Structure–function relationships of the antibacterial activity of phenolic acids and their metabolism by lactic acid bacteria

Authors


Michael Gänzle, Department of Agricultural, Food and Nutritional Science, University of Alberta, 4-10 Ag/For Centre, Edmonton, AB, Canada T6G 2P5. E-mail: mgaenzle@ualberta.ca

Abstract

Aims:  To determine structure–function relationships of antibacterial phenolic acids and their metabolites produced by lactic acid bacteria (LAB).

Methods and Results:  Minimum inhibitory concentrations (MICs) of 6 hydroxybenzoic and 6 hydroxycinnamic acids were determined with Lactobacillus plantarum, Lactobacillus hammesii, Escherichia coli and Bacillus subtilis as indicator strains. The antibacterial activity of phenolic acids increased at lower pH. A decreasing number of hydroxyl groups enhanced the activity of hydroxybenzoic acids, but had minor effects on hydroxycinnamic acids. Substitution of hydroxyl groups with methoxy groups increased the activity of hydroxybenzoic, but not of hydroxycinnamic, acid.

Metabolism of chlorogenic, caffeic, p-coumaric, ferulic, protocatechuic or p-hydroxybenzoic acids by L.  plantarum, L. hammesii, Lactobacillus fermentum and Lactobacillus reuteri was analysed by LC-DAD-MS. Furthermore, MICs of substrates and metabolites were compared. Decarboxylated and/or reduced metabolites of phenolic acids had a lower activity than the substrates. Strain-specific metabolism of phenolic acids generally corresponded to resistance.

Conclusions:  The influence of lipophilicity on the antibacterial activity of hydroxybenzoic acids is stronger than that of hydroxycinnamic acids. Metabolism of phenolic acids by LAB detoxifies phenolic acids.

Significance and Impact of the Study:  Results allow the targeted selection of plant extracts for food preservation, and selection of starter cultures for fermented products.

Introduction

Phenolic compounds are secondary plant metabolites that possess aromatic rings with one or more hydroxyl or methoxy groups. Phenolic acids are a subclass of phenolic compounds, encompassing hydroxybenzoic acids (C6–C1 structures) and hydroxycinnamic acids (C6–C3 structures) (Schieber and Aranda Saldana 2009; Dai and Mumper 2010). Phenolic acids have antimicrobial activity and hold promise for application as preservatives in food and food-packing materials. Phenolic acids or plant extracts containing these compounds gave satisfactory results when added to beef and food-packing materials such as hand sheets and pea starch (Ejechi and Akpomedaye 2005; Elegir et al. 2008; Corrales et al. 2009).

The antimicrobial activity of phenolic acids is determined by their chemical structure, in particular the number and position of substitution in the benzene ring, and the saturated chain length (Cueva et al. 2010). Phenolic acids had lower antimicrobial activity compared with their butyl and methyl esters (Cueva et al. 2010). The antimicrobial effect increased with increasing length of the alkyl chain (Merkl et al. 2010). Oligomers show higher activity than the corresponding phenolic acid monomers (Elegir et al. 2008). Hydroxybenzoic and hydroxycinnamic acids occurring in plants exhibit diversity with respect to the number of hydroxyl or methoxy groups (Fig. 1); however, current knowledge on structure–function relationships of the antimicrobial activity of phenolic acids does not account for this diversity of compounds. Moreover, additional derivatives of these compounds are produced from bacterial metabolism (Fig. 1).

Figure 1.

 Hydroxybenzoic and hydroxycinnamic acids and their decarboxylated and reduced metabolites used in this study.

Some lactic acid bacteria (LAB) such as Lactobacillus brevis, Lactobacillus fermentum and Lactobacillus plantarum metabolize phenolic acids by decarboxylation and/or reduction. The products of phenolic acid decarboxylases are vinylcatechol, vinylphenol, vinylguaiacol, pyrogallol and catechol; reduction of hydroxycinnamic acids yields dihydrocaffeic and dihydroferulic acids (Van Beek and Priest 2000; Curiel et al. 2010; Svensson et al. 2010) (Fig. 1). Lactobacilli that are capable of phenolic acid metabolism were isolated predominantly from fermented products with a high content of phenolic compounds such as olives, whisky, wine and sorghum (Rozes and Peres 1998; Van Beek and Priest 2000; Campos et al. 2009; Svensson et al. 2010).

Lactobacillus spp. are more resistant to phenolic compounds when compared to other groups of bacteria such as Clostridium spp. and Bacteroides spp. (Lee et al. 2006). The tolerance of lactobacilli to phenolic acids and their ability to metabolize phenolic acids are strain or species specific (Van Beek and Priest 2000; Cueva et al. 2010; Curiel et al. 2010; Svensson et al. 2010). However, the strain-specific tolerance of LAB to phenolic acids has not been related to the metabolism of phenolic acids. Moreover, the antibacterial activity of the products of phenolic acid metabolism by LAB remains unknown. Knowledge on structure–function relationships of phenolic acids is important for application of these compounds as food preservatives as well as the selection of starter cultures for food fermentations. Therefore, this study aimed to elucidate the structure–function relationships of phenolic acids with model organism isolated from food. Moreover, it was determined whether phenolic acid metabolism by LAB is a mechanism of detoxification of noxious compounds that LAB encounter in their natural habitat.

Materials and Methods

Chemicals

Caffeic, p-coumaric, ferulic, gallic, protocatechuic and p-hydroxybenzoic acids were obtained from Extrasynthèse (Genay, France). Chlorogenic, cinnamic, syringic, benzoic, dihydrocaffeic and phloretic acids and catechol were purchased from Sigma (St Louis, MO, USA). MRS and LB media were obtained from BD (Mississauga, ON, Canada).

Bacterial strains and culture conditions

The following strains and incubation conditions were used to evaluate the antibacterial activity of phenolic acids: L. plantarum TMW 1.460 (MRS medium, 30°C, microaerophilic conditions), isolated from spoiled beer (Ulmer et al. 2000); Lactobacillus hammesii DSM 16381 (MRS medium, 30°C, microaerophilic conditions), isolated from sourdough (Valcheva et al. 2006); L. fermentum FUA 3165 and Lactobacillus reuteri FUA 3168 (MRS medium, 34°C, microaerophilic conditions), both isolated from fermented sorghum (Sekwati-Monang et al. 2011); Bacillus subtilis FAD 110 (LB medium, 37°C, aerobic conditions), isolated from ropy bread (Röcken and Spicher 1993); and Escherichia coli AW1.7 (LB medium, 37°C, aerobic conditions), isolated from beef (Dlusskaya et al. 2011).

Antibacterial activity

The minimum inhibitory concentration (MIC) of phenolic acids was determined by a critical dilution assay as described by Parente et al. (1995) and Gänzle et al. (1996). Stock solutions of each phenolic acid (20 g l−1) were prepared in 50% methanol (v/v) and 50% phosphate buffer (50 mmol l−1, pH 6·5). Caffeic acid was dissolved in 100% methanol. Media and stock solutions were adjusted to pH 4·0, 5·5 and 7·0 with 2 N HCl or NaOH to evaluate the effect of pH on the activity of phenolic acids. The stock solutions (100 μl) and media (100 μl of LB or MRS) were mixed, and twofold serial dilutions of the phenolic acids were prepared on microtitre plates (Parente et al. 1995; Gänzle et al. 1996). Solvents were evaporated under a sterile flow of air.

Indicator strains were subcultured twice in liquid media under conditions listed earlier. Media were inoculated with 10% of stationary cultures obtained after overnight incubation, and 50 μl of inoculated media was added to microtitre plates. This procedure resulted in phenolic acid concentrations in growth media ranging from 6·7 g l−1 to 6·5 mg l−1 and an initial cell count of indicator strains of about 107 CFU ml−1. Microtitre plates were incubated for 24 h. After incubation, bacterial growth was determined by optical density at 630 nm. MIC was defined as the lowest concentration of phenolic acids inhibiting bacterial growth (Gänzle et al. 1996). MICs are expressed as means ± standard deviations of at least two independent experiments. For evaluation of structure–function relationships of antibacterial phenolic acids, MICs are reported as means ± standard deviations of four independent experiments and significance was determined by Student’s t-test using sigmaplot software (Systat Software, Inc., San Jose, CA, USA), where P was <0·05.

Metabolism of phenolic acids by LAB strains

Modified MRS media (Svensson et al. 2010) were supplemented with caffeic, ferulic, p-coumaric, protocatechuic or p-hydroxybenzoic acids at a concentration of 1 mmol l−1 to match conditions employed in previous investigations related to phenolic acid metabolism by LAB (Svensson et al. 2010). Chlorogenic acid was added at a concentration of 1·5 mmol l−1. Media were inoculated with 5% overnight cultures of L. plantarum, L. hammesii, L. fermentum or L. reuteri and incubated for 24 h at 30 or 34°C. Sterile media containing the corresponding phenolic acids served as a control. After incubation, cells were removed by centrifugation, and the supernatant was acidified with hydrochloric acid to pH 1·5. Liquid–liquid extraction with ethyl acetate was carried out three times. The extracts were analysed by LC-DAD-MS using a Shimadzu UFLC system with a Kinetex PFP column (100 × 3·0 mm, 2·6 μm) and a SPD-M20A Prominence diode array detector. The mobile-phase solvents consisted of (A) 0·1% (v/v) formic acid in water and (B) 0·1% formic acid in water/acetonitrile (10:90, v/v). The gradient was as follows: 0–20% B (1·5 min), 20% B (4·5 min), 20–90% B (7·5 min), 90% B (8 min) and 90–0% B (14 min). The injection volume was 5 μl, and the flow rate was 0·9 ml min−1. The analytes were quantified at 280 nm. Bacterial conversion of phenolic acids during incubation was calculated as percentage of the initial substrate concentration. Metabolism is expressed as means ± standard deviation of duplicate independent experiments analysed in duplicate.

Results

Effect of pH on the antibacterial activity of phenolic acids

To evaluate the effect of pH on the antimicrobial activity of phenolic acids, the MICs of five phenolic acids (chlorogenic, protocatechuic, caffeic, ferulic and p-coumaric acids) were determined at pH 4·0, 5·5 and 7·0. Two strains of LAB, L. plantarum and L. hammesii, were used as indicator strains. B. subtilis and E. coli did not grow at a pH of 4·0 independent of the presence of phenolic acids (data not shown). The MIC of chlorogenic acid was higher than the highest concentration used, 6·7 g l−1. The antimicrobial activity of phenolic acids increased at lower pH (Fig. 2). This effect was more pronounced for L. plantarum than for L. hammesii.

Figure 2.

 MIC of protocatechuic acid (○), caffeic acid (•), ferulic acid (△) and p-coumaric acid (bsl00001) against Lactobacillus plantarum (a) and Lactobacillus hammesii (b) at different pH values. Symbols on the upper axis indicate a MIC that is higher than the highest concentration used (6·7 g l−1).

Structure–function relationship of phenolic acids: Effect of number of hydroxyl (–OH) and methoxy (–OCH3) groups in the aromatic ring

The effect of hydroxyl (–OH) and methoxy (–OCH3) groups on the antimicrobial activity of phenolic acids was studied by comparison of the MICs of benzoic acid, cinnamic acid, hydroxybenzoic acids (p-hydroxybenzoic, protocatechuic, gallic and syringic acids) and hydroxycinnamic acids (p-coumaric, caffeic and ferulic acids) against L. plantarum and L. hammesii as well as E. coli and B. subtilis. The antimicrobial activity of hydroxycinnamic acids was comparable or higher than that of hydroxybenzoic acids with same number of hydroxyl groups (Tables 1 and 2). The antimicrobial activity of hydroxybenzoic acids decreased significantly with increasing number of hydroxyl groups (Table 1). The activity of gallic acid was approximately two- to tenfold lower compared to other hydroxybenzoic acids. In comparison, the effect of the number of hydroxyl groups on the antimicrobial effect of hydroxycinnamic acids was relatively minor (less than fourfold change in activity). Caffeic acid was only about half as active as cinnamic and p-coumaric acids (Table 2).

Table 1.   MICs of hydroxybenzoic acids according to the number of hydroxyl and methoxy groups in the aromatic ring
Number of –OH groupsNumber of –OCH3 groupsMIC (g l−1)
Lactobacillus plantarumLactobacillus hammesiiEscherichia coliBacillus subtilis
  1. MIC values in the same column with different superscripts differ significantly (P < 0·05).

 Hydroxybenzoic acids
001·45 ± 0·04A0·97 ± 0·65A0·07 ± 0·01A0·04 ± 0·02A
101·57 ± 0·15A1·12 ± 0·64A0·12 ± 0·08A0·13 ± 0·02B
203·87 ± 0·02C1·19 ± 0·41A0·31 ± 0·06B0·35 ± 0·01C,D
303·74 ± 0·10B4·56 ± 0·86B0·49 ± 0·30B0·64 ± 0·20D
 Methoxy-hydroxybenzoic acids
121·75 ± 0·42A1·15 ± 0·65A0·39 ± 0·08B0·26 ± 0·12B,C
Table 2.   MICs of hydroxycinnamic acids according to the number of hydroxyl and methoxy groups in the aromatic ring
Number of –OH groupsNumber of –OCH3 groupsMIC (g l−1)
Lactobacillus plantarumLactobacillus hammesiiEscherichia coliBacillus subtilis
  1. Values are mean ± standard deviation (n = 4).

  2. MIC values in the same column with different superscripts differ significantly (P < 0·05).

 Hydroxycinnamic acids
000·79 ± 0·11A0·86 ± 0·11A0·11 ± 0·01A0·07 ± 0·04A
101·21 ± 0·20B1·04 ± 0·63A0·12 ± 0·02A0·25 ± 0·13B
201·52 ± 0·14B,D0·63 ± 0·38A0·23 ± 0·12A0·30 ± 0·11B
 Methoxy-hydroxycinnamic acids
111·68 ± 0·14D0·89 ± 0·59A0·16 ± 0·08A0·38 ± 0·25A,B

The antimicrobial activity of syringic acid with one hydroxyl and two methoxy groups was higher than the activity of gallic acid but comparable to other hydroxybenzoic acids. Thus, the antimicrobial activity of hydroxybenzoic acids increased through substitution of a hydroxyl group with a methoxy group (Table 1). In contrast, the presence of methoxy groups did not significantly affect the antimicrobial activity of cinnamic acids (Table 2). The MICs of ferulic acid were comparable to MICs of other hydroxycinnamic acids, except for L. plantarum.

Metabolism of phenolic acids by lactobacilli

In comparison with E. coli and B. subtilis, lactobacilli were substantially more tolerant to phenolic acids. The metabolism of phenolic acids by lactobacilli was investigated to determine whether metabolic conversion detoxifies phenolic acids and thus contributes to tolerance. The concentration of 1 mmol l−1 matched conditions employed in previous investigations (Svensson et al. 2010) and corresponds to 0·13 to 0·2 g l−1, well below the MIC of these phenolic acids against lactobacilli (Tables 1 and 2). Metabolites of six phenolic acids (chlorogenic, caffeic, p-coumaric, ferulic, protocatechuic and p-hydroxybenzoic acids) by four strains (L. plantarum, L. hammesii, L. fermentum and L. reuteri) were quantified using LC-DAD-MS (Svensson et al. 2010). The metabolites were identified by comparing their mass spectra to literature data (Table 3). None of the strains metabolized p-hydroxybenzoic acid (data not shown). Lactobacillus plantarum consumed all phenolic acids except chlorogenic acid. L. hammesii consumed more than 80% of caffeic acid, p-coumaric acid and protocatechuic acid. Lactobacillus fermentum metabolized p-coumaric and ferulic acids, and L. reuteri metabolized only chlorogenic acid.

Table 3.   Metabolism of phenolic acids by lactic acid bacteria. Shown are the percentage of phenolic acids remaining after incubation and major metabolites produced during incubation
Phenolic acidsLactobacillus plantarumLactobacillus hammesiiLactobacillus fermentumLactobacillus reuteri
Rem. %MetabolitesRem.%MetabolitesRem. %MetabolitesRem. %Metabolites
  1. Values are mean ± standard deviation (n = 2) of the remaining amount of phenolic acids after 24 h of incubation with each LAB strain.

  2. *NI, not identified; ND, not detected.

Chlorogenic acid96 ± 7·4Caffeic acid98 ± 2·3Caffeic acid75 ± 1·0Caffeic acid64 ± 4·5Caffeic acid
Caffeic acid0·2 ± 0·07Vinyl catechol9 ± 1·5Vinyl catechol63 ± 26Dihydrocaffeic acid Vinyl catechol94 ± 16NI*
p-Coumaric acid0·2 ± 0·01p-Vinyl phenol16 ± 2·2p-Vinyl phenol1 ± 0·1Phloretic acid p-Vinyl phenol85 ± 10NI*
Ferulic acid1 ± 0·1Dihydro ferulic acid94 ± 7·2ND*1 ± 0·3Dihydroferulic acid86 ± 6·0NI*
Protocatechuic acid0·2 ± 0·06Catechol0·5 ± 0·05Catechol82 ± 4·2NI*91 ± 3·4ND*

Caffeic acid results from chlorogenic acid hydrolysis. Lactobacillus plantarum, L. hammesii and L. fermentum decarboxylated caffeic acid to vinylcatechol. Lactobacillus fermentum alternatively reduced caffeic acid to dihydrocaffeic acid. These compounds were identified based on mass spectra reported by Svensson et al. (2010). Comparable to the metabolism of p-coumaric acid by other lactobacilli (Van Beek and Priest 2000; Rodriguez et al. 2008), decarboxylation of p-coumaric acid by L. plantarum, L. hammesii and L. fermentum gave p-vinylphenol, which was identified by the [M–H] ion at m/z 119. Lactobacillus fermentum also reduced p-coumaric acid to phloretic acid, which was previously identified as a metabolite of p-coumaric acid (Barthelmebs et al. 2000). Lactobacillus plantarum and L. fermentum reduced ferulic acid to dihydroferulic acid as previously described by Svensson et al. (2010). Decarboxylation of protocatechuic acid by L. plantarum and L. hammesii generated catechol, which was identified by the [M–H] ion at m/z 109 (Albarran et al. 2010).

Antimicrobial activity of phenolic acids compared to their metabolites produced by LAB

The MIC of phenolic acids was compared to the MIC of their metabolites to determine whether metabolism reduces their antimicrobial activity. Caffeic and p-coumaric acids were compared with their reduced products, dihydrocaffeic and phloretic acids, respectively. Protocatechuic acid was compared with its decarboxylated product, catechol. Six bacterial strains (L. plantarum, L. hammesii, L. fermentum, L. reuteri, E. coli and B. subtilis) were used as indicator strains. The antimicrobial activity of reduced metabolites was two- to fivefold lower than that of the substrates, except for B. subtilis (Fig. 3). Decarboxylation of protocatechuic acid decreased the antimicrobial activity against L. fermentum, E. coli and B. subtilis, but did not alter the MIC against L. plantarum, L. hammesii and L. reuteri.

Figure 3.

 Comparison of the antibacterial activity of phenolic acids and their metabolites. (a) Effect of the reduction of the double bond of hydroxycinnamic acids. Shown is the MIC of caffeic acid (inline image) and p-coumaric acid (bsl00001) and the corresponding reduced metabolites dihydrocaffeic acid (bsl00020) and phloretic acid (bsl00023), respectively. (b) Effect of decarboxylation. Shown is the MIC of protocatechuic acid (inline image) and the decarboxylated metabolite, catechol (bsl00020).

Discussion

Despite the substantial body of literature related to the antibacterial activity of phenolic acids (Table 4), a systematic evaluation of the effect of hydroxyl and methoxy groups on the antibacterial activity of phenolic acids has not been reported. Moreover, different studies used different methodologies for the determination of antibacterial activity, and organisms of the same species or genus exhibit substantial strain-to-strain variation in the sensitivity to phenolic acids (Table 4). This study determined structure–function relationships of hydroxycinnamic and hydroxybenzoic acids by taking into account the effects of hydroxyl and methoxy groups as well as the contribution of carboxyl groups and the double bond in hydroxycinnamic acids. Our data generally agree with MICs of phenolic acids that were previously published (Table 4). Our results particularly confirm and extend for a large selection of structurally diverse phenolic acids that lactobacilli are more resistant against their antibacterial activity compared to E. coli, B. subtilis and other bacteria relevant in food (Table 4, Lee et al. 2006).

Table 4.   Antibacterial activity of phenolic acids
Phenolic acidLactobacillus spp.MIC (g l−1)References
Escherichia coliBacillus spp.
  1. *Benzoic and t-cinnamic acids were included in this study to investigate the effects of aromatic substitution on the antibacterial activity.

  2. †MIC corresponding to growth inhibition by 50% or more.

  Hydroxybenzoic acids 
Benzoic acid*1–1·80·500·05–0·12Chipley (2005) and Cueva et al. (2010)
p-Hydroxybenzoic acid0·125–13·80·34–0·550·40–0·69Campos et al. (2003), Tuncel and Nergiz (1993), Landete et al. (2008), Cueva et al. (2010) and Merkl et al. (2010)
Protocatechuic acid7·7–14·10·55–2·670·4–2·67Tuncel and Nergiz (1993), Taguri et al. (2006), Landete et al. (2008) and Merkl et al. (2010)
Gallic acid< 0·5†0·600·02–1·6Campos et al. (2003), Taguri et al. (2006) and Wansi et al. (2010)
Syringic acid4·95–4·990·550·40Tuncel and Nergiz (1993), Landete et al. (2008)
  Hydroxycinnamic acids 
Chlorogenic acid 0·10 Xia et al. (2010)
Caffeic acid0·5†–10·32–2·670·22–1·60Stead (1993), Wen et al. (2003), Lee et al. (2006), Taguri et al. (2006), Almajano et al. (2007) and Merkl et al. (2010)
p-Coumaric acid*0·5†–10·450·40Stead (1993) and Tuncel and Nergiz (1993)
t-Cinnamic acid7·41·33 Landete et al. (2008) and Rastogi et al. (2008)
Ferulic acid0·5†–10·45–1·940·40–1·94Stead (1993), Tuncel and Nergiz (1993) and Merkl et al. (2010)

In analogy to other weak organic acids, benzoic acid and hydroxybenzoic acids exert antimicrobial activity by diffusion of the undissociated acid across the membrane, resulting in acidification of the cytoplasm and, eventually, cell death (Herald and Davidson 1983; Ramos-Nino et al. 1996; Phan et al. 2002; Campos et al. 2009). Consequently, the pKa and the lipophilicity were proposed to determine the solubility of phenolic acids in bacterial membranes and thus their antimicrobial activity (Herald and Davidson 1983; Ramos-Nino et al. 1996; Campos et al. 2009). Hydroxybenzoic acids and hydroxycinnamic acids are weak organic acids but differ in their lipophilicity. Factors that affect the lipophilicity of phenolic acids include pH, which determines the charge of the carboxyl group, ring substitutions (hydroxyl and methoxy groups) and the saturation of the side chain of cinnamic acids.

A decrease in the pH increased the antibacterial activity of phenolic acids. The same trend has been reported for hydroxycinnamic acids and benzoic acid (Otto and Conn 1944; Herald and Davidson 1983; Wen et al. 2003; Almajano et al. 2007). The concentration of undissociated, more lipophilic phenolic acids increases with decreasing pH. The activity of undissociated phenolic acids is higher compared to dissociated phenolic acids, because they are more soluble in the cytoplasmic membrane (Ramos-Nino et al. 1996). However, our results demonstrate that dissociation of phenolic acids does not fully account for the effect of pH on their activity. The pH also had a strong effect on the MIC of phenolic acids when these were calculated on the basis of the undissociated portion (data not shown). Thus, dissociation is not the only factor responsible for their antibacterial activity.

The number of hydroxyl groups altered the antibacterial activity of hydroxybenzoic acids but did not affect the activity of hydroxycinnamic acids. Likewise, an increase in the lipophilicity by substitution of hydroxyl groups with methoxy groups increased the activity of hydroxybenzoic acids, but not of hydroxycinnamic acids. This result contrasts previous estimations that 80% of the antibacterial activity of phenolic acids is determined by their pKa and lipophilicity (Herald and Davidson 1983; Ramos-Nino et al. 1996). Hydroxycinnamic acids are more lipophilic than hydroxybenzoic acids because of their unsaturated chain (Campos et al. 2003). It is thus possible that their lipophilicity is less affected by substitutions of the aromatic ring. However, the double bond of the side chain, which is the main difference in the structure of hydroxybenzoic and hydroxycinnamic acids, likely contributes to the antibacterial activity of hydroxycinnamic acids.

The reduction of the double bond of hydroxycinnamic acids substantially decreased the antibacterial activity against LAB. This unexpected result further confirms that the double bond of hydroxycinnamic acids plays an important role in their mode of action. The reduction of the double bond, which strongly affects their antibacterial activity, has only a minor effect on the lipophilicity of the overall molecule. In contrast, the number of hydroxyl groups did not affect the antibacterial activity of hydroxycinnamic acids but has a more pronounced effect on the lipophilicity. Decarboxylation of protocatechuic acid decreased the antibacterial activity against some indicator strains (L. fermentum FUA3168, E. coli AW1.7 and B. subtilis FAD110). However, the MIC against other lactobacilli (L. plantarum TMW 1.460, L. hammesii DSM13681 and L. reuteri FUA3168) remained unchanged. The role of the carboxylic group in the activity of protocatechuic acid was thus not more pronounced than the role of hydroxyl groups.

LAB exhibit a strong strain-to-strain variation with respect to their tolerance to phenolic acids (Campos et al. 2003). Because the antibacterial activity of phenolic acid metabolites was generally lower when compared to the original substrates (Fig. 3), this variability likely relates to the strain-specific metabolism. In keeping with prior observations, lactobacilli metabolized phenolic acids by strain-specific decarboxylation and/or reduction (Van Beek and Priest 2000; De las Rivas et al. 2009; Svensson et al. 2010). Lactobacillus plantarum TMW 1.460 and L. fermentum FUA3165 produced decarboxylases and reductases. However, metabolism of L. fermentum FUA3165 differed from L. plantarum TMW 1.460 as the former strain also reduced caffeic and p-coumaric acids. L. hammesii DSM13681 only produced decarboxylases. Chlorogenic acid was hydrolysed by L. reuteri FUA3165 and L. fermentum FUA3165 to produce caffeic acid, indicating esterase activity of these two strains. Caffeic acid was also identified in supernatants of other strains, but the low conversion of chlorogenic acid by these strains may be attributable to factors other than enzyme activity. Among LAB, chlorogenic acid esterase activity was previously shown only for Lactobacillus gasseri (Coteau et al. 2001). Lactobacillus reuteri FUA3168 had the highest esterase activity among the strains tested in this work and converted more than 50% of the caffeic acid. This strain or its esterase could serve as suitable catalyst for the enzymatic conversion of chlorogenic acid and other phenolic acid esters for food and pharmaceutical purposes.

Although the assay systems for the determination of tolerance to phenolic acids and phenolic acid metabolism differed, L. plantarum TMW1.460, the strain with the highest metabolic activity towards phenolic acids, was also the most tolerant strain. Lactobacillus reuteri FUA3168 was the most sensitive among the four LAB strains tested in this work (Fig. 3; data not shown) and also exhibited the lowest metabolic activity towards phenolic acids. Lactobacillus fermentum FUA3165 and L. hammesii DSM13681 exhibited intermediate sensitivity and intermediate potential for the metabolism of phenolic acids. This relationship between metabolic capacity and sensitivity to phenolic acids further indicates that metabolism of phenolic acids by LAB contributes to their detoxification.

In conclusion, the antibacterial mode of action of hydroxybenzoic and hydroxycinnamic acids differs. The antibacterial activity of hydroxybenzoic acids decreases with an increasing number of hydroxyl groups and is thus primarily correlated with their lipophilicity. The antibacterial activity of hydroxycinnamic acids, particularly their activity against lactobacilli, depends to a much lesser extent on the substitutions of the aromatic ring with hydroxyl or methoxy groups but is strongly dependent on the double bond of the side chain. LAB metabolism of phenolic acids by decarboxylation and/or reduction thus likely is primarily a mechanism for detoxification of noxious compounds encountered by LAB in plant substrates. This knowledge on structure–function relationships of antibacterial phenolic acids facilitates the selection of phenolic acids or plant extracts containing phenolic acids for use as food preservatives as well as the selection of starter cultures for the fermentation of substrates that are rich in phenolic acids, such as sorghum or olives.

Acknowledgements

We would like to thank the Alberta Agriculture Funding Consortium, Alberta Innovates-BioSolutions, for financial support. Alma Fernanda Sánchez-Maldonado acknowledges support from the Mexican National Council for Science and Technology; Andreas Schieber and Michael Gänzle acknowledge support from the Canada Research Chairs Program.

Ancillary