High-resolution within-sewer SARS-CoV-2 surveillance facilitates informed intervention

To assist in the COVID-19 public health guidance on a college campus, daily composite wastewater samples were withdrawn at 20 manhole locations across the University of Colorado Boulder campus. Low-cost autosamplers were fabricated in-house to enable an economical approach to this distributed study. These sample stations operated from August 25th until November 23rd during the fall 2020 semester, with 1512 samples collected. The concentration of SARS-CoV-2 in each sample was quantified through two comparative reverse transcription quantitative polymerase chain reactions (RT-qPCRs). These methods were distinct in the utilization of technical replicates and normalization to an endogenous control. (1) Higher temporal resolution compensates for supply chain or other constraints that prevent technical or biological replicates. (2) The data normalized by an endogenous control agreed with the raw concentration data, minimizing the utility of normalization. The raw wastewater concentration values reflected SARS-CoV-2 prevalence on campus as detected by clinical services. Overall, combining the low-cost composite sampler with a method that quantifies the SARS-CoV-2 signal within six hours enabled actionable and time-responsive data delivered to key stakeholders. With daily reporting of the findings, wastewater surveillance assisted in decision making during critical phases of the pandemic on campus, from detecting individual cases within populations ranging from 109 to 2048 individuals to monitoring the success of on-campus interventions.


Introduction
On March 11, 2020, the World Health Organization (WHO) declared coronavirus disease 2019 (COVID-19) a pandemic (WHO, 2020). As of June 15, 2021, 176 million confirmed cases have resulted in over 3.8 million deaths World Health Organization (WHO). Clinical testing of individuals is crucial for identifying infected persons, understanding infection prevalence, and containing the disease, but supply chain limitations and logistical challenges limit clinical testing capacity. Testing is therefore generally reserved for individuals either showing symptoms or likely exposed to the disease Centers for Disease Control and Prevention (CDC). However, the etiologic agent responsible for COVID-19, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has a significant asymptomatic percentage (with some estimates of 50%) (Mizumoto et al., 2020;Nishiura et al., 2020;Oran and Topol, 2020) and can be transmitted by pre-symptomatic and asymptomatic persons Hu et al., 2020;Bai et al., 2020;Wei et al., 2020;Rothe et el., 2020). Further, symptoms may take between approximately two days and two weeks (with a pooled median of 95th percentile estimates in a comprehensive meta-analysis reporting 13.1 days (Xin et al., 2021)) to develop post-infection Backer et al., 2020;Lauer et al., 2020), and even when symptomatic, individuals may not self-report. As a result, clinical testing alone fails to identify many infected individuals before they transmit the disease to others and under-represents caseload numbers utilized by officials to inform public health directives. The need to address these shortcomings with a supplementary epidemiological tool was recognized early in the pandemic with a global collaborative of researchers advocating for wastewater-based epidemiology (WBE) .
WBE efficiently and non-invasively monitors community metrics by sampling generated wastewater and screening for chemical and biological entities, with previous success demonstrated in tracking community drug use (Daughton 2001;Choi et al., 2018) and poliovirus circulation (Asghar et al., 2014;WHO 2003;Hovi et al., 2012). Wastewater networks can be sampled at points at which discharges from community members have combined, aggregating a semi-anonymous signal representative of the upstream community. Analyzing aggregated wastewater for SARS-CoV-2 RNA therefore provides an opportunity to test entire communities within a single sample. As SARS-CoV-2 RNA is present in the feces of both symptomatic (Jiehao et al., 2020;Wang et al., 2020;Xiao et al., 2020;Zhang et al., 2020a;Holshue et al., 2020;Zhang et al., 2020b;Wu et al., 2020b;Wölfel et al., 2020) and asymptomatic Lo et al., 2020;Han et al., 2020) COVID-19-infected individuals, analysis of those samples offers rapid insight into infection prevalence potentially unhindered by factors such as symptom onset and the healthcare-seeking behavior of individuals. Further, whereas this aggregated testing cannot pinpoint infected individuals, it can allow for more effective use of clinical testing resources. For example, WBE can quickly identify the regions and communities with the most infections and allow for the targeted allocation of resources to those "hotspots'' for early and comprehensive testing of symptomatic and asymptomatic persons on a localized level (Hart and Halden, 2020). Although SARS-CoV-2 RNA fecal shedding behavior has been reported as erratic (Walsh et al., 2020), wastewater represents a complex mixture of all liquid-conveyed waste exiting a premise, including urinary, respiratory, oral, and hygiene-based discharges, compositing multiple potential sources of viral RNA.
An international network of researchers has detected SARS-CoV-2 RNA in their local wastewaters (Westhaus et al., 2021;Ahmed et al., 2020;La Rosa et al., 2020;Kumar et al., 2020;Medema et al., 2020;Lodder and de Roda Husman, 2020;Randazzo et al., 2020;Wu et al., 2020a;Nemudryi et al., 2020). Their efforts establish proof of concept for WBE in the context of COVID-19 monitoring and work to validate the utility of the approach. These previous studies primarily sampled from wastewater treatment plants/water resource recovery facilities, efficient locations for obtaining population-level signals. Monitoring the upstream sewer network and sampling at the building-or microsewershed-scale, however, enhances spatial resolution and enables more targeted surveillance and response efforts. A number of building-level campaigns have been implemented in response to COVID-19, particularly at colleges, but more experience and guidance are desired to inform their further implementation (Harris-Lovett et al., 2021).
Here, we report the WBE campaign conducted at the University of Colorado Boulder (Colorado, USA). We sampled up to 20 manhole locations seven days per week between August 25th and November 23rd to monitor on-campus residential buildings for the presence of COVID-19. To obtain economical 24-h composite samples, we designed, assembled, and deployed low-cost autosamplers. We tracked the concentrations of SARS-CoV-2 RNA and control species in the wastewater using reverse transcription quantitative polymerase chain reaction (RT-qPCR) assays following best practices (Ahmed et al., 2021) and reported to campus decision makers daily. This campaign was coupled with weekly individualized saliva-monitoring RT-qPCR testing of all asymptomatic on-campus residents (Yang et al., 2021;Bjorkman et al., 2021). The comprehensiveness of both the WBE campaign and the clinical and saliva-monitoring testing services provides a unique opportunity in evaluating the utility of building-level wastewater surveillance.

Sample locations
Twenty sample locations targeting the wastewater outfall captured within surface-accessible manholes were prioritized to discriminate the SARS-CoV-2 signals originating from the on-campus residential buildings at the University of Colorado Boulder (Fig. 1a, Table 1). Twentythree pumps (autosamplers) operated at these locations, with three sites discriminating two separate flows within a single manhole. Overall, then, 23 flows from 20 manhole locations were monitored. Each flow roughly corresponded to an individual residential structure (except for the flow that originated from the singular monitored administrative building). The university housed over 6200 students in on-campus residential buildings during the semester, and each flow on average accounted for the wastewater generated by 450 residents (range extending from 109 to 2048 residents), with select residents being monitored at multiple sites. The targeted manholes ranged from approximately 1 to 7 m in depth. To protect the privacy of residents, the presented data was anonymized with a unique label assigned to each residential structure indicating its position and any other structure contributing to its associated flow in parentheses:

Composite autosampler
The composite autosampler was assembled from readily available materials. The main components were a 24-V Stenner pump (E10VXG; Stenner Pump Company, Jacksonville, FL, USA; designed by DEWCO Pumps, Denver, CO, USA), a 300-Wh portable DC/AC power bank (R300; GoLabs Inc., Carrolton, TX, USA), a 5-gal jerrycan (Uline, Pleasant Prairie, WI, USA), a 9-gal cooler, gel ice packs, insulation, ¼-in. O.D. PVC tubing, and exterior casing (Fig. 1b, Supplemental Figure 1, Supplemental Table 1). The samplers were positioned above ground and next to each manhole (Fig. 1c). The wastewater inflow and overflow tubing lines were fed through the D-pick of the manhole cover, and the inlet strainer that resided in the underground wastewater stream was constructed from either ¼-in. O.D. copper or ¼-in. O.D. steel tubing, with 0.157-in. holes drilled into the side.
To achieve daily composite samples representative of flushes that occurred throughout the entire day, the pumps (operating at around 33% of full capacity) slowly but continuously pulled and deposited wastewater into the 5-gal jerrycans. Collection of approximately 50-mL sub-samples occurred daily between 7 AM and 12 PM, with power banks replaced every 48 h and ice packs replaced daily when ambient average temperatures exceeded 4 • C. These collection and maintenance tasks each only required the pump to be off for about 10 to 30 min and thus did not significantly detract from continuous pump operation. Whereas each pump was scheduled to draw about 10 L of wastewater per day, the actual wastewater withdrawn was monitored at the time of sample collection by weighing the jerrycan mass with a luggage scale (Supplemental Figure 2, Supplemental Table 2). The following three subsamples were collected from each sampler daily: one for RNA extraction and viral detection, one for the determination of basic water quality parameters, and one backup. Samplers were only turned off for entire days for the following three events: a September blizzard, a supply chain disruption in October, and a cold weather event in October. More details on the sample collection procedure are provided in Supplemental File 1.

Sample processing
All samples were processed in the laboratory the same day as collection. Samples that could not be processed upon immediate arrival at the laboratory were temporarily stored at 4 • C. Backup and RNA extraction/viral detection samples were processed within 4 h of arrival; water quality samples were processed within 9 h of arrival. All samples were spiked with a known amount of bovine coronavirus (Bovilis® Coronavirus; Merck Animal Health, NJ, USA) serving as the internal process control (three different spike-in levels were explored during the campaign corresponding to concentrations of 2 × 10 6 , 2 × 10 7 , and 2 × 10 9 genome copies/L wastewatersee Supplemental Figure 9). Sample collection tubes were pre-loaded with Tween TM 20 detergent (Thermo Fisher) to inactivate infectious wastewater agents for worker safety. Additionally, processing workstations were cleaned with both ethanol and RNaseZap TM RNase Decontamination Solution (Thermo Fisher).

Water quality parameter samples
The pH (measured with accumet TM AB150 pH meter, Thermo Fisher) and total suspended solids (TSS) values of each sample were measured following standard methods (Supplemental Figures 3, 4; Supplemental Tables 3, 4).

. RNA extraction/viral detection samples
Samples collected for the detection of SARS-CoV-2 RNA were weighed and then centrifuged at 4500 x g for 20 min at 4 • C. Viruses were concentrated from approximately 35 mL of each sample's supernatant using ultrafiltration pipettes (CP-Select™ using Ultrafiltration PS Hollow Fiber Concentrating Pipette Tips; InnovaPrep, Drexel, MO, USA) following the manufacturer's guidelines. Concentrate eluted from the ultrafiltration pipettes was captured and weighed in pre-weighed 15-mL tubes. RNA was then extracted from the concentrate using RNA PureLink Mini Kits (Thermo Fisher) according to the manufacturer's protocol (except Wash Buffer I was not used). A 1-μL aliquot was taken from the extracted RNA of each sample and analyzed on a Qubit TM 4 Fluorometer (Q33238, Thermo Fisher) using the High Sensitivity RNA Kit to quantify the total RNA extracted and roughly assess the success of the extraction process (Supplemental Table 5).

RT-qPCR
Two separate RT-qPCR pipelines were then used to detect and quantify SARS-CoV-2 RNA. Full details fulfilling the MIQE guidelines are provided as supplemental (Supplemental File 1) (Taylor et al., 2010). In brief, the first RT-qPCR pipeline, entitled SURV1, was executed by the University of Colorado Boulder's COVID Surveillance Laboratory simultaneously with the sampling campaign. SURV1 employed a multiplex assay targeting the SARS-CoV-2 nucleocapsid (N) and envelope (E) genomic regions as well as the human RNaseP transcript (Supplemental Figures 5, 6; Supplemental Tables 6, 7) (Yang et al., 2021). From August 28th to September 29th, the N1 primer and probe set was used to detect the nucleocapsid region. After September 30th, the N2 primer and probe set was used instead because of supply-chain constraints associated with the primer/probe master mixes; similar efficiencies were noted between N1 and N2. Multiple technical replicates were not run. The second RT-qPCR pipeline, entitled SENB+, was executed in December 2020 after the fall sampling campaign had ended. In this second pipeline, the extracted RNA samples (frozen at -80 • C) were reevaluated for SARS-CoV-2 RNA using a wastewater-specific multiplex assay detecting the following targets: SARS-CoV-2 N (N2), SARS-CoV-2 E, the spiked internal control bovine coronavirus, and genogroup II F+ RNA bacteriophage (Supplemental Table 8). Genogroup II F+ RNA bacteriophage was targeted to serve as a human fecal indicator (Cole et al., 2003). To quantify the SENB+ data, standard curves were established for each run from ten-fold serial dilutions ranging from 10 5 , 10 6 , and 10 6 copies to 1, 10, and 10 copies of SARS-CoV-2, bovine coronavirus, and F+ bacteriophage standard per reaction, respectively (Supplemental Figure 7; Supplemental Tables 9, 10). The lower standard amount established the limit of quantification (LOQ). The limit of detection (LOD) was set at amplification occurring before the 40th cycle and above the background amplifications in the extraction blank (Supplemental Figure 8) and negative control reactions. Select runs included an additional standard dilution containing 1.53 × 10 7 copies of bovine coronavirus standard per reaction, assisting quantification when the bovine coronavirus spike-in amount was increased from approximately 50,000 to 500,000 copies per reaction (Supplemental Figure 9). SARS-CoV-2 E and N2, bovine coronavirus, and F+ bacteriophage amplicons recovered from wastewater collected from the primary isolation building on September 17th, 2020 were subjected to Sanger sequencing at the University of Colorado Anschutz Molecular Biology Core to confirm that the expected sequences were amplified; the recovered amplicons matched the expected targeted sequences. Overall, detailed protocols are provided in Supplemental File 1 for both the SURV1 and SENB+ pipelines.

Data normalization
SARS-CoV-2 data from SURV1 was normalized by subtracting (because these values are logarithmic in nature) the RNaseP Ct value from the SARS-CoV-2 E Ct value, a comparative Ct method common in diagnostic PCR assays (Dahdouh et al., 2021;Rowan et al., 2021). The N gene was utilized to confirm trends. Data from SENB+ was processed by calculating copies per liter of wastewater using the recorded masses of sample collected and concentrated (Supplemental Table 11) and the following equation: The bovine coronavirus and F+ bacteriophage signals were used to track sample variability but not to transform the concentration of SARS-CoV-2 RNA. The bovine coronavirus recovery efficiency was determined by comparing the RT-qPCR-obtained concentration value to the extracted-RNA concentration of the spiked-in control. Throughout the campaign, the recovery efficiency averaged 53 ± 30% S.D. October 2nd samples are masked from this analysis because they were frozen prior to extraction and a key intermediate weight was not recorded.

Incorporation of medical services and isolation space utilization data
On-campus medical services in Wardenburg Health performed nasalswab Lyra® Direct SARS-CoV-2 assays (Quidel Corporation, San Diego, CA, USA) to confirm suspected cases within the community. Individuals tested for suspected infection included symptomatic students, students identified through contact tracing, and students presumed positive based on their weekly saliva-monitoring RT-qPCR test results. The affirmative medical services data are considered as "positive detections" within the residential structures, and the date of each positive is used to denote the case (though that date is not the date of actual infection) (Supplemental Table 12). Isolation space utilization data tracked the number of beds in designated isolation spaces occupied on a given day (Supplemental Table 13).

Performance of the composite samplers
In general, the composite samplers performed well, reliably withdrawing sample mass ( Table 1). The design achieved the objectives and provided an economical sampling unit. Additionally, if a source of electricity is near the sample point, then the cost decreases with removing the necessity of the power bank. Throughout the campaign, concerns were noted over leakage through the small sampling port on the jerrycan and the inlet strainer either clogging or being knocked offline because of toilet paper accumulation during low-flow conditions. To prevent further leakage, a short PVC tube was epoxied to the small sampling port and positioned such that the free end of the tube sat (with a removable cap) above the jerrycan. Several redesigns of the inlet strainer suffered similar issues as the primary design, exacerbated by the increasing prevalence of "flushable" wipes. These issues occasionally disabled continuous wastewater collection at a site for numerous hours until a team member could manually correct them; unclogging and redeploying the inlet strainers remained the primary maintenance demand. Future surveillance campaigns may consider more permanent modifications to the flow path to enable ease of sample collection. While inlet strainer issues contributed to variability in the daily withdrawn wastewater (Supplemental Table 2), we do not believe they impaired our results. Clogging was uncommon among the collective set of sites, and a similar WBE campaign that collected grab samples representative of a single time point rather than composite samples successfully monitored the presence of SARS-CoV-2 (Betancourt et al., 2021). Additionally, some of the variability in the withdrawn amounts is attributable not to these inlet issues but to air bubbles getting pumped with wastewater during periods of low flow.

Dataset summary
Prior to resumption of on-campus activities, incoming on-campus residents were required to test five days prior to the scheduled movein (August 17th-21st), establishing the baseline. An initial increase in SARS-CoV-2 RNA wastewater concentrations was detected near the beginning of the campaign (Fig. 2). This event occurred two weeks after the Labor Day holiday in the USA, with many traced large off-campus gatherings. The wastewater concentrations plateaued the week of September 15th and were in decline prior to Boulder County enacting aggressive social distancing policies on September 24th (Fig. 2a). That concentrations were already decreasing before these policies were enacted likely reflects the success of on-campus testing, tracing, and isolation efforts. The September 24th orders were enforced until October 13th and prohibited (1) anyone aged 18 to 22 years old in the City of Boulder from engaging in gatherings and (2) residents in 36 nearby offcampus buildings from leaving their place of residence to the maximum extent possible ("stay-at-home" order) (Boulder County Public Health, 2020). Well after the expiration of those public health orders, another increase in wastewater concentrations was detected after October 31st (the Halloween holiday in the USA). Clinical services detected fewer cases on-campus during this event (119 individuals from Halloween until the end of monitoring) as compared to the September event (530 individuals up until October 1st), and this lower caseload was reflected in the nearly halved average SARS-CoV-2 wastewater concentration (3.6 × 10 6 ± 7.2 × 10 5 compared to 6.6 × 10 6 ± 5.6 × 10 5 SARS-CoV-2 copies/L). The similar dynamics emphasize a quantitative relationship, not simply a presence or absence of viral RNA correlation, between the SARS-CoV-2 caseload and the wastewater concentrations. Finally, students vacated campus prior to November 23rd for the scheduled end of in-person instruction. The SARS-CoV-2 wastewater concentration averages taken over the monitoring campaign (Fig. 2c) largely reflect on-campus prevalence (Fig. 2g) (number of reported infections within a residential structure divided by the initial census data, Fig. 2e) when masking data from wastewater samplers that were activated later in the semester.
Both the SURV1 data and the SENB+ data reflected the medical services data throughout the campaign (Fig. 2b, d, f, h; Supplemental  Tables 7, 10, 12). Overall, the E data from the SURV1 and SENB+ pipelines are consistent when SURV1's E to RNaseP ratio is compared to SENB+'s log 10 SARS-CoV-2 copies/L in a linear model (lm function in R); the slope of the linear model is 0.17 with an intercept of 4.88 and a residual standard error of 0.63, a multiple R 2 of 0.37, and a p-value < 2.2 × 10 − 16 . This result suggests that a single technical replicate (the SURV1 dataset) is admissible when performing a daily monitoring campaign and when resources become limited due to either supply chain disruptions or rapid campaign expansions designed to meet the pace of emerging pandemics. Technical replicates are still recommended when feasible (the SENB+ dataset) to avoid false reporting. Additionally, the SARS-CoV-2 E and N2 targets displayed a linear correlation of 0.97 in the SENB+ data (Supplemental Figure 10), confirming that both are suitable to track the prevalence of the virus. The quantitative range of the predicted concentrations (in terms of genome copies per liter of wastewater) was also similar for both targets. We still recommend monitoring multiple targets of different genes encoded in the genome of SARS-CoV-2 to ensure that mutations within viral variants detectable in wastewater (Crits-Christoph et al., 2021) do not immediately produce an overall false negative by interfering with primer or probe sites.
The reported medical services data is mirrored in the SARS-CoV-2 wastewater concentrations detected on the same day, likely reflecting the high frequency with which both wastewater sampling (daily) and compulsory individualized saliva-monitoring testing (weekly) were executed, but these datasets are not perfectly matched. To analyze the co-occurrence of an individual's positive test result with the detection of a high SARS-CoV-2 concentration in their residential structure's wastewater on either the same day or in near temporal proximity, we considered the 16 wastewater samplers not impacted by flow originating from isolating individuals. A total of 1103 wastewater samples were collected from those 16 samplers. Over the course of the campaign, 556 individuals tested positive within the residential structures served by those samplers. For 547 of those positive tests, the individual's test result was reported the same day a wastewater sample was collected for their residential structure (corresponding to 246 wastewater samples). Notably, the wastewater signal from structure H is duplicated in the I(H) signal; to account for this duplication, we increase the considered numbers by the number of duplicated events (i.e., the total number of positive individuals from structure H, 22, and the number of wastewater samples collected from structure I(H) over the days those individuals' positive tests were reported, 15). Therefore, 261 of the 1103 wastewater samples represent structures with a corresponding 569 positive population equivalents. When setting the wastewater concentration that must be exceeded to indicate contribution by an infected individual at 10 5 , 10 6 , and 10 7 SARS-CoV-2 copies/L, a total of 521, 450, and 146 expected cases (92%, 79%, and 26% of all cases), respectively, are effectively captured in the wastewater sample on the day of positive testing. However, the specificity of using the day-of 10 5 SARS-CoV-2 copies/L as a cutoff is low, with 741 of all 1,103 (67%) wastewater samples exceeding this threshold, and only 229 of these samples directly mapped to a positive. This exceedance results from individuals shedding SARS-CoV-2 RNA prior to testing positive and individuals requiring time post-identification to relocate from the residential building into the isolation space.
Expanding the timeframe considered around a positive detection in the medical services data may lead to better association with the SARS-CoV-2 wastewater concentrations. When considering a time window of five total days surrounding the date of positive testing (-48 to 72 h), 67% and 92% of the samples exceeding 10 5 and 10 7 SARS-CoV-2 copies/L, respectively, correspond to a detected medical services positive (Supplemental Table 14). In total, 600 of the 1103 considered wastewater samples were collected within a positive's 5-day time window (92 of these samples with SARS-CoV-2 signals below LOQ), 468 samples were not (214 of these samples with SARS-CoV-2 signals below LOQ), and 35 were excluded due to an overlap with the termination of the sampling period. The average wastewater concentrations for the time windows associated with a positive and the time windows not associated with a positive were 7.0 × 10 6 ± 1.5 × 10 6 and 1.2 × 10 6 ± 3.3 × 10 5 SARS-CoV-2 copies/L (± 95% confidence interval), respectively, with a significant Kruskal-Wallis chi-squared difference equal to 77.8 (p-value = 2.2 × 10 − 16 ). This result supports the hypothesis that samples taken temporally nearer to an infection will display a higher concentration of SARS-CoV-2 copies/L wastewater than samples taken temporally farther from an infection. However, only 12 of the considered 16 samplers display a significant difference between their respective means (timewindows-associated-with-a-positive mean versus time-windows-notassociated-with-a-positive mean) when considered individually, with samplers D1, D2, H, and M displaying no significant difference (Supplemental Figure 11). Over the campaign, 14, 13, 22, and 61 infections were detected in the populations served by these four samplers, respectively. The D1, D2, and H samplers were established after the main peak of September cases (October 15th, November 1st, and September 16th, respectively), resulting in a lower overall infection incidence and, in the case of D2, relatively distributed caseloads in which 19 of the 21 possible wastewater datapoints fall within the -48-to-72-h time window of a detected positive.
When relating the concentration of SARS-CoV-2 RNA in wastewater to the medical services data, the isolation strategy enabled further estimation of the SARS-CoV-2 shed by infected individuals into wastewater. Infected individuals received temporary housing in the primary isolation building up until September 18th and after October 6th and were assigned to a secondary building in the interim (Fig. 3). Two residential structures (A and B) are apartment complexes which enabled infected residents to instead isolate in place. Isolation in these alternate structures complicated the signal in their associated wastewater flows as well as in the combined G(FEDCBA) flow (the E2(CBA) flow is not noted given that sampler began operating after October 6th). It is further noteworthy that such combined flows bias the median data (Fig. 2a), with the contribution of a single infection potentially detected at multiple sites. These complications emphasize the importance of quantifying the signal for these locations rather than relying on binary presence/absence of virus determinations. Quantification enables the detection of temporal trends such as increasing SARS-CoV-2 RNA concentrations above the expected baseline. Additionally, students do not proceed through the entire course of infection profiled within a given residence and shift their contribution to the isolation structures (Fig. 3).
The isolation buildings maintained a unique wastewater in which the flow did not represent a fluctuating proportion of infected individuals amongst non-infected individuals. These buildings were instead occupied entirely by individuals progressing through the course of the viral infection, remaining empty otherwise. The wastewater concentrations from the primary isolation building peaked in mid-September (4.6 × 10 7 SARS-CoV-2 copies/L on September 11th) and again in mid-November (2.7 × 10 7 SARS-CoV-2 copies/L on November 16th; Fig. 3). These peaks resulted from the co-occurrence of disease progression and virus/ viral RNA shedding in stool, explaining the two peaks' nearly identical wastewater concentrations despite different numbers of infected residents (75 and 25, respectively). In this building, the viral wastewater inputs of a smaller number of infected individuals were not diluted by the inputs of a corresponding larger proportion of non-infected residents; the isolated individuals' wastewater mixed only with idle plumbing and appliance flows. The peaks noted over October reflect the progression of viral shedding from individual contributions. Although this value will vary with the underlying characteristics of the idle flow emanating from each building, from the presented data, the expected maximum concentration of detectable SARS-CoV-2 within domestic wastewater in the USA should be near 10 7 genome copies/L. Considering that individuals in residences are expected to produce between 100-250 L of wastewater per day, the maximum shedding per person is on the order of 10 10 SARS-CoV-2 genome copies/day, in agreement with Schmitz et al. (2021). This number additionally aligns with the upper-end of identified fecal concentration ranges, suggesting that individuals within these structures likely produce between 100-1000 mL of feces per day (5 × 10 3 -10 7.6 copies/mL feces) (Foladori et al., 2020).
Notably, this analysis is corroborated by the concentrations noted in the secondary isolation building during the peak of infections in September (1.7 × 10 7 SARS-CoV-2 copies/L on September 27th from 158 infected isolating residents ; Fig. 3).
The concentration of fecal matter becomes more critical when considering wider communities with industrial, infiltration, and other diluting contributions to wastewater. In initial attempts to normalize to the varying concentrations of fecal matter within the wastewater samples, the genogroup II F+ bacteriophage was selected as an internal reference marker for the fall campaign to align with other sampling efforts ongoing within Colorado. At the microsewershed level, the F+ bacteriophage signal displayed inconsistent geographical and temporal trends (Fig. 4). Select sites (e.g., R, Q, and O) displayed consistently low signals, within the range of 10 4 to 10 6 copies/L, whereas other sites (e.g., G(FEDCBA), J, and L(Admin)) displayed signals often approaching 10 9 copies/L. Even more concerning, sites such as C, F, H, M, and N displayed inconsistent temporal trends, fluctuating over five orders of magnitude during the campaign. These shifts potentially result from either changes in resident diet or interpersonal fluctuations in the gut virome (Minot et al., 2011). The F+ bacteriophage was therefore replaced by the pepper mild mottle virus (PMMoV) for the spring 2021 monitoring campaign (Rosario et al., 2009).

Utility and consideration of the data
Throughout the fall monitoring campaign, the interpretation and utility of the data varied with the prevalence of SARS-CoV-2 within the community. Considering six scenarios in which the concentration of SARS-CoV-2 is (1) absent, (2) low and stable, (3) low and increasing, (4) high and increasing/stable, (5) high and decreasing, and (6) decreasing to absent, the daily monitoring campaign provided varying levels of support to the pandemic response.
WBE's utility as an early warning signal is primarily experienced in scenarios (1), (2), and (3), in which early detections are the most critical for either preventing or halting community spread. Our campaign demonstrated this utility within the first two weeks of monitoring when high wastewater concentrations were detected for structures K, P, and Q. Residents and employees in these structures were specifically contacted within 12 h of the reporting of the wastewater results and asked to submit a saliva sample for RT-qPCR testing as soon as possible. This testing revealed individual cases not yet discovered by routine salivamonitoring that were then successfully isolated. This prevention response required the deployment of the robust and well-connected testing, contact tracing, and isolation infrastructure. When entering either scenario (4) or (5), the primary utility in WBE is in monitoring the effectiveness of public health intervention strategies employed. The overall fall campaign provided an example, in which the peak in SARS-CoV-2 wastewater concentrations occurred before the social distancing order imposed by the county. Therefore, the on-campus mechanism of testing, tracing, and isolation was demonstrated as effective prior to a more robust and stringent social distancing order being put in-place. The wastewater concentration data was directly used to assess the success and viability of the interventions on campus, both internally and with public health officials. This monitoring better equips public health officials to determine appropriate responses with the infrastructure at hand, notably with more stringent control measures likely leading to migration from campus and potentially transporting viral infections abroad and/or allowing reentry of the virus from broader communityacquired infections when social distancing requirements ease. After the public health orders were enacted in Boulder, Wi-Fi connections within residence halls decreased by 33%. Additionally, clinical testing data later in the semester identified cases with low viral loads without an active infection, highlighting cases in which progression through the disease profile occurred off campus. During this recovery phase (scenario (6)), wastewater data also monitors individuals who may still shed viral RNA even after exiting the infectious period. On campus, students were permitted to leave the isolation structure and return to their residences after ten days. These reentry events could be detected in the wastewater (e.g., see site O, Fig. 2). Reentries thus must also be taken into consideration to prevent shifts in policy based on a true detected signal that is not reflective of a case of concern.
Across university campuses globally and reported within this study, the utility of wastewater monitoring to support public health has been demonstrated. For example, coupling building-level wastewater sampling two-to-three times per week with clinical testing has been demonstrated as an effective approach at diverse institutions such as the University of Arizona (Arizona, USA) (Betancourt et al., 2021), University of North Carolina at Charlotte (North Carolina, USA) (Gibas et al., 2021), and Kenyon College (Ohio, USA) (Barich and Slonczewski, 2021). At Hope College (Michigan, USA), Travis et al. (2021) implemented a higher collection frequency by sampling from nine on-campus residential zones every weekday. In these campaigns, wastewater samples indicating infection prevalence led to clinical testing of all individuals associated with the flagged buildings/populations (subject to university-specific decision frameworks). All found WBE to be a valuable tool for disease containment, often noting wastewater surveillance's utility for identifying and isolating asymptomatic individuals.

Conclusions
With the wide range of fluctuations in daily habits surrounding toilet flushes and personal hygiene behaviors combined with rapid changes in viral loads, daily monitoring becomes critical to track the prevalence of pathogens within building-scale wastewater. The demonstrated monitoring campaign informed on the emergence of likely new infections within given residential structures, notably during the first two weeks of operation, and the effectiveness of on-campus interventions. The utility of these data relied on being in concert with robust medical services and individualized saliva-monitoring testing, providing the ability to translate from community monitoring to intervention. Resulting from the success of the operation of this wastewater monitoring network, this study concluded that (1) economical solutions are readily assembled for operating composite samplers, (2) daily samples enable informed decisions and monitoring of the success of interventions on campus, and (3) wastewater data provides substantial and unique benefit when surveying community health at multiple stages in a disease outbreak. Combined, wastewater monitoring provides a flexible and effective public-health technique when deployed at the building level.

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.