Download pdf - Indigenous bacteria with antagonistic and plant-growth-promoting activities improve slow-filtration efficiency in soilless cultivation

Indigenous bacteria with antagonistic and plant-growth-promoting activities improve slow-filtrationefficiency in soilless cultivation

F. Déniel, P. Rey, M. Chérif, A. Guillou, and Y. Tirilly

Abstract: In tomato soilless culture, slow filtration allows one to control the development of diseases caused by patho-genic microorganisms. During the disinfecting process, microbial elimination is ensured by mechanical and biologicalfactors. In this study, system efficacy was enhanced further to a biological activation of filter by inoculating thepozzolana grains contained in the filtering unit with 5 selected bacteria. Three strains identified as Pseudomonas putidaand 2 as Bacillus cereus came from a filter whose high efficiency to eliminate pathogens has been proven over years.These 5 bacteria displayed either a plant growth promoting activity (P. putida strains) or antagonistic properties (B. cereusstrains). Over the first months following their introduction in the filter, the bacterial colonisation of pozzolana grainswas particularly high as compared to the one observed in the control filter. Conversely to Bacillus spp. populations,Pseudomonas spp. ones remained abundant throughout the whole cultural season. The biological activation of filter unitvery significantly enhanced fungal elimination with respect to the one displayed by the control filter. Indeed, the 6-month period needed by the control filter to reach its best efficacy against Fusarium oxysporum was shortened for thebacteria-amended filter; in addition, a high efficacy filtration was got as soon as the first month. Fast colonization ofpozzolana grains by selected bacteria and their subsequent interaction with F. oxysporum are likely responsible for filterefficiency. Our results suggest that Pseudomonas spp. act by competition for nutrients, and Bacillus spp. by antibiosisand (or) direct parasitism. Elimination of other fungal pathogens, i.e., Pythium spp., seems to differ from that ofFusarium since both filters demonstrated a high efficacy at the experiment start. Pythium spp. elimination appears tomainly rely on physical factors. It is worth noting that a certain percentage of the 5 pozzolana-inoculated bacteriafailed to colonise the filter unit and were, thus, driven to the plants by the nutrient solution. Their contribution to theestablishment of a beneficial microbial community in the rhizosphere is discussed.

508Key words: Pythium spp., Fusarium oxysporum, Bacillus cereus, Pseudomonas putida.

Résumé : La filtration lente permet de lutter contre le développement des maladies en cultures hors-sol de tomate. Du-rant le processus de désinfection, l’élimination des micro-organismes est assurée par des facteurs mécaniques et biolo-giques. Pour améliorer l’efficacité du système, au cours de la présente étude, nous avons biologiquement activé unfiltre en inoculant les grains de pouzzolane contenus dans la colonne filtrante avec 5 bactéries sélectionnées. Trois dessouches identifiées comme étant Pseudomonas putida et 2 autres comme étant Bacillus cereus proviennent d’un filtredont l’efficacité élevée pour éliminer les microorganismes a été vérifiée pendant plusieurs années. Les souches deP. putida montrent une activité favorisant la croissance de plantes et celles de B. cereus des propriétés antagonistes.Dés les premiers mois suivant leur introduction dans le filtre, la colonisation des grains de pouzzolane par les bactériesétait particulièrement élevée en comparaison de celle du témoin. A l’inverse des populations de Bacillus spp., celles dePseudomonas spp. furent particulièrement abondantes tout au long de la saison culturale. Comparé au filtre témoin,l’activation biologique de la colonne filtrante induit une augmentation significative du niveau d’élimination fongique.La période de 6 mois nécessaire au témoin pour atteindre son niveau maximal d’efficacité contre Fusarium oxysporumest raccourcie dans le cas du filtre ensemencé en bactéries ; en effet, une très grande efficacité est observée dès le pre-mier mois de filtration. Une colonisation rapide des grains de pouzzolane par les bactéries sélectionnées et leurs inte-ractions avec les F. oxysporum sont certainement responsables de l’efficacité du système. Ces observations suggèrentune action par compétition nutritive dans le cas des Pseudomonas spp., et par antagonisme (antibiose et/ou parasitisme)pour les Bacillus spp. L’élimination des autres champignons, i.e., Pythium spp., diffère de celle décrite ci-dessus carune efficacité élevée a été obtenue dans les 2 filtres dés le premier mois d’expérimentation. Les facteurs physiques sont

Can. J. Microbiol. 50: 499–508 (2004) doi: 10.1139/W04-034 © 2004 NRC Canada

499

Received 14 November 2003. Revision received 26 March 2004. Accepted 31 March 2004. Published on the NRC Research PressWeb site at http://cjm.nrc.ca on 8 September 2004.

F. Déniel, P. Rey,1 and Y. Tirilly. Laboratoire de Biodiversité et Ecologie Microbienne, ESMISAB, Université de BretagneOccidentale-Brest, Technopôle Brest-Iroise, 29280, Plouzané, France.M. Chérif. Laboratoire de Phytopathologie, Institut National Agronomique de Tunisie, 43 Avenue Charles Nicolle, 1082 CitéMahrajène, Tunis, Tunisie.A. Guillou. CATE, Station Expérimentale de Vézendoquet, 29250 Saint-Pol-de-Léon, France.

1Corresponding author (e-mail: [email protected]).

certainement responsables de l’élimination des Pythium spp. On peut noter qu’ un certain pourcentage des bactériesinoculées n’a pas colonisé le support dans la colonne filtrante, mais a été véhiculé par la solution nutritive jusqu’auxplantes. Leur contribution à l’établissement d’une microflore bénéfique dans la rhizosphère est discutée.

Mots clés : Pythium spp., Fusarium oxysporum, Bacillus cereus, Pseudomonas putida.

Déniel et al.

Introduction

In soilless cultivation, the water, which comes fromsources such as lakes, rivers, and wells, is generally colo-nized by numerous bacteria and fungi, some of which arepathogenic to plants (Stanghellini and Rasmussen 1994).Once introduced, these microorganisms are easily spreadthrough the greenhouse by recirculated solutions. Closed hy-droponic systems minimize pollution by reusing the run-off,however, they concomitantly increase the risks of pathogenattacks by using water contaminated with pathogenic micro-organisms (McPherson et al. 1995; Van Os 1999). Findingmethods that prevent such disinfection has become a majorchallenge.

Several effective methods, such as heat treatment,ozonization, ultraviolet radiation, and chlorination, havebeen proposed for the disinfection of nutrient solutions(Ehret et al. 2001; Goldberg et al. 1992; Rey et al. 2001;Runia 1995; Steinberg et al. 1994). Using ultraviolet irradia-tion on recirculating solution has been proven to controlPythium spp.–induced root rot in tomato and cucumberplants (Postma et al. 2001; Zhang and Tu 2000). Unfortu-nately, this active method affects the total microflora by de-stroying not only the target pathogen, but also nontargetmicroorganisms. (Zhang and Tu 2000). Similarly, Poncet etal. (2004) demonstrated that chlorine reduced bacterial di-versity in the rhizosphere. Postma et al. (2000) looked at therole of natural microflora in suppressing certain diseases bycomparing systems with and without their originalmicroflora. In fact, natural microflora have often shown acertain ability to suppress diseases (Berger et al. 1996; Chenet al. 1998). Tu et al. (1999) observed that a large bacterialpopulation in the rhizosphere can limit the extent of Pythiumroot rot, which led them to speculate about the involvementof resident bacteria in disease biosuppression. Generally, ac-tive disinfecting methods are unable to preserve nonpatho-genic microflora (McPherson et al. 1995) because theynegatively affect the suppressing potential of naturalmicroflora against certain pathogens, such as Pythiumspp. and Phythophthora spp.

To prevent this undesirable effect, the attention of re-searchers, over the last decade, has been directed on a prom-ising method for soilless cultivation, the slow-filtrationtechnique. During the disinfection process, the nutrient solu-tion flows slowly through a filter unit, which is filled withdifferent substrates, such as fine sand, rockwool flocks, orpozzolana grains. This passive method eliminates pathogenswithout destroying the natural microflora (Van Os andPostma 2000). Among the pathogens eliminated atsubstantial rates with this technique are zoosporic fungi (e.g.,Phytophthora spp.), bacteria (e.g., Xanthomonas campestris),nematodes, and even viruses (Ehret et al. 2001; Van Os et al.1999). Analysis of the total microflora has identified a clear

change in the bacterial community after the nutrient solutionhas passed through the filter unit (Postma et al. 1999). Inter-estingly, slow filtration keeps a part of the natural microfloraalive; it has been proven harmless to specific groups of bac-teria. Mechanical and biological factors are thought to be re-sponsible for the effectiveness of the system. However, untilnow, experiments conducted to improve system effective-ness have focused on determining flow rates through the fil-ter unit and on the nature and optimal depth of substrates infilter tubes (Wohanka et al. 1999). Brand and Wohanka(2001) showed that the formation of bacterial microcoloniesor biofilms on substrates is a key factor in enhancing effi-ciency. Brand (2000) isolated and identified a large numberof these bacteria, and showed that the dominating genus,Pseudomonas, contributed to more than 50% of all isolates.Of the other isolates, 10.2% were identified were assignedas Bacillus.

This study was designed to optimize biofiltration using se-lected bacteria. A new filter system often needs severalmonths to reach peak efficiency; our aim was to shorten thistime by inoculating the filter with specific bacteria. To dothis, bacteria were isolated from an effective filter, identi-fied, screened for their efficiency against fungal pathogens,and then a second, new, filter was inoculated with the fivemost abundant bacteria. The development of bacteria onpozzolana grains in the filter was also studied, and the effi-ciency of our procedure was assessed against a control filter.

Materials and methods

FiltersTwo filter units were used: one was inoculated with five

selected bacteria at the beginning of the cultivation season,and the other one served as a control. Each filter consisted ina plastic pipe (220 cm long, with an inner diameter of40 cm), and was filled with pozzolana grains (1–4 mm diam-eter), which acted as the filtering medium. Nutrient solution(Kemira, France) flowed through a layer of pozzolana 100-cm thick, which was deposited above three successive layersof graded gravel (2–8, 8–16, and 16–32 mm) that had anoverall thickness of 40 cm. The upper water layer was regu-lated by a float switch, placed 40–50 cm above thepozzolana surface. The filtration rate ranged from 100–150 L·h–1·m–2. The filter units were set, at room temperature,in 2 separate areas of an experimental tomato greenhouse atroom temperature. For each filter, the same nutrient solutionfed plants throughout the cultivation season.

Identification of bacteria used to inoculate filtersBacteria were selected from the top layer of pozzolana

grain from a three-year-old filter, chosen because of itshighly efficient elimination of fungi, such as Pythiumspp. and Fusarium oxysporum and bacteria, such as total

© 2004 NRC Canada

500 Can. J. Microbiol. Vol. 50, 2004

bacteria microflora and fluorescent Pseudomonas, from thenutrient solution (Rey et al. 1999).

To isolate bacteria, 7 g of pozzolana grains were washedthree times with 63 mL of a solution of physiological water(0.85% NaCl) and Tween 80. Pozzolana grains were thensonicated in 63 mL of the same solution for 90 s at the high-est setting, using a VibraCell (Bioblock Scientific, Illkirch,France) bench sonicator. The solution sample was platedwith a spiral plater on plate-count agar (PCA) plates. We se-lected the three most abundant bacteria, and chose two oth-ers for their typical features; all of them were purified andidentified to the species level using biochemical tests, suchas catalase and oxydase activities and growth on selectivemedium, for example, King B. Pseudomonas was identifiedwith API 20 NE galleries, and Bacillus with API 20 E andAPI 50CH (API galleries, BioMérieux, France). For eachisolate, the profiles issued from API tests were analyzedwith APILAB software (BioMérieux, France), and identifi-cation down to the species level was expressed as a percent-age of probability.

Effect of the five selected bacteria on plant growthPlants were inoculated with the five selected bacteria us-

ing the following procedure.

Plant materialTomato seeds (Lycopersicon esculentum Mill. cv. Prisca)

were sterilized by immersing them in 70% ethanol for5 min, soaking them in 5% aqueous sodium hypochlorite for10 min, and rinsing them thoroughly three times in steriledistilled water. They were then placed on water-soaked filterpaper (Whatmann No. 1) in Petri dishes, and kept in the darkat 25 °C. One week later, all the germinated seeds were indi-vidually transferred to sterile glass tubes (20 × 150 mm) thatcontained 12 mL of the nutrient solution described byHoagland and Arnon (1938). A sterile piece of filter paper(90 × 180 mm), perforated in its centre, was placed in eachtube to provide a solid support for the growing roots.Plantlets were grown under a 16-h light (at 25 °C) : 8-h dark(at 16 °C) photoperiod in a Sanyo growth cabinet (Antony,France). Light was provided by Mazdafluor Incandia 830tubes (Mazdafluor, France), which had an intensity of40 µmol·m–2·s–1.

Plant inoculation with bacteriaEach bacterium was precultured in tubes that contained

9 mL of nutrient broth, for 18 to 24 h at 30 ± 1 °C. Theywere then grown in 250-mL Erlenmeyer flasks, which werefilled with 125 mL of nutrient broth. The cultures wereincubated for 24 h on a rotary shaker (120 r/m) in the dark, at30 ± 1 °C. Three-week-old plantlets were inoculated bypouring 1 mL of each bacterial strain in the plant nutrientsolution at 2 different concentrations: 106 CFU(colony form-ing units)·mL–1 and 109 CFU·mL–1. For each experiment, 18to 23 plantlets were used. Control plants were treated with1 mL of distilled water. Root systems and shoots of plantswere weighed 3 weeks after inoculation with bacteria. Statis-tical analyses were performed using Duncan’s test at 95%confidence. The statistical program used for analysis wasStatGraphics, release 4.0 Manugistic Inc.).

Effect of the five selected bacteria on fungal growth invitro

This study used the dual-culture technique. The selectedbacteria were tested against the fungal species frequently de-tected in soilless tomato cultivation. Some of these fungi arepathogenic, such as Colletotrichum coccodes, F. oxysporumf.sp. radicis lycopersici (FORL), Pythium group F, Pythiumirregulare, and Rhizoctonia solani; others are nonpatho-genic, such as Acremonium sp., Aspergillus ochraceus,F. oxysporum, Pythium oligandrum, and Trichoderma viride.To produce bacterial inoculum, each bacterium wasprecultured for 24 h at 30 ± 1 °C in tubes filled with 9 mLof nutrient broth. A single colony of the bacterial isolate wasused to inoculate the dual-culture plates of potato dextroseagar and nutrient agar by placing a single droplet (20 µL) ofbacterial inoculum at each side of the plate (4 droplets/plate).The dual-culture plates were also inoculated with plugs ofmycelium from six-day-old plates of fungi (1 plug per fun-gus). Five plates were used for each bacteria/fungus combi-nation. The plates inoculated with only one fungus served ascontrols. The plates were incubated at 25 °C, and the micro-bial interactions were analysed by measuring the radius ofthe zones of inhibition produced when the control platesreached full development. Our notation system was adaptedfrom the one described by Walker et al. (1998).

Inoculation of filter with bacteriaBefore inoculating the filter, the five selected bacteria

were precultured for 18 to 24 h, at 30 ± 1 °C, in tubes filledwith 9 mL of tryptic soy broth (TSB). For each bacterium,1 mL of preculture was poured into 250-mL Erlenmeyerflasks, which were filled with 150 mL of TSB. ThreeErlenmeyer flasks were used for strain L1, 18 for L2, 3 forL3, 11 for L4, and 5 for L5 (the number of Erlenmeyersused per strain corresponds to the proportion of each bacte-rium estimated at isolation). The cultures were incubated for24 h on a rotary shaker (120 r/m) in the dark at 30 ± 1 °C.Filters were inoculated by pouring the contents of theErlenmeyer flasks (total volume of 12 L) over the pozzolanagrains. The first inoculation was performed at the end ofFebruary, and the second one two weeks later. The bacterialconcentrations introduced into the filters were 3.21 × 1011

and 1.86 × 1011 CFU/mL for the first and the second inocu-lation, respectively.

Nutrient-solution samplingTo determine the effectiveness of the two filters, regular

samples were taken throughout the cultivation season, fromApril to September. Each month, three samples (500 mLeach) of the nutrient solution were collected just before itpassed through the filter, and three others (500 mL each)were taken from the filter effluent. Fungi, such as Pythiumspp. and F. oxysporum, and total mesophylic bacteria werelooked for; these microorganisms are key components ofroots and nutrient solution microflora in soilless cultivationin greenhouses (Berkelman et al. 1995; Rey et al. 1997).

The concentration of Pythium spp. and F. oxysporum wasdetermined using 150 and 100 mL, respectively, of nutrient-solution samples filtered through a 0.45 µm membrane filter.The filters were plated on selective media, coded CMA-PARPfor Pythium spp., and coded Komada for F. oxysporum.

© 2004 NRC Canada

Déniel et al. 501

Pythium thalles were counted after the plates were incubatedfor 48 h at 25 °C in the dark; F. oxysporum colonies werecounted 5 and 7 d after incubation under the same condi-tions.

To count bacterial colonies, 50 µL of each nutrient-solution sample were plated with a spiral plater on a selec-tive medium (2 subsamples per sample). The media usedwere either PCA for the total mesophylic bacteria, or King Bagar for fluorescent Pseudomonas. The number of bacteriacolonies was determined after plates were incubated in thedark for 48 h. Results are expressed as the percentage ofeliminated microorganisms.

Assessment of the bacteria colonies plated on pozzolanagrains

Pozzolana grains were sampled monthly, from April toSeptember, to study the bacterial populations responsible fortheir colonization. These populations were assessed as previ-ously described: the mesophylic aerobe bacteria werecounted on PCA, Bacillus spp. on antibiotics-amendedglucose-agar medium, Pseudomonas spp. on cetrimide-fucidine-cefaloridine (CFC) medium, and fluorescent Pseu-domonas on King B medium. Bacterial species were ex-pressed as the number of CFUs per gram of pozzolana grain.Statistical analyses were performed using Duncan’s test at95% confidence. The statistical program used for analysiswas StatGraphics, release 4.0.

Tomato cultivation in a soilless greenhouseExperiments were performed with seeds of tomato cv.

Tradiro (De Ruiter Graines, France) sown in rockwool cubesin a nursery greenhouse, and fertilized daily with a nutrientsolution. After four weeks, the plants were transferred tococo-fibre slabs (4 plants per slabs), and were set in a green-house on January 15th. Each slab was enveloped in a plasticbag to isolated it from the others. The nutrient solution wasdelivered to each plant through a capillary system installedat its collar, and circulated in closed loop inside the green-house; its pH was regularly controlled and was always be-tween 5.5 and 6.2. Greenhouse temperature was alsoregularly measured, and ranged from 16 °C at night to 25 °Cduring the day. The greenhouse was divided in two areas,each of which contained a filter unit and 136 tomato plants.

In one of the greenhouse areas, selected bacteria wereadded to the filter at the start of the cultivation season; in thesecond area, the filter unit served as a control. The tomato

fruits were collected every week, from the end of March tothe end of September; the yield per square meter was as-sessed, and the data were statistically analysed using theNewman–Keuls test, at the 95% confidence limit.

Results

Identification of the five selected bacteriaAs shown in Table 1, the identification of each isolate

down to the species level is expressed as a percentage proba-bility, in accordance with the software analysis. Positiveidentification was obtained for the five strains of bacteria:two were identified as Bacillus cereus (L1 and L3), andthree (L2, L4, and L5) were Pseudomonas putida (Table 1).

Effect of selected bacteria on plant growthThe introduction of 109 CFU/mL bacteria to the nutrient

solution induced different rates of plant growth. In fact, theplants inoculated with one of the three P. putida isolates(L5) resulted in a more developed root system, whichweighed more than those of control and B. cereus–colonizedplants (Fig. 1). This weight increase was associated withnormal white roots and green shoots, similar those of thecontrol plants. On the other hand, roots colonized by the twoB. cereus isolates (L1 and L3) had typical necrotic areas andwithered shoots; statistically, they weighed less than those ofthe control plants. The plants that had been fed a nutrient so-lution inoculated with 106 CFU/mL bacteria has neithershoot withering nor necrotic symptomsin the root. Only theL5 strain of P. putida was associated with an increase in rootweight. The shoot weights of all plants were within the samerange (data not shown).

Effect of selected bacteria on fungal growth in vitroThe interaction of bacteria with fungi in this dual culture

varied, according to the bacterium used. The L3 isolate ofB. cereus inhibited the development of nearly all fungi (Ta-ble 2). The zones of inhibition against the pathogenic(FORL, R. solani) and the nonpathogenic (F. oxysporum,Acremonium spp.) fungi were particularly well developed.Nevertheless, this isolate seemed to be ineffective againstP. oligandrum and T. viride. In contrast, the L4 and L5 iso-lates of P. putida failed to inhibit the growth of any of thetested fungi, except P. oligandrum. In 8 of the 12 combina-tions tested, fungal mycelium completely covered the bacte-rium. As far as L1 (B. cereus) and L2 (P. putida) are

© 2004 NRC Canada


Strains Species name % identification Test against Comments

L1 Bacillus cereus 99.4 Arbutine (85%), citrate (78%),N-acetyl-glucosamine (97%),Nitrate reduction (78%)

Very goodidentification

L2 Pseudomonas putida 93.0 Maltose (8%) Excellent identificationL3 B. cereus 98.2 Citrate (78%) Excellent identificationL4 P. putida 99.1 — Very good

identificationL5 P. putida 99.1 — Very good

identification

Note: Biochemical tests (e.g., catalase and oxydase activities) and API galleries (API 20 NE for Pseudomonas; API 20 E and API 50 CH for Bacillus)were used to identify bacteria.

Table 1. Identification of the five selected bacteria strains to the species level.

concerned, no inhibition zones were observed in presence ofthe tested fungi; these bacterial isolates were usually over-grown with the tested fungi (Table 2).

Elimination of fungi and bacteria contained in nutrientsolution after slow filtration

The recirculating solutions used in the soilless cultivationcontained various fungi, such as Pythium spp. andF. oxysporum, which were regularly detected in the influentsolution during the cultivation season (Tables 3 and 4). Dur-ing our six-month experiment, these fungi were efficientlyeliminated by slow filtration. When the nutrient solutionflowed through two filters, generally 99% of Pythiumspp. were eliminated; the lowest values, 92% and 98.1%,were observed in June for the bacteria-amended filter, and inJuly for the control filter, respectively (Table 3). Eliminationof F. oxysporum varied significantly, according to the usedfilter. In the control filter, during the first 3 months of exper-

iment, elimination ranged from none to 50%; this rose to69.5% in July, and finally reached 98.3% in September. In thebacteria-amended filter, elimination of F. oxysporum was atleast 98% during the entire cultivation season (Table 4).

High densities of the mesophylic aerobe microflora bacte-ria were recovered from the recirculating nutrient solutionsduring the six-month sampling period. Once the solutionshad passed through the filter units, the population rangeswere within 1.3 × 103 and 4.2 × 104 CFU/mL. Whatever theexperimental conditions, the percentage of bacteria elimina-tion was relatively low (Table 5).

Evolution of bacterial populations on pozzolana grainsin bacteria-amended and control filters

Regardless of the sampling month, the pozzolana grainssampled from both filters were regularly and highly colo-nized by various mesophylic aerobe microflora bacteria(Fig. 2). These bacteria were identified as Bacillus spp.,

© 2004 NRC Canada

Déniel et al. 503

Fig. 1. Effect of five selected bacteria on root growth of tomato plantlets. Plantlets were grown in sterile glass tubes containing nutri-ent solution and under a 16-h light (at 25 °C) : 8-h dark (at 16 °C) photoperiod; 18 to 23 plantlets were inoculated with each bacteria(109 CFU/mL). Root systems were weighed 3 weeks after inoculation with bacteria. The same letter on the bars indicates no signifi-cant difference at P ≥ 0.05 (Duncan’s test).

Bacteria strains

FungiL1B. cereus

L2P. putida

L3B. cereus

L4P. putida

L5P. putida

Acremonium sp. – – – – +++ – – – –Aspergillus ochraceus – – – + – –Colletotrichum coccodes – – ++ – – –Fusarium oxysporum – – – – +++ – – –FORL – – – – +++ – – – –Pythium group F – – – – ++ – – – –Pythium irregulare – – – – + – – – –P. oligandrum – – – – – ++ ++P. ultimum – – – – + – – – –Rhizoctonia solani – – – – +++ – – – –Trichoderma viride – – – – – – – – –

Note: +++, strong inhibition of fungal growth; ++, good inhibition of fungal growth; +, some inhibition of fungal growth; –, no zone of inhibition; and– –, the fungus grows over the bacteria.

Table 2. Antagonism of the five selected bacteria against fungi in vitro.

Pseudomonas spp., and fluorescent Pseudomonas. With theexception of Bacillus spp., colonization of pozzolana grainsthroughout the cultural season was always higher in the fil-ter inoculated with the five bacteria (L1 to L5) than it was inthe control filter. The largest population was detected duringthe first month in the bacteria-amended filter, after whichthe size of the populations decreased regularly, and leveledoff in August and September (Figs. 2 and 3). In contrast,Bacillus spp. populations regularly increased in both theamended and control filters, from the beginning of the ex-periment (April) to its end (September) (Fig. 4).

Tomato yields in a soilless greenhouseFrom March to the end of April, there was no difference

in tomato yields in from the two areas of the greenhouse.After this period, from May to the end of the cultivation sea-son, the yield was always slightly lower in the area equippedwith the bacteria-amended filter. This is illustrated by theyield of tomatos at the end of September: 42.46 with the

bacteria-amended filter and 43.64 kg·m–2 with the controlfilter. However, the Newman–Keuls test at 95% confidenceshowed no statistical difference between these data (notshown).

Discussion

In soilless cultivation, disinfection of recycled water iscrucial for the control of plant pathogens. The technique ofslow filtration, used for more than 100 years to disinfect mu-nicipal water supplies (Graham and Collins 1996; Ellis1985), has been adapted to horticulture over the last decade(Ehret et al. 2001). During the slow filtration process, waterflows slowly through a bed of susbtrate, such as sand,rockwool, or pozzolana; mechanical and biological factorsare thought to be responsible for the effectiveness of the sys-tem (Ellis 1985; Weber-Shirk and Dick 1997). Although themicroflora colonizing the filters seem to play an importantrole in slow sand filtration, very few studies have investi-

© 2004 NRC Canada


April May June July August September

Control filterInfluent solution 5 43 3 103 82 102Effluent solution 0 0 0 2 0 0% elimination 99.9 99.9 99.9 98.1 99.9 99.9

Filter amended with bacteriaInfluent solution 24 42 38 110 83 88Effluent solution 0 0 3 1 0 1% elimination 99.9 99.9 92 99.1 99.9 98.9

Note: Pythium spp. thalles were counted on selective media, coded CMA-PARP, after incubating the plates for 48 h at 25 °C in the dark. Results areexpressed as the percentage of eliminated microorganisms.

Table 3. Efficiency with which control and bacteria-amended filters eliminated Pythium spp.


Control filterInfluent solution 1113 337 692 383 1130 2217Effluent solution 553 1817 1717 117 182 37% elimination 50.3 0 0 69.5 83.9 98.3

Filter amended with bacteriaInfluent solution 2040 1531 150 7667 225 567Effluent solution 17 8 3 32 2 0% elimination 99.2 99.5 98 99.6 99.1 99.9

Note: F. oxysporum thalles were counted on selective media, coded Komada, after incubating the plates for 48 h 25 °C in the dark. Results are ex-pressed as the percentage of eliminated microorganisms.

Table 4. Efficiency with which the control and bacteria-amended filters eliminated Fusarium oxysporum.


Control filterInfluent solution 4.2×104 1.2×105 1.3×105 4.8×104 5.2×104 1.9×105

Effluent solution 2.1×104 2.0×104 4.2×104 9.7×103 1.4×104 1.0×103

% elimination 50 83 67.7 79.8 73 99.5Bacteria-amended filter

Influent solution 4.3×104 2.1×105 3.4×104 3.2×104 3.3×104 3.4×104

Effluent solution 1.1×104 1.3×104 1.02×104 1.04×103 4.0×103 1.3×103

% elimination 74.4 93.8 69 96.7 88 96.2

Note: The number of bacteria in the solution were counted (in CFU/mL) on selective media, coded PCA, after incubating the plates for 48 h at 30 °Cin the dark. Results are expressed as the percentage of eliminated microorganisms.

Table 5. Efficiency with which control and bacteria-amended filters eliminated mesophylic bacteria microflora.

gated this role (Alsanius 2001; Brand and Wohanka 2001).Our work demonstrates, for the first time, that the inocula-tion of filters with five selected bacteria enhanced the bio-logical activity of slow filtration, and therefore increased theeffectiveness of the system.

The five bacteria used in this study were isolated from thetop layer of a filter proven to eliminate microorganisms (Reyet al. 1999). They were collected from this area, as they hadbeen in the studies conducted by Visscher et al. (1987),which reported a high volume of biological activity in a40-cm-deep area of the filter. In our study, the threeP. putida isolates displayed different properties than the twoB. cereus isolates. In the dual culture, where they were con-fronted with pathogenic and nonpathogenic fungi, only oneB. cereus isolate inhibited fungal development by antibiosis;the other isolates had nearly no impact on fungal growth. To

ascertain the degree to which our five bacteria would effectsoilless tomato cultivation, we tested their plant pathogenic-ity: at 106 CFU/mL, neither B. cereus nor P. putida isolatesshowed pathogenicity on tomato plantlets. However, at 109

CFU/mL, B. cereus induced root necrosis, whereas P. putidapromoted the growth of the root system. It should be notedthat the bacterial concentrations in the nutrient solutionsfound in our study, and in the study by Berkelman et al.(1995), were always been far below this level. The fivestrains we looked at can, therefore, be used to inoculate fil-ters used for cultivation.

After inoculating the filter with these unlabelled bacteria,the question of whether the detected bacteria are identical tothe introduced ones was raised. The fact that the biologicalactivity was higher in the bacteria-amended filter than in thecontrol one suggests the involvement of these bacteria in the

© 2004 NRC Canada

Déniel et al. 505

Fig. 2. Effect over time of mesophylic bacteria populations on pozzolana grains from bacteria-amended and control filters. Samples ofpozzolana grain were taken monthly, from April to September, from the top of the filter unit. Each month, bacteria were assessed inthree 7-g samples of pozzolana grain. A plate-count agar medium was used to count the mesophylic bacteria populations; results areexpressed as CFUs per pozzolana gram. The same letter on bars indicates no significant difference at P ≥ 0.05 (Duncan’s test).

Fig. 3. Effect over time of Pseudomonas spp. populations on pozzolana grains from bacteria-amended and control filters. Samples ofpozzolana grains were taken monthly, from April to September, from the top of the filter unit. Each month, bacteria were assessed inthree 7-g samples of pozzolanagrains. A cetrimide-fucidine-cefaloridine medium was used to count the Pseudomonas spp. populations;results are expressed as CFUs per pozzolana gram. The same letter on the bars indicates no significant difference at P ≥ 0.05(Duncan’s test).

high microbial colonization of the pozzolana grains col-lected from the top of the filter. The organic substancestrapped at the top of the filter may provide a continuous nu-trient supply for microorganisms, which might explain themaximal biological activity detected in the upper layer ofthe filter. During our cultivation season, the inoculated filterwas populated by more bacteria than the control one, andPseudomonas and fluorescent Pseudomonas were particu-larly abundant. Nevertheless, in both filters, these microbialpopulations decreased regularly from April to September, bywhich the time colonization of the pozzolana grains levelledoff. Bacillus spp. colonized pozzolana grains in a quite dif-ferent way: in both the bacteria-amended and control filters,Bacillus spp. populations were very limited in April, andsubsequently and regularly increased as the months passed. Thissuggests that Bacillus spp. are strong competitors, able to colo-nize grains already invaded by other bacteria. Pseudomonasspp. and Bacillus spp. likely act in different ways: nutrientcompetition seems to be the major mode of action for theformer, whereas direct parasitism seems to be of key impor-tance to the latter. This assumption is supported by the factthat Pseudomonads are able to develop in organic com-pound-rich areas. Our group recently observed that Pseudo-monas spp. from slow filters can specifically metabolizeamino acids (unpublished data). Alsanius et al. (1998) re-ported the same phenomenon with Pythium spp. Therefore,Pseudomonas spp. and Pythium spp. may compete for nutri-ents from amino acids in filters. Bacillus spp. show antago-nistic activity, in addition to their ability to induce a shift inbacteria colonies inside filters. Indeed, many Bacillus spe-cies are known to produce molecules with antimicrobial ac-tivities (Chérif et al. 2002; Edwards et al. 1994). In ourstudy, one B. cereus strain used to inoculate the filter wasable to greatly inhibit the growth of plant-pathogenic fungi.Further investigations will have to be performed to deter-mine the extent of this kind of interaction in filters.

Inoculating the new filter with five selected bacteria ele-vated the level of fungal elimination very significantly. Two

distinct effects on F. oxysporum and Pythium spp. should benoted. It took six months before the control filter reached itsmost efficient elimination of F. oxysporum (98.3% in Sep-tember); the bacteria-amended filter shortened this time, andeliminated F. oxysporum efficiently in the first month. Thisrate of elimination remained nearly constant, at about 98%to 99.9% during the entire cultivation season. In a previousexperiment, Brand and Wohanka (2001) found no differencebetween filters that were biologically activated and thosethat were sterilized against F. oxysporum f. sp. cyclaminis,and concluded that no biological factors are involved in thecontrol of this fungus. In contrast, our data show that the bi-ological activity against F. oxysporum, induced by the bacte-ria, was greatly enhanced. One possible explanation forthese contradictory results is that the respective properties ofthe different microorganisms eliminate different pathogensduring the filtration process. This hypothesis is supported byBrand and Wohanka (2001), who found that the eliminationof F. oxysporum was not enhanced, despite the effective re-moval of X. campestris pv. pelargonii. This suggests thattheir filter allowed efficient competition between microor-ganisms and bacteria, but not F. oxysporum. In our experi-ment, the fast colonization of the filter by the fiveintroduced bacteria and the establishment of colonies on thesurface of the pozzolana grain likely increased competitionamong microorganisms, which may explain the highly effi-cient elimination of F. oxysporum. The duration of operationof the slow filtration device also affects its effectiveness;F. oxysporum were eliminated most efficiently by the con-trol filter only after six months. Such a result can only beexplained by an evolution in biological factors. According toWohanka (1995), the filter-composing medium is biologi-cally activated by the recirculating nutrient solutions. Wenoticed that after the nutrient solution was highly colonizedby microorganisms, the recycling favoured the establishmentof bacterial colonies on pozzolana grains. In the control fil-ter, this resulted in more and more acute competition amongmicroorganisms and after six months of operation, the elimi-

© 2004 NRC Canada


Fig. 4. Effect over time of Bacillus spp. populations on pozzolana grains from bacteria-amended and control filters. Samples ofpozzolana grains were taken monthly, from April to September, from the top of the filter unit. Each month, bacteria were assessed inthree 7-g samples of pozzolana grains each. Antibiotic-amended glucose agar was used to count the Bacillus spp.

nation of F. oxysporum. Further investigation into the devel-opment and evolution of microorganism populations infilters is needed to confirm these two assumptions.

The process by which slow filtration eliminates Pythiumspp. likely differs from the one described for F.oxysporum.Whatever the sampling month, the level of Pythiumspp. elimination was very high (99.9%), both in bacteria-amended and control filters; these similar results suggestthat the effect that the biological factors of the microflorahave on Pythium spp. elimination, is reduced. According toVan Os et al. (1999), slow filtration enables the fixation ofPhythophthora cinnamomi zoospores to sand grains, whichsubsequently leads to fungal death. Because Pythium spp.,especially the Pythium group F, which are ubiquitous in to-mato soilless cultivation (Rey et al. 1998), produce highquantities of zoospores, their elimination should rely on aprocess very similar to the one for P. cinnamomi. However,in the system we used, the attachment of zoospores topozzolana grains makes them interact with bacterial colo-nies, which suggests the involvement of micoorganisms inPythium spp. control. Further experiments will have to beconducted to determine the role of physical and microbialfactors in Pythium spp. elimination.

Our study demonstrated that a certain proportion of thefive selected bacteria failed to colonize the filter. These bac-teria were driven to the plants by the nutrient solution,which raises questions about their ability to colonize therhizosphere. The bacteria possess either plant-growth-promoting activities or antagonistic properties; their benefi-cial effect on the root system is, therefore, a matter of specu-lation. Studies will have to be conducted to determine theirrole in filters and their effect on the root system of plants.

In conclusion, inoculating filters with selected bacteriaseems a promising way to control diseases in soilless culti-vation. Studies of bacterial colonization on pozzolana grainsand interaction among microorganisms will help improvethis technique.

Acknowledgements

This research work was financially supported by theBrittany and Pays de Loire Regional Councils (GIS-LBIOprogram) and the French Research Department (“Directionde la Technologie” No. 01B0419). We thank Dr. M.P.Friocourt for critical discussion of this work.

References

Alsanius, B.W., Khalil, S., and Hultberg, M. 1998. Biochemicaland chemical characterization of Pythium ultimum. Med. Fac.Landbouww. Univ. Gent. 63(3): 891–897.

Alsanius, B.W., Nilsson, L., Jensén, P., and Wohanka, W. 2001.Microbial communities in slow filters. Acta Hortic. 548: 591–601.

Berger, F., Li, H., White, D., Frazer, R., and Leifert, C. 1996. Ef-fect of pathogen inoculum, antagonist density, and plant specieson biological control of Phytophthora and Pythium damping-offby Bacillus subtilis Cot 1 in high-humidity fogging glasshouses.Phytopathology, 86: 428–433.

Berkelman, B., Wohanka, W., and Wolf, G.A. 1995. Characteriza-tion of the bacterial flora in circulating nutrient solutions of ahydroponic system with rockwool. Acta Hortic. 361: 371–381.

Brand, T. 2000. Beurteilung der bakterienflora von langsamfilternals biologische filtrationskomponente in geschlossenenbewässerungssystemen des gartenbaus. Geisenheimer berichte41. Ph.D. thesis, Technischen Universität München,Geisenheim. 112 p.

Brand, T., and Wohanka, W. 2001. Importance and characterizationof the biological component in slow filters. Acta Hortic. 554:313–321.

Chen, C., Belanger, R.R., Benhamou, N., and Paulitz, T.C. 1998.Induced systemic resistance (ISR) by Pseudomonas spp. impairspre- and post-infection development of Pythium aphaniderma-tum on cucumber roots. Eur. J. Plant Pathol. 104: 877–886.

Chérif, M., Sadfi, N., Benhamou, N., Boudabbous, A., Boubaker,A., Hajlaoui, M.R., and Tirilly, Y. 2002. Untrastructure andcytochemistry of in vitro interactions of the antagonistic bacteriaBacillus cereus X16 and B. thuringiensis 55T with Fusariumroseum var. sambucinum. J. Plant Pathol. 84: 83–93.

Edwards, S.G., McKay, T., and Seddon, B. 1994. Interaction of Ba-cillus species with phytopathogenic fungi — Methods of analy-sis and manipulations for biocontrol purposes. In Ecology ofplant pathogens. Edited by J.P. Blakeman and B. Williamson.CAB International. Wallingford, UK. pp. 101–118.

Ellis, K.V. 1985. Slow sand filtration. Crit. Rev. Environ. Control,15: 315–354.

Ehret, D.L., Alsanius, B., Wohanka, W., Menzies, J.G., andUtkhede, R. 2001. Disinfestation of recirculating nutrient solu-tions in greenhouse horticulture. Agronomie, 21: 323–339.

Goldberg, N.P., Stanghellini, M.E., and Rasmussen, S.L. 1992. Fil-tration as a method for controlling Pythium root rot of hydro-ponically grown cucumbers. Plant Dis. 76: 777–779.

Graham, N., and Collins, R. 1996. Advances in slow sand and al-ternative biological filtration. Wiley, Chichester, UK.

Hoagland, D.R., and Arnon, D.I. 1938. The water culture methodof growing plants without soil. Univ. Calif. Coll. Agric. Exp.Sta. Circ. Berkeley, Calif. pp. 347–352.

McPherson, G.M., Harriman, M.R., and Pattison, D. 1995. The po-tential for spread of root diseases in recirculating hydroponicsystems and their control with disinfection. Medische FaculteitLandbouwuniversiteit Gent, 60(2b): 371–379.

Poncet, C., Chave, M., Hericher, D., Julien, P., Brun, R., andBarthelemy, L. 2004. Chlorine disinfection of recyclingeffluents of rose crops: consequences on the rhizosphere bacte-rial communities. Acta Hortic. In press.

Postma, J., Bonants P.J.M., and Van Os, E.A. 2001. Population dy-namics of Pythium aphanidermatum in cucumber grown inclosed systems. Medische Faculteit Landbouwuniversiteit Gent,66(2): 47–59.

Postma, J., Van Os, E.A., and Kritzman, G. 1999. Prevention ofroot diseases in closed soilless growing systems by microbialoptimization. Medische Faculteit Landbouwuniversiteit Gent,64(3b): 431–440.

Postma, J., Willemsen-de-Klein, M.J., and Van Elsas, J.D. 2000.Effect of the indigenous microflora on the development of rootand crown rot caused by Pythium aphanidermatum in cucumbergrown in rockwool. Phytopathology, 90: 125–133.

Rey, P., Nodet, P., and Tirilly, Y. 1997. Pythium F induce a minorbut ubiquitous disease in tomato soiless cultures. J. Plant Pathol.79: 173–180.

Rey, P., Benhamou, N., and Tirilly, Y. 1998. Ultrastructural andcytochemical investigation of asymptomatic infection byPythium spp. Phytopathology, 88: 234–244.

Rey, P., Picard, K., Déniel, F., Benhamou, N., and Tirilly, Y. 1999.Development of an IPM system in soilless culture by using slow

© 2004 NRC Canada

Déniel et al. 507

sand filtration and a biocontrol agent, Pythium oligandrum.IOBC/WPRS Bull. 22: 205–208.

Rey, P., Déniel, F., Vasseur, V., Benhamou, N., and Tirilly, Y. 2001.Evolution of Pythium spp. populations in soilless cultures andtheir control by active disinfecting methods. Acta Hortic. 554:342–348.

Runia, W.Th. 1995. A review of possibilities for disinfection ofrecirculation water from soilless cultures. Acta Hortic. 382:221–229.

Stanghellini, M.E., and Rasmussen, S.L. 1994. Hydroponics, a so-lution for zoosporic pathogens. Plant Dis. 78: 1129–1138.

Steinberg, C., Moulin, F., Gaillard, P., Gautheron, N., Stawiecki,K., Bremeersch, P., and Alabouvette, C. 1994. Disinfection ofdrain water in greenhouses using a wet condensation water.Agronomie, 14: 627–635.

Tu, J.C., Papadopoulos, A.P., Hao, X., and Zheng, J. 1999. The re-lationship of Pythium root rot and rhizosphere microorganismsin a closed circulating and an open system in rockwool cultureof tomato. Acta Hortic. 481: 577–583.

Van Os, E.A. 1999. Closed soilless growing systems: a sustainablesolution for dutch greenhouse horticulture. Water Sci. Technol.39: 105–115.

Van Os, E.A., Amsing, J.J., van Kuik, A.J., and Willers, H. 1999.Slow sand filtration: a potential method for the elimination ofpathogens and nematodes in recirculating nutrient solutionsfrom glasshouse-grown crops. Acta Hortic. 481: 519–526.

Van Os, E.A., and Postma, J. (2000) Prevention of root diseases inclosed soilless systems by microbial optimisation and slow sandfiltration. Acta Hortic. 532: 97–102.

Visscher, J.T., Parmasivan, R., Raman, A., and Heijnen, H.A. 1987.Slow sand filtration for community supply: planning, designconstruction, operation and maintenance. International Rever-ence Centre for Community Water Supply and Sanitation, TheHague.

Walker, R., Powell, A.A., and Seddon, B. 1998. Bacillus isolatesfrom the spermosphere of peas and dwarf French beans withantifungal activity against Botrytis cinerea and Pythium species.J. Appl. Microbiol. 84: 791–801.

Weber-Shirk, M.L., and Dick, R.I. 1997. Physical-chemical mecha-nisms in slow sand filters. Am. Water Works Assoc. J. 89: 87–100.

Wohanka, W. 1995. Disinfection of recirculating nutrient solutionby slow sand filtration. Acta Hortic. 382: 246–255.

Wohanka, W., Luedtke, H., Ahlers, H., and Luebke, M. 1999. Opti-mization of slow filtration as a means for disinfecting nutrientsolutions. Acta Hortic. 481: 539–544.

Zhang, W., and Tu, J.C. 2000. Effect of ultraviolet disinfection ofhydroponic solutions on Pythium root rot and non-target bacte-ria. Eur. J. Plant Pathol. 106: 415–421.

© 2004 NRC Canada
