Pnrc2 regulates 3′UTR-mediated decay of segmentation clock-associated transcripts during zebrafish segmentation

Abstract

Vertebrate segmentation is controlled by the segmentation clock, a molecular oscillator that regulates gene expression and cycles rapidly. The expression of many genes oscillates during segmentation, including hairy/Enhancer of split-related (her or Hes) genes, which encode transcriptional repressors that auto-inhibit their own expression, and deltaC (dlc), which encodes a Notch ligand. We previously identified the tortuga (tor) locus in a zebrafish forward genetic screen for genes involved in cyclic transcript regulation and showed that cyclic transcripts accumulate post-splicing in tor mutants. Here we show that cyclic mRNA accumulation in tor mutants is due to loss of pnrc2, which encodes a proline-rich nuclear receptor co-activator implicated in mRNA decay. Using an inducible in vivo reporter system to analyze transcript stability, we find that the her1 3′UTR confers Pnrc2-dependent instability to a heterologous transcript. her1 mRNA decay is Dicer-independent and likely employs a Pnrc2-Upf1-containing mRNA decay complex. Surprisingly, despite accumulation of cyclic transcripts in pnrc2-deficient embryos, we find that cyclic protein is expressed normally. Overall, we show that Pnrc2 promotes 3′UTR-mediated decay of developmentally-regulated segmentation clock transcripts and we uncover an additional post-transcriptional regulatory layer that ensures oscillatory protein expression in the absence of cyclic mRNA decay.

INTRODUCTION

Ultradian oscillatory circuits, with periods of minutes or hours, are pervasive in biological systems (Levine et al, 2013; Purvis and Lahav, 2013; Sonnen and Aulehla, 2014). Oscillatory expression encodes an enormous amount of potential information; for example, there can be critical information in the number, amplitude, duration, or frequency of oscillations, as well as signal integration among multiple oscillators that collectively determine cellular response. A well-studied example of biological oscillation is the segmentation clock, a rapid molecular oscillator that generates periodic expression in developing embryos (Hubaud and Pourquié, 2014; Oates et al, 2012; Pourquié 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 and ribs, body musculature, and dermis. Molecular oscillations during vertebrate segmentation were first described for c-hairy1, a chick homolog of the Drosophila pair rule gene hairy. In chick embryos, c-hairy1 expression cycles in the presomitic mesoderm (PSM) and the period of each cycle corresponds with segment formation (Palmeirim et al, 1997). Since its discovery, c-hairy1 orthologs have been identified in many vertebrate species. Mouse orthologs Hes1 and Hes7 and zebrafish orthologs her1 and her7 cycle dynamically in the PSM (2 hours in mouse and 30 minutes in zebrafish) and are required for proper segmentation (Bessho et al, 2001; Bessho et al, 2003; Gajewski et al, 2003; Harima et al, 2014; Henry et al, 2002; Hirata et al, 2002; Holley et al, 2000; Oates et al, 2002; Takke and Campos-Ortega, 1999). In zebrafish, overexpression of her mRNA is associated with severe segmentation defects (Giudicelli et al, 2007; Takke and Campos-Ortega, 1999), and more recent work has confirmed that oscillatory expression is important for somite formation (Soza-Ried et al, 2014).

Several studies have explored activation and negative feedback inhibition of oscillatory transcription (e.g., Bessho et al, 2003; Giudicelli et al, 2007; González et al, 2013; Hirata et al, 2002; Lewis, 2003; Schwendinger-Schreck et al, 2014). More recently, studies have also investigated post-transcriptional mechanisms regulating transcript processing and clearance (Cibois et al, 2010; Fujimuro et al, 2014; Hanisch et al, 2013; Nitanda et al, 2014). Notable are studies that indicate splicing is a critical parameter (Harima et al, 2013; Takashima et al, 2011), mRNA export is a rate-limiting step (Hoyle and Ish-Horowicz, 2013), translational delays contribute to traveling waves of expression (Ay et al, 2014), oscillatory protein turnover is required for transcriptional and post-transcriptional clock function (Williams et al, 2016), cyclic transcript 3′UTRs can promote decay (Delaune et al, 2012, Fujimuro et al, 2014; Giudicelli et al, 2007), and miRNAs regulate decay of some cyclic transcripts (Bonev et al, 2012; Riley et al, 2013; Tan et al, 2012; Wong et al, 2015). Rapid clearance of cyclic transcripts likely occurs using mRNA decay machinery that promotes deadenylation, 5′ cap removal, and/or exonucleolytic cleavage of natural, non-aberrant transcripts (Garneau et al, 2007; Ghosh and Jacobson, 2010; Houseley and Tollervey, 2009; Lykke-Andersen and Jensen, 2015; Schoenberg and Maquat, 2012), though how cyclic transcripts are efficiently targeted and cleared remains largely unknown.

In a forward genetic screen, we discovered a zebrafish mutant, tortuga (tor), with post-transcriptional accumulation of clock-associated transcripts, such as her1, her7, deltaC (dlc), and deltaD (dld) (Dill et al, 2005). For simplicity, we refer collectively to clock-associated transcripts as cyclic through the body of this work, although deltaD (dld) expression does not oscillate (Holley et al, 2000). Normally, cyclic expression appears as dynamic stripes of expression in the anterior PSM due to rapid oscillatory transcription followed by rapid mRNA decay; in tortuga mutants, the accumulation of cyclic transcripts obscures the striped expression pattern even though cyclic transcription appears normal (Dill et al., 2005). Although many genes are deleted in the tortuga deficiency allele, we hypothesized that loss of pnrc2 specifically leads to accumulation of segmentation clock transcripts in tortuga mutants. PNRC2 was first identified in a yeast two-hybrid screen of a human mammary gland cDNA library using mouse steroidogenic factor 1 (SF1) as bait (Zhou and Chen, 2001) and subsequently shown to interact with several classes of steroid hormone receptors in vitro (Hentschke and Borgmeyer, 2003; Zhou and Chen, 2001; Zhou et al, 2006). More recently, Pnrc2 has been described as an adapter protein of mRNA decay machinery that promotes decay of reporter mRNA containing a premature termination codon (PTC) (Cho et al, 2009; Cho et al, 2012; Cho et al, 2013a; Cho et al, 2013b; Cho et al, 2015; Lai et al, 2012; Mugridge et al, 2016). We show here that Pnrc2 is required for rapid turnover of cyclic transcripts during vertebrate segmentation. We demonstrate that the her1 3′UTR confers Pnrc2-dependent instability, extending previous work in cultured cells that shows that Pnrc2 affects mRNA stability of synthetic PTC-containing reporters via NMD (Cho et al, 2009; Lai et al, 2012). We find that Pnrc2-mediated decay of her1 transcripts does not require Dicer-dependent miRNAs and likely occurs via interaction with the mRNA decay factor Upf1. Our work identifies novel targets regulated by Pnrc2 in a developmental context and implicates the existence of an additional post-transcriptional regulatory mechanism that ensures proper oscillatory protein expression.

MATERIALS & 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 tortuga (tor) mutant allele, b644, was isolated in a screen designed to identify mutations that disrupt segmental gene expression (Dill et al, 2005). The segmentation clock reporter line, Tg(her1:her1-Venus)^bk15, was generated previously to visualize cyclic gene expression (Delaune et al, 2012; Shih et al, 2015). The stable hsp70l:Venus-her1 3′UTR reporter line and the pnrc2^oz22 allele, described below, were generated in this study. Animal experiments were performed in accordance with institutional and national guidelines and regulations and were approved by the UC Berkeley and Ohio State University Animal Care and Use Committees.

Recombination mapping

Initial recombination mapping of the tor^b644 deletion allele was performed using bulk segregant analysis (Postlethwait et al, 1994) to identify polymorphic CA-repeat microsatellite markers showing biased representation in pooled genomic DNA from haploid progeny derived from an F1 AB/SJD hybrid female carrying the tor^b644 allele. Recombination frequency for linked markers was calculated by analyzing marker segregation among many mutant and wildtype F1 hybrid individuals. The extent of the tor^b644 deletion was defined as described in Results. Mapping marker locations and sequences are available at the Zebrafish Model Organism Database (ZFIN), University of Oregon, Eugene, OR 97403-5274; URL: http://zfin.org/.

BAC injection

A total of 4 BACs spanning the tortuga^b644 deletion were injected at doses of 0.4–30 pg directly into 1-cell stage embryos from a cross between heterozygous tor^b644 carriers. At 18 hpf, embryos in each BAC-injected clutch were sorted for neural degeneration (a visible tor phenotype), and then fixed in 4% PFA for 5 hours at room temperature, processed for her1 in situ hybridization, scored for her1 expression phenotype, and PCR genotyped. Only BAC AL844887 restored normal her1 expression in tor^b644 mutants (Fig 1A). BAC map position and sequence (CR848819, CR936374, AL844887, and BX649265) are available at http://zfin.org.

Figure 1. The tortuga^b644 allele is a 1.46 Mb deficiency that includes the pnrc2 gene.

Haploid-based mapping revealed that the tortuga lesion lies 0.30 cM to the left (4/1434 recombinants) and 0.37 cM to the right (4/1083 recombinants) of the SSLP markers z13511 and z9511, respectively (A). The extent of the deletion was refined by PCR-based screening of regions within and around the deletion interval from pooled genomic DNA samples of 10 wild-type (WT) and 10 tortuga mutant embryos (B); regions that fail to amplify in mutants are deleted in the tor^b644 allele (A, B). Among BACs spanning the deletion interval, only BAC AL844887 restores her1 expression when injected into tor mutants (A; Fig S1A–D). In wild-type embryos, her1 is expressed in a striped pattern (n=54/54) (C). In contrast, embryos injected with 6 ng of pnrc2 splice-blocking morpholino (sbMO) have a tortuga-like her1 expression defect (n=37/39) (D). When pnrc2 mRNA is co-injected with 4 ng pnrc2 sbMO, her1 expression is partially restored in a dose-dependent manner (n=3/19 WT her1 expression, 150 pg pnrc2 mRNA; n=7/17 WT her1 expression, 600 pg pnrc2 mRNA) (E; Table 1). Alignment of vertebrate Pnrc2 amino acid sequences reveals a conserved 10 amino acid N-terminus (purple) and conserved C-terminal SRC-Homology 3 (SH3) (red) and Nuclear Receptor (NR box) domains (green) (F). Clustal alignments were performed in consultation with published alignments for vertebrate Pnrc2 (Valkov et al, 2016).

CRISPR/Cas9 mutagenesis

An optimal target site, 5′-GGGCACCCCTAAGGCTCCTG-3′, in the 5′ coding sequence of pnrc2 was identified using the ZiFit Targeter software package (Sander et al, 2007; Sander et al, 2010). pnrc2-targeting gRNA (5′-CAGGAGCCTTAGGGGTGCCC-3′) and Cas9 mRNA (Jao et al, 2013) were synthesized and co-injected into 1-cell stage embryos (135 ng and 200 ng, respectively) as described (Talbot and Amacher, 2014). At 24 hpf, a subset of injected embryos were individually screened by high-resolution melting analysis (HRMA) to assess target site mutation efficiency in somatic cells. Remaining embryos were raised and crossed to AB wild-type adults; F1 adults were screened for germline transmission of CRISPR-induced mutations using HRMA. HRMA revealed two unique pnrc2 mutant alleles transmitted by an individual F0 founder at a transmission rate of ~13% (2 of 15 F1 individuals). We recovered one allele, pnrc2^oz22, and outcrossed pnrc2^oz22 heterozygotes to the AB wild-type strain for two generations before intercrossing for phenotypic analyses. Primer sequences are listed in Table S4.

DNA extraction and pnrc2^oz22 genotyping strategy

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 (BP1700, ThermoFisher), 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 reaction 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). Primer sequences are listed in Table S4.

Morpholino injection

The pnrc2 splice-blocking morpholino (sbMO) sequence is: 5′-ACTGGATGTCACctagcagaagaca-3′ (uppercase, sequence complementary to exon 3; lower case, sequence complementary to intron 2) (Gene Tools, LLC). The upf1 sbMO, 5′-TTTTGGGAGTTTATACCTGGTTGTC-3′, was published previously (Wittkopp et al, 2009). The rbfox1l sbMO, 5′-GCATTTGTTTTACCCCAAACATCTG-3′, and rbfox2 sbMO, 5′-TATAATGCTTTATATACCCCGAACA-3′, was published previously (Gallagher et al, 2011; Berberoglu et al, 2017). Morpholinos were diluted to 0.1–2 ng/nl in 0.2M KCl and 0.1% phenol red and injected into the yolk of 1-cell stage embryos. pnrc2 sbMO dose was optimized by determining the highest dose that gave reproducible and rescuable phenotypic defects with no toxicity. Primers used to assess efficacy of pnrc2 sbMO injection (Fig S1K–L) are listed in Table S4. upf1, rbfox1l, and rbfox2 sbMO doses were performed according to published methods using doses that gave reproducible phenotypic defects matching published results (Wittkopp et al, 2009; Gallagher et al, 2011; Berberoglu et al, 2017). Embryos were incubated at 28.5°C until 6 hours post fertilization (hpf) and then transferred to 25°C thereafter, except for a subset of rbfox1l/rbfox2 double-injected and uninjected control embryos that were incubated at 28.5°C until 24 hpf and subsequently scored for ability to move.

mRNA injection

Full length pnrc2 cDNA was amplified by RT-PCR and subcloned into expression vector pCS2+ (Rupp et al, 1994; Turner and Weintraub, 1994) to generate plasmid SP6-pnrc2-cDNA (TLG109). For rescue experiments, pnrc2 mRNA was synthesized using the SP6 mMessage Machine Kit (Life Technologies), diluted in 0.2M KCl with 0.1% phenol red, and injected into 1-cell stage embryos (150–600 pg mRNA per embryo). Primer sequences are listed in Table S4.

In situ hybridization

Whole mount in situ hybridization was performed as previously described (Broadbent and Read 1999; Jowett 1999) using DIG-labeled antisense probes. The full length pnrc2 cDNA was amplified by RT-PCR and subcloned into pBSKS+ (Stratagene), linearized with BamHI, and transcribed using T7 RNA polymerase to make DIG-labeled antisense pnrc2 riboprobe (Roche Life Science). The same construct was linearized with XhoI and transcribed using T3 RNA polymerase to make DIG-labeled sense pnrc2 riboprobe. Riboprobes for her1, her7, dlc, dld, and Venus were made as previously described (Dill et al, 2005; Delaune et al, 2012). In situ hybridization chain reaction (HCR-ISH) was performed using a combination of five anti-sense 50-nt probes spanning the her1 transcript according to published procedures (Choi et al, 2010; Choi et al, 2014) and a zebrafish-specific protocol provided by Molecular Instruments. Probe targeting sequences (5′ to 3′) were:

1. GGGTTTTGAAGTCGCGAATCTAAAGTATTATCCAGAAGAAGCGTTCGCAG,

2. CGCCTTGATCTCTCGCAGTCGCGGTTTTAGTCCTAATATACTCAACAGCC,

3. GAGAATGGAGGAGAGCTGCTTGAAAAGCCTGGAGACGGCGGAGGAGAAAT,

4. TCACCTGAAGATGAGGTCCTGGGACGACCGGTAATGAAGTCGTTGAGAGA, and

5. TCGTCTCAGAGTCCGTGGTTGAGAGGATTGAACAGAGCCACTAAACCGCA.

RNA analysis

Whole embryos (n=20 per time point or condition) were solubilized in Trizol for RNA extraction (Life Technologies). 1 ug total RNA was purified and reverse transcribed into cDNA with random primers and Superscript III reverse transcriptase (RT) according to the manufacturer’s instructions (Life Technologies). Expression analysis of pnrc2 using primer pairs spanning constitutive exons 2 and 3 were used for RT-PCR-based detection of pnrc2 transcript (Fig 4G, S1K and Table S4 for primer sequences). Splicing of pnrc2 using primer pairs spanning within and across each of three constitutive exons of the 3146 nt pnrc2 mRNA were used for RT-PCR-based detection of spliced and unspliced pnrc2 transcript (Fig S1K–L and Table S4 for primer sequences).

Figure 4. pnrc2 is broadly expressed during segmentation stages.

pnrc2 transcripts are detected during early embryonic stages and throughout segmentation stages by in situ hybridization (n>15 per time point) (A–F). As expected, there is no detectable staining with a pnrc2 sense probe (n=15/15) (data not shown). RT-PCR expression analysis for pnrc2 at 2 hours post fertilization (2h) through mid-segmentation (18h) is consistent with in situ detection of pnrc2 transcript (see Methods and Results) (G).

Plasmid construction and Transgenesis

The heat-shock reporter hsp70l:Venus-her1 3′UTR was assembled by PCR amplification and restriction digestion of the hsp70l promoter from Tol2kit construct #222 (entry plasmid p5E-hsp70l) (Kwan et al, 2007), in parallel with restriction digestion of the Venus-her1 3′UTR sequence from the her1:her1-Venus plasmid (Delaune et al, 2012), followed by ligation of both fragments into a modified version of pBSKS+ plasmid containing flanking I-SceI meganuclease recognition sites (Thermes et al, 2002). The Venus-her1 3′UTR fragment isolated from plasmid her1:her1-Venus contains the Venus coding sequence followed by 1.1 kb of her1 3′ noncoding sequence that includes the annotated 724 nt her1 3′UTR and native her1 pA signal sequence (Delaune et al, 2012). Constructs were sequence confirmed; primers used for cloning are listed in Table S4. Transgenic lines were generated as previously described using I-SceI-based transgenesis (Thermes et al, 2002).

Stably transgenic heatshock assay

Adult fish carrying the stable hsp70l:Venus-her1 3′UTR transgene that transmits as a single Mendelian locus were crossed to AB wild-type fish and resulting progeny were either injected at the 1-cell stage with 6 ng splice-blocking pnrc2 morpholino (sbMO) or were set aside as uninjected control siblings. Progeny were raised to mid-segmentation, heat-shocked at 37°C for 15 minutes, and fixed in 4% PFA at 0, 20, and 30 minutes post-heat-shock and processed for Venus in situ hybridization.

Immunohistochemistry

All embryos described below were immunostained following standard protocols using 4% PFA fixation, dehydration and rehydration in a methanol series, and incubation in blocking solution for 1 hour. Tg(her1:her1-Venus)^bk15 embryos were immunostained in 2% BSA/2% goat serum/1% DMSO/0.1% Tween-20/PBS blocking solution with 1:1000 dilution chicken anti-GFP that recognizes Venus protein (A10262, Life Technologies), and 1:400 dilution goat anti-chicken Alexa-Fluor-488 (A11039, ThermoFisher). Mid-segmentation embryos from wild-type and pnrc2^oz22 crosses were immunostained in 2% BSA/5% goat serum/0.1% Tween-20/PBS blocking solution with 1:200 dilution anti-zdc2 that recognizes DeltaC protein (ab73336, Abcam) according to previously published methods (Giudicelli et al, 2007) or immunostained in 2% BSA/10% goat serum/0.5% Triton X-100/PBS blocking solution with 1:100 anti-zdd2 that recognizes Dld protein according to previously published methods (Wright et al, 2011) (ab73331, Abcam), followed by 1:800 dilution goat anti-mouse Alexa-Fluor-488 (A11001, ThermoFisher). Nuclear counter-staining was performed by transferring and mounting dissected in situ-hybridized embryos from 80% glycerol into SlowFade Gold Antifade Mountant with DAPI (S36939, Thermo Fisher) and incubation at 4°C overnight prior to imaging.

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 dissected and flat mounted or whole mounted in 80% glycerol and imaged at 10x, 20x, and 60x magnification using MetaMorph software (Molecular Devices) on an Andor™ SpinningDisc Confocal Microscope (Oxford Instruments) with iXon Ultra EMCCD and Nikon Neo cameras; laser wavelength and intensity were set at 488 nm and 100% for Venus protein detection, 488 nm and 50% for Dlc protein detection, 488 nm and 100% for Dld protein detection, 561 nm and 30% for her1 mRNA detection, 405 nm and 40% for DAPI detection, respectively, and bit depth at 16-bit. Maximum intensity projections using MetaMorph software are shown for Venus, Dlc, and Dld protein detection (Fig 7D–E, K–N). Single z-sections are shown for her1 HCR-ISH and DAPI (Fig 7C–C′, F–F′; Fig S6A–B‴).

Figure 7. Both reporter and endogenous cyclic transcripts accumulate in Pnrc2-depleted embryos, but protein expression appears normal.

Embryos carrying the her1:her1-Venus^bk15 transgenic clock reporter were injected with pnrc2 splice-blocking morpholino (sbMO) and processed to detect Venus transcripts (A, B) and Venus protein (D, E) at mid-segmentation stages. Representative embryos are shown in A (n=21/21); B (n=18/18), D (n=32/32), and E (n=29/29). Venus immunofluorescence panels (D, E) are at slightly higher magnification than Venus in situ panels (A, B). Detection of her1 mRNA by in situ hybridization chain reaction (HCR-ISH) (C, F) is consistent with chromogenic NBT/BCIP-based in situ detection of endogenous her1 transcript in wild-type and pnrc2 mutant embryos (Fig 2F, H′), with substantial cytoplasmic localization revealed by DAPI counter staining in 500X magnified view (C′, F). Because relative intensity of her1 HCR-ISH in pnrc2 mutants to wild-type embryos is high, levels have been reduced in pnrc2 mutant panels (F–F; see Fig S6A‴, B‴). Misexpression of dlc and dld mRNA is detected throughout the presomitic mesoderm (PSM), formed somites and neurons in the expected one-quarter of embryos in a pnrc2^oz22 intercross, n=5/28 (X²=0.76, p=0.4) and n=7/40 (X²=1.2, p=0.3), respectively (G–J). In contrast, Dlc and Dld protein expression is indistinguishable among siblings of the same pnrc2^oz22 heterozygote intercross (K–N). Dlc and Dld immunolabeled embryos were genotyped prior to imaging and a subset of wild-type and mutant siblings were imaged by confocal microscopy with representative embryos shown (K–N). Total genotyped individuals per representative panel: n=5 (K), n=6 (L), n=12 (M), n=4 (N). Scale bars = 50 um (D, F), 50 nm (F′), 100 um (K).

RESULTS

tortuga^b644 is a Chromosome 16 deficiency allele

The tortuga^b644 allele is an ENU-induced deletion that leads to the post-transcriptional accumulation of segmentation clock transcripts (Dill et al, 2005). Using genetic markers that distinguish wild-type AB and SJD mapping strains, we found that b644 is a deficiency that maps to a 1.46 Mb region on Chromosome 16 spanning an interval of at least 20 known or predicted protein-coding RefSeq-annotated genes in genome assembly GRCz10/danRer10 (Howe et al, 2013). Haploid-based mapping revealed that the tortuga lesion lies 0.30 cM to the left and 0.37 cM to the right of the SSLP markers z13511 and z9511, respectively (Fig 1A). Using PCR-amplification of genomic regions (mostly in protein-coding genes) that lie between the two SSLP markers, we characterized the extent of the deletion in diploid embryos derived from heterozygous b644 intercrosses by identifying genes that fail to amplify in tortuga homozygous mutant versus wild-type sibling embryos (Fig 1B). To better map deletion breakpoints, we analyzed presence or absence of amplicons near the presumptive ends of the tortuga deletion region of Chromosome 16. At the end near z13511, an intergenic region located ~124 kb upstream of pou3f2b failed to amplify, indicating the break lies in a ~206 kb interval between mms22l and the intergenic region (Fig 1B). Similar analysis at the other end indicates that the other breakpoint lies in the ~2.9 kb genomic interval between exon 1 and intron 3 of snip1 (Fig 1B).

Injection of a BAC that includes pnrc2 rescues the her1 expression defect in tortuga mutants

To narrow the list of relevant candidate gene(s) in the tortuga deficiency, we injected four BACs spanning regions of the deletion interval between pou3f2b to snip1 into zebrafish embryos at the 1-cell stage and assessed her1 expression phenotype. Because tortuga mutants do not survive beyond larval stages, BACs were injected into embryos from a heterozygous intercross, from which ~25% are homozygous for the tor deletion. Injection of zebrafish BAC clone AL844887 spanning nine full-length open reading frames rescues the her1 expression defect in tor mutants in a dose-dependent manner (Fig 1A, Fig S1A–D; Table S1).

Loss of pnrc2 is associated with accumulation of her1 mRNA in tortuga mutants

To identify the gene or genes on BAC AL844887 that restore proper her1 expression in tortuga mutants, we first analyzed candidate gene expression by RT-PCR before and during segmentation. Of nine candidates present on BAC AL844887, only seven are expressed at relevant time points (Fig S1E; data not shown for me1). Of these seven, only pnrc2, rragca, and to a lesser extent, akirin1, are detectably expressed from BAC AL844887 when injected into tortuga^b644 mutants (Fig S1F). We injected antisense morpholinos (MOs) into 1-cell stage wild-type embryos to determine whether knockdown of any of the three candidates recapitulated the tor-like her1 expression defect. Injection of translation-blocking MOs targeting akirin1 and rragca does not cause overt morphological or her1 expression defects (Fig S1I–J), and injection of maximal non-toxic doses of akirin1 or rragca mRNA into tortuga mutants does not rescue her1 expression defects (data not shown). In contrast, injection of pnrc2 splice-blocking MOs (sbMOs), that effectively disrupt proper pnrc2 splicing (Fig S1K–L), disrupts her1 expression just as in tortuga^b644 mutants (Fig 1C–D). Injection of a second pnrc2-targeting translation-blocking MO gave the same phenotype (Fig S1H; see Methods). Importantly, co-injection of MO-resistant pnrc2 mRNA with pnrc2 sbMO restores normal her1 expression (Fig 1C–E; Table 1).

Table 1.

Injection of pnrc2 mRNA partially restores normal her1 expression in pnrc2 morphants

Condition	her1 expressiona
	Normal	Partial tor-like	tor-like
Uninjected WT	54/54 (100%)	0/54 (0%)	0/54 (0%)
4 ng pnrc2 sbMO	2/39 (5%)	28/39 (72%)	9/39 (23%)
4ng pnrc2 sbMO + 150 pg pnrc2 mRNAb	3/19 (16%)	12/19 (63%)	4/19 (21%)
4 ng pnrc2 sbMO + 600 pg pnrc2 mRNAb	7/17 (42%)	9/17 (53%)	1/17 (6%)

^aWild-type embryos were injected at the 1-cell stage with 4 ng pnrc2 sbMO with and without increasing doses of MO-resistant pnrc2 mRNA, raised to 16–18 hpf, and processed by her1 in situ hybridization. Normal, her1 is expressed in visible stripes separated by regions of low expression; Partial tor-like, her1 stripes are distinguishable, but there is substantial her1 detected between stripes; tor-like, her1 expression is strong throughout the PSM with no obvious peaks or troughs.

^bChi-square analysis indicates a significant difference in her1 expression between pnrc2 morphants with and without co-injection of pnrc2 mRNA. Co-injection of 150 pg and 600 pg of pnrc2 mRNA significantly restores her1 expression in morphants (p<2.1 × 10⁻³ and p<8.8 × 10⁻⁸, respectively).

Pnrc2 restores proper her1 expression in tortuga mutants

To determine whether Pnrc2 can also restore normal her1 expression in tortuga mutants, we injected pnrc2 mRNA into embryos from a tor^b644 heterozygote intercross. To discriminate “rescued” tor mutants and wild-type siblings among injected intercross progeny, we developed a visual assay to unambiguously identify fish homozygous for the tor deletion. The pou3f1 gene lies within the tortuga deletion (Fig 1A) and thus is not expressed in tor^b644 mutant embryos, whereas in wild-type siblings, pou3f1 mRNA is expressed anteriorly in a pattern easily distinguished from that of her1 mRNA expression. By co-hybridizing pou3f1 and her1 antisense probes, tor^b644 mutants are readily identified and assessed for rescue of her1 expression. Using this assay, we find that pnrc2 mRNA-injected tortuga mutants exhibit wild-type her1 expression (Fig 2A–D′; Table 2), indicating that Pnrc2 can restore proper her1 expression in tor mutants.

Figure 2. tor mutant embryos are rescued by injection of pnrc2 mRNA and phenocopied by frame-shifting mutation of the pnrc2 locus.

Injection of 100 pg pnrc2 mRNA has no effect in wild-type sibling embryos (n=15/15 non-mutant siblings) (A, A′ vs B, B′) but restores striped her1 expression in tor mutant embryos (n=10/10 mutants) (C, C′ vs D, D′; Table 2). In these experiments, rescued tor^b644 mutant embryos were distinguished from wild-type siblings by lack of expression of pou3f1, a gene located in the tor^b644 deficiency interval. Asterisks (*) mark the neural pou3f1 expression domain. Arrows mark the her1 expression domain (A–D), magnified in dorsal view to the right of each embryo (A′–D′). Using CRISPR-based mutagenesis, we induced a 17 bp deletion within the pnrc2 coding sequence, creating an early frameshift allele, designated pnrc2^oz22 (E). Predicted mutant protein sequence is based on sequenced genomic DNA (E). The pnrc2^oz22 allele fails to complement the her1 accumulation phenotype of the tor^b644 deletion allele (n=4/15 embryos from a heterozygote intercross with tor-like her1 accumulation) (F, G). Similar to tor^b644 mutants, pnrc2^oz22 mutant embryos lack a striped her1 expression pattern (H, ′H).

Table 2.

Injection of pnrc2 mRNA restores normal her1 expression in tor^b644 mutants

mRNA dose (pg)	Live Embryonic Phenotype		WT her1 ISH/ WT genotypeb	WT her1 ISH/ tor genotypeb	tor her1 ISH/ tor genotypeb
	WTa	Neural degena
0	103	31	103/103	0/31	31/31
75	64	20	64/64	19/20c	1/20d
100	15	10	15/15	10/10c	0/10d

^aAt the 16–18 somite stage, pnrc2 mRNA-injected embryos from a tor^b644 heterozygote intercross were scored for neural degeneration, a non-somitic phenotype observed in tor mutants.

^bAfter sorting by live phenotype, injected embryos were processed by her1 in situ hybridization to assess phenotypic rescue and with pou3f1 to identify individuals homozygous for the tor deletion. Representative embryos are shown in Fig 2.

^cBoth 75 and 100 pg doses correlate with restoration of normal her1 expression. At the 100 pg dose, all tor mutants have a normal her1 expression pattern.

^dChi-square analysis indicates a significant difference in her1 expression among pnrc2 mRNA-injected tor mutants compared to uninjected homozygous siblings (p<4.7 × 10⁻⁷ and p<3.9 × 10^{− 3}, for 75 pg and 100 pg doses, respectively).

Targeted mutation of pnrc2 recapitulates the cyclic transcript accumulation phenotype of tor^b644 mutants

Because the tor^b644 allele is a multi-gene deficiency, we used CRISPR/Cas9 mutagenesis (Hruscha et al, 2013; Hwang et al, 2013; Talbot and Amacher, 2014) to generate a nonsense pnrc2 allele. We isolated a 17 bp deletion allele, pnrc2^oz22 that causes an early frame shift and likely results in a truncated Pnrc2 protein. Compared to wild-type Pnrc2 (148 amino acids), the predicted mutant protein determined from sequenced genomic DNA contains the N-terminal 35 amino acids followed by 11 aberrant residues (Fig 2E), terminating well before the two highly conserved SH3 and NR box regions (Fig 1F). To determine whether the pnrc2^oz22 frame-shifting allele fails to complement the tor^b644 deletion allele, we crossed pnrc2^oz22 and tor^b644 heterozygotes and found that all pnrc2^oz22/tor^b644 trans-heterozygote progeny display the tor^b644 her1 expression defect (Fig 2F–G). As expected, homozygous pnrc2^oz22 mutants phenocopy the tor^b644 her1 expression defect (Fig 2H–H′). Additionally, expression of segmentation clock-associated her7, dlc, and dld transcripts is also abnormal in pnrc2^oz22 mutants (Fig 3) and in the cases of her1 and dlc, arise due to post-transcriptional accumulation of mRNA (Fig S2), all consistent with previous observations in tor^b644 mutants using intronic and exonic in situ probes that distinguish nascent from processed transcripts (Dill et al, 2005). Unlike tor^b644 mutants, pnrc2^oz22 mutants and pnrc2^oz22/tor^b644 trans-heterozygotes do not have neural degeneration and somite shape defects, suggesting these tor mutant phenotypes are caused by loss of function of another gene or genes in the tor deletion interval. pnrc2^oz22/oz22 embryos appear morphologically normal and survive through 6 days post fertilization, but very few survive to adulthood (data not shown). Reduced survivorship of zebrafish pnrc2^oz22/oz22 mutants contrasts with Pnrc2-null mice that survive to adulthood and are indistinguishable from wild-type littermates up to 12 months of age (Zhou et al, 2008).

Figure 3. Segmentation clock transcripts accumulate in pnrc2 mutant embryos.

Like her1, other segmentation clock genes, dlc (A, B), dld (C,D), and her7 (E, F), are misexpressed in pnrc2 mutant embryos, with expression detected throughout the presomitic mesoderm (PSM) in the expected one-quarter of embryos in a pnrc2^oz22 intercross, n=8/25 (X²=0.65, p=0.4), 6/25 (X²=0.013, p=0.9), and 8/25 (X²=0.65, p=0.4), respectively (A–F).

pnrc2 is broadly expressed during segmentation

To characterize embryonic pnrc2 expression, we performed whole mount in situ hybridization and found that pnrc2 is broadly expressed, with slight enrichment in somites and neural tissue during segmentation stages (Fig 4A–F). RT-PCR analysis of mRNA extracted from wild-type embryos confirms that pnrc2 is expressed across similar stages examined by in situ (Fig 4G). Expression at the 8-cell stage (Fig 1A) and at the 1- and 2-cell stages (data not shown) suggests that pnrc2 transcripts are maternally provided.

The her1 3′UTR confers instability to transcripts in a Pnrc2-dependent manner

In previous studies, heat-shock-induced reporter transcripts containing her1 or her7 coding and 3′UTR sequences decayed rapidly post-induction (Giudicelli et al, 2007). Because 3′UTR sequences can influence mRNA stability (Ghosh and Jacobson, 2010; Chen and Shyu, 2011; Schoenberg and Maquat, 2012), we hypothesized that the her1 3′UTR alone might be sufficient to trigger Pnrc2-mediated decay. We therefore developed a stable transgenic heat-shock reporter line to drive expression of a Venus transcript followed by the her1 3′UTR. We compared reporter expression after heat-shock of uninjected and pnrc2 sbMO-injected hsp70l:Venus-her1 3′UTR transgenic siblings across multiple time points. Heat-shock-induced Venus transcripts are almost completely absent by 30 minutes post heat-shock (pHS) in the reporter line (Fig 5A–C). However, injection of pnrc2 sbMO into hsp70l:Venus-her1 3′UTR embryos negates the destabilizing effect of the her1 3‘UTR (Fig 5D–F; Table 3). These results support our hypothesis that Pnrc2-mediated decay of her1 transcripts occurs through destabilizing features of the her1 3′UTR.

Figure 5. Pnrc2-mediated decay functions via 3′UTR recognition and does not require Dicer-dependent miRNAs.

Stably transgenic embryos carrying the hsp70l:Venus her1 3′UTR reporter were injected at the 1-cell stage with pnrc2 sbMO or reserved as uninjected controls. Embryos were then raised to mid-segmentation stage, heat-shocked for 15 minutes, then collected at the indicated minutes post-heat-shock (pHS) and processed by Venus in situ hybridization (n=65, 6 ng pnrc2 sbMO; n=55, uninjected controls) (A–F). Venus transcripts are not detected in the absence of heat-shock (n=10) (not shown). To determine whether Dicer-generated miRNAs contribute to her1 mRNA decay, MZdicer mutants (n=11) and wild-type controls (n=10) were raised to mid-segmentation stages (16–18 hpf) and processed by her1 in situ hybridization (G, H). At this timepoint, MZdicer mutants are developmentally delayed relative to wild-type controls (Giraldez et al, 2005), and thus have a different overall shape.

Table 3.

Pnrc2 is required for the transcript destabilizing effect of the her1 3′UTR

hsp70l:Venus-her1 3′UTR	pHS recovery timeb	Expression Levelc			Number of embryos	Mean Expression
Conditiona		3	2	1
Uninjected	0	20	0	4	24	2.7d
6 ng pnrc2 MO	0	27	0	0	27	3.0d
Uninjected	20	0	0	6	6	1.0d
6 ng pnrc2 MO	20	11	0	0	11	3.0d
Uninjected	30	0	8	17	25	1.3d
6 ng pnrc2 MO	30	27	0	0	27	3.0d

^aA stably transgenic male carrying a hsp70l:Venus-her1 3′UTR reporter was crossed to a wild-type female; half of the progeny were injected with pnrc2 sbMO and the other half used as uninjected sibling controls.

^bAt 14–16 hpf, embryos were heat-shocked at 37°C for 15 minutes, followed by fixation at indicated post-heat shock (pHS) recovery times.

^cFixed embryos were processed and scored as described in Table 1. Embryos completely lacking Venus expression were not included because we could not discriminate the expected 50% of embryos that did not inherit the transgene from transgenic embryos lacking detectable Venus expression.

^dAverage expression levels reveal that pnrc2 MO-injection into transgenic embryos dramatically stabilizes Venus transcript levels post heat-shock across the recovery time course of 30 minutes when compared to uninjected controls.

her1 expression is unaffected in embryos lacking maternal and zygotic Dicer function

It is well established that microRNAs (miRNAs) play an essential role in post-transcriptional gene regulation in developing zebrafish embryos (Bazzini et al, 2012; Giraldez, 2010; Giraldez et al, 2005; Mishima and Tomari, 2016; Mishima et al, 2006; Jonas and Izaurralde, 2015). Using TargetScan Fish (Ulitsky et al, 2012), we find that predicted miRNA target sites are present throughout the her1 3′UTR and other cyclic transcript 3′UTRs (Table S2). The 3′UTRs of her1, her7, dlc, and dld lack a common predicted target site, however some target sites are present in at least two of the four 3′UTRs analyzed. We therefore reasoned that miRNAs might influence the decay of her1 and other transcripts. Using maternal-zygotic dicer (MZdicer) mutants that lack Dicer-dependent miRNA processing (Giraldez et al, 2005), we find that segmenting MZdicer mutants, despite having severe morphogenesis defects (Giraldez et al, 2005), have a normal striped her1 expression pattern (n=11/11), demonstrating that proper her1 expression is independent of Dicer-dependent miRNA function (Fig 5G–H).

Pnrc2 and Upf1 may genetically interact to promote her1 transcript decay

Human PNRC2 binds directly to nonsense-mediated decay (NMD) factors, including UPF1, and these interactions are required for decay of reporter mRNA in cultured cells (Cho et al, 2009; Lai et al, 2012). We therefore predicted that Pnrc2 and Upf1 might also interact to promote decay of non-aberrant, cyclic transcripts during segmentation. To determine whether Pnrc2 and Upf1 cooperatively regulate cyclic mRNA decay, we co-injected MOs targeting both transcripts and examined her1 expression. As expected, injection of pnrc2 sbMO at the optimal dose (“moderate”, 6 ng) causes strong her1 misexpression, but has no effect at a sub-optimal (“low”) dose (2 ng) (Fig 6A–C). Strikingly, co-injection of a sub-optimal dose of pnrc2 sbMO (2 ng) with a sub-optimal dose of upf1 sbMO (0.25 ng) results in a her1 expression defect markedly similar to that observed with optimal doses of pnrc2 sbMO alone (6 ng) (Fig 6B–E, quantified in F). Interestingly, single upf1 knockdown at published (0.65 ng) or higher doses (2 ng) does not alter her1 mRNA expression, but does induce the expected published phenotypes including neural necrosis and abnormal segment formation (data not shown) (Wittkopp et al, 2009; Anastasaki et al, 2011). As a control, we injected a sub-optimal dose of pnrc2 sbMO (2 ng) with an unrelated MO, rbfox1l sbMO, at the published 6 ng dose (Gallagher et al, 2011; Berberoglu et al, 2017), and observed that her1 mRNA expression is normal (Fig S3), suggesting that her1 misexpression in embryos co-injected with sub-optimal doses of pnrc2 and upf1 sbMOs is a specific effect. Taken together, these results reveal that depletion of Upf1 sensitizes embryos to partial loss of Pnrc2. Although depletion of Upf1 alone does not affect her1 expression, it is possible that low levels of Upf1 protein persist in upf1 morphants and may be sufficient for cyclic mRNA clearance in the presence of normal levels of Pnrc2.

Figure 6. pnrc2 and the nonsense-mediated decay effector Upf1 promote decay of cyclic mRNA.

Injection of low dose pnrc2 sbMO (2 ng) has little to no effect on her1 expression (A, C), contrasting with the expected her1 misexpression observed after injection of moderate dose pnrc2 sbMO (6 ng) (B). Low dose injection of upf1 sbMO (0.25 ng) also has little effect on her1 expression (D), but when combined with a low dose of pnrc2 sbMO (2 ng), her1 misexpression is observed in about 50% of injected embryos (E). The proportion affected in each condition is plotted on a bar graph that indicates significant differences between single morphants, double morphants, and controls (F). sbMO = splice-blocking MO; **** = p<0.0001.

Unlike mRNAs, cyclic proteins do not accumulate upon Pnrc2 depletion

Our data indicate that Pnrc2 triggers decay of reporter mRNAs and natural cyclic transcripts like her1. Previous work has shown that embryos injected with her1 mRNA at the 1-cell stage have severe somite patterning and boundary defects, as well as decreased expression of Her1 transcriptional targets like dlc and her7 (Takke and Campos-Ortega, 1999; Giudicelli et al, 2007). The morphological phenotype of embryos constitutively overexpressing her1 contrasts sharply with that of pnrc2^oz22 mutant embryos, which have normal somite numbers and boundaries (Table 4 and data not shown) despite dramatic accumulation of endogenous her1 and other cyclic transcripts. We therefore hypothesized that accumulated her1 transcripts in pnrc2^oz22 mutants do not result in increased Her1 protein. Because a Her1 antibody is lacking, we tested this idea by depleting pnrc2 function in a validated transgenic cyclic reporter line, Tg(her1:her1-Venus)^bk15, that infers her1 transcript and Her1 protein dynamics via detection of her1-Venus reporter mRNA and protein (Delaune et al, 2012; Shih et al, 2015). As expected, her1-Venus transcripts are misexpressed in pnrc2 morphants carrying the cyclic reporter transgene, mirroring what is observed for endogenous her1 transcripts (Fig 7A–B). Her1-Venus reporter protein expression, however, is indistinguishable between uninjected and pnrc2 sbMO-injected transgenic embryos (Fig 7D–E). Her1-Venus protein expression is similarly unaffected in transgenic reporter embryos injected with sub-optimal doses of pnrc2 sbMO (2 ng) and upf1 sbMO (0.25 ng) as well as transgenic reporter embryos co-injected with an optimal dose of pnrc2 sbMO (6 ng) together with a higher dose of upf1 sbMO (2 ng) (Fig S4; Table S3). Together, the observed misexpression of her1-Venus mRNA, but not Her1-Venus protein, is consistent with our hypothesis that pnrc2 mutants segment normally because accumulated her1 transcripts do not result in abnormal levels of Her1 protein.

Table 4.

pnrc2 mutants segment normally

Feature	Genotype
	Wild type	pnrc2+/−	pnrc2−/−
Segment number	34.7 ± 0.7 (n = 14)	35.3 ± 0.3 (n = 14)	34.3 ± 0.4 (n = 14)

Sibling embryos from a pnrc2o^z22 heterozygote intercross were raised to 36 hpf and processed by cb1045 (xirp2a) in situ hybridization that reliably marks segment boundaries. PCR-based genotyping was performed for individuals after in situ hybridization and segment counts. No significant differences were observed between genotypes. Segment numbers are given as mean ± 95% confidence interval (Student’s t test).

Discordant transcript and protein expression in pnrc2 mutants and morphants might be due to nuclear retention of accumulated mRNA, so we employed in situ hybridization chain reaction (HCR-ISH) (Choi et al, 2010; Choi et al, 2014) coupled with DAPI nuclear counter staining to assess sub-cellular localization of her1 mRNA. As a control for probe set specificity, we first performed her1 HCR-ISH on wild-type and Df(Chr05:her1,her7,ndrg3a)^b567 homozygote embryos that lack the linked her1 and her7 genes (Henry et al, 2002) and find that b567 homozygotes lack detectable signal (Fig S5). We then performed HCR-ISH on both wild-type and pnrc2 mutants and find substantial cytoplasmic localization of her1 mRNA in both conditions, suggesting that accumulated transcripts in pnrc2 mutants are not retained in the nucleus and are properly exported to the cytoplasm (Fig 7C–C′, F–F′; Fig S6A–A″, B–B″). Because relative intensity of her1 HCR-ISH in pnrc2 mutants to wild-type embryos is high, levels have been reduced in pnrc2 mutant panels (Fig 7F–F′; Fig S6B–B″). Intensity maps of raw her1 HCR-ISH signal reflect the extent of her1 mRNA accumulation in pnrc2^oz22 mutants (Fig S6A‴, B‴). Overall, despite the increase in her1 mRNA levels, accumulated transcripts in pnrc2 mutants are exported from the nucleus, consistent with our hypothesis that Pnrc2 promotes decay of cyclic mRNA, a process that likely employs cytoplasmic decay machinery (Garneau et al, 2007; Schoenberg and Maquat, 2012).

Having established that her1-Venus reporter mRNA, but not protein, is misexpressed in pnrc2 morphants, we next asked whether discordant expression occurs for other cyclic transcripts and proteins. Siblings from a pnrc2^oz22 heterozygote intercross were split and processed in parallel for mRNA and protein expression analysis. While pnrc2^oz22 homozygous mutants were readily distinguished from unaffected siblings based on dlc and dld mRNA misexpression (Fig 7G–J), there were no discernible differences in Dlc and Dld protein expression among sibling embryos (Fig 7K–N). Overall, we provide evidence that despite the accumulation of cyclic transcripts in pnrc2^oz22 mutants, cyclic protein expression is unaffected, which may help explain the absence of obvious segmentation phenotypes.

DISCUSSION

In this work, we present evidence that Pnrc2 promotes the decay of cyclic mRNAs and that the her1 3′UTR is sufficient to trigger Pnrc2-mediated decay. Our work builds upon previous studies of Pnrc2-regulated mRNA decay and reveals natural, developmentally-regulated transcripts that are targets of Pnrc2-mediated decay. Zebrafish Pnrc2 shares significant similarity with human PNRC2. Residues in human PNRC2 that are required for decay of reporter mRNA containing a premature termination codon (PTC) are conserved in zebrafish, and include K109 (K119 in zebrafish), W114 (W124 in zebrafish), and the C-terminal KTLLK nuclear receptor domain (NR box) (KSLLK in zebrafish) (Fig 1F) (Cho et al, 2009; Lai et al, 2012). PNRC2 does not bind mRNA directly in human cultured cells, but instead functions via interactions with decay factors SMG6, DCP1A, UPF1, and STAU1 to regulate decay of PTC-containing reporter transcript (Cho et al, 2009; Cho et al, 2012; Cho et al, 2013a; Cho et al, 2013b; Cho et al, 2015; Lai et al, 2012). Our work shows that in segmenting zebrafish embryos, pnrc2 and upf1 may also interact to promote decay of natural, developmentally-regulated transcripts. Future identification of factors that participate in the Pnrc2-mediated decay process will define this key aspect of oscillatory gene regulation and to what extent it employs existing, well-characterized RNA decay machinery.

Cyclic transcript accumulation is due to loss of Pnrc2 in tortuga and pnrc2 mutants

Because the tor^b644 allele is a multi-gene deficiency, we generated a nonsense pnrc2 allele. As expected, pnrc2^oz22 mutants show misexpression of cyclic transcripts including the cyclic transcripts her1, her7, and dlc, as well as dld (Figs 2–3), and in the cases of her1 and dlc, accumulate post-transcriptionally (Fig S2), all consistent with previous observations in tor^b644 mutants (Dill et al, 2005). Because an early frame shift pnrc2 mutation has the same impact on her1 post-transcriptional regulation as complete deletion of the pnrc2 locus, it is unlikely that accumulation of her1 transcripts is due to loss or mutation of miRNA- or lncRNA-encoding genes at or near the pnrc2 locus, but is instead due to loss of Pnrc2 protein.

Pnrc2 translation may be developmentally regulated

pnrc2 transcripts are maternally provided and broadly expressed (Fig 4), however pnrc2 may be subject to translational regulation. Using global ribosomal profiling experiments in developing zebrafish embryos, Giraldez and colleagues showed that the pnrc2 transcript contains a conserved upstream open reading frame (uORF) that lies out of frame with the Pnrc2-coding ORF and is preferentially associated with translational machinery during early embryonic stages (Lee and Bonneau et al, 2013). The presence of two ORFs, an uORF and the Pnrc2-coding ORF, suggests that Pnrc2 translation may be affected by translation at the uORF, as has been demonstrated in other organisms and contexts (Gebauer and Hentze, 2004; Hood et al, 2009; Medenbach et al, 2011; Morris and Geballe, 2000; Sachs and Geballe, 2006; Somers et al, 2013; Wethmar 2014). An uORF can also act as an NMD-inducing feature (Barbosa et al, 2013; Peccarelli and Kebaara, 2014), adding another layer of possible post-transcriptional regulation. It is also interesting that transcripts encoding several NMD factors, like UPF2 and SMG factors, themselves contain uORFs that confer post-transcriptional regulation via feedback control (Popp and Maquat, 2013). Future development of antibodies for detection of zebrafish Pnrc2 and other decay factors will help to advance our understanding of Pnrc2-mediated decay in a developmental context.

Expression patterns of cyclic proteins appear normal despite accumulated cyclic transcripts in pnrc2-deficient embryos

It is remarkable that the dramatic accumulation of Pnrc2-regulated cyclic transcripts like her1 in tor and pnrc2 mutant embryos does not confer phenotypes typically associated with excess Her1, like somite boundary defects and down-regulation of Her1 transcriptional targets including dlc and her7 (Takke and Campos-Ortega, 1999; Giudicelli et al, 2007). Similarly, accumulation of dlc transcripts in tor and pnrc2 mutants does not confer somite boundary defects associated with excess Dlc protein (Soza-Ried, 2014). We find that pnrc2 mutants lack such phenotypes and hypothesize that this is due to normal cyclic protein expression despite mRNA accumulation. It remains unclear why accumulated her1 transcripts in pnrc2 mutants fail to produce defects associated with excess Her1, particularly because previous her1 overexpression experiments included the full length her1 3′UTR (Takke and Campos-Ortega, 1999; Giudicelli et al, 2007) that we now show confers Pnrc2-dependent instability. Overall, because Her1-Venus reporter protein and endogenous Dlc and Dld proteins are expressed normally despite misexpression of mRNA in pnrc2-deficient embryos (Fig 7; Fig S4; Table S3), we suggest that Pnrc2-mediated mRNA decay acts in addition to a regulatory layer that controls oscillatory protein expression. Although the her1 transcript does not contain obvious NMD-inducing features such as an uORF and accumulated her1 transcripts in pnrc2 mutants do not appear retained in nuclei (Fig 7F′; Fig S6B″), there are other mechanisms that may function to repress or prevent translation (Mishima et al, 2012; Takeda et al, 2009; Yasuda et al, 2010; Takahashi et al, 2014; Subtelny et al, 2014; Mishima and Tomari, 2016). Alternatively, normal cyclic protein expression in pnrc2 mutants might occur because protein decay machinery compensates for increased cyclic expression. Treatment of cultured mouse cells with ubiquitin-proteasome inhibitors stabilizes Hes1 protein (closely related to zebrafish Her1) (Hirata, 2002) and mathematical modeling predicts short half-lives for zebrafish Her1 and Her7 proteins (Lewis, 2003). More recently, it has been shown that translational fusion of Her1 to Venus reporter protein confers protein instability (Delaune et al, 2012) and measurements of Her7 protein half-life reveal rapid turnover of Her7 protein within 3.5 minutes (Ay et al, 2013). Future investigation into destabilizing features that trigger rapid protein decay might provide insight into the contribution of protein decay in maintaining oscillatory expression.

Potential cis-regulatory elements reside in the her1 3′UTR

Our data show that the her1 3′UTR is sufficient to trigger Pnrc2-dependent decay of reporter mRNA. It is well known that 3′UTR sequences can influence mRNA stability by length-dependent, sequence, and/or structural feature recognition (Ghosh and Jacobson, 2010; Chen and Shyu, 2011; Schoenberg and Maquat, 2012). We favor the latter two possibilities because the her1 3′UTR is of average length with respect to the post-gastrulation zebrafish transcriptome (Li et al, 2012; Ulitsky et al, 2012; Mishima and Tomari, 2016) and analysis of the her1 3′UTR using the mfold Web Server (Zuker, 2003) predicts a stable stem-loop structure (data not shown). Future biochemical methods of structure probing coupled with deletion reporter assays will help determine if mfold-predicted structures are functionally relevant.

mRNA decay pathways include AU-rich element (ARE)-mediated (von Roretz and Gallouzi, 2008), Staufen1 (Stau1)-mediated (Park and Maquat, 2013), nonsense-mediated (Schoenberg and Maquat, 2012), and miRNA-mediated (Bazzini et al, 2012; Giraldez, 2010; Mishima and Tomari, 2016; Mishima et al, 2006; Jonas and Izaurralde, 2015) decay. Normal her1 expression in MZdicer mutants (Fig 5G–H) suggests that miRNA-mediated decay does not contribute to her1 oscillatory expression despite the presence of numerous predicted miRNA 3′UTR sites (Table S2), although we have not ruled out the possibility that Dicer-independent miRNAs contribute to her1 mRNA decay. Additional motif searches of the her1 3′UTR using AREsite2 (Fallmann et al, 2016) yield four ARE core motif ATTTA sequences in addition to numerous other ARE-related motifs for her1 and other cyclic 3′UTRs that might promote mRNA decay (data not shown). Experiments using deletion reporter assays are underway that will functionally test the role of these and other potential regulatory elements.

We favor nonsense mediated decay (NMD) and/or Stau1-mediated decay (SMD), both of which are Upf1-dependent, as the most likely pathway(s) utilized to clear natural cyclic transcripts during segmentation. Upf1 and Pnrc2 promote decay of cyclic transcripts during segmentation (Fig 6; Fig S4; Table S3) and others have shown that human PNRC2 acts as a decay adapter that physically interacts with decay factors DCP1 and UPF1 and can promote decapping activity of Dcp2 in vitro (Cho et al, 2009; Cho et al, 2012; Cho et al, 2013a; Cho et al, 2013b; Cho et al, 2015; Lai et al, 2012; Mugridge et al, 2016). Transcripts annotated in zebrafish EST and GenBank databases (GRCv10) for her1, dlc, and dld lack canonical NMD-inducing features (retained introns, premature termination codons, upstream ORFs, or intron-containing 3′UTRs), although there exists an intron-retained her7 isoform. Global analyses of UPF1-binding sites in murine embryonic stem cells (mESCs) uncovered a large class of UPF1-3′UTR-bound mRNAs that undergo repression by NMD despite lacking canonical NMD-inducing features, although cyclic transcripts were not among those analyzed (Hurt et al, 2013). Because upf1 knockdown alone does not affect her1 expression, but partial knockdown of pnrc2 together with upf1 leads to accumulated her1 mRNA (Fig 6; Fig S4; Table S3 and data not shown), it is unlikely that Upf1 is absolutely required for Pnrc2-mediated decay of cyclic transcripts (although maternally-provided upf1 function may mask such a function). Instead, it may be that Pnrc2 promotes decay through a combination of Upf1-dependent and -independent mechanisms. Alternatively, low levels of Upf1 protein may be sufficient for Pnrc2-mediated cyclic mRNA decay. Substantial depletion of Upf1 protein is achieved with splice-blocking morpholinos, however, Upf1 is faintly detected by immunoblot (Wittkopp et al, 2009) and this may be sufficient for cyclic mRNA clearance when Pnrc2 levels are normal. Future biochemical and genetic interaction studies with known factors of NMD and SMD will further enhance our understanding of mechanisms that drive rapid decay of cyclic transcripts during segmentation.

CONCLUSIONS

Overall, we propose that cyclic mRNA accumulation in tortuga and pnrc2 mutants results from misregulation of 3′UTR-mediated mRNA decay. This decay process occurs independently of Dicer-generated miRNAs and instead employs decay machinery associated with nonsense-mediated decay. pnrc2 mutants have normal segment number and this is likely due to normal expression of cyclic proteins despite misexpression of cyclic transcripts. Future biochemical, molecular, and genetic studies will provide a deeper understanding of the segmentation clock and post-transcriptional mechanisms that regulate oscillatory expression.

Summary statement.

Pnrc2 regulates 3′UTR-mediated decay of segmentation clock transcripts during vertebrate segmentation. Pnrc2 loss results in accumulation of cyclic transcripts, but not cyclic proteins.

Highlights.

• ❖pnrc2 promotes decay of cyclic transcripts during segmentation.
• ❖A cyclic 3′UTR confers Pnrc2-dependent instability to a heterologous transcript.
• ❖Cyclic mRNA decay is Dicer-independent.
• ❖Cyclic mRNA decay likely employs a Pnrc2-Upf1-containing mRNA decay complex.
• ❖Cyclic mRNA accumulates in pnrc2 mutants, but cyclic protein is expressed normally.