Cold-temperature induction of Escherichia coli polynucleotide phosphorylase occurs by reversal of its autoregulation †‡



When Escherichia coli cells are shifted to low temperatures (e.g. 15°C), growth halts while the ‘cold shock response’ (CSR) genes are induced, after which growth resumes. One CSR gene, pnp, encodes polynucleotide phosphorylase (PNPase), a 3′-exoribonuclease and component of the RNA degradosome. At 37°C, ribonuclease III (RNase III, encoded by rnc) cleaves the pnp untranslated leader, whereupon PNPase represses its own translation by an unknown mechanism. Here, we show that PNPase cold-temperature induction involves several post-transcriptional events, all of which require the intact pnp mRNA leader. The bulk of induction results from reversal of autoregulation at a step subsequent to RNase III cleavage of the pnp leader. We also found that pnp translation occurs throughout cold-temperature adaptation, whereas lacZ+ translation was delayed. This difference is striking, as both mRNAs are greatly stabilized upon the shift to 15°C. However, unlike the lacZ+ mRNA, which remains stable during adaptation, pnp mRNA decay accelerates. Together with other evidence, these results suggest that mRNA is generally stabilized upon a shift to cold temperatures, but that a CSR mRNA-specific decay process is initiated during adaptation.


When Escherichia coli cells growing at 37°C are shifted to 20°C or lower, a phenomenon called the cold shock response (CSR) initiates (Jones and Inouye, 1994; Graumann and Marahiel, 1996; Panoff et al., 1998). When shifted to 10°C, growth halts for about 4 h, during which time the levels of so-called CSR proteins induce to abnormally high levels and then readjust to new steady-state levels, after which growth resumes with a generation time of ≈ 25 h (Jones et al., 1987). Some CSR proteins, including CspA, CspB, CspG, CsdA (DeaD) and RbfA, are transiently induced more than 10-fold, whereas others, including RecA, NusA, DNA gyrase subunit A and polynucleotide phosphorylase (PNPase), are induced to lesser extents. For example, CspA, often referred to as the ‘major cold shock protein’, induces more than 200-fold during the CSR, whereas PNPase induces only ≈ sevenfold (Jones et al., 1987; Goldstein et al., 1990; Mitta et al., 1997).

It is proposed that the primary purpose of the CSR is to ensure adequate rates of translation at low temperatures, but this has not been established (Brandi et al., 1996; Graumann and Marahiel, 1996; Jones and Inouye, 1996; Farewell and Neidhardt, 1998). Moreover, the molecular signal for cold-temperature induction is unknown, and the mechanisms by which individual genes are induced remain unclear, although both transcriptional and post-transcriptional changes have been documented (Brandi et al., 1996; Bae et al., 1997; Fang et al., 1997). At least one CSR protein, CspA, appears to respond more generally to stress, as it induces transiently at 37°C when cells are shifted from a saturated culture into fresh medium (Brandi et al., 1999).

Interestingly, five CSR genes, all of which are involved in post-transcriptional processes, are clustered at one location on the E. coli chromosome (Fig. 1). These genes encode transcription antitermination factor NusA (nusA;Greenblatt et al., 1981), translation initiation factor 2 (infB;Laalami et al., 1991), RbfA, a ribosome-binding factor possibly required for normal cold-temperature adaptation (rbfA;Dammel and Noller, 1995; Jones and Inouye, 1996), CsdA, an ATP-dependent RNA helicase (csdA, previously named deaD;Toone et al., 1991; Jones et al., 1996), and polynucleotide phosphorylase (PNPase), which is one of two 3′-exoribonucleases in E. coli and a component of the RNA degradosome (Carpousis et al., 1994; Py et al., 1994; reviewed by Coburn and Mackie, 1999).

Figure 1.

Genetic structure of the E. coli metY–mtr region and pnp–lacZ fusions. All known genes lying between metY and mtr of the E. coli chromosome are depicted (not to scale), along with relevant promoters (bent arrows), transcription terminators (filled lollipops; t) and RNase III cleavage sites (open lollipops). P1 and P2 are the principal promoters for pnp (stippled). RNase III cleaves a stem–loop structure in the untranslated pnp mRNA leader at 5′ and 3′ sites (R1 and R2 respectively), as depicted in the inset. See text for further detail. Structures (not to scale) of the pnp′–′lacZ translational and P1/P2–lacZ+ transcriptional fusions borne on λRS741 and λRS880 are also shown, aligned with corresponding elements in the chromosome (lacZ sequences are hatched). The latter fusion junction disrupts the stem–loop RNase III cleavage site. Grey and black boxes highlight the translation initiation sites for pnp and lacZ respectively. The lacZ leader is dashed. Approximate regions complementary to the PNP1 and LAC2 oligonucleotides, used for primer extension analysis of intact and fused pnp mRNAs, respectively, are indicated with split arrows.

There are a number of promoters and transcriptional terminators in the metY–mtr region and at least two stem–loop structures in the expressed transcripts that are cleaved by ribonuclease III (RNase III), a double-stranded RNA-specific endoribonuclease (Regnier and Portier, 1986; Regnier and Grunberg-Manago, 1989; 1990; Court, 1993; Nicholson, 1997). These features suggest that gene expression in this region is complex. Indeed, both rpsO and pnp are subject to post-transcriptional control (Regnier and Portier, 1986; Portier et al., 1987; Robert-Le Meur and Portier, 1992; 1994; Hajnsdorf et al., 1994), the latter being of particular interest to the work described here. Under normal growth conditions, the pnp gene is transcribed from two promoters: P1, which precedes rpsO; and P2, which precedes pnp itself (Regnier and Portier, 1986). More than 80% of pnp transcription derives from P1, showing that transcription termination at the t1 terminator is not absolute (Evans and Dennis, 1985; Regnier and Portier, 1986). Whether transcribed from P1 or P2, pnp mRNAs are efficiently cleaved by RNase III (Regnier and Portier, 1986; see Fig. 1), encoded by the unlinked rnc gene. Thus, most pnp transcripts extend from the downstream RNase III cleavage site (R2) in the pnp leader to a termination site (t2) located ≈ 2250 nucleotides (nt) downstream (and occasionally ≈ 500 nt beyond) (Portier et al., 1987).

In cells growing at 37°C, PNPase inhibits its own translation, possibly by binding the pnp mRNA and blocking ribosome binding (Robert-Le Meur and Portier, 1994). This, in turn, decreases pnp mRNA stability by an unknown mechanism. Importantly, PNPase autoregulation is efficient only when the pnp leader is cleaved by RNase III. Thus, RNase III cleavage produces the RNA ‘operator’ at which PNPase exerts translational repression. In rnc or pnp cells, pnp expression is ≈ sevenfold higher than in wild-type cells (Robert-Le Meur and Portier, 1992). Because pnp mRNA stability is apparently coupled to pnp translation, the level of pnp mRNA is also higher in rnc and pnp cells (Robert-Le Meur and Portier, 1994). Here, we show that cold-temperature induction of pnp expression occurs at several different post-transcriptional levels, including reversal of pnp autoregulation at a step subsequent to RNase III cleavage, all of which require an intact pnp mRNA leader.


PNPase expression is derepressed at low temperature

To explore whether pnp autoregulation plays a role in PNPase cold-temperature induction, we compared PNPase levels in wild-type and rnc14::mini-Tn10 cells (rnc14; this insertion mutation abolishes RNase III expression; Takiff et al., 1989). In Fig. 2A, total cell protein was examined by SDS–PAGE and Coomassie blue staining. Bands corresponding to the 85 kDa PNPase polypeptide are indicated as described previously (Portier et al., 1987). At 37°C, PNPase levels are elevated severalfold in rnc14 cells compared with the isogenic rnc+ strain (Fig. 2A, lanes 1 and 2), as expected. After a 3 h incubation at 15°C, PNPase levels increase a few fold in rnc+ cells (Fig. 2A, lane 3), but there is less apparent increase in cells lacking RNase III. To help quantify these changes and to identify PNPase unambiguously, we examined total cell protein by chemiluminescent Western blot analysis, using polyclonal antibody specific for PNPase (Fig. 2B). Although chemiluminescence is not linear with the amount of PNPase blotted (not shown), it is clear that the same pattern of PNPase expression is seen. As the rnc14 mutation derepresses PNPase expression at 37°C, but prevents further derepression at 15°C, it appears that PNPase cold-temperature induction in wild-type cells occurs by derepression of pnp autoregulation.

Figure 2.

Cold-temperature induction of PNPase in wild-type and rnc14 cells.

A. Total cell protein extracted from P90C (wt) or RS6521 (rnc14) cells growing at 37°C or incubated for 3 h at 15°C, separated by SDS–PAGE on a Novex 4–12% gradient gel for 18 h at 90 V and 4°C, and then stained with Coomassie blue, essentially as described (arrow identifies the 85 kDa PNPase band, as shown previously) (Portier et al., 1987).

B. PNPase levels were quantified by Western blot analysis with antibodies specific to PNPase and chemiluminescent reagents from Pierce Biochemicals (used according to the supplier's instructions). Relative chemiluminescence values (averages are shown; variation < 10%) were obtained with an AlphaInnotech ChemiImager 4400 gel imaging and analysis apparatus and alphaimager software package (numbered boxes show the areas quantified). Protein concentrations were determined with Bio-Rad protein assay dye.

PNPase derepression at low temperature requires the pnp untranslated leader

Because pnp autoregulation at 37°C requires cis-acting elements in the 5′ untranslated leader of the pnp gene (Robert-Le Meur and Portier, 1994), we examined cold-temperature induction of a pnp′–′lacZ translational fusion. λRS741 is a phage-borne fusion containing all known cis-acting elements required for pnp expression and autoregulation: the P1 and P2 promoters, the sites of RNase III cleavage, the pnp ribosome binding site and the first 61 codons of the pnp gene fused in frame to lacZ (Robert-Le Meur and Portier, 1992; see Fig. 1). An isogenic set of strains bearing single-copy λRS741 prophages was therefore constructed, and fusion expression was analysed under several relevant conditions (Table 1A).

Table 1. Effects of rnc and pnp mutations on cold-temperature induction of pnp′–′lacZ fusion expression.
 Relevant genotypesIncubation conditionsaCold-temperature induction
ratio (15°:37°)
BacterialMulticopyMid-log4.5 h g = 4  
strainbplasmidc37°C15°C15°C4.5 hg = 4
  • a

    . Aliquots were taken from: mid-log cultures growing at 37°C; 4.5 h after those mid-log cultures had been shifted to 15°C (during which time no appreciable growth occurred); and after cold-adapted cultures that had resumed mid-log growth at 15°C and undergone four cell doublings (g = 4).

  • b

    . Part A: RS8872 (wild type); RS8942 (rnc14); RS8873 (pnp::Tn5). Part B: RS9028 (wild type); RS9029 (rnc14); RS9030 (pnp::Tn5).

  • c . pRS2813 (null); pRS1398 (rnc +); pBP111 (pnp+).

  • d . Averages of at least three experiments (in Miller units), determined as described previously ( Simons et al., 1987).

A. pnp′–′lacZ translational fusion expression (units of β-galactosidase activityd)
 1Wild typenull365171518854.75.2
 2Wild type rnc + 27514805.4
 3Wild type pnp + 5051010.2
 4 rnc14 null1000182016951.81.7
 5 rnc14 rnc + 21011855.6
 6 pnp::Tn5null2235354031201.61.4
 7 pnp::Tn5 pnp + 1009359.3
B. P1/P2–lacZ+ transcriptional fusion expression (units of β-galactosidase activityd)
 8Wild type(none)134019151.4
 9 rnc14 (none)114018151.6
 10 pnp::Tn5(none)129018501.4

At 37°C, pnp′–′lacZ fusion expression is about threefold higher in an rnc14 strain than in wild-type cells (Table 1A, lines 1 and 4) and ≈ sixfold higher in a strain bearing the pnp::Tn5 insertion mutation (Table 1A, lines 1 and 6), as seen previously (Robert-Le Meur and Portier, 1992). These mutant effects are complemented by multicopy plasmids bearing the rnc+ and pnp+ genes respectively (Table 1A, lines 5 and 7). We note, in particular, that the multicopy pnp+ gene not only complements the pnp::Tn5 mutation, but reduces pnp′–′lacZ fusion expression ≈ sevenfold in otherwise wild-type cells (Table 1A, lines 1 and 3). This ‘super-repression’ is almost certainly caused by elevated levels of PNPase expressed from this plasmid. In contrast, a multicopy rnc+ gene exhibits little or no super-repression (Table 1A, lines 1 and 2). Even though a multicopy rnc+ gene overexpresses RNase III (Matsunaga et al., 1996), super-repression was not expected in this case, as the pnp leader is fully cleaved in wild-type cells (as shown below).

When these same strains are shifted to 15°C, cold-temperature induction of pnp′–′lacZ fusion expression is seen in all cases, but to different degrees. The induction ratio (15°:37°) after 4.5 h at 15°C is ≈ fivefold in wild-type cells (Table 1A, line 1), but less than twofold when the cells contain either the rnc14 or the pnp::Tn5 mutation (Table 1A, lines 4 and 6). These absolute values and induction ratios correlate reasonably well with the results shown in Fig. 2. As expected, a multicopy rnc+ gene restores essentially normal cold-temperature induction in rnc14 cells (Table 1A, line 5), but does not increase the induction ratio in either rnc+ or rnc14 cells (Table 1A, lines 2 and 5) (the pnp leader is fully cleaved in rnc+ cells at 15°C; see below). In contrast, the multicopy pnp+ gene, which super-represses fusion expression at 37°C, also increases the cold-temperature induction ratio to ≈ 10-fold (Table 1A, lines 3 and 7). Thus, in general, the greater the level of pnp′–′lacZ fusion repression at 37°C, the greater the magnitude of cold-temperature induction at 15°C.

After four cell generations at 15°C (g = 4), which requires ≈ 20 h for wild-type and rnc14 cells and much longer for the pnp::Tn5 mutant (see below), the levels of pnp′–′lacZ fusion expression are not much different than at the 4.5 h point (see Table 1A). Thus, pnp cold adaptation is complete in about 4–5 h. Nevertheless, reversal of pnp autoregulation is not maximal even after four cell generations, as higher levels of pnp′–′lacZ fusion expression are observed in the pnp::Tn5 mutant than in wild-type or rnc14 cells (Table 1A, cf. lines 1, 4 and 6). Moreover, the super-repressing multicopy pnp+ gene suppresses cold-temperature induction as much as threefold (Table 1A, lines 1 and 3). These results clearly suggest that some degree of autoregulation persists at low temperatures.

Both the rnc14 and the pnp::Tn5 mutations have pleiotropic effects on gene expression (Portier et al., 1987; Robert-Le Meur and Portier, 1992) and could therefore affect pnp′–′lacZ fusion expression at a level other than autoregulation. To address this and other questions, we constructed λRS880, a phage-borne fusion closely related to λRS741, in which P1, P2 and a proximal portion of the pnp leader are transcriptionally fused to an intact lacZ+ gene (see Fig. 1). This effectively replaced the portion of the pnp leader downstream of the R2 site, as well as the early coding sequence of the pnp gene, with the untranslated leader, ribosome binding site and early coding region of lacZ+ (without recreating an RNase III cleavage site). When this P1/P2–lacZ+ fusion was examined, no more than ≈ 1.5-fold cold-temperature induction was seen, and there was no effect of either the rnc14 or the pnp::Tn5 mutation (Table 1B.). Therefore, in the absence of the cis-acting autoregulatory sites present in the pnp leader, little or no cold-temperature induction of fusion expression is observed. Together, these results show that cold-temperature induction of pnp expression occurs primarily by reversal of its autoregulation.

pnp cold-temperature derepression occurs at a step subsequent to RNase III cleavage

As RNase III cleavage of the pnp leader is required for efficient pnp repression by PNPase, cold-temperature reversal of autoregulation might occur by inhibition of that cleavage. We examined this question directly through primer extension analysis of total cellular RNA extracted from relevant cell cultures (Fig. 3). When a primer complementary to the pnp mRNA (PNP1; see Fig. 1) is used to analyse RNA extracted from wild-type cells growing at 37°C, a single predominant band is seen (Fig. 3, lane 1). This signal corresponds to the 5′ terminus produced by RNase III cleavage of the pnp leader (R2; see Fig. 1), as shown previously (Regnier and Portier, 1986). When rnc14 cells growing at 37°C are analysed, four bands are consistently observed (Fig. 3, lane 2; bracketed and marked by *), which presumably correspond to the partially characterized 5′ termini of pnp mRNAs previously observed in RNase III-deficient cells (Portier et al., 1987). When these same strains are analysed 1–4 h after incubation at 15°C, the overall pattern remains unchanged (Fig. 3, lanes 3–10), showing that RNase III cleavage of the pnp leader is not perturbed. When a primer complementary to proximal lacZ mRNA sequences (LAC2; see Fig. 1) is used to analyse RNA extracted from strains bearing the pnp′–′lacZ fusion, the same overall pattern is observed (not shown). Together, these results show that RNase III cleavage of the pnp leader is as efficient at 15°C as it is at 37°C, and that there is no substantial difference between the processing or fate of intact pnp and fused pnp′–′lacZ mRNAs during cold-temperature incubation. There are, however, changes in the levels of these signals during the course of cold-temperature adaptation, especially in wild-type cells, which we address in greater detail below.

Figure 3.

Primer extension analysis of pnp mRNA in wild-type and rnc14 cells incubated at 37°C and 15°C. Total cellular RNA was extracted from wild-type (+, P90C) and rnc14 (–, RS6521) cells immediately before shifting from 37°C to 15°C (0 h) and at 1 h intervals thereafter and analysed by extension of the PNP1 primer (Fig. 1), essentially as described previously (Matsunaga et al., 1996), except that aliquots were immediately frozen in a dry ice–ethanol bath to prevent further cold shock induction. The resulting phosphorimage (Molecular Dynamics) is shown. The site of RNase III cleavage in the pnp leader (R2; Fig. 1) is marked. The bracket (with *) indicates the principal bands consistently observed in rnc14 cells (see text).

Pnp derepression commences upon shift to cold temperatures

To define further cold-temperature induction of pnp expression, we examined the time course of pnp′–′lacZ fusion activity in various cells after their shift to 15°C (Fig. 4A). During this time course, there is no appreciable change in cell density, so the results represent an accumulation of the β-galactosidase reporter. At 37°C, β-galactosidase levels are about three- and fivefold higher in rnc14 and pnp::Tn5 cells, respectively, than in the wild type and ≈ sevenfold lower in the presence of the super-repressing, multicopy pnp+ gene. After a shift to 15°C, β-galactosidase levels rise in all cells over the course of cold-temperature adaptation, but the apparent rates of synthesis differ (cf. the slopes in Fig. 4A). In wild-type cells, the β-galactosidase level rises to that seen in the rnc14 mutant, but not to that in pnp::Tn5 cells. When super-repressed, the apparent rate of synthesis is slowed. Nevertheless, in all cases, β-galactosidase levels increase steadily during adaptation, showing that the pnp gene is translated immediately upon (or very soon after) the shift to 15°C. In wild-type cells, this rate appears to accelerate at later times.

Figure 4.

pnp–lacZ fusion expression during cold-temperature adaptation.

A. β-Galactosidase levels were determined as described previously (Simons et al., 1987) in relevant strains bearing single-copy λRS741 (pnp′–′lacZ) prophages immediately before a shift from 37°C to 15°C (0 h) and at 45 min intervals thereafter. Values are expressed in Miller units. Strains used were RS8872 (wild type), RS8942 (rnc14), RS8873 (pnp::Tn5) and RS8872 transformed with pBP111 (pnp++).

B. β-Galactosidase levels were similarly determined in strains bearing single-copy λRS880 (P1/P2–lacZ+) prophages at 1 h intervals after shifting to 15°C. Strains used were RS9028 (wild type), RS9029 (rnc14) and RS9030 (pnp::Tn5).

In contrast, when the time course of P1/P2–lacZ+ fusion expression is examined under the same conditions (Fig. 4B), β-galactosidase activity does not begin to increase until about 2 h after the shift to 15°C, and there is no large effect of either the rnc14 or the pnp::Tn5 mutations. As the pnp′–′lacZ+ and P1/P2–lacZ+ fusions have the same promoters in common, but differ with regard to their sites of translation initiation, these results show that the pnp untranslated leader and/or early translated region bear determinants that enable its translation early in cold-temperature adaptation.

pnp mRNA levels rise transiently during cold-temperature adaptation

The results in Fig. 3 suggest that the relative levels of pnp mRNA increase in wild-type cells after a cold-temperature shift (Fig. 3, odd-numbered lanes), but do so less in rnc14 cells (Fig. 3, even-numbered lanes). As rnc14 cells are defective for pnp autoregulation, these results also suggest that the increase in primer extension signal seen in wild-type cells results from reversal of autoregulation or some other effect. To explore this observation more carefully, we used phosphorimage analysis to quantify pnp mRNA levels in appropriate cells at various times during the course of cold-temperature adaptation. Differences in primer extension termination signal intensities reflect differences in transcript levels per se (Matsunaga et al., 1996). In the case of rnc+ cells, we quantified the RNase III cleavage signal (R2; Figs 1 and 3). For rnc14 cells, the multiple bands detected in the absence of RNase III (*; Fig. 3) were quantified individually and then summed. The results are shown in Fig. 5. There are important technical differences in how these results were obtained. In Fig. 5A, the LAC2 primer was used to detect the RNase III-cleaved pnp′–′lacZ fusion mRNA in the presence or absence of normal or super-repressing levels of PNPase. As the fusion construct contains only the P1/P2 promoters, transcription from native upstream promoters is not accounted for. In Fig. 5B, the PNP1 primer was used to detect the (summed) multiple 5′ termini of uncleaved native pnp mRNAs in the absence of RNase III, and R2 in its presence. The summed bands are not directly comparable with the R2 band. Moreover, primer extension analysis will probably not detect the 5′ termini of uncleaved mRNAs originating at far-upstream promoters. In Fig. 5C, the PNP1 primer was used to detect RNase III-cleaved pnp mRNAs that either contain or do not contain the pnp::Tn5 insertion mutation. It is unknown whether disruption of normal pnp mRNA sequences by Tn5 alters pnp mRNA decay directly. For these technical reasons, the results shown in Fig. 5A–C are not directly comparable. Nevertheless, the overall trend remains quite similar in all cases. Specifically, pnp mRNA levels rise during the first 2 h of cold-temperature adaptation and then decline.

Figure 5.

pnp and pnp′–′lacZ mRNA levels during cold-temperature adaptation. Total cellular RNA was analysed essentially as described in the legend to Fig. 3, with either the PNP1 or LAC2 primers (Fig. 1). Appropriate bands were quantified by phosphorimage analysis and plotted versus time of extraction. For rnc+ strains, the R2 band was quantified; for rnc14 strains, the multiple bands bracketed (*) in Fig. 3 were quantified individually and summed. Band volumes are relative values uniquely obtained in each phosphorimage quantification.

A. The LAC2 primer was used to quantify pnp′–′lacZ fusion mRNAs in RS8872 (wild type), RS8873 (pnp::Tn5) and RS8872 transformed with pBP111 (pnp++).

B. The PNP1 primer was used to quantify pnp mRNAs in P90C (wild type) and RS6521 (rnc14).

C. The PNP1 primer was used to quantify pnp mRNAs in P90C (wild type) and RS7195 (pnp::Tn5). Note that, in RS7195, pnp mRNAs are disrupted by Tn5 sequences, which begin far downstream of the region complementary to PNP1.

Within 4 h, this rise and fall is essentially complete in wild-type cells. The decline is delayed somewhat in rnc14 and pnp::Tn5 cells, in which pnp autoregulation is disrupted, and accelerated slightly in super-repressed cells, suggesting that PNPase plays a minor role in the decay of these transcripts. Nevertheless, by the time two complete cell doublings at 15°C have occurred (requiring ≈ 10 h), the relative levels of pnp′–′lacZ mRNA in wild type, rnc14 and pnp::Tn5 mutants, as well as the rnc14 pnp::Tn5 double mutant, differ by no more than 40% (Fig. 6). Thus, whereas the rnc14 and pnp::Tn5 mutations disrupt pnp autoregulation at 37°C and retard cold-temperature adaptation (rise and fall) of pnp mRNA levels at 15°C, they ultimately have little effect on pnp mRNA or PNPase levels once adaptation is achieved.

Figure 6.

Phosphorimage analysis of pnp′–′lacZ mRNAs after cold-temperature adaptation has occurred. Total cellular RNA was extracted, analysed and quantified with the LAC2 primer as described in the legend to Fig. 5, except that cultures had doubled twice at 15°C (taking > 20 h). Relative phosphorimage volumes (the means of three independent quantifications ± SD) are shown at the bottom. Strains used were RS8872 (wild type), RS8942 (rnc14), RS8873 (pnp::Tn5) and RS8917 (rnc14 pnp::Tn5).

When we similarly examined the lacZ+ mRNA expressed from the P1/P2–lacZ+ fusion in wild-type cells, its level rose ≈ twofold in the first 2 h but then levelled off (not shown). Therefore, the rise and fall observed with the intact pnp mRNA must result, in part, from determinants in its 5′ leader.

Combined with other results described above, these time course experiments at the mRNA level reveal several important insights. It is clear that both transcription and translation of the pnp gene continue immediately or soon after the shift to 15°C and at times thereafter, as evidenced by the accumulation of pnp mRNA, β-galactosidase reporter and PNPase polypeptide (in cells that are not dividing). The lack of P1/P2–lacZ+ fusion expression in the first 2 h of adaptation (Fig. 4B) is therefore probably due to an initial block of lacZ+ translation initiation. This in turn implies that pnp translation initiation determinants are poised for cold-temperature function, whereas those of lacZ+ are not. Moreover, as these effects are seen in the context of the pnp′–′lacZ fusion, it is clear that the cis-acting determinants of pnp cold-temperature translation reside within the untranslated leader and/or early coding region of the pnp mRNA.

pnp mRNA is strongly but transiently stabilized during cold-temperature adaptation

The rise and fall of pnp mRNA levels seen during cold-temperature adaptation could result from changes in transcription initiation (and/or antitermination) or from changes in pnp mRNA stability. To examine this question directly, we determined the rates of pnp mRNA decay at various points during cold-temperature adaptation after adding rifampicin to block further mRNA synthesis (see Experimental procedures for details). Typical results are shown in Fig. 7, and the half-lives determined in different cells are listed in Table 2. In all cases, decay was logarithmic.

Figure 7.

Decay of the pnp–lacZ mRNA in wild-type cells at different times during cold-temperature adaptation. A culture of RS8872 growing at 37°C was shifted to 15°C. At 10 s, 2 h and 4 h intervals, portions of the culture were removed, combined with rifampicin (rif; 400 µg ml−1 final concentration) to inhibit further transcription and returned to 15°C. Immediately after combination with rifampicin, and at 30, 60 and 120 min intervals thereafter, aliquots were removed in each case and immediately frozen in a dry ice–ethanol bath (to terminate RNA decay). Total cellular RNA was then extracted and quantified by primer extension analysis (with oligonucleotide LAC2) and phosphorimaging, as described in the legends to Figs 3 and 5. The pnp′–′lacZ mRNA half-life at each interval (0, 2 and 4 h after shift) was then determined from a plot of the log (band volume) versus time after combination with rifampicin.

Table 2. mRNA half-lives during cold-temperature adaptation.


Half-livesc (in h) after cold-
temperature adaptation for
10 s2 h4 h
  1. a . Bacterial strains used (see Table 4) were as follows. wild type (P90C, RS8872 and RS9028); rnc14 (RS6521 and RS8942); pnp::Tn5 (RS8873 and RS9030); multicopy pnp+ (RS8872 transformed with pBP111).

  2. b . Intact pnp mRNAs were detected by primer extension analysis with the PNP1 oligonucleotide and pnp′–′lacZ fusion and P1/P2–lacZ + mRNAs with the LAC2 oligonucleotide (see Fig. 1).

  3. c . Half-lives were determined after the addition of rifampicin as described in the legend to Fig. 8, after cells had been incubated at 15°C for the time shown. Appropriate primer extension signals (R2 in rnc+ cells; summed multiple bands in rnc14 cells; see Figs 1 and 3) were quantified by phosphorimage analysis. All decay rates were approximately logarithmic.

Wild type pnp
pnp′–′lacZ > 40.90.35
rnc14 pnp 31.50.8
pnp′–′lacZ 31.10.6
pnp::Tn5 pnp′–′lacZ
Multicopy pnp+ pnp′–′lacZ
Wild typeP1/P2-lacZ+

At 37°C, pnp and pnp′–′lacZ mRNAs are quite unstable in the presence of autoregulation, with a chemical half-life of ≈ 90 s (Portier et al., 1987). In contrast, upon a shift to 15°C, these mRNAs are stabilized dramatically, exhibiting half-lives of the order of 3 h. This stabilization is essentially instantaneous, as it is established within the first 10 s. Clearly, the pnp mRNA decay processes operating at 37°C are essentially inactivated at 15°C. After 2 h of cold-temperature adaptation, stability decreased about twofold (to a half-life of ≈ 1.5 h) and by 4 h to ≈ 30 min. These differential effects on stability account for the rise and fall in pnp and pnp′–′lacZ mRNA levels seen in Fig. 5 and suggest that little or no induction of transcription occurs. Importantly, as decay at each of these points is logarithmic (once rifampicin has been added), the acceleration in decay seen over the course of cold-temperature adaptation must depend on the synthesis (or activation) of some degradative enzyme or other function. Zangrossi et al. (2000) have also documented transient stabilization of the pnp mRNA during cold-temperature adaptation.

As there is no substantial difference between the decay of native pnp and pnp′–′lacZ fusion mRNAs, the rate-limiting, cis-acting determinants of pnp mRNA decay probably lie wholly within the portion of the untranslated pnp leader downstream of R2 and/or in the early pnp coding region. It was therefore of interest to examine decay of the P1/P2 fusion mRNA, which lacks part or all of these determinants. In this case, the mRNA is stabilized throughout cold-temperature adaptation (Table 2). These effects on stability account for the rise and levelling off of lacZ+ mRNA levels noted above and again show that little or no induction of pnp transcription is occurring. Additionally, they show that the intact pnp leader is required for differential decay of the pnp mRNA during cold-temperature adaptation, as well as for autoregulation and early cold-temperature translation. Finally, Table 2 also shows that the rnc14 and pnp::Tn5 mutations have no more than an ≈ twofold effect on pnp, pnp′–′lacZ and P1/P2 mRNA stability, indicating that neither of these ribonucleases plays a significant role in pnp mRNA decay at 15°C.

PNPase is required for optimal cell growth and normal cspA mRNA adaptation at low temperatures

Although the autorepressor activity of PNPase is apparently altered at 15°C, three observations suggest that its ribonuclease activity is largely normal. First, whereas the pnp::Tn5 mutation had little effect on cell growth at 37°C, it severely retarded growth at 15°C and all but stopped growth at 10°C (Table 3). It has been shown previously that PNPase is required for the low-temperature growth of Escherichia coli, Bacillus subtilis and Yersinia enterocolitica (Luttinger et al., 1996; Goverde et al., 1998). Secondly, we found that PNPase is required for normal cold-temperature adaptation of the cspA mRNA. In wild-type cells, cspA mRNA levels rise and fall dramatically during the first few hours after a shift to cold temperatures (Fig. 8; Goldenberg et al., 1996). When pnp::Tn5 cells were examined under the same conditions, cspA mRNA levels rose more quickly and to a much higher level and then fell more slowly than in isogenic wild-type cells (Fig. 8). These mutant effects are complemented by a multicopy pnp+ gene. The rnc14 mutation had only very minor effects on cspA mRNA levels at 37°C or 15°C (not shown). Interestingly, PNPase was recently shown to have the same effect on cspA mRNA levels during cold-temperature adaptation by Y. enterocolitica (Neuhaus et al., 2000). These observations are consistent with the notion that the ribonuclease activity of PNPase is active at 15°C, where it contributes to cspA mRNA decay. Thirdly, Zangrossi et al. (2000) have shown that polynucleotide phosphorylase activity per se is induced two- to threefold at low temperatures.

Table 3. Effects of the pnp::Tn5 mutation on low-temperature growth.

Multicopy plasmidsColony diameter relative to wild type when incubated atb
NameGenotype37°C (18 h)15°C (9 days)10°C (30 days)
  • a

    . Strains were P90C (wild type) and RS7195 (pnp::Tn5).

  • b

    . Relative colony diameters after growth on LB plates at the indicated temperatures were scored by eye. NG, no growth; ND, not determined.

Wild typeNone101010
Wild typepRS2813Null1010ND
Wild typepBP111 truB + rpsO + pnp + yhbM + 10910
Wild typepRS3474 truB + rpsO + pnp + 1010ND
Wild typepRS3480 truB + rpsO + 101ND
pnp::Tn5pBP111 truB + rpsO + pnp + yhbM + 10109
pnp::Tn5pRS3474 truB + rpsO + pnp + 109ND
pnp::Tn5pRS3480 truB + rpsO + 91ND
Figure 8.

Cold-temperature adaptation of the cspA mRNA in cells lacking PNPase. Total cellular RNA was extracted from wild-type (P90C) and pnp::Tn5 (RS7195) cells immediately before shifting from 37°C to 15°C (0 h) and at 1 h intervals thereafter and analysed as described in the legends to Figs 3 and 5, except that a primer (CSPA2) complementary to the early coding region of the cspA mRNA was used.


Cold-temperature induction of PNPase involves reversal of pnp autoregulation

The work described here establishes that cold-temperature induction of PNPase occurs at several post-transcriptional levels. Importantly, the pnp untranslated leader and/or early coding region contain determinants necessary for all these events. Little or no increase in transcription is apparent. Most of the induction occurs through a mechanism involving reversal of the pnp translational autorepression that operates at higher growth temperatures. This is seen most clearly in the effects of rnc and pnp mutations, which derepress pnp expression at 37°C but prevent further derepression at cold temperatures. Partial deletion of the pnp mRNA leader has essentially the same effect.

Other important post-transcriptional aspects of pnp cold-temperature induction were revealed during careful studies on the course of this adaptive process. First, pnp translation continues immediately after the shift to cold temperatures, as evidenced by the early and steady accumulation of PNPase polypeptide and expression of the pnp′–′lacZ fusion at 15°C. Zangrossi et al. (2000) have drawn essentially the same conclusion. This immediate pnp translation is in contrast to the lacZ+ mRNA expressed from the P1/P2–lacZ+ fusion, which is not translated during the first 2 h of cold-temperature adaptation, but is at later times. The functional difference between the pnp and lacZ+ mRNAs must therefore lie in their untranslated leaders or early coding sequences. Nevertheless, there is no obvious conserved sequence in these regions of the pnp mRNAs from different organisms, despite conservation of the RNase III cleavage site (Goverde et al., 1998; R. W. Simons and R. K. Beran, unpublished observation). Importantly, reversal of pnp autoregulation early in cold-temperature adaptation would not be possible were it not for the intrinsic, cold-temperature translational competence of the pnp mRNA.

These observations also show that cells growing at 37°C have the ability to translate the pnp mRNA at 15°C, but can translate the lacZ+ mRNA only after the induction (or activation) of a new translational component. The pnp mRNA may experience this adaptation as well, as its translation accelerates late in adaptation at a time when its level is declining. Goldenberg et al. (1997) arrived at the same general conclusion, having shown that translation of the lacZ and cat (chloramphenicol acetyltransferase) genes is delayed during cold adaptation, whereas that of the cold shock genes cspA, csdA and infB is not. It is likely that other genes will fall appropriately into these two functional classes.

Another aspect of pnp induction revealed during adaptation occurs at the level of mRNA stability. In wild-type cells growing at 37°C, the pnp mRNA is very short-lived, decaying with an ≈ 1.5 min half-life (Portier et al., 1987). When cells are shifted to 15°C, however, there is an immediate and profound stabilization of this mRNA, such that it now decays with an ≈ 3 h half-life. The decay rate then increases gradually over the next 4 h, resulting in an ≈ 0.5 h half-life by the time cell growth resumes. In contrast, the lacZ+ mRNA, which lacks the 5′ determinants necessary for both autoregulation and cold-temperature translational competence, is strongly stabilized throughout cold-temperature adaptation.

Interestingly, the stability of the cspA (Goldenberg et al., 1996) and csdA (R. K. Beran, unpublished observation) mRNAs is differentially regulated during cold-temperature adaptation in the same general way as is the pnp mRNA. In contrast, our preliminary studies show that a number of non-CSR mRNAs are stabilized throughout adaptation like the lacZ+ transcript. These observations suggest that the machinery that degrades mRNAs efficiently at 37°C does so only inefficiently at 15°C. They also suggest that, during the course of cold-temperature adaptation, a new CSR mRNA-specific decay process is apparently induced (or activated). Whether key functional elements of the 37°C machinery are rescued by a cold-adaptive component or entirely replaced by a new process remains an important question.

It has been argued that translation of the pnp mRNA enhances its stability at 37°C (Robert-Le Meur and Portier, 1994). Such coupling may be a general property of mRNAs at 37°C. It might then be argued that derepression of pnp translational control at 15°C is responsible for the increased pnp mRNA stability seen at cold temperatures. This cannot be true, however, as pnp translation accelerates during cold-temperature adaption, during which time pnp mRNA stability declines. Rather, differential regulation of pnp mRNA stability appears to dampen the effects of increased pnp translation. Moreover, the lacZ+ mRNA, which is not translated until well into adaptation, remains stable throughout. We suggest that coupling between translation and mRNA stability is suspended at low growth temperatures. With these considerations in mind, the general differences between pnp mRNA levels in wild-type and mutant cells in the midst of cold-temperature adaptation must largely reflect the higher transcript levels in mutant cells before the shift and the need to overcome pnp autoregulation in wild-type cells. Once adaptation has been achieved, pnp transcript levels differ by no more than ≈ 40% in mutant and wild-type cells.

Upstream transcription antitermination is probably not involved in pnp induction

Bae et al. (2000) recently concluded that CspA and other CspA-like proteins antiterminate transcription in the metY–mtr region and suggested that this may be the means by which pnp and certain other genes in that region are induced during cold-temperature adaptation. Several observations suggest that antitermination of pnp transcription plays only a small role at best. First, the metY–pnp readthrough transcript identified by Bae et al. (2000) undergoes very little cold temperature induction compared with the induction of putative readthrough transcripts bearing metY–infB or metY–rpsO (refer to Fig. 1). Secondly, in rnc+pnp::Tn5 cells (in which autoregulation is absent), we observe that pnp mRNA levels increase only ≈ twofold, reaching a maximum level only slightly higher than that observed in wild-type cells, at least as assessed by the level of primer extension signal corresponding to the R2 cleavage site (Fig. 5). As we have established that RNase III is fully active during cold-temperature adaptation, and it is reasonable to assume that all transcripts containing the rpsO–pnp intergenic region are efficiently cleaved by RNase III regardless of their site of initiation or level of antitermination, this observation strongly suggests that pnp mRNA levels increase no more than ≈ twofold during adaptation. Thirdly, we and others find that the level of PNPase polypeptide induction is of the order of five- to 10-fold (Jones et al., 1987; Fig. 2). This is the same magnitude of induction seen with pnp′–′lacZ fusions, which lack promoters upstream of the rpsO promoter, suggesting that most cold-temperature pnp transcription derives from the nearby P1/P2 promoters.

From Northern analysis, Zangrossi et al. (2000) suggested that downstream transcription antitermination may play a role in pnp induction. We cannot rule this out, as our experimental approach focuses on the 5′-proximal determinants of pnp induction. They also suggest that cold-temperature induction of the csdA gene, which lies downstream of pnp, depends to some extent on antitermination of upstream transcription. Our primer extension analysis reveals a substantial induction of csdA transcription at 15°C (R. K. Beran, unpublished observations). However, the 5′-terminus of this csdA transcript maps downstream of the pnp gene. Moreover, this transcript is seen in both wild-type and pnp::Tn5 mutant cells and induces to similar levels in both cell types. These observations suggest that cold-temperature induction of csdA mRNA is not coupled to that of pnp, as its character is not perturbed by the intervening Tn5 insertion.

What is the mechanism by which pnp autoregulation is reversed?

How PNPase represses its own translation at 37°C and what changes occur during cold-temperature induction to reverse this regulation are important questions. Several mechanisms can be entertained. Robert-Le Meur and Portier (1994) have suggested that PNPase binds the RNase III-processed pnp mRNA at or near its Shine–Dalgarno sequence to block ribosome binding directly during translation initiation. As RNase III cleavage of the pnp leader is normal at 15°C, reversal of autoregulation must occur at a subsequent step. This could be through an inhibition or alteration of the RNA-binding properties of PNPase. However, such a mechanism seems unlikely, as PNPase is required for growth at low temperatures, where its ribonuclease activity appears to be functional (at least with regard to cspA mRNA decay).

It seems more likely that reversal of autoregulation involves changes in competition between PNPase and 30S ribosomal subunits for the cleaved pnp leader. At 37°C, where the pnp mRNA must compete with all mRNAs for a limited number of free 30S subunits, PNPase appears to compete effectively with the 30S subunits for pnp mRNA binding, thereby establishing autoregulation. However, when cells are shifted to 15°C, the pnp mRNA is (at least initially) one of only a few transcripts apparently competent for 30S binding and, under such conditions, it is likely that 30S subunits can compete more effectively with PNPase for pnp mRNA binding, thereby circumventing autoregulation. This mechanism is supported by the observation that a multicopy pnp+ gene super-represses pnp′–′lacZ expression much more severely at 37°C than at 15°C.

PNPase function at cold temperatures

At low temperatures, PNPase is required for optimal growth of E. coli (Table 3), as well as B. subtilis and Y. enterocolitica (Luttinger et al., 1996; Goverde et al., 1998). Nevertheless, the biological function of PNPase at low temperatures remains unclear. In both organisms, PNPase is required for normal induction and decline of the cspA mRNA (Fig. 8; Neuhaus et al., 2000), suggesting an indirect role in the process of cspA cold-temperature adaptation. However, as CspA itself is not essential for growth at low temperatures (Bae et al., 1997), it appears that PNPase plays a broader role in cold-temperature growth.

At 37°C, PNPase is a component of the E. coli degradosome, a multienzyme ‘machine’ that plays an important role in RNA degradation (Carpousis et al., 1994; Py et al., 1994). PNPase may play a similar role at low temperatures. Interestingly, the degradosome isolated from cells growing at 37°C contains, in addition to PNPase and ribonuclease E, an ATP-dependent RNA helicase called RhlB (Py et al., 1996), which enables the complex to degrade highly structured RNAs (Coburn et al., 1999). CsdA is homologous to RhlB, and we found recently that it can fully replace RhlB when degradosomes are assembled in vitro (R. K. Beran, A. Prud'homme Genereux, G. Mackie and R. W. Simons, unpublished observations). We predict that PNPase and CsdA are important at low temperatures, in part so that a cold-adapted degradosome can be assembled. However, as ribonuclease E mutants unable to assemble the 37°C degradosome grow normally at 15°C, assembly of any such cold-adapted degradosome cannot account for the cold-temperature essentiality of pnp and csdA in a simple way.

Experimental procedures


Restriction and DNA modification enzymes were from New England Biolabs. Reverse transcriptase was from Promega. Radioisotopes were from New England Nuclear. Chemiluminescent reagents were from Pierce Chemical. All other materials were from Fisher or Sigma. Oligonucleotides, synthesized on a Beckman Oligo 1000 or purchased from Gibco BRL or New England Biolabs, were (5′−3′): LAC2, GTTTTCCCAGTCACGAC; PNP1, CGATCGGATTAAGCAATG; R1, CCGGAATTCAACCGTCTTGCGATAACAGG; R2, GCGGATCCATTAGCCGCGCGAACCTCTG; CSPA2, CCATTTTACGATACCAGTCAT.

Strains, media and growth conditions

Plasmid, phage and bacterial strains are listed in Table 4. pMLM100, previously named pGF, was kindly provided by Dr C. Portier. λRS741 (kindly provided by Dr J. Matsunaga) was constructed by in vitro recombination between pMLM100 and λRS45, as described previously (Simons et al., 1987). pRS3474 was constructed in vitro by deleting the yhbM-bearing EcoRI–KpnI fragment from pBP111 (identical to pB15-6; Plumbridge and Springer, 1983). pRS3480 was constructed in vitro by deleting the pnp-yhbM-bearing, EcoRI–SmaI fragment from pBP111. To construct pRS3478, an ≈ 540 bp fragment bearing rpsO, the P1 and P2 promoters and a proximal portion of the pnp leader (ending at the position corresponding to the R2 cleavage site; see Fig. 1) was prepared by polymerase chain reaction (PCR) amplification (with primers R1 and R2), cleaved at upstream EcoRI and downstream BamHII sites (incorporated during PCR), ligated into the corresponding cloning sites of pRS1553 and confirmed by DNA sequencing after recombinant plasmid recovery. pRS1553 is a specifically designed plasmid bearing an α-complementing fragment of the promoterless lacZ gene (Pepe et al., 1997). λRS468 (Pepe et al., 1997) is a specifically designed phage bearing a lacZ gene that is slightly truncated at its 5′ end. λRS880 was constructed by in vivo recombination between pRS3478 and λRS468, essentially as described previously (Simons et al., 1987), such that the rpsO–P1/P2 fragment was fused to an intact lacZ gene. RS8917 was constructed by transducing RS8873 with phage P1vir grown on RS6521, selecting for TcR. All other strains were obtained from the sources indicated or constructed by routine methods (Simons et al., 1987; Miller, 1992). All bacteria were grown at 37°C unless otherwise indicated. Liquid and solid LB media were used throughout. When needed, antibiotics were supplied at (µg ml−1): ampicillin (Ap), 150; kanamycin (Km), 35; tetracycline (Tc), 15.

Table 4. Bacteria, bacteriophage and plasmids.
NameRelevant genotypeaReference/source
  1. a . See Experimental procedures for details of strain construction. All plasmids are based on pBR322. Tc R, KmR and ApR, tetracycline, kanamycin and ampicillin resistance respectively.

 P90CΔ(lac-pro) thi ara Miller (1992)
 RS6521P90C rnc14::mini-Tn10 (TcR) Matsunaga et al. (1996)
 RS7195P90C pnp::Tn5 (KmR) Pepe et al. (1994)
 RS8872P90C λRS741This work
 RS8942RS6521 λRS741This work
 RS8873RS7195 λRS741This work
 RS8917RS8873 rnc14This work
 RS9028P90C λRS880This work
 RS9029RS6521 λRS880This work
 RS9030RS7195 λRS880This work
 λRS741 pnp′–′lacZ translational fusionThis work
 λRS880P1/P2–lacZ+ transcriptional fusionThis work
 λRS468 ′lacZ Pepe et al. (1997)
Multicopy plasmids (ApR)
 pBP111 truB + rpsO + pnp + yhbM + Plumbridge and Springer (1983)
 pBP13–1 truB + rpsO + pnp + yhbM + csdA + Plumbridge and Springer (1983)
 pRS1398 rnc + ApR Matsunaga et al. (1996)
 pRS1553 lacZ′ (α-complementing) Pepe et al. (1997)
 pRS2813pBR322 TcS Johnstone et al. (1999)
 pRS3474 truB + rpsO + pnp + This work
 pRS3480 truB + rpsO + This work
 pRS3478P1/P2–lacZ+ transcriptional fusionThis work
 pMLM100 pnp′–′lacZ translational fusion Robert-Le Meur and Portier (1992)


The authors thank Dr Claude Portier for providing strains and unpublished results, Drs A. J. Carpousis and C. Portier for PNPase antibodies, Dr Jim Matsunaga for preliminary strain constructions, and all members of the laboratory for helpful discussion and reading of the manuscript. This work was supported by grants from the National Institutes of Health and the National Science Foundation. R.K.B. was the recipient of a Pauley Predoctoral Fellowship.


  1. TempEdited by Dr R. P. Gunsalus.