To directly assess the consequences of CAG expansion mutation on HTT physiological function we compared the phenotypes of a HD hESC line, where one HTT allele carries a 72CAG expansion mutation (HTT ^72/20, HD), with isogenic lines producing half of the physiological level of HTT protein due to monoallelic inactivation (HTT ^20/-, heterozygous) or none due to biallelic inactivation of HTT gene (HTT ^−/−, hereafter referred as HTT-KO) in hESC that we previously generated (Ruzo et al., 2018) (Figures 1A–C). In wild-type (HTT ^20/20, WT) hESC confluent micropattern culture, following 1 h stimulation with activin A (Figure 1D), cells at the edge of the colonies were activated and accumulate SMAD2/3 in the nucleus in response to apical activin A induction (Figure 1E, F), while cells at the center of the micropattern colonies did not show nuclear translocation of SMAD2/3. On the other hand, and as previously shown (Galgoczi et al., 2021), micropattern colonies formed by HD lines lost the spatial restriction of sensitivity to activin A, allowing the response to the morphogen in cells residing at the center region of the micropatterns (Figure 1E, F). In this experimental setting, the heterozygous line (HTT ^20/-) showed a very subtle increase in SMAD2/3 nuclear signal in the center of the colony and overall significant increase in the total number of SMAD2/3⁺ nuclei (Figure 1F, G), while lack of HTT expression (HTT-KO) resulted in the nuclear translocation of SMAD2/3 throughout the micropattern colony. This indicates that a reduced level of HTT expression results in an HD-signature phenotype, albeit significantly milder than complete lack of HTT expression or monoallelic HD mutation.

The consequences of lower HTT expression were also explored in the neuruloid formation assay (Figure 1H) (Haremaki et al., 2019). Here, HD hESC (HTT ^72/20) showed a loosened central neuroectodermal domain, measured as an increased PAX6⁺ area and neural-cadherin (N-CAD) lumen, along with a contraction of the neural crest lineage, quantified as a decrease of the area occupied by SOX10⁺ cells (Figures 1I, J) (Haremaki et al., 2019). The neuruloids derived from the heterozygous (HTT ^20/-) line, that express half of the physiological HTT level, displayed an HD-like phenotype in all three measurements. However, the extent of this phenotype was less severe than the phenotype observed in HD lines, which grossly failed to compact the central neuroectodermal area, a phenotype only worsened by complete lack of expression of HTT, as in the HTT-KO cell line (Figures 1F, G).

Taken together, these results show that reduced expression of HTT recapitulates aspects of the phenotypes observed in HD samples, suggesting that HD mutation might result in a loss of HTT function. However, when compared to the more severe HD phenotype, the milder HD-like phenotypes displayed by heterozygous cells provides evidence supporting the idea that monoallelic CAG expanded HTT might also exert deleterious activity that results in more severe phenotypes.

To dissect the effect of HD mutation on HTT function and its role in these platforms, we designed a transgenic system that allows us to modulate the expression of wild-type (wtHTT) and mHTT in the various isogenic backgrounds of the RUES2 HD allelic series (Ruzo et al., 2018). For this purpose, we engineered two PiggyBac vectors containing the coding sequence of full-length human HTT carrying a normal trinucleotide stretch (wtHTT, 20CAG), or an expanded CAG stretch (mHTT, 56CAG), expressed under the regulation of a constitutive promoter. At the 3′ end of the HTT coding sequence, an in-frame sequence coding a Glycine-Serine flexible linker and the HaloTag coding sequence were introduced, resulting in a C-terminal tagging of the transgenic full-length HTT proteins (Figure 2A). These vectors (wtHTT or mHTT) were stably transfected into the HTT-KO cell line, and successful in-frame full-length expression was verified by immunoblot, with detection of the HTT protein as well as the HaloTag amino acid sequences, confirming that the abundance of recombinant HTT protein resulting from the two transgenes in the generated cell lines were equivalent (Figure 2B; Supplementary Figure S3). To determine whether the HaloTag 3′-end/C-terminal fusion affects HTT activity, we tested for its ability to rescue the phenotype of HTT-KO hESCs in the two micropatterned-based assays. First, we performed the 1h activin A induction protocol where expression of wtHTT was able to rescue the phenotype of HTT-KO cells, resulting in a SMAD2/3 nuclear translocation restricted to edge of the micropatterns (Figure 2C, D). In contrast, expression of mHTT failed to rescue the HD-like phenotype of edge restricted SMAD2/3 activation (Figure 2C, D). Similarly, exogenous expression of wtHTT was also able to rescue the HD-like phenotype of HTT-KO hESC in the neuruloid assay, allowing for proper compaction of the central neuroectodermal region, measured as reduced area of the PAX6⁺ domain and the N-CAD⁺ lumina (Figures 2E, F), as well as rescue of the neural crest SOX10⁺ area. However, exogenous expression of full-length mHTT in HTT-KO hESC only led to partial rescue of its HD-like phenotype shown by a less pronounced reduction in the area PAX6⁺ and N-CAD⁺ domains, and milder increase in SOX10⁺ domain, indicating that in this more complex model, mutant HTT retains some activity for at least a partial rescue of the phenotype in cells devoid of HTT expression.

These results indicate that CAG-expansion might directly impair the function of HTT.

After assessing the functionality of wtHTT and mHTT transgenic insertion in HTT-KO hESCs, we exogenously expressed them in WT hESCs to directly test the effect of mHTT on the two micropattern assays (Figures 3A, B; note that recombinant HTT-Halotag proteins are ∼33 kDa larger than endogenous wtHTT due to the C-Terminal tag). As a control, overexpression of transgenic wtHTT in WT hESCs did not affect the ability of WT cells to spatially restrict sensitivity to activin A upon 1 h of stimulation, with nuclear translocation of SMAD2/3 limited to the edges of their colonies (Figures 3C, D). Similarly, WT hESCs overexpressing wtHTT were able to generate neuruloids that properly form a compact central neuroectodermal structure (PAX6/N-CAD), which was surrounded by a well-formed neural crest structure (SOX10) (Figures 3E, F).

On the other hand, exogenous expression of mHTT profoundly affected the ability of WT hESCs to perform in the two micropatterns assays. Upon 1 h of activin A stimulation, large patches of cells exhibiting nuclear translocation of SMAD2/3 appeared toward the center of the colony, reminiscent of the HD phenotype previously shown (Figures 3C, D). Likewise, in the neuruloid assay, a similar effect was observed with the micropatterns presenting grossly misshapen and enlarged central domains (as shown by PAX6 and N-CAD signal) as well as a reduction of the neural crest lineage (SOX10), recapitulating the phenotype previously characterized for the HD line (Figures 3E, F).

These observations show that the expression of a mutant form of HTT in a fully competent WT (HTT^20/20) background is sufficient to elicit HD-like phenotypes. This indicates that CAG expansion mutation on the HTT gene results in some kind of detrimental activity that might interfere with physiological processes performed by endogenous HTT in WT cells.

Since exogenous expression of mHTT induces HD-like phenotypes in WT hESCs, while only partially rescuing the HD-like phenotypes of HTT-KO hESCs, we hypothesized that CAG expansion mutation might result in a reduced function of HTT and in some gained activity that somehow interferes with the physiological function of wild-type HTT gene products. To test this hypothesis, we sought to genetically complement an HD hESC line, carrying monoallelic 72CAG expansion mutation (HTT^72/20), with our construct encoding full length wtHTT and determine whether excess of wtHTT protein would rescue its HD-signature phenotypes (Supplementary Figure S4A). Stable protein translation of HTT-HaloTag was verified by HTT and HaloTag protein immunodetection (Supplementary Figures S4B, S4C).

Differently from the HTT-KO cell line, complementation of HD hESC lines with wtHTT partially, but significantly, rescued the HD phenotype on the activin A assay (Supplementary Figures S4D, S4E). Likewise, on the neuruloid assay a similar partial rescue was observed, with the central neuroectoderm domain displaying an improved level of compaction, as shown by a smaller area of PAX6+ cells and N-CAD⁺ lumen, when compared to the parental HD cell line (Supplementary Figures S4F, S4G).

The partial improvement of the HD-signature phenotypes observed in these experiments provides evidence in support of the hypothesis of a dominant negative effect associated with HD mutation.

In the experimental context described above, the inability to fully rescue these phenotypes could be the result of a heterogeneous expression level of the exogenous wtHTT transgene or that a certain amount of extra wtHTT gene products might be required to overcome the dominant effect of endogenous mHTT. To address this possibility, we took advantage of the fused HaloTag in our recombinant HTT proteins (Figure 4A) to enrich for cells homogenously producing low or high levels of wtHTT-HaloTag protein by fluorescence activated cell sorting (FACS) (Figure 4B). Thus, HD cells complemented with wtHTT-HaloTag were sorted for low levels of HTT (hereafter named as Low), equivalent to endogenous HTT protein level (determined by using a WT knock-in HTT-HaloTag reporter cell line as reference). A population of HD hESC complemented with wtHTT-HaloTag was also sorted for high level of the wtHTT recombinant protein and hereafter named as High (Figure 4B). HaloTag labelling confirmed that Low and High homogeneously express the transgene at two different levels (Figure 4C), also shown by immunoblot analysis (Figure 4D). Additionally, linear quantification by MesoScale Discovery Analysis showed that the amount of full-length wtHTT recombinant protein in Low is estimated as expected at ∼2x the endogenous amount of HTT protein, and High at >10x the level measured in control WT (Figure 4E).

In the HD line complemented with both the Low and High wtHTT protein levels, SMAD2/3 nuclear translocation in response to activin A induction in micropattern cultures resulted in a pronounced WT-like phenotype (p = ns vs. WT cells) (Figures 5A, B). Quantification of the percentage of SMAD2/3 nuclei showed that the number of cells responding to activin A stimulus decreased with overexpression of wtHTT, but without reaching the lower level of activated cells observed in WT colonies, regardless of the expression level of the transgene (Figure 5C). Additionally, the degree of rescue was equivalent for Low and High wtHTT levels, indicating that in the context of this assay, complementing an HD cell line with wtHTT only partially rescues the HD phenotype, independent of expression level.

In the neuruloid assay, complementation of the HD cell line with low levels of wtHTT was enough to elicit a rescue towards WT-like phenotype, with improved compact neuroectodermal (PAX6⁺) and N-CAD domains, as well as a larger neural crest SOX10⁺ domain than the parental HD line (Figures 5D, E). On the other hand, complementation with wtHTT protein at High levels resulted in a more pronounced rescue, both in extent of the compaction of the central neuroectoderm domain as well as its morphology, with colonies reaching full closure of the central area and rescue of the area occupied by neural crest cells.

Taken together, these results confirm the hypothesis that, at least partially, the HD signatures observed in a spatially confined pluripotent monolayer in response to activin A stimulation and in the neuruloid self-organization of ectoderm domains are the result of mHTT that acts in a dominant negative fashion. We show that CAG-expanded HTT expression impairs the normal function of wtHTT, and that mHTT impairment can be partially rescued with increased expression of wtHTT, although never to a full extent even with levels as high as 10 times greater than the endogenous level.

Cell lines used in this study were derived from the RUES2 cell line, part of the NIH Human Embryonic Stem Cell Registry (RUES2-NIHhESC-09-0013). hESCs were maintained in mouse embryonic fibroblast conditioned HUES Media in feeder-free conditions (conditioned media—CM), supplemented with bFGF 20 ng/mL, and replaced daily. hESCs were passaged at 70%–80% confluency every 3–4 days into Geltrex-coated dishes (Life Technologies) using Gentle Cell Dissociation Reagent (STEMCELL Technologies). Cells were tested for Mycoplasma spp. at 2-monthly intervals.

Human ESCs were nucleofected using an Amaxa nucleofector II in Nucleofector Solution L (all Lonza) using B-016 preset. A total of 10⁶ hESCs were nucleofected with 1 µg of PiggyBac HTT-HaloTag plasmid (wtHTT or mHTT according to each experiment) and 1 µg of the PiggyBac transposase (Lacoste et al., 2009). Cells were plated on Geltrex coated plates in CM supplemented with bFGF 20 ng/mL and 10 µM ROCK-inhibitor (Y-27632) for 24 h. Cells were then selected with puromycin 1 µg/mL for up to 10 days, and stable transfection validated by in situ labelling, flow cytometry, and immunoblot to confirm detection of HTT-HaloTag fused recombinant protein.

Protein was extracted on ice using ice cold RIPA Lysis and Extraction Buffer (Thermo Scientific 89900) supplemented with 1x Halt™ Protease and Phosphatase Inhibitor Cocktail (Thermo Scientific 78441). Cell lysates were briefly sonicated using a Branson Sonifier 450 (VWR; output 4, duty-cycle 10%, allowing 5 bursts) and incubated on ice for 15 min. Lysates were cleared by centrifuging at 12,000 rpm for 5 min at 4°C and protein concentration determined using the Pierce™ BCA Protein Assay (Thermo Scientific 23225). Samples were prepared for SDS-PAGE in 1x NuPAGE™ LDS Sample Buffer containing 50 mM DTT, boiled at 95°C for 5 min and loaded into a NuPAGE 3%–8% Tris-Acetate Gel (Thermo Scientific) along with HiMark™ Pre-stained Protein Standard (Thermo Scientific LC5699). Electrophoresis was run on ice and under stirring for 1h30 at 150 V. Protein bands were transferred to a Trans-Blot Turbo 0.2 µm PVDF membrane (Bio-Rad Cat.#: 1704156) using a Trans-Blot Turbo Transfer System (Bio-Rad) and membranes blocked in Tris-buffered saline containing 0.1% Tween-20 (TBS-T) and 5% milk for 30 min. Primary antibodies were incubated overnight at 4°C (see Table S1 for antibody information) and HRP secondary antibodies for 30 min at RT. All incubations were carried under constant agitation and washed thoroughly between steps. HRP was detected using either Clarity or Clarity Max ECL Western blotting Substrates (Bio-Rad) and chemiluminescence detected in a Chemidoc MP system (Bio-Rad).

Spatially confined hESC response to activin A was performed as previously described (Galgoczi et al., 2021). Briefly, custom micropatterned glass coverslips presenting 500 µm growth areas (CYTOOCHIPS Arena 500A) were covered with 800 µL of 10 μg/mL recombinant human laminin 521 (BioLamina) diluted in PBS^+/+(Gibco) for 3 h at 37°C. Coated coverslips were transferred to 35 mm dishes containing 5 mL of PBS^+/+ and serially washed 5 times, avoiding full removal to avoid drying, followed by one final complete wash with PBS^+/+. Cells lines were precultured in parallel to ensure similar growth upon the start of the assay. hESCs were dissociated in single cells using Accutase (STEM CELL Technologies) and 0.8 × 10⁶ cells seeded per micropattern in 3 mL of CM with 20 ng/mL bFGF, 100 μg/mL Normocin (InvivoGen), ROCK inhibitor and Y27632 (10 μM; Abcam ab120129). Micropatterns were left unperturbed for 10 min to ensure homogenous distribution across the patterns and then transferred to 37°C. ROCK inhibitor was removed 3 h after seeding and upon 18 h incubation, the monolayers were induced with 100 ng/mL of activin A (R&D Systems). After 1 h incubation, samples were washed, fixed, and processed for immunofluorescence.

Neuruloid culture (Haremaki et al., 2019) was performed in coated 500 µm micropatterns as described above. Briefly, 0.5 × 10⁶ cell were seeded per micropattern and incubated for 3 h at 37°C in HUES medium supplemented with 20 ng/mL bFGF, 10 μM ROCK inhibitor Y27632 and 100 μg/mL Normocin. Cultures were kept in Normocin for the whole experiment. The seeded micropatterns were then washed once with PBS^+/+ and day 0 initiated in HUES with 10 μM SB431542 and 0.2 μM LDN 193189. On day 3, media was replaced with HUESM containing 10 μM SB431542 and 3 ng/mL BMP4, with media replenished at day 5. Micropattern cultures were then fixed at day 7 and processed for immunofluorescence analysis.

Cells were fixed with 4% paraformaldehyde (Electron Microscopy Sciences 15713) for 30 min at RT, washed thrice with PBS and permeabilized and blocked DPBS containing 0.5% Triton-X 100 (Sigma 93443) and 3% Normal Donkey Serum for 30 min. Primary antibody staining was performed in DPBS with 0.1%TX-100 (PBST) for 1 h 30 at room temperature, washed thrice with PBST for 5 min each, and secondary antibody staining was performed for 1 h in PBST at RT. After 2 washes with PBST and 1 PBS wash, samples were counterstained with 0.1 μg/mL DAPI (Thermo Fisher Scientific D1306) for 10 min and washed. Micropattern coverslips were mounted on slides using ProLong Gold antifade mounting medium (Molecular Probes P36934). hESCs grown in IBIDI slides were mounted in IBIDI mounting medium and stored at 4°C before imaging. Antibodies used can be found in Table S1.

Cells grown under confluency were incubated in CM+bFGF media containing 1 μM cell permeable Janelia Fluor^® 549 HaloTag ligand complex (Promega, GA1110) for 1 h at 37°C. Cells were washed 3 times with PBS^+/+ and incubated with fresh media to allow unbound ligand to be fully removed for 3 h. Monolayers were washed again with PBS−/−, dissociated with Accutase and resuspended in PBS −/−, containing HEPES 10 mM, EDTA 5 mM, BSA 0.5%. Duplets were excluded based on FSC-W and SSC-W gating and DAPI was used at 20 ng/mL for cell death exclusion. Cells were sorted in a FACS Aria Flow Cytometer (BD) using a 100 µm/20psi nozzle. An endogenous C-terminal tagged CRISPR HTT reporter with HaloTag (unpublished) was used to define the gating for Low levels of HTT protein, and High levels were defined as all cells above that threshold. The parental untagged cell line was treated with HaloTag ligand in parallel and used as negative control. Cells were allowed to recover after sorting in Normocin and the obtained populations analyzed by flow cytometry, Western blot and HTT MSD quantification.

To quantify total HTT protein levels on the generated cell lines, about 3 × 10⁶ cells were pelleted and flash frozen in triplicates and sent for contracted analysis at Evotech with MSD assay CHDI_HTT_144 (antibody pair 2B7/D7F7) following SOP. Briefly, samples were lysed in ice cold MSD Tris Lysis Buffer and incubated 30 min at 4°C by rotating. After lysis, the samples were centrifuged (15700 rcf, 4°C; 6 min) and protein concentration determined with BCA assay and adjusted to 0.2 mg/mL, further diluted to 0.1 mg/mL in blocking buffer. Final buffer conditions for HTT quantitation assays were 20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton-X, 1x phosphatase inhibitor I and II (Sigma), complete mini EDTA free protease inhibitor cocktail tablet (Roche), 0.5 mM PMSF, 10 mM NaF. Purified recombinant full-length 1-3144 protein of human Huntingtin protein with 23 polyglutamine (CHDI-90001858) was used as standard, ranging between 0.001 and 10 fmol/μL.

For imaging, hESCs were grown on glass surfaces previously coated with 10 μg/mL recombinant human laminin 521 (BioLamina) diluted in PBS^+/+(Gibco) for 3 h at 37°C. Freshly thawed Janelia Fluor^® 549 HaloTag^® Ligand (Promega, GA1110) was diluted to 200 nM in warmed CM and cells incubated for 30 min. After washing thrice with PBS^+/+, cells were incubated for 30 min to allow unbound ligand to wash. Cells were then fixed in 4% PFA, nuclei counterstained with DAPI and imaged by confocal microscopy.

Confocal microscope images of hESCs and neuruloid IF, were acquired with a 10x or 20×/0.8 NA dry objective lens (Zeiss), using 405, 488, 561, and 633 nm laser lines, and a combination of PMT and GaAsP detectors (LSM 780, Zeiss). Large tilled imaged were obtained to sample the neuruloid assay. Activin A assay was imaged acquiring large tiled areas with an inverted wide-field epifluorescence microscope using 10×/0.45 NA dry objective lens (Zeiss Axio Observer Z1).

Fiji (2.0.0/1.53t) was used to stitch tiled images and to generate maximum projections of the generated confocal z-stacks. SMAD2/3 radial intensity profiling was performed as previously described with slight changes (Galgoczi et al., 2021). Briefly, using a custom Python script, a foreground mask was created to detect colonies by thresholding the DAPI channel and calculating alpha shapes with respect to predicted colony size, and colonies extracted into individual files. Ilastik (1.4.0b27, Berg et al., 2019) was then used to segment individual nuclei based on the DAPI channel. Identified nuclei were then filtered according to expected size to exclude debris and mitotic cells, and the center of the nuclear mask was determined for further watershed segmentation (scikit-image). The generated nuclear mask was then applied to the SAMD2/3 channel and median nuclear intensities were extracted, and SMAD2/3 radial profile was plotted according to each nucleus distance to the center of the colony. Profiles were normalized to 1 using the maximum value obtained in the profile.

To assess the total SMAD2/3 number, individual nuclei were segmented using Stardist (Schmidt et al., 2018; https://arxiv.org/abs/1806.03535), and the corresponding nuclear masks were used to calculate the mean intensity for each individual nucleus (scikit-image). Images were manually thresholded for SMAD2/3 to classify nuclear translocation as binary. The percentage of SMAD2/3 positive nuclei was then calculated on the total nuclei number.

Individual neuruloid images were obtained as explained above. Ilastik was used to segment the areas occupied by each marker and DAPI was used to determine the total neuruloid area. The pixel area for each marker was then determined and normalized as a fraction of the total DAPI area. Python libraries numpy, pandas, matplotlib and seaborn were used to organize the data and for plotting.

All data presented in this study were obtained from at least two independent experiments. Statistically significant differences between conditions were determined using one-way analysis of variance (ANOVA) followed by Dunnett’s post hoc test for correction of multiple comparisons. Statistical analyses were performed in Prism 9 (GraphPad). Results were considered not significant (ns) when the observed differences when p >0.05, and significance levels were denoted as follows: *p <0.05, **p <0.01, ***p <0.001 and ****p <0.0001.