Allele-selective inhibition of huntingtin expression by switching to an miRNA-like RNAi mechanism

SUMMARY

Inhibiting expression of huntingtin (HTT) protein is a promising strategy for treating Huntington’s Disease (HD), but indiscriminant inhibition of both wild-type and mutant alleles may lead to toxicity. An ideal silencing agent would block expression of mutant HTT while leaving expression of wild-type HTT intact. We observe that fully complementary duplex RNAs targeting the expanded CAG repeat within HTT mRNA block expression of both alleles. Switching the RNAi mechanism towards that used by miRNAs by introducing one or more mismatched bases into these duplex RNAs leads to potent (< 10 nM) and highly selective (>30-fold relative to wild-type HTT) inhibition of mutant HTT expression in patient derived cells. Potent, allele selective inhibition of HTT by mismatched RNAs provides a new option for developing HD therapeutics.

INTRODUCTION

Huntington’s disease (HD) is a neurodegenerative disorder with an incidence of 5-10 per 100,000 individuals in Europe and North America (Walker, 2007). HD is characterized by progressive neurodegeneration resulting in chorea, dystonia, and cognative or psychiatric disturbances. Symptoms worsen until death, 10-20 years after disease onset. Genetic tests and palliative treatments exist, but there are no curative therapies. Effective therapeutic strategies that delay the onset of HD remain a major unmet need.

HD is caused by a dominant heterozygous expansion of a CAG trinucleotide repeat that encodes multiple glutamine residues within the gene encoding huntingtin (HTT) protein (MacDonald et al., 1993). The mutant HTT allele in HD patients contains tracts of more than 39 consecutive glutamine codons. Most patients have 50 repeats or less, but some patients can have as many as 120. Higher numbers of repeats correlate with earlier onset of disease (Kremer et al., 1994; Duyao et al., 1993). While the exact molecular defects caused by mutant HTT expression are not definitively known, aggregation of poly-glutamine fragments of HTT induces plaque formation (Sanchez et al., 2003). HTT is expressed in many tissues, but the central nervous system and brain are especially sensitive to expression of mutant HTT, with striatal neurons suffering the most degeneration (Reddy et al., 1998).

Developing agents to lower levels of mutant HTT in patients offers one strategy for developing medications to treat HD (Pfister and Zamore, 2009). Lower HTT levels should delay or prevent aggregation of poly-glutamine fragments, slow plaque formation, and reduce damage to cells. This hypothesis has been tested in several animal models using both antisense oligonucleotides and duplex RNAs and reducing levels of mutant HTT has been shown to alleviate symptoms and prolong survival (Harper et al., 2005; DeFiglia et al, 2007; Boudreau et al., 2009; Gagnon 2010). These data suggest that controlling mutant HTT in patients would be a major therapeutic advance.

Typically, agents that silence gene expression block expression of both alleles of the target gene. Wild-type HTT, however, may play a role in normal adult function (Nasir et al., 1995; White et al. 1997) and strategies that reduce expression of both the wild-type and mutant alleles may cause side-effects in patients. While it is possible that an agent that inhibits both wild-type and mutant HTT might be both safe and effective (Drouet et al. 2009; Boudreau et al. 2009), an ideal agent would selectively reduce mutant HTT expression but not alter wild-type protein levels. Because of the lack of curative treatments for HD it is essential that both allele-selective and nonallele-selective approaches for reducing mutant HTT levels be pursued.

Strategies have been developed that achieve allele-selective inhibition of HTT expression. One approach takes advantage of the fact that the mutant HTT allele often contains single-nucleotide polymorphisms that distinguish it from the wild-type allele, allowing short interfering RNAs to selectively target the mutant gene (Miller et al, 2003; Schwarz et al. 2006; van Bilsen et al. 2008; Pfister et al. 2009; Zhang et al., 2009). The identity of SNPs varies with 86 % of European Caucasian patients sharing 26 different SNPs (Lombardi et al., 2009), but some SNPs are common and only five duplex RNAs would be sufficient to treat 75% of HD patients (Schwartz et al., 2007). While targeting SNPs is promising, the need to develop several different agents complicates development and suggests a continuing need for new strategies for silencing HD.

Another strategy involves single stranded oligomers complementary to the expanded CAG repeat within HTT mRNA that achieve potent and selective inhibition of HTT expression (Hu et al. 2009; Hu et al. 2009b; Hu et al. 2009c). This strategy would require just one oligomer to treat the entire HD patient population. Oligomers that target the long (250 repeat) CUG repeat within the 3’ untranslated region (3’-UTR) of DM protein kinase, the cause of myotonic dystrophy, have also been shown to be biologically active (Wheeler et al., 2009). However, after testing several different chemically modified oligomers complementary to the CAG repeat in HTT mRNA, it has been difficult to improve upon the potencies and selectivities that were observed initially (Gagnon et al., unpublished).

Developing highly potent and selective methods for silencing the mutant allele of HTT remains challenging. Based on the potency of duplex RNA and the length of CAG repeat, we asked whether we could achieve the selectivity by targeting CAG repeat region using siRNA. We observed that these siRNAs were potent selective inhibitors of HTT but showed little selectively between the mutant and wild type alleles (Hu et al. 2009a; Hu et al. 2009b).

RNAi can function through two distinct mechanisms (Filipowicz et al. 2008; Kurreck 2009). If the duplex RNA is fully complementary to its mRNA target, human argonaute 2 (AGO2) protein cleaves the mRNA at the phosphodiester linkage opposite siRNA bases 10 and 11 (Fig. 1a) (Liu et al. 2004). This mechanism is used by the short interfering RNAs (siRNAs) commonly used in laboratories for gene silencing.

Figure 1. Allele-selective inhibition by mismatch-containing duplex RNAs.

Schemes showing recognition of a (a) perfectly complementary duplex leading to cleavage of mRNA and (b) mismatch-containing duplex leading to inhibition of translation. Effect on HTT expression from adding duplex RNAs (c) P9, (d) REP (e) PM3 to GM04281 cells or f) P9 to GM04717 cells. Data describing the selectivities and potencies of other duplexes can be found in Table 1. See also Figure S1.

Mismatches between the duplex RNA and mRNA target, however, disrupt interactions at the catalytic center of AGO2 and make cleavage less efficient (Fig. 1b) (Du et al. 2005; Wang et al. 2008). Gene silencing can still occur, but may involve inhibition of ribosome-mediated translation or reduced mRNA stability rather than cleavage of the mRNA target. This mechanism is commonly used by naturally occurring miRNAs to inhibit expression of endogenous genes and regulate cell physiology. In addition, the expanded CAG repeat offers multiple target sequences for finding complementary duplex RNAs, and it is known that the binding of multiple mismatch-containing duplex RNAs to mRNA can yield synergistic increases in potency for miRNAs (Grimson et al. 2007). The miRNA-like mechanism is similar to the mechanism of single-stranded PNAs which also block translation rather than cause mRNA cleavage (Knudsen et al. 1996), raising the possibility that miRNA-like duplexes might be better candidates for achieving allele selectivity. We hypothesized that deliberately switching the RNAi mechanism would be advantageous for discriminating among closely related RNA targets and achieving allele-selective regulation of disease genes.

Here we test mismatch-containing duplex RNAs complementary to the CAG repeat. We find that they are potent, versatile, and selective agents for blocking expression of the mutant HTT allele.

Results and Discussion

We transfected duplex RNA P9 (single mismatch at base 9 within the guide strand/antisense relative to the mRNA target) (Table 1) into patient-derived fibroblast cell line GM04281 (69 CAG repeats/mutant; 17 repeats/wild-type allele). P9 possessed high selectively, >31 fold, for inhibition of mutant versus wild-type HTT (Fig. 1c). Fully complementary RNA REP and sister compounds REPC and REPU that were shifted to start with C or U showed selectivities of less than 2-3 fold (Fig. 1d) (Supplementary Fig. 1a). These data demonstrated that mismatch-containing RNAs widen the window for achieving selective inhibition.

Table 1.

siRNAs targeting the HTT CAG repeat region.

RNA	Sequence	Position of mismatch	Tm, °C	mut IC50 (nM)	wt IC50 (nM)	Selectivity
siRNA fully complementary to CAG repeat region (only guide strand is showed)
REP	GCUGCUGCUGCUGCUGCUGTT		86.8	5.0 ± 3.0	13 ± 5.0	3
REPC	CUGCUGCUGCUGCUGCUGCTT		86.6	5.4 ± 1.6	12 ± 2.6	2
REPU	UGCUGCUGCUGCUGCUGCUTT		86.5	5.6 ± 1.1	9.8 ± 1.2	2
siRNA containing one mismatched base on both strands (only guide strand is showed)
P4	GCUACUGCUGCUGCUGCUGTT	4	84.0	4.1 ± 2.0	16 ± 7.4	4
P5	GCUGAUGCUGCUGCUGCUGTT	5	83.2	14 ± 2.9	>100	>7
P6	GCUGCAGCUGCUGCUGCUGTT	6	85.1	No inhibition	No inhibition	-
P7	GCUGCUACUGCUGCUGCUGTT	7	82.7	13 ± 5.6	53 ± 13	4
P8	GCUGCUGAUGCUGCUGCUGTT	8	83.8	5.1 ± 0.7	>100	>20
P9	GCUGCUGCAGCUGCUGCUGTT	9	86.7	3.2 ± 0.9	>100	>31
P10	GCUGCUGCUACUGCUGCUGTT	10	83.5	4.0 ± 2.8	>100	>25
P10R	GCUGCUGCUUCUGCUGCUGTT	10	78.0	3.5 ± 2.2	>100	>27
P11	GCUGCUGCUGAUGCUGCUGTT	11	83.7	2.7 ± 0.8	>100	>37
P12	GCUGCUGCUGCAGCUGCUGTT	12	85.6	3.7 ± 1.9	33 ± 12	9
P13	GCUGCUGCUGCUACUGCUGTT	13	82.8	5.3 ± 1.1	26 ± 8.4	5
P16	GCUGCUGCUGCUGCUACUGTT	16	76.4	1.0 ± 0.8	7.0 ± 1.3	7
siRNA containing multiple mismatched bases on both (only guide strand is showed)
P910	GCUGCUGCAACUGCUGCUGTT	9,10	83.5	4.5 ± 0.9	>100	>22
PM3	GCUGCUGCAAAUGCUGCUGTT	9,10,11	79.9	3.2 ± 1.0	>100	>31
PM4	GCUGCUGAAAAUGCUGCUGTT	8,9,10,11	76.4	4.8± 1.6	>100	>21
RM3	GCUACUGCUACUGCUACUGTT	4,10,16	83.0	63 ± 42	>100	>2
RM4	GCAGCUGUUGCUACUGUUGTT	3,8,13,17	78.0	No inhibition	No inhibition	-
siRNA containing one or two mismatched bases on only one strand
P9a	GCUGCUGCUGCUGCUGCUGTTTTCGACGACGUCGACGACGAC	9	79.4	2.0 ± 0.6	8.9 ± 1.3	4
P10a	GCUGCUGCUGCUGCUGCUGTTTTCGACGACGAUGACGACGAC	10	80.8	5.4 ± 1.4	27 ± 12	5
P910a	GCUGCUGCUGCUGCUGCUGTTTTCGACGACGUUGACGACGAC	9,10	77.6	2.7 ± 0.5	9.9 ± 2.0	4
P9b	GCUGCUGCAGCUGCUGCUGTTTTCGACGACGACGACGACGAC	9	80.0	2.3 ± 1.5	>100	>43
P10b	GCUGCUGCUACUGCUGCUGTTTTCGACGACGACGACGACGAC	10	77.6	6.8 ± 1.5	>100	>15
P10R -b	GCUGCUGCUUCUGCUGCUGTTTTCGACGACGACGACGACGAC	10	75.4	2.3 ± 0.5	>100	>43
P11b	GCUGCUGCUGAUGCUGCUGTTTTCGACGACGACGACGACGAC	11	77.5	10± 2.4	>100	>10
P910b	GCUGCUGCAACUGCUGCUGTTTTCGACGACGACGACGACGAC	9,10	74.1	6.6 ± 0.6	>100	>15
REPC-b	CUGCUGCUGAUGCUGCUGCTTTTGACGACGACGACGACGACG	10	77.2	7.5 ± 2.1	>100	>13
REPU-b	UGCUGCUGCAGCUGCUGCUTTTTACGACGACGACGACGACGA	10	80.2	4.0 ± 1.2	>100	>25
Non-complementary control siRNA
CM	GCUAUACCAGCGUCGUCAUTT		80.0	-	-	-
siRNA tested on HD patient fibroblasts with shorter CAG repeat
P9	* in GM04719, 44mut/15wt			4.5 ± 1.3	>100	>22
P910	* in GM04719, 44mut/15wt			8.9 ± 2.2	>100	>11
P9	* in GM04717, 41mut/20wt			2.6 ± 0.6	71 ± 43	27
P910	* in GM04717, 41mut/20wt			4.2 ± 0.8	57 ± 16	13

The guide strand is shown from 5’ to 3’. Mismatched bases are underlined, in red, and in italics. Both strands of the duplexes containing two TT overhang. siRNAs were tested in HD patient fibroblasts GM04281 (mutant allele/69 CAG, wild type/17 CAG repeats) unless otherwise noted. Selectivity is calculated by dividing the IC₅₀ for inhibition of wild-type HTT versus the IC₅₀ for inhibiting expression of the mutant HTT protein.

We tested duplexes containing mismatches placed sequentially throughout the RNA duplex (Table 1, Supplementary Fig 1bc). siRNAs P8, P9, P10, P10R, and P11 containing centrally-located mismatches yielded greater selectivity for inhibition of mutant versus wild-type HTT expression (> 20 to >37 fold) than duplexes P4, P5, P6, P7, P12, P13, and P16 with mismatches closer to the 5’ or 3’ termini (4-9 fold). Mutations within the RNA duplex disrupt cleavage of the mRNA by RISC (Du et al. 2005; Wang et al. 2008) and our data indicated that shifting the mechanism for RNAi away from mRNA cleavage improves selectivity.

We tested duplexes containing several mismatches to establish the breadth of modifications that might be tolerated. When multiple mismatches are spread throughout the RNA duplex, potencies decrease substantially (Supplementary Fig. 1d). However, when two, three or four contiguous mismatches are clustered within the central portion of the siRNA, we observed potent and selective inhibition of mutant HTT (Fig. 1e, Supplementary Fig. 1e). The ability to introduce as many as four mismatches provides great flexibility for optimizing selective compounds for in vivo investigation and offers additional evidence for a mechanism that does not depend on RISC-mediated cleavage.

Half of HD patients have mutant alleles with > 45 CAG repeats (Duyao et al., 1993; MacDonald et al., 1993). To determine whether our strategy might be applied to a broad range of HD patients we tested mismatch-containing duplex RNAs in GM04719 (44 repeats/mutant; 15 repeats/wild-type allele) and GM04717 (41 repeats/mutant; 20 repeats/wild-type allele) patient-derived fibroblast cells. Previously, we had observed that single-stranded PNAs could achieve >1.8 and >1.5 fold selectivities for GM04719 and GM04717 cells respectively. Using mismatch-containing duplex RNAs we observed robust selectivities (> 10 to 20-fold) and potencies for inhibiting mutant HTT (< 10 nM) in both cell lines (Fig. 1f, Table 1, Supplementary Figs. 1fg), demonstrating substantial allele-selective inhibition across the full spectrum of CAG repeat lengths encountered in patients.

Duplex RNAs consist of a guide strand (antisense relative to the mRNA) and a passenger strand (sense relative to the mRNA). The guide strand is complementary to the target mRNA, is loaded into the RISC complex, binds to the mRNA target, and should be more sensitive to the introduction of mutations than the passenger strand. Duplexes P9a, P10a, and P910a with mutations on the passenger strand possessed relatively low selectivities (4 or 5 fold) (Fig. 2a, Supplementary Fig. 2a). Duplexes P9b, P10b, P10R/b, P11b, and P910b with mutations on the guide strand had selectivities up to > 43 fold (Fig. 2b, Supplementary Fig. 2bc), indicating that the mechanism of selectivity requires mutation of the guide strand.

Figure 2. Mechanism of allele-selective RNAs.

Effect on mutant and wild-type HTT expression after treatment with duplex RNAs that contain a mismatched base in either the (a) passenger strand, P9a, or (b) guide strand, P9b. (c) qPCR analysis of effect on HTT mRNA levels after treatment of increasing amount of siRNAs siHdh1, REP, and P9. (d) Effect of varied siRNAs (12 nM) on HTT mRNA levels measured by qPCR. (e) Effect of allele-selective duplex RNAs (25 nM) on expression of other genes containing regions with CAG repeats. CM: a duplex RNA not complementary to HTT mRNA. siHdh1: an siRNA that is complementary to HTT mRNA outside of the CAG repeat region (Omi et al., 2005). See also Figure S2.

We used quantitative PCR (qPCR) to evaluate levels of HTT mRNA after addition of mismatch-containing siRNA P9, fully complementary siRNA REP, and siRNA siHdh1 (Omi et al. 2005) that recognizes HTT mRNA at a sequence outside of the CAG repeat region. Relative to addition of REP or siHdh1, addition of P9 caused a significantly lower reduction in HTT mRNA levels (Fig. 2c). Mismatch-containing RNAs P910, PM3, and PM4 also resulted in little alteration of HTT mRNA levels (Fig. 2d). The finding that mismatch-containing duplex RNAs yield little decrease of HTT mRNA is consistent with a mechanism of action that involves blocking translation rather than mRNA cleavage.

Several physiologically important genes contain CAG repeats and inhibiting their expression would be detrimental. We examined expression two important repeat-containing genes, TATA box binding protein (TBP) (19 CAG repeats) and FOXP2 (40 glutamines encoded by a mixed CAG and CAA repeats). We observed that expression of these proteins was not inhibited by a panel of duplex RNAs that were added at 25 nM, a concentration substantially above their IC₅₀ values for inhibiting mutant HTT (Fig. 2e). We had previously observed that single-stranded oligomers targeting the CAG repeat that reduce levels of mutant HTT protein also did not affect expression of a panel of CAG repeat containing genes (Hu et al., 2009a).

There are no curative treatments of HD and strategies for reducing expression of HTT offer a direct approach towards mitigating or preventing the course of the disease. Improved delivery technologies (Smith et al. 2006; De Souza et al. 2009) are making nucleic acids a promising approach for neurodegenerative diseases. A Phase I clinical trial for amyotrophic lateral sclerosis (ALS) using an antisense oligonucleotide delivered directly into the central nervous system has recently been initiated (isispharm.com). Given the obvious potential for benefits to patients from reducing HTT levels and improving methods for nucleic acid delivery to the brain, approaches that use duplex RNAs or antisense oligonucleotides to silence HTT expression are a promising approach towards meeting a major unmet therapeutic need.

It is possible that non-allele specific inhibition of both wild-type and mutant HTT protein may be an adequate therapeutic approach. There may be a window where inhibition of the mutant allele will be sufficient to prevent or reverse formation of harmful aggregation of polyglutamine fragments while equivalent inhibition of the wild-type protein preserves sufficient HTT for essential normal functions. Agents that preferentially inhibit expression of mutant HTT, however, might widen this therapeutic window and improve treatment outcomes.

In our previous studies we examined inhibition of HTT expression by single-stranded oligomers targeting the CAG repeat (Hu et al. 2009a; Hu et al. 2009b; Hu et al. 2009c). Our data were promising, with significant allele-selectivity being achieved by several different chemically modified oligomers. Locked nucleic acid oligomers (LNAs), a chemistry already being tested in the clinical Phase II trials for treatment of other diseases (Koch et al. 2008), showed up to 6-fold allele-selectivity and provide a promising starting point for further development (Hu et al. 2009a). However, after exhaustively testing peptide nucleic acid (PNA), LNA, (S)-cET bridged nucleic acid, carba-LNA, ethylene nucleic acid (ENA), altriol nucleic acid (ANA), 2’-O-methoxyethyl (MOE), 2’-fluoro, and 2’-fluoroarabino nucleic acid (FANA) oligomers we were unable to identify any compound with better than 8-fold allele-selectivity (Gagnon, unpublished).

Our most effective single-stranded oligonucleotides inhibited expression by blocking gene translation rather than causing degradation of HTT mRNA. Since mismatched RNAs can also function through the RNAi pathway by blocking translation, we tested them and observed high levels of allele selectivity. Importantly, this selectivity was not a property of only one or two compounds. Several different arrangements of mismatched bases were potent and selective inhibitors of gene expression, including one compound with four contiguous mismatched nucleotides. These RNA duplexes demonstrate an ample supply of lead compounds to optimize for the potency, allele-selectivity, low toxicity, and biodistribution necessary for successful in vivo application.

SIGNIFICANCE

Developing curatives treatments for HD remains an urgent unmet medical need. Reducing the amount of mutant HTT would directly target the agent causing HD and duplex RNAs and antisense oligonucleotides have proven to be effective in engineered animal models. The challenge now is to improve the properties of anti-HTT nucleic acids sufficiently to enable clinical trials in humans. One important goal for improvement is identification of compounds that preferentially reduce levels of mutant HTT while preserving physiologically essential levels of wild-type HTT.

Mismatch-containing RNAs reduce expression of mutant HTT robustly, with low nM potencies and selectivities greater than 30-fold. Selective inhibition can be achieved in cell lines with mutant alleles containing low numbers of repeats, extending the approach to cell culture models representing almost all of the human patient population. The fact that several different compounds are potent and selective offers several starting points for the medicinal chemistry necessary to create molecules that will have optimal potency, selectivity, and biodistribution.

Beyond potential therapeutic application to HD, our work has implications for understanding recognition of RNA sequences during RNAi. The basic target sequence – multiple CAGs – is the same regardless of whether the repeat is expanded. The RISC complexes formed by duplex RNAs must sense structural variations or relatively modest differences in repeat length. This sensitivity can be translated into substantial improvements in recognition by a diverse group of mismatch-containing RNA duplexes. Introducing mismatches may allow gene silencing to be fine-tuned to achieve optimal results at specific RNA target sequences and structures.

Experimental Procedure

Full details of materials and experimental procedures used in this study can be found in the supplemental information that accompanies this Brief Communication at: http://www.xxxxxxxxxxx.xxxx