Allele-Selective Suppression of Mutant Huntingtin in Primary Human Blood Cells

Post-transcriptional gene silencing is a promising therapy for the monogenic, autosomal dominant, Huntington’s disease (HD). However, wild-type huntingtin (HTT) has important cellular functions, so the ideal strategy would selectively lower mutant HTT while sparing wild-type. HD patients were genotyped for heterozygosity at three SNP sites, before phasing each SNP allele to wild-type or mutant HTT. Primary ex vivo myeloid cells were isolated from heterozygous patients and transfected with SNP-targeted siRNA, using glucan particles taken up by phagocytosis. Highly selective mRNA knockdown was achieved when targeting each allele of rs362331 in exon 50 of the HTT transcript; this selectivity was also present on protein studies. However, similar selectivity was not observed when targeting rs362273 or rs362307. Furthermore, HD myeloid cells are hyper-reactive compared to control. Allele-selective suppression of either wild-type or mutant HTT produced a significant, equivalent reduction in the cytokine response of HD myeloid cells to LPS, suggesting that wild-type HTT has a novel immune function. We demonstrate a sequential therapeutic process comprising genotyping and mutant HTT-linkage of SNPs, followed by personalised allele-selective suppression in a small patient cohort. We further show that allele-selectivity in ex vivo patient cells is highly SNP-dependent, with implications for clinical trial target selection.


Results
Allele-selective HTT suppression in HD patient cells using siRNA targeted to rs362331. We utilised siRNAs designed against the SNPs rs362331 in exon 50 (39.4% heterozygous C/T), rs362273 in exon 57 (35.2% heterozygous A/G) and rs362307 in the 3′-UTR (48.6% heterozygous C/T) of the HTT transcript. The siRNAs targeted to rs362273 and rs362307 have previously demonstrated good selectivity using luciferase assays in HeLa cells 8 . However, the siRNAs designed against rs362331 have not been previously published. As the T allele of rs362307 is linked to mHTT 8 , testing siRNA targeting the C allele was unnecessary. Diagnostic DNA samples were analysed to determine individual SNP genotypes for a cohort of HD patients; we identified fifty-seven individuals who were heterozygous for rs362331, thirty-eight who were heterozygous for rs362273 and fifty-four who were heterozygous for rs362307. Monocyte-derived macrophages were then isolated from repeat peripheral blood samples donated by a subset of the heterozygous individuals. Transfection was carried out using glucan-encapsulated siRNA particles (GeRPs), which are taken up by phagocytosis before releasing their siRNA contents into the cytoplasm of the target cell. Cultures were harvested 72 h after transfection and mRNA expression was analysed by allele-specific qPCR. Previously validated non-selective anti-total HTT siRNA was used as a positive control 12 .
Targeting rs362331 with anti-U siRNA resulted in 74% on-target U allele suppression, compared to 17% off-target C allele suppression (Fig. 1a). Targeting this SNP with anti-C siRNA resulted in 63% on-target suppression, with 30% off-target suppression. Post-hoc testing revealed the discrimination between alleles to be significant for each siRNA (p < 0.05). Indeed, in both cases on-target knockdown was equivalent to non-selective siRNA, while off-target knockdown was not statistically significant compared to nonsense control.
Unfortunately, comparable selectivity was not achieved when testing siRNAs targeted to either rs362273 or rs362307 (Fig. 1b,c). Although post-hoc testing demonstrated statistically significant allelic discrimination following treatment with the rs362273 anti-G siRNA, mean off-target knockdown of this degree is clearly too great to consider the particular siRNAs to be truly selective. These data are consistent with the poor discrimination previously seen when testing siRNA against rs362307 in cell lines and animal models 11,13 . Selective suppression of mHTT protein in HD patient cells using siRNA targeted to rs362331. We next validated that total (polyQ-independent) and mHTT (polyQ-expanded) protein levels were similarly reduced; analysis was again carried out 72 h after transfection with siRNAs targeted to either allele of rs362331. As the siRNAs targeted to rs362273 and rs362307 did not achieve significant mRNA selectivity, it was decided not to advance their use to protein studies. To determine which siRNA was targeting mHTT, each patient's SNP alleles were linked to either wild-type or mHTT using the SNP linkage by circularization technique 12 ; the T allele of rs362331 was located on the mHTT allele in each subject used for protein analysis. As expected, we observed equivalent suppression of total HTT following treatment with each allele-selective siRNA, demonstrating that their effects on total HTT levels are independent of mHTT-linkage (Fig. 2a). However, we only observed significant suppression of mHTT compared to nonsense control with mHTT-targeted anti-U siRNA, with no significant reduction in mHTT following treatment with wild-type targeted anti-C siRNA (Fig. 2b). This result shows that the selectivity seen when targeting mRNA at rs362331 also affects the protein level.
Allele-selective suppression of wild-type and mHTT reverses the hyper-reactive phenotype of human HD myeloid cells. Previous work has shown that HD myeloid cells are hyper-reactive compared to control cells in response to stimulation with LPS and IFNγ 14 . This phenotype was found to be reversible following non-selective HTT suppression with siRNA. To address whether allele-selective suppression has similar effects, we stimulated monocyte-derived macrophages with LPS and IFNγ 72 h after treatment with allele-selective GeRPs targeting either allele of rs362331. Production of IL-6, IL-8 and TNFα was measured 24 h later. Since we observed no significant differences between the mRNA knockdown provided by the anti-U and anti-C siRNAs, the samples were grouped for analysis into anti-wild-type and anti-mHTT following SNP linkage to mHTT. Interestingly, allele-selective suppression of either wild-type or mHTT produced a significant reduction in the levels of each cytokine compared to nonsense control (Fig. 3). No differences in cytokine production were found when selectively suppressing mutant compared to wild-type HTT.

Discussion
We demonstrate the first example of a complete sequential workflow comprising SNP genotyping, SNP linkage, and personalised allele-selective suppression with phenotypic reversal of an HD-related cellular deficit in heterozygous primary ex vivo patient cells. We achieved this through repeated sampling of a small patient cohort, using siRNA targeted to rs362331 in exon 50 of the HTT gene. These genotyping techniques would form the basis of allele-selective treatment in the clinic, and our study demonstrates their first use in combination, instead of individually as part of a technical study 8,15 . Indeed, blood cells are likely to be used for genotyping in clinical trials due to their easy accessibility, and provide a convenient screening tool for novel therapeutics. Although we saw reduced suppression of protein relative to mRNA, we consider this is likely due to the longer protein half-life allowing more pre-existing species to persist up to 72 h.
These data reinforce previous studies demonstrating that the potency and selectivity of mHTT suppression varies considerably depending on the SNP being targeted 9,11 . While we observed significant allele-selectivity using siRNAs targeted to both alleles of rs362331, we did not see similar effects when targeting either rs362273 or rs362307; this is likely due to additional factors including the nucleotide sequence and tertiary structure of the transcript surrounding the base mismatch. rs362307 is a particularly challenging target for allele-selective suppression as the surrounding sequence is extremely GC rich, with the SNP comprising a pyrimidine mismatch (purine mismatches typically confer greater selectivity) 7 . Indeed, this SNP has now demonstrated poor selectivity across multiple research groups and studies 11,13 . While previous work has shown that adding an additional Monocyte-derived macrophages were isolated from SNP-genotyped HD patients and treated with GeRPs containing either nonsense, anti-total HTT or allele-selective siRNA targeted to (a) rs362331 in exon 50, (b) rs362273 in exon 57 or (c) rs362307 in the 3′-UTR of the HTT transcript. Expression of each SNP allele was then measured 72 h later by qPCR. Data show mean expression of each SNP allele ± SEM (n = 3 for rs362331, n = 4 for rs362273 and rs362307), statistical analysis carried out using two-way ANOVA with Bonferroni posttests. Asterisks above the horizontal line at the top of each graph show significance on ANOVA across the whole experiment before post-hoc testing. Asterisks above individual bars represent post-test significance for the mRNA expression of that allele compared to the same allele in the nonsense-treated samples. Asterisks above horizontal lines between two bars denote allele expression differences within the same treatment on post-hoc testing. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. N = individual biological repeats.
Scientific RepoRts | 7:46740 | DOI: 10.1038/srep46740 base mismatch in the siRNA sequence can improve selectivity (often at the cost of reduced potency) 8 , the results seen with our anti-rs362307 siRNA suggest that the benefits of this approach may be limited. These data further demonstrate the importance of testing potential therapeutics in primary patient cells containing endogenous expression of full-length HTT, as these siRNAs achieved promising selectivity during preliminary testing with a short construct attached to a luciferase reporter 8 .
We further utilised allele-selective suppression to investigate the cytokine response of HD myeloid cells. HD myeloid cells produce increased levels of proinflammatory cytokines compared to control following stimulation with LPS 14 . This phenotype is due to an exaggerated NFκ B signalling response caused by a direct interaction of mHTT with IKK, a key cytoplasmic regulator of NFκ B 16 . NFκ B signalling has recently been shown to be upregulated even in resting HD myeloid cells, and contributes to an abnormal proinflammatory transcriptional profile in the absence of any stimulation 17 . However, very little is known about whether wild-type HTT also has a role in innate immune function. If HTT's effects on cytokine production were limited exclusively to the mutant form, it would be expected that selective mHTT-lowering would reduce cytokine production more than selective wild-type lowering. However, we saw no significant differences in cytokine production between the allele-selective treatment groups. These results suggest that wild-type HTT has an as yet undefined role in normal immune cell function, and is consistent with previous work showing that cytokine production by control myeloid cells is also reduced following HTT-lowering 14 . Although none of the differences reached statistical significance, the mean IL-6 and TNFα levels measured following total HTT-lowering were less than 50% of those Monocyte-derived macrophages were isolated from HD patients heterozygous for rs362331 with linkage of the T allele to mHTT, before treating with GeRPs containing either nonsense, anti-total HTT or allele-selective siRNA targeted to each allele of rs362331. Expression of (a) total and (b) mutant HTT protein was measured after 72 h using MSD assays, before normalisation to total cellular protein content as measured by BCA assays. Data show mean protein levels ± SEM (n = 5), statistical analysis carried out using one-way ANOVA with Tukey post-hoc testing. Asterisks above the horizontal line at the top of each graph show significance on ANOVA across the whole experiment before post-hoc testing. Asterisks above bars show posttest significance for each siRNA compared to the nonsense-treated samples. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. N = individual biological repeats.
Scientific RepoRts | 7:46740 | DOI: 10.1038/srep46740 measured following selective lowering of either the mutant or wild-type allele. This result suggests that total cellular HTT load may be an important factor influencing cytokine production, as the sparing of one allele by the allele-selective siRNAs will result in less overall knockdown than that achieved by anti-total HTT siRNA. However, a full investigation of this hypothesis is beyond the scope of this study.

Figure 3. Allele-selective suppression of wild-type and mHTT reverses the hyper-reactive phenotype of human HD myeloid cells.
Monocyte-derived macrophages were isolated from SNP-genotyped HD patients and treated with GeRPs containing either nonsense, anti-total HTT or allele-selective siRNA targeted to each allele of rs362331. Linkage of each SNP allele to either wild-type or mHTT was determined using the SNP linkage by circularization protocol. After 72 h the macrophages were stimulated with LPS and IFNγ , before the culture supernatants were collected after a further 24 h and analysed using MSD assays. Cytokine values were normalised to total protein content as measured by BCA assays. Data show mean cytokine levels ± SEM (n = 5 for IL-6 and TNFα , n = 4 for IL-8), statistical analysis carried out using one-way ANOVA with Tukey post-hoc testing. Asterisks above the horizontal line at the top of each graph show significance on ANOVA across the whole experiment before post-hoc testing. Asterisks above bars show post-test significance for each siRNA compared to the nonsense-treated samples. *P < 0.05, **P < 0.01, ***P < 0.001. N = individual biological repeats.
Scientific RepoRts | 7:46740 | DOI: 10.1038/srep46740 Finally, as rs362307 is the most prevalent SNP in Caucasian HD patients with 48.6% heterozygosity 8 , it should not be assumed that the optimal SNPs will be targetable when estimating what percentage of HD patients may benefit from allele-selective therapeutics. As a result, it may be necessary to target less prevalent SNPs to improve selectivity, with a subsequent reduction in the treatable proportion of the HD patient population. Non-selective therapeutics are therefore likely to be used to treat the majority of adult HD patients, as partial HTT-lowering has been shown to be well tolerated in the mature brain 4,5 . However, the deleterious effects associated with HTT knockout in the developing brain 6 suggest that an allele-selective strategy may be more useful for treating presymptomatic young adults. We demonstrate that rs362331 may be a promising candidate for this approach.
To test the siRNAs, monocytes were isolated from peripheral blood samples by density centrifugation and magnetic cell sorting, before differentiation into macrophages as previously described 14 . Transfection was carried out on day three of the differentiation protocol by changing to fresh media containing GeRPs at a 10:1 particle to cell ratio. GeRPs were removed after 24 h by complete media change; this protocol achieves 90% transfection efficiency. Huntingtin mRNA and protein levels were then measured after a further 48 h in culture.
HTT protein quantification assays. Total (polyQ-independent) and mHTT (polyQ-dependent) protein levels were measured by Meso Scale Discovery (MSD) assays using a protocol adapted from published methods 19 . 2B7 and biotinylated-4C9 were used as the antibody pair for detection of total HTT 20 , while 2B7 and biotinylated MW1 were used as the antibody pair for detection of mHTT 21 . Readings were normalised to total protein content of the sample as measured by BCA assays.
MSD cytokine assays. Monocyte-derived macrophages were stimulated by changing to fresh media containing 2 μ g/ml LPS and 10 ng/ml IFNγ 72 h after GeRP treatment. After a further 24 h the supernatants were analysed using the V-PLEX Human Proinflammatory Panel II (4-Plex) Kit. IL-1β was not included in the analysis as previous work has shown that its production is not upregulated in HD myeloid cells 14 . Readings were normalised to total protein content of the sample as measured by BCA assays.

Statistical analysis.
Statistical analysis was carried out using GraphPad Prism 6 (GraphPad). Analysis of mRNA knockdown was performed using two-way ANOVAs with Bonferroni post-hoc multiple comparison testing. Protein knockdown and cytokine profiling experiments were analysed using one-way ANOVAs with Tukey post-hoc multiple comparison testing. All error bars represent standard error of the mean.