Sensitivity of RNAi-deficient mutants to ribosomal inhibitors
Utilizing the disruption mutant strains corresponding to the five genes encoding the members in the RNAi machinery (17), we extended the search for phenotypic consequence from the disruption. We found that, compared to the wild type, RNAi mutants were hypersensitive to inhibitory chemicals of eukaryotic ribosomes, e.g., hygromycin B, G418 and anisomycin, suggesting impairments in the ribosomes (Fig. 1A-C). The mutant strains exhibited significantly delayed growth on the plates supplemented with the drugs. This result indicates that there were defects in the ribosome function resulted from the loss of RNAi machinery. Considering the action of RNAi machinery, we turned to examine the rRNA level to find out whether sirRNAs were produced in the yeast.
Abundant sRNAs generated in C. neoformans and a novel class of rRNA-derived sRNAs
Using PAGE, we observed that JEC21 actually produced a large amount of 20-25-nt sRNAs (WT, Fig. 1D). This unique capacity of sRNAs production is rarely seen, even in some other RNAi-proficient fungi (Fig. S1B). Relying on the RNAi-deficient mutants of JEC21, we clearly witnessed that the biogenesis of 20-25-nt sRNAs in the yeast was dependent on the function of RNAi machinery (Fig. 1D). Disruption of the components of the pathways led to a significant drop of sRNAs. Thus they are siRNAs in general. The largest decrease in the amount of siRNAs was seen in three mutants, rdp1Δ, ago1Δ (CNJ00490) and dcr1Δ. Thus, Rdp1, Ago1 and Dcr1 possibly played a major part in the generation of endogenous siRNAs in JEC21. This pathway is divergent from the conventional transgene-evoked RNAi pathway in JEC21 i.e. Dcr2, Rdp1 and Ago1 play a major role (14), which is also the SIS pathway in var. grubii strain H99 (27). Notably, a trace, yet discernable amount of siRNAs was left in the double mutant, dcr1△dcr2△, suggesting unidentified pathways, e.g. the RNase IIIs apart from the Dicers, such as Rnt1, were capable to generate sRNAs as well (4,21).
To identify sirRNAs from the entire population of siRNAs, we conducted HT-seq followed by bioinformatics covering all 18-30-nt sRNAs with emphasis on sirRNAs in that range. A chart on relative percentages of siRNAs sorted by origination sources was depicted in Figure 2A, which displayed a diversity of siRNA originations in C. neoformans. A group of siRNAs from transposable elements took the largest percentage, up to 72.59%, of all groups (Left panel), similar to a previous observation in var. grubii H99 (12,14,27). Surprisingly, a group of putative sirRNAs delimited by structural characteristics took the second position, up to 12.88%, of the entire sRNA population. In the control rdp1Δ (Right panel), the percentage of all the non-sirRNAs dropped sharply to only 13.08%, whereas sirRNAs climbed to 86.92%. The shifting of percentages attests in consistence with the PAGE data that RNAi machinery plays a critical role in the biogenesis of endogenous siRNAs.
A routine analysis for the regular siRNAs (non-sirRNAs hereafter), in which sirRNAs were excluded, was separately conducted. The size allocation of them exhibited a range from 20 to 24 nt with a reads peak at 22 nt (~37%) (Left panel). A preference of nucleotide U at 5’ end was obvious (Fig. 2C). These archetypal structural features demonstrate they are siRNAs of C. neoformans. The features were lost in rdp1Δ, suggesting they were largely the products of Rdp-dependent RNAi. In parallel, we conducted bioinformatics analysis for sirRNAs (Fig. 2B&C, Right panels). They were distinct from siRNAs in structure. The size allocation peaked at 19 and 20 nt, shorter than the siRNAs. They had a nucleotide preference of C at the 5’ end (more than 95%), suggesting sirRNAs are a novel class of siRNAs. Such characteristic features disappeared in rdp1Δ, and again manifesting a role of Rdp1 in sirRNA biogenesis (Fig. 2, Fig.S2 and S3). Depiction on reads of sirRNAs verse the location on rRNA sequences revealed a pattern of clustered distribution, ascertained they were generated by RNAi machinery but not random degradation (Fig. 2D). We also found a large number of classic siRNAs corresponding to the antisense sequence of rDNA, suggesting they were derived from the natural antisense transcripts of rDNAs (NAT-rRNAs for short), thus, designated as a-sirRNAs for antisense sirRNAs (Fig. 2E). The presence of a-sirRNAs suggested that RNAi occurred on NAT-rRNAs. Taken together, we demonstrated the existence of RNAi-generated sirRNAs in C. neoformans, which are new in C. neoformans. And a-sirRNAs from NAT-rRNAs were also found. The co-existence of two classes of siRNAs derived from the complementary transcripts of the same genetic loci manifests complexity of RNAi in this yeast.
Characterization of sirRNAs 001/006 and their biogenesis by RNAi machinery
By HT-seq, we discovered two sirRNAs, 001 and 006, had the largest number of reads. They each had three isoforms sharing a similar sequence (Table 1). Alignment confirmed they were originated from 25S rRNA and largely Rdp-dependent (Fig. 2D, S4A). Interestingly, they were located in juxtaposition with a spacer of 4 nucleotides (Fig. 3A, S4A), and likely formed intracellular duplex to make substrate of the dicers (Fig. 3A). Computation predicted a possible stem-loop structure within a sequence of approximately 108-nt flanking the sirRNAs, 001/006, with an initial ΔG = -58.00 Kcal/mol (a minimal value of ΔG is 25 Kcal/mol for stem-loop formation). In the prediction, sequences of 001 and 006 could form perfect duplex to each other in inverted orientation, leaving a loop of the 4-nucleotide spacer. Non-Watson-Crick base pairing was employed in the stem, e.g. eight G-U and two A-A pairings in the twenty base-paired 001-006 duplex. G-U and A-A pairings are common in RNA molecules (13). However, whether this duplex could be a suitable substrate for dicing by RNase IIIs to generate sirRNAs 001/006 needs proof.
To demonstrate the existence of sirRNAs, we believed the HT-seq and bioinformatics data should further be confirmed by molecular approaches. We reasoned that bona fide sirRNAs should meet two criteria: 1) They should be at a stable concentration sufficient to promote RNAi reaction. 2) They should be processed into an approximately uniform length, i.e, 20-25 nt, by RNAi, which made them distinguishable from degradation products. Therefore they could form clear bands that were visible in Northern blotting. To verify the results of HT-seq, we performed Northern blotting for sirRNAs 001/006. As expected, we saw strong signals of 001/006 in the wild type, especially in an extremely high and stable level (WT lane, Fig. 3B). This result clearly verified the presence of 001/006. In RNAi mutants, signals of 001/006 decreased greatly, confirming that the majority of them were produced by RNAi machinery (Fig. 3B). In the control blots for two rsRNAs 033 and 065 (locus shown in Fig.S4A), two randomly chosen small rRNA fragments (20-25 nt) also obtained by HT-seq, only smeared signals were detected, suggesting they were degraded rRNA fragments (Fig. S4C). Thus, rRNA-degraded sRNAs were indeed present and probably in a large number in the sequenced sRNAs. And they were difficult to be distinguished from the RNAi-dependent sRNAs only by HT-seq. Even more, a clear remainder band of 001 or 006 in rdpΔ (Fig. 3B), suggesting that a fair portion of sirRNAs 001/006 were made through an Rdp-independent pathway, likely by Dcr1/2 or other pathways. This result was consistent with the HT-seq data which showed that in rdpΔ, the reads of sirRNAs 001 dropped by 70.57%, while sirRNA-006 by 84.02%, verse the reads in the wild type JEC21 (Table 1). Blotting result still suggested that each individual member in the pathways, Dcr1, Dcr2, Ago1 or Ago2, was critical for the biogenesis of 001/006, as well as a precursor (approximately 33 nt for 001 and ~35 nt for 006, arrowed, Fig. 3B, 3C). Disruption of any one led to a significant decrease of 001/006 and the precursor, suggesting that the components of the machinery act in a manner of cooperation in the biogenesis process of sirRNAs. In sum, HT-seq and Northern blots clearly demonstrate the presence of a novel class of rRNA-derived sRNAs, sirRNAs in C. neoformans, and they are generated by RNAi machinery.
Unique mechanism of the biogenesis of sirRNAs
The above Northern blots suggested that the two Dcrs and the two Agos, plus Rdp could form a long pathway in the biogenesis of the majority of sirRNAs 001/006 (Fig. 3B). Reads data confirmed that a fair amount of sirRNA 001 (29.43%) and 006 (15.98%) remained in Δrdp1 (Table 1). In other words, this portion of sirRNAs 001/006 could be produced by a short pathway without Rdp, e.g.. probably by the Dcrs with Agos (Lane rdp1Δ, Fig. 3B). Two other blots with double mutant strains, dcr1dcr2 and ago1ago2, confirmed both Dcrs and Agos are involved in the short pathway (Fig.3C). Even more, when we disrupted two copies of Dcrs or Agos, we could still see a trace amount of sirRNAs 001 and 006 in the blots, suggesting a third pathway was acting (Fig.3C), which might contain other RNase IIIs in addition to the known Dcrs. In short, our data suggests at least three pathways could independently generate sirRNAs in the yeast.
Unlike mammalian cells, fungi have no counterparts of the microprocessor which consists of different set of proteins for the generation of miRNA precursors (1,5). To elicit the early steps leading to the production of sirRNAs, we explored the initial rRNA substrates and intermediate products. We constructed double mutants for the genes of the Dcrs and Agos, designated as dcr1dcr2Δ and ago1ago2Δ, respectively. Northern blots to detect the intermediate products in these strains were performed (Fig. 3C). In Fig. 3C (Lane WT, right panels), a band near 120 nt (by calculating the mobility with the markers) below the 25S rRNA band was respectively detected by the probes for sirRNAs 001 and 006, suggesting this band was a shared precursor by sirRNAs 001 and 006, thus, designated as pro-sirRNAs. Below this band, processing of the precursor diverges into two distinct pathways toward the final sirRNAs. Generation of this 120-nt precursor was apparently independent of RNAi machinery, i.e. it was not a process of RNAi. As a matter of fact, disruption of Dcrs or Agos actually led a discernibly increase of the 120-nt precursor, rather than a decrease (indicated by arrow, Lanes dcr1dcr2 and ago1ago2, Fig. 3C). Nonetheless, the RNAi machinery obviously participated in the following steps. This was evident by band pattern of intermediate products in-between of the 120-nt precursor and the sirRNAs in the mutants (Fig. 3C, Right panels). Along the route to sirRNAs 001, there were four bands, approximately, 80 nt, 70 nt, 55 nt and 33 nt (Lane WT), indicating four corresponding steps of processing. Whereas there were only two intermediates, 60 nt and 35 nt, leading to the generation of sirRNAs 006 (Lane WT, Fig. 3C). In RNAi mutants, these intermediate products significantly decreased or disappeared, demonstrating that Dcrs (1/2) and/or Agos (1/2) participated in generating these intermediates (Fig. 3B, C). Notably, there were still trackable amount of the intermediate products in the double mutants (Fig. 3C, Lanes dcr1dcr2Δ and ago1ago2Δ), suggesting that other pathways, e.g., RNase IIIs, participate in the formation of the intermediate products. As the 55-nt and 60-nt bands for 001 and 006, respectively, exhibited as the major ones during this stage, thus designated them as pri-sirRNAs. Subsequently, we could see that Dcrs and Agos processed the pri-sirRNAs into two products, the 33-nt one for 001, and the 35-nt for 006 (Fig. 3B, C, Right two panels). We hereby designate these two small rRNA fragments as pre-sirRNAs. Last, the RNAi machinery was definitely required for the formation of the mature sirRNAs 001 and 006. In sum, our data suggests that the biogenesis of sirRNAs 001/006 could be roughly divided into three stages (Fig. 6C). The primary stage gives rise to a 120-nt rRNA precursor, the pro-sirRNAs, from rRNAs with an unknown mechanism. The second stage is the generation of pri-sirRNAs, the 55-nt product for 001 and the 60-nt for 006, by split of the 120-nt precursor. In this stage, RNAi machinery and a certain unidentified RNase IIIs are involved. The last stage is responsible for the formation of pre-sirRNAs and the final products sirRNAs 001 and 006. And this stage is accomplished solely by the RNAi machinery. Disruption of the two Dcrs or Agos led to a nearly complete loss of pre-sirRNAs and most sirRNAs (Fig. 3C). Still, we can see that Rdp played a critical role in the third stage. In rpdΔ, pre-sirRNAs were hardly detectable as in the double mutants (Fig. 3B), suggesting that sirRNAs were mainly generated from the products of Rdp amplification that were subject to processing by the Dcrs and Agos, rather than from a stem-loop precursor, or the double-stranded RNA molecules formed by rRNA and NAT-rRNAs (Fig. 6C).
In short, our experiments delimited that three potential pathways could contribute to the biogenesis of sirRNAs 001/006, the long one including Rdp, Dcrs and Agos, a short one containing merely Dcrs and Agos, and a third unidentified pathway. Rdp vastly boosts the yield of sirRNAs 001/006. This biogenesis mechanism of sirRNAs 001/006 is distinct from the known strategies of the biogenesis of siRNA or miRNA (1,5,7).
Silencing activity of sirRNAs 001/006
We reasoned whether an endogenous sirRNA had the capacity to effectively trigger RNAi reaction on targets might rely on two conditions, the first was the intracellular concentration of the sirRNA and secondly their accessibility to RNAi machinery. We then tested whether sirRNAs 001/006 could guide RNAi machinery to knock down the expression of reporter genes. To this end, we employed a reporter cassette that was previously designed for silencing assay in C. neoformans JEC21 (28). The yeast genes URA5 and CLC1 were inserted with a complementary sequence of 001 or 006 to form the targets (Fig.4A). As expected, the transformants carrying the cassette of URA5-sirRNAs displayed retarded growth on minimal media YNB (Fig. 4B). Meanwhile they showed an apparent tolerance to the toxin 5-FOA (the right panel), demonstrating that expression of URA5 was suppressed. The control strain which contained only the plasmid (URA5) grew properly on YNB, but failed to grow on FOA-containing plate (right panel). A following qPCR measurement confirmed a deep fall of URA5 mRNA in the transformants (Fig.4C). Similarly, in the test for the cassette of CLC1-sirRNAs (Fig. 4D, Bottom panels), transformants of silencing cassette produced less pigments on nor-epinephrine (NE)-containing plates, which was similar to a clc-disrupted mutant TX1 (clc1Δ) (23), verifying a silencing effect on CLC1. A qPCR confirmed a significant decrease of CLC1 mRNA in the transformants (Fig. 4E). Thus, sirRNAs, 001 and 006, could trigger RNAi reaction on the complementary targets and caused silencing effect. Also, this assay implies that sirRNAs could initiate RNAi against invasive nucleic acids.
RNAi machinery maintains 25S rRNA level by suppressing NAT-rRNAs
Considering the fact that a large sum of sirRNAs 001/006 were synthesized from 25S rRNAs, we reasoned that the level of 25S rRNA might rise when RNAi pathways were disrupted. To test this hypothesis, we applied qPCR to measure the level of 25S rRNA in RNAi mutants. Contrary to our expectation, the level of 25S rRNA decreased apparently in all mutants (Fig. 5A). For instance, the amount of 25S rRNA in ago1Δ was only ~56% of that in the wild type. Thus, the machinery rather played a positive role in maintaining the level of 25S rRNA. To explain this paradoxical result, we conceived that natural antisense transcripts (NAT-rRNAs) might be expressed from the repetitive rDNA loci that could form RNA duplex with rRNAs to cause rRNAs to be degraded by RNase IIIs or/and exosomes.
We searched for NAT-rRNAs arisen around the locus of sirRNAs 001/006 and detected them by reverse transcription PCR (Fig. S4D). We did Northern blots and qPCR to track the variation of the Nat-rRNA fragments in RNAi mutants. We knew that it was risky to expect a concentrated band of NAT-rRNAs in the blots, as NAT-rRNAs might be transcribed into fragments of various length. However, it is proven that RNAi in this yeast generates a 70-nt precursor of the targets which may accumulate in RNAi mutants (18). Hence, we probed instead the 70-nt precursor of NAT-rRNAs in the mutants. We conducted two blots in parallel with two probes, one located at the 001/006 locus (Table S2, Fig. S4E). The blots were shown in Fig. 5B. In the wild type, the NAT-rRNA band was barely seen, suggesting it was in a state of suppression by RNAi, if taking into account of a-sirRNAs in the wild type (Fig. 2E). In contrast, NAT-rRNAs remarkably mounted in all the mutants in dual blots. This result clearly verified that RNAi mediates the suppression of NAT-rRNAs in the yeast. Further, de-suppression of NAT-rRNA in RNAi mutants was validated by qPCR (Fig.5C). For instance of ago2Δ, NAT-rRNAs increased by nearly 7-fold in qPCR which was consistently evident in Northern blots. Similar increase was seen for dcr2Δ. This Dcr2-Ago2 pathway deviated from the main pathway taken by regular siRNAs biogenesis and transgene-evoked RNAi, in which Dcr1/Ago1 was the major player (Fig.1D). Taken together, we demonstrated a seesawing relationship between 25S rRNA and its complementary NAT-rRNAs in RNAi mutants, which supports the view that rRNA duplexes might form between rRNAs and the NAT-rRNAs to cause rRNA degradation in RNAi mutants via likely RNase IIIs or exosomes. Therefore, RNAi machinery forms a defense system for rRNAs from the formation of RNA duplex by knocking down the NAT-rRNAs.
Functional defects in ribosomes in the RNAi mutants
Disruption of 25S rRNA level in RNAi-deficient mutants may cause functional impairments in ribosomes. This may explain the result shown in Fig. 1A-C, i.e. RNAi mutants exhibited hypersensitivity to ribosome inhibitors. We found more defective phenotype of the mutants related to ribosome function, e.g. the synthesis of proteins remarkably dropped in RNAi mutants (Fig.6A, S5A). Further, in E. coli, S. cerevisiae and A. thaliana, ribosomal damages result in sensitivity to cold temperature (2,15,32). We then test whether our RNAi mutant yeasts had a similar phenotype. We found that RNAi mutants grew severely slower at low temperature (4℃) than the wild type, though they had a similar growth at 30℃ (Fig. 6B, S1A). When inhibitors and low temperature were jointly applied, the growth was substantially worsened (Fig. S5B). These results confirm again that loss of RNAi machinery causes impairments in ribosome functions.