (fetal or neonatal) can result in widespread gene transfer due to the disseminated distribution of the vector in the growing brain. This mediates widespread distribution of enzyme and correction of lesions throughout the brain.^15,23,27 In more developed brains, the situation is considerably different, with the spread of vector from the injection site being much more limited and the number of injections that can be performed in an individual animal being constrained.^4,5,20 One strategy to expand the distribution of an LSE within the developed brain is to increase enzyme secretion from a limited number of transduced cells.³ Further distribution can be achieved in some regions by axonal transport of the vector genome and/or enzyme to distal sites.^4,5,20,28,29

Many virus promoters have been used to mediate strong transcription from AAV vectors.¹¹ The LAT sequence we used included regions encoding the LAT promoter and long-term expression element, which is thought to keep the LAT promoter active during latency in vivo.^30-33 However, to our knowledge, LAT sequences have not been tested as an independent promoter in a non-herpesvirus vector. Our data with AAV indicate that the LAT promoter might be effective for long-term expression in other classes of vectors as well. Eukaryotic promoters can be used to mediate gene expression in specific types of differentiated cells, such as dentate granule cells in the hippocampus.³⁴ The NSE promoter has very high activity in several structures of the rat brain compared to viral promoters¹¹ and was effective in expressing high levels of LSE in the brain of a mouse model of Sanfilippo disease, another form of MPS.¹⁰ In the present experiments, the NSE promoter mediated sustained gene expression, but the LAT and hGUSB promoters were just as good with respect to overall, long-term levels. Nevertheless, the LAT and NSE promoters may be preferred to hGBp for some applications because they produced a moderate but significant (P<0.05) increase in enzyme activity on the contralateral side. This was probably a result of all of the expression occurring in neurons, which is more likely to result in transport to contralateral sites through axonal transport.^{5,20,25,28,35}

The hGBp structure was not predicted to express as high a level as it does from within a vector because the native gene is expressed at a very slow rate in vivo (estimated to be 1 transcript per cell per 36–40 h).¹⁸ The GUSB promoter is unusual in being significantly different between the mouse and human: the mouse promoter has a TATA and CAATT box structure, whereas the human promoter has Sp1 and Ap1 sites.¹³ Nevertheless, our data show that the minimal promoter sequence from the human gene is very effective as a pan-cellular promoter in viral vectors. The small size of the hGBp promoter is a significant advantage for AAV vectors because it was able to deliver high levels of gene expression and enzyme production.

The rationale for using a pan-cellular promoter in genetic diseases is to express a therapeutic cDNA in as many cells as possible because in most genetic diseases most or all cells of a specific type are affected.¹ The hGBp has been shown to be functional for GUSB expression in vivo with retrovirus, lentivirus and AAV vectors; in several cell types including fibroblasts, neurons, neural stem cells, astrocytes and oligodendrocytes; and in at least three species, mice, dogs and cats.^{3-5,16,17,19,36} Thus, it appears to be potentially useful in settings where widespread expression is desired. The hGBp promoter should thus be useful in any type of brain cell where continuous high-level gene expression is needed or where a small promoter is required in the vector to accommodate other genetic elements.