Understanding human health risks caused by antibiotic resistant bacteria (ARB) and antibiotic resistance genes (ARG) in water environments: Current knowledge and questions to be answered

Abstract

Aquatic environments are identified as an ideal setting for acquisition and dissemination of antibiotic resistance, and human exposure to antibiotic resistant bacteria (ARB) and antibiotic resistance genes (ARGs) in aquatic environments may pose an additional health risk. Quantitative microbial risk assessment (QMRA) has been suggested as a suitable method to evaluate and quantify this health risk. However, information about the exposure to ARB and ARGs in aquatic environments is lacking for many scenarios and dose-response models regarding the ARB infections are not developed yet. This review summarizes the current knowledge regarding the ARB and ARGs in aquatic environments and highlights the challenging questions remaining to be answered to better forecast the health risks caused by ARB and ARGs in water environments. The questions include what are the missing information needed to quantify the human health risks caused by exposing to ARB and ARGs in aquatic environments? what are the suitable markers to evaluate the ARB/ARGs contamination in aquatic environments? how frequently do the ARG selection and propagation occur in aquatic environments? and are there any unknown hot spots? Studies on the above topics will contribute to better management of antibiotic resistance dissemination in water environments and its risks on human health.

Abbreviations
3GC	=	3rd generation cephalosporins
ARB	=	Antibiotic resistant bacteria
ARG	=	Antibiotic resistance gene
CFU	=	Colony forming unit
DBP	=	Disinfection by-products
eDNA	=	Extracellular DNA
EPS	=	Extracellular polymeric substances
HGT	=	Horizontal gene transfer
ISCR	=	Insertion sequence common region
MAR	=	Multiple antibiotic resistant
MIC	=	Minimum inhibitory concentration
MGE	=	Mobile genetic elements
MSW	=	Municipal solid waste
QMRA	=	Quantitative microbial risk assessment
VBNC	=	Viable but non-culturable
WWTP	=	Wastewater treatment plant

Graphical Abstract

Keywords:

• Antibiotic resistant bacteria
• antibiotic resistance genes
• quantitative microbial risk assessment

1. Introduction

Antibiotic resistance is the ability of bacteria to resist the effects of an antibiotic which they were previously sensitive. Bacteria become antibiotic resistant by either genetic mutations or by acquiring antibiotic resistance genes (ARGs). Antibiotic resistant bacterial (ARB) infections have higher mortality and morbidity rates and result in longer hospital stays (Cosgrove, 2006). ARB infections are estimated to cause 10 million deaths worldwide by year 2050 (O'Neill, 2016). Antimicrobial resistance poses a challenge in achieving the universal health coverage and hinder the achievement of sustainable development goals related to health, food security, clean water and sanitation (Interagency Coordination Group on Antimicrobial Resistance, 2019). In view of these facts, World Health Organization (WHO) published a list of antibiotic resistant priority pathogens. The list consists of 12 bacteria families which are considered to pose greatest threat to the human health (World Health Organization, 2017b). Since the drivers of antimicrobial resistance lie in humans, animals, plants, food and the environment, importance of One Health approach is emphasized in overcoming this threat (Interagency Coordination Group on Antimicrobial Resistance, 2019).

Aquatic environments are recognized as one of the reservoirs and transmission routes for the dissemination of antibiotic resistance (Karkman, Do, Walsh, & Virta, 2018; Suzuki, Pruden, Virta, & Zhang, 2017; Wellington et al., 2013). Drinking water and wastewater treatment processes are shown to be incapable of completely removing antibiotic resistance genes (ARGs) (Li, Cheng, Xu, Strong, & Chen, 2015; Pruden, Pei, Storteboom, & Carlson, 2006; Rodriguez-Mozaz et al., 2015). Wastewater treatment plant (WWTP) effluents containing ARB and ARGs can end up in aquatic environments like rivers and lakes. Usage of reclaimed wastewater in beneficial purposes like agricultural irrigation and recreational activities can introduce ARB and ARGs to the particular environment (Ben et al., 2017; Pruden et al., 2006; Rodriguez-Mozaz et al., 2015). Therefore, in addition to drinking, humans can be exposed to ARB and ARGs through activities like bathing, aquatic sports, occupational exposure during agricultural irrigation, and consumption of food produce irrigated with reclaimed water (Leonard, Yin, Zhang, Hui, & Gaze, 2018; Leonard, Zhang, Balfour, Garside, & Gaze, 2015; Leonard, Zhang, et al., 2018; O’Flaherty, Borrego, Balcázar, & Cummins, 2018; Søraas, Sundsfjord, Sandven, Brunborg, & Jenum, 2013).

However, potential human health risks caused by exposing to ARB and ARGs present in aquatic environments have not been completely evaluated yet, because specific information like the dose-response curves and exposure assessment data related to ARB and ARGs in different water usage scenarios needed to conduct a quantitative microbial risk assessment (QMRA) are scarce (Ashbolt et al., 2013; Pepper, Brooks, & Gerba, 2018). This review summarizes the information reported in recent studies on the human health risk assessment of ARB and ARGs, and highlight the required additional information to perform a refined QMRA for different scenarios. We further suggest important research questions related to the occurrence, proliferation and dissemination of ARGs and ARB in aquatic environments which needs to be answered for a better evaluation of human health risks caused by ARB and ARGs in aquatic environments.

2. Sources of antibiotics, ARB and ARG in water environments

2.1. Freshwater and wastewater related environments

Human and animal microbiota is a reservoir for ARGs (Salyers, Gupta, & Wang, 2004). Bacteria passing through the intestinal tract can acquire antibiotic resistance via conjugation and end up in human and animal feces (Anderson, Whitlock, & Harwood, 2006; Salyers et al., 2004). Consumption of antibiotics can provide a selective pressure (a biotic or abiotic factor which alters the behavior and fitness of an organism in a given environment) inside the gut, leading to the development of antibiotic resistance in enteric bacteria. Penicillin resistance was developed in dairy calves fed with milk containing penicillin G (0–50 µl/kg) in a dose dependent manner and persisted over the course of the experiment (Langford, Weary, & Fisher, 2003). Chicken manure compost collected from farms which use amoxicillin, kanamycin, gentamicin and cephalexin contained ARB between 21% and 66% of the total bacterial isolates. Among the ARB isolates, 8–20% were resistant to multiple antibiotics (Q. Yang et al., 2014).

Wastewater or other biological substances generated in hospitals, farms, homes and other facilities contain antibiotics, ARBs and ARGs. According to a recent study, tylosin administered sheep excreted the antibiotic in feces and urine up to four days. On average 11% of the administered tylosin was discharged in feces and urine (Ishikawa et al., 2018). A recent metagenomics analysis performed in 79 wastewater treatment plants (WWTPs) located in 60 countries reported that abundance and diversity of ARGs have differences between Asia/Africa/South America and North America/Europe and Oceania. North America/Europe and Oceania cluster contained limited number of ARGs encoding macrolide resistance in high abundance while Asia/Africa/South America cluster contained ARGs related to sulfonamides and phenicols. The study also reported Vietnam, India and Brazil having the most divergent ARG distribution and suggested them as possible hotspots for the emergence of new antibiotic resistance mechanisms (Hendriksen et al., 2019). Table 1 provides a representative summary of ARGs and mobile genetic elements (MGEs) detected in different water environments.

Table1. Examples on various ARGs, MGEs and antibiotics found in different water environments.

Environment	ARGs*	MGEs*	Antibiotics*	Remarks	Reference
South China Sea	macB, acrB			Pristine environment with minimal human activity	(Chen et al., 2013)
Polar marine sediments	sul1, tet(D), qnrB	intI1		sul gene abundance in polar sediments were 2-5 orders lower than the human impacted Haihe river, China.	(Tan et al., 2018)
Marine environment coastal areas of Bohai bay, China	blaTEM, sul1, sul2, tet(B),	intI1			(Zhang, Niu, Zhang, & Zhang, 2018)
Sediments from 18 estuaries along over 4,000 km of coastline in China	aacC, aadA5, mphA, oprD, oprJ, qacEΔ1	intI1	Macrolides, TC		(Zhu et al., 2017)
Manilla Bay water	sul1, sul2, sul3			Bay water contained sul genes with high copy numbers, even though the inlet river water didn’t, suggesting marine non-culturable bacteria are the reservoir of the three sul genes	(Suzuki et al., 2013)
Maricultural site sediments	bacA, mexF, mexB, sul1, tet(C)		CIP, ENR, OTC, SDZ, SMX, TC	Samples were taken from sites to represent the complete Chinese coastline	(Gao et al., 2018)
Maricultural site water and sediments	sul2>sul1, tet(W), tet(Q)>tet(O)		SDZ (water) OTC (sediments)		(Chen, Zheng, Zhou, & Zhao, 2017)
Estuarine lake water	sul1		NOR, OTC, SDZ, SMX, TC		(Yuan, Ni, Liu, Zheng, & Gu, 2019)
Estuarine lake sediments	tet(C)
Urban sewage pipes	FCA and MLSB resistance genes	tnpA (IS6 group)	SMX		(Huang et al., 2019)
Activated sludge reactors	aadA, bacA, chloramphenicol exporter, sul1, tet(C)		ERY, KAN, LIN, SMT, TC, TMP	Minimum antibiotic concentration of 20 mg /L was used in this study	(Zhao et al., 2019)
Full-scale municipal WWTP influent	blaTEM, ermf, qnrS	intI1		Several treatment process chains were used including activated sludge process, oxidation ditch and A2O treatment	(Tong et al., 2019)
WWTP (activated sludge) effluent	sul1, tet(O), tet(W)		AMX, TC		(Proia et al., 2018)
WWTP (activated sludge + nutrient removal) effluent	sul1, sul2, qnrS, tet(O), tet(W)		AMX, TC
activated sludge TP influent	aac(6')-1b, blaCTX-M, ermB, sul1, sul2, tet(M)	intI1	AMX, CTC, CIP, OTC, TC, VAN		(Le, Ng, Ngoc, Chen, & Gin, 2018)
Membrane bioreactor influent	aac(6')-1b, blaCTX-M, dfrA, ermB, sul1, sul2, tet(M)
Hospital WWTP effluent	aacC2, blaSHV, floR, mefA, sul1, sul2, tet(A), tet(B), tet(C), tet(W)	tnpA	AMP, FEP, CAZ, TMP, VAN		(Szekeres et al., 2017)
WWTP effluent	sul1, sul2, tet(A), tet(C), tet(G), tet(M), tet(O), tet(W), tet(X)	intI1, intI2			(Ben et al., 2017)
Urban WWTP effluent	FCA and Aminoglycoside resistance genes	tnpA (IS6 group), IS613, Tp614, tnpA (IS4 group)	SMX		(Huang et al., 2019)
Waste stabilization pond effluent			CEF, CIP, ERY, SMX		(Azanu, Styrishave, Darko, Weisser, & Abaidoo, 2018)
Recycled water	vanA, ermF, sul2			A broader range of ARGs were detected after the reclaimed water passed through the distribution systems	(Fahrenfeld, Ma, O’Brien, & Pruden, 2013)
Full-scale municipal WWTP sludge	blaTEM, ermF	intI1		Several treatment process chains were used including activated sludge process, oxidation ditch and A2O treatment	(Tong et al., 2019)
Dewatered sludge	ermC, qnrB, sul1, sul2, tet(A), tet(B), tet(E), tet(G), tet(H), tet(S), tet(T), tet(X)		Sulfonamides, TC	Concentrations are higher than the influent values	(Mao et al., 2015)
Urban river discharge	FCA and MLSB resistance genes	tnpA (IS6 group), IS613, Tp614, tnpA (IS4 group)	SMX		(Huang et al., 2019)
Freshwater biofilms	ermB			After notable runoff event	(Winkworth, 2013)
	vanB, tet(B)			Swimming spot
River water from upstream until the estuary	sul1, sul2, tet(W)				(Jiang, Zhou, Yang, et al., 2018)
River surface water	blaOXA-21, emrA, ereA, mdtK, tet(G)	52 integrons and 39 plasmids		Unique ARGs to different water systems were marked	(Jiang, Zhou, Zhang, et al., 2018)
River water			CEF, CIP, ERY, SMX		(Azanu et al., 2018)
River sediments	lsa, vanM, vanS, vanY, vanZ	18 integrons and 105 plasmids			(Jiang, Zhou, Zhang, et al., 2018)
River water and sediments	sul1, sul2			Sediments contained 120 to 2000-fold higher concentrations of sul1 and sul2 genes compared to river water	(Luo et al., 2010)
Recreational lake water	blaSHV, blaTEM, ampC, ermA, ermB, cat, cmr				(Fang, Wang, Cui, Rogers, & Dong, 2018)
Lake surface sediments	blaOXY, blaSHV, blaTEM, sul1, sul3, tet(O)	intI1	NOR, TC		(Ohore, Addo, Zhang, Han, & Anim-Larbi, 2019)
Recreational water and sediments	blaTEM, sul1, sul2, tet(X), tet(W)	intI1			(Dong, Cui, et al., 2019)
Irrigation water			CIP, ERY		(Azanu et al., 2018)
Fish intestine, Finland	aadA1, aadA2, dfrA1, sul1, tet32, tet(M), tet(O) tet(W),	intI1, tnpA01, tnpA04, tnpA06, tnpA07		Antibiotics are not used in the fish farm	(Muziasari et al., 2017)
Fish gills and intestinal contents, Concepcion Bay, Chile			AMP, CHL, STR, TET	ARB mainly belonged to Vibrionaceae and Enterobacteriaceae	(Miranda & Zemelman, 2001)
Spleen tissue of Carp, Czech Republic	tet(A), tet(E), qnrS1, qnrS2, sul2,	intI1	OTC	OTC was used to treat carp erythrodermatitis	(Syrova et al., 2018)
Intestinal mucus of wild freshwater fish species, Spain	blaTEM			Brown trout and Ebro barbel from La Llosa del Cavall reservoir	(Marti et al., 2018)
Intestinal mucus of wild freshwater fish species, Spain	qnrS, sul1			Common carp from Foix reservoir
Seafood products, Japan	blaTEM-1, blaCMY-2, blaCMY-13, blaCMY-39, qnrB2, qnrB6, qnrS1	intI1		Products were made in Japan or imported from Chile, China, India, Indonesia or North Pacific	(Ahmed, Maruyama, Khalifa, Shimamoto, & Shimamoto, 2015)
Active and closed down landfills			CEP, CIP, ERY, SMX, TMP	Antibiotics in leachate from active landfills were significantly greater than in closed landfills in the wet season	(Chung, Zheng, Burket, & Brooks, 2018)
Landfill leachate	ampC, blaCTX-M, qnrA, qnrB, qnrS	intI1, intI2, tnpA01, tnpA03	AMX, INN-d4, CEP, ENR, NOR, NOR-d5 OFX, PEF		(You, Wu, Wei, Xie, & Lu, 2018)
Neglected MSW landfill leachate	sul1, sul2	intI1, tnpA, traA, IS-CR3, IS-26			(Yu et al., 2016)
MSW landfill groundwater	171 ARGs	tnpA01, intI1			(Chen et al., 2017)
Urban residential land discharge	FCA and beta-lactamase resistance genes	tnpA (IS6 group), Tp614, tnpA (IS4 group)	OTC		(Huang et al., 2019)
Urban public land discharge	FCA and beta-lactamase resistance genes	tnpA (IS6 group), Tp614, tnpA (IS4 group)	SMX		(Huang et al., 2019)
Urban tap water	Aminoglycoside and MLSB resistance genes	IS613, tnpA (IS4 group)	SMX		(Huang et al., 2019)
Tap water (Extracellular ARG)	blaTEM, sul1, sul2, tet(C)			Monthly variations of ARGs has been performed	(Hao et al., 2019)
Tap water (Intracellular ARG)	sul1, sul2
Groundwater (Closer to urban areas)	sul1, tet(C), tet(O), tet(W)	intI1	ERY, TMP		(Szekeres et al., 2018)
Piggery WWTP effluent	ampC, blaOXA-1, blaTEM-1			Higher abundance of ARGs in winter	(Yang et al., 2019)
Aquaculture/ livestock wastewater			SMT, OTC, LIN, SMX	Piggery, shrimp culture and other livestock wastewater from Thailand and Vietnam	(Shimizu et al., 2013)
Hospital sewer outlet	blaTEM, sul1, tet(W)		SMX, TC		(Proia et al., 2018)
Hospital wastewater	sul1, tet(O)	intI1			(Qiang Wang et al., 2018)
Hospital wastewater			CIP, ERY, CEF, SMX		(Azanu et al., 2018)
Hospital effluents	blaSHV > blaNDM-1> blaCTX> blaKPC			Predominant ARB taxa were Pseudomonas spp., Klebsiella spp., Enterobacter spp. and Citrobacter spp.	(Haller et al., 2018)
Pharmaceutical manufacturing plant A WWTP effluent	lmrA		CLI, LIN		(Guo et al., 2018)
Pharmaceutical manufacturing plant B WWTP effluent	aac, aadD		PM, RSM
Pharmaceutical manufacturing plant CWWTP effluent	qnrD		LVX
Pharmaceutical manufacturing plant D WWTP effluent	ermB		AZM, CIP, CLR, ENR, ROX
Pharmaceutical manufacturing plant E WWTP effluent	blaOXA		CEP, INN

* Most abundant ARGs, MGEs and antibiotics were reported when data are available. AMX: Amoxicillin, AMP: Ampicillin, AZM: Azithromycin, CAZ: Ceftazidime, CEF: Cefuroxime, CEP: Cephalexin, CHL: Chloramphenicol, CIP: Ciprofloxacin, CLR: Clarithromycin, CLI: Clindamycin, CTC: Chlortetracycline, ENR: Enrofloxacin, ERY: Erythromycin, FEP: Cefepime, INN: Cefradine, KAN: Kanamycin, LIN: Lincomycin, LVX: Levofloxacin, NOR: Norfloxacin, OFX: Ofloxacin, OTC: Oxytetracycline, PEF: Pefloxacin, PM: Paromomycin, ROX: Roxithromycin, RSM: Ribostamycin, SDZ: Sulfadiazine, SMT: Sulfamethazine, SMX: Sulfamethoxazole, STR: Streptomycin, TET: Tetracycline, TMP: Trimethoprim, VAN: Vancomycin. A2O: Anaerobic-Anoxic-Oxic, FCA: fluoroquinolone, quinolone, florfenicol, chloramphenicol, and amphenicol, MLSB: Macrolide-Lincosamide-Streptogramin B resistance genes, MSW: Municipal solid waste, WWTP: Wastewater treatment plant.

Wastewater treatment plants (WWTPs) receive microbial and chemical loads from different sources. Inside WWTPs, under the selection pressure of antibiotics, metals, and other chemical substances, horizontal gene transfer (HGT) can occur (Karkman et al., 2018). Many studies have confirmed that WWTPs are not capable of completely removing antibiotics, ARB and ARGs (Ben et al., 2017). Therefore, WWTP effluent discharge to environments including surface water, groundwater, marine environments and soil where environmental microorganisms can interact with ARB and ARGs can lead to development of antibiotic resistance in the microorganisms from those environments. Detection of antibiotics, ARGs, and ARB in different aquatic environments are summarized in Table 1.

2.2. Marine environments

Occurrence of antibiotics, ARB and ARGs in marine environments can have different mechanisms than the fresh water and wastewater. According to a recent study, only 28% of the ARGs identified in marine environments were previously known (Hatosy & Martiny, 2015). Coastal runoff of ARB from terrestrial sources has been identified as one mechanism for antibiotic resistance occurrence in marine environments (Hatosy & Martiny, 2015). Secondly, anthropogenic antibiotic runoff related selection can occur in marine aquaculture related activities. For example, Chilean marine salmonid farms have used 363.4 tons of antibiotics in 2016 alone which can act as a selective pressure for antibiotics resistance development in marine environments (Miranda, Godoy, & Lee, 2018). According to a metagenomic study on the intestinal contents of fish from the farms located in Baltic Sea, ARGs found in fish intestines were same as the ARG enriched in farm sediments, suggesting fish feces contributes to the ARG enrichment even when there are no concurrent antibiotic treatments. Usage of antibiotics during hatching and rearing of juvenile fish before entering the sea farms is suggested as a possible reason for the presence of ARGs in fish intestines. However, it was also suggested the possibility of transferring ARG derived in water environment to the fish intestine (Muziasari et al., 2017). Selection for the antibiotic resistance as a result of the generation of antibiotics in marine environments by marine microorganisms is the third mechanism. (Baam, Gandhi, & Freitas, 1966; Rosenfeld & Zobell, 1947; Tortorella et al., 2018). Among 58 marine microorganisms tested by Rosenfeld & Zobell (1947), 4 Bacillus species, 3 Micrococcus species, one Actinomyces and Serratia species demonstrated the antibiotic activity against non-marine forms. Among them, Bacillus and Micrococcus were the most active antagonistic marine genera (Rosenfeld & Zobell, 1947). Fluoroquinolone resistance (qnr) genes, qnrA, qnrB, qnrS, are thought to have originated from water-dwelling bacterial chromosome (Takasu, Suzuki, Reungsang, & Viet, 2011), which are transferred to other species. Recently discovered sulfonamide resistance gene (sul4) from Indian river bacteria is accompanied by class 1 integrons (Razavi et al., 2017). Thus, some ARGs are originated in aquatic environment and might have invaded to human environment.

In conclusion, major sources of ARB and ARGs should be human and veterinary clinical settings where intestinal bacteria are exposed to high concentrations of antibiotics, WWTPs, and related land wastes. However, recent reports presenting global antibiotic contamination in water environments (Fekadu, Alemayehu, Dewil, & Van der Bruggen, 2019; Segura et al., 2015; Shimizu et al., 2013) suggest development of ARB in water environments including open oceans (Hatosy & Martiny, 2015). Bacteria can evolve to be antibiotic resistant via mutations at very low antibiotics concentrations in natural aquatic environments (100-fold below minimum inhibitory concentration (MIC)) (Kohanski, DePristo, & Collins, 2010). Even though the fraction of resistant mutants is very low (10⁻⁴), selection of ARB can occur over generations. Sub-MIC selection contributes to prolonged persistence of ARB since low concentrations of antibiotics are enough to maintain the existing resistant bacteria (Gullberg et al., 2011) and attention should be paid on water environments as an ARB and ARGs risk origin.

3. Acquisition of antibiotic resistance by horizontal gene transfer and mutation

3.1. Antibiotic resistance acquisition via horizontal gene transfer (HGT)

Antibiotic resistance of environmental bacteria can be intrinsic, acquired via spontaneous mutations (de novo) or horizontal gene transfer (HGT) (J. Davies & Davies, 2010; Sharma, Johnson, Cizmas, McDonald, & Kim, 2016). Antibiotic resistance hotspots are the locations where susceptible bacteria can develop antibiotic resistance and disseminate (Marti, Variatza, & Balcazar, 2014). Horizontal gene transfer can occur through transformation, transduction and conjugation (Aminov, 2011).

It is shown that as high as 3% of the Pseudomonas aeruginosa genome may contribute to intrinsic resistome (Fajardo et al., 2008; Martinez, 2014). Spontaneous mutation rate of ciprofloxacin and rifampin resistance in Helicobacter pylori was 1 × 10⁻⁸ to 2 × 10⁻⁸ per cell per division (G. Wang, Wilson, Jiang, & Taylor, 2001).

In the transformation process, bacteria will uptake and integrate the extracellular DNA present in the environment (Thomas & Nielsen, 2005). Environmental ARG pool is formed under grazing pressure by protists, which is found to be constant (Bien, Thao, Kitamura, Obayashi, & Suzuki, 2017). A large number of naturally transformable prokaryotes in the environment have been reported (de Vries & Wackernagel, 2005; Johnsborg, Eldholm, & Håvarstein, 2007; Lorenz & Wackernagel, 1994). Natural transformation of Pseudomonas stutzeri strains isolated from soil and marine sediments to rifampicin resistance by chromosomal DNA have been observed (Stewart & Sinigalliano, 1991). Transformation of naturally competent Acinetobacter calcoaceticus by plasmid and chromosomal DNA adsorbed on sand and ground water/ground water aquifer microcosms was observed; chromosomal DNA transformation frequency was 10³-10⁴ times higher than plasmid DNA transformation frequency (Chamier, Lorenz, & Wackernagel, 1993). Transformation of environmental bacteria found in other aquatic environments like river water, spring water and groundwater have also been reported (Davison, 1999).

In the transduction process, cellular DNA packaged to the bacteriophages during their replication will be incorporated to the host bacterial genome after the infection by a transduced bacteriophage (Calero-Cáceres, Ye, & Balcázar, 2019). Extended spectrum beta-lactamase (ESBL) genes bla_TEM, bla_CTX-M, bla_CMY-2, bla_KPC, bla_OXA-48, bla_PSE, conferring resistance to β-lactam antibiotics were detected in DNA of bacteriophages isolated from canal water, urban sewage and river water samples collected from Washington State whereas bla_NDM-1 gene was detected from bacteriophage DNA fraction isolated from river and wastewater samples (A. Zhang et al., 2019). Analysis of viromes from seven marine habitats including Arctic Ocean, Indian Ocean and Gulf of Mexico confirmed the presence of phage associated ARGs (Calero-Cáceres & Balcázar, 2019). Muniesa et al. (2013) have provided a detailed review on the contribution of environmental bacteriophages on spreading ARGs (Muniesa, Colomer-Lluch, & Jofre, 2013b).

During the conjugation process, MGEs like plasmids, transposons, integrons, insertion sequences, insertion sequence common region (ISCR) elements, genomic islands and integrating conjugative elements (ICE) are moved from the donor cell to the recipient cell (Marti et al., 2014).

3.2. Antibiotic resistance development by mutagenesis

Mutagenesis can lead to development of antibiotic resistance in environmental bacteria via antibiotic target alteration, increased drug efflux, gene amplification, reduced target expression and drug modification enzyme alteration (Andersson & Hughes, 2010). Mutagenic disinfectant by-products (DBPs), antibiotics, non-antibiotic microbials, antidepressants and biofilms have shown to play an important role in this regard (Ford, 2006; Jin et al., 2018; Lu, Jin, et al., 2018; Lv, Yu, Xu, & Ye, 2015; Savage, Chopra, & O'Neill, 2013). Sub-inhibitory concentrations of antibiotics can produce reactive oxygen species (ROS) which consequently induce mutagenesis. As a result, some mutant strains which are sensitive to the applied antibiotic may be resistant to other antibiotics (Kohanski et al., 2010). In the presence of DBPs at sub-inhibitory concentrations and near-inhibitory concentrations, resistance mechanism selections were different. Under high DBP concentrations, overexpressed efflux pump systems, impermeability of the membrane and depletion of replication error repair protein leading to increased mutation rate were the main mechanisms. Under low NaClO₂ and IAA concentrations, substitution mutations in gyrA and deletion mutation in proS likely to contribute to ciprofloxacin resistance development (Li, Zeng, He, & Gu, 2016). After exposing to mutagenic DBPs, P. aeruginosa PAO1 and E. coli K12 displayed increased resistances to individual and multiple antibiotics (Lv, Jiang, Zhang, & Yu, 2014; Lv et al., 2015). E. coli K12 exposed to triclosan (0.2 mg/l) for 30 culture cycles induced oxidative stress leading to the development of mutations in major genes like frdD, marR, acrR and soxR. These mutations cause overexpression of adjoining genes or enhanced the expression of repressed genes like ampC, acrAB, soxS and marAB. This may have initiated the translation of AmpC and multidrug efflux pump AcrAB inducing multiple antibiotic resistance (MAR) (Lu, Jin, et al., 2018; Lu, Wang, et al., 2018). Antidepressant fluoxetine overproduced ROS which lead to transcriptional regulator gene mutation. These mutations increased the expression of efflux pump system and decreased abundance of outer membrane protein OmpF leading to resistance development (Jin et al., 2018). Antibiotic resistance development by mutation at sub-inhibitory levels of different selection pressures are summarized in Table 2. Comprehensive reviews on the enhancement of antibiotic resistance of Gram-negative and Gram-positive bacteria caused by biocidal agents are available elsewhere (Kampf, 2018, 2019).

Table 2. Sub-inhibitory concentrations of selection pressures induce antibiotic resistance.

Bacterial strain	Selection pressure	Concentration	MIC	Remarks	Reference
(a) Mutation
E. coli MG1655	Ampicillin	1 µg/ml		Significant increases in the mutation rate relative to an untreated control associated with increased production of reactive oxygen species. Multidug cross-resistance was observed	(Kohanski et al., 2010)
	Kanamycin	3 µg/ml
	Norfloxacin	15 ng/ml
	Norfloxacin	50 ng/ml
Staphylococcus aureus	Ampicillin	35 ng/ml
E. coli J93 and derived strains	Ciprofloxacin	1/5 of MIC		Some derived strains, result in a loss of fitness relative to that of the wild type in the absence of antibiotic	(Liu et al., 2011)
	Tetracycline	1/20 of MIC
Pseudomonas aeruginosa, Pseudomonas protegens	Ciprofloxacin	1/10 of MIC		Genome architucture chnages were observed in as low as 5 serial passages in the presence of sub-MIC antibiotics. After 4 passages, MIC of aome P. protegens strains was increased	(Chow, Waldron, & Gillings, 2015)
	Kanamycin	1/10 of MIC
	Tetracycline	1/10 of MIC
E. coli MG1655	Ag(I)	1.79 mg/l	8.95 mg/l	Mutation rate of E. coli resistant to ciprofloxacin increased by 14.6 1538.2 folds for Ag(I), 4 -27 fold for Zn (II) and 7 -20 fold for Cu(II). Resistance to ciprofloxacin occurred via adaptive mutations	(Li et al., 2019)
	Zn (II)	26.81 mg/l	134 mg/l
	Cu (II)	11.5 mg/l	575 mg/l
E. coli MG1655	Sodium Chlorite (NaClO2)	10 mg/l		MIC to ciprofloxacin was increased by 4 – 8-fold	(D. Li et al., 2016)
	Iodoacetic acid	10 mg/l
E. coli MG1655 wild type	Ciprofloxacin		0.023 µg/ml		(Gullberg et al., 2011)
E. coli MG1655 mutants			0.047–0.38 µg/ml	Ratio of resistant: susceptible strains changed as a function of number of generations and the antibiotic concentration. Minimum selective concentration (fitness cost of the resistance and antibiotic-conferred selection for the resistant mutant are in balance), varied between 1/4 (Streptomycin) and 1/230 (Ciprofloxacin)
Salmonella enterica serovar Typhimurium LT2	Tetracycline		1.5 µg/ml
	Streptomycin		4 µg/ml
S. enterica serovar Typhimurium LT2 wild type	Streptomycin	1 mg/l	4 mg/l	After 900 generations, high-level resistance up to >1024 mg/l was observed. Several small-effect resistance mutations combined and conferred high-level resistance via alteration of ribosomal RNA target, reduction in aminoglycoside uptake and induction of an aminoglycoside modifying enzyme.	(Wistrand-Yuen et al., 2018)
(b) Horizontal gene transfer
From bacterial community to E. coli CV 601	Gentamicin	0.1 mg/l	0.125 mg/l		(Jutkina et al., 2018)
	Sulfamethoxazole	1mg/l	16 mg/l
	Chlorhexidine digluconate	24.4 µg/l	4.8 mg/l
	Triclosan	0.1 mg/l	2 mg/l	Two-fold increase in relative transfer frequency
From E. coli K12 LE392 to E. coli K12 MG1655	Triclosan	0.02 µg/l–2 mg/l		2 µg/l Triclosan increased the intra-genera conjugation frequency by 6.2-fold	(Lu, Wang, et al., 2018)
From E. coli K12 LE392 to Pseudomonas putida KT 2440				3 µg/l Triclosan increased the inter-genera conjugation frequency by 2.3 - fold
From E. coli S17-1 to E. coli K12 MG1655	Free chlorine	0.1–1 mg/l		Conjugative transfer efficiency increased by 3.4–6.9-fold	(Zhang, Gu, He, Li, & Chen, 2017)
	Chloramine	0.1–1 mg/l		Conjugative transfer efficiency increased by 1.9–7.5-fold
	Hydrogen peroxide	0.24–3 mg/l		Conjugative transfer efficiency increased by 1.4–5.4-fold
From E. coli S17-1 to Salmonella typhimurium TA1535	Free chlorine	0.1–1 mg/l		Conjugative transfer efficiency increased by 1.4–2.3-fold
	Chloramine	0.1–1 mg/l
	Hydrogen peroxide	0.24–3 mg/l
From E. coli S17-1 to E. coli K12 MG1655	Ag (I)	0–0.1 mg/l	0.39 ± 3.54 mg/l	All ions promoted the conjugation frequency. The ability of heavy metal ions to facilitate conjugative transfer varied in Cu (II) > Ag (I) > Cr (VI) > Zn (II) order. Reactive oxygen species formation, SOS response, increased cell membrane permeability and altered expression of conjugation-relevant genes contributed to the increased conjugation efficiency.	(Ye Zhang et al., 2018)
	Zn (II)	0–100 mg/l	135 ± 4.50 mg/l
	Cu (II)	0–0.5 mg/l	410 ± 6.65 mg/l
	Cr (VI)	0–20 mg/l	77 ± 5.93 mg/l
Acinetobacter baylyi ADP1	Bromoacetic acid (BAA)	600 µM		Natural transformation rates were increased by 2 -fold as a result of BAA induced DNA damage.	(Mantilla-Calderon et al., 2019)

3.3. Selective pressures for the antibiotic resistance acquisition by HGT

Disinfectants, antibiotics, metals and DBPs in water and wastewater, even at low concentrations, can affect the HGT efficiency (Karkman et al., 2018) (Table 2). Rusin and Gerba reviewed the earlier studies on the development of antibiotic resistance by chlorination of water and wastewater (Rusin & Gerba, 2001). Disinfection with free chlorine at low doses (0.05–0.2 mg/l) didn’t affect the transfer efficiency of RP4 plasmid while high doses (0.3–0.5 mg/l) caused a decrease in conjugative transfer efficiency. Reduced cell viability due to cell damage was the major contribution factor (Lin, Li, Zhang, & Yu, 2016). Chlorination enriched six out of 125 ARGs and one out of 13 MGEs present in a municipal WWTP secondary effluent (Lin, Zhang, Zhang, & Yu, 2016).

Accumulation of metals like Hg, Zn, Cu and Cd have been recognized as a major factor for the co-selection of antibiotic resistance in aquatic environments (Seiler & Berendonk, 2012). Abundance of MAR bacteria in drinking water was increased from 20.4% to 36.4% upon entering the distribution system after chlorine percolation treatment. A significant proportion of MAR bacteria isolated from drinking water distribution systems displayed tolerance to different combinations of Cu²⁺, Pb²⁺ and Zn²⁺ where 20.9% of isolates were tolerant to all three metals. A positive association between MAR and metal tolerance was observed in the point of use compared to point of entry suggesting that bacteria became antibiotic resistant inside the galvanized iron and concrete pipes during the distribution (Calomiris, Armstrong, & Seidler, 1984). Presence of Cu²⁺ ions and CuO nanoparticles (1–100 µmol/l) enhanced the conjugative transfer efficiency of ARGs across bacterial genera. Cu²⁺ ions and CuO nanoparticles caused over production of reactive oxygen species (ROS). Increased ROS induced the SOS response as shown by the upregulation of DNA damage related genes which ultimately lead to the conjugative transfer of ARG. In addition, Cu²⁺ ions and CuO nanoparticles increased the conjugative transfer by increased expression of pilin production genes, increased cell membrane permeability and by conjugative regulator alteration leading to accelerated plasmid replication (Shuai Zhang et al., 2019). V and Ca accelerated HGT by conjugation in experimental and field studies (Suzuki, Kimura, Agusa, & Rahman, 2012). In an activated sludge plant, concentration of heavy metals was correlated with the enrichment of several ARGs because heavy metals can induce efflux pumps (Mao et al., 2015). There was a significant correlation (p < .05) between ARGs and heavy metals (Zn, Cu and Co) in MSW landfill groundwater (Yu, He, Shao, Zhang, & Lü, 2016). Cu and Cr concentrations in mariculture drainage ditch sediments were positively correlated with the presence of fluoroquinolone resistance gene acc(6)-lb-cr, β-lactamase resistance genes imp5, imp12 and qnrS (Z. Zhao et al., 2017).

Number of studies have shown that antibiotics can act as a selective pressure for the antibiotic resistance development in aquatic environments. Gentamicin and sulfamethoxazole significantly increased the transfer frequencies of antibiotic resistance at sub-inhibitory antibiotic concentrations (1/16 for sulfamethoxazole and right below the MIC for gentamicin) (Jutkina, Marathe, Flach, & Larsson, 2018). Remaining ARG concentrations in the effluent were significantly correlated with the remaining antibiotics concentrations in an activated sludge plant (p < .05) (Mao et al., 2015). Raw wastewater tetracycline concentration correlated with the antibiotic resistance prevalence in the treated effluent at high temperatures (Novo, André, Viana, Nunes, & Manaia, 2013). There were significant correlations between sulfamerizine concentration and the proportion of tetracycline resistance genes tet(B) and tet(C) genes in a wastewater treatment plant (Li et al., 2015). Conjugative transfer efficiency of pB10 plasmid with P. aeruginosa PAKexoT and activated sludge increased at ppb (µg/l) levels of tetracycline and sulfamethoxazole (Kim et al., 2014). Number of kanamycin resistance (Kr) genes acquired by the sediments lacking Kr genes after 15 days were significantly higher (p < .05) in the presence of kanamycin compared to the control condition which lacked kanamycin (Mao et al., 2014). In mesocosms treated with different sub-inhibitory oxytetracycline concentrations, tet(M) and transposon Tn916/1545 concentrations were correlated for all antibiotic concentrations and the selection patterns for total tet genes and Tn916/1545 relative to the oxytetracycline levels were also similar (Knapp et al., 2008).

O-xylene, ethylbenzene, trioxymethylene, styrene, 2,4-dichloroaniline and malachite green found in a mixed wastewater (Domestic and dyeing) promoted the conjugative transfer of RP4 plasmid from Escherichia coli HB101 to E. coli NK5449 by 4 to 200-fold compared to the control (Jiao, Chen, Gao, Zhu, & Rensing, 2017). A 60-fold increase in RP4 plasmid transfer efficiency was observed in the presence of ionic liquid [BMIm][PF6] (Wang, Mao, Mu, & Luo, 2015). Carbamazepine also promoted the HGT of plasmid-borne MAR genes in a similar manner to Cu²⁺ ions and CuO nanoparticles (Wang et al., 2019).

4. Removal/enrichment of ARB and ARGs in water supply systems and wastewater treatment unit processes

Water and wastewater treatment plants are designed according to the multiple-barrier concept where a stipulated pathogen log reduction is achieved by a combination of different treatment unit processes (Sano, Amarasiri, Hata, Watanabe, & Katayama, 2016). Several studies report the achievable log reductions for different pathogens by different treatment unit processes (Amarasiri, Kitajima, Nguyen, Okabe, & Sano, 2017; Sano et al., 2016). Observed log reductions had intra-process and interprocess variations depending on the pathogen of concern (Amarasiri et al., 2017). However, currently there are no stipulated minimum reductions for antibiotics, ARB and ARGs in water/wastewater treatment guidelines (Hong et al., 2018).

Membrane bioreactor treatment achieved significantly higher removals of ARB and ARGs (range of removal: 2.57–7.06 log₁₀) compared to conventional treatment methods like activated sludge, rotating biological contactor or oxidation ditch (range of removal: 2.37–4.56 log₁₀) (p < .05)(Munir, Wong, & Xagoraraki, 2011; O’Flaherty & Cummins, 2017). Sand filtration and sedimentation effectively removed ARGs from drinking water (Su et al., 2018). Retention in the sand filter medium with low nutrient conditions may have caused the ARB starvation and permanent loss of antibiotic resistance due to plasmid degradation (Griffiths, Moyer, Caldwell, Ye, & Morita, 1990; Tan et al., 2019). Drinking water treatment by conventional (flocculation + sedimentation + sand filtration + chlorination) and advanced (preozonation + flocculation + sand filtration + postozonation + Granular Activated Carbon + chlorination) treatment plants achieved 0.03-2.4 ARG log reductions. However, different ARGs displayed different log reduction values. For example, class 1 integron-integrase gene (intI1) had 2.4 log reduction by conventional process whereas advanced process provided 1.8 log reduction. Macrolide-lincosamide-streptogramin B (MLSB) resistance gene ermB removal by conventional process was negligible and advanced process achieved 1.1 log reduction. Flocculation + sedimentation step in the conventional treatment process contributed to the ARG reduction by the removal of ARB (Hu, Zhang, Jiang, Luo, et al., 2019).

Chromosomally located β-lactamase resistance gene mecA and plasmid-borne tet(A) genes (10⁵ copies/µl) could be removed below detection limit using free chlorine. However, tetA was able to tolerate double the concentration of free chlorine compared to mecA and similar results were obtained with UV. Differences in the distribution of adjacent thymine sites which can rapidly react with free chlorine is suggested as a possible reason for the observation. By using short and long amplicon detection, the study also confirmed that free chlorine doesn’t entirely destroy ARGs (M. Zhang, Chen, Yu, Vikesland, & Pruden, 2019). A study on the antibiotics, ARB and ARGs reduction by peroxymonosulphate (PMS) oxidation, UV irradiation and UV/PMS combination treatment revealed that sulfonamide resistant bacterial strain HLS-6 can be reduced by 5.3 log using UVC (100 μW/cm²)/PMS (1 mg/L) treatment. HO^・ and SO₄^・─ radicals generated by breaking the PMS peroxide bond under UV radiation was suggested to be responsible for ARB inactivation. However, a higher PMS dose of UVC/PMS (100 μW/cm²)/(20 mg/L) was needed to achieve 2.9 ± 0.1 log reduction of sul1 and 3.4 ± 0.1 log reduction of intI1 within 30 mins (Hu, Zhang, Jiang, Yao, et al., 2019). Even though ARB were degraded immediately in the presence of sulfamethoxazole, ARGs reduction was inhibited after some time. Competitive consumption of free radicals by sulfamethoxazole and ARGs was suggested as the possible reason (Hu, Zhang, Jiang, Yao, et al., 2019).

Many studies have shown that treatment processes are incapable of completely eliminating ARB and ARGs from source water. Furthermore, in some cases, WWTPs are shown to increase the abundance of specific ARGs in the plant effluent (Chu et al., 2018; Karkman et al., 2018; Rizzo et al., 2013; Jian Xu et al., 2015). According to a metagenomic study, there were no significant differences in the ARGs detected in potable source water and in the treated water at the extraction point confirming that ARGs were not removed during the drinking water treatment (Garner et al., 2018). Wastewater treatment lead to the reductions of ermB and tet(W) relative concentrations while relative concentrations of bla_TEM, sul1 and qnrS were increased. Spread of some ARGs among bacterial cells in activated sludge was suggested as a plausible reason (Rodriguez-Mozaz et al., 2015). Granular activated carbon filtration enhanced the abundance of ARGs in the filtered water. Formation of biofilms on the biological activated carbon surfaces where antibiotics and ARB can adhere to and grow is suggested as the possible reason. (Su et al., 2018; Tan et al., 2019; Xu et al., 2016). Twelve ARGs were discharged through the dewatered sludge and plant effluent at higher rates than influent values from two municipal WWTPs, indicating overall proliferation of resistant bacteria. There were statistically significant correlations (p < .05) between the effluent ARG concentration and residual antibiotic concentration, As, Zn, Pb and Hg concentration, suggesting the selective pressure for some ARG enrichment (Mao et al., 2015). Chlorination and ozone treatment increased the propagation of some ARGs in conventional and advanced drinking water treatment plants as a result of resistance development due to mutagenesis and cell damage repair mechanisms (Hu, Zhang, Jiang, Luo, et al., 2019).

In addition to WWTPs, sewer pipes containing inner surface biofilms are suggested as possible hotspots for accumulation and dispersal of ARGs. Auguet et al. (2017) observed no significant differences in the concentration of ARGs in wastewater entering and leaving sewer pipes. However, concentrations of blaTEM, qnrS, sul1, sul2 and IntI1 in the outlet biofilm were significantly different from the inlet biofilm concentrations (Auguet et al., 2017). Sul1 and sul2 were the most abundant ARGs in the biofilm with an approximate resistance gene/16S rRNA gene ratio of 0.1.

As shown here, different water and wastewater treatment systems have shown differences in the ARG removal capability; depending on the unit process and ARG of concern. In some cases, same treatment unit process has shown different removal efficiencies of ARGs depending on the treatment flow. For example, sand filtration contributed to ARG reduction in the advanced drinking water treatment process while it led to ARG propagation during the conventional process. In the advanced process, reduction of organic matter by preozonation of water before sand filtration reduced the possibility of biofilm development and ARG dissemination (Hu, Zhang, Jiang, Luo, et al., 2019). Since, present drinking water treatment plants, WWTPs and collection systems act as ARG reservoirs and HGT hot spots, further technological advances are necessary in order to reduce the release and propagation of ARGs in water environments.

5. Challenging questions regarding the ARB and ARGs in aquatic environments

5.1 How to quantify the health risks caused by antibiotics, ARB and ARGs present in aquatic environments?

Quantitative microbial risk assessment (QMRA) approach combines the information on occurrence, exposure and dose-response to evaluate the potential health risks posed by a particular pathogen (Figure 1). It has been applied in calculating the health risks caused by E. coli, Campylobacter, Legionella and other pathogens in different scenarios (Evers, Blaak, Hamidjaja, de Jonge, & Schets, 2016; Hamilton & Haas, 2016; Strachan, Dunn, Locking, Reid, & Ogden, 2006). QMRA has been suggested as a suitable method to quantitatively evaluate the additional human health risk caused by the ingestion of ARB in the environment (Ashbolt et al., 2013; Hong et al., 2018; Manaia, 2017), and Ashbolt et al. (2013) analyzed the feasibility and complexities of applying QMRA framework to evaluate the health risks caused by ARB (Ashbolt et al., 2013).

Figure 1. Quantitative microbial risk assessment (QMRA) procedure for assessing the human health risks by ARB and ARGs in aquatic environment. The diagram briefly describes the process and provide summarizes the current knowledge.

One major assumption in the approach was that pathogenic ARB will display similar behavior in the environment with nonpathogenic ARB and the pathogenicity will also be similar to antibiotic susceptible bacteria (Ashbolt et al., 2013; O’Flaherty, Solimini, Pantanella, De Giusti, & Cummins, 2019; Pepper et al., 2018). Based on a meta-analysis of E. coli log reduction by UV lamps (259 antibiotic sensitive E. coli data points and 44 antibiotic resistant E. coli data points), a level of 7.3 mJ/cm² UV was required for antibiotic sensitive E. coli and 7.5 mJ/cm² UV for antibiotic resistant E. coli with a medium- pressure lamp to achieve a 6 log reduction. For low-pressure lamps, 8.1 mJ/cm² UV was required for antibiotic sensitive E. coli and 8.4 mJ/cm² UV for antibiotic resistant E. coli. The study confirmed there was no statistical difference between the inactivation of antibiotic sensitive and resistant E. coli by UV and supports the idea that the behavior of ARB may be similar to antibiotic susceptible bacteria (O’Flaherty, Membré, & Cummins, 2018).

For the exposure assessment step of QMRA process, it is necessary to quantify the ARB or ARGs taken by an individual via ingestion, inhalation or contact during different processes like drinking or recreational activities (Manaia, 2017). A recent study evaluated the removal of antibiotic resistant E. coli using 5 different drinking water treatment process chain scenarios and calculated the human exposure level to antibiotic resistant E. coli by consuming drinking water treated by those systems in the summer and winter. According to the results, mean adult human exposure to antibiotic resistant E. coli from tap water consumption ranged between 3.44 × 10⁻⁷ and 2.95 × 10⁻¹ CFU/day depending on the scenario and treatment method (O’Flaherty, Borrego, et al., 2018). Another study reported that the risk of carrying an antibiotic resistant E. coli is 1.26 times higher for the consumers of drinking water contaminated with resistant E. coli (Coleman et al., 2012). Human exposure potential to antibiotic resistant E. coli via the consumption of lettuce irrigated with surface water has been evaluated. Study design included river water from rural and urban areas and lake water from rural areas. Antibiotic resistant E. coli were categorized in to single antibiotic resistant, multiple antibiotic resistant and potentially pathogenic antibiotic resistant. Modeling process also included the influence of E. coli decay, postharvest processes, time of consumption and consumer washing on the quantity of antibiotic resistant E. coli internalized by consuming 100 g of lettuce irrigated with surface water. According to the results, antibiotic resistant E. coli quantities between 1.00 × 10⁻² and 1.35 × 10⁶ colony forming units (CFU) per 100 g of surface water irrigated lettuce can be internalized based on different scenarios. The study also revealed that postharvest treatments of washing with water, rapid cooling with water and washing with a chlorine solution can clean lettuce cultivated using irrigation water contaminated with antibiotic resistant E. coli up to 2.7, 3.1 or 4.8 log CFU/ml (O’Flaherty, Solimini, Pantanella, De Giusti, et al., 2019). Among the E. coli detected in coastal surface waters from England and Wales (average = 233.3 CFU/100 ml, SD = 409.5), 0.12% were third generation cephalosporins (3GCs) resistant. Quantity of 3GCs resistant E. coli ingested during water sports varied depending on the sport and the water quality of the particular location. Highest quantity of 3GCs resistant E. coli were ingested by surfers (Leonard et al., 2015; Leonard, Zhang, et al., 2018). Based on the abundance of resistance genes in bathing waters, a mean number of 1.24 resistance genes per E. coli has been estimated (Leonard, Yin, et al., 2018). Human exposure to antibiotic resistant E. coli by recreational activities performed in bathing sites located in close proximity to WWTPs were predicted to be between 0.45 and 345.09 CFU/100 ml (O’Flaherty, Solimini, Pantanella, & Cummins, 2019). ARB colonization by exposure to methicillin resistant Staphylococcus aureus and vancomycin resistant Enterococcus spp. was not detected among spray irrigation workers (Rosenberg Goldstein et al., 2014).

Hunter et al. (2008) performed a meta-analysis and estimated that ARGs transfer rates vary from 10⁻² to 10⁻⁹ related to either donor or recipient bacterial counts and calculated the probability of ARG transfer to potential pathogens following oral ingestion (Hunter, Wilkinson, Catling, & Barker, 2008). Based on infectious doses of different bacterial pathogens, in some cases, 10 E. coli particles are enough to infect an individual (Schmid-Hempel & Frank, 2007). By exposing to coastal waters in England and Wales, it is possible to ingest more than 10 3GCs resistant E. coli. There were 6.3 million water sport sessions resulting in exposure to 3GCs resistant E. coli. These data evince that the beachgoers are at risk of an ARB infection and if the exposed people have compromised immune systems, the risks will be elevated (Leonard et al., 2015). Moreover, colonization of ARB can occur even with a lesser number of bacteria than the infectious dose (Manaia, 2017). Therefore, even though information on the exact quantity of ARB necessary to colonize or infect the humans is unavailable, it will be reasonable to assume an extremely low dose of ARB can pose human health risks (Manaia, 2017).

Limited availability of exposure and dose-response data of ARB and ARGs for different scenarios hamper the usage of QMRA approach in quantitatively evaluating the additional human health risks posed by ARB and ARG in aquatic environments (Manaia, 2017; Pepper et al., 2018). For example, exposure to ARB via the spray irrigation of reclaimed water was found to be statistically insignificant compared to the control group, because of the smaller sample size (Rosenberg Goldstein et al., 2014). Further studies on exposure and dose-response assessment following the ingestion of antibiotics, ARGs and ARB will be of utmost importance since the exposure levels and HGT efficiencies can be different among different bacterial strains.

5.2. How frequently do the ARG selection and propagation occur in aquatic environments? Are there any unknown hot spots?

Transfer of resistance genes to environmental bacteria needs to overcome several bottlenecks including ecological connectivity, founder effect, fitness costs and maintenance of the resistance genes in the absence of selection pressures (Bengtsson-Palme, Kristiansson, & Larsson, 2018; Martínez, 2012; Melnyk, Wong, & Kassen, 2015). Therefore, presence of ARGs, ARB and MGEs in a particular aquatic environment may not provide conclusive evidence on the transfer of antibiotic resistance to environmental strains.

Acquisition of resistance plasmids or chromosomal mutations are expected to cost the fitness of bacterial cell while chromosomal mutations carry a larger fitness cost compared to the acquisition of antibiotic resistance via a plasmid as explained by the high frequency of plasmid acquired evolution of antibiotic resistance (Silva et al., 2011; Vogwill & Maclean, 2015). However, compensatory mutations are shown to arise and alleviate the negative effects of resistance mutations (Andersson & Hughes, 2010; Woodford & Ellington, 2007). Ecological connectivity bottleneck for resistance gene transfer can be eased by bacteriophages which can infect the bacteria not belonging to the same genetic exchange communities (Muniesa, Colomer-Lluch, & Jofre, 2013a). Bacteriophage mediated HGT is suggested to be important for transferring chromosomally located ARG from environmental bacteria to other susceptible strains when there is no selective pressure (Muniesa et al., 2013a).

The ability of environmental resistome to induce the antibiotic resistance in human pathogens needs to be confirmed by direct evidence. However, such studies are limited because the origin of ARGs identifiable in human pathogens may have been a taxonomically and ecologically distant bacterial population and they may have diversified within the environmental bacterial community over time (Aminov, 2013; Aminov & Mackie, 2007; Perry & Wright, 2013). A class-A bla_CTX-M found in pathogenic bacteria was detected in environmental Kluyvera spp. whereas qnrA like genes were found in Shewanella algae (Humeniuk et al., 2002; Poirel, Kämpfer, & Nordmann, 2002; Poirel, Rodriguez-Martinez, Liard, Mammeri, & Nordmann, 2005). qnrA1, qnrB1, qnrS1, and aac(6’)-Ib-cr sequences isolated from marine bacteria in an aquaculture region were identical to the genes isolated from patients in neighboring hospitals with quinolone-resistant uropathogenic E. coli infections (Aedo, Ivanova, Tomova, & Cabello, 2014; Tomova, Ivanova, Buschmann, Godfrey, & Cabello, 2018; Tomova et al., 2015). Bacteria isolated from farmed fish contained tet genes which are 92–100% identical to the tet genes isolated from clinical bacteria suggesting the same origin (Furushita et al., 2003).

To confirm the selection and propagation of ARGs in aquatic environments, it is important to evaluate the transfer efficiency of ARGs by conjugation, transformation and transduction and the factors governing them. Inside a membrane bioreactor, average RP4 plasmid transfer frequency of 2.76 × 10⁻⁵/recipient was observed between E. coli K12 and indigenous bacteria. Number of transconjugants observed with Shewanella spp., Photobacterium spp., Pseudomonas spp., Proteus spp. and Vibrio spp. were higher than other genera, confirming these genra are more likely to acquire plasmid RP4 (D. Yang et al., 2013). Transfer rate of tet(M) gene from Gram negative marine bacteria to E. coli JM109 was 10⁻³, but to Enterococcus was below detection limit, whereas the tet(M) transfer rate between Gram positive marine bacteria and E. coli was below detection limit and that to Enterococcus was 10⁻⁵ (Neela, Nonaka, Rahman, & Suzuki, 2009). Integron mediated transfer frequency of oxytetracycline and tetracycline resistance genes between Shiga toxin-producing E. coli O157:H7 and E. coli K12 strain MG1655 CA32 in the stormwater at 28^°C were 1.04 × 10⁻⁶ and 2.50 × 10⁻⁷ respectively (Nagachinta & Chen, 2008). Transfer frequency between the microbial community in a WWTP and two donor strains, namely E. coli MG1655 and Pseudomonas putida containing one of multidrug resistant plasmids pKJK5, pB4 or RP4 have shown to be between 3 × 10⁻⁵ and 5 × 10⁻⁴/recipient (Li et al., 2018). Maximum transfer efficiencies of trimethoprim resistant plasmid pKJK5 and activated sludge bacteria were in the range of 1 × 10⁻³–4.3 × 10⁻²/recipient, and transconjugants with Aeromonas, Acinetobacter and Enterobacter were dominant. Biofilm bacteria had a higher conjugative potential compared to suspended activated sludge samples (Li, Qiu, Zhang, Liang, & Huang, 2019). Multiple antibiotic and heavy metal resistance plasmid pSCL located in Pseudomonas fluorescens was transferred to marine bacteria in natural seawater. Conjugation efficiencies ranged from 10⁻⁴ in day one and reduced till 10⁻⁷ in day 18 when the experiment was continued (Chandrasekaran, Venkatesh, & Lalithakumari, 1998). Further studies on evaluating HGT frequencies in different scenarios related to aquatic environments to supplement the currently available limited knowledge will significantly contribute to the prediction of human health risks related to ARB and ARGs in water environments.

Natural transformation of environmental bacteria depends on the exposure to extracellular DNA (eDNA) (Blum, Lorenz, & Wackernagel, 1997; Thomas & Nielsen, 2005). eDNA is broadly defined as DNA that is not associated with living biomass and eDNA is assimilated in the environment due to bacterial transformation (Corinaldesi, Beolchini, & Dell’Anno, 2008; Mao et al., 2014). eDNA adsorbs to negatively charged particles like clay, quartz, feldspar and refractory organic molecules by phosphates and cation bridging (Torti, Lever, & Jørgensen, 2015). Reported eDNA concentrations in sediments are 3-4 orders higher than in water (Corinaldesi, Danovaro, & Dell’Anno, 2005; England, Vincent, Trevors, & Holmes, 2004; Torti et al., 2015). Due to the interaction with wastewater solids like clay and sand, eDNA degradation rates are lower than intracellular DNA (Torti et al., 2015). Microfiltration membranes commonly used for wastewater treatment are not capable of removing eDNA. However, eDNA attached to wastewater solids were removed efficiently by membrane filtration (Slipko et al., 2019). Report by Bien et al. (2017) suggests that ARGs conveying eDNA is formed and remained by biological interaction within microbial loop (Bien et al., 2017). In a study performed in Haihe River basin in China, sul1, sul2, tet(W) and tet(T) concentrations in the sediment were significantly higher compared to water and these ARGs were mainly concentrated in eDNA. (Luo et al., 2010; Mao et al., 2014). By using kanamycin resistant gene located in pEASY plasmid as eDNA, Mao et al. (2014) showed that eDNA has the potential to transform indigenous bacteria (Mao et al., 2014). Moreover, eDNA attached to sludge has shown to preferentially couple with competent cells rather than free eDNA (Dong, Wang, Fang, Wang, & Ye, 2019). However, studies analyzing the transformation of bacterial strains using the eDNA extracted from environmental samples are scarce. Since eDNA and more than 90 species of naturally transformable bacteria including human pathogens can be found in water environments, it is possible to expect the development of antibiotic resistance due to natural transformation in aquatic environments (Johnsborg et al., 2007; Mao et al., 2014). Future studies on quantifying the efficiency of environmental bacteria transformation by eDNA containing ARGs extracted from aquatic samples may provide valuable insights on the contribution of eDNA on the acquisition of antibiotic resistance.

Wastewater treatment plants are identified as hotspots for dissemination of antibiotic resistance (Karkman et al., 2018). However, it is possible to have unidentified hot spots for antibiotic resistance inside the wastewater treatment plant. One possible location is the surface of biofilms. Biofilm matrix mainly consists of extracellular polymeric substances (EPS) (Flemming, 2016). There are studies describing the adsorption of antibiotics to the EPS in activated sludge systems (Khunjar & Love, 2011; Juan Xu, Sheng, Ma, Wang, & Yu, 2013). It is possible that some fraction of the antibiotics gets degraded while adsorbed in the EPS. However, adsorption of antibiotics in the EPS may increase the concentration of antibiotics in the biofilm surface (Figure 2a). Increased concentrations of antibiotics in the biofilm surface layer may result in the selection for antibiotic resistance. However, this hypothesis has to be tested for conclusive evidence.

Figure 2. Hypothesis for new antibiotic resistance hotspots. (a) Attachment of antibiotics to EPS can increase the antibiotic concentration in the biofilm surface layer and can act as a selective pressure, (b) Elevated concentration of chemicals in the vicinity of ultrafiltration and reverse osmosis treatment systems membrane surface can act as mutagens and can lead to the development of antibiotic resistance in the bacteria in gel/cake layer and surroundings via mutation.

Another possible hot spot in wastewater treatment plants is the gel/cake layer in the ultrafiltration and reverse osmosis treatment systems used for tertiary treatment. Concentration of chemicals in the vicinity of the membrane surface can increase by 10 to 20 times compared to the bulk concentration due to concentration polarization. If the chemical is a mutagen, elevated concentrations of the chemical in the biofilm may act as a selective pressure to develop antibiotic resistance in the bacteria in gel/cake layer and surroundings via mutation (Figure 2b). Even if there is a breach of the membrane, ARB will be inactivated by disinfectants. However, if the genetic material containing ARGs which is shown to be not completely removable by disinfection are leaked to the aquatic environments, environmental bacteria can uptake the ARGs and become resistant. Therefore, in addition to monitoring membrane rupture, monitoring of ARG concentrations in reclaimed wastewater may be required in some cases.

5.3. What are the suitable markers for evaluation of ARB/ARG contamination?

Due to the large number of ARB, ARGs and antibiotics present in aquatic environments, quantification of each bacteria, gene or antibiotic concentration is time, cost and Labor intensive. A suitable ARB/ARG contamination marker can contribute to minimize the workload involved in evaluating the contamination. However, until now no consensus exists on the suitability of particular antibiotics, ARB and/or ARGs which can act as a molecular/chemical marker to evaluate the contamination of aquatic environments by antibiotics, ARB and ARGs. A recent review emphasizes there is no one-size-fits-all type of solution for this question (Huijbers, Flach, & Larsson, 2019). The study defines five different environmental surveillance objectives; risk of transmission, risk of evolution, risk of ecological effects, population-level resistance prevalence and population-level antibiotic use and critically discusses the suitability of markers like antibiotics, metals and biocides, known ARGs and MGEs, emerging ARGs, fecal contamination markers and ARB number to achieve the above objectives. It concludes that a combination of markers for fecal contamination and antibiotic resistance, together with selective agent measurement can be of use to achieve all five objectives (Huijbers et al., 2019).

By analyzing crAssphage genome sequence data obtained from wastewater samples and human samples (from 67 countries in 6 continents), Edwards et al. concluded that crAssphage is stable in individuals and common throughout the human population (Edwards et al., 2019). Karkman et al. (2019) analyzed the correlations between crAssphage levels with ARG abundance levels in sewage (Karkman, Pärnänen, & Larsson, 2019). In an urban stream contaminated by human sewage, sul1, sul2, tet(O), tet(W) and ermF concentrations had significant correlations with crAssphage when quantified using digital droplet PCR (>0.8 from Spearman’s rank correlation coefficient) (Stachler, Crank, & Bibby, 2019). However, it has been suggested that the abundance of ARGs in sewage can be simply due to the fecal contamination instead of selection pressure (Karkman et al., 2019). Therefore, crAssphage quantification may only provide information about the risk of ARG transmission in aquatic environments (Huijbers et al., 2019).

Study specific results have been reported for the most abundant ARG in many studies conducted in different geographical locations. Sul genes were reported as the most detected ARGs in wastewater effluents from several countries including China, Sweden and USA (Ben et al., 2017; Berglund, Fick, & Lindgren, 2015; Mao et al., 2015; Munir et al., 2011). Others report tet(M) as the most commonly detected ARG in both Gram-positive and Gram-negative bacteria in wastewater lagoons, and cattle feedlots having different antibiotic usage strategies have complimented the result, suggesting tet(M) as a suitable candidate to confirm the contamination by antibiotics (Peak et al., 2007; Roberts, 2005). It is clear that the detectable ARGs and their quantities are location specific. It can be further confirmed by Table 1 where different aquatic environments have shown the presence of different ARGs. However, detection of only ARGs may provide limited information regarding the risk of transmission, evolution and prevalence of resistance at population-level (Huijbers et al., 2019).

ARGs in aquatic environments can be transferred through mobile genetic elements (MGEs) and can later be assimilated by the environmental bacteria to become antibiotic resistant. MGEs and ARGs in aquatic environments have shown correlations (Park, Kim, Kim, & Seo, 2018; Pruden et al., 2006; Tao et al., 2016). IntI1 in groundwater had significant correlations (p < .01) with ermA, sul1, tet(O), and tet(W) (Szekeres et al., 2018). Positive correlations were observed between intI1 and qnrD, intI2 and sul3, intI3 and tet(X), Tn916/Tn1545 and sul2, and ISCR1 and sul3 in untreated hospital wastewater (Qiang Wang, Wang, & Yang, 2018). In a pharmaceutical WWTP, abundance of ARGs correlated with the abundance of MGEs (R²=0.52, p < .05). Among the ARGs, aminoglycoside and sulfonamide resistance genes displayed higher correlations with MGE abundance (R² > 0.74) (Tao et al., 2016). In recreational water samples, IntI1 had a statistically significant positive correlation with sul1, bla_TEM and tet(X) (Dong, Cui, Fang, Huang, & Wang, 2019). In a recent metagenomics analysis combined with proximity-ligation assay on wastewater, IntI1 exhibited links with 39 clusters within the Beta- and Gammaproteobacteria displaying the broad host range (Stalder, Press, Sullivan, Liachko, & Top, 2019). Therefore, detection of MGEs in aquatic environments may be considered as an indicator for the presence of ARGs. However, similar to ARGs, it may only provide limited information regarding the risk of transmission, evolution and population-level prevalence of resistance.

Antibiotics are identified as selective agents for antibiotic resistance. Based on the data from Mekong river delta and Jiulongjiang river basin, sulfamethazine (SMT) has been suggested as a molecular-marker for aquatic environment contamination by livestock sources while Haihe river basin data suggests sulfachloropyridazine (SCP) as a suitable marker for same purpose (Hongyou Jiang, Zhang, Xiao, Geng, & Zhang, 2013; Luo et al., 2011; Managaki, Murata, Takada, Bui, & Chiem, 2007). These results confirm that antibiotics also display the location specificity. Since apramycin is used only for swine and tylosin is exclusively used for animals (Kemper, 2008), detection of these antibiotics in a particular aquatic environment can lead to the conclusion of livestock related contamination. However, antibiotics like oxytetracycline is used on both animals and humans which makes it difficult to identify the contamination source if detected in municipal wastewater (Kemper, 2008; Shimizu et al., 2013). If it is possible to identify and regulate the antibiotics usable in particular situations (e.g. Swine farms only, aquaculture only, human consumption only etc.), there is a high possibility of identifying a molecular/chemical marker to identify antibiotic related contamination. These information can be of help in evaluating the risks of ecological effects and the evolution of antibiotic resistance (Huijbers et al., 2019).

Berendonk et al. (2015) suggested an array of 6 bacterial groups (E. coli, Klebsiella pneumoniae, Aeromonas spp., P. aeruginosa, Enterococcus faecalis and Enterococcus faecium) and 13 ARGs (intI1, sul1, sul2, bla_CTX-M, bla_NDM-1, bla_VIM, bla_KPC, qnrS, aac-(6’)-lb-cr, vanA, mecA, ermB, ermF, tet(M) and aph) which can be used to assess the antibiotic resistance status in environmental settings (Berendonk et al., 2015). WHO suggests an integrated surveillance system under One health approach to monitor extended spectrum ESBL-E.coli in humans, food chain and the environment which can be performed in many countries with low or minimal resources (Tricycle project) and expects that the data will be comparable, statistically valid and robust (World Health Organization, 2017a, 2018). In a recent study, in vivo proximity ligation method Hi-C (physically linking ARGs, plasmids and integrons to their respective host chromosomes) was used to evaluate the plasmid-host association in a wastewater sample. According to the results, IncQ plasmids and class I integrons had the broadest host range. Among all the detected bacterial taxa, Aeromonadaceae was associated with the highest number of ARGs (Stalder et al., 2019). Therefore, it may be possible to use the bacteria from Aeromonadaceae as potential markers for the contamination of aquatic environments with ARB and ARGs.

Types and concentrations of antibiotics used in different countries and geographical regions vary and as a consequence, detectable antibiotics, ARB and ARGs in aquatic environments will also be location specific with varying compositions (Shimizu et al., 2013). A systematic review on the abundant ARB in different water environments can be of great help in narrowing down the candidate markers for antibiotic contamination. Proximity-ligation can be optimized using methods like abundance normalization to obtain the connections between the plasmids carrying ARGs and their hosts (Stalder et al., 2019). These narrow-down methods combined with qPCR-based quantification may provide valuable information regarding the ARB which can be used as contamination markers. In terms of evaluating the human health risks, the absolute number of ARB present in a particular aquatic sample is essential. Therefore, the usage of ARB contamination markers will significantly contribute to the evaluation of human health risks caused by antibiotic resistance in aquatic environments.

5.4. What is the effect of VBNC status on antibiotic resistance?

Majority of the bacterial strains including human enteric bacteria can’t be cultured, are obligate anaerobes or viable but non-culturable (VBNC) (Amann, Ludwig, & Schleifer, 1995; Miura et al., 2013). Up to now, most of the evaluations on ARB has been conducted with culturable fraction of bacteria. However, several studies have attempted to answer the contribution of VBNC state bacteria on the antibiotic resistance. Vancomycin-resistant enterococci maintained the antibiotic resistance during the VBNC state and the resuscitated cells also were resistant to vancomycin (Lleo, Bonato, Signoretto, & Canepari, 2003). According to Lin et al. (2017), low dosage of free chlorine disinfection (0.5 mg/L) effectively reduced the cultivability of E. coli and induced a VBNC state. VBNC bacteria displayed higher persistence to 9 antibiotics compared to their culturable counterparts. Transcript expression levels of folA, acrA and acrB (multidrug efflux) genes were increased more than 2-fold in the VBNC E. coli compared to culturable bacteria (Lin, Ye, Chen, Zhang, & Yu, 2017). Even though E. coli became VBNC during the low nutrient and energy phase of carcass leachate treatment, other bacteria present in the leachate took up the plasmid indicating the nutrient deprivation and VBNC bacteria may not hinder the plasmid transfer (Salcedo & Kim, 2017). Plasmid transfer efficiency between VBNC P. fluorescens and marine bacteria was 10⁻⁵ at the initiation of VBNC state and decreased with the number of days passing. After 15 days from the beginning of VBNC state, transfer efficiency was stabilized at 10⁻⁷ range (Chandrasekaran et al., 1998). sul3 was detected only in the non-culturable fraction of bacteria in the water collected from Japanese marine aquaculture environments, Manila Bay and Taiwanese aquaculture water (Suzuki et al., 2013, 2019).

Since disinfection can induce VBNC state in bacteria, VBNC bacteria can be antibiotic resistant, and the VBNC bacteria carrying ARGs can survive and resuscitate after the removal of stress conditions, it is important to include the VBNC fraction of bacteria in evaluating the human health risks caused by ARB and ARGs. However, it is not always possible to separately quantify the culturable and VBNC bacterial fraction in each sample. To address this problem, several studies have evaluated and modeled the induction of VBNC state in culturable bacteria under disinfectants like chlorine, chloramine (S. Chen et al., 2019; Oliver, Dagher, & Linden, 2005) and UV (Shenghua Zhang, Ye, Lin, Lv, & Yu, 2015). As the majority of environmental bacteria belong to VBNC, these models combined with the data on culturable ARB quantities may be usable in approximating the VBNC fraction of ARBs in a sample and in turn in evaluating the health risks.

6. Conclusions

Ingestion of antibiotic resistant bacteria and antibiotic resistance genes present in aquatic environments can pose a human health risk. However, currently it is not possible to quantitatively evaluate this human health using QMRA, because of the limited availability of necessary information including exposure assessment and dose-response assessment data for different scenarios. Furthermore, there are unknown hotspots contributing to the propagation of antibiotic resistance in aquatic environments. Role of eDNA which is ubiquitous in water environments on the acquisition of antibiotic resistance by environmental bacteria through natural transformation needs to be further elucidated. Identification of suitable antibiotic resistance markers also is a challenge because there is no one-size-fits-all solution. However, identifying an ARB as a contamination marker will be beneficial in terms of human health risk evaluation. Contribution of the VBNC fraction of bacteria on the propagation of antibiotic resistance and their effects on the human health also is a point of concern.

Disclosure statement

No potential conflict of interest

Additional information

Funding

This work was supported by the Japan Society for the Promotion of Science under Grants18K13855 and 16H01782, and by “The Sanitation Value Chain: Designing Sanitation Systems as Eco-Community Value System” Project from Research Institute for Humanity and Nature (RIHN, Project No. 14200107).