Cis-regulatory elements driving motor neuron-selective viral payload expression within the mammalian spinal cord

Abstract

Spinal motor neuron (MN) dysfunction is the cause of a number of clinically significant movement disorders. Despite the recent approval of gene therapeutics targeting these MN-related disorders, there are no viral delivery mechanisms that achieve MN-restricted transgene expression. In this study, chromatin accessibility profiling of genetically defined mouse MNs was used to identify candidate cis-regulatory elements (CREs) capable of driving MN-selective gene expression. Subsequent testing of these candidates identified two CREs that confer MN-selective gene expression in the spinal cord as well as reduced off-target expression in dorsal root ganglia. Within one of these candidate elements, we identified a compact core transcription factor (TF)-binding region that drives MN-selective gene expression. Finally, we demonstrated that selective spinal cord expression driven by this mouse CRE is preserved in non-human primates. These findings suggest that cell-type-selective viral reagents in which cell-type-selective CREs drive restricted gene expression will be valuable research tools in mice and other mammalian species, with potentially significant therapeutic value in humans.

Keywords: AAV; cis-regulatory element; dorsal root ganglion; gene therapy; motor neuron.

Fig. 1.

Experimental strategy and motor neuron CRE identification. (A) Schema of overall experimental design. (B) Representative IHC image of a L2 spinal cord section from a 10-week-old Sun1-ChAT mouse, demonstrating cytoplasmic ChAT (magenta) and GFP-labeled MN nuclear envelope (green). DAPI nuclei labeled in blue. (Scale bar, 250 μm.) (C) Representative ATAC-seq genome browser tracks (normalized counts per location) across three bioreplicates (each bioreplicate comprises combined spinal cord tissue from two animals) derived from immunopurified spinal motor neuron (IP, green) and flowthrough (FT, yellow) nuclei near the 5’ Chat gene terminus. An example CRE that is selectively accessible in MNs (CRE98, gray highlight) and a nearby nonselective CRE (blue highlight) are shown. Sequence conservation across the Placental mammalian clade (PhyloP) is also shown. (D) Fixed-line-plots showing mean ATAC-seq signal strength (Left and Middle) and sequence conservation z-score (PhyloP, placental mammals, Right) in immunopurified spinal MN (green) and flowthrough (yellow) nuclei, sorted by descending signal strength, averaged across three bioreplicates (each bioreplicate comprises combined spinal cord tissue from two animals). Each ATAC-seq site is represented as a single horizontal line centered at the peak summit with flanking 1 kb. Color denotes ATAC-seq signal intensity [displayed in log₂(counts per million + 1)]. Sites are classified by differential accessibility across the immunoprecipitated spinal MN and flowthrough populations as follows: both, accessible in both cell populations and not significantly different (FDR-corrected q > 0.05); ChAT^NEG only, significant enrichment of accessibility in the flowthrough population (ChAT^POS/ChAT^NEG fold-change ≤ 0.5, FDR-corrected q < 0.05, DESeq2); ChAT^POS only, significant enrichment of accessibility in the spinal MN population (ChAT^POS/ChAT^NEG fold-change ≥ 2, FDR-corrected q < 0.05, DESeq2); Background (n = 2,000), randomly selected genomic loci included for visual comparison. (E) Aggregate ATAC signal (Left) and conservation (Right) across peaks described in Fig. 1D plotted as mean ATAC signal by distance from peak center. (F) Left, MA plot of MN-enrichment [log2(ChAT^POS/ChAT^NEG ATAC-seq signal)] as a function of mean ATAC-seq signal for each TSS-distal peak. Sites that exhibit >32-fold enriched accessibility in MNs with FDR-corrected q < 0.01 are denoted in green. Right, histogram of MN enrichment for all peaks, with MN-enriched peaks denoted by green rug plot (>32-fold enriched, q < 0.05). (G) Left, scatter plot of MN-enrichment [log2(ChAT^POS/ChAT^NEG ATAC-seq signal)] as a function of conservation z-score (PhyloP, placental mammals) for each MN-enriched (>32-fold enriched accessibility in MNs, FDR-corrected q < 0.01) TSS-distal peak (light green hollow points). Sites with conservation z-score > 1 denoted in dark green filled points. Top and Right, corresponding histograms across conservation z-score and MN-enrichment, respectively.

Fig. 2.

Initial candidate CRE screening by confocal microscopy. (A) Upper panel: volcano plot of –log₁₀(FDR-corrected q) versus MN-enrichment [fold change, log2(ChAT^POS/ChAT^NEG ATAC-seq signal)] for all TSS-distal putative CREs. Lower panel: conservation z-score versus MN-enrichment [fold change, log2(ChAT^POS/ChAT^NEG ATAC-seq signal)] for all TSS-distal putative CREs. The vertical dotted line denotes 32-fold MN enrichment threshold; the horizontal dotted line denotes conservation z-score threshold of 1. Putative MN-enriched (>32-fold enriched accessibility in MNs, FDR-corrected q < 0.01) and conserved (z-score > 1) CREs plotted in green. CREs selected from this subpopulation for final screening are highlighted and labeled by their relevant selection criteria: lowest q-value in blue, highest MN enrichment in yellow, most conserved in purple, and controls in gray. (B) Table of candidate CREs selected for testing, showing internal identification (CRE), genomic position [Chromosome, Start, End; (mm10)], element length (defined as 500 by peak-calling algorithm), MN ATAC-seq enrichment [calculated as log2(ChAT^POS/ChAT^NEG ATAC-seq signal, labeled as log2(FC)], FDR-corrected q-value (q), and conservation z-score. (C) AAV screening construct viral genome maps (not drawn to scale). ITR, inverted terminal repeats; pBG, minimal beta-globin promoter; i, intron; NLS, nuclear localization sequence; GFP, green fluorescent protein. WPRE, woodchuck hepatitis virus posttranscriptional response element; pA, poly-A tail. (D) Representative confocal native fluorescence images of T1-L4 spinal cord sections from wild-type mice 14 d after P0 ICV injection of candidate CRE reporter AAVs (DAPI, cyan; GFP, yellow). CRE selection criteria denoted by line above image labels (lowest MN differential ATAC-seq q-value in blue, highest MN ATAC-seq enrichment in yellow, most conserved in purple, controls in gray, and CAG in black). (Scale bar, 250 μm.) Combined ventral horn images also shown at higher magnification in inset (denoted by dashed box, 4.7× magnification). (E) Quantification of GFP signal intensity fold-change in ventral compared to dorsal horn for all constructs tested (ventral enrichment). Mean enrichment across three spinal cord sections per animal denoted by points, mean across three animals denoted by bar, SD across three animals denoted by line. CRE selection criteria denoted by point color (lowest MN differential ATAC-seq q-value in blue, highest MN ATAC-seq enrichment in yellow, most conserved in purple, controls in gray, and CAG in black). Null hypothesis (ventral enrichment = 1) plotted as dotted line.

Fig. 3.

Quantification of CRE specificity by immunohistochemistry. (A) Representative IHC images of T1-L4 spinal cord sections 14 d after P0 ICV injection of candidate CRE reporter AAVs (NeuN, cyan; ChAT, magenta; GFP, yellow). (Scale bar, 250 μm.) GFP and Merge images of ventral horn also shown as magnified inset). (B) Quantification of mean GFP signal intensity in off-target spinal neurons (NeuN^POS ChAT^NEG, blue) and on-target spinal MNs (NeuN^POS ChAT^POS, purple) for combined CRE57/∆CRE control (n = 6 mice), CRE98 (n = 3 mice), and CRE119 (n = 2 mice) across all animals and (C) across all cells. Mean enrichment across all spinal cord sections per animal denoted by each point, mean across all animals per condition denoted by bar plot, SD across all animals denoted by line. (D) Percentage of GFP^POS cells in off-target spinal neurons (NeuN^POS ChAT^NEG, blue) and on-target spinal MNs (NeuN^POS ChAT^POS, purple) for combined CRE57/∆CRE control (n = 6 animals), CRE98 (n = 3 animals), and CRE119 (n = 2 animals) constructs. Mean percentage across all spinal cord sections per animal denoted by each point, mean across all animals per condition denoted by bar plot, SD across three animals denoted by line. (E) Selectivity (on-target NeuN^POS ChAT^POS/off-target NeuN^POS ChAT^NEG mean signal intensity across all neurons per animal) for combined CRE57/∆CRE control (n = 6 animals), CRE98 (n = 3 animals), CRE119 (n = 2 animals), and CAG (n = 3 animals) constructs. Mean percentage across all spinal cord sections per animal denoted by each point, mean across all animals per condition denoted by bar plot, SD across all animals denoted by line. (F) Representative IHC images of T1-L4 DRG simultaneously dissected from previously described animals 14 d after P0 ICV injection of candidate CRE reporter AAVs (DAPI, cyan; GFP, yellow). (Scale bar, 100 μm.) (G) Percentage of GFP^POS off-target DRG neurons (visually identified by nuclear DAPI morphology and size) for combined CRE57/∆CRE control (n = 6), CRE98 (n = 3), CRE119 (n = 2), and CAG (n = 3) constructs. Mean percentage across all DRG sections per animal denoted by each point, mean across all animals per condition denoted by bar plot, SD across all animals denoted by line. All images acquired at pBG promoter-optimized acquisition parameters. Significance measured by Bonferroni-corrected two-tailed t test. *** q < 0.001, ** q < 0.01, * q < 0.05, ns q > 0.05.

Fig. 4.

A core TF-binding region mediates CRE98 function. (A) Top: Native genomic context of mouse CRE98 (mCRE98) showing TF-binding sites (TFBS) along genomic coordinates of chromosome 14 (mm10). Middle: Raw conservation score by base-pair conservation (PhyloP, placental clade). Bottom: CRE98 core region (mCORE), truncation constructs (A–F), and 5KO targeted deletion construct. (B) Percentage GFP^POS cholinergic cells (ChAT^POS) and (C) mean GFP signal intensity across all animals in off-target spinal neurons (NeuN^POS ChAT^NEG, blue) and on-target spinal MNs (NeuN^POS ChAT^POS, purple) for saline-injected control (n = 2 animals), murine CRE98 (n = 3 animals), and truncation and 5KO constructs (n = 2 to 3 animals per construct). Mean across all spinal cord sections per animal denoted by each point, mean across all animals per condition denoted by bar plot, SD across three animals denoted by line. Images acquired and quantified at pBG promoter-optimized acquisition parameters. (D) Scatter plot of statistical significance [−log10 (Benjamini–Hochberg corrected TF motif enrichment q-value in CRE98)] as a function of MN-enrichment [log2(ChAT^POS/ChAT^NEG RNA-seq signal)]. Motif enrichment q-values derived from FIMO. Motifs that exhibit motif-enrichment q < 0.01 and whose associated TFs are >2-fold more highly expressed in ChAT^POS immunoprecipitate over ChAT^NEG flowthrough (FDR-corrected q < 0.01) are denoted in green and labeled by associated transcription factor.

Fig. 5.

Evaluation of synthetic and human CRE98-derived constructs. (A) Synthetic AAV screening construct vector genome maps (not drawn to scale). ITR, inverted terminal repeats; pBG, minimal beta-globin promoter; i, intron; NLS, nuclear localization sequence; GFP, green fluorescent protein. WPRE, woodchuck hepatitis virus posttranscriptional response element; pA, poly-A tail. (B) Native genomic context of human CRE98 (hCRE98), showing TF-binding sites (TFBS) along genomic coordinates of chromosome 10 (hg38). (C) Representative IHC images of T1-L4 spinal cord sections 14 d after P0 ICV injection of synthetic CRE reporter AAVs (NeuN, cyan; ChAT, magenta; GFP, yellow). mCRE98 and CAG images reproduced from Fig. 3A. (Scale bar, 250 μm.) (D) Quantification of mean GFP signal intensity and (E) percentage of GFP^POS cells in off-target spinal neurons (NeuN^POS ChAT^NEG, blue) and on-target spinal MNs (NeuN^POS ChAT^POS, purple) for saline-injected and ∆CRE controls (n = 3 animals each), CRE98 (n = 3 animals), synthetic constructs (n = 2 to 3 animals), and CAG (n = 3 animals). Mean enrichment across all spinal cord sections per animal denoted by each point, mean across all animals per condition denoted by bar plot, SD across all animals denoted by line. ∆CRE, mCRE98, and CAG data replotted from Fig. 3.

Fig. 6.

CRE98 drives spinal cord-selective expression in the macaque. (A) Experimental design for NHP study. NGS, next-generation sequencing. (B) AAV screening construct vector genome maps (not drawn to scale). ITR, inverted terminal repeats; prom, CAG, or hSYN1 promoter; pBG, minimal beta-globin promoter; mCORE, mCRE98 core sequence; eGFP, enhanced green fluorescent protein; nanoLuc, luciferase reporter; WPRE, woodchuck hepatitis virus posttranscriptional response element; pA, poly-A tail. (C) Relative (normalized to dosing material) expression of hSYN1-, mouse CRE98 (mCRE98)-, mouse core CRE98 (mCORE)-, and CAG-driven AAV9 reporter constructs across tissues (on-target spinal cord in purple, off-target CNS-derived tissues in yellow, DRG in orange, and other in blue) harvested 28 d postdosing from AAV9 neutralization-naive cynomolgus monkeys (n = 3 animals). Normalized expression value of five denoted by dotted line across all constructs.