Mangoes were harvested from eight locations of Reunion Island, over a period of 10 weeks (Table 1). Reunion Island is characterized by a tropical climate, with average annual temperature of 25˚C and ferrallitic soils [44]. Annual rainfall largely varies depending on the location: from 500 - 1000 mm for West locations like Saint Paul, Grand Fond and Savana, to 1250 - 2000 mm for Bassin Plat and 2000 - 3000 mm for East locations like Cambuston, Quartier Français and Sainte Suzanne [45]. Mature mangoes, from 17 cultivars, were selected and manipulated without direct hand contact. Bacteria were collected from the fruit surface with a sterile cotton swab soaked in 10 mL of buffered peptone water [46]. By using a sterile pipette, the peptone water was collected and poured into a sterile

Table 1. Mango sampling and bacteria isolation. Both number and name of isolates are indicated depending on the medium of isolation, either MRS at 30˚C or NM at 12˚C.

Continued

tube. The microbial solution obtained was streaked over MRS (de Man, Rogosa and Sharpe) agar and Nutritive Medium (NM) plates and placed at 30˚C for 72 h and at 12˚C for 10 days, respectively. Colonies with different aspects were isolated on the same growth medium after microscopy examination. Isolated strains were stored at −80˚C in 20% glycerol.

Bacillus subtilis AvoGreen was used as a reference biocontrol strain [38].

Before use, bacteria were cultivated in broth, either MRS or NM depending on the isolation medium, at 25˚C during 72 h in an incubator with agitation at 100 rpm.

The MUCL 43868 strain of C. gloeosporioides (Penzig) from the pathogen collection of the Catholic University of Leuven (Leuven, Belgium) was used and cultivated on potato dextrose agar (PDA) medium at 27.5˚C. This strain was isolated from Mexican mangoes by GL Hennebert [47]. Strains CG Aust Mango 3-3 and CG Avocado 23-703 were obtained from the collection of Pr. Korsten.

An inoculum of C. gloeosporioides was grown on PDA plates over 10 days at 27˚C. Conidia were collected by pouring 10 mL of sterilized buffered peptone water on the plates and recovered by pipetting and adjusted to a final concentration of 10⁵ conidia per mL using a Malassez cell.

In a 96-well microplate, 180 µL of isolation medium broth and 20 µL of a 72 h bacterial culture were deposited. Each bacterial culture was loaded into 3 wells. Controls corresponded to 200 µL of isolation medium broth, distributed in 3 wells. Six batches of microplates were prepared and placed in incubator, with a lid on, at 12˚C, 25˚C, 30˚C, 37˚C and 42˚C, over 72 h without agitation. The absorbance was read for each microplate three times a day until 72 h of incubation with the microplate reader. For each time point, the plate was shaken for 15s, then the reading was performed at a wavelength of 600 nm and the collected OD was the mean of 10 reads of the same well. For each well, the growth curve was plotted over time and maximal growth rate (µ_max) was determined from OD slope over the growth phase. The optimal growth temperature was defined as the temperature leading to the highest µ_max.

After 72 h, each microplate incubated at 25˚C was emptied and rinsed in clear water, then 125 µL of a 0.1% of crystal violet solution was added in each well. After 10 min, the content of the wells was dropped off, rinsed again with clear water and then dried at open air. A volume of 200 µL of 30% acetic acid solution was added and 125 µL were taken off and poured in a new microplate. The microplate was read on a 550 - 680 nm wavelength range on the microplate reader and the wavelength rendering the maximal absorbance (590 nm) was chosen for the exploitation of the results. Biofilm strength is given by the ratio of absorbance between the bacteria and the control wells.

DNA extraction from bacterial strains was performed using the InstaGene Matrix commercial kit (Bio-Rad Laboratories, Hercules, CA, USA) [48]. For the amplification of the 16S rDNA region, the DNA primer pair was used: FD1-mod 5’-3’: AGAGTTTGATCHTGGCTCAG and RD1-mod 5'-3': GGMTACCTTGTTACGAYTTC [49]. The reaction volume was composed of 5 μl of purified DNA and 45 μl of a mixed solution composed of: 10 μl 5 × PCR buffer (Green Go Taq), 1 μl dNTP mixture (10 mM), 4 μl MgCl₂ (25 mM), 1.0 μl of each primers (FD1-mod and RD1-mod at 1 μM), 0.25 μl enzyme Go Taq DNA polymerase (5 U/µL) and qs of water. PCR amplification was carried out using Applied BiosystemsVeriti ™ Thermal Cycler. The thermal cycling program was: 3 min at 94˚C, followed by 35 cycles of [40 sec at 94˚C, 40 sec at 55˚C, 60 sec at 72˚C], and further for a time extension of 10 min at 72˚C. The quality of the amplification products was analysed on 2% TAE agarose gels after electrophoresis for 90 min at 110 V and staining with ethidium bromide.

2.5.1. Plate Assay for Mycelial Growth Inhibition

To assay mycelial growth inhibition, C. gloeosporioides MUCL 43868, CG Aust Mango 3-3 and CG Avocado 23-703 strains were used. Mycelium (0.5 mm × 0.5 mm) was spotted in the middle of a plate containing PDA medium and incubated for 21 days at 30˚C. Bacterial isolates were grown on nutrient agar (NA) incubated at 30˚C for 48 h. A suspension was created by gently scraping the bacterial lawn from NA. The OD at 600 nm of the suspension was adjusted to 1 unit (per mL) as assessed with PowerWave™ microplate spectrophotometer (BioTek). A volume of 100 µL was used for inoculation.

Two mycelium pieces were deposited on opposite sides of the PDA plates. The bacterial strains were inoculated as a central streak 48 h after inoculation of the fungus. Plates were incubated for 10 - 15 days at 30˚C, until the complete colonization of the plate surface by the fungus in the control plate (without bacteria). Thereafter, diameters of mycelium were measured to calculate a percentage of radial inhibition. Percentage of inhibition was calculated from the diameter of Colletotrichum colony in the presence of an isolate compared to the diameter without any inhibitor (0% of inhibition).

2.5.2. Inhibition of the Germination of Conidia

A spore suspension of MUCL 43868 strain of C.gloeosporioides was prepared by placing 10 mL of peptone water (or sterile distilled water) in the Petri dish containing a 15-day fungus culture. After filtration, the concentration of spores was adjusted to 5 log spores/mL using a Malassez cell.

Conidia germination inhibitory bacterial activity was evaluated by microscopy on special sterile slides. The first step was the deposit of 100 µL of melted PDA medium on the slides, followed by drying under a laminar flow hood for 1 hour. To carry out the test, 10 μL of spore suspension were deposited on the slide and 10 μL of each bacterial strain were added. The slide was then incubated at 27˚C inside a Petri dish lined with a double layer of moistened filter paper. Each treatment was repeated twice. The control corresponded to the suspension of spores in the absence of bacteria.

A conidiospore was considered to have germinated when a germinating tube of at least half the length of the conidiospore was observed under optical microscope. The qualitative data for the inhibition of germination of C. gloeosporioides MUCL 43868 were described as; (+): the germination of less than 25% of conidia was inhibited; (++): the germination of 25% - 50% of conidia was inhibited; (+++): the germination of 50% - 75% of conidia was inhibited; (++++): more than 75% of conidia germination was inhibited. Negative control showed no inhibition of conidia germination which corresponded to a germination rate close to 100%.

The ˚C. gloeosporioides strain MUCL 43868 was cultivated in Petri dishes for 21 days on PDA solid medium at 27.5˚C in the dark.

Mangoes cv. José were harvested and treated on the same day. A batch of 36 mangoes cv. José collected from the same orchard and with same maturity level (yellow point) was used. Fruit with no disease symptoms were selected, and the surface to be inoculated was washed with 70% ethanol and air-dried at room temperature. Subsequently, fruit inoculation for curative treatment was performed according to [50] by uniformly wounding (a cross: 2 mm deep and 10 mm wide) a relatively flat area in the middle of the fruit with a sterilized cork-borer and inoculating it with 20 µL of a spore suspension of C. gloeosporioides (10⁵ spores/mL). To ensure that anthracnose development was due to MUCL 43868, the peels from non-inoculated and inoculated fruits were placed on PDA amended with chloramphenicol and left for 8 days at 25˚C. The identification of the re-isolated fungi was based on morphological criteria. After fungal inoculation, fruits were maintained at room temperature for 24 h (21˚C, 85% relative humidity).

Bacteria Enterobacter sp. H222 and Leuconostocmesenteroides H255 were separately grown in nutrient broth for 3 days at 27.5˚C in a nutrient broth. Two suspensions corresponding to OD at 600 nm of 0.5 and 1.0 were prepared for each isolate.

From all 36 mangoes inoculated with C. gloeosporioides, six fruits were submerged in distilled water, to be used as non-treated positive controls. Three treatments were used to assess the impact of bacteria on anthracnose development. The thirty remaining mangoes were separated into three batches of 10 fruits corresponding to [H222], [H255], and [H222 + H255] treatments. Pre-inoculated mangoes were immersed in each bacterial suspension for two minutes. Fruit were stored at 20˚C with high relative humidity. After 10 days of storage, black spots on the surface of mangoes were measured using a digital caliper.

The experiment was repeated twice with independent batches of mango.

XLSTATsoftware (Addinsoft, Paris, France) was used for statistical analysis. K-means clustering and hierarchical clustering analysis (HCA) were performed with maximal growth rate, optimal growth temperature, ability to form biofilm and inhibition of mycelial growth as variables and isolates as observations. Variables were centred and normalized, and Euclidian distances were used. For k-means clustering, trace (W), i.e. pooled SSCP matrix, was used as classification criterion. For HCA, Ward aggregation method was used. Dendrogram was built through the Ward’s minimal distance algorithm. To compare inhibition of fungal growth, ANOVA was used with the REGWQ test and the bilateral Dunnet’s test was used forin vivo assay.

Bacteria were collected from 17 mango cultivars grown in eight locations on Reunion Island in order to cover the widest diversity of epiphytic mango bacteria available locally.

As mangoes are generally stored under refrigerated conditions, isolation of bacteria was performed from NM incubated at 12˚C to select psychrotrophic bacteria. Colonies of LAB were also recovered as these bacteria have a long history of use in food and some of them are already used to preserve foods. From mango surface, 305 bacteria were isolated: 99 were isolated from MRS, and 206 from NM (Table 1).

The most appropriate candidates for biocontrol of mango anthracnose were then selected without prejudice of the bacterial species and through a funnel-shaped approach.

The maximal growth rate, the optimal temperature for growth, the ability to form biofilms and the level of inhibition of C. gloeosporioides MUCL 43868 mycelial growth were determined for the 305 isolates. Two classification methods, k-means and HCA, were applied, resulting in similar results. Isolates were clustered into six classes, leading to an inter-classes’ variance of 79% of the total variance (Table 2 and Figure 1). This high inter-classes’ variance value implies that classes were clearly defined and well differentiated.

The most homogeneous class (class 6), i.e. with the lowest intra-class variance, gathered 24 isolates, characterized by low optimal growth temperature and low inhibition level. Class 5 included the highest number of isolates (133) and was also characterized by a low inhibition level of mycelial growth. Class 4, which gathered 18 isolates, harboured the same characteristics, but with the highest ability to form biofilm. Classes 1, 2 and 3 were the most interesting regarding the inhibition of mycelial growth, with an inhibition level of the gravity centre of 18.3%, 25.4% and 17.0% respectively. Those three classes differed by the number of isolates, class 1 being the largest. Class 1 was characterized by the highest µmax and optimal growth temperature. Class 2 was characterized by the highest mycelial growth inhibition level, an intermediate optimal growth temperature of 28.0˚C and the lowest ability to form biofilm. Class 3 gathered psychrophilic bacteria, with the highest ability to form biofilms. Among the three interesting classes, inhibition level was the lowest in class 3.

Figure 1. Dendrogram of 305 isolates classified into six classes by hierarchical clustering analysis (HCA) based on the variables “maximal growth rate”, “optimal growth temperature”, “ability to form biofilm” and “inhibition of mycelial growth”.

Table 2. Characteristics of classes of isolates, according to maximal growth rate (µmax), optimal growth temperature (˚C), biofilm strength and inhibition activity (%) against Colletotrichum gloeosporioides MUCL 43868 hyphal growth.

From the three classes showing the highest inhibition ability, 26 isolates were selected and identified by sequencing of the chromosomal region encoding 16S rRNA (Table 3). From the first class, 11 isolates were selected, representing 15% of the class size, whereas 12 isolates (43%) were selected from class 2 and 3 (10%) from class 3. No isolates from class 4 were selected in spite of the highest ability to form biofilm which is considered as an advantage to help the antagonist to colonize fruit surface [26], because of lower ability to inhibit the fungal pathogen.

Among those, 19 isolates were identified as Enterobacterales: eight belonged to the genus Enterobacter and seven were identified as Kosakoniacowanii. The other genera were Pantoea and Serratia. Besides, one isolate was identified as a Microbacteriaceae, Curtobacteriumluteum, and two Staphylococcus species were represented. More interestingly because of their safety for food use, three isolates were identified as L. mesenteroides and one as Gluconobacter sp.

Most of the species hereby identified were previously described as part of plant microbiomes. Enterobacter, Pantoea and Curtobacterium were identified from salad leaves [51]. Enterobacter, Pantoea, Leuconostoc and Curtobacterium were identified on tomato fruit or leaves surface [52][53][54]. The same genera, plus Staphylococcus and Gluconobacter were identified on mango tree leaves [55]. Gluconobacter was identified on grape surface, but Sphingomonas was the most abundant genus in this niche [56]. Interestingly, Kosakoniacowanii was here for the first time reported as isolated from carposphere. K.cowanii is the type species of Kosakonia, and was isolated from clinical and environmental samples, especially soil and trees [57].

Serratia marcescens is essentially studied for its role in nosocomial infections and for chitinase production [58][59]. For these reasons, isolates from the present work were not further studied. Four of the Enterobacter isolates belong to the Enterobacter cloacae complex, which gathers 12 species which cannot be easily differentiated from 16S rRNA coding region sequence [60][61]. Many

Table 3. Isolate identification and inhibition of C. gloeosporioides MUCL 43868 hyphal growth (%) and conidia germination.

¹mean ± standard deviation; ²(+): the germination of less than 25% of conidia was inhibited; (++): the germination of 25% - 50% of conidia was inhibited; (+++): the germination of 50% - 75% of conidia was inhibited; (+++): more than 75% of conidia germination was inhibited.

isolates from this complex originate from clinical samples [61][62][63], and their multidrug-resistance capacity raises concerns about their pathogenicity and virulence. However, many other isolates come from plant, possibly plant pathogens and isolates from this complex might play a role in biocontrol [64]. The Pantoea genus is widely distributed in nature and many species are described as epiphytes, endophytes or plant pathogens [65]. Many isolates, some being from the species Pantoeadispersa, have been used as biocontrol agents against post-harvest rots of fruit, onion or sweet potato [66]-[71]. Pantoeaagglomerans strain CPA-2 is an effective biocontrol agent (BCA) against the major postharvest pathogens present on pome and citrus fruits [72][73][74][75]. K.cowanii was classified in this new genus in 2013 from Enterobacter and based on genomic polymorphism analysis [55][76]. This species can act as a plant growth promoter, especially for sugar cane [77].

Curtobacterium was found as an endophytic bacterium of many plants including rambutan fruit [78][79][80]. It has been studied as a biocontrol agent for fungal brown rot of plum [27].

Among other LAB, L.mesenteroides is described for its antifungal activities and potential use for biocontrol [81][82]. This bacterium is commonly found in fermented foods, especially from vegetables and fruit [83][84].

Mycelial growth inhibition was in the range 20% - 40% for most isolates, the highest inhibition being observed for two Enterobacter isolates, two Kosakonia isolates and P. dispersa. Complementary to hyphal growth inhibition, isolates were tested for their ability to inhibit MUCL 43868 conidia germination (Table 3 and Figure 2). The results were listed in four classes according to isolate capability to inhibit germination, from low to strong inhibition [<25%, 25% - 50%, 50% - 75% and >75% of conidia inhibited]. Contrarily, conidia germination

Figure 2.In vitro slide test of inhibition of conidial germination of C. gloeosporioides MUCL 43868. The device used, germination of conidia under control condition, and germination of conidia in presence of bacteria isolates H182, H185, H186, H222, H232, H255 and H270 are shown. A magnification time of 400 was used.

inhibition was clearly more marked for Enterobacter and Leuconostoc (Table 3). Six isolates inhibited the germination of more than 75% of conidia: five Enterobacter isolates (H028, H068, H232, H222, H223), and L. mesenteroides H255. The K. cowanii H191 isolate inhibited the germination of 50% - 75% of conidia. Six isolates, four from K. cowanii (H182, H184, H192, H188), P. dispersa H186 and Serratia marcescens H177 inhibited 25% - 50% of conidia. The other isolates inhibited the germination of <25% of conidia.

From these results, isolates from species exhibiting good inhibition ability were selected to further assay inhibition capacity: Kosakonia H182 and H191, Pantoea H186, Enterobacter H222 and H232, and Leuconostoc H255. L. mesenteroides H270 was added to this pool of isolates as it has a safe use status as a LAB. B. subtilis AvoGreen was used as a reference. The eight isolates were tested against the hyphal growth of two other strains of C. gloeosporioides, one from South Africa collected from avocado and the other one from Australian mango (Table 4). Most of the isolates showed a hyphal growth inhibition activity in the range 19.3% to 24.9%. The isolate Kosakonia H182 exhibited a lower inhibition activity (10.0%) on the strain CG Avocado 23-703. AvoGreen and Enterobacter H222 showed inhibition activities above 30.4% against the two fungal strains, and Kosakonia H191 a 34.9% inhibition of hyphal growth of strain CG Aust Mango 3-3.

Whatever the fungal strain assayed, Enterobacter H222 was clearly the most efficient.

An in vivo assay was performed on a local variety of mango (cv. José), purposely injured and inoculated with fugal conidia. Selected bacteria, Enterobacter H222 and Leuconostoc H255, were added 24 h later and mangoes were stored at 20˚C.

Table4. Inhibition activity (%) of C. gloeosporioides CG Aust Mango 3-3 and CG Avocado 23-703 hyphal growth. Different letter in a column indicates a significant difference (p-value < 0.001).

¹mean ± standard deviation.

Symptom diameter observed after 10 days was slightly lower for bacteria treated spots, especially with isolate H222 (21.1 mm ± 0.9 mm for H222 versus 23.4 mm ± 1.3 mm for the control condition), but the significance was low (p-value = 0.141) (Figure 3). Combination of the two isolates did not enhance the observed effect.

Conditions applied before addition of bacteria, i.e. wounding of mature fruit and incubation with C.gloeosporioides for 24 h, are drastic as they strongly favour fungal development and activate fruit defence mechanisms.

Moreover, the influence of temperature is a crucial parameter for microbiome, as well as the maturity level of mango [85]. Enterobacter H222 was isolated at 12˚C but its optimal growth temperature is 27˚C. L. mesenteroides can also grow at low temperatures [86][87] and optimal growth temperature of isolate H255 is 30˚C. It is thus likely that a different effect would have been observed with less mature fruit stored at 12˚C.

Eventually, a pre-harvest application of fruit or application on unripe fruit and optimization of the bacterial population to spray on fruit surface should be considered. A more accurate identification and a careful examination of H222 isolate should be performed prior to consider any further development of biocontrol product in order to assess the safety of the use.

Figure 3. Severity of anthracnose caused by C. gloeosporioides MUCL 43868 in wounded cv. José mangoes treated with H222 [E. cloacae complex], H255 [L. mesenteroides], and H222 + H255 [E.cloacae complex and L. mesenteroides] and stored 10 days at 20˚C. Severity is expressed as diameter of symptom (mm) on mango.