Fate and persistence of antibiotic-resistant bacteria and genes through a multi-barrier treatment facility for direct potable reuse


 Given the availability of technological solutions and guidelines for safe drinking water, direct potable reuse of reclaimed water has become a promising option to overcome severe lack of potable water in arid regions. However, the growing awareness of the presence of antibiotic-resistant bacteria (ARB) and antibiotic resistance genes (ARG) in corresponding raw wastes has led to new safety concerns. This study investigated the fate of ARB and intracellular and extracellular ARG after each treatment step of an advanced water treatment facility in Windhoek, Namibia. The New Goreangab Water Reclamation Plant (NGWRP) produces drinking water from domestic secondary wastewater treatment plant effluent and directly provides for roughly a quarter of Windhoek's potable water demand. Procedures to study resistance determinants were based on both molecular biology and culture-based microbiological methods. TaqMan real-time PCR was employed to detect and quantify intracellular resistance genes sul1, ermB, vanA, nptII and nptIII as well as extracellular resistance gene sul1. The NGWRP reduced the amount of both culturable bacterial indicators as well as the resistance genes to levels below the limit of detection in the final product. The main ozonation and the ultrafiltration had the highest removal efficiencies on both resistance determinants.


GRAPHICAL ABSTRACT INTRODUCTION
Population growth in various countries of the world together with the effects of climate change lead to a continuous increase of the pressure on the availability of water resources for irrigation of agricultural crops and potable water (EPA 2012). The use of reclaimed water can help address the resulting challenges by regarding wastewater (WW)/used water (UW) as a valuable key resource that can be recovered rather than be considered a waste product.
An alternative to relying on traditional raw water sources for water supply therefore is direct potable reuse (DPR), a process in which the reclaimed water is not re-introduced into the environment (groundwater or surface water) but directly reused for drinking water supply and further human consumption (USEPA 2018). DPR is a suitable method to augment water supplies when necessary, and technologies exist that reliably produce safe potable water that continuously meets drinking water regulations (du Numerous studies show that WW is laden with antibiotic-resistant bacteria (ARB) and associated antibiotic resistance genes (ARG). This is because partially metabolized antibiotic remnants as well as ARB themselves are excreted from humans, collected in sewer systems and further concentrated at wastewater treatment plants (WWTP) along with all other biological and chemical UW (WW) constituents (Kreuzinger ). Resistant heterotrophic and faecal coliform bacteria are present in raw and treated WW and may reach the environment and drinking water sources through effluent discharge. If they are not eliminated during drinking water treatment, there is a risk they spread into the human microbiome through drinking water consumption (Cooke ; Vaz-Moreira et al. ).
A number of previous studies detected ARB and ARG in drinking water (Xi et al. ; Sanganyado & Gwenzi ). Removal efficiencies of conventional WWTP for antibiotic traces, ARBs and ARGs vary and are dependent on the physiology of microorganism and operating conditions of the treatment system (Michael et al. 2013). Therefore, it is necessary to address the concern of AR in the DPR context and expand the scientific knowledge on advanced water treatment technologies to provide a safeguard against AR dissemination during subsequent reuse. For this reason, one part of this study is concerned with the detection of resistant phenotypes using sulphamethoxazole (SMX) resistance as a high abundance lead parameter. This allows to follow the removal of ARB throughout the treatment train without losing the signal in an early stage treatment process. For the same reason, SMX resistance genes are used to follow the abundance of ARG through the technological steps of DPR. Resistance genes are cell-associated within the host's genome or plasmids. However, microbial death, as can occur during water treatment processes leads to the release of the genetic material. Occurrence and persistence of free DNA in water and soil environments have been reported (Nielsen et al. ), but the associated risks for the spread of AR coming from free resistance genes are still unknown. The concern is that free DNA facilitates transformation by direct uptake of extracellular DNA (Zhang et al. ). Therefore, this study includes an assessment of resistance genes on free environmental DNA within the DPR process.
Within that frame, this study investigates the efficiency of the treatment train at New Goreangab Water Reclamation Plant (NGWRP) and its treatment steps on the reduction of ARB and ARG by culture-dependent methods and quantification of ARG by TaqMan real-time PCR. It aims to address the following research questions: • Does the treatment process at NGWRP as a whole reduce the abundance of (i) resistant bacteria, (ii) intracellular resistance genes and (iii) extracellular resistance genes?
• To what extend do the individual treatment technologies at NGWRP reduce the abundance of (i) resistant bacteria, (ii) intracellular resistance genes and (iii) extracellular resistance genes? Therefore, the study focuses on life bacterial indicators resistant to SMX, the intracellular indicator resistance genes sul1 (sulphonamides), ermB (macrolides), vanA (vancomycin), npt II and npt III (kanamycinaminoglycosides) and extracellular resistance gene sul1.

Study area and site
Namibia's capital, Windhoek, has encountered reoccurring periods of limited water supply and suffered from great droughts (Lahnsteiner & Lempert ). In order to augment the available water supply, the city covers a quarter of its drinking water demand through DPR. The NGWRP produces purified drinking water from pre-treated municipal WW at a maximum capacity of 21,000 m 3 /day and includes a series of advanced biological and chemical purification steps. An overview on the treatment train applied at the NGWRP system is shown in Figure 1

Sampling
For this study, water samples at nine sampling points representing the effluents of various treatment steps at NGWRP were collected in the third week of September 2018 by the facility's routine sampling team. A reducing agent (sodium thiosulphate) that neutralizes potential free chlorine was added to the sample 9-EF final product as well as 10-POU to maintain the sample's original bacterial state.  Table 1 lists all sampling points and their abbreviations used throughout this work.

Selection of AR determinants
The following antimicrobial resistance genes have been selected for our analyses due to their clinical relevance   With the exception of nptII which is characterized by a low prevalence in natural environments and in clinical pathogens (Woegerbauer 2014(Woegerbauer , 2015, all remaining selected resistance genes are frequently present in gut bacteria and, thus, supposed to be prevalent constituents of an average WW microbiome.

Cultivation of bacterial indicators
The number of viable heterotrophs as well as coliform bacteria in the samples was estimated by the filtration-based  We herein define the term 'free DNA' solely to extracellular DNA, which is DNA of any size that is found to be freely dispersed in the water column after a filtration The same procedure was carried out for Wash Buffer 2 and 3. Upon removal of Wash Buffer 3, the tubes were set to air dry for 20 min before adding 50 μL of heated Elution Buffer.
For the elution, the microcentrifuge tubes were shaken at 250 rpm for 10 min at 40 C. Finally, the Elution Buffer containing the extracted free DNA was separated from the beads.
The same procedure was carried out with 5 mL molecular water as a no-sample extraction negative control (NCex). 50 μL of a heterogenous effluent culture grown on agar that was supplemented with 0.5 mg/L SMX was added to 5 mL of molecular grade water and served as a positive control (PCex) for extraction. Samples were extracted in duplicates and transported to Vienna by aeroplane in an insulated cooler box. They were frozen at À20 C immediately upon arrival in Vienna and stored there until further analyzed by PCR and qPCR.

DNA quantitation
Concentrations of the extracted DNA were measured by fluorescence staining of double-stranded DNA (Quant-it PicoGreen dsDNA Assay Kit, Thermo Fisher Scientific). A low-range standard curve from 2 to 1,000 ng/mL was prepared from the kit's lambda DNA standard. The LOQ of this method is 1 ng/mL.

Quantitative gene analysis
Quantitative TaqMan real-time PCR was used to estimate the copy number of intracellular bacterial 16S rRNA gene and five different ARGs inactivating four major classes of antibiotics: sul1 (sulphonamides), vanA (vancomycin), ermB (macrolides), nptII and nptIII (aminoglycosides). All genes were quantified in samples 1-12 (see Table 1

Log removal values
For the evaluation of removal efficiencies of the individual treatment units, LRV were calculated as follows: LRV ¼ Log 10 copies=mL before treatment copies=mL after treatment The number does not reflect the type of removal, which may be physical removal or destruction of the gene. Where the treatment reduced the abundance below detectable levels, LRV were calculated with the respective LOD.

Statistics
Results were analyzed with statistic tools available in Sigma-

RESULTS AND DISCUSSION
As the NGWRP applies a set of state-of-the-art technologies into one multi-barrier treatment train, it is not possible to directly compare the total LRV with single-step technologies. Hence, only the individual treatment technologies that have been assessed for their reduction efficiencies can be discussed and compared with the literature.

Bacterial counts
Addressing research question part (i) in regard to the reduction of the abundance of ARB, the following section will present CFU results and discuss the effects of the sequential treatment units of the NGWRP treatment train on the abundance of SMX-resistant heterotrophic bacteria and coliforms, respectively. Figure 2 gives a summary of the results in CFU/mL. The raw CFU data and individual graphs can be found in the Supplementary Material.

Total HPC and coliforms
The NGWRP reduced the HPC as well as coliforms from 1.6 × 10 4 and 1.9 × 10 3 CFU/mL, respectively, to below the LOD in the final product (9-EF) with corresponding total log reductions of 6.5 and 3.8 or higher. The pre-ozonation treatment only decreased the HPC slightly to 3.1 × 10 4 CFU/mL and had no significant effect on the total coliforms resulting in 1.2 × 10 3 CFU/mL at 2-Pre-O 3 . As the results after 3-DAF are likely to be influenced by permanganate dosing in close proximity of the sampling spot and therefore being not representative for the whole process, they are excluded from the evaluation. There were HPC and coliforms of 2.2 × 10 3 and 2.3 × 10 1 CFU/mL detected at 4-SF, which then decreased by 4.7 LRV to 4.1 × 10 À2 CFU/mL and 3.2 LRV to 1.4 × 10 À2 CFU/mL at 5-M- sources, POU, BH and BD, showed HPC of 1.4 × 10 À2 , 3.0 × 10 À1 and 6.5 × 10 À3 , respectively.

Resistant HPC and coliforms
The NGWRP reduced the SMX-resistant HPC as well as coliforms from 3.5 × 10 1 and 5.6 × 10 1 CFU/mL at 1-IF, respectively, to below the LOD in the final product (9-EF) with corresponding log reductions of 5.3 and 4 or higher.
Of the total heterotrophs in the influent, 0.2% showed SMX resistance. This ratio was considerably higher for the coliforms with 2.9% SMX resistance.  at 0.9% and is considerably higher. Of the ratios that could be obtained, the coliforms have a proportionally higher share and occurrence of resistance to SMX compared with the HPC.

Quantification of intracellular resistance genes and absolute copy numbers
The ARG sul1, ermB, vanA, nptII and nptIII were quanti- 4-SF is in the range of 1 LRV per treatment step, the removal of sul1 copies is steadily increasing from below 1 LRV after 2-Pre-O 3 to close to 2 after sand filtration indicating a continuous gradual decrease of the relative abundance of sul1 resistance genes.

Main ozonation
The highest removal for sul1 of at least 2.5 was achieved by the main-ozonation treatment reducing the sul1 gene abun- This is 3-7.5 times higher than in aforementioned advanced WW treatment. The reason is that the primary aim in DPR is the reliable inactivation of virus and protozoa needing higher specific ozone concentrations. During ozonation disinfection, oxygen radicals interact with the cell surface inactivating its function but rarely oxidize the cell's interior content (in contrast to UV disinfection). This may be the underlying reason for the observation that the LRV for bacterial counts is 3.2 compared with 2.6 for 16S rRNA genes and 2.5 for sul1 in the same sample.
BAC and GAC. The absolute abundance of the sul1 gene (2.49 × 10 2 copies/mL) and the 16S rRNA genes (1.88 × 10 5 copies/mL) increased significantly after the biological activated carbon step (6-BAC LRV 16S À4.1, LRV sul1 À2.8) whereas the relative abundance stayed in the same order of magnitude as before the main-ozonation treatment (see Table 14 in the Supplementary Material). This is also in accordance with the CFU results that show resistant HPC and coliforms under the LOD after the BAC even though the total bacterial counts increased (LRV HPC À5.3, The opposite was reported by Xu et al. () in their study of a DWTP that employs similar technologies as the ones at NGWRP. They suggest that their observed increase in relative abundance may have been due to selection pressure that was acting on the bacteria growing in the activated carbon from adsorbed antibiotic micropollutants. In conclusion, the BAC does not remove ARG but instead gene abundance returned to the value upstream of the main ozonation. This is likely due to the bacterial community proliferating on the carbon particles. Subsequently, the 7-GAC results in a log reduction of 0.8 to 3.56 × 10 1 copies/mL. Experiments on ozonation and ultrafiltration reduction efficiencies for ARB and ARG with real WW by Hembach et al.
() showed that the ultrafiltration was able to reduce sul1 abundance to below the LOD (LRV 6-7). This was achieved by a membrane cut-off (20 nm) half of that at NGWRP. The reduction efficiency of this technology strongly depends on the cut-off/nominal pore size of the applied membrane (Hiller et al. ). A considerable disadvantage of this technology is its non-destructive nature. Hembach et al. () found that the retentate water contained 2 log units higher concentrations of ARB and ARG than the influent. Considering ARB and ARB removal in DWT facilities, treating the backwash water separately could be an option to increase the removal efficiency, yet the potential of this needs to be further investigated.
The effects of the final treatment stepschlorination and pH-stabilizationon the genes cannot be deduced since even for sul1 results were below LOD after ultrafiltration. Stabilization by chlorination is an important measure to prevent the regrowth of pathogens (including ARB) in storage tanks and the drinking water distribution system.
Generally, chlorination has been suggested to have a relatively consistent reduction efficiency on sul1 resistance genes of below 1.5 LRV (Hiller et al. ).
All reductions were statistically significant except for the pre-ozonation treatment at 2-Pre-O 3 . There were no outliers identified for all gene quantification results.
Relative abundance of the sul1 resistance gene   show that the percentage of particle-associated sul1 resistance genes in the secondary effluent was roughly 90% of the total sul1 abundance. According to this, the separation between the two fractions would not have made essential differences for this study.

CONCLUSION
In order to gain an understanding of the impact of combined water treatment technologies on the removal of ARB and ARG, this study quantified viable bacteria possessing resistance to SMX and ARG sul1, ermB, vanA, nptII and nptIII after each treatment step at NGWRP and of three additional drinking water sources in Windhoek, Namibia. The resulting observations aim to provide a suggestion of possible combinatory treatment technologies that can be employed to reduce the discharge of ARB and ARG into potable reuse systems or the environment.
Regarding the research questions, the main findings of this study can be summarized as follows: • The NGWRP statistically significantly reduced the abundance of total and SMX-resistant heterotrophic bacteria as well as total and SMX-resistant coliforms down to undetectable levels below LOD in the final effluent. The main ozonation and the ultrafiltration had the highest reduction efficiencies.
There were no SMX-resistant heterotrophs or coliforms detected downstream the main-ozonation treatment nor at POU.
• The NGWRP reduced the abundance of intracellular sul1 resistance genes to undetectable levels below LOD.
• All treatment technologies decreased the abundance of intracellular sul1 resistance genes except for the pre-ozonation and the biological activated carbon.
The main ozonation and the ultrafiltration had the highest reduction efficiencies.
• Despite removal below LOQ was observed for extracellular sul1, methodological limitations regarding efficiency and yield of extracellular DNA recovery from samples did not allow to follow the removal of ARGs located on free extracellular DNA throughout the whole treatment train.
The results show that the advanced multi-barrier system for potable reclamation at NGWRP reduced SMX-resistant bacteria and the investigated resistance genes to undetectable levels below LOD (LODs for individual sampling sites, see Supplementary Material). This shows that an advanced technical treatment train is able to remove ARB as well as ARG to levels below 1 copy per 100 mL and with this it may be possible to reach recommended prospective removal targets in the future.

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