Role of Computationally Evaluated Target Specificity in the Hepatotoxicity of Gapmer Antisense Oligonucleotides

Abstract

Gapmer antisense oligonucleotides (gapmers) sometimes cleave nontarget pre-mRNAs by recognizing target-like intronic/exonic portions. This off-target RNA cleavage could be a major cause of the hepatotoxicity that is induced by gapmers. In line with these findings, we hypothesized that gapmers with higher specificity have less hepatotoxicity, and that those with lower specificity have greater toxicity. To examine this concept, we investigated various Malat1-targeting gapmers with various computationally evaluated target specificities. We had expected that higher specificity gapmers would have lower hepatotoxicity, but these factors were not significantly related. In silico analysis of gapmer sequences does not always contribute to mitigating the risk of hepatotoxicity. Transcriptome analysis indicated that nontoxic gapmers do not cleave off-target RNAs, although they have many target-like RNA sequences. The present results shed light on the mechanism of the hepatotoxicity of gapmers.

Introduction

Gap-modified antisense oligonucleotides, called gapmers, are promising therapeutics, but they sometimes cause severe hepatotoxicity [1 –4], for which four explanations have been considered: contamination of toxic by-products, knockdown-independent chemical/physical toxicity, knockdown-dependent on-target effects, and knockdown-independent off-target effects. We previously reported that much of the hepatotoxicity of gapmers could be derived from the off-target RNA cleavage mediated by intranuclear ribonuclease H1 (RNaseH1), which catalyzes the knockdown effect of gapmers. Chemical modification in the central region of toxic gapmers completely suppressed their hepatotoxicity, and gene silencing of hepatic RNaseH1 dramatically reduced the hepatotoxicity of gapmers [5].

Other groups also reported that hepatotoxic gapmers tended to cleave off-target RNAs. They found that the expression levels of many unintended transcripts, especially long pre-mRNAs, were downregulated by administration of toxic gapmers [6]. These findings led us to consider that gapmers with higher specificity effect less hepatotoxicity, and that those with lower specificity have greater toxicity. An ultimately target-specific gapmer that does not have any presumptive target in the entire genome is expected to show no hepatotoxicity.

One of the most simple and easy way to evaluate target specificity of a gapmer is counting mismatch-containing off-targets in whole transcripts, utilizing genome database. To examine if this simple in silico approach could contribute to estimating the hepatotoxicity of gapmers, we investigated the hepatotoxicity of several locked nucleic acid (LNA)-modified, Malat1-targeting gapmers with various computationally evaluated target specificities.

First, we synthesized two gapmers: Malat1-naïve, which is a previously reported gapmer [7], and Malat1 gapmer containing one mismatch (Malat1-1MM), the presumptive off-target genes of which there were fewer than for Malat1-naïve. We administered these gapmers to mice at a high dose (≤200 mg/kg), but neither naïve nor 1MM showed significant toxicity. We could not detect significant off-target knockdown in the liver of 1MM-administered mice using microarray.

Next, we selected four novel Malat1 gapmers with low to high specificity. We expected that the gapmer with lower specificity would cause more toxicity; however, only the gapmer with the highest specificity caused hepatotoxicity in vivo. We found that one of the presumptive off-target genes was downregulated by the hepatotoxic gapmer. Thus, the computationally predicted sequence specificity of gapmers is not always related to their hepatotoxicity. The present results shed light on the mechanism of the hepatotoxicity of gapmers.

Materials and Methods

Chemical compounds

LNA-modified, fully phosphorothioated gapmers were supplied from GeneDesign (Osaka, Japan). Some gapmers were designed using siDirect (http://sidirect2.rnai.jp). Lyophilized gapmers were dissolved in saline (Otsuka Pharmaceuticals, Tokyo, Japan), and utilized for in vitro and in vivo experiments. RNAlater was purchased from QIAGEN (Hilden, Germany). All other reagents and solvents were biochemical-grade commercial products.

Cell culture and transfection

The Hepa1c1c7 murine hepatoma cell line was obtained from ATCC (Manassas, VA), and maintained at 37°C and 5% CO₂ in Minimum Essential Medium Eagle-Alpha Modification (α-MEM; Invitrogen, Tokyo, Japan) that was supplemented with 10% heat-inactivated fetal bovine serum and antibiotics. Hepa1c1c7 was seeded at 2.0 × 10⁴ cells/well in 96-well culture plates. After 24 h, gapmers were transfected using Lipofectamine3000 (Invitrogen) according to the manufacturer's instructions. Cell lysates were prepared using the CellAmp Direct RNA Prep Kit for RT-PCR (Takara Bio, Otsu, Japan) 24 h after transfection and stored at −80°C.

Animal experiment

Male C57BL6/j mice (7-weeks old; CLEA Japan, Tokyo, Japan) were purchased and maintained per the Association for Assessment and Accreditation of Laboratory Animal Care International. All experimental protocols were approved by the Institutional Animal Care and Use Committee of Shionogi Co., Ltd., and all experiments were performed in accordance with the committee's guidelines.

Gapmers, dissolved in saline, were administered subcutaneously into the back of mice at the indicated doses. Plasma was collected from the vena cava, and livers were extracted under isoflurane anesthesia at the time when the mice were sacrificed. Plasma markers for hepatotoxicity—aspartate transaminase (AST) and alanine transaminase (ALT)—were measured using the Transaminase CII-test Wako Kit (Wako Pure Chemical Industries, Osaka, Japan) according to the manufacturer's instructions for 96-well format. Approximately 40 mg of liver sample was stored in RNAlater (QIAGEN) at 4°C overnight, and then at −80°C. Total RNA was purified using the RNeasy 96 Universal Kit (QIAGEN) according to the manufacturer's instructions. In brief, liver tissue was homogenized in 750-μL Qiazol reagent. The homogenate was mixed with 150-μL chloroform, and the supernatant, containing the RNA, was transferred to an RNeasy 96-well plate. The eluted RNA solutions were stored at −80°C.

Quantitative reverse-transcription PCR

Quantitative reverse-transcription PCR (qRT-PCR) was performed with One Step SYBR primescript (Takara Bio) on an ABI7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA) using 2 μL of cell lysate (for in vitro experiments) or 100 ng of total RNA (for in vivo experiments) as template. Quantified RNA levels were normalized to Gapdh, and are expressed relative to a saline or untreated control.

Microarray analysis and database screening

Total RNA isolated from mouse liver was analyzed using a whole-mouse genome (8 × 60K) Oligo Microarray Kit (Agilent Technology, Santa Clara, CA) according to the manufacturer's instructions. RNA transcripts containing a sequence that was similar to the target sequence of the gapmers (containing no or one mismatch) were screened using the ultrafast sequence searching database GGGenome (https://gggenome.dbcls.jp/en).

Results and Discussion

Sequence and target specificity of Malat1-naïve and -1MM

A previous report has shown that a Malat1 gapmer (Malat1-naïve) has potent knockdown effects at ≥3 mg/(kg·week) [7]. First, we focused on Malat1-naïve as a highly active gapmer model to investigate the relationship between target specificity and hepatotoxicity. Malat1-naïve possesses 5 target sequences that contain no mismatch and 89 one-mismatch-containing target sequences in the mouse genome. We redesigned Malat1-naïve by substituting a single nucleotide in the 3′ terminal region (G→T), (Malat1-1MM). Malat1-1MM showed higher specificity, with 1 presumptive target sequence with no mismatch and 76 presumptive 1-mismatch-containing target sequences in the mouse genome (Table 1).

Table 1.

Sequence and Target Specificity of Malat1-Naïve and Malat1-1MM

			No. of target sequences in mouse genome
	Sequence	Specificity	1 MM (RNA-coding genes)	0 MM (RNA-coding genes)
Malat1-naïve	CTAgttcactgaaTGC	Mid-Low	89 (37)	5 (3)
Malat1-1MM	CTAgttcactgaaTTC	Mid-High	76 (32)	1 (0)

Uppercase: LNA, lowercase: DNA. Underline indicates mismatch.

In vivo knockdown activity and hepatotoxicity

We compared the knockdown activity and hepatotoxicity of Malat1-naïve and -1MM in vivo. We had anticipated that Malat1-naïve would cause hepatotoxicity, whereas Malat1-1MM would show less hepatotoxicity than Malat1-naïve or none at all. Mice were sacrificed 7 days after gapmer injection, plasma and liver samples were collected, and knockdown activity and hepatotoxicity were examined. Both gapmers showed high knockdown activity; >95% of Malat1 expression was suppressed by only 10 mg/kg of the gapmers, and dose escalation did not increase the knockdown effect (Fig. 1A). Unexpectedly, neither of them showed hepatotoxicity at 200 mg/kg (Fig. 1B, C). This result was surprising, because all of the potent gapmers that we have ever examined were hepatotoxic at high doses (<100 mg/kg). Thus, Malat1-naïve and -1MM could be extraordinarily effective and nontoxic gapmers.

FIG. 1.

(A) Knockdown activity of Malat1-naïve and Malat1-1MM. (B, C) Plasma AST and ALT levels of mice that received gapmers. (D) qPCR analysis of presumptive off-target genes for MALAT1-1MM (n = 3, mean ± SD, N/D). ALT, alanine transaminase; ASO, antisense oligonucleotide; AST, aspartate transaminase; N/D, not detected; SD, standard deviation.

Transcriptome analysis of Malat1-1MM

The major cause of gapmer-derived hepatotoxicity is considered to be off-target knockdown. Thus, we examined whether off-target knockdown was induced by Malat1-1MM. Liver samples of the mice that received 20 mg/kg Malat1-1MM at 24 h before sacrifice were analyzed using microarray. We sacrificed mice at 24 h after gapmer administration because hepatotoxic side effect does not occur at 24 h (it appears after 48–72 h) [5]. Before the analysis, presumptive targets that contained <2 mismatches were identified using the GGGnome database. We analyzed the expression of these genes and found that eight genes were reduced by Malat1-1MM, although none reached statistical significance (Table 2); four of these genes harbored multiple target-like sequences in their pre-mRNA. No transcripts were upregulated by Malat1-1MM. Next, qRT-PCR was performed to confirm off-target knockdown in livers of mice that received 200 mg/kg of the gapmer 7 days before sacrifice. If the change of gene expressions (found by microarray analysis after 20 mg/kg administration) were caused by off-target pre-mRNA cleavage, 200 mg/kg of gapmer administration would cause much more drastic downregulation of those genes and the knockdown effect should continue until day 7. However, none of these eight genes was downregulated, indicating that off-target knockdown of these genes did not occur in the livers of mice that received even 200 mg/kg of Malat1-1MM (Fig. 1D). We cannot conclude that off-target knockdown did not occur with 200 mg/kg of Malat1-1MM, but these results imply that hepatotoxicity could not have developed without off-target RNA cleavage.

Table 2.

Presumptive Off-Target Candidate Genes Found by Microarray Analysis

	Gene expression (%)
Gene	Malat1-1MM	Negative control	Chr.	Position (from)	Position (to)	Sequence	MM	Intron/exon
Malat1	56	88	19	5801341	5801356	GcATTCAGTGAACTAG	1	Exon
P4ha3	51	60	7	100314159	100314173	aAATTCAGTGA_CTAG	2	Intron
March1	61	102	8	65906628	65906644	GAATaCAGTGgAACTAG	2	Intron
March1	61	102	8	66096279	66096294	GAATTtAGTtAACTAG	2	Intron
March1	61	102	8	66452302	66452318	GAATTCtGTGaAACTAG	2	Intron
Pitpnc1	62	83	11	107213826	107213842	GAATTttAGTGAACTAG	2	Intron
Pitpnc1	62	83	11	107333514	107333529	GcATgCAGTGAACTAG	2	Intron
Aff3	64	87	1	38443237	38443251	GAAgTCAG_GAACTAG	2	Intron
Aff3	64	87	1	38644310	38644324	GAATTCAGaGAACT_G	2	Intron
Oprm1	64	90	10	7034859	7034872	GAATTCAGT_AACTAa	2	Intron
Pdzrn3	65	91	6	101165688	101165703	GAATTtCAGTGAACTAt	2	Intron
Lrrc4c	65	91	2	96567636	96567651	GAATTCAGTGtgCTAG	2	Intron
Lrrc4c	65	91	2	96623137	96623151	GAATTCAGTGAgCTAc	2	Intron
Lrrc4c	65	91	2	96652385	96652400	GAgcTCAGTGAACTAG	2	Intron
Lrrc4c	65	91	2	97134081	97134098	GAATTCAGTGAcACTtAG	2	Intron
Lrrc4c	65	91	2	97197907	97197921	GAATTCA_TGAgCTAG	1	Intron
Lrrc4c	65	91	2	97405699	97405713	GAATTCAGTGAAaT_G	2	Intron
Golph3l	66	91	3	95594239	95594253	GAAcTCAGT_AACTAG	2	Intron

Malat1, shown in bold, is the ‘target gene’, not off-target candidate.

Knockdown activity and hepatotoxicity of the four gapmers

Next, we examined 10 newly designed gapmers to investigate the relationship between target specificity and hepatotoxicity. We examined their in vitro knockdown activity by transfecting them into the Hepa1c1c cell line (Fig. 2A). Six gapmers showed effective (>75%) knockdown activity, and we selected four of them that had varying computationally evaluated target specificities (Table 3). For example, #8 had the lowest specificity, harboring 2 target transcripts with no mismatches and 171 with target sequences containing 1 mismatch (68 of the 171 encode RNA transcripts) in the mouse genome. Surprisingly, #4 showed much higher specificity than the others, nearly targeting Malat1 exclusively; only five target sequences containing one mismatch were in the mouse genome, and only two of these five encoded RNA transcripts.

FIG. 2.

Knockdown activity and hepatotoxicity of newly selected MALAT1 gapmers. (A) In vitro knockdown activity of 10 candidate gapmers. (B) In vivo knockdown activity of the four gapmers. (C) Plasma AST and ALT levels of the mice that received gapmers. (D) Off-target knockdown of the two genes (n = 5 for A; n = 3 for B, C, and D; mean ± SD, *P < 0.01, t-test, vs. saline).

Table 3.

Sequence and Target Specificity of Four Newly Selected Gapmers

			No. of target sequences in mouse genome
No.	Sequence	Target specificity	1 MM (RNA-coding genes)	0 MM (RNA-coding genes)
#2	CCCtagatgtttaGCC	Mid-high	36 (19)	1 (1)
#4	ACTaccagcaattCCG	High	5 (2)	1 (1)
#8	ATCttatcttcttTGC	Low	171 (68)	2 (2)
#10	CAAacatactataCCC	Mid-low	108 (49)	2 (1)

Then, we examined the knockdown activity and hepatotoxicity of the four gapmers. The mice that received 40 mg/kg of the gapmers were sacrificed 7 days after injection, and plasma and liver samples were collected. As a result, all of the gapmers had potent (>90%) knockdown effects (Fig. 2B). Next, we analyzed ALT/AST levels in plasma and found that contrary to our expectation, only #4 showed hepatotoxicity, increasing ALT and AST levels fivefold to sixfold over the control (Fig. 2C). To investigate whether off-target knockdown had occurred, we performed qRT-PCR analysis of two presumptive targets that harbor one-mismatch-containing target sequences (Table 4). There were two candidate genes, Gpr107 and Phospho1, of which Gpr107 was significantly downregulated (Fig. 2D).

Table 4.

Sequence of Two Presumptive Off-Target RNAs That Contain 1 Mismatch for #4

Gene	Chr.	Position (from)	Position (to)	Sequence	Intron/exon
Malat1	19	5800409	5800424	CGGAATTGCTGGTAGT	—
Gpr107	2	31186646	31186661	GGGAATTGCTGGTAGT	Intron
Phospho1	11	95765194	95765209	CGTAATTGCTGGTAGT	Intron

Relationship between knockdown activity and hepatotoxicity

In summary, we first examined the relationship between knockdown activity and hepatotoxicity by redesigning a previously reported Malat1 gapmer, but it showed no hepatotoxicity, even at very high doses (200 mg/kg). We performed microarray analysis but could not detect off-target knockdown, although the specificity of these gapmers was not high. Conversely, the extremely specific gapmer, #4, which possesses only 2 one-mismatch-containing RNA-coding off-target candidates, showed significant hepatotoxicity in vivo, and one of the presumptive targets was significantly downregulated. These results indicate that in silico analysis of gapmer sequences does not always select for nontoxic gapmers. The computationally evaluated target specificity of gapmers might not be related directly to hepatotoxicity or off-target knockdown.

All gapmers have the potential to cause off-target knockdown and hepatotoxicity. The most important approach for the development of gapmers as drugs is to identify those that have potent knockdown activities without causing side effects at the appropriate dose. In terms of their effect on the transcriptome, gapmers do not differ from already approved small drugs; therefore, hybridization-mediated side effects must be carefully investigated [8]. Focusing on genome-wide, computationally evaluated target specificity is considered to be an important approach for avoiding the toxicity of gapmers but is insufficient to mitigate the risk of the side effects. In silico sequence analysis is quick and simple way, but we should not rely too much on it to predict sequence specificity and hepatotoxic potential. Both experimental and computational investigations could be required to assess the side effects of gapmers [9 –11]. Several effective, nontoxic gapmers have been identified, and some remain in clinical trials [7,12,13]. Furthermore, recently, novel modifications have been reported to lower the toxicity while maintaining or increasing the knockdown effect of the original gapmers [14 –16]. Analysis of sequence motifs has contributed to lowering the toxicity of gapmers [17]. Comprehensive studies that focus on target specificity and other factors should be performed to select gapmer candidates for clinical trials.

Footnotes

Acknowledgments

The authors thank Dr. Shin-ichiro Hori, Dr. Masatomo Rokushima, and Mr. Toru Yanagimoto (Shionogi & Co., Ltd.) for their technical supports and helpful comments. They also thank Edanz Group () for editing a draft of this article.

Author Disclosure Statement

No competing financial interests exist.

References

Bennett

and Swayze

. (2010). RNA targeting therapeutics: molecular mechanisms of antisense oligonucleotides as a therapeutic platform. Annu Rev Pharmacol Toxicol, 50:259–293.

Swayze

, Siwkowski

, Wancewicz

, Migawa

, Wyrzykiewicz

, Hung

, Monia

and Bennett

. (2007). Antisense oligonucleotides containing locked nucleic acid improve potency but cause significant hepatotoxicity in animals. Nucleic Acids Res, 35:687–700.

Stanton

, Sciabola

, Salatto

, Weng

, Moshinsky

, Little

, Walters

, Kreeger

, DiMattia

, et al. (2012). Chemical modification study of antisense gapmers. Nucleic Acid Ther, 22:344–559.

van Poelgeest

, Swart

, Betjes

, Moerland

, Weening

, Tessier

, Hodges

, Levin

and Burggraaf

. (2013). Acute kidney injury during therapy with an antisense oligonucleotide directed against PCSK9. Am J Kid Diş, 62:796–800.

Kasuya

, Hori

, Watanabe

, Nakajima

, Gahara

, Rokushima

, Yanagimoto

and Kugimiya

. (2016). Ribonuclease H1-dependent hepatotoxicity caused by locked nucleic acid-modified gapmer antisense oligonucleotides. Sci Rep, 6:30377.

Burel

, Hart

, Cauntay

, Hsiao

, Machemer

, Katz

, Watt

, Bui

, Younis

, et al. (2015). Hepatotoxicity of high affinity gapmer antisense oligonucleotides is mediated by RNase H1 dependent promiscuous reduction of very long pre-mRNA transcripts. Nucleic Acids Res, 44:2093–2109.

Hung

, Xiao

, Peralta

, Bhattacharjee

, Murray

, Norris

, Guo

and Monia

. (2013). Characterization of target mRNA reduction through in situ RNA hybridization in multiple organ systems following systemic antisense treatment in animals. Nucleic Acids Ther, 23:369–378.

Hagedorn

, Hansen

, Koch

and Lindow

. (2017). Managing the sequence-specificity of antisense oligonucleotides in drug discovery. Nucleic Acids Res, 45:2262–2282.

Lindow

, Vornlocher

, Riley

, Kornbrust

, Burchard

, Whiteley

, Kamens

, Thompson

, Nochur

, et al. (2012). Assessing unintended hybridization-induced biological effects of oligonucleotides. Nat Biotechnol, 30:920–923.

10.

Kamola

, Kitson

, Turner

, Maratou

, Eriksson

, Panjwani

, Warnock

, Douillard Guilloux

, Moores

, et al. (2015). In silico and in vitro evaluation of exonic and intronic off-target effects form a critical element of therapeutic ASO gapmer optimization. Nucleic Acids Res, 43:8638–8650.

11.

Kamola

, Maratou

, Wilson

, Rush

, Mullaney

, McKevitt

, Evans

, Ridings

, Chowdhury

, et al. (2017). Strategies for in vivo screening and mitigation of hepatotoxicity associated with antisense drugs. Mol Ther Nucleic Acids, 15:383–394.

12.

Burel

, Han

, Lee

, Norris

, Lee

, Machemer

, Park

, Zhou

, He

, et al. (2013). Preclinical evaluation of the toxicological effects of a novel constrained ethyl modified antisense compound targeting signal transducer and activator of transcription 3 in mice and cynomolgus monkeys. Nucleic Acid Ther, 23:213–227.

13.

Shen

and Corey

. (2017). Chemistry, mechanism and clinical status of antisense oligonucleotides and duplex RNAs. Nucleic Acid Res, 46:1584–1600.

14.

Nishina

, Piao

, Yoshida-Tanaka

, Sujino

, Nishina

, Yamamoto

, Nitta

, Yoshioka

, Kuwahara

, et al. (2015). DNA/RNA heteroduplex oligonucleotide for highly efficient gene silencing. Nat Commun, 6:7969.

15.

Seth

, Siwkowski

, Allerson

, Vasquez

, Lee

, Prakash

, Wancewicz

, Witchell

and Swayze

. (2009). Short antisense oligonucleotides with novel 2′-4′ conformationaly restricted nucleoside analogues show improved potency without increased toxicity in animals. J Med Chem, 52:10–13.

16.

Seth

, Allerson

, Berdeja

, Siwkowski

, Pallan

, Gaus

, Prakash

, Watt

, Egli

and Swayze

. (2010). An exocyclic methylene group acts as a bio-isostere of the 2′-oxygen atom in LNA. J Am Chem Socs, 132:14942–14950.

17.

Hagedorn

, Yakimov

, Ottosen

, Kammler

, Nielsen

, Høg

, Hedtjärn

, Meldgaard

, Møller

, et al. (2013). Hepatotoxic potential of therapeutic oligonucleotides can be predicted from their sequence and modification pattern. Nucleic Acid Ther, 23:302–310.