Expression of Upf1-Nos inhibits activity of the anterior determinants Hb and Bicoid (Bcd) and causes the degradation of bcd mRNA

With the goal of identifying mRNA regulatory targets of Nos, we prepared transgenic flies that express either a Upf1-Nos chimera, or one of 3 control proteins: Upf1; Nos; or Upf1-Nos^L7, which bears the C-terminal tail deletion that abrogates recruitment to hb mRNA. Transcription of each transgene is dependent on the GAL4 driver that is introduced by appropriate crosses, and each transgenic mRNA bears the 3′ UTR of wild-type nos mRNA to provide the posttranscriptional regulation conferred on native nos.

We first drove expression of Upf1-Nos and the 3 control proteins described above with Nos GAL4-VP16, which directs transcription in the germarium and subsequent stages of oogenesis. Females expressing Upf1-Nos produce essentially no eggs and bear rudimentary ovaries; in contrast, females expressing Upf1, Nos, or Upf1-Nos^L7 appear to produce normal numbers of eggs. As explained in more detail below, we wished to harvest eggs at nuclear cycle 9 to take advantage of the prolonged transcriptional quiescence flanking the oocyte–egg transition (Fig 1A) and therefore asked whether use of another GAL4 driver would permit egg development. We found this to be the case with a maternal alphaTub>GAL4 driver, which directs expression throughout much of oogenesis and was used in all subsequent experiments. To determine the extent to which Upf1-Nos is overexpressed relative to endogenous Nos, we prepared transgenes encoding HA-epitope tagged versions of each protein. Expression of Upf1-HA-Nos was driven by the alphaTub>GAL4 driver referred to above; expression of HA-Nos was under the control of native signals in an otherwise wild-type fragment of genomic DNA that rescues the nos mutant phenotype. Comparison of the level of maternally derived protein in 0- to 2-hour embryonic extracts reveals that Upf1-Nos is overexpressed approximately 3-fold (S1 Fig and S1 Data).

Although females expressing Upf1-Nos lay eggs, virtually none hatch. In a pilot study, 0/245 Upf1-Nos eggs hatched, whereas between 80% and 95% of Upf1, Nos, or Upf1-Nos^L7 control eggs hatched. Since expression of Upf1-Nos is exclusively maternal, we expected that defects might be evident early in embryonic development. Therefore we harvested 0- to 3-hour-old embryos and examined the distribution of nuclei by staining with DAPI to assess developmental progress. Unlike the 3 control genotypes, a large fraction of Upf1-Nos embryos have a single nucleus and are either not fertilized or fail to commence nuclear division cycles; we have not investigated this further and restricted subsequent analysis to fertilized embryos between nuclear cycles 2 to 13. Almost all the Upf1, Nos, and Upf1-Nos^L7 embryos exhibit normal distributions of syncytial cleavage stage nuclei (94% to 100%; see S1 Data for details). In contrast, 63% of Upf1-Nos embryos have abnormal nuclear distributions, with irregularly spaced nuclei and variable DNA content, based on the intensity of DAPI fluorescence (e.g., Fig 1C).

We next asked whether the Upf1-Nos embryos that fail to hatch exhibit interpretable developmental defects by examining their body plan, as revealed by pattern in the epidermal cuticle secreted late in embryonic development. About 97% of Upf1-Nos eggs were “empty,” without detectable cuticle, probably because they die earlier in development (Fig 1B). However, among the 3% of Upf1-Nos embryos that did secrete cuticle, 2 had normal segmentation and 14 revealed a striking bicaudal body plan, in which normal head and thoracic segments are replaced by a mirror image duplication of posterior abdominal segments (Fig 1B). Bicaudal development is diagnostic for simultaneous inhibition of Bcd and Hb activity [33–35]. None (0/862) of the Upf1, Nos, or Upf1-Nos^L7 control embryos exhibited segmental polarity reversals, much less a mirror image duplication of the abdomen. As shown in Fig 1B, most (80% to 95%) of the control embryos secreted cuticle that revealed body plans with normal antero-posterior polarity and, at most, minor defects; the remaining embryos failed to secrete cuticle.

Even though only a few Upf1-Nos exhibited bicaudal body plans, the phenotype is sufficiently characteristic to strongly suggest that Upf1-Nos inhibits activity of both Hb and Bcd. bcd is transcribed exclusively maternally, while hb is transcribed both maternally and, later, zygotically (under the control of Bcd). For simplicity, we therefore focused on examining the effects of Upf1-Nos on bcd mRNA by in situ hybridization in early cleavage stage embryos. Most (82%) of the Upf1-Nos embryos that were either unfertilized or in the first nuclear division cycles have a normal level of bcd mRNA at the anterior (Fig 1C); this was also the case for most (95%) control Upf1-Nos^L7 embryos in cycles 0 to 2. But we observed a novel phenotype in Upf1-Nos embryos that we judged to be approximately in cycles 9 to 10 based on nuclear density: an absence of detectable bcd mRNA (Fig 1C). In contrast, Upf1-Nos^L7 embryos in cycles 9 to 10 have normal levels and distributions of bcd mRNA at the anterior pole.

In summary, the vast majority of control Upf1, Nos, and Upf1-Nos^L7 embryos develop normally, which is consistent with subsequent analysis of their transcriptomes (below). In contrast, Upf1-Nos embryos are abnormal: Many fail to either fertilize or commence nuclear division, subsequent nuclear divisions are abnormal, only a handful escape the deleterious early effects of Upf1-Nos and progress through to late development, and none hatch. Most importantly, 2 phenotypes—the absence of bcd mRNA during early development and the bicaudal body plan of Upf1-Nos embryos—support our premise that the chimeric protein would degrade Nos-regulated mRNAs.

Upf1-Nos depletes 39% of maternal mRNAs

To identify targets of Upf1-Nos in the entire transcriptome, we prepared triplicate samples from 4 experimental genotypes (Upf1, Nos, Upf1-Nos, and Upf1-Nos^L7) as well as wild-type embryos as a control. Embryo collections were soft-fixed in methanol, stained with DAPI, and manually sorted in the fluorescence stereomicroscope to identify embryos at nuclear cycle 9 to 10, or in the case of Upf1-Nos, embryos that appeared to be approximately the same age despite an abnormal distribution of nuclei (as in Fig 1C). Because transcription in the germ line ceases at stage 10B of oogenesis and resumes only during the MZT beginning at nuclear cycle 10, preexisting Upf1-Nos (or the other ectopically expressed regulators) in the collected embryos have had the opportunity to act on the maternal transcriptome for approximately 9 to 10 hours without synthesis of new mRNA. Prior to the transcriptional “shutoff,” Upf1-Nos might cause changes to the transcriptome via both direct and indirect mechanisms; we hoped that direct action of Upf1-Nos during the period of transcriptional quiescence would have a predominant effect on transcriptome composition in embryos harvested at nuclear cycle 9. RNA was prepared from all the genotypes, prepared for RNAseq, and analyzed by standard methods.

We first assessed the quality of our samples. The absence of detectable mRNAs encoded by genes that comprise the first zygotic response to maternal patterning gradients (e.g., Kr, gt, kni) supports the idea that embryos were in fact collected prior to the MZT (see S2 Data). Reducing the complexity of all 15 samples by Multidimensional Scaling reveals that biological replicates are similar to each other and that the control samples are also similar to each other (S2 Fig). In contrast, the Upf1-Nos samples are markedly different from all 3 UAS controls and wild type, consistent with their unusual embryonic phenotypes (Fig 1).

We next compared the transcriptomes of each experimental genotype to wild type to identify genes with altered expression, arbitrarily setting a minimum 2-fold threshold change and a false discovery rate (FDR) < 0.05. For simplicity, we describe data in which all mRNA isoforms for a gene are binned; analysis of individual mRNA isoforms yields broadly similar results (S3 Fig). We initially found that a small fraction of the maternal transcriptome is depleted (2.4%) or enriched (2.1%) upon expression of Upf1 alone (Fig 2A). Because we were interested primarily in mRNAs that are targeted by the Nos portion of the Upf1-Nos chimera, we excluded Upf1-regulated genes from subsequent analysis of Upf1-Nos and Upf1-Nos^L7.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Upf1-Nos targets a large fraction of the maternal transcriptome.

(A) For each of the misexpressed proteins, the plot shows the difference in mRNA abundance in comparison to wt (log₂ fold change) as a function of the average mRNA abundance (log₂ transcripts per million reads) for embryonic samples collected at nc 9–10. Each dot represents a binned collection of mRNA isoforms for a single gene; analysis of individual isoforms is in S3 Fig. Genes altered by expression of Upf1 are excluded from the analysis of Upf1-Nos and Upf1-Nos^L7 data. Gene-binned mRNAs significantly depleted and enriched are indicated in blue and orange, respectively. The green lines mark log₂ ± 1, and the numbers at the right indicate the fraction of the (gene-binned) transcriptome that is significantly changed and above or below the arbitrary 2-fold cutoff. hb is highlighted in the Upf1-Nos plot. (B) Correlation between the difference in mRNA abundance (Upf1-Nos vs. wt) measured by RNAseq and by RT-qPCR for a collection of genes, which span an approximately 1,000-fold range in expression level in wt. (C) The results of DAVID analysis reveal gene sets enriched among Upf1-Nos-targeted mRNAs (red) and nontargeted mRNAs (blue), which are plotted vs. their ES. Shown are annotation clusters with highest ES for which at least 1 Category has a Benjamini adjusted p-value < 0.05. For clarity, 3 clusters that satisfy these criteria and are enriched among nontargeted mRNAs with ES between 2.3 and 1.9 were omitted (see S2 Data for details). (D) Examples of enriched gene sets, highlighted on the Upf1-Nos vs. wt plot from (A). Underlying data for the figure are in S2 Data. ES, enrichment score; hb, hunchback; nc, nuclear cycle; Nos, Nanos; RT-qPCR, quantitative reverse transcription PCR; wt, wild type.

https://doi.org/10.1371/journal.pbio.3002840.g002

As shown in Fig 2A, the expression of Upf1, Nos, or Upf1-Nos^L7 has relatively little effect on the transcriptome. The proportion of significantly up- and down-regulated genes is approximately the same for Upf1 and Nos, and the proportion of the transcriptome depleted by Upf1-Nos^L7 is only modestly greater than the proportion of up-regulated genes. In contrast, preliminary analysis of mRNAs in Upf1-Nos embryos suggested that many mRNAs were significantly down-regulated (S4 Fig). The large, asymmetric effect of Upf1-Nos on the transcriptome results in oversampling of nontargeted mRNAs. To account for this effect, we adjusted the normalization factors of Upf1-Nos samples in analyzing differential expression using an average derived from 20 highly expressed mRNAs (most of which encode ribosomal proteins), on the assumption they are not differentially expressed (as described [36]). Subsequent analysis reveals that >39% (n = 2,596) of the genes encoding maternal mRNAs are depleted more than 2-fold by Upf1-Nos (Fig 2A).

To test the normalization described above and obtain an independent measure of changes in abundance due to expression of Upf1-Nos, we measured the levels of 16 mRNAs by quantitative reverse transcription PCR (RT-qPCR); these were chosen to span a log₂ = 15-fold change in abundance. As shown in Fig 2B, the 2 methods show excellent agreement, with a Pearson’s correlation R value of 0.98. The absence of a significant population of enriched mRNAs in the Upf1-Nos data supports a premise of our experiment—that tethering Upf1 to mRNAs via Nos during the transcriptionally quiescent period preceding sample collection would lead to their degradation via NMD and that this degradation would predominate to shape the composition of the transcriptome. We assume that indirect effects (i.e., up-regulation of some mRNAs as a secondary consequence of the action of Upf1-Nos on other, directly targeted mRNAs) are minimized in the absence of concurrent transcription.

Several observations support the idea that, although Upf1-Nos targets a large portion of the transcriptome, it does so specifically. First, the inactivity of Upf1-Nos^L7 argues that the wild-type chimera acts in collaboration with Pum to specifically target NRE-bearing mRNAs, a point we return to in the following section. Second, 3 of the 4 mRNAs previously shown to be bound and regulated by Nos—hb, bcd, and alpha-Importin [37,38]—are targeted by Upf1-Nos; moreover, hb, which is the critical Nos target in the early embryo, is one of the most highly depleted mRNAs (highlighted in Fig 2A). The exception is CycB mRNA, which is bound by Nos in collaboration with Pum and repressed in the PGCs [1,7], but not targeted by Upf1-Nos. (We note that, for reasons that are not clear, Nos is necessary but not sufficient to repress most of the CycB mRNA in these experiments, which derives from the prospective somatic cytoplasm [1].) Third, if Upf1-Nos acted promiscuously, due to high nonspecific binding (for example), it would preferentially target mRNAs with longer 3′ UTRs; in fact, mRNAs targeted by Upf1-Nos have on average slightly shorter 3′ UTRs than nontargeted mRNAs (S5 Fig).

The depletion of such a large fraction of the transcriptome by Upf1-Nos suggests that Nos regulates thousands of mRNAs in Drosophila, in contrast to the hundreds regulated by Nos in C. elegans [10]. Since such a large fraction of the transcriptome is targeted by Upf1-Nos, it is unclear whether it is meaningful to identify preferentially targeted pathways. Nevertheless, clustering analysis of gene function via DAVID [39,40] reveals that a number of pathways are significantly enriched among the pool of genes targeted by Upf1-Nos as well as the pool of untargeted genes (Fig 2C). Among Upf1-Nos targeted genes, only 4 clusters have gene subsets with significant Benjamini-adjusted p-values; the most highly enriched of these subsets encodes factors involved in DNA replication and repair (enrichment score [ES] = 3.1; Fig 2C), perhaps accounting for the aberrant replication seen in Upf1-Nos embryos (Fig 1C). An independent search for Drosophila genes associated with the Gene Ontology terms “DNA replication” and “DNA repair” revealed that 79/79 are depleted, as shown in Fig 2D. Among genes not targeted by Upf1-Nos, constituents of the ribosome (ES = 20.7) and spliceosome (ES = 9.2) are even more significantly enriched (Fig 2C and 2D). We conclude that while Upf1-Nos selectively targets a few biological processes/pathways with modest preference, it acts globally to deplete maternal mRNAs that encode factors distributed across most biological functions.

Upf1-Nos acts primarily in conjunction with Pum

Three lines of evidence support the idea that, in targeting a large fraction of the maternal transcriptome, Upf1-Nos acts in collaboration with Pum, much as native Nos and Pum collaboratively repress hb mRNA.

First, Upf1-Nos^L7 is almost inert, changing the transcriptome to essentially the same minor extent as does Upf1 (Fig 2A). The C-terminal tail of Nos that is altered in the L7 mutant is specifically required to promote cooperative binding of Pum and Nos to the NREs in hb mRNA.

Second, a set of 538 ovarian mRNAs that copurify with affinity-tagged Pum [27] is enriched among mRNAs depleted by Upf1-Nos (p-value = 0.03 by the CAMERA competitive test of [41]; Fig 3A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Pum mediates most of the activity of Upf1-Nos.

(A) Barcode plot showing enrichment among mRNAs depleted by Upf1-Nos of mRNAs bound by transgene-encoded Pum RNA-binding domain in ovarian extracts identified by Gerber and colleagues [27]. Each mRNA is represented by a line; mRNAs depleted and enriched > log₂ = 1 are highlighted in blue and red, respectively. Here and in subsequent barcode analyses, the x-axis is nonlinear. (B) A modified NRE motif (shown above) is enriched among the 3′ UTR sequences of mRNAs depleted by Upf1-Nos. The plot below is of adjusted p-values (Hommel) for the significance of the enrichment on the y-axis of each possible motif variant displayed along the x-axis; enrichment of the canonical NRE motif (e.g., +4A) is shown to the right. Enrichment was calculated as described in the text, and each boxplot displays the distribution of p-values from 1,000 trial comparisons between the fraction of 3′ UTRs bearing ≥1 motif in targeted vs. nontargeted mRNAs. (C) Bar graph showing mean number of various sites (shown above) per kb of 3′ UTR in targeted and nontargeted mRNAs, with error bars representing the SEM. The first pair of columns on the left shows enrichment of the modified NRE motif, and subsequent columns show enrichment of variants at position +4 of this motif, with p-values calculated in the Wilcoxon test. The data are also displayed by local regression in S6 Fig. Underlying data for the figure are in S2 Data. Pum, Pumilio; Nos, Nanos; NRE, Nanos Response Element.

https://doi.org/10.1371/journal.pbio.3002840.g003

Third, the 3′ UTRs of mRNAs targeted by Upf1-Nos are enriched for a modified Nos+Pum binding motif, UGU , in which position +4 is either U or A. This enrichment emerged from 2 analyses.

First, we asked whether the canonical Nos+Pum motif ( UGUA) is enriched among 3′ UTRs targeted by Upf1-Nos, comparing the fraction of 3′ UTRs bearing the motif between randomly selected subsets of targeted and nontargeted mRNAs, chosen to have the same number of sequences and similar distributions of 3′ UTR lengths. As shown at the right of Fig 3B, the fraction of targeted 3′ UTRs bearing at least 1 canonical site (with A at position +4) is significantly greater than the fraction of nontargeted UTRs (median p-value of 1,000 trials = 1.9 × 10⁻⁵). As controls, we asked whether singly substituted mutant derivatives at each of the 7 motif positions are also enriched. All of the derivatives but one appear to be “mutant” sites not significantly enriched in targeted 3′ UTRs; the exception is UGUU, which, in fact, is enriched to a greater extent than the canonical site (Fig 3B).

Second, we measured the density of Nos+Pum motifs, comparing 3′ UTR sequences from mRNAs targeted and not-targeted by Upf1-Nos. As shown in Figs 3C and S6, the average number of UGU sites is significantly enriched in targeted mRNAs, unlike UGUG or UGUC sites (Fig 3C). The results shown in Fig 3B and 3C are consistent with work described elsewhere showing that Nos+Pum bind efficiently to NREs with either U or A at position +4 [42].

Taken together, the evidence described above supports the idea that Upf1-Nos is primarily recruited to its targets by Pum in a manner similar to their joint recruitment to the hb NRE. We were unable to test Pum dependence directly, as pum mutant flies that also express Upf1-Nos are subviable, and females produce very few eggs.

Upf1-Nos may be recruited to a minor fraction of its targets by Bruno

Although Upf1-Nos^L7 is largely inactive (as described above), a small cohort of 178 genes is depleted >2-fold upon expression of this mutant chimera. We find no significant enrichment of NREs in the 3′ UTR sequences of mRNAs targeted by Upf1-Nos^L7 (Figs 4A and S7, p-value = 0.087 by Wilcoxon test). Thus, depletion is unlikely to be the result of inefficient recruitment of the mutant Nos moeity to NREs via Pum. We therefore considered the possibility that another RNA-binding protein might recruit Upf1-Nos^L7 to this minor fraction of the transcriptome.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Upf1-Nos targets a small portion of the transcriptome via Bru.

(A) Similar to Fig 3C, the mean 3′ UTR density of NRE and BRE sites is compared between mRNAs targeted vs. not-targeted by Upf1-Nos^L7 (p-values calculated via Wilcoxon test). The data are also displayed by local regression in S7 Fig. (B) Barcode enrichment plot displaying mRNAs derepressed upon inhibition of Bru expression in the ovary [43] plotted vs. the log₂ (fold change) in Upf1-Nos^L7 vs. wt (e.g., Fig 2A). (C) DAVID analysis reveals enrichment of gene sets among mRNAs targeted by Upf1-Nos^L7. (D-F) Yeast interaction assays show that both Nos and Nos^L7 can be recruited to RNA-bound Bru. In each panel, the binding of AD fusions to various factors (y-axis) to a DNA-bound bait (title above each bar graph) is measured in arbitrary light units of β-galactosidase (x-axis, log₁₀ plot). For the “control” entry in each bar graph, yeast were transformed with plasmids that express the bait indicated in the title and an empty vector encoding AD only; the background level of β-galactosidase is also indicated with a vertical dashed line. (D) Two-hybrid experiments showing protein–protein interactions between Nos and Bru (above) and Bru and Nos (below). We are unable to assess interaction with DBD fused to the Nos NTD, as the resulting fusion “autoactivates” transcription of the reporter without an AD-fusion partner. (E) Three-hybrid experiments showing specific binding of Bru to RNA bearing wt BREs. Below the horizontal dashed line is an additional control showing no significant binding of Bru to NREs. (F) Four-hybrid experiments showing binding of Nos and Nos^L7 to Bru-bound RNA. Below the horizontal dashed line are additional controls showing no binding of Nos in the absence of Bru to RNA bearing either wt or mutant BREs. Note that the background level of β-galactosidase is higher in the yeast strain used for 2-hybrid experiments than in the strain used for 3- and 4-hybrid experiments. Underlying data for the figure are in S3 Data. AD, activation domain; BRE, Bruno Response Element; Bru, Bruno; DBD, DNA-binding domain; Nos, Nanos; NRE, Nanos Response Element; NTD, N-terminal domain; wt, wild type.

https://doi.org/10.1371/journal.pbio.3002840.g004

To identify such a putative novel Nos cofactor, we used the DREME tool [44] to search for motifs overrepresented among 3′ UTR sequences of mRNAs depleted by Upf1-Nos^L7. The search revealed enrichment of the motif UAU U UA, which is similar to one of the binding sites for Bruno (Bru): the Bruno Response Element (BRE) U U U U [45]. The average density of BREs is greater in mRNAs depleted ≥2-fold by Upf1-Nos^L7 (Figs 4A and S7).

Bru is a translational repressor that binds a number of RNA motifs via its 3 RNA recognition motifs (RRMs) [46]. mRNA regulatory targets of Bru in the ovary have been identified by virtue of increased polysome association following RNAi-mediated inhibition of Bru expression [43]. If Bru mediates the depletion of mRNAs by Upf1-Nos^L7 in our experiments, we would expect significant overlap with the Bru-regulated mRNAs identified by Rangan and colleagues. This is indeed the case: mRNAs depleted by Upf1-Nos^L7 are enriched for Bru-regulated targets (p-value = 7.2 × 10⁻⁶, CAMERA competitive test), as revealed in a barcode enrichment plot (Fig 4B). Functional clustering of gene analysis reveals a number of enriched gene sets among Upf1-Nos^L7 targets; these encode membrane cytoskeleton factors or regulators (PDZ, SH3, pleckstrin domain proteins, spectrin repeat proteins), at least one of which (β-spectrin) plays a key role in GSCs and differentiation of their progeny (Fig 4C and S3 Data).

We next asked whether Bru can recruit Nos and Nos^L7 to RNA. Previous experiments have shown that Nos and Bru proteins interact [47,48]. To further characterize the Bru–Nos interaction, we performed yeast 2-hybrid experiments, the results of which are shown in Fig 4D. We find that a fragment of Bru interacts with full-length Nos, stimulating LacZ reporter activity to a slightly greater extent than does interaction with either of 2 known Nos effectors, Cup and NOT4 (Fig 4D, above) [1,49]. In reciprocal experiments with DNA-binding domain (DBD) and transcriptional activation domain (AD) fusions swapped, we observe binding of both Nos and Nos^L7 to Bru, demonstrating that the Bru–Nos interaction does not depend on an intact Nos C-terminal tail. Consistent with this observation, Bru interacts with the Nos N-terminal domain (NTD) but not its C-terminal RBD (Fig 4D, below).

To directly test whether RNA-bound Bru can recruit Nos to RNA, we next used yeast 3-hybrid RNA-binding experiments to show that Bru binds specifically to a synthetic RNA sequence with 3 copies of the canonical BRE, but not to RNAs with either 3 copies of a mutant BRE or 2 copies of a NRE (Fig 4E). Then, in yeast 4-hybrid experiments similar to those used to show Pum-dependent recruitment of Nos to NREs [23], we find that RNA-bound Bru can recruit either Nos or Nos^L7 to the RNA (Fig 4F). Recruitment is dependent on both the NTD and RBD of Nos, raising the possibility that protein–protein interactions between Bru and Nos as well as protein–RNA interactions between Nos and the RNA are required for formation of the ternary Bru/Nos/RNA complex.

Although we initially identified Bru as a mediator of the activity of Upf1-Nos^L7, the interaction experiments in Fig 4D–4F show that Bru interacts similarly with the L7 mutant and wild-type Nos proteins. It therefore seems likely that Bru mediates targeting of the same small subset of mRNAs by both Upf1-Nos^L7 and Upf1-Nos but that Pum-mediated targeting of the much larger set of mRNAs by the latter obscures the contribution by Bru. Consistent with this idea, neither BREs (p-value = 0.129 by Wilcoxon test) nor Bru-repressed mRNAs (p-value = 0.084 by CAMERA competitive test; S8 Fig) are enriched among mRNAs targeted by Upf1-Nos.

In summary, we have shown that Bru can recruit Nos or Nos^L7 to specific RNA sequences, supporting the idea that Bru mediates targeting of a minor subset of the maternal transcriptome by Upf1-Nos^L7 (and presumably contributing to depletion of the same minor subset by Upf1-Nos). The mechanism by which the Nos-Bru-BRE complex forms is independent of the integrity of the Nos C-terminal tail and thus different from cooperative binding of Nos+Pum to the hb NRE.

mRNAs targeted by Upf1-Nos in transgenic animals are repressed in wild-type ovaries and embryos by Nos

We next asked whether mRNAs depleted by Upf1-Nos might be regulated by endogenous Nos in wild-type animals. As a first step to addressing this question, we reanalyzed 2 studies of maternal mRNAs in late oogenesis and early embryogenesis, partitioning these into mRNAs either targeted or not targeted by Upf1-Nos.

In the first study, Tadros and colleagues [50] identified maternal mRNAs that are degraded in unfertilized eggs; by definition, the regulation they observed occurs in the absence of zygotic gene expression. Tadros and colleagues [50] identified 1,069 mRNAs whose abundance drops at least 1.5-fold as unfertilized eggs age. This set of destabilized maternal mRNAs is significantly enriched among our Upf1-Nos targets (p-value = 1.6 × 10⁻¹¹; Fig 5A). The embryos used in their experiments were unfertilized but otherwise wild type, and, therefore, destabilization must be mediated by endogenous maternal factors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Maternal mRNAs targeted by Upf1-Nos are translationally repressed and degraded in wild-type eggs.

(A) Barcode plot displaying enrichment of maternal mRNAs degraded in unfertilized eggs [50] vs. log₂ (fold change) in Upf1-Nos vs. wt (e.g., Fig 2A). (B) Maternal mRNAs expressed at nc 9–10 in the RNAseq experiments of Fig 2 were binned into those targeted (red) and not targeted (teal) by Upf1-Nos. Then, the poly(A) tail length and relative translational efficiency of each mRNA was extracted from the data of Eichhorn and colleagues [51] at the 4 stages of development indicated below: stage 11 and 14 oocytes, activated eggs, 0–1-hour embryos (before the onset of zygotic transcription). The plot at the right is of the change in abundance for each maternal mRNA between each of the 3 developmental transitions indicated below, again using the data of Eichorn et al. (2016). Statistical analysis in S4 Data. Note that zygotic transcription contributes significantly to mRNA abundance in the 2–3-hour embryo pool, but all other time points analyze purely maternal mRNAs. (C) Local regression analysis of binding motifs in mRNA 3′ UTRs as a function of TE at egg activation, from the data of Eichorn et al. (2016). For this analysis, we chose the most highly expressed maternal isoform for each gene. Underlying data for the figure are in S4 Data. nc, nuclear cycle; Nos, Nanos; TE, translational efficiency; wt, wild type.

https://doi.org/10.1371/journal.pbio.3002840.g005

In the second study, Eichhorn and colleagues [51] characterized mRNAs in staged samples from wild-type ovarian egg chambers and timed embryo collections up to 6 hours after egg deposition, measuring abundance, poly(A) tail length, and translational efficiency (TE). Their work reveals comprehensive changes to the maternal transcriptome during an extended span of development that includes the oocyte–egg transition (e.g., the vertical dashed line in Fig 1A) and the MZT, a period from approximately 1.5 to 3.0 hours of embryonic development during which many maternal mRNAs are degraded and zygotic transcription is activated. Analyzing their data, we find that mRNAs targeted by Upf1-Nos are regulated differently than are nontargeted mRNAs. Nontargeted maternal mRNAs undergo lengthening of their poly(A) tails at egg activation and concomitant enhanced translation, consistent with the observation that poly(A) tail length and TE are coupled until the onset of gastrulation (which occurs at 3 hours of embryonic development) [51]. In contrast, Upf1-Nos targets undergo neither poly(A) tail lengthening nor translational enhancement (Fig 5B). The repressed Upf1-Nos targets are stable until embryogenesis but degraded between the 0- to 1- and 2- to 3-hour windows of development (Fig 5B). The data likely underestimate the extent of degradation during the MZT, since some maternally transcribed mRNAs are retranscribed zygotically, as discussed below.

Taken together, the data in Fig 5 suggest that, late in oogenesis, the large set of approximately 2,600 mRNAs targeted by Upf1-Nos (in transgenic animals) is translationally repressed in wild-type animals during late oogenesis and subsequently degraded in the early embryo. We suggest that these regulatory events are a delayed response to the burst of Nos expression that peaks at stage 10B of oogenesis (Fig 1A), which represses targeted mRNAs and primes them for subsequent degradation in the embryo.

As a first test of the idea that Nos is responsible for repressing a large cohort of maternal mRNAs, we asked whether the modified NRE motif that mediates Nos+Pum binding is overrepresented among maternal mRNAs that are translated inefficiently at egg activation. As shown in Fig 5C, localized regression analysis reveals a correlation between the translational efficiency and the density of modified NRE motifs in the 3′ UTR across the maternal transcriptome. In comparison with the 3′ UTRs of mRNAs that are not repressed, the modified NRE is enriched in the 3′ UTRs of repressed mRNAs (defined as TE < 1, p-value = 1.1 × 10⁻⁸ by two-sided Wilcoxon test). As controls, canonical Pum binding sites (PREs) and BREs are present at lower density (Figs 5C and S9) and are enriched to a lesser extent (PREs) or not at all (BREs) in repressed mRNAs (p-values of 7.3 × 10⁻⁶ and 0.22, respectively).

All the work described above is based on the ectopic expression of Upf1-Nos, which behaves like a hyperactive version of Nos that efficiently degrades bound mRNAs. Part of the rationale for using Upf1-Nos is that, in the Drosophila embryo, Nos acts primarily to block translation and only secondarily to promote the degradation of targeted mRNAs [35,37,52,53]. In C. elegans, RNAseq analysis of nos mutant PGCs successfully identified a cohort of maternal mRNAs that are destabilized by Nos in wild-type PGCs [10]. This observation led us to ask (1) if we could detect Nos-dependent destabilization of maternal mRNAs and (2) whether the same mRNAs are targeted by Upf1-Nos and by Nos. To address these issues, we collected embryos from wild-type and nos⁻ females at 0 to 1 and 2 to 3 hours of development (i.e., before and after the onset of zygotic transcription, respectively) and measured the abundance of maternal mRNAs via RNAseq. In analyzing the data, we ignored mRNAs that are transcribed zygotically, but not maternally, since these are not exposed to and thus cannot be regulated by Nos.

As shown in Fig 6A, at 0 to 1 hours of development, the wild-type and nos^- mRNA populations are similar in composition, with only 3.5% of the maternal transcriptome significantly enriched >log₂ = 0.5-fold in nos^- embryos. A somewhat greater fraction of the maternal transcriptome (5.6%) is depleted in nos^- embryos at this stage of development, presumably due to indirect effects of Nos on gene expression earlier, during ovarian development (since Nos is thought to act primarily as a repressor). However, after the onset of the MZT, 18.5% of the maternal transcriptome is stabilized in 2- to 3-hour nos^- embryos (Fig 6A). We have not examined where in embryos Nos-dependent degradation takes place; however, we note that 35% of targeted maternal mRNAs are up-regulated >2-fold in the absence of Nos, suggesting that in wild-type animals, it acts throughout the cytoplasm rather than just in the posterior, as in the case of maternal hb. A rank-ordered barcode display of mRNAs stabilized in the absence of Nos reveals a correlation with mRNAs targeted by Upf1-Nos (p-value = 3.8 × 10⁻¹⁴; Fig 6B). In fact, 86% (1,019/1,185) of the mRNAs stabilized in the absence of Nos are depleted by Upf1-Nos (Fig 6C). Taken together, the complementary loss- and gain-of-function results of Figs 6 and 2 indicate that Nos regulates at least these 1,019 common mRNA targets.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Regulation of maternal mRNA abundance by endogenous Nos in wild-type oocytes and embryos.

(A) Results of RNAseq analysis, comparing the abundance of maternal mRNAs between wild-type and nos^- embryos at 0–1 and 2–3 hours of embryonic development. Only maternal mRNAs (based on the manually sorted embryos used in the RNAseq experiments of Fig 2 are shown. The green lines mark log₂ ± 0.5, and the figures at the right indicate the fraction of the (gene-binned) transcriptome that is significantly changed and above or below the arbitrary cutoff, which is smaller than the 2-fold cutoff used in Fig 2 since the magnitude of destabilization by Nos is generally smaller than the magnitude of degradation by Upf1-Nos. mRNAs stabilized or destabilized in the absence of Nos at an FDR < 0.05 are orange and blue, respectively. (B) The barcode shows mRNAs depleted by Upf1-Nos (Fig 2) plotted vs. the log₂ (fold change) in nos^- vs. wt at 2–3 hours (e.g., Fig 6A). Each depleted mRNA is a hashmark, the length of which indicates the extent of depletion by Upf1-Nos; this plot shows a poor correlation between the extent of degradation by Upf1-Nos and by Nos in wt embryos (see also the scatterplot in D). (C) A Venn diagram shows that most of the maternal mRNAs stabilized in the absence of Nos are degraded by Upf1-Nos. (D) The scatterplot shows the correlation between the extent of degradation by Upf1-Nos and stabilization in the absence of Nos. The canonical Nos substrate hb mRNA is highlighted as an extreme example, in which zygotic reexpression between 2–3 hours almost completely obscures Nos-mediated regulation of maternal hb mRNA. In all these comparisons, we analyze only the 6,399 mRNAs common to the 2 datasets (e.g., Figs 2A and 6A). Underlying data for the figure are in S5 Data. FDR, false discovery rate; hb, hunchback; Nos, Nanos; wt, wild type.

https://doi.org/10.1371/journal.pbio.3002840.g006

What then of the 1,528 maternal mRNAs that are regulated by Upf1-Nos but apparently not by Nos (Fig 6C)? One possibility is that they are repressed but not degraded by Nos. Another possibility is that they are repressed and degraded by Nos and then reexpressed zygotically during the 2- to 3-hour window of embryonic development. At this stage, Nos is confined to the PGCs (which constitute only a small fraction of the volume of the embryo), and, therefore, essentially no zygotically reexpressed mRNAs will suffer regulation by Nos even if they might be susceptible. This is not a confounding issue for our analysis of Upf1-Nos activity, since mRNA was harvested before the onset of transcription. However, zygotic reexpression likely does mask regulation of mRNA stability by Nos, thereby accounting, at least in part, for the modest correlation between the extent of regulation by Upf1-Nos and by Nos (R² = 0.25; Fig 6D). For example, the canonical Nos target hb (highlighted in the scatter plot in Fig 6D), which is among the most sensitive Upf1-Nos targets, is retranscribed in the 2- to 3-hour window such that the net level of hb mRNA at 2 to 3 hours of development is not significantly different in the presence or absence of Nos.

To estimate the potential influence of zygotic reexpression on the 2- to 3-hour samples ± Nos, we first analyzed data from Lott and colleagues (see S5 Data) [54]. Their measurements of RNA abundance in precisely staged single embryos allow us to determine the fraction of maternal mRNAs in our experiments that are reexpressed zygotically within 3 pools: mRNAs targeted by both Upf1-Nos and Nos (pool A; 105/1,019 = 10%); mRNAs targeted by Upf1-Nos but not Nos (pool B; 655/1,528 = 43%); and mRNAs targeted by neither (pool C; 760/3,852 = 20%). There is a striking enrichment of zygotically reexpressed mRNAs in pool B (p-values < 2.2 × 10⁻¹⁶ versus pools A or C, respectively, by Fisher exact test). Taken together, these observations support the idea that, for many of the genes in pool B, Nos-dependent degradation of the maternal mRNA component might be masked by new transcription at 2 to 3 hours of embryonic development (e.g., as for hb).

We reasoned that we might observe a reduction in poly(A) tail length and TE of the 1,528 mRNAs in pool B at earlier stages of development, before the potentially confounding contribution of zygotic transcription. To test the idea, we returned to the resource described by Eichhorn and colleagues [51], examining the fate of maternal mRNAs at various stages from late oogenesis through the first hour of embryonic development. Data were parsed into the 3 pools described above: (A) the 1,019 mRNAs targeted both by Upf1-Nos and Nos; (B) the 1,528 mRNAs targeted only by Upf1-Nos; and (C) the remaining 3,852 mRNAs not targeted by Upf1-Nos. As shown in Fig 7A, the 2 pools containing Upf1-Nos-targeted mRNAs (A and B) are regulated in a similar fashion, particularly at egg activation: mRNAs in both pools are translationally repressed and suffer shortening of their poly(A) tails (albeit, to slightly different extents; S5 Data). In contrast, the mRNAs not targeted by Upf1-Nos (pool C) are translationally activated and undergo lengthening of their poly(A) tails.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Regulation of maternal mRNA polyadenylation, translation, and degradation by Upf1-Nos and Nos.

(A) Maternal mRNAs were divided into 3 pools: down in Upf1-Nos and up in nos^- at 2–3 hours (Pool A); down in Upf1-Nos only (Pool B); and the remaining 3,852 maternal mRNAs untargeted by Upf1-Nos (Pool C). These bins were used to parse the data of Eichhorn and colleagues [51], as described for Fig 5. Statistical analysis in S5 Data. (B) Analysis of unstable maternal mRNAs degraded in early embryogenesis. At the left is shown the overlap between maternal mRNAs that are degraded in the first 3 hours of embryogenesis (60% of all maternal mRNAs; [55]) with mRNAs targeted by Upf1-Nos. To the right are Venn diagrams showing the distribution of mRNAs among various degradation pathways. Down arrows denote degradation by either Mat or Zyg pathways, and the single upward arrow indicates zygotic reexpression. These correspond to mRNA classes II-V in [55] and an “other” category, which consists primarily of mRNAs absent from Thomsen and colleagues’ data, as well as minor contributions from stable maternal mRNAs (class I) and purely zygotic mRNAs (class VII). Underlying data and statistical analysis in S6 Data. Mat, maternal; Nos, Nanos; Zyg, zygotic.

https://doi.org/10.1371/journal.pbio.3002840.g007

Overall, we conclude that, at or around the oocyte–egg transition, endogenous Nos likely regulates most of the maternal mRNAs targeted by Upf1-Nos.

Nos is a major contributor to maternal mRNA degradation in the early embryo

As in many organisms [56], a large proportion of the maternal transcriptome is degraded during early Drosophila embryonic development. Our analysis in Fig 5B suggests that mRNAs targeted by Upf1-Nos are degraded during the MZT. But we could not easily gauge the overall contribution of Upf1-Nos, or Nos, to maternal degradation in general from these data. We therefore turned to experiments in which maternal mRNA degradation has been measured more directly [55]. Thomsen and colleagues measured degradation in both unfertilized eggs and in embryos before and after onset of the MZT using microarrays. Their experiments reveal that 60% of maternal transcripts are degraded in the early embryo and that degradation of each mRNA is dependent on the activity of maternal factors, zygotic factors, or a mixture of the two. The initial wave of degradation has recently been verified via analysis of single embryo transcriptomes using RNAseq [57].

We asked whether the subsets of degraded maternal mRNAs and mRNAs targeted by Upf1-Nos overlap. The striking result is that 25% of degraded maternal mRNAs are targeted by Upf1-Nos and by Nos, and an additional 34% of degraded maternal mRNAs are targeted by Upf1-Nos but not Nos (S6 Data); conversely, the overwhelming majority of Upf1-Nos targets (90%) are degraded in wild-type embryos (Fig 7B). These observations support the ideas that (1) Nos contributes to the degradation of between 25% and 59% of maternal mRNAs in wild-type embryos and (2) Upf1-Nos does not adventitiously target otherwise stable mRNAs. Parsing the data of Thomsen and colleagues reveals that degradation of mRNAs targeted by both Upf1-Nos and Nos is dependent on both maternal and zygotic factors (Fig 7B). Their data further show that at least 19% of mRNAs targeted by Upf1-Nos but not Nos are zygotically reexpressed (Fig 7B), supporting our idea that zygotic expression masks Nos-dependent degradation in the experiments of Fig 6.

Overall, we conclude that Nos regulates the poly-adenylation, translation, and stability of a large portion of the maternal transcriptome.

Drosophila strains and methods

Plasmids used in this study are described in S7 Data. Transgenic lines were constructed by microinjection of w¹¹¹⁸ embryos by standard methods. The following strains were used: w¹¹¹⁸ as the wild-type reference in RNAseq experiments; the maternal alpha-tubulin GAL4-VP16 driver (stock # 7062 from the Bloomington Drosophila Stock Center [BDSC]); trans-heterozygotes of nos^BN and Df(3R)Exel6183 (stock # 7662, [BDSC]) for the RNAseq experiment of Fig 6. We initially tested 3 or more transgenic lines with the GAL4-VP16 driver for each construct and observed only minor phenotypic differences among the 3 lines. We screened for lines expressing Upf1-Nos and Upf1-Nos^L7 at essentially the same level by harvesting total RNA from collections of 0- to 2-hour embryos and measuring the chimeric mRNA by RT-qPCR using the appropriate primer pair in S2 Data. Fly stocks were maintained by standard methods and grown at 25°C. Embryos were collected on apple juice agar plates smeared with yeast paste.

Early embryos for assessment of nuclear morphology (Fig 1) were prepared by dechorionating in bleach, harvesting by filtration, fixing in 4% formaldehyde under heptane for 20 to 30 minutes, removal of the aqueous phase and shocking into methanol. Embryos were then rehydrated into phosphate-buffered saline (PBS) containing 0.1% Tween-20 (PBT) and incubated with 5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) for 10 minutes, washed repeatedly with PBT, mounted in PBST containing 25% glycerol, and examined by epifluorescence using a Zeiss Axiophot.

Embryonic cuticle was examined by harvesting embryos from apple juice plates 24 hours after removing adults, dechorionating with bleach, and mounting in Hoyers/lactic acid. Slides were cleared by heating for several hours and then examined by dark field microscopy using a Zeiss Axiophot.

In situ hybridization was performed essentially as described [73], with probe prepared by nick translation of a double-stranded PCR product bearing the bcd ORF using alkali-stable digoxigenin-11-2′-deoxyuridine-5′-triphosphate (Roche), visualizing with alkaline-phosphate coupled sheep anti-digoxigenin Fab fragments (Roche) and subsequent reaction with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium before examination using Nomarski optics on a Zeiss Axiophot.

For hand-sorting of embryos to identify those in nc 9 to 10, we followed the protocol of [74] with the following modifications. Embryos were dechorionated in 100% bleach for 1 minute, rinsed well, and then simultaneously devitellinized and fixed in a 1:1 ratio of ice-cold methanol and heptane for 5 minutes with gentle rocking. The vials were vortexed for 30 seconds, heptane and embryos at the interface removed, and the remaining embryos washed with ice-cold methanol 3 times. Embryos were then rehydrated into PBT with washing for 10 minutes and then stained with 0.1 μg/ml DAPI for 10 minutes. After washing in PBT, embryos were sorted under UV illumination using a Leica MZFLIII stereomicroscope.

For the western blot of S1 Fig, 0- to 2-hour embryo collections from transgenic females were homogenized in sample buffer and transferred to 0.2 μm nitrocellulose following SDS-PAGE. We analyzed samples from 2 independent transgenic HA-Nos and HA-Upf1-Nos lines. The influenza hemagglutinin (HA) epitope was detected with monoclonal C29F4 rabbit anti-HA (3274, Cell Signaling Technology) and the α-tubulin loading control with monoclonal B-5-1-2 (T5168, Sigma), followed by goat anti-rabbit IRDye 800 IgG secondary antibodies (LiCor 926–32211) and goat anti-mouse IRDye 680 IgG secondary antibodies (LiCor 926–68070), with detection and quantitation using a LI-COR Odyssey CLx.

RNAseq and RT-qPCR

Total RNA was extracted from sorted embryos using TRIzol reagent (Thermo Fisher Scientific). Samples were submitted to the Ohio State University Cancer Center Genomics Shared Resource for poly(A) selection and library preparation with TruSeq (Illumina) and subsequently sequenced on a HiSeq4000 (Illumina) sequencer. Paired-end 150-nucleotide (nt) raw reads were first processed and demultiplexed at the sequencing facility and subsequently processed as follows. Adapter sequences were trimmed followed by trimming low-quality bases from the 3′ end of reads based on a minimum Phred score of 15. Reads of less than 60 nt were removed, as were reads mapping to mtRNA, tRNA, and rRNA. STAR (v2.5.3a) was used to map the reads either to the fly genome or the transcriptome [75]. Mapped reads were counted using RSEM (v1.3.0) [76]. The resulting outputs were analyzed in R (v4.2.1) using the edgeR package (v3.40.2) to find differentially expressed genes and transcripts [77]. Low abundance genes and transcripts were filtered out by setting min.count = 30 and min.total.count = 30 in filterByExpr function in the edgeR package, removing genes and transcripts with approximately <2 counts per million mapped reads. All statistical analyses including differential expression and locally estimated scatterplot smoothing (LOESS) were performed using R (v4.2.1). R code is available at https://zenodo.org/doi/10.5281/zenodo.13337141 with updates at https://github.com/marhabaie/dmel-nanos-targets.

To account for oversampling caused by the large depletion of mRNAs targeted by Upf1-Nos, correction factors for these data were calculated essentially as described [36]. We assumed that the most abundantly expressed mRNAs, which are transcribed from housekeeping genes, are not regulated by Upf1-Nos and thus can be used to calculate a correction factor. Rather than relying on a single highly expressed gene (e.g., eEF1α1), an average correction factor was calculated using the 20 most abundantly expressed genes for each Upf1-Nos replicate. Among genes with the highest average read abundance in wild type samples, we selected 20 with a similar rank order among Upf1-Nos and wild-type samples. For each gene, the ratio of read abundance in the Upf1-Nos replicate to its average abundance across the 3 wild-type replicates was calculated. The resulting ratios for all 20 genes were averaged, yielding a correction factor for the replicate (S2 Data). These correction factors were used as normalization factors for the Upf1-Nos replicates in edgeR. A similar approach was used to calculate 3 correction factors for the analysis of individual transcripts (S2 Data).

For RT-qPCR, cDNA was synthesized from 250 ng RNA using random hexamers and qPCR performed using SYBR Green using a 7500 Real Time PCR machine (Applied Biosystems). Triplicate measurements for each primer pair were obtained (for each of the 3 biological replicates). Abundance of each RT product was calculated relative to the abundance of RT product from RpS2 mRNA. Primer efficiencies for all primer pairs were 90% to 100%.

Motif enrichment analysis

Motif enrichment analysis was done using the following DREME command: “dreme -rna -norc -png -e 0.1 -m 10 -o DEOUT -p input.fa”. Parameters were set per the user manual. To prevent outlier large 3′ UTRs (e.g., the approximately 10.5-kb headcase 3′ UTR) from biasing the analysis, only UTRs <2,600 nt (98% of all 3′ UTRs) were analyzed.

Gene ontology enrichment analysis

GO enrichment analysis was performed using DAVID (v6.8) with the default setting and categories (COG_ONTOLOGY, UP_KEYWORDS, UP_SEQ_FEATURE, GOTERM_BP_DIRECT, GOTERM_CC_DIRECT, GOTERM_MF_DIRECT, KEGG_PATHWAY, INTERPRO, SMART).

Yeast interaction assays

Yeast were transformed with the 2-μ plasmids described in S7 Data by a standard lithium acetate/PEG protocol. For 2-hybrid experiments, we used the PJ69-4A strain [78], which is MATa, trp1-901, leu2-3, 112, ura3-52, his3-200, gal4Δ, gal80Δ, LYS2::GAL1-HIS3, GAL2-ADE2, met2::GAL7-lacZ. Double transformants were obtained and grown on minimal SD dropout medium lacking tryptophan and leucine. For 3- and 4-hybrid experiments, we used the YBZ1 strain [79], which is MATa, ura3-52, leu2-3, 112, his3-200, trp1-1, ade2, LYS2::(LexAop)-HIS3, ura3::(lexA-op)-lacZ, LexA-MS2 coat (N55K). For 3-hybrid experiments, double transformants were grown on minimal SD dropout medium lacking uracil and leucine, and for 4-hybrid experiments, triple transformants were grown on SD dropout medium lacking uracil, leucine, and tryptophan.

β-galactosidase activity was measured using Beta-Glo (Promega) essentially as described [79]. Briefly, transformants were grown by diluting saturated overnight cultures into appropriate selective media to early-log phase (OD₆₀₀ = 0.2 to 0.3), and then 40 μl of culture was incubated with the same volume of Beta-Glo reagent for 60 minutes at room temperature in 96-well microplates. Three transformants were grown separately and assayed for each experiment. Samples were analyzed in a Veritas luminometer (Turner Biosystems). The output signal from the luminometer was divided by the OD₆₀₀ to normalize for the number of cells in each sample, and by 1,000 (by convention) to generate a reading of beta-galactosidase activity in arbitrary light units.