Regionally distinct GFAP promoter expression plays a role in off-target neuron expression following AAV5 transduction

Parts used

Key findings

• Off-target neuronal expression from the human GFAP promoter is regionally heterogeneous within the cortex and grows over time in some regions but not others. After AAV5-hGFAP-Cre injection into Ai14 tdTomato reporter mice, the proportion of transduced cells that are neurons in the medial prefrontal cortex rises from 17.60 ± 5.53% at day 5 to 51.85 ± 6.09% at day 21 (p = 0.038) while the proportion that are GFAP+ astrocytes falls from 69.29 ± 6.61% to 24.95 ± 8.29% (p = 0.007). In the motor cortex neither changes significantly over the same period — astrocytes hold at 77.17 ± 0.15% to 73.51 ± 4.46% and neurons at 2.41 ± 1.40% to 11.38 ± 4.215% (p = 0.986) — leaving roughly fivefold more transduced neurons in the medial prefrontal cortex than the motor cortex at day 21 (p = 0.010) and roughly threefold fewer transduced astrocytes (p = 0.003).
• Viral titre does not explain regional differences in off-target neuronal expression from the hGFAP promoter. Across AAV5-hGFAP-Cre doses from 1 × 10^9 to 1 × 10^13 GC/mL, doses below 1 × 10^11 GC/mL produce no detectable tdTomato, and above that threshold the percentage of transduced cells that are neurons does not change with increasing titre within either region. The medial prefrontal cortex exceeds the motor cortex at every dose tested, by 2.2 ± 0.2-fold (p = 0.040), 2.7 ± 0.3-fold (p = 0.001) and 4.6 ± 0.5-fold (p = 0.0002) at 1 × 10^11, 1 × 10^12 and 1 × 10^13 GC/mL.
• AAV5 capsid tropism is equivalent between the motor cortex and medial prefrontal cortex, so the promoter rather than the capsid creates the regional difference. Driving Cre from the ubiquitous CAG promoter in the same AAV5 capsid produces no significant regional difference in either total transduced cells (245.7 ± 18.59 versus 343.1 ± 44.53 tdTomato+ cells per section, p = 0.078) or the percentage of transduced cells that are neurons (50.69 ± 4.42% versus 47.35 ± 2.81%, p = 0.541).
• Cortical neurons transcribe Gfap at low levels, and they do so more in the medial prefrontal cortex than the motor cortex, giving those neurons the transcriptional machinery to drive expression from an hGFAP promoter. RNAscope in uninjected mice detects Gfap mRNA in NeuN+ cells in both regions, with significantly less colocalization in the motor cortex (41.49 ± 0.72%) than the medial prefrontal cortex (50.85 ± 3.03%, p = 0.040), and Gfap-expressing neurons of the medial prefrontal cortex carry more transcript puncta each — 2.450 ± 0.1637 versus 1.717 ± 0.1092 per cell (p = 0.0003).
• The human GFAP promoter carries binding sites for transcription factors present in cortical neurons, providing a sequence-level basis for its neuronal activity. Of 225 transcription factor binding sites catalogued in the aligned human GFAP promoter, corresponding to 131 transcription factors, 13 factors are at least modestly expressed in subpopulations of mouse cortical neurons and account for 15 binding sites. Twelve of those 15 cluster at just two loci in the promoter, spanning 160–216 bp and 531–568 bp.
• Intercellular material transfer moves both Cre recombinase and tdTomato protein from astrocytes to neurons in vivo, but too rarely to account for most off-target neuronal reporter expression. Astrocytes isolated from postnatal day 4 Ai14 mice, transduced with AAV5-hGFAP-Cre and transplanted into lox-stop-lox YFP reporter mice survive equally in both regions (184.00 ± 70.12 versus 202.25 ± 30.58 tdTomato+ cells per section, p = 0.819). tdTomato protein transfer yields 1.00 ± 0.71 NeuN+tdTomato+ cells per section in the motor cortex and 8.00 ± 3.49 in the medial prefrontal cortex, and Cre protein transfer yields 4.50 ± 2.40 and 4.25 ± 1.49 NeuN+YFP+ cells per section respectively (p > 0.999).

Abstract

Astrocyte to neuron reprogramming has been performed using viral delivery of neurogenic transcription factors in GFAP expressing cells. Recent reports of off-target expression in cortical neurons following adeno-associated virus (AAV) transduction to deliver the neurogenic factors have confounded our understanding of the efficacy of direct cellular reprogramming. To shed light on potential mechanisms that may underlie the neuronal off-target expression of GFAP promoter driven expression of neurogenic factors in neurons, two regionally distinct cortices were compared-the motor cortex (MC) and medial prefrontal cortex (mPFC)-and investigated: (1) the regional tropism and astrocyte transduction with an AAV5-GFAP vector, (2) the expression of Gfap in MC and mPFC neurons; and (3) material transfer between astrocytes and neurons. Using a Cre-based system (AAV5-hGFAP-Cre; Rosa26R-tdTomato reporter mice), regional differences were observed in tdTomato expression between the MC and mPFC. Interestingly, this correlated with the presence of a greater expression of Gfap mRNA in neurons in the mPFC. Additionally, intercellular material transfer of Cre and tdTomato was observed between astrocytes and neurons in both regions, albeit at very low frequencies. Our study highlights regionally distinct variation in neurons that warrants consideration when designing genetic constructs for gene therapies targeting astrocytes including astrocyte to neuron reprogramming.