A toolbox of astrocyte-specific, serotype-independent adeno-associated viral vectors using microRNA targeting sequences

Abstract

Astrocytes, one of the most prevalent cell types in the central nervous system (CNS), are critically involved in neural function. Genetically manipulating astrocytes is an essential tool in understanding and affecting their roles. Adeno-associated viruses (AAVs) enable rapid genetic manipulation; however, astrocyte specificity of AAVs can be limited, with high off-target expression in neurons and sparsely in endothelial cells. Here, we report the development of a cassette of four copies of six miRNA targeting sequences (4x6T) which triggers transgene degradation specifically in neurons and endothelial cells. In combination with the GfaABC1D promoter, 4x6T increases astrocytic specificity of Cre with a viral reporter from <50% to >99% in multiple serotypes in mice, and confers astrocyte specificity in multiple recombinases and reporters. We also present empty vectors to add 4x6T to other cargo, independently and in Cre/Dre-dependent forms. This toolbox of AAVs allows rapid manipulation of astrocytes throughout the CNS, is compatible with different AAV serotypes, and demonstrates the efficacy of using multiplexed miRNA targeting sequences to decrease expression in multiple off-target cell populations simultaneously.

Fig. 1. Low astrocytic viral specificity can be enhanced with miRNA targeting sequences.

a Systemic delivery of two different GFAP-based promoters (GFAP, n = 3 mice; and GfaABC1D, n = 4 mice), co-injected with PHP.eB::CAG-flex-GFP, yield high levels of transduction of non-astrocytic cells. Sox9, astrocytes; NeuN, neurons. b Characterization of PHP.eB::GfaABC1D-Cre delivery in transgenic (Tg) Ai14-tdTomato mice: while the majority of tdTomato⁺ cells are astrocytes, there are high numbers of transduced neurons (NeuN⁺) and low numbers of endothelial cells (CD31⁺; yellow arrow); n = 4 mice. c Schematic diagram of miRNA targeting approach. d Changes in astrocytic selectivity by the addition of multiple copies of single miR targeting sequences (n = 3 mice); compared to GfaABC1D-Cre (n = 4 mice) (one-way ANOVA, Dunnett’s multiple comparisons test, P < 0.0001, F = 24.76, df=18. All vs no miR: miR124T.4x, four copies, ****P < 0.0001; miR137T.2x, two copies, *P = 0.0263; miR329T.2x, two copies, P = 0.4348; miR369T.2x, two copies, P = 0.2680; miR431T.2x, two copies, *P = 0.0118.) The dotted line denotes astrocyte specificity in GfaABC1D-Cre with no miR cassettes. e Schematic diagram of final AAV::GfaABC1D-Cre-4x6T plasmid. f Enhanced astrocyte specificity with PHP.eB::GfaABC1D-Cre-4x6T (n = 4 mice) vs PHP.eB::GfaABC1D-Cre-miR124T.4x (n = 3 mice), using a PHP.eB::CAG-flex-GFP reporter (two-tailed unpaired t test, *P = 0.0103, t = 4.002, df=5). Non-astrocytes expressing GFP after transduction with PHP.eB::GfaABC1D-Cre-miR124T.4x include both neurons (red arrowhead) and endothelial cells (yellow arrows). g Further enhancement of astrocyte specificity using PHP.eB::GfaABC1D-Cre-4x6T with a transgenic Ai14-tdTomato reporter (n = 4 mice) or a PHP.eB::CAG-flex-lck-smV5-4x6T reporter (n = 3 mice) rather than PHP.eB::CAG-flex-GFP (n = 4 mice). Using a transgenic mouse line or including the 4x6T cassette on the flex reporter increases astrocyte specificity (one-way ANOVA, Tukey’s multiple comparisons test, P = 0.0144, F = 7.548, df=10; flex-GFP vs Ai14, *P = 0.0249; flex-GFP vs flexV5-4x6T, *P = 0.0259). Mice were injected retroorbitally at 2–5 months old and euthanized 2 weeks post injection. Titers: GFAP-Cre, 5 × 10¹⁰ vg/mouse + CAG-flex-GFP, 2 × 10¹¹ vg/mouse.; GfaABC1D-Cre, GfaABC1D-Cre-(all miRs): 5 × 10¹¹ vg/mouse + CAG-flex-GFP or CAG-flex-smV5, 5 × 10¹¹ vg/mouse. Source data are provided as a Source Data file. All data are presented as mean ± SEM. Scale bars: 40 μm.

Fig. 2. 4x6T cassette increases astrocyte specificity at high titers and across serotypes.

a High-titer (3 × 10¹² vg/mouse) systemic delivery of PHP.eB::GfaABC1D-Cre-4x6T in Ai14-tdTomato mice yields 99.67% astrocyte specificity (% of tdTomato⁺ cells that were Sox9⁺); the efficiency of transduction (% of Sox9⁺ cells transduced) was 65.66% of Sox9⁺ cells in the cortex. Scale bar: 40 μm. b Schematic of viral delivery approaches: retroorbital injection leads to lower levels of brain-wide transduction, while direct intracortical injection leads to higher levels of transduction within a narrow region around the injection site. c High-titer direct intracortical injection (500 nl of 1 × 10¹² vg/ml per virus) of PHP.eB::GfaABC1D-Cre-4x6T and PHP.eB::CAG-flex-lck-smV5-4x6T shows high levels of neuronal contamination. Scale bar: 30 μm. d High-titer direct intracortical injection of PHP.eB::GfaABC1D-Cre-4x6T and AAV2/5::CAG-flex-lck-smV5-4x6T into Ai14-tdTomato mice shows predominantly astrocytic expression of the tdTomato reporter but more astrocyte-specific expression of the AAV2/5::flexV5-4x6T reporter. Two-tailed unpaired t test, *P = 0.0340; t = 3.165, df=4. Scale bars: 30 μm. e High-titer direction intracortical injection of GfaABC1D-Cre-4x6T packaged in different serotypes, co-injected with AAV2/5::CAG-flex-lck-smV5-4x6T into Ai14-tdTomato mice. Serotype impacts astrocyte specificity, with highest specificity with AAV2/5. The presence of the 4x6T on Cre, on the reporter virus, or both increases astrocytic specificity, with the highest specificity seen when both elements of a Cre/reporter system are tagged with 4x6T. Within serotypes: GfaABC1D-Cre vs all 4x6T conditions, one-way ANOVA with Tukey’s multiple comparisons test, ****P < 0.0001. AAV2/1: F = 320.2, df=11. AAV2/5: F = 124.4, df=11. AAV2/9: F = 259.0, df=11. Source data are provided as a Source Data file. Scale bars: 30μm. All data presented as mean ± SEM; n = 3 mice per condition; all mice 2–5 months old, euthanized 2 weeks post injection.

Fig. 3. 4x6T cassette confers astrocyte specificity across the lifespan and in reactive astrocytes.

a Systemic injection in mice as young as postnatal days 1–2 (temporal vein injection; 1 × 10¹⁰ vg/mouse PHP.eB::GfaABC1D-Cre-4x6T + 2 × 10¹⁰ vg/mouse PHP.eB::CAG-flex-lck-smV5-4x6T; n = 3 mice) and as old as 28 months (retroorbital injection; 2 × 10¹¹ vg/mouse PHP.eB::GfaABC1D-Cre-4x6T and PHP.eB::CAG-flex-lck-smV5-4x6T; n = 4 mice) show highly astrocyte-specific viral expression patterns. Scale bars: 40 μm. b Systemic injection in P1-2 pups yields transduced radial glia (yellow arrows) in the dentate gyrus, which is not readily observed in animals injected at 2–5 months; similar observations in four mice/age group. Dotted gray lines, granule cell layer. Scale bars: 100 μm. c High-titer retroorbital PHP.eB::GfaABC1D-Cre-4x6T injection (3 × 10¹² vg/mouse) in adult Ai14-tdTomato mice (3–4 months old, n = 3 mice) results in high astrocyte specificity in the dentate gyrus. The majority of Aldh1l1^-tdTomato⁺ cells are found in the SGZ and granular layers (combined totals, 3 mice, 2 sections/mouse; total = 58 cells). Scale bars: left, 100 μm; right, 20 μm. d 4x6T cassette maintains high levels of astrocytic specificity after stroke (n = 3 mice), although morphologically distinct V5⁺Aldh1l1⁻ can be found near the infarct border. Left, 50 μm scale bar; right, 20 μm scale bar. Yellow box denotes area of higher magnification at right; yellow arrows indicate V5⁺Aldh1l1⁻ cells. V5⁺Aldh1l1⁻ cells: mean distance of 55 μm ± 14.81 from infarct border. Source data are provided as a Source Data file. Bar graphs are presented as mean ± SEM. All mice were euthanized 2 weeks post injection.

Fig. 4. Astrocyte specificity of 4x6T cassette is preserved for long time periods and across CNS regions.

a Astrocyte specificity 6 months after injection (retroorbital injection, 5 × 10¹¹ vg/mouse PHP.eB::GfaABC1D-Cre-4x6T; 5 × 10¹¹ vg/mouse PHP.eB::CAG-flex-lck-smV5-4x6T) into young adult (2–5-month-old) mice is preserved in the cortex (scale bar: 40 μm) and to a slightly lesser degree in the hippocampus. Hippocampus, left: entire structure (scale bar: 150 μm); yellow box shows the region of higher magnification on the right (scale bar: 20 μm). Yellow arrow: example of neuron in the dentate gyrus. Astrocyte specificity is higher in the cortex than the dentate gyrus sub-region (*P = 0.0134); Kruskal–Wallis test, Dunn’s multiple comparisons test, P = 0.0024, Kruskal–Wallis statistic 8.346). Mean ± SEM; n = 4 mice per brain region. Source data are provided as a Source Data file. b Sagittal section of mouse brain 6 months after virus injection (scale bar: 1 cm), with higher magnification examples of the cerebellum, thalamus, and olfactory bulb (scale bars: 100 μm), showing high levels of colocalization of V5 with Sox9. c Coronal section of mouse spinal cord 6 months after virus injection (scale bar: 100 μm); yellow box denotes area of higher magnification section on the right (scale bar: 20 μm), showing high V5/Sox9 colocalization; similar observations in four mice in each CNS region.

Fig. 5. Transcriptional analysis of astrocytes with viral transduction with and without 4x6T cassette (TRAP).

a Immunohistochemical analysis of RiboTag⁺ cells (hemagglutinin HA⁺ ribosomal tag), colocalized with astrocytic marker Aldh1l1 and astrocytic reactivity marker GFAP. Astrocytic specificity is high in Aldh1l1-CreERT2 transgenic mice and systemic delivery of PHP.eB::GfaABC1D-Cre-4x6T with no overt evidence of astrocyte reactivity (GFAP% coverage). Specificity remains high with intracortical delivery of PHP.eB::GfaABC1D-Cre-4x6T and decreases with intracortical delivery of AAV2/5::GfaABC1D-Cre; this route of viral delivery shows some evidence of astrocyte reactivity by GFAP immunoreactivity. Scale bars: 50 μm. Mean ± SEM; n = 4 mice per cohort. Astrocyte specificity, HA⁺Aldh1l1⁺/HA⁺: one-way ANOVA, Holm–Sidak’s multiple comparisons test, P < 0.0001, F = 77.01, df=15; Aldh1l1-CreERT2 vs AAV2/5::GfaABC1D-Cre ****P < 0.0001. GFAP % area coverage: Aldh1l1-CreERT2 = 2.33% coverage ± 0.64; PHP.eB::GfaABC1D-Cre-4x6T systemic, 2.77% coverage ± 0.55; PHP.eB::GfaABC1D-Cre-4x6T intracortical, 13.61% coverage ± 4.71; AAV2/5::GfaABC1D-Cre, 12.52% coverage ± 2.90; one-way ANOVA, Holm–Sidak’s multiple comparisons test, P = 0.0011, F = 10.54, df=15; Aldh1l1-CreERT2 vs PHP.eB::GfaABC1D-Cre-4x6T intracortical *P = 0.0431; Aldh1l1-CreERT2 vs AAV2/5::GfaABC1D-Cre *P = 0.0481. Source data are provided as a Source Data file. b Relative levels of enrichment and de-enrichment of canonical genes for astrocytes, neurons, oligodendrocytes, microglia, and endothelial cells in IP-vs-input samples. c Differentially expressed genes in IP samples: different viral cohorts vs Aldh1l1-CreERT2 IP samples, and systemic vs cortical 4x6T samples. FDR < 0.05, average FPKM across all IP samples >1. d Cell-type specificity of differentially expressed genes in IP samples: genes in (c) that are represented in the top 1000 specific genes for astrocytes, endothelial cells, microglia, neurons, or oligodendrocytes. e Venn diagram showing the overlap of which neuron-specific genes from (d) are upregulated in different viral cohort IP samples vs Aldh1l1-CreERT2 samples. f Top ten most significantly upregulated and downregulated Gene Ontology gene sets (FDR < 0.05) in ranked IP transcriptomes of each of the viral cohorts, calculated with Gene Set Enrichment Analysis; some gene sets overlapped, particularly for 4x6T cohorts, and only seven gene sets passed FDR < 0.05 for intracortical 4x6T. In cases where the FDR was 0, scores were reassigned as 0.0001 to visualize the relative -logFDR. g Numbers of genes potentially regulated by miRNAs that comprise the 4x6T cassette (5409 total) that show evidence of 4x6T-specific regulation in IP samples (FDR < 0.05 in either 4x6T cohort vs Aldh1l1-CreERT2 cohort; not differentially regulated in AAV2/5 cohort vs Aldh1l1-CreERT2 cohort). Note: none of the 5409 genes show evidence of 4x6T-specific regulation in input samples. n = 4 mice per cohort; 2–4 months old, euthanized 2 weeks after final tamoxifen administration and 18 days postvirus injection. Titer: retroorbital PHP.eB::GfaABC1D-Cre-4x6T, 1 × 10¹² vg/mouse; intracortical, PHP.eB::GfaABC1D-Cre-4x6T or AAV2/5::GfaABC1D-Cre: 500 nl of 1 × 10¹² vg/ml in each of three sites.

Fig. 6. Non-recombinase astrocyte specificity with and without the 4x6T cassette.

a PHP.eB::GfaABC1D driving a spaghetti monster reporter (smMyc) shows high levels of non-astrocytic contamination, while adding a 4x6T cassette significantly increases astrocyte specificity (n = 4 mice per cohort, two-tailed t test, ****P < 0.0001, t = 10.73, df=6). The addition of the 4x6T cassette does not affect astrocyte transduction efficiency (two-tailed t test, P = 0.8314, t = 0.2224, df=6). b Astrocytic specificity remains high with spaghetti monster reporter smV5 constructs delivered either via direct intracortical injection or systemically (n = 4 mice per cohort, two-tailed t test, P = 0.2099, t = 1.404, df=6). Adding the 4x6T cassette to other smFPs (FLAG, Myc) also show high astrocyte specificity, although PHP.eB::GfaABC1D-smFLAG-4x6T shows surprisingly low transduction efficiency when delivered systemically. Source data are provided as a Source Data file. All data presented as mean ± SEM. Scale bars: 40 μm. Retroorbital virus delivery: each virus 2 × 10¹¹ vg/mouse; intracortical: 500 nl of 1 × 10¹² vg/ml in each of two sites; 2–4-month-old mice, euthanized 2 weeks post injection. Note: all GfaABC1D-smFP constructs included a WPRE regulatory element to boost transgene expression.

Fig. 7. Astrocyte specificity of the 4x6T cassette with inducible and alternative recombinases.

a Tamoxifen- (ERCreER) and light- (iCreV) inducible forms of Cre show high levels of neuronal background without induction when co-injected with a flexV5 reporter with no 4x6T cassette (PHP.eB::CAG-flex-smV5) but much less background when co-injected with a flexV5-4x6T reporter (PHP.eB::CAG-flex-smV5-4x6T). Scale bars: 100 μm. Astrocyte specificity with a flexV5-4x6T reporter is high with both inducible forms of Cre, but higher in ERCreER (two-tailed unpaired t test, **P = 0.0055, t = 4.233, df=6; n = 4 mice with tamoxifen or light; two mice with no tamoxifen or no light). Light induction of iCreV shows expression across a broad area of cortex ipsilateral to light placement and limited contralateral expression. Light placement: yellow arrow. Scale bar: 500 μm. All: 2–5-month-old mice, euthanized 2 weeks after Cre induction, 4–4.5 weeks after retroorbital virus injection. Titers: ERCreER-4x6T, iCreV-4x6T: 5 × 10¹¹ vg/mouse + CAG-flex-smV5-4x6T: 5 × 10¹¹ vg/mouse. b PHP.eB::GfaABC1D-Dre-4x6T and Dre-dependent reporter PHP.eB::CAG-dDIO-smMyc-4x6T can be used orthogonally with Cre/flex viral systems and show similar levels of astrocytic specificity (scale bar: 50 μm); yellow box shows the region of higher magnification on the right (scale bar: 10 μm); n = 4 mice per recombinase, 2–3-month-old mice, euthanized 2 weeks after retroorbital virus delivery. Titers: Dre-4x6T: 2 × 10¹¹ vg/mouse; Cre-4x6T, dDIO-smMyc-4x6T, flex-smV5-4x6T: 1 × 10¹¹ vg/mouse. Source data are provided as a Source Data file. All data are presented as mean ± SEM. c Schematic diagrams of empty vectors with multiple cloning sequences (MCS) for insertion of other cargo: GfaABC1D-MCS--4x6T for non-recombinase-dependent expression; CAG-flex-MCS--4x6T for Cre-dependent expression; and CAG-dDIO-MCS--4x6T for Dre-dependent expression. Note the presence of a woodchuck hepatitis virus post-transcriptional regulatory element (WPRE); while we omitted this element in the case of recombinase vectors, where high levels of transgene expression were neither wanted nor needed, we have included it in these more general vectors. More detail on transgene components and the full sequences can be found on Addgene.