Acid- and multistress-resistant mutants of Lactococcus lactis : identification of intracellular stress signals

Authors


Emmanuelle Maguin, E-mail maguin@biotec.jouy.inra.fr; Tel. (+33) 1 34 65 25 18; Fax (+33) 1 34 65 25 21.

Abstract

Lactococcus lactis growth is accompanied by lactic acid production, which results in acidification of the medium and arrest of cell multiplication. Despite growth limitation at low pH, there is evidence that lactococci do have inducible responses to an acid pH. In order to characterize the genes involved in acid tolerance responses, we selected acid-resistant insertional mutants of the L. lactis strain MG1363. Twenty-one independent characterized mutants were affected in 18 different loci, some of which are implicated in transport systems or base metabolism. None of these genes was identified previously as involved in lactococcal acid tolerance. The various phenotypes obtained by acid stress selection allowed us to define four classes of mutants, two of which comprise multistress-resistant strains. Our results reveal that L. lactis has several means of protecting itself against low pH, at least one of which results in multiple stress resistance. In particular, intracellular phosphate and guanine nucleotide pools, notably (p)ppGpp, are likely to act as signals that determine the level of lactococcal stress response induction. Our results provide a link between the physiological state of the cell and the level of stress tolerance and establish a role for the stringent response in acid stress response regulation.

Introduction

A distinctive feature of lactic acid bacteria (LAB) is that their growth is accompanied by lactic acid production, which results in acidification of the medium and, eventually, transition to stationary phase. For eubacteria, an acid environment inhibits cell multiplication and can result in cell death. As their own growth results in an acidic environment, LAB would be expected to have efficient acid stress resistance mechanisms to allow their survival.

Lactococcus lactis is a mesophilic LAB commonly used as a starter in cheese production. As neutrophiles, lactococci exhibit optimal growth rates within an external pH range of 6.3–6.9 (Hutkins and Nannen, 1993). During fermentation, the external pH (pHout) decreases. For pHout values above 5, L. lactis can maintain a ΔpH of around 1. Below this value, the ΔpH begins to collapse. At the end of milk fermentation, the pHout is around 4.5, and the ΔpH is maintained at around 0.6 for at least 72 h (Nannen and Hutkins, 1991a). Growth arrest is caused by the low intracellular pH (pHin), despite non-limiting nutrient concentrations. However, lactococci (but not all L. lactis ssp. cremoris strains; see Kim et al., 1999) appear to have inducible responses to an acid pH. First, they are more resistant to acid stress after carbon starvation (Hutkins and Nannen, 1993; Hartke et al., 1994). Second, L. lactis induces an adaptive response in mildly acidic media (Hartke et al., 1996; Rallu et al., 1996; O'Sullivan and Condon, 1997). Although induction involves protein synthesis, the requirement for de novo protein synthesis is controversial. In an L. lactis ssp. lactis strain (IL1403), development of acid tolerance is reportedly not protein synthesis dependent, but acid adaptation does induce 33 proteins, including a few heat shock and UV-inducible proteins (Hartke et al., 1996; 1997). For L. lactis ssp. cremoris strains (MG1363 and 712), acid adaptation needs de novo protein synthesis (Rallu et al., 1996) and results in increased survival from heat, ethanol, H2O2 and NaCl stress (strain 712; O'Sullivan and Condon, 1997). These observations suggest that acid stress adaptation overlaps with other stress responses and results in cross-protection.

Acid-adapted cells reportedly maintain a slightly higher pHin (around 0.2 pH unit) than non-adapted cells during an acid challenge (O'Sullivan and Condon, 1997), suggesting that a physical maintenance of pH homeostasis is involved. To date, three functions possibly implicated in pH homeostasis have been characterized in lactococci: (i) an H+-ATPase; (ii) the arginine deiminase pathway; and (iii) a glutamate decarboxylase. The H+-ATPase expels protons out of the cell via ATP hydrolysis. This activity, which increases as the pH decreases, consumes ATP and is thus detrimental to growth (Suzuki et al., 1988; Nannen and Hutkins, 1991b). Mutants defective in the H+-ATPase are impaired for survival at low pHout (Yokota et al., 1995). The arginine deiminase pathway (ADI) converts arginine to ammonia, ornithine and carbon dioxide and generates 1 mol of ATP per mol of arginine. Ammonia production (contingent upon arginine availability) may contribute to survival at low pHout by neutralizing the pH (Marquis et al., 1987; Casiano-Colon and Marquis, 1988; Poolman and Konings, 1988; Curran et al., 1995). Nevertheless, most L. lactis ssp. cremoris strains (with the exception of MG1363; Poolman et al., 1987a; Crow and Thomas, 1982) seem to lack a functional ADI pathway. Glutamate decarboxylase (GadB), which converts glutamate plus a proton to γ-aminobutyrate and carbon dioxide, may contribute to pH homeostasis and has been characterized in strain MG1363 (Sanders et al., 1998). The gadB gene forms an operon with an antiporter, gadC, which expels γ-aminobutyrate and internalizes glutamate. Mutants of the gad operon are impaired in acid stress survival in the presence of chloride and glutamate (Sanders et al., 1998). All three mechanisms may reduce acidification of the internal compartment and thus be important in maintaining acid resistance. Also, an acid- and stationary phase-inducible promoter controlling the expression of an unknown function has been characterized recently (Madsen et al., 1999). However, regulation of these and other stress response factors remains essentially unknown in lactococci.

Acid stress responses seem to be inducible in bacteria (Hall et al., 1995; Slonczewski and Foster, 1996), thus implying that regulatory functions exist. One means of identifying such regulators is by selecting acid-resistant mutants. Few acid-resistant mutants have been described, and only one is in a LAB (Oenococcus oeni ). The mutant was selected as a survivor of long-term incubation at pH 2.6, and biochemical studies suggest that its high H+-ATPase activity might be the basis for acid resistance (Drici-Cachon et al., 1996). In Salmonella typhimurium, mutants that survived incubation at pH 3.3 were also isolated (Foster and Hall, 1991): 18% were auxotrophs for either a base or an amino acid (among which half required glutamate). Also in S. typhimurium, Tn10 inactivation of atbR (pgi ) and mviA (rssB ), which are involved in the regulation of 10 acid-inducible proteins and of rpoS respectively, increased the acid tolerance of the strain (Foster and Bearson, 1994; Bearson et al., 1996; 1998). Although these mutants were not initially selected as acid resistant, their phenotypes suggested that it must be possible to use that selection to identify regulators.

In this work, we have developed a strategy for identifying factors involved in stress regulatory networks in lactococci. We have isolated and characterized 21 acid-resistant insertional mutants of L. lactis (called arl for acid-resistant locus). Theoretically, such mutants could be specific for a stress condition or could confer a more general stress resistance. Both types of mutants were obtained. Our data indicate that the intracellular pools of phosphate and guanine nucleotides (GP), notably(p)ppGpp, are involved in the regulation of acid stress tolerance in L. lactis.

Results

Selection of acid-resistant mutants

We observed that the combination of a mild, non-lethal acid stress (pH 5 or 5.5) and elevated temperature (37.5°C) is lethal to L. lactis strain MG1363. We performed insertional mutagenesis, selecting for survivors under these conditions. Mutagenesis using the pG + host9:ISS1 plasmid (pGh9:IS; Maguin et al., 1996) was carried out in L. lactis strain MG1363. For these studies, the standard rich medium (M17) was modified to maintain cultures above ≈ pH 6.8 (referred to as M17/7, Experimental procedures ) or at different acid pH values (referred to as M17/pHvalue). Cultures were grown initially in M17/7 at 30°C, and acid-resistant mutants were selected at 37.5°C on M17/5 or M17/5.5 plates containing erythromycin (Experimental procedures ). The plating efficiency of MG1363 on low-pH plates (M17/5 or M17/5.5) compared with M17/7 plates was ≈10−4 at 37.5°C (whereas it was 1 at 30°C). Forty putative mutants obtained by our selection procedure showed higher acid plating efficiencies, suggesting that they were indeed all altered in acid resistance. Twenty-one mutants that exhibited acid plating efficiencies varying from 1 to 10−1 were named arl (for acid-resistant locus) and characterized further.

Identification of the transposition target

ISS1 transposition into the chromosome results in insertion of duplicated ISS1 flanking the pGh9 sequence (Maguin et al., 1996). Unique restriction sites, EcoRI and HindIII, present on pGh9, were used to clone chromosomal sequences flanking the transposed structure (Experimental procedures ). Twenty-one clones were characterized by sequencing the junctions. Mutations in 18 different loci were obtained, of which 10 showed homology with known genes, as described in Table 1 and below.

Table 1. . Characterization of the arl mutants.Thumbnail image of
  • a

    a. Organism, gene name and, in parentheses, length of the protein sequence giving highest homology score using the tblastn program are shown. A result is given only if the p(N) probability is lower than 10−10.b. Numbers correspond to the amino acid positions in the protein showing the highest homology score.c. Numbers correspond to the deduced amino acid position of ISS1-generated interruption in the protein showing the highest homology.d. Results from growth in M17/7. In this condition, the generation time of the wild-type strain is 32 ± 3 min.e. Only one junction cloned.f. Sequence analysed length (in nucleotides, nt) for the mutants showing no homology.ND, not determined; RBS, ribosome binding site.

  • Glutamate/glutamine transporter.

    Four mutants (arl11arl14 ) were clustered in an operon putatively encoding the glutamate/glutamine ABC transporter of L. lactis (Table 1). These mutants were at least 200-fold more resistant than the parental strain to 5 mM glutamic acid hydroxamate, a toxic glutamate analogue, indicating that glutamate transport was altered in all mutants and confirming the mutation assignment.

    High-affinity phosphate transporter.

    Two mutants (arl15 and arl16 ; Table 1) were affected in genes homologous to yzmB and yzmE of Bacillus subtilis, which correspond to the high-affinity phosphate transport pstS and pstB genes respectively (Takemaru et al., 1996). As the pstB gene is duplicated in L. lactis (data not shown), only the pstS mutant was studied further.

    Purine metabolism.

    Several mutants were affected in genes putatively involved in different aspects of purine metabolism (Zalkin, 1993; Zalkin and Nygaard, 1996): deoB (arl21arl22, phosphopentamutase; Duwat et al., 1997), guaA (arl19–arl20, GMP synthetase; Mantsala and Zalkin, 1992), hpt (arl17, hypoxanthine guanine phosphoribosyltransferase; Nilsson and Lauridsen, 1992), relA (arl18, pppGpp synthetase; Mechold et al., 1993) and possibly arl7 (see below). The deoB gene is part of the deoB–orfC–deoD operon (Duwat et al., 1997), and it is likely that, in both deoB mutants (arl21 and arl22 ), ISS1 insertion inactivates the entire operon. The guaA, hpt and relA genes constitute single transcriptional units. Although hpt is followed by the heat shock protease gene ftsH (Nilsson and Lauridsen, 1992; Nilsson et al., 1994), these genes appear to have independent promoters; it is thus unlikely that ftsH is affected by the hpt mutation. The RelA protein is responsible for (p)ppGpp synthesis during the stringent response and is conserved in Gram-positive and Gram-negative bacteria (Cashel et al., 1996). In Gram-positive bacteria, it has been shown that RelA is probably implicated in both the synthesis and the degradation of (p)ppGpp (Mechold et al., 1996; Mechold and Malke, 1997; Martinez-Costa et al., 1998; Wehmeier et al., 1998).

    Genes with homologues of unknown function.

    One mutant (arl23 ) exhibited significant homology with the B. subtilis recN gene, which is implicated in DNA repair (Van Hoy and Hoch, 1990; Alonso et al., 1993; Walker, 1996). However, to our knowledge, its role in DNA metabolism is not clearly documented. Two other mutations (arl8 and arl10 ) map in a homologue of an unknown B. subtilis gene, yybT (Kunst et al., 1997).

    Sequences with no homologies.

    No gene assignment based on sequence homologies could be made for six mutants (arl1arl7 ). However, our results suggest that arl7 is involved in guanine metabolism as it cannot grow in chemically defined SA medium (Jensen and Hammer, 1993) unless supplemented with guanine.

    To summarize, the 21 characterized independent mutants were altered in 18 different loci, none of which were previously associated with acid resistance in LAB.

    Mutant growth rates

    Growth rate and acidification are key factors for industrial use of LAB. In M17/7, 10 out of the 14 tested mutants exhibited growth rates similar to that of the parental MG1363 strain (doubling time around 32 min). The remaining four mutants exhibited longer doubling times (43 min; Table 1). These results show that most of the mutants grow well in M17 media. Acid tolerance may thus be related to an auxotrophy in some of the mutants, but does not result from slow growth.

    Acid resistance of the mutants at 30°C

    To study mutants at 30°C, the optimal growth temperature of L. lactis, the thermosensitive pGh9:IS plasmid was excised from the transposed structure, thus leaving a single copy of ISS1 at the mutated locus (Experimental procedures ). When multiple insertions in a given gene were obtained, all mutants were characterized; as they all displayed similar phenotypes, a single mutant of each type is presented. Single cross-over integration was performed to reconstruct mutants in the hpt, guaA, relA, pstS and recN genes (Experimental procedures ); in all cases, phenotypes were indistinguishable from those of the original mutants.

    Tolerance to a lethal acid shift.

    We compared the acid tolerance of mutants and wild-type (wt) strain at 30°C under two physiological states, i.e. exponential phase or early stationary phase (2 h after cessation of growth at pH 6.8; Fig. 1, first two columns). As cells are less acid tolerant in the exponential phase than after carbon starvation, different pH conditions were used in order to obtain a similar survival of the wt strain in both tests.

    Figure 1.

    . Relative survival of arl mutants and wt strain under acid stress conditions. 1, class 1 mutants; 2, class 2 mutants; 3, class 3 mutants and 4, class 4 mutants. Tolerance to lethal acid stress was determined using exponentially growing cells, after 1 h at pH 3.7 (first column), or using early stationary cells, after 2 h at pH 3.0 (second column). As cells are less acid tolerant in exponential phase than after carbon starvation, the acid challenge was performed at a different pH, such that similar survival rates of the wt strain were obtained under both conditions. Long-term survival, i.e. tolerance to glucose starvation, was estimated (i) under acidic conditions after 2 days of starvation (third column, grey bars) and (ii) for class 4 mutants, at neutral pH after 4 days of starvation (1.4, third column, striped bars); in this test, wt survival rate was around 1.6 × 10−3. Survival rate is calculated as the ratio of the number of colonies obtained on M17 plates after (T1) and before (T0) challenge. The relative survival is the ratio of the survival rate of the mutant to that of the wt strain. Standard deviations of three experiments are shown. Note that, as mentioned in the text, the pstB (arl16 ) was not characterized.

    All mutants were more acid resistant than MG1363 under one or both tested conditions. This result confirmed that all mutants were acid resistant independently of growth temperature. We distinguished three classes of mutants depending upon their phenotypes. Class 1 mutants, glnQ (arl14 ) and pstS (arl15 ), were more resistant than wt to an acid shift only during exponential growth (Fig. 1.1). Class 2 mutants, glnH (arl12 ), glnP (arl13 ), arl11 and arl3, exhibited improved survival in early stationary phase (Fig. 1.2). The other 11 mutants survived the acid shift better than the wt (up to 1000-fold increased survival) and independently of growth phase; these were tentatively grouped in class 3 (Figs 1.3 and 1.4).

    Long-term survival.

    The stationary phase in LAB can result from nutrient exhaustion and/or extensive acidification of the medium. Long-term (2 days) survival of acid-resistant mutants and the wt strain was compared under conditions in which growth arrest resulted from glucose starvation at low pH (Fig. 1, third column, grey bars). Long-term survival of mutants belonging to classes 1 and 2, as well as five mutants grouped in class 3, was equivalent to that of MG1363 (i.e. 10−4 after 2 days). In contrast, survival of six mutants initially assigned to class 3 was 120- to more than 1000-fold better than that of MG1363 (Fig. 1.4); these were assigned to class 4. Four of these mutants are affected in purine metabolism (relA, hpt, guaA and arl7 ), and the other two, arl5 and arl8 (yybT ), have unknown functions. Survival of class 4 mutants was also monitored after glucose starvation at neutral pHout (Fig. 1.4, third column, striped bars). After 4 days at 30°C, survival of arl8 and MG1363 was similar (around 6 × 10−5), indicating that arl8 survival advantage was restricted to an acidic environment. However, the relA, hpt, guaA, arl7 and arl5 mutants were all more resistant to carbon starvation at pH 6.8 (20- to 110-fold). Considering that survival during carbon starvation often involves global stress resistance, we predict that many class 4 mutants are multistress resistant.

    Tolerance to other lethal stresses

    We tested the arl mutants for survival under stress conditions other than acid, i.e. oxidative and heat shock stress.

    Oxidative stress.

    Five of the tested clones (arl2, arl4, relA, hpt and guaA) displayed slightly increased survival (two- to 10-fold) compared with wt in the presence of H2O2. In contrast, pstS and arl1 mutants were markedly more resistant (2000- and 800-fold respectively) to oxidative stress (Fig. 2A).

    Figure 2.

    . Relative survival of arl mutants and the wt strain under oxidative stress and heat shock conditions. A. Oxidative stress tests, carried out by incubating exponentially growing cultures with 1 mM H2O2 for 30 min (see Experimental procedures ), resulted in a survival rate of 10−4 for the wt strain. The relative survival rate of mutated strains was calculated as described in the legend to Fig. 1. B. Heat shock was performed by shifting cultures from 30°C to 55°C for 5 min (non-adapted, grey bars) or by incubating the cultures for 15 min at 37°C before transfer to a 15 min challenge at 55°C (adapted, striped bars). The survival rate of the wt strain is around 8 × 10−4 in both cases. Standard deviations of three experiments are shown.

    Heat shock.

    Seven mutants (recN, relA, deoB, hpt, guaA, arl7 and arl5 ) showed improved survival to a heat shock compared with MG1363 (Fig. 2B). If cells were first adapted at an intermediate temperature, mutants relA, guaA, arl7 and arl5 remained more resistant, whereas recN, deoB and hpt were about as tolerant as the wt strain. This result suggests that, for the latter three mutants, increased heat shock survival may have resulted from elevated levels of heat shock proteins at 30°C. Western blots performed on extracts of the recN, guaA and wt strains seemed to confirm this hypothesis (data not shown). Taken together, the above results show that 11 of the mutants isolated as acid resistant are also resistant to other stress conditions.

    Some of the mutations resulting in multistress resistance are predicted to alter intracellular levels of particular metabolites: phosphate in the pstS mutant, (p)ppGpp in the relA strain and guanosine monophosphate (GMP) in guaA and hpt. We therefore studied the effects of altering metabolite pools on acid tolerance phenotypes in the mutated and wt strains.

    Phosphate is an internal stress signal

    Many bacteria have both high- and low-affinity phosphate transport systems (Wanner, 1996; Qi et al., 1997). The pstS insertion presumably inactivates the high-affinity transporter, which is used at low phosphate concentrations (< 1 mM). An observable phenotype for the pstS mutant in M17/7 suggests that phosphate concentration in this medium is low; the addition of 20 mM phosphate to the culture medium should allow efficient phosphate transport via the low-affinity phosphate transporter. Indeed, the addition of phosphate to the medium of the pstS strain abolished acid resistance (Fig. 3), whereas no such effect was observed for guaA, hpt and relA strains. This result shows that acid resistance of the pstS mutant results from a decreased internal phosphate concentration.

    Figure 3.

    . Relative survival of acid-tolerant mutants and wt strain under acid stress in the presence of phosphate. Tolerance to lethal acid shift was estimated on exponentially growing cells (see Experimental procedures ) without (grey bars) or with (striped bars) 20 μmol ml−1 potassium phosphate (Pho) added during growth and acid challenge. Survival rate of the wt strain was 7 × 10−5 and 8 × 10−4, without and with Pho respectively. Standard deviations of three experiments are shown.

    GP/GMP pools are internal stress signals

    The effect of adding different GMP precursors on the acid tolerance phenotype of mutant strains was examined. Cultures were grown and acid challenged in M17 supplemented with guanine, guanosine or hypoxanthine (Fig. 4B). As expected, the pstS and relA mutants were unaffected by base additions (Fig. 4B). However, acid resistance of the hpt clone was suppressed when medium was supplemented with hypoxanthine, but not with the two other guanine nucleotide precursors tested (Fig. 4B), indicating that acid resistance of hpt is related to low levels of the intracellular nucleotide pool. Note that, in Escherichia coli, the lactococcal hpt gene seemed to be active on guanine and hypoxanthine substrates (Nilsson et al., 1994). In contrast, suppression of acid resistance by hypoxanthine, but not by guanine, suggests that, in lactococci, the hpt gene product only acts on guanine. For the guaA mutant, acid tolerance was abolished when grown with added guanine or guanosine, but remained resistant in the presence of hypoxanthine (Fig. 4B). These observations suggest that full complementation of the GMP biosynthesis defect in the guaA mutant suppresses the stress resistance phenotype.

    Figure 4.

    . Purine nucleotide biosynthesis and acid tolerance. A. Schematic representation of GP and (p)ppGpp biosynthesis pathways in Gram-positive bacteria. Functions of mutants studied here are indicated in italics. In Gram-positive bacteria, RelA both synthesizes and degrades (p)ppGpp, although the degradation products are not clearly identified. B. Relative survival of acid-tolerant mutants and wt strain under acid stress in the presence of bases. Tolerance to lethal acid shift was examined in exponentially growing cells (see Experimental procedures ) without base addition (grey bars) or with guanine or guanosine (both gave similar results, striped bars) or hypoxanthine (dotted bars) added during growth and acid challenge. The survival rate of the wt strain was 7 × 10−5 without added bases and 7 × 10−4 with added guanine, guanosine or hypoxanthine. C. Survival of wt strain during acid stress in the presence of decoyinine. The tolerance to a lethal acid shift was estimated in exponentially growing cells (see Experimental procedures ) without (circles) or with decoyinine. Decoyinine was added either during growth and acid challenge (diamonds) or during acid challenge only (triangles). Percentage survival was estimated by plating on M17 before (T0) and after 1 and 2 h of acid challenge. Standard deviations of three experiments are shown.

    To demonstrate further that GP pools determine the degree of acid resistance, we treated MG1363 with decoyinine, an inhibitor of the guaA gene product (Mitani et al., 1977), before an acid pH shift (Fig. 4C). Decoyinine-treated cells survived a pH downshift 100-fold better than the non-treated culture. Treatment with decoyinine concomitant with acid challenge had no effect on bacterial survival (Fig. 4C). This experiment shows that alteration of the GP biosynthesis flux before a lethal acid challenge leads to improved cell survival.

    (p)ppGpp pools control stress tolerance

    Sequence analysis showed that the L. lactis relA mutant (arl18 ) potentially synthesizes a 451-amino-acid truncated protein (Table 1). In E. coli and in Streptomyces coelicolor, truncated RelA′ proteins of 455 and 489 amino acids respectively, synthesize (p)ppGpp (Schreiber et al., 1991; Svitil et al., 1993; Martinez-Costa et al., 1998). Therefore, we could not exclude the possibility that acid tolerance of the relA mutant was caused by the expression of a truncated but active form of (p)ppGpp synthetase. Chromatography experiments revealed the presence of low levels of (p)ppGpp in the relA451 mutant (Fig. 5A). To test the role of (p)ppGpp pools in acid tolerance further, we treated the wt strain with serine hydroxamate or α-methylglucoside, two inducers of the stringent response (Cashel et al., 1996). Both treatments improved acid tolerance of the wt strain about 100-fold (Fig. 5B). This result strongly suggests that stringent response induction is a signal for acid resistance induction.

    Figure 5.

    . (p)ppGpp and acid tolerance. A. Detection of (p)ppGpp in wt and relA451 strains. Spots were identified by comparison of 32P-GTP-labelled E. coli (CF1648) extracts prepared from non-induced cells or after stringent response induction. B. Acid tolerance of MG1363 in the presence of drugs that induce (p)ppGpp. Tolerance to a lethal acid shift was determined on cells grown to exponential phase in SA without or with SH or MG. To mimic a relA phenotype (i.e. good growth), we used SH and MG concentrations that did not significantly affect strain growth. Without drugs, survival of MG1363 to acid treatment was 10−4. Relative survival was estimated as above.

    Discussion

    We determined a combination of non-lethal conditions, i.e. low pH, elevated temperature and aerobiosis, under which L. lactis is poorly viable. Twenty-one insertional mutants (affected in 16 different loci) able to survive under these conditions were selected. All of them were acid resistant, regardless of temperature. Moreover, 11 mutants (68%) were multistress resistant. We propose that this selective procedure based on the association of various stresses favoured the isolation of multiresistant clones and can simplify identification of global regulatory systems. Duwat et al. (1999) used a combination of stress conditions to select thermoresistant mutants in a recA background: 86% of them were multistress resistant. Note that mutations in pstS, pstB, guaA and deoB were obtained in both studies. These data emphasize the importance of the phosphate and purine nucleotide pools in the expression of multistress tolerance in L. lactis.

    The existence of four classes of arl mutants (based on their acid tolerance phenotypes) strongly suggests that, in L. lactis, acid tolerance is a complex process involving several mechanisms. Possibly, these mechanisms (i) improve pH homeostasis at low pH (by modification of buffering capacity) and/or (ii) protect cellular components against acid (and possibly other stress) damage (for example through expression of repair functions; Hahn et al., 1999). Therefore, we propose that mutants belonging to classes 1 and 2 are likely to be altered in pH homeostasis; note that transporters are prevalent in these classes. On the other hand, expression of protective functions may lead more often to a general acid (i.e. growth phase independent) and multistress tolerance; consequently, mutants of classes 3 and 4 may be of the second type.

    Transport functions and acid tolerance

    Several mutants affect ABC transport functions and are involved in either glutamate (arl11–14 ) or phosphate (arl15, arl16 ) transport. Biochemical studies have shown that activity of both transporters decreases with lower pH (Poolman et al., 1987c,d). Our results also show that inactivation of these transporters is beneficial for survival upon a pH downshift. Both glutamate and phosphate transporters hydrolyse ATP to internalize anions as electroneutral species. As the transport defect results in decreased ATP consumption, proton uptake and/or anion uptake, it may improve pHin maintenance at acid pH. Alternatively, particular metabolites may act as stress sensors in the cell; decreased levels in a transport mutant may induce a stress response. This has been well documented for phosphate ion limitation (Spira et al., 1995; Eymann et al., 1996; Wanner, 1996). In E. coli, a link between phosphate concentration and acid tolerance has also been reported (Rowbury et al., 1992). Note that these two hypotheses are not mutually exclusive.

    Purine metabolism, stringent response and stress resistance

    Eleven mutants (one of class 1, all of class 3 and five out of six class 4 mutants) are more resistant not only to acid but also to at least one other stress condition tested. Thus, mutations that confer acid tolerance in a particular physiological state (i.e. exponential growth or stationary phase) can be distinguished from those that lead to general acid tolerance (grouped in classes 3 and 4). Our results also reveal the importance of purine metabolism in stress signalling, as five (relA, deoB, hpt, guaA, arl7 ) of the 11 multistress-resistant clones are altered in this pathway.

    In Streptococcus mutans, characterization of acid-sensitive mutants might suggest a link between purines and acid tolerance (Yamashita et al., 1993; Crowley et al., 1997). In addition, a previous study in E. coli suggested that the stringent response is implicated in acid resistance under high salt conditions (Rowbury, 1993; 1997). Our results in L. lactis under normal salt conditions show clearly that GP and (p)ppGpp pools affect resistance to acid stress. The involvement of stringent response in starvation and stress tolerance has been studied in E. coli, S. typhimurium and Vibrio cholerae (Nyström, 1994; Munro et al., 1995; Ostling et al., 1995; 1996; Aizenman et al., 1996; Cashel et al., 1996) and also in B. subtilis (Ikehara et al., 1984; Hecker et al., 1989; Wendrich and Marahiel, 1997). From these studies and our present results in L. lactis, GP and (p)ppGpp are implicated in multistress tolerance response. In all cases, variations in metabolite levels (or ratios) appear to induce functions that protect cellular components from stress damage.

    Do the guaA and relA mutants induce multistress resistance by the same pathway? Analysis of known pathways and our data lead us to propose that both mutants act via induction of the stringent response: (i) both GP and (p)ppGpp participate in the same metabolic pathway; (ii) we observed a striking similarity between the relA and guaA mutant phenotypes; (iii) for the relA mutant, detection of (p)ppGpp is consistent with a direct involvement of the stringent response; and (iv) lower GTP levels predicted in the guaA mutant would decrease translation efficiency (GTP is an important cofactor in translation; Karimi et al., 1999; Luchin et al., 1999), and thereby might induce a stringent response. In addition, it is known that, in L. lactis, acidification inhibits various amino acid and oligopeptide transport systems (Poolman et al., 1987b). Presumably, this can lead to reduced translation and, consequently, induction of the stringent response (Cashel et al., 1996). The above consideration leads us to propose that the guaA and relA mutants have equivalent effects on cell physiology leading to multistress resistance.

    Conclusion

    In this work, the genetics of acid tolerance were studied on an acidifying fermenting bacterium. A combination of mild stress conditions resulted in lethality for L. lactis. Selection of surviving mutants led to the identification of several, previously unidentified genes involved in stress response. The variety of mutant phenotypes obtained suggests that this acidifying bacterium has several acid tolerance responses and/or regulatory networks. Mutants with altered GP and (p)ppGpp pools display the most striking acid-resistant phenotypes. Our data also suggest that modifications of flux through the purine–nucleotide pathway and/or increased (p)ppGpp concentrations are perceived as intracellular stress signals in L. lactis, leading to multistress tolerance. These results provide a link between the physiological state of the cell and the level of stress tolerance and establish a role for the stringent response in acid stress response regulation.

    Experimental procedures

    Bacterial strains and growth conditions

    E. coli TG1 (Sambroock et al., 1989) [supE hsd-5 thiΔ(lac-proAB ) F′(traD6 proAB+ lacI q lacZΔM15 )] and EC101 containing a chromosomal copy of the pWV01 repA gene (Leenhouts, 1995) were grown in Luria–Bertani medium (Miller, 1972) at 37°C, to which 150 mg ml−1 erythromycin (Em) was added as required. L. lactis MG1363 (Gasson, 1983) was grown in M17 medium (Terzaghy and Sandine, 1975) supplemented with 1% glucose at 30°C or 37.5°C, as specified. M17/7 corresponds to M17 supplemented with 200 mM MOPS, instead of β-glycerophosphate, and 0.25% glucose and adjusted to pH 7.2 with NaOH. This medium was used to maintain the culture at a pH above 6.8 throughout growth. Stress tests were performed with logarithmic cultures unless otherwise specified. Routinely, dilutions of a resuspended fresh colony were used to inoculate overnight cultures in M17/7. The overnight culture with an OD600 below 0.3 (pH above 6.9) was diluted to a final OD6000.02 in fresh M17/7 medium and incubated at 30°C until OD6000.1. Stress challenge experiments were performed under such conditions.

    For solidified media, agar was added at 15 g l−1. The following components were added as needed: 2.5 μg ml−1 Em, 20 μmol ml−1 potassium phosphate, 15 μg ml−1 guanine or hypoxanthine and 30 μg ml−1 guanosine. Decoyinine, serine hydroxamate or α-methylglucoside were added to cultures at 400 μg ml−1, 0.5 mM and 1% (w/v) respectively.

    Mutagenesis and mutant selection

    Two independent mutagenesis experiments were performed as has been described previously (Maguin et al., 1996), with MG1363 containing pG+host9:ISS1. Cells were grown overnight at 30°C in M17/7 with Em. Saturated cultures were diluted 100-fold (mutagenesis 1) or 1000-fold (mutagenesis 2) in M17/7 without selection, incubated for 150 min at 30°C and then shifted to 37.5°C for 150 min. Samples were then diluted and plated at 37.5°C. Mutant selection was performed on M17 acidified plates, i.e. M17 without β-glycerophosphate, containing 0.5% glucose, 15 g l−1 agar and acidified to pH 5.5 or 5.0 with HCl (indicated in the text as M17/pH value).

    In mutagenesis 1, the culture was glucose starved before plating (OD600nm2 with a final pH of 6.86), and transposition frequency was ≈8 × 10−2 EmR per total cells plated. Acid-tolerant mutants were selected on M17/5 plates with a frequency of about 3.3 × 10−3 per mutant. In the other mutagenesis, the culture was not saturated before selection (OD600nm0.76 and pH 7.06). The transposition frequency was around 10−1 EmR per total cells plated and the frequency of clones growing on M17/5.5 plates was 2.3 × 10−3 per mutant.

    Characterization of transposition targets

    The transposition event results in integration of the entire plasmid, flanked by duplicate copies of ISS1 (Maguin et al., 1996). The ISS1 flanking sequences were recovered by cloning in the EC101 strain as has been described previously (Duwat et al., 1997), and proximal HindIII and EcoRI junctions were sequenced, using primers CTACTGAGATTAAGGTC-3′ and ATAGTTCATTGATATATC-3′ respectively. Accession numbers are as follows: AF188108 for the gln operon; AF184611 for arl15 (pstS ); AF184610 for arl16 (pstB ); AF188107 for arl18 (relA); AF188105 for arl23 ; AF188102 for arl8 and arl10 (yybT ); AF184608 for arl1 ; AF188103 forarl2 ; AF188106 for arl4 ; AF188109 for arl5 ; and AF188104 for arl7.

    Excision of the integrated plasmid from ISS1 mutants

    Before studying a mutant at 30°C, the pGh9:IS plasmid present in the transposed structure was excised as has been described previously (Maguin et al., 1996). The resulting clones contain a single ISS1 sequence at the mutated locus. The absence of any other chromosomal rearrangements involving ISS1 was checked using HindIII and HincII digests and Southern hybridization in each case. Furthermore, the plating efficiencies of initial and corresponding excised mutants at 37.5°C on M17/5.5 and on M17 plates were essentially identical (data not shown).

    Single crossing-over inactivation

    Polymerase chain reaction (PCR)-amplified fragments of the recN, pstS, hpt, relA and guaA genes were cloned in plasmid pRV300 (Leloup and others, 1997) and introduced into MG1363. Single cross-over integrations were verified using Southern hybridization. recN : primers TTGTCGCTCTGTCTGTCG-3′ and CAGCAGTTTTGTCAGCAAG-3′ were used to amplify a DNA fragment encoding amino acids 102–230. pstS : Primers GGAAACTGCAGGGTCACAGCTGGTGGCTCAACTGC-3′ and CAAATCTAGACACGCGTTCCTGAACCAGCAG-3′ were used to amplify a DNA fragment encoding amino acids 28–157. hpt : primers CTTAGGAATTCTTCGCGG-3′ and CCGACATATGGAAGATTACG-3′ were used to amplify a DNA fragment encoding amino acids 43–174. relA : primers TGAATTAGTCGAATTGCGCG-3′ and CTTAGGAATTCTTCGCGG-3′ were used to amplify a DNA fragment encoding amino acids 367–495. guaA ; positions of primers are as in GenBank sequence (AF58326); forward primer positions 870 to 889 and reverse primer positions 1724 to 1703.

    DNA manipulations

    Cloning, plasmid and chromosomal DNA extractions, restriction analysis and Southern hybridization (Projan et al., 1983; Langella and Chopin, 1989; Sambrook et al., 1989), as well as E. coli transformation (Hanahan, 1985) and L. lactis electroporation (Holo and Nes, 1989) were performed according to published methods.

    Tolerance to a lethal acid shift

    Tests were performed on exponentially growing cells (as described above, OD600 0.1 and pH ≈7) or early stationary cells (2 h after cessation of growth in M17/7; OD6002 and pH 6.8). Cultures were harvested using centrifugation and resuspended in M17 without β-glycerophosphate, containing 0.5% glucose and acidified to pH 3.7 or 3.0 with HCl (indicated in the text as M17/pH value). Exponentially growing cells were shifted to M17/3.7, and viable cell counts were estimated by plating on M17 at time 0 (T0) and after 1 h (T1). In tests using early stationary phase cultures, viable cell counts were determined by plating on M17 at either time 0 (T0) or 2 h (T1) after transfer to M17/3.0. In each test, strain survival was calculated as the ratio of viable cell counts at T1 to that at T0, and relative survival corresponds to the ratio of the mutated strain survival compared with MG1363 survival.

    Long-term survival

    Tolerance to glucose starvation at pH 4.3 was performed using M17 containing 40 mM MOPS (instead of β-glycerophosphate), 0.5% glucose and adjusted to pH 7.2 with NaOH (M17/7-4.3). Cell survival was followed by plating on M17, 2 h (T0) and 48 h (T1) after growth arrest. To test for tolerance to glucose starvation at neutral pH, cells were grown until saturation in M17/7. Survival was evaluated by plating on M17 2 h (T0) and 96 h (T1) after the growth arrest. Strain survival and relative survival were calculated as in the previous section.

    Oxidative stress

    Cells were grown to logarithmic phase (OD6000.1) in M17/7 depleted of beef extract. H2O2 was added to cultures at OD6000.1, at a final concentration of 1 mM. Viability was estimated by plating on M17 at 30°C just before (T0) or 30 min (T1) after H2O2 addition. Strain survival and relative survival were calculated as described above.

    Heat shock

    A logarithmic phase culture (OD6000.1) was transferred to 55°C for 5 min or adapted for 15 min at 37°C before transfer to 55°C for 15 min. Cell survival was measured by plating on M17 at 30°C before (T0) and after (T1) incubation at 55°C. Strain survival and relative survival were calculated as described above.

    (p)ppGpp measurements

    Strains were grown in SA media (Jensen and Hammer, 1993) at low phosphate concentration (30 mM final) to exponential phase. At OD6000.05, 150 μCi ml−1 [32P]-H3PO4 (ICN) was added, and cells were grown for two doublings at 30°C. After the addition of one volume of 13 M formic acid, cells were lysed by three freeze–thaw cycles. Samples were centrifuged (13 000 r.p.m. for 5 min) and 10 μl of supernatant was spotted on a polyethyleneimine-cellulose plate (Machery-Nagel) for separation by thin-layer chromatography in 1.5 M KH2PO4 of the phosphorylated guanosine nucleotides.

    Tolerance to a lethal acid shift with stringent response inducers

    Tests were performed on cells in exponential phase grown in SA medium buffered with 120 mM MOPS (Jensen and Hammer, 1993; OD600 0.1 and pH ≈7). When appropriate, serine hydroxamate (SH) or α-methylglucoside (MG) were added to cultures at 0.5 mM and 1% (w/v) respectively, 5 min before the acid challenge. Cultures were harvested using centrifugation and resuspended in SA, without MOPS, adjusted with HCl to pH 3.5, with or without SH and MG. Viable cell counts were estimated by plating on M17 at time 0 (T0) and after 1 h (T1) Strain survival was calculated as the ratio of viable cell counts at T1 to that at T0, and relative survival corresponds to the ratio of the mutated strain survival compared with MG1363 survival.

    Note added in proof

    The ahrC mutant (Rallu et al., 1996) was not confirmed.

    Acknowledgements

    We thank J. Foster, K. Hammer and our team colleagues for their interest in this work and their helpful comments, M. Champommier-Vergès for discussion and helpful technical assistance. We are grateful to S. Sourice, L. Hissler, V. Lehman, B. Quinquis and R. Dervyn for their technical help with DNA sequencing. This work was financed in part by the BIO4-CT96-0498 programme.

    Ancillary

    Advertisement