Antisense suppression of the nonsense mediated decay factor Upf3b as a potential treatment for diseases caused by nonsense mutations
© The Author(s). 2018
Received: 1 June 2017
Accepted: 29 December 2017
Published: 15 January 2018
About 11% of all human genetic diseases are caused by nonsense mutations that generate premature translation termination codons (PTCs) in messenger RNAs (mRNA). PTCs not only lead to the production of truncated proteins, but also often result in decreased mRNA abundance due to nonsense-mediated mRNA decay (NMD). Although pharmacological inhibition of NMD could be an attractive therapeutic approach for the treatment of diseases caused by nonsense mutations, NMD also regulates the expression of 10–20% of the normal transcriptome.
Here, we investigate whether NMD can be inhibited to stabilize mutant mRNAs, which may subsequently produce functional proteins, without having a major impact on the normal transcriptome. We develop antisense oligonucleotides (ASOs) to systematically deplete each component in the NMD pathway. We find that ASO-mediated depletion of each NMD factor elicits different magnitudes of NMD inhibition in vitro and are differentially tolerated in normal mice. Among all of the NMD factors, Upf3b depletion is well tolerated, consistent with previous reports that UPF3B is not essential for development and regulates only a subset of the endogenous NMD substrates. While minimally impacting the normal transcriptome, Upf3b-ASO treatment significantly stabilizes the PTC-containing dystrophin mRNA in mdx mice and coagulation factor IX mRNA in a hemophilia mouse model.
Furthermore, when combined with reagents promoting translational read-through, Upf3b-ASO treatment leads to the production of functional factor IX protein in hemophilia mice.
These data demonstrate that ASO-mediated reduction of the NMD factor Upf3b could be an effective and safe approach for the treatment of diseases caused by nonsense mutations.
About 11% of all human genetic diseases are caused by nonsense mutations that generate premature translation termination codons (PTCs) in messenger RNAs (mRNA) [1, 2]. PTCs often inactivate gene function due to the production of truncated proteins and usually lead to a significant decrease in mRNA abundance due to degradation by the nonsense-mediated mRNA decay (NMD) pathway [3, 4]. In human diseases, the role of NMD can be twofold: on one hand, if the PTC results in production of a dominant-negative, truncated protein, NMD can protect cells by eliminating the aberrant transcripts; on the other hand, if the truncated protein is partially functional, NMD can lead to a more severe disease [5–7].
The development of therapeutic approaches for diseases caused by nonsense mutations has focused on small-molecule read-through agents [8, 9]. The goal of this type of therapy is to induce the translation machinery to recode a PTC into a sense codon so that translation continues in the correct reading frame to complete the synthesis of a full-length, potentially functional protein. As of today, no read-through therapy has received final approval for the treatment of diseases caused by nonsense mutations due to the lack of sufficient efficacy and/or safety. This is in part because of the low abundance of mRNA substrates available for translational read-through due to their degradation by NMD [8–11]. It has been shown that NMD attenuation by small molecule inhibitors or small interfering RNAs (siRNA) targeting NMD factors significantly enhances the efficacy of translational read-through drugs in cell lines derived from cystic fibrosis patients carrying a nonsense mutation in the CFTR gene , as well as in a mouse model of the lysosomal storage disease mucopolysaccharidosis I-Hurler (MPS I-H) caused by a PTC in the Idua gene locus . Inhibition of NMD alone also partially restores protein function by stabilizing PTC-containing mRNAs when the truncated proteins are functional as shown in Ullrich disease patient-derived fibroblasts [13, 14] and in a mouse model for neuronal ceroid lipofuscinosis [15, 16].
More than twenty proteins have been reported to play a role in NMD [4, 17–19]. The recognition and degradation of mRNAs with PTCs is mediated by sequential remodeling of protein–RNA complexes [17–19]. In mammals, the current model suggests that a PTC is recognized when the stop codon is distant from the poly(A) tail so that the translation termination factor ERF3 is recruited to the ribosome at a PTC, but binds UPF1 instead of PABP as during normal translation termination [18, 20]. This forms the SMG1–UPF1–eRF1–eRF3 (SURF) complex that then interacts with UPF2 and/or UPF3B, which, in some cases, is facilitated by the exon junction complex (EJC), to trigger UPF1 activation by phosphorylation [18, 20]. The phosphorylation of UPF1 is mediated by the kinase SMG1, which is regulated by SMG8 and SMG9 [18, 20]. Once UPF1 is activated, the mRNA is tagged for degradation. Phosphorylated UPF1 then recruits SMG6, which cleaves the mRNA near the PTC. The 3΄ RNA fragment is then rapidly degraded by XRN1 and the 5΄ fragment may be digested by the exosome [18, 20]. In addition, UPF1 also recruits the SMG5–SMG7 heterodimer that in turn recruits the CCR4-NOT complex to induce mRNA deadenylation-dependent decapping and subsequent XRN1-mediated degradation [18, 20].
Beyond its role in RNA surveillance, NMD is a post-transcriptional regulatory pathway that regulates 10–20% of the normal transcriptome across many species [4, 17–19]. Therefore, inhibition of the NMD pathway could have catastrophic effects on an organism, which is supported by the fact that several NMD factors are essential for early embryonic development in mouse [21–25]. Several lines of evidence suggest that NMD is not a single biochemical pathway in higher eukaryotes, but rather a pathway with several branches . Three branches of the NMD pathway diverging at the stage of PTC recognition were reported—UPF2-independent, EJC-independent, and UPF3B-independent branches—each of which only regulates a subset of the endogenous NMD substrates [26–28]. At the step of RNA destruction, several studies show that NMD substrate RNAs can be degraded through either SMG6-mediated endonucleolytic degradation or SMG5-SMG7-mediated degradation [29–32]. These branch-specific NMD factors could be potential therapeutic targets for diseases caused by nonsense mutations. However, it remains unclear if NMD can be effectively inhibited to stabilize disease-causing PTC transcripts with minimum impact on the normal transcriptome, resulting in an acceptable therapeutic index.
Here, we sought to identify those NMD components that could be depleted to effectively inhibit NMD to alleviate the phenotype of PTC-related genetic diseases, while simultaneously causing minimum toxicity to the organism. We used antisense oligonucleotides (ASOs) as tools to address this question. ASOs bind specifically to their RNA targets through Watson-Crick base pairing to form DNA–RNA heteroduplexes. These DNA–RNA heteroduplexes are substrates for the ubiquitous endonuclease RNase H1, which mediates the degradation of the target RNA strand [33, 34]. ASOs have proven to be specific, potent, and well tolerated treatment approaches for cardiovascular, metabolic, neurological, and severe genetic diseases and cancer .
In this study, we developed ASOs to specifically deplete mRNAs encoding mouse core NMD factors, the UPF and SMG proteins, to evaluate the efficacy and safety of NMD inhibition. Among the ten NMD factors targeted, we found that the ASO-mediated depletion of Upf3b efficiently suppressed NMD on certain disease-causing mRNAs and had a minimal global impact on the transcriptome. Our results suggest that targeting UPF3B with ASOs might be a viable approach for inhibiting NMD to ameliorate human diseases caused by nonsense mutations.
Identification of three categories of NMD regulators using ASOs
The second category contains the “modest” NMD regulators Upf3b, Smg5, and Smg7. Only when these NMD factors were depleted by > 70% was there an increase in signal from the PTC-containing reporter (Fig. 1 and Additional file 1: Figure S1). These factors also regulated fewer endogenous NMD substrates and to a lesser extent than the robust regulators (Fig. 2e–g). We categorized Smg8 and Smg9 as “passive” NMD regulators, as their depletion did not significantly alter the luciferase signal from the PTC-containing reporter (Fig. 1 and Additional file 1: Figure S1) and only slightly affected the levels of a few endogenous NMD substrates tested (Fig. 2h and i). Finally, depletion of Upf3a by > 99% did not inhibit reporter NMD and did not alter the abundance of the endogenous NMD substrates tested (Figs. 1 and 2j). While these results are consistent with a recent report that UPF3A acts primarily as an NMD inhibitor , we did not observe NMD activation with ASO-mediated Upf3a depletion in mouse MHT cells (Figs. 1 and 2j).
ASO-mediated Upf3b depletion is well tolerated in normal mice
Several NMD factors are essential for mammalian development. Complete knockout of Upf1, Upf2, Upf3a, Smg1, and Smg6 in mouse causes early embryonic lethality [21–25]. Heterozygote animals are normal, however, with fully functional NMD. Depletion of these NMD factors in adult animals has not been fully investigated. To address the question of whether there is a tolerable level of NMD inhibition in adult animals that is sufficient to alleviate disease phenotypes caused by nonsense mutations, we first conducted dose-response experiments in normal mice with ASOs targeting mRNAs encoding each of the ten NMD factors. Animals were dosed subcutaneously with ASOs twice a week for four weeks. Body weights were monitored throughout the study. At four weeks, necropsy was conducted and organ weights, as well as blood chemistry analyses including alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (T. Bil), albumin (Alb), and blood urea nitrogen (BUN) levels, parameters indicative of liver and kidney function, were measured (Additional file 2: Table S1) . The efficacies of NMD factor depletions in mouse livers were evaluated at both the RNA and protein levels (Additional file 2: Table S1 and data not shown).
As expected, ASO-mediated depletion of the robust NMD regulators was less well tolerated compared to the depletion of the modest and passive NMD regulators. Tolerability somewhat correlated with the degree of NMD pathway inhibition upon their depletion, as measured by upregulation of endogenous NMD substrates (Additional file 2: Table S1 and data not shown). Among all the NMD factors, the depletion of the robust NMD regulators Upf2 and Smg1 were least tolerated. Reduction of Upf2 mRNA by 80–90% caused mild to severe toxicity, demonstrated by less body weight gain and elevated plasma ALT and AST levels compared to control-ASO treated mice (Additional file 2: Table S1). This is consistent with the observation that Upf2 liver-specific deletion leads to liver injury and steatosis . Up to 60% depletion of Upf2 mRNA in mouse liver was tolerated (Additional file 2: Table S1). For Smg1, 60–70% mRNA depletion induced elevation of plasma ALT and AST levels (Additional file 2: Table S1). For Upf1, the target reduction in the liver plateaued at approximately 50%, as animals dosed weekly at 12.5, 25, or 50 mg/kg/week all had Upf1 mRNA and protein levels reduced to similar extents of about 50% of control levels (Additional file 2: Table S1 and data not shown). Although multiple additional ASOs were evaluated, none reduced Upf1 levels by > 50% in mouse liver (data not shown). Depletion of the final robust NMD factor, Smg6, was relatively well tolerated in normal mice compared to the depletion of the other robust NMD regulators, as > 90% mRNA reduction only resulted in slightly increased liver weights (by 30% compared to control animals) and ALT/AST levels (by approximately threefold compared to control animals) (Additional file 2: Table S1). This could be because SMG6 only functions in one of the two potentially redundant RNA degradation pathways that destruct mRNAs targeted for NMD .
Overall, we found that the depletion of the robust NMD factors was less well tolerated than the depletion of the modest or passive NMD factors. We hypothesized that the depletion of the robust NMD factors would elicit strong inhibition of NMD and significantly stabilize disease-associated PTC-containing mRNAs, but would likely lead to significant changes in the normal transcriptome and therefore would not be well tolerated as a therapeutic approach. In contrast, the depletion of the modest NMD factors might be sufficient to stabilize the disease-associated PTC mRNAs while having less impact on the normal transcriptome and therefore could be potential therapeutic targets for diseases caused by nonsense mutations. To test this hypothesis, we evaluated ASOs targeting mRNAs encoding NMD factors in mouse disease models.
Upf3b-ASO treatment stabilizes dystrophin PTC-containing mRNA
We did not detect full-length dystrophin protein in either Upf3b-ASO or Smg6-ASO treated mice (Additional file 1: Figure S4D). It is likely that the low level of the basal translational read-through that occurs is insufficient to produce detectable amounts of the full-length proteins in this mouse model. A combination of a read-through agent with ASOs targeting NMD factors might result in full-length protein production. However, in our hands, the small-molecule read-through drugs gentamicin and PTC124 [45, 46] had no effect in this model (data not shown), so we were unable to test this hypothesis.
Upf3b-ASO treatment stabilizes hF9_R29X mRNA and has minimum impact on the normal liver transcriptome in a mouse model of hemophilia B
Next, we tested the effect of the Upf3b-ASO in a hemophilia B mouse model. Hemophilia B is an X-linked bleeding disorder that results from a defect in the gene encoding coagulation factor IX (FIX), a serine protease that is critical for blood clotting . Persons with severe hemophilia B have functional FIX levels that are < 1% of normal values and have frequent bleeding episodes that are associated with crippling arthropathy and early death . Current treatment involves frequent intravenous injections of FIX protein concentrate; however, this treatment is prophylactic rather than curative and is associated with inhibitor formation . A rise in circulating FIX to 1% of normal levels can substantially ameliorate the bleeding phenotype . In this hemophilia mouse model, a human FIX minigene with a nonsense mutation (hFIX-R29X) is expressed in mice that lack the endogenous mouse FIX gene . These mice recapitulate the phenotype of patients carrying the R29X mutation (CGA to TGA) who have severe hemophilia with no detectable circulating FIX protein . The hFIX transgenes are driven by the human transthyretin promoter and thus are primarily expressed in mouse hepatocytes, which are the main cell type for endogenous factor IX production . The mRNA expressed from the hFIX-R29X minigene was shown to be degraded by the NMD pathway in HepG2 cells . The level of hFIX-R29X mRNA in these hemophilia mice is approximately 10% of the level of hFIX mRNAs in control mice that express a WT hFIX minigene [51, 52].
To improve ASO delivery to hepatocytes and minimize NMD inhibition in other cell types, we used an Upf3b-ASO conjugated to triantennary N-acetyl galactosamine (GalNAc). GalNAc is a high-affinity ligand for the hepatocyte-specific asialoglycoprotein receptor (ASGPR) [53–55]. GalNAc-conjugation results in enhanced ASO delivery to hepatocytes relative to non-parenchymal cells and potency that is six- to tenfold higher in mouse liver than an ASO of the same sequence without the GalNAc conjugation . Due to these altered pharmacokinetic and dynamic properties, we first optimized the dosing regimen for the GalNAc-conjugated Upf3b-ASO (Upf3b-GalNAc-ASO) in normal mice. Normal mice were dosed every five days with a total of six doses of parent ASOs at previously optimized concentrations, or with five- or tenfold lower concentrations of the Upf3b-GalNAc-ASO. We found that the Upf3b-GalNAc-ASO had at least fivefold higher potency than the parent ASO, and was well tolerated (Additional file 1: Figure S5). Similar to the Upf3b-GalNAc-ASO, an Upf1-GalNAc-ASO had more than fivefold higher potency than the parent ASO (Additional file 1: Figure S5D). Moreover, the Upf1-GalNAc-ASO was better tolerated than the parent Upf1-ASO in mouse. Treatment with the parent Upf1-ASO at 50 mg/kg led to less body weight gain than observed in controls, whereas mice treated with 10 mg/kg Upf1-GalNAc-ASO, which results in better Upf1 depletion, had weights similar to controls over the time course of the experiment (Additional file 1: Figure S5).
A Upf3b-GalNAc-ASO in combination with translational read-through molecules improved FIX coagulation activity in hemophilia B mice
To evaluate if the inhibition of the NMD pathway by Upf3b-GalNAc-ASO improves the efficacy of read-through therapy, we tested combination treatments of Upf3b-GalNAc-ASO and read-through agents in hemophilia B mice. Previously, treatment of hFIX-R29X mice with the small-molecule read-through drug geneticin alone did not result in detectable full-length protein in plasma . This could be because: (1) the hFIX-R29X mRNA level was insufficient for translational read-though to produce detectable amounts of full-length protein; and (2) this small molecule read-through drug is not efficient in promoting translational read-through of hFIX-R29X mRNA. Since the depletion of the translation termination factor eRF3a leads to efficient translational read-through of a reporter gene in human cells , we developed an ASO targeting Gspt1 mRNA, the mouse homolog of eRF3a, as a second approach in addition to geneticin to promote efficient read-through in vivo. In order to maximize the read-through efficiency, we also tested the triple combination of Upf3b-GalNAc-ASO, Gspt1-GalNAc-ASO, and geneticin.
hFIX-R29X mice were treated with Upf3b-GalNAc-ASO (10 mg/kg), Gspt1-GalNAc-ASO (5 mg/kg) or a combination of Upf3b-GalNAc-ASO (10 mg/kg) and Gspt1-GalNAc-ASO (5 mg/kg) every five days with six total doses for 4.5 weeks. During the last week of treatment, one group of each Upf3b-GalNAc-ASO-treated, Gspt1-GalNAc-ASO-treated or Upf3b-GalNAc-ASO/Gspt1-GalNAc-ASO-treated mice were also treated with 28 mg/kg geneticin daily for seven days. Animals were sacrificed 48 h after the last ASO treatments and 9 h after the last dose of geneticin.
In this study, we generated a catalog of efficacious and safe antisense oligonucleotides targeting each component of the NMD pathway. Testing these ASOs in vitro, we classified NMD factors into three categories: “robust,” “modest,” and “passive” based on their regulation of NMD. ASOs targeting the robust regulators Upf1, Upf2, Smg1, and Smg6 efficiently inhibited degradation of a reporter-gene-based NMD target and regulated most endogenous NMD substrates tested (Figs. 1 and 2). ASOs targeting the modest NMD regulators Upf3b, Smg5, and Smg7 only moderately affected the reporter and upregulated a few endogenous NMD substrates tested (Figs. 1 and 2). ASOs targeting the passive NMD regulators Smg8 and Smg9 did not alter reporter degradation and stabilized few endogenous NMD substrates (Figs. 1 and 2). In line with this, ASO-mediated depletion of Smg8 and Smg9 in mouse liver were well tolerated (Additional file 2: Table S1). Also, as expected, ASO-mediated depletion of the robust NMD regulators were less well tolerated compared to the depletion of the modest NMD regulators in vivo (Additional file 2: Table S1). These results are consistent with reports that the complete knock out of robust NMD regulators Upf1, Upf2, Smg1, and Smg6 result in early embryonic lethality in mice [21–24], whereas Upf3b-null mice are viable and fertile [38, 59].
Interestingly, depletion of the robust NMD factor Smg6 was better tolerated than the depletion of other robust NMD factors. SMG6 mediates endonucleolytic degradation of PTC-containing transcripts, which may also be degraded by the SMG5-SMG7 mediated exonucleolytic decay pathway. Recently, a transcriptome analysis revealed that SMG6 and SMG7 act on essentially the same transcripts, indicating extensive redundancy between the endo- and exonucleolytic decay routes . These results may underlie the tolerability of ASO-mediated Smg6 depletion in vivo. Both in vitro (Figs. 1 and 2) and in vivo (data not shown) results indicate that ASO-mediated Smg6 depletion induced a much more robust NMD inhibition compared to either Smg5 or Smg7 depletion. This suggests that, at least in liver cells, the SMG6-mediated endonucleolytic decay pathway could be the primary decay pathway for NMD and that the SMG5-SMG7-mediated exonucleolytic decay pathway could function as a secondary decay mechanism. In support of this, ASOs targeting Smg5 and Smg7 are much better tolerated in mouse liver than the Smg6-ASO (Additional file 2: Table S1).
The depletion of the third modest NMD regulator Upf3b to > 95% was well tolerated in both normal and diseased mice. This is probably because UPF3B only regulates a subset of the NMD substrates, as it has been reported that NMD can occur in the absence of UPF3B [28, 38, 59]. Using whole transcriptome analysis of mouse livers treated with ASOs, we confirmed that only ~ 2.8% of detected transcripts are affected by ASO-mediated Upf3b depletion, which is a very small subset compared to the ~ 11.4% of the liver transcriptome affected by ASO-mediated Upf1 depletion (Fig. 6).
To our surprise, the depletion of Upf3b elicited significant upregulation of the PTC-containing mutant mRNAs to levels equivalent to 60–70% of WT transcript expression in both mdx and hemophilia mouse models (Figs. 4 and 5). UPF3B levels increase during myogenesis, resulting in an increase in the efficiency of UPF3B-dependent NMD . This could contribute to the sensitivity of Upf3b depletion in the muscles of mdx mice. In the hFIX-R29X mice, the nonsense codon is located in the last exon of the minigene. This transcript, due to a lack of exon junctions downstream of the PTC, should be subjected to EJC-independent NMD. UPF3B protein was originally thought to function as a bridge between the EJC and UPF1. Recently, it has become clear that UPF3B also functions in the EJC-independent NMD pathway. In the absence of a downstream EJC, the recruitment of UPF3B to PTC-containing mRNAs occurs slowly, and therefore these PTC-containing transcripts are more sensitive to the depletion of UPF3B . Overall, our results suggested that the UPF3B-dependent NMD pathway plays an important role in degrading PTC-containing mutant transcripts in both mdx and hemophilia mouse models, which seems to be in accordance with a recent report that assigned a central role for UPF3B in the NMD pathway . Using a fully reconstituted in vitro translation system, Neu-Yilik et al. demonstrated that UPF3B plays dual role in both early and late translation termination and therefore is involved in the crosstalk between NMD machinery and the PTC-bound ribosome, a central mechanistic step of NMD . However, this raises the question of why the UPF3B-dependent NMD branch appears to regulate a small subset of endogenous substrates. The answer might lie in the fundamental differences in translation termination between PTCs in mutant transcripts and NMD-induceing stop codons in normal transcripts.
Although it significantly increased the PTC-containing transcript levels, Upf3b-ASO treatment alone did not ameliorate disease phenotypes in either mdx or hemophilia mouse model. This is likely due to two reasons: first, the truncated proteins produced in these two mouse models have no residual function; and second, the basal read-through levels are low for both of the transcripts. Importantly, Upf3b-ASO-mediated NMD inhibition did significantly improve the efficacy of the small molecule read-through drug geneticin (Fig. 7). Furthermore, when combined with Gspt1-ASO treatment to further promote translational read-through, plasma hFIX protein level rose to ~ 2–3% of that in the WT mice (Fig. 7d). The increased hFIX protein was functional, as shown by significant improvement in hFIX activity in plasma from mice treated with the triple combination of Upf3b-ASO, Gspt1-ASO, and geneticin (Fig. 7e). Therefore, inhibition of NMD with Upf3b-ASO could be part of a viable therapeutic approach for diseases caused by this type of nonsense mutation. In future studies, it will be important to evaluate the therapeutic effect of ASO-mediated Upf3b depletion as a standalone treatment in a disease model where the mutant protein remains at least partially functional.
Of note, NMD is a cell type- and tissue-specific process . Different cell types differ in their sensitivity to NMD factor depletion . For example, Smg8 was shown to be essential for the degradation of a PTC-containing mRNA encoding a mutant collagen in fibroblasts isolated from patients with Ullrich congenital muscular dystrophy . In this system, siRNAs targeting Smg8 restored defective mRNA and protein levels without affecting cell growth . Therefore, it will be important to test the effects of depleting each NMD factor in relevant models in vivo to identify the optimal therapeutic target. The ASOs we generated provide tools to enable these experiments in murine systems.
In this study, we took advantage of antisense technology to systematically target each component of the nonsense-mediated mRNA decay pathway to investigate whether NMD inhibition has the potential as a therapeutic strategy for treating human diseases caused by nonsense mutations. We demonstrated that depletion of the modest NMD regulator Upf3b was well tolerated in both normal and diseased mice. Importantly, we showed that ASO-mediated depletion of Upf3b significantly stabilized PTC-containing mutant transcripts and had a minimal impact on the normal transcriptome. Further, we demonstrated that Upf3b-ASO treatment significantly enhanced the efficacy of read-through therapy and led to improved coagulation activity in a hemophilia mouse model. Our results suggest that ASO-mediated depletion of the NMD factor UPF3B is potentially a safe and effective approach for the treatment of diseases caused by nonsense mutations.
Antisense oligonucleotides used in this study were chemically modified with phosphorothioate in the backbone and constrained ethyl (cET) modifications in the wings with a central 10-nucleotide deoxy gap (3-10-3 gapmer). Oligonucleotides were synthesized using an Applied Biosystems 380B automated DNA synthesizer (PerkinElmer Life and Analytical Sciences) and purified as previously described [34, 64]. ASO sequences are provided in Additional file 2: Table S1. Lyophilized ASOs were dissolved in sterile DPBS (without calcium or magnesium) and quantified by ultraviolet spectrometry, diluted to the desired concentration, sterilized through a 0.2-μm filter.
Cell culture and transfection
Mouse liver MHT cells  were cultured in DMEM containing 10% fetal bovine serum, in 5% CO2 at 37 °C. Cells were co-transfected with WT or mutant β-GLOBIN-luciferase construct encoding the Renilla luciferase with a Firefly luciferase control construct  using Effectene transfection reagent following the manufacturer’s protocol (Qiagen). Transfected cells were then selected by G418 and single colonies were isolated for further analysis. For ASO treatment, stably transfected cells were seeded at 96-well plate at concentration of 5000 cell/well. ASOs were added to the culture media 5–12 h after seeding cells at the indicated concentrations. Cells were harvested 72 h after ASO-treatment.
All the animals of wild-type (purchased from JAX), mdx  (purchased from JAX), and hemophilia mice  (licensed from The Children’s Hospital of Philadelphia and maintained in Taconic) genotypes were housed under standard conditions in a pathogen-free mouse facility. All animal studies were approved by Institutional Animal Care and Use Committees at Ionis Pharmaceuticals and were conducted in accordance with the United States Public Health Service’s Policy on Human Care and Use of Laboratory Animals. ASOs were administrated subcutaneously at a volume of 10 μL/g. Geneticin (G418) (5.6 mg/mL) was administrated subcutaneously at a volume 5 μL/g.
Plasma chemistry analysis
Blood samples were collected by cardiac puncture at time of sacrifice. Plasma chemistry values were measured on the AU480 Clinical Chemistry Analyzer (Beckman Coulter).
Cultured cells were lysed and the total RNA was extracted with Qiagen RNeasy columns. Animal tissues were homogenized in guanidine isothiocyanate solution (Invitrogen) supplemented with 8% 2-mercaptoethanol (Sigma-Aldrich). Total RNA was prepared using the RNeasy Mini Kit (Qiagen). Quantitative real-time PCR (qRT-PCR) was performed using an ABI step-one sequence detector. Taqman primer probe sequences are listed in Additional file 2: Table S4.
NMD factor protein levels were measured using western blot. Cultured cells were lysed with RIPA buffer (Thermo Fisher Scientific) containing Halt Protease Inhibitor Cocktail (Life Technologies). Animal tissues were homogenized in the same buffer. Protein concentrations were determined using the BioRad DC protein assay, and protein was loaded (25 ug for cell samples and 40 ug for tissue samples) onto a 4–15% Criterion™ TGX™ Precast Midi Protein Gel (BioRad). Western blot membranes were probed with primary anti-UPF1 antibody, anti-UPF2 antibody (both generously provided by Dr. Jens Lykke-Anderson (UCSD), anti-UPF3B antibody (Boster Immunoleader PB9843), anti-SMG1 antibody (Bethyl Laboratory A301-535A), anti-SMG6 antibody (Abcam ab87539), anti-SMG9 antibody (Abcam ab85659), or anti-GSPT1 antibody (Abcam ab49878). An antibody against β-Actin (Sigma A5316) was used as a loading control. Membranes were then incubated with IRDye secondary antibodies (Li-COR) and scanned using an Odyssey infrared system (Li-COR). Images were quantified using Image Studio (Li-COR). The western blot for dystrophin was performed using NuPAGE Novex 3–8% Tris-Acetate Protein Gel (Life Technologies). Blot was incubated with anti-Dystrophin (Leica NCL-DYS2; 1:100) and anti-alpha actinin (Abcam EP2529Y; 1:20,000) followed by anti-mouse IgG, HRP-linked antibody (CST#7076) anti-rabbit IgG, HRP-linked antibody (CST #7074) and then detected using the Amersham ECL Prime Western Blotting Detection Reagent (GE RPN2232 Life Sciences).
hFIX protein level in mouse plasma was measured by ELISA using Human Factor IX ELISA Kit (Abcam ab188393) following manufacture instructions.
RNA-sequencing (RNA-seq) was performed by sequencing fragmented libraries from purified total RNA using Illumina Tru-seq protocol. Samples were sequenced on an Illumina HiSeq2500 average depth of approximately 30 × 10^6 reads and had an average alignment rate of 70%. Read data were demultiplexed and transcript quantitation was performed using Salmon-Ver-0.7.2 with the quasi-mapping based mode and default parameters (Automated libType detection: -l A) . A Salmon gene model index was built using complementary DNA sequences from Ensembl Mus musculus Build-81 (genome build GRCm38) and used in read alignment. Expression values are reported as TPM. Genes were identified as differentially expressed assuming a model of gene expression variance derived from a negative binomial distribution and based upon gene expression levels in the reference cohorts. Gene-specific p values were computed for each biological replicate and median-aggregated. Genes whose median aggregate p value were ≤ 0.01 and had an average log2-fold-change magnitude > 1 within a group were considered significant and used in downstream analysis.
FIX activity assay
FIX activity assay was done at UCSD Murine Hematology and Coagulation Core Laboratory. In brief, clotting times are determined in duplicate with an ST4 semi-automated mechanical coagulation instrument (Diagnostica Stago, NJ). A total of 30 μL of citrated sample plasma diluted 1/10 in HN/BSA buffer are incubated with 30 μL of APTT reagent and 30 μL of human citrated plasma deficient of factor IX at 37 °C for 5 min, followed by the addition of 30 μL of 25 mM 37 °C CaCl2 to initiate clotting. Time until clot formation is measured and interpolated on a standard curve of serial dilutions citrated normal (BL/6 pool) mouse plasma tested as described to give reported result in % BL/6.
The authors thank Dr. Miles F. Wilkinson from UCSD, Dr. David Bedwell from UAB, and Dr. Stanley T. Crooke and Dr. C. Frank Bennett from Ionis Pharmaceuticals for their helpful discussions.
The work was supported by Ionis Pharmaceuticals in the form of salaries for authors LH, AL, SSD, MK, SK, SFM, BPM, and SG, but the company did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Availability of data and materials
The datasets generated and/or analyzed during the current study are available in the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106436 ).
Conceived and designed the experiments: LH, AL, SFM, BPM, SG. Performed the experiments: LH, AL, SSD, SK. Analyzed the data: LH, AL, SSD, SFM, SG. Wrote the paper: LH, MK, SG. All authors read and approved the final manuscript.
All animal studies were approved by Institutional Animal Care and Use Committees at Ionis Pharmaceuticals and were conducted in accordance with the United States Public Health Service’s Policy on Human Care and Use of Laboratory Animals.
The authors LH, AL, SSD, MK, SK, SFM, BPM, and SG are employees of Isis Pharmaceuticals, the funder of this study. This does not alter the authors’ adherence to all the Genome Biology policies on sharing data and materials.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Frischmeyer PA, Dietz HC. Nonsense-mediated mRNA decay in health and disease. Hum Mol Genet. 1999;8(10):1893–900.View ArticlePubMedGoogle Scholar
- Mort M, Ivanov D, Cooper DN, Chuzhanova NA. A meta-analysis of nonsense mutations causing human genetic disease. Hum Mutat. 2008;29(8):1037–47.View ArticlePubMedGoogle Scholar
- Nickless A, Bailis JM, You Z. Control of gene expression through the nonsense-mediated RNA decay pathway. Cell Biosci. 2017;7:26.View ArticlePubMedPubMed CentralGoogle Scholar
- Ottens F, Gehring NH. Physiological and pathophysiological role of nonsense-mediated mRNA decay. Pflugers Arch. 2016;468(6):1013–28.View ArticlePubMedGoogle Scholar
- Khajavi M, Inoue K, Lupski JR. Nonsense-mediated mRNA decay modulates clinical outcome of genetic disease. Eur J Hum Genet. 2006;14(10):1074–81.View ArticlePubMedGoogle Scholar
- Holbrook JA, Neu-Yilik G, Hentze MW, Kulozik AE. Nonsense-mediated decay approaches the clinic. Nat Genet. 2004;36(8):801–8.View ArticlePubMedGoogle Scholar
- Miller JN, Pearce DA. Nonsense-mediated decay in genetic disease: friend or foe? Mutat Res Rev Mutat Res. 2014;762:52–64.View ArticlePubMedGoogle Scholar
- Keeling KM, Bedwell DM. Suppression of nonsense mutations as a therapeutic approach to treat genetic diseases. Wiley Interdiscip Rev RNA. 2011;2(6):837–52.View ArticlePubMedPubMed CentralGoogle Scholar
- Nagel-Wolfrum K, Moller F, Penner I, Baasov T, Wolfrum U. Targeting nonsense mutations in diseases with translational read-through-inducing drugs (TRIDs). BioDrugs. 2016;30(2):49–74.View ArticlePubMedGoogle Scholar
- Linde L, Boelz S, Neu-Yilik G, Kulozik AE, Kerem B. The efficiency of nonsense-mediated mRNA decay is an inherent character and varies among different cells. Eur J Hum Genet. 2007;15(11):1156–62.View ArticlePubMedGoogle Scholar
- Linde L, Boelz S, Nissim-Rafinia M, Oren YS, Wilschanski M, Yaacov Y, et al. Nonsense-mediated mRNA decay affects nonsense transcript levels and governs response of cystic fibrosis patients to gentamicin. J Clin Invest. 2007;117(3):683–92.View ArticlePubMedPubMed CentralGoogle Scholar
- Keeling KM, Wang D, Dai Y, Murugesan S, Chenna B, Clark J, et al. Attenuation of nonsense-mediated mRNA decay enhances in vivo nonsense suppression. PLoS One. 2013;8(4):e60478.View ArticlePubMedPubMed CentralGoogle Scholar
- Usuki F, Yamashita A, Shiraishi T, Shiga A, Onodera O, Higuchi I, et al. Inhibition of SMG-8, a subunit of SMG-1 kinase, ameliorates nonsense-mediated mRNA decay-exacerbated mutant phenotypes without cytotoxicity. Proc Natl Acad Sci U S A. 2013;110(37):15037–42.View ArticlePubMedPubMed CentralGoogle Scholar
- Usuki F, Yamashita A, Kashima I, Higuchi I, Osame M, Ohno S. Specific inhibition of nonsense-mediated mRNA decay components, SMG-1 or Upf1, rescues the phenotype of Ullrich disease fibroblasts. Mol Ther. 2006;14(3):351–60.View ArticlePubMedGoogle Scholar
- Miller JN, Kovacs AD, Pearce DA. The novel Cln1(R151X) mouse model of infantile neuronal ceroid lipofuscinosis (INCL) for testing nonsense suppression therapy. Hum Mol Genet. 2015;24(1):185–96.View ArticlePubMedGoogle Scholar
- Miller JN, Chan CH, Pearce DA. The role of nonsense-mediated decay in neuronal ceroid lipofuscinosis. Hum Mol Genet. 2013;22(13):2723–34.View ArticlePubMedPubMed CentralGoogle Scholar
- Kurosaki T, Maquat LE. Nonsense-mediated mRNA decay in humans at a glance. J Cell Sci. 2016;129(3):461–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Karousis ED, Nasif S, Muhlemann O. Nonsense-mediated mRNA decay: novel mechanistic insights and biological impact. Wiley Interdiscip Rev RNA. 2016;7(5):661–82.View ArticlePubMedGoogle Scholar
- Hug N, Longman D, Caceres JF. Mechanism and regulation of the nonsense-mediated decay pathway. Nucleic Acids Res. 2016;44(4):1483–95.View ArticlePubMedPubMed CentralGoogle Scholar
- Fatscher T, Boehm V, Gehring NH. Mechanism, factors, and physiological role of nonsense-mediated mRNA decay. Cell Mol Life Sci. 2015;72(23):4523–44.View ArticlePubMedGoogle Scholar
- Medghalchi SM, Frischmeyer PA, Mendell JT, Kelly AG, Lawler AM, Dietz HC. Rent1, a trans-effector of nonsense-mediated mRNA decay, is essential for mammalian embryonic viability. Hum Mol Genet. 2001;10(2):99–105.View ArticlePubMedGoogle Scholar
- Weischenfeldt J, Damgaard I, Bryder D, Theilgaard-Monch K, Thoren LA, Nielsen FC, et al. NMD is essential for hematopoietic stem and progenitor cells and for eliminating by-products of programmed DNA rearrangements. Genes Dev. 2008;22(10):1381–96.View ArticlePubMedPubMed CentralGoogle Scholar
- McIlwain DR, Pan Q, Reilly PT, Elia AJ, McCracken S, Wakeham AC, et al. Smg1 is required for embryogenesis and regulates diverse genes via alternative splicing coupled to nonsense-mediated mRNA decay. Proc Natl Acad Sci U S A. 2010;107(27):12186–91.View ArticlePubMedPubMed CentralGoogle Scholar
- Li T, Shi Y, Wang P, Guachalla LM, Sun B, Joerss T, et al. Smg6/Est1 licenses embryonic stem cell differentiation via nonsense-mediated mRNA decay. EMBO J. 2015;34(12):1630–47.View ArticlePubMedPubMed CentralGoogle Scholar
- Shum EY, Jones SH, Shao A, Dumdie J, Krause MD, Chan WK, et al. The antagonistic gene paralogs Upf3a and Upf3b govern nonsense-mediated RNA decay. Cell. 2016;165(2):382–95.View ArticlePubMedPubMed CentralGoogle Scholar
- Gehring NH, Kunz JB, Neu-Yilik G, Breit S, Viegas MH, Hentze MW, et al. Exon-junction complex components specify distinct routes of nonsense-mediated mRNA decay with differential cofactor requirements. Mol Cell. 2005;20(1):65–75.View ArticlePubMedGoogle Scholar
- Buhler M, Steiner S, Mohn F, Paillusson A, Muhlemann O. EJC-independent degradation of nonsense immunoglobulin-mu mRNA depends on 3′ UTR length. Nat Struct Mol Biol. 2006;13(5):462–4.View ArticlePubMedGoogle Scholar
- Chan WK, Huang L, Gudikote JP, Chang YF, Imam JS, MacLean 2nd JA, et al. An alternative branch of the nonsense-mediated decay pathway. EMBO J. 2007;26(7):1820–30.View ArticlePubMedPubMed CentralGoogle Scholar
- Loh B, Jonas S, Izaurralde E. The SMG5-SMG7 heterodimer directly recruits the CCR4-NOT deadenylase complex to mRNAs containing nonsense codons via interaction with POP2. Genes Dev. 2013;27(19):2125–38.View ArticlePubMedPubMed CentralGoogle Scholar
- Colombo M, Karousis ED, Bourquin J, Bruggmann R, Muhlemann O. Transcriptome-wide identification of NMD-targeted human mRNAs reveals extensive redundancy between SMG6- and SMG7-mediated degradation pathways. RNA. 2017;23(2):189–201.View ArticlePubMedPubMed CentralGoogle Scholar
- Eberle AB, Lykke-Andersen S, Muhlemann O, Jensen TH. SMG6 promotes endonucleolytic cleavage of nonsense mRNA in human cells. Nat Struct Mol Biol. 2009;16(1):49–55.View ArticlePubMedGoogle Scholar
- Schmidt SA, Foley PL, Jeong DH, Rymarquis LA, Doyle F, Tenenbaum SA, et al. Identification of SMG6 cleavage sites and a preferred RNA cleavage motif by global analysis of endogenous NMD targets in human cells. Nucleic Acids Res. 2015;43(1):309–23.View ArticlePubMedGoogle Scholar
- Crooke ST. Molecular mechanisms of antisense oligonucleotides. Nucleic Acid Ther. 2017;27(2):70–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Bennett CF, Swayze EE. RNA targeting therapeutics: molecular mechanisms of antisense oligonucleotides as a therapeutic platform. Annu Rev Pharmacol Toxicol. 2010;50:259–93.View ArticlePubMedGoogle Scholar
- Bennett CF, Baker BF, Pham N, Swayze E, Geary RS. Pharmacology of antisense drugs. Annu Rev Pharmacol Toxicol. 2017;57:81–105.View ArticlePubMedGoogle Scholar
- Boelz S, Neu-Yilik G, Gehring NH, Hentze MW, Kulozik AE. A chemiluminescence-based reporter system to monitor nonsense-mediated mRNA decay. Biochem Biophys Res Commun. 2006;349(1):186–91.View ArticlePubMedGoogle Scholar
- Mendell JT, Sharifi NA, Meyers JL, Martinez-Murillo F, Dietz HC. Nonsense surveillance regulates expression of diverse classes of mammalian transcripts and mutes genomic noise. Nat Genet. 2004;36(10):1073–8.View ArticlePubMedGoogle Scholar
- Huang L, Lou CH, Chan W, Shum EY, Shao A, Stone E, et al. RNA homeostasis governed by cell type-specific and branched feedback loops acting on NMD. Mol Cell. 2011;43(6):950–61.View ArticlePubMedPubMed CentralGoogle Scholar
- Harrison Jr SD, Burdeshaw JA, Crosby RG, Cusic AM, Denine EP. Hematology and clinical chemistry reference values for C57BL/6 X DBA/2 F1 mice. Cancer Res. 1978;38(8):2636–9.PubMedGoogle Scholar
- Thoren LA, Norgaard GA, Weischenfeldt J, Waage J, Jakobsen JS, Damgaard I, et al. UPF2 is a critical regulator of liver development, function and regeneration. PLoS One. 2010;5(7):e11650.View ArticlePubMedPubMed CentralGoogle Scholar
- Blake DJ, Weir A, Newey SE, Davies KE. Function and genetics of dystrophin and dystrophin-related proteins in muscle. Physiol Rev. 2002;82(2):291–329.View ArticlePubMedGoogle Scholar
- Gao QQ, McNally EM. The dystrophin complex: structure, function, and implications for therapy. Compr Physiol. 2015;5(3):1223–39.View ArticlePubMedPubMed CentralGoogle Scholar
- Dent KM, Dunn DM, von Niederhausern AC, Aoyagi AT, Kerr L, Bromberg MB, et al. Improved molecular diagnosis of dystrophinopathies in an unselected clinical cohort. Am J Med Genet A. 2005;134(3):295–8.View ArticlePubMedGoogle Scholar
- Sicinski P, Geng Y, Ryder-Cook AS, Barnard EA, Darlison MG, Barnard PJ. The molecular basis of muscular dystrophy in the mdx mouse: a point mutation. Science. 1989;244(4912):1578–80.View ArticlePubMedGoogle Scholar
- Barton-Davis ER, Cordier L, Shoturma DI, Leland SE, Sweeney HL. Aminoglycoside antibiotics restore dystrophin function to skeletal muscles of mdx mice. J Clin Invest. 1999;104(4):375–81.View ArticlePubMedPubMed CentralGoogle Scholar
- Welch EM, Barton ER, Zhuo J, Tomizawa Y, Friesen WJ, Trifillis P, et al. PTC124 targets genetic disorders caused by nonsense mutations. Nature. 2007;447(7140):87–91.View ArticlePubMedGoogle Scholar
- DeLoughery TG. Hemostasis and thrombosis. Cham: Springer; 2015.View ArticleGoogle Scholar
- Nathwani AC, Tuddenham EG. Epidemiology of coagulation disorders. Baillieres Clin Haematol. 1992;5(2):383–439.View ArticlePubMedGoogle Scholar
- Kessler CM, Mariani G. Clinical manifestations and therapy of the hemophilias. In: Colman RW, Marder VJ, Clowes AW, George JN, Goldhaber SZ, editors. Hemostasis and thrombosis: basic principles and clinical practice. 5th ed. Philadelphia: Lippincott-Raven; 2006. p. 887–904.Google Scholar
- Nathwani AC, Reiss UM, Tuddenham EG, Rosales C, Chowdary P, McIntosh J, et al. Long-term safety and efficacy of factor IX gene therapy in hemophilia B. N Engl J Med. 2014;371(21):1994–2004.View ArticlePubMedPubMed CentralGoogle Scholar
- Sabatino DE, Armstrong E, Edmonson S, Liu YL, Pleimes M, Schuettrumpf J, et al. Novel hemophilia B mouse models exhibiting a range of mutations in the Factor IX gene. Blood. 2004;104(9):2767–74.View ArticlePubMedGoogle Scholar
- Yang C, Feng J, Song W, Wang J, Tsai B, Zhang Y, et al. A mouse model for nonsense mutation bypass therapy shows a dramatic multiday response to geneticin. Proc Natl Acad Sci U S A. 2007;104(39):15394–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Spiess M. The asialoglycoprotein receptor: a model for endocytic transport receptors. Biochemistry. 1990;29(43):10009–18.View ArticlePubMedGoogle Scholar
- Stockert RJ. The asialoglycoprotein receptor: relationships between structure, function, and expression. Physiol Rev. 1995;75(3):591–609.View ArticlePubMedGoogle Scholar
- Steirer LM, Park EI, Townsend RR, Baenziger JU. The asialoglycoprotein receptor regulates levels of plasma glycoproteins terminating with sialic acid alpha2,6-galactose. J Biol Chem. 2009;284(6):3777–83.View ArticlePubMedPubMed CentralGoogle Scholar
- Prakash TP, Graham MJ, Yu J, Carty R, Low A, Chappell A, et al. Targeted delivery of antisense oligonucleotides to hepatocytes using triantennary N-acetyl galactosamine improves potency 10-fold in mice. Nucleic Acids Res. 2014;42(13):8796–807.View ArticlePubMedPubMed CentralGoogle Scholar
- Morrissy AS, Morin RD, Delaney A, Zeng T, McDonald H, Jones S, et al. Next-generation tag sequencing for cancer gene expression profiling. Genome Res. 2009;19(10):1825–35.View ArticlePubMedPubMed CentralGoogle Scholar
- Chauvin C, Salhi S, Le Goff C, Viranaicken W, Diop D, Jean-Jean O. Involvement of human release factors eRF3a and eRF3b in translation termination and regulation of the termination complex formation. Mol Cell Biol. 2005;25(14):5801–11.View ArticlePubMedPubMed CentralGoogle Scholar
- Huang L, Shum EY, Jones SH, Lou CH, Dumdie J, Kim H, et al. A Upf3b-mutant mouse model with behavioral and neurogenesis defects. Mol Psychiatry. 2017. doi.10.1038/mp.2017.173.Google Scholar
- Gong C, Kim YK, Woeller CF, Tang Y, Maquat LE. SMD and NMD are competitive pathways that contribute to myogenesis: effects on PAX3 and myogenin mRNAs. Genes Dev. 2009;23(1):54–66.View ArticlePubMedPubMed CentralGoogle Scholar
- Metze S, Herzog VA, Ruepp MD, Muhlemann O. Comparison of EJC-enhanced and EJC-independent NMD in human cells reveals two partially redundant degradation pathways. RNA. 2013;19(10):1432–48.View ArticlePubMedPubMed CentralGoogle Scholar
- Neu-Yilik G, Raimondeau E, Eliseev B, Yeramala L, Amthor B, Deniaud A, et al. Dual function of UPF3B in early and late translation termination. EMBO J. 2017;36(20):2968–86.View ArticlePubMedPubMed CentralGoogle Scholar
- Huang L, Wilkinson MF. Regulation of nonsense-mediated mRNA decay. Wiley Interdiscip Rev RNA. 2012;3(6):807–28.View ArticlePubMedGoogle Scholar
- Zong X, Huang L, Tripathi V, Peralta R, Freier SM, Guo S, et al. Knockdown of nuclear-retained long noncoding RNAs using modified DNA antisense oligonucleotides. Methods Mol Biol. 2015;1262:321–31.View ArticlePubMedGoogle Scholar
- Koller E, Vincent TM, Chappell A, De S, Manoharan M, Bennett CF. Mechanisms of single-stranded phosphorothioate modified antisense oligonucleotide accumulation in hepatocytes. Nucleic Acids Res. 2011;39(11):4795–807.View ArticlePubMedPubMed CentralGoogle Scholar
- Bulfield G, Siller WG, Wight PA, Moore KJ. X chromosome-linked muscular dystrophy (mdx) in the mouse. Proc Natl Acad Sci U S A. 1984;81(4):1189–92.View ArticlePubMedPubMed CentralGoogle Scholar
- Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods. 2017;14(4):417–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Huang L, Damle S. Hemophilia mice treated with antisense oligonucleotides targeting either Upf1 or Upf3b. Gene Expression Omnibus. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106436. Accessed 3 Nov 2017.