Heparin-binding correlates with increased efficiency of AAV1- and AAV6-mediated transduction of striated muscle, but negatively impacts CNS transduction

Abstract

Gene delivery vectors derived from adeno-associated virus (AAV) have great potential as therapeutic agents. rAAV1 and rAAV6, efficiently target striated muscle, but the mechanisms that determine their tropism remain unclear. It is known that AAV6, but not AAV1, interacts with heparin-sulfate proteoglycans (HSPG). HSPGs are not primary receptors for AAV6, but heparin interactions may affect tissue tropism and transduction. To investigate these possibilities, we generated rAAV1 and rAAV6 capsids that do or do not bind heparin. We evaluated the transduction profile of these vectors in vivo across multiple routes of administration, and found that heparin binding capability influences tissue transduction in striated muscle and neuronal tissues. Heparin-binding capsids transduce striated muscle more efficiently than non-binding capsids, via both intramuscular and intravenous injection. However, rAAV6 achieved greater muscle transduction than the heparin-binding rAAV1 variant, suggesting that there are additional factors that influence differences in transduction efficiency between AAV1 and AAV6. Interestingly, the opposite trend was found when vectors were delivered via intracranial injection. Non-binding vectors achieved robust and widespread gene expression, whereas transduction via heparin-binding serotypes was substantially reduced. These data indicate that heparin-binding capability is an important determinant of transduction that should be considered in the design of rAAV-mediated gene therapies.

Introduction

Viral vector gene delivery and the field of gene therapy have evolved tremendously over the past decade. The use of vectors derived from viruses is transitioning from the bench to the clinic and may lead to an effective treatment for multiple genetic disorders. Among the viral vectors currently being investigated, recombinant adeno-associated viral (rAAV) vectors have demonstrated great promise as therapeutic agents due to low pathogenicity and the capability of efficient transgene delivery and expression in multiple tissues (reviewed by Schultz, 2008).¹

AAV is a single-stranded, non-enveloped DNA virus that is a member of the parvovirus family. The AAV genome contains two open reading frames that encode four replication (rep) proteins, three capsid (cap) proteins¹, and a recently identified assembly-activating protein.² The three separate capsid proteins (VP1, VP2, and VP3) are generated from the same genomic sequence via alternative translation start codons. Numerous AAV capsid variants have been isolated, and to date, at least twelve primate serotypes are recognized.¹ The overall level of capsid amino acid sequence identity varies between serotypes. AAV serotypes 1-9 are approximately 45% identical, with the highest degree of divergence seen in AAV4 and AAV5.³ In contrast, AAV1 and AAV6 are 99% identical and differ by only six amino acids.^4,4a,5

Different serotypes of AAV demonstrate unique profiles of tissue tropism. AAV1 and AAV6 are among the serotypes that can transduce striated muscle with relatively high efficiency.^6-13 In addition, AAV1 has been shown to exhibit a high affinity for neuronal tissues.^14,15 The affinity of AAV1 and AAV6 for striated muscle makes these serotypes primary candidates for the treatment of neuromuscular disorders, including Duchenne muscular dystrophy (DMD). DMD is a fatal genetic disorder that leads to severe muscle wasting and death in young adulthood.¹⁶ Currently, no curative treatment has been established, but gene replacement strategies may prove to be an effective treatment for this devastating disease.¹⁶ Successful gene therapy of DMD will necessitate transduction of large volumes of striated muscle, emphasizing the need for robust, muscle-tropic vectors.

The use of vectors with a high degree of efficiency and tissue specificity will likely be a core component of any successful viral vector-mediated gene therapy. However, the factors governing tropism of AAV1 and AAV6 for striated muscle and neuronal tissue have not yet been clearly defined. Both serotypes are remarkably similar, but they have different affinities for heparin. AAV6 exhibits an affinity for heparin, permitting purification of AAV6 capsids via heparin-affinity chromatography.¹⁷ In contrast, AAV1 does not interact with heparin. This difference is mediated by a single lysine residue at position 531 in the AAV6 capsid protein.^18,19 Mutation of this residue in the AAV1 capsid to match that of AAV6 confers heparin-binding capabilities on AAV1. The reciprocal mutation abolishes heparin binding in AAV6 (Fig. 1).

Figure 1.

Salt elution profile of AAV2, AAV6, and AAV1 chimeric capsids. Capsids were passed through a heparin column and eluted with increasing concentrations of NaCl. Focus forming units (ffu) per ml were determined using various dilutions of fractions added to HT1080 cells seeded at 9 ×104 cells per well of a 12 well plate 24 hours pre-transduction. At 72 hours post-transduction cells were fixed and stained for hPLAP activity.

Numerous viruses exploit heparan sulfate proteoglycans for cell binding and entry. It has previously been shown that the heparin affinity of AAV6 modestly increases transduction of hepatocytes, both in vitro and in vivo.¹⁹ However, the influence of heparin binding on transduction of other tissues was not evaluated, and administration of vector in vivo was limited to low-dose portal vein infusion. In this study, we examined the influence of heparin-binding capability on transduction efficiency and tissue tropism in wild-type mice. We generated rAAV1 and rAAV6 vectors that do and do not bind heparin and evaluated their performance in vivo across multiple routes of delivery, including intramuscular, intravenous, and intracranial injection. Our results demonstrate that heparin-binding capability is correlated with increased efficiency of striated muscle transduction. In contrast, heparin binding capability correlates with reduced transduction of neuronal tissue and restricted vector dissemination within the central nervous system.

Results

Mosaic AAV6 capsids demonstrate unique elution profiles during heparin-affinity chromatography purification

Both AAV2 and AAV6 are known to bind heparin, and AAV2 has been shown to utilize HSPG to facilitate transduction.²⁰ However, the residues mediating heparin binding in AAV2 and AAV6 are found at different positions of the AAV capsid protein^19,21, suggesting that the heparin-binding sites in these two serotypes may not be qualitatively or functionally equivalent. To further characterize the heparin-binding site of AAV2 in comparison to AAV6, pair-wise combinations of AAV1, AAV2, and AAV6 capsid expression plasmids were used to generate chimeric rAAV vector preparations that expressed the human placental alkaline phosphatase (hPLAP) reporter gene. The elution profiles of these recombinant capsids were evaluated using heparin-affinity chromatography (Fig. 1), and transduction titers were utilized to identify the fractions that contained viral particles. In agreement with published results^21a, rAAV1/2 chimeric vector preparations (Figs. 1a,b) behave similarly to rAAV2 with regard to heparin binding, eluting at 400 to 500 mM NaCl. In contrast, a significant fraction of rAAV1/6 chimeric capsids lost the ability to bind to heparin columns (Fig. 1e). Approximately 30% of alkaline phosphatase units were detected in column flow-through fractions. In contrast, rAAV2/6 capsids (Fig. 1c) bound heparin columns, but required higher concentrations of NaCl than either parental vector (>700 mM) for elution, suggesting that the unique heparin-binding sites in AAV2 and AAV6 capsids can function cooperatively to enhance heparin affinity.

Attenuation of heparin binding dramatically reduces transduction efficiency following intramuscular administration of vector

Heparin binding in AAV6 is dependent on a unique lysine residue at position 531 of the viral capsid protein. This residue is not conserved in AAV1. To investigate the influence of heparin-binding affinity on AAV1 and AAV6 transduction efficiency, we generated recombinant capsids that do or do not bind heparin. rAAV1 E531K encodes a glutamate-to-lysine substitution that confers heparin binding on AAV1 (Fig. 1f). rAAV6 K531E encodes the reciprocal lysine-to-glutamate substitution that abolishes heparin-binding in AAV6. No differences were observed in the elution profiles of rAA1 E531K and rAAV6 on heparin affinity columns (Fig. 1f). Each of the four rAAV heparin-binding variants was used to generate pseudotyped vectors expressing the hPLAP reporter gene under control of the CMV promoter.

We first performed a series of IM injections into the tibialis anterior of wild-type mice (Fig. 2). Vector was administered at three different doses: 10¹⁰ vector genomes (vg), 10⁹ vg, and 10⁸ vg per muscle. Muscles were collected two weeks post-injection and stained for hPLAP expression. At the highest dose, the majority of the muscle was transduced by each serotype, and differences in transduction efficiency were not readily apparent. However, at the two lower doses, dramatic differences in transduction efficiencies were discernable. At a dose of 10⁹ vg, muscles treated with rAAV6 demonstrated the highest level of transduction, followed by rAAV1 E531K, and rAAV6 K531E. Those treated with rAAV1 had the least number of positive fibers. At the lowest dose, hPLAP staining was only evident in the cohort treated with rAAV6. Interestingly, the transduction efficiency of rAAV1 was greatly improved when modified to bind heparin, but was still significantly less than that of rAAV6. These data indicate that the ability to bind heparin correlates positively with the efficiency of skeletal muscle transduction, but is not the sole determinant of skeletal muscle transduction efficiency.

Figure 2.

IM injection of rAAV vectors into tibialis anterior. Images are representative fields from cross-sections of tibialis anterior stained for alkaline phosphatase activity in weight-matched, adult mice. Injected serotype is indicated horizontally. Total vector genome dose is indicated in the left vertical column. vg, vector genomes. Scale bar: 50 μm.

Heparin binding increases efficiency of striated muscle transduction following systemic delivery

The route of vector administration can influence tissue tropism and dissemination of AAV in vivo.³ Intramuscular injection delivers vector directly to muscle fibers, but efficient transport of vector from the injection site remains a significant obstacle for non-secreted proteins. In contrast, systemic vascular delivery permits dissemination of vector to a larger volume of tissue and is an effective method for achieving widespread transduction.⁷ However, transduction patterns following systemic delivery are highly dependent on capsid serotype³ and have the potential to be influenced by interactions of the viral capsid with both extracellular and intracellular proteins.¹

To determine the influence of heparin binding on the transduction profile of rAAV1 and rAAV6 following systemic delivery, we injected cohorts of mice intravenously with one of the four heparin-binding variants (Fig. 3). Tissues were collected and analyzed for hPLAP expression and vector genome copy number. hPLAP expression was primarily seen in striated muscle and liver. The pattern of hPLAP expression in striated muscle across different serotypes was similar to that seen in IM injected animals. In heart, rAAV6 generated the highest level of hPLAP expression (Figs. 3a,b), followed by heparin-binding rAAV1 E531K and non-binding rAAV6 K531E. Skeletal muscle hPLAP expression exhibited a similar trend (Figs. 3b,c). In all striated muscle analyzed, rAAV1 generated the least hPLAP expression. hPLAP expression was also detectable at low levels in kidney and intestine (data not shown).

Figure 3.

Systemic injection: hPLAP protein expression profile and vector genome quantification. (a) Representative cross-sections of hearts collected from injected animals and stained for alkaline phosphatase activity. Injected serotype is indicated above each image. The levels of hPLAP activity [in relative light units (RLU) per μg protein] (b) and persistence of vector genomes [vg per μg total DNA] (c) were measured in select tissues following intravenous injection of 2 x1012 vector genomes into adult mice. Data are presented as mean values ±SEM. (n=3) Scale bar: 1 mm. *, p<0.05 **, p<0.01

Heparin binding decreases transduction efficiency and restricts vector dissemination following direct delivery to CNS

Transduction of the central nervous system (CNS) is an important component of gene therapies designed to target neuronal tissue for treatment of neuromuscular and CNS disorders. However, the blood-brain barrier is a significant obstacle to viral entry into the CNS via the vasculature. Thus, it is important to evaluate AAV-mediated transduction of neuronal tissues using a method of delivery that effectively bypasses this barrier. Therefore, we injected mice bilaterally into the dorsal striatum with one of each of the four heparin-binding variants and assayed for hPLAP expression and vector genome copy number (Fig. 4 and Supplemental Fig. 1). We observed a transduction profile that was opposite of that seen in striated muscle. rAAV1 exhibited the highest level of transduction, followed by rAAV6 K531E and rAAV1 E531K. rAAV6 achieved the lowest levels of transduction within the striatum (Fig. 4). As seen previously in the case of striated muscle transduction, attenuation of heparin binding dramatically altered the transduction efficiency of rAAV6. Interestingly, the serotypes that bound heparin exhibited a spatial pattern of transduction that was restricted to the vicinity of the terminal portion of the needle tract (Supplemental Fig. 1). In contrast, the pattern of transduction seen with the non-binding serotypes was diffuse and spread well beyond the path of the needle. This suggests that within the CNS, an affinity for heparin may restrict vector dissemination and reduce transduction efficiency.

Figure 4.

Intracranial injection: hPLAP protein expression profile and vector genome quantification. Animals received 1 ×109 vector genomes in 0.75 μl, administered bilaterally within the dorsal striatum via stereotaxic injection. hPLAP activity [in relative light units (RLU) per μg protein] (a) and quantification of vector genomes [vg per μg total DNA] (b) are reported. Data are presented as mean values ±SEM. (n=5) Scale bar: 50 μm. * indicates statistically significant difference compared to AAV6 (p<0.05).

Discussion

AAV1 and AAV6 have demonstrated a strong tropism for skeletal muscle and neuronal tissue that makes them particularly suitable for treatment of neuromuscular disorders. However, they differ significantly in their affinity for heparin, which impacts vector isolation and ease of purification. In addition, it has been suggested that the ability to bind heparin may influence interactions with antigen-presenting cells.²⁶ These factors are important considerations that have generated significant interest in clarifying whether or not heparin binding is advantageous or detrimental in potential gene therapy vectors.

The utility of heparin interactions in relation to viral infectivity has been well established. Numerous viruses exploit the binding potential of HSPG, a close analog of heparin, to facilitate cellular attachment and entry. Examples include herpes simplex virus types 1 and 2²⁷, human immunodeficiency virus²⁸, dengue virus²⁹, and cytomegalovirus.³⁰ AAV2 directly interacts with membrane-bound HSPG as a primary receptor for viral entry and demonstrates a strong affinity for heparin.²⁰ AAV6 has also been shown to bind heparin, but at a lower affinity than AAV2.¹⁷ Interestingly, the transduction efficiency of AAV2 is inhibited by soluble heparin, but that of AAV6 is not.¹⁷ In addition, the residues that mediate heparin binding in AAV2 and AAV6 are found at different positions of the AAV capsid protein.^19,21 These factors suggest that the heparin-binding sites of AAV2 and AAV6 may possess distinct functional characteristics, and this is further supported by our results from the elution profile of rAAV2/6 chimeric capsids. rAAV2/6 chimeric capsids elute at a higher salt concentration that is distinct from either of the parental capsids, suggesting that the unique heparin-binding sites are acting cooperatively to increase heparin affinity. This observation may be related to the structural organization of the capsid. The heparin-binding region of AAV2 is thought to form a continuous, basic patch on the surface of the capsid, which facilitates electrostatic interaction with the negatively charged sulfate and carboxyl groups of HSPG.^19,21 Cryoelectron microscopy and three-dimensional image reconstruction examining the interaction between AAV2 and heparin suggest that heparin binding may induce a conformational change in the AAV2 capsid that facilitates interaction with co-receptors.^31,32 It is possible that the heparin-binding sites within the assembled chimeric capsid are positioned in a manner that increases the strength of the electrostatic interaction. Further structural analysis and crystallography studies of chimeric capsids will be needed to clarify the mechanism behind these observations.

In this study, we sought to characterize the influence of heparin binding on the transduction profile of rAAV1 and rAAV6. Thus far, the majority of research has focused on heparin binding in AAV2 and has demonstrated the importance of HSPG in this serotype. Non-binding mutants of rAAV2 exhibit reduced infectivity in vitro.^21,33 Low infectivity correlated with reduced cell binding to the viral capsid, which is consistent with previous data regarding the role of HSPG as a known receptor for AAV2.²⁰ Significant differences in tissue tropism between heparin-binding and non-heparin-binding rAAV2 were also observed in vivo²¹, and loss of heparin binding has been shown to significantly reduce tropism for striated muscle in rAAV2 mutants. Far less is known about the role that heparin binding serves in AAV6. Our results show that the unique lysine residue at position 531 in the AAV6 capsid plays a critical role in mediating high-efficiency transduction of striated muscle following both IM and IV administration of vector. Both rAAV6 and heparin-binding rAAV1 E531K demonstrate robust transduction of striated muscle in comparison to their non-heparin-binding counterparts. This is in contrast to prior studies conducted with AAV2, in which injection of wild-type capsids preferentially transduced both heart tissue and liver, but non-heparin-binding mutants transduced cardiac tissue to a much higher degree.²¹ Interestingly, rAAV6 achieved higher transduction in striated muscle than rAAV1 E531K, indicating that the other unique amino acids within the AAV6 capsid contribute to enhanced muscle transduction. These residues may also account for some of the other differences in biodistribution seen between the different vectors. The mechanism behind these data remains to be determined, but may be related to differences in extravasation, cell internalization, or intracellular processing of the viral capsid and genome.¹

Our results demonstrate that heparin-binding capsids consistently transduce striated muscle with higher efficiency when compared with non-binding capsids, whether delivered by IM or IV injection. However, important differences were observed in regards to CNS transduction. Following intracranial injection, non-binding capsids achieved a higher level of transduction when compared to heparin-binding capsids (Fig. 4). Furthermore, rAAV6 and rAAV1 E531K transduction was restricted to the terminal portion of the needle tract (Supplemetal data), suggesting that heparin binding correlates with restricted vector dissemination within the CNS. This is consistent with previous data suggesting that dissemination of rAAV2 within the CNS can be enhanced with co-infusion of heparin.^33a

A variety of HSPGs are expressed extensively within the CNS and other tissues.^34,35 It is possible that patterns of glycosaminoglycan sulfation unique to the dorsal striatum may promote interaction between extracellular matrix proteins and heparin-binding rAAV capsids in a manner that prevents diffusion of the vector and reduces neuronal transduction. However, HSPGs are also found extensively within striated muscle.³⁵ Thus, it is unclear how a model based purely on diffusion of vector would explain the advantageous effects of heparin binding in striated muscle transduction. A thorough characterization of the diverse array of HSPGs in striated muscle versus the dorsal striatum may shed more light on this topic and perhaps identify potential HSPG motifs that are permissive or restrictive to AAV transduction and dissemination. However, the contrasting results obtained in the CNS versus skeletal muscle emphasize that transduction efficiency can be both delivery- and tissue-dependent. These data underscore the importance of considering heparin-binding capability in the rational design of targeted vectors, as a single amino-acid change can be quite advantageous under certain conditions, but deleterious in others.

In addition to evaluating hPLAP expression, we also quantified the number of vector genomes in surveyed tissues. It is interesting to note that the pattern and ratio of hPLAP expression to number of vector genomes was not always equivalent, indicating that the efficiency of genome expression differed across serotypes and tissues. In striated muscle, there was a general trend towards increased expression efficiency from genomes packaged in rAAV6 capsids. The opposite trend was apparent in neuronal tissue. These serotype- and tissue-dependent differences in expression efficiency have been shown in previous serotype comparison studies and may reflect differences in internalization, intracellular trafficking, and uncoating of vector genomes.^3,36 It has been clearly established that intact capsid can persist intracellularly^36,37, and serotype-specific patterns of intracellular localization have been demonstrated. It is intriguing to consider the possibility that single amino acid changes within the heparin-binding site could alter intracellular processing and kinetics of genome expression in rAAV6.

Materials and Methods

Cloning of AAV1 and generation of AAV6 mutants

pDG6 is an AAV helper plasmid that contains the rep gene from AAV2 and the cap sequences from AAV6. The cap sequence is flanked by a unique 5’ SwaI site and a 3’ ClaI site. pDG6 was digested by these enzymes and the 2.3 Kb fragment obtained was subcloned into a modified pBluescript plasmid containing a SwaI site.³⁸ The resulting construct was named AQ1 AAV6. pH21-AAV1, a plasmid containing the rep sequence from AAV2 and the cap sequence from AAV1 was kindly provided by Dr. W Xiao. Both AAV6 and AAV1 cap open reading frames (ORFs) present a unique 5’ StuI site and a unique 3’ AgeI site. Thus, AQ1 AAV6 and pH21 were digested with StuI and AgeI. The 2 Kb fragment from pH21, which encodes for the 6 different amino acids between the AAV1 and 6 cap sequences was then cloned into the AQ1 AAV6 backbone. The resulting construct was named AQ2 AAV1. AQ2 AAV1 was digested with SwaI and ClaI and cloned back into the pDG6 vector. The resulting plasmid contains the AAV2 rep ORF and AAV1 cap ORF.

Plasmid AQ1 AAV6 was used as the template for all mutant constructions. Mutagenesis was achieved by using the Quicksite mutagenesis kit (Stratagene, La Jolla, CA) following the manufacturer's instructions. Briefly, a pair of primers containing the desired mutation and 10-20 flanking bases was annealed at 45°C (20-mer) or 55°C (40-mer). PCR was performed to generate mutant complementary strains and PCR products were then treated with Dpn1. Parental strain-free plasmids were transformed into XL-1 Gold Escherichia coli and plated in Luria-Bertoni broth supplemented with ampicillin. Plasmid DNA from selected colonies was sequenced to assure it contained only the desired mutation. Once verified, cap ORF was cloned into the helper plasmid.

AAV vector production

All vectors were prepared as described previously, using a double co-transfection method into HEK-293 cells, followed by CsCl gradient purification.³⁹ Chimeric capsids were generated via co-transfection of the corresponding packaging plasmids at a 1:1 ratio. Vector titer was determined by Southern blot.

Studies in vivo

Weight-matched, male C57Bl/6 mice approximately 3 months in age were utilized for all injections in this study. For IM injections, mice (n = 3) were anesthetized with isofluorane and injected bilaterally into the tibialis anterior with either 1×10⁸, 1 ×10⁹, or 1 ×10¹⁰ vector genomes diluted in 25 μl of PBS. For systemic injection, mice (n = 3) were administered 2 ×10¹² vector genomes in a total volume of 200 μl intravenously via the tail vein. For intracranial injection, animals (n = 5) were anesthetized with isoflurane and injected with 1 ×10⁹ vector genomes in 0.75 μl of PBS. Bilateral injections were performed using stereotaxic coordinates (x = ±2.00 mm, y = 0.8 mm, and z = −3.60 mm from bregma). Animals were allowed two weeks to recover from all procedures before they were euthanized, and tissues were collected for histochemical analysis and vector genome quantification.

Tissue preparation and analysis of hPLAP expression

Tissues from IM or IV injected animals were either snap frozen in blocks of OCT for histochemical analysis or frozen in liquid nitrogen for quantification of hPLAP expression. Blocked tissues were cryosectioned and stained for hPLAP expression as previously described.⁴⁰ Liquid nitrogen frozen tissues were ground in using a mortar and pestle on dry ice. Homogenized tissues were re-suspended in tissue lysis buffer (0.05% Sodium deoxycholate, 50 mM Tris pH 7.5, 150 mM NaCl, 1% protease inhibitor cocktail [Sigma]). Lysates were analyzed for hPLAP expression using a Phospha-Light chemiluminescence kit (Applied Biosciences, Foster City, CA) and quantified using a Victor-3 luminometer. Statistical analysis was performed via ANOVA followed by a Tukey post-test using GraphPad Prism Software (La Jolla, CA).

Intracranially-injected animals were anesthetized with pentabarbitol and transcardially perfused with PBS followed by 4% paraformaldehyde (PFA). For histochemical analysis, whole brains (n=2 injection sites per serotype) were post-fixed overnight in PFA, cryoprotected in 30% sucrose, blocked in OCT, sectioned, and stained for hPLAP expression, as described above. From the remaining animals in each cohort, striatum was dissected and standardized punch biopsies were taken from the site of injection within the dorsal striatum. Biopsies were homogenized in tissue lysis buffer for chemiluminescent analysis and vector genome quantification. Both protein expression and vector genomes were quantified from each biopsy sample.

Vector genome quantification

DNA was isolated from tissue lysates using a DNeasy blood and tissue kit (Qiagen, Valencia, CA) according the manufacturer's guidelines. Genome quantification was performed utilizing a SV40 polyA-specific probe, as previously described.⁷