Introduction

Fungicide application is an indispensable part of postharvest fruit processing, contributing significantly to the extension of shelf life of fruits during storage (Nguyen Van Long et al. 2016). However, the application of fungicides in fruit packaging plants results in the production of large wastewater volumes, containing fungicides like thiabendazole, imazalil, and fludioxonil (Campos-Mañas et al. 2019). Direct environmental disposal of these effluents is prohibited due to high aquatic toxicity of all the above compounds (European Food Safety Authority 2007; European Commission Report 2013; Commission Directive 2007/76/EC 2007).

Conventional municipal wastewater treatment plants are not effective in the removal of these fungicides (Masiá et al. 2013), resulting in the contamination of surface water systems (Fonseca et al. 2019). Various methods for the treatment of these effluents have been tested, including (i) filtration through activated carbon, which was effective against triazole compounds (Morin-Crini et al. 2018) and thiabendazole (Garcia-Portillo et al. 2004), (ii) oxidation with Fenton or photo-Fenton processes in the presence or absence of TiO2 (Jiménez et al. 2015; Santiago et al. 2016, 2018), and (iii) combined membrane bioreactor (MBR) process/Fenton (Sánchez Peréz et al. 2014) or photo-Fenton followed by a microbial degradation process (Loveira et al. 2019). Despite their high removal efficiency, advanced oxidation processes (AOPs) are not advocated by the industries, due to huge operating and construction costs and the production during treatment of several oxidized transformation products of unknown toxicity (Sirtori et al. 2014). Biological treatment of these effluents has been also proposed and biobeds, generic on-farm biopurification systems, showed high pesticide dissipation efficiency (Omirou et al. 2012; Karas et al. 2016). However, the limitation of biobeds to receive high wastewater loads points to the need of more engineered approaches for the treatment of those effluents.

All the above treatment methods have been evaluated for the removal of thiabendazole and imazalil, while little is known about their efficiency against fludioxonil. Fludioxonil (4-(2,2-difluoro-1,3-benzodioxol-4-yl)-1H-pyrrole-3-carbonitrile) is a phenylpyrrole fungicide recently introduced for the control of infestations by Penicillium, Botrytis, Rhizopus, Colletotrichum, and Monilinia (Schirra et al. 2005; Errampalli et al. 2006; Zhao et al. 2010; D’Aquino et al. 2013; Wiedemann et al. 2016) on apples (Xiao and Boal 2009), pears (Sugar and Basile 2011), citrus (Abad-Fuentes et al. 2014), and nectarines, apricots, and peaches (D’Aquino et al. 2007). It may be cytotoxic and cause endocrine disruptions (Teng et al. 2013; Brandhorst and Klein 2019; Lee et al. 2019). Mixtures of fludioxonil with difenoconazole or metalaxyl-M have been reported to induce toxic effects on the earthworm Eisenia andrei and the amphibian Rhinella arenarum (Svartz et al. 2018; Velki et al. 2019). Fludioxonil is considered persistent in the environment. Thomas and Hand (2012) reported a slow degradation of fludioxonil in surface water bodies. Papazlatani et al. (2019) assessed the dissipation of fludioxonil in soil at various application rates (10, 20, and 150 mg/kg) and noted a dose-dependent increase in its 50% dissipation time (DT50) (93, 120, and 276 days, respectively). Only a few studies have explored the removal of fludioxonil from wastewaters of variable origin. Hence, Fenoll et al. (2011) reported the photocatalytic oxidation of a fungicide mixture consisting of cyprodinil and fludioxonil in the presence of ZnO and TiO2. Moreover, a bioreactor system operating for a period of 30 days was used for the treatment of vineyard effluent containing fludioxonil (Esteve et al. 2009).

The high off-target toxicity and environmental persistence of fludioxonil, combined with the lack of validated methods for its removal from agro-industrial effluents, highlight the urgent need for the development and implementation of novel, preferentially biobased engineered systems for the treatment of fludioxonil-contaminated effluents. In this context, the main aim of the study was to develop and test a biological treatment method for the effective removal of fludioxonil from agro-industrial effluents. This was achieved in an immobilized cell bioreactor treating heavily contaminated fludioxonil-based wastewater. In addition, the succession in the bacterial community colonizing the bioreactor during bioprocessing was thoroughly investigated by using high-throughput amplicon sequencing analysis.

Materials and methods

Bioreactor setup

Biological treatment of fludioxonil was carried out in an upflow immobilized cell bioreactor of 500-mL working volume (total volume of 550 mL), which was filled with 150-mL porous glass beads. An oxygen diffuser linked to an air pump was installed inside the bioreactor, on the headspace of the unit, whereas upflow recirculation of the wastewater was performed by the use of a peristaltic pump (30 L/h) to provide adequate aeration to the immobilized cells (Supplementary Material Fig. S1). The bioreactor was inoculated with 50 mL activated sludge and operated under a hydraulic retention time (HRT) of 10 days, whereas dissolved oxygen (DO) concentration was always greater than 4.5 mg/L. A solution consisting of 250 mg/L fludioxonil, which served as the sole carbonaceous and nitrogenous feeding substrate, minerals, and trace elements (Atlas 2010), i.e., 61 mM Na2HPO4, 39 mM KH2PO4, 5 mM KCl, 2 mM MgSO4·7H2O, 2 mM CaCl2·2H2O, 0.13 mM MnSO4·4H2O, 1.65 mM FeSO4·7H2O, 3 μM NiCl2·6H2O, 1 μM Na2SeO3, 3 μM CoCl2·6H2O, 3 μM NaMoO4·2H2O, 2 μM ZnSO4·7H2O, and 1 μM H3BO4, was daily prepared to feed the bioreactor system.

Physicochemical analyses

Biochemical oxygen demand (BOD5) and chemical oxygen demand (COD) values were determined in the influent and effluent of the bioreactor according to the “Standard Methods for the Examination of Water and Wastewater” (Clesceri et al. 1998). Electrical conductivity (EC), pH, and DO values were measured through the use of Crison CM35, HANNA HI 98191, and WTW Oxi 320 m meters, respectively. Determination of total Kjeldahl nitrogen (TKN) and ammonium nitrogen (NH4+-N) was performed by employing the Kjeldahl and ammonium distillation method. A Cd-copperized column was used for nitrate reduction to nitrite, whereas nitrites were estimated colorimetrically at 453 nm in the presence of an sulfanilamide/(1-naphthyl)ethylenediamine-dihydrochloride indicator (Clesceri et al. 1998). Total nitrogen (TN) was measured by determining the concentration of TKN in the influent and the sum of TKN, NO3−-N, and NO2−-N concentrations in the effluent.

Analytical determination of fludioxonil and fluoride

Fludioxonil concentration in the influent and the effluent of the bioreactor was determined isocratically by high-performance liquid chromatography with photodiode array detection (ECOM, Czech Republic) using a 5 UniverSil C18 250 × 4.6 mm column (Fortis Technologies Ltd., UK). A mixture of acetonitrile/H2O (3/1, v/v) under a flow of 0.8 mL/min was used as the mobile phase. Fluoride concentration potentially released by degradation of fludioxonil was measured by ion chromatography using method APHA 4110 C (Clesceri et al. 1998).

DNA extraction, amplicon sequencing, and ecological analysis

Porous glass beads were collected at 3-week intervals to follow bacterial community succession in the bioreactor. The immobilized biomass was initially frozen and pestled to dust and used for DNA extraction by using the “NucleoSpin® Microbial DNA” kit (Macherey-Nagel GmbH & Co. KG, Germany), following manufacturer’s instructions. Universal primers new515F (5′-GTG YCA GCM GCC GCG GTA A-3′) and 909R (5′-CCC CGYC AAT TCM TTT RAG T-3′) were used to amplify the V4-V5 region of the 16S rRNA gene (Siles and Margesin 2017). Amplification of the 16S rRNA gene was carried out at “Mr DNA” (USA) through the use of “Qiagen HotStarTaq Plus Master Mix Kit” (Qiagen, USA) by employing a 3-min DNA denaturation step at 94 °C, followed by 30 cycles of 30-s DNA denaturation at 94 °C, 40-s annealing at 53 °C, and 1-min DNA elongation at 72 °C, after which a 5-min DNA extension step at 72 °C was performed. Polymerase chain reaction (PCR) products were purified through the use of Ampure XP beads (Pacific Biosciences, USA). Illumina sequencing was performed in a MiSeq platform producing 2 × 300 bp paired-end reads. Amplicons were subjected to demultiplexing and trimming, removing reads with Ν(s), abnormal length size, or low-quality score reads (Masella et al. 2012). Quality improvement of the assembled sequences, selection of the unique reads, clustering of the selected amplicons, and exclusion of chimera were carried out by using the -fastq_filter, -fastx_uniques, -cluster_otus, and -unoise3 options of USEARCH v.11, respectively (Edgar 2016, 2017). A total of 2,244,351 non-chimeric bacterial reads (five samplings of three replicates each) were sequenced and deposited in National Center for Biotechnology Information (NCBI) database under the BioProject PRJNA604457.

Permutational multivariate analysis of variance (PERMANOVA) (Anderson 2017) was performed through the calculation of the Bray and Curtis (1957) dissimilarity matrix. Principal coordinate analysis (PCoA) was used to illustrate beta diversity. The Shannon diversity index (H′) and the Shannon evenness index (SEI) were calculated, according to Magurran (1988), and one-way analysis of variance (ANOVA), followed by Duncan’s test, was used for evaluating the ecological indices and the relative abundances of bacterial taxa over time. Pearson’s correlation tests were performed to extract relationships among operating characteristics, whereas the Student’s t test at p < 0.05 or p < 0.01 was performed to identify statistically significant differences between influent and effluent physicochemical values. The construction of correlation matrix and dendrogram using Illumina sequencing data was performed through the use of “MicrobiomeAnalyst” tool (Dhariwal et al. 2017).

Results and discussion

Performance of the immobilized cell bioreactor treating fludioxonil wastewater

Influent and effluent pH values were stable at 7.0 ± 0.1 and 7.3 ± 0.2 throughout the bioprocessing. Influent EC was high (12.8 ± 0.2 mS/cm) due to the presence of mineral and trace elements in the solution added in the bioreactor (used to sustain microbial growth). Effluent EC gradually increased, reaching values as high as 20 mS/cm. Effluent EC patterns can be divided into three periods (i) first 2 months where EC increased up to 16 mS/cm, (ii) third month where EC increased up to 20 mS/cm, and (iii) from third month until the end of the operation where EC was stabilized at 18–18.5 mS/cm (Fig. 1a). This gradual but clear increase of EC in the effluent could be associated with the release of the fluoride ion during degradation of fludioxonil (see data below). Alexandrino et al. (2020) suggested the release of fluoride as an indication for the biological transformation of fludioxonil.

Fig. 1
figure 1

Electrical conductivity (EC) (a), COD (b), and fludioxonil (c) profiles during biotreatment of fludioxonil wastewater in the upflow immobilized cell bioreactor

Full size image

Influent COD was gradually reduced during the first 20 days, as the consequence of activated sludge adaptation to bioreactor’s operating conditions. Then, a significant COD decrease from 435 ± 31 mg/L in the influent to 119 ± 29 mg/L in the effluent was observed (p < 0.01, in Student’s t test), corresponding to COD removal of 72.6 ± 6.8% (Fig. 1b). Interestingly, COD removal exceeded 80%, i.e., 81.0 ± 2.4%, at the end of biotreatment.

Fludioxonil concentration was significantly decreased (p < 0.01) from 250 to 11.43 mg/L in the immobilized cell bioreactor, corresponding to fludioxonil reduction of 95.4 ± 4.0% (Fig. 1c). Determination of fluoride concentration by ion chromatography at the end of the biological treatment process (day 117) revealed that 94.0 ± 5.2% of fludioxonil-F was released in the effluent as fluoride ion, in accordance with the mean fludioxonil removal efficiency (95%). These results indicated that the F-containing moiety of fludioxonil was degraded during bioprocessing in the immobilized cell bioreactor, releasing fluoride in the effluent. In contrast to our findings, previous non-thermal plasma treatment of fludioxonil did not result in the cleavage of F-containing moiety of fludioxonil, suggesting its stability to non-biological processes (Misra et al. 2014).

There is limited knowledge about the biological dissipation of fludioxonil to date. In a recent study, Alexandrino et al. (2020) isolated microbial consortia capable of defluorinating 10 mg/L of fludioxonil within a period of 14–21 days. Esteve et al. (2009) reported the effective removal of 9.5 mg/L of fludioxonil from vineyard effluents treated in an activated sludge system operating for 30 days. Vischetti et al. (2012) reported a removal efficiency of 92.5% when 2.1 g of fludioxonil was treated in a biobed system for eight depuration cycles (approximately 90 days). Coppola et al. (2011) investigated the removal of a mixture of azoxystrobin, cyprodinil, and fludioxonil (50, 30, and 20 mg/kg, respectively) in a bio-organic mixture composed of wheat straw/pruning residues and reported a 50% dissipation time (DT50) of 70 days. In contrast to these studies, our biotreatment process was efficient in removing high concentration of fludioxonil (250 mg/L), relevant to the fungicide levels commonly present in agro-industrial effluents, in a shorter time period (HRT of 10 days). This should be attributed to the advantages of immobilized cell bioreactors, which prevent cell washout at low HRT, sustain high cell density, and operate under prolonged biomass retention time, a fact that favors the attachment of specialized slow-growers into immobilization carriers (Melidis et al. 2003; Gundogdu et al. 2013; Girijan and Kumar 2019; Navrozidou et al. 2019).

Despite their effectiveness, the use of fixed-bed bioreactors for the removal of other fungicides contained in agro-industrial effluents is limited. In particular, activated sludge bioaugmented with the triazole-degrading bacteria Shinella sp. NJUST26 and Sphingomonas sp. NJUST37 resulted in the depuration of 1H-1,2,4-triazole and tricyclazole under a hydraulic retention time of 4 days (Wu et al. 2018). Moreover, a biofilm reactor operating under the plug-flow mode was employed for the biodegradation and removal of carbendazim (Alvarado-Gutiérrez et al. 2017). Similarly, an effluent containing the pyrrole insecticide chlorfenapyr (15 mg/L) was recently reported to be effectively treated within a period of 5 days in a photo-reactor inoculated with a Rhodopseudomonas capsulata strain (Wu et al. 2020). Comparative evaluation of the efficiency of a microbial consortium to degrade pesticides in suspension vs. tezontle-packed upflow reactor documented the superiority of the latter in removing 25 mg/L methyl-parathion and 10 mg/L tetrachlorvinphos (Yáñez-Ocampo et al. 2011).

Effluent TKN concentration was gradually reduced during bioprocessing. In particular, TKN removal efficiency gradually increased from 43.0 ± 4.9% during the first 2 months to 80.1 ± 2.8% during the last month of bioreactor operation (Fig. 2). Despite that adequate aeration was provided, nitrite and nitrate concentrations in the effluent were negligible, possibly due to the assimilation of the nitrogen contained in fludioxonil. Regarding nitrification, inhibition of nitrifiers cannot be excluded. According to the European Food Safety Authority (2007), fludioxonil at concentration in soil of 1.3 mg/kg can result in a 20% inhibition of the nitrogen mineralization process. Moreover, a strong relationship among COD removal, TKN/TN removal, and effluent EC increase was identified (COD removal versus effluent EC; Pearson’s correlation coefficient r = 0.70, p < 0.01; COD removal versus TKN/TN r = 0.55/0.54, p < 0.05; TKN/TN versus effluent EC r = 0.55/0.54, p < 0.05).

Fig. 2
figure 2

Profiles of nitrogenous compounds during fludioxonil biotreatment in the upflow immobilized cell bioreactor

Full size image

Bacterial community succession in the immobilized cell bioreactor treating fludioxonil wastewater

The bacterial community succession during fludioxonil biotreatment was monitored by Illumina amplicon sequencing. A significant decrease (p < 0.05) in both Shannon diversity and evenness indices occurred with time, due to the acclimatization of the activated sludge to the selection pressure applied by the high fludioxonil concentration (Table 1). Robust beta diversity clusters were observed (Fig. 3), demonstrating the distinct bacterial community structure at the beginning and the end of fludioxonil biotreatment (Supplementary Material Fig. S2). Based on bacterial community profiles, three distinct periods were denoted (Fig. 4a; Supplementary Material Fig. S2): (i) the period of the first 2 months (period I, days 25 and 53), (ii) the third month (period II, day 77), and (iii) the last month of the experiment (period III, days 94 and 117). A shift in bacterial communities was observed at phylum/class level along fludioxonil biotreatment (Fig. 4a; Supplementary Material Table S1). Bacteroidetes and Gammaproteobacteria abundances showed complementary patterns throughout periods I, II, and III (Fig. 4a). This trend probably indicates changes in the composition of the lower taxa of these phyla. In comparison to period I, the relative abundances of Betaproteobacteria, Chloroflexi, Planctomycetes, and Firmicutes were reduced significantly (p < 0.05) during periods II and III. On the other hand, Alphaproteobacteria and Actinobacteria abundances increased significantly (p < 0.05) along the biotreatment from 11.07 ± 0.64% and 5.48 ± 1.72% at day 25 to 40.46 ± 1.17% and 9.77 ± 2.19% at day 117, respectively. Interestingly, Wu et al. (2018) reported a bacterial population shift, where Chloroflexi and Firmicutes decreased, while Actinobacteria increased, during the bioprocessing of carbendazim in an activated sludge tank supplemented with polyhedral hollow balls for the immobilization of the bioaugmented strains NJUST26 and NJUST37.

Table 1 Changes in ecological indices during biotreatment of fludioxonil wastewater in the upflow immobilized cell bioreactor
Full size table
Fig. 3
figure 3

Visualization of beta diversity clusters during biotreatment of fludioxonil wastewater in the upflow immobilized cell bioreactor

Full size image
Fig. 4
figure 4

Bacterial community shifts at phylum/class (a) and genus (b) levels during biotreatment of fludioxonil wastewater in the upflow immobilized cell bioreactor

Full size image

At the genus level, Thauera, Bellilinea, Pirellula, Phycisphaera, Flavobacterium, and Chitinophaga relative abundances decreased significantly (p < 0.05; Fig. 4b; Supplementary Material Table S2). On the other hand, a significant increase in the abundance of Empedobacter, Sphingopyxis, and Rhodopseudomonas from 0.67 ± 0.13% at day 25 to 34.34 ± 1.60% at day 117 (p < 0.05) was observed during bioprocessing (Fig. 4b; Supplementary Material Table S2). Figure 5 illustrates the main changes in bacterial communities during bioprocessing of fludioxonil wastewater, indicating the prevalence of Empedobacter, Sphingopyxis, and Rhodopseudomonas at the end of biotreatment. Bacteria of the genus Empedobacter (ex. Wauterseilla) have been identified (i) as members of the symbiome of Teleogryllus occipitalis crickets involved in the degradation of the insecticide chlorpyrifos (He et al. 2018), (ii) as degraders of polycyclic aromatic hydrocarbons (PAHs) in a terrestrial ecosystem (Cheng et al. 2017), and (iii) as carriers of various plasmids conferring resistance to antimicrobial agents (Hugo et al. 2006). Previous studies have identified bacteria of the genera Sphingopyxis and Rhodopseudomonas as pesticide-degraders (You et al. 2016; Lu and Lu 2018). Members of the family Sphingomonadaceae, to which the genus Sphingopyxis belongs, are versatile degraders of xenobiotic organic pollutants (Aylward et al. 2013), a capacity owed to their reservoir of plasmids encoding a multitude of xenobiotic degradation pathways for the catabolism of various pollutants, including pesticides (Stolz 2009; Kaminski et al. 2019). Similarly, Rhodopseudomonas strains, which are able to degrade a range of pesticides and toxicants (Wu et al. 2019), possess the capacity to mobilize and transfer plasmid-encoded and chromosomally encoded catabolic genes to other microbial strains (Sistrom 1977). We propose that the selection pressure induced to the bacterial community of the bioreactor by the high fludioxonil concentration shifted the composition of the bacterial community from common inhabitants of activated sludge like Thauera (Seviour and Nielsen 2010) to taxa that acquired the genetic mechanisms to cope and degrade the high concentrations of fludioxonil.

Fig. 5
figure 5

Heatmap illustrating the bacterial succession during biotreatment of fludioxonil wastewater in the upflow immobilized cell bioreactor

Full size image

Sphingobacterium, Pseudoxanthomonas, Candidatus phytoplasma and Microbacterium abundances also increased significantly, but to a lesser extent (Fig. 4b; Supplementary Material Table S2). Microbacterium spp. are capable of degrading fungicides, e.g., iprodione (Cao et al. 2018), while Sphingobacterium spp. are well-known pesticide-degraders (Abraham and Silambarasan 2013). Moreover, a Pseudoxanthomonas suwonensis strain HNM was able to rapidly degrade profenofos (Talwar and Ninnekar 2015). In addition, a strong relationship was revealed among the dominant taxa Sphingopyxis, Rhodopseudomonas, and Pseudoxanthomonas as well as Empedobacter, Afipia, and Microbacterium (Fig. 6).

Fig. 6
figure 6

Relationships among the major bacterial taxa identified in the upflow immobilized cell bioreactor during biotreatment of fludioxonil wastewater

Full size image

Based on the amplicon sequencing data obtained, the relative abundance of ammonia-oxidizing bacteria, consisting mainly of Nitrosomonas and Nitrosococcus and to a lesser extent from Nitrosospira and Nitrosovibrio, decreased significantly from 2.48 ± 0.65% to 0.11 ± 0.01% (Table 2), whereas the respective abundance of nitrite-oxidizing bacteria, comprising of the genera Nitrospira, Nitrobacter, Nitrolancea, and Candidatus Nitrotoga, remained at low level, even though Nitrolancea increased over Nitrospira population. Such reduction in the proportion of ammonia oxidizers could be possibly attributed to the inhibitory effects of the fungicide applied and/or its break-down compounds. Ammonia-oxidizing microorganisms were identified as the most sensitive functional microbial group to pesticide exposure in soil (Karas et al. 2018). Furthermore, studies with fungicides used in fruit packaging industry suggested a strong toxicity of the metabolic products of the pesticides ethoxyquin and iprodione on soil ammonia oxidizers (Papadopoulou et al. 2016; Vasileiadis et al. 2018).

Table 2 Ammonia-oxidizing bacteria (AOB) and nitrite-oxidizing bacteria (NOB) relative abundances over time. Letter in common within the same row denotes non-statistical differences (in Duncan’s test)
Full size table

Indeed, a specialized part of the activated sludge microbiota favored during bioprocessing of fludioxonil wastewater in the immobilized cell bioreactor, due to biomass retention on porous beads. It is worth noting that Esteve et al. (2009) reported the dispersion of biomass, due to the damage of the activated sludge flocs from the pesticide application during treatment of vineyard effluent containing pesticide residues in a suspended solid bioreactor.

Conclusions

Biotreatment of fludioxonil in the immobilized cell bioreactor resulted in high removal efficiency, exceeding 95%, which concurred with the release of 94% of the fludioxonil-fluoride in the effluent, suggesting a degradation of the fluoride-containing moiety of the fungicide during bioprocessing. Moreover, a strong statistical relationship among COD removal, TKN/TN removal, and effluent EC increase was identified. Betaproteobacteria, Chloroflexi, Planctomycetes, and Firmicutes abundances were decreased, whereas Alphaproteobacteria and Actinobacteria proliferated along biotreatment. A significant increase in the abundance of Empedobacter, Sphingopyxis, and Rhodopseudomonas was denoted, implying their involvement in the degradation of fludioxonil during biotreatment.