Distinct Poly(A) nucleases have differential impact on sut-2 dependent tauopathy phenotypes.

Aging drives pathological accumulation of proteins such as tau, causing neurodegenerative dementia disorders like Alzheimer's disease. Previously we showed loss of function mutations in the gene encoding the poly(A) RNA binding protein SUT-2/MSUT2 suppress tau-mediated neurotoxicity in C. elegans neurons, cultured human cells, and mouse brain, while loss of PABPN1 had the opposite effect (Wheeler et al., 2019). Here we found that blocking poly(A) tail extension with cordycepin exacerbates tauopathy in cultured human cells, which is rescued by MSUT2 knockdown. To further investigate the molecular mechanisms of poly(A) RNA-mediated tauopathy suppression, we examined whether genes encoding poly(A) nucleases also modulated tauopathy in a C. elegans tauopathy model. We found that loss of function mutations in C. elegans ccr-4 and panl-2 genes enhanced tauopathy phenotypes in tau transgenic C. elegans while loss of parn-2 partially suppressed tauopathy. In addition, loss of parn-1 blocked tauopathy suppression by loss of parn-2. Epistasis analysis showed that sut-2 loss of function suppressed the tauopathy enhancement caused by loss of ccr-4 and SUT-2 overexpression exacerbated tauopathy even in the presence of parn-2 loss of function in tau transgenic C. elegans. Thus sut-2 modulation of tauopathy is epistatic to ccr-4 and parn-2. We found that human deadenylases do not colocalize with human MSUT2 in nuclear speckles; however, expression levels of TOE1, the homolog of parn-2, correlated with that of MSUT2 in post-mortem Alzheimer's disease patient brains. Alzheimer's disease patients with low TOE1 levels exhibited significantly increased pathological tau deposition and loss of NeuN staining. Taken together, this work suggests suppressing tauopathy cannot be accomplished by simply extending poly(A) tails, but rather a more complex relationship exists between tau, sut-2/MSUT2 function, and control of poly(A) RNA metabolism, and that parn-2/TOE1 may be altered in tauopathy in a similar way.

Introduction

Abnormal intracellular inclusions composed of the microtubule-binding protein tau are seen in a number of age-associated neurodegenerative diseases called tauopathies (Irwin, 2016). The most prevalent tauopathy, Alzheimer’s disease (AD), is the most common cause of dementia (Gaugler et al., 2016) and it is the burden (extent) of tau tangles, rather than the burden (density or distribution) of amyloid plaques, that most strongly correlates with cognitive decline in AD (Bierer et al., 1995; Riley et al., 2002; Tiraboschi et al., 2004). Additionally, multiple mutations in the MAPT gene, which encodes tau protein, have been directly linked or associated with frontotemporal dementia with parkinsonism-17 (Irwin, 2016; Wang and Mandelkow, 2016). Not only are there still no pharmacological treatments that reverse or cure tauopathies, the molecular disease mechanisms underlying tauopathy are still unclear (Khanna et al., 2016; Wang and Mandelkow, 2016).

To further understand mechanisms of tauopathy and to discover novel genes and pathways that regulate tauopathy, we previously developed a C. elegans model that expresses human tau in all its neurons (Kraemer et al., 2003). Phenotypes seen with this model include uncoordinated locomotion, accumulation of insoluble tau, neuron loss, and shortened life span (Kraemer et al., 2003). We have used this model to screen for novel modifiers of tauopathy including forward mutagenesis screens (Guthrie et al., 2009; Kraemer and Schellenberg, 2007), whole genome RNAi screening (Kraemer et al., 2006), and chemical drug library screening (McCormick et al., 2013).

sut-2 was identified as a novel suppressor of tauopathy in forward mutagenesis screening of tau transgenic C. elegans (Guthrie et al., 2009). Loss of function mutations reversed locomotor deficits, decreased insoluble tau accumulation, and reduced neurodegeneration. The mammalian homolog, MSUT2, was also shown to modulate tauopathy in mouse models of tauopathy (Wheeler et al., 2019). Additionally, MSUT2 protein levels are altered in Alzheimer’s disease and associated with changes in tau pathology, inflammation, and age of disease onset (Wheeler et al., 2019).

Not much is known about the molecular function of MSUT2; MSUT2 has been shown to bind to poly(A) RNA and to the protein PABPN1, and together MSUT2 and PABPN1 localize to nuclear speckles (Wheeler et al., 2019). MSUT2 and PABPN1 have opposing effects on poly(A) tail length of mRNAs and tauopathy. MSUT2 knockdown by siRNA increases overall poly(A) tail lengths of mRNAs while PABPN1 knockdown decreases poly(A) tail length (Kelly et al., 2014). MSUT2 knockdown reduces insoluble and phosphorylated tau in a human cell culture model of tauopathy while PABPN1 knockdown increases insoluble and phosphorylated tau (Wheeler et al., 2019).

Poly(A) polymerases and poly(A) deadenylases also regulate poly(A) tail length of mRNAs (Eckmann et al., 2011; Zhang et al., 2010). Poly(A) polymerase activity increases poly(A) tail length while poly(A) deadenylase activity reduces poly(A) tail length. Controlling poly(A) tail length is important for mRNA stability and translation; while classically long poly(A) tails are thought to stabilize mRNA and stimulate translation of mRNA (Eckmann et al., 2011), a recent study has shown that shorter poly(A) tails are more common among highly-expressed genes in somatic cells and suggests that appropriate pruning of poly(A) tails is important for efficient translation (Lima et al., 2017). Since loss of MSUT2 leads to lengthened poly(A) tails, loss of poly(A) deadenylase expression could lead to the equivalent effect on poly(A) tails as well as on tauopathy. Various poly(A) deadenylases such as CCR4-NOT, PAN2/PAN3, and PARN are conserved from yeast to C. elegans to mammals and have distinct cellular localizations, regulatory mechanisms, and functions (Yan, 2014). Two genes that encode poly(A) deadenylases have been previously shown to modulate tauopathy in the C. elegans tauopathy model (Kraemer et al., 2006; Wheeler et al., 2019). Interestingly, loss of these genes enhanced tauopathy. To better understand how these various poly(A) deadenylases might modulate tauopathy, we screened a collection of poly(A) deadenylase gene mutants against the C. elegans tauopathy model. We found that some deadenylases could modulate tauopathy but did so differently than sut-2.

Materials and methods

Cell culture, cordycepin treatment, and RNA interference

HEK293 and HEK/tau cells (100 μg/mLL Zeocin) were cultured under standard tissue culture conditions (DMEM, 10% defined fetal bovine serum, Penicillin (1000 IU/mL) Streptomycin (1000 μg/mL) as previously described (Wheeler et al., 2019). Small interfering RNAs (siRNAs) were obtained from Integrated DNA Technologies (Coralville, IA). RNA interference and plasmid transient transfections were conducted following manufacture’s protocol (RNAiMAX, Invitrogen). Cells were fixed 96 h after transfection. For cordycepin treatment, cells were dosed with 100 mg/mL cordycepin (Tocris, catalog# 2294) 2 h prior to fixation.

C. elegans strains, genome editing, and transgenics

N2 (Bristol) was used as wild type C. elegans (Brenner, 1974). C. elegans alleles used in this study are listed in Table 1. bk3002, bk3045, and bk3046 were generated using CRISPR. All other strains containing desired alleles were obtained from the National Bioresource Project (Japan) or the C. elegans Genetics Center (CGC). C. elegans mutants were crossed into the tau transgenic C. elegans strain CK10 (bkIs10[aex-3p::tau-V337M], myo-2p::GFP) expressing human 4R1N tau with a V337M mutation (Kraemer et al., 2003). The presence of alleles of interest was verified with PCR genotyping. All strains were maintained at 20 °C on standard NGM plates containing OP50 Escherichia coli as previously described (Brenner, 1974). For protein studies, worms were grown on NGM plates containing 5× more peptone (5XPEP) prior to collection.

Table 1

List of C. elegans genes of interest and alleles tested.

C. elegans gene	Closest human homolog	Alleles tested	Molecular change
angl-1	ANGEL	tm1264	Deletion
ccr-4	CNOT6	tm1312	Deletion
panl-2	PAN 2	gk368072	Q474Ochre
panl-3	PAN 3	tm1182	Deletion
parn-1	PARN	ok1994	Deletion
		tm869	Deletion
parn-2	TOE1	tm1339	Deletion
		bk3045	Deletion
		bk3046	Deletion
sut-2	MSUT2	bk3002	Deletion

Two independent null sut-2 alleles (bk3001, bk3002) and two independent parn-2 alleles (bk3045, bk3046) were generated using CRISPR-Cas9 genome editing technology by introducing purified active recombinant Cas9 protein and synthetic CRISPR guide RNAs (gRNAs) as previously described (Paix et al., 2015).

Cas9 gRNAs were prepared by mixing equimolar amounts (0.5 μg/uL) of Alt-R crRNA and Alt-R tracrRNA (Integrated DNA Technologies) and then the gRNAs were complexed to equimolar quantities of Cas9 protein (Alt-R S.p. Cas9 Nuclease V3, Integrated DNA Technologies as described). dpy-10 was utilized as a gene co-conversion marker as previously described (Arribere et al., 2014) and removed by outcrossing. For each genome editing effort, two gRNAs were introduced targeting the promoter/start codon and 3’UTR/stop codon sequences of the sut-2 and parn-2 genomic loci respectively to generate null alleles.

Behavior in C. elegans

NGM plates containing 4-day-old tau transgenic C. elegans or 3-day-old non-transgenic C. elegans, corresponding to roughly day 1 adult C. elegans, were flooded with 1 mL of M9 buffer. Swimming worms were pipetted onto a 35 mm unseeded NGM plate. 1 min following the addition of M9 buffer videos of swimming worms were recorded for 1 min at 14 frames per second. Videos were analyzed using WormLab 2019 (MBF Bioscience). The frequency of body bends, defined as “turns” by the software, was quantified. A turn was defined as a change in the body angle defined by the quarter points and midpoint of the worm that was at least 20 degrees positive or negative from a straight line. Worms that were tracked for less than 20 s were omitted from analysis. At least 50 worms were counted per strain when comparing non-transgenic C. elegans to deadenylase mutants and at least 100 worms from at least 3 independent samples were counted per strain when comparing tau transgenic C. elegans to tau transgenic C. elegans containing deadenylase mutations.

Poly(A) tail length assay

Staged young adult tau transgenic C. elegans were grown from eggs at 20 °C for 3 days on 5XPEP plates, washed off plates in M9 buffer, and collected by centrifugation. RNA was extracted from ~100 μL of packed worms per sample using Tri-Reagent (Molecular Research Center) as per manufacturer instructions. Poly(A) tail lengths were assayed using the USB Poly(A) Tail-Length Assay Kit (Affymetrix) as per manufacturer instructions with the following modification to the PCR amplification step. The PCR reaction consisted of 10 μL of HotStart Taq, 7 μL of nuclease-free water, 0.5 μL of 10 μM gene-specific primer (see below), 0.5 μL of 10 μM Universal PCR Reverse Primer, and 2 μL of the diluted reverse transcription sample. A three-step PCR reaction with 40 cycles of amplification was used to amplify the signal. PCR products were run on a QIAexcel Advanced capillary electrophoresis instrument (Qiagen) using a QIAxcel DNA high resolution Kit with QX DNA Size Marker pUC18/HaeIII (catalog #929550) at 20 ng/uL and QX alignment marker 15 bp/600 bp (catalog #929530). Samples were processed using the 0M400 method with a 20 s sample injection time and analyzed with the QIAxcel ScreenGel Software for major peak, median signal, and concentration.

Tau extraction and immunoblotting

Staged young adult tau transgenic C. elegans were grown from eggs at 20 °C for 3 days on 5XPEP plates, washed off plates in M9 buffer, and collected by centrifugation. Worms were snap frozen in liquid nitrogen and stored at −70 °C. tau fractions were obtained as described previously (Kow et al., 2018) with some modifications. 2 μL of RAB Hi-Salt buffer (0.1 M MES, 1 mM EGTA, 0.5 mM MgSO4, 0.75 M NaCl, 0.02 M NaF, pH 7.0) containing PMSF and protease inhibitors was added per mg of worm pellet and homogenized by sonication. A portion of the sample was saved for immunoblotting (for total tau and tubulin) while the rest was centrifuged at 40,000 xg, 4 °C for 40 min. The supernatant was saved as the RAB/soluble fraction. The pellet was extracted by adding 1 μL of RIPA buffer (50 mM Tris, 150 mM NaCl, 1% Nonidet P-40, 5 mM EDTA, 0.5% deoxycholate, 0.1% SDS, pH 8.0) containing PMSF and protease inhibitors per mg of original worm pellet weight and centrifuged at 40,000 xg, 4 °C for 20 min. The supernatant was saved as the RIPA/detergent soluble fraction. The pellet was extracted again by adding 2 μL 70% formic acid (FA) per mg of original worm pellet weight and centrifuged at 13,000 xg, 4 °C for 15 min. The supernatant was the FA/detergent insoluble fraction. This fraction was dried in a speed vacuum concentrator until only a dry film remained in the tube.

Protein samples were diluted with 5× sample buffer (0.046 M Tris, 0.005 M EDTA, 0.2 M dithiothreitol, 50% sucrose, 5% sodium dodecyl sulfate, 0.05% bromophenol blue), boiled for 5 min., and centrifuged at 13,000 xg for 2 min. Prior to being loaded onto 4–15% pre-cast Criterion sodium dodecyl sulfate polyacrylamide gel electrophoresis gradient gels and transferred to PVDF membranes as recommended by the manufacturer (Bio-Rad). Total protein samples were diluted 2-fold with 5× sample buffer. FA/insoluble fractions were resuspended with 2× sample buffer at 2 μL per mg of original worm pellet weight. All other protein samples were diluted 5-fold with 5× sample buffer. 5–20 μL of diluted sample was loaded for each analysis. Primary antibodies used were rabbit monoclonal anti-tau antibody (Rockland) at 1:5000 and mouse anti-tubulin antibody E7 (Developmental Studies Hybridoma Bank) at 1:5000. Secondary antibodies used were anti-rabbit HRP (Jackson Immuno Research) and anti-mouse HRP (Jackson Immuno Research) at 1:5000. ECL substrate (Bio-Rad) was added to the membrane and chemiluminescence signals were detected with ChemiDoc-It Imager (UVP) and measured with UVP Software.

Immunohistochemistry

Immunohistochemistry for TOE1 was performed on 5-μm paraffin-embedded frontal cortex sections from 15 AD cases and 16 normal control cases. Brain sections were deparaffinized, rehydrated through ethanols and microwaved in citrate buffer for antigen retrieval. Sections were treated for endogenous peroxidases, blocked in 5% milk, incubated with anti-TOE1 polyclonal antibody (Invitrogen, PA5-30948) overnight at 4 °C, followed by biotinylated rabbit secondary antibody. Sections were incubated with an avidin-biotin complex (Vector, Vectastain Elite ABC kit) and the reaction product was visualized with 0.05% diaminobenzidine (DAB)/0.01% hydrogen peroxide. Our AD cohort was previously characterized as being MSUT2 positive or MSUT2 depleted and pTau burden (AT180) and neuronal loss (reduced NeuN immunoreactivity) was correlated with loss of MSUT2 (Wheeler et al., 2019). In the current study, the AD cohort was re-classified based on TOE1 immunostaining as AD, TOE1 positive and AD, TOE1 depleted, and it was determined that TOE1 expression corresponded with MSUT2 expression. Re-analysis of the quantitation of AT180 and NeuN immunoreactivity based on this TOE1 classification was performed to determine if pTau burden and neuronal loss also correlated with TOE1 depletion. Data were averaged and are represented as means +/− SEM. A two-tailed Student’s t-test was used to assess differences in staining intensity between the TOE1 positive and TOE1 depleted groups.

Fluorescent immunohistochemistry

Cells were grown on poly-D-lysine coated 12 mm round coverslips and fixed in 4% formaldehyde solution. Cells were washed 3 × 5 min in PBS/Ca2+/Mg2+, then blocked in antibody buffer (PBS, 0.5% Triton X-100, 1 mM EDTA, 0.1% BSA, 0.05% NaN3) with 10% normal goat serum. Primary antibodies were applied and incubated for 1 h at room temperature. Primary antibodies used were: pTau = AT180 (Thermo Fisher), TOE1 = Thermo Fisher (Catalog# PA5-30948), PARN = Protein Tech (Catalog# 13799-1AP), PAN2 = Protein Tech (Catalog# 16427-1AP), CNOT6 = Cell Signal (Catalog# 13415). Cells were washed 3 × 5 min in PBS/Ca2+/Mg2+, then re-blocked for 10 min. Appropriate Alexa dye-labeled secondary antibodies (Invitrogen) were applied and incubated for 20 min at room temperature. Cells were again washed 3 × 5 min in PBS/Ca2+/Mg2+, counterstained with 300 nM DAPI and mounted with ProLong Gold antifade (Molecular Probes). Microscopy was performed on a Delta Vision microscope (GE, Inc) using a 100× oil immersion objective, a sCMOS camera, and 2 × 2 binning. Image analysis was performed using softWoRx 6.0 Beta software (GE, Inc).

Human tissue

We obtained samples of postmortem tissue from the University of Washington Alzheimer’s Disease Research Center (ADRC) Neuropathology Core (PI, Dr. C. Dirk Keene) after receiving human subjects’ approval (University of Washington Human Subjects Division approval: HSD# 06-0492-E/A 01). We selected AD cases on the basis of having autopsy-confirmed AD (Braak stages V or VI with CERAD score of moderate or frequent). Control samples came from age matched neurologically healthy control participants with low levels of AD pathology (Braak stage III or less and CERAD scores of none or sparse).

Results

General inhibition of Poly(A) addition provokes pathological tau accumulation in an MSUT2 dependent fashion

Since loss of MSUT2 expression causes lengthening of poly(A) tails, we investigated whether generally reducing poly(A) tail lengths might affect tauopathy. The drug Cordycepin inhibits the addition of poly(A) to mRNAs (Rose et al., 1977; Wong et al., 2010). When we treated cultured human HEK293 cells that overexpress human tau with the compound cordycepin, we found a significant increase in phosphorylated tau (Fig. 1). This effect was blocked completely by co-treatment with RNAi against MSUT2 but not by PABPN1, confirming that MSUT2 specifically mediates the increase in pathological tau caused by inhibiting polyadenylation.

Fig. 1.

General inhibition of Poly(A) addition by cordycepin increases pathological tau in a MSUT2-dependent manner. Immunocytochemistry staining of HEK293 cells overexpressing tau protein with DAPI (blue), AT180 (detects phospho Thr231 of tau protein) (green), and actin (red). (a) HEK293 cells overexpressing human tau protein. (b) HEK293 cells overexpressing human tau protein treated with 100 μg/mL cordycepin. (c) HEK293 cells overexpressing human tau protein treated with 100 μg/mL cordycepin and MSUT2 RNAi. (d) HEK293 cells overexpressing human tau protein treated with 100 μg/mL cordycepin and PABPN1 RNAi. The RNAi knockdown efficiency for MSUT2 and PABPN1 was similar to that published previously (Wheeler et al., 2019). Scale bars = 15 μm.

A reverse genetic survey of mRNA deadenylase function in a C. elegans model of tauopathy

We used our C. elegans model of tauopathy to examine whether loss of any of the poly(A) deadenylases might phenocopy the effects of loss of sut-2 on tauopathy. Seven deadenylases have been identified in C. elegans via homology, belonging to the Ccr4-Not complex, the Pan2-Pan3 complex, or orthologs of individual deadenylase family members PARN, Angel, and 2′ phosphodiesterase (Nousch et al., 2013). In C. elegans, loss of the Ccr4-Not components caused significant lengthening of poly(A) tails, while loss of Pan2-Pan3 complex components or PARN genes did not cause longer poly(A) tails but only reduced the abundance of short poly(A) tails (Nousch et al., 2013). We obtained loss of function alleles (Table 1) that affected all known deadenylase complexes except for pde-12, whose homologs in other species are localized exclusively to mitochondria (Yan, 2014). We crossed them into our C. elegans tauopathy model and evaluated effects on tau-induced behavioral deficits (Fig. 2). We previously reported that loss of ccr-4 enhanced tau-induced behavior deficits (Wheeler et al., 2019). Here we confirmed with a second allele that loss of ccr-4 enhanced tau-induced behavior deficits. We also found that loss of panl-2 enhanced tau-induced behavior deficits, loss of parn-2 partially suppressed tau-induced behavior deficits, and angl-1 and panl-3 had no effect. Most of the alleles tested caused reduction in locomotor ability of ~10–20% in non-transgenic C. elegans (Supplemental Fig. 1), but the effects of each allele were significantly less on non-transgenic C. elegans than on tau transgenic C. elegans, suggesting that loss of these genes caused effects on tau-induced behavioral primarily through affecting tau toxicity. Although angl-1 was evaluated due to its similarity to the mammalian Angel, it is unclear if this gene is important for poly(A) tail regulation in C. elegans (Nousch et al., 2013). The fact that panl-2 and panl-3 did not have the same effect on tauopathy was surprising, since loss of either gene caused the same reduction in brood size of C. elegans raised at elevated temperatures (Nousch et al., 2013). Additionally, despite having different effects on tau, these deadenylases had similar effects on polyA tails in tau transgenic C. elegans (Supplemental Figure 2, Supplemental table 1). This suggests that the effects these deadenylases have on tauopathy may not correlate with their effects on other phenotypes.

Fig. 2.

Loss of function in various poly(A) deadenylase genes have varied effects on tau-induced behavior deficits in tau transgenic C. elegans. Swimming assays in M9 buffer were performed on 4-day-old tau transgenic (tau Tg) C. elegans (roughly day 1 adulthood). At least 100 worms per strain were analyzed by video software and frequency of body bends (turns) were calculated and normalized to control (tau Tg) for each double mutant. Error bars are means ± SEM. Student’s t-tests were used to determine significance of double mutant versus tau Tg. **** p < < 0.0001 when compared to tau Tg.

The C. elegans PARN homolog parn-1 is epistatic to parn-2 with respect to tauopathy

Since we observed that only parn-2 had significant effects on tau-induced behavioral deficits, we evaluated the effect of loss of both parn-1 and parn-2 on the same tau-induced behavioral deficits (Fig. 3). While loss of parn-1 and parn-2 had similar effects on poly(A) tail lengths in C. elegans (Nousch et al., 2013, Supplemental Figure 2), in human cell culture studies, the human homologs of parn-1 and parn-2, PARN and TOE1 respectively, have shown sometimes overlapping and sometimes non-redundant effects on poly(A) tails of non-coding RNAs (Deng et al., 2019; Son et al., 2018). We found that loss of parn-1 blocked the suppression of tauopathy caused by loss of parn-2. Interestingly this is opposite of the effect seen on brood size in C. elegans, where loss of parn-2 blocked the fertility deficit caused by loss of parn-1 in C. elegans raised at higher temperatures (Nousch et al., 2013). However, loss of parn-2 does not block the lengthening of piwi-interacting RNAs (piRNA) induced by loss of parn-1 (Tang et al., 2016). Our results suggest that in the modulation of tauopathy, parn-1 is dominant over parn-2.

Fig. 3.

Loss of parn-1 blocks the suppression of tau-induced behavioral deficits seen with loss of parn-2. Swimming assays in M9 buffer were performed on 4-day-old tau transgenic (tau Tg) C. elegans (roughly day 1 adulthood). At least 100 worms per strain were analyzed by video software and frequency of body bends (turns) was calculated. Error bars are means ± SEM. **** p < < 0.0001 when compared to tau Tg. Statistical significance was determined by two-way ANOVA followed by Tukey’s multiple comparisons tests. There was no significant difference between tau Tg;parn-1 and tau Tg;parn-1;parn-2 worms.

Loss of deadenylase function does not significantly alter tau aggregation

Next we evaluated whether loss of ccr-4, panl-2, parn-1, or parn-2 had any effect on total tau protein levels or tau aggregation. We performed sequential extractions of tau protein and found no significant changes in total tau or insoluble tau levels by any of these genes (Fig. 4). This contrasts with sut-2, MSUT2, and PABPN1, which have been previously shown to alter levels of insoluble tau (Guthrie et al., 2009; Wheeler et al., 2019). However, the effects of these deadenylase genes on tauopathy behavior, which varied between a ~ 50% reduction and ~ 50% increase in locomotor ability, was less significant than loss of sut-2 on tauopathy behavior which results in near complete rescue in locomotor ability (Guthrie et al., 2009); it is therefore possible that a significant change in tau protein aggregation was not detectable. Alternatively, the modulation of tau-induced behavioral phenotypes by these deadenylase genes may occur independent of modulating tau protein levels.

Fig. 4.

Loss of deadenylase genes did not significantly alter levels of total tau or insoluble tau protein. Lysates of 3-day-old tau transgenic C. elegans and double mutants were subjected to sequential extraction and total tau, insoluble formic-acid (FA) tau, and tubulin were detected by immunoblot. (a) Comparison of tau protein levels in tau Tg and tau Tg;ccr-4. (b) Comparison of tau protein levels in tau Tg and tau Tg;panl-2. (c) Comparison of tau protein levels in tau Tg and tau Tg;parn-1, tau Tg;parn-2, and tau Tg;parn-1;parn-2. Error bars are means — SEM. At least 3 replicates were analyzed per comparison. Student’s t-test was used for tau Tg vs tau Tg;ccr-4 or tau Tg;panl-2 comparisons, while one-way ANOVA for tau Tg vs tau Tg;parn-1, tau Tg;parn-2, and tau Tg;parn-1;parn-2 to determine statistical significance. No comparisons were significant.

SUT-2 modulation of tauopathy is epistatic to parn-2 and ccr-4

Although the various deadenylases we examined did not have the same effects on tauopathy as sut-2, since they all can interact with poly (A), they might affect how sut-2 modulates tauopathy. We investigated whether loss of parn-2, which suppresses tauopathy, could block enhancement caused by overexpression of SUT-2 and whether loss of ccr-4, which enhances tauopathy, could block suppression caused by loss of sut-2. We found that loss of parn-2 did not rescue the enhancement of tauopathy caused by overexpression of SUT-2 protein (Fig. 5a). We also found that loss of ccr-4 did not block suppression of tauopathy by loss of sut-2 (Fig. 5b). This shows that sut-2 is epistatic to both parn-2 and ccr-4 in regulation of tauopathy.

Fig. 5.

Loss of parn-2 and ccr-4 do not modulate effects of SUT-2 overexpression and sut-2 deletion respectively in tau transgenic C. elegans. Swimming assays in M9 buffer were performed on 4-day-old tau transgenic (tau Tg) C. elegans (roughly day 1 adulthood). At least 100 worms per strain were analyzed by video software and frequency of body bends (turns) were calculated. Error bars are means ± SEM. *** p < 0.005 **** P < < 0.0001 when compared to tau Tg. (a) Comparison of tau Tg C. elegans to tau Tg;parn-2, tau Tg;SUT-2:GFP, and tau Tg;parn-2;SUT-2:GFP worms. There was no significant difference between tau Tg;SUT-2:GFP and tau Tg;parn-2; SUT-2:GFP worms. (b) Comparison of tau Tg C. elegans to tau Tg;ccr-4, tau Tg;sut-2, and tau Tg;ccr-4;sut-2 worms. Statistical significance was determined by two-way ANOVA followed by Tukey’s multiple comparisons tests.

MSUT2 does not colocalize with poly(A) deadenylases

Because sut-2 was epistatic to parn-2 and ccr-4 in regulating tauopathy, we investigated whether this could be due to close interactions between sut-2/MSUT2 and poly(A) deadenylases in our tauopathy model. In cells, SUT-2/MSUT2 localizes to nuclear speckles (Guthrie et al., 2009; Wheeler et al., 2019), while various poly(A) deadenylases are found in the nucleus or cytoplasm, in Cajal bodies and/or P bodies (Yan, 2014). However, the localization of these deadenylases could be altered in tauopathy. Using cultured human HEK293 cells overexpressing human tau protein, we coimmunostained for MSUT2 and the human homologs of ccr-4 (CNOT6), panl-2 (PAN2), parn-1 (PARN), and parn-2 (TOE1) (Fig 6). Staining for CNOT6, PAN2, PARN, and TOE1 was similar to what was reported in previous literature (Bett et al., 2013; Deng et al., 2019; Mittal et al., 2011; Son et al., 2018). We found little co-localization of immunostaining between MSUT2 and any of these deadenylases. While immunostaining studies do not exclude the possibility of transient or limited interactions between MSUT2 and these deadenylases, it does suggest that the majority of MSUT2 protein exists in separate cellular compartmentsrelative to the deadenylases. This suggests that SUT-2/MSUT2 may not be blocking the effects of deadenylases on tauopathy via direct interactions.

Fig. 6.

Poly(A) deadenylases do not colocalize with MSUT2. Immunocytochemistry of HEK293 cells overexpressing human tau protein with DAPI (blue), MSUT2 (red), and deadenylase of interest (green). Colocalization (CLC) plots indicate amount of signal overlap between MSUT2 and deadenylase of interest. (a) Comparison of CNOT6 and MSUT2. (b) Comparison of PAN2 and MSUT2. (c) Comparison of PARN and MSUT2. (d) Comparison of TOE1 and MSUT2. Scale bars = 15 μm.

TOE1 expression correlates with MSUT2 expression in Alzheimer’s disease brains

Since loss of parn-2/TOE1 suppresses tau toxicity in our C. elegans model but its effect is blocked by overexpression of SUT-2, we investigated whether expression of TOE1 correlated with MSUT2 expression in Alzheimer’s disease. We previously demonstrated MSUT2 depletion correlates with disease severity in AD patients (increase in phosphorylated tau and increased neuronal loss) (Wheeler et al., 2019). In the current study, we found that neurons in the frontal cortex of postmortem brain tissue from these same AD brain donors with depleted MSUT2 also exhibited reduced TOE1 immunoreactivity. In contrast, AD donor brains with normal levels of MSUT2 demonstrated robust levels of TOE1, comparable to TOE1 immunoreactivity in normal control cases (Table 2, Fig. 7a,b). This TOE1 staining was primarily cytoplasmic in both normal and Alzheimer’s disease patient brains with no gross differences in intensity (Fig. 7b). Thus, compared to cases with normal levels of TOE1, cases with reduced TOE1 expression exhibited greater accumulation of pathological tau (Fig 7c,d) and increased loss of NeuN positive neurons (Fig. 7e,f). Taken together these data suggest that the mechanisms altering MSUT2 expression in Alzheimer’s disease also affect TOE1 expression.

Table 2

Characteristics of human AD cases.

Case ID	NPDX	AGE	SEX	ONS	DU	PMI	CERAD	Braak	MSUT2	TOE1
UWA730	AD	100	F	82	18	3:08	Moderate	VI	High	High
UWA763	AD	91	M	82	9	7:00	Moderate	VI	High	High
UWA771	AD	86	M	81	5	4:25	Frequent	VI	High	High
UWA1501	AD	77	M	71	6	6:32	Frequent	VI	High	High
UWA1518	AD	77	M	63	14	5:20	Frequent	IV	High	High
UWA1550	AD	94	M	82	12	4:25	Frequent	VI	High	High
UWA2385	AD	86	F	76	10	8:32	Frequent	VI	High	High
UWA3556	AD	89	M	86	3	6:25	Frequent	V	High	High
UWA797	AD	62	M	55	7	4:32	Frequent	VI	High	Low
UWA732	AD	89	F	82	7	5:23	Moderate	V	Low	Low
UWA1443	AD	84	F	76	8	4:47	Frequent	V	Low	Low
UWA3554	AD	77	M	57	20	7:25	Frequent	VI	Low	Low
UWA6511	AD	90	F	76	14	6:23	Frequent	VI	Low	Low
UWA6542	AD	70	F	62	8	5:00	Frequent	VI	Low	Low
UWA6589	AD	68	M	58	10	9:25	Frequent	VI	Low	Low

NPDX = neuropathology diagnosis.

ONS = onset.

DU = duration (years).

PMI = post-mortem interval (hours:minutes).

Fig. 7.

TOE1 expression levels parallel MSUT2 expression levels in Alzheimer’s disease brain, suggesting the mechanisms altering MSUT2 expression and tauopathy may also affect TOE1 expression. Representative brain sections of postmortem brain frontal cortex from a neurologically normal control and an MSUT2 positive or MSUT2 depleted AD case stained with MSUT2 (a) or TOE1 (b). AD cases that were positive for MSUT2 had robust levels of TOE1, while AD cases depleted of MSUT2 were also deficient in TOE1 immunoreactivity. Representative sections of postmortem cerebral cortex from TOE1 positive or TOE1 depleted AD cases stained with AT180 (detects phospho Thr231 of tau protein) (c) or NeuN (e). The AD brains with depleted levels of mSUT2 (a) and TOE1 (b) demonstrated increased pTau burden (c,d) and increased neuronal loss as indicated by a decrease in NeuN immunoreactivity (e,f). (d,f) Densitometry analysis of AT180 and NeuN positive reactivity in TOE1 positive AD cases (n = 8) compared to TOE1 depleted AD cases (n = 7). (AT180, *p = 0.015 and NeuN, **p = 0.0056 by two-tailed Student’s t-test). Scale bars = 100 μm.

Discussion

Cordycepin has been previously demonstrated to limit polyadenylation by acting as a poly(A) polymerase inhibitor (Rose et al., 1977; Wong et al., 2010). Although inhibiting polyadenylation with the chemical inhibitor cordycepin exacerbated pathological tau accumulation, if poly(A) tail length regulation was the only mechanism by which sut-2 modulated tauopathy, then loss of poly(A) deadenylase activity, which increases poly(A) tail length, would all have the same effect on tauopathy as sut-2. By screening various poly(A) deadenylases for modulation of pathological tau, we have found both deadenylases that enhance tau toxicity when lost (ccr-4, panl-2) and suppress tau toxicity when lost (parn-2) even though they had similar effects on poly(A) tail length. Thus poly(A) tail length regulation alone fails to explain the impact of any deadenylases on tauopathy. Furthermore, despite different mechanisms and cellular localizations, sut-2 was epistatic to ccr-4 and parn-2 in modulating tauopathy. Our data suggest that poly(A) deadenylases and sut-2 regulate tauopathy indirectly through poorly understood non-poly(A) mediated mechanisms.

It is unclear how deadenylases modulate behavioral deficits caused by tau overexpression in tau transgenic C. elegans when they cause no detectable change in tau protein levels or solubility. One possibility is that there is still a change in the conformation of tau protein that affects its toxicity despite no change in solubility. Another possibility is that toxic tau interacts with substrates that are upregulated when deadenylase activity is lost, causing a change in toxicity. For example, perhaps loss of deadenylases changes the availability of tau binding partners permitting additional tau-tau self-association and initiation of a toxic aggregation cascade.

Interestingly, even though parn-2/TOE1 differs from sut-2/MSUT2 in terms of function and cellular localization, TOE1 and MSUT2 expression varied in the same way in Alzheimer’s disease brains. This suggests a similar mechanism may lead to TOE1 and MSUT2 downregulation in earlier onset Alzheimer’s disease. However, whether TOE1 and MSUT2 expression becomes downregulated simultaneously, sequentially, or independently is impossible to discern from study of postmortem human tissue.

It is unclear mechanistically how TOE1 works in aging and neurodegenerative diseases of aging. Familial loss of function mutations in TOE1 have been shown to cause pontocerebellar hypoplasia type 7, where neurodegeneration in patient brains occurs so early it overlaps with neurodevelopment (Lardelli et al., 2017). Otherwise in mitotic cells, TOE1 has been shown to modulate small nuclear RNA and telomerase RNA maturation (Deng et al., 2019; Son et al., 2018). The fact that PARN has a similar function in regulating non-coding RNA (ncRNA) and telomerase RNA function but acts as a mild enhancer of tau toxicity instead indicates that general defects in ncRNA and telomerase RNA function do not explain suppression of tau toxicity by parn-2 in tau transgenic C. elegans. However, TOE1 knockdown in HeLa cells affected some ncRNA that PARN knockdown did not (Deng et al., 2019; Son et al., 2018), so perhaps the regulation of specific targets leads to suppression versus enhancement of tau toxicity. Additionally, while specific regulation of piRNA length has been ascribed to parn-1 in C. elegans (Tang et al., 2016), no unique roles have been identified yet for parn-2. Further investigation into the distinct functions of parn-1 and parn-2 as well as better understanding of the role of parn-2/TOE1 in post-mitotic cells like adult neurons may shed more light on its ability to modulate tauopathy and bring a more precise mechanistic understanding of why parn-2/TOE1 has an effect on tauopathy while the related deadenylase parn-1/PARN did not, and why it has opposite effects on tauopathy compared to other deadenylases panl-2 and ccr-4.

With triple mutant analysis we observed that sut-2 is epistatic to parn-2 and ccr-4. This suggests that sut-2 may interact downstream of parn-2 and ccr-4 pathways in modulating tau toxicity. Perhaps this is through reversing the effects of parn-2 and ccr-4 on tau-induced toxicity. Another possibility is that instead of acting downstream of parn-2 and ccr-4 in modulating tau toxicity, sut-2 interacts with tau upstream of the pathways that allow parn-2 and/or ccr-4-related pathways to interact with tau, diverting tau away from interactions with parn-2 or ccr-4.

Evidence that could support the latter idea is as follows. Previous work has suggested a variety of polyanions as a disease relevant molecular driver of aggregation seeding necessary to overcome charge repulsion between tau monomers (Friedhoff et al., 1998; Kuret et al., 2005). Indeed, recent work on tau fibril structure has shown that an uncharacterized polyanion resides in the core of corticobasal degeneration tau fibrils (Zhang et al., 2020). More specifically, in vitro tau forms complexes with RNA that potently seed tau fibril formation (Kampers et al., 1996). Furthermore, tau’s interaction with and polymerization of microtubules is disrupted by poly(A) RNA (Bryan et al., 1975; Kampers et al., 1996). Thus, one hypothesis is that MSUT2 is changing the availability of free RNA available to tau in the cytoplasm. Under this scenario, loss of sut-2/MSUT2 could reduce interactions between cellular poly(A) RNA and tau while overexpression of SUT-2/MSUT2 could promote interactions between RNA and tau that preclude interference by poly(A) polymerases. Therefore, the work described here can be viewed as a first probing of the in vivo mechanisms of RNA triggered tau aggregation.

In summary, we found that modulation of tauopathy by various poly (A)-related genes is not as simple as regulating bulk polyadenylation state. We found multiple deadenylases can modulate tauopathy, meriting further investigation into poly(A) regulating genes and tauopathy. We also identified a new link between parn-2/TOE1 and tauopathy with evidence that TOE1 expression correlates with disease severity in Alzheimer’s patients. Additional work is now necessary to determine how parn-2/TOE1 and sut-2/MSUT2 modulates disease mechanisms in tauopathy and how the many cellular deadenylases might differentially impact accumulation of pathological tau This includes investigating parn-2/TOE1 and sut-2/MSUT2 functions and molecular binding partners. While MSUT2 may be a promising target for tauopathy treatments (Wheeler et al., 2019), this study suggests expanding the scope of tau related therapeutic targets into other RNA metabolism regulating genes may be fruitful.

S.p. crRNA name	Protospacer	Relative Position in gene
parn-2 null 5′	AAAAATCGATAAAATTAGGG	parn2 start codon
parn-2 null 3′	TAACAATCAAGATCAGGTGG	parn2 stop codon
sut-2 null 5′	AGAGACTAGTAAAAAGGAGG	sut-2 promoter
sut-2 null 3′	GGACTAGTTAGCTGTGGCGG	sut-2 stop codon

Gene ofinterest	Primer sequence	Distance from end of 3′UTR
act-1	ACCACCAGCTTTCTATTTCTTGTTTGC	149
snb-1	TGATCCCACATCCACATGAGCTTATC	132
unc-54	AAAAATTGTGCTCCCTCCCC	103