Prior studies showed that the RNA binding protein FOX-1 determines sex by repressing xol-1 via a post-transcriptional mechanism, but the molecular basis of this regulation was not established (Carmi and Meyer, 1999; Nicoll et al., 1997; Skipper et al., 1999). We therefore devised an assay to identify regions of xol-1 necessary for repression by FOX-1. Our prior experiments showed that overexpression of FOX-1 by itself is sufficient to repress endogenous xol-1 activity, causing XO embryos to adopt the hermaphrodite sexual fate and die from reduced X-chromosome expression triggered by binding of the DCC to the single X (Nicoll et al., 1997). Hence, our strategy to identify FOX-1 regulatory sites was to assay deletion derivatives of a xol-1 transgene controlled by the native xol-1 promoter for responsiveness to FOX-1 repression in strains lacking the endogenous xol-1 gene.

The wild-type xol-1 transgene included all xol-1 genomic sequences, and an insertion of gfp sequences in frame at the first ATG codon of xol-1. Expression of wild-type transgenes and deletion derivatives was monitored in xol-1(null) mutant strains by functional assays of XOL-1 activity. XX animals are very sensitive to the dose of xol-1, and extra-chromosomal arrays carrying wild-type xol-1 transgenes could only be established using a 20–30-fold lower concentration of xol-1 DNA than typical for routine markers (see Materials and methods). Wild-type transgenes in all seven independent arrays exhibited proper sex-specific regulation: XO animals were rescued from lethality caused by the endogenous xol-1(null) mutation, and XX animals were viable (Figure 2A). The proportion of XX versus XO animals in each line was in agreement with the expected ratio (2:1) from the male-producing mutation him-5(e1490) present in all lines. However, the xol-1(+) transgenes were expressed in XX animals at a somewhat higher level than the endogenous xol-1 gene: the seven XX array lines could only be maintained if both endogenous copies of fox-1 were wild type. The observation that fox-1 mutations kill XX animals carrying wild-type xol-1 transgenes shows the need for stringent xol-1 repression in hermaphrodites by both transcriptional and post-transcriptional mechanisms.

For XO animals with wild-type xol-1 transgenes, excess FOX-1 protein expressed from an integrated array [yIs44(fox-1)] carrying multiple copies of the fox-1(+) gene was lethal (Figure 2A; Nicoll et al., 1997). No XO animals were viable in the seven lines that carried wild-type xol-1 transgenes and expressed high FOX-1 levels. Although GFP fluorescence was XO-specific when produced from wild-type transgenes, it proved to be too insensitive a monitor for changes in xol-1 activity and was not used as part of our subsequent assays.

Deletion derivatives of xol-1 transgenes lacking FOX-1 regulatory sequences are predicted to be insensitive to repression by excess FOX-1, allowing xol-1(null) XO animals to be rescued and fully viable. Deletion derivatives lacking FOX-1 regulatory regions are also expected to kill XX animals or cause visible dosage compensation defects (Dumpy and Egg-laying defective phenotypes) due to lack of repression by FOX-1. The XX lethality and other dosage compensation phenotypes should not be suppressed by excess FOX-1. Thus, the phenotypic consequences of wild-type and deletion-derivative xol-1 transgenes act as sensitive monitors of regulation by FOX-1.

We first assayed the effects on XX and XO animals of xol-1 transgenes with different combinations of intron deletions to define FOX-1 regulatory regions (Figure 2A–E). The deletion derivatives were made by replacing genomic regions of xol-1 with corresponding xol-1 cDNA to remove introns without altering the protein coding sequence. Extra-chromosomal arrays carrying deletion derivatives of xol-1 transgenes were created in him-5; xol-1 strains capable of producing both XX and XO embryos and then crossed into yIs44(fox-1); him-5; xol-1 strains that produce excess FOX-1. The DNA concentrations used for the distinct deletion-bearing transgenes were the same as for xol-1(+) transgenes. Even if a transgene deletion derivative causes XX-specific lethality, the array can be recovered and maintained through array-bearing XO animals.

The three array lines created from xol-1 transgenes lacking introns II–VI (Δ introns II–VI) rescued xol-1(null) XO males, but killed all xol-1(null) XX hermaphrodites, even though the starting DNA concentration was the same as for xol-1(+) arrays (Figure 2B). Hundreds of array-bearing males were maintained for each independent array line through genetic crosses but no array-bearing hermaphrodite progeny survived, indicating that essential FOX-1 regulatory sequences had been deleted. Excess FOX-1 also failed to suppress the XX-specific lethality and failed to kill XO animals, demonstrating the necessity of intronic sequences for xol-1 repression by FOX-1 (Figure 2B).

The essential role of introns in causing the death of XO animals with elevated FOX-1 levels could reflect the specific contribution of intron VI, which undergoes alternative splicing to yield at least three different xol-1 transcript variants (Figure 2F; Rhind et al., 1995). Only the 2.2 kb variant, which lacks intron VI and includes all of exon 7, encodes a functional XOL-1 protein that has full XOL-1 activity essential for male development (Figure 2F). This 2.2 kb variant is both necessary and sufficient for full XOL-1 function in XO embryos (Rhind et al., 1995). It is the most abundant transcript of the three, accumulating to a level 10-fold higher in XO than XX embryos. A 1.5 kb variant is made from the same 5' donor in exon 6 as used for the 2.2 kb variant but a different 3' splice acceptor, one in the 3' UTR. This splicing event eliminates the coding region of exon 7, an essential exon (Rhind et al., 1995). The 1.5 kb variant does not encode xol-1 XO activity. A 2.5 kb variant results from the failure to remove intron VI. In this variant, an in-frame UAA stop codon within intron VI terminates translation prematurely, precluding production of the male-determining protein (Rhind et al., 1995).

Indeed, removal of only the alternatively spliced intron VI from the xol-1 transgene (Δ intron VI) in all three independent lines permitted all XO animals to be viable and to escape lethality caused by high levels of FOX-1 (Figure 2C). Two of the lines caused complete XX-specific lethality that was not suppressed by excess FOX-1; lines were maintained through genetic crosses with array-bearing XO males (Figure 2C). A third line caused milder dosage compensation defects in XX animals that were not suppressed by excess FOX-1 and caused the animals to be sterile, likely reflecting a lower copy number of the transgene and hence less expression. The line had to be maintained through array-bearing XO animals. Thus, intron VI is essential for FOX-1 repression of xol-1.

Further analysis confirmed that introns II–V are dispensable for FOX-1 regulation. All lines carrying transgenes lacking these introns (Δ introns II–V) behaved like lines of the wild-type xol-1 transgene (Figure 2D). XX and XO xol-1 animals in each line were viable in the relative proportion expected for the him-5 mutation, and XO xol-1 males were killed by excess FOX-1.

When intron VI was removed from its normal location in xol-1 and relocated to the 3' UTR without splice junctions, FOX-1 could not repress xol-1 (Figure 2E). In two independent arrays of this transgene derivative, XO xol-1 animals were viable, despite excess FOX-1, and greater than 99% of XX animals were inviable, even with excess FOX-1. The rare viable XX animals were Dpy and Egl. Thus, the presence of intron VI does not simply enable FOX-1 to repress xol-1 by reducing mRNA stability, blocking nuclear transport of xol-1 RNA or preventing translation in the cytoplasm. Furthermore, experiments quantifying levels of all xol-1 RNA splice variants showed that high levels of FOX-1 did not diminish transcription from xol-1 or cause transcript degradation (Figure 2—source data 1). Instead, FOX-1 likely regulates xol-1 pre-mRNA splicing.

If FOX-1 represses xol-1 by promoting alterative splicing, elevated levels of FOX-1 that cause male lethality should change the distribution of xol-1 RNA splice variants. Specifically, the 2.2 kb splice variant that encodes male-determining activity should be reduced, while the inactive 2.5 kb variant with intron VI retention and/or the inactive 1.5 kb variant with partial exon 7 deletion should be increased. Using RNase protection assays and sequence analysis of cloned cDNAs, we show that elevated FOX-1 levels cause precisely these results (Figure 3 and Figure 3—figure supplement 1).

For RNase protection assays, RNAs from two genetically distinct embryo populations were tested for the relative abundance of specific xol-1 splice variants within each population: one RNA from him-5 mixed-stage XX and XO embryos with wild-type FOX-1 levels and one RNA from yIs44(fox-1); him-5 mixed-stage XX and XO embryos with high levels of FOX-1. These RNAs were assayed initially using an act-1 control gene probe to assess the quality and quantity of RNAs (Figure 3—figure supplement 1D). The first set of xol-1 RNase protection assays with these quantified RNAs used a xol-1 antisense probe that spanned the 3' splice junction of intron VI–exon 7 and distinguished the 2.2 kb and 2.5 kb transcripts (Figure 3A). The ratio of 2.2 kb and 2.5 kb splice variants was calculated from transcripts within each RNA population, and ratios from different populations were then compared to assess the change in transcript ratios caused by varying the FOX-1 level. XX and XO him-5 mixed embryo populations with wild-type levels of FOX-1 showed a fourfold accumulation of active 2.2 kb transcript relative to the unspliced 2.5 kb transcript. In contrast, XX and XO populations that overexpressed FOX-1 [yIs44(fox-1); him-5] showed a drastic reduction in accumulation of the 2.2 kb transcript and a corresponding sixfold higher accumulation of the 2.5 kb transcript (Figure 3A). This experiment and repetitions (Figure 3—figure supplement 1A) established that FOX-1 promotes intron VI retention.

Transcript analysis using RT-PCR next showed that intron VI was the only intron retained in the presence of high FOX-1 levels. Using an oligonucleotide primer set that spans all six introns, RT-PCR was performed on cDNA made from the same RNAs as the protection assays. Only two PCR products were observed, one that corresponded to the 2.5 kb transcript and the other to the 2.2 kb transcript (see Materials and methods).

A second set of xol-1 RNase protection assays utilized a different antisense probe that not only distinguished between the 2.2 kb and 2.5 kb transcripts, but also detected the 1.5 kb transcript and all alternatively spliced variants that contained the 3' end of exon 6 (Figure 3—figure supplement 1B). The probe included the 3' end of exon 6 and the 5' end of exon 7. These protection experiments showed a decrease in accumulation of active 2.2 kb transcript relative to inactive 2.5 kb transcript in the presence of excess FOX-1 (2.4 to 1 with low FOX-1 and 1 to 5 with high FOX-1). They also showed a dramatic decrease in 2.2 kb transcripts compared to all spliced variants. With low FOX-1 levels, the 2.2 kb transcript was present at a ratio of 1:3, but with excess FOX-1 its accumulation decreased more than 10-fold, to a ratio of 1:38.

A third xol-1 probe differentiated the inactive 1.5 kb transcript from the combination of 2.2 kb and 2.5 kb transcripts and from all alternatively spliced transcripts that included the 3' end of exon 6 (Figure 3—figure supplement 1C). Excess FOX-1 caused the 1.5 kb transcript to increase in accumulation compared to the 2.2 kb and 2.5 kb transcripts, from a ratio of 1:14 with low FOX-1 to a ratio of 1:4 with high FOX-1. The 1.5 kb transcript also increased in accumulation compared to all splice variants present in excess FOX-1, from a 1:35 ratio to a 1:21 ratio.

This series of protection experiments demonstrated that FOX-1 represses xol-1 by controlling its pre-mRNA splicing, promoting both intron retention and also deletion of exon 7 coding sequences via alternative 3' acceptor site choice. They also reveal that excess FOX-1 causes the production of more splice variants than previously mapped. Abundance of the universal protected fragment from the 3' end of exon 6 (79 nt) relative to the protected fragments from known splice variants (191 nt and 144 nt) (Figure 3—figure supplement 1A) indicates the occurrence of unidentified splice variants caused by excess FOX-1. As a consequence, we took an alternative approach to identify splice variants by synthesizing cDNA from total embryonic RNA made from both the him-5 and the yIs44; him-5 strains and then selectively cloning and sequencing cDNAs that contained the 3' end of exon 6 (Figure 3B).

Sequences of cDNA clones from xol-1 transcripts in low FOX-1 conditions revealed only the expected: properly spliced active 2.2 kb transcripts, inactive 2.5 kb transcripts that retained intron VI, and inactive alternatively spliced 1.5 kb transcripts that deleted part of exon 7 through use of a 3' acceptor site in the 3' UTR (Figure 3B). Clones of cDNAs from high FOX-1 conditions revealed both expected and unexpected transcripts (Figure 3C). As expected, no active 2.2 kb transcripts were found, but splice variants that retained intron VI or deleted exon 7 were found. Unexpectedly, several cDNAs corresponded to xol-1 transcripts that had been trans-spliced to unrelated genes. This splicing involved the 5' donor at the exon 6–intron VI splice junction, permitting inclusion of exon 6, but then utilized an naturally occurring 3' acceptor site at an exon from an unrelated gene on chromosome II, either flcn-1, polyg-1, or K02E7.12, thus achieving accurate trans-splicing. Thus, the process by which FOX-1 enhances intron retention during RNA processing also promotes the use of alternative 3' acceptor sites, causing deletion of exon coding sequences and enabling trans-splicing.

Promiscuous fox-1-mediated trans-splicing that fuses partially spliced xol-1 transcripts onto transcripts from unrelated genes may result from the normal 3' acceptor site in xol-1 exon 7 being unavailable to receive the primed 5' splice donor in exon 6, perhaps because the 3' site is blocked or not yet synthesized. The primed 5' site then fuses with a nearby available 3' acceptor site, whether or not the site is in xol-1 or an unrelated gene. Analysis of recovered promiscuous splicing events revealed xol-1 transcripts spliced not only to transcripts from chromosome II genes, but also, and even more frequently, to transcripts from adjacent genes encoded on transgenic arrays that include xol-1. Although trans-splicing between a common 22 nucleotide SL-1 or SL-2 leader sequence and the 5' end of nascent transcripts has been well documented in C. elegans (Blumenthal, 2012), trans-splicing of exons from two different protein-coding genes has not been reported previously. In contrast, developmentally programmed trans-splicing has been observed in Drosophila for neighboring genes (Büchner et al., 2000; Dorn et al., 2001; Gabler et al., 2005; Horiuchi et al., 2003; Labrador et al., 2001).

To understand the mechanism by which FOX-1 regulates intron splicing, we asked whether intron VI is sufficient to confer FOX-1 repression upon a heterologous transcript (Figure 4A,B). We placed xol-1's 466 bp intron VI into the fifth exon of a lacZ reporter gene driven by the xol-1 promoter and tested whether excess FOX-1 from yIs44(fox-1) could prevent its expression (Figure 4B). The reporter gene also included the 3' UTR from the unc-54 myosin gene to assess whether intron VI by itself, without the 3' UTR from xol-1, can confer FOX-1-mediated repression. Four independently derived lines with extra-chromosomal arrays carrying the Pxol-1::lacZ::intron VI::unc-54 3' UTR reporter exhibited proper sex-specific regulation, expressing β-galactosidase at high levels in XO embryos (Figure 4B, see XX and XO images) and low levels in XX embryos (Figure 4B, see XX image), as did five independently derived lines with control Pxol-1::lacZ::xol-1 3' UTR reporter arrays that lacked intron VI and had a 3' UTR from xol-1 (Figure 4A, see XX vs. XX and XO images). Excess FOX-1 caused a marked decrease in the frequency and intensity of embryonic β-galactosidase expression in all four intron-VI-containing lines (Figure 4B, see yIs44 image) but had no effect on the five lines carrying the control lacZ reporter without intron VI (Figure 4A, see yIs44 image). At least 1000 embryos were scored for each genotype derived from each of the nine independent arrays. Therefore, intron VI by itself is sufficient to confer FOX-1 repression. These results also show that xol-1's 3' UTR is neither necessary nor sufficient for FOX-1 repression.

To determine whether repression of the lacZ reporter occurs by the same mechanism as repression of xol-1, we examined the splicing pattern of the reporter transcripts. cDNA was synthesized from total embryonic RNA made from intron VI-containing lacZ reporters expressed in both him-5 and yIs44(fox-1); him-5 strains. lacZ-specific PCR primer sets were used to clone lacZ transcripts from the two strains and determine the splicing pattern. Clones from XX animals with low FOX-1 levels revealed the same classes of RNA processing events in lacZ transcripts as found from RNA processing events of endogenous xol-1 RNA, consistent with intron VI conferring FOX-1 repression (Figure 4C). They included a class with proper intron VI removal, one with intron VI retention, one with proper intron VI removal plus an additional splice in lacZ sequences corresponding to the DNA region between AgeI and Pvul restriction sites, and one with the correct 5' donor site usage at the lacZ–intron VI junction but an alternative 3' splice acceptor in an exon of an unrelated gene, unc-76 (chromosome V), F58D5.5 (chromosome I), or H05L03 (chromosome X).

Clones from XX animals with high FOX-1 levels revealed five classes of transcripts consistent with transcripts from endogenous xol-1 in the presence of high FOX-1, confirming that intron VI is sufficient for FOX-1 regulation (Figure 4C). The classes included one that had proper intron VI removal, one that retained intron VI, one that used the correct 5' donor at the lacZ–intron VI junction but an alternative 3' splice acceptor in lacZ, one that had aberrant 5' and 3' junctions in lacZ, and one that used the correct 5' donor but a naturally occurring 3' splice acceptor at an exon of an unrelated gene, most commonly unc-76, but also to zen-4 (chromosome IV) and T28F4.4 (chromosome I), an ortholog of human ARMC5. The high frequency of trans-splicing, particularly to unc-76, likely results from both lacZ and unc-76 being transcribed from the same extra-chromosomal array. We conclude that repression by FOX-1 can be conferred in vivo onto a heterologous gene simply by the insertion of intron VI. Moreover, repression occurs by promoting either intron retention or use of alternative 3' acceptor sites, resulting in deletion of exon coding sequences and enabling trans-splicing.

Because intron VI is both necessary and sufficient for xol-1 repression by FOX-1, we developed an in vitro assay to determine whether FOX-1 regulates RNA splicing by binding directly to intron VI. Using purified FOX-1 protein and equimolar amounts of ³²P-labeled RNA probe for full-length intron VI and for intron III, we performed initial cross-linking binding assays to find conditions that could permit highly specific FOX-1 binding. In experiments with increasing concentrations of FOX-1, we found that FOX-1 bound robustly to the intron VI probe, but not to the intron III negative control probe (Figure 5—figure supplement 1A). Also, intron III RNA served as a better nonspecific competitor than tRNA to inhibit nonspecific binding (Figure 5—figure supplement 1A). In competition experiments for which ³²P-labeled intron VI was challenged with increasing concentrations of either cold intron III RNA or cold intron VI RNA, intron III did not compete for FOX-1 binding. In contrast, cold intron VI severely reduced FOX-1 binding to intron VI probe (Figure 5—figure supplement 1B). These results show that FOX-1 binds directly and specifically to intron VI. All subsequent binding experiments were performed in the presence of cold intron III RNA to inhibit nonspecific binding.

Direct binding assays with ³²P-labeled RNA probes to five subregions of intron VI demonstrated specific binding to three, fragments B, C, and E (Figure 5—figure supplement 1C). Competition experiments using intron VI RNA as probe and cold RNA fragments as competitors confirmed that FOX-1 binds to multiple sites in intron VI (Figure 5—figure supplement 1D). Sequence analysis of the FOX-1 binding fragments revealed common RNA sequences within the three fragments, as delineated by B-37, C-35, and E-35 in Figure 5A.

To determine whether these common sequences are responsible for FOX-1 binding, we first used three RNA oligonucleotides of different sizes to fragment B in direct competition experiments against ³²P-labeled fragment B probe and ³²P-labeled intron VI probe (Figure 5B,C). FOX-1 binding to the B probe was eliminated not only by cold B RNA, but also by the cold 45 nt and 37 nt RNA oligonucleotides that covered the entire common fragment B sequence (Figure 5B). Thus, the common sequence is sufficient for FOX-1 binding. In addition, FOX-1 binding to ³²P-intron VI probe was eliminated by both cold intron VI RNA and cold 45 nt and 37 nt fragment B RNA oligonucleotides, indicating the common sequence supports high-affinity binding that might account for all FOX-1 binding to intron VI (Figure 5C,D). Consistent with that interpretation, FOX-1 binding to ³²P-intron VI was eliminated by a cold 35 nt RNA oligonucleotide to fragment C sequence and a cold 35 nt RNA oligonucleotide to fragment E sequence (Figure 5D).

In contrast, a 15 nt RNA oligonucleotide (CAUUUGAUCGUUAUG) from the middle of sequences common to all three FOX-1 binding fragments was incapable of competing for FOX-1 binding to either a fragment B probe or an intron VI probe, indicating that FOX-1 utilizes one or both of the small motifs, GCAUG and GCACG (Figure 5B,C). Both GCAUG and GCACG are in fragments B and C, but only GCACG is in fragment E.

A 25 nt RNA oligonucleotide that includes GCACG and the center of the common sequence in fragment E was sufficient to eliminate binding to an intron VI probe, indicating that GCACG promotes strong FOX-1 binding, and GCAUG is not essential for FOX-1 binding (Figure 5D). This result does not exclude the interesting possibility that GCAUG might substitute for GCACG or enhance binding to RNA that also includes GCACG.

In a final series of competition experiments, we asked whether FOX-1 binding to intron VI utilizes all three separate regions of common sequence. We compared a cold intron VI RNA competitor that lacked the common sequences in fragments B and C (∆37 ∆35) with a cold intron VI RNA competitor that lacked the common sequences in fragments B, C, and E (∆37 ∆35 ∆26) (Figure 5D). While the ∆37 ∆35 intron was a very poor competitor for the intact intron VI probe, the ∆37 ∆35 ∆26 intron was even worse; it lacked the ability to compete (Figure 5D). Thus, the three regions of common sequence contribute to FOX-1 binding in vitro and suggest they might all contribute to FOX-1 repression in vivo. The competition experiments reinforce the model that direct FOX-1 binding to intron VI facilitates intron VI retention and also causes deletion of exon 7 coding sequences by promoting use of an alternative 3' splice acceptor site.

Identification of FOX-1 binding sites in vitro led us to analyze the function of these sites in vivo for regulating xol-1 splicing and to determine the impact of xol-1 splicing regulation on X-signal activity during normal nematode development (Figure 6). Thus far, our experiments identified intron VI as the target of xol-1 splicing regulation in the context of elevated xol-1 expression caused by multiple copies of xol-1. Because elevated xol-1 expression is lethal to XX animals, these results cannot be extrapolated to disclose the full contribution of splicing regulation to X-signal activity during normal embryogenesis. The impact of splicing regulation in vivo can be determined by editing the endogenous xol-1 gene to eliminate sites used in vivo for regulation by FOX-1.

The approach of removing cis-acting sites in xol-1 by CRISPR/Cas9-mediated genome editing confers three additional advantages over mutating the fox-1 gene itself for analyzing pre-mRNA splicing regulation. It cleanly separates the role of FOX-1 in sex determination from its roles in other developmental processes, revealing a more precise understanding for the contribution of alternative splicing regulation to the X signal. Furthermore, eliminating intron VI blocks all RNA binding proteins and accessory factors from participating in xol-1 splicing, potentially revealing a larger role for splicing regulation in communicating the X signal than simply achieved by FOX-1 alone. Lastly, disrupting individual FOX-1 binding sites allows us to determine the specific sites used for splicing regulation and the number of sites needed to convey the effect of a twofold difference in FOX-1 dose between XO and XX embryos to specify sex. Repression through multiple sites has the potential to amplify the small change in FOX-1 concentration between sexes by minimizing aberrant splicing with one dose of fox-1 to allow high xol-1 activity for male development and by increasing aberrant splicing with two doses of fox-1 to allow low xol-1 activity for hermaphrodite development.

We used multiple assays to judge the impact on xol-1 splicing regulation caused by disrupting endogenous FOX-1 binding sites. First, we assessed the viability of XX hermaphrodites carrying xol-1 mutations in FOX-1 regulatory regions. This assay measures the contribution of xol-1 splicing regulation toward X:A signal assessment in the context of full transcriptional repression by other X signal elements, SEX-1 (nuclear hormone receptor) and CEH-39 (homeodomain protein). Second, we assessed the viability of XX xol-1 mutant hermaphrodites in the context of reduced SEX-1 activity, and hence elevated xol-1 transcription, to measure synergy between transcriptional and splicing regulation. This sensitized XSE mutant condition was achieved using RNA interference (RNAi) against sex-1. Third, we assessed the viability of XO xol-1 mutant males in the context of high FOX-1 levels that are sufficient to kill all otherwise wild-type XO animals by causing non-productive alternative splicing. In these XO animals, the single dose of sex-1 and ceh-39 does not repress xol-1 expression. This sensitive assay measures the efficacy of single and multiple wild-type FOX-1 binding sites on splicing regulation under conditions in which FOX-1 levels are not limiting.

In initial experiments, we eliminated the non-productive alternative splicing mode of xol-1 repression using CRISPR/Cas9 editing to fuse exon 6 in frame with exon 7 at the endogenous xol-1 locus (y810) and thereby exclude intron VI from the pre-mRNA. In XO animals, removing intron VI blocked the XO-specific lethality caused by overexpressing FOX-1. Viability of XO males increased from 0% to 89% (p<10⁻⁵) for y810, indicating that splicing regulation was severely disrupted, as predicted (Figure 6A,C).

In XX animals, loss of intron VI did not reduce either the viability of mutant (y810) versus wild-type animals (99% vs. 101%, respectively) or the average brood size per hermaphrodite (240 ± 4 vs. 267 ± 32 embryos) (Figure 6—figure supplement 1A,C). Consistent with this intron VI deletion result, a null mutation of fox-1(y793) created by a Cas9-induced deletion of the entire endogenous gene also resulted in insignificant XX lethality and no reduction in brood size (Figure 6B and Figure 6—figure supplement 1B). However, just as eliminating FOX-1 activity by gene deletion killed XX hermaphrodites sensitized by reduced activity of the XSE transcription repressor SEX-1, eliminating xol-1 intron VI killed all XX hermaphrodites with reduced SEX-1 activity (Figure 6C). XX viability decreased from 57% to 2% (p<10⁻⁵) for deletion of fox-1(y793) and from 57% to 0% (p<10⁻⁵) for deletion of the intron (y810) (Figure 6C and Figure 6—figure supplement 1C). Thus, repression of xol-1 by splicing regulation in XX animals becomes critical primarily in the context of compromised transcriptional repression.

One other FOX-1 family member, the autosomal protein ASD-1, binds GCAUG motifs and controls alternative splicing of other C. elegans developmental regulators (Kuroyanagi et al., 2013; Kuroyanagi et al., 2007; Kuroyanagi et al., 2006). Our genetic evidence suggested it does not regulate xol-1 either in the presence or absence of FOX-1 (Figure 6—source data 1). That possibility could not have been fully eliminated until intron VI mutations removed the opportunity for any RNA binding factors to regulate xol-1 alternative splicing. No greater benefit to XO males or greater detriment to XX hermaphrodites occurred when all potential sources of splicing regulation were eliminated by removing intron VI rather than by deleting fox-1 alone.

To determine whether FOX-1 binding sites identified in vitro by biochemical analysis do indeed function in vivo to regulate xol-1 RNA splicing, we used two strategies to mutate endogenous DNA encoding these sites in intron VI. First, we deleted the 37 bp, 35 bp, and 26 bp regions corresponding to the in vitro FOX-1 binding sites (y805) (Figure 6D). Second, we altered the sequence of all individual putative binding motifs within each binding site (y804) (Figure 6E). Endogenous DNA was edited to convert the RNA motif GCAUG to AUAUA and GCACG to AUACA.

The nucleotide deletions (y805) and substitutions (y804) of all FOX-1 binding sites and motifs, respectively, within intron VI had the same effect as eliminating the intron: nearly complete loss of XX viability in the sensitized sex-1(RNAi) XSE mutant condition and suppression of male lethality caused by FOX-1 overexpression (Figure 6D,E). XX viability with sex-1(RNAi) was reduced from 57% to 0% (p<10⁻⁵) by the binding site deletions, and from 57% to 1% (p<10⁻⁵) by the motif substitutions. Male viability with high FOX-1 levels increased from 0% to 88% (p<10⁻⁵) with the deletions and from 0% to 89% (p<10⁻⁵) with the substitutions. In contrast, the viability and brood sizes of XX animals bearing only deletions or nucleotide substitutions of all FOX-1 binding sites and motifs in a sex-1(+) condition were not different from wild-type XX animals (Figure 6D,E and Figure 6—figure supplement 1D,E). These experiments indicate that FOX-1 binding sites and motifs identified in vitro do function in vivo to mediate xol-1 splicing repression, and confirm that splicing regulation is essential for hermaphrodite viability primarily in the context of reduced transcriptional repression.

Genome editing of fox-1 binding sites also allowed us to determine the efficacy in vivo of GCAUG motifs versus GCACG motifs in splicing-mediated xol-1 repression in XX and XO animals and to determine how many FOX-1 binding sites are required for full splicing regulation. Mutating the three GCACG motifs while retaining the two GCAUG motifs (y808) reduced the viability of sex-1(RNAi) XX animals to about half the level of sex-1(RNAi) XX animals with five wild-type motifs (24% vs. 57%) (p<10⁻⁵), but permitted more XX viability than with mutant versions of all five GCAUG and GCACG motifs (y804) (24% vs. 1%) (p<10⁻⁵) (Figure 6F and Figure 6—figure supplement 1F). These results indicate that GCACG motifs function in vivo for splicing repression in XX animals, but by themselves are not sufficient for full repression; the GCAUG motifs are also required. The two GCAUG motifs remaining in y808 also severely reduced the viability of XO animals with high FOX-1 levels compared to those with five mutant motifs (y804) (8% vs. 89%) (p<10⁻⁵), demonstrating that GCAUG motifs contribute to FOX-1-mediated repression in XO animals (Figure 6F).

Reciprocally, mutating the two GCAUG motifs while retaining the three wild-type GCACG motifs (y809) reduced the viability of XSE-sensitized XX animals by about half compared to those sensitized XX animals having all five wild-type motifs (33% vs. 57%) (p<10⁻⁴), but permitted more XX viability than all five mutant motifs (y804) (33% vs. 1%) (p<10⁻⁵) (Figure 6G and Figure 6—figure supplement 1G). These results indicate that GCAUG motifs function in vivo for splicing repression in XX animals but are not sufficient for full repression; GCACG motifs are also important. The three GCACG motifs remaining in y809 reduced the viability of XO animals with high FOX-1 levels by half compared to XO animals with all mutant motifs (y804) (41% vs. 89%) (p<10⁻⁵) (Figure 6G), indicating that GCACG motifs contribute to splicing regulation in XO animals. Thus, multiple FOX-1 binding sites are required in vivo for xol-1 repression by alternative splicing, and both GCACG and GCAUG motifs are important.

To determine whether GCACG motifs alone or GCAUG motifs alone could be sufficient for full splicing-mediated repression when present at all five sites, we converted all five motifs to either GCACG (y807) or GCAUG (y806) (Figure 6H,I and Figure 6—figure supplement 1H,I). When present in all sites, either GCACG or GCAUG motifs were sufficient to confer complete splicing repression, just like a wild-type intron VI. Virtually all XO animals were killed by high levels of FOX-1 (1% or 3%, respectively, vs. 0%), and the viability of XSE-sensitized XX animals was equivalent to that achieved with a wild-type intron VI (49% or 57%, respectively, vs. 57%) (Figure 6H,I). These results indicate that the number of binding sites is more important than whether the sequences are GCAUG or GCACG.

The significance of motif number in splicing repression is further illustrated by additional experiments. First, XX viability was compared directly between edited animals that have only GCAUG motifs but differ in GCAUG number. Viability of sex-1(RNAi) XX animals with five GCAUG motifs (y806) was 57%, but viability with two GCAUG motifs was 24% (p<10⁻⁴) (Figure 6I and Figure 6—figure supplement 1I). Second, viability of sex-1(RNAi) XX animals was relatively normal with four of five wild-type motifs (y811) (52% vs. 57%) (Figure 6J and Figure 6—figure supplement 1J). However, viability of sex-1(RNAi) XX animals was reduced when xol-1 had just two or three wild-type motifs: 24% (p<10⁻⁵) for two GCAUG motifs in y808% and 33% (p<10⁻⁴) for three GCACG motifs in y809 (Figure 6F,G). Third, mutation of a single motif, the terminal GCACG motif (y811), was sufficient to reduce splicing-mediated repression of xol-1 in XO animals by high FOX-1 levels (Figure 6J). More XO animals with excess FOX-1 were viable with only four of five wild-type motifs (y811) than with all five (36% vs. 0%). Fourth, in a reciprocal experiment, a single GCACG motif (y803) was sufficient to decrease viability of XO animals with high FOX-1 levels compared to XO animals with no wild-type sites (y804) (22% vs. 89%) (p<10⁻⁵) (Figure 6K). Thus, in the context of high FOX-1 levels, a single GCACG motif functions in splicing repression, but increasing the number of either GCACG or GCAUG motifs causes progressively greater repression and less XO viability.

To further assess the effect of only a single FOX-1 binding site in xol-1, we examined XX animals in which only the terminal GCACG motif (y803) was present (Figure 6K and Figure 6—figure supplement 1K). We found severe XX lethality in the XSE-sensitized background (6% viable), indicating that the single GCACG site by itself is not sufficient for robust regulation in XX animals. Multiple sites are required, as exemplified from the increased viability of sex-1(RNAi) XX animals with three GCACG motifs (33%, p<10⁻⁵) (Figure 6G) and five GCACG motifs (49%, p=0.01) (Figure 6H).

The fact that 6% of XX animals were viable with just the single site (y803) instead of 1% (p=0.002) when all sites were altered (y804) indicates that the GCACG site by itself permits sufficient alternative splicing to rescue some XX animals (Figure 6E,K). Utility of the single site is also reflected in the brood size of these XSE-sensitized animals. Hermaphrodites with one GCACG site had an average brood size of 100 ± 25 embryos, while those with no sites had an average brood size of 11 ± 4 embryos (Figure 6—figure supplement 1E,K). Consistent with the efficacy of a single GCACG site, viability of XO animals in the presence of high FOX-1 levels was reduced from 89% when all sites were absent (y804) to only 22% (p<10⁻⁵) when a single GCACG site (y803) was present (Figure 6E,K). Thus, while one binding site can achieve some repression, multiple sites are needed for full repression, thereby creating a sensitive mechanism for xol-1 regulation by FOX-1.

No GCAUG or GCACG motifs are present in any xol-1 intron other than intron VI, consistent with intron VI being sufficient for FOX-1 regulation. Two GCACG motifs are present in the 3' UTR of the 2.2 kb transcript, but this 3' UTR is not necessary for repression by FOX-1 (Figure 4B).

Recent studies demonstrated that mammalian Rbfox can utilize secondary, low-affinity binding motifs such as GCUUG to promote alternative splicing in vivo if these motifs are present in multiple copies, and if Rbfox reaches a concentration higher than necessary for binding to GCAUG motifs (Begg et al., 2020). Therefore, we tested whether replacing all GCACG and GCAUG motifs in intron VI with GCUUG motifs would promote sufficient non-productive splicing in the presence of high FOX-1 levels to cause the death of XO males and, reciprocally, the survival of XX embryos exposed to sex-1(RNAi). In XO animals with high FOX-1 levels, the low-affinity GCUUG binding sites were sufficient to cause complete XO lethality (y820 in Figure 6L).

In XX animals with wild-type GCUCG and GCACG motifs in intron VI, high levels of FOX-1 could suppress the death caused by sex-1(RNAi) (24% vs. 63% viable p=0.004) (Figure 6—figure supplement 2A,B), demonstrating that enhancing post-transcriptional repression of xol-1 can compensate for the deficiency in repression caused by reducing transcriptional repression. Although high FOX-1 levels in XX animals with mutated GCACG and GCAUG motifs could not suppress any death caused by sex-1(RNAi) (Figure 6—figure supplement 2D), high FOX-1 levels could rescue XX animals with low-affinity GCUUG motifs from sex-1(RNAi)-induced death (p=0.006) (Figure 6—figure supplement 2C), as could high FOX-1 levels with normal GCAUG and GCACG motifs.

In contrast, GCUUG motifs were not adequate to suppress the lethality caused by sex-1(RNAi) when only the two wild-type doses of fox-1 were present in XX embryos. Only 8% of y820 sex-1(RNAi) XX animals carrying GCUUG motifs survived compared to 57% survival when all sites had the original high-affinity GCAUG or GCACG motifs (p<10⁻³) (Figure 6L and Figure 6—figure supplement 1L). These results show that although multiple low-affinity GCUUG binding sites, coupled with high FOX-1 levels, are sufficient to kill XO animals or to suppress the death of sex-1(RNAi) XX animals, the low-affinity GCUUG sites are inadequate to repress xol-1 in XX embryos with only the normal two doses of fox-1. The degree of non-productive splicing necessary to repress xol-1 in XX embryos with two copies of fox-1 is only reached if multiple high-affinity motifs are present in intron VI.

The need for multiple high-affinity binding sites to enable two doses of FOX-1 to repress xol-1 in XX animals suggested that xol-1 splicing control would be sensitive to FOX-1 dose. We took two approaches to evaluate the dose-dependence of FOX-1 action in determining sex. In XX animals with reduced sex-1 activity caused by RNAi directed against sex-1, we compared the impact on xol-1 regulation of reducing the dose of FOX-1 from two copies to one to the impact of mutating combinations of FOX-1 binding sites in one endogenous copy of xol-1. Both approaches revealed dose-sensitivity of FOX-1 action.

Viability of the sex-1(RNAi) XX animals declined more than 10-fold, from 44% to 3% (p=0.002), when the dose of fox-1 was reduced by half, from two copies to one copy, demonstrating dose-dependence of FOX-1 function in XX animals (Figure 7A,B). Similarly, viability of sex-1(RNAi) XX animals declined from 44% to 7% (p=0.006) when all FOX-1 binding sites in intron VI were mutated in one copy of xol-1 (Figure 7A,C). As expected, viability of sex-1(RNAi) XX animals declined to an intermediate level when either the three GCACG motifs (18%) (p=0.018) or the two GCAUG motifs (13%) (p=0.012) were mutated in one copy of xol-1 (Figure 7A,D,E). Viability of sex-1(RNAi) XX animals with five low-affinity GCUUG motifs in one copy of xol-1 was better (36%) than that with either two GCAUG (p=0.023) or three GCACG (p=0.017) mutated motifs in one xol-1 copy (Figure 7F). Hence, the dose-dependence of FOX-1 function in regulating alternative xol-1 splicing in XX animals is evident both from reducing the dose of the trans-acting FOX-1 protein or by mutating different combinations of cis-acting FOX-1 binding sites in only one copy of xol-1. Thus, FOX-1 acts as an XSE to convey X-chromosome number by regulating alternative xol-1 splicing in a dose-dependent manner.

Multiple high-affinity FOX-1 binding motifs are necessary to restrict the non-productive mode xol-1 splicing to XX embryos over the small FOX-1 concentration range that distinguishes XX from XO embryos. Among numerous experiments, the need for multiple high-affinity RNA binding sites to repress xol-1 was well exemplified by the comparison in XX viability of engineered xol-1 strains that carried only GCACG motifs in intron VI, but in different numbers, and were subjected to sex-1(RNAi). The viability of the XX strain with five GCACG motifs (49%) was greater than that with three GCACG motifs (33%) or with one motif (6%). Moreover, replacing all GCUAG and GCACG motifs with low-affinity GCUUG motifs permitted only minimal survival (8% vs. 57%).

In other cases of splicing regulation by C. elegans FOX-1, specifically the unc-32 and egl-15 gene targets, robust regulation is achieved through a single GCAUG binding site using a different strategy from the one for xol-1. FOX-1 binds the unc-32 pre-mRNA at the single GCAUG site in concert with the neuronally expressed CELF RNA binding protein UNC-75 to promote skipping of the upstream exon (Kuroyanagi et al., 2013). To do so, FOX-1 acts redundantly with ASD-1, another FOX-1 family member. Either FOX-1 or ASD-1 can bind the GCAUG site with UNC-75, and together regulate unc-32. Similarly, FOX-1 and ASD-1 regulate splicing of egl-15 pre-mRNA in combination with the muscle-specific RNA binding protein SUP-12 using a single GCAUG site (Kuroyanagi et al., 2007). For both gene targets, two FOX-1 family members bind a single binding site to help ensure proper splicing in conjunction with key tissue-specific RNA binding proteins. Loss of either fox-1 or asd-1 activity causes partial loss-of-function phenotypes for egl-15 and for unc-32, revealing that asd-1 and fox-1 function in a collaborative fashion to reinforce splicing control of two different pre-mRNA targets, each with one binding site (Kuroyanagi et al., 2013; Kuroyanagi et al., 2007). In contrast, our results showed that ASD-1 is not required for regulation of alternative xol-1 splicing to repress xol-1 activity in XX embryos.

For many mammalian genes, Rbfox can bind to an intron and control pre-mRNA splicing using a single high-affinity GCAUG motif. A tyrosine-rich, low-complexity domain at the C-terminus then nucleates aggregation of Rbfox to attain the concentration of bound proteins necessary to drive alternative splicing (Ying et al., 2017). Rbfox aggregation is essential for inclusion of specific exons into mature mRNA and has the potential to recruit additional splicing factors. Rbfox aggregation may also be necessary for concentrating Rbfox complexes in nuclear speckles, where RNA synthesis occurs (Ying et al., 2017). Such concentrated localization may promote binding to and alternative splicing of newly synthesized RNAs. C. elegans FOX-1 lacks a tyrosine-rich low-complexity domain to cause protein aggregation, but during xol-1 regulation, recruitment of FOX-1 through multiple intronic binding sites using GCAUG and GCACG motifs appears to substitute for protein aggregation in achieving robust alternative splicing. In this case, availability of multiple sites would increase the probability of any site being occupied.

Genome-wide binding studies of Rbfox revealed that many mammalian introns have more than one primary GCAUG or GCACG motif (Begg et al., 2020; Lambert et al., 2014). While experiments have not yet addressed whether the multiple motifs enhance splicing regulation at these endogenous sites, other experiments using mammalian reporter constructs have demonstrated that three GCAUG motifs are more effective at regulating splicing than a single motif (Modafferi and Black, 1999).

The importance of multiple motifs in mammalian pre-mRNA splicing regulation is further exemplified by a recent study showing that a cluster of six Rbfox low-affinity secondary binding motifs, such as GCUUG or GUUUG, but not a single low-affinity motif, controls splicing regulation during neuronal differentiation in vivo, as the Rbfox concentration naturally increases during development (Begg et al., 2020). Higher Rbfox concentrations are required for splicing regulation using the clustered low-affinity sites than those required for splicing regulation by a single high-affinity GCAUG motif. Hence, the clustered motifs create tissue and temporal specificity in splicing regulation in response to changes in Rbfox concentration during development (Begg et al., 2020).

A cluster of intronic xol-1 motifs could tune splicing regulation over the twofold increase in FOX-1 concentration between XO and XX embryos, as Rbfox appears to do during neuronal differentiation (Begg et al., 2020). The cluster of xol-1 motifs likely enhances the probability that the higher FOX-1 concentration in XX embryos versus XO embryos reaches the threshold necessary for FOX-1 binding and function, thereby restricting splicing regulation to XX embryos. Our studies showed that a single FOX-1 binding site becomes sufficient to induce some non-productive xol-1 splicing in XO embryos and thereby kill them when the concentration of FOX-1 is increased artificially by integrating numerous extra copies of fox-1 into the genome. However, multiple high-affinity motifs are necessary to achieve the non-productive mode of xol-1 splicing to XX embryos. Even for xol-1, though, secondary GCUUG motifs can achieve splicing regulation if the FOX-1 concentration is elevated beyond the normal level in XX embryos. Five GCUUG motifs are sufficient to permit high FOX-1 levels to suppress the XX-specific lethality caused by sex-1(RNAi) or to kill all XO embryos. These results converge with and extend those found for mammalian Rbfox.

Like C. elegans, the fruit fly D. melanogaster utilizes the combination of transcriptional and pre-mRNA splicing regulation to enhance the precision of X:A counting by controlling Sxl (Sex-lethal), the master sex-determination switch gene and direct target of the X:A signal. This sex switch gene dictates female development when active and permits male development when inactive (Cline and Meyer, 1996). Sxl encodes an X-linked RNA binding protein with an RRM. It directs female development by regulating alternative pre-mRNA splicing of itself and downstream sex-determination genes. Sxl is activated in 2X:2A embryos, but not 1X:2A embryos, by a set of feminizing XSEs that stimulate transcription of Sxl in a dose-dependent manner. Once produced, SXL protein functions in a positive autoregulatory loop to control splicing of its own pre-mRNA in a dose-dependent manner and thereby promote continued production of female-specific SXL protein (Bell et al., 1991; Cline and Meyer, 1996; Horabin and Schedl, 1993; Sakamoto et al., 1992). By binding to two neighboring Sxl introns, SXL prevents inclusion of the intervening male-specific exon 3, which encodes an in-frame stop codon that prevents translation of the full-length female SXL protein when incorporated into mature RNA.

SXL also directs pre-mRNA splicing of its downstream sex-determination target gene called tra (transformer), a switch gene that directs female sexual differentiation (Inoue et al., 1990; Sosnowski et al., 1989; Valcárcel et al., 1993). SXL controls tra by regulating 3' splice site selection. By binding to an upstream 3' splice site in the first intron of tra pre-mRNA, SXL competes with binding of splicing factor U2AF and thereby diverts splicing to a distal 3' splice acceptor site in that intron. The resulting shorter RNA isoform encodes a full-length TRA protein.

Thus, SXL dictates female development by directing formation of RNA isoforms for itself and downstream targets that encode essential female-specific proteins. In contrast, FOX-1 promotes hermaphrodite development by dose-dependent binding to intron VI of the male-determining gene xol-1, causing either intron retention or deletion of exon coding sequences, and thereby blocking formation of the RNA variant that encodes male-specific XOL-1 protein. While fox-1 acts as a worm XSE to communicate X-chromosome dose, Sxl responds to fly XSEs to determine sex, but its location on X and its autoregulatory feature allow it to serve as both signal and target.

The parental transgene plasmid (pMN45) that Nicoll et al., 1997 used to construct deletion derivatives diagrammed in Figure 2A is a genomic xol-1 rescuing clone that has gfp coding sequences inserted at the first ATG of xol-1. In extra-chromosomal arrays made by injecting 5 μg/ml plasmid DNA, the multi-copy transgenes produce a bifunctional GFP::XOL-1 protein that rescues the lethality of xol-1 XO null mutants and expresses GFP in a sex-specific manner, high in XO and low in XX embryos. Although arrays of xol-1(+) transgenes from 5 μg/ml injections exhibited sex-specific regulation, the levels of XOL-1 were higher in XX animals than those produced by two copies of the endogenous gene. This elevated level of xol-1 expression made XX animals dependent on endogenous FOX-1 activity to repress the transgenes and achieve full XX viability. Because xol-1(+) transgenic arrays cause XX lethality when fox-1 is inactive, reporters with deletion derivations that eliminate nucleotides required for FOX-1 regulation are expected to cause XX lethality or at least milder dosage compensation phenotypes such as a dumpy (Dpy) body shape and an egg-laying defect (Egl). Prior experiments showed that excess FOX-1 is unable to repress a translational xol-1::gfp reporter that lacks sequences downstream of the first two introns and 89 codons of xol-1, allowing us to keep intron I in the reporters. Because we found that the phenotypic analysis of XOL-1 activity (Dead, Viable, Dpy Egl) was a much more sensitive indicator of xol-1 activity than GFP fluorescence, we reported only the phenotypic analysis. As an example of the difference in sensitivity, we found that derepression of the gfp::xol-1 transgene caused XX-specific lethality and/or a Dpy phenotype even for cases in which GFP fluorescence did not appear to increase as measured by a fluorescence dissecting microscope.

Arrays of xol-1(+) transgenes could only be established using a xol-1(+) DNA injection concentration at 5 μg/ml. Any higher xol-1(+) DNA injection concentration killed XX animals, and extra-chromosomal lines could not be established. All deletion derivatives of xol-1 transgenes were also injected with a DNA concentration of 5 μg/ml. The xol-1 plasmids were co-injected with 150 μg/ml of unc-76(+) marker plasmid p76-16b into him-5(e1490) unc-76(e911); xol-1(y9) hermaphrodites. him-5 unc-76; xol-1 hermaphrodites produce 67% viable Unc XX hermaphrodites and 33% Unc XO males that die as embryos or L1 larvae in the absence of xol-1(+) from a transgenic array. Extra-chromosomal arrays have variable transmission through meiosis and can only be followed when the arrays carry DNA encoding a wild-type genetic maker, such as unc-76(+), and they are established in animals carrying an unc-76 mutation, which causes the animals to be Unc. An array resulting in only Unc hermaphrodites and both Unc and non-Unc males indicate the transgenic array causes XX-specific lethality. Lines were maintained through non-Unc animals in each generation either by cloning non-Unc hermaphrodites when the arrays permitted XX viability or crossing non-Unc males into him-5 unc-76; xol-1 hermaphrodites and recovering non-Unc XO progeny bearing the arrays, if the arrays caused XX lethality.

Once array lines were made with xol-1 transgene deletion derivatives in the him-5 unc-76; xol-1 strain, they were crossed into yIs44(fox-1); him-5 unc-76; xol-1 hermaphrodites. yIs44 has multiple copies of the fox-1-containing cosmid R04B3 and the rol-6 plasmid pRF4, which causes the array animals to have a roller phenotype. xol-1 activity was scored by the presence of non-Unc males, an indication of rescue of the XO-specific lethality caused by the xol-1(y9) mutation. DNA sequences and details of transgene constructions are available upon request for pMN66 (Δ introns II-VI), pMN67 (Δ intron VI), pMN60 (Δ introns II-V), and pCP5 (intron VI in 3' UTR).

The lacZ reporters Pxol-1::lacZ::xol-1 3'UTR (pMN27) and Pxol-1::lacZ::intron VI::unc-54 3' UTR (pMN110) were co-injected at 20 μg/ml with 150 μg/ml of unc-76(+) marker plasmid p76-16b into him-5(e1490) unc-76(e911) hermaphrodites. Array lines were crossed into yIs44(fox-1); him-5 unc-76 hermaphrodites. pMN27 was made from the promoterless lacZ vector pRD95.11 from A. Fire. pMN110 was made by inserting intron VI into pMN21, a Pxol-1::lacZ transcriptional fusion made from the intron-rich lacZ plasmid pPD95.03 from A. Fire. DNA sequences and details of reporter transgene constructions are available upon request. Embryonic β-galactosidase expression was assayed as previously described using 4 μl of 4% 5-bromo-4-chloro-3-indolyl-D-galactopyranoside staining solution and incubation overnight at room temperature (Nicoll et al., 1997). More than 1000 embryos of each genotype were scored for β-galactosidase expression.

Total RNA was isolated from 0.2 to 0.5 ml of washed, packed embryos that had been stored at −80°C. Frozen embryos were ground to a very fine powder using a liquid-nitrogen-cooled mortar and pestle, adding more liquid nitrogen as needed to prevent the material from thawing. Ground, frozen material was vortexed in two microfuge tubes, each with 1 ml Trizol, for 30 s to 1 min and then incubated for 5 min at room temperature. Chloroform (0.2 ml) was added to each tube, vortexed for 30 s, and incubated at room temperature for 2–3 min. After centrifugation at 13,000 × g for 5 min at 4°C, the upper aqueous phases (0.75 ml/tube) were transferred to a new tube, and 0.5 ml of the material was distributed into each of three tubes. An equal volume of phenol was added to each tube, vortexed, and 0.2 ml CHCl₃ was added and vortexed. Following centrifugation at 13,000 × g for 5 min at 4°C, the aqueous phases were transferred to new tubes, and an equal volume of CHCl₃ was added and vortexed. Following a 13,000 × g spin in a microfuge for 5 min at 4°C, the aqueous phases were transferred to new tubes and incubated a room temperature with one volume of 2-propanol for 10 min. Following centrifugation at 13,000 × g spin for 10 min at 4°C, the supernatants were removed, and the pellets were washed with 70% ethanol and dried in a Spin-Vac. Each pellet from the three tubes was resuspended in 100 μl water, and all volumes were combined into one tube (~0.3 ml total). One volume of 4 M LiCl was added and incubated overnight at 4°C. After a 13,000 × g microfuge spin for 5 min at 4°C, the aqueous phase was transferred to a new tube. The RNA was incubated with 0.1× volume of 3 M NaOAc and 2.5× volume of ethanol on ice for more than 10 min. After centrifugation at 13,000 × g for 10 min at 4°C, the aqueous phase was removed, the pellet was washed with 70% ethanol, and dried in a Spin-Vac. The RNA was resuspended in 100 μl water, and the O.D. was read after diluting a sample 250-fold.

Quantitative RT-PCR was used to measure transcript levels in RNA isolated from three independent growths of the four strains listed in Figure 2—source data 1 as previously described (Gladden and Meyer, 2007; Van Gilst et al., 2005). Three different sets of RT-PCR experiments were performed with each of the three independent growths of worms. Worms were grown on egg plates (http://www.wormbook.org) prior to isolating the mixed-stage embryos, and the total RNA was isolated as described above. A complete list of primer sequences is in Supplementary file 1. The xol-1 primers were designed to measure all splice variants simultaneously, not just the active ones.

Transcript levels for xol-1 and nhr-64 were normalized to the transcript level of fasn-1, which is expressed constitutively throughout embryogenesis, by adjusting the Ct (cycle threshold) value of fasn-1 measured in each strain to equal the Ct value of fasn-1 measured in the control embryos, either him-5 or wild type. This adjustment equalizes the small variations in concentration of the starting material added to each PCR reaction from different RNA preparations. The transcript level of each mutant strain was then expressed as fold change relative to the control embryos, either him-5 or wild-type (ΔCt). The normalized Ct value for each transcript measured in each strain was subtracted from the normalized Ct value of the same transcript measured in control embryos. The difference between these values corresponds to the change in transcript levels relative to those in control animals. Ct values are expressed as PCR cycle numbers. Each PCR cycle increases the concentration of the template by twofold. Therefore, to convert the difference in Ct values to a relative change in concentration, the expression 2^ΔCt was used. The same protocol was used to measure xol-1 and fasn-1 transcript levels in the four strains of Figure 2—source data 1 when they were normalized to the nhr-64 transcript.

Riboprobes were made from linearized pBluescript SK plasmids in which DNA encoding the RNA region of interest had been cloned adjacent to the T3 promoter. For a 20 μl T3 RNA polymerase reaction mixture, the following reagents were combined: 4 μl of 5× transcription buffer (Stratagene), 2 μl each of 50 mM ATP, GTP, and CTP, 10 μl of α-³²P-UTP (Amersham, 400 Ci/mmol, 10 μCi/μl), 1 μl linearized DNA (0.4 μg), 1 μl of 0.2 M DTT, 1 μl RNasin (Promega 40 U/μl), and 1 μl T3 RNA Polymerase (Stratagene, 50 U/μl). The reaction was incubated for 60 min at 40°C. The following reagents were added to the transcription reaction and incubated for 15 min at 37°C: 1 μl RNasin, 2.5 μl vanadyl ribonucleoside complex (200 mM), 20 μl water, 6 μl 5× transcription buffer, and 1 μl RNase free DNase I (1 mg/ml). The reaction was extracted with 50 μl phenol. The phenol was back extracted with 50 μl TE. The two aqueous phases were combined and extracted with 100 μl of chloroform:isoamyl alcohol (24:1). Two ethanol precipitation steps were then performed. For the first, the following reagents were added and the mixture was incubated at room temperature for 5 min and then spun: 1 μl glycogen, 66 μl of 5 NH₄Ac, and 450 μl of 100% ethanol. The pellet was resuspended in 100 μl water. For the second, 10 μl of 3 M NaOAc and 250 μl of 100% ethanol were added, and the mixture was spun and dried. The pellet was resuspended in 100 μl water.

RNase protection assays were performed using Ambion's RPA (RNase Protection Assay) III Kit.

To determine how many introns were retained in xol-1 when FOX-1 was at high levels, we performed PCR on cDNA prepared from total RNA isolated as for the RPA. Using primers overlapping the first ATG of xol-1 (5'-gcaggttgaagcaaattctgagagaag-3') and within the 7th exon (5'-cactcttcatcctcatcatacgtg tc-3'), only two bands corresponding to the size of the 2.5 kb and 2.2 kb transcripts were amplified from wild-type, him-5, and yIs44; him-5 animals. The primer set does not detect cDNA from the alternatively spliced 1.5 kb transcript.

To characterize splice variants of xol-1 transcripts made in low and high levels of FOX-1, we synthesized cDNA and then cloned and sequenced xol-1 specific transcripts. To capture full-length transcripts, the first strand of cDNA was made from total embryonic RNA made from him-5 and yIs44(fox-1); him-5 strains using a Clontech Advantage RT-for-PCR Kit and an oligo(dT)₁₈ primer. We then performed two sets of PCR reactions to capture transcripts that included exon 6 of xol-1 and the downstream splice variants. In the first PCR reaction, the xol-1-specific primer in exon 6 (CSNP-8, 5'-GAGTTTGATAGCCAAGTTGCTCTTG-3') was used in combination with a 3' RACE primer with a unique tag on it (5'-aagcagtggtatcaacgcagagTAC(T)₃₀N-1 N-3', where N-1 is A,C, or G and N is A, C, G, or T). In the second PCR reaction, a nested exon 6 primer closer to the 3' end of the exon (CSNP-9, 5'-CATGAGCAAGTAGAAGGTTTCGAAG-3') was used in combination with a primer to the unique tag (5'-aagcagtggtatcaacgcagagT-3'). The PCR products were TOPO cloned (Thermo Fisher Scientific) and sequenced.

To characterize the effect of different FOX-1 levels on splice variants from the Pxol-1- driven lacZ reporter carrying intron VI of xol-1, we made total embryonic RNA from two strains: TY2882, him-5(e1490) unc-76(3911) V; yEx280 and TY3082, yIs44; him-5(e1490) unc-76(3911) V; yEx280. yEx80 is an extra-chromosomal array that carries the unc-76 co-injection marker and the reporter Pxol-1::lacZ::xol-1 intron VI::unc-54 3' UTR (pMN110). The first strand of cDNA was made using a Clontech Advantage RT-for-PCR Kit and the 3' RACE primer with the unique tag from above (5'-aagcagtggtatcaacgcagagTAC(T)₃₀N-1 N-3'). In the first PCR reaction the lacZ-specific primer just before intron VI (CSNP-12, 5'-AATCAGGCCACGGCGCTAATCACG-3') was used in conjunction with a primer to the unique tag (5'-aagcagtggtatcaacgcagagT-3'). In the second PCR reaction, a nested lacZ-specific primer closer to intron VI (CSNP-13, 5'-CGCTGGATCAAATCTGTCGATCC-3') was used in conjunction with primer to the unique tag. PCR products were TOPO cloned and sequenced. In the second PCR, because the sequence from the beginning of intron VI to the poly A tail was too long to pick up transcripts with intron VI properly spliced out in strain TY2882, a third set of PCR reactions was performed using exclusively lacZ sequences that flank intron VI. DNA from the first PCR reaction was used to perform PCR with the nested lacZ-specific primers CSNP-14 (5'-CTAATCACGACGCGCTGTATCG-3') and CSNP-15 (5'-GTCGGCAAAGACCAGACCGTTC-3'). PCR products were TOPO cloned and sequenced. The expected PCR product with no intron VI splicing was 900 bp, and the expected product with correct intron VI splicing was 434 bp.

A cocktail of 2 μl normalized riboprobe, 2 μl of water or RNA competitor, and 1 μl non-competing cold background RNA (intron III, 100 molar excess, 217 ng) was added to 1.5 ml tubes on ice. Riboprobe was used in RNA excess, and intron VI riboprobe was determined empirically to be in excess when 8 × 10⁶ cpm riboprobe was cross-linked with 32 ng FOX-1 protein. Riboprobes of intron VI subregions (A–E) or intron VI with small deletions were cross-linked at the same molar concentration as full-length intron VI (determined by percent of labeled Us in each fragment vs. Us in intron VI). A cocktail of the following reagents was added to the chilled tubes: 11 μl of cross-linking buffer [20 mM HEPES (pH 7.9), 20% glycerol, 100 mM KCl, 0.2 mM EDTA, 0.2 mM PMSF, 0.5 mM DTT, 0.091 ng/μl BSA], 2 μl of 20 mg/ml heparin, and 2 μl of 16 ng/μl FOX-1 protein (or a series of FOX-1 concentrations). The 20 μl reactions were incubated for 25 min in a room-temperature water bath. The tubes were placed in an aluminum block pre-chilled on ice and then UV irradiated for 12 min (model UVP-54G shortwave UV lamp: 254 nm, 1 cm from tube). Reactions were digested for 15 min at room temperature with RNase (1 μl of 3.5 mg/ml RNase A, final concentration of 175 ng/μl and 1 μl of 10 U/μl RNase T1, final concentration of 0.5 U/μl). Then 7 μl of SDS sample buffer and 1 μl of β-mercaptoethanol (0.5 M final) were added to the samples. The samples were boiled for 5 min, spun down in a microfuge, and half of the sample (15 μl) was loaded onto a NOVEX 10% Tris-glycine gel, which was run at 150 V for 90 min. The gel was dried for at least 30 min and exposed to a phosphor screen. The ³²P RNA-protein bands were imaged using a Fuji Phosphorimager and quantified using the MacBASV2.5 software package.

For RNA competition experiments using ³²P-labeled intron VI as the riboprobe, the probe included all 466 nt of intron VI and was transcribed from pMN147. The xol-1 intron III non-competing cold background RNA included all 456 nt of intron III and was transcribed from pKA1. The RNA oligos B-37, C-35, E-35, and E-25 were purchased from Dharmacon Research, Inc The sizes of the cold RNA competing fragments for each region in Figure 5A and their fraction of Us relative to intron VI are the following: A, 124 nt (0.27); B, 104 nt (0.17); C, 116 nt (0.23); D, 89 nt (0.16); E, 137 nt (0.21); intron VI with ΔB-37 and ΔC-35, 426 nt (0.87); intron VI with ΔB-37, ΔC-35, and ΔE26, 400 nt (0.83); and intron III, 456 nt (0.88).

FOX-1 protein was tagged at the N-terminus with maltose binding protein (MBP), expressed in bacteria, and purified with an amylose resin column, using 10 mM maltose to elute FOX-1 from the resin.

The endogenous fox-1 and xol-1 genes (Figure 6) were edited using microinjected Cas9 ribonucleoprotein complexes (RNPs), as described previously (Farboud et al., 2019). Target-specific CRISPR guide RNAs (crRNA) (Supplementary file 2) and trans-activating crRNA (tracrRNA) were obtained from Dharmacon. crRNAs have a 2xMS modification to improve nuclease resistance. For all experiments, a co-CRISPR strategy was used that relied on the dpy-10 co-conversion marker (Supplementary files 2–4) to enrich for edited worms, based on a Dpy or Rol phenotype, prior to PCR screening and DNA sequencing to identify desired mutations (Arribere et al., 2014; Kim et al., 2014a; Ward, 2015).

To edit fox-1, two crRNAs (Supplementary file 2) targeted Cas9-dependent DSBs to two sites flanking fox-1. A single-stranded DNA oligonucleotide (ssDNA) (Supplementary file 3) with homology 5’ of one DSB and 3’ of the other served as the homology-directed repair template to join the broken DNA ends and thereby delete fox-1. The ssDNA repair template was injected with Cas9 RNPs at a final concentration of 0.5 μM. PCR using the three primers BF-2492, BF-2493, and BF-2494 in a single reaction (Supplementary file 4) permitted the simultaneous detection of an unedited wild-type fox-1 gene, a heterozygous 10 kb fox-1 deletion, and a homozygous 10 kb fox-1 deletion. PCR amplification with BF-2493, which anneals to sequences targeted for deletion, and BF-2494, which anneals to sequences downstream of the deletion, yields a 307 bp amplicon only from an unedited fox-1 gene. In contrast, PCR amplification with BF-2492 and BF-2494, which anneals to sequences upstream of the deletion, yields a 475 bp amplicon only from a deleted fox-1 gene, named fox-1(y793). Precise excision was confirmed by PCR amplifying the edited locus with primers BF-2492 and BF-2494 and performing Sanger sequencing with primer BF-2492.

To edit xol-1 DNA encoding intron VI, two crRNAs targeted Cas9-dependent DSBs to sites flanking intron VI (Supplementary file 2). The broken DNA ends were joined by homology-directed repair that was templated from exogenously provided double-stranded DNA (dsDNA) fragments, either 1553 bp for the deletion allele xol-1(y810) or 1651 bp for all others. The dsDNA repair templates with intron VI mutations used in Figure 6 were synthesized as gBlocks by Integrated DNA Technologies. The templates also included silent mutations to eliminate the PAM (AGG to AGA at codon 330 associated with crispr_bf69 RNA) or to mutate a key nucleotide (GCA to GCT at codon 255 associated with crispr_bf68 RNA). These mutations prevented Cas9 from cleaving the repair template and from re-cleaving the precisely edited genomic locus. They did not alter the primary amino acid sequence of xol-1. The dsDNA repair template also included approximately 500 bp of uninterrupted homology on either side of the silent mutations. Sequences of dsDNA repair templates are available upon request. The gBlocks were subcloned into pCR-Blunt II vectors, sequenced, and then PCR amplified using Phusion High-Fidelity DNA polymerase with primers BF-2507 and BF-2508 (Supplementary file 4). The resulting amplicon was purified and concentrated using Qiagen MinElute columns. dsDNA repair templates were injected with Cas9 RNPs at a final concentration of 350 ng/μl. Resulting co-converted Dpy or Rol F1 animals were lysed, and the xol-1 intron VI region examined. To screen for all xol-1 intron VI mutations except the y810 deletion, primers BF-2518 and BF-2519 were used to PCR amplify the targeted locus, and BF-2518 was used to perform Sanger sequencing to confirm precise editing. To screen for the xol-1(y810) deletion, F1 Dpy or Rol worms were screened by PCR using primers BF-2301, BF-2676, and BF-2746 in a single reaction to detect an unedited gene and heterozygous or homozygous deletion variants lacking the intron. Strains carrying homozygous deletions were reexamined by amplifying xol-1 using BF-2518 and BF-2519, and performing Sanger sequencing with primer BF-2301 to confirm precise editing. All oligonucleotides used to screen for CRISPR/Cas9 induced mutations and to determine sequences of resulting mutations are listed in Supplementary file 4.

Sequences within intron VI of xol-1 include five potential FOX-1 binding sites (Figure 6), each having GCAUG and/or GCACG motifs. The motifs are distributed throughout the intron and start at the following positions relative to the 5' end of the intron: 124 nt, 152 nt, 231 nt, 257 nt, and 440 nt.

Adjacent and overlapping the GCACG motif at position 152 is a GCUUG sequence resembling a FOX-1 binding motif. The overlap is one nt: GCUUGCACG. In vitro binding studies of Rbfox to GCAUG and GCACG sites showed that GCACG is a higher affinity binding site than GCUUG. Since this site can only be occupied by a single FOX-1 protein, and FOX-1 is more likely to bind GCACG than GCUUG, we classify this entire GCUUGCACG sequence as a GCACG site (Figure 6 and Figure 6—figure supplement 1). Even though the GCUUG motif is highly unlikely to bind FOX-1, we mutated it in several strains in case it can bind FOX-1 when GCACG is altered. To disrupt it in xol-1(y803) and xol-1(y804), the GCUUG motif was mutated to AUUUA in addition to changing the GCACG motif, making the final sequence AUUUAUACA. For xol-1(y808), the GCUUG motif was mutated to GCUUA in addition to mutating the GCACG motif, making the final sequence GCUUAUACA. For xol-1(y809), the GCUUG motif was mutated to AUUUG while preserving the adjacent GCACG motif, making the final sequence AUUUGCACG. For xol-1(y807), the GCUUG site was mutated to GCUCG, while preserving the GCACG motif, making the final sequence GCUCGCACG. The GCUUG motif was not mutated by itself to measure the effect of GCUUG loss because changing that motif to AUUUA would mutate the adjacent higher affinity GCACG motif.