Processing of the Escherichia coli leuX tRNA transcript, encoding tRNA^Leu5, requires either the 3′→5′ exoribonuclease polynucleotide phosphorylase or RNase P to remove the Rho-independent transcription terminator

Abstract

Here we report a unique processing pathway in Escherichia coli for tRNA^Leu5 in which the exoribonuclease polynucleotide phosphorylase (PNPase) removes the Rho-independent transcription terminator from the leuX transcript without requiring the RhlB RNA helicase. Our data demonstrate for the first time that PNPase can efficiently degrade an RNA substrate containing secondary structures in vivo. Furthermore, RNase P, an endoribonuclease that normally generates the mature 5′-ends of tRNAs, removes the leuX terminator inefficiently independent of PNPase activity. RNase P cleaves 4–7 nt downstream of the CCA determinant generating a substrate for RNase II, which removes an additional 3–4 nt. Subsequently, RNase T completes the 3′ maturation process by removing the remaining 1–3 nt downstream of the CCA determinant. RNase E, G and Z are not involved in terminator removal. These results provide further evidence that the E. coli tRNA processing machinery is far more diverse than previously envisioned.

INTRODUCTION

The 86 transfer RNA (tRNA) genes in Escherichia coli are transcribed either as part of complex operons containing multiple tRNAs, ribosomal RNAs (rRNAs) or messenger RNAs (mRNAs) or as monocistronic transcripts. Irrespective of their physical organization, each precursor tRNA contains extra sequences at both the 5′- and 3′-ends that are removed to generate a mature, functional tRNA. Either RNase E or RNase P are required for the separation of tRNAs that are transcribed as part of polycistronic transcripts to generate pre-tRNAs that become substrates for RNase P, which cleaves them endonucleolytically to generate the mature 5′ termini (1).

While 5′-end processing of all tRNAs has been shown to be catalyzed by only endonucleases, 3′-end maturation is believed to be a multistep process involving both endo- and exoribonucleases (2). For example, for tRNA transcripts terminating with a Rho-independent transcription terminator, an initial endonucleolytic cleavage can remove the stem-loop followed by exonucleolytic trimming to generate the mature 3′ terminus (3–5). In contrast, 3′-end processing can be all exonucleolytic for tRNA transcripts that are terminated in a Rho-dependent fashion (6). Of the eight known 3′→5′ exoribonucleases in E. coli, RNase T and RNase PH are the most important for final 3′-end maturation (7).

Interestingly, polynucleotide phosphorylase (PNPase), a major 3′→5′ exoribonuclease encoded by pnp gene, has only been shown to be involved in shortening of long 3′ trailer sequences (>15 nt) generated after intergenic endonucleolytic cleavages of polycistronic tRNA operons (3,8) or Rho-dependent transcription termination (6). Thus, although PNPase plays a significant role in the processing and degradation of a variety of RNA species (7), its role in tRNA processing and maturation has been thought to be minimal.

Of the eight leucine tRNA genes encoded in E. coli, seven are part of five different polycistronic operons that depend on either RNase E or RNase P or both for initial processing and maturation (3–6). Here we describe a unique processing pathway for the eighth leucine tRNA, leuX that encodes tRNA^Leu5. Specifically, PNPase initiates processing by removing the the Rho-independent transcription terminator and stops 1–3 nt downstream of the CCA determinant. This is the first in vivo demonstration of processing of an RNA substrate containing a stem-loop structure by PNPase, even though the enzyme has been shown to be inhibited by such structures in vitro (9). Endonucleases such as RNase E, RNase G, RNase LS or RNase Z are not involved. Our data further show that RNase P, the endonuclease responsible for generating mature 5′ termini, also removes the terminator from ∼10% of primary transcripts by cleaving 4–7 nt downstream of the CCA determinant, generating substrates for RNase II, which removes an additional 3–4 nt. Subsequently, RNase T completes the 3′ maturation process by removing the last 1–3 nt downstream of the CCA determinant left by either PNPase or RNase II.

MATERIALS AND METHODS

Bacterial strains

The E. coli strains used in this study were all derived from MG1693 (thyA715) (E. coli Genetic Stock Center, Yale University). The rne-1 and rnpA49 alleles encode temperature sensitive RNase E and RNase P proteins, respectively, that do not support cell viability at 44°C (10–12). SK5665 (rne-1) (11), SK2525 (rnpA49 rbsD296::Tn10) (4), SK2534 (rne-1 rnpA49 rbsD296::Tn10) (4), SK10019 (pnpΔ683) (13), CMA201 (Δrnb) (14), SK5726 (pnp-7 rnb-500) (15), SK10443 (rnpA49 rbsD296::Tn10 pnpΔ683) (6), SK7988 (ΔpcnB) and SK10148 (Δrnt) (4) have been previously described. A P1 lysate grown on SVK53 (MC1061/ΔrhlB) was used to transduce both MG1693 and SK10019 (pnpΔ683) to construct SK10553 (ΔrhlB) and SK10554 (pnpΔ683 ΔrhlB), respectively. SVK53 is similar to SVK1 (MC1061/ΔrhlB) (16) but has a nearby mini Tn10 that was used for tetracycline selection. A P1 lysate grown on JW2798 (Keio Collection, Japan) was used to transduce MG1693 and SK2525 (rnpA49) to construct SK4390 (ΔrppH::kan) and SK4395 (ΔrppH:kan rnpA49), respectively. SK2059 (rnr::kan) was constructed by P1 mediated transduction using CA265R⁻ (rnr::kan) as the donor strain.

Growth of bacterial strains, isolation of total RNA and northern analysis

Bacterial strains were grown (6) and total RNA was extracted as described previously (17). The RNAs were quantified by measuring the OD₂₆₀ using a Nanodrop (ND1000) apparatus. Total RNA was separated in 6% polyacrylamide gels containing 8 M urea in Tris–Borate–EDTA (TBE) buffer as described previously (6).

Primer extension

Primer extension analysis of the leuX transcripts was carried out as described previously (6) with the following modifications. The leuX nucleotide sequence was obtained from a PCR DNA product (containing leuX genomic sequences) using the Promega fmol sequencing kit and the primer LEUX (primer b, Figure 1A) that was also used for the reverse transcription. The sequences were analyzed on a 6% PAGE containing 8M urea.

Figure 1.

Analysis of the processing of leuX transcript. (A) Schematic presentation of leuX coding sequences (not drawn to scale). Relative positions of the oligonucleotide probes (a, LEUX-UP, 5′ACTGAAGAGGTATGCGGAAA3′; b, LEUX, 5′CGTATTTCTACGGTTGATTTTG3′; and c, LEUX-TER, 5′ACCTCAAGTTGCGGTCTATTTAC3′) used in the northern analysis are shown below the cartoon. (B–G) Northern analysis of leuX transcript. Total RNA (12 µg/lane) was separated on 6% PAGE, transferred to nylon membrane and probed multiple times with probes a–c as described in the ‘Materials and Methods’ section. The blots probed with b are shown. The genotypes of the strains used are noted above each lane. The deduced structures of the processing intermediates of leuX transcript are shown in between panels B and C. The full-length primary transcripts, which were identical, are not shown in panel G. The apparent difference in the electrophoretic mobility of the mature leuX transcript seen in panel C is due to partial bending of the gel during transfer to the membrane.

RT–PCR cloning and sequencing of 5′–3′ ligated transcripts

The 5′- and 3′-ends of leuX transcripts were identified by cloning and sequencing the RT–PCR products obtained from 5′→3′ end-ligated circular RNAs following the methods described previously (17). The 5′–3′ junctions of the cDNAs were amplified with a pair of gene-specific primers using GoTaq®Green Master Mix (Promega).

In vitro transcription and RNase E digestion of leuX transcripts

The full-length leuX runoff transcript was obtained by in vitro transcription using a PCR amplified DNA template that contained a modified T7 RNA polymerase promoter. The leuX DNA template was amplified using a 5′-primer (T7-LEUX, 5′GGATCCTAATACGACTCACTATAGTTTTCCGCATACCTCTTCA3′) and a 3′-primer (LEUX-442, 5′AACACTGGATTTCAGGCATAA3′) that was designed to generate the leuX transcript with identical in vivo 5′ (+1) end and 3′ Rho-independent transcription terminator. The 5′-primer contained a modified T7 RNA polymerase promoter (underlined) in order to start the transcription at G (bold). The PCR DNA template was purified using PCR purification kit (Qiagen) and used in the in vitro transcription reaction with T7 RNA polymerase (Promega) to generate either unlabeled or uniformly labeled (with α-³²P-UTP) RNAs. Tri- or monophosphorylated transcripts were generated by adding either GTP or GMP (10-fold in excess), respectively. All transcription reactions generated a single band and were purified using PCR purification column (Qiagen). The unlabeled transcript was used to confirm the 5′ (+1) end by primer extension analysis (data not shown).

His-tagged (carboxy terminus) RNase E was purified from SK10154 (thyA715 ompT1::kan) and used in the in vitro reaction as described previously (18). 9S rRNA precursor substrates were used as control to test the RNase E activity as described before (Chung and Kushner, manuscript in preparation). Two ng of the purified RNase E protein was required to process >50% of 9S rRNA substrate in 4 min at 37°C (data not shown). Either 5′-tri- or 5′ monophosphorylated leuX transcripts were assayed in presence of 1–40 ng of RNase E at 37°C. Samples (5 µl) were removed at 0, 2, 4, 8, 16 and 32 min and the reaction was stopped by adding equal volume (5 µl) of 2 × RNA loading dye (17). The samples were heated at 65°C for 15 min before separating on a 20% denaturing PAGE/7 M urea. Oligo (dA)_10–30 (5′-end-labeled with γ-³²P-ATP) were used as the molecular weight size standards.

RESULTS

RNase E is not involved in the processing of leuX transcript

As part of our efforts to identify every tRNA processing pathway in E. coli, we examined the processing of leuX, encoding tRNA^Leu5. This is the only leucine tRNA that is transcribed as a monocistronic transcript and terminates with a putative Rho-independent transcription terminator (Figure 1A) (19). Since previous studies have shown that the Rho-independent transcription terminators associated with many primary tRNA transcripts are removed endonucleolytically by RNase E (3–5) (Table 1), we hypothesized that RNase E would be required for processing of this transcript.

Table 1.

Analysis of the Rho-independent transcription terminators associated with tRNA transcripts

Transcripts tRNA	Nucleotide sequence downstream of CCA determinant	Terminator
		Stem length (nts)	Percentage (GC)	Free energy (kcal)
leuX	CCAAAAGUAUGUAAA	7	57	−9.9
trpT	CCAGAAUCAUCCUUA	9	44	−14
lysQ	CCAAUGUAAAAAAGC	11	45	−14.5
lysV	CCAAUGUAAAAAAGC	11	45	−14.5
proL	CCAAAAAUCCCAAGA	13	31	−15.3
leuZ	CCAUGGGAAAGAUAA	8	50	−15.7
asnV	CCAAAUUCAAAAAAG	12	42	−15.7
proK	CCAAAUUCGAAAAGC	12	42	−16.2
pheU	CCAAAUUCAUAUAAA	11	45	−16.3
glnX	CCACAUUAAAAAAGC	11	45	−16.4
pheV	CCACUAAUUCUUAAG	11	55	−16.5
proM	CCAAUUUUGAACCCC	9	67	−17.0
aspU	CCACUUAUUAAGAAG	14	43	−17.1
alaX	CCAAAUUUCCAACCC	10	70	−18.2
leuU	CCAAAUUCCAGAAAA	13	38	−18.7
serX	CCAUAUUUAAAGAAG	15	33	−19.8
serW	CCAUAUUUAAAGAAG	16	31	−21
aspV	CCACUAUUCACUCAU	12	33	−21.3

The Rho independent transcription terminators associated with the genes shown in bold have been shown to be removed endonucleolytically by RNase E or RNase G.

Accordingly, northern analysis was carried out in strains deficient in RNase E, RNase P or both enzymes, since we anticipated that both enzymes would be required to generate the mature tRNA. Initial probing with a tRNA^Leu5 specific probe b (Figure 1A) revealed two bands in both the wild-type and rne-1 strains (Figure 1B, lanes 1–2). The nucleotide composition of both bands was determined by using additional probes that were complementary to the 5′ upstream and 3′ downstream of immature sequences (a,c, Figure 1A). Since probes a and c only hybridized to high molecular weight band (data not shown), this species (∼140 nt) was the full-length primary transcript, while the more abundant smaller species represented the 85 nt mature tRNA.

Surprisingly, the relative quantities (RQ) of the primary transcript as well as the mature tRNA^Leu5 were identical in the wild-type and rne-1 strains (Figure 1B, lanes 1–2; Table 2). Furthermore, the processed fractions (PFs, defined as the percentage of mature species relative to the total amount of both processed and unprocessed tRNA transcript) were also identical. Since the Rne-1 protein retains significant RNase E activity on tRNA substrates at the nonpermissive temperature (5), we also tested an rne deletion strain (rneΔ1018) (Chung and Kushner, manuscript in preparation). The complete absence of RNase E also had no effect on the processing of leuX transcript (Figure 1B, lane 5; Table 2).

Table 2.

Analysis of leuX processing intermediates

Strain genotype	Relative quantity of primary transcripta	Relative quantity of mature transcripta	PFb (%)
Wild-type	1	1	99 ± 1
rne-1	1 ± 0.1	1 ± 0.1	99 ± 1
rnpA49	1 ± 0.1	0.4 ± 0.1	87 ± 3
rne-1 rnpA49	1 ± 0.2	0.4 ± 0.1	85 ± 3
Δrne rng-219	1 ± 0.1	1 ± 0.1	99 ± 1
Δrnb	1 ± 0.1	0.5 ± 0.1	87 ± 3
Δpnp	10 ± 2	1 ± 0.1	80 ± 3
Δpnp rnpA49	10 ± 2	0.1 ± 0.05	40 ± 3
ΔpcnB	2 ± 0.5	0.9 ± 0.1	98 ± 1
ΔrhlB	1 ± 0.1	1 ± 0.1	99 ± 1
Δrnr	1 ± 0.1	1 ± 0.1	99 ± 1
Δrnt	1 ± 0.1	0.3 ± 0.1	30 ± 5

^aRelative quantities of primary and mature transcripts compared to the wild-type control.

^bPF is the percentage of mature species relative to the total amount of both processed and unprocessed tRNA transcripts. The data presented represents the average of at least three independent determinations.

However, in a strain carrying the rnpA49 allele, thermal inactivation of RNase P led to ∼60% reduction in RQ of the mature tRNA, although there was no change the RQ of the primary transcripts (Figure 1B, lane 3, Table 2). In addition, a smear of bands larger than the mature tRNA species appeared under these circumstances (Figure 1B, lane 3, asterisk). The largest of these species was found to contain the 5′ unprocessed end based on its hybridization to probe a and not to probe c (Figure 1A, data not shown). This result was consistent with the RNase P generating the mature 5′ terminus of tRNA^Leu5. Interestingly, inactivation of both RNase P and RNase E led to disappearance of the majority of the heterogeneous pre-tRNA precursor species (Figure 1B, lane 4, asterisk) except for the species containing the intact 5′-end and a processed 3′-end. In fact, the intensity of this transcript increased ∼2.5 ± 0.5-fold compared to the rnpA49 single mutant (Figure 1B, lanes 3–4) indicating a role for limited RNase E activity within the 5′ leader region resulting in the increased amounts of the heterogeneous tRNA processing intermediates seen in the rnpA49 single mutant (Figure 1B, compare lanes 3 and 4). Nevertheless, within experimental error, the PF of the mature leuX tRNA in both rnpA49 single and rnpA49 rne-1 double mutants remained unchanged (Table 2).

The leuX Rho-independent transcription terminator is removed independent of RNase G, RNase Z or RNase LS activity

Since we have previously shown some involvement of RNase G in the removal of the terminator from the secG leuU transcript (5), we tested a series of isogenic strains carrying various combinations of mutations in RNase G, RNase Z, RNase LS, RNase E and RNase P in case there were redundant processing pathways for leuX. Accordingly, we analyzed rng, rne-1 rng, rnz, rne-1 rnz, rnlA, rne-1 rnlA, rne-1 rng rnz rnlA and rne-1 rng rnz rnlA rnpA49 strains. To our surprise, the RQs of the full-length primary transcript and mature tRNA were identical, within experimental error, in all of the strains tested (data not shown).

The leuX primary transcript is a substrate for RppH but 5′ dephosphorylation is not required for processing

Recently, it has been shown that the RppH protein converts the 5′ triphosphate terminus of many E. coli transcripts to a 5′ phosphomonoester (20), making these RNAs more susceptible to degradation by 5′-end-dependent ribonucleases such as RNase E and RNase G (21–23). Even though neither of these enzymes was involved in leuX tRNA processing, we still wanted to investigate if failure to remove the 5′ triphosphate might interfere with leuX processing. Total RNA isolated from wild-type, ΔrppH, rnpA49 and ΔrppH rnpA49 strains was analyzed by Northern analysis. The RQ and PF of the mature tRNA were comparable in the wild-type and ΔrppH strains as well as in the rnpA49 and ΔrppH rnpA49 strains (data not shown).

Even though the absence of RppH did not affect the processing of leuX, we tested to see if the leuX primary transcript was a substrate for RppH. Since all primary transcripts contain a 5′ triphosphate, the absence of RppH will inhibit self-ligation unless dephosphorylation occurred by an unidentified RNA pyrophosphohydrolase, the transcript was cleaved near the 5′ terminus by an endoribonuclease or was degraded by an E. coli functional equivalent of the B. subtilis 5′→3′ exonuclease RNase J1 (24). In order to distinguish these possibilities, total RNA isolated from the wild-type and ΔrppH strains was self-ligated using T4 RNA ligase. As a control, half of each RNA sample was treated with tobacco acid pyrophosphatase (TAP) to convert the 5′ triphosphate to monophosphate before ligation. The 5′–3′ junctions were reverse transcribed, cloned and sequenced (Figure 2) as described in the ‘Materials and Methods’ section. Without TAP treatment, only ∼11% (3/27) of the wild-type clones ligated at the first nucleotide (G) of the primary transcript (19) as compared to 63% (10/16) after TAP treatment. In contrast, none of the clones (0/21) were ligated at the first nucleotide in the ΔrppH strain in the absence of TAP treatment. However, ∼65% (13/20) of the clones were ligated at the first nucleotide in the ΔrppH strain after TAP treatment.

Figure 2.

Identification of 5′- and 3′-ends of leuX transcripts using RT–PCR cloning of 5′–3′ self-ligated transcripts. The 5′ triphosphate ends of RNAs from selected strains were converted to monophosphate by treating with TAP. (A) Schematic presentation of leuX coding sequences showing the 5′ upstream and 3′ trailer sequences. A potential Rho-independent transcription terminator at the 3′-end is shown. The nucleotide at the beginning of the transcription start (as identified in Figure 5) is numbered as 1. (B) The 5′- and 3′-ends of leuX transcripts in wild-type, ΔrppH, rnpA49, rne-1 rnpA49 and Δrnb strains. Each arrow (both black and red) before the nucleotides at 5′-end and after the nucleotides at 3′-end represent a 5′- and 3′-end, respectively. A red arrow indicates the presence of 1–5 non-templated A’s after the nucleotide. The numbers directly above or below each arrow represents the number of clones identified with these respective 5′- or 3′-ends.

A few of the clones in both genetic backgrounds had 5′-ends upstream of the 5′ mature terminus (Figure 2), which might have arisen from the RNase E cleavages described in Figure 1B. In order to confirm that these species were derived from RNase E cleavages, the 5′–3′ ligated junctions of leuX transcripts derived from rnpA49 single and rne-1 rnpA49 double mutants were cloned without TAP treatment and sequenced (Figure 2). In the rnpA49 single mutant ∼83% (19/23) of the clones had leuX transcripts whose 5′-ends fell between the transcription start site and the mature terminus. In contrast, only ∼10% (3/25) of the leuX transcripts in rne-1 rnpA49 double mutant had 5′ truncated ends. In addition, it should be noted that ∼88% (22/25) of the leuX transcripts isolated from the double mutant ligated at the first nucleotide without TAP treatment indicating the presence of a 5′ monophosphate terminus, which was presumably generated by the action of RppH.

PNPase removes the leuX Rho-independent transcription terminator

Since none of the endoribonucleases tested (RNase E, RNase G, RNase LS, RNase Z or RNase P) seemed to have any significant effect on the removal of the leuX transcription terminator, we turned our attention to the major 3′→5′ exonucleases present in E. coli. While the activities of PNPase, RNase II and RNase T are all strongly inhibited by secondary structures, RNase R has been shown to degrade structured RNA substrates (9,25,26). Northern analysis of Δrnb, Δrnt, and Δrnr mutant strains showed no effect on the level of the full-length leuX transcripts (Figure 1C, Table 2, data not shown). In contrast, the level of the full-length transcript increased 10 ± 2-fold in the Δpnp mutant compared to the wild-type control, indicating a significant role of PNPase in the removal of the terminator (Figure 1D, Table 2). In addition, a new minor species that had been processed at the 5′-end but retained the unprocessed 3′ terminator was also identified in the Δpnp strain (Figure 1D, lane 2).

The RhlB RNA helicase is not required for PNPase to remove the leuX transcription terminator

It has been shown that the RhlB RNA helicase binds directly to PNPase both in vivo and in vitro and helps promote the degradation of double-stranded RNA degradation by PNPase in vitro (27). Accordingly, we anticipated that the RhlB RNA helicase assisted PNPase in the removal of leuX transcription terminator. We predicted that the full-length leuX primary transcript would accumulate in a ΔrhlB strain, comparable to what was observed in the Δpnp mutant (Figure 1D, E, lane 2). As shown in Figure 1E (lanes 1–3) and Table 1, the wild-type and ΔrhlB strains had almost identical levels of the full-length transcripts. Furthermore, the leuX processing pattern in an ΔrhlB Δpnp double mutant (Figure 1E, lane 4) was identical to Δpnp single mutant (Figure 1E, lane 2).

Polyadenylation by PAP I only has a minor effect on leuX tRNA processing

It has been shown that repeated rounds of polyadenylation by PAP I helps PNPase degrade RNA species containing secondary structures in vitro (28,29). Since PNPase appeared to be the primary ribonuclease responsible for the removal of the leuX Rho-independent transcript terminator, we tested to see if PAP I was required for this reaction. As shown in Figure 1F (lanes 1–2), there was ∼2-fold increase in the level of the full-length transcript in the ΔpcnB single mutant compared to the wild-type strain (Table 2). However, the inactivation of both PNPase and PAP I did not increase the level of the full-length transcript compared to the Δpnp single mutant (data not shown). While polyadenylation was not required for the efficient removal of the Rho-independent transcription terminator, it was of considerable interest that 33% of the sequenced leuX transcripts showed the presence of poly(A) tails (1–5 nt in length) at the 3′ termini of both the processing intermediates and the full-length transcripts in all genetic backgrounds (Figures 2 and 4), but none were found at the mature 3′ terminus.

Figure 4.

Identification of 5′- and 3′-ends and 3′ Rho-independent transcription terminator of leuX transcript in SK10019 (Δpnp) using RT–PCR cloning of 5′–3′ self-ligated transcripts. (A) Schematic presentation of leuX coding sequences as described in Figure 2A. (B) The 5′- and 3′-ends of leuX transcripts in Δpnp strain. Arrows are as described in legends to Figure 2B. (C) The secondary structure of the Rho-independent transcription terminator as determined by RNA-STAR program (31). The CCA terminus of tRNA^Leu5 is underlined. P indicates the putative RNase P cleavage site as determined by the in vitro studies of Nomura and Ishihama (30).

RNase P also plays a role in the 3′-end processing of leuX

Since PNPase was involved in the removal of the 3′ Rho-independent transcription terminator of the leuX transcript (Figure 1D, lane 2), we were surprised that the RQ of the mature tRNA was identical in the Δpnp and wild-type strains (Table 2). Accordingly, we hypothesized the presence of other ribonuclease(s) that could inefficiently remove the 3′ terminator either in the absence of or in combination with PNPase. As described above, we had not seen any significant effect on terminator removal with RNase II, RNase T, RNase R, RNase E, RNase G, RNase Z, RNase LS or RNase P (Figure 1, data not shown). However, because PNPase was primarily responsible for terminator removal, the role played by a back-up ribonuclease could easily have been overlooked.

Since Nomura and Ishihama (30) suggested, based on in vitro experiments, that RNase P might be involved in the processing of both the 5′- and 3′-ends of the leuX tRNA, we hypothesized that if RNase P processed both ends of the leuX transcript in vivo, we should observe a dramatic drop in both the RQ and PF of the mature leuX species in an rnpA49 Δpnp double mutant compared to either of the single mutants. In fact, both the RQ and PF of the mature tRNA were significantly reduced in the double mutant compared to the single mutants (Figure 1B, lane 3, 1D, lanes 2–3; Table 2). Thus, in the absence of both PNPase and RNase P, the RQ of the mature tRNA was lowered to 10% of the wild-type level whereas only 40% of the tRNA (PF) was matured (Table 2). In contrast, both the Δpnp and rnpA49 single mutants had significantly higher RQ and PF levels of mature tRNA (Table 2).

In order to provide further support for our conclusion that either PNPase or RNase P was required for the removal of the transcription terminator, we measured the half-life of the leuX primary transcript. As shown in Figure 3, no full-length transcripts were detected in the wild-type strain indicating its half-life was less than 30 seconds (6). In contrast, the half-life of the primary transcript in the pnp deletion strain was 4.9 ± 0.5 min (Figure 3). The half-life was further increased to 6.7 ± 0.5 min in the Δpnp rnpA49 double mutant (data not shown).

Figure 3.

Determination of the half-life for the primary leuX transcript. Total RNA (12 µg/lane) from MG1693 (wild-type) and SK10019 (Δpnp) was isolated at times (minutes after rifampicin addition) indicated at the top of the blot and was separated on 6% PAGE, transferred to nylon membrane and probed with probe b (Figure 1A) as described in the ‘Materials and Methods’ section. The intensity of the bands corresponding to the full-length transcript was quantified with a PhosphorImager (Amersham Bioscience) and the values (log of the percentage of transcript remaining) were plotted as a function of time to calculate the half-life.

RNase T and RNase II are also required for leuX processing

While the RQ of the mature tRNA in the pnp deletion strain was identical to the wild-type strain (Figure 1D, lanes 1–2), the PF of the mature tRNA decreased to ∼80%, which was lower than in the rnpA49 single mutant (Table 2). In contrast, both the RQ and PF of the mature tRNA in the Δrnt strains were reduced to 30% of the wild-level (Table 2). These data were consistent with the significant accumulation of 1–3 nt longer precursor products (Figure 1G), presumably dependent on the known role of RNase T in the final 3′-end maturation.

The RQ and PF of the mature tRNA in the Δrnb strain were also reduced to 50% and ∼87%, respectively, of the wild-type strain (Table 2), with the concomitant appearance of new processing intermediates which were 3–4 nt longer than the mature tRNA (Figure 1C, lane 2, arrow). These longer processing intermediates were ∼11 ± 2% of the total leuX tRNA transcripts in the Δrnb strain and were larger than the processing intermediates seen in the Δrnt strain (Figure 1G). Furthermore, the levels of the mature tRNA and the 3–4 nt longer unprocessed tRNA did not change in a pnp-7 rnb-500 double mutant compared to the rnb deletion strain (data not shown). This suggested no involvement of PNPase in the generation of the longer unprocessed tRNA in the Δrnb strain. However, the full-length transcript level in the double mutant increased ∼8 ± 2-fold compared to the wild-type and Δrnb strains (data not shown), which was consistent with the role played by PNPase in removing the transcription terminator.

In order to characterize the larger processing intermediates, we cloned the 3′–5′ junction of the self-ligated leuX tRNA transcripts from the RNase II deletion strain. Eighty-two percent (28/34) of the clones had 3′-ends between the second and seventh nt downstream of CCA, while 75% (21/28) of these ended between fourth and seventh nt (Figure 2). In contrast, majority of leuX transcripts with immature 3′ termini in rnb⁺ strains contained 1–3 extra nt downstream of CCA (Figure 2). Specifically, in the absence of RNase P (both rnpA49 and rnpA49 rne-1) ∼77% (36/49) of the transcripts had immature 3′ termini containing 1–3 nt downstream of the CCA (Figure 2).

The leuX transcription terminator has a low predicted free energy

Since PNPase is thought to be strongly inhibited by secondary structures (9), its ability to remove the leuX Rho-independent transcription terminator was surprising. In order to confirm that the predicted Rho-independent transcription terminator (Figure 1A) was functional in vivo, we cloned and sequenced 5′–3′ junctions of leuX transcripts isolated from a Δpnp strain. Consistent with the northern analysis, ∼57% (13/23) of the clones contained the full-length leuX transcript (Figure 4B) compared to only ∼11% (3/27) in the wild-type control (Figure 2). Furthermore, ∼65% (15/23) of the transcripts terminated in between nucleotides 146–148 which was preceded by a run of 7 Us, consistent with a typical Rho-independent transcription terminator. The rest of the transcripts were processing intermediates, probably derived from the combined action of RNase P and RNase II, with 3′-ends upstream of the terminator.

The predicted secondary structure of the leuX 3′ trailer sequence that terminated at nucleotides 146–148 was determined using RNA-STAR program (31). It showed the features of a typical Rho-independent transcription terminator containing a seven nt stem-loop structure with a dyadic stem pairing high in guanine and cytosine residues followed by a uracil-rich stretch (Figure 4C). Since no transcript longer than the 148 nts was detected in any genetic backgrounds (Figures 1,2 and 4), we conclude that the 3′ secondary structure described in Figure 4C serves as an efficient Rho-independent transcription terminator in vivo.

In order to determine why the leuX Rho-independent transcription terminator is a substrate for PNPase while other tRNA associated Rho-independent transcription terminators are removed endonucleolytically by either RNase E or RNase G, we compared their relative strength (based on available free energy calculated with RNA-STAR), stem length and percent GC content. Surprisingly, the free energy associated with leuX transcription terminator was the lowest of all the terminators tested (Table 1). Furthermore, the lower energy level of the leuX terminator seemed to correlate with its shorter stem length rather than its net GC content (Table 1).

RNase E cleaves inefficiently at multiple locations within the 5′ leader region of leuX primary transcript when normal processing is impaired by the absence of either RNase P or PNPase

The fact that inactivation of both RNase P and RNase E led to the disappearance of the majority of the smaller pre-tRNA species observed in the rnpA49 single mutant and 2.5-fold increase in the 5′ unprocessed transcripts (Figure 1B, lanes 3–4, asterisk) suggested that RNase E cleaved inefficiently in the 5′ upstream region of leuX precursor tRNA. To provide further support for this hypothesis, primer extension analysis was carried out to identify the precise 5′-ends. Two distinct primer extension products (I and II, Figure 5) were obtained in the wild-type strain. The smaller of the products (II, G) represented the mature 5′-end of leuX, while the larger one (I, G) 22 nts upstream of the mature 5′ terminus was identified as the transcription initiation site of leuX primary transcript. This result was in agreement with the previous identification of the leuX transcription initiation site (19). Consistent with the northern analysis (Figure 1B), the primer extension products in wild-type and rne-1 strains were at identical locations, but the level of the primer extension product II (mature tRNA) was reduced significantly in the rnpA49 and rne-1 rnpA49 mutants (Figure 5). However, the level of the primer extension product I (primary transcripts) increased ∼4 ± 1-fold in the rnpA49 single and ∼6 ± 1-fold in the rne-1 rnpA49 double mutants compared to the wild-type control. Furthermore, several additional primer extension products upstream of the mature 5′ terminus (Figure 5, asterisk) were detected in the rnpA49 single mutant. Most of these termini either disappeared or were significantly reduced in the rne-1 rnpA49 double mutant.

Figure 5.

Primer extension analysis of the 5′ upstream region of leuX transcript. The reverse transcription products of total RNA (10 μg/lane) isolated from wild-type (MG1693), rne-1 (SK5665), rnpA49 (SK2525) and rne-1 rnpA49 (SK2534) strains using the primer b (Figure 1A) was separated on a 6% PAGE as described in the ‘Materials and Methods’ section. The CTAG sequencing reactions were carried out using the same primer and a leuX PCR DNA fragment as template. The transcription start site (I) and the 5′ mature terminus (II) of leuX transcript are indicated. The nucleotide sequences between transcription start (G) and the 5′ mature terminus (G) are shown to the left of the autoradiogram. Asterisk indicates the 5′-ends arising from RNase E cleavages.

Analysis of more than 250 5′- to 3′-end-ligated leuX transcripts derived from various genetic backgrounds (Figures 2 and 4) showed that many transcripts had truncated 5′-ends that were distributed between 2–16 nt downstream of the transcription start site (Figure 6). Consistent with northern and primer extension analysis (Figures 1 and 5), these 5′ truncated ends were more common in the rnpA49 and Δpnp strains (Figures 2 and 4), but were significantly reduced in an rne-1 genetic background (Figure 2). Although there were numerous minor cleavages, the major cleavage sites were at 9 and 15 nt downstream of the transcription start site, respectively (Figure 6).

Figure 6.

A schematic presentation of putative RNase E cleavage sites in the 5′ upstream region of leuX transcript. The sites were compiled from sequencing data of more than 250 independent clones from various genetic backgrounds. The thickness of each arrow indicates the relative abundance of each cleavage site.

The presence of so many closely spaced cleavage sites in the 5′ upstream region of the transcript (Figure 6) raised the question if RNase E was working exonucleolytically, similar to RNase J1 of B. subtilis (24). Even though we did not observe any in vivo cleavages after first nucleotide in the cloning and sequencing experiments (compiled in Figure 6), we carried out an in vitro kinetic analysis to determine if purified RNase E could degrade a uniformly ³²P-labeled leuX transcript that had exact 5′ (+1) end of the in vivo transcript (see ‘Materials and Methods’ section). The reaction mixtures were separated on 20% denaturing PAGE/7M urea gels in order to detect the release of any mononucleotides (data not shown). The leuX transcript was found to be a very poor substrate for RNase E, since it required 10–20-fold more protein and longer reaction times to observe cleavages compared to a good RNase E substrate (9S rRNA). The major cleavage products were 10, 6 and 4 nt in length and consistent with the in vivo findings, a 2 nt fragment was the smallest product that was detected on the gels (data not shown).

DISCUSSION

We describe here a unique PNPase mediated processing pathway for tRNA^Leu5 (Figure 7) that is distinct from all previously reported RNase E and RNase P dependent tRNA processing pathways (3–6). PNPase degrades the 3′ leuX trailer sequences that include the Rho-independent transcription terminator but stops 1–3 nt downstream of the CCA terminus (Figure 7). Alternatively, terminators from ∼10% of the primary transcripts are removed by RNase P cleavage at 4–7 nt downstream of the CCA terminus (Figure 7). RNase II cleaves the extra nucleotides at the 3′-end of the RNase P processed transcripts to within 1–3 nt of the CCA terminus. Finally, RNase T removes the last 1–3 nt to generate the mature 3′-end (Figure 7). All the pre-tRNAs, whether they are generated by PNPase or RNase P, have their 5′ termini matured by RNase P. Furthermore, terminator removal precedes 5′ processing.

Figure 7.

Processing pathways for tRNA^Leu5 in E. coli. PNPase initiates the processing by removing the 3′ trailer sequences that include the Rho-independent transcription terminator up to 1–3 nt downstream of the CCA terminus. Alternatively, RNase P removes the terminator of ∼10% of the transcripts by cleaving at 4–7 nt downstream of the CCA terminus. The RNase P cleavage intermediates are substrates for RNase II, which removes 3–4 additional nucleotides. Subsequently, the 5′-end is matured by RNase P and the 3′-end is matured, primarily by the activity of RNase T. However, RNase PH, RNase D, RNase BN and RNase II may substitute in the absence of RNase T (8).

There are a number of features that make this tRNA processing pathway very unusual. In the first place, since the Rho-independent terminators associated with tRNA transcripts are usually removed endonucleolytically, by either RNase E or RNase G (3–5), it was unexpected that the leuX terminator was degraded primarily by a 3′→5′ exonuclease, PNPase. However, the ∼10-fold increase in the steady-state level and an almost 10-fold increase in the half-life of the full-length transcript in the Δpnp strain clearly demonstrated the requirement of PNPase for terminator removal (Figures 1 and 3; Table 2). Furthermore, the cloning and sequencing of leuX species from a Δpnp strain confirmed the presence of significantly more transcripts with intact terminators compared to the wild-type control (Figures 2 and 4), while the data derived from the rnpA49 strains showed that the majority of the PNPase processed species contained an extra 3 nt downstream of the CCA (Figure 2). Furthermore, while the leuX terminator (Figure 4C) has a relatively low predicted free energy (Table 1), the sequencing of over 250 leuX clones (Figures 2 and 4) demonstrated that this structure functions very efficiently in vivo.

Thus, our data represent the first documented example of PNPase degrading through a region of double-stranded RNA in vivo. Although PNPase has been shown to be involved in the removal of Rho-dependent transcription terminators (6), its ability to degrade the leuX terminator without significant assistance from multiple rounds of polyadenylation of the substrate or through the unwinding activity of the RhlB RNA helicase was surprising because the enzyme is generally inhibited by secondary structures (9,32). However, since the thermodynamic stability of the 7 bp leuX stem-loop, which only contains four G/C base pairs (Figure 4C), is relatively low, there may not be a requirement for the unwinding activity of RhlB RNA helicase. In fact, our in vivo observations are consistent with previous in vitro data showing that stem-loops containing five G/C base pairs are not effective barriers against PNPase degradation (9). Furthermore, Blum et al. (29) demonstrated that PNPase could not efficiently degrade the lpp Rho-independent transcription terminator (containing a 13 bp stem that includes six G/C base pairs) in vitro unless a poly(A) tail longer than five nt was added. Thus, the need for 3′ polyadenylation may be minimized by the presence of 6–8 nt single-stranded region following the leuX stem-loop (Figure 4C).

Another interesting feature of this pathway relates to the requirement of RNase II (Figure 1C, lane 2). Many transcripts ended 4–7 nt downstream of the encoded CCA in the RNase II deficient strain (Figure 2) even though PNPase levels are increased 2–3-fold in this genetic background (33). Since PNPase had no difficulty in degrading through the stem-loop structure of the terminator, there was no reason to believe that it stopped at 4–7 nt downstream of the encoded CCA in the absence of RNase II, particularly since most of the immature 3′ termini in the RNase P deficient strains were mapped at 1–3 nt downstream of the CCA terminus (Figure 2). In contrast, a large drop in the RQ and PF of the mature tRNA and the dramatic increase in the half-life of the primary transcript in the Δpnp rnpA49 strain compared to the Δpnp single mutant and wild-type control (Figures 1D,3; data not shown) suggested that the RNase P cleavages at 3′ the ends, left a single-stranded region (∼10-11 nt), sufficient for RNase II (34) to bind and remove the next 3–4 nt, but were not substrates for PNPase.

It should also be noted that a previous in vitro study mapped a 3′ RNase P cleavage site at 7 nt downstream of CCA (Figure 4C) (30). Thus, it is apparent that the longer processing intermediates in the RNase II deficient strain (Figure 1C) were in fact processed by RNase P, although not as precisely as in vitro. While the exact mechanism of how RNase P recognizes cleavage sites in the 3′ trailer sequences is not known, one possible explanation is that a structure different from that shown in Figure 4C forms in a limited number of leuX transcripts providing access to RNase P (35). It is not clear if the same RNase P molecule cleaves both the 3′- and 5′-ends of a single transcript. Furthermore, while no other exoribonuclease could replace RNase II in these experiments, it has been shown that all MG1655 derivatives are deficient in RNase PH (36). The final 3′ maturation of tRNA^Leu5 still requires RNase T, since more than 70% of the precursor tRNA were 1–3 nt longer than the mature tRNA in the Δrnt strain (Figure 1G, Table 1).

The detection of some leuX primary transcripts in a wild-type strain [Figure 1, (19)] suggests less efficient processing by PNPase compared to terminator removal by either RNase E or RNase G, since primary transcripts processed endonucleolytically are not readily detectable in wild-type strains. In fact, this may be the rate-limiting step during tRNA^Leu5 processing because 3′-end processing appears to be necessary before RNase P cleavage at the 5′-end (Figure 1). Thus, in the absence of PNPase the major unprocessed species is the full-length leuX transcript (Figure 1D, lane 2) with only a minor fraction (<1%) of the processing intermediates having a mature 5′ terminus containing the 3′ terminator sequences. While this result is in contrast to the in vitro study of leuX processing (19), it is consistent with inhibition of processing of both the secG leuU and metT operons by RNase P prior to endonucleolytic removal of their respective Rho-independent transcription terminators in vivo (5). In contrast, RNase P processing of tRNA transcripts is not inhibited by Rho-dependent transcription terminators (6).

Finally, it should also be noted that previous experiments have shown that RNase E can inefficiently cleave at a limited number of locations within the 5′ upstream region of tRNA transcripts dependent on RNase P for processing (6). Thus, it was not entirely surprising that RNase E inefficiently cleaved the 5′ upstream region of leuX transcripts when RNase P was inactivated (Figures 1B and 2). However, what was different in the case of leuX was the large number of cleavage sites (Figures 1, 5 and 6) such that it initially appeared as if the enzyme possessed a 5′→3′ exonucleolytic activity similar to what has been observed with RNase J from B. subtilis (24). However, a careful combination of the sequencing data (Figures 2 and 4) and in vitro analysis of RNase E activity on the full-length leuX transcript, demonstrated that this activity is indeed endonucleolytic. In addition, the absence of a role for RppH in tRNA^Leu5 maturation (data not shown), even though the transcript was a substrate for the enzyme in vivo, was consistent with no involvement of the 5′-end-dependent endoribonucleases (RNase E, RNase G) during processing.

Taken together, we have identified a novel tRNA processing pathway that involves primary removal of a Rho-independent transcription terminator by the 3′→5′ exonuclease PNPase. Furthermore, this is the first in vivo demonstration that RNase P acts at both the 5′- and 3′-ends of transcripts during tRNA maturation. Surprisingly, when we examined the sequences downstream of the CCA determinant of the Rho-independent transcription terminators presented in Table 1, there was no obvious reason as to why the leuX terminator was not a substrate for RNase E. Even though most of the terminators known to be cleaved by RNase E (Table 1, bold type) primarily contain As and Us downstream of the CCA determinant, there are three Gs in the leuZ sequence. It thus seems possible that the recognition of individual terminators involves more than primary sequences. If this hypothesis is correct, there could be additional Rho-independent transcription terminator sequences that are not substrates for endonucleolytic cleavage. Taken together, the tRNA processing pathways in E. coli are far more diverse than previously envisioned.

FUNDING

National Institutes of General Medical Sciences (GM75220 and GM81554 to S.R.K.). Funding for open access charge: National Institutes of Health grant.

Conflict of interest statement. None declared.