Identification of a pyrimidine repeat motif in mRNAs encoding yeast secretome proteins

RNA codons for the major hydrophobic residues (e.g. valine, isoleucine, leucine, methionine, and phenylalanine) are enriched in uracil in their second position [31], although others like proline, alanine, serine and threonine, which can also be found in TMDs, contain cytosine at this position. We speculated, therefore, that triplet repeats of the form, NYN (where N = any nucleotide and Y = pyrimidine—U or C), might be common to proteins destined for translation at the ER (i.e. secreted and membrane proteins). We further speculated that uninterrupted repeats of this nature might be indicative of an RNA localization motif that could, potentially, exist in any frame. Thus, we examined mRNAs encoding yeast secretome proteins for the presence of consecutive pyrimidine repeats every third nucleotide (i.e. YNN, NYN, or NNY) in both the coding and UTR regions, using computational analysis.

We first determined how many such repetitive nucleotide triplets might best differentiate mSMPs from non-mSMPs (i.e. other nuclear-encoded mRNAs). For that, the number of consecutive triplet repeats along an mRNA transcript was scored according to a defined threshold (e.g. 5, 7, 10, 12, and 15 repeats). For a random motif we expected to see a linear correlation between the probability of its appearance(s) in a gene with gene length, as shown in Fig 1A. We examined SECReTE lengths between 5 and 15 (e.g. nucleotide triplet lengths of 5,7,10,12, and 15) and indeed observed a direct correlation between SECReTE count and gene length for SECReTE5 and SECReTE7 (Fig 1A). However, the dependency on gene length becomes significantly weakened for SECReTE10 and above, where motif occurrence at ≥10 triplets correlates poorly with gene length (Fig 1A). This implies the presence of ≥10 consecutive repeats is not a random phenomenon and may be important.

Download:

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

Fig 1. Determination of the number of NNY repeats to use as a threshold for SECReTE.

(A) Correlation between SECReTE count and transcript length. The total SECReTE count was calculated for the coding region of each yeast gene (5904 genes scored) by counting the number of consecutive NNY repeats present in the transcript sequence according to the indicated threshold, and in all three frames. Scatter plots represent the correlation between the SECReTE count and gene length. The SECReTE count does not correlate with gene lengths above a threshold of 10 NNY repeats (SECReTE10). r represents the Pearson correlation coefficient. (B) SECReTE motifs are more abundant in the mRNAs coding for secretome proteins than for non-secretome proteins. SECReTE presence, according to the indicated threshold, was counted in mRNAs coding for secretome (blue) and non-secreted (gray) proteins. Bars represent the fraction of SECReTE positive transcripts at the indicated threshold. SECReTE abundance is significantly higher in secretome mRNAs. *p ≤ 2.28E-13, Chi-squared test. (C) SECReTE10 maximizes the ability to distinguish secretome transcripts. ROC curves were plotted for each of the indicated thresholds. Secretome transcripts were used as the “true positive” set, while non-secretome transcripts were used as the “true negative” set. The AUC (area under the curve) of SECReTE10 was the highest.

https://doi.org/10.1371/journal.pgen.1008248.g001

If SECReTE repeats ≥10 (e.g. termed here as “SECReTE10”) play a role in protein secretion, we expect them to be more abundant in mRNAs encoding secretome proteins, as defined according to Ast et al. [19]. To test this possibility, we divided the complete yeast genome into two gene sets: secretome and non-secretome, and calculated the fraction of transcripts that contain at least one instance of SECReTE per gene in each gene set. We found transcripts coding for secretome proteins are enriched with SECReTE motifs of length >7 (Fig 1B), in comparison to transcripts encoding non-secretome proteins. To test the motif length that gives the most significant separation between secretome and non-secretome transcripts, we evaluated the different thresholds for their ability to classify mSMPs using receiver operator characteristics (ROC) analysis. Bona fide secretome protein-encoding transcripts were used as a true positive set and non-secretome protein-encoding transcripts were defined as true negatives. As seen (Fig 1C), motifs with ≥10 repeats (i.e. SECReTE10 and above) maximally differentiated (in terms of ROC area under the curve) secretome transcripts from non-secretome transcripts. As the occurrence SECReTE10 did not show a dependency upon gene length and gave the most significant separation between secretome and non-secretome transcripts, we used it as the threshold by which to define motif presence in subsequent analyses. Previous studies have used high throughput analyses to quantify the level of enrichment of transcripts on yeast ER-bound ribosomes and ER membranes [22,23]. By comparing the cumulative distribution of the ER enrichment value of SECReTE10-containing transcripts to transcripts lacking SECReTE10, we could verify that a higher fraction of SECReTE10-containing transcripts is indeed enriched on ER-bound ribosomes (S1A and S1B Fig) and ER membranes (S1C Fig). In contrast, SECReTE10-containing transcripts are not enriched on mitochondrial ribosomes, in comparison to transcripts lacking SECReTE10 (S1D Fig).

SECReTE abundance in mSMPs is not dependent on the presence of a TMD

To ascertain whether SECReTE enrichment in mSMPs is not merely due to its presence in encoded TMDs, we determined at which position of the nucleotide triplet in SECReTE10 elements is the pyrimidine (Y) is located: i.e. first (YNN); second (NYN); or third (NNY). We calculated SECReTE10 abundance separately for each position using only the coding sequences (i.e. from start codon to the stop codon; CDS) and without the UTRs. While the signal is present in the second position (Fig 2A; NYN), as expected, it is also abundant in the third position of the codon (Fig 2A; NNY). In contrast, the SECReTE10 element is poorly represented in the first position, YNN (Fig 2A). Importantly, the abundance of NNY-based motifs suggests that the TMD is probably not the only element that confers SECReTE enrichment in the coding sequences of yeast mSMPs. A list of all SECReTE motifs (i.e. ≥10 repeats, either NYN- or NNY-based) present in the coding sequences of the yeast genome and their position within the mRNA is listed in S3 Table.

Download:

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

Fig 2. SECReTE abundance in mSMPs is TMD-independent.

(A) SECReTE is abundant in the second position of the codon. SECReTE abundance was calculated for each codon position separately. SECReTE abundance in mSMPs is most significant in the second codon position (NYN), but significant differences were also detected in in the third position (NNY). *p≤ 9.9E-10, Chi-squared test. (B) SECReTE is also highly abundant in the mRNAs encoding soluble secretome proteins. SECReTE10 presence was examined separately for TMD-containing proteins and soluble secreted proteins. A higher fraction of mRNAs coding for soluble secreted proteins (Secretome without TMD; cyan) contains SECReTE in comparison to non-secretome transcripts, either with or without a TMD (Non-secretome with TMD; dark gray, Non-secretome without TMD; light gray). In the third codon position (NNY), the fraction of soluble secreted proteins is even larger than TMD-containing secretome proteins and is significant. *p≤3.03E-3, Chi-squared test. (C) SECReTE is abundant at the third position after removal of the TMD sequence. SECReTE10 presence was scored in mRNAs coding for membrane proteins after the encoded TMD was removed. SECReTE10 is significantly more abundant in the third position (NNY) in mRNAs encoding secretome proteins (blue) than non-secretome proteins (gray), even after removal of the TMD sequence. *p = 0.01, Chi-squared test. (D) SECReTE is highly abundant in mSMPs after the removal of the SSCR sequence. SECReTE10 presence in the different positions was scored after the regions encoding signal peptides were removed. Similar to the results shown in B, mRNAs coding for soluble secreted proteins (cyan) are enriched with SECReTE, in comparison to non-secretome transcripts without a TMD (light gray). In the third codon position (NNY), the fraction of soluble secreted proteins is larger than for TMD-containing secretome proteins and is significant. *p≤5.93E-3, Chi-squared test.

https://doi.org/10.1371/journal.pgen.1008248.g002

Next, we checked for the presence of SECReTE10 in mRNAs coding for TMD-containing proteins and soluble secreted proteins separately. As expected, more transcripts encoding TMD-containing secretome proteins contain SECReTE10 (i.e. ≥1 SECReTE10) in the second position (NYN) than transcripts that encode soluble secreted proteins (Fig 2B). However, the fraction of transcripts coding for soluble secreted proteins that contain at least one SECReTE10 in the third position (NNY) is even higher. This provides compelling evidence for SECReTE10 enrichment in transcripts that is independent of encoded TMD regions. Correspondingly, when we removed the TMD sequences from mRNAs encoding membrane proteins, we found that these transcripts were no longer enriched with the NYN-based form of SECReTE10 (Fig 2C). In contrast, SECReTE10 remained abundant at the third position, NNY, after TMD removal (Fig 2C). Thus, TMD sequences contribute to NYN-based SECReTE abundance. Finally, we note that the removal of TMD sequences from genes encoding secretome and non-secretome proteins did not alter the overall enrichment of SECReTE in those secretome messages versus non-secretome messages, nor did it change the threshold for differentiating between these groups (S1E Fig).

As contiguous stretches of codons for hydrophobic amino acids (i.e. TMDs) foster SECReTE abundance, we also examined whether the removal of signal sequence coding regions (SSCRs), which encode signal peptides, from secretome genes had an effect upon the computational analysis of the SECReTE score (Fig 2D). However, the results indicate that no significant change in the overall fraction of secretome genes bearing SECReTE is incurred upon SSCR removal and this is not altered by TMD presence or motif position (compare Fig 2D to 2B). Thus, SSCRs do not contribute extensively to SECReTE abundance.

Since SRP depletion does not block the co-translational ER targeting of mRNAs encoding predicted SRP-independent proteins [24], we examined whether SRP-independent transcripts on yeast ER-bound ribosomes are more enriched with SECReTE10 than SRP-dependent ones (S2A–S2D Fig). Using this dataset [24], we found that both SRP-dependent and -independent transcripts contain SECReTE10 (S2B Fig). However, SRP-dependent (i.e. TMD-containing) transcripts essentially bear only the NYN-based motif, whereas both NYN- and NNY-based motifs appear in SRP-independent transcripts (S2B and S2C Fig). Moreover, the subset of mRNAs that remain ER-bound after SRP depletion all appear to be enriched in NNY-based SECReTE (S2D Fig).

SECReTE is the principal repetitive sequence motif in secretome transcripts

Repetitive nucleotide triplet sequence elements [i.e. NNX; where X = is any of five other dinucleotide combinations: K (T/G), M (C/A), R (A/G), S (G/C), or W (A/T)] other than NNY might exist in secretome transcripts and distinguish them from non-secretome transcripts. Thus, we examined for the presence of ≥10 uninterrupted NNX repeats in the coding sequences of secretome and non-secretome proteins. Examination of the yeast genome did not reveal any significant repetitive NNX-based triplet repeats in mSMPs, except for that seen for pyrimidine (i.e. NNY). Moreover, this did not change upon removal of the TMD sequences from the analysis, indicating that concatenation of the regions flanking the TMD does not create SECReTE motifs de novo (S3A and S3B Fig). In contrast, we did find that R (purine) and W repeats are enriched in the third position of a large fraction of non-secretome transcripts, especially after TMD removal (S3A and S3B Fig). This indicates that SECReTE is the principal, if not sole, nucleotide triplet motif of ≥10 repeats in the secretome protein-encoding genes of yeast, although a repetitive NNR motif could be identified in the third position of a subset of the non-secretome genes. No additional motifs in either the first or second positions (i.e. XNN, NXN) were identified in non-secretome genes.

Next, we examined the distribution of SECReTE in the different regions (i.e. 5’UTR, CDS, 3’UTR) of yeast genes. We found that the large majority (>90%) of SECReTE motifs (8211 out of 9003) are present in the CDS regions (S4A Fig, left), however, the overall distribution is biased to the 5’ and 3’UTRs, when normalized for the mean length of these smaller regions (S4A Fig, right). Secretome transcripts (1144) contained ~35% of the total SECReTE motifs, an amount proportionally larger than that of non-secretome transcripts (4760), although the motif is more or less evenly represented in both the CDS and separate UTR regions after normalization for length. Mapping of motif distribution along the entire gene length (after normalization) revealed a uniform distribution in TMD-containing transcripts, but also showed that a number of transcripts encoding soluble proteins (~60) have a preference for SECReTE at the 5’ end, as this distribution could be eliminated upon SSCR removal (S4B Fig). Thus, despite the fact the SSCRs do not highly contribute to overall SECReTE abundance (Fig 2D), this does not exclude the possibility that the motif cannot be present therein. Examination of both NYN- and NNY-based SECReTE motifs in the coding regions showed that both contribute to motif presence at the 5’ end of the same subset of transcripts encoding soluble secreted proteins (S4B Fig).

When comparing SECReTE motifs residing in the CDS to those in the UTRs, we found that CDS motifs tend to consist of RRY repeats rather than NNY-based repeats (S4C Fig). On the other hand, UTR-residing SECReTE motifs are statistically more pyrimidine-rich and, thus, are biased towards the NNY pattern of repeats (S4C Fig). Next, we checked if the UTRs of the secretome transcripts are enriched with pyrimidines in general. Indeed, we found that secretome transcripts have a slightly higher Y content in their UTR (S4D Fig), however, this enrichment disappears after removal of the SECReTE-containing UTRs from the analysis (e.g. 43 and 99 transcripts for the 5’ and 3’UTRs, respectively) (S4E Fig). This implies that SECReTE motifs contribute to the pyrimidine-enrichment of UTRs in secretome transcripts.

SECReTE abundance is not dependent upon codon usage

There is a possibility that SECReTE enrichment results from codon usage of the transcript. To check this possibility, we performed permutation test analysis. In this case, each gene sequence was randomly shuffled (1000 times), while codon usage remained constant. We then calculated the Z-score (i.e. number of standard deviations from the mean) of SECReTE10 for each gene to evaluate the probability of the signal to appear randomly. By looking at Z-score distribution in secretome and non-secretome genes, it can be concluded that SECReTE enrichment in mSMPs is not a random consequence of codon usage (S5A Fig). This conclusion is valid for mSMPs encoding both membrane and soluble proteins (S5B Fig). We also conducted the analysis for each nucleotide position of the codon separately (i.e. for the YNN, NYN, and NNY versions of the motif). For that, we calculated the fraction of genes with a significant Z-score (≥1.96) for each position separately. The fraction of genes with a significant Z-score was larger in secretome genes than in the non-secretome genes at both the second and third positions of the codon (S5C Fig), strengthening the notion that SECReTE is significantly more enriched in those positions. This finding is not dependent on the presence of TMDs, since the fraction of genes with a significant Z-score was larger for both soluble and TMD-containing secretome transcripts, rather than for soluble and TMD-containing non-secretome transcripts (S5D Fig).

Cell wall, signal peptide- and TMD-containing yeast proteins in yeast are enriched with SECReTE

To determine those gene categories that are overrepresented with SECReTE-containing genes, gene ontology (GO) enrichment analysis was conducted. When genes that contain at least one occurrence of SECReTE10 in any of its YNN-, NYN- or NNY-based forms were searched for GO enrichment (using all yeast genes as a background), unsurprisingly, membrane proteins were found to have a high enrichment score (fold enrichment = 1.67) (Fig 3A). The most SECReTE-enriched gene category was that comprising cell wall proteins (fold enrichment = 1.8) (Fig 3A). When 15 NNY repeats served as a threshold, the fold-change enrichment of the cell wall protein category increased to 4.8-fold (Fig 3B). To further characterize the mRNAs enriched with SECReTE, we divided the secretome and non-secretome into subgroups and calculated the fraction of transcripts containing SECReTE10 in each category. In agreement with the GO analysis, more than 90% of mRNAs coding for cell wall proteins possess SECReTE10 (and above) motifs and the cell wall proteins were the most SECReTE-rich overall (Fig 3C). Interestingly, this group also comprised the principal set of transcripts that remain associated with ER-bound ribosomes after SRP depletion (S2E and S2F Fig). In addition, we found that 86% of mRNAs of proteins encoding both TMD and signal-sequence (SS) regions, as well as 84% of TMD-encoding secretome mRNAs, contain SECReTE10 (Fig 3C). Of these, mRNAs encoding tail-anchored (TA) proteins contain the lowest number of transcripts with SECReTE10 in the secretome (Fig 3C). TA proteins are known to translocate to the ER through an alternative pathway (GET) after being translated in the cytosol [32–34], and their transcripts are not enriched on ER membranes [22,23] either before or after SRP depletion [24]. This implies that SECReTE is more abundant in mRNAs undergoing translation on the ER. In contrast, transcripts for non-secretome proteins (i.e. mitochondrial and cytonuclear) have the lowest abundance of SECReTE elements (Fig 3C).

Download:

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

Fig 3. Cell wall proteins are highly enriched with SECReTE.

(A) Cellular component ontology analysis for genes containing SECReTE10. Genes encoding cell wall proteins, as well as membrane proteins, show the highest and most significant enrichment score. (B) Cellular component ontology analysis for genes containing SECReTE15. Genes encoding cell wall proteins are the most enriched with SECReTE. (C) SECReTE10 abundance in different groups of genes. More than 90% of mRNAs encoding proteins annotated to localize to the cell wall contain SECReTE. High SECReTE abundance was also noticed in other secretome groups except tail-anchored (TA) proteins. Mitochondrial mRNAs (Mito) have low SECReTE abundance. Numbers above bars represent the number of genes in each group. (D) MEME analysis of cell wall transcripts. A motif similar to SECReTE was revealed in cell wall transcripts using MEME. Numbers on the x axis indicate base number.

https://doi.org/10.1371/journal.pgen.1008248.g003

Since SECReTE is highly enriched in mRNAs coding for cell wall proteins, we wanted to check if it could be discovered using an unbiased motif search tool. For that, we analyzed the mRNA sequences of cell wall proteins using MEME to identify enriched mRNA motifs. The most significant result obtained highly resembled the SECReTE10 repeat with either U or C (Fig 3D). Importantly, we did not detect a protein motif within this mRNA motif, eliminating the possibility that the SECReTE element is dependent on a specific protein sequence.

SECReTE enrichment in secretome transcripts occurs in both prokaryotes and higher eukaryotes

Conservation or convergence in evolution are often strong indications of functional significance. To check whether SECReTE enrichment in mSMPs is found in additional organisms (e.g. humans and B. subtilis) we analyzed other genomes. In humans, as in S. cerevisiae, SECReTE10 gave the most significant separation between RNAs encoding secretome and non-secretome proteins, based on ROC analysis (Fig 4A). After verifying that SECReTE10 does not correlate with gene length, SECReTE10 served as a threshold to define presence of the SECReTE motif. As in yeast, SECReTE is enriched in the second and third codon positions of secretome transcripts, in comparison to non-secretome transcripts (Fig 4B). Also, a larger fraction of secretome transcripts that lack TMDs contain the NNY-based SECReTE, as compared to non-secretome transcripts bearing TMDs (Fig 4C). Thus, the SECReTE motif is present in higher organisms. A list of all SECReTE10 and higher motifs found in human genes is given in S4 Table. Interestingly, unlike yeast, we found that a disproportionally large majority of motifs (29753 out of 52,047) are present in the UTRs instead of being in the CDS, especially after normalization for length, and this phenomenon is observed for both secretome and non-secretome transcripts (S6A Fig). Therefore, in contrast to yeast, the UTRs and especially the 3’UTR are preferential sites for SECReTE location in human transcripts. Like yeast, however, RRY enrichment is observed for SECReTE motifs in the CDS regions, while high pyrimidine content is observed in the UTRs (S6B Fig). Therefore, although both yeast and human share the same SECReTE motifs, their distribution over gene region appears different. Interestingly, human transcripts encoding glycophosphatidylinositol (GPI)-anchored proteins, which are equivalent to cell wall proteins, were found to be highly enriched with SECReTE. In fact, a SECReTE-like motif was previously shown to confer the translation-independent localization of a transcript encoding human GPI-anchored protein, placental alkaline phosphatase, to the ER [35] In contrast, tail-anchored genes, as well as mitochondrial and cytonuclear genes, have a low SECReTE abundance as seen in yeast (Fig 4D). Finally, we also detected a high abundance of SECReTE10 in genes encoding secretome proteins from B. subtilis, in comparison to those encoding non-secretome proteins (Fig 4E). Thus, SECReTE motifs are also present in prokaryotic genomes.

Download:

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

Fig 4. SECReTE enrichment in the secretomes of yeast, bacteria, and humans.

(A) SECReTE10 maximizes the ability to classify secretome genes in human. ROC curves were plotted for each of the indicated thresholds. Secretome genes were used as the true positive set and non-secretome genes as the true negative set. The AUC (area under the curve) of SECReTE10 was the highest. B. SECReTE is highly abundant in the mRNAs of human secretome proteins. SECReTE10 abundance was calculated for each codon position separately. SECReTE abundance in human mSMPs is most significant in the second position of the codon, but highly significant differences were also detected in the third position. *p≤ 3.73E-68, Chi-square test C. SECReTE is highly abundant in mRNAs coding for soluble secretome proteins in humans. SECReTE10 presence was examined separately for TMD-containing proteins and soluble secreted proteins. A higher fraction of mRNAs coding for soluble secreted proteins (Secretome without TMD; cyan) contains SECReTE in comparison to non-secretome transcripts without a TMD (light gray). The fraction of soluble secreted proteins having SECReTE in the third position is larger than that of TMD-containing non-secretome proteins (NNY) and is significant. n represent the number of genes in each group. *p ≤ 3.49E-12, Chi-square test. (D) SECReTE10 abundance in different groups of human genes. High SECReTE abundance was observed for other secretome protein groups, except tail-anchored (TA) proteins. Mitochondrial mRNAs (Mito) have low SECReTE abundance. Numbers above bars represent the number of genes in each group. (E) SECReTE10 abundance in B. subtilis. SECReTE10 abundance was scored and was observed to be higher in mRNA coding for genes encoding secretome proteins (i.e. SS&TMD, TMD, and SS) as compared to those encoding non-secretome (Non-Sec) proteins. Numbers under bars represent the number of genes in each group. (F) SECReTE10 abundance in S. pombe SECReTE10 abundance was calculated for each codon position separately for TMD-containing proteins and soluble secreted proteins. A higher fraction of mRNAs coding for soluble secreted proteins (Secretome without TMD; cyan) contains SECReTE in comparison to non-secretome transcripts, either with or without a TMD (Non-secretome with TMD; dark gray, Non-secretome without TMD; light gray). The fraction of soluble secreted proteins having SECReTE in the third position is larger than that of TMD-containing non-secretome proteins (NNY) and is significant. n represent the number of genes in each group. n represent the number of genes in each group. * p≤ 5.63E-3, Chi-square test.

https://doi.org/10.1371/journal.pgen.1008248.g004

We next asked whether SECReTE is conserved evolutionarily via inheritance. To differentiate between conservation and possible convergence we analyzed the genome of the fission yeast, S. pombe, for the presence and position of SECReTE in secretome and non-secretome transcripts. As found for S. cerevisiae, SECReTE is enriched (in both NYN- and NNY-based forms) in a larger fraction of S. pombe mSMPs that lack TMDs, as compared those containing TMDs or to non-secretome transcripts that either bear or lack TMDs (Fig 4F). Next, we aligned orthologous genes encoding secretome proteins from S. cerevisiae to those of S. pombe (457 genes total), and examined whether SECReTE is found in the same (i.e. aligned) position within the gene. We found that the coordinates of SECReTE motifs in the large majority (e.g. 393 out of 457) of ortholog pairs were non-aligned. This might imply that the majority of SECReTE motifs arose through convergent evolution, although we cannot rule out drift of the motif after species divergence. Nonetheless, it is clear that SECReTE is present in all species examined by us, from prokarya to eukarya, the latter including yeasts and mammals.

Design of mutations in SECReTE to examine the effects upon protein secretion

To further understand the significance of SECReTE and validate its importance to yeast cell physiology, we examined its relevance by elevating or decreasing the signal in selected genes. Three representative genes were chosen, based on their relatively short gene length, a detectable phenotype upon their deletion, and their function in different physiological pathways. These genes included: SUC2, which encodes a soluble secreted periplasmic enzyme; HSP150, which encodes a soluble media protein; and CCW12, which encodes a GPI-anchored cell wall protein. The overall SECReTE signal of the genes was increased by substituting any A or G found in the third codon position with a T or C, respectively, thereby enriching SECReTE presence along the entire gene [(+)SECReTE]. The reverse substitution, converting T to A or C to G, decreased the overall SECReTE signal [(-)SECReTE]. We note that we added or removed only NNY-based triplet motifs, in order not to change the amino acid sequence of the encoded protein. The number of motifs present in each gene before and after SECReTE addition/reduction is shown in S5 Table. Thus, in the case of HSP150 several NYN-based SECReTE motifs remain in the (-)SECReTE mutant. Changes in the stability of the mRNA secondary structure (free energy) and the Codon Adaptation Index (CAI) [32] were kept to within a similar range (S5 Table). SECReTE mutations in SUC2, HSP150, and CCW12 are shown along the length of the gene, using a minimum threshold of either 1 NNY repeats or 10 NNY repeats, as shown in S7 Fig (S7A–S7C Fig; upper and lower parts, respectively).

SECReTE mutations in SUC2 alter invertase secretion

SUC2 codes for different forms of invertase translated from two distinct mRNAs, short and long, which differ only at their 5’ ends. While the longer mRNA codes for a secreted protein that contains a signal sequence, the signal sequence is omitted from the short isoform, which codes for a cytoplasmic protein. Secreted Suc2 expression is subjected to glucose repression; however, under inducing conditions (i.e., glucose depletion), Suc2 is trafficked through the secretory pathway to the periplasmic space of the cell. There, it catalyzes the hydrolysis of sucrose to glucose and fructose, this enzymatic activity being responsible for the ability of yeast to utilize sucrose as a carbon source and can be measured by a biochemical assay (i.e. invertase activity), both inside and outside of the cell. The effect of SECReTE mutations on Suc2 function was tested by examining the ability of mutants to grow on sucrose-containing media by drop-test. Interestingly, the growth rate of SUC2(-)SECReTE on sucrose plates was decreased, while the SUC2(+)SECReTE mutant exhibited better growth in comparison to WT cells (Fig 5A), even though no growth change was detected on YPD plates. These findings suggest that SECReTE strength affects the secretion of Suc2. These changes in Suc2 secretion could result from changes in SUC2 transcription, Suc2 production, and/or altered rates of secretion. To distinguish between possibilities, WT cells, suc2Δ, and SUC2 SECReTE mutants were subjected to invertase assays. The invertase assay enables the quantification of secreted Suc2, as well as internal Suc2, by calculating the amount of glucose produced from sucrose. As expected, under glucose repressing conditions (e.g. 2% glucose) the levels of both secreted and internal Suc2 were very low. When cells were grown on media containing low glucose (e.g. 0.05% glucose) to promote the expression of the secreted enzyme, secreted Suc2 levels were altered due to changes in SECReTE. Corresponding to the drop-test results (Fig 5A), a significant decrease in secreted invertase was detected with SUC2(-)SECReTE cells, while a significant increase was detected with SUC2(+)SECReTE cells, in comparison to WT cells (Fig 5B). Importantly, SUC2(+)SECReTE cells were found to secrete nearly 2-fold (92.2±9.2%, p <0.016) more invertase than SUC2(-)SECReTE cells, while no Suc2 secretion was detected from suc2Δ cells (Fig 5B, secreted). If SECReTE mutations affect the efficiency of Suc2 secretion, but not its synthesis, then Suc2 should accumulate in SUC2(-)SECReTE cells corresponding to the difference secreted from SUC2(+)SECReTE cells. However, this was not the case as the internal amount of Suc2 decreased in SUC2(-)SECReTE cells and slightly increased in SUC2(+)SECReTE cells (Fig 5B, internal). These findings suggest that SECReTE alterations in SUC2 might affect the level of protein production. We next examined the rate of invertase secretion for WT, SUC2(+)SECReTE, and SUC2(-)SECReTE cells shifted to low glucose medium for varying amounts of time (Fig 5C). The results show that the average maximal rate of secretion from SUC2(+)SECReTE cells is slightly higher than for WT cells (i.e. 0.479±0.016 vs. 0.432±0.013 units per min per O.D.₆₀₀ unit of cells; ±standard deviation, n = 3 experiments), and was significantly (62.4±7.8%; p <0.0001) higher than of SUC2(-)SECReTE cells (0.295±0.022 units/min per O.D.₆₀₀ unit of cells; ±standard deviation, n = 3 experiments). In contrast, the time required to achieve half-maximal secretion between SUC2(+)SECReTE and SUC2(-)SECReTE cells was relatively unchanged under the experimental conditions (i.e. ~74 min; R² values = >94). Thus, the presence of SECReTE affects not only invertase production and overall secretion, but also its rate of secretion from yeast.

Download:

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

Fig 5. The levels of secretion of endogenous and exogenous proteins are affected by SECReTE strength.

(A) SECReTE enhances the ability to grow on sucrose. The ability of WT, suc2Δ, SUC2(+)SECReTE and SUC2(-)SECReTE yeast to grow on sucrose was examined by drop-test. Cells were grown to mid-log on glucose-containing YPD medium, prior to serial dilution and plating onto sucrose-containing synthetic medium or YPD. Cells were grown for 2 days prior to photo-documentation. The SUC2(-)SECReTE mutant exhibited reduced growth than WT cells, while SUC2(+)SECReTE cells exhibited better growth. suc2Δ cells were unable to grow on sucrose-containing medium. (B) SECReTE enhances invertase secretion. The indicated strains from A were analyzed using an invertase secretion assay. Both internal and secreted invertase activity was measured in units (1 U = 1 μmol glucose released/min per O.D.₆₀₀ unit) after glucose de-repression. Both activities were reduced in SUC2(-)SECReTE cells and elevated in SUC2(+)SECReTE cells. Error bars represent the standard deviation from three experimental repeats. *p<0.0161 (t-test). (C) SECReTE presence enhances the rate of invertase secretion. The indicated strains from A were incubated in low glucose medium for varying times (0-150min) and analyzed using the invertase secretion assay, as in B. Secreted invertase activity was measured in units (1 U = 1 μmol glucose released/min per O.D.₆₀₀ unit) after glucose de-repression. Error bars represent the standard deviation from three experimental repeats. Linear regression of the data (dotted lines) from inverse reciprocal plots was used to determine the T_1/2 for half-maximal accumulation and rate; “x” mark on line indicates time of half-maximal accumulation. The rate of invertase secretion is enhanced in (+)SECReTE cells and reduced in (-)SECReTE cells relative to WT. (D) SECReTE enhances the ability to grow on calcofluor white. The ability of WT, hsp150Δ, HSP150(+)SECReTE and HSP150(-)SECReTE cells to grow on CFW was examined by drop-test. Cells were grown to mid-log on YPD, prior to serial dilution and plating on YPD alone or YPD plates containing CFW, and incubated at 30°C. Cells were grown for 2 days prior to photodocumentation. The HSP150(-)SECReTE mutant exhibited hypersensitivity in comparison to WT cells, while HSP150(+)SECReTE cells were less sensitive. hsp150Δ cells grew poorly on medium containing CFW. (E) SECReTE enhances Hsp150 secretion. The indicated strains from D were subjected to the Hsp150 secretion assay. Cells were grown to mid-log phase at 37°C for 4hrs and examination in cell lysates (internal) or medium (external) by Western analysis using anti-Hsp150 antibodies. External Hsp150 was decreased in HSP150(-)SECReTE cells in comparison to WT, while it was increased in the HSP150(+)SECReTE strain. Internal Hsp150 was decreased in HSP150(-)SECReTE cells and also slightly in HSP150(+)SECReTE cells, in comparison with WT cells. No internal nor external Hsp150 was detected in the lysate or medium derived from hsp150Δ cells, respectively. Band intensity was quantified using ImageJ and presented in the histogram below. A representative experiment is shown in the top panels for both external and internal Hsp150 secretion. The graphs below represent the ratio of the intensity of all samples relative to that of WT for three biological repeats, *p<0.05 (t-test). Error bars represent the standard deviation. (F) SECReTE enhances the ability to grow on hygromycin B. The ability of WT, ccw12Δ, and CCW12(-)SECReTE cells to grow on HB was examined by drop-test. Cells were grown to mid-log on glucose-containing YPD medium, prior to serial dilution and plating onto HB-containing YPD or YPD alone. Cells were grown for 2 days prior to photodocumentation. The CCW12(-)SECReTE strain was more sensitive to HB stress in comparison to WT cells. ccw12Δ cells were unable to grow on medium containing HB. (G) SECReTE enhances secretion of an exogenous protein, SSGAS1-GFP. Yeast expressing SSGAS1-GFP-3’GAS1^UTR(+)SECReTE, SSGAS1-GFP, SSKAR2-GFP, GFP, and SSGAS1-LacZ from plasmids were grown to mid-log phase on synthetic medium containing 2% raffinose and shifted to 3% galactose-containing medium for 4hrs. Proteins expressed from the different strains were TCA precipitated from the medium and the precipitates resolved by SDS-PAGE. GFP was detected with an anti-GFP antibody, while Hsp150 was detected with an anti-Hsp150 antibody and was used as a loading control. Band intensity was quantified using ImageJ; intensity was scored relative to SSGAS1-GFP secretion. Addition of the GAS1 3’UTR mutated to contain SECReTE improved the secretion of SS-Gas1 and was comparable to that of SSKAR2-GFP. GFP lacking a signal sequence (GFP) was not secreted and SSGAS1-LacZ was used as a negative control for GFP detection.

https://doi.org/10.1371/journal.pgen.1008248.g005

SECReTE mutations alter Hsp150 secretion and cell wall stability

Next, we wanted to study the importance of SECReTE in HSP150. Hsp150 is a component of the outer cell wall and while the exact function of Hsp150 is unknown, it is required for cell wall stability and resistance to cell wall-perturbing agents, such as Calcofluor White (CFW) and Congo Red (CR). While hsp150Δ cells are more sensitive to cell wall stress, the overproduction of Hsp150 increases cell wall integrity [36]. Hsp150 is secreted efficiently into the growth media and its expression is increased upon heat shock [37,38]. The effect of modifying SECReTE in HSP150 was examined via drop-test by testing the sensitivity of HSP150(-)SECReTE and HSP150(+)SECReTE cells to added CFW, in comparison to WT and hsp150Δ cells. As can be seen from Fig 5D while the HSP150(-)SECReTE strain was more sensitive to CFW as compared to WT cells, the HSP150(+)SECReTE strain was more resistant to CFW. As expected, hsp150Δ cells are the most susceptible to CFW (Fig 5D). HSP150 strains were also subjected to Western blot analysis to measure levels of the mutant proteins. Since HSP150 secretion is elevated upon heat-shock [37,38], cells were shifted to 37°C before protein extraction. Protein was extracted from both the growth medium and cells to detect both external and internal protein levels, respectively. The amount of Hsp150 secreted to the medium was decreased in HSP150(-)SECReTE cells and elevated in HSP150(+)SECReTE cells, in comparison to WT cells (Fig 5E). Similar to Suc2, the internal amount of Hsp150 was decreased in HSP150(-)SECReTE cells, relative to WT cells, and showed a greater reduction than that seen in the external form (Fig 5E), despite the fact that several NYN-based motifs remain in the gene. As the internal level of Hsp150 in HSP150(+)SECReTE cells was more or less unchanged relative to WT cells, we concluded that SECReTE alteration in HSP150 may also affect protein production.

SECReTE mutations in CCW12 alter cell wall stability

CCW12 encodes a GPI-anchored cell wall protein that localizes to regions of the newly synthesized cell wall and maintains wall stability during bud emergence and shmoo formation. Deletion of CCW12 was shown to cause hypersensitivity to cell wall destabilizing agents, like hygromycin B (HB) [39,40]. Since the SECReTE score is very high in CCW12, it was not possible to further increase the signal. Therefore, we generated only CCW12(-)SECReTE cells and tested their ability to grow on HB-containing plates. As seen with HSP150(-)SECReTE (Fig 5D), we found that the CCW12(-)SECReTE mutation rendered cells sensitive to cell wall perturbation, in comparison to WT cells (Fig 5F).

SECReTE addition affects secretion of an exogenous naïve protein

The ability of SECReTE addition to improve the secretion of an exogenous protein would not only be substantial evidence for its importance, but also could be a useful, low-cost, industrial tool to improve the secretion of recombinant proteins without changing protein sequence. To test that, we employed a GFP transcript construct bearing the encoded signal sequence (SS) of Gas1 (SSGAS; SSGas1) at the 5’ end. SSGas1 addition enables the secretion of GFP protein to the medium, although its secretion was not as efficient in comparison to other SS-fused GFP proteins, such as SSKar2 (Fig 5G). To potentially improve the secretion of SSGAS-GFP, we added an altered 3’UTR sequence of GAS1 that contained SECReTE [i.e. in which all A’s and G’s were replaced with T’s and C’s, respectively; SSGAS-GFP-GAS1^3’UTR(+)SECReTE]. We then tested the effect of SECReTE addition upon the secretion of GFP into the media. We found that the addition of SECReTE to the 3’UTR of SSGAS-GFP improved the secretion of GFP secretion into the media, in comparison to SSGAS-GFP, and was similar to that of SSKar2-GFP construct (Fig 5G). GFP expression without the signal sequence was unable to be secreted (Fig 5G).

The effect of SECReTE mutations on mRNA levels

As protein levels may be altered by (-)SECReTE and (+)SECReTE mutations (Fig 5B, 5E and 5G), we examined whether changes in gene transcription or mRNA stability are involved. Quantitative real-time (qRT) PCR was employed to check whether mRNA levels of SUC2, HSP150, and CCW12 are affected by SECReTE strength. We found that SUC2(–)SECReTE mRNA levels were almost 30% lower than in SUC2 WT cells, while SUC2(+)SECReTE levels were ~50% higher than WT (S8A Fig). This change in mRNA levels might contribute to the ability of SUC2(+)SECReTE mutant to increase protein production and, therefore, grow better on sucrose-containing medium (Fig 5A–5C).

The effect of SECReTE mutation on HSP150 mRNA levels was also studied. We found that the mRNA level of HSP150(-)SECReTE was similar to WT, while that of HSP150(+)SECReTE was slightly decreased (S8B Fig). Thus, the change in Hsp150 protein levels and sensitivity to CFW due to SECReTE alteration (Fig 5D and 5E) is not entirely explained by changes in mRNA levels. SECReTE mutations in CCW12(-)SECReTE did not cause a significant change in its mRNA level (S8C Fig), therefore, the increased sensitivity of CCW12(-)SECReTE to HB (Fig 5F) is probably not due to a decrease in CCW12 mRNA.

The addition of SECReTE motifs increases mRNA localization to the ER

To test whether SECReTE has a role in dictating mRNA localization, we visualized the SUC2 and HSP150 mRNAs by single-molecule FISH (smFISH) using specific fluorescent probes and tested the influence of SECReTE alteration on the level of mRNA co-localization with the ER. We used Sec63-GFP as an ER marker and calculated the percentage of mRNA granules (spots) per cell that co-localized with cortical and perinuclear ER (cER and nER, respectively) or were not localized to the ER. We note that probes to the native gene sequences were used to measure the level of mRNA localization in WT cells as well as in (+)SECReTE or (-)SECReTE cells. The number of FISH spots per cell was variable for both mRNAs, with (-)SECReTE cells having less spots than WT cells, while (+)SECReTE cells had more spots per cell than WT cells (S9A Fig). This largely reflects the results obtained by qRT-PCR for SUC2 (S8A Fig), however, we cannot discount the possibility that SECReTE alterations lessen the level of mRNA hybridization with the probe set and, thus, underestimate RNA localization to some degree.

We found that the level of co-localization between SUC2(-)SECReTE mRNA granules and Sec63-GFP was significantly less in comparison to SUC2(+)SECReTE mRNA granules (e.g. 56.7±1.5% vs. 74.1±1.7% co-localization, respectively; p = 2.0E-13) (Figs 6A, 6B and S9B), while being similar to native SUC2 (e.g. 56.3±2.4% ER co-localization). This finding suggests that the number of SECReTE motifs influences mRNA localization to the ER, in addition to enhancing secretion. We also found that there were fewer granules present in SUC2(-)SECReTE cells than observed in either SUC2(+)SECReTE or WT cells (S9A Fig), which corresponds with the qRT-PCR results (S8A Fig). Finally, we note that no specific ER subdomain (i.e. cER or nER) was preferentially labeled upon the increase in SECReTE motifs (Fig 6B).

Download:

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

Fig 6. SECReTE abundance enhances mRNA localization to the ER.

(A) Visualization of endogenously expressed SUC2(+)SECReTE and SUC2(-)SECReTE mRNAs using smFISH. Yeast endogenously expressing WT SUC2, SUC2(+)SECReTE, or SUC2(-)SECReTE and Sec63-GFP from a plasmid were grown to mid-log phase on SC medium containing 2% glucose prior to shifting cells to low glucose-containing medium (0.05% glucose) to induce SUC2 expression. Cells were processed for smFISH labeling using non-overlapping, TAMRA-labeled, FISH probes complementary to SUC2, prior to labeling with DAPI (shown in merge). Representative images are shown. Line = 1μm. B. Quantification of SUC2 and HSP150 (+)SECReTE and (-)SECReTE mRNA localization to the ER. mRNA granule scoring was performed using the FISH-quant algorithm and co-localization of the granules to the ER (both cER and nER) was performed (see Materials and Methods). The percentage of granules that either co-localized or did not co-localize with Sec63-GFP-labeled cER or nER was scored for each cell. The histogram shows the average score for at least ~50 cells and ~250 granules for each strain examined. SUC2 p values: (+)SECReTE vs. (-)SECReTE, p < 3.1E-8; WT vs. (+)SECReTE, p < 8.1E-10; HSP150 (+)SECReTE vs. (-)SECReTE, p < 0.008 (t-test). C. Analysis of SUC2 mRNA localization using cell fractionation and qRT-PCR. WT SUC2, SUC2(+)SECReTE, or SUC2(-)SECReTE yeast cells were fractionated to membrane and cytosol fractions. RNA was extracted from both fractions and subjected to qRT-PCR analysis to quantify the levels of SUC2 mRNA in each fraction. Primers were used to amplify the long transcript of SUC2, which encodes the secreted protein. Primers for actin were used for normalization. SUC2(-)SECReTE cells exhibited lower SUC2 mRNA levels in both fractions than WT and SUC2(+)SECReTE, indicating a loss in stability as well as localization. Error bars represent the standard deviation of three biological repeats.

https://doi.org/10.1371/journal.pgen.1008248.g006

We next examined the level of HSP150 mRNA localization to the ER in HSP150(+)SECReTE or HSP150(-)SECReTE cells. We found that as with SUC2, the addition of SECReTE motifs increased the level of ER localization (Fig 6B and S9C Fig) from 63.7±2.0% to 77.9±1.6% over native HSP150 localization (p = 7.0E-8). In contrast, no change in the level of HSP150(-)SECReTE mRNA co-localization was observed (e.g. 64.0±1.6%), which perhaps reflects presence of the NYN-based motifs that could not be mutated without altering the amino acid sequence. Overall, however, both sets of results show that SECReTE addition to an mRNA increases the pattern of ER localization. To substantiate the smFISH results for SUC2, we also performed the subcellular fractionation of cells expressing native SUC2, SUC2(-)SECReTE, or SUC2(+)SECReTE to obtain crude membrane (containing ER) and cytosolic fractions and quantified the distribution of mRNA using qRT-PCR (Fig 6C). After normalization using actin mRNA as a control, the results indicated that SUC2(-)SECReTE mRNA is less abundant overall (as observed above in S8A Fig and S9A Fig) and appeared to be less membrane-associated than either native or SUC2(+)SECReTE mRNA by ~40%. Taken altogether, our results imply that SECReTE presence/addition stabilizes secretome mRNAs, increases mRNA localization to the ER, and enhances both protein production and secretion.

Identification of potential SECReTE-binding proteins

To further elucidate the role of SECReTE it is essential to identify its binding partners, presumably RBPs. Large-scale approaches were previously used to identify mRNAs that are bound >40 known RBPs in yeast [41–43]. To obtain a list of potential SECReTE-binding proteins (SBPs) we searched the datasets for RBPs that bind mRNAs highly enriched with SECReTE. For each RBP, we calculated what fraction of its bound transcripts contain SECReTE10. RBPs found to bind large fractions of SECReTE10-containing mRNAs included Bfr1, Whi3, Puf1, Puf2, Scp160, and Khd1 (Fig 7A), and were all previously shown to bind mSMPs [41–43]. To test which of these candidates bind SECReTE, each of the genes these RBPs was deleted in either WT or HSP150(+)SECReTE cells. We hypothesized that the deletion of a genuine SBP might confer hypersensitivity to CFW and eliminate the growth rate differences between WT and HSP150(+)SECReTE cells observed on CFW-containing plates (Fig 5D). When PUF1, PUF2, or SHE2 were deleted we found that HSP150(+)SECReTE strain was still more resistant to CFW than WT cells (S10 Fig). One possible explanation for this lack of effect is that these RBPs either do not bind HSP150 or that they are redundant with other SBPs. However, we did find that the deletion of either WHI3 or KHD1 eliminated the differences between WT and HSP150(+)SECReTE strains on CFW-containing plates (Fig 7B). This suggests Whi3 and Khd1 bind HSP150 mRNA and possibly other secretome mRNAs, and even WT cells alone were rendered more sensitive to CFW in their absence (Fig 7B).

Download:

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

Fig 7. Identification of potential SECReTE-binding proteins.

(A) Identification of SECReTE10-containing transcripts in RNA-binding protein pulldown studies. The number and fraction of SECReTE10-containing mRNAs from the total mRNAs bound to the indicated RBPs is shown. The microarray analysis data used to generate the histogram was published in references [41–43]. (B) Identification of potential SECReTE-binding partners. WT cells and either WT or HSP150(+)SECReTE cells deleted for genes encoding the indicated RBPs (e.g. Whi3, and Khd1) were grown to mid-log phase on YPD at 30°C, prior to serial dilution and plating onto either solid YPD medium or YPD containing CFW. Yeast were grown 2 days prior to photodocumentation.

https://doi.org/10.1371/journal.pgen.1008248.g007

Yeast strains and plasmids used

The yeast strains and plasmids used in this study are listed in S1 Table and S2 Table, respectively.

Yeast strains, genomic manipulations, and growth conditions

Yeasts were grown at the indicated temperature either in a standard growth medium (1% Yeast Extract, 2% Peptone, 2% Dextrose) or synthetic medium containing 2% glucose [e.g., synthetic complete (SC) and selective SC dropout medium lacking an amino acid or nucleotide base] [63]. Deletion strains using the NAT antibiotic resistance gene in WT (BY4741) cells were created using standard LiOAc transformation procedures and with nourseothricin (100μg/ml) for selection on synthetic solid medium. For the creation of SECReTE mutant strains, SECReTE gene fragments were designed with the appropriate modifications, from the first to the last mutated base, and synthesized either as a gBlock (Integrated DNA Technologies, Inc., Coralville, IA, USA) or cloned into a pUC57-AMP vector (Bio Basic Inc.). Both (-)SECReTE and (+)SECReTE strains were generated. SUC2(-)SECReTE, SUC2 (+)SECReTE and CCW12(-)SECReTE strains were constructed in the BY4741 background genome using the delitto perfetto method for genomic oligonucleotide recombination [64], in which the CORE cassette from pGKSU [64] was integrated first into the genomic region corresponding to site of the SECReTE gene fragment. The CORE cassette contains the URA3 selection marker with an I-SceI homing endonuclease site and a separate inducible I-SceI gene. The SECReTE gene fragment for CCW12(-)SECReTE was amplified from the synthetic gBlock using primer sequences containing 20 bases of homology to both the region outside of the desired genomic locus and the CORE cassette. The amplified SECReTE gene fragment subsequently replaced the CORE cassette in the desired genomic site through an additional step of integration. CRISPR/Cas9 was utilized instead to generate the HSP150 mutant strains. HSP150(-)SECReTE and HSP150(+)SECReTE were created in the BY4741 genome. The CRISPR/Cas9 procedure involved deletion of the native genomic region corresponding to the SECReTE gene fragment, using the NAT cassette from pFA6-NatMX6. A CRISPR/Cas9 plasmid vector was designed to express the Cas9 gene, a guide RNA that targets the NAT cassette, and the LEU2 selection marker. The CRISPR/Cas9 plasmid was co-transformed with the amplified SECReTE gene fragment to replace the NAT cassette. Standard LiOAc-based protocols were employed for transformations of plasmids and PCR products into yeast. Transformed cells were then grown for 2–4 days on selective media. Correct integrations were verified at each step using PCR and, at the final step, accurate integration of the (-)SECReTE or (+)SECReTE sequences was confirmed by DNA sequencing.

Quantitative RT-PCR (qRT-PCR)

RNA was extracted and purified from overnight cultures using a MasterPure Yeast RNA Purification kit (Epicentre Biotechnologies). For each sample, 2μg of purified RNA was treated with DNase (Promega, Madison, WI, USA) for 2hrs at 37°C and subjected to reverse transcription (RT) using Moloney murine leukemia virus RT RNase H(-) (Promega) under the recommended manufacturer conditions. Primer pairs were designed, using NCBI Primer-Blast [65], to produce only one amplicon (60-70bp). Standard curves were generated for each pair of primers and primer efficiency was measured. All sets of reactions were conducted in triplicate and each included a negative control (H₂O). qRT-PCR was performed using a LightCycler480 device and SYBR Green PCR Master Mix (Applied Biosystems, Waltham, Massachusetts, USA). Two-step qRT-PCR thermocycling parameters were used as specified by the manufacturer. Analysis of the melting curve assessed the specificity of individual real-time PCR products and revealed a single peak for each real-time PCR product. The ACT1 or UBC6 RNAs were used for normalization and fold-change was calculated relative to WT cells.

Drop test growth assays

Drop test assays were performed by growing yeast strains in YPD medium to mid-log phase and then performing serial dilution five times (10-fold each) in fresh medium. Cells were spotted onto plates with different conditions and incubated for 48hrs, prior to photo-documentation. Calcofluor White (CFW) or Hygromycin B (HB) sensitivity was tested by spotting cells onto YPD plates containing either 25μg/ml HB or 50μg/ml CFW (dissolved in DMSO, prepared as described) [66], following the protocol as mentioned above.

Hsp150 and GFP secretion assays

For the induction of Hsp150 secretion, strains were grown in YPD overnight at 26°C, diluted in YPD medium to 0.2 O.D.₆₀₀ units, and then incubated at 37°C and grown until log-phase. For GFP secretion, yeast were grown O/N to 0.2 O.D.₆₀₀ at 30°C in synthetic selective medium containing raffinose as a carbon source, then diluted to 0.2 O.D.₆₀₀ units in YP-Gal and grown to mid-log phase (0.6–0.8 O.D.₆₀₀) at 30°C. Next, 1.8ml of the culture was taken from each strain and centrifuged for 3mins at 1900 x g at room temperature. Trichloroacetic Acid (100% w/v) protein precipitation was performed on the supernatant and protein extraction, using NaOH 0.1M, was performed on the pellet [67]. Samples were separated on SDS-PAGE gels, blotted electrophoretically onto nitrocellulose membranes, and detected by incubation with rabbit anti-Hsp150 [1:10,000 dilution; gift from Jussi Jäntti (VTT Research, Helsinki)] or monoclonal mouse anti-GFP (Roche Applied Science, Penzberg, Germany) antibodies followed by visualisation using the Enhanced Chemiluminescence (ECL) detection system with anti-rabbit peroxidase-conjugated antibodies (1:10,000, Amersham Biosciences). Protein markers (ExcelBand 3-color Broad Range Protein Marker PM2700, SMOBiO Technology, Inc., Hsinchu, Taiwan) were used to assess protein molecular mass.

Invertase assay

Invertase secretion was measured as described previously [68]. Cell preparation for the invertase assay was performed as described in [69]. The protocol was optimized based on previous work [70]. Internal and external activities were expressed in units based on absorption at 540 nm (1 U = 1 μmol glucose released/min per OD unit).

Single-molecule FISH

Yeast cells expressing Sec63-GFP were grown to mid-log phase and shifted to low glucose–containing medium [0.1% glucose] for 1.5 h to induce SUC2 expression. Cells were fixed in the same medium upon the addition of paraformaldehyde (4% final concentration) and incubated at room temperature (RT) for 45min with rotation. Cells were gently washed three times with ice cold Buffer B (0.1M potassium phosphate buffer, pH 7.5 containing 1.2M sorbitol), after which cells were spheroplasted in 1ml of freshly prepared spheroplast buffer [Buffer B supplemented with 20mM ribonucleoside vanadyl complexes (Sigma-Aldrich, St. Louis, MO), 20mM β-mercaptoethanol, and lyticase (Sigma-Aldrich, St. Louis, MO) (25 U per O.D.₆₀₀ unit of cells)] for 10min at 30°C. The spheroplasts were centrifuged for 5min at 1300 × g at 4°C and washed twice in ice cold Buffer B. Spheroplasts were then resuspended in Buffer B and approximately 2.5 O.D.₆₀₀ units of cells were placed on poly-L-lysine coated coverslips in 12-well plates and incubated on ice for 30 min. Cells were carefully washed once with Buffer B, then incubated with 70% ethanol for several hours to overnight at -20°C. Afterwards, cells were washed once with SSCx2 (0.3M sodium chloride, 30mM sodium citrate), followed by incubating with Wash buffer (SSCx2 with 10% formamide), for 15 min at room temperature (RT; ~23°C). Next, 45μl of hybridization buffer (SSCx2, 10% dextran sulfate, 10% formamide, 2mM ribonucleoside vanadyl complexes, 1mg/ml E. coli tRNA, and 0.2mg/ml BSA) containing 250nM of the TAMRA-labeled Stellaris probe mix for native SUC2 or HSP150 (Biosearch Technologies, Novato, CA)) was placed on parafilm in a hybridization chamber. Coverslips with the immobilized cells were placed face down on top of the hybridization buffer and were incubated overnight at 37°C in the dark. After probe hybridization, cells were incubated twice in Wash buffer for 15min at 37°C. Cells were then washed once with SSCx2 containing 0.1% Triton X-100 and incubated with SSCx2 supplemented with 0.5μg/ml DAPI for 1min at RT and finally washed with SSCx2 for 5min at RT. Cells were mounted with Prolong Glass (Thermo Scientific) mounting media on clean microscope slides. Samples were imaged using a Zeiss AxioObserver Z1 DuoLink dual camera imaging system equipped with an Illuminator HXP 120 V light source, PlanApo 100× 1.4 NA oil immersion objective, and Hamamatsu Flash 4 sCMOS cameras. Incremental (0.2μm) z-stack images were taken using a motorized XYZ scanning stage 130x100 PIEZO and ZEN2 software at 0.0645μm/pixel. Images were processed by deconvolution. At least 50 cells each showing both cER and nER labeling, as well as mRNA spots, were scored for ER localization per cell type [e.g. native, (+)SECReTE, and (-)SECReTE] examined. Scoring of mRNA granules (spots) was performed using the FISH-quant program [71] (https://bitbucket.org/muellerflorian/fish_quant) to analyze deconvolved images of single cells. mRNA co-localization to cER or nER was scored manually and was defined by overlap between the deconvolved signals.

Subcellular fractionation

Yeast cells were grown to mid-log phase (O.D.₆₀₀ = 0.6–0.8) and shifted to low glucose–containing medium (0.1% glucose) for 1.5hrs to induce SUC2 expression. Cultures (400ml) were centrifuged for 3min at 500 × g, resuspended in a buffer solution containing 50mM Tris (pH 7.6), 150mM NaCl, 200U of RNasin Ribonuclease Inhibitor (Promega, Madison, WI), Complete Protease Inhibitor Cocktail (Roche Diagnostics, Basel, Switzerland), and cycloheximide 100μg/ml, and disrupted using glass beads and vortexing for 10min at 4°C. Crude lysates were centrifuged for 10min at 1000 × g to remove cell debris, and then 1ml of each lysate was subjected to ultracentrifugation for 1 h at 48,000 × g. The resulting pellet was then resuspended in 500μl of buffer containing 50mM Tris (pH 7.6), 150mM NaCl, 80U/ml RNasin Ribonuclease Inhibitor, Complete Protease Inhibitor Cocktail, and 100μg/ml cycloheximide. Next total RNA from both the membrane fraction (resuspended pellets) and the cytosolic fraction (supernatants) was isolated using the MasterPure Yeast RNA Purification Kit (including DNase I treatment) according to the manufacturer's recommendations.

Computational analyses

Assignment of SECReTE and permutation analysis

To calculate the SECReTE count for each gene, the length of an uninterrupted run of NNY triplet repeats was counted separately for each of the three different codon positions (i.e. YNN, NYN, NNY, where N is any nucleotide and Y is a pyrimidine) for every gene. A SECReTE motif defined by the number of NNY triplet repeats where the threshold is ≥10 (see Results). Note that a gene may have more than one SECReTE motif along its length, each potentially in a different frame (relative to the coding sequence). The SECReTE count is defined as the number of SECReTE motifs present in a transcript.

For permutation analysis, the codon order of each gene sequence was randomly shuffled 1,000 times to evaluate the statistical significance of the SECReTE count in random sequences with unchanged codon usage. A Z score was calculated for each motif according to the formula: Z = (Observed–mean)/STD. Observed is the length of the motif in the real gene sequence. Mean is the average SECReTE length for all shuffled sequences of the gene. STD is the standard deviation of the SECReTE length from all shuffled sequences of the gene.

Calculation of RRY motif scores

In order to differentiate between SECReTE motifs that follow an NNY pattern and those that are simply pyrimidine-rich, an RRY [R (purine); G or A)] score was calculated for each defined motif. Since poly-Y conforms to NNY, RRY was used instead. The score is calculated for every motif by representing each nucleotide of the motif that follows the RRY pattern as ‘1’ and each nucleotide that defies this pattern as ‘0’, and taking the average score over the entire length of the motif. The score is calculated for all three possible frames and the maximal score between them is taken for each motif. The UTR sequences for the yeast genome were obtained from: https://downloads.yeastgenome.org/sequence/S288C_reference/.

Identification of a motif in cell wall proteins

A motif search was performed by MEME suites [73] (http://meme-suite.org/tools/meme) to identify RNA motifs in genes encoding cell wall proteins in yeast.