Pivotal Roles for Ribonucleases in Streptococcus pneumoniae Pathogenesis

ABSTRACT RNases perform indispensable functions in regulating gene expression in many bacterial pathogens by processing and/or degrading RNAs. Despite the pivotal role of RNases in regulating bacterial virulence factors, the functions of RNases have not yet been studied in the major human respiratory pathogen Streptococcus pneumoniae (pneumococcus). Here, we sought to determine the impact of two conserved RNases, the endoribonuclease RNase Y and exoribonuclease polynucleotide phosphorylase (PNPase), on the physiology and virulence of S. pneumoniae serotype 2 strain D39. We report that RNase Y and PNPase are essential for pneumococcal pathogenesis, as both deletion mutants showed strong attenuation of virulence in murine models of invasive pneumonia. Genome-wide transcriptomic analysis revealed that the abundances of nearly 200 mRNA transcripts were significantly increased, whereas those of several pneumococcal small regulatory RNAs (sRNAs), including the Ccn (CiaR-controlled noncoding RNA) sRNAs, were altered in the Δrny mutant relative to the wild-type strain. Additionally, lack of RNase Y resulted in pleiotropic phenotypes that included defects in pneumococcal cell morphology and growth in vitro. In contrast, Δpnp mutants showed no growth defect in vitro but differentially expressed a total of 40 transcripts, including the tryptophan biosynthesis operon genes and numerous 5′ cis-acting regulatory RNAs, a majority of which were previously shown to impact pneumococcal disease progression in mice using the serotype 4 strain TIGR4. Together, our data suggest that RNase Y exerts a global impact on pneumococcal physiology, while PNPase mediates virulence phenotypes, likely through sRNA regulation.

RNase Y functions as a broadly pleiotropic regulator whose absence significantly impacts the pneumococcal mRNA transcriptome, growth, virulence, and stability and function of conserved pneumococcal Ccn (CiaR-controlled noncoding RNA) sRNAs. In contrast, PNPase impacts the abundance of several important transcripts, including riboswitches that were previously implicated in pneumococcal virulence control. The absence of PNPase consistently resulted in a strong virulence defect in vivo while displaying no obvious phenotypes in vitro. Together, our work has uncovered for the first time the crucial roles of two wellconserved RNases in regulating pneumococcal physiology and virulence.

RESULTS
RNase Y is required for normal pneumococcal growth and cell morphology. Prior studies showed that deletion of rny, the gene encoding RNase Y, from B. subtilis and C. perfringens caused a drastic reduction in growth, but the effect of removal of this gene on S. pyogenes and S. aureus growth was modest (25,(27)(28)(29). However, deletion of pnp led to a cold-sensitive phenotype in B. subtilis, similar to what was observed for Escherichia coli (30,31). Therefore, we assessed the effects of clean deletion in rny or pnp (Table S1) on pneumococcal growth at both optimal (37°C) and lower (32°C) temperatures. We found that at 37°C in brain heart infusion (BHI) broth, the Drny mutant exhibited a significant reduction in growth rate and yield compared to the wildtype (WT) strain ( Fig. 1A; Table S2). The average doubling time and growth yield for the Drny mutant were ;69 min and 0.37, compared with ;39 min and 0.96 for the WT strain. The observed growth defect of the Drny mutant was restored by expressing rny from a constitutive P mal(c) promoter at the ectopic CEP (chromosomal expression platform) locus or by repairing the mutation to the WT allele at the native locus ( Fig. 1A; Fig. S1A to D; Table S2). We also observed that the growth deficiency of the Drny mutant became more pronounced in 15-and 25-day-old BHI compared to freshly prepared (#5-day-old) BHI, whereas the isogenic parental strain grew similarly in fresh and aged BHI (Fig. S1 to D; Table S2). In contrast, the Dpnp mutant grew like the WT strain in BHI broth at 37°C ( Fig. 1A; Fig. S1A to D). In contrast to E. coli and B. subtilis (30,31), the pneumococcal Dpnp mutant did not show a cold-sensitive (CS) phenotype ( Fig. 1B) at 32°C, the lowest temperature at which S. pneumoniae D39 grows well. Finally, the Drny mutant was not cold sensitive, as the growth rate differences between the WT and Drny mutant were not significantly different at 32°C compared to 37°C ( Fig. 1B; Fig. S1E to G; Table S2).
To gain insight into the growth impairment of Drny mutants, we examined cells from early-exponential-phase cultures (optical density at 620 nm [OD 620 ] % 0.1 to 0.15) of the WT and Drny mutant by phase-contrast microscopy. We found that the Drny mutant exhibits significant morphological defects. Occasionally, Drny mutants formed minicells at the ends or in the middle of a chain, indicating a possible cell division defect (Fig. 1C). The abnormalities in cell morphology that we observed in the encapsulated Drny mutant were even more pronounced in a Dcps mutant lacking capsule (Fig. 1C). This observation is consistent with previous findings that capsule tends to dampen pneumococcal cell shape and division phenotypes (32). In addition, the Drny Dcps mutant formed longer chains comprising 4 to 12 cells/chain, in contrast to the WT parent, which occurred mainly as diplococci (Fig. 1C). The observed chaining effect of the Drny mutant was reversed by expressing rny from a constitutive P mal(c) promoter at the ectopic CEP locus (Fig. 1C). Finally, we did not observe any morphological differences between the Dpnp mutant and the WT parent in either the cps 1 or the Dcps background (Fig. 1C). We conclude that RNase Y, but not PNPase, is required for S. pneumoniae D39 normal growth and cell morphology.
Lack of RNase Y or PNPase attenuates S. pneumoniae D39 pathogenesis. Lack of RNase Y in S. pyogenes and S. aureus resulted in virulence attenuation (28,33,34). Therefore, we determined the consequences of the rny and pnp deletions on S. pneumoniae D39 pathogenesis using a murine invasive pneumonia model (see Materials and Methods). Both the Drny and Dpnp mutants were substantially attenuated for virulence compared to the WT parent ( Fig. 1D and E). Of the mice inoculated with the Dpnp or Drny and Dpnp (IU4883) strains, and complemented Drny//rny 1 (IU4834) and Dpnp//pnp 1 (IU5510) strains, grown statically at 37°C and at 32°C in 5-day-old BHI broth in an atmosphere of 5% CO 2 . Growth curves represent data from three independent replicates for each strain at 37°C or 32°C. Average growth rates and growth yields are listed in Table S2. (C) Representative phase-contrast images of the D39 wild-type strain (WT; IU1781), its derived Dcps (IU1824), Drny (NRD10092), Dpnp (IU4883), Dcps Drny (NRD10109), and Dcps Dpnp (NRD10108) mutants, and Drny//rny1 (NRD10388) and Dcps Drny// rny1 (NRD10389) complemented strains in early exponential growth phase. Distributions of chain lengths were based on 100 to 200 chains from at least two independent cultures of each strain. Bars, 2 mm. (D and E) Survival curve analysis showing disease progression in an invasive murine model of pneumonia. ICR male mice were inoculated intranasally with ;10 7 CFU in 50-ml inocula of the D39 parent expressing a lux luminescence cassette (D39 (Continued on next page) Sinha et al. mutant, 75% and 87%, respectively, survived the course of the experiment (;170 h), whereas the median survival time for the WT parent strain ranged from 42 h to 64 h ( Fig. 1D and E; Fig. S2A). To determine if the attenuated virulence observed in each case was correlated with loss of rny or pnp function, we repaired Drny or Dpnp back to the rny 1 or pnp 1 allele, respectively, by allelic exchange (see Materials and Methods). The rny 1 and pnp 1 repaired strains displayed median survival times of 60 h and 67 h, respectively, indicative of full virulence. Taking these results together, we conclude that both RNase Y and PNPase contribute to pneumococcal pathogenesis.
RNase Y and PNPase impact the pneumococcal mRNA transcriptome differently. To identify target transcripts of RNase Y or PNPase that influence pneumococcal physiology, next, we compared the genome-wide transcriptome profiles of Drny or Dpnp mutant relative to the WT parent grown in matched batches of BHI broth at 37°C in an atmosphere of 5% CO 2 using mRNA-sequencing (mRNA-seq) analysis (see Materials and Methods). mRNA-seq of the Drny mutant revealed that 185 transcripts were significantly upregulated compared to the WT parent strain, using a cutoff of a .1.8-fold change and a P value adjusted for multiple testing (P adj ) of ,0.05. In contrast, only 28 genes were significantly downregulated in the Drny mutant compared to the WT strain (Table 1; Fig. 2A). The upregulated transcripts encode proteins that are involved in diverse cellular functions, including translation; transcription; transport and metabolism of carbohydrates, amino acids, nucleotides, coenzymes, and inorganic ions; cell wall and envelope biogenesis; and stress response (Table 1). In particular, several transcripts that were upregulated in the Drny mutant are under the regulatory control of the WalRK, LiaFSR, PnpRS, or CiaRH two-component system (TCS) ( Table 1). Notably, relative transcript abundance for genes encoding important cell division and cell wall proteins, including mapZ, cozE, gpsB, lytB, licB, licC, licA, tarI, tarJ, spd_0703, and spd_0104, were increased by ;2-5-fold in the Drny strain. Lack of RNase Y also increased the relative expression of genes involved in stress response (clpL, 9-fold; dnaK, 6-fold; dnaJ, 3-fold; hptX, 2-fold) and pavB (;2-fold), which encodes a fibronectin-binding protein involved in pneumococcal virulence.
Deletion of pnp had substantially less impact on relative mRNA transcript amounts, with significant changes in abundance of only 20 transcripts (Table 2; Fig. 2C). Interestingly, a majority of mRNA transcripts that were differentially regulated in the Dpnp mutant were shown by a previous transposon insertion sequencing (Tn-seq) screen of serotype 4 strain TIGR4 to be important for colonization of the nasopharynx and/or infection of the lungs in murine infection models (35) ( Table 2). In particular, the relative abundance of transcripts corresponding to the tryptophan biosynthesis operon (trpABCDEFG), including the upstream gene spd_1604, were maximally downregulated by ;3to 4-fold. In addition, the relative transcript amount of alaS, which encodes alanyl-tRNA-synthetase, was downregulated by ;2-fold in the Dpnp mutant compared to the WT strain (Table 2; Fig. 2C).
Results from mRNA-seq analyses were confirmed by reverse transcriptase droplet digital PCR (RT-ddPCR) as described in Materials and Methods. Consistent with the RNA-seq results, the relative transcript amounts of mapZ (;2-fold), spd_0703 (;3-fold), clpL (;4-fold), and dnaK (;2-fold) increased in the Drny mutant compared to WT strain (Fig. 2B). In the Dpnp mutant, RT-ddPCR showed that the relative transcript amounts of spd_1604-trpD-trpA-trpE and alaS decreased by ;4-fold and 2.4-fold, respectively, whereas the relative amount of spd_0437 (ribU) increased by ;6-fold (Fig. 2D), again consistent with the RNA-seq results. Together, these data confirm the relative changes in steady-state mRNA transcript amounts caused by lack of RNase Y or PNPase in S. pneumoniae.   RNase Y and PNPase mediate the sRNA transcriptome of S. pneumoniae. Previous studies demonstrate that RNase Y directly and indirectly impacts sRNA levels in several important Gram-positive pathogens (25,36,37), whereas PNPase promotes the stability of some sRNAs and degrade others in E. coli (38,39). A recent Grad-seq study indicates that S. pneumoniae PNPase binds to several sRNAs, including CcnA, CcnB, CcnC, CcnD, and Spd_sr34 (23). To further understand how RNases modulate the stability and function of sRNAs expressed by S. pneumoniae D39, we sought to identify the sRNAs targeted by RNase Y and PNPase using sRNA sequencing (sRNA-seq) (see Materials and Methods). At least 112 distinct sRNAs have been identified in S. pneumoniae D39 (40)(41)(42)(43)(44)(45).
sRNA-seq analysis revealed that 11 sRNAs (;10% of total sRNAs) showed a .1.8fold change in relative amount between the Drny mutant and WT strain (Table 3; Fig. 3A). Seven sRNAs were upregulated in the Drny mutant compared to the WT strain, whereas only 4 were downregulated. The putative regulatory RNAs impacted by Drny fall into all five categories of sRNAs based on their location relative to previously annotated genes in D39 (Fig. 3B). Three of the sRNAs differentially expressed in the Drny mutant contain regulatory elements; Spd-sr12 and Spd-sr32 contain T-box riboswitches, and Spd-sr48 has an L20 leader sequence that regulates the expression of  (Table S1). Fold changes (1.8-fold cutoff) and adjusted P values (P , 0.05) are based on three independent biological replicates. Boldface indicates genes mentioned in the text. b Member of the CbpRS two-component system regulon (82). c Member of the SaeRS two-component system regulon (83). d Member of the WalRK two-component system regulon (55). e Member of the TCS07/YesMN two-component system regulon (84). f Member of the LiaFSR two-component system regulon (58). g Member of the CiaRH two-component system regulon (85). h Member of the PnpRS two-component system regulon (67).
downstream ribosomal genes. Interestingly, among the significantly upregulated sRNAs in the Drny mutant are 2 Ccn sRNAs (CcnA and CcnE) ( Table 3; Fig. 3A), which are among the five homologous, highly conserved intergenic pneumococcal sRNAs under positive transcriptional control of the CiaR response regulator and function to inhibit competence development via base-pairing with the precursor of the competence stimulatory peptide-encoding mRNA comC (86, 87). Seven of 11 sRNAs that were differentially expressed in the Drny mutant relative to the WT strain were experimentally validated using Northern blotting. We found that four sRNAs (CcnE, CcnA, Spd-sr12, and Spd-sr32) are significantly upregulated, while for the sRNAs Spd-sr100 and Spd-sr116, the annotated full-length transcripts could not be detected in the Drny mutant ( Fig. 4; Fig. S3 and S5). However, we did observe that a higher-molecular-weight   (Table S1). Fold changes (1.8-fold cutoff) and P values (P adj , 0.05) are based on three independent biological replicates. sRNAs validated in this study are in bold ( Fig. 4; Fig. S3 and S5). b 59 regulatory element present. c Spd-sr108 levels were comparable between the wild type and a Drny mutant on Northern blots (Fig. S5).
band corresponding to Spd-sr116 was increased in abundance only in a Drny mutant (Fig. S3). Spd-sr108 was the only sRNA for which we observed a significant difference in abundance between the Drny mutant and WT strain by RNA-seq but not by Northern blotting analysis (Table 3; Fig. S3 and S5). In addition to these sRNAs, we probed for 12 additional sRNAs that were not significantly differentially expressed in the Drny mutant relative to the WT strain in the RNA-seq analysis. Northern blots revealed that eight of these sRNAs (Spd-sr43, Spd-sr44, Spd-sr73, Spd-sr74, Spd-sr80, Spd-sr83, Spd-sr88, and Spd-sr114) were upregulated in the Drny mutant relative to the wild-type strain, whereas 4 others (Spd-sr70, Spd-sr54, Spd-sr82, and Spd-sr96) were unaffected by Drny ( Fig. 4; Fig. S3 and S5). Together, these data confirm that the cellular amounts of a relatively small number of sRNAs are changed in the Drny mutant.
In contrast to the Drny mutant, 21% of the pneumococcal sRNA transcriptome was significantly altered in the Dpnp mutant. Twenty-three sRNAs exhibited .1.8-fold differences in relative expression in the Dpnp mutant (Table 4; Fig. 3C), where 17 and 6 sRNAs were significantly up-and downregulated, respectively. Notably, approximately half of the sRNAs upregulated in a Dpnp mutant relative to the WT strain are riboswitch RNAs. Spd-sr32, Spd-sr70, Spd-sr74, Spd-sr80, and Spd-sr88 are characterized by the   Fig. S4. Quantification of signal intensity for each full-length sRNA normalized to 5S rRNA amount is displayed in Fig. S5, and the probes used are listed in Table S3. presence of a T-box riboswitch, while Spd-sr43, Spd-sr44, and Spd-sr114 contain a thiamine pyrophosphate (TPP) riboswitch element (Table 4; Fig. 3C and D). The riboswitch RNAs Spd-sr44 and Spd-sr88 are particularly interesting, because Tn-seq screens with the serotype 4 strain (TIGR4) of S. pneumoniae indicated that the loss of spd-sr44 or spd-sr88 results in reduced pneumococcal fitness in murine blood and lung infection, respectively (46). 59-intergenic and 39-intergenic sRNAs are in the overrepresented category of sRNAs that showed differential regulation in Dpnp compared to the WT strain (Fig. 3D). Finally, we validated the expression of a total of 14 of 23 sRNAs that were significantly differentially expressed in the Dpnp mutant relative to the WT strain ( Fig. 4; Fig. S3 and S5). Taken together, these data suggest that both RNase Y and PNPase play important roles in regulating the relative amounts of different sets of pneumococcal regulatory RNAs.
PNPase and RNase Y play important roles in riboswitch RNA decay and processing. T-box-containing riboswitch RNAs that are upregulated in the Dpnp mutant include Spd-sr88 and Spd-sr70, which are located within the 59 untranslated regions (UTRs) of the trp operon (encoding enzymes involved in tryptophan biosynthesis) and alaS (encoding alanyl-tRNA synthetase) operon, respectively (Table 4; Fig. 4A). Northern blotting confirmed increases in spd-sr88 and spd-sr70 in the Dpnp mutant compared to the WT strain determined by RNA-seq analysis (Table 4) and showed accumulations of decay products of these sRNAs ( Fig. 4B and C). Concurrently, relative transcript amounts of both alaS and the entire trp operon, including the upstream gene spd_1604, are decreased by ;2-fold and ;2to 4-fold, respectively, in the Dpnp mutant in mRNA-seq analysis ( Table 2; Fig. 3C and D). Based on these observations, we a RNA extraction and sRNA-seq analyses were performed as described in Materials and Methods. RNA was prepared from cultures of the encapsulated parent strain IU1781 (wild-type parent; D39 rpsL1 pnp 1 ) and its derived mutant IU4883 (D39 rpsL1 Dpnp) (Table S1). Fold changes (1.8-fold cutoff) and P values (P adj , 0.05) are based on three independent biological replicates. sRNAs validated in this study are in bold ( Fig. 4; Fig. S3 and S5). b 59 regulatory element and T-box element present. c 59 regulatory element and TPP riboswitch element present. d CcnA sRNA levels were comparable between the wild type and the Dpnp mutant on Northern blots (Fig. S5).
further tested the expression profiles of six other TPP or T-box riboswitch RNAs that also showed increased relative expression in the Dpnp mutant (Table 4). We observed a similar pattern of accumulation of decay intermediates for Spd-sr32, Spd-sr43, Spd-sr44, Spd-sr74, Spd-sr80, and Spd-sr114 in the Dpnp mutant, but not in the WT or Drny strain ( Fig. 4D and E; Fig. S3). In addition, we noticed that the full-length transcripts for all six of these riboswitch RNAs were more abundant in a Drny mutant than that in a WT strain (Fig. 4B and D; Fig. S3 and S4). While Spd-sr32 was identified by sRNA-seq analysis to be significantly upregulated by ;4-fold in the Drny mutant compared to WT (Table 3), the increased relative transcript steady-state levels of the other riboswitch RNAs (Spd-sr43, Spd-Sr44, Spd-sr74, Spd-sr88, and Spd-sr114) in a Drny mutant were not detected by this transcriptomics-based approach but were detected independently by Northern blot analysis (Fig. S3). Taking these results together, we conclude that both RNase Y and PNPase jointly function in the processing and decay of riboswitch regulatory RNAs. RNase Y regulates Ccn sRNA stability and function. After validating by Northern blotting, the ;3-fold increases in relative steady-state levels of CcnA and CcnE in the Drny mutant ( Fig. 4G and H; Fig. 5A, B, D, and E; Fig. S6) in accordance with our sRNAseq data (Table 3), we sought to further define the mechanism by which RNase Y regulates of the abundance of the Ccn sRNAs in S. pneumoniae D39. Therefore, we measured the stability of CcnA and CcnE in exponentially growing cultures of a Drny mutant or a WT strain after blocking transcription initiation by adding rifampin (Fig. 5C and F).  (Table S4). These findings prompted us to test the impact of RNase Y on the stability and the corresponding steady-state levels of the remaining three Ccn sRNAs. The relative transcript levels for CcnB, CcnC, and CcnD, were similarly upregulated by ;2-fold in the Drny mutant compared to WT (Fig. 5G). Consistent with these increased amounts, CcnB and CcnC were significantly stabilized in the Drny mutant ( Fig. 5H and I; Table S4). We were unable to accurately determine the relative stability of CcnD, because it was extremely unstable in the WT strain following rifampin addition (data not shown).
Finally, we investigated the role of RNase Y in Ccn-mediated comC target regulation. To this end, we constructed a translational fusion in which the 59 untranslated region and the first 12 codons of comC are fused in-frame with the truncated E. coli b-galactosidase gene lacZ. The comC'-'lacZ translational fusion, driven from the constitutive vegetative promoter vegT (derived from the vegII promoter of Bacillus subtilis [Tables S1 and S2]), was integrated in the chromosomal bgaA locus in strain D39 (thereby knocking out pneumococcal b-galactosidase). Consistent with previous reports, deletion of all 5 Ccn sRNAs (DccnA-E) relieved ComC translational repression and led to increased relative expression of b-galactosidase specific activity (;3.5-fold) from comC'-'lacZ (Fig. 5J). Conversely, Drny led to decreased (;3.8-fold) relative b-galactosidase specific activity from the comC'-'lacZ fusion (Fig. 5J), consistent with increased stabilization of Ccn sRNAs (Fig. 4) and increased translational repression of ComC. To further test this idea, we attempted to measure comC'-'lacZ expression in a Drny DccnA-E mutant. Unexpectedly, the Drny DccnA-E mutant exhibited a synthetic phenotype with severely impaired growth and low growth yield compared to the WT (data not shown). In contrast, a Drny DccnACDE mutant did not exhibit a strong synthetic phenotype (data not shown). Relative expression of comC'-'lacZ is less elevated in the DccnACDE than the DccnA-E mutant and is reduced further in the Drny DccnACDE mutant to near the WT level (Fig. 5J), consistent with stabilization of remaining CcnB in the Drny background. However, the transformation frequency (TF) of a Drny mutant was comparable to that of a WT strain (TF Drny = 0.38% versus TF WT = 0.32%), using a spontaneous competence assay (Fig. S2B). Together, these results indicate that RNase Y-mediated regulation of Ccn sRNA stability has a consequential impact on Ccn-mediated target regulation in S. pneumoniae D39.

DISCUSSION
This paper is the first report of the global roles of two highly conserved Gram-positive RNases, RNase Y and PNPase, in the human pathogen S. pneumoniae (summarized in Fig. 6). The loss of RNase Y significantly impacts gene expression by affecting ;10% of the pneumococcal transcriptome and thereby causing pleiotropic phenotypes (Fig. 1A, C, and D; Fig. 2A and B; Fig. 3A and B; Tables 1 and 3). In contrast, PNPase exerts a relatively smaller impact on the transcriptome compared to RNase Y but, interestingly, regulates the expression of specific transcripts previously implicated in pneumococcal virulence control (Tables 2 and 4; Fig. 2C and D; Fig. 3C and D). Accordingly, the loss of PNPase severely attenuates S. pneumoniae virulence in vivo ( Fig. 1E; Fig. S2A). This study also revealed that both RNase Y and PNPase work in concert to regulate the processing and decay of several regulatory RNAs, in particular, those characterized by the presence of 59 cis-acting regulatory elements (Fig. 3B and D and 4; Fig. S3; Tables 2 and 4). In addition, RNase Y stabilizes the conserved pneumococcal trans-acting sRNAs CcnA to -E, further impacting Ccn-mediated target gene regulation (Fig. 5).
RNase Y is a pleiotropic regulator in S. pneumoniae D39. Deletion of rny in S. pneumoniae leads to a ;2-fold increase in doubling time in vitro ( Fig. 1A; Table S2), similar to prior observations with B. subtilis and C. perfringens (25,27), and interferes with pneumococcal cell division (Fig. 1C). We identified several important pneumococcal cell wall and division regulators, including mapZ (encoding a midcell anchor protein), cozE (encoding a coordinator of zonal division), and gpsB (encoding a regulator that balances septal and elongation peptidoglycan synthesis), as being significantly upregulated in a Drny mutant (Fig. 2B; Table 1). In S. pneumoniae, MapZ guides tubulin-like FtsZ protein from midcell rings of dividing cells to the equators of new daughter cells (47,48), whereas GpsB and CozE are major peptidoglycan (PG) biosynthesis regulators that play distinct but crucial roles at the midcell to maintain the normal ovococcus shape of pneumococcus by modulating the activities of different penicillinbinding proteins (PBPs), which catalyze peptide cross-link formation in peptidoglycan (49)(50)(51). Accordingly, DmapZ mutants exhibit a variety of abnormal cell shapes and sizes, decreased cell viability, increased doubling time, and aberrant FtsZ movement (47,48,52), while cells depleted of gpsB or cozE form elongated cells that are unable to divide or form chains that round up and lyse, respectively (49,51,53).
Several transcripts under the control of the essential TCS WalRK and the TCSs LiaFSR and CiaRH are impacted in the Drny mutant (Table 1), again consistent with cell wall and surface stress in cells lacking RNase Y, as numerous proteins in these regulons are known to impact cell morphology and chaining (54)(55)(56)(57)(58). In this regard, the defects in cell shape and morphology observed for B. subtilis Drny mutants were attributed to the upregulation of several PG biosynthesis genes, including rodA (27). It remains to be determined what cell wall stress is caused by absence of pneumococcal RNase Y and whether induction of certain proteins in these multiple surface stress TCS regulons can account for the defects in growth and morphology of the S. pneumoniae Drny mutant.
Besides responding to cell wall stress, the CiaRH TCS has been implicated in pneumococcal biofilm formation (59), competence (60), and virulence (61). In particular, the five conserved pneumococcal base-pairing sRNAs (CcnA to -E) negatively regulate translation of comC, which encodes the competence stimulatory peptide (42)(43)(44). We show here that RNase Y functions as a critical regulator of Ccn sRNA stability and impacts Ccn-mediated negative regulation of competence development in S. pneumoniae (Fig. 5). Interestingly, the recent Grad-seq analysis indicated possible stable RNA-protein complexes between the 39-to-59 exonuclease YhaM/Cbf1 and the Ccn sRNAs that were confirmed in pulldown experiments with Ccn sRNAs as bait in S. pneumoniae TIGR4 strain. In addition, CcnA to -E pulled down several degradosome components, including RNase J1/ J2 and PNPase (23). The Gram-positive specific Cbf1 exonuclease has been implicated in trimming single-stranded RNA tails at the 39 ends of Rho-independent terminated transcripts (16,17,23), thereby preventing decay by other exoribonucleases, such as PNPase and RNase R, that require an unstructured tail of 7 to 10 nucleotides (nt) for binding (17,62). Although data presented here suggest that the Ccn sRNAs are targeted by RNase Y (Fig. 5), RNase Y was not identified as a strong Ccn sRNA interactor by Grad-seq (23), perhaps indicating complex dissociation during gradient centrifugation. Future experiments will determine whether Ccn sRNAs are direct substrates of RNase Y and whether Cbf1mediated 39 trimming impacts Ccn sRNA decay by RNase Y. Moreover, results in this paper raise the important question of whether RNase Y functions similarly to RNase E in mediating decay of trans-acting sRNAs that form sRNA-mRNA base pairs in S. pneumoniae and other Gram-positive bacteria.
PNPase is a key regulator of S. pneumoniae D39 virulence. In contrast to the highly pleiotropic effects caused by the absence of RNase Y, the lack of PNPase minimally affected growth or morphology in vitro, but remarkably, caused strong attenuation in vivo (Fig. 1A, B, C, and E). The lack of phenotypes of the Dpnp mutant in vitro may suggest that the pneumococcal 39-59 exoribonuclease RNase R can functionally bypass PNPase under certain experimental conditions. Notably, 10 of 20 protein-coding transcripts that were either upregulated (ribU [;4-fold], fruR [;2-fold], and galE-2 [;2-fold]) or down-regulated (trpACDGE [;2.5-to 4-fold]) in the Dpnp mutant included metabolic genes implicated in nasopharyngeal colonization and/or lung infection in a mouse model (35) (Fig. 2C and D; Table 2). The relative level of the full-length transcript of the T-box riboswitch Spd-sr88 located within the 59 UTR of the trp operon (Fig. 4A) also decreased by ;2-fold in the Dpnp mutant, with concomitant accumulation of spd-sr88-derived decay intermediates ( Fig. 4B; Table 4). These decay products are likely generated by RNase Y cleavage, since the relative full-length Spd-sr88 transcript levels increase by ;11-fold in a Drny mutant (Fig. 4B; Fig. S5).
We do not yet know how PNPase positively regulates the trp operon in S. pneumoniae, but in general, trp operon regulation is important and complex in different bacteria and often involves RNA-based posttranscriptional mechanisms (68). For example, in B. subtilis under tryptophan-replete conditions, trp expression is repressed as a consequence of TRAP regulator protein-mediated transcription termination of the trp leader, which is subsequently degraded by RNase Y and/or J1 and PNPase. (69). In E. coli, tryptophan synthesis is regulated by a classical transcription attenuation mechanism, where under tryptophan-replete conditions, the upstream trpL leader peptide (TrpL) is translated efficiently, allowing formation of a terminator stem-loop that stops transcription before the downstream trp genes (70). Recently, the terminated trpL RNA generated by this attenuation mechanism was shown to function in Sinorhizobium meliloti as a base-pairing sRNA to destabilize several transcripts, including that of the trp biosynthesis genes (71). Likewise, previous studies in important Gram-positive pathogens, including Listeria monocytogenes and Enterococcus faecalis, show that terminated riboswitches are not "junk RNA" but function as mRNA-or protein-binding regulatory RNAs (72,73). S. pneumoniae does not possess obvious homologs of TRAP or TrpL, but its trp operon instead contains the T-box (tRNA-sensing structure) riboswitch Spd-sr88. Whether the RNA decay products derived from spd-sr88 (Fig. 4B) function as regulatory RNAs to destabilize the trp operon transcript in a Dpnp mutant awaits further investigation. These combined results show that PNPase controls the transcript amounts of numerous genes required for pneumococcal pathogenesis, including the trp operon and the riboswitches Spd-sr88 and Spd-sr44 ( Fig. 4B and E; Fig. S5; Tables 2 and 4) (35,46), supporting the notion that PNPase is a key regulator of S. pneumoniae pathogenesis.
RNase Y and PNPase play roles in sRNA processing and decay in S. pneumoniae D39. Riboswitch turnover is important for recycling of the ligands to which they respond, and a role for RNase Y in this process was reported in B. subtilis (11,74) and S. aureus (21). Here, we show that pneumococcal RNase Y mediates the initial endoribonucleolytic cleavage of 59 cis-acting regulatory elements, which are subsequently degraded by PNPase. Likewise, in a Dpnp mutant, we found that decay intermediates of eight riboswitch RNAs accumulated, while their corresponding full-length transcripts increased in abundance in the absence of RNase Y ( Fig. 4; Fig. S3 and S5; Tables 3 and  4). These observations are consistent with a recent study in S. pyogenes showing that the coordinated actions of RNase Y and PNPase play a crucial role in the decay of riboswitches (20). In addition, our data indicate that RNase Y likely generates some sRNAs by cleaving larger transcripts, as observed for Spd-sr88 and Spd-sr116 ( Fig. 4; Fig. S3 and S5). We conclude that RNase Y and PNPase work in tandem to degrade pneumococcal cis-acting regulatory RNAs, while RNase Y also plays an important role in sRNA processing and maturation. Whether RNase Y and PNPase interact together in a degradosome-like complex to impact regulatory RNA levels in S. pneumoniae will be resolved in future experiments.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Bacterial strains used in this study were derived from encapsulated S. pneumoniae serotype 2 strain D39W and are listed in Table S1. Strains were grown on plates containing Trypticase soy agar II (modified; Becton Dickinson [BD]) and 5% (vol/vol) defibrinated sheep blood (TSAII BA) at 37°C in an atmosphere of 5% CO 2 . Liquid cultures were grown statically in BD brain heart infusion (BHI) broth at 37°C in an atmosphere of 5% CO 2 . Bacteria were inoculated into BHI broth from frozen cultures or single colonies. For overnight cultures, strains were first inoculated into 17mm-diameter polystyrene plastic tubes containing 5 ml of BHI broth and then serially diluted 100-fold into five tubes; these cultures were then grown for 10 to 16 h. Cultures with an OD 620 of 0.1 to 0.4 were diluted to a starting OD 620 between 0.002 and 0.005 in 5 ml of BHI broth in 16-mm glass tubes. Growth was monitored by measuring OD 620 using a Spectronic 20 spectrophotometer. For antibiotic selections, TSAII BA plates and BHI cultures were supplemented with 250 mg/ml kanamycin or 150 mg/ml streptomycin.
Construction and verification of mutants. Mutant strains were constructed by transformation of competent S. pneumoniae strains with linear PCR amplicons as described previously (75). DNA amplicons containing antibiotic resistance markers were synthesized by overlapping fusion PCR. S. pneumoniae cells were induced to competence by the addition of synthetic competence stimulatory peptide 1 (CSP-1; Anaspec, Inc.). Markerless deletions and replacements of target genes were constructed using the Kan r rpsL 1 (Janus cassette) allele replacement method as described previously (76). In the first step, the Janus cassette was used to disrupt target genes in an rpsL1 (Str r ) strain background, and transformants were screened for kanamycin resistance and streptomycin sensitivity. In the second step, the Janus cassette was replaced by a PCR amplicon containing the desired mutation or replacement lacking antibiotic markers, and the resulting transformants were screened for streptomycin resistance and kanamycin sensitivity. Final transformants were isolated as single colonies three times on TSAII BA plates containing antibiotics listed in Table S1 and subsequently grown for storage in BHI containing the appropriate antibiotic. All constructs were confirmed by PCR amplification and sequencing.
Microscopy. After cultures reached an OD 620 of ;0.1 to 0.2, 1 ml was removed and centrifuged at 16,000 Â g for 2 min at room temperature. Pellets were suspended in 50 ml of BHI broth. Cells were examined using either a Nikon E200 or a Leica DM 1000 LED phase-contrast microscope, and images were captured using a Nikon DS-Fi3 or a Leica ICC50W camera, respectively. A total of over 100 chains from each of two independent cultures of each strain were counted to determine distributions of numbers of cells per chain.
RNA extraction. RNA for high-throughput sequencing was prepared as described previously (77). Briefly, strains were grown in 30 ml of BHI starting at an OD 620 of 0.002 in 50-ml conical tubes. RNA was extracted from exponentially growing cultures of IU3116 (wild-type parent; D39 rpsL1 CEP::kan rpsL 1 ) and its derived isogenic mutants IU5498 (D39 rpsL1 Dpnp CEP::kan rpsL 1 ) and IU5504 (D39 rpsL1 Drny CEP::kan rpsL 1 ) at an OD 620 of ;0.1 from matched batches of BHI broth for mRNA-seq analysis or from IU1781 (wild-type parent; D39 rpsL1) and its derived markerless mutants IU4883 (D39 rpsL1 Dpnp) and NRD10092 (D39 rpsL1 Drny) at an OD 620 of ;0.15 for sRNA-seq analysis using the FastRNA Pro Blue kit (MP Bio) according to the manufacturer's guidelines. RNA extracted for mRNA-seq analysis was purified with an miRNeasy minikit (Qiagen), which included an on-column DNase I (Qiagen) treatment. For sRNA-seq analysis, RNA was alcohol precipitated following extraction and subsequently subjected to DNase treatment (Turbo DNase; Ambion) following the manufacturer's protocol. Sample mixtures (total reaction volume of 50 ml) were incubated with Turbo DNase for 30 min at 37°C, and each reaction was stopped by addition of 150 ml of diethyl pyrocarbonate (DEPC)-treated water and 200 ml of neutral phenol-chloroform-isoamyl alcohol (Fisher). DNase-treated RNA samples were phenol extracted and alcohol precipitated. To isolate RNA for droplet digital PCR, RNA was extracted from exponential-growth-phase cultures following the procedure described above for sRNA-seq analysis. The amount and purity of all RNA samples isolated were assessed by NanoDrop spectroscopy (Thermo Fisher). RNA integrity of the samples used for RNA-seq library preparation was further assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies).
Library preparation and mRNA-seq. cDNA libraries were prepared from total RNA by the University of Wisconsin-Madison Biotechnology Center as described previously (40). Briefly, total RNA was subjected to rRNA depletion using a RiboZero rRNA removal kit (Epicentre, Inc., Madison, WI, USA). Double-stranded cDNA synthesis was performed with rRNA-depleted mRNA using a ScriptSeq v2 RNA-seq library preparation kit (Epicentre, Inc., Madison, WI, USA) in accordance with the manufacturer's standard protocol. The amplified libraries were purified using Agencourt AMPure XP beads. Quality and quantity were assessed using an Agilent DNA 1000 chip (Agilent Technologies, Inc., Santa Clara, CA, USA) and a Qubit double-stranded DNA (dsDNA) High Sensitivity assay kit (Invitrogen, Carlsbad, CA, USA), respectively. Libraries were standardized to 2 mM and cluster generation was performed using standard Cluster kits (v3) and Illumina Cluster Station. Single-end 100-bp sequencing was performed using standard SBS (sequencing by synthesis) chemistry (v3) on an Illumina HiSeq2000 sequencer. Images were analyzed using the standard Illumina pipeline, version 1.8.2.
Library preparation and sRNA-seq. sRNA libraries were prepared from total RNA as described previously (40). Briefly, 5 mg of DNase-treated total RNA was subjected to rRNA removal (RiboZero rRNA removal for Gram-positive bacteria; Illumina). rRNA-depleted samples were then subjected to RNA fragmentation using the Ambion RNA fragmentation kit (AM8740). Fragmented RNA was subjected to RNA 59-polyphosphatase (Epicenter) treatment, which was performed to facilitate the 59-adapter ligation step. Small RNA libraries were generated by Macrogen using the TruSeq small RNA library kit (Illumina). Then, 100-bp paired-end read sequencing was performed using an Illumina HiSeq2000 sequencer.
RNA-seq analysis. Raw sequencing reads from mRNA-seq were quality and adapter trimmed using Trimmomatic version 0.17 (78) with a minimum length of 90, while those corresponding to sRNA-seq were preprocessed for alignment with Cutadapt. The trimmed reads were mapped on the Streptococcus pneumoniae D39 (RefSeq NC_008533) genome and D39 plasmid pDP1 sequence (RefSeq NC_005022) using Bowtie2 (79). mRNA-seq and sRNA-seq analysis were performed as described previously using DESeq2 (77). Genes were defined as differentially expressed if their P adj (P value adjusted for multiple testing) was ,0.005. Primary data from mRNA-seq and sRNA-seq analyses was submitted to the NCBI Gene Expression Omnibus (GEO). The accession numbers for the sRNA-seq data corresponding to wild-type samples used for comparison of Drny and Dpnp mutants are GSE148867 and GSE123437, respectively. ddPCR analysis. One microgram of DNA-free RNA was reverse transcribed using random hexamers and Superscript III reverse transcriptase (RT) (Invitrogen) following the manufacturer's protocol. For each sample, a no-RT (NRT) control reaction was performed. cDNA samples were diluted 1:10, 1:10 2 , 1:10 3 , or 1:10 6 , and 2 ml of each diluted RT and NRT PCR sample was added to a 22-ml reaction mixture containing 11 ml of QX200 ddPCR EvaGreen Supermix (Bio-Rad) and 1.1 ml of each ddPCR primer, each at 2 mM (Table S3). A single no-template control (NTC) for each ddPCR primer pair used in this study was included. Droplet generation from each reaction mixture was achieved via the QX200 automated droplet generator (Bio-Rad), and endpoint PCR was performed using a thermal cycler following the instructions from the manufacturer. A QX200 droplet reader (Bio-Rad) was used to analyze droplets from each individual reaction mixture, where PCR-positive and PCR-negative droplets were counted to provide absolute quantification of the target transcript. Data analysis was performed with QuantaSoft software (Bio-Rad), and the concentration of each target is expressed as copies per microliter. Reactions were performed using cDNA from at least three independent biological replicates, and transcript copies were normalized to 16S rRNA (internal control). Normalized transcript copy numbers were used to calculate fold changes of transcripts corresponding to target genes in different sets of mutants relative to the WT parent. Statistical analysis was performed using Student's t test in GraphPad Prism version 7.0.
RNA stability assay. To determine RNA stabilities, cultures were grown in BHI to exponential phase (OD 620 % 0.15) as described above, and a culture sample (0 h after the end of log-phase growth [T 0 ]) was collected. Rifampin was added to inhibit transcription, and additional samples were collected 5, 10, 20, and 30 min after rifampin addition. All samples were subjected to hot phenol lysis as described previously (80). Briefly, 700 ml of sample was added to a mixture containing 800 ml of acid phenolchloroform-isoamyl alcohol, pH 4.3 (Fisher Scientific), and 100 ml of lysis buffer (320 mM sodium acetate [pH 4.6], 8% [wt/vol] SDS, and 16 mM EDTA) equilibrated to 65°C. Samples were mixed at 65°C for 5 min and centrifuged for 30 min at 4°C to separate phases. The upper aqueous phase was extracted a second time with an equal volume of neutral phenol-chloroform-isoamyl alcohol, pH 6.7 (Fisher Scientific). RNA was ethanol precipitated and resuspended in DEPC-treated water. RNA concentration was measured using a NanoDrop 2000 (Thermo Fisher Scientific).
Northern blot analysis. Two micrograms of each RNA sample was loaded on 10% polyacrylamide gels containing 7 M urea or loaded onto 10% Criterion TBE-urea precast gels (Bio-Rad) and electrophoresed at 85 V. RNA samples were transferred to a Zeta-Probe GT membrane (Bio-Rad) using a Trans-Blot SD semidry transfer apparatus (Bio-Rad) following the manufacturer's guidelines. Transferred RNA was UV cross-linked and hybridized overnight with 100 ng/ml of 59 biotinylated DNA probe (Table S3) in Ultrahyb (Ambion) hybridization buffer at 42°C. Blots were developed using a BrightStar BioDetect kit protocol (Ambion), imaged with a ChemiDoc MP imager (Bio-Rad), and quantified using Image Lab software version 5.2.1 (Bio-Rad). Signal intensity corresponding to each sRNA was normalized to that of 5S rRNA, which served as an internal loading control. Decay curves corresponding to RNA stability time course experiments were generated by using GraphPad Prism version 7.0.
Mouse models of infection. All procedures were approved in advance by the Bloomington Institutional Animal Care and Use Committee (BIACUC) or UTHealth Animal Welfare Committee and were performed according to recommendations of the National Research Council. Experiments were performed as described in reference 76, with the following changes. Male ICR mice (21 to 24 g; Harlan) were anesthetized by inhaling 4% isoflurane (Butler Animal Health Supply) for 8 min. In two independent experiments, a total of 8 mice were intranasally inoculated with each bacterial strain to be tested. Bacteria were grown exponentially in BHI broth in an atmosphere of 5% CO 2 to an OD 620 of ;0.1. Ten milliliters of culture was centrifuged for 5 min at 14,500 Â g and then suspended in 1 ml 1Â PBS to yield ;10 7 CFU ml 21 . CFU counts were determined by serial dilution and plating. Fifty microliters of suspensions was administered intranasally as described previously (75). Mice were monitored visually at 4-to 8-h intervals, and moribund mice were euthanized by CO 2 asphyxiation followed by cervical dislocation (IU-Bloomington), which was used as the time of death in statistical analyses. Alternatively, isoflurane-anesthetized moribund mice were euthanized by cardiac puncture-induced exsanguination followed by cervical dislocation (UTHealth). Kaplan-Meir survival curves and log-rank tests were generated using GraphPad Prism 7.0 software.
b-Galactosidase assays. Strains containing the comC'-'lacZ translational fusion were grown in BHI broth to exponential phase (OD 620 % 0.15). Samples were taken from each culture and assayed for b-galactosidase activity as described by Miller (81), with slight modifications. Briefly, 1 ml of culture was removed and centrifuged at 16,000 Â g for 2 min at 4°C. Pellets were resuspended in 1 ml of Z-buffer containing 2b-mercaptoethanol at a final concentration of 0.27%. Each sample mixture was lysed by subsequent incubation at 37°C for 10 min following the addition of 10 ml of 5% (vol/vol) Triton. One hundred microliters of lysed culture samples was then assayed for b-galactosidase specific activity as described previously (81).
Competence assays. Overnight cultures of strains were diluted into 5 ml of C1Y medium (caseinbased medium supplemented with yeast extract), pH 8 (7), to a starting OD 620 of ;0.002. Starting from the initial inoculation and at 1-h intervals thereafter, 1 ml of a cell suspension was removed and mixed with 50 ng of amplicon DNA carrying a kanamycin resistance marker. After incubation for 90 min at 37°C in an atmosphere of 5% CO 2 , samples were serially diluted and plated on blood agar plates containing 250 mg/ ml kanamycin and on blood agar plates without antibiotics to determine transformant CFU and total viable CFU, respectively. The transformation frequency (TF) was determined as the ratio of Kan r CFU to total CFU per unit volume of cell suspension. Under these culture conditions, the natural transformation frequency of the wild-type strain followed a reproducible pattern with time in culture, with a high peak (2 Â 10 25 ) about 2 h after inoculation (OD 620 = 0.02 to 0.03). Accordingly, the natural transformation frequencies of the Drny mutant and wild-type strains were determined around an optical density of ;0.02 to 0.03.
Data availability. The sRNA-seq data corresponding to Drny and Dpnp mutants and the mRNA-seq data corresponding to all strains have been deposited in GEO under the accession number GSE173392.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.