Efficient and Precise Processing of the Optimized Primary Artificial MicroRNA in a Huntingtin-Lowering Adeno-Associated Viral Gene Therapy In Vitro and in Mice and Nonhuman Primates

Constructs and vectors

AAV vectors (Supplementary Fig. S1) were produced by Voyager Therapeutics or the University of Massachusetts Chan Medical School Viral Vector Core. The transgene DNA comprised AAV2-derived inverted terminal repeats flanking a chicken β-actin (CBA) promoter, a pri-amiRNA scaffold containing a miRNA targeting human HTT mRNA, a rabbit beta-globin polyadenylation (pA) signal, and a stuffer sequence (iHtt). The positive control pri-amiRNA targeting HTT mRNA comprised our modifications of the previously published sequence mi2.4, ¹⁴ with the modifications consisting of the removal of two adenine nucleotides at the 3′ end of the seed sequence in the passenger strand, creating a bulge due to the remaining thymine nucleotides in the guide strand, and the addition of an adenine nucleotide at the 3′ end of the passenger strand and a guanine nucleotide at the 3′ end of the guide strand.

The transgene DNA was packaged in a recombinant AAV1 or AAV2 capsid. Recombinant AAV vectors were produced either by triple transfection of adherent HEK293 cells or with the Sf9/baculovirus system and purified using previously described protocols. ^{35 –39} In brief, recombinant vectors were purified from clarified lysates using iodixanol gradients or by the cesium chloride gradient sedimentation method. Vectors were formulated in phosphate-buffered saline with 0.001% (w/v) Pluronic F-68^®, pH 7.2–7.4, with or without 55–110 mM sodium chloride, and stored at less than or equal to −60°C. Vector titers were measured by droplet digital polymerase chain reaction (ddPCR) and/or quantitative polymerase chain reaction (qPCR). For vectors used for in vivo studies, the relative purity of the capsid proteins was confirmed using polyacrylamide gel electrophoresis (PAGE) followed by silver staining analysis. Genome integrity was evaluated by denaturing 1% agarose gel electrophoresis followed by SYBER Gold staining. In addition, the percentage of full capsids by cesium chloride gradient was at least 70%. Aggregation was <25%, and endotoxin levels were <2 endotoxin units (EU)/mL by the limulus amebocyte lysate test.

Vector formulation and aliquoting were generally performed 1 day before use. AAV vectors were removed from storage at less than or equal to −60°C and allowed to thaw at room temperature or on refrigerated cold packs, then maintained at ∼4°C. Dilutions were made with formulation buffer (192 mM sodium chloride, 10 mM sodium phosphate, 2 mM potassium phosphate, 2.7 mM potassium chloride, and 0.001% Pluronic F-68; pH 7.4) and then stored at ∼4°C until use. Vehicle groups received formulation buffer. For studies in NHPs, ProHance^® (gadoteridol; Cat. No. 111104; Bracco Diagnostics, Monroe Township, NJ) was added at a 1:250 dilution to all dosing solutions. Dosing solutions were brought to room temperature before intracranial infusion.

For in vivo comparison studies in both YAC128 mice and NHPs, dosing solutions were blinded and unblinded after the data had been analyzed.

Cell culture and treatments

HeLa, U251 MG, and SH-SY5Y cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM)/F12-GlutaMAX^™ medium (Cat. No. 10565-018; Thermo Fisher Scientific, Waltham, MA) supplemented with 10% fetal bovine serum (FBS; Cat. No. 16000-044; Thermo Fisher Scientific), MEM Non-Essential Amino Acid Solution (Cat. No. 11140-050; Thermo Fisher Scientific), and HEPES.

For evaluation of endogenous HTT mRNA suppression, HeLa, U251MG, or SH-SY5Y cells were seeded in a 96-well plate at 1 × 10⁴, 2 × 10⁴, or 0.3 × 10⁴ cells/well, respectively, before transduction with Treatments A–P or AAV.mCherry (negative control) at a multiplicity of infection (MOI) of 1 × 10⁵ vector genomes (VG)/cell. HeLa cells were transduced with vector ∼24 h after plating. U251MG cells were mixed with vector just before plating. Before vector transduction, SH-SY5Y cells were neuronally differentiated. SH-SY5Y cells were plated on tissue culture plastic coated with 100 μg/mL Poly-d-Lysine (Cat. No. A-003-E; MilliporeSigma, St. Louis, MO) and 10 μg/mL laminin (Cat. No. 23017-015; Thermo Fisher Scientific) and cultured in differentiation medium comprising DMEM/F12-GlutaMAX medium supplemented with MEM Non-Essential Amino Acids Solution, sodium pyruvate, 0.5% FBS, 50 nM human recombinant IGF-1 (Cat. No. AF-100-11; Peprotech, Cranbury, NJ), and HEPES for 4 days before addition of vector. Each vector was tested in triplicate. At 48, 24, or 36 h post-transduction of HeLa, U251MG, or neuronally differentiated SH-SY5Y cells, respectively, the cells were harvested and lysed for analysis of endogenous HTT mRNA levels.

For the dual-Luc assay, HeLa cells were seeded in a 96-well plate at 8 × 10³ or 1 × 10⁴ cells/well, transfected with 1.25 ng/μL HTT pri-amiRNA plasmid in triplicate using FuGENE^® HD (Cat. No. E2311; Promega, Madison, WI), and lysed at 48 h post-transfection. For small RNA sequencing, HeLa and neuronally differentiated SH-SY5Y cells were cultured in 10 or 15 cm plates.

HEK293T cells were cultured in complete DMEM/F12-GlutaMAX medium supplemented with 10% FBS. For evaluation of endogenous HTT mRNA and protein suppression, the cells were seeded in a 96-well plate at 2 × 10⁴ cells/well 4 h before AAV.EGFP or transduction with Treatment P. The AAV vector was added to cultured HEK293T cells at a MOI of 3.0 × 10² to 3.0 × 10⁵ VG/cell. Each vector dilution was tested in duplicate for HTT mRNA determination and on a single culture for HTT protein determination. After 24, 48, 72, or 96 h, the cells were harvested and lysed for analysis of HTT mRNA and protein levels. For the dual-Luc assay, HEK293T cells were seeded in a 96-well plate at 2.5 × 10⁴ cells/well, transfected with 1.25 ng/μL HTT pri-amiRNA plasmid in triplicate using FuGENE HD, and lysed at 48 h post-transfection.

HeLa S3 cells used for the AAV neutralizing antibody (NAb) assay were cultured in complete DMEM supplemented with 10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. One day before AAV transduction, the cells were seeded in a 96-well plate at 1 × 10⁴ cells/well. AAV1.eGFP was preincubated with the NHP serum sample at 37°C for 1 h and then added to cultured HeLa S3 cells at a MOI of 4 × 10⁴ VG/cell. Each condition was tested at dilutions of 1:16 and 1:64. After 48 h, the relative percentage of GFP positive cells was determined by flow cytometry.

Anti-AAV1 NAb assay

To screen NHPs for low or no detectable levels of anti-AAV1 NAb activity in the serum before inclusion in the study, an in vitro anti-AAV1 NAb assay was used to measure the ability of NHP serum to inhibit AAV1.eGFP transduction of HeLa S3 cells. AAV1.eGFP was preincubated with the NHP serum sample at 37°C for 1 h and then added to cultured HeLa S3 cells at a MOI of 4.00 × 10⁴ VG/cell and a NHP serum dilution of 1:16 or 1:64. After 3 days, the cells were dissociated with 0.25% trypsin-EDTA to generate a single cell suspension for analysis on a flow cytometer (BD Accuri C6; Becton Dickinson, Franklin Lakes, NJ). Fixable Viability Stain-660 (Cat. No. 564405; Becton Dickinson) and then Cytofix^™ solution (Cat. No. 554655; Becton Dickinson) were added to the cells, to analyze viable cells for GFP signal. The percentage of GFP-positive versus GFP-negative cells (FL-1 channel) was determined. The relative number of GFP-positive cells was then calculated by normalizing the percentage of GFP-positive cells with AAV1-eGFP treatment in the presence of NHP serum to the percentage of GFP-positive cells with AAV1-eGFP treatment in the absence of NHP serum. NHPs with sera that resulted in <50% inhibition of AAV1.eGFP transduction in HeLa S3 cells at 1:16 dilution (i.e., at least 50% GFP-positive cells) and <20% inhibition at 1:64 dilution (i.e., at least 80% GFP-positive cells) were defined as animals with no detectable or low levels of serum anti-AAV1 NAb activity and selected for inclusion in the study.

Animals and treatments

General tolerability assessments

Tissue collection

Dual-Luc assay

The dual-Luc assay (Dual-Luciferase^® Reporter 1000 Assay System; Cat. No. E1980; Promega) was conducted in HEK293T and HeLa cells. Reporter plasmids for evaluating guide and passenger strand activity contained the target regions of the tested set of guide or passenger strands in tandem, separated by 2–6 nucleotides. HTT pri-amiRNA and reporter plasmids were transfected into the cells at 1.25 and 0.025 ng/μL, respectively, in triplicate using FuGENE HD, and then 48 h after transfection, the cells were lysed and the Renilla and firefly luciferase activities measured, according to the manufacturer's instructions. The relative activity was obtained by normalizing the Renilla luciferase level to the control firefly luciferase level.

HTT mRNA levels

HTT protein levels

Human HTT protein levels in vitro were measured by western blot. Cells were pelleted and lysed in MSD Tris Lysis Buffer (Cat. No. R60TX; Meso Scale Discovery, Rockville, MD) supplemented with protease and phosphatase inhibitor cocktail. Cell lysates were then cleared by centrifugation and supernatants evaluated for total protein by the BCA assay. A sample containing 20 μg protein was denatured at 70°C and run on a 3–8% Tris Acetate NuPAGE^™ gel. Proteins were transferred onto PVDF membrane using the iBlot^™ 2 transfer device (Invitrogen) according to the manufacturer's instructions. The PVDF membrane was treated with LI-COR Odyssey^® blocking buffer and then incubated with primary antibodies against β-actin (Cat. No. ab8227; Abcam, Cambridge, MA) or HTT protein (Cat. No. MAB2166; EMD Millipore, Danvers, MA), followed by secondary antibodies comprising goat anti-mouse IRDye 800CW conjugated secondary antibody (LI-COR; 800 nm, green) for HTT protein and rabbit goat anti-rabbit IRDye 680RD conjugated secondary antibody (LI-COR; 680 nm, red) for β-actin. After extensive washing, the PVDF membrane was scanned with the LI-COR Odyssey imaging system at 680 and 800 nm, and the fluorescence intensities for the HTT and β-actin protein bands were quantified using the LI-COR Odyssey imaging software. For each sample, the HTT protein signal was normalized to the corresponding β-actin signal.

VG levels

The number of VG copies in putamen samples was measured with the ddPCR assay after DNA extraction. An aliquot of the frozen lysates prepared in QuantiGene Plex homogenization buffer with proteinase K as described above was thawed for whole cell DNA extraction using the DNeasy Blood & Tissue DNA Purification Kit (Cat. No. 69581; Qiagen, Germantown, MD). After combining the aliquot with an equal volume of ATL buffer, DNA was purified according to the manufacturer's instructions. The concentration of whole cell DNA was determined by NanoDrop (Thermo Fisher Scientific).

ddPCR was performed using a FAM labeled TaqMan probe set specific for the CBA promoter common to AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3 and a VIC labeled probe set specific for the RNaseP gene of Rhesus macaque (Cat. No. 4403328; Thermo Fisher Scientific). These primer-probe sets (IDT, Coralville, IA), Supermix (Cat. No. 1863024; Bio-Rad), and HindIII-HF (Cat. No. R3104L; New England Biolabs, Ipswich, MA) were added to wells containing whole cell DNA. Droplets were generated using the Automatic Droplet Generator (Bio-Rad), and the reaction plate was then run on a thermocycler and analyzed using a ddPCR reader (QX-200; Bio-Rad). Values with ≤5 positive droplets per reaction were defined as below the limit of detection (LOD), based on no template controls which had no positive droplets or sporadic counts of 1–2 droplets per reaction, and excluded from analysis.

The number of VGs per diploid genome was calculated by dividing the copies/μL obtained with the CBA probe set by half the copies/μL obtained with the RNaseP probe set. The lower limit of quantification (LLOQ) of the assay was 0.01 VG copies/diploid genome.

Small RNA sequencing and data analysis

Precandidate pri-amiRNA designs lower reporter and endogenous HTT mRNA levels efficiently in vitro

To identify pri-amiRNA candidates worthy of in vivo assessment, we first measured the in vitro activities of 16 precandidate pri-amiRNAs in human cell lines using a reporter assay designed to separately detect guide- and passenger-strand activity.

We conducted guide and passenger strand reporter assays in both HeLa and HEK293T cells following transfection of the precandidate pri-amiRNAs. From the starting pool of pri-amiRNA sequences, 16 precandidate pri-amiRNAs, Treatments A–P, were selected, primarily based on the guide strand reporter assay in which they were among the most effective pri-amiRNAs out of the starting pool in lowering the luciferase signal, and lowered the luciferase signal by more than 70% on average (Supplementary Fig. S2). In addition, these 16 precandidate pri-amiRNAs exhibited less than ∼30% lowering on average of the luciferase signal in the passenger strand reporter assay (data not shown). These findings demonstrate that the guide strands of these 16 precandidate pri-amiRNAs lowered their targets substantially, whereas the passenger strands of these precandidate pri-amiRNAs did not. Notably, all 16 precandidate pri-amiRNAs reduced the guide strand reporter signal similarly under these conditions.

Next, we evaluated the 16 precandidate pri-amiRNAs, Treatments A–P, for lowering of endogenous HTT expression. To determine useful in vitro conditions for comparing target lowering, we characterized the time course (Supplementary Fig. S3A) and dose dependence (Supplementary Fig. S3B, C) of HTT mRNA and protein lowering in HEK293T cells using Treatment P. HTT mRNA lowering demonstrated maximum reduction at approximately 24–48 h post-transduction with slight recovery of HTT mRNA levels at 72 and 96 h postdosing. HTT protein lowering slightly lagged HTT mRNA lowering with initial, partial HTT protein lowering at 24 h, and maximal HTT protein lowering attained at approximately 48–72 h post-transduction. At the 48-h timepoint post-transduction, which provided approximately maximum HTT mRNA and protein lowering, HTT mRNA and protein levels relative to vehicle showed similar dose–responses across MOIs of 3.0 × 10² to 3.0 × 10⁵ VG/cell. Graded levels of HTT mRNA and protein lowering were observed at MOIs of 3.0 × 10² to 9.5 × 10³ VG/cell with maximum levels of HTT mRNA and protein lowering attained at MOIs of 3.0 × 10⁴ to 3.0 × 10⁵ VG/cell. These results confirm that an MOI of 1.0 × 10⁵ VG/cell and a timepoint of 24–48 h post-transduction comprise optimal conditions for assessing HTT mRNA reduction in vitro.

To evaluate the precandidate pri-amiRNAs for lowering of endogenous HTT mRNA, three different human cell lines were used. Since the primary target cell types for a HTT mRNA lowering therapy are neurons and glia, neuronally-differentiated SH-SY5Y cells, derived from a human neuroblastoma, and U251MG cells, derived from a human glioblastoma, were selected. In addition, in vitro testing was conducted in HeLa cells, which have been used widely to evaluate HTT mRNA lowering. Substantial differences between the 16 pri-amiRNAs emerged in these 3 cell lines when lowering of endogenous HTT mRNA was assessed (Fig. 2). For these studies, the 16 precandidate pri-amiRNAs were packaged in AAV2, which is more effective in transducing cells in vitro than AAV1 (used later for in vivo studies). All vectors were identical in construction, production, and purification, the only difference between the 16 vectors being the pri-amiRNA sequence. At an MOI of 1 × 10⁵ VG/cell, HTT mRNA lowering ranged from 27% to 70% in HeLa, 19% to 75% in U251MG, and 18% to 65% in neuronally differentiated SH-SY5Y cells, relative to the AAV2.mCherry control.

OPEN IN VIEWER

Figure 2.

Human HTT mRNA lowering and pri-amiRNA processing in vitro. HeLa, U251MG, and neuronally differentiated SH-SY5Y cells were transduced with AAV2-packaged pri-amiRNAs targeting HTT mRNA at a MOI of 1.0 × 10⁵ VG per cell and harvested approximately 24–48 h later. Sixteen precandidate pri-amiRNAs (A–P) and a positive control were evaluated. (A) Human HTT mRNA levels, as well as mRNA levels of human XPNPEP1 as the endogenous reference gene, were measured by RT-qPCR. Human HTT mRNA levels were normalized to XPNPEP1 mRNA levels and then further normalized to the negative control group (AAV2.mCherry). All treatments were tested in triplicate and then samples were run in duplicate for qPCR. The group mean ± standard deviation is shown for each treatment. (B–D) Pri-amiRNA processing was evaluated by small RNA sequencing. N = 1 per treatment. (B) Level of mature exogenous miHTT relative to the total endogenous pool of miRNAs. (C) Ratio of guide to passenger strands of the mature exogenous miHTT. (D) Percentage of mature exogenous miHTT guide strands with precise 5′ ends. HTT, huntingtin; miRNA, microRNA; MOI, multiplicity of infection; mRNA, messenger RNA; RT-qPCR, reverse transcription-quantitative polymerase chain reaction; VG, vector genome; XPNPEP1, X-prolyl aminopeptidase 1. Color images are available online.

Pri-amiRNAs that lower the target mRNA to the greatest extent across multiple cell lines are more likely to be most effective in subsequent screens. The rank order of the 16 different precandidate pri-amiRNAs for lowering endogenous HTT mRNA was similar but not identical across the three cell lines (Fig. 2A). In HeLa cells, the rank order (most to least lowering of HTT mRNA on average) of precandidate pri-amiRNAs was: Treatment A, P > L > G > B > F > K > I > O > J > E > M > D > N > C > H. In U251MG cells, the rank order for HTT mRNA lowering was: Treatment L > P > A > G > I > F > E, O > B > M > N > J > C > H > K > D. In neuronally differentiated SH-SY5Y cells, the rank order for HTT mRNA lowering was: Treatment P > L > A > G > E, I > F > B > O > J > N > M > D > C, K > H. Thus, treatment with precandidate pri-amiRNAs P, L, A, and G resulted in the most HTT mRNA lowering on average in all three cell lines, with 59–75% lowering by precandidates L, P, and A and 52–60% lowering by precandidate G. The next (fifth) best precandidate pri-amiRNA in HeLa cells, Treatment B, resulted in more modest lowering of HTT mRNA in U251MG and neuronally differentiated SH-SY5Y cells than in HeLa cells. The fifth best precandidate pri-amiRNA in U251MG and neuronally differentiated SH-SY5Y cells, Treatment I was only ranked 8th in HeLa cells. The sixth best precandidate was Treatment F in all three cell lines. Interestingly, Treatment K performed well in HeLa cells, ranking seventh and producing 54% HTT mRNA lowering but was relatively ineffective at lowering HTT mRNA in U251MG and neuronally differentiated SH-SY5Y cells (by 22% or 21%, respectively), ranking 15th (second worst) in these two cell lines. These results show that the rank order of pri-amiRNAs in lowering endogenous target mRNA can vary substantially across different cell lines in vitro. Thus, HTT mRNA lowering in a single cell line may not be representative for selecting the most effective pri-amiRNA. These results demonstrate the importance of using multiple cell lines in vitro when assessing pri-amiRNA candidates.

Extent of harnessing of cellular miRNA biogenesis pathway and guide/passenger strand ratio provide dramatic differentiation of precandidate pri-amiRNAs in vitro

We used high-throughput small RNA deep sequencing to evaluate the efficiency and precision of pri-amiRNA processing of the 16 precandidate pri-amiRNAs in HeLa and neuronally differentiated SH-SY5Y cells. These characteristics were compared to the reduction of HTT mRNA in these same cell lines (See Precandidate pri-amiRNA designs lower reporter and endogenous HTT mRNA levels efficiently in vitro section), using the same AAV vectors and MOIs.

Since minimal harnessing of the endogenous RNAi pathway is less likely to perturb the endogenous pool of miRNAs and more likely to leave their associated cellular functions intact, potent target lowering concomitant with low harnessing of the endogenous RNAi pathway are key characteristics for safety and efficacy considerations. In this study we evaluated the level of mature exogenous miHTT relative to the total endogenous pool of miRNAs in HeLa and neuronally differentiated SH-SY5Y cells (Fig. 2B) to monitor our precandidate's usage of the endogenous miRNA biogenesis pathway. The samples from HeLa and neuronally differentiated SH-SY5Y cells each had 13.7–33.8 million raw reads with the correct adapter sequence. First, we aligned the reads to annotated human miRNAs to identify the reads corresponding to endogenous miRNAs. We then aligned the remaining reads to our exogenous miHTT sequences. The level of mature exogenous miHTT relative to the total endogenous pool of miRNAs in HeLa and neuronally differentiated SH-SY5Y cells was then calculated. The abundance of mature exogenous miHTTs relative to the total endogenous pool of human miRNAs varied greatly across the 16 precandidate pri-amiRNAs, spanning a 1004-fold range from 0.2% for Treatment K in neuronally differentiated SH-SY5Y cells to 200.8% for Treatment A in HeLa cells. The expression levels of mature exogenous miHTTs from Treatments A and O exceeded the sum of all endogenous miRNA levels in the same sample (>100% of the endogenous miRNA pool), suggesting that these mature exogenous miHTTs out-competed endogenous miRNAs for loading into Argonaute. Although similar levels of HTT mRNA lowering, 59–65% on average, were achieved by precandidates A, L, and P in neuronally differentiated SH-SY5Y cells, the level of mature exogenous miHTT relative to the total endogenous pool of miRNAs varied from 0.6% for Treatment P to 168.2% for Treatment A, with Treatment L intermediate at 3.6%. Our data demonstrate that potent target lowering concomitant with low harnessing of the endogenous RNAi pathway is a key characteristic that provides strong differentiation in vitro between different pri-amiRNAs.

Since the guide strand is the strand intended to mediate HTT mRNA lowering using RNAi, while the passenger strand has no intended activity, higher G/P strand ratios are more favorable for potency and selectivity. Moreover, in the event that the passenger strand has unintended off-target activity, lower levels of the P strand would result in less off-target activity and, presumably, enhanced safety.

The G/P strand ratio varied substantially across the 16 precandidate pri-amiRNAs, from 0.1 to 317.6 in HeLa cells, and 0 to 834.4 in neuronally differentiated SH-SY5Y cells (Fig. 2C). Twelve of the 16 precandidate pri-amiRNAs exhibited G/P strand ratios >10 in neuronally differentiated SH-SY5Y cells (Treatments A–C, E, and I–P), as well as the positive control. In HeLa cells, transduction with 9 precandidate pri-amiRNAs packaged in AAV2 resulted in G/P strand ratios >10 (Treatments A, B, E, I–K, M–O). Treatment C was not evaluated for miHTT processing in HeLa cells, and Treatments L and P, as well as the positive control, were transfected using plasmids into HeLa cells so the results cannot be compared directly to those obtained with AAV transduction. Treatments D, F, G, and H showed G/P strand ratios <10 (range 0–1.0), including three precandidate pri-amiRNAs (Treatments D, F, and H) with ratios of <0.5, indicating that there were more than twice as many P strands as G strands. The G/P strand ratios were remarkably similar in HeLa and SH-SY5Y cells for a given pri-amiRNA, and the rank order for the precandidate pri-amiRNAs was likewise very similar in these two cell lines. Thus, the G/P strand ratio is a parameter that provides strong differentiation of different pri-amiRNAs and appears to be similar for a given pri-amiRNA across different cell lines in vitro.

Since the seed region is critical for recognition of the target mRNA and begins at a fixed position relative to the 5′ end of the guide strand, accurate processing of the 5′ end of the guide strand is required for the guide strand to bind with its intended target. Any shift in the 5′ end of the guide strand would result in an altered seed sequence that could interact with unintended mRNAs and potentially mediate off target effects. We assessed the precision of 5′ end processing of miRNA guide strands in HeLa and neuronally differentiated SH-SY5Y cells (Fig. 2D). In HeLa cells, seven precandidate pri-amiRNAs showed at least 90% correct 5′ end processing of the guide strand, whereas eight pri-amiRNAs exhibited <80% correct 5′ ends. Notably, four of these latter eight pri-amiRNAs were from the same scaffold, suggesting that the secondary structure of this scaffold may not favor accurate Dicer cleavage in HeLa cells. In neuronally differentiated SH-SY5Y cells, only one precandidate pri-amiRNA (Treatment B) showed <80% of guide strands with correct 5′ ends. All of the other 15 precandidate pri-amiRNAs were processed with at least 80% of guide strands with correct 5′ ends. These results indicate that all of our pri-amiRNA designs generated the desired seed sequences with good precision, but some pri-amiRNAs clearly outperformed others.

Precandidate AAV1-iHtt vectors are well tolerated in YAC128 mouse

For in vivo screening of precandidate pri-amiRNAs, we selected the YAC128 mouse because it is a model of HD that would allow us to readily evaluate a substantial number of pri-amiRNAs for lowering of human HTT expression and pri-amiRNA processing, with the potential for future correlation of lowering of human HTT expression with phenotypic improvement. The YAC128 mouse is a yeast artificial chromosome animal model of HD with the entire human HTT gene containing 128 CAG repeats. ⁴⁰ This mouse model of HD is well characterized, has been used previously for evaluating HTT mRNA and protein lowering, and exhibits age-dependent striatal neurodegeneration and progressive motor and cognitive deficits. ^40,52,53 We compared the 16 precandidate pri-amiRNAs in the YAC128 mouse, with bilateral intrastriatal injection at total doses of 2.65 × 10¹⁰ to 6.00 × 10¹⁰ VG per animal, which were the maximum possible doses given the stock concentrations of vector. For in vivo studies, the pri-amiRNAs were packaged in AAV1, which distributes more broadly in the brain than AAV2 (used in vitro). Due to the large number of animals in the experiment, the study was divided into two cohorts, with each cohort having its own vehicle group.

All 16 precandidate AAV1-iHtt vectors (7–8 animals per AAV treatment) were well tolerated over the 4 weeks in-life period postdosing. Cage-side observations showed no abnormalities for all animals, and there were no significant differences in body weights, body weight changes, or brain weights measured at necropsy across groups (data not shown). An extensive tolerability evaluation is ongoing in the YAC128 mouse with much longer in-life periods postdosing that includes detailed histopathological and immunohistochemical examination, as well as efficacy in this model of HD; these results will be reported elsewhere.

Precandidate pri-amiRNA designs lower HTT mRNA efficiently in YAC128 mouse striatum

We first evaluated HTT mRNA reduction in striatum samples from these YAC128 mice. Treatment with the 16 precandidate AAV1-iHtt vectors resulted in different levels of human mutant HTT mRNA reduction relative to their corresponding vehicle group at 4 weeks postdosing (Fig. 3A). Treatments A, B, F, J, L, M, and P, as well as the positive control, suppressed striatal human mutant HTT mRNA levels to a statistically significant extent (p < 0.05 by one-way analysis of variance [ANOVA] with Dunnett's multiple comparisons post hoc test), ranging from 22% to 35% lowering on average relative to the corresponding vehicle group. However, no significant human mutant HTT mRNA reduction was seen for treatments C–E, G–I, K, N, and O.

OPEN IN VIEWER

Figure 3.

Human mutant HTT mRNA lowering and pri-amiRNA processing in YAC128 mouse striatum 4 weeks after bilateral intrastriatal injection of AAV1-packaged pri-amiRNAs targeting HTT mRNA at a dose of 2.6 × 10¹⁰ to 6.0 × 10¹⁰ VG per animal. Sixteen precandidate pri-amiRNAs (A–P), as well as a positive control and vehicle, were administered across two mouse cohorts (cohort 1: A–H; cohort 2: I–P), (A) Human mutant HTT mRNA levels, as well as mRNA levels of mouse XPNPEP1 and Hypoxanthine Guanine Phosphoribosyltransferase (HPRT) (endogenous reference genes), were measured by RT-qPCR. Human HTT mRNA levels were normalized to the geometric mean of XPNPEP1 and HPRT mRNA levels and then further normalized to the vehicle group of the cohort. Each symbol represents the average relative HTT mRNA level in the striatum of one animal. N = 8 for all groups except F, G, and M (N = 7). (B–D) Pri-amiRNA processing was evaluated by small RNA sequencing. N = 3 per group. The positive control samples are from cohort 1. (B) Level of mature exogenous miHTT relative to the total endogenous pool of miRNAs. (C) Ratio of guide to passenger strands of the mature exogenous miHTT. (D) Percentage of mature exogenous miHTT guide strands with precise 5′ ends. For all panels, the group mean ± standard deviation is shown for each treatment. Color images are available online.

These results illustrate that greater discrimination across pri-amiRNAs is provided by comparison of human HTT mRNA reduction in YAC128 mice than in vitro, even with three different cell lines evaluated, including neuronal and glial cell types. The 16 precandidate pri-amiRNAs lowered human HTT mRNA in vitro by approximately 20–80% but only 7 significantly reduced human HTT expression in vivo. The most effective pri-amiRNAs in vitro that lowered human HTT mRNA by at least 55% in all three cell lines at an MOI of 1 × 10⁵ (i.e., Treatments A, L, and P and positive control) also lowered human HTT mRNA significantly in YAC128 mouse striatum. In addition, other pri-amiRNAs that were less effective in vitro (i.e., Treatments J and M) with as little as 25% human HTT mRNA lowering showed significant human HTT mRNA reduction in vivo. Conversely, there are examples of pri-amiRNAs that were substantially active in vitro with as much as 50% human HTT mRNA lowering (i.e., Treatment I) but not active in vivo. Notably, pri-amiRNAs with similar pharmacological activity in vitro such as Treatments F and I showed different pharmacological activity in vivo, indicating that target lowering in vitro may not completely predict target lowering efficiency in vivo for pri-amiRNAs packaged in AAV. These results highlight the importance of conducting an in vivo screen to identify the most effective pri-amiRNAs for target lowering. Nonetheless, given the much larger screening capacity in vitro than in vivo and the predictive power of certain pri-amiRNA processing parameters in vitro for in vivo results (See Pri-amiRNA processing efficiency and precision in YAC128 mouse striatum is similar to in vitro pri-amiRNA processing section), in vitro screens are useful for eliminating some precandidate pri-amiRNAs from further consideration.

Extent of harnessing of cellular miRNA biogenesis pathway and guide/passenger strand ratio provide dramatic differentiation of precandidate pri-amiRNAs in YAC128 mouse striatum

Next, we evaluated the efficiency and precision of pri-amiRNA processing in striatum samples from 3 YAC128 mice per group (1 positive control, 16 precandidate pri-amiRNAs, and 2 vehicle groups), using small RNA deep sequencing. The data comprised 15.5–46.5 million raw reads per sample with the correct adapter sequence.

To evaluate endogenous miRNA profiles in these samples, the percentage of reads aligned to annotated M. musculus miRNA hairpin sequences was assessed (Supplementary Fig. S4). In the vehicle groups for cohorts 1 and 2, an average of 67.7% ± 4.4% and 57.0% ± 0.2% (N = 3, mean ± standard deviation) of reads, respectively, was aligned to annotated M. musculus miRNA hairpin sequences and thus represented endogenous miRNAs; these served as the baseline endogenous miRNA profile for the two cohorts in this study. In animals treated with Treatments B–H (cohort 1) or Treatments I–N and P (cohort 2), 60.9–68.3% and 50.8–59.8% of reads on average, respectively, were aligned to endogenous M. musculus miRNA hairpin sequences, with no significant difference from the baseline from the corresponding vehicle group (p > 0.05 by one-way ANOVA with Dunnett's multiple comparison post hoc test). However, the endogenous miRNA levels after administration of Treatment A (cohort 1), Treatment O (cohort 2), and the positive control (cohort 1), 29.2% ± 7.3%, 30.0% ± 8.2%, and 54.6% ± 1.8%, respectively, were significantly lower than the baseline in the corresponding vehicle group (p < 0.0001, p < 0.0001, and p = 0.03, respectively, by one-way ANOVA with Dunnett's multiple comparison post hoc test). These results suggest that the endogenous miRNA profiles in the striatum were significantly affected by administration of Treatments A and O, as well as the positive control, but not Treatments B–N and P.

The level of precandidate miRNA relative to the total endogenous pool of miRNAs was evaluated (Fig. 3B). For the 6 samples from vehicle-treated animals, the detectable miHTT read number (with all 16 precandidate pri-amiRNAs used as reference) was <85 rpm, resulting in expression levels of 0.0% ± 0.0% (mean ± standard deviation) relative to the total detected endogenous pool of miRNAs. This represents the background signal for the level of precandidate miRNA relative to the total endogenous pool of miRNAs in this study. While the abundance of mature exogenous miHTTs relative to the total endogenous pool of miRNAs varied greatly across the 16 AAV1-iHtt vectors, spanning a wide, ∼1,100-fold range, 13 of the 16 AAV1-iHtt vectors had a relative abundance of <10%. The mean expression levels of mature exogenous miHTTs from Treatments A and O exceeded the sum of all endogenous miRNA levels in the same striatal tissue sample, which were significantly reduced relative to vehicle, suggesting that these precandidate pri-amiRNAs became dominant in Argonaute loading and overharnessed the endogenous miRNA biogenesis pathway. Although very similar levels of HTT mRNA lowering, 29–35% on average, were achieved with Treatments A, J, L, and P and the positive control, the level of mature exogenous miHTT relative to the total endogenous pool of miRNAs varied widely from 0.2% for Treatment P to 143.8% for Treatment A. Treatments J and L and the positive control showed intermediate levels of mature exogenous miHTT relative to the total endogenous pool of miRNAs at 5.7%, 0.8%, and 32.2%, respectively. These results demonstrate the importance of evaluating the extent to which the endogenous miRNA biogenesis pathway is harnessed in selecting a therapeutic pri-amiRNA, in addition to levels of target lowering in vivo. The level of mature exogenous miHTT relative to the total endogenous pool of miRNAs at a given level of HTT mRNA lowering is a key parameter that provides strong differentiation of different pri-amiRNA candidates in vivo, with lower levels of mature exogenous miHTT preferred.

As discussed in detail above, the G/P strand ratio varied substantially across the 16 AAV1-iHtt vectors, spanning a very wide range, from 0.0 to 2384 (Fig. 3C). Twelve of the 16 precandidate miHTTs exhibited G/P strand ratios >10 (Treatments A–C, E, and I–P, as well as the positive control), whereas 4 precandidate miHTTs (Treatments D, F, G, and H) showed G/P strand ratios <3, including 3 (Treatments D, F, and H) with ratios of <0.5, indicating that there were more than twice as many P strands as G strands for these 3 precandidate miHTTs. Thus, the G/P strand ratio is a parameter that provides strong differentiation of different pri-amiRNAs in vivo, with higher G/P strand ratios preferred.

When we assessed the precision of 5′ end processing of miHTT guide strands, we observed a batch effect when the two cohorts were compared (Fig. 3D). In the first cohort, the positive control showed 96.8% ± 3.0% (N = 3, mean ± standard deviation) of miHTT guides with correct 5′ ends, and six of eight precandidate pri-amiRNAs in this cohort had >90% on average of miHTT guide strands with precise 5′ ends. Only two precandidate pri-amiRNAs (Treatments B and F) exhibited slightly lower percentages of miHTT guide strands with precise 5′ ends (83.2% ± 2.2% and 85.7% ± 21.4%, respectively). In the second cohort, however, none of the precandidate pri-amiRNAs showed mean percentages of more than 90% of miHTT guide strands with precise 5′ ends. Nevertheless, the percentages of correct 5′ end start of the guide strand were in the range from 82.9% ± 0.0% to 89.8% ± 0.3%. These results indicate that the precision of 5′ end processing of miRNA guide strands in vivo provides modest differentiation of our precandidate pri-amiRNAs.

Combinatorial effect of G strand and P strand sequences together with scaffold on pri-amiRNA processing in the YAC128 mouse striatum

As expected, the scaffold was a determinant of pri-amiRNA processing. Different scaffolds with the same G and P strand sequences resulted in different levels of mature exogenous miRNA relative to the total endogenous pool of miRNAs (Fig. 3B) and different G/P strand ratios (Fig. 3C) in the YAC128 mouse striatum. For example, Treatments A and M comprised the same G and P strand sequences inserted into different scaffolds and within the context of each of these scaffolds exhibited highly distinct levels of mature exogenous miRNA relative to the total endogenous pool of miRNAs (143% for Treatment A vs. 1.8% for Treatment M) and different G/P strand ratios (423 for Treatment A vs. 135 for Treatment M).

However, the scaffold was not the sole determinant of pri-amiRNA processing. There was a dramatic interaction of the G and P strand sequences with the scaffold that resulted in a given scaffold producing very different levels of mature miRNA relative to the total endogenous pool of miRNAs and G/P strand ratios, depending on the specific G and P strand sequences within the scaffold. For example, although Treatments A and B comprised the same scaffold but contained different G and P strand sequences inserted into this scaffold, the level of mature exogenous miRNA relative to the total endogenous pool of miRNAs was 143% for Treatment A but only 0.1% for Treatment B. In addition, the G/P strand ratio for Treatment A was 423 but only 24 for Treatment B. As another example, Treatments D and J comprised the same scaffold, which was different from that of Treatments A and B, with different G and P sequences inserted into the scaffold. The level of mature exogenous miRNA relative to the total endogenous pool of miRNAs was substantially lower for Treatment D (0.1%) than for Treatment J (5.7%), and the G/P strand ratio was much higher for Treatment J (629) than for Treatment D (0.3). These results demonstrate the dramatic impact of the G and P strand sequences on pri-amiRNA processing for a given scaffold and support a combinatorial effect of scaffold and G and P strand sequences on pri-amiRNA processing.

Pri-amiRNA processing efficiency and precision in YAC128 mouse striatum is similar to in vitro pri-amiRNA processing

A comparison of pri-amiRNA processing in HeLa cells and neuronally differentiated SH-SY5Y cells in vitro versus YAC128 mice in vivo shows excellent concordance of the relative level of mature exogenous miHTT relative to the total endogenous pool of miRNAs, relative G/P strand ratios, and precision of 5′ end processing of miHTT guide strands.

The abundance of mature exogenous miHTTs relative to the total endogenous pool of miRNAs spanned a wide range in YAC128 mouse striatum (Fig. 3B), HeLa cells (Fig. 2B), and neuronally differentiated SH-SY5Y cells (Fig. 2B). Treatment K exhibited the lowest relative levels of 0.2% and 0.4% in neuronally differentiated SH-SY5Y and HeLa cells, respectively, and <0.05% in vivo, whereas Treatment A exhibited the highest relative levels of 168.2% and 200.8% in neuronally differentiated SH-SY5Y and HeLa cells, respectively, and 143.8% in vivo. Treatment O also exhibited very high relative levels, similar to Treatment A both in vitro and in vivo. In general, the treatments that exhibited low levels of precandidate miHTTs relative to the total endogenous pool of miRNAs in vitro also showed relatively low levels in YAC128 mouse striatum. For example, Treatments B, H, K, and P resulted in the lowest relative levels of mature exogenous miHTTs (<1% of the total endogenous miRNA pool) both in vitro and in vivo. Treatments D and L resulted in the next lowest relative levels of mature exogenous miHTTs both in vitro and in vivo, with <4% relative levels in vitro and <1% relative levels in vivo. Treatments C, E, F, G, I, J, M, and N provided intermediate relative levels in vitro (5–32%) and in YAC128 mouse striatum (1–11%). Thus, the rank order of precandidate pri-amiRNAs with respect to harnessing the endogenous RNAi pathway in YAC128 mouse striatum was predicted well by their rank order in vitro.

For a given precandidate miHTT, the G/P strand ratio differed by 1.5- to 5.1-fold compared across YAC128 mouse striatum (Fig. 3C), HeLa cells (Fig. 2C), and neuronally differentiated SH-SY5Y cells (Fig. 2C). In both YAC128 mice striatum and the two in vitro systems, G/P strand ratios were very high (>100) for Treatments A, C, E, I, J, M, and N (except that for Treatment N, the G/P strand ratio was 97.9 in HeLa cells), high (8–100) for Treatments B, K, L, and P, and low (<3) for Treatments D, F, G, and H. For Treatment O, the G/P strand ratio was 163.1 in YAC128 mouse striatum, 74.7 in HeLa cells, and 56.3 in neuronally differentiated SH-SY5Y cells.

The precision of 5′ end processing of the miHTT guide strand spanned a similar range in YAC128 mouse striatum, HeLa cells, and neuronally differentiated SH-SY5Y cells, ranging from 73% to 98% in vitro (Fig. 2D) and 83% to 98% in vivo (Fig. 3D). Treatments B and F were consistently on the lower end of these ranges in all three test systems. In contrast, other treatments such as Treatment O exhibited 94.5% and 92.1% precision of 5′ end processing of the miHTT guide strand in HeLa and neuronally differentiated SH-SY5Y cells, respectively, but only 82.9% precision of 5′ end processing of the miHTT guide strand in YAC128 mouse striatum.

Importance of efficient and precise pri-amiRNA processing in addition to target lowering in YAC128 mice and in vitro systems for selecting precandidates to evaluate in NHP

The in vivo results from YAC128 mice, taken together, led to our ranking of the precandidate pri-amiRNAs and selecting the subset to move forward into NHP. The requirement for statistically significant lowering of human mutant HTT mRNA in the striatum by at least 20% resulted in 7 of the 16 precandidate pri-amiRNAs remaining under consideration: Treatments A, B, F, J, L, M, and P. Treatments A, F, and J produced relatively high utilization of the endogenous RNAi pathway (5.7–143.8%) and thus were eliminated from further consideration. In addition, Treatment F exhibited a G/P strand ratio of <0.05, indicating a relatively high proportion of P strands generated relative to G strands, further supporting its removal from further consideration. Given the capacity to evaluate three of these AAV1-iHtt precandidate vectors in an NHP study, and the lower G/P strand ratio and consistently lower precision of 5′ end processing of the miHTT guide strand for Treatment B, Treatments L, M, and P were prioritized for the NHP study to select the lead candidate.

Precandidate AAV1-iHtt vectors are generally well tolerated in NHP

New batches of AAV1-iHtt vectors (AAV1-iHtt-1 corresponding to Treatment P, AAV1-iHtt-2 corresponding to Treatment M, and AAV1-iHtt-3 corresponding to Treatment L) were produced for the NHP study. Biochemical and biophysical characterization of these batches included titer, relative purity of the capsid proteins, the percentage of full capsids, aggregation and endotoxin levels, as well as genome integrity (Supplementary Fig. S5). Before conducting the study in NHP, the pharmacological activities of these newly produced batches of AAV1-iHtt vectors were confirmed in YAC128 mice (Supplementary Fig. S6). Human HTT mRNA reduction was evaluated by RT-qPCR, using mouse XPNPEP1 as the endogenous reference mRNA. Animals were necropsied at 4 weeks following bilateral intrastriatal injection of AAV1-iHtt-1 (corresponding to Treatment P), AAV1-iHtt-2 (corresponding to Treatment M), or AAV1-iHtt-3 (corresponding to Treatment L) at a dose of 3.6 × 10¹⁰ VG per animal, with seven to eight animals per group, and striatum punches were collected. As expected, significant HTT mRNA reduction in the striatum was observed for each of the three AAV1-iHtt vectors relative to the vehicle group (44%, 42%, and 43% lowering for AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3, respectively, compared with 35% HTT mRNA lowering for the positive control).

NHPs with no detectable or low levels of anti-AAV1 NAb activity in the serum underwent acute neurosurgical infusion of AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3 into the putamen at a dose of 9.00 × 10¹⁰ (one putamen in each of six animals) or 2.70 × 10¹¹ VG (one putamen in each of six animals) or vehicle (both putamens in each of three animals). All animals were followed for 36 ± 3 days. Assessment of MRI images collected during infusion procedures indicates that all infusions were successfully placed within the left and right putamen.

Postoperative assessments, including daily health observations, body weights, food consumption, and clinical pathology, all indicated that there were no significant alterations in the animals' health. There were no abnormalities observed in any of the major organs during gross necropsy. Importantly, there were no histopathologic findings in the putamen that were considered adverse, based on hematoxylin and eosin staining. In general, observations such as minimal to mild gliosis were noted in all groups, including the vehicle control, and were consistent with injection into the putamen and a needle track. In addition, administration of AAV was associated with minimal to mild mononuclear cell infiltrates. Thus, intraputaminal administration of AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3 was generally well tolerated in NHP over the 5-week in-life period.

Dose-dependent lowering of HTT mRNA in NHP putamen by precandidate AAV1-iHtt vectors

To evaluate HTT mRNA reduction in NHP putamen after intraputaminal injection of precandidate AAV1-iHtt vectors, animals were necropsied at 5 weeks following administration, and ten punches were collected from each putamen. HTT transcript levels were assessed using a bDNA assay for HTT mRNA and the endogenous reference mRNAs TBP, AARS, and XPNPEP1. Each of the three AAV1-iHtt vectors resulted in significant HTT mRNA reduction in the putamen relative to the vehicle group that was dose dependent (two-way ANOVA with Tukey's multiple comparisons post hoc test, p < 0.05; Fig. 4A). At the high dose (2.7 × 10¹¹ VG per putamen), there was 52%, 33%, and 53% HTT mRNA lowering on average for AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3, respectively, whereas at the low dose (9 × 10¹⁰ VG per putamen), there was 42%, 27%, and 40% HTT mRNA lowering on average, respectively. A two-way ANOVA main effects model with Tukey's multiple comparisons test was used and showed a statistically significant difference both by dose (F = 10.68, p = 0.0023) and by vector treatment (F = 31.89, p < 0.0001). There was significantly more reduction of HTT mRNA with AAV1-iHtt-1 and AAV1-Htt-3 than with AAV1-iHtt-2 at either the low or high dose (p < 0.01). However, there was no significant difference between AAV1-iHtt-1 and AAV1-iHtt-3 in reducing HTT mRNA at either dose level. The high dose of AAV1-iHtt-2 achieved the same level of HTT mRNA lowering as a threefold lower dose of AAV1-iHtt-1 or AAV1-iHtt-3. These results indicate that AAV1-iHtt-1 and AAV1-iHtt-3 are more potent than AAV1-iHtt-2 in lowering HTT mRNA in the NHP putamen.

OPEN IN VIEWER

Figure 4.

HTT mRNA lowering and pri-amiRNA processing in NHP putamen 5 weeks after intraputaminal administration of low (9 × 10¹⁰ VG per putamen) or high (2.7 × 10¹¹ VG per putamen) dose of AAV1-iHtt-1, AAV1-iHtt-2, or AAV1-iHtt-3. (A) Relative HTT mRNA levels in the NHP putamen. Tissue punches were measured by the bDNA assay for HTT, TBP, AARS, and XPNPEP1 mRNA levels. HTT mRNA levels were normalized to the geometric mean of the mRNA levels of the three housekeeping genes TBP, AARS, and XPNPEP1 and subsequently compared to the average normalized HTT mRNA level of the vehicle group. Each symbol represents the average relative HTT mRNA level in the ten punches from one putamen. N = 6 per AAV treatment per dose, N = 6 for vehicle. (B–D) Pri-amiRNA processing in the NHP putamen was evaluated by small RNA sequencing. N = 6 per AAV treatment group, one punch per putamen. (B) Level of mature exogenous miHTT relative to the total endogenous pool of miRNAs. (C) Ratio of guide to passenger strands of the mature exogenous miHTT. (D) Percentage of mature exogenous miHTT guide strands with precise 5′ ends. For all panels, the group mean ± standard deviation is shown for each treatment. AARS, alanyl-tRNA synthetase; bDNA, branched DNA; TBP, TATA-box binding protein. Color images are available online.

The importance of sufficient tissue sampling is illustrated in scatter plots of data from all punches (Supplementary Fig. S7). Substantial variation in HTT mRNA lowering was observed across the 10 putamen punches from a given animal at a given dose. This variability in HTT mRNA lowering across putamen punches is likely due to the variability in vector distribution within the putamen (as reflected in the variation in VG copies/diploid genome across putamen samples, shown in Supplementary Fig. S8B) and the relatively small size of the tissue punches from the putamen which reveal and highlight this heterogeneity. Of a total of 60 putamen punches analyzed per AAV1-iHtt dose group, the percentage of putamen punches with at least 30% HTT mRNA reduction at the high dose was 80%, 52%, and 77% for AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3, respectively, whereas the percentage of putamen punches with at least 30% HTT mRNA reduction at the low dose was 58%, 45%, and 62%, respectively.

Thus, the pharmacological potency for HTT mRNA reduction is comparable for AAV1-iHtt-1 and AAV1-iHtt-3, and greater than that for AAV1-iHtt-2, based on the dose levels evaluated.

Dose-dependent levels of VGs in NHP putamen after administration of precandidate AAV1-iHtt vectors

To quantify VG copy numbers in NHP putamen after intraputaminal injection of AAV1-iHtt vectors, whole cell DNA was purified from aliquots of lysates from the tissue punches used for mRNA measurements and were subsequently evaluated by ddPCR. Each of the three AAV1-iHtt vectors resulted in substantial numbers of VG copies in the putamen (Supplementary Fig. S8A): 272 and 517 VG copies/diploid genome for AAV1-iHtt-1, 230 and 527 VG copies/diploid genome for AAV1-iHtt-2, and 264 and 554 VG copies/diploid genome for AAV1-iHtt-3, at low and high dose levels, respectively, on average. VG copy number levels were below the lower level of quantification (BLLOQ) for most putamen punches from the vehicle group. When the BLLOQ samples were assigned a value of 0 and then a mean for each animal determined, the averages for the vehicle groups corresponding to the low and high dose ddPCR runs were 0 and 0.1 VG copies/diploid genome, respectively.

For all three AAV1-iHtt vectors tested, VG copy numbers were dose dependent but less than dose proportional. There was no significant difference in VG copy numbers between AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3 at the low dose, nor at the high dose (one-way ANOVA with Tukey's multiple comparisons test).

To evaluate the relationship between VG level and HTT mRNA lowering in the putamen, the relative levels of remaining HTT mRNA were plotted as a function of VG levels (Supplementary Fig. S8B). The number of VG copies detected for each vector correlated with HTT mRNA lowering in the putamen, with more VG copies per diploid genome resulting in more HTT mRNA lowering. Since the data appeared sigmoidal, a nonlinear regression model comprising a four-parameter curve fit was used, with the maximum and minimum asymptotes constrained at 100% and 20%, respectively. The effective concentration for producing 50% of the maximal response relative to the minimum level (EC₅₀)'s for HTT mRNA lowering calculated from these curve fits represent relative HTT mRNA levels of 60%, midway between the two asymptotes. These EC₅₀'s suggest that AAV1-iHtt-1 and AAV1-iHtt-3 are more potent than AAV1-iHtt-2, in that lower VG levels are required for HTT mRNA lowering for AAV1-iHtt-1 and AAV1-iHtt-3 than for AAV1-iHtt-2. Specifically, ∼40% HTT mRNA suppression was achieved with 40 VG copies/diploid genome after treatment with AAV1-iHtt-1, 273 VG copies/diploid genome after treatment with AAV1-iHtt-2, and 68 VG copies/diploid genome after treatment with AAV1-iHtt-3. Spearman correlation coefficients were −0.8968, −0.9087, and −0.8783 for the AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3 curves, respectively, indicating strong negative correlations between relative HTT mRNA levels and VG levels for all three AAV1-iHtt vectors.

Pri-amiRNA processing in NHP putamen after administration of precandidate AAV1-iHtt vectors is precise and efficient

Putamen samples from the same NHPs used to evaluate HTT mRNA lowering and VG levels were collected and submitted for small RNA deep sequencing. The 18 samples from the high dose groups had 23.1–39.6 million raw reads with the correct adapter sequence per sample, whereas the 18 samples from the low dose groups had 14.8–37.8 million raw reads with the correct adapter sequence per sample. In the vehicle group, there were 15.9–41.8 million raw reads with the correct adapter sequence per sample. These data indicate that all samples showed sufficient sequencing depth for evaluation of pri-amiRNA processing.

To evaluate endogenous miRNA profiles in putamen samples following AAV1-iHtt administration, the percentage of reads aligned to annotated rhesus miRNA hairpin sequences was assessed. In the vehicle group, an average of 54.1% of reads (range 47.7–61.1%) was aligned to annotated rhesus miRNA hairpin sequences (data not shown) and, thus, represented endogenous miRNAs; this served as the baseline endogenous miRNA profile in this study. In animals treated with AAV1-iHtt-1, AAV1-iHtt-2, or AAV1-iHtt-3 at either high or low dose, 53.4–55.8% of reads were aligned to endogenous rhesus miRNA hairpin sequences (data not shown), similar to the baseline of 54.1% from the vehicle group. These results suggest that the endogenous miRNA profiles for all putamen samples following AAV1-iHtt-1, AAV1-iHtt-2, or AAV1-iHtt-3 administration were not affected by treatment.

The level of mature exogenous miHTT relative to the total endogenous pool of miRNAs was evaluated (Fig. 4B). For the three animals (two putamen punches per animal) in the vehicle group, the detectable miHTT read number (with all three precandidate miRNAs used as reference) was <10 rpm, resulting in expression levels of 0% ± 0.00% relative to the total detected endogenous pool of miRNAs. As with the small RNA sequencing conducted on YAC128 mouse striatal samples, this represents the background signal for the level of precandidate miHTT relative to the total endogenous pool of miRNAs in this study. In NHPs treated with AAV1-iHtt-1 or AAV1-iHtt-3, but not AAV1-iHtt-2, the relative mean putaminal expression level of mature exogenous miHTT was dose related. At the low dose, the level of miHTT relative to the total endogenous pool of miRNAs was approximately fivefold lower for AAV1-iHtt-1 (0.35% on average) than AAV1-iHtt-3 (1.53% on average), indicating that AAV1-iHtt-1 treatment achieved a similar level of HTT mRNA lowering as AAV1-iHtt-3 at this dose, but with substantially less harnessing of the endogenous miRNA biogenesis pathway. The high dose of AAV1-iHtt-2 achieved a similar level of HTT mRNA lowering as the low dose of AAV1-iHtt-1 (Fig. 4A) with a comparable level of precandidate miHTT relative to the total endogenous pool of miRNAs. Thus, the level of mature exogenous miHTT relative to the total endogenous pool of miRNAs at a given dose and at a given level of HTT mRNA lowering is a parameter that differentiates these AAV1-iHtt vectors.

For each AAV1-iHtt vector, the G/P strand ratio was similar across the two dose levels administered (Fig. 4C). However, the G/P strand ratio varied substantially across AAV1-iHtt vectors. AAV1-iHtt-1 showed the highest G/P ratios (423.8 and 375.5 on average at the low and high doses, respectively), which were ∼10-fold higher than the G/P strand ratios for AAV1-iHtt-3 (46.0 and 43.6 on average at the low and high doses, respectively). The G/P strand ratios for AAV1-iHtt-2 were intermediate between those for AAV1-iHtt-1 and AAV1-iHtt-3. Thus, the G/P strand ratio is a parameter that differentiates these AAV1-iHtt vectors in NHP putamen, with AAV1-iHtt-1 exhibiting the most favorable G/P strand ratio.

The precision of 5′ end processing of miHTT guide strands was assessed. All three AAV1-iHtt vectors at the two dose levels administered showed high mean percentages (95–98%) of miHTT guide strands with precise 5′ ends (Fig. 4D), indicating that the 5′ end of the miHTT guide strands is primarily processed correctly. In addition, these results indicate that the precision of 5′ end processing of miHTT guide strands was very similar across the three AAV1-iHtt vectors, as well as across the two dose levels evaluated. Thus, the precision of 5′ end processing of miHTT guide strands does not differentiate these AAV1-iHtt vectors in NHP putamen.

Pri-amiRNA processing of three top precandidates in NHP putamen differs from in vitro cells and YAC128 mouse striatum

A comparison of pri-amiRNA processing of precandidates L (AAV1-iHtt-3), M (AAV1-iHtt-2), and P (AAV1-iHtt-1) in NHP putamen, YAC128 mouse striatum, HeLa cells, and neuronally differentiated SH-SY5Y cells shows excellent concordance of relative precision of 5′ end processing of miHTT guide strands (Table 1), given the RNA extraction and small RNA library preparation in different laboratories with different methods. Levels of mature exogenous miHTT relative to the total endogenous pool of miRNA were fairly similar in NHP putamen versus YAC128 striatum across the three candidates evaluated in NHP (Table 1). However, in vitro (Table 1), one of the three precandidates (Treatment M, corresponding to AAV1-iHtt-2) showed substantially higher levels of mature exogenous miHTT relative to the total endogenous miRNA pool, indicating that in vitro systems do not always predict this parameter well in YAC128 mice or in NHP. In addition, G/P strand ratios differed substantially in NHP putamen versus YAC128 mouse striatum and neuronally differentiated SH-SY5Y cells (Table 1) for one of the three precandidates (Treatment P, corresponding to AAV1-iHtt-1), indicating that neither in vitro systems nor YAC128 mice reliably predict this parameter well in NHP.

The precision of 5′ end processing of miHTT guide strands was high (>87%) and indistinguishable across the vectors AAV1-iHtt-1, AAV1-iHtt-2, and AAV1-iHtt-3 in NHP putamen at the low and high doses, in YAC128 mouse striatum, and in HeLa and neuronally differentiated SH-SY5Y cells. In addition, the absolute precision of 5′ end processing of miHTT guide strands was consistent across NHP putamen (95–98%), YAC128 mouse striatum (87–90%), and in vitro systems (93–97%), given the relatively lower 5′ end precision of miHTT guide strands measured in the corresponding entire cohort of YAC128 mouse striatum samples. Thus, in all systems evaluated—NHP putamen, YAC128 mouse striatum, HeLa cells, and neuronally differentiated SH-SY5Y cells—the 3 AAV vectors were indistinguishable with respect to the precision of 5′ end processing of miHTT guide strands.

The level of mature miHTT relative to the total endogenous pool of miRNAs was <4% in NHP putamen at both dose levels and in YAC128 striatum. At doses that provided the same degree of HTT mRNA lowering in NHP putamen (low dose of AAV1-iHtt-1, high dose of AAV1-iHtt-2, and low dose of AAV1-iHtt-3), AAV1-iHtt-1 and AAV1-iHtt-2 harnessed the endogenous RNAi pathway to a lower extent on average than AAV1-iHtt-3. In YAC128 mouse striatum, AAV1-iHtt-1 harnessed the endogenous pathway to the lowest extent on average followed by AAV1-iHtt-3 and then AAV1-iHtt-2. Taken together, these in vivo data from NHP putamen and YAC128 striatum suggest that AAV1-iHtt-1 harnesses the endogenous pathway to the lowest extent among the 3 AAV1-iHtt vectors. Similarly, in HeLa and neuronally differentiated SH-SY5Y cells, Treatments P and L (which correspond to AAV1-iHtt-1 and AAV1-iHtt-3, respectively) exhibited low levels of mature exogenous miHTT relative to the total endogenous pool of miRNA. However, in contrast, Treatment M (which corresponds to AAV1-iHtt-2) exhibited noticeably higher relative levels of mature exogenous miHTT in vitro (27.9% and 9.2% in HeLa and neuronally differentiated SH-SY5Y cells, respectively) than in vivo (1.8%).

The relative and absolute G/P strand ratios for the precandidate pri-amiRNAs in NHP putamen contrasted substantially with these ratios in YAC128 mouse striatum and in neuronally differentiated SH-SY5Y cells, especially for one of the precandidate pri-amiRNAs. For AAV1-iHtt-2, the average G/P strand ratios were indistinguishable, comprising 135.7, 147.3, 135.3, and 126.3 for NHP putamen at low dose, NHP putamen at high dose, YAC128 striatum, and neuronally differentiated SH-SY5Y cells, respectively. However, for AAV1-iHtt-1, the average G/P strand ratios were ∼6- to 10-fold higher in NHP putamen (423.8 and 375.5 at low and high doses, respectively) than in YAC128 striatum (39.9) and neuronally differentiated SH-SY5Y cells (70.7). In contrast, for AAV1-iHtt-3, the average G/P strand ratios were approximately twofold lower in NHP putamen (46.0 and 43.6 at low and high doses, respectively) and neuronally differentiated SH-SY5Y cells (38.3) than in YAC128 striatum (82.9). Thus, for AAV1-iHtt-1, the G/P strand ratio in NHP putamen was substantially superior to either neuronally differentiated SH-SY5Y cells or YAC128 striatum, whereas for AAV1-iHtt-2 and AAV1-iHtt-3, the G/P strand ratios in NHP putamen, neuronally differentiated SH-SY5Y cells, and YAC128 striatum were more similar.

Selection of optimized pri-amiRNA based on efficient and precise pri-amiRNA processing, in addition to HTT lowering and general tolerability

Taken together, the in vivo results in NHP putamen that comprised HTT mRNA lowering and pri-amiRNA processing were the primary basis of selecting the pri-amiRNA sequence in AAV1-iHtt-1 as the optimized pri-amiRNA for the treatment of HD. At the doses used, AAV1-iHtt-1 showed target levels of HTT mRNA lowering in the NHP putamen of 40% and 53% at the low and high doses, respectively, concomitant with: (A) minimal harnessing of the endogenous miRNA biogenesis pathway (0.35% and 2.20% at the low and high doses, respectively), (B) a high G/P strand ratio (423.8 ± 195.9 and 375.5 ± 118.4 at the low and high doses, respectively), and (C) precise 5′ end processing of the miHTT G strand (97.3% and 97.6% at the low and high doses, respectively). Furthermore, the dose levels used for the 5-week NHP study were well tolerated based on daily health observations, body weights, food consumption, clinical pathology, and lack of abnormalities in any of the major organs during gross necropsy. In addition, the NHP results were supported by YAC128 mouse and in vitro studies that demonstrated similar robust HTT mRNA lowering accompanied by efficient and precise processing of the pri-amiRNA to the miHTT G strand without perturbing the endogenous miRNA pool.