Autogenous Regulation of Escherichia coli Polynucleotide Phosphorylase Expression Revisited

Abstract

The Escherichia coli polynucleotide phosphorylase (PNPase; encoded by pnp), a phosphorolytic exoribonuclease, posttranscriptionally regulates its own expression at the level of mRNA stability and translation. Its primary transcript is very efficiently processed by RNase III, an endonuclease that makes a staggered double-strand cleavage about in the middle of a long stem-loop in the 5′-untranslated region. The processed pnp mRNA is then rapidly degraded in a PNPase-dependent manner. Two non-mutually exclusive models have been proposed to explain PNPase autogenous regulation. The earlier one suggested that PNPase impedes translation of the RNase III-processed pnp mRNA, thus exposing the transcript to degradative pathways. More recently, this has been replaced by the current model, which maintains that PNPase would simply degrade the promoter proximal small RNA generated by the RNase III endonucleolytic cleavage, thus destroying the double-stranded structure at the 5′ end that otherwise stabilizes the pnp mRNA. In our opinion, however, the first model was not completely ruled out. Moreover, the RNA decay pathway acting upon the pnp mRNA after disruption of the 5′ double-stranded structure remained to be determined. Here we provide additional support to the current model and show that the RNase III-processed pnp mRNA devoid of the double-stranded structure at its 5′ end is not translatable and is degraded by RNase E in a PNPase-independent manner. Thus, the role of PNPase in autoregulation is simply to remove, in concert with RNase III, the 5′ fragment of the cleaved structure that both allows translation and prevents the RNase E-mediated PNPase-independent degradation of the pnp transcript.

Posttranscriptional regulation at the level of mRNA stability and translatability is acknowledged as an important step in gene expression control (for reviews, see references 8 and 25). An interesting system for the study of the interplay between mRNA translation and decay may be the regulation of pnp gene expression. The pnp gene encodes polynucleotide phosphorylase (PNPase), one of the major players in Escherichia coli mRNA degradation. The enzyme is widely conserved among members of the class Bacteria, and homologous genes have been found in plants and human cells (reviewed in references 3, 29, and 51). In E. coli, PNPase is a homotrimeric protein of a 711-amino-acid polypeptide with a structural core, which contains the catalytic domain, and two C-terminal RNA-binding domains, KH and S1 (5, 38). The three subunits associate via trimerization interfaces of the core domain, forming a central channel (55). In vitro, PNPase catalyzes the processive 3′-to-5′ phosphorolytic degradation of RNA, the reverse polymerization reaction, and the exchange reaction between free phosphate and the β-phosphate of ribonucleoside diphosphates (7, 19, 59). PNPase can also bind RNA via its two RNA-binding domains (34, 47, 54). In vivo, the enzyme has been shown to be implicated in both degradation and polyadenylation of RNAs (13, 36, 37) and may be found as a component of a multiprotein machine, the RNA degradosome, together with the endonuclease RNase E, the DEAD-box RNA helicase RhlB, and enolase (6, 35, 42).

The pnp gene is located downstream of rpsO, which codes for the ribosomal protein S15 (40). Transcription of pnp can originate from two different promoters, rpsOp (also referred to as P1), upstream of rpsO, and pnp-p (P2), located between rpsO and pnp (41). Both primary transcripts are processed by RNase III, which cleaves the RNAs 37 (site 1) and 75 (site 2) nucleotides downstream of the pnp-p transcription start point within a long hairpin structure in the 5′-untranslated region (5′-UTR) (39, 43, 46, 58). This originates a monocistronic mRNA 2.25 kb long, which extends from RNase III site 2 to the pnp-t Rho-independent terminator, and longer transcripts terminating downstream (39, 57, 62). Transcripts originating at rpsOp are cleaved, in addition, by RNase E in the rpsO-pnp intergenic region, and the RpsO-encoding mRNAs undergo independent processing and degradation (21, 44-46; see Fig. S1 in the supplemental material).

PNPase posttranscriptionally regulates its own expression by controlling the stability of the RNase III-processed mRNA, as the processed transcript, in the presence of PNPase, becomes very unstable (48, 49). Two models have been proposed by C. Portier and collaborators to mechanistically explain PNPase autogenous regulation. The former one, proposed in 1994 (49), postulates that PNPase can bind determinants in the 5′-UTR of pnp mRNA; this would inhibit translation, thus promoting degradation of the transcript (39, 48, 49). The current model proposed in 2001 maintains that the duplex formed by the 37-nucleotide (nt)-long RNA (RNA37) upstream of the proximal RNase III cleavage site 1 and the new 5′-monophosphate end of the mature pnp mRNA protects the transcript from degradation. PNPase would trigger pnp mRNA decay by degrading the small RNA37 from the short 3′ overhang (23).

In our view, the first model has not been completely ruled out by the second one, as the issue of translational repression was not further addressed by Jarrige et al. (23). It thus remains possible that these two non-mutually exclusive mechanisms operate in concert to autoregulate PNPase expression.

In this work, through the analysis of mutants in the leader region, we provide further support to the 2001 model and provide evidence that the putative translational control by PNPase does not play a role in autoregulation. Moreover, we show that the native pnp mRNA with its long 5′-UTR is stable and translatable, whereas the processed transcript devoid of its 5′ double-strand structure is poorly, if at all, translated and is rapidly degraded by RNase E, the major endoribonuclease implicated in the degradation of mRNAs.

MATERIALS AND METHODS

Bacterial strains and plasmids.

The bacteria and plasmids are listed in Tables 1 and 2, respectively. The mutant pnp alleles created in this work (Table 3) were obtained either directly by PCR or by conventional in vitro recombination in an intermediate plasmid as detailed below. A linear mutant DNA fragment (either as a purified PCR product or recovered from the intermediate plasmid) was transferred by electroporation into the recipient strain harboring pKD46, a replication-thermosensitive plasmid expressing bla (Amp^r) and the λ Red recombination system, as described previously (9). Coordinates of the pnp locus are given, throughout, relative to the pnp-p transcription start point (23), conventionally defined as +1, corresponding to coordinate G7947 in NCBI accession no. ECAE000397 and to coordinate 3309346 in the complete E. coli sequence (NCBI accession no. U00096.2) on the complementary strand. pGEX104, the parental plasmid in which all the mutations in the pnp 5′ and 3′ regions have been constructed, harbors a DNA fragment delimited by EcoRI-KpnI that contains the pnp regions −201 to +249 and 2239 to 2589 separated by a cat cassette (Δpnp-871::cat allele). In Δpnp-871::cat, an amplicon obtained by PCR of plasmid pKD13 with primers FG1701 and FG1702 replaces the 248-2238 pnp region. The mutations in the 5′ and 3′ regions have been obtained by replacing, respectively, the EcoRI-DraIII and XbaI-KpnI regions of the pGEX104 pnp insert with equivalent DNA fragments harboring the specific mutations obtained by three-step PCR (30) using appropriate oligonucleotides (Table 4). Construction of the primary single and double pnp mutants is outlined in Fig. 1. To obtain the 5′-end mutants in both the leader and the coding regions, C-5705/pKD46, which contains the rpsOt::aph (Kan^r) allele (an insertion of the aph cassette at −60) was transformed with the entire EcoRI-KpnI mutant inserts. Cam^r transformants were screened for the loss of the rpsOt::aph marker. FLP-mediated excision of resistance cassettes from the mutants transformed with pCP20 was performed as described previously (9). The deletion of the pnp-p terminator was obtained by selecting Cam^r recombinants of C-1a/pKD46 transformed with the pGEX1010 insert, which harbors the Δpnp-1010t deletion (positions 2309 to 2328) of the pnp Rho-independent terminator obtained by three-step PCR (30) using oligonucleotides FG1802 and FG1803. Cam^r transformants were screened by PCR to detect pnp-t deletion. The presence of each mutation was confirmed by sequencing suitable DNA fragments obtained by PCR.

TABLE 1.

Genotypes of the bacterial strains used in this study

Strain	Genotype	Source or reference
BZ31 zce-726::Tn10	his ΔtrpE5 mukB106 smbB-rne-131 zce-726::Tn10 (λ)	Kindly provided by M. Dreyfus
C-1a	E. coli C, prototrophic	52
C-5684	Δrnc38::kan	By P1*SK7622 transduction into C-1a (this work)
C-5686	Δrne-131∼Tn10	By P1*BZ31 zce-726::Tn10 transduction into C-1a (this work)
C-5705	rpsOt::aph	From C-1a/pKD46 by λ Red-mediated recombination with DNA fragment obtained by PCR of pKD13 with primers FG1701 and FG1702; aph cassette inserted at −60 from pnp-p into rpsO transcription terminator (this work)
C-5729	Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX104 insert (this work)
C-5731	Δpnp-871	From C-5729/pCP20 by FLP-mediated excision of cat cassette (this work)
C-5801	ΔpnpL1001 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1001 insert (this work)
C-5802	ΔpnpL1002 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1002 insert (this work)
C-5803	ΔpnpL1003 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1003 insert (this work)
C-5804	ΔpnpL1004 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1004 insert (this work)
C-5805	ΔpnpL1005 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1005 insert (this work)
C-5806	ΔpnpL1006 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1006 insert (this work)
C-5807	pnpL1007 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1007 insert (this work)
C-5808	pnpL1008 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1008 insert (this work)
C-5809	pnpL1009 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1009 insert (this work)
C-5810	Δpnp-1010t Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1010 insert (this work)
C-5811	ΔpnpL1011 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1011 insert (this work)
C-5812	ΔpnpL1012 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1012 insert (this work)
C-5813	ΔpnpL1013 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1013 insert (this work)
C-5814	ΔpnpL1014 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1014 insert (this work)
C-5815	ΔpnpL1015 Δpnp-871::cat	From C-5705/pKD46 by λ Red-mediated recombination with pGEX1015 insert (this work)
C-5851	ΔpnpL1001 Δpnp-871	From C-5801/pCP20 by FLP-mediated excision (this work)
C-5852	ΔpnpL1002 Δpnp-871	From C-5802/pCP20 by FLP-mediated excision (this work)
C-5853	ΔpnpL1003 Δpnp-871	From C-5803/pCP20 by FLP-mediated excision (this work)
C-5854	ΔpnpL1004 Δpnp-871	From C-5804/pCP20 by FLP-mediated excision (this work)
C-5855	ΔpnpL1005 Δpnp-871	From C-5805/pCP20 by FLP-mediated excision (this work)
C-5856	ΔpnpL1006 Δpnp-871	From C-5806/pCP20 by FLP-mediated excision (this work)
C-5857	pnpL1007 Δpnp-871	From C-5807/pCP20 by FLP-mediated excision (this work)
C-5858	pnpL1008 Δpnp-871	From C-5808/pCP20 by FLP-mediated excision (this work)
C-5859	pnpL1009 Δpnp-871	From C-5809/pCP20 by FLP-mediated excision (this work)
C-5860	Δpnp-1010t Δpnp-871	From C-5810/pCP20 by FLP-mediated excision (this work)
C-5861	ΔpnpL1011 Δpnp-871	From C-5811/pCP20 by FLP-mediated excision (this work)
C-5862	ΔpnpL1012 Δpnp-871	From C-5812/pCP20 by FLP-mediated excision (this work)
C-5863	ΔpnpL1013 Δpnp-871	From C-5813/pCP20 by FLP-mediated excision (this work)
C-5864	ΔpnpL1014 Δpnp-871	From C-5814/pCP20 by FLP-mediated excision (this work)
C-5865	ΔpnpL1015 Δpnp-871	From C-5815/pCP20 by FLP-mediated excision (this work)
C-5868	rne-3071∼Tn10	By P1*KG241 transduction into C-1a, Tetr selection, screening for TS; presence of rne-3071 confirmed by sequencing
C-5869	rne+∼Tn10	By P1*KG241 transduction into C-1a, Tetr selection, screening for TS+; presence of rne+ confirmed by sequencing
C-5870	Δpnp-871 rne-3071∼Tn10	By P1*KG241 transduction into C-5731, Tetr selection, screening for TS; presence of rne-3071 confirmed by sequencing
C-5871	Δpnp-871 Tn10	By P1*KG241 transduction into C-5731, Tetr selection, screening for TS+; presence of rne+ confirmed by sequencing
DH10B	F−mcrA Δ(mrr-hsdRMS-mcrBC) φ80dlacZΔM15 ΔlacX74 deoR recA1 araD139 Δ(ara-leu)7697 galU galK rpsL endA1 nupG	32
KG241	rne-3071∼Tn10	By P1*BZ31 into N3431, Tetr selection, screening for TS; presence of rne-3071 confirmed by sequencing
MRE600	rna	60
N-3431	lacZ43 λ−relA1 spoT1 rne-3071	18
N-3433	lacZ43 λ−relA1 spoT1 thi-1	18
SK7622	thyA715 rph-1 rncΔ38::kan	1 (kindly provided by S. Kushner)

TABLE 2.

Characteristics of the plasmids used in this study

Plasmid	Relevant characteristics	Reference
pAZ76	pGM743 derivative; harbors 1-380 pnp operon region under T7 promoter	This work
pAZ77	pGM743 derivative; harbors 76-380 pnp operon region under T7 promoter	This work
pAZ101	pGZ119HE derivative; pnp+	47
pAZ133	pAZ101 derivative; harbors Δpnp-833 allele that encodes KHS1-truncated PNPase	4
pAZ133L1001	Harbors ΔpnpL1001 Δpnp-833 double mutation	This work
pAZ133L1002	Harbors ΔpnpL1002 Δpnp-833 double mutation	This work
pCP20	Thermosensitive replication, thermoinducible FLP (FRT-specific) recombinase; Ampr Camr	9
pGEX-1	Cloning vector; Ampr	53
pGEX104	pGEX-1 derivative; harbors pnpL+ Δpnp-871::cat	This work
pGEX1001	pGEX104 derivative; harbors ΔpnpL1001 Δpnp-871::cat	This work
pGEX1002	pGEX104 derivative; harbors ΔpnpL1002 Δpnp-871::cat	This work
pGEX1003	pGEX104 derivative; harbors ΔpnpL1003 Δpnp-871::cat	This work
pGEX1004	pGEX104 derivative; harbors ΔpnpL1004 Δpnp-871::cat	This work
pGEX1005	pGEX104 derivative; harbors Δpnp-1005 Δpnp-871::cat	This work
pGEX1006	pGEX104 derivative; harbors Δpnp-1006 Δpnp-871::cat	This work
pGEX1007	pGEX104 derivative; harbors pnp-1007 Δpnp-871::cat	This work
pGEX1008	pGEX104 derivative; harbors pnp-1008 Δpnp-871::cat	This work
pGEX1009	pGEX104 derivative; harbors pnp-1009 Δpnp-871::cat	This work
pGEX1010	pGEX104 derivative; harbors Δpnp-1010t Δpnp-871::cat	This work
pGEX1011	pGEX104 derivative; harbors ΔpnpL1011 Δpnp-871::cat	This work
pGEX1012	pGEX104 derivative; harbors ΔpnpL1012 Δpnp-871::cat	This work
pGEX1013	pGEX104 derivative; harbors ΔpnpL1013 Δpnp-871::cat	This work
pGEX1014	pGEX104 derivative; harbors ΔpnpL1014 Δpnp-871::cat	This work
pGEX1015	pGEX104 derivative; harbors ΔpnpL1015 Δpnp-871::cat	This work
pGM743	Promoterless pGZ119EH derivative	62
pGZ119HE	oriVColD Camr	28
pKD13	oriRγ bla; harbors the aph (Camr) cassette flanked by FRT sites	9
pKD46	oriR101 repA101(Ts) araC araBp-gam-bet-exo Ampr	9

TABLE 3.

New pnp alleles constructed in this work

Allele	Description
Δpnp-871	Deletion 250-2238, in-frame insertion of 78-nt-long SCAR871 sequence
Δpnp-871::cat	Deletion 250-2238, insertion of cat cassettea
ΔpnpL1001	Deletion 38-73 (5′-UTR)
ΔpnpL1002	Deletion 1-75 (5′-UTR)
ΔpnpL1003	Deletion 1-108 (5′-UTR)
ΔpnpL1004	Deletion 1-127 (5′-UTR)
Δpnp-1005	Deletion 162-167 (within 5′-coding sequence)
Δpnp-1006	Deletion 178-180 (within 5′-coding sequence)
pnpL1007	Transversion GG 147-148 CC (Shine-Dalgarno)
pnp-1008	Transversion G159T (start codon)
pnp-1009	Transversion T157A (start codon)
Δpnp-1010t	Deletion 2309-2328 (Rho-independent terminator)
ΔpnpL1011	Deletion 4-75 (5′-UTR)
ΔpnpL1012	Deletion 4-108 (5′-UTR)
ΔpnpL1013	Deletion 114-145 (5′-UTR)
ΔpnpL1014	Deletion 125-145 (5′-UTR)
ΔpnpL1015	Deletion 132-145 (5′-UTR)
rpsOt::aph	Insertion of aph cassettea at −60

^aSee the text.

TABLE 4.

Oligonucleotides used in this study

Oligonucleotide no.	5′→3′ sequence	Coordinatesa
FG1080/38b	CTAATACGACTCACTATAGGGGTCACGGTGTGTTGGCC	206-190
FG1081/20	GTCGCGAGGATGCGCAGAAG	79-98
FG1626/40b	CTAATACGACTCACTATAGGGCATTAGCCGCGCGAACCTC	37-19
FG1627/21	GAATGATCTTCCGTTGCAGAG	1-21
FG1682/41b	CTAATACGACTCACTATAGGGGTAGGGTGGAGCTGCTTCGA	33-52
FG1683/20	ATTCCGGGGATCCGTCGACC	1315-1334
FG1701/59c	CTGCGTCGCTAATTCTTGCGAGTTTCAGAAAAGGGGCCTGGCTTTTGAAGCTCACGCTG	−101-61/ 1267-1249
FG1702/60c	CCAGTGAATTGCTGCCGTCAGCTTGAAAAAAGGGGCCACTCAAGATCCCCTTATTAGAAG	−21-60/98-117
FG1802/28	CATTTGCCCTTAACCGGGCAGGACGCCT	2299-2308/ 2329-2346
FG1803/30	CCTGCCCGGTTAAGGGCAAATGGCAACCTT	2340-2329/ 2308-2291

^aCoordinates are from pnp-P +1 if not otherwise indicated.

^bBoldface letters indicate T7 promoter sequence.

^cBoldface letters and coordinates are from the pKD13 plasmid.

FIG. 1.

Construction of mutants in the pnp 5′ end. C-5705 (upper drawing) harboring an aph (Kan^r) cassette and the plasmid pKD46, which expresses the λ Red system, was transformed with a DNA fragment (second drawing) harboring a mutant pnp 5′-end (pnpL^mut, depicted by an x) and the Δpnp-871::cat allele. After λ Red-mediated recombination, chloramphenicol-resistant clones were selected and screened for the loss of Kan^r and pKD46. Such clones, putatively pnpL Δpnp-871::cat (third drawing), were transformed at 30°C with the thermosensitive plasmid pCP20, which confers ampicillin resistance and harbors a thermoinducible FLP recombinase. After growth at 42°C without antibiotic selection, PCR amplicons of Cam^s Amp^s clones (bottom drawing) were then sequenced to verify the presence of the desired pnpL Δpnp-871 mutations.

Bacterial strains were grown in LD broth (17) supplemented with chloramphenicol (30 μg/ml), ampicillin (100 μg/ml), kanamycin (50 μg/ml), 0.4% glucose, or 1% l-arabinose when needed.

Northern blotting and data processing.

Bacterial cultures were grown at 37°C in LD broth supplemented with antibiotics when needed and grown up to an optical density at 600 nm (OD₆₀₀) of 0.8, if not otherwise indicated. For RNA decay analysis, rifampin and nalidixic acid (final concentrations of 800 μg/ml and 60 μg/ml, respectively) were added and samples for RNA extraction were taken at different time points. RNA extraction was performed using the RNeasy (or miRNeasy, for small RNA analysis) minikit (Qiagen). Basic procedures for Northern blot analysis and synthesis of radiolabeled riboprobes by in vitro transcription with T7 RNA polymerase were previously described (4, 12). The specific riboprobes were obtained by in vitro transcription with T7 RNA polymerase of DNA fragments obtained by PCR amplification as follows. The template for the SCAR871 riboprobe, complementary to the Δpnp-871 scar, was obtained by PCR amplification of genomic C-5731 DNA with primers FG1682 and FG1683. The template for ANTI37, complementary to the small RNA37, was obtained by PCR of pAZ101 with primers FG1626 and FG1627. The template for the pnp mRNA riboprobe (PNP5, complementary to 79-206) was obtained by PCR of pAZ101 with primers FG1080 and FG1081.

Autoradiographic images and densitometric analysis of Northern blots were obtained by phosphorimaging using ImageQuant software (Molecular Dynamics). The intensities of pnp mRNA signals were normalized to the intensity of the pnpL⁺ signal in the absence of PNPase (relative abundance, as reported in Table 5 and in the histogram of Fig. 4). The ratio of the abundance in the presence of PNPase to the abundance in its absence thus represents the fraction of pnp mRNA escaping autogenous regulation (fraction escaping PNPase [FEP]), whereas (1-FEP) is the mRNA fraction subject to autoregulation in each strain (raw autogenous regulation index [RAI]). A normalized autogenous regulation index (AI) was then calculated as the ratio of the RAI of each mutant to the RAI obtained in pnpL⁺ signal. In experiments of mRNA decay analysis, mRNA half-lives were estimated by regression analysis of curves obtained by plotting the percentage of mRNA remaining (calculated as the densitometric signal at a given time after rifampin addition divided by the signal at time 0) versus time after rifampin addition.

TABLE 5.

pnp mRNA abundance and stability in the presence and absence of PNPase

Mutationa	Coordinate		Relative abundanceb		AIc	Half-life (min)
	Start	End	−PNPase	+PNPase		−PNPase	+PNPase
None (pnpL+)	0	0	1.00	0.06	1.00	4	NDd
Δ1001	38	73	1.42	0.81	0.46	5	8
Δ1002	1	75	0.15	0.11	0.28	ND	ND
Δ1011	4	75	0.24	0.21	0.12	ND	ND
Δ1003	1	108	0.57	0.37	0.37	4	8
Δ1012	4	108	0.54	0.52	0.04	ND	ND
Δ1004	1	127	0.37	0.39	−0.06	ND	ND
Δ1013	114	145	0.24	0.05	0.84	5	ND
Δ1014	125	145	0.20	0.04	0.85	ND	ND
Δ1015	132	145	0.20	0.04	0.86	ND	ND
Δ1005	162	167	0.58	0.20	0.70	ND	ND
Δ1006	178	180	0.68	0.18	0.79	ND	ND
Δ1010	2309	2328	0.90	0.13	0.92	ND	ND
1007	147	148	0.53	0.08	0.92	ND	ND
1008	159	159	0.22	0.07	0.74	2	ND
1009	157	157	1.08	0.17	0.90	7	ND

^aMutations are listed in the same order as in Fig. 2.

^bRatio to pnpL⁺ in the absence of PNPase.

^cNormalized AI.

^dND, not determined.

FIG. 4.

Relative abundance and stability of Δpnp-871 mRNA. Relative abundance and stability were estimated by phosphorimaging quantification of Northern blot signals as described in Materials and Methods. The intensity of all discrete signals has been summed up to evaluate the total abundance of Δpnp-871 scar-containing transcript. Data have been normalized to the signals obtained in the pnpL⁺ strain in the absence of PNPase. Data from the pnpL⁺ strain in the presence of PNPase are the average of four experiments, and the error bar is shown. The normalized AI (see text) and half-lives (t_1/2), when evaluated, are also reported. nd, not determined. Cross-hatched bars, strains harboring pAZ101 expressing PNPase; open bars, strains harboring the empty vector pGZ119HE. In the wild-type C-1a strain (with pnp in a single copy), the half-life is about 1 min (62; unpublished observation).

Preparation of E. coli crude extracts, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) of proteins and immunoassays.

E. coli crude extracts were obtained as described previously (47). Protein content was determined using Coomassie Plus protein assay reagent (Pierce). Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was performed as described previously (26) on 10% resolving gels containing 0.1% SDS. High-molecular-weight markers (Amersham Biosciences, England) were used as size references. For immunological detection of PNPase, the gels were blotted onto a nitrocellulose (Hybond ECL) sheet and incubated with polyclonal anti-PNPase antibodies (15, 50). Immunoreactive bands were revealed using the ECL Western blotting reagent (Amersham GE Healthcare).

In vitro translation assay.

Escherichia coli MRE600 cells were grown in LB medium supplemented with 0.5% glucose at 37°C up to an OD₆₀₀ of 1.2, washed with 0.9% NaCl, and harvested immediately. Cell-free (S30) extracts were prepared as described previously (27), except that extracts were prepared in 10 mM Tris-HCl (pH 7.1), containing 10 mM Mg acetate, 60 mM NH₄Cl, and 6 mM β-mercaptoethanol and the preincubation step was omitted.

In vitro translation assays were carried out using in vitro-transcribed pnp mRNAs and S30 extracts (normalized for their ribosome content) in a mixture containing 28 mM Tris-HCl (pH 7.7), 10 mM Mg acetate, 50 mM NH₄Cl, 2 mM dithiothreitol, 2 mM ATP, 0.4 mM GTP, 10 mM phosphoenolpyruvate, 0.025 μg of pyruvate kinase/μl reaction mixture, 200 mM of each amino acid (excepting Met), 0.1 μM [³⁵S]Met, 1 μg/μl E. coli MRE600 tRNA mixture, and 0.12 mM citrovorum (Serva). After 30 min of incubation at 37°C, samples were spotted onto 3MM paper discs in order to establish the trichloroacetic acid-insoluble radioactivity incorporated.

Full-length and 5′-deletion pnp mRNAs were obtained by in vitro runoff transcription of 80 μg/ml SmaI-linearized pAZ76 and pAZ77, respectively, 2,000 U of T7 RNA polymerase in a transcription mixture containing 40 mM Tris-HCl (pH 8.1), 3.75 mM NTPs, 22 mM MgCl₂, 10 mM spermidine, 5 mM dithiothreitol, 50 μg bovine serum albumin, and 60 U RNAsin [Amersham]. After 4 h of incubation at 37°C, the mRNA was purified by LiCl precipitation followed by ethanol precipitation.

RESULTS

cis-acting elements controlling PNPase autogenous regulation: experimental plan.

To identify specific cis-acting determinants of PNPase autoregulation and discriminate between the two proposed models, we constructed several pnpL Δpnp-871 mutants and one Δpnp-1010t Δpnp-871 chromosomal double mutant. In the Δpnp-871 mutants, a 78-nt-long scar sequence replaces in frame a 1,989-bp pnp region from codons 83 to 743. Expression of the Δpnp-871 mutant allele can thus be specifically monitored by Northern blotting even in the presence of an ectopically expressed pnp⁺ gene using a riboprobe (SCAR871) complementary to the unique scar sequence. Δpnp-1010t is a deletion of the Rho-independent transcription terminator immediately downstream of pnp, whereas the pnpL mutants harbor deletions or point mutations in the leader (either 5′-UTR or 5′ coding) region of Δpnp-871.

The general strategy for the 5′ mutant construction is described in Materials and Methods and outlined in Fig. 1. The pnp double mutant strains (Fig. 2 and Table 1) were then transformed with either pAZ101, which expresses PNPase from its natural promoter, pnp-p, or the parental plasmid pGZ119HE. The abundance of the chromosomally encoded Δpnp-871 mRNA in the different mutants in the presence and absence of ectopically provided PNPase was then monitored by Northern blotting using a radiolabeled SCAR871 riboprobe.

FIG. 2.

Map of the pnpL Δpnp-871 mutations. The pnp 5′- and 3′-UTRs are drawn to scale; the Δpnp-871 allele, which is 225 bp long, is on an arbitrary scale. Coordinates shown above the vertical bars are nucleotides from the pnp-p transcription start point (bent arrow). The coordinates of pnp codons retained in the mutant allele are shown below the upright arrowheads. Inverted arrows indicate the large and the small stem-loops (SL1and SL2, respectively). RIII₁ and RIII₂ are the two RNase III cut points. Checkerboard region, in-frame 78-nt-long scar (SCAR871) in Δpnp-871 allele; lollipop, pnp-t Rho-independent terminator. The Shine-Dalgarno and start codon sequences are also shown. The bars indicate the extent of the deletions, and boldface letters represent the bases substituted in the different mutants.

An example of such analyses is shown in Fig. 3A (see Fig. S2 in the supplemental material). The intensity of pnp mRNA signals was measured by phosphorimaging, and the results were normalized to the intensity of the pnpL⁺ signal in the absence of PNPase (Table 5 and Fig. 4, histogram). A normalized AI was calculated as described in Materials and Methods. We assumed that AI values of ≥0.7 and <0.5 indicate autoregulation and lack thereof, respectively (see below).

FIG. 3.

An example of Northern blot analysis of abundance and stability of Δpnp-871 mRNA with different pnpL mutations. RNA extracted from E. coli strains harboring the pnpL Δpnp-871 mutations in the presence or absence of ectopically expressed PNPase, as indicated in the figure, was analyzed by Northern blotting using a radiolabeled SCAR871 riboprobe. (A) Abundance of RNA extracted from cultures grown up to OD₆₀₀ of 0.8. (B) Stability of RNA extracted from cultures grown to an OD₆₀₀ of 0.8 and treated with rifampin (rif) as described in Materials and Methods. Time points (min) after rifampin addition and the half-lives (t_1/2) estimated by phosphorimaging quantification of the signals are indicated. The lengths of the pnpL⁺ Δpnp-871 transcripts (which are 1,911 bp shorter than their corresponding pnp⁺ mRNAs; see Fig. S1 in the supplemental material for correspondence) are indicated in kb. The signals from the mutants are in fair agreement with the mRNAs expected from the harbored deletions. Moreover, mutants affected in RNase III processing also show some minor signals deriving from rpsOp. This complex signaling profile has been verified in part by hybridization with specific radiolabeled DNA oligonucleotide probes and an rpsO-specific riboprobe. The intensity of all discrete signals has been summed up to evaluate the total abundance of Δpnp-871 scar-containing transcript. The greater stability of the lowest band in panel B could be imputed to processing of the higher-molecular-weight transcripts.

To measure mRNA half-life (Table 5, last two columns), rifampin was added to cultures so as to inhibit transcription initiation and RNA was extracted at different time points for Northern blot analysis (see Fig. 3B as an example and see Fig. S3 in the supplemental material). The signal intensities of pnp mRNA remaining were measured and normalized to the intensity detected before transcription inhibition. Half-lives could not be estimated under conditions of very low transcript abundance.

A processable pnp mRNA double-stranded 5′ stem is the unique determinant of PNPase autoregulation.

Deletion ΔpnpL1001, which removes the upper (central) part of the large stem-loop (SL1) that serves as a substrate for RNase III, abolishes the endonucleolytic processing, as assessed by Northern blotting (data not shown). The primary transcript conserves a stem-loop of the same length as the double-stranded stem generated by RNase III digestion. The former, however, cannot be degraded by PNPase as it lacks the dangling 3′ end. As shown in Table 5 and Fig. 4, the abundance and stability of the pnp transcript in the presence of PNPase were over 10-fold higher than those of the wild type, thus indicating that in the ΔpnpL1001 mutant autoregulation was abolished. Although the AI of this mutant was relatively high (0.46), in the presence of PNPase its mRNA was greatly stabilized with a half-life of 8 min. These data are in agreement with previous results with RNase III-deficient strains or point mutations that prevent RNase III processing of the stem-loop (23, 49) and show that a double-stranded stem-loop as short as the stem produced by RNase III processing is sufficient for pnp mRNA stabilization.

Deletions ΔpnpL1002 and ΔpnpL1011 completely remove the large stem-loop. In the ΔpnpL1002 mutant, the transcript 5′-end corresponds to that of the RNase III-processed pnp mRNA (starting with a U), whereas the ΔpnpL1008 mutant conserves three nucleotides (GAA) at the natural pnp-p transcription start point so as to minimize effects of the deletion on transcription initiation efficiency. The abundance of pnp mRNA in these mutants was substantially decreased in both the presence and the absence of PNPase to a level comparable to that of the wild type in the presence of PNPase, and the AI was low. Thus, removal of SL1 generates a transcript that is unstable independently of PNPase. Similar results were obtained with mutants harboring longer deletions (ΔpnpL1003, ΔpnpL1012, and ΔpnpL1004) of the pnp 5′-UTR. Although the abundance of the pnp mRNA expressed from these latter mutants was slightly higher than that of the ΔpnpL1002 and ΔpnpL1008 mutants, it did not decrease in the presence of PNPase, giving low autoregulation indexes. Thus, the instability of pnp transcripts missing the double-stranded region at the 5′-UTR appears to be independent of PNPase autogenous control.

Other mutations (ΔpnpL1013, -1014, and -1015) that removed the small stem-loop (SL2) in the pnp 5′-UTR and/or the region immediately upstream of the Shine-Dalgarno sequence decreased to some extent the pnp mRNA abundance in the absence of PNPase relative to the wild-type allele. However, PNPase further reduced the amount of pnp mRNA so that the AI remained greater than 0.7. Likewise, deletions ΔpnpL1005 and ΔpnpL1010, which remove, respectively, a 5′ region of the coding sequence previously implicated in autogenous control (49) and the Rho-independent transcription terminator, did not affect autogenous regulation.

Overall, these data suggest that removal of the double-stranded 5′-UTR, which under wild-type conditions is removed by the concerted RNase III and PNPase activities, is necessary and sufficient to destabilize pnp mRNA.

Translation is not a direct target of PNPase autoregulation.

It has been proposed that PNPase indirectly induces its own mRNA decay by binding to the pnp 5′-UTR and preventing translation (49). A prediction of this model is that mutations preventing translation would greatly destabilize pnp mRNA in a PNPase-independent manner. We therefore constructed an AGGA→ACCA double transversion in the pnp Shine-Dalgarno sequence (pnpL1007), and a TTG→TTT transversion (pnp-1008) that destroys the natural UUG start codon of pnp. As a control, TTG was replaced with the canonical (ATG; pnp-1009) start codon sequence.

In the absence of PNPase, the pnp mRNA abundance of such constructs correlated with their expected translatability. All of these constructs, however, were subject to autoregulation (AI of >0.7) as PNPase drastically reduced their pnp transcript levels. Therefore, although in the absence of PNPase pnp mRNA translation seems to affect its stability, translation inhibition is not sufficient to fully destabilize the pnp mRNA at a level comparable to that observed in the presence of PNPase. This indicates that PNPase cannot control the stability of its mRNA by simply inhibiting translation of the processed transcript and suggests that an additional stabilizing factor needs to be directly removed by PNPase.

The RNase III-processed pnp mRNA with a single-stranded 5′-UTR is degraded by RNase E in a PNPase-independent manner and is not translatable.

RNase E-dependent degradation is thought to be the major pathway for mRNA decay and has been implicated in RNase III-processed pnp mRNA destabilization (20, 23). A major RNase E cut site was previously found 178 to 180 nt within the pnp coding sequence (20). An in-frame deletion of such a site (Δpnp-1006), however, did not affect autogenous regulation (Fig. 4), suggesting that additional RNase E sites could be used for pnp mRNA decay.

We thus tested pnp mRNA abundance and stability in an rne-thermosensitive (rne-3071) (18) mutant. As shown in Fig. 5A (upper panel), a temperature shift from 30 to 44°C led to a remarkable increase of pnp mRNA abundance in the rne-3071 strain, whereas it decreased in the isogenic wild-type strain. The half-lives at 30 and 44°C in the mutant were about 2 and 10 min, respectively (Fig. 5B), whereas in the wild type at 44°C, half-life could not be adequately assessed given the low abundance of the mRNA. Interestingly, the abundance of the small RNA37 in the rne-3071 mutant at 44°C did not concomitantly increase with the mature pnp mRNA (Fig. 5A, middle panel), suggesting that at the nonpermissive temperature its turnover was not affected and, therefore, at 44°C a larger proportion of the pnp mRNA 5′-UTR was mostly single stranded.

FIG. 5.

Stabilization of pnp mRNA and PNPase expression upon RNase E inactivation. Strains C-5869 (rne⁺; wild type [wt]), C-5868 (rne-3071; RNase E-thermosensitive mutant [rne^ts]), and C-5684 (Δrnc38::kan; RNase III-null mutant [Δrnc]) were grown at 30°C to an OD₆₀₀ of 0.8 and then shifted at 44°C. At different time points from the temperature shift, samples were taken for RNA and protein extraction. Northern and Western blotting were then performed as described in Materials and Methods. (A, upper panels) Northern blotting of a denaturing agarose gel electrophoresis hybridized with radiolabeled PNP5 riboprobe that reveals monocistronic (2.25, arrowhead) and polycistronic (upper bands) pnp mRNAs. (Middle panels) Northern blotting of a denaturing acrylamide gel electrophoresis hybridized with radiolabeled ANTI37 riboprobe that reveals a ladder of PNP37 polyadenylated forms. (Lower panels) Western blotting with anti-PNPase serum as described in Materials and Methods. (B) Decay of C-5686 pnp mRNA at 30 and 44°C. Rifampin (rif) was added to an aliquot of the 30°C culture at an OD₆₀₀ of 0.8 and 40 min after the shift at 44°C. Samples were taken at different time points for RNA extractions, and half-lives (t_1/2) were assessed as described. The strain and sample time points after either temperature shift or rifampin addition (in min) are indicated on the figure. The apparent molecular size (in kb) of the major transcripts detected is also shown (see Fig. S1 in the supplemental material). Some variability in the relative intensity of the 2.25- and 5.4-kb mRNAs has been observed with the rne mutant in different experiments.

It may be noticed that the RNA37 appears as a ladder of discrete bands (Fig. 5A). These forms derive from polyadenyl polymerase (PAP)-dependent oligoadenylation, as they disappeared in pcnB mutants (see Fig. S4, lower panel, in the supplemental material). Nevertheless, PAP does not seem to play any role in autoregulation, since this is not affected in pcnB mutants (23; see Fig. S4, upper panel, in the supplemental material).

We also tested PNPase abundance under such conditions (Fig. 5A, lower panel). Like in the wild-type strain, in the rne-3071 strain the PNPase abundance remained substantially unchanged upon temperature upshift, notwithstanding the dramatic (over fivefold) increase in pnp mRNA. This suggests that pnp mRNA with a single-stranded 5′-UTR is poorly, if at all, translated. On the contrary, in an RNase III mutant a greater protein abundance corresponded to the higher pnp mRNA level (Fig. 5A).

To investigate on the relationships between the structure of pnp mRNA and its translatability, we transformed the rne-3071 and its wild-type isogenic strain with plasmids expressing a truncated PNPase (Pnp-ΔKHS1 from the Δpnp-833 allele, a mutant that does not exhibit autoregulation) (4) under the pnp-p promoter and a wild-type or mutant 5′-UTR. This allowed us to distinguish the endogenous from the ectopically expressed pnp mRNAs and PNPases by Northern and Western blotting, respectively. The data obtained are shown in Fig. 6. In the wild-type strain, the amounts of both Pnp and Pnp-ΔKHS1 proteins only slightly decreased at 44°C relative to 30°C, although the mRNA level decreased at the higher temperature. This could be dependent on the long turnover time of PNPase. In the rne(Ts) strain, however, the increased mRNA abundance at 44°C was not accompanied by an increase in Pnp-ΔKHS1—rather, it slightly decreased. This suggests that the pnp mRNA, coordinately processed by RNase III and PNPase, is not (efficiently) translated. The highest level of Pnp-ΔKHS1, corresponding to the highest abundance of mRNA, was attained with the ΔpnpL1001 leader sequence in both wild-type and rne(Ts) strains at both temperatures. This indicates that a pnp mRNA with a nonprocessable double-stranded stem may be efficiently translated. Curiously, the abundance of the plasmid-encoded ΔpnpL1001 mRNA was lower in the rne(Ts) strain and appeared to diminish at 44°C. It might be that in the absence of RNase E-dependent pathways, other decay pathways may take over. On the other hand, both the plasmid-encoded ΔpnpL1002 Δpnp-KHS1 mRNA and ΔPnp-KHS1 were undetectable in the rne⁺ host and in the rne(Ts) mutant at 30°C. In the rne(Ts) strain at 44°C, however, the mRNA signal, albeit weak, was present, whereas the protein could not be detected. This suggests that the RNase III-processed mRNA missing a double-stranded stem is not only very unstable but also poorly, if at all, translated.

FIG. 6.

Expression of PNPase upon RNase E inactivation in different 5′-UTR mutants. Strains C-5869 (rne⁺) and C-5868 (rne-3071; RNase E-thermosensitive mutant [rne^ts]) were transformed with plasmids pAZ133, pAZ133L1001, and pAZ133L1002, which express a KHS1-truncated PNPase under a pnpL⁺, ΔpnpL1001, and ΔpnpL1002 5′-UTR, respectively. Cultures were grown at 30°C and shifted at 44°C to inactivate the thermosensitive RNase E, and samples for RNA and protein extraction were taken before (0), and at different time points (indicated in min) after the temperature shift. Northern blotting with radiolabeled PNP5 riboprobe (upper panel) and Western blotting with anti-PNPase (middle panel) and anti-ribosomal protein S1 (lower panel) antisera were then performed as described in Materials and Methods. The extra bands in the S1 panel are remnants of prior labeling with anti-PNPase antibodies.

This conclusion was supported by the in vitro translation experiment shown in Fig. 7. Incorporation of [³⁵S]Met in an in vitro translation system was about 14-fold higher using a pnp mRNA starting from the natural pnp-p at +1 than with an mRNA starting at +76, which corresponds to the RNase III-processed pnp mRNA.

FIG. 7.

In vitro translation of full-length and 5′-end-truncated pnp mRNA. Incorporation of [³⁵S]methionine in an in vitro translation system at increasing concentration of pnp native (circles) and 5′-truncated (triangles) mRNAs was performed as described in Materials and Methods.

From the above results, it appears that RNase E acts in concert with and downstream of the step controlled by PNPase for the destabilization of pnp mRNA (Fig. 8). To test whether the RNA degradosome could be implicated in autoregulation, we analyzed the pnp mRNA in an RNase E mutant with the C-terminal domain required for the assembly of the degradosome deleted (31). However, pnp abundance was only slightly increased and the half-life was doubled (3 min in the mutant versus 1.5 min in the wild-type strain; data not shown). This could be imputed to a lower activity of the mutant RNase E. Moreover, rhlB deletion mutants did not affect autoregulation (data not shown). Thus, the RNA degradosome appears to be marginally implicated, if at all, in PNPase autogenous regulation.

FIG. 8.

A model for PNPase autogenous regulation. Control of PNPase expression occurs at three sequential steps. Both native (upper drawing) and RNase III-processed (second drawing) pnp mRNAs are stable and translatable. RNase III creates the substrate for PNPase that degrades the small RNA37, thus destroying the double-stranded 5′ stem (third drawing). The processed pnp mRNA with a single-stranded 5′ end (bottom drawing) is not translated and is targeted for RNase E-dependent decay.

DISCUSSION

Early work on PNPase expression control had shown that PNPase posttranscriptionally regulates its own expression by controlling the stability of pnp mRNA. RNase III makes a staggered double-strand cleavage about in the middle of a long hairpin in the 5′-UTR and generates a double-stranded stem in which a small RNA with a protruding 3′ end (RNA37) is paired to the new 5′-monophosphate end of the processed pnp mRNA (48, 58) (Fig. 8). It has been proposed that PNPase would act as a translational repressor by binding the RNase III-matured pnp mRNA (48) and that translation inhibition by PNPase would target the pnp transcript to RNA decay pathways (49). This model was based mainly on analysis of reporter gene expression in wild-type and mutant pnp 5′-UTR-lacZ translational fusions and the mRNA decay rate of such constructs by hybridization of pulse-labeled RNA to DNA-specific probes.

More recently, the same group proposed an elegant new model that replaced the former one. This latest model maintains that the primary pnp transcript and the RNase III-processed mRNA are very stable by virtue of the double-stranded region at the 5′-UTR, whereas the transcript with a single-stranded 5′ end would be unstable. The stabilizing hairpin would be removed by the concerted action of RNase III and PNPase itself. RNase III, by cleaving the hairpin and generating the small RNA37 with a 3′ overhang, would create the substrate for PNPase exonucleolytic activity that would degrade the small RNA37, thus destroying the double-stranded stabilizing structure at the 5′ end (23) (Fig. 8).

However, a possible role of translational repression by PNPase had not been directly addressed in the context of the second model, and in our view, it was not completely ruled out. It thus remained possible that these two non-mutually exclusive mechanisms, removal of a stabilizing structure and translational repression, could operate in concert to autoregulate PNPase expression. For example, it could be assumed that PNPase binds to the pnp translation initiation region of the RNase III-processed pnp mRNA and both acts as a translational repressor and degrades the stabilizing small RNA37. Moreover, PNPase binding to the pnp translation initiation region could be required for RNA37 degradation. The latter hypothesis would be consistent with the observation that both phosphorolytic and RNA binding activities are necessary for autoregulation (4, 16, 22, 34). It thus remained unclear whether PNPase-dependent translation inhibition was implicated in destabilization of this transcript. Moreover, which decay pathway would degrade the processed pnp RNA had not been established.

The data we have presented support the most recent model as the only mechanism for PNPase autogenous regulation. In addition, we show that the RNase III-processed pnp transcript devoid of the double-stranded stem is both very unstable and does not seem to be translated and suggest that the high instability of this RNA is not a direct consequence of its poor translatability. Finally, we identify RNase E-dependent decay as the major pathway for degradation of mature pnp mRNA (see model in Fig. 8).

Our search for cis-acting determinants of pnp mRNA posttranscriptional regulation has been performed by Northern blotting of pnp mRNA of an internal pnp deletion mutant in its natural chromosomal context associated with cis mutations, in the presence and absence of ectopically expressed PNPase. This approach, unlike the use of protein-encoding reporter genes, avoids interference by variations of translation efficiency that may be associated with some mutations. Moreover, Northern blotting allows direct visualization of full-length (both native and processed) transcripts, thus avoiding to some extent possible interference by intermediate decay products. Some discrepancies between our results and previously published work (23, 48, 49) may be mostly imputed to the different experimental approaches.

PNPase regulates its own expression by transforming a stable mRNA precursor in an unstable transcript simply by degrading the small RNA37.

The data we have presented are in full agreement with the Jarrige et al. (23) model only. All conditions that prevent formation of a single-stranded 5′ end (mutation in the 5′ leader and/or mutations in RNase III and PNPase) abolish autogenous regulation, i.e., increase the abundance (and stability, when tested) of pnp mRNA in the presence of PNPase. Conversely, deletions that remove the double-stranded 5′ end destabilize the pnp mRNA independently of PNPase. Mutations affecting translation by destroying the Shine-Dalgarno sequence or the start codon also have an adverse effect on mRNA stability in the absence of PNPase. However, the pnp mRNA abundance and half-life are still higher than in the presence of PNPase, indicating that translational repression cannot be the only mechanism by which PNPase targets its own mRNA to decay and that removal of the double-strand structure at the 5′ end is required to fully destabilize the transcript. Thus, our data do not support the hypothesis of translational repression by PNPase previously proposed by Robert Le Meur and Portier (48, 49). Likewise, a 5′ region of the coding sequence previously implicated in autogenous control (49) and the Rho-independent transcription terminator do not appear to be implicated in this process, as deletions ΔpnpL1005 and ΔpnpL1010, which remove the 5′ region and Rho-independent transcription terminator, respectively, did not affect PNPase autoregulation.

Stability and translatability of processed pnp mRNA.

RNase III processing of pnp mRNA facilitates endonucleolytic degradation at several specific cleavage sites (56). Hajnsdorf et al. (20) implicated RNase E in this process and identified in vitro a high-affinity primary site and additional secondary sites for this endonuclease. We have shown in vivo that deletion of this high-affinity site did not affect autogenous regulation. However, inactivation of a thermosensitive RNase E stabilizes pnp mRNA in the presence of PNPase (20; this work). Therefore, it appears that RNase E is the major endoribonuclease responsible for the fast decay of processed pnp mRNA and that the secondary RNase E cleavage sites are sufficient to promote endonucleolytic degradation. Interestingly, the abundance of the small RNA37 did not increase at 44°C, whereas when autoregulation was abolished by lack of PNPase or transiently suppressed during cold acclimation (2, 62), both pnp mRNA abundance and RNA37 abundance increased concomitantly (unpublished data). This suggests that the RNA37 plays a major role in protecting the RNase III-processed pnp mRNA from RNase E.

Current models claim that RNase E-mediated mRNA decay in E. coli is triggered by translation inhibition, whereas translating ribosomes protect mRNA from RNase E endonucleolytic attack (10). We have shown both in vivo and in vitro a tight correlation between translatability and stability of pnp mRNA in PNPase autogenous regulation. However, rapid decay and inefficient translation appear to be two at least partially independent features of the RNase III-processed pnp mRNA lacking the 5′-end duplex. In fact, on the one hand pnp mRNAs harboring mutations that alter its translatability remain more stable in the absence of PNPase, thus suggesting that disruption of the 5′-end RNA duplex is sufficient to further destabilize the transcript; on the other hand, such an mRNA seems to be very poorly translated both in vivo when RNase E is inactivated and in vitro.

The protective role of the 5′-end hairpin can be easily reconciled with current models on mRNA decay mechanisms. It is known that a 5′-end triphosphorylated RNA is not a good substrate for RNase E (14, 24, 33, 61). It may be that in the RNase III-processed mRNA, the dangling RNA37 in the duplex region hinders the new monophosphate 5′ end and thus remains, like the native mRNA, inaccessible to RNase E. Recently, however, an RNA pyrophosphohydrolase (RppH) that removes a pyrophosphate from native RNAs has been found in E. coli (11). Since disruption of the double-strand stem via degradation of RNA37 seems to be required to trigger RNase E-mediated decay, it may be assumed that the 5′-end hairpin protects not only the recessed monophosphate 5′ end of the RNase III-processed transcript, but also, to some extent, the triphosphate 5′ end of native pnp mRNA from RppH activity, thus preventing RNase E-mediated decay until the RNase III-processed 5′-triphosphate stem is destroyed by PNPase degradation. Alternatively, the 5′ double-strand structure protects the pnp mRNA from RNase E irrespective of the 5′-end phosphate status.

Another puzzling issue is how a double-stranded 5′ stem makes the pnp mRNA translatable, whereas the processed single-stranded 5′ end is not. Comparison of secondary structures of native and RNase III-processed pnp mRNA predicted by MFOLD (63) did not show differences that could obviously suggest differential translation inhibition between the two forms. Structural studies of the alternative configurations that could be adopted by the pnp 5′-UTR will be essential to hint at reliable models. In addition, to our knowledge, nothing is known about the influence of 5′-end phosphorylation status on translation in E. coli. We cannot exclude that this feature could independently influence both stability and translatability of pnp mRNA.