Volatile-Mediated Killing of Arabidopsis thaliana by Bacteria Is Mainly Due to Hydrogen Cyanide

Abstract

The volatile-mediated impact of bacteria on plant growth is well documented, and contrasting effects have been reported ranging from 6-fold plant promotion to plant killing. However, very little is known about the identity of the compounds responsible for these effects or the mechanisms involved in plant growth alteration. We hypothesized that hydrogen cyanide (HCN) is a major factor accounting for the observed volatile-mediated toxicity of some strains. Using a collection of environmental and clinical strains differing in cyanogenesis, as well as a defined HCN-negative mutant, we demonstrate that bacterial HCN accounts to a significant extent for the deleterious effects observed when growing Arabidopsis thaliana in the presence of certain bacterial volatiles. The environmental strain Pseudomonas aeruginosa PUPa3 was less cyanogenic and less plant growth inhibiting than the clinical strain P. aeruginosa PAO1. Quorum-sensing deficient mutants of C. violaceum CV0, P. aeruginosa PAO1, and P. aeruginosa PUPa3 showed not only diminished HCN production but also strongly reduced volatile-mediated phytotoxicity. The double treatment of providing plants with reactive oxygen species scavenging compounds and overexpressing the alternative oxidase AOX1a led to a significant reduction of volatile-mediated toxicity. This indicates that oxidative stress is a key process in the physiological changes leading to plant death upon exposure to toxic bacterial volatiles.

Bacteria interact with plants in many different ways. Over the past few years, the significance of volatile compounds as mediators of bacterium-plant interactions has become increasingly evident (14, 18, 33, 46, 50, 54). Using compartmental petri dish assays with a bacterial culture on one side and Arabidopsis plants on the other, bacterial volatiles have been shown to either promote the growth of plants (14, 36) or to reduce it (16, 18, 43, 46). Despite the strong effects observed, very little is known at present about the identity of the compounds involved in this volatile-mediated impact of bacteria on plants. In addition to carbon dioxide (17), 2,3-butanediol and its precursor acetoin are, to our knowledge, the only organic volatiles which have been put forward as candidates to explain the beneficial effects of bacterial volatiles on plants (14, 36). Likewise, the compounds responsible for the “killing” effect of some bacterial strains on Arabidopsis thaliana are yet to be discovered.

In the course of a study to assess the volatile-mediated impact of a collection of rhizosphere bacterial isolates on the model plant A. thaliana, we observed contrasting effects, ranging from 6-fold growth promotion to plant killing. We noticed that the most virulent strains were known producers of hydrogen cyanide (e.g., Pseudomonas or Chromobacterium species). We therefore hypothesized that hydrogen cyanide (HCN), a potent inhibitor of cytochrome c oxidase and of other metal-containing enzymes, might be responsible for the observed plant-killing effects.

Few bacterial species are known to produce cyanide, and these are restricted to members of the genera Pseudomonas, Chromobacterium, and Rhizobium (5). Cyanide is usually produced at the end of exponential phase, when the oxygen concentration is reduced and cells have reached a density at which quorum sensing is activated. The involvement of quorum sensing in the control of cyanogenesis is not a general phenomenon but appears to be strain specific: in the extensively studied clinical strain Pseudomonas aeruginosa PAO1, both acyl-homoserine lactone (AHL)-based quorum-sensing systems (RhlI/R and LasI/R) are necessary for HCN production (31). Similarly, the disruption of quorum sensing in Chromobacterium violaceum CV0 was reported to abolish cyanogenesis (44). Conversely, quorum sensing does not seem to regulate cyanogenesis in Pseudomonas fluorescens 2P24, although other biocontrol properties are diminished when the gene encoding the AHL synthase is inactivated (48). Likewise, the biocontrol strain P. fluorescens CHA0 produces very high levels of HCN but does not possess any known AHL-based quorum-sensing system.

Although bacterial cyanogenesis is considered to be damaging to animals (10), cyanide production by rhizosphere bacteria is traditionally regarded as a plant-growth-promoting trait. This is mainly due to the antifungal activity cyanogenesis confers to P. fluorescens CHA0 (47) and other strains used in biocontrol against phytopathogenic fungi. However, early reports proposed that the observed phytotoxicity of P. aeruginosa strains might be due to HCN production (1, 3, 4, 9, 23, 37). More recently, Rudrappa et al. (35) analyzed the effect of cyanogenic pseudomonads on A. thaliana seedlings. These authors observed reduced root growth in the presence of two Pseudomonas strains (P. aeruginosa PAO1 and P. fluorescens CHA0) but normal growth in the presence of the respective HCN-defective mutants. This was a further indication that bacterial cyanogenesis might be a key factor responsible for the plant-killing effects observed by us and others (43, 46) when growing Arabidopsis plants in contact with bacterial volatiles.

To test this hypothesis, we used a collection of environmental strains including Chromobacterium, Serratia, and Pseudomonas species and assessed both HCN production and volatile-mediated effects on A. thaliana. In addition, we investigated the strain specificity of HCN production, and its consequences on plant growth. We compared various P. aeruginosa strains, including clinical strains and the environmental plant-growth-promoting strain PUPa3 (21). Since HCN production is reported to be controlled by quorum sensing in P. aeruginosa PAO1 and C. violaceum CV0, we hypothesized that the quorum-sensing-impaired mutants of these strains would show reduced volatile-mediated toxicity on A. thaliana. Moreover, we investigated the effect of quorum sensing in the control of cyanogenesis in C. violaceum CV0, in two P. aeruginosa strains, and in Pseudomonas chlororaphis ATCC 13985, a strain closely related to the biocontrol strain P. chlororaphis subsp. aureofaciens 30-84. To obtain first insights into the mode of action of the deleterious volatiles on plant growth, we tested the volatile-mediated effects of our cyanogenic strains on Arabidopsis mutants with altered expression levels of the alternative oxidase AOX1a (45).

MATERIALS AND METHODS

Plant accession numbers and bacterial strains.

The bacterial strains used in the present study are listed in Table 1. A. thaliana Col-0 was used for all plant experiments. Transgenic lines in which the AOX1a gene was overexpressed or silenced were obtained from the European Arabidopsis stock center NASC (Table 2).

TABLE 1.

Strains used in this study

Species	Strain name	Strain no.	Genotype and/or plasmid	Origina	Source or reference
Agrobacterium tumefaciens	A136		pCF218 + pCF372		56
Escherichia coli	S17-1		pMLBAD-aiiA-Gmr		24
	HB101		pRK600	Laboratory strain
Pseudomonas chlororaphis subsp. aureofaciens		ATCC 13985	Wild type		30
			pMLBAD-aiiA-Gmr		This study
Pseudomonas aeruginosa	PA01a		Wild type	Wound	7
	PA01b	MH340	Wild type	Wound	East Carolina University
		MH694	ΔlasI		15
		MH698	ΔrhlI		15
		MH710	ΔlasI/ΔrhlI		15
	TB		Wild type	CF patient	39
	TBCF10839		Wild type	CF patient	34
	PA14		Wild type	Burn wound	51
	PUPa3		Wild type	Rice rhizosphere soil	21
			ΔlasI		41
			ΔrhlI		41
			ΔlasI/ΔrhlI		41
			ΔrhlR/ΔlasR		41
Pseudomonas fluorescens	CHA0		Wild type	Suppressive soil	42
	CHA77		ΔhcnABC		22
Chromobacterium violaceum	CV0	ATCC 31532	Wild type	Soil	49
	CV026		AHL-negative strain		44
Serratia marcescens	MG-1		Wild type	Liquefied plant tissues	11
Serratia plymuthica	IC14		Wild type	Soil	19

^aCF, cystic fibrosis.

TABLE 2.

Arabidopsis thaliana cell lines used in this study

Background	ID no.	Genotype	Reference
Columbia	Col-0	Wild type
Columbia	N6590	Empty-vector (pBI1.4t) transformed control line for stocks CS6591 to CS6599 and CS6707	45
Columbia	N6595	Overexpressor of AOX1a (At3g22370) with Cys-127 mutated to Glu to confer activity without the activator pyruvate and to prevent formation of the intersubunit disulfide bond; CaMV 35S promoter-driven expression	45
Columbia	N6599	Antisense line with transgene expression construct; exhibits decreased AOX1a (At3g22370) protein in mitochondria due to presence of antisense AOX1a construct under CaMV 35S promoter control	45

Chemicals, culture media, and growth conditions.

Chemicals were purchased from Sigma-Aldrich, Buchs AG, Switzerland, unless specified otherwise. Bacteria were routinely cultured on LB (per liter: 10 g of Bacto tryptone, 5 g of Bacto yeast extract, 4 g of NaCl [AppliChem/Axon], pH adjusted to 7.4 and supplemented with 16 g agar [European Bacteriological Agar, Chemie Brunschwig] when needed). When required, N-acyl homoserine lactones (AHLs) were added to the medium to a final concentration of 200 nM. The AHLs used were: hexanoyl homoserine lactone (HHL) for C. violaceum CV026, oxo-dodecanoyl homoserine lactone (OdDHL) for P. aeruginosa lasI mutants and butanoyl homoserine lactone (BHL) for P. aeruginosa rhlI mutants. Pseudomonas isolation agar (PIA) medium contained (per liter) 45 g of Difco PIA supplemented with 20 ml of glycerol and 3 g of agar. ABM soft agar contained (per liter) 2 g of (NH₄)₂SO_4, 6 g of Na₂HPO_4, 3 g of KH₂PO_4, 3 g of NaCl, 1 ml of 0.1 M CaCl₂, 1 ml of 1 M MgCl₂, 1 ml of 0.003 M FeCl₃, 0.2% glucose (wt/vol), and 7 g of agar. Half-strength MS agar contained (per liter) 2.2 g of MS basal medium, 15 g of sucrose, and 8 g of agar; the pH was adjusted to 5.7. When required, ascorbate was added as sodium ascorbate to the MS medium to a final concentration of 0.1 mM. The antibiotics kanamycin, chloramphenicol, spectinomycin, and tetracycline were used at concentrations of 50, 10, 100, and 5 μg ml⁻¹, respectively. Bacterial strains were stored at −80°C in LB broth containing 16% glycerol.

Plant-bacterial dual growth experiments.

Experiments were performed in three-compartment dishes (Greiner) with three 6-day-old seedlings placed in one compartment containing half-strength MS agar. The second compartment contained LB agar, and the third compartment was left empty in all experiments described in the present study. Three (instead of two)-compartment dishes were chosen in case they would be needed at some point of the study (e.g., supplying active charcoal or other trapping material). Arabidopsis seeds were sterilized in 1.5-ml Eppendorf tubes by adding 1 ml of 70% ethanol and shaking them for 2 min on an IKA Vortex Genius 3 with an adapter at force 4. The supernatant was removed after centrifuging for 1 min at 6,000 rpm and 1 ml of a 1% NaOCl (Fluka) solution containing 0.03% Triton X-100 (Fluka) was added and shaken for 20 min as described above. After centrifugation, the supernatant was removed, and the pellet was washed four times with sterile MilliQ water. The seeds were stratified in a sterile 0.15% agarose solution at 4°C overnight in the dark and then plated on 12-cm square petri dishes containing half-strength MS agar. The plates were incubated in a climate chamber with a 12-h/12-h day/night alternation at 20°C, 50% relative humidity, and 100 μmol m⁻² s⁻¹ light for 6 days. After 6 days, three seedlings were transferred to one compartment of the divided petri dishes containing half-strength MS agar. In the next compartment, 20 μl of a liquid culture of each strain to be investigated (grown overnight at 30°C under shaking) were spotted on LB agar. Sterile LB broth was used as a control. The plates containing both the plants and the bacterial inoculum (or an uninoculated control) were sealed with Parafilm and incubated in the plant growth chamber described above. Pictures were taken after 14 and 21 days, and the plants were harvested after 21 days. Shoots were cut and weighed. The results were expressed as a percentage of plant biomass (fresh weight) relative to the biomass of control plants (growing next to uninoculated LB). All experiments were performed using four replicates.

Chemical exposure of plants to HCN.

In this experiment, one petri dish compartment contained three 6-day-old seedlings on half-strength MS agar, and the next contained on one side a 20-μl drop of a 0.09 M NaOH solution with 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2, 5, or 10 μmol of KCN and on the other side a 20-μl drop of 1 M HCl. The plates were sealed with Parafilm and tapped to make the two drops meet, thereby releasing HCN. After 21 days, pictures were taken, and the plant's shoots were weighed.

Collection and measurement of HCN.

HCN produced by bacteria was measured used a method modified from that of Guilbaul and Kramer (13). Bacteria were grown in three-compartment dishes on LB agar as described above. HCN was trapped in 1 ml of 4 M NaOH dropped in an empty compartment, and the petri dish was sealed with Parafilm. After incubation, a sample was taken and diluted to a concentration of 0.09 M NaOH. Additional dilutions of the samples in 0.09 M NaOH were made if needed to keep the cyanide concentration in the linear range of the measuring method (0.5 to 10 μM). We prepared 0.1 M o-dinitrobenzene (Fluka) and 0.2 M p-nitrobenzaldehyde solutions in 2-methoxyethanol; for each measurement, a fresh 1:1 mixture of these solutions was made. Then, 23-μl portions of the diluted sample were added to 77 μl of this mixture, followed by incubation for 30 min at room temperature. Next, 900 μl of 2-methoxyethanol was added, and the optical density was measured at 578 nm. Concentrations were obtained by comparison with a calibration curve computed using serial dilutions of a KCN stock solution.

Statistical analyses.

Differences between treatments were compared by using a two-sided Student t test (n = 3 to 4, P < 0.05).

RESULTS

Phytotoxicity of bacterial volatiles is dependent on cyanogenesis.

In order to assess the impact of bacterial cyanogenesis on the growth of A. thaliana, we selected 11 bacterial strains previously observed to have volatile-mediated deleterious effects on A. thaliana (Table 1). We then quantified their HCN production after different incubation times (see Fig. S1 in the supplemental material). In parallel, we grew A. thaliana in a compartmental petri dish in the presence of each strain and compared the biomass after 3 weeks to the biomass obtained in uninoculated control petri dishes. We observed a strong negative correlation between plant biomass and the total levels of HCN produced by the strains (Fig. 1). P. fluorescens CHA0 and P. chlororaphis subsp. aureofaciens both produced extremely high quantities of HCN (17 and 12 μmol, respectively) and led to plant death (CHA0) or to a drastic growth reduction (P. chlororaphis). The P. aeruginosa strains tested showed similar HCN production (2 to 4 μmol), causing a reduction of plant growth to values ranging from 5 to 18% relative to the uninoculated control. The environmental P. aeruginosa isolate, PUPa3, produced less HCN (<2 μmol) and was less virulent than most of the other P. aeruginosa strains (30% of the control plant biomass). C. violaceum CV0 showed similar levels of HCN production and phytotoxicity. Finally, Serratia plymuthica IC14 was found to produce HCN, albeit in very low amounts (ca. 90 nmol), whereas Serratia marcescens MG1 did not produce any detectable HCN but resulted in a similar inhibition of plant growth.

FIG. 1.

Plant growth inhibition and cyanogenesis by different bacterial strains. White bars indicate plant biomass (fresh weight) after 3 weeks of exposure to bacterial volatiles, expressed as a percentage of control plant biomass (not exposed to volatiles). Black bars indicate total HCN production (μmol) over 3 weeks. P.fluo: Pseudomonas fluorescens CHA0; P.chloro, Pseudomonas chlororaphis subsp. aureofaciens; C.viol, Chromobacterium violaceum CV0; S.plym, Serratia plymuthica IC14; S.marc, Serratia marcescens MG1. Averages of three to four replicates and standard errors are shown. Different lowercase letters (a to g) indicate significant differences (Student t test, P < 0.05).

An HCN-negative mutant is attenuated in A. thaliana virulence.

The observed correlation between HCN production and plant killing supported our hypothesis that HCN constitutes a major factor in the observed volatile-mediated deleterious effects of bacteria on plants. In order to test this hypothesis, we investigated the volatile-mediated effect of a HCN-negative mutant of P. fluorescens CHA0 on A. thaliana. As expected, the HCN-negative mutant CHA77 exhibited drastically reduced toxicity to A. thaliana compared to the wild-type CHA0 (Fig. 2). Although the plant biomass upon exposure to CHA77 volatiles was variable and in average only ca. 40% of the uninoculated control biomass (suggesting that HCN is not the only deleterious volatile produced by CHA0), plants were still alive and growing, in contrast to the plants exposed to the volatiles from the wild-type strain (Fig. 2, pictures).

FIG. 2.

Plant growth inhibition and cyanogenesis by P. fluorescens wild type (CHA0) and an HCN-negative mutant (CHA77). White bars indicate the plant biomass (fresh weight) after 3 weeks of exposure to bacterial volatiles produced by CHA0 or CHA77, expressed as a percentage of control plant biomass (not exposed to volatiles). Black bars indicate the total HCN production (μmol) over 3 weeks. CHA0, P. fluorescens wild type; CHA77, P. fluorescens hcnABC mutant. The averages of three to four replicates and standard errors are shown. The pictures are representative examples of control plants (upper picture), CHA0-treated plants (middle picture), and CHA77-treated plants (lower picture) after 3 weeks of incubation.

The plant growth inhibitory effect of bacterial volatiles can be mimicked by supplying HCN.

If HCN production by bacteria accounts to a large extent for the observed inhibitory effects of bacterial volatiles, exposing plants to HCN should give rise to similar symptoms. The observed effects were strongly dose-dependent: no effect was observed when the supplied quantities were below 1 μmol, but above this threshold, plant growth was severely inhibited: to 30% of the control with 1 μmol and 6% with 2 μmol. Plant death occurred very rapidly when 10 μmol of HCN was supplied (Fig. 3).

FIG. 3.

Plant growth inhibition by chemical addition of HCN. (A) Plant biomass (fresh weight) after 3 weeks of exposure to different amounts of HCN, expressed as a percentage of control plant biomass (not exposed to HCN). Averages of three to four replicates and standard errors are shown. Different lowercase letters (a to c) indicate significant differences (Student t test, P < 0.05). (B) Representative pictures of plants grown for 3 weeks after treatment with different amounts of HCN.

Role of quorum-sensing in HCN-mediated plant killing.

In agreement with previous studies, we noticed that the QS-deficient strain C. violaceum CV026 produced greatly reduced levels of HCN (Fig. 4A) (43). We concomitantly observed an increase in plant biomass from 30% biomass reduction to 160% biomass increase relative to the untreated control. This suggests that not only toxic substances, but also other, plant-growth-promoting, volatiles, whose effects may be masked by HCN in the wild-type, are produced by the strain. This is in line with the results described above (see Fig. 1): despite similar HCN production rates, the volatile-mediated virulence to A. thaliana was less severe for C. violaceum than for most P. aeruginosa strains. This is also corroborated by our observation that C. violaceum significantly promoted plant growth when cultured on MR-VP medium, which does not sustain HCN production (data not shown). Supplementing the mutant with C6-AHL restored both cyanogenesis and volatile-mediated toxicity. To assess whether HCN production was also under the control of quorum sensing in P. chlororaphis subsp. aureofaciens, we applied a quorum-quenching approach. To this end, a lactonase that degrades AHLs was introduced into the wild-type strain, resulting in strong decrease of AHL production (see Fig. S2 in the supplemental material). In contrast to the two P. aeruginosa strains and to C. violaceum, inactivation of quorum sensing increased the rate of HCN production, although the total amount produced did not differ significantly (Fig. 4B). No significant difference in plant killing was observed between the wild-type strain and the AHL-deficient transconjugant. In P. aeruginosa PAO1, single gene deletion mutants of the AHL synthase genes (lasI or rhlI) showed reduced levels of HCN production compared to the wild type, and inactivating both quorum-sensing systems was necessary to completely abolish cyanogenesis (Fig. 4C). The double mutant gave rise to significantly higher plant biomass compared to plants treated with the wild-type strain. Supplementing the double-knockout mutants with the corresponding AHL signal molecules partially restored cyanogenesis and also restored plant toxicity to wild-type levels. In the environmental strain P. aeruginosa PUPa3, simultaneous inactivation of lasI and rhlI was also required to abolish HCN production and reduce phytotoxicity (Fig. 4D). However, single gene deletion mutants (lasI or rhlI) produced higher levels of HCN (relative to the wild type) and were consequently more toxic to plants. Supplementing the two single mutants with their respective AHLs resulted in partial recovery of the wild-type phenotypes, while the supplemented double-knockout mutant recovered HCN production and phytotoxicity to a higher level than the wild type.

FIG. 4.

Effect of quorum sensing on cyanogenesis and phytotoxicity. White bars indicate the plant biomass (fresh weight) after 3 weeks of exposure to bacterial volatiles, expressed as a percentage of control plant biomass (not exposed to volatiles). Black bars give the total HCN production (μmol) over 3 weeks. (A) Chromobacterium violaceum CV0; (B) Pseudomonas chlororaphis subsp. aureofaciens; (C) Pseudomonas aeruginosa PAO1b; (D) Pseudomonas aeruginosa PUPa3. See Table 1 for more details. sup., supplemented with AHLs. Averages of three to four replicates and standard errors are shown. Different lowercase (a to e) letters indicate significant differences (Student t test, P < 0.05).

Bacterial volatiles induce oxidative stress in A. thaliana.

In order to investigate whether the alternative oxidase (AOX) is involved in plant tolerance to bacterial cyanogenesis, we grew lines of A. thaliana in which the AOX1a gene was either silenced or overexpressed (Table 2) in the presence of C. violaceum CV0, P. aeruginosa PAO1, P. aeruginosa PUPa3, and P. fluorescens CHA0. These four strains differ both in the total levels of HCN produced and in the kinetics of cyanogenesis. Furthermore, the effect of ascorbate addition in counteracting putative oxidative stress was assessed in the different plant lines. The addition of ascorbate alone did not improve plant growth; neither did overexpression of AOX1a (Fig. 5). However, a combination of supplying ascorbate and overexpressing AOX1a restored plant growth in the presence of the two P. aeruginosa strains. Surprisingly, plant growth reduction caused by C. violaceum CV0 was not alleviated by this combined treatment. Silencing of AOX1a led to higher susceptibility to bacterial volatiles in all four strains tested, including C. violaceum CV0. Overexpressing AOX1a and supplying ascorbate did not significantly change the growth of the plants when exposed to the volatiles of the HCN-deficient mutant CHA77 (data not shown). Interestingly, when plants were challenged with 2 μmol of chemical HCN, overexpression of the AOX1a alone (without ascorbate addition) was sufficient to restore plant growth from 2% (empty vector) to 98% of the unexposed AOX1-expressing plant (data not shown).

FIG. 5.

Effect of ascorbate addition and alternative oxidase (AOX) overexpression on A. thaliana's tolerance to deleterious bacterial volatiles. Black bars, no ascorbate addition; white bars, addition of 0.1 mM ascorbate as sodium ascorbate to the plant's culture medium. ctrl., empty vector plants (N6590); AOX+, plants overexpressing AOX (N6595); AOX−, plants silencing AOX (N6599). See Table 2 for more details. Averages of three to four replicates and standard errors are shown. Within each group (bacterial strain), different letters indicate significant differences (Student t test, P < 0.05).

DISCUSSION

The volatile-mediated impact of bacteria on plants has gained increasing attention over the last few years (14, 16, 33, 46, 50, 54). Despite the magnitude of the effects observed in both positive (14, 36) and negative cases (16, 43, 46), the compounds responsible for the effects and the mechanisms involved in plant growth alteration are still poorly understood. We have presented evidence for one candidate deleterious volatile, hydrogen cyanide (HCN). HCN inhibits several metal-containing enzymes, most significantly cytochrome c oxidase, one of the key enzymes of the respiratory electron transport chain. Cyanogenesis occurs in few bacteria genera, mainly Pseudomonas, Chromobacterium, Rhizobium (31), and, as we show here for the first time, Serratia. S. plymuthica IC14 produced approximately 100 nmol of HCN, while cyanogenesis was not detected in S. marcescens MG1. We observed large differences in HCN production between the strains tested (Fig. 1): the highest levels of HCN were observed for P. fluorescens CHA0 and P. chlororaphis subsp. aureofaciens, two strains of environmental origin, followed by the P. aeruginosa strains and C. violaceum CV0. Interestingly, the two PAO1 strains tested differed significantly in the amounts of HCN produced. It was recently reported that an originally identical PAO1 strain kept for decades in different laboratories displayed significant genomic variability (20). It is thus not surprising that the two PAO1 strains we analyzed, which were obtained from different sources, also displayed phenotypic variability. In addition to the different P. aeruginosa strains of clinical origin, we assessed cyanogenesis and volatile-mediated impact of the closely related environmental strain P. aeruginosa PUPa3 on plants. PUPa3 was notably the least deleterious of all P. aeruginosa strains tested, even though it produced similar quantities of HCN to the three strains of clinical origin: PA14, TB, and TBCF10839. It should be noted that P. aeruginosa PUPa3, first described in 2005 as an antifungal bacterium isolated from the rhizosphere of rice, was reported to be noncyanogenic (21). However, even if PUPa3 produces less HCN than some clinical strains, we show here that it is definitely cyanogenic. Differences in the sensitivity of the detection method (qualitative versus quantitative), as well as in the time and temperature of incubation, might account for this discrepancy. Under our experimental conditions, PUPa3 only started to produce significant amounts of HCN after 1 week.

The observed differences in cyanogenesis between the 11 strains tested were correlated with the volatile-mediated effects of these strains on A. thaliana (Fig. 1), from rapid plant death when growing next to P. fluorescens CHA0 (>17 μmol) to a reduction to 30% of the control plant biomass when growing next to P. aeruginosa PUPa3 (<2 μmol). While the production of other deleterious volatiles is obvious from the plant growth inhibition observed upon exposure to the HCN-negative S. marcescens MG1 (30% of the control) or to the HCN-deficient mutant P. fluorescens CHA77 (40%), the difference between reduced growth and death is likely due to cyanogenesis in P. fluorescens CHA0 (Fig. 2, pictures). Moreover, the symptoms observed upon contact with the volatiles of cyanogenic bacteria (chlorosis, reduced growth, or even death) were mimicked when plants were challenged with pure HCN. Interestingly, at similar doses, plants were more severely affected by the chemical supply of HCN than by bacterial HCN (Fig. 1 and 3): 2 μmol of chemically supplied HCN resulted in 6% of growth compared to the control, but when plants were grown in the presence of strains producing a similar quantity of HCN (PUPa3 and CV0), plant biomass was five times higher (30% of the control biomass). Similar observations were made in the field, where external application of HCN proved much more plant growth inhibitory than cyanogenesis by inoculated rhizosphere pseudomonad populations (53). In addition to the putative production of growth promoting volatiles by the bacterial strains, which could counteract the negative effect of HCN, the different timing dynamics of application is likely to account in great part for the differences observed in plant responses. While chemical HCN was supplied all at once to young seedlings, bacterial HCN was supplied to plants continuously as they grew for 3 weeks, and in most cases, started only once a critical population density (the so-called “quorum”) had been reached.

Quorum sensing has been shown to regulate many phenotypes associated with virulence (reviewed in references 2 and 8). However, until now, only a single study by Müller et al. (27) investigated the involvement of quorum sensing in volatile-mediated virulence and showed that the volatiles produced by a quenched strain of S. plymuthica HRO-C48 were more effective in limiting the growth of two phytopathogenic fungi than the volatiles of the wild type. We were therefore interested in assessing whether volatile-mediated phytotoxicity would also be quorum sensing dependent. In the well-studied model strain PAO1, which possesses two quorum-sensing systems, the RhlI/R and LasI/R systems, quorum sensing plays a role in cyanogenesis: the involvement of the RhlI/R quorum-sensing system in the regulation of HCN biosynthesis was first reported by Winson et al. in 1995 (52), while it became clear 5 years later that both AHL-based quorum-sensing systems were required for HCN production (31). If the main compound responsible for volatile-mediated phytotoxicity were indeed HCN, we would expect the quorum-sensing mutants to lose their volatile-mediated virulence to A. thaliana. This is precisely what we observed with P. aeruginosa PAO1 (Fig. 4C): deleting both AHL synthase genes led to loss of cyanogenesis and resulted in the restoration of plant growth to levels similar to those observed in control plants. In contrast to P. aeruginosa PAO1, regulation of cyanogenesis in the environmental P. aeruginosa strain PUPa3 had not yet been investigated. We show that in PUPa3, disruption of both quorum-sensing systems was also necessary to abolish cyanogenesis and to reduce phytotoxicity. This is consistent with previous findings, where the virulence of the PUPa3 lasI/rhlI double mutant was shown to be significantly reduced in both animal and fungal models (41). Although the mechanisms underlying virulence have not yet been assessed, it is tempting to speculate that reduced virulence might be due, at least in part, to the loss of cyanogenesis in the PUPa3 double mutant. Both the lasI and the rhlI single mutants of PUPa3 produced more HCN, which led to an increased virulence on plants. This contrasts with observations made in PAO1, where single mutants showed reduced cyanogenesis relative to the wild type. Another difference compared to PAO1 is that in the absence of HCN production, plants reached only 50% of their control biomass, suggesting that additional, as-yet-unidentified toxic compounds are present in the volatiles of PUPa3. In C. violaceum CV0, cyanogenesis has also been reported to be under the control of quorum sensing (44). We provide here quantitative data showing that inactivation of the AHL synthase in C. violaceum abolished both cyanogenesis and virulence on A. thaliana. Surprisingly, the quorum-sensing mutant CV026 promoted plant growth, suggesting the production of plant-growth-promoting volatiles, whose beneficial effects on plants were masked by HCN in the wild type. We are currently investigating the chemical nature of these volatiles. Finally, we discovered that P. chlororaphis subsp. aureofaciens ATCC 13985, a close relative of P. chlororaphis subsp. aureofaciens 30-84 (30), produced very high levels of HCN, similar to those observed with P. fluorescens CHA0. P. chlororaphis 30-84 is a well-studied biocontrol strain that produces phenazines in addition to HCN. Two quorum-sensing systems have been identified in P. chlororaphis 30-84, the PhzI/R system, which is responsible for phenazine biosynthesis, and the CsaI/R system, which controls biofilm formation. Both systems are needed for proteolytic activity and rhizosphere competence (32, 55). Cyanogenesis in this organism has been reported to be independent of quorum sensing (55), a finding that was supported by the results of the quorum-quenching experiment performed on a closely related strain in the present study.

Besides the inhibitory effect of HCN when applied exogenously in high amounts to plants, recent studies suggest roles for HCN in planta, far beyond its being solely a by-product of ethylene biosynthesis (29): in addition to its function in plant defense against herbivory (through the wound-induced hydrolysis of cyanogenic glucosides), HCN has been shown to break seed dormancy in various species and to be involved in induced resistance against viruses (12, 28, 38). However, it is not clear whether HCN itself is the signaling molecule, or rather the burst of reactive oxygen species (ROS) that follows treatment with HCN. In sunflower (28) and apple (12) seeds, stimulation of seed germination by application of KCN was mediated by ROS. It has further been shown that the gene expression changes occurring upon treatment with either KCN or with the ROS-generating methyl viologen were very similar (6). We therefore hypothesized that the phytotoxicity of bacterial volatiles might be linked to oxidative stress. The results we obtained upon overexpression of the alternative oxidase AOX1 and concomitant supplementation of ascorbate as a ROS-trapping compound confirmed this hypothesis, and indicated that oxidative stress is a key process in the volatile-mediated negative impact of cyanogenic bacteria on plants. Interestingly, in a study investigating the impact of truffle volatiles on plant growth, deleterious fungal volatiles (other than HCN) were shown to induce oxidative stress in Arabidopsis (40). This suggests that oxidative stress might be a general response of plants to detection of microbial volatiles (25, 26). In summary, we present evidence that HCN, when produced in high amounts by bacteria, can kill plants. We suggest that this accounts to a great extent for the deleterious impact of bacterial volatiles observed in previous studies using cocultivation of Arabidopsis and bacterial strains. Moreover, we report for the first time that cyanogenesis is not restricted to members of the Pseudomonas, Chromobacterium, and Rhizobium genera but that Serratia species, e.g., Serratia plymuthica IC14, can also produce HCN. We further demonstrate that the volatile-mediated phytotoxicity of C. violaceum CV0, P. aeruginosa PAO1 and P. aeruginosa PUPa3 is quorum sensing regulated. The environmental strain PUPa3 showed less phytotoxicity and lower cyanogenesis than the clinical isolate PAO1, but in both strains, cyanogenesis was dependent on functional quorum-sensing systems. In contrast, quorum-sensing was not required for HCN production in P. chlororaphis subsp. aureofaciens. Our data also provide initial insights into the mechanism of action of these bacterial volatiles leading to growth reduction or plant death: supplying ascorbate to alternative oxidase overexpressing lines of Arabidopsis increased their tolerance to cyanogenic pseudomonads drastically, while silenced lines showed higher susceptibility. This indicates that volatile-mediated phytotoxicity involves oxidative stress and that the perception of deleterious bacterial volatiles causes, like many other biotic and abiotic stresses, an oxidative burst in the plant. Future studies are required for a more detailed understanding of the metabolic changes occurring in plants upon exposure to toxic microbial volatiles.