Pumilio response and AU-rich elements drive rapid decay of Pnrc2-regulated cyclic gene transcripts

Abstract

Vertebrate segmentation is regulated by the segmentation clock, a biological oscillator that controls periodic formation of somites, or embryonic segments, which give rise to many mesodermal tissue types. This molecular oscillator generates cyclic gene expression with the same periodicity as somite formation in the presomitic mesoderm (PSM), an area of mesenchymal cells that give rise to mature somites. Molecular components of the clock include the Hes/her family of genes that encode transcriptional repressors, but additional genes cycle. Cyclic gene transcripts are cleared rapidly, and clearance depends upon the pnrc2 (proline-rich nuclear receptor co-activator 2) gene that encodes an mRNA decay adaptor. Previously, we showed that the her1 3’UTR confers instability to otherwise stable transcripts in a Pnrc2-dependent manner, however, the molecular mechanism(s) by which cyclic gene transcripts are cleared remained largely unknown. To identify features of the her1 3’UTR that are critical for Pnrc2-mediated decay, we developed an array of transgenic inducible reporter lines carrying different regions of the 3’UTR. We find that the terminal 179 nucleotides (nts) of the her1 3’UTR are necessary and sufficient to confer rapid instability. Additionally, we show that the 3’UTR of another cyclic gene, deltaC (dlc), also confers Pnrc2-dependent instability. Motif analysis reveals that both her1 and dlc 3’UTRs contain terminally-located Pumilio response elements (PREs) and AU-rich elements (AREs), and we show that the PRE and ARE in the last 179 nts of the her1 3’UTR drive rapid turnover of reporter mRNA. Finally, we show that mutation of Pnrc2 residues and domains that are known to facilitate interaction of human PNRC2 with decay factors DCP1A and UPF1 reduce the ability of Pnrc2 to restore normal cyclic gene expression in pnrc2 mutant embryos. Our findings suggest that Pnrc2 interacts with decay machinery components and cooperates with Pumilio (Pum) proteins and ARE-binding proteins to promote rapid turnover of cyclic gene transcripts during somitogenesis.

Summary statement:

We show that her1 and dlc 3’UTR regulatory sequences confer Pnrc2-mediated decay to reporter transcripts and that a Pumilio response element (PRE) and an AU-rich element (ARE) in the terminal her1 3’UTR drive transcript destabilization. Our work suggests that Pnrc2 interacts with decay machinery components and cooperates with Pumilio (Pum) proteins and ARE-binding proteins to promote rapid cyclic gene transcript turnover during somitogenesis.

INTRODUCTION

Genetic oscillations underlie many cellular events and function in the regulation of critical developmental processes. A well-studied example of genetic oscillation is the segmentation clock, a rapid ultradian oscillator that generates periodic expression in developing embryos (Hubaud and Pourquie, 2014; Oates et al., 2012; Pourquie, 2011). The segmentation clock controls vertebrate somitogenesis, the process by which the mesoderm is sequentially divided into segmental units called somites that later give rise to vertebrae, ribs, body musculature, and dermis. Molecular evidence for the segmentation clock was first uncovered with the characterization of the c-hairy gene, a chick homolog of the Drosophila pair rule gene hairy (Palmeirim et al., 1997). c-hairy encodes a member of the Hairy/Enhancer of split-related (Hes)/Hes-related (her) family of helix-loop-helix transcription factors that oscillate through a negative feedback loop in which the Hes/Her protein inhibits its own transcription (Bessho et al., 2003; Chen et al., 2005; Hirata et al., 2002; Lewis, 2003). Homologs of hairy have since been identified in vertebrate models such as mouse, fish, frog, and snake (Bessho et al., 2001; Gomez et al., 2008; Holley et al., 2000; Li et al., 2003), and in each species, members of the Hes/her family undergo oscillatory expression in the presomitic mesoderm (PSM). Genetic studies have shown that the mouse hairy1 homolog Hes7 and the zebrafish hairy1 homologs her1 and her7 are core segmentation genes that cycle in the PSM of each species with the same periodicity of segment formation (Bessho et al., 2001; Delaune et al., 2012; Gajewski et al., 2003; Harima et al., 2013; Henry et al., 2002; Hirata et al., 2002; Holley et al., 2000; Holley et al., 2002; Oates and Ho, 2002; Shih et al., 2015; Takashima et al., 2011; Takke and Campos-Ortega, 1999; Williams et al., 2016). Maintenance of oscillation periodicity requires many levels of regulation as a somite pair develops rapidly in developing embryos (30 minutes in zebrafish, 120 minutes in mouse) (Kageyama et al., 2012; Oates et al., 2012). Studies have explored transcriptional activation and the effect of negative feedback inhibition on oscillatory expression (Bessho et al., 2003; Giudicelli et al., 2007; Gonzalez et al., 2013; Hirata et al., 2002; Lewis, 2003; Schwendinger-Schreck et al., 2014), and have emphasized the importance of post-transcriptional regulation in maintaining proper oscillatory periodicity (Cibois et al., 2010; Fujimuro et al., 2014; Hanisch et al., 2013; Nitanda et al., 2014). For example, splicing (Harima et al., 2013; Takashima et al., 2011) and mRNA export (Hoyle and Ish-Horowicz, 2013) are rate-limiting steps of oscillatory expression. Additionally, cyclic gene transcript 3’UTRs can promote rapid decay of oscillating transcripts (Delaune et al., 2012; Fujimuro et al., 2014; Gallagher et al., 2017; Giudicelli et al., 2007). miRNAs regulate decay of some cyclic gene transcripts (Bonev et al., 2012; Riley et al., 2013; Tan et al., 2012; Wong et al., 2015), but not others (Gallagher et al., 2017; Zhang et al., 2011). Thus, specific mechanisms that govern cyclic gene transcript turnover are still not well-understood.

In a forward genetic screen, we discovered a zebrafish mutant, tortuga (tor), with disrupted cyclic gene expression (Dill and Amacher, 2005). Over 20 genes are deleted in the tor deficiency allele, but loss of a single gene in the deletion interval, pnrc2, is responsible for defects in cyclic gene expression (Gallagher et al., 2017). Recent work in cell culture systems has shown that human PNRC2 interacts with factors such as UPF1, DCP1A, and STAU1, revealing that Pnrc2 can function as a decay adaptor in nonsense-mediated mRNA decay (NMD) and STAU1-mediated mRNA decay (SMD) (Cho et al., 2013a; Cho et al., 2013b; Cho et al., 2012; Cho et al., 2009; Cho et al., 2015; Lai et al., 2012; Mugridge et al., 2016). Tethering of PNRC2 to reporter mRNAs confers instability (Cho et al., 2009; Lai et al., 2012; Nicholson et al., 2018), suggesting that PNRC2 can recruit decay machinery when directly associated with transcripts. However, recent work reported that destabilization of reporters containing NMD-inducing premature termination codons (PTCs) in HeLa cells is unaffected by PNRC2 knockdown, suggesting that PNRC2 is not necessary for NMD-induced decay in all cases (Nicholson et al., 2018).

We have shown that pnrc2 functions in clearance of cyclic gene transcripts such as her1, her7, and deltaC (dlc) during zebrafish segmentation (Gallagher et al., 2017). By in situ hybridization, cyclic gene transcript expression appears striped in the anterior PSM of wild-type embryos due to rapid oscillatory transcription followed by rapid mRNA decay. In pnrc2^oz22 mutants, the accumulation of cyclic gene transcripts obscures this dynamic striped expression (Gallagher et al., 2017). We also previously showed that pnrc2 is maternally-provided and zygotically-expressed throughout somitogenesis (Gallagher et al., 2017) and show here that maternally-provided pnrc2 partially compensates for zygotic pnrc2 in regulating cyclic gene transcript turnover during somitogenesis. Using a series of inducible transgenic reporter lines driving expression of a reporter mRNA fused to various portions of the her1 3’UTR, we find that the last 179 nucleotides (nts) of the 725 nt her1 3’UTR is necessary and sufficient to confer Pnrc2-dependent instability to reporter transcripts. We find the dlc 3’UTR also confers Pnrc2-dependent instability, demonstrating 3’UTR instability elements are present in both cyclic gene transcripts. Motif analysis comparing the last 179 nts of the her1 3’UTR to the dlc 3’UTR uncovered two potential cis-regulatory motifs: a Pumilio response element (PRE) and an AU-rich element (ARE). Mutation of the PRE or ARE motif partially disrupts the destabilizing effect of the her1 3’UTR on reporter mRNA, and mutation of the PRE and ARE severely disrupts the destabilizing effect, suggesting that the PRE and ARE both contribute to rapid turnover of endogenous her1 transcripts. Finally, we show that mutation of Pnrc2 residues and domains that are known to facilitate interaction of human PNRC2 with decay factors DCP1A and UPF1 eliminate or severely reduce the ability of Pnrc2 to restore normal cyclic gene expression in pnrc2 mutant embryos. Together, this work suggests that Pumilio proteins, ARE-binding proteins (ARE-BPs), and/or Pnrc2 interact with decay machinery components to regulate 3’UTR-mediated turnover of cyclic gene transcripts during vertebrate segmentation.

METHODS

Animal stocks and husbandry

Adult zebrafish strains (Danio rerio) were kept at 28.5°C on a 14 hour (h) light/10h dark cycle and obtained by natural spawning or in vitro fertilization, and were staged according to Kimmel et al (1995). The pnrc2^oz22 line has been described previously (Gallagher et al., 2017). The stable transgenic reporter lines generated and analyzed in this study are: Tg(hsp70l:Venus-her1 3’UTR-SV40 pA)oz44, oz45, oz46; Tg(hsp70l:Venus-her1 3’UTRΔ1–362-SV40 pA)oz47, oz48, oz50; Tg(hsp70l:Venus-her1 3’UTRΔ363–725-SV40 pA)oz51, oz54, oz96; Tg(hsp70l:Venus-her1 3’UTRΔ1–546-SV40 pA)oz55, oz57, oz58; Tg(hsp70l:Venus-her1 3’UTRΔ1–362;Δ547–725-SV40 pA)oz60, oz61; Tg(hsp70l:Venus-disrupted PRE her1 3’UTR-SV40 pA)oz69, oz70, oz71, oz72, oz73; Tg(hsp70l:Venus-disrupted ARE her1 3’UTR-SV40 pA)oz74, oz75, oz77, oz78, oz79; Tg(hsp70l:Venus-SV40 pA)oz64, oz65, oz66, oz67, oz68; Tg(hsp70l:Venus-dlc 3’UTR-SV40 pA)oz80, oz81, oz83; and Tg(hsp70l:Venus-disrupted PRE & ARE her1 3’UTR-SV40 pA)oz93, oz94, oz95. To control for potential locus-specific effects on transgene expression, all of the above independent lines were analyzed by in situ hybridization and showed consistent Venus reporter decay across all heat shock experiments except for Tg(hsp70l:Venus-disrupted ARE her1 3’UTR-SV40 pA)oz79 which had abnormally high Venus induction levels compared to the three other lines and was therefore excluded from the analysis. For the reporters hsp70l:Venus-her1 3’UTR-SV40 pA, hsp70l:Venus-disrupted PRE her1 3’UTR-SV40 pA, hsp70l:Venus-disrupted ARE her1 3’UTR-SV40 pA, and hsp70l:Venus-disrupted PRE & ARE her1 3’UTR-SV40 pA, three independent lines per reporter were analyzed by qPCR across three biological replicates and exhibited comparable Venus decay dynamics for each reporter (Figs. S2–S5). Animal experiments were performed according to institutional and national guidelines and regulations and were approved by the Ohio State University Animal Care and Use Committee.

DNA extraction and pnrc2^oz22 and Venus genotyping

Individual embryos and adult fin tissue were lysed in 50 ul 1X ThermoPol Buffer (NEB) at 95°C for 10 minutes, digested at 55°C for 1–4 hours using 25–50 ug Proteinase K (Thermo Fisher), followed by Proteinase K inactivation at 95°C for 10 minutes. 1 ul of DNA extract was used as template in a standard 25 ul PCR with Taq polymerase according to manufacturer’s protocol (NEB). To molecularly identify pnrc2^oz22 carriers after PCR amplification, samples were digested with 20 units NsiI-HF (NEB) to distinguish cleavable wild-type from un-cleavable mutant amplicons. Reaction products were analyzed on a 2% agarose gel stained with Gel Red (Biotium). To identify carriers of the heat shock inducible reporter transgenes, embryos were either screened post-heat shock (pHS) for Venus fluorescence or molecularly identified by PCR amplification of Venus coding sequence. Genotyping was performed with 1 ul of DNA extract as template in a standard 25 ul reaction with Taq polymerase according to manufacturer’s protocol (NEB). Primers were designed to amplify presence of Venus coding sequence (Table S2) and reaction products were analyzed on a 2% agarose gel stained with Gel Red (Biotium).

Plasmid construction and Transgenesis

The heat shock reporter construct hsp70l:Venus-her1 3’UTR-SV40 pA was assembled using standard restriction digestion-based cloning and replacement of the 1.1 kilobase (kb) her1 3’ noncoding sequence present in construct hsp70l:Venus-her1 3’UTR (Gallagher et al., 2017) with a synthetic 725 nt her1 3’UTR sequence directly fused to an SV40 polyadenylation (pA) sequence synthesized by GeneArt^® Gene Synthesis (Thermo Fisher). Derivative her1 3’UTR constructs, were generated by restriction digestion or PCR amplification of the hsp70l:Venus-her1 3’UTR-SV40 pA plasmid, in parallel with removal of the full-length her1 3’UTR from the hsp70l:Venus-her1 3’UTR-SV40 pA plasmid by restriction digestion, followed by ligation of the truncated her1 3’UTR sequence into the digested hsp70l:Venus-her1 3’UTR-SV40 pA plasmid. Modification of either the her1 3’UTR PRE or ARE sequence was performed by site-directed mutagenesis of the hsp70l:Venus-her1 3’UTR-SV40 pA reporter using KOD polymerase (EMD Millipore) with mutagenic primers (Table S2), followed by DpnI digestion and transformation into E. coli. To generate the construct with disruptions in both the PRE and ARE sequences, we performed site-directed mutagenesis of the reporter hsp70l:Venus-her1 3’ UTR with disrupted ARE-SV40 pA using KOD polymerase (EMD Millipore) with primers that mutate the PRE sequence without affecting the mutated ARE sequence (Table S2). The full-length 1327 nt dlc 3’UTR was cloned by extracting total RNA from wild-type embryos at mid-segmentation with Trizol Reagent according to manufacturer’s protocol (Thermo Fisher Scientific), followed by reverse-transcription with Superscript III (Thermo Fisher) using a dlc-specific reverse primer (Table S2) followed by PCR amplification of the dlc 3’UTR using gene-specific primers containing restriction enzyme sites for cloning. In parallel, construct hsp70l:Venus-her1 3’UTR-SV40 pA was digested to remove the her1 3’UTR followed by replacement with the dlc 3’UTR. All constructs were sequence confirmed. Transgenic lines were generated as previously described using I-SceI-based transgenesis (Thermes et al., 2002).

Heat shock assay

Adult fish carrying the stable reporter transgenes were crossed to AB wild-type fish. Progeny were raised to mid-segmentation, heat shocked at 37°C for 15 minutes, and fixed in 4% PFA at defined intervals post-heat shock. All transgenic lines were analyzed for Venus expression post-heat shock by colorimetric in situ hybridization or qPCR analysis in parallel (each method described below).

In situ hybridization

Whole mount in situ hybridization was performed as previously described (Broadbent and Read, 1999; Jowett, 1998) using DIG-labeled antisense probes. Riboprobes for her1, dlc, and Venus were made as previously described (Delaune et al., 2012; Dill and Amacher, 2005).

RNA extraction

Whole embryos at mid-segmentation (n=10 per time point or condition) were solubilized in Trizol for RNA extraction and purified following standard procedures (Thermo Fisher). 0.5 –1 ug total RNA was reverse transcribed using random primers or gene-specific reverse primers and Superscript III reverse transcriptase (RT) according to the manufacturer’s instructions (Thermo Fisher).

Quantitative PCR and half-life calculations

Quantitative PCR was performed using PowerUp SYBR Green Master Mix (Thermo Fisher) and 4.5 ul cDNA (diluted 1:50) in 10 ul reactions, following manufacturer’s procedures. Negative controls lacking template were included for each primer set. All reactions were subjected to thermal melting to confirm that each reaction gave single peaks. For each transgenic line and time point, three biological replicates were performed, and transcript levels were normalized to mobk13 (mob4) (Hu et al., 2016; Gangras et al, 2019). Cycle thresholds (C_t) were determined using Bio-Rad CFX manager software. Changes in mRNA expression were calculated by ΔΔC_t = ΔC_{t target} − ΔC_{t control}. Relative changes in mRNA expression levels are represented graphically as fold change, where relative mRNA fold change = 2^−ΔΔCt. All graphs were generated using Prism 8.1 (GraphPad). For half-life calculations, 30 minutes pHS was set as “time 0” because our experiments revealed that heat shock induction continues temporarily after the downshift from 37ºC to 28.5ºC. To determine half-lives, we fitted exponential decay equations to normalized Venus mRNA levels across time for each heat shock experiment in the form y = ae^−(bx), with y representing fold change, a representing fold change at the y intercept, b representing a decay constant, and x representing time. We determined the value for x when y = 0.5 using the online algebraic tool MathPapa at http://www.mathpapa.com. Three biological replicates per reporter line were used to calculate average half-lives and standard deviation.

Polyadenylation site determination

Polyadenylation (pA) site use for each reporter was determined using 3’-rapid amplification of cDNA ends. Briefly, 0.5 ug total RNA was reverse transcribed using an oligo-dT adapter primer and Superscript III reverse transcriptase (Thermo Fisher), followed by amplification using Taq polymerase (NEB) and a Venus-specific forward primer and universal reverse adapter primer (Table S2). Amplified products were gel purified and TOPO cloned according to manufacturer’s instructions (Thermo Fisher). pA site use was determined by plasmid restriction digestion with BamHI-HF (New England Biolabs) which generates an additional cleaved product of unique size when the SV40 pA rather than the endogenous her1 pA site is used. A subset of clones utilizing the SV40 and her1 pA sites were sequence analyzed to confirm that the digestion strategy accurately distinguishes between clones derived from use of the her1 or SV40 pA sites.

Microscopy and Imaging

In situ hybridized embryos were mounted in Permount and imaged using an Axiocam HRc digital camera with AxioPlan2 microscope (Zeiss). Immunofluorescent embryos were mounted in 80% glycerol and imaged at 20x magnification using MetaMorph software (Molecular Devices) on an Andor™ SpinningDisc Confocal Microscope (Oxford Instruments) with a Nikon Neo camera. Laser wavelength and intensity were set at 488 nm and 50%, respectively, and bit depth at 16-bit. Maximum intensity projections are shown (Fig. S1).

Immunohistochemistry

Standard immunohistochemistry protocols were followed using 4% PFA fixation, dehydration and rehydration in a methanol series, and incubation in blocking solution for 1 hour. Mid-segmentation embryos from wild-type and pnrc2^oz22 crosses were incubated in 2% BSA/5% goat serum/0.1% Tween-20/PBS blocking solution containing 1:200 anti-zdc2 that recognizes Dlc protein (ab73336, Abcam) according to previously published methods (Giudicelli et al., 2007), followed by goat anti-mouse Alexa-Fluor-488 (1:800) (Thermo Fisher).

Plasmid construction and mRNA injection

Full-length pnrc2 cDNA was generated as previously described (Gallagher et al., 2017) to create plasmid SP6-pnrc2-cDNA (pTLG109). The Cerulean coding sequence was translationally fused to pnrc2 by overlap extension PCR (Horton et al., 2013) using Phusion polymerase (NEB), pTLG109 and pCS-H2B-Cerulean (Megason, 2009) as template in separate reactions, and primers that incorporate a flexible linker between Cerulean and pnrc2 coding sequences. Stitching of fragments was performed using Phusion polymerase (NEB), Cerulean and pnrc2 overlap fragments as templates, and outer primers complementary to the 5’ and 3’ ends of the Cerulean and pnrc2 coding sequences, respectively, containing restriction sites for subcloning into expression vector pCS2+ (Rupp et al., 1994; Turner and Weintraub, 1994) to generate SP6-Cerulean-pnrc2-cDNA (pTLG149). To generate mutant pnrc2 versions, site-directed mutagenesis (Hutchison et al., 1978) of pTLG109 was performed using KOD polymerase (EMD Millipore) and mutagenic primers, followed by DpnI (New England Biolabs) digestion and transformation into E. coli, creating SP6-pnrc2^F138→stop-cDNA (pTLG137), SP6-pnrc2 ^K119→A-cDNA (pTLG138), and SP6-pnrc2 ^W124→A-cDNA (pTLG139). Mutant pnrc2 versions from pTLG137–139 were subcloned into pTLG149 to replace the wild-type pnrc2 coding sequence, creating SP6-Cerulean-pnrc2^F138→stop-cDNA (pTLG151), SP6-Cerulean-pnrc2 ^K119→A-cDNA (pTLG152), and SP6-Cerulean-pnrc2 ^W124→A-cDNA (pTLG153). All plasmids were sequence validated. Primer sequences are listed in Table S2.

For rescue experiments, wild-type and mutant Cerulean-pnrc2 mRNAs were synthesized using the SP6 mMessage Machine Kit (Thermo Fisher), diluted in 0.2M KCl with 0.1% phenol red, and injected into 1-cell stage embryos (40 pg mRNA per embryo). To determine the minimal dose required for consistent rescue, we performed dose response experiments with wild-type Cerulean-pnrc2 mRNA and found that 40–100 pg doses rescued almost all MZpnrc2 mutants, whereas doses between 5–20 pg doses were less penetrant.

RESULTS

Zygotically-expressed and maternally-provided pnrc2 promotes cyclic gene transcript decay

We previously showed that zygotically-expressed pnrc2 promotes decay of cyclic gene transcripts, including her1 and dlc, during somitogenesis (Gallagher et al., 2017). Because pnrc2 transcript is detected in wild-type (WT) embryos at early time points including the 8-cell stage (Gallagher et al., 2017), we hypothesized that maternally-provided pnrc2 transcript and/or Pnrc2 protein may partially compensate for loss of zygotic pnrc2 function during somitogenesis. To assess maternal contribution, we examined cyclic gene transcript expression in maternal-zygotic pnrc2^oz22 (MZpnrc2) mutant embryos that lack both maternal and zygotic pnrc2 function, maternal pnrc2^oz22 (Mpnrc2) mutant embryos that have only zygotic pnrc2 function, and zygotic pnrc2^oz22 (Zpnrc2) mutant embryos that have only maternal pnrc2 function (Fig. 1). As shown previously (Gallagher et al., 2017), her1 and dlc transcripts are misexpressed in Zpnrc2 embryos (Fig. 1A vs B, E vs F). Although her1 and dlc expression appear normal in Mpnrc2 embryos (Fig. 1A vs C, E vs G), MZpnrc2 embryos show an enhanced phenotype compared to Zpnrc2 embryos (Fig. 1B vs D, F vs H). To confirm that Pnrc2 functions post-transcriptionally to negatively regulate cyclic gene expression, we quantitatively assessed her1 and dlc transcript levels in Mpnrc2 and MZpnrc2 mutant and wild-type sibling embryos using quantitative PCR (qPCR). To distinguish unspliced from spliced transcripts, we used primers that selectively amplify one or the other form of transcript, with the expectation that only spliced transcripts would be elevated in MZpnrc2 mutants. Indeed, spliced her1 and dlc transcript levels are elevated almost 4-fold in MZpnrc2 mutant embryos compared to Mpnrc2 mutant and wild-type sibling embryos (Fig. 1I and J). In contrast, unspliced her1 and dlc transcripts are down-regulated ~2-fold in MZpnrc2 compared to wild-type embryos (Fig. 1K and L), suggesting that cyclic expression might be transcriptionally reduced despite accumulation of cyclic gene transcripts or that splicing efficiency is increased in pnrc2-deficient embryos. Surprisingly, Mpnrc2 mutant embryos also show reduced unspliced cyclic gene transcript levels compared to wild-type embryos despite normal levels of spliced cyclic gene transcript, demonstrating that post-transcriptional accumulation of cyclic gene transcripts in MZpnrc2 mutant embryos is unrelated to transcriptional downregulation. Together, these data strongly support that Pnrc2 regulates cyclic gene expression at the post-transcriptional level, and extend our previous qualitative work (Dill and Amacher, 2005; Gallagher et al., 2017). Overall, these results suggest maternal and zygotic pnrc2 function promotes rapid turnover of her1 and dlc transcripts during somitogenesis.

Figure 1. Maternal and zygotic pnrc2 promotes proper her1 and dlc expression.

Wild-type (WT), zygotic pnrc2^oz22 (Zpnrc2), maternal pnrc2^oz22 (Mpnrc2), and maternal-zygotic pnrc2^oz22 (MZpnrc2) mutant embryos were raised to mid-segmentation stage (16–18 hpf) and probed for her1 (A–D) and dlc expression (E–F) by in situ hybridization (n ≥ 7 each). WT, Mpnrc2 mutant, and MZpnrc2 mutant embryos (n = 10 per biological replicate) were analyzed by qPCR using primers to amplify across exon-exon boundaries to detect spliced her1 (I) and dlc (J) transcripts or primers to amplify across intron-exon boundaries to detect her1 (K) and dlc (L) unspliced transcripts. MZpnrc2 mutant embryos have ~4-fold higher levels of spliced her1 and dlc mRNA than wild-type or Mpnrc2 mutant embryos, which have comparable levels (I and J). Both MZpnrc2 and Mpnrc2 mutant embryos have ~2-fold less unspliced her1 and dlc transcripts compared to WT embryos (K and L). hpf = hours post-fertilization.

Unlike cyclic gene transcript, cyclic gene protein does not accumulate in MZpnrc2 mutant embryos

We previously showed that expression of the cyclic gene protein Dlc is unaffected in zygotic pnrc2 mutants (Gallagher et al., 2017). To assess whether the same is true for MZpnrc2 mutants, we examined endogenous Dlc expression by immunofluorescence. While dlc transcript levels are almost 4-fold higher in MZpnrc2 mutants than in wild-type embryos (Fig. 1J), there is no obvious corresponding increase in Dlc protein expression (Fig. S1). Thus, despite accumulation of cyclic gene transcripts in MZpnrc2 mutants, cyclic gene protein expression appears normal, which may explain why mutants lack a morphological segmentation phenotype.

The terminal 179 nucleotides of the her1 3’UTR are necessary and sufficient for Pnrc2-mediated decay of reporter transcripts

Previous studies have shown that reporter transcripts containing the her1 3′UTR are rapidly degraded (Gallagher et al., 2017; Giudicelli et al., 2007), and that rapid decay requires Pnrc2 function (Gallagher et al., 2017). To identify regulatory elements within the her1 3’UTR that are required for Pnrc2-mediated decay, we modified our transgenic heat shock-inducible reporter (Gallagher et al., 2017), generated stable transgenic lines driving expression of Venus transcript followed by various portions of the her1 3’UTR, and performed heat shock induction assays (Fig. 2A). Because a subset of reporter lines lacks the endogenous her1 polyadenylation (pA) signal, we included an SV40 pA signal on all reporters, including the full-length her1 3’UTR, to ensure that all reporters contained at least one bona fide pA signal. As a control for Venus transcript stability, a transgenic line driving inducible expression of Venus transcript with an SV40 pA signal (Venus-SV40 pA) was tested in parallel. Reporter mRNA containing the full-length her1 3’UTR (Venus-her1 3’UTR-SV40 pA) is rapidly decayed between 30 to 90 minutes post-heat shock (pHS) when compared to reporter mRNA lacking the her1 3’UTR sequence (Venus-SV40 pA) by in situ hybridization (Fig. 2B–I), confirming that the full-length her1 3’UTR is sufficient to confer rapid instability. Consistent with in situ hybridization data, qPCR analysis reveals that reporter half-life between 30 to 90 minutes pHS is >3-fold lower when the her1 3’UTR is present (Fig. 2R) and that this effect is consistent across three independent lines (Fig. S2). To test whether Pnrc2 is required for rapid reporter decay, we performed the same heat shock experiments in Mpnrc2 and MZpnrc2 mutant embryos carrying transgenic reporters. Wild-type and Mpnrc2 mutant embryos expressing Venus-her1 3’UTR-SV40 pA reporter mRNA show comparable expression patterns across all pHS time points (Fig. 2F–M). In contrast, reporter mRNA in MZpnrc2 mutant embryos perdures (Fig. 2N–Q), indicating that Pnrc2-mediated decay of her1 transcripts occurs through destabilizing features of the her1 3’UTR.

Figure 2. The her1 3’UTR confers Pnrc2-mediated instability to reporter transcripts.

(A) Diagram illustrating the heat shock protocol used for transgenic lines in this study. (B–I) Transgenic embryos carrying the hsp70l:Venus-her1 3’UTR-SV40 pA reporter (line oz44) or hsp70l:Venus-SV40 pA reporter (line oz68) were raised to mid-segmentation stage, heat shocked for 15 minutes, then collected at the indicated minutes pHS and processed by Venus in situ hybridization (n ≥ 7 embryos per time point). Venus transcript is not detected in the absence of heat shock (n = 10 per reporter line) (data not shown). (J–Q) Mid-segmentation stage Mpnrc2 mutant embryos and MZpnrc2 mutant embryos carrying the hsp70l:Venus-her1 3’UTR-SV40 pA reporter (line oz44) were heat shocked and processed by Venus in situ hybridization as above (n ≥ 8 embryos per time point). Representative embryos were genotyped post-imaging to confirm genotype. (R) qPCR analysis comparing Venus transcript fold change from 30 minutes pHS to 60 and 90 minutes pHS for the Tg(hsp70l:Venus-her1 3’UTR-SV40 pA)oz44 and Tg(hsp70l:Venus-SV40 pA)oz68 reporter lines (n = 10 embryos per time point across three biological replicates). Three independent lines carrying the hsp70l:Venus-her1 3’UTR-SV40 pA reporter and five independent lines carrying the hsp70l:Venus-SV40 pA reporter were analyzed in wild-type embryos by in situ hybridization and exhibited comparable Venus decay across all lines carrying the same reporter (data not shown); one representative line for each is shown (see Methods for details). For the hsp70l:Venus-her1 3’UTR-SV40 pA reporter, three independent lines were analyzed by qPCR and exhibited comparable decay (Fig. S2). pHS = post-heat shock; hpf = hours post-fertilization; t_1/2 = half-life; ± = standard deviation; pA = polyadenylation sequence.

To define regions of the her1 3’UTR that are necessary and sufficient for Pnrc2-dependent decay, we conducted a deletion analysis of the her1 3’UTR. Transgenic lines were generated with hsp70l promoter sequence driving expression of Venus coding sequence fused to the first 362 nts or the last 363 nts of the her1 3’UTR followed by an SV40 pA. Reporter induction was performed and analyzed as in Fig. 2A. By in situ hybridization, reporter transcripts containing the first 362 nts of the her1 3’UTR display minimal destabilization by 90 minutes pHS, whereas reporter transcripts containing the last 363 nts of the her1 3’UTR are rapidly degraded (Fig. 3A–F). Correspondingly, qPCR analysis reveals that the reporter containing the terminal half of the her1 3’UTR has a short half-life that is very similar to that of the reporter containing the full-length her1 3’UTR and that the reporter containing the first half of the her1 3’UTR is >6-fold more stable (Figs. 2R and 3G). These data indicate that the terminal half of the her1 3’UTR is necessary and sufficient to confer instability to reporter transcripts.

Figure 3. The terminal 179 nucleotides of the her1 3’UTR is sufficient for Pnrc2-mediated decay of reporter transcripts.

(A–F) Transgenic embryos carrying the hsp70l:Venus-her1 3’UTRΔ363–725-SV40 pA reporter (line oz54) or hsp70l:Venus-her1 3’UTRΔ1–362-SV40 pA reporter (line oz47) were raised to mid-segmentation stage and heat shocked for 15 minutes, then collected at the indicated minutes pHS and processed by Venus in situ hybridization (n ≥ 6 per time point). Venus transcript is not detected in the absence of heat shock (n = 10 per reporter line) (data not shown). (G) qPCR analysis comparing Venus transcript fold change from 30 minutes pHS to 60 and 90 minutes pHS for the Tg(hsp70l:Venus-her1 3’UTRΔ363–725-SV40 pA)oz54 or Tg(hsp70l:Venus-her1 3’UTRΔ1–362-SV40 pA)oz47 reporter lines (n = 10 embryos per time point across three biological replicates). (H–M) Mid-segmentation stage wild-type (WT) and MZpnrc2 mutant embryos carrying the hsp70l:Venus-her1 3’UTRΔ1–546-SV40 pA reporter (line oz50) were heat shocked and processed by Venus in situ hybridization (n ≥ 7 embryos per time point). Venus transcript is not detected in the absence of heat shock (n = 10 wild-type embryos) (data not shown). Representative embryos were genotyped post-imaging to confirm genotype. For each reporter, three independent lines were analyzed in wild-type embryos by in situ hybridization and exhibited comparable Venus decay across all lines carrying the same reporter (data not shown); one representative line for each is shown (see Methods for details). pHS = post-heat shock; t_1/2 = half-life; ± = standard deviation; pA = polyadenylation sequence.

To elucidate specific destabilizing features of the her1 3’UTR, we generated heat shock inducible transgenic lines containing the last quarter of the her1 3’UTR. Reporter transcripts containing the last 179 nts of the her1 3’UTR are rapidly degraded, similar to reporters containing the full-length or the last half of the her1 3’UTR (Fig. 2F–I and Fig. 3D–F vs Fig. 3H–J). Collectively, all of our her1 3’UTR deletion reporters indicated that instability elements are located near the end of the her1 3’UTR (Fig. 4). To investigate whether the terminal 179 nt her1 3’UTR is Pnrc2-dependent, we examined reporter decay in MZpnrc2 mutant embryos and observed that the reporter persists longer than in wild-type embryos (Fig. 3H–M) or Mpnrc2 mutant controls (data not shown). Even though rapid decay of the reporter containing the terminal 179 nts is similar to that containing the full-length her1 3’UTR (Figs. 2F–I and 3H–J), the two reporters are differentially affected upon loss of Pnrc2 function (Figs. 2N–Q and 3K–M). Loss of Pnrc2 function stabilizes full-length reporter transcripts to a greater extent than that observed for reporters containing the terminal 179 nts, suggesting that additional pnrc2-dependent destabilizing elements lie upstream of the terminal 179 nts. Because the terminal 179 nts are necessary and sufficient to trigger reporter decay in wild-type embryos (Fig. 4), such upstream pnrc2-dependent elements could only elicit decay when the terminal 179-nt region is present.

Figure 4. Transgenic reporters reveal that the terminal her1 3’UTR is necessary and sufficient to confer transcript destabilization.

Summary of reporter destabilization in transgenic embryos carrying various derivatives of the her1 3’UTR (A–E). All lines were raised to mid-segmentation stage and heat shocked for 15 minutes, then collected and processed by Venus in situ hybridization at 0, 30, 60, and 90 minutes pHS. (A) hsp70l:Venus-her1 3’UTR-SV40 pA reporter line (see Figs. 2 and 6). (B) hsp70l:Venus-her1 3’UTRΔ1–362-SV40 pA reporter line (see Fig. 3). (C) hsp70l:Venus-her1 3’UTRΔ363–725-SV40 pA reporter line (see Fig. 3). (D) hsp70l:Venus-her1 3’UTRΔ1–546-SV40 pA reporter line (see Fig. 3). (E) hsp70l:Venus-her1 3’UTRΔ1–362; Δ547–725-SV40 pA reporter line (data not shown). Three independent lines were analyzed in wild-type embryos by in situ hybridization for each reporter, except for the hsp70l:Venus-her1 3’UTRΔ1–362; Δ547–725-SV40 pA reporter (E) for which two independent lines were analyzed. Each reporter exhibited comparable Venus decay across all independent lines (data not shown); see Methods for details. pHS = post-heat shock.

An alternative possibility is that Pnrc2-independent stabilizing features are present in the first 556 nts of the her1 3’UTR that are suppressed or overcome by a Pnrc2-dependent destabilizing element in the terminal 179 nts. The activity of such stabilizing features could only be unmasked when Pnrc2 is absent, leading to greater stabilization of the full-length versus the terminal 179 nt reporter in MZpnrc2 mutant embryos.

The dlc 3’UTR confers instability to reporter transcripts in a Pnrc2-dependent manner

Pnrc2 function is important for proper expression of cyclic genes her1 and dlc (Dill and Amacher, 2005; Gallagher et al., 2017). To determine if the dlc 3’UTR, like the her1 3’UTR, contains features that promote Pnrc2-dependent decay, we used our heat shock-inducible system to drive expression of Venus transcripts containing the full-length 1327 nt dlc 3’UTR and SV40 pA sequence (Venus-dlc 3’UTR-SV40 pA) and compared reporter decay among wild-type, Mpnrc2 mutant, and MZpnrc2 mutant embryos at 0, 30, and 60 minutes pHS (Fig. 5). We find that reporters in wild-type and Mpnrc2 mutant embryos are rapidly decayed (Fig. 5A–C vs D–F). In contrast, reporter expression in MZpnrc2 mutants is stabilized relative to wild-type and Mpnrc2 mutant embryos (Fig. 5A–F vs G–I). These data reveal that the dlc 3’UTR promotes Pnrc2-dependent transcript decay and suggests cyclic gene transcripts may share common Pnrc2-dependent 3’UTR decay features.

Figure 5. The dlc 3’UTR confers Pnrc2-mediated decay to reporter transcripts.

(A–C) Transgenic embryos carrying the hsp70l:Venus-dlc 3’UTR-SV40 pA reporter (line oz81) were raised to mid-segmentation stage and heat shocked for 15 minutes, then collected at the indicated minutes pHS and processed by Venus in situ hybridization (n ≥ 6 embryos per time point). Venus transcript is not detected in the absence of heat shock (n = 10 embryos) (data not shown). (D–I) Mpnrc2 (D–F) and MZpnrc2 mutant embryos (G–I) carrying the hsp70l:Venus-dlc 3’UTR-SV40 pA reporter (line oz81) were also heat shocked and processed for Venus transcript (n ≥ 10 embryos per time point). Representative embryos were genotyped post-imaging to confirm genotype. (J) RBPmap motif analysis (Paz et al., 2014) identifies a canonical PRE (5’UGUAAAUA; yellow) and a canonical ARE (5’UAUUUAU; white) near the end of the her1 3’UTR and two PREs and one ARE near the end of the dlc 3’UTR. The indicated PRE and ARE are the only such motifs in the 725 nt full-length her1 3’UTR, whereas the 1327 nt dlc 3’UTR contains an additional PRE and two additional AREs. Three independent lines carrying the hsp70l:Venus-dlc 3’UTR-SV40 pA reporter were analyzed in wild-type embryos by in situ hybridization and exhibited comparable Venus decay across all three lines (data not shown); one representative line is shown (see Methods for details). pHS = post-heat shock; nts = nucleotides; PRE = Pumilio response element; ARE = AU-rich element; pA = polyadenylation sequence.

Pumilio response and AU-rich elements in the her1 3’UTR promote transcript decay

To determine if her1 and dlc 3’UTRs contain shared motifs, we compared the terminal 179 nt her1 3’UTR with the dlc 3’UTR using RBPmap motif analysis (Paz et al., 2014) and manual inspection. Both share two perfect matches to motifs that are not present in the first 75% (1–546 nts) of the her1 3’UTR: a Pumilio response element (PRE) and an AU-rich element (ARE) (Fig. 5J). The terminal her1 3’UTR contains a single 5’UGUAHAUA PRE and a single 5’UAUUUAU ARE while the full-length dlc 3’UTR contains three 5’UGUAHAUA PREs and three 5’UAUUUAU AREs. Because PREs and AREs are associated with mRNA instability (Goldstrohm et al., 2018; Schoenberg and Maquat, 2012), we determined whether they contribute to cyclic gene transcript turnover by introducing mutations within the context of the full-length her1 3’UTR reporter and comparing transcript stability to that of the unmodified reporter. The PRE was disrupted by changing PRE base positions 2 and 3 from GU to CC, a replacement that disrupts the ability of human PUMILIO proteins to bind target mRNAs (Miles et al., 2015). By in situ hybridization, reporter mRNA with the PRE mutation appears partially stabilized relative to unmodified reporter mRNA (Fig. 6A–F). Correspondingly, qPCR analysis reveals that the PRE mutation increases reporter half-life by ~1.6-fold when compared to the full-length, unmodified reporter (Fig. 6G) and that these results are consistent across three independent lines (Fig. S3). To determine if the her1 ARE promotes decay of reporter transcript, ARE positions 3–5 were changed from UUU to CCC, a replacement that disrupts ARE-binding protein association with target mRNAs (Lai et al., 2005). qPCR analysis reveals that the ARE mutation increases reporter half-life, although this affect is modest relative to the PRE mutation (Fig. 6G). Reporter stabilization due to ARE mutation is consistent across three independent lines (Fig. S4).

Figure 6. The Pumilio response and AU-rich elements in the her1 3’UTR contribute to reporter transcript turnover.

(A–F) Transgenic embryos carrying the hsp70l:Venus-her1 3’UTR-SV40 pA reporter (line oz44) or the hsp70l:Venus-her1 3’UTR with disrupted PRE-SV40 pA reporter (line oz71) with a 2 nt mutation in the PRE sequence were raised to mid-segmentation stage and heat shocked for 15 minutes, then collected at the indicated minutes pHS and processed by Venus in situ hybridization (n ≥ 11 embryos per time point). Venus transcript is not detected in the absence of heat shock (n = 10 per reporter line) (data not shown). (G) qPCR analysis comparing Venus transcript fold change from 30 minutes pHS to 60 and 90 minutes pHS for the reporter lines Tg(hsp70l:Venus-her1 3’UTR-SV40 pA)oz44, Tg(hsp70l:Venus-her1 3’UTR with disrupted PRE-SV40 pA)oz71, and Tg(hsp70l:Venus-her1 3’UTR with disrupted ARE-SV40 pA)oz75 (n = 10 embryos per time point across three biological replicates). The PRE mutation changes the 5’UGUAAAUA site to 5’UCCAAAUA and the ARE mutation changes the 5’UAUUUAU site to 5’UACCCAU. Both mutations extend reporter half-life; the PRE-mutated reporter half-life is increased 1.7-fold and the ARE-mutated reporter half-life is increased 1.6-fold. Five independent lines carrying the hsp70l:Venus-her1 3’UTR with disrupted PRE-SV40 pA reporter and four independent lines carrying the hsp70l:Venus-her1 3’UTR with disrupted ARE-SV40 pA reporter were analyzed in wild-type embryos by in situ hybridization and exhibited comparable Venus decay across all lines carrying the same reporter (data not shown); one representative line for each is shown (see Methods for details). For each reporter, three independent lines were chosen for qPCR analysis and exhibited comparable Venus decay (Figs. S3–S4). pHS = post-heat shock; PRE = Pumilio response element; ARE = AU-rich element; t_1/2 = half-life; ± = standard deviation; pA = polyadenylation sequence.

We next investigated the combinatorial activity of the PRE and ARE motifs on reporter decay by introducing both PRE and ARE mutations in the full-length her1 3’UTR reporter and comparing transcript stability to that of the unmodified reporter. By in situ hybridization, reporter mRNA with both PRE and ARE mutations appear dramatically stabilized (Fig. 7A–C). Correspondingly, qPCR analysis reveals that combined PRE and ARE mutations increase reporter half-life ~7-fold when compared to the full-length, unmodified reporter (Fig. 7D), and that these results are consistent across three independent lines (Fig. S5).To address the possibility that differences in pA site use between reporters might influence mRNA stability, we analyzed the 3’ ends of reporter transcripts and find that the unmodified and PRE-mutated her1 3’UTR reporters primarily use the SV40 pA, whereas reporters with disrupted ARE and disrupted PRE & ARE use either the SV40 or natural her1 pA with similar frequency (Table S1). Because reporters with a disrupted ARE motif exhibit a shift in pA site preference, we quantified Venus transcripts that use the SV40 pA and find no differences in decay when compared to quantification of all Venus transcripts regardless of pA site use for each reporter (Fig. S6). Taken together, our analysis of single- and double-mutated reporters suggests that the PRE and ARE in the her1 3’UTR both contribute to transcript decay and together these motifs dramatically destabilize reporter mRNA.

Figure 7. The Pumilio response and AU-rich elements in the her1 3’UTR are both required for rapid reporter transcript turnover.

(A–C) Transgenic embryos carrying the hsp70l:Venus-her1 3’UTR with disrupted PRE & ARE-SV40 pA reporter (line oz93) with a 2 nt mutation in the PRE sequence and 3 nt mutation in the ARE sequence were raised to mid-segmentation stage and heat shocked for 15 minutes, then collected at the indicated minutes pHS and processed by Venus in situ hybridization (n ≥ 11 embryos per time point). Venus transcript is not detected in the absence of heat shock (n = 10 embryos) (data not shown). (D) qPCR analysis comparing Venus transcript fold change from 30 minutes pHS to 60 and 90 minutes pHS for the reporter lines Tg(hsp70l:Venus-her1 3’UTR-SV40 pA)oz44 and Tg(hsp70l:Venus-her1 3’UTR with disrupted PRE & ARE-SV40 pA)oz93 (n = 10 embryos per time point across three biological replicates). The presence of both mutations extends reporter half-life by ~7-fold compared to the unmutated control. Three independent lines carrying the hsp70l:Venus-her1 3’UTR with disrupted PRE & ARE-SV40 pA reporter were analyzed by in situ hybridization and qPCR and each line exhibited comparable Venus decay dynamics (data not shown and Fig. S5); one representative line is shown (see Methods for details). pHS = post-heat shock; PRE = Pumilio response element; ARE = AU-rich element; t_1/2 = half-life; ± = standard deviation; pA = polyadenylation sequence.

Pnrc2 may promote cyclic gene transcript decay by interacting with known decay pathway components

While our analysis has identified cis-elements and potential trans-factors that promote cyclic gene transcript decay, it is still unclear how they interact with transcript decay machinery. Pnrc2 lacks an obvious RNA-binding domain, but does contain C-terminal SRC-Homology 3 (SH3) and Nuclear Receptor-box (NR-box) domains that are 100% identical to human PNRC2 (Gallagher et al., 2017). Human PNRC2 interacts with the mRNA decapping factor DCP1A through the SH3 domain and with the nonsense-mediated mRNA decay (NMD) effector UPF1 through the NR-box domain (Albers et al., 2005; Cho et al., 2009; Lai et al., 2012; Loh et al., 2013; Mugridge et al., 2016). Mutation of W114A in the SH3 domain abrogates PNRC2 binding to DCP1A and deletion of the NR-box abrogates PNRC2 binding to UPF1, and these interactions are required to elicit decay of reporter mRNA (Lai et al., 2012; Nicholson et al., 2018). We hypothesized the highly conserved SH3 and/or NR-box domains of Pnrc2 might also be essential for Pnrc2-mediated decay of cyclic gene transcripts. We injected mRNA encoding wild-type or mutant Pnrc2 translationally fused to Cerulean fluorescent protein into MZpnrc2 mutants and assayed rescue (Table 1). Whereas injection of Cer-pnrc2 mRNA restores proper her1 expression in MZpnrc2 mutants, injection of Cer-pnrc2 mRNA containing a W124A mutation (orthologous to the human W114A mutation) into MZpnrc2 mutants did not rescue her1 expression (Table 1). Cer-pnrc2 lacking the NR-box retains partial function and restores her1 expression in 63% of injected MZpnrc2 mutant embryos (Table 1). Overall, these results suggest that interaction of Pnrc2 with known decay machinery is necessary for turnover of cyclic gene transcripts.

Table 1.

Pnrc2 SH3 domain and NR-box are important for cyclic gene transcript decay

Condition	her1 expression a
	Normal b	Accumulated b	Percent affected
Uninjected	14/29	15/29	51.7%
Cerulean-pnrc2	48/49	¼9	2.0%
Cerulean-pnrc2ΔNR (c)	22/27	5/27	18.5%
Cerulean-pnrc2W->A (d)	26/50	24/50	52.0%

^aAt 16–18 hpf, pnrc2 mRNA-injected embryos from a cross of a pnrc2 homozygous female to a pnrc2 heterozygous male were processed by her1 in situ hybridization. The cross yields 50% Mpnrc2 mutants with normal her1 expression and 50% MZpnrc2 mutants with accumulated her1 expression unless there is phenotypic rescue by the injected pnrc2 mRNA.

^bChi-square analysis indicates a significant difference in her1 expression between Cerulean-pnrc2 mRNA-injected and uninjected MZpnrc2 mutant embryos (p < 0.0001) and between Cerulean-pnrc2^ΔNR mRNA-injected and uninjected MZpnrc2 mutant embryos (p = 0.001) and no significant difference between Cerulean-pnrc2^W->A mRNA-injected and uninjected MZpnrc2 mutant embryos (p = 0.7773).

^cΔNR = deletion of NR box

^dW->A = W124A mutation

DISCUSSION

In this work, we show that Pnrc2 promotes 3’UTR-mediated mRNA decay of cyclic gene transcripts her1 and dlc. This work builds upon previous studies demonstrating that Pnrc2 promotes decay of PTC-containing transcripts (Cho et al., 2013a; Cho et al., 2009; Lai et al., 2012) and non-PTC-containing transcripts (Gallagher et al., 2017; Nicholson et al., 2018) and identifies mRNA elements that promote Pnrc2-mediated decay. We show here that the terminal 179 nts of the her1 3’UTR are necessary and sufficient for Pnrc2-dependent decay of non-PTC-containing reporter mRNA in vivo. We also show that the full-length dlc 3’UTR confers Pnrc2-mediated decay to reporter transcripts, demonstrating that elements within cyclic gene transcript 3’UTRs are sufficient for Pnrc2-mediated decay. Both the her1 and dlc 3’UTRs contain at least one PRE and ARE and disruption of either or both motifs in the her1 3’UTR stabilizes reporter mRNA, raising the possibility that Pnrc2 promotes decay via Pumilio proteins and/or ARE-BPs. Human PNRC2 is implicated in mRNA turnover through interactions with decay factors SMG5, DCP1A, UPF1, and STAU1 (Cho et al., 2013a; Cho et al., 2013b; Cho et al., 2012; Cho et al., 2009; Cho et al., 2015; Lai et al., 2012; Mugridge et al., 2016; Nicholson et al., 2018), and our work suggests Pnrc2 interacts with Dcp1a and Upf1 to promote cyclic gene transcript turnover.

Maternal pnrc2 contributes to cyclic gene transcript turnover

Zebrafish utilize a large contribution of maternally-provided mRNA and protein (Tadros and Lipshitz, 2009). We show here that maternal deposition of pnrc2 partially compensates for loss of zygotic pnrc2 during somitogenesis. Embryos without maternal and zygotic pnrc2 function display enhanced accumulation of her1 and dlc transcripts compared to embryos lacking only zygotic pnrc2 function (Fig. 1). Interestingly, loss of maternal pnrc2 function alone has no consequence on her1 and dlc transcript levels, showing that zygotic pnrc2 expression is sufficient to clear cyclic gene transcripts during somitogenesis (Figs. 1, 2, and 5). Incomplete compensation of Zpnrc2 mutants by maternal pnrc2 function is likely caused by eventual depletion of maternal stores during somitogenesis, but differences in post-transcriptional processing of zygotic pnrc2 and maternal pnrc2 transcripts, such as splicing and/or translation, may also contribute.

Other cyclic gene transcript 3’UTRs contain Pnrc2-dependent instability elements

Our previous work demonstrated that the full-length her1 3’UTR is sufficient to confer Pnrc2-mediated decay (Gallagher et al., 2017), but it was unknown if other cyclic gene transcript 3’UTRs similarly confer Pnrc2-mediated decay. We show here that the full-length dlc 3’UTR is also capable of destabilizing transcripts in a Pnrc2-dependent manner (Fig. 5), suggesting a conserved mechanism of Pnrc2-dependent destabilization may regulate cyclic gene transcripts through 3’UTR interactions. The 1327 nt dlc 3’UTR is almost twice the length of the 725 nt her1 3’UTR, and more rapid destabilization of reporter transcripts by the dlc 3’UTR may be attributed to decay factors such as UPF1 that display preference for longer 3’UTRs (Hogg and Goff, 2010), or to the presence of additional decay elements. Identification of additional Pnrc2-regulated destabilizing 3’UTRs will facilitate the identification and analysis of conserved cyclic gene 3’UTR decay features and determine if 3’UTR length is a factor in Pnrc2-mediated decay.

PRE and ARE motifs are found in multiple cyclic gene transcripts

Transcriptome-wide analyses show that the PRE is among the features most strongly correlated with mRNA instability (Schwanhausser et al., 2011; Sharova et al., 2009; Yang et al., 2003), and a recent study found global enrichment of PREs and AREs in the 3’UTRs of maternal transcripts rapidly degraded during zebrafish embryogenesis (Rabani et al., 2017). Our work shows that disruption of the PRE and ARE in the her1 3’UTR affects destabilization of reporter transcripts (Fig. 6 and Fig. 7), suggesting these motifs may regulate instability of endogenous her1 transcripts. Cyclic gene transcripts such as dlc and her7 and the clock-associated transcript dld also contain PRE and ARE motifs (Fig. 5J and data not shown), suggesting PRE and ARE motifs may function in the regulation of multiple cyclic gene transcripts during somitogenesis. The dlc 3’UTR appears to confer more rapid destabilization of reporter transcripts than the her1 3’UTR, and this may be due to the presence of multiple PREs and AREs in the dlc 3’UTR, allowing for more efficient recruitment of decay factors. Additionally, unidentified decay-promoting elements, such as sequence-specific RNA binding protein motifs or decay-inducing secondary structures, may be present. The terminal 179 nt her1 3’UTR and the dlc 3’UTR both contain a canonical 5’UGCUGU Muscleblind-like 1 (MBNL1) binding motif, and MBNL1 is known to promote mRNA turnover through 3’UTR interactions (Masuda et al., 2012; Wang et al., 2015). Additionally, human PNRC2 interacts with STAU1, a well-known RNA binding protein that recognizes RNA hairpin structures, rather than sequence-specific elements, to promote STAU1-mediated decay (Cho et al., 2012). Further her1 3’UTR mutagenesis may identify additional features, and precise mutagenesis of such 3’UTR regulatory features within the endogenous gene will reveal their respective contributions to cyclic gene expression.

Pumilio proteins and ARE-binding proteins may regulate cyclic gene transcript expression during somitogenesis

We show that disruption of either the her1 3’UTR PRE or ARE extends reporter transcript half-life (Fig. 6G) and that disruption of both PRE and ARE motifs dramatically extends reporter transcript half-life (Fig. 7D). Because mutation of the PRE or ARE alone is not sufficient to fully stabilize reporter mRNA, these motifs likely promote decay in parallel. PREs are well-studied binding sites for Pumilio proteins across many species (Filipovska et al., 2011; Morris et al., 2008; Van Etten et al., 2012; Wang et al., 2002). Pumilio proteins function in diverse biological processes (Chen et al., 2012; Gennarino et al., 2018; Gennarino et al., 2015; Mak et al., 2016; Siemen et al., 2011; Xu et al., 2007; Zahr et al., 2018; Zhang et al., 2017), and Pumilio dysfunction has been linked to diseases such as neurodegeneration and cancer (Gennarino et al., 2018; Gennarino et al., 2015; Kopp et al., 2019; Miles et al., 2016; Naudin et al., 2017). With rare exceptions, Pumilio-regulated transcripts are destabilized by Pumilio (Goldstrohm et al., 2018). In contrast, of the many characterized ARE-BPs, some function to stabilize target transcripts and others to destabilize transcripts. Examples of well-studied destabilizing ARE-BPs are the ARE/poly(U)-binding/degradation factor 1 (AUF1), tristetrapolin (TTP), and KH-type splicing regulatory protein (KSRP) (Briata et al., 2005; Gratacos and Brewer, 2010; Lykke-Andersen and Wagner, 2005; Petryszak et al., 2016; Sanduja et al., 2011). Zebrafish pumilio orthologs, pum1 and pum2, and the ARE-BP orthologs auf1, ksrp, ttp, and tia-1 are expressed throughout somitogenesis (Petryszak et al., 2016), and are thus candidate regulators of cyclic gene transcript decay.

One way that Pumilio proteins promote transcript decay is by recruiting the major deadenylation machine, the Ccr4-Not (CNOT) complex, to transcript 3’UTRs (Goldstrohm et al., 2006; Joly et al., 2013; Lau et al., 2009; Van Etten et al., 2012; Weidmann et al., 2014; Arvola et al, 2019). Similar to Pumilio-mediated repression, deadenylation is also the first and rate-limiting step in degradation of many ARE-containing mRNAs (Brewer and Ross, 1988; Laird-Offringa et al., 1990; Lieberman et al., 1992; Peppel and Baglioni, 1991; Shyu et al., 1991; Wilson and Treisman, 1988). A recent zebrafish study found that inhibiting the CNOT complex causes increased her1 and her7 transcripts, suggesting that her1 and her7 3’UTRs may recruit the CNOT deadenylase complex (Fujino et al., 2018). Together, our work and the work of others suggests that Pumilio proteins and/or ARE-BPs recruit the CNOT deadenylase complex to the her1 3’UTR and other cyclic gene transcript 3’UTRs.

In addition to promoting transcript decay, Pumilio proteins can also repress translation by antagonizing poly(A)-binding protein (PABP) function (Chritton and Wickens, 2011; Van Etten et al., 2012; Weidmann et al., 2014). ARE-BPs such as TIA-1 and TIAR have also been found to inhibit translation (Dixon et al., 2003; Gueydan et al., 1999; Piecyk et al., 2000). Translational repression conferred by PRE- and/or ARE-binding proteins may explain why cyclic gene protein levels are normal in Zpnrc2 and MZpnrc2 mutant embryos (Fig. S1) (Gallagher et al., 2017).

Cyclic gene transcript 3’UTR elements may confer Pnrc2-mediated decay by influencing decapping

In addition to deadenylation, removal of the 5’ cap facilitates mRNA destabilization by providing access to 5’ to 3’ exonucleases (Schoenberg and Maquat, 2012). PNRC2 has previously been shown to interact with DCP1A in human cells (Cho et al., 2009; Lai et al., 2012; Mugridge et al., 2016), which interacts directly with DCP2 to remove the 5’ cap (She et al., 2008). Our work shows that the orthologous residue in the SH3 domain, which was shown to promote binding to DCP1A in human cells, is important for promoting Pnrc2-mediated turnover of her1 mRNA. This suggests that zebrafish Pnrc2 may recruit decapping factors to cyclic gene transcripts to initiate degradation. Studies also suggest that Pumilio proteins and ARE-BPs can regulate gene expression through interactions with the 5’ cap. In Xenopus, Pum2 has been shown to regulate translation by directly interacting with the 5’ cap (Cao et al., 2010). In Drosophila cells, knockdown of the decapping factor Dcp2 abrogates Pum-mediated mRNA decay and translational repression (Arvola, 2019). In HeLa cells, the ARE-binding protein TTP associates with decapping factors, and the presence of an ARE in mRNA stimulates decapping (Gao et al., 2001; Lykke-Andersen and Wagner, 2005). Our data suggests interaction of Pnrc2 with Dcp1a is necessary for decay of cyclic gene transcripts and that PREs and AREs in cyclic gene transcript 3’UTRs may also regulate cyclic gene transcript expression through recruitment of decapping factors such as Dcp1a.

CONCLUSIONS

We propose Pnrc2 regulates cyclic gene transcript expression through 3’UTR-mediated mRNA decay. The last 179 nucleotides (nts) of the her1 3’UTR, as well as the full-length dlc 3’UTR, are necessary and sufficient to confer Pnrc2-mediated decay to reporter transcripts. We have shown that the PRE and ARE motifs in the last 179 nts of the her1 3’UTR each contribute to reporter transcript destabilization and that together, these motifs are potent drivers of decay. Future biochemical, molecular, and genetic studies of Pnrc2, Pum1, Pum2, and ARE-BPs, and further investigation of how these factors interact with decay machinery, will provide a deeper understanding of regulators of the segmentation clock and post-transcriptional mechanisms that regulate cyclic gene expression.

• Maternally deposited pnrc2 contributes to cyclic gene transcript turnover

• The terminal her1 3’UTR confers Pnrc2-mediated instability to reporter transcripts

• The deltaC 3’UTR also confers Pnrc2-dependent instability

• PRE and ARE motifs in the her1 3’UTR drive reporter transcript destabilization

• Pnrc2 residues required for interaction with UPF1 and DCP1A contribute to activity