Spatially resolved qualified sewage spot sampling to track SARS-CoV-2 dynamics in Munich - One year of experience

Abstract

Wastewater-based epidemiology (WBE) is a tool now increasingly proposed to monitor the SARS-CoV-2 burden in populations without the need for individual mass testing. It is especially interesting in metropolitan areas where spread can be very fast, and proper sewage systems are available for sampling with short flow times and thus little decay of the virus. We started in March 2020 to set up a once-a-week qualified spot sampling protocol in six different locations in Munich carefully chosen to contain primarily wastewater of permanent residential areas, rather than industry or hospitals. We used RT-PCR and sequencing to track the spread of SARS-CoV-2 in the Munich population with temporo-spatial resolution.

The study became fully operational in mid-April 2020 and has been tracking SARS-CoV-2 RNA load weekly for one year. Sequencing of the isolated viral RNA was performed to obtain information about the presence and abundance of variants of concern in the Munich area over time.

We demonstrate that the evolution of SARS-CoV-2 RNA loads (between <7.5 and 3874/ml) in these different areas within Munich correlates well with official seven day incidence notification data (between 0.0 and 327 per 100,000) obtained from the authorities within the respective region. Wastewater viral loads predicted the dynamic of SARS-CoV-2 local incidence about 3 weeks in advance of data based on respiratory swab analyses. Aligning with multiple different point-mutations characteristic for certain variants of concern, we could demonstrate the gradual increase of variant of concern B.1.1.7 in the Munich population beginning in January 2021, weeks before it became apparent in sequencing results of swabs samples taken from patients living in Munich.

Overall, the study highlights the potential of WBE to monitor the SARS-CoV-2 pandemic, including the introduction of variants of concern in a local population.

Graphical abstract

1. Introduction

Severe acute respiratory syndrome virus 2 (SARS-CoV-2) was first described in the city of Wuhan in December 2019 and has quickly spread around the world resulting in the current pandemic situation with more than 141,754,944 confirmed Coronavirus disease 2019 (COVID-19) cases and 3,025,835 deaths as of April 20th, 2021 (World Health Organization, 2021). In order to mitigate the spread of the virus and to avoid the collapse of the health care sector, governments have resorted to different policies including strict lockdowns or even curfews. The world is still struggling to adapt to the unprecedented economic, behavioural and societal changes associated with the pandemic. On a global scale, the COVID-19 pandemic caused by the SARS-CoV-2 virus can be regarded as the most serious economic crisis since World War II (Alradhawi et al., 2020; Nicola et al., 2020; Shubber et al., 2020). Transmission of SARS-CoV-2 occurs primarily through aerosols and droplets excreted from infected individuals while breathing, sneezing or coughing. Pre- and oligosymptomatic patients are unaware of their disease and can also be infectious, therefore the spread is very difficult to control. As viruses of the family Coronaviridae are enveloped single-strand, positive-sense RNA viruses, infected individuals can be identified by reverse transcription PCR (RT-qPCR) of naso-oropharyngeal swabs. This is considered the gold standard of laboratory-based detection of SARS-CoV-2 (Oliveira et al., 2020). For population surveillance however, this approach is not feasible, due to the number of swabs and RT-PCR reactions needed. In resource limited settings, the situation is even more challenging, and mass testing is neither feasible nor affordable.

Nevertheless, governments and health authorities need data to adapt their actions to the current local epidemiology. Many countries use mandatory notification of newly diagnosed cases and the number of hospital beds occupied by COVID-19 patients to get indications for disease activity. Due to the large fraction of untested oligo-/asymptomatic courses of disease and the often late onset of complications requiring hospital treatment, actions are often taken too late and outbreaks are detected only in advanced states.

Thus, efficient surveillance systems delivering unbiased information about the local disease burden are indispensable. It was reported that SARS-CoV-2 can be found in the feces and urine of symptomatic and asymptomatic infected subjects (Jafferali et al., 2021). Subsequently, RT-PCR from wastewater was successfully performed demonstrating SARS-CoV-2 RNA in sewage systems (Kitajima et al., 2020; Kitamura et al., 2021; Larsen and Wigginton, 2020). Even infections related to direct contact with human excretions have been described and warrant further investigation (Liu et al., 2020). This is certainly a less prominent transmission path compared to the airborne route, and the duration of infectivity of the virus in sewage water is so far unclear. Still, wastewater-based epidemiology (WBE) could be a potentially useful complementary tool for the environmental surveillance of local SARS-CoV-2 outbreaks (Aguiar-Oliveira et al., 2020; Chan et al., 2020; Gonzalez et al., 2020; Wang et al., 2020). Wastewater-based surveillance overcomes the need to test a large proportion of the population avoiding the biases of other epidemiological indicators, but still tracking the infectious agents in communities (Aguiar-Oliveira et al., 2020; Wang et al., 2020; Contreras et al., 2021). Recently, several reports have demonstrated a significant correlation between SARS-CoV-2 RNA concentration from wastewater samples and actual prevalence of SARS-CoV-2 infections over a defined period of time, complementing conventional screening and notification approaches (Shubber et al., 2020; Kitamura et al., 2021; Wang et al., 2020; Peccia et al., 2020; Tanhaei et al., 2021). We started in March 2020 to investigate the feasibility of WBE-approaches for SARS-CoV-2 in Munich, Bavaria, southern Germany. We chose to perform qualified spot sampling of sewage in the morning once weekly in six different positions in the Munich sewage system, covering close to one third of the population of the city (504,807/1,560,042 inhabitants; 32.4%). Emphasis was placed on short flow times from sink to sampling and drainage areas including primarily residential areas without larger hospitals, industrial complexes or significant foreign water influx.

To keep the protocol convenient and cost efficient, filtration with 1 mm stainless steel strainer and subsequent ultracentrifugation was used to concentrate SARS-CoV-2 for RNA isolation. Viral load measurements were performed with RT-qPCR and digital droplet RT-PCR (dd RT-PCR) against two independent target genes. A subset of RNA eluates was also sequenced after PCR amplification using the ARTIC-protocol primer sets (ARTIC Network, 2020), to analyse viral genomic information.

2. Materials and methods

2.1. Wastewater sampling

Untreated wastewater (sewage) samples were collected weekly since mid-April 2020 from six different locations of the sewage system in Munich, Germany. The drainage areas were selected from geographically distinct regions within the city (Fig. 1 ). Maximum flow times from sink to sampling were selected to be 5 h. The drainage area should not include industrial complexes, and common foreign water influx should be below 20%. There should also be no significant collateral sewage draining pipes. Five of the six regions (Numbers 1, 3, 4, 5, 6 in Table 1 and Fig. 1) were chosen to be residential, without significant influx from hospitals. The sixth selected region (Number 2 in Table 1 and Fig. 1) corresponded to an area with relatively few permanent inhabitants, but comprising the LMU university hospital at the Campus Grosshadern, the largest tertiary care facility of the region. Qualified spot sampling was performed during the morning flush surge. Sewage pH and temperature were measured in situ and the samples were consecutively transported to the laboratory. Care was taken to cool the samples to 4–6 °C in a cooler box immediately. Water was transported in 500 ml cleaned and autoclaved amber glass bottles. Most samples were processed immediately upon arrival in the laboratory. Some samples had to be stored at −80 °C until further analysis. Comparison between fresh and −80 °C stored samples was performed to evaluate possible RNA degradation and signal losses due to the freeze-thaw cycle.

Fig. 1.

City map of Munich with the sampling areas highlighted. Sampled neighbourhoods are spread across the city and include about 1/3rd of the total population. The small region 2 includes the largest university clinics of Munich. For details of the respective sampling areas see Table 1; Geographic North is indicated by the arrow; size bar represents 10 km.

Table 1.

Basic characteristics of the six drainage areas chosen for the study. In area 2 (*) the university hospital Grosshadern is locally connected besides relatively few permanent residents. To the right of the table, there are calculations of the total number of subjects newly infected per week in each respective area based on the weekly incidence rate of 25 and 100/100,000 inhabitants.

Nr.	Drainage area	Permanent inhabitants	Maximum flow [L/s]	Drainage area size [ha]	Maximum time from sink to sampling [h]	Infected subjects at 7 day incidence of 25/100,000	Infected subjects at 7 day incidence of 100/100,000
1	Langwieder Bach	9471	30	200	2.5	2.4	9.5
2	Großhadern*	6781	60	85	1	1.7	6.8
3	Schmidbartlanger	66,914	180	670	5	16.7	66.9
4	Schenkendorfstr.	118,304	580	1050	5	29.6	118.3
5	Gyßlingstr.	157,876	630	1150	5	39.5	157.9
6	Savitsstr.	145,461	870	2300	4	36.4	145.4
Sum		504,807	2350	5455	NA	NA	NA

2.2. Virus concentration and nucleic acid extraction

Upon arrival in the laboratory, sewage was drained through a 1 mm stainless steel strainer and collected in disposable 50 ml centrifuge tubes (Corning 430829). For direct nucleic acid extraction after the sewage collection, one 50 ml tube from each location was centrifuged with 3000g at 4 °C for 20 min to pellet the debris, the other sewage samples were immediately stored at −80 °C for further analysis. 38 ml of the debris-free sewage was transferred to high-speed centrifuge tubes (Nalgene 3138-0050). Samples were then centrifuged with 26,000g at 4 °C for 1 h. Pellets were resuspended in 200 μl nuclease-free water (Ambion A9937/VWR 436912C). For nucleic acid extraction from frozen samples, two 50 ml centrifuge tubes from each location were placed at room temperature for about 3 h to slowly thaw. Afterwards, sample debris was pelleted as mentioned above. 76 ml (twice 38 ml) of the debris-free sewage was transferred to high-speed centrifuge tubes for ultracentrifugation. Pellets from the same location were homogenized and re-suspended in a total of 200 μl nuclease-free water. In direct comparison, using double the volume corrected for freeze-thaw losses encountered. An internal positive control was added to the concentrate before extraction to verify the efficacy of the RNA isolation. As a negative control, clear tap water was treated in parallel throughout the process. Due to severe delays in the delivery of the AllPrep PowerViral DNA/RNA Kit (Qiagen), we also used the RNeasy PowerMicrobiome Kit (Qiagen) which contains the same buffer solutions also found in the other kit. The performance of the two kits was compared and found to be similar, when RNA isolation was performed following the manufacturer's recommendations for the AllPrep PowerViral DNA/RNA Kit for both kits. RNA was eluted in 50 μl and reloaded in the spin column to increase sample concentration.

2.3. RT-PCR and DD-PCR

SARS-CoV-2 RNA was detected using the nucleocapsid (N1) primer/probe combination described in the CDC protocol (Centers for Disease Control and Prevention, 2021). PCR reactions were performed using the one-step QuantiNova Multiplex RT-PCR Kit on a Roche LightCycler 480 II as described in more detail previously (Muenchhoff et al., 2020). For quantification, standard curves were generated in multiple replicates using a commercially available standard for calibration containing the nucleocapsid gene with defined copy numbers (2019-nCoV-N-PositiveControl, IDT). The lower limit of detection (LOD) of the PCR reaction was extrapolated in a probit regression analysis using the same standard as described previously (Tsai et al., 2020). Viral loads of sewage samples are calculated as N1 copy numbers per 100 ml of sewage. Additionally, samples were quantified using digital droplet PCR as described previously (Muenchhoff et al., 2020).

2.4. Sequencing and bioinformatic analyses

Amplicon pools spanning the SARS-CoV-2 genome were prepared for a subset of samples, converted to barcoded sequencing libraries and sequenced on an Illumina HiSeq1500 following the ARTIC network nCoV-2019 sequencing protocol v2 (Protocols .io, 2020). Sequenced reads were demultiplexed and mapped to the SARS-CoV-2 reference genome (NC 045512.2) with bwa-mem (HJae-p, 2013). The sequenced amplicons were assembled using the iVar package (Grubaugh et al., 2019). Briefly, the package trims the primer sequences from the mapped reads, filters them by a base quality >20 and a minimal read length of 30 nt. Pileup files were generated using samtools and used for consensus sequence generation within the iVar package setting a minimum frequency threshold of 90% (-t 0.9) and a minimal read depth of 20 (-m 20). SNPs and Indels were called from the mapped reads with freebayes (Garrison and GJae-p, 2012) using a ploid of 1 (-p 1). To address potential issues of contamination, negative controls (PCR-grade water) were included in all sequencing runs. The consensus sequences and metadata for the samples were uploaded to the GISAID repository.

2.5. Data analysis

For analyses and visualization, we used the software R, version 4.0.2. We analysed the relationship between two time series (x _t = viralload and y _t = incidencerate), where the series may be related to past lags of the x-series. We used the sample cross correlation function (CCF), a simple and useful exploratory tool for identifying lags of the x-variable that might be useful predictors of y _t. We could identify regions where the most dominant cross correlations occur. Assuming independent repetitions of independent variables x _t and y _t, the standard deviation of a sample correlation coefficient is approximately the classical value, and approximates the 5% significance limits. The mutation heatmaps were generated with the R package pheatmap (https://CRAN.R-project.org/package=pheatmap (2012)) with variant frequencies obtained from freebayes.

3. Results and discussion

Real-time monitoring of how SARS-CoV-2 spreads in the population is challenging. One of the biggest challenges is that the majority of SARS-CoV-2-infected subjects are oligo/asymptomatic (Huang et al., 2021). The symptom severity also varies greatly depending on the demographics of the affected population, making it very hard to evaluate the current disease burden. Sewage is a source of information on human health and habits that is not affected as much by the clinical variation of disease courses, because in human excreta viral particles and -RNA are shed, more or less regardless of their symptomatology status. Thus, tracking the sewage system might provide complementary information to the clinical testing, offering near-real-time outbreak data. In this study, we evaluated if weekly qualified spot sampling of the sewage system of six neighbourhoods in Munich is sufficient to detect and track the local distribution of SARS-CoV-2. The population of Munich was over 1.5 million as of the most recent dataset from December 2019. We probed the sewage of roughly one third of the population within this study, spread across the city (Fig. 1, Table 1). To investigate the sewage of hospitals, we used one area (number 2) which contains to a large fraction the wastewater from the LMU university hospital in Grosshadern. The six chosen regions in Munich detailed in Table 1 and Fig. 1 align with the aforementioned criteria (see Section 2.1) to ensure that the concentration of virus RNA is representative for residential areas.

SARS-CoV-2 RNA copy numbers in sewage align well with the notification numbers in all individual regions (Fig. 2). Variation of notification numbers and the SARS-CoV-2 RNA-load in the sewage over time are considerable. Thus, a cumulative incidence for the seven days prior to the wastewater sampling was calculated, removing most artefacts due to delayed notifications or closed screening facilities/holidays. The fitted chronological sequences of wastewater SARS-CoV-2 RNA concentrations were plotted with the cumulative seven-day incidences of the respective areas including 95% confidence intervals (Fig. 2). It can be appreciated that the sewage viral load precedes the notification numbers by roughly three weeks (Fig. 3, Supplemental Fig. 1A and B). Other reports have identified intervals between 4 days and three weeks, which is in line with our observations (Karthikeyan et al., 2021; Wu et al., 2020).

Fig. 2.

Incidence rates (cumulative seven day incidence per 100,000 inhabitants) over the last year for the drainage areas under investigation (areas 1–6, black). Copy numbers of the SARS-CoV-2 N1-gene target (CDC protocol) are indicated in red, expressed as the number of copies per 100 ml of sewage. The lower limit of detection (LOD) of the PCR reaction (equivalent to 7.58 copies per 100 ml sewage) was used to plot negative samples. The solid line represents the LOESS (locally estimated scatterplot smoothing or local regression) modelling the incidence rates (in black) and the viral loads (in red). The grey regions represent the 95% CI of the LOESS estimate. The viral load in sewage precedes the incidence based on notifications by roughly 3 weeks. For reference the cumulative seven day incidence per 100,000 inhabitants for the same timeframe for the whole of Munich (MUC) and for the non-sampled areas (MUC-R), interpolated with a LOESS. The dashed vertical line separates 2020 from 2021. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3.

Cross correlation values, for identifying lag time (in weeks) of the viral load-variable predictors of incidence rate. Regions where the most dominant cross correlations occur above or below horizontal lines. A lag of roughly three weeks (dashed vertical line) is observed, between sewage viral load and notification data. The solid line represents the LOESS (locally estimated scatterplot smoothing or local regression) modelling of the cross correlation values over the different sampling areas. The grey regions represent the 95% CI of the LOESS estimate. See Supplemental Fig. 1B for specific sampling site information.

Cross correlation values, for identifying lag time (in weeks) of the viral load-variable predictors of incidence rate. Regions where the most dominant cross correlations occur above or below horizontal lines. A lag of roughly three weeks (dashed vertical line) is observed, between sewage viral load and notification data. The solid line represents the LOESS (locally estimated scatterplot smoothing or local regression) modelling of the cross correlation values over the different sampling areas. The grey regions represent the 95% CI of the LOESS estimate. See Supplemental Fig. 1B for specific sampling site information.

Of note, area 2 (Grosshadern) includes the university hospital. Here, notification numbers regarding SARS-CoV-2 incidence are artificially too high, hence the marked difference between the seven-day incidence and the viral load in the respective sewage which was not observed in the other regions. Many COVID-19 patients in the hospital are treated as inpatients in the facility for longer durations, which are not reflected in the notification numbers covering only new diagnoses. On the other hand, many subjects will be diagnosed with mild SARS-CoV-2 and subsequently sent home. Only severe courses are admitted. In these cases, many of the patients' excretions will be disposed of in the waste rather than the sewage due to the use of diapers and similar products. Positively tested personnel and contacts are also quarantined at home and therefore do not contribute to viral load in the hospital's sewage. All this results in surprisingly low concentrations of SARS-CoV-2 RNA detected in sewage from area 2. There is no special waste water treatment within the hospital for regular sewage of wards or intensive care units including SARS-CoV-2 units, the only specialized facility relates to radioactive waste of patients receiving radiotherapy and was not used for SARS-CoV-2 patients.

Besides the quantification of the genomic equivalents of SARS-CoV-2 in the wastewater samples, we performed sequencing to obtain information about the prevalence of variants of concern within the population. We highlighted exemplarily the dynamic over time of a subset of signature mutations characteristic for the variant of concern B1.1.7 in drainage area 6 (Fig. 4 A). For B.1.1.7, we detected sustained levels of key mutations in wastewater beginning around week 3 of 2021 and reached an average representation of about 60% around week 7. Similar proportions were only reached in sequenced patients' swabs around week 10 (Fig. 4B). It can be easily appreciated that the sewage sequencing was able to predict the subsequent increase of B.1.1.7 in the population, similarly to what was observed with case numbers in our study and others (Karthikeyan et al., 2021; Wu et al., 2020).

Fig. 4.

A: Selected signature mutations for the variant of concern lineage B.1.1.7 are exemplarily shown over time in a heatmap for the sampling site 6 (Savitsstr.) between July 2020 and March 2021. Percentages (starting at 0) represent the fraction of the indicated single nucleotide polymorphisms detected at the respective time point. Grey fields (NA) depict no coverage (less than 20 mapped reads) at the respective genome position in the sample. No sustained signature mutation signals were detected before mid of January (19th). Subsequently, the proportion of key mutations increased rapidly and reached high levels by the beginning of March 2021. Signature mutations for P1 and B.1.351 were not detected over the study period. B: Baggtitr chart of reported percentage of B.1.1.7 in sequenced SARS-CoV-2 swabs between calendar weeks 1 and 9 of 2021 in Munich. Dots represent frequencies of B.1.1.7 mutations detected in sewage as plotted in A. The solid line represents the LOESS (locally estimated scatterplot smoothing or local regression) modelling the different signature mutations. The grey region represents the 95% CI of the LOESS estimate. Blue bars represent confirmed sequenced B.1.1.7 cases, black and yellow represent B.1.351 and P1 respectively. Red represents other variants than the three aforementioned variants of concern; green bar is identification of S1 mutants by hybridization assays without definitive confirmation by sequencing. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.1. Limitations/challenges

Munich received a sewage system elaborate for the time, by the end of the 19th century, vast parts of which are still in use today. This system includes many collateral connections which ensure functionality in case of blockage or regional overload. Further, the sewage layout is based on gravity flow from south to north of the city due to a height difference of about 90 m, without the general use of pumps. This is associated with several challenges for the detection of SARS-CoV-2 in the wastewater. First, it is difficult to assess the drainage area of a certain sampling point if collaterals exist, thus, we chose drainage areas which can be clearly characterized. Second, gravity flow based sewage is less homogenized and also slower regarding flow rates, due to the general absence of pumps. Third, not the whole sewage system is separated in surface water carrying pipes and fecal/sewage systems, leading to dilution artefacts and changes in pH and composition, potentially altering the concentration and decay characteristics of the virus in the samples taken.

One of the main limitations of the study is the use of qualified spot sampling. This leads to larger variation in the data. One of the main reasons thereby is the lack of homogenization prior to sampling, and the fact that the sample is drawn only during a short time period in the morning once a week and there may be different factors influencing one specific sample. On the other hand, inhabitants are also more likely to be home during the night and morning and go to work thereafter despite increasing rates of home office work recently due to the pandemic. Therefore, sampling in the morning might also give an advantage over integrated sampling over the whole day. The latter applied to residential areas might bias the data towards those inhabitants quarantined for known disease or being first degree contacts as well as white collar workers able to work from home. The practical influence of these factors is currently unknown and thus cannot be controlled for, they are also vastly ignored in most works on the topic currently. We have also omitted to try and calculate back the number of infected persons in each area, as the variation is too large using the spot sampling protocol applied here.

Another effect of the sample variation and the population size in a respective drainage area is a varying limit of detection for SARS-CoV-2 RNA. As the sewage is not sampled over the whole day and homogenized, averaging effects are achieved either by larger numbers of patients in the drainage area, or more measurements by PCR. We were also restricted to six sampling sites within the city due to capacity limitations and the layout of the sewage system as previously detailed. Other sites would have incorporated larger hospitals, industry or significant fractions of foreign water, obscuring proper analysis. Due to prolonged flow times of the wastewater resulting in significant decay of the virus particles in the wastewater until the wastewater treatment plant (WWTP) is reached, a systematic bias towards regions closer to the WWTP would be expected if samples are drawn there. Thus, samples must be obtained in the areas where the wastewater is produced. Therefore, we chose sampling sites with a maximum of 5 h of flow time from sink to sampling, also restricting the possible sampling sites. Due to the influence of rain on the sewage system with spill over of surface water, not all time points could be analysed. Precipitation amounts of <5 mm do not lead to significant changes in sewage water composition to hamper the analysis performed in this study. During the one year study duration, only 4 time points were analysed within the study where >5 mm of precipitation in the 24 h preceding the sample taking was recorded, as 2020 was a very dry year overall in Munich. This was June 29th, 2020 (5.1 mm), October 27th, 2020 (16.7 mm), January 26th, 2021 (9.9 mm) and March 16th, 2021 (5.8 mm). Thereby, on January 26th, 2021 and March 16th, 2021 the 12 h before sample taking were 3 mm of precipitation or less. We performed the modelling analysis including (presented above) and excluding these four time points, and there was no change.

There is also a notification bias, which is caused by the dark field of unreported cases in the population. This might obscure the correlation between notification numbers and virus RNA load in sewage. Due to the seroprevalence and incidence studies performed in parallel in the department, we were able to estimate the dark field as well (Pritsch et al., 2021; Radon et al., 2020). Overall, we expect the real case numbers to be about 4 times higher than the reported incidences at the beginning of the spring of 2020 wave, decreasing over time. It is as low as roughly a factor 2 in the winter 2020/21 wave due to increased screening efforts. These dynamic effects cannot be fully taken into account in the analysis; however, wastewater concentrations are not affected by the notification dark field.

4. Conclusion

This study is one of the first and longest studies to follow SARS-CoV-2 RNA loads in wastewater over time in Germany (Agrawal et al., 2021; Westhaus et al., 2021). With a start in April 2020, the last part of the first SARS-CoV-2 wave of 2020 is covered as is the winter wave 20/21. The data demonstrates that even a simple qualified spot sample taken once weekly was able to foreshadow the SARS-CoV-2 incidence derived from notification data by 3 weeks. This study also demonstrates that sequencing of RNA isolated from raw sewage is a reliable tool to detect the introduction of SARS-CoV-2 variants of concern in the local population weeks prior to their appearance in relevant numbers in clinical or screening swab samples. More frequent sampling as well as homogenization of the sewage would be desirable to decrease variability and increase data quality.

The following are the supplementary data related to this article.

Supplementary Fig. 2

A: Selected signature mutations over time in a heatmap for the sampling site 1 (Langwieder Bach). B: Baggtitr chart of reported percentage of mutations in sequenced SARS-CoV-2 swabs between weeks 1 and 9 of 2021 in Munich. Dots represent frequencies of mutations detected in sewage as plotted in A.

Supplementary Fig. 3

A: Selected signature mutations over time in a heatmap for the sampling site 2 (Grosshadern). B: Baggtitr chart of reported percentage of mutations in sequenced SARS-CoV-2 swabs between weeks 1 and 9 of 2021 in Munich. Dots represent frequencies of mutations detected in sewage as plotted in A.

Supplementary Fig. 4

A: Selected signature mutations over time in a heatmap for the sampling site 3 (Schmidbartlanger). B: Baggtitr chart of reported percentage of mutations in sequenced SARS-CoV-2 swabs between weeks 1 and 9 of 2021 in Munich. Dots represent frequencies of mutations detected in sewage as plotted in A.

Supplementary Fig. 5

A: Selected signature mutations over time in a heatmap for the sampling site 4 (Schenkendorfstr.). B: Baggtitr chart of reported percentage of mutations in sequenced SARS-CoV-2 swabs between weeks 1 and 9 of 2021 in Munich. Dots represent frequencies of mutations detected in sewage as plotted in A.

Supplementary Fig. 6

A: Selected signature mutations over time in a heatmap for the sampling site 5 (Gyßlingstr.). B: Baggtitr chart of reported percentage of mutations in sequenced SARS-CoV-2 swabs between weeks 1 and 9 of 2021 in Munich. Dots represent frequencies of mutations detected in sewage as plotted in A.

CRediT authorship contribution statement

AW designed the study with great support from MSR, BB, MH; RRA, JB, IP, AG, HB, SK, AO, AW and MM established and performed laboratory analyses and sequencing. MM, MMB, AG, ZNK, MSR, MH, KR, BS, CN, BB, NC, AW performed data analysis and interpretation. RRA, JB, ZNK, NC and AW wrote the manuscript. All authors corrected and approved the manuscript. KR, OTK and MH provided funding and critically revised the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Editor: Frederic Coulon