Background

Polyglutamine (polyQ) diseases are a family of neurodegenerative disorders that include spinocerebellar ataxia type 3 (SCA3) and Huntington’s disease (HD), caused by a CAG repeat expansion in the coding region of specific genes [1, 2]. Due to an autosomal dominant inheritance pattern, most of polyQ disease patients carry a normal (wild type, WT) and mutant (MUT) allele. Normal alleles usually contain 5–30 CAG repeats, while mutant alleles, encoding proteins with long stretches of glutamines, are characterized by more than 39 CAG repeats in HD and above 60 CAG repeats in the case of SCA3 [2, 3].

SCA3 is caused by a CAG repeat expansion in the exon 10 of the ataxin-3 (ATXN3) gene and typically manifests with dysfunction and degeneration of neurons in cerebellum and spinocerebellar tracts [4]. HD, in turn, is caused by a CAG repeat expansion in the first exon of the huntingtin (HTT) gene. In HD, the mutation leads mostly to degeneration of striatal and cortical neurons. However, other brain regions may also be substantially affected in both SCA3 and HD [5].

A mutant protein is generally recognized to be the main pathogenic factor in polyQ diseases. Expanded polyQ tracts induce the formation of misfolded protein aggregates that disturb cellular homeostasis and lead to neuronal death [2, 6,7,8,9]. The exact role of mutant transcripts in pathogenic mechanisms in polyQ diseases is being unraveled [10,11,12,13]. RNA-related studies have demonstrated so far: incomplete splicing of HTT mRNA [14], abnormal interactions of RNAs containing expanded CAG repeats with proteins [15, 16], and RNA foci formation [17, 18], and potential importance of alternative polyadenylation of various polyQ transcripts [19]. Furthermore, mutant polyQ disease transcripts have been shown to be promising therapeutic targets in strategies that aim to downregulate mutant gene expression [2, 9, 20].

Molecular examination of endogenous polyQ transcripts is a challenging task. PolyQ disease genes are expressed at low levels across all tissues, including the brain regions that are mostly affected by the disease [21]. Any method that seeks to inspect polyQ disease transcripts needs to be highly accurate and sensitive. Moreover, if method of choice requires discrimination between WT and MUT alleles, the CAG sequence region cannot be used to design a primer or probe. One possible solution to overcome these limitations is to utilize heterozygous SNP variants present in transcripts. Droplet digital PCR (ddPCR) utilizes droplet-based microfluidics and compartmentalization to provide absolute quantification of analyzed molecules and has been used for allele-selective quantification of transcripts [22,23,24,25].

In this study, we first identified heterozygous SNP variants in ATXN3 and HTT cDNAs from patient-derived cell lines. Next, we adapted ddPCR to perform accurate and quantitative determination of the WT/MUT allele ratio and estimation of the number of the ATXN3 and HTT transcripts per cell. We performed our analyses on fibroblasts and induced pluripotent stem cells (iPSCs), as well as neural stem cells (NSCs) and neurons, differentiated from iPSCs. Using these cell lines, we demonstrated changes in WT/MUT allele expression ratios emerging from a neuronal differentiation process. Additionally, we determined the number of HTT WT and MUT transgene copies and their expression ratio in the brain tissue of biallelic HD mice. The obtained results provide a starting point for further investigation of the relevance of expression of WT and MUT alleles to the pathology of polyQ diseases.

Results

Examination of endogenous ATXN3 transcripts

For the ddPCR assay measuring the expression of ATXN3 WT and MUT alleles, we used a set of patient-derived cell lines with the SCA3 genetic background: fibroblasts, iPSCs, NSCs, and neurons (characterized by 16/67 CAG repeats in ATXN3) (Fig. 1a).

Fig. 1
figure 1

SNP-based allele-specific ddPCR assays for ATXN3 transcript. a SCA3 cell lines used; R—reprogramming; D—differentiation. b Scheme of ATXN3 transcript with marked CAG repeat tract (green triangle) and SNP variants identified in SCA3 cell lines (red arrows—SNPs used in ddPCR assays; black arrows—other identified SNPs); light orange box—UTRs. Table presents all SNP variants identified in analyzed SCA3 cell lines. Bolded SNPs were selected to be targets for ddPCR assays, and they will be hereinafter referred to as ATXN3_SNP2 and ATXN3_SNP5 in the text. CDS—coding sequence. RefSNP number according to the Single Nucleotide Polymorphism Database (dbSNP). c Results from ddPCR analysis of ATXN3_SNP2 assay specificity performed on seven samples with predefined ratios of WT/MUT ATXN3 plasmids (samples: I—100% WT and 0% MUT; II—90% WT and 10% MUT; III—75% WT and 25% MUT; IV—50% WT and 50% MUT; V—25% WT and 75% MUT; VI—10% WT and 90% MUT; VII—0% WT and 100% MUT). Precise values are indicated on WT/MUT bars ± poisson error. d Percentage ratio of ATXN3 WT allele obtained with ddPCR ATXN3_SNP5 assay using gDNA from indicated SCA3 cell lines. Three biological replicates were performed. These data are presented as means ± SD

Full size image

Identification of SNP variants in ATXN3 transcripts

To identify SNP variants, we cloned and sequenced ATXN3 cDNA from SCA3 NSCs and identified 5 heterozygous SNPs in exons of the ATXN3 gene (Fig. 1b). Two of these ATXN3 SNPs (rs1048755 and rs910369, referred to as ATXN3_SNP2 and ATXN3_SNP5) were selected to design ddPCR assays. These SNPs either lead to missense mutation in N-terminal ubiquitin-interacting motif 1 (UIM1) of ataxin-3 [26] or are considered to cause alterations of RNA-binding protein (RBP)-binding sites in ATXN3 mRNA [27].

Validation of the ATXN3 ddPCR assay specificity and accuracy

To examine specificity of allelic discrimination of ATXN3 ddPCR assays, we performed ATXN3_SNP2 and ATXN3_SNP5 ddPCRs on predefined mixtures of plasmids containing either WT or MUT sequences with respective SNPs. We observed that the ratio of the signal originating from WT and MUT probes was extremely consistent with the ratio of premixed WT/MUT plasmids for both SNPs: SNP2 (Fig. 1c) and SNP5 (Additional File 1: Fig. S1A). To further validate the ATXN3 ddPCR accuracy, we examined gDNA isolated from SCA3 fibroblasts, iPSCs, NSCs, and neurons. We used the ATXN3_SNP5 assay that utilizes the SNP variant located in the 3′UTR of the ATXN3 gene, rendering the assay suitable for gDNA analysis. We found that in all analyzed gDNA samples, the WT/MUT allele ratio was remarkably close to the expected 50:50 (Fig. 1d).

Determination of the ATXN3 mRNA allele ratio in SCA3 cell lines

We examined allele-specific expression patterns of ATXN3 in SCA3 cell lines: patient-derived fibroblasts, iPSCs reprogrammed from fibroblasts, as well as NSCs and neurons differentiated from iPSCs (Fig. 1a). To confirm cellular identity of generated neurons, we performed immunostaining and observed strong expression of MAP 2 and TUJ1 neuronal markers (Additional File 2: Fig. S2). When we quantified the ATXN3 allelic expression ratios in SCA3 cell lines, we observed 54.7% WT/45.3% MUT ratio in fibroblasts, 53.2% WT/46.8% MUT ratio in iPSCs, 53.9% WT/46.1% MUT in NSCs, and 53.8% WT/46.2% MUT ratio in neurons (Fig. 2a–d, precise values for each ATXN3 assay are given in Additional File 3: Fig. S3). We detected significantly higher expression of the WT allele, as compared to MUT allele, in almost all analyzed cell types. Two-way ANOVA showed that there are no statistically significant changes in the WT abundance between cell types (Fig. 2e); therefore, we conclude that the ratio of endogenous ATXN3 alleles is not influenced by the cell type (in analyzed set of cell lines).

Fig. 2
figure 2

Determination of endogenous WT and MUT ATXN3 allele expression ratios in SCA3 cell lines. a–d Results from ddPCR are presented as a mean WT and MUT ATXN3 transcript allele abundance, calculated based on results from both ATXN3 assays (ATXN3_SNP2 and ATXN3_SNP5), in patient-derived fibroblasts (a), iPSCs (b), NSCs (c), and neurons (d). These data were analyzed using unpaired t test. e Corresponding results to a–d presented for ATXN3_SNP2 and ATXN3_SNP5 assays separately and analyzed with two-way ANOVA with Tukey’s multiple comparison test for identification of any cell type-specific changes. In this graph, only WT allele abundance is presented, the MUT allele abundance equals a remainder to the sum of 100%). f Estimation of the total number (WT + MUT) of endogenous ATXN3 transcripts per diploid genome using ATXN3_SNP5 assay. Data were analyzed using one-way ANOVA with Tukey’s multiple comparison test. For all experiments presented in this figure n=3. Two-tailed p value < 0.05 was considered significant and is depicted in the figure by: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All data are presented as means ± SD. Individual data values are available in Additional File 13

Full size image

Estimation of the number of endogenous ATXN3 transcripts per diploid genome

To examine whether there are differences in the total number of ATXN3 transcripts among analyzed cell lines, we split cells into two tubes and performed concurrent RNA and DNA isolations. We used the 3′UTR-located ATXN3_SNP5 assay to analyze both cDNA and gDNA. By dividing the total copy number of ATXN3 transcripts by the total copy number of gDNA amplicons, we estimated a number of ATXN3 transcripts per diploid genome. We documented approximately 12, 10, and 54 copies of ATXN3 mRNA in fibroblasts, iPSCs, and NSCs, respectively (Fig. 2f). This result suggests the increase of ATXN3 expression in the process of neural differentiation. Moreover, the number of ATXN3 transcripts calculated for fibroblasts corresponds well with our previous results generated using single-molecule FISH, for this cell line (approximately 18 copies of ATXN3 mRNA per cell) [28].

Examination of endogenous HTT transcripts in cell lines

We used a set of patient-derived cell lines with the HD genetic background: iPSCs, NSCs, and neurons (characterized by 19/109 CAG repeats in HTT) for ddPCR assay discriminating between expression of HTT WT and MUT allele (Fig. 3a).

Fig. 3
figure 3

Assessing the specificity of HTT ddPCR assays. a HD cell lines used; D—differentiation. b Scheme of HTT transcript with marked CAG repeat tract (green triangle) and SNP variants identified in HD cell lines (red arrows—SNPs used in ddPCR assays; black arrows—other identified SNPs); light orange box—UTRs. Table presents all SNP variants identified in analyzed HD cell lines. Bolded SNPs were selected to be targets for ddPCR assays and they will be hereinafter referred to as HTT_SNP2, HTT_SNP5, and HTT_SNP7 in the text. CDS—coding sequence, ∆—deletion. RefSNP number according to the Single Nucleotide Polymorphism Database (dbSNP). c Results from ddPCR analysis of HTT_SNP2 assay specificity performed on seven samples with predefined ratios of WT/MUT plasmids (samples: I—100% WT and 0% MUT; II—90% WT and 10% MUT; III—75% WT and 25% MUT; IV—50% WT and 50% MUT; V—25% WT and 75% MUT; VI—10% WT and 90% MUT; VII—0% WT and 100% MUT). Precise values are indicated on WT/MUT bars ± poisson error. d Percentage ratio of HTT WT allele obtained with ddPCR HTT_SNP7 assay using gDNA from indicated HD cell lines. Three biological replicates were performed. These data are presented as means ± SD

Full size image

Identification of SNP variants in HTT transcripts

To identify SNP variants located in HTT transcripts, we replicated our approach implemented for ATXN3. First, through the cloning and sequencing process, we identified 7 heterozygous SNPs in HTT exons in HD patient-derived NSCs (Fig. 3b). Three of them (rs362331, rs362306, and rs362307, named HTT_SNP2, HTT_SNP5 and HTT_SNP7, respectively) were selected for design of ddPCR assays (Fig. 3b). Those SNPs were described as associated with the most common HTT haplotype [29], correlated with the presence of mutation in HTT allele [30], and/or were targets in clinical testing of therapeutic approaches using ASOs.

Validation of specificity and utility of the HTT ddPCR assays

To assess allelic specificity of the HD ddPCR assay, we prepared plasmid mixtures of predefined WT/MUT ratios. We confirmed that ddPCR output accurately reflects tested plasmid ratios, using all three assays: HTT_SNP2 assay (Fig. 3c), as well as for HTT_SNP5 and HTT_SNP7 assays (Additional File 1: Fig. S1B and C). To determine the reliability of the HD ddPCR, we performed HTT_SNP7 assay (located in 3′UTR) on gDNA isolated from HD iPSCs, NSCs, and neurons. Amplification of HTT gDNA alleles from HD cells showed a canonical 50/50 WT/MUT ratio (Fig. 3d).

HTT WT/MUT ratio determination in HD cell lines

To verify allelic expression of the HTT gene, we examined HD iPSCs, NSCs, and neurons using HTT_SNP2, HTT_SNP5, and HTT_SNP7 ddPCR assays. Cellular identity of generated neurons was confirmed by immunostaining for expression of DARPP32, TUJ1, MAP 2, and GAD67 neuronal markers (Additional File 2: Fig. S2). Allelic expression ratio in iPSCs averaged 68.7% WT/31.3% MUT (Fig. 4a; precise values for each HTT assay are given in Additional File 4: Fig. S4). In NSCs, we observed a slightly lower WT allele contribution, resulting in 64% WT/36% MUT ratio (Fig. 4b). In HD neurons, WT allele fraction dropped even further to 62.4% WT/37.6% MUT (Fig. 4c). The differences in allelic ratios between various HD cell lines were statistically significant (Fig. 4d). Comparing results obtained from HD and SCA3 cell lines, we documented greater differences between WT/MUT allele ratio in HD cells, than in SCA3 cells. We also noticed that the abundance of the MUT HTT allele increases with neural differentiation process (Fig. 4d).

Fig.  4
figure 4

Determination of endogenous WT and MUT HTT allele expression ratios in HD cell lines. a–c Results from ddPCR are presented as a mean WT and MUT HTT transcript allele abundance, calculated based on results from all three HTT assays (HTT_SNP2, HTT_SNP5 and HTT_SNP7), in iPSCs (a), NSCs (b), and neurons (c). These data were analyzed using unpaired t test. d Corresponding results to a–c presented for HTT_SNP2, HTT_SNP5, and HTT_SNP7 assays separately and analyzed with two-way ANOVA with Tukey’s multiple comparison test for identification of any cell type-specific changes. In this graph, only WT allele abundance is presented, the MUT allele abundance equals a remainder to the sum of 100%). e Estimation of the total number (WT + MUT) of endogenous HTT transcripts per diploid genome using HTT_SNP5 assay. Data were analyzed using unpaired t test. For all experiments presented in this figure n=3. Two-tailed p value < 0.05 was considered significant and is depicted in the figure by: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All data are presented as means ± SD. Individual data values are available in Additional File 13

Full size image

To directly compare ddPCR results with information obtained from RNA-Seq, we extracted HTT-relevant reads from data obtained for HD iPSCs and NSC transcriptomes. We looked at HTT reads that span the sequences of three SNP variants selected for ddPCR design. The allelic expression ratio in iPSCs averaged 74.7% WT/25.3% MUT (Additional File 5: Fig. S5A), while in NSCs we obtained the ratio of 73.8% WT/26.2% MUT (Additional File 5: Fig. S5B), and these results were roughly consistent with ddPCR results. However, when each SNP region (SNP2, SNP5 or SNP7) was analyzed separately, we observed substantial differences in WT/MUT allele ratio, in both cell lines (Additional File 5: Fig. S5C-E), which generated higher error bars during statistical analysis. This implies that ddPCR is a more accurate and precise method for such approach.

Estimation of the number of endogenous HTT transcripts per diploid genome

To estimate an absolute number of HTT transcripts per diploid genome, we analyzed cDNA and gDNA from iPSCs and NSCs using HTT_SNP5 and HTT_SNP7 assays. As in SCA3 experiments, we divided the total number of HTT transcripts by the number of gDNA amplicons and, as a result, ~37 HTT transcripts in iPSC and ~16 HTT transcripts in an NSC were calculated for both HTT_SNP5 (Fig. 4e) and HTT_SNP7 assays (Additional File 6: Fig. S6). The difference between the number of HTT mRNAs in iPSCs and NSCs was statistically significant. Furthermore, the number of HTT transcripts calculated for NSCs corresponds well with our previous estimation obtained for this cell line using smFISH (~17 HTT mRNA molecules per cell) [28].

Exploration of WT/MUT HTT allelic expression after gene editing in iPSCs

We previously used a CRISPR-Cas9 method to correct the CAG expansion in HD iPSCs, and we generated isogenic control lines C39 and C105 [31]. Here, we inspected these lines to precisely determine expression of WT and corrected MUT (corrMUT) alleles. Additionally, we used a non-HD-related iPSC line derived from a healthy individual (C7522) that is also heterozygous for SNP5. When we examined gDNA of all three cell lines, we found nearly perfect 50/50 allelic ratio, which indicates that in the case of C39 and C105 lines CRISPR-Cas9 editing did not affect analyzed genomic sequences of the HTT gene (Fig. 5a). Our analyses of HTT allelic expression performed in C39 and C105 lines showed significantly altered expression in comparison with the initial iPSC line. While the initial iPSC line averaged 68.7% WT/31.3% MUT (Fig. 4a), C39 showed the average ratio of 75.9 WT/24.1% corrMUT (Fig. 5b; precise values for each HTT assay are given in Additional File 7: Fig. S7) and C105 revealed a ratio of 99.5 WT/0.5% corrMUT (Fig. 5c). All three SNP assay measurements for C39 and C105 were consistent (Additional File 7: Fig. S7), confirming almost total silencing of expression of the corrMUT allele in C105 line. Results obtained for C105 line suggest that there are off-target effects of gene editing that affected the expression of corrected allele, whereas the change in ratio of WT/corrMUT alleles in C39 line may result from CAG repeat tract shortening or, also, from some off-target effects. In control C7522 iPSC line, HTT_SNP5 assay showed 55.6% [WT-1]/44.4% [WT-2] ratio (Fig. 5d), suggesting more equal ratio in case of two WT alleles.

Fig. 5
figure 5

Determination of endogenous HTT allele ratios in genetically corrected HD iPSCs and non-HD iPSCs. a Percentage ratio of HTT WT allele obtained with ddPCR HTT_SNP7 assay (or HTT_SNP5 regarding C7522 cells) using gDNA from indicated iPSC lines. b,c Results from ddPCR are presented as a mean WT or corrMUT HTT transcript allele abundance, calculated based on results from all three HTT assays (HTT_SNP2, HTT_SNP5, and HTT_SNP7), in genetically corrected C39 (b) and C105 (c). d Ratio of WT alleles of HTT transcript (WT-1 and WT-2) in control C7522 cell line. Data were analyzed using unpaired t test. For all experiments presented in this figure n=3. Two-tailed p value < 0.05 was considered significant and is depicted in the figure by: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All data are presented as means ± SD. Individual data values are available in Additional File 13

Full size image

Silencing of endogenous HTT expression in HD NSCs

To verify the HTT ddPCR assay in assessing the efficacy of allele-selective gene silencing, we transfected HD NSCs with selected oligonucleotides, such as allele-selective miRNA-like oligonucleotide (A2) targeting expanded CAG tracts [32], non-allele-selective siRNA (siHTT) targeting HTT mRNA [33], and non-target control siRNA (siRL) targeting Renilla luciferase [28]. A2 oligo and siHTT reduced overall HTT expression by around 35% and around 50%, respectively, as measured by ddPCR (Fig. 6a). When we looked at HTT allelic expression, we observed equal downregulation of WT and MUT alleles in cells transfected with siHTT. However, silencing of the MUT allele expression was more prominent, as compared to WT allele, in cells transfected with A2 (Fig. 6b). This confirms that siHTT and A2 functions, in gene silencing mechanisms, in a non-allele-selective and an allele-selective manner, respectively [28], and highlights the utility of the HD ddPCR assay as a precise tool for determining the efficacy of transcript-reducing therapeutic strategies.

Fig. 6
figure 6

Silencing of endogenous HTT expression in HD NSCs. a Endogenous HTT expression level after HD NSCs transfection with indicated oligonucleotides (siRL, A2, siHTT) assessed using ddPCR. b HTT WT and MUT allele abundance after transfection with indicated oligonucleotides. Results in a and b were normalized to HTT expression level in cells transfected with control siRNA targeting Renilla luciferase (siRL). Mean values from results obtained using HTT_SNP2, HTT_SNP5, and HTT_SNP7 are presented. Two-way ANOVA with Tukey’s multiple comparison test was used to analyze changes in HTT expression level [WT (*) or MUT (#) allele], while unpaired t test was used to determine significance between WT and MUT HTT allele downregulation for each oligonucleotide (†). For all experiments presented in this figure n=3. Two-tailed p value < 0.05 was considered significant and is depicted in the figure by: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All data are presented as means ± SD. Individual data values are available in Additional File 13

Full size image

Examination of HTT transgenes in HD mice brain tissue

Determination of the HTT transgene copy number

To demonstrate the robustness of our HD ddPCR assay, we examined brain tissue from Hu128/21 mice, the HD mouse model created by crossing BAC21 and YAC128 mouse lines [34]. Hu128/21 mice carry HTT transgenes with heterozygous SNP variants for rs362331 (HTT_SNP2) and rs362306 (HTT_SNP5), allowing us to determine copy number and expression of mutant and wild type alleles. In the first step, we analyzed gDNA isolated from Hu128/21 mice using HTT_SNP5 assay. As BAC and YAC HTT transgenes are present in various copy numbers, the assay in mice did not show 50/50 ratio. Instead, we documented a mean 64.5% WT/35.5% MUT ratio (Fig. 7a). Next, to determine precise transgene copy number in Hu128/21 mice, we performed a copy number variation (CNV) assay. We performed ddPCR for Rpp30 as a reference and used HTT_SNP7 assay targeting HTT transgene. Since the rs362307 (HTT_SNP7) SNP variant is homozygous in alleles of the transgene in Hu128/21 mouse model, this assay recognizes total HTT, without WT/MUT discrimination. Our results showed that HTT transgene was around 8.5 times more abundant than a reference gene (represented by two copies); thus, each mouse had around 17 copies of HTT integrated into its genome (Fig. 7b). Taking WT/MUT ratio into consideration, Hu128/21 mice had around 11 copies of BAC21 transgene and around 6 copies of YAC128 transgene.

Fig. 7
figure 7

Analyses on brain tissues of Hu128/21 mice aged 4 and 10 months. a Abundance of HTT transgene WT allele determined using HTT_SNP5 ddPCR assay on gDNA obtained from three Hu128/21 mice. b Copy number of HTT transgene in Hu128/21 mice, calculated using HTT_SNP7 assay, compared to the reference gene Rpp30. c,d HTT WT transcript abundance in midbrain (MB), olfactory bulb (OF), cortex (CTX), and striatum (STR) of 4-month-old (c) and 10-month-old (d) Hu128/21 mice. e,f Comparison of HTT WT transcript abundance in analyzed brain tissues between 4- and 10-month-old mice according to HTT_SNP2 assay (e), and according to HTT_SNP5 assay (f). Data were analyzed using unpaired t test (b), ordinary one-way ANOVA (c, d), or two-way ANOVA (e, f) with Sidak multiple comparison test. In e, we present results obtained by two-way ANOVA analysis, where all 4 tissues were grouped together and analyzed by the significance of the time point factor. For all experiments presented in this figure n=3. Two-tailed p value < 0.05 was considered significant and is depicted in the figure by: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. All data are presented as means ± SD. Individual data values are available in Additional File 13

Full size image

HTT WT/MUT ratio determination in selected brain tissues of two age groups of Hu 128/21 mice

Next, we analyzed WT/MUT allelic expression in striatum (STR), cortex (CTX), midbrain (MB), and olfactory bulb (OF) of presymptomatic and symptomatic Hu128/21 mice. In presymptomatic 4-month-old mice, we detected 63.4% WT/36.6% MUT ratio in STR, 64.5% WT/35.5% MUT ratio in CTX, 63.0% WT/37.0% MUT ratio in MB, and 66.5% WT/33.5% MUT ratio in OF (Fig. 7c; precise values for each HTT assay are given in Additional File 8: Fig. S8). When we quantified allelic expression in symptomatic 10-month-old mice, we obtained 65.9% WT/34.1% MUT ratio in STR, 65.8% WT/34.2% MUT ratio in CTX, 64.6% WT/35.4% MUT ratio in MB, and 69.1% WT/30.9% MUT ratio in OF (Fig. 7d). The allelic expression ratios matched the allelic distribution found using genomic DNA, suggesting that differences in transgene expression result directly from their copy numbers. Moreover, performed quantifications of HTT transcripts are consistent with reported WT/MUT huntingtin protein levels in the original paper describing Hu128/21 mice [34]. Using two age groups of Hu128/21 mice, we aimed to investigate any age-related changes in HTT transgene expression ratios in selected tissues (Fig. 7e, f). Two-way ANOVA performed on the data obtained for HTT_SNP2 assay and all four tissues considered/analyzed together showed that WT HTT mRNA abundance in 10-month-old mice was slightly higher, when compared to 4-month-old-mice (Fig. 7e), suggesting consistent increase of WT/MUT expression ratio in time in analyzed tissues. However, when all tissues were analyzed individually, post hoc test did not show a significance.

There was no statistically significant time point-dependent changes of HTT allele expression ratio in test performed for particular tissue. However, according to the two-way ANOVA performed on the data obtained for HTT_SNP2 assay and all four tissues considered together, WT HTT mRNA abundance in 10-month-old mice was slightly higher, as compared to 4-month-old mice (Fig. 7e), suggesting consistent increase of WT/MUT expression ratio in time in analyzed tissues.

Discussion

Three decades of extensive research have uncovered polyQ toxicity as a main driver of CAG repeat expansion diseases. However, a growing body of evidence suggests that CAG expanded transcripts also contribute to the pathogenesis and progression of these disorders [35, 36]. Investigation of RNA toxicity in dominantly inherited diseases is challenging since mRNA molecules transcribed from two nearly identical alleles of the causative genes cannot be easily distinguished. Since the polyQ disease genes are typically expressed at low levels, an ideal allele differentiating analytical method should demonstrate high accuracy, precision, sensitivity, and specificity. Because only a limited number of allele discrimination studies have been conducted in the context of transcripts related to polyQ diseases [37, 38], here, we developed an SNP-based ddPCR approach for allele-selective analyses of endogenous HTT and ATXN3 transcripts. We performed a series of control amplifications using newly designed sets of primers and probes to show that our method is reliable and quantitative for analyzing low abundant transcripts with relatively complex GC-rich sequences. Analyses of the ATXN3 and HTT transcripts in cells and tissues relevant to pathology of examined diseases, using independent SNP assays, demonstrated excellent reproducibility of transcript detection and discrimination.

The occurrence of differences in expression between two alleles of one gene is a known phenomenon [39], and several factors responsible for affecting mRNA allelic ratios were already identified using high-throughput analyses [40]. Interestingly, the presence of specific SNP variants itself can affect the expression level of alleles, e.g., through regulation of promoter activity, as shown for HTT [41], or changes in miRNA-binding sites in 3′UTR, which was recently shown for ATXN3 expression [27]. Our allelic expression analysis in several HD and SCA3 cellular and mouse models revealed significant differences between WT and MUT alleles. These differences in selected patient-derived cell lines were generally lower in SCA3 cells (~54% WT/46% MUT) than in HD cells (62–69% WT/31–38% MUT) (Figs. 2e and 4d), but more cell lines should be used to draw general conclusions about ATXN3 and HTT expression in patients’ cells. Such broader analyses were performed on postmortem HD brain samples previously, using standard quantitative RT-PCR, rather than more precise ddPCR [37]. In that study, in the majority of tested samples, MUT HTT RNA was more abundant than the WT; however, a precise WT/MUT allele ratio seems to be an individual case. Another interesting issue to examine in the future is alternative transcript variants of both genes, ATXN3 and HTT. In this research, we wanted to focus on the main variants of each transcript, but analyses on, e.g., HTTintron1 variant [14, 42] or ATXN3 transcripts which does not contain exon 11 (which lead to the translation of ataxin3-a long and ataxin3-a short proteins) [26], might be interesting in terms of revealing new facts regarding the pathogenesis of both diseases.

For both cell lines, SCA3 and HD, we showed that the total number of the ATXN3 and HTT transcripts, respectively, changes during neuronal differentiation in cell culture conditions (Figs. 2f and 4e). Estimations of examined transcripts per diploid genome showed diverse trends, which we confirmed using various calculation approaches. Moreover, these results corresponded well with smFISH-based detection of ATXN3 and HTT in SCA3 fibroblasts and HD NSCs, respectively [28]. Interesting observations were made by Didiot et al., concerning normal HTT mRNA abundance and localization [43]. Not only they showed that the number of HTT mRNAs decreased with reprogramming of human fibroblasts into neuron-like cells, but also the localization of HTT mRNA changed—in neuron-like cells, HTT transcript was detected more often in the nuclear fraction, than in the cytoplasmic one. This is an interesting issue for further examination using both SCA3 and HD cells, together with discrimination of normal and mutant mRNA alleles. Additionally for HTT, we showed that abundance of MUT allele vs. WT allele increases with neuronal differentiation (Fig. 4d), what was also shown previously at a protein level in mouse cell lines obtained from HD YAC128 model [44, 45]. Again, these observations require confirmation in a larger set of cell lines, e.g., for investigation of processes that might be related to the phenomenon of neuronal vulnerability to mutations causing polyQ diseases [46]. However, these results support the idea that presented approach can be applied in studies aimed at the investigation of RNA biology of mutant transcripts leading to a better understanding of the pathogenesis of polyQ disorders [4, 5, 47,48,49]. Additionally, our approach could be also used for analyses of non-neuronal cells, as these cells were reported to be implicated in the pathogenesis of both SCA3 and HD [4, 50, 51]. Moreover, a joint characteristic for polyQ diseases is a regional vulnerability of specific brain regions to neurodegeneration. Hence, our assays could be used in determining whether specific cells (e.g., cortical neurons, Purkinje cells) show varied WT/MUT ratios or absolute numbers of transcripts per cell.

We applied our ddPCR-based approach in the evaluation of genetic engineering methods. The CRISPR-Cas9 technology provides a powerful tool for cell line generation, but off-target effects may affect gene expression and have an impact on the obtained results [52]. During the examination of potential off-target effects, an analysis performed on gDNA may provide only limited information about how the introduced modification affects the expression of the gene-of-interest. On the other hand, qPCR on cDNA is not as precise as ddPCR—since the amplification is exponential (meaning a 2-fold difference in number of molecules per cycle), qPCR is prone to even minor pipetting errors or incorrect choice of a reference gene [53]. Using previously generated isogenic control iPSC lines [31], we showed that quantitation using ddPCR can overcome this potential limitation. Additionally, based on RNA-Seq data from HD iPSCs and NSCs, we were able to estimate the ratio of WT/MUT alleles, but the accuracy and precision of ddPCR-based quantitation was much higher.

The analyses performed with the use of Hu128/21 mice tissues confirmed that the approach can be very useful for the investigation of polyQ disease-related transcripts in animal models. Unlike cells, transgenic Hu128/21 mice carried unequal distribution of WT and MUT gene copies already in gDNA. Interestingly, we showed differences in HTT mRNA levels originating from WT vs. MUT copies between analyzed tissues in the 10-month-old mice. Despite the fact that mostly striatal and cortical neurons are affected in HD [5], we have not observed significant differences in HTT mRNA WT/MUT ratios in those tissues, as compared to other two analyzed. Statistical analysis also showed that there was a slightly higher WT mRNA allele abundance in 10-month-old mice in comparison to 4-month-old animals. This observation is consisted with results from human HD brain tissues, where lower abundance of WT mRNA allele (relative to MUT allele) was reported in samples from patients at early disease stages, as compared to samples from patients at late disease stages [37]. However, more testing is required to verify whether those observations are general or model-specific. So far, time course of mutant and total HTT protein levels was investigated in detail in BACHD rats, showing that the ratio between mutant and total huntingtin can significantly change in time [54].

There is still no available treatment for polyQ diseases. The experimental approaches which are considered as the most promising (as the most efficient and safe) for HD and SCA3 are based on allele-selective targeting of mutant alleles [55,56,57,58]. Preferential targeting of mutant transcripts requires precise evaluation of efficiency and selectivity of the used strategy, e.g., through quantitative analysis of WT and MUT protein levels [59,60,61]. However, protein assays usually require more tissue or CSF, which is often limited. The solution can be a highly sensitive transcript assay, like the one described in this study. Since measurements of the mutant gene expression level are also considered as important biomarker in polyQ diseases [37, 62, 63], our approach can be considered as a powerful tool in clinical research.

Conclusions

We adapted a SNP-based ddPCR method for the evaluation of WT/MUT allele expression ratio in SCA3 and HD cells, as well as in the HD mouse model. Our results clearly show that the allele-specific ddPCR methods presented in this work is suitable for the detection of slight variations in the expression levels of tested alleles in both cell cultures and mice tissues implicated in polyQ diseases. We examined various applications where our approach can be used and, most importantly, we showed its usefulness for accurate and precise analysis of allelic expression during cell differentiation, and after gene editing, as well as in assessing the efficacy of polyQ therapies.

Methods

Cell lines and cell culture

SCA3 fibroblasts (GM06153, 16/67 CAG repeats in ATXN3) were obtained from the NIGMS Repository (Coriell Institute for Medical Research). They were cultivated in Minimum Essential Medium (Gibco), containing 10% fetal bovine serum (Biowest), a penicillin-streptomycin solution (Sigma-Aldrich), and 2mM L-glutamine (Sigma-Aldrich). SCA3 iPSCs were generated (from GM06153 fibroblasts) previously [64]. HD iPSCs (ND42222, 19/109 CAG repeats in HTT) were obtained from the NIGMS Repository. Additional cell lines were used: iPSC clonal lines (C39 and C105) previously generated via CRISPR-Cas9-based modification of the ND2222 line [31], and a non-HD-related iPSC line (C7522) generated from a healthy individual-derived fibroblast line (GM07522) obtained from the NIGMS Repository. iPSCs were cultured in StemFlex (Gibco) medium on Geltrex (Gibco). NSCs were grown in STEMdiff Neural Progenitors Medium (NPM) (STEMCELL Technologies).

Differentiation of HD and SCA3 iPSCs to NSCs

iPSCs were differentiated to NSCs using the STEMdiff SMADi Neural Induction Kit (STEMCELL Technologies) and a monolayer protocol, following the manufacturer’s instructions. Briefly, iPSCs were grown on a 6-well plate until 70–80% confluence was reached, and then dissociated to single cells by incubation with Accutase (STEMCELL Technologies) for 4 min. Cells were counted using TC20 Automated Cell Counter (Bio-Rad) and resuspended at 1 × 106 cells/ml density for seeding in STEMdiff Neural Induction Medium with SMADi and 10 nM Y-27632 (all from STEMCELL Technologies). After third passage, they were grown in STEMdiff NPM.

Differentiation of SCA3 NSCs to neurons

SCA3 NSCs were differentiated to neurons according to protocol published by Hansen et al. [65] with slight modifications. Briefly, NSC were grown until full confluence and then dissociated to single cells by incubation with Accutase. Collected cells were passaged in 1:10 ratio onto Geltrex-coated 6-well and 12-well (for ICC) plates with NPM and 10 nM Y-27632. Next-day NPM was exchanged for neural maturation medium N2B27 with 20 ng/mL recombinant human BDNF and 10ng/mL recombinant human GDNF (both STEMCELL Technologies) and 300 ng/μL cAMP (Sigma-Aldrich). Medium was changed every 2 days for 3 weeks.

Differentiation of HD NSCs to MSNs

HD NSC neural differentiation was performed according to protocol published by M. Fjodorova and M. Li [66] with slight modifications. Briefly, NSCs were grown in STEMdiff NPM until full confluence and then dissociated to single cells by incubation with Accutase. Collected cells were passaged in 1:5 ratio onto a Geltrex-coated 6-well plate with NPM and 10 nM Y-27632. Next-day NPM was exchanged for LGE pattering medium with 25 ng/mL recombinant human Activin A (STEMCELL Technologies). Medium was changed every day until day 10 when LGE progenitors were passaged for terminal differentiation into medium spiny neurons (MSNs): cells were passaged with Accutase in ratio 1:5 onto Geltrex precoated 6-well and 12-well plates (for ICC) in LGE pattering medium and 10 nM Y-27632. Next-day LGE pattering medium was exchanged for a terminal differentiation medium N2B27 with 10 ng/mL recombinant human BDNF and 10 ng/mL recombinant human GDNF (both STEMCELL Technologies). Medium was changed every 2 days for 3 weeks.

Immunocytochemistry (ICC)

NSC and neuronal cells were fixed firstly for 5 min by adding 2% PFA directly to a cell culture medium and then gently washed in PBS followed by fixation in 4% PFA in PBS for 10 min. Cells were permeabilized with 0.5% Tween, blocked in 1% bovine serum albumin, and incubated with primary antibodies and fluorescent-dye conjugated secondary antibodies (listed in Additional File 9: Table S1). DAPI was used for DNA staining. Images were captured with a Leica DMI6000 microscope (Additional File 10: Fig. S2).

HD mice

Hu128/21 HD mice model was described previously [34]. Briefly, the Hu128/21 mouse model was created by crossing Hdh−/− BAC21 mice (Hu21) and Hdh−/− YAC128 mice (Hu128), and then genotyped. Mice were housed in the animal facility of the Center for Advanced Technology (CAT) in Poznań and kept under standard conditions with a 12-h light/dark cycle with water and food ad libitum. All procedures and handling of the animals were performed to minimize stress to the animals and according to a protocol approved and monitored by the Local Ethical Commission for Animal Experiments in Poznan (Poland). Mice were sacrificed and brain tissues harvested and snap frozen in liquid nitrogen and stored at −80°C. Number and age of mice used in all experiments are reported in the accompanying figure legends.

SNP variant identification

For SNP variant identification, the TOPO XL-2 PCR Cloning Kit (Invitrogen) was used. Briefly, RNA isolated from SCA3 and HD cell lines (iPSCs) was reversely transcribed using High-Capacity cDNA Reverse Transcription kit (Applied Biosystems) with random hexamer primers (for ATXN3) or SuperScript IV Reverse Transcriptase (Thermo Fisher) with a gene-specific primer (HTT_RT) (for HTT). Then, cDNA served as a template in PCR reactions using Platinum SuperFi polymerase included in a cloning kit. In case of HTT, PCRs were performed using PrimeSTAR GXL DNA Polymerase (Takara Bio). Primers in those PCR reactions were designed in a way to amplify whole CDSs, with fragments of both 5′ and 3′UTRs (ATXN3_Fwd and ATXN3_Rev; HTT_Fwd and HTT_Rev, sequences given in Additional File 10: Table S2). PCR products were then cloned into pCR-XL-2-TOPO vector. Generated vectors were transformed into bacterial cells provided with the kit. Transformants were analyzed through colony PCR and restriction enzyme digestion to select clones with either WT or MUT versions of analyzed genes. Positive clones were then sequenced, and SNPs allowing for discrimination between WT and MUT alleles, after additional confirmation on genomic DNA (gDNA), were identified. All identified SNPs are listed in Figs. 1b and 3b.

RNA isolation and reverse transcription

For RNA isolation, the Total RNA Zol-Out D kit (A&A Biotechnology) was used, according to the manufacturer’s protocol. In case of cells, cell pellets were suspended in 800 μL of TRI Reagent (Thermo Fisher), while mouse tissues were homogenized in 300 μL of TRI Reagent. The concentration of isolated total RNA was calculated by measurement at 260 nm using DeNovix spectrophotometer. One microgram of RNA was then used for reverse transcription using the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems) with random hexamer primers.

Genomic DNA isolation

gDNA was isolated using the Genomic DNA Isolation kit (Norgen Biotek Corp.) according to the manufacturer’s protocols designed for cells growing in a monolayer and animal tissues. The concentration of isolated gDNA was calculated by measurement at 260 nm using the DeNovix spectrophotometer.

Droplet digital PCR

Among various ddPCR applications, Rare Mutation and Sequence Detection were chosen. To specifically target sequences differing by only one nucleotide, two probes labelled with different fluorophores were designed together with a pair of primers. Because of the high dilution of the sample, probes bind to templates carrying specific SNP variants with a high specificity. After the readout based on the fluorescence, data were analyzed using Poisson statistics to determine the concentration of template carrying each SNP variant (and its flanking sequence) in the original sample. ddPCRs were prepared using DG8 cartridges and gaskets, Droplet Generation Oil for Probes, and ddPCR Supermix for Probes (no dUTP) (Bio-Rad), and they were performed on the QX200 Droplet Digital PCR System (Bio-Rad), according to the manufacturer’s protocols and established MIQE guidelines for ddPCR [67]. ddPCR assays targeting identified SNP variants were designed by a Bio-Rad’s technician and their sequences are unknown; however, their assay IDs are listed in Additional File 10: Table S2. As a reference, ddPCR assay targeting β-actin was used, while for determination of HTT transgene copy number, an assay targeting mouse Rpp30 was used.

NSC transfection with oligonucleotides

All siRNA oligonucleotides (Additional File 11: Table S3) were synthesized by FutureSynthesis, dissolved in water to 100 μM concentration, and stored at −80 °C. To obtain 20 μM duplexes, sense and antisense strands were diluted in water, heated for 1 min in 90 °C, and kept for gradual cooling at room temperature for 45 min. HD NSCs were then transfected using 2 μL of siPORT Amine (Ambion) and 100 nM oligonucleotides per well of a 6-well plate in 1 mL of complete medium. After 3 h, the medium was replaced with a fresh one, and after the next 24 h, cells were harvested and total RNA was isolated and then used for reverse transcription and ddPCRs, as previously described.

RNA-Seq

RNA-Seq was performed in the CeNT’s Genomics Core Facility in the Centre of New Technologies, University of Warsaw, Poland. Total RNA isolated from HD iPSCs and NSCs after ribodepletion was sequenced using NovaSeq 6000 system with a pair-end 2x100 cycles procedure. Fastq files were mapped to the reference genome (GENCODE release 42, GRCh38.p13) using STAR (v. 2.7.10b) algorithm. Only the reads mapped to the forward strand were subsetted (samtools 1.16.1). The statistics for reads mapped to the investigated SNPs in HTT were retrieved using samtools mpileup (QC parameters: -q 20 -Q 20).

Statistical analysis

Three biological replicates were included for each experiment. Analyses of obtained results were performed using GraphPad Prism software (version 8.0.1.). Depending on the experimental setup, specific statistical tests were used and are indicated in figure legends. All data are presented as means. Error bars always represent standard deviations.