Wastewater surveillance for antibiotic resistance genes during the late 2020 SARS-CoV-2 peak in two different populations

The first cases of coronavirus disease 2019 (COVID-19) in Portugal (PT) were reported to the World Health Organization (WHO) on March 2, 2020 (WHO 2020). The Portuguese government declared a state of emergency on March 19, 2020, which began the first lockdown (RTP 2020). The number of cases escalated rapidly after this, with 600–800 daily cases reported on April 1, 2020 (WHO 2020). Daily reported cases remained at about this level over the summer until they started to rise again in September and October. Within 1 year of the first cases in the country, there were over 700,000 cases and 16,000 deaths reported in PT (WHO 2020). SARS-CoV-2, the causative virus for COVID-19, was detected in the United States earlier than in PT, with the first confirmed COVID-19 case identified in Snohomish County, Washington from a sample collected on January 18, 2020 (NPR 2020). There were over 80,000 confirmed cases and 1,000 deaths in King County, Washington within 1 year of the first confirmed case (King County Public Health 2020).

A clinical diagnostic test was quickly developed and distributed around the world (Sheridan 2020), but the demand for tests outpaced testing capacity. This created a need for disease surveillance separated from clinical testing such as wastewater surveillance (WS). WS has been successfully used to supplement clinical surveillance for other infectious diseases, most notably for poliovirus elimination and eradication (The World Health Organization 2003; Hovi et al. 2012). Some of the earliest reports in March 2020 detected SARS-CoV-2 in wastewater all over the world (Ahmed et al. 2020; Medema et al. 2020; Peccia et al. 2020). A Portuguese effort to conduct SARS-CoV-2 WS began on April 27, 2020, and continued until December 2, 2020 (Monteiro et al. 2022). In King County, wastewater was collected for SARS-CoV-2 WS throughout 2020 and 2021 (Philo et al. 2021; Philo et al. 2022).

Although existing anti-viral medications and antibiotics such as remdesivir and azithromycin, respectively, were being investigated to treat SARS-CoV-2 (Chibber et al. 2020), no medication had been shown to have clinical efficacy. In the absence of better treatments, antibiotics were widely prescribed to treat COVID-19 and secondary infections. A cross-sectional survey carried out from April 7 to 28, 2020 indicated that 60% of doctors in PT routinely prescribe antibiotics to patients admitted to the hospital, and β-lactams with a macrolide was the most common medication combination (Beović et al. 2020). In the same cross-sectional survey, all doctors surveyed in North America routinely prescribed antibiotics to patients (Beović et al. 2020). Other data suggest as many as three-quarters of hospitalized patients were prescribed antibiotics (Langford et al. 2021; Russell et al. 2021; Centers for Disease Control and Prevention et al. 2022), even though a meta-analysis suggests that bacterial co-infection was low (Langford et al. 2021). Additionally, patients who were mechanically ventilated were more likely to be prescribed antibiotics (Langford et al. 2021). Antibiotics were also routinely used on surfaces, with over half of the surface disinfectants approved to disinfect SARS-CoV-2 containing a quaternary ammonium compound (Hora et al. 2020), which have been shown to upregulate efflux pumps associated with multidrug resistance in bacteria (Merchel Piovesan Pereira & Tagkopoulos 2019). The wide use of antibiotics and surface disinfectants suggests that bacteria would have experienced increased resistance during this time (Centers for Disease Control and Prevention 2019a; Hutchings et al. 2019).

Almost every organism threatened by antimicrobial resistance (AMR) and tracked by the US Centers for Disease Control and Prevention (CDC) showed an increase in resistance compared to 2019 (Centers for Disease Control and Prevention et al. 2022). Bacteria such as carbapenem-resistant Enterobacterales exhibited decreased resistance before the pandemic, but hospital-onset infections increased by 35% after the pandemic (Centers for Disease Control and Prevention et al. 2022). Changes in abundance during the pandemic for antibiotic resistance genes (ARGs) causing clinical resistance are not well understood.

Given the apparent increase in AMR and the widespread adoption of WS for SARS-CoV-2, there have been numerous calls to include AMR targets in these programs (Pruden et al. 2021; Jin et al. 2022). The CDC recently named AMR as a future target for the National Wastewater Surveillance System (NWSS) (Centers for Disease Control and Prevention 2022). Yet, very little research has been done to assess the impacts of COVID-19 on AMR in wastewater. A study carried out in Las Vegas, Nevada was able to detect increases in ARGs in wastewater collected from the wastewater treatment plant (WWTP) corresponding to the antibiotics used to treat COVID-19 infections in late 2020 and early 2021 (Harrington et al. 2022). There are very few, if any, studies looking at this problem in multiple communities, both inside and outside of the United States.

Samples originally collected from PT and King County, Washington, USA (WA) for SARS-CoV-2 WS were tested for different ARGs targeting some of the most threatened antibiotics. These two study locations were chosen because they represent two geographically distinct areas with different scales of the COVID-19 pandemic and different antibiotics prescribing behavior. During the study period, the average number of daily incident COVID-19 cases in PT was 1,940 (maximum: 6,297) (WHO 2020), but the average daily incident cases in WA was 88 (maximum: 295) (King County Public Health 2020). The goal of this research is to validate that methods used for SARS-CoV-2 WS apply to ARGs and to understand how ARGs in wastewater changed during the COVID-19 pandemic in late 2020.

Methods carried out in PT and WA were not identical because the samples used in this project were originally collected for different projects using different concentration methods. Primary influent wastewater and effluent wastewater were not frozen for all locations and could therefore not be reconcentrated and extracted using the same method. Additionally, different quantitative polymerase chain reaction (qPCR) reagents and platforms were available in the two labs. To account for these differences, all qPCR results are adjusted using the effective volume assayed to allow for greater consistency in data outputs (Crank et al. 2023). Because the goal was to assess changes in ARGs contemporaneous with changes in SARS-CoV-2, sample dates were selected for additional analysis around the peaks of the two countries' first COVID-19 cases, resulting in slightly different sampling periods.

Wastewater concentrate, originally collected in PT to conduct SARS-CoV-2 WS from August 18, 2020, through November 25, 2020, was used in this study (Monteiro et al. 2022) (SI Table 1). Samples were collected from five WWTPs and from three hospitals where patients were referred for COVID-19 care. Two WWTPs are in the North and three are in the South. The three hospitals are in the water catchment areas of some of the WWTPs. Sampling dates were selected around the first major pandemic peak in November 2020. Starting August 18, 2020, the first weekly sample collected every other week was chosen for ARG analysis (SI Table 1). The last samples processed were collected on November 25, 2020.

Twenty-four-hour composite samples were collected using autosamplers (ISCO, Inc., Lincoln, NE, USA) at all sampling locations except the hospitals in the north, where grab samples were collected because there were no secure locations to store the autosampler. All samples were transported with refrigeration at 5 °C (+/ − 3 °C) to the Laboratório Análises do Instituto Superior Técnico (LAIST) at the Instituto Superior Técnico in Lisbon, PT within 8 h of collection. Samples were processed within 24 h upon arrival at the LAIST.

Wastewater was concentrated according to Monteiro et al. (2022) using Inuvai R180 hollow fiber filters with a molecular weight cut-off of ≤ 18.8 kDa (Inuvai, Fresenius Medical Care, Bad Homburg, Germany). The step-by-step protocol is included in Supplemental Information Section A.1. Briefly, the filters were first primed using a 0.9% v/v NaCl solution. After priming, the wastewater sample was filtered at a flow of 200–250 mL/min. Next, the sample was eluted using 1× PBS (pH = 7.4) supplemented with 0.01% sodium polyphosphate, 0.01% Tween 80, and 0.001% antifoam. The total elution volume was 300 mL.

The elution was further concentrated using polyethylene glycol (PEG) precipitation. PEG 8000 was added for a 20% v/v final concentration (60 g per 300 mL of eluate) with NaCl (7.17 g per 300 mL). The sample bottles were agitated until the PEG completely dissolved and were then placed on a rocking table at 4 °C overnight. Samples were next centrifuged at 10,000 × g for 30 min and resuspended in 5.0 mL of 1× PBS (pH = 7.4). Samples were stored at −80 °C until extraction using the QIAamp Fast DNA Stool Mini Kit (QIAGEN, Hilden, Germany) following the protocol for human DNA analysis with the following modifications. The input material was 220 μL of wastewater concentrate, and the DNA was eluted into 100 μL of the included elution buffer (buffer AVE). Extracts were kept at −20 °C until analysis.

Wastewater concentrate previously collected from two WWTPs in WA serving King County was used in this study to test for SARS-CoV-2 (Philo et al. 2021 , 2022). Samples were selected around increased COVID-19 cases from November 2020 to January 2021. Samples collected every other week from September 1, 2020 to February 16, 2021 were tested for ARGs (SI Table 2). Primary influent wastewater was grab sampled in the early morning. All samples were transported on ice to the Environmental and Occupational Health Microbiology Lab (EOHML) at the University of Washington Seattle campus on the same day as sample collection. Samples were stored at 4 °C and processed within 48 h upon arrival at the EOHML.

Wastewater was concentrated following the skimmed milk flocculation protocol according to Philo et al. (2021). A step-by-step protocol is included in Supplemental Information Section A.2. Briefly, a 5% skimmed milk solution was added to 250 mL of wastewater (1% v/v final). The pH of the wastewater was dropped to 3–4 using 5 M HCl and shaken at about 200 RPM for 2 h. After shaking, the wastewater was centrifuged at 3,500 × g and 4 °C for 30 min, and pellets were resuspended in 3.0 mL of sterile 1× PBS (pH = 7.4). Resuspended pellets were stored at −80 °C until extraction.

Genetic material was extracted using the Viral RNA Mini Kit (QIAGEN, Germantown, MD, USA). Per the manufacturer, both RNA and DNA will be extracted using this kit. Each sample was extracted in duplicate with a doubled input volume for a total extraction volume of 560 μL using the necessary volume adjustments described in the manufacturer's protocol. Each column was eluted in 60 μL of buffer AVE for a total elution volume of 120 μL. Duplicate extracts were combined and re-aliquoted before being stored at −20 °C for later molecular analysis.

Specific gene targets were selected because they confer resistance against some of the last-line antibiotics and/or have been identified as threats to public health by the US CDC and the WHO (Tacconelli et al. 2018; Centers for Disease Control and Prevention 2019b). Samples were tested for genes conferring resistance to extended spectrum β-lactams and cephalosporins (bla_CMY, bla_SHV, and bla_CTX-M) (Roschanski et al. 2014), colistin (mcr1) (Donà et al. 2017), vancomycin (vanA) (He et al. 2020), and methicillin (mecA) (Sabet et al. 2007). Samples were assessed for ARGs using pre-existing qPCR assays (SI Table 3). All qPCRs in PT were carried out on either the Applied Biosystems QuantStudio™ 5 System or the 7300 Fast Real-Time PCR System (Applied Biosystems Corp., Waltham, MA, USA). DNA was amplified using the Luna^® Universal qPCR Master Mix for SYBR assays and the Luna^® Universal Probe qPCR Master Mix for TaqMan probe assays (New England Biolabs, Ipswich, MA, USA) in a total reaction volume of 15 μL with 1.0 μL of the DNA extract, 7.5 μL of the respective master mix, and variable amounts of water, primers, and probes according to their respective publications. Reaction conditions and primer and probe concentrations are provided in SI Table 4. All qPCRs in Seattle were carried out using the iTaq Universal Probes Supermix or the iTaq Universal SYBR Green Supermix and the Bio-Rad CFX qPCR systems (Bio-Rad Laboratories, Hercules, CA, USA). The total reaction volume was 15 μL with 7.5 μL of the respective master mix, 2.0 μL of the extract, and variable amounts of water, primers, and probes. Reaction conditions and primer and probe concentrations are provided in SI Table 4. All samples were run in duplicates with duplicate 10-fold dilutions. A negative control of nuclease-free water was included with each run. Samples were quantified using standard curves generated using 10-fold dilutions of positive controls. Positive controls in WA were gBlocks obtained from IDT (Integrated DNA Technologies, Inc., Coralville, IA, USA). Positive controls in PT were created following the methods described in Section A.3 of the SI. All positive controls were quantified using the digital PCR (dPCR) as described in Section A.3 of the SI and SI Table 5.

To assess trends in ARG abundance across the study period, random-effect models were fit for the PT and WA datasets, each regressing log-transformed gc on day with a random effect for each sampling site. This assumes that gc within a site are more similar than gc between sites. To determine if metro areas with hospital effluent and WWTP influent trended together in PT, Pearson's correlation coefficients were calculated for samples collected in the same week. To test for differences in ARG abundance between the qPCR and the dPCR, Wilcoxon signed-rank tests were carried out on paired samples using the built-in R function (wilcox.test (paired = TRUE)). Significance was assessed at p < 0.05.

The most abundant gene target, on average, during the sampling period in both PT and WA is bla_SHV, with 1.85 × 10⁸ gc/L and 4.26 × 10¹⁰ gc/person/day, respectively (Table 1). The least abundant gene targets in PT are bla_{CTX-M Groups 2, 8, 25}, with 4.76 × 10³ gc/L (Table 1). However, bla_{CTX-M Groups 2, 8, 25} is highly abundant in WA, with 1.73 × 10¹⁰gc/person/day. Both bla_SHV and bla_CTX-M are among the most identified extended spectrum beta-lactamase (ESBL) genes (Poirel et al. 2012; Roschanski et al. 2014). Because bla_CTX-M genes are frequently associated with mobile genetic elements, detection at any abundance suggests that the genes can move easily (Cantón et al. 2012). Both WA and PT had high bla_cmy concentrations on average, with 1.07 × 10⁹ gc/person/day and 1.03 × 10⁵ gc/L, respectively (Table 1). This is particularly concerning because bla_CMY is plasmid-associated and can move easily between bacteria (Roschanski et al. 2014).

ARG abundance in wastewater in other studies varies widely. For example, bla_SHV is highly abundant (Quach-Cu et al. 2018) and one of the least detected genes in wastewater (Harnisz & Korzeniewska 2018). Additionally, while it is common to describe the number of wastewater isolates resistant to a certain antibiotic or containing a certain ARG at a given time (Kwak et al. 2015; Zhang et al. 2019), it is less common to quantify the concentration of ARGs in influent wastewater over time. Furthermore, when ARGs are detected, the genetic context around them, and whether they are intra- or extracellular, is not always known. While intracellular ARGs are more common in wastewater than extracellular ARGs, extracellular ARGs are more likely to be picked up by bacteria in stressful environments such as wastewater (Zarei-Baygi & Smith 2021). Surveilling ARGs in wastewater can help give a better idea of the underlying prevalence in the population.

To determine if antibiotic resistance in wastewater changed during the late 2020 surge in COVID-19 cases before medications had been developed, random-effect models were fit for PT and WA data. Log-transformed gc were regressed on sampling day with a random effect for each site to allow for different baseline ARG abundances by the site. In PT, the only two genes that significantly increased over the course of the sampling period are mecA and bla_{CTX-M Groups 2, 8, 25} (Table 2, SI Figure 1D and E). Across the study period in PT, the average mecA concentration is 1.05 times higher than the previous day (95% CI: 1.04–1.06, p < 0.001). Additionally, mecA concentrations tend to be higher in hospital effluent compared to WWTP influent, particularly in the north regions (SI Figure 1E). This suggests that the hospital signal is still detectable at the WWTPs but has been diluted throughout the system. However, in the south 2 region, mecA is higher in the WWTP influent than the hospital effluent. This region is a large metro area, and WWTP influent contains wastewater from several other hospitals and clinics that could be increasing the signal. In WA, mecA also increased during the study period (Table 2, SI Figure 2E). Average daily concentrations are 1.01 times higher than the previous day (1.00–1.01, p < 0.001). However, these results are likely heavily influenced by relatively higher concentrations in several of the samples at the end of the study. Despite the statistical evidence of a trend, changes, such as bla_{CTX-M Groups 2, 8, 25} in PT and mecA in WA, are small. Likely, the qPCR is not sensitive enough and these results should be assessed critically. There is no statistical evidence that any of the other genes in PT or WA changed over time (Table 2, SI Figures 1 and 2).

There is strong clinical evidence to back up the rise in mecA detected in PT and WA over the course of the study period. First, Staphylococcus aureus is the most isolated organism in intensive care units, and the incidence of methicillin-resistant S. aureus (MRSA) ventilator-associated pneumonia (VAP) rises to over 30% after 5 days of mechanical ventilation (Lam & Wunderink 2006). Because COVID-19 patients are often ventilated for over a week (Hazard et al. 2020), they are also at increased risk for MRSA infections. A review by Bassetti et al. (2022) found that MRSA is the most common secondary infection in ventilated COVID-19 patients. Additionally, the US CDC showed that hospital-onset MRSA cases increased by 13% in 2020 compared to 2019 (2022). Rising MRSA infections associated with COVID-19 patients likely led to the increase in mecA wastewater abundance detected in this study. Data on AMR infections, VAP in the hospitals, and antibiotics used in the water catchment area would further support these data showing changes in gene targets.

The US CDC identified increased AMR during the first year of the pandemic in all the organisms they tracked for which data were available in 2022. This includes bacteria with some of the ARGs tested for in this study, such as ESBL-Enterobacterales, MRSA, and vancomycin-resistant Enterococcus (Centers for Disease Control and Prevention et al. 2022). While we were not able to detect changes for most of the ARG targets in wastewater, we demonstrated that qPCR can be used for general surveillance and to assess changes in gene abundance over time in certain settings. There is little other work assessing changes in environmental AMR during the pandemic using molecular methods. Harrington et al. (2022) collected primary influent wastewater weekly from a WWTP in Nevada from November 2020 through the end of January 2021 and found significant increases in β-lactam and fluoroquinolone ARGs in December 2020 associated with rising COVID-19 cases (Harrington et al. 2022). Harrington et al. (2022) were also able to show significant decreases in ARGs from December to January when COVID-19 cases again dropped. It is important to validate and implement WS programs to better understand how widespread antibiotic use affects entire communities of bacteria.

Seasonality has been observed for AMR in other studies. A meta-analysis found that respiratory bacteria had higher odds of being resistant to antibiotics in the winter than the rest of the year (Martinez et al. 2019). Resistance in Escherichia coli and Klebsiella pneumoniae urine isolates have also been shown to vary seasonally, with resistance to different antibiotics peaking in winter months in the Netherlands (Martínez et al. 2020) and K. pneumoniae resistance peaking in the summer in Michigan (Cassone et al. 2021). As with the current study, seasonal changes in clinical isolates in the Netherlands were very small, with the fraction of resistance isolates among total isolates being less than 1% different across seasons (Martínez et al. 2020). In both the Netherlands and Michigan, antibiotic use peaked several months before resistance peaked, suggesting a delay in resistance patterns in clinical samples (Martínez et al. 2020; Cassone et al. 2021). It is unclear if resistance trends in wastewater would follow a similar delay. However, this highlights the importance of conducting continued AMR environmental surveillance to better detect changes over time rather than in response to a global pandemic.

Pearson's correlation coefficients were calculated for log-transformed ARG abundances in hospital effluent and WWTP influent samples collected in the same week from the same water catchment areas in PT (Table 3). There was little statistical evidence that ARGs within metro areas trended together, except bla_SHV, which showed strong correlations in concentrations in the North 1 (r = 0.95, 95% CI: 0.59–0.99, p < 0.001) and South 2 (r = 0.80, 95% CI: 0.13–0.95, p = 0.03) regions (Table 3). No genes were found to be correlated in the North 2 region. This is likely because of the small number of paired samples available in this region. It is important to note that not all samples within a metro area were collected on the same day, with some samples being collected 1 day after the other (SI Table 1). This reduces our ability to associate ARG concentrations within water catchment areas. Future work should model the residence time of the wastewater conveyance system to aim to collect samples from different points that are more closely associated with each other.

This study describes antibiotic resistance in wastewater during the pandemic in two different countries and attempts to fill the research gap of changes in environmental AMR due to the SARS-CoV-2 pandemic. The described work is novel because there are very few other publications attempting to understand the effects of the COVID-19 pandemic on AMR in wastewater. While the current study did not detect changes in most ARGs assayed, the data showed significant increases in mecA in PT and WA, which aligns with clinical reports of increased MRSA infections in ventilated COVID-19 patients. This indicates that qPCR can be used to supplement hospital data for AMR in certain settings.

Future work should aim to collect samples more frequently and to better understand AMR infections at the time of sample collection. Data on the number and types of antibiotics used, as well as how patients are typically treated, will help shed light on the utility of WS for ARGs. Additionally, molecular methods should be used that are less subject to inhibitors. Concentrating wastewater using skimmed milk flocculation results in significant inhibition of the PCR assays, and the protocol was optimized for isolating SARS-CoV-2 RNA, not bacterial DNA. This could result in an underestimation of ARG abundance in King County. Future work should identify MRSA infections and antibiotics used in the hospitals sampled in this study to better describe the temporal relationship. Finally, future studies should compare detected ARGs with the numbers of hospitalized and ventilated patients to better understand the relationship between the clinics and environmental ARGs.

Study data are limited in several ways. First, the samples were initially collected, concentrated, and extracted for the surveillance of an RNA virus, SARS-CoV-2. Particularly for WA, DNA abundance in these samples could be systematically underestimated due to the use of the ViralRNA extraction kit. Second, samples were selected to cover a longer time rather than samples as frequently as possible. This could have obscured true changes in ARG abundance over time. Finally, this study does not attempt to describe which organisms are harboring the detected ARGs and the genetic context around the ARGs or to isolate MRSA from the wastewater.

Statistical advice for this project was provided by students and faculty through the Department of Biostatistics Consulting Program at the University of Washington, Seattle. Sample collection was facilitated by the West Point Process Lab, the South Treatment Plant collection team, and collaborators in Portugal. This work could also not have been carried out without the support of all members of the Environmental and Occupational Health Microbiology lab in Seattle, WA and the Laboratório Análises do Instituto Superior Técnico in Lisbon, Portugal. The US Fulbright Student Program in Portugal helped make the project in Lisbon possible.

S.E.P. conceived the project, carried out lab experiments, interpreted data, and wrote the manuscript. S.M. advised on experimental design and lab experiments in Portugal. E.R.F., R.S., and J.S.M. helped with study conception and experimental design. S.M., E.R.F., R.S., and J.S.M. revised the manuscript.

Ethical approval was not needed for this study because it did not involve research with human subjects, and all SARS-CoV-2 case data are publicly available online.

Research reported in this publication was supported by the National Institute of Environmental Health Sciences of the National Institutes of Health under award number T32ES015459. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Research activities in Portugal were supported by Programa Operacional de Competitividade e Internacionalização (POCI) (FEDER component), Programa Operacional Regional de Lisboa, and Programa Operacional Regional do Norte (Project COVIDETECT, ref. 048467), Fundação para a Ciencia e Tecnologia (FCT), through the Joint Programming Initiative on Antimicrobial Resistance (JPIAMR) program, project Surveillance of Emerging Pathogens and Antibiotic Resistances in Aquatic Ecosystems (SARA), grant number Aquatic/0006/2020, and Fundação para a Ciência e Tecnologia (FCT) through the Programa ‘Testar com Ciência e Solidariedade’ – COVID-19, project CAPTURA – Uso de partículas funcionalizadas para enriquecimento e detecção eficiente de SARS-CoV-2 em amostras clínicas e ambientais, grant number SAICTCOVID/72536/2020 da FCT.

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

The authors declare there is no conflict.