Short-term exposure to benzalkonium chloride in bacteria from activated sludge alters the community diversity and the antibiotic resistance profile

The use and release of antimicrobial substances into the environment from urban uses (houses, hospitals, factories, for example) and agricultural activities (horticulture, aquaculture, and livestock production) cause concern and require urgent attention. Large amounts of disinfectants were released into the environment before the COVID-19 pandemic. Over 450,000 kg per year of cleaning products, as quaternary ammonium compounds (QAC), were manufactured or imported alone in the USA (Hora et al. 2020). Among the QACs, benzalkonium chloride (BAC) is the most common surfactant, composed of a mix of chlorides of alkyldimethylbenzylammonium. The primary mechanism of action involves a general perturbation of lipid bilayers of membranes leading to a generalized and progressive leakage of cytoplasmic materials to the environment (Gilbert & Moore 2005). Owing to its action mechanism, BAC is one of the most recommended disinfectants against the SARS-CoV-2 virus (US Environmental Protection Agency 2020).

BAC concentration in domestic wastewater treatment plants (WWTPs) was estimated at 9.9 μg/L, and 2.2–2.8 ml/L in effluents from hospitals (Martínez-Carballo et al. 2007), but currently, higher BAC concentrations are expected due to the COVID-19 pandemic (Hora et al. 2020). Biodegradation and adsorption are the major removal pathways at WWTPs, but some studies have revealed incomplete degradation and complete compound adsorption in the activated sludge (Clarke & Smith 2011). The activated sludge (AS) constitutes the principal agent used for the biological purification of industrial and domestic effluents; some AS functions include nitrogen fixation, nitrification, ammonification, and other biochemical processes (Gernaey & Sin 2008). Also, the microbial structure of AS influences receiving water bodies (Numberger et al. 2019). The microbial composition of each AS is shown to be specific for each geographical area and the organic substrates of each sewage; however, the main phylum found in AS is Proteobacteria, followed by the phyla Firmicutes, Bacteroidetes, and Actinobacteria (Zhang et al. 2012).

Some studies in temperate regions have shown that the environmental concentrations of BAC can alter the microbial communities from activated sludge and interfere with the depuration process; concentrations higher than 2.0 mg BAC per gram of solids inhibited enzyme activity, and a long-term exposure reduced the microbial community diversity and selected for BAC-resistant bacteria as Pseudomonas (Chen et al. 2018). A recent review about disinfectant resistance includes qac genes as important efflux pumps that can export harmful molecules as disinfectants and antibiotics outside the bacterial cell. These genes might be either chromosomally encoded or in plasmids and can be eventually transferred to other bacteria by conjugation and transduction process and are involved in the spread of antibiotic resistance genes in the environment. Moreover, disinfectant and antibiotic resistance genes can be integrated by site-specific recombination in an aatI site of class I integrons, composed by an integrase gene (intI) followed by disinfectant (qacE), and antibiotic resistance genes as sul family genes (McCarlie et al. 2020). The WWTPs receive disinfectants, antibiotics, and heavy metals; they generate a selection pressure for antibiotic-resistant microorganisms. For this reason, they are well recognized as hot spots of antibiotic resistance spread (Karkman et al. 2018).

In Costa Rica, the most common system for wastewater treatment is AS, since tertiary or quaternary disinfection steps are not mandatory in national legislation (Ruiz Fallas 2012; Mora-Alvarado & Portuguez-Barquero 2016). This study explores the impact of a higher BAC exposure to a bacterial community from activated sludge from a municipal WWTP in Costa Rica, regularly exposed to low BAC concentrations. Additionally, we analyzed changes in bacterial composition (using 16S rRNA gene sequencing) and the antibiotic resistance genes load (quantitative PCR detection of intI1, sul2, and qacE/qacEΔ genes).

Benzalkonium chloride (BAC) (≥95.0% Fluka 12060 Sigma), methanol (MeOH) (>99.8% grade HPLC, Lot 1687318324, Merck 1.06018.4000 (DS228)), hydrochloric acid (HCl) (37% Merck GR 37.2500 133 K16502817), dimethanechloride (DMC) (Merck 1.06054.4000, Lot 1584154 114, G.C. grade, purity >99.8%), sodium sulfate (NaSO4), acetonitrile (ACN), methanol and formic acid, Optima LC-MS grade was purchased from Fisher Chemical, and water was purified by using a Thermo Scientific system (OH, USA). The benzalkonium chloride (BAC) standard was obtained from Sigma Aldrich (≥95.0% Fluka 12060).

As a representative of the most common type of WWTPs in Costa Rica, we selected a small size residential plant (serving less than 4,000 inhabitants, see operative details in the supplementary material S1) located in the Costa Rican Central Valley (9°55′14″N, 84°14′34″W, 1,400 m above sea level). We collected a sample of the AS (4 L) from an aeration tank (4 m deep) using a metal bucket at approximately 50 cm of depth. The sample was then transferred to a sterile amber glass container, kept at a temperature of 4 °C, and immediately transported to the laboratory. Later, the sample was divided into three portions corresponding to three treatments. The first portion (T0) was frozen immediately and used as a baseline. For replication purposes, the second portion (T1) was homogenized, distributed in 3 aliquots of 500 mL in Erlenmeyers of 1 L, and kept at 20 °C (environmental temperature) with aeration for 12 hr until the enrichment test was performed. T1 sample was enriched with a nominal concentration of 10 mg BAC/L; for the sludge dosing, we used a solution of 100 mg/L of BAC diluted with ultrapure water. For dosed concentration and exposition time selection, previous pilot studies were carried out using oxidation substrates rates’ changes as a parameter for measuring global BAC-induced variations in bacterial community behavior (data are not shown). The three containers were enriched and incubated with aeration at a temperature of 20 °C for 96 hr. The third portion (T2) was used as process control, and followed the same protocol as T2 except that it did not have enrichment with BAC. The general experiment procedure is described in Figure 1.

BAC extraction was carried out from all samples enriched (T1) and without dosing (T0 and T2). Twenty mL (triplicates) of each sample were filtered using vacuum equipment, retaining the solids with a 47 mm fiberglass filter (VWR). Further details of the extraction are described in the supplementary material. The last concentration step was carried out under N₂ flow. The residue was re-suspended in 1 mL of Acer (these tubes had been previously weighed and were weighed again after adding the Acer). The extracts were filtered through a PDVF syringe of 0.22 μm and transferred to previously weighed vials. For BAC quantification, all samples were analyzed using an Acquity Ultra Performance Liquid Chromatography system (UPLC), consisting of Waters Acquity binary solvent manager, autosampler, and Photodiode Array Detector (PDA) coupled with a Quadrupole Time of Flight (Q-ToF) (Waters Synapt G1), (Waters Corp., Milford, MA, USA) in series. The details of the chromatographic process are described in the supplementary material. The three BAC homologs were identified according to their retention time and high-resolution molecular mass determination, and comparisons with reference standards were done.

Three independent DNA extractions were performed from enriched sludge (T1) and the non-enriched control sludge (T2) samples after 96 hr experiment, and each of the three sub-samples was analyzed (T1 n=9 and T2 n=9). Also, six independent DNA extractions were carried out to the baseline sludge (T0 n=6). All DNA extractions (T0, T1, and T2 samples) were conducted in parallel. We used the NucleoSpin^® Tissue Kit (Macherey Nagel^®, Germany) to extract DNA, following the manufacturer's protocol using 1 mL of sludge sample previously homogenized. The V1–V3 regions of the 16S rRNA gene were sequenced using primers 27F and 518R primers with a Roche GS FLX. Sequencing services were provided by Macrogen (Seoul, Republic of Korea).

For antibiotic resistance gene assays, quantitative PCR (qPCR) for qacE/qacEΔ1, intI1, sul2, and 16S rRNA were assessed using StepOnePlus™ Real-Time PCR thermocycler (Thermo Fisher, USA). Standard calibration curves were carried out using each gene's purified, quantified amplicon as previously described (Di Cesare et al. 2013). Each gene amplicon was visualized by electrophoresis (60 min at 60 V, 1.5% agarose gel), the amplicon extraction was carried out with NucleoSpin Gel PCR Clean-Up (Macherey-Nagel, USA) according to the manufacturer's instructions and quantified by NanoDrop 2000c spectrophotometer (ThermoFisher, USA). With the absolute quantity of DNA for each amplicon, a gene copy number estimation was conducted using the theoretical molecular mass of each amplicon sequence, according to the sequences deposited in the NCBI database. Multi-resistant Escherichia coli PTA-A0653-2 (GeneBank WAAM01000041.1) was used as a positive control for all studied genes. For qPCR assays, PowerUp™ SYBR™ Green Master Mix (Applied Biosystems™, USA) was used according to the manufacturer's instructions, 5 μL of DNA of each sample was used. The qPCR program, primers, primers’ concentration, and expected amplicon are detailed in Table 1. Melt curve analysis was performed from 60 °C to 95 °C with a continuous increment of 3% °C. Efficiencies and R² averages for the tested genes were 90.78% and 0.999 for qacE/qacEΔ1, 80.81% and 0.989 for intI1, 96.02% and 0.992 for sul2, and 103.64% and 0.976 for 16S rRNA. The limits of quantification for each gene were the minimum concentration detected with a standard linear curve (Di Cesare et al. 2015), for all assays were ten copies μL⁻¹. The abundances of the different genes were expressed as gene copies mL⁻¹.

The 16S rRNA gene sequences were paired and quality filtered using Mothur v1.39.5 (Kozich et al. 2013). Subsequent processing was performed with the SILVA NGS v1.3 pipeline (Quast et al. 2013), including the alignment against the SILVA SSU rRNA SEED using SINA v1.2.10 (Pruesse et al. 2012), operational taxonomic unit (OTU) clustering at a 0.03 distance cut-off with Cd-hit v3.1.2 (Li & Godzik 2006), and taxonomic classification by local nucleotide BLAST search against SILVA SSU Ref dataset 132 using blastn (Camacho et al. 2009). This process resulted in 274,368 bacterial sequences (sample average=23,833, range=5,522 to 27,966).

The statistical analyses and visualizations were performed in R (R Core Team 2019). We used Vegan (Oksanen et al. 2019) to calculate alpha diversity estimators (richness and Shanon), the non-metric multidimensional scaling analyses (NMDS), and to perform the permutational analysis of variance (Permanova) on OTU tables normalized to the relative abundance of each sample. Indicator Species package (De Cáceres 2013) was used for indicator species analysis. Additionally, to analyze differences between the abundance of each studied gene and diversity indexes, a Kruskal–Wallis non-parametric test was done. Previously, each antimicrobial resistance gene load was normalized by 16S rRNA gene load to obtain a relative gene load used in the Kruskal–Wallis test (Thorsten 2021).

We confirmed the presence of BAC in the three types of samples of analyzed AS (treatments T0, T1, and T2). The background concentration of the compound in the sludge before treatment (T0) was 1.46 mg/L, and after 96 hr incubation (T2) the concentration was 3.14±1.5 mg/L. The BAC concentration of the enriched sample (T1) was 14.19 mg/L immediately after exposure, and after 96 hr of exposure, the concentration decreased to 2.27±0.67 mg/L.

Significant changes occurred in antibiotic resistance-associated genes present in the sludge after treatment with BAC (Figure 2, Table S1, supplementary material). qacE/qacEΔ1 was significantly higher in the treated sludge (T1) compared to the T0 and T2 (ρ=0.0066). Conversely, the load of intI1 was higher in the original sludge (T0) in comparison with both the samples that were exposed to BAC (T1) and the laboratory experiment control (T2) (ρ=0.0096). No differences in sul2 gene load were found among samples (ρ=0.8061).

The microbial composition analyses identified 1,127 bacterial OTUs from the 274,368 sequences analyzed. In general, Proteobacteria was the most abundant phylum in the sludge samples analyzed, representing around 53% of the sequences. This group was followed by Bacteroidetes (11%), Chloroflexi (7%), Planctomycetes (7%), and Acidobacteria (3%), while nearly 10% of the bacterial sequences could not be assigned to any phylum (Figure 3). Within Proteobacteria, Alphaproteobacteria represented 26.6% of the total sequences while Gammaproteobacteria represented 23.7% and Deltaproteobacteria 2.7%. Some differences can be observed between each treatment (Table 2). For example, in dosed samples (T1 treatment), the presence of Alphaproteobacteria is higher than other treatments; meanwhile, other phyla such as Bacteroidetes, Patescibacteria, and Planctomycetes decreased.

Richness diversity index revealed a significant diversity decrease between the baseline sludge (T0, 362±53) and the samples from the controlled exposure T1 (259±41) and T2 (263±62) (ρ=0.0129). Shannon diversity presents a tendency to decrease between treatments, T0 (4.64±0.05), T1 (4.59±0.07), and T2 (4.54±0.10); however, non-statistical differences were found (ρ=0.1062). In this context, the application of indicator species analysis corroborated differences between the three sample groups (Table 3). For example, T1 samples only present Methyloversatilis sp. as indicator species of the BAC exposure, T2 samples showed Aminobacter sp. as an indicator, and T0 samples showed high specie number. Table 3 summarizes the identified indicator species.

Finally, NMDS analysis showed differences between the three analyzed microbial communities. The samples within the experimental treatments clustered together more than the baseline sample, but still T0 (baseline community), T1 (enriched with 10 mg BAC/L), and T2 (non-enriched control) were separated from each other (Figure 4). The clustering was consistent with the Permanova analysis, which determined significant differences (p=0.001, α=0.001) between the three treatments, showing the impact of BAC exposure on the structure of the bacterial communities. Concomitantly, we also observed a reduction in the values of α-diversity indexes in both T1 and T2 compared to the baseline community (Figure 5).

The study results showed that BAC concentration in activated sludge samples induced changes in the antibiotic resistance-associated genes studied. It is worth mentioning that BAC was detected in the baseline samples (T0) indicating this compound's frequent and extensive use. A significant increase in the copy load of qacE/qacEΔ1 was determined in the samples exposed to BAC (T1, Figure 2) concerning the other treatments. This gene copy increase suggests a specific selection process by the presence of BAC. Previously, qacE/qacEΔ1 was related to resistance to quaternary ammonium compounds as it encodes for an efflux pump of the SMR family (Bay et al. 2008). This gene usually is present in mobile genetic elements such as integrons and plasmids (Chuanchuen et al. 2007). Accordingly, other studies have shown a similar response in microbial communities exposed to BAC (Kim et al. 2018; Yang & Wang 2018).

Concerning the intI1 gene, its load was higher in T0 samples (baseline), whereas the sul2 gene did not show differences between treatments; although class 1 integrons carrying intI1 gene are usually associated with sul genes conferring resistance to sulfonamides. There are some differences among the sul genes: sul1 usually is linked to other antibiotic resistance genes in class 1 integrons, while sul2 is generally located on small nonconjugative or large transmissible multi-resistance plasmids (Sköld 2000; Antunes et al. 2005). The prevalence of sul genes may explain, at least partially, the results observed in this study. In addition, other factors should be considered for our findings, including: (1) the reduction in the diversity of microbial communities could be related to the reduction of the antibiotic resistance gene loads; (2) the surviving mechanisms related to BAC's response are general and, in some situations, are not related to antibiotic resistance response; and, (3) the microbial communities are highly resilient to antimicrobial substances like BAC.

In the studied Costa Rican municipal WWTPs, we found antimicrobial resistance genes in a similar relative load to that found previously in other latitudes (Di Cesare et al. 2016). The intI1 gene has been used as a marker of anthropogenic pollution since it is commonly linked to disinfectants, antibiotics, and heavy metal resistance genes, it has penetrated pathogenic and commensal bacteria from humans and animals, and its abundance can rapidly change in response to environmental pressures (Gillings et al. 2015). Our results indicate the possible spread of the intI1 gene in AS and the associated risk of resistance genes’ horizontal transference in this ecosystem.

The sludge community composition obtained in this study is consistent with previous studies where Proteobacteria (and, more specifically, Alphaproteobacteria), followed by Bacteroidetes, Acidobacteria, and Chloroflexi are predominant phyla in AS (Xia et al. 2018). In addition, previous studies have shown a lower microbial diversity in BAC-degrading communities than non-exposed communities (Oh et al. 2013). Those findings are consistent with our results, showing a reduction of the richness and α-diversity in the samples exposed to BAC (Figure 4).

The most abundant genera in all treatments (Table 2) are associated with nitrogen fixation, nitrifiers, denitrifiers, methylotrophs, and others associated with bulking in WWTP. However, no functional or genetic analyses were performed to assess microbial nitrogen metabolism in the studied samples. For example, Nitrospira, a well-known nitrifier (Tian et al. 2017), decreased its relative abundance in BAC-enriched sludges. In this context, nitrification has been found to decrease at BAC concentrations of 2 mg/L (Hajaya & Pavlostathis 2013). On the other hand, Rhodobacter and Methylorosura (Alphaproteobacteria) increased in the BAC-treated sludge. Rhodobacter is a genus with photosynthetic capacity and nitrogen fixation and assimilation capabilities (Mackenzie et al. 2007), while Methylorosula is a methylotrophic bacterium associated with low temperatures (Berestovskaya et al. 2012).

Rhodobacter and Rhodopseudomonas (also Alphaproteobacteria) are involved in the denitrification process during AS treatment (Lu et al. 2014). A previous study showed that the denitrification process efficiency was reduced linearly to 64% in a mixed culture of nitrate-reducing bacteria exposed to 50–100 mg/L of BAC (Hajaya & Pavlostathis 2013). Previous studies have shown that these genera also share BAC resistance features in their annotated genomes: sugE gene presence in Rhodopseudomonas palustris genome (Larimer et al. 2004) and SMR efflux transporters’ genes in Rhodobacter sphaeroides and R. capsulatus genomes (Kontur et al. 2012; Ding et al. 2014). As previously mentioned, the SMR pumps such as sugE and qacE are described as transporters that confer resistance to BAC and other quaternary ammonium compounds (Zou et al. 2014).

Using indicator species analysis (Table 3), we identified the genus Methyloversatilis (Rhodocyclaceae, Proteobacteria) as an indicator in BAC's exposed samples. This genus can be responsible for denitrification, nitrogen fixation, and the assimilation of single carbon compounds (Smalley et al. 2015). In the context of our study, the role of Methyloversatilis could be associated with processes related to the degradation of quaternary amines (such as benzalkonium chloride in AS, mainly when BAC is the primary source of nitrogen and carbon resources). Methyloversatilis sp. as indicator species is consistent with previous studies showing the impact of BAC exposure in the nitrogen cycle (Hajaya & Pavlostathis 2013). Additionally, the whole genome sequence of Methyloversatilis universalis FAM5 shows 20 annotated genes related to efflux pumps and a sugE gene (Kittichotirat et al. 2011).

We hypothesize that the increase in the relative abundance of putative denitrifiers is part of the adaptability response of the AS microbial communities when they are exposed to BAC in concentrations below 50 mg/L. Further measurements of the nitrification-denitrification potential under these concentrations and BAC resistant phenotype confirmation should be performed in parallel to confirm this hypothesis. Finally, we found some genera previously described as BAC degraders, such as Pseudomonas and Achromobacter (Ertekin et al. 2016); however, these were found in abundances below 1%.

Our findings confirm that the use of cleaning products containing BAC, at a domestic level, can alter bacterial communities in activated sludges of a WWTP, as that studied in a tropical country such as Costa Rica. Furthermore, it suggests that BAC can alter some antibiotic resistance genes of the bacterial community and select some bacterial groups that can replace traditional microorganisms to maintain the nitrogen cycling in the microbial community. Nevertheless, the ultimate effects of these disinfectants on the ecology and evolution of tropical aquatic communities should be further studied.

This work was funded by CONARE and Vicerrectoría de Investigación of Universidad de Costa Rica. We thank Dr José Bonilla from CIBCM-UCR for his valuable suggestions.

All relevant data are included in the paper or its Supplementary Information.