The AAV9 receptor and its modification to improve in vivo lung gene transfer in mice (2024)

Initial attempts to identify cellular receptors that promote AAV9 transduction focused on a potential role of glycans with terminal SA. Previous studies have shown that α-2,3 and α-2,6 N-linked SA facilitates binding and transduction by AAV1 and -6 (4), while α-2,3 N-linked SA is important for transduction by AAV5 (6). Several previously described AAVs (serotypes 1, 2, 5, and 6) and the novel AAV serotypes 7, 8, 9, and -rh32.33 (8–10) were screened for binding to the CHO cell line Pro-5 with and without treatment with exo-α-sialidase from Vibrio cholerae (neuraminidase [NA]). This NA is capable of cleaving almost all known CHO, murine, and human SA linkages. Results with the first-generation AAVs were as expected, in that NA pretreatment (Figure 1A) had no effect on binding of AAV2, and binding of AAV1, AAV5, and AAV6 was diminished by at least 2 logs (Figure 1B). Studies with the novel AAVs failed to show an effect of NA treatment on binding of AAV7, -8, and -rh32.33, while AAV9 showed an unexpected 2.5-log increase in binding (Figure 1B). The impact of NA treatment on other cell lines commonly used in AAV studies, i.e., HEK293 (Figure 1C) and Huh-7 (Figure 1D), was evaluated. For both cell lines, NA had no effect on AAV2 binding, while it decreased binding of AAV6 and increased binding of AAV9 to the same extent as observed with Pro-5 cells. The electrophoretic mobility of the 3 primary capsid proteins of AAV9 — VP1, VP2, and VP3 — was unaffected when purified vector was treated with various glycosidases (Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI57367DS1). This suggests that the capsid is not glycosylated, which is consistent with studies of other AAV serotypes (16) and supports the hypothesis that the NA-mediated effect is due to modification of a cellular receptor, not the capsid.

We speculated that the presence of terminal SA on an oligosaccharide of a receptor either inhibited binding to AAV9 or that the removal of SA exposed a molecule that enhanced binding. The most likely candidate would be β-1,3 or β-1,4 galactose, which is the penultimate monosaccharide on most SA-rich glycans (Figure 1A). These hypotheses were initially explored using somatic cell mutants of the parent CHO cell line, Pro-5, which are deficient in various enzymes involved with glycosylation by virtue of lectin resistance (17). Lec-2 is deficient in cytidine monophosphate–SA (CMP-SA) Golgi transporter and should have a full complement of N- and O-glycans that are missing terminal SA (Figure 1A). Lec-8 is deficient in uridine diphosphate galactose (UDP-galactose) Golgi transporter and should produce N- and O-glycans that are missing both SA and galactose saccharides (Figure 1A). Vectors based on AAV2, -6, and -9 expressing firefly luciferase (ffLuc) were incubated with Pro-5, Lec-2, and Lec-8 cells and analyzed for binding and transduction (Figure 2, A and B). Binding and transduction of AAV2 was the same in all 3 cell lines and decreased with AAV6 in Lec-2 and Lec-8 cells relative to Pro-5 cells, which was expected based on previous studies demonstrating the importance of SA in facilitating AAV6 entry (4). For AAV9, binding and transduction increased substantially in Lec-2 but decreased to baseline levels in Lec-8 cells, which is more consistent with the hypothesis that binding to terminal β-galactose facilitates uptake, rather than that SA inhibits uptake. The concurrence of binding with transduction suggests that binding to terminal galactose is an important step in AAV9 transduction. The impact of eliminating SA was greater on binding than on transduction, suggesting that post-entry steps may also limit transduction.

The role of terminal galactose in AAV9 binding was further studied in Pro-5 cells that were pretreated with NA and then cocultured with lectins of different specificities (Figure 2C, binding; Figure 2D, transduction). The only lectins that blocked binding of AAV9 were Erythrina cristagalli lectin (ECL), which recognizes β-1,4 galactose, and Ricinus communis agglutinin I (RCA I), which recognizes several types of β-galactose linkages, with β-1,4 galactose showing the highest affinity; neither affected binding of AAV2. Lectins that recognize α-galactose (Griffonia simplicifolia lectin I isolectin B₄ [GSL B₄]), α-1,3 and α-1,6 mannose (Hippeastrum hybrid lectin [HHL]), and α-fucose (Lotus tetragonolobus lectin [LTL]) did not interfere with binding of AAV2 or AAV9. Wheat germ agglutinin (WGA) had a slight effect on AAV9 binding, likely due to its interaction with N-acetylglucosamine, which is commonly bound to galactose. Lectin inhibition of AAV2 and AAV9 transduction confirmed the binding results, except in the case of RCA, which was not informative since it was toxic to cells.

A final confirmation of the glycan specificity of AAV9 uptake was performed in Pro-5 cells pretreated with NA to cleave terminal SA or a combination of NA and β-gal that would remove both NA and β-galactose saccharides (Figure 1A). Binding and transduction with AAV2 was unaffected by enzyme pretreatment, while NA enhanced transduction of AAV9 as described earlier, an effect that was reversed by subsequent treatment with β-gal (Figure 2, E and F).

The glycan-capsid interactions of AAV9 were further interrogated using a glycan microarray (GMA) composed of 465 different natural and synthetic mammalian glycans that contains 6 replicates for each glycan. A similar strategy was used to verify the binding of AAV1 to sialylated glycans as determined by biochemical and molecular approaches (4). In our analysis, each glycan was evaluated for binding as measured by relative fluorescence units (RFU) in terms of the mean of 4 replicates within the array ± 1 SD (the highest and lowest binding within the 6 replicates were eliminated). Variation within the 4 replicates was assessed as the coefficient of variation (%CV), which was considered low if it was less than 20. The selection of glycans with specific binding affinity for AAV9 capsids was based on two independent criteria: high overall total binding as measured by RFU and low variation among the 4 replicates as assessed by %CV. The highest binding glycans were arbitrarily defined as those with mean RFU values that fell within 3 SD of the highest binding glycan, 415. Of the 4 glycans that fulfilled the criteria of high binding affinity, 3 showed high specificity as indicated by a %CV less than 20 (Figure 3). These 3 glycans, 415, 297, and 399, demonstrated sufficient affinity and specificity to be considered potential receptors for AAV9 binding. The 3 glycans selected as being specific for binding to AAV9 based on total binding and specificity of binding were all shown to have terminal galactose (Gal) with either β-1,4 linkages (i.e., 415 and 399) or a β-1,3 linkage (297, Figure 3). Each glycan contains Galβ1-4(Fucα1-3)GlcNAc linked via β1-3 or β1-6 to GalNAc in its structure, suggesting a structural context of the SA-deficient glycan for binding to AAV9. Reconciling the binding data with cell-based binding/transduction experiments confirmed the role of terminal β-galactose linkages in tropism of AAV9.

The relevance of the novel β-galactose–mediated uptake pathway for AAV9 to in vivo gene delivery was evaluated in the context of lung-directed gene transfer. Previously we showed that delivery of AAV9 vectors into lungs of mice resulted in efficient transduction of epithelial cells of the alveoli but no transduction of epithelial cells of the conducting airway, which is the preferred target for the treatment of cystic fibrosis (15). Studies were performed in mice to determine whether transduction of conducting airway with AAV9 could be enhanced by pretreating mouse airways with a formulation that contained NA. Comparisons were made to mice administered AAV6, which in previous studies showed efficient in vivo transduction of conducting airway (18). Based on in vitro studies with cell lines, this in vivo transduction may be dependent on binding to SA (4). Figure 4, A–D, shows representative histochemical analyses of lung tissue, while Figure 4E presents a morphometric analysis of lacZ-positive cells in conducting airway. Lung delivery of AAV9 in the absence of NA demonstrated the expected pattern of alveolar-restricted transduction (Figure 4, A, B, and E). Administration of NA into the lung 1 hour prior to or at the time of AAV9 administration yielded a very different pattern, with high-level transduction in both the alveolar and conducting airway epithelial cells (Figure 4, A, B, and E). The high-level transduction of conducting airway obtained with AAV6 was completely eliminated when animals were treated with NA, confirming the dependence of in vivo transduction on binding to SA (Figure 4, C–E). The consequences of airway delivery of NA on the abundance and distribution of glycan structure were studied directly by staining lung sections with fluorescence-labeled lectins that bind terminal galactose residues with β linkages (RCA) or terminal SA residues (SNA). Prior to NA treatment, staining with RCA was limited to alveolar cells and the basolateral region of conducting airways (Figure 5, A and C). A similar pattern was observed with SNA, although the apical surface of the conducting airway was also stained (Figure 5, B and D). Treatment with NA had the most dramatic effect on the staining of conducting airway epithelial cells, which was substantially increased with respect to RCA (Figure 5, E and G) and moderately reduced with respect to SA (Figure 5, F and H). Higher-resolution microscopic studies were performed to determine whether RCA binding localized to the plasma membrane or the overlying mucus layer. Lung sections from animals pretreated with NA were evaluated for colocalization of RCA staining with immune localization of α-tubulin, which delineates the cilia from the apical surface of airway epithelial cells (Figure 5, I–K). These studies demonstrated binding of RCA to the cell surface of epithelial cells in the conducting airway of NA-pretreated lungs.

A major impediment to transduction following intravascular (IV) administration of AAV vectors is the physical barrier of the contiguous endothelium and basal lamina of the microcirculation. The one exception is the liver, which contains fenestrated endothelia, allowing direct access of vector in blood to the hepatocytes. AAV9 is unique in that it partially overcomes this barrier, allowing targeting of skeletal and cardiac muscle and, to a lesser extent, cells of the central nervous system (13, 14, 19). We analyzed lacZ transduction following IV administration of AAV9 vector at doses of 10¹¹ and 10¹² genome copies (GC)/mouse (Figure 6, A–C). This was correlated with the presence of terminal β-galactose through staining with RCA (Figure 6, A–C). As has been described, liver was efficiently transduced even at low doses of vector (Figure 6C and ref. 20); it was possible to efficiently transduce cardiac and skeletal muscle at higher vector doses (Figure 6, A and B, and ref. 19). Surfaces of muscle fibers from skeletal and cardiac tissues demonstrated high levels of terminal β-galactose, as evidenced by binding to RCA (Figure 6, A and B). Hepatocytes demonstrated lower levels of RCA binding, although endothelial surfaces of hepatic vessels stained brightly (Figure 6C).

Several AAVs such as AAV5 and AAVrh.10 are capable of efficient transduction of cells of the CNS following direct injection, although transduction is limited to regions in proximity to the needle injection site (7, 21). AAV9 has demonstrated the unique properties of detectable transduction of the CNS following intravenous injection (13, 14). We evaluated brain for the presence of β-galactose terminal glycans via binding to RCA, which showed high levels in small vessels as demonstrated by colocalization with the endothelial-specific marker CD31 (Figure 6D).