Mechanisms of Acetoin Toxicity and Adaptive Responses in an Acetoin-Producing Species, Lactococcus lactis

We tested for toxicity of glucose catabolic products (acetoin, lactate, acetate) on L. lactis strain MG1363 (15) (Table 1). Among them, acetoin had the strongest inhibitory effect (data not shown). In M17Glucose 1% medium, addition of acetoin (70–200 mM) decreased final biomass yields in a dose-dependent manner after either fermentation (static or aerated liquid medium) or respiration growth (aerated liquid medium supplemented with 5 μM haem) (Fig. S1). To understand the mechanisms underlying acetoin toxicity, we first tested its effects on a ΔrecA mutant. RecA is a main DNA recombination and repair protein that contributes to bacterial protection survival upon DNA damage (16). Compared to growth in non-supplemented medium, addition of 0.2 M acetoin (∼ 6-fold higher of acetoin production under respiration in laboratory M17 broth (4–6)) led to a slight reduction in growth rate of the wild-type (WT) strain compared to a marked reduction of the ΔrecA strain (Fig. 2); these differences were accentuated upon plating (3-log and 6-log decreases of WT and ΔrecA strains, respectively; Fig. S2A). We suspected that acetoin is mutagenic. To test this hypothesis, we compared mutation frequencies of the WT strain with and without acetoin using a rifampin assay. The number of rifampin resistant clones was ∼ 15-fold higher in the presence of 0.2 M acetoin than in its absence (Fig. 3A). Acetoin was also found to inhibit DNA synthesis using a PCR assay (Fig. 3B). We observed a net decrease of polymerization efficiency when the enzyme was exposed to physiological concentrations of acetoin (L. lactis synthesized up to 35 mM acetoin in respiration growth from 1% glucose). At 33 mM acetoin we detected no PCR product. As Taq polymerase was sensitive to acetoin we tested effects of acetoin on a ΔclpP mutant. ClpP is an intracellular housekeeping protease that is required for protein turnover and stability (17). As observed for a ΔrecA mutant, a ΔclpP mutant was more sensitive to 0.2 M acetoin than in the WT strain (Fig. S2B and S3). Altogether, these observations suggest that acetoin leads to mutations and inhibition of protein stability and/or activity including that of DNA polymerase.

To investigate the mechanism of acetoin toxicity, we compared sensitivity of a ΔrecA mutant to molecules having a similar structure to acetoin: diacetyl, 2,3-butanediol, and methylglyoxal (MG) (Fig. 4A). Like acetoin, diacetyl and 2,3-butanediol are used as flavor additives, whereas MG is a well-known toxic by-product of carbon metabolism in bacteria (18). The reduced form of acetoin (2,3-butanediol) was not toxic, in contrast to its oxidized form (diacetyl), which arrested growth (Fig. 4B). Moreover, diacetyl completely inhibited ΔrecA strain growth even at very low concentrations, as observed with MG (0.5 mM, data not shown). Using the PCR assay, DNA polymerization efficiency was not affected by the presence of 2,3-butanediol (Fig. 4C). In contrast, no PCR product was detected when the enzyme was exposed to diacetyl or MG. Acetoin treatment led to intermediate effects (Fig. 4B and C). Significant effects of these compounds were also observed in a ΔclpP mutant (Fig. S3). Based on these data, we conclude that the keto group is responsible for acetoin toxicity.

To gain insight on the consequences of acetoin on L. lactis protein expression, we performed proteomic analyses in cells non-treated or treated with acetoin (Table 2, Fig. S4; see Materials and Methods). Compared to the control condition, acetoin treatment resulted in decreased detection of ∼30 proteins belonging to main categories: fatty acid, stress, DNA and translation. No over-produced proteins were detected upon acetoin treatment. The observation that DNA metabolism and translation were affected might be related to acetoin sensitivity of the ΔrecA and ΔclpP mutants and the DNA polymerase. Three fatty acid synthesis (FASII) proteins were decreased: AccC, a subunit of the ACC FASII initiation complex involved in malonyl-CoA synthesis from acetyl-CoA, and FASII elongation enzymes FabF (3-oxoacyl-acyl carrier protein synthase) and FabZ (3-hydroxy-acyl-[acyl-carrier-protein] dehydratase). The decrease of these protein amounts might suggest that acetoin affects membrane fatty acid composition.

The observed decrease in FASII enzyme amounts in acetoin-treated cells, and the fact that acetoin is membrane-miscible led us to analyze how this molecule affects membrane fatty acid composition (Table 2). To test this, cells grown without or with 0.2 M acetoin were collected during exponential growth phase (OD₆₀₀ = 0.5), and fatty acids were analyzed by gas chromatography. Non-treated cells produced mainly 4 fatty acids (FAs): myristic acid (C14:0, 10%), palmitic acid (C16:0, 26.5%), oleic acid (C18:1, 36%), and cyclic methylenoctadecenoic acid (cycC19:0, 22%) (Fig. 5A). Addition of acetoin led to a marked increase in the proportion of cycC19:0 FA, to the detriment of its precursor C18:1 FA (2-fold increase and >3-fold decrease, respectively).

CycC19:0 FA is produced from C18:1 FA via the cfa gene product, a cyclopropane FA synthase (19, 20). To determine the role of C18:1/cycC19:0 shift in acetoin stress resistance, we compared bacterial survival between the WT strain and a Δcfa mutant treated or not with 0.35 M acetoin (Fig. 5B). These strains grew similarly in the absence of treatment. In contrast, acetoin sensitivity was exacerbated in the cfa deletion strain (30-fold greater acetoin sensitivity than the WT strain). We conclude that C18:1/cycC19:0 shift contributes to acetoin resistance in L. lactis.

Our results indicated that L. lactis growth and survival may be limited by accrued acetoin production, particularly in respiration conditions where acetoin concentrations reach high levels. One way to overcome toxicity is to select for acetoin-resistant mutants. To this purpose, and to identify functions involved, we conducted transposon mutagenesis in L. lactis strain MG1363 (8). We constructed a mutant library and screened cells growing in the presence of the minimum lethal acetoin concentration, which was determined to be 0.45 M under respiration conditions at 30°C on agar plates (see Materials and Methods). After 48 h of incubation at 30°C, colonies appeared at a frequency of ∼10⁻⁶. Remarkably, out of 5 selected stable colonies, all transposon insertion sites mapped to the pstABCDEF operon (llmg_1896 to llmg_1901 (15)), which encodes a high affinity phosphate transport complex (Fig. 6A and B) (8). This selection was also performed on L. lactis but growing under aerobic fermentation (agar medium with no haem addition, with a minimum lethal acetoin concentration of 0.65 M). Remarkably, all transposon insertion sites mapped again to the same operon (in 8 mutant colonies tested), indicating that energy mode did not impact the functions giving rise to acetoin resistance. A pstA mutant, previously constructed in our lab (8), was ∼100-fold more resistant to acetoin than the WT strain (Fig. 6C). As expected, phosphate supplementation (20 mM) reversed acetoin resistance of the mutant, while it did not affect the WT strain (data not shown).

We determined above that acetoin affected L lactis on three levels: DNA damage repair, protein turnover, and membrane lipid function, seen respectively as growth sensitivity of the ΔrecA, ΔclpP. and Δcfa mutants (Fig. 2, Fig. S2 and S5). We asked whether pst inactivation rescues survival to an acetoin challenge by constructing the respective double mutants (pstE-ΔrecA, pstA-ΔclpP, or pstA-Δcfa). The pstA mutation partially rescued both ΔclpP and Δcfa mutants from acetoin toxicity; however, it did not rescue the ΔrecA strain (Fig. 7).

To obtain insight into acetoin resistance in the pstA mutant we performed a comparative (phospho)-proteome analysis of the WT and pstA strains treated or not with acetoin (Table 2, Fig. 8, Fig. S4, and Tables S1, S2, and S3). Although we observed differences in some protein amounts and phosphorylation states between the two strains, we failed to identify specific proteins or metabolic pathways that would explain acetoin resistance in the pstA mutant. Furthermore, fatty acid analysis showed that the pstA mutant behaved like the WT strain in response to acetoin stress (Fig. S5). These results suggest that pstA rescue affects cell physiology via mechanisms that may not directly affect the interactions of acetoin with its targets. This conclusion is supported by the fact that pst mutations arose in independent selections, including dithiothreitol, acidification, and tellurite stress, or in a DNA repair defective strain, combined with high temperature (8, 21, 22).

Strains and plasmids are described in Table 1. L. lactis strains were grown in reconstituted M17 broth supplemented with 0.5 or 1% glucose and riboflavin 5 μM, referred to as M17Glucose. Fresh medium was inoculated at OD₆₀₀ = 0.025 from overnight precultures, and cells were grown under static, aeration (shaking, 200 rpm), or respiration (shaking, 200 rpm; 5 μM haem in broth) conditions. Cultures were usually incubated at 30°C and harvested at the indicated cell densities (OD₆₀₀). For assays of acetoin sensitivity, acetoin at the indicated concentrations is added with inoculum and growths were performed in microplates. Cell densities were measured using a plate reader (Sunrise, TECAN), or otherwise as specified in text. When appropriate, erythromycin (Ery) was added at 1 μg ml⁻¹.

Cells were grown in liquid M17Glu1% in the presence of 0.2 M acetoin or not. After overnight growth, cultures were diluted in M17 broth and 0.1 ml of each dilution was spread on solid growth medium supplemented or not with 50 μg ml⁻¹ rifampin. Bacterial counts were determined after 48 h of incubation.

A fragment (1.7 kb of size) was PCR amplified with primer pairs: for 5’CCGGAATTCTGGTTCGCTTCAATTGATCG3’, Rev 5’CCGCTCGAGTAATCTAAAGACCATTATACC3’, and L. lactis MG1363 chromosome as a matrix in a reaction mixture containing 1X buffer, 0.25 mM each dNTP, 1 μM each primer, and 1 unit of Phusion DNA polymerase (New England Biolabs). PCR was run for 20 cycles.

Insertional mutagenesis was performed using MG1363 carrying the thermosensitive plasmid pGhost9::ISS1 as described (36). Cell dilutions were plated on M17Glu agar plates containing Ery 1 μg ml⁻¹ and incubated at 37°C for 48 h. Clones (about 10,000) were scraped from plates and resuspended in M17 medium with glycerol 15% and stocked at –80°C. The cells were screened on plates on M17Glu containing the minimum lethal concentration of acetoin, (defined as no cell growth for 72 h): 0.65 M acetoin in aerobic fermentation conditions, and 0.45 M acetoin in respiration permissive conditions in medium containing 10 μM haem. Plates were incubated at 30°C. Colonies that appeared at 24 and 48 h were streaked on M17Glu plus acetoin 0.45 or 0.65 M for fermentation or respiration growth and incubated at 30°C before identification of ISS1 transposon insertion sites.

Mutant chromosomes were purified, digested by SspI, and ligated by T4 ligase. The circulated products containing ISS1 transposon were PCR amplified using primers (ISS1-pEcoRI, ISS1-pHindIII, Table 1) and sequenced (Eurofins & GATC society). Sequencing results were blasted against the genome of L. lactis MG1363 (37) to identify transposon insertion site.

Overnight cultures of L. lactis strains were diluted 1/100 in fresh M17Glu0.5% broth in static conditions. At OD₆₀₀ = 0.1, acetoin 0.2 M was added. At OD₆₀₀ = 0.5, cells were harvested by centrifugation at 4°C, and resuspended in 30 mM Tris pH 7, DTT 1 mM. Cells were disrupted with glass beads (FastPrep FP120; MP Biomedical) by three cycles of 45 sec at a speed of 6 m s⁻¹. Extracts were centrifuged at 8,000 rpm for 20 min at 4°C. Crude extracts were ultracentrifuged for 1 h at 10°C to separate membrane and cytosolic fractions. Protein concentrations were measured according to the Bradford procedure with bovine serum albumin as the standard (Bio-Rad, France).

Five hundred μg cytosolic protein samples were used for analyses. They were treated with nuclease (purified NucA from Streptococcus agalactiae (38)) for 30 min at 37°C, and 2D gels were performed as described (39). Phosphoproteins were stained by Pro-Q Diamond dye (Invitrogen, France), which detects P-tyrosine, P-serine, and P-threonine, and visualized with Chemidoc MP Imaging System (Bio-Rad, France) with a UV standard filter. Gels were then washed twice in water (20 min, at room temperature) and stained by Instantblue dye (Expedeon, United Kingdom). Two independent experiments were performed. Spots selected by visual inspection were excised from gels for mass spectrometry identification (see the supplemental material).

Fatty acid extractions were performed on cells grown to OD₆₀₀ = 0.5 ± 0.05. Acetoin (0.2 M) was added when indicated at OD₆₀₀ = 0.1. Membrane fatty acid extraction and analysis were performed as described (40).

We proceeded as follows. Two DNA fragments covering the upstream (cfa For × cfa intRev) and downstream (cfa intFor × cfa Rev) regions of the cfa gene (Table 1) were PCR-amplified and then fused by a second PCR. The resulting fragment was ligated to pBR322-pGhost8 into EcoRI and BamHI sites. The modified plasmid was established in E. coli strain DH5α and transferred to L. lactis strain MG1363. Deletion by temperature shift was performed as described (8) and checked by PCR with primer pair outside of cfa locus.