Impacts of vegetable processing and cheese making effluent on soil microbial functional diversity, community structure, and denitrification potential of land treatment systems

The cheese making and vegetable processing industries generate immense volumes of high‐nitrogen wastewater that is often treated at rural facilities using land applications. Laboratory incubation results showed denitrification decreased with temperature in industry facility soils but remained high in soils from agricultural sites (75% at 2.1°C). 16S rRNA, phospholipid fatty acid (PLFA), and soil respiration analyses were conducted to investigate potential soil microbiome impacts. Biotic and abiotic system factor correlations showed no clear patterns explaining the divergent denitrification rates. In all three soil types at the phylum level, Actinobacteria, Proteobacteria, and Acidobacteria dominated, whereas at the class level, Nitrososphaeria and Alphaproteobacteria dominated, similar to denitrifying systems such as wetlands, wastewater resource recovery facilities, and wastewater‐irrigated agricultural systems. Results show that potential denitrification drivers vary but lay the foundation to develop a better understanding of the key factors regulating denitrification in land application systems and protect local groundwater supplies.


INTRODUCTION
Cheese making and vegetable processing are major components of global food production, but these operations generate immense wastewater volumes that must be managed (Asgharnejad et al., 2021).Wisconsin leads the United States in total cheese production, with 1.6 Â 10 9 kg produced in 2022 (United States Department of Agriculture [USDA], 2023).Each kilogram of cheese produced generates an average of 9.4 L of effluent (European Dairy Association, 2018), or 15.0 Â 10 6 m 3 annually statewide.In Europe, it is estimated that 94.3 Â 10 6 m 3 of cheese making wastewater is produced annually (Stasinakis et al., 2022).In 2022, the Wisconsin (US) vegetable processing industry produced 2.8 Â 10 5 mt of snap beans, 4.4 Â 10 5 mt of sweet corn, and 4.9 Â 10 4 mt of green peas (USDA, 2023).It is estimated that for each metric ton of specialty crop processed, between 98 and 23 m 3 of wastewater are generated (Ölmez, 2014)-in 2022, generating between 7.5 Â 10 6 and 17.7 Â 10 6 m 3 of wastewater effluent in Wisconsin for these three crops alone.Discharge of poorly treated industrial wastewaters has been identified as an important anthropogenic source of groundwater nitrate pollution (Gutiérrez et al., 2018;Tokazhanov et al., 2020).
Frequently found in rural locations worldwide, cheese making and vegetable processing operations often cannot dispose of wastewater at large-scale wastewater resource recovery facilities, and direct discharge to waterways is prohibited.Instead, many utilize comparatively simple land application systems designed to optimize microbial denitrification soil processes and which operate at much lower costs than highly engineered techniques (Crites et al., 2010;Smith, 2006).Long-term wastewater disposal using land application can be conducted in a manner protective of the environment with careful characterization and monitoring of wastewater and application site parameters (Isosaari et al., 2010;Siemering et al., 2024).
With the identification of high-quality drinking water supplies as a United Nations (2015) Sustainable Development Goal and recent estimates that in the United States, almost 50% of drinking water and approximately 40% of irrigation water come from aquifers (USGS, 2018), successful treatment of cheese making and vegetable processing wastewater is imperative to prevent contamination of nearby groundwater sources.In Wisconsin, facility nitrogen (N) discharges are limited under state regulation NR214 (Wisconsin Legislative Reference Bureau, 2014) to the sum of that taken up by cover crops (easily quantified and difficult to increase), ammonia volatilization (assumed minimal), and denitrification (substantial, variable, and difficult to measure).Denitrification is a major pathway of nitrogen loss in terrestrial ecosystems in which soil microbial processes reduce nitrate (NO 3 À ) to nitrous oxide (N 2 O) and dinitrogen (N 2 ) (Pan et al., 2022).Soil microbial communities are complex components of the soil ecosystem (Young & Crawford, 2004), drive the recycling and transformation of many major and trace elements in soil (Dang et al., 2019), and are crucial to soil ecological function (Tedersoo et al., 2014;Zornoza et al., 2015).Soil microbial communities are known to be affected by a range of factors, including soil type (Lian et al., 2022;Obayomi et al., 2021), management (Bier et al., 2024), and climatic conditions (Levy et al., 2018;Norris et al., 2023;Orwin et al., 2018).Denitrification is also known to be impacted by system management (Pastorelli et al., 2011;Sun et al., 2021;Wang et al., 2018).
Wastewater application facilitates complex interactions that affect soil microbial activity, diversity, and biomass, which Yalin et al. (2023) found could be crudely divided into positive, neutral, and mixed effects.Variables inherent in real-world systems, the relatively recent widespread low-cost availability of DNA sequencing, and the complexity of the soil microbiome make definitive conclusions challenging.With further development, microbiome analysis may provide insights to allow for system design optimization to denitrify nitrate containing wastewater quickly and completely, thereby protecting groundwater resources.
Research on wastewater application impacts on soil microbial communities is largely focused on municipal wastewater simply due to the volume available for application (Yalin et al., 2023), with a limited focus on industrial wastewaters (Dang et al., 2019).To date, no research has been published focusing solely on cheese making and vegetable processing wastewater application impacts on the soil microbiome.Wastewater from these industries is higher in chemical and biological oxygen demand and can be somewhat higher in total nitrogen than treated municipal wastewater (Mordechay et al., 2018;Yalin et al., 2023;Table S1), but because of its source in food production, it does not contain microbial pathogens, potentially toxic metal elements, or other contaminants (e.g., pharmaceuticals and per-and polyfluoroalkyl substances [PFAS]), which degrade the soil.Prior to application, vegetable processing wastewater passes through settling tanks or ponds to remove organic particulate matter from the waste stream.Cheese making wastewater is filtered to remove whey protein from the waste stream for valorization, thereby removing milk solids, which could plug soil pores and prevent water infiltration.The sodium content in both systems is managed in the manufacturing process.If the system controls operate as designed and hydraulic loading is managed, soil functionality is not compromised, and the systems can operate without the need for soil treatment or replenishment.
The work presented here is part of a larger industry and regulatory agency-funded project investigating cheese making and vegetable processing land-application wastewater treatment system efficiency.An automated N fate monitoring system was developed and deployed at industry land application sites (Siemering et al., 2022).Annual industry system denitrification was determined and compared with values from laboratory incubation studies mimicking field sites.Results confirmed that land application systems were denitrifying wastewater at up to 2129 kg N/ha year, far more than common regulatory limits equal to system cover crop uptake rates (Siemering et al., 2024).
To explore the potential impacts of long-term wastewater application on soil microbial communities and how these communities might be managed to increase denitrification efficiency or used to evaluate the denitrification potential of new treatment facility locations, phospholipid fatty acid (PLFA), ribosomal RNA, and soil respiration analysis were conducted on soil samples collected from agricultural research stations (ARS) and industry facilities (cheese and vegetable).Incubation studies were also conducted using ARS samples to evaluate denitrification in soils not previously receiving wastewater applications.To our knowledge, this is the first use of these techniques to investigate cheese making and vegetable processing wastewater land application systems.We hypothesized that (1) the ARS denitrification rates would be lower than the cheese and vegetable rates because the long-term, frequent application of high nitrate wastewater would alter the treatment systems and microbial communities to favor an increased population of denitrifying microorganisms, and (2) there would be a higher correlation between industry system denitrification and biotic factors compared with the ARS sites.We further hypothesized that abiotic system factors would show a correlation to biotic factors (e.g., microbial community richness and composition) and that these correlations could be used as a tool to evaluate additional industry wastewater system treatment efficiency.

Sample collection and incubation study
Soil was collected at four industry facilities (two cheese making and two vegetable processing facilities) and two ARS in different regions of Wisconsin.One vegetable processing site has sandy soil, while the other sites have silty loam soil.All industry facilities operated under normal permitted conditions and applied wastewater to these systems for a minimum of 8 years prior to sampling.The cheese making sites utilized ridge and furrow systems: a series of closed-end asymmetrical furrows approximately 90 m long and 5 m wide planted with water-tolerant grasses (e.g., Echinochloa crusgalli L., Polygonum persicaria, Polygonum coccineum, Cyperus esculentus L., and Amaranthus spinosus).Wastewater, in a 13-or 30-day cycle, is pumped into one furrow per day to a depth of 10-20 cm and allowed to percolate into the soil.Cheese plant effluent averages 150 mg N/L but is highly variable depending on facility processes (Watkins & Nash, 2010).Vegetable processing facilities apply wastewater via center pivot irrigation systems on 12-or 48-h cycles at rates calibrated to not exceed site soil field moisture capacity to multi-hectare spray fields planted with a mix of perennial forage grasses (Lolium perenne, Bromus inermis, Poe annua, and Elymus repens).Vegetable processing facility wastewater nitrogen levels are moderate (averaging 12-50 mg N/L), but volumes are high and constant for the entire processing season (June-November) (Miller et al., 2008;Wei et al., 2011).Vegetable system grasses are harvested two to three times per growing season for use as animal forage; cheese system grasses are burned annually.The ARS sites are farmed for corn (Zea mays) grain production for 6 years before sampling using standard production techniques for the region.Prior to corn, the agricultural fields were in a corn-soybean (Glycine max) rotation.
The four industry facilities were sampled using an 8.3-cm-diameter soil hand auger at four to six random locations in each treatment cell (cheese) or field (vegetable processing) used in the field monitoring component (Siemering et al., 2024) at 0-30 and 30-60 cm depth layers.Due to COVID research station site restrictions at the time of study, the ARS samples were composited from equal amounts of excess 0-30 cm layer soil collected from experimental plots for a separate study, but 30-60 cm samples were unavailable.Each sample ($30 L total) was composited, kept moist in sealed 19-L plastic buckets, and stored at 2.1 C prior to incubation and microbial analysis.At the time of soil collection, 10-L facility wastewater grab samples were collected from the wastewater streams or holding tank mixtures in polyethylene carboys and stored at 2.1 C until use during typical production and not a cleaning or disinfection cycle.

Soil microbial analysis
For soil microbial analysis, from each homogenized sample (i.e., four to six auger cores from each industry site and layer collected or composite ARS sample), three subsamples were analyzed to account for the high variability in microbial communities, sensu Kane et al. (2020).In all, three replicate samples from the four industry sites (each with two soil layers) and the two ARS facilities (one soil layer) were analyzed, resulting in 30 total samples.The analysis was conducted to explore the variation of microbial communities between facility type (cheese, vegetable, and research station) and sample depth (for cheese and vegetable samples).Correlation analyses of biotic versus abiotic factors and biotic factors versus denitrification were conducted on ARS versus industry soils for the 0-30 cm layer (control vs. wastewater treated) for comparison to incubation study data.
Basal and glucose-induced soil respiration were measured as indicators of general soil microbial activity and biomass, respectively.Respiration was measured by trapping carbon dioxide released in a 24-h period of incubation following rewetting of an air-dried sample, followed by another 24-h period after glucose addition (Batterman et al., 2022).An ANOVA comparing soil sources (industry vs. ARS) was conducted using JMP to evaluate statistical variability in basal and induced respiration.
Microbial communities were also assessed in each sample by amplicon sequencing of the 16S ribosomal RNA gene for bacteria and archaea and a portion of the internal transcribed spacer (ITS) region for fungi at the University of Minnesota Genomics Center.DNA was extracted with the PowerSoil Pro DNA extraction kit (Qiagen, Germantown, MD, USA), and sequencing was performed on an Illumina MiSeq (Illumina, Madison, WI, USA) using 2 Â 300-bp chemistry as described in Gohl et al. (2016).The total number of sequences obtained for the 16S gene and ITS region were 3,650,076 and 2,481,678, respectively.Forward and reverse reads were denoised and merged using DADA2 (Callahan et al., 2016).For the ITS gene, merging the reads resulted in a loss of more than half sequences due to lower quality scores on the reverse reads, so we proceeded with the forward reads only.The resulting sequences were clustered at 99% sequence identity into operational taxonomic units (OTUs) via open reference clustering using the Silva v. 138 (bacterial) and Unite v. 27.10.2022(fungal) databases with vsearch in the QIIME2 platform (Bolyen et al., 2019;Nilsson et al., 2019;Quast et al., 2013;Rognes et al., 2016).The OTUs were exported into R, and further analysis included the use of the phyloseq, vegan, lme4, car, emmeans, and ggplot2 packages (Bates et al., 2015;Fox & Weisberg, 2023;Lenth, 2021;McMurdie & Holmes, 2013;Oksanen et al., 2022;R Core Team, 2021;Wickham, 2024).
The microbial data were analyzed for differences in species richness (Chao1; Chao, 1984), community composition, and the abundance of bacteria associated with denitrification.Analyses were conducted separately for 16S and ITS.Prior to analysis, OTUs were rarefied to 21,002 and 23,592 sequences per sample for 16S and ITS, respectively.Good's (1953) coverage was also calculated to estimate the proportion of total richness that was accounted for by the generated sequencing depth.OTU counts were transformed to relative abundance per sample for community composition and species abundance analyses.Denitrifying prokaryotes were identified in the 16S dataset following Martin et al. (2022) and Walkup et al. (2020), which are the putative ammonia oxidizing taxa (Candidatus Nitrososphaera and Nitrosomonadaceae g.) and the nitrite oxidizing bacteria (Nitrospira sp., Nitrospiraceae g., Nitrospirales FW 4-29, and Nitrospirales 0319-6a21 g.).Community composition was defined as Bray-Curtis dissimilarity, and differences between treatments were evaluated using permutational multivariate ANOVA (PERMANOVA; Anderson, 2001) implemented as adonis2 in the vegan package (Oksanen et al., 2022).For the univariate response metrics (e.g., denitrifying bacterial abundance and Chao1 diversity), tests of depth Â wastewater type were performed with mixed effects ANOVA with Type III sum of squares due to the source (SS).The test of the upper layer of all samples was a one-way ANOVA.For the PERMANOVA of depth Â wastewater type, the nested design was specified in the permutational scheme.Anywhere pairwise contrasts are reported, the p values are corrected for multiple testing via the Bonferroni method.

Basic soil parameters
The basic parameters for all layers of the three soil types used in the incubations are shown in Table 1.In general, soils from the different sites were similar, with the exception of P, K, and NO 3 À at the ARS sites, which were generally lower than at industry sites.

Incubation denitrification rates
At each incubation temperature, the percent N loss was higher and rate variability was lower for the 0-30 cm ARS soils compared with the cheese and vegetable soils (Figure 1).The denitrification rate difference and sample variability between all soil sources were lowest at 35 C. The ARS soils showed overall much higher denitrification rates, except at 35 C, compared with the other sites.
At the lowest incubation temperature (2.1 C), ARS soil denitrification was estimated to be 75% (±3%) compared with 11% (±8%) for cheese making site soils and 14% (±15%) for vegetable processing facility soils.Statistical analysis showed that the combined effect of soil source and temperature was significant (p = 0.0057).Individually, both temperature and soil source variability produced highly statistically significant (ANOVA; p ≤ 0.0001) differences in denitrification rates (Table 2).

Microbiome variation by site and depth
Bacterial and fungal community composition analysis showed significant differences (PERMANOVA; p = 0.001; Table S2) in the composition of both the bacterial and fungal communities among the individual sites in the 0-30 cm layer (Figure 2a,b).Good's coverage was >99 in all samples, indicating that our sequencing depth was appropriate to capture most of the microbial diversity present in these soils.
Differences between soil layers at the cheese and vegetable sites were also explored.Only one soil layer was available for analysis at the ARS sites, precluding layer comparisons between all sites.Pairwise comparisons of bacterial data indicated that depth was not an important factor structuring the soil bacterial community at the cheese sites but was statistically significant at the vegetable sites (Figure 2 and Table 3).The composition of the soil fungal communities did not differ significantly by depth within the wastewater type (Table 3).

Microbiome indices analysis
Bacterial OTUs and PLFA richness were significantly different (p < 0.05; Table 4) between ARS and both cheese and vegetable sites.ARS bacterial richness was significantly higher only at cheese sites, and fungal diversity was higher at ARS versus vegetable sites.Basal and glucose respiration were significantly higher in both the cheese ($2.3Â) and vegetable (3.4Â) soils versus ARS (Tables 4 and S4).PLFA diversity and denitrifier relative abundance showed no difference between the ARS and industry sites, indicating that wastewater treatment affected microbial activity without concomitant changes to community composition or biomass.Significant differences were also observed between depth layers in the cheese and vegetable samples (Tables 4 and S5).Vegetable soil respiration was significantly higher ($1.45) than cheese soils in the 0-30 cm layer, but cheese respiration was higher in the 30-60 cm layer.The 0-30 cm vegetable layer also had significantly higher fungal diversity, PLFA diversity, and PLFA richness than the 30-60 cm vegetable layer.The cheese soil layers were more similar, with only the 0-30 cm layer having a significantly higher PLFA diversity compared with the 30-60 cm layer.

Biotic and abiotic factor correlation to % N loss
There are significant high correlations at 20.4 C, 30 C, and 35 C between basal and glucose respiration and % GramÀ for the ARS soils (Table 5), but significant negative correlations for PLFA richness at 20.4 C, 30 C, and 35 C. For the cheese and vegetable soils, there were  multiple significant positive and negative correlations to PLFA factors, but without patterns that would explain the difference in denitrification rates.For cheese soils, there were no significant correlations to PLFA factors at 8 C, while at other temperatures, there were significant positive and negative correlations.Bacterial denitrifier relative abundance (Table 6) was highly correlated to denitrification in vegetable systems at 20.4 C and 30 C but highly negatively correlated at 2.1 C. Bacterial diversity was correlated with ARS at 8 C, 30 C, and 35 C. Bacterial richness showed no correlation with denitrification.Fungal diversity was positively correlated at 20.4 C (cheese and vegetable) and 30 C (vegetable).
ARS sites showed no significant correlations with any bacterial community factors but significant positive and negative correlations between Nitrosphaeria, Acidobacteriae, Thermoleophilia, Anaerolineae, and Vicinamibacteria and various abiotic factors.Cheese sites showed a significant correlation between bacterial denitrifier relative abundance versus NH 4 + and bacterial richness versus C/N ratio.Actinobacteria, Bacilli, Bacteroidia, Gammaproteobacteria, and Planctomycetes showed significant positive and negative correlations with multiple abiotic factors at cheese sites.Vegetable sites showed significant positive and negative correlations between bacterial denitrifier relative abundance, fungal diversity, and multiple abiotic factors.There were also many more bacterial classes that showed significant correlation abiotic factors in vegetable systems than cheese and ARS sites, including Nitrosphaeria, Myxococcia, Gemmatimonadetes, Actinobacteria, Bacilli, KD4-96, Alphaproteabacteria, Thermoleophilia, Verrucomicrobiae, and Vicinamibacteria.ARS were positively correlated with Acidobacteriae and Bacilli at 35 C, 30 C, and 20.4 C; KD4-96, Vicinamibacteria, Nitrososphaeria, Acidimicrobiia, Anaerolineae, Bacteroidia, and Gammaproteobacteria at 8 C; and Myxococcia and Bacteroidia at 2.1 C. Cheese sites were positively correlated with Alphaproteobacteria and Planctomycetes at 35 C; Blastocatellia, Bacilli, and Polyangia at 20.4 C and 30 C; Bacteroidia and Planctomycetes at 8 C and 2.1 C; and Gammaproteobacteria at 2.1 C. Vegetable sites were positively correlated with Actinobacteria, Bacilli, and Alphaproteobacteria at 35 C; Nitrososphaeria, Myxococcia, Gemmatimonadetes, KD4-96, Thermoleophilia, and Vicinamibacteria at 20.4 C and 30 C; and Actinobacteria, Bacilli, Alphaproteobacteria, and Verrucomicrobiae at 8 C and 2.1 C.
Correlations between soil microbial, soil respiration, and PLFA factors and bacterial class >2% relative abundance to abiotic factors were calculated (Tables S5-S8).Correlations with abiotic factors exhibited no obvious T A B L E 4 Mean and SD of bacterial, fungal, and PLFA community richness and diversity indices and respiration between the three soil systems and two soil layers.
Soil system and layer (cm)  patterns.Soil respiration and PLFA factors showed significant positive and negative correlations at all three site types.More noteworthy are the respiration and PLFA factors that showed no correlation: cheese sites, metabolic stress ration and glucose respiration; vegetable sites, PLFA diversity and PLFA richness; and ARS sites, metabolic stress ratio, % Gram+, and PLFA fungal:bacterial.

DISCUSSION
Basic soil parameters (Table 1) show that the C/N ratio, pH, OM%, and organic N values at all sites are in the range that supports microbial denitrification processes (Li et al., 2022;Subbarao et al., 2006).P and K values in the industry facility soils are considered excessively high, with the ARS sites at optimal P and low K values for plant growth (Laboski et al., 2012).System P levels were all sufficiently high to not limit nitrogen cycling activities, with Wu et al. (2022)

Incubation denitrification rates
The overall high denitrification rates at 35 C (Figure 1) correspond to the increase in soil denitrification between 30 C and 40 C in Lai et al. (2021), who observed a positive correlation between temperature, extracted DNA, and the nrfA and nosZ genes.Lai et al. (2021)  incubated at 30 C and 20.4 C is likely due to variability in soil microbial composition caused by differences in wastewater chemical composition.Siemering et al. (2024) showed this same temperature effect for the 30-60 cm soil layer at these same cheese and vegetable site soils, indicating that depth is not a major factor in system denitrification rates.
Conversely, microbial metabolic suppression, community composition, and activity at the 8 C and 2.1 C incubation temperatures are the likely causes of the lower denitrification rates in industry soils (Guo et al., 2022;Pang et al., 2015;Song et al., 2024).Meng et al. (2020) found that community diversity may decrease under low temperatures, thereby limiting denitrification in a riverbank filtration system with TN removal efficiency of 53.7% at 10.5-14.5 C, compared with 75% in ARS and 11%-14% in industry soils.In a more controlled environment, Xu et al. (2019) found that nitrogen removal efficiency in an inoculated sludge municipal wastewater bioreactor gradually declined as operating temperature decreased with denitrification at 30 C 5Â that at 10 C. In comparison, the cheese and vegetable systems studied here decreased to approximately 50% efficiency at 20.4 C, only showing a 3Â difference in denitrification between 30 C and 10 C. A review by Zhou et al. (2018) identified bioaugmentation (the addition of pure strains or mixed cultures of microorganisms) as a possible method to improve system efficiency.A future study may indicate that bioaugmentation may improve land application systems as well.
As the cheese making facilities apply wastewater year-round, sufficient denitrification at 8 C (spring/fall) and 2.1 C (winter) is critical for system success.Siemering et al. ( 2024) calculated that even in the cold seasons, the facilities operate with sufficient efficiency to protect groundwater supplies, likely because heated discharge water creates denitrification hotspots even at low ambient temperatures similar to those observed by Hattori et al. (2019).
The decrease in cheese and vegetable system performance with temperature contrasts with approximately 90% efficiency at 20.4 C, 80% at 8 C, and 75% at 2.1 C in the ARS systems.One possible explanation for the divergence in denitrification rates is the difference in aerobic status between systems.ARS systems are aerobic virtually all the time, while cheese systems are intermittently aerobic between flowing events, and the vegetable systems are aerobic only at the surface during the 4-to 6-month warm season operating cycle.Although the incubations were conducted with the soil just below saturation, these initial system conditions may have altered the microbial community composition.Ji et al. (2015) noted that aerobic denitrification can be performed by various genera of microorganisms (primarily α-, β-, and γ-Proteobacteria) and tends to work efficiently at 25-37 C, pH 7-8, 3-5 mg/L dissolved oxygen, and C/N load ratio 5-10.The data presented here may show that the ARS systems are dominated by aerobically efficient Proteobacteria, thereby possibly leading to their high incubation study denitrification rates.
The reason for the uniformly higher ARS denitrification rates with lower variability was not immediately obvious.The soils were collected similarly, stored, and incubated identically.It is plausible that the lengthy history of soil wastewater application at industry sites and its potential "conditioning" of the soil microbial community would be the most likely factor in the denitrification rate difference.

Composition of microbial communities
Figure 3 indicates that high-nitrogen cheese making and vegetable processing wastewater appear to change the relative abundances of soil microbial phyla both compared with ARS soils and to each other.The data here mirror the findings of multiple studies on microbial communities in systems receiving wastewater with elevated nitrogen levels and suggest a "core" wastewater soil microbial community.For the denitrifier stricto sensu communities, the potential NO and N 2 O reducers have been found to differ between the ecosystems, with soil denitrification rates dependent on multiple environmental conditions such as adequate moisture and inorganic N availability (Butterbach-Bahl et al., 2013;Hallin et al., 2018).
The dominant Proteobacteria, Actinobacteria, and Acidobacteria phyla observed in all three soil types correspond with the results of several other studies: Qin et al. (2022), constructed wetlands; Kim et al. (2015), agricultural region groundwater; and Guo et al. (2022), nitrate-laden groundwater.Meng et al.'s (2020) analysis of microbial communities in a river-bank system in Harbin, China, detected the dominant phyla to be Proteobacteria, Actinobacteriota, Acidobacteria, Firmicutes, Gemmatimonadetes, Verrucomicrobia, Nitrospirae, and Bacteroidetes, which matches the results of this study.In the ephemeral wetland-like cheese making facilities, Firmicutes were present in significantly greater numbers than in the ARS or vegetable processing soils, which corresponds to the findings of Ishii et al. (2009), who identified Firmicutes as playing a leading role in denitrification in rice paddy soil.The fluctuating nutrient environments of the industry systems may be partially dominated by Acidobacteria, as these are able to thrive in diverse oxygen conditions and utilize various carbohydrates, as well as inorganic and organic nitrogen sources (Eichorst et al., 2018).
The long Wisconsin cold season (October-April) appears to impact microbial community diversity in both the industry and ARS soils, with the results here matching Pessi et al.'s (2022) work on arctic tundra soils.In both cases, Actinobacteria, Proteobacteria, and Acidobacteria phyla dominated, with Chloroflexota and Verrucomicrobiota also prevalent but at smaller percentages.Pessi et al. (2022) further concluded that dominant phyla had truncated denitrification pathways with denitrification-associated genes found in only 13.8% of the metagenomically assembled genomes and affiliated with these phylla.
Figure 4 shows that the dominant prokaryotic classes in the studied systems here (α-and γ-Proteobacteria) differ substantially from the incubation studies of Coyotzi et al. (2017), which found Betaproteobacteria comprising 71% of soil taxa across multiple sites and time points in agricultural systems treated with N fertilizer.This difference may be because Coyotzi et al. (2017) starved their soil incubation samples of oxygen for 22 days, followed by feeding with potassium nitrate and glucose, as opposed to our use of facility wastewater as a C and N source with no microbial population manipulation before incubation.
The results here correspond more closely to the real-world conditions of a managed wetland system, where Zhang et al. (2021) found system denitrification dominated by the classes Gammaproteobacteria, Deltaproteobacteria, Bacteroidia, and Anaerolineae and C cycling controlled by the phyla Chloroflexi, Actinobacteria, Firmicutes, Spirochaetes, and Bacteroidetes and the class Alphaproteobacteria. Notably, all the systems here are dominated by the class Nitrososphaeria, an ammonia oxidizer (Stieglmeier et al., 2014) within the globally dominant Archaea, with wide evolution reflecting adaptation to variations in environmental conditions (Zhao et al., 2023).Zhao et al. (2023) further concluded that predominant Nitrososphaeria lineages exhibit a patchwork of gene inventory and expression profiles for ammonia, urea, and phosphate utilization with their carbon fixation, respiration, and adenosine triphosphate (ATP) synthesis-associated genes conserved and expressed consistently in soil and that this combination indicates resource-based patterns and complementary ecophysiological niches associated with differing nutrient availability.
Dominant microbial phyla and classes in denitrifying systems are well documented, and the findings of this study largely agree, but little literature exists showing correlations between denitrification rates and the relative abundance of specific classes at different temperatures.Shan et al. (2023) detailed how bioaugmentation of water resource recovery facilities at low operating temperatures with mixed microorganism consortia leads to the enrichment of functional taxa such as Proteobacteria, Bacteroidota, and Actinobacteria and may represent a potentially useful approach to nitrogen removal under low temperatures.Future work would benefit from an assessment of not only those microbes whose DNA or cellular material are present but also the subset of the microbial community that is active at the time of sampling, for example, through an RNA-based (Freedman et al., 2015;Romanowicz et al., 2016) or stable isotope-informed analysis (Piñeiro et al., 2024), which may enable further insight into the association between the abundance and composition of putative denitrifying microorganisms and their functions.

Biotic factor correlation to %N loss
Statistically significant positive and negative correlations were found between soil respiration, PLFA-derived factors, and incubation temperatures (Table 5), but without obvious patterns or trends.This differs from Feng and Simpson (2009), who found via PLFA analysis that increased incubation temperatures enhanced microbial activity, concluding that at higher temperatures, easily degradable structures were quickly exhausted.In all cases, correlations at 2.1 C and 35 C paired and were in opposition (positive to negative) to the correlations at 20.4 C and 30 C. A similar pattern is seen in the vegetable soils, but with only the 2.1 C correlations in counterpoint to the 20.4 C and 30 C data.The cause of this pattern could not be determined but could be related to the predominance of complex C and N structures in the wastewater.
Among microbial biotic factors (Table 6), fungal diversity was positively correlated in industry systems at 20.4 C and 30 C, but not at lower temperatures nor in the ARS systems.Hayatsu et al. (2008) found fungi to be major contributors in the production of N 2 O and N 2 in grasslands using O 2 respiration, denitrification, and ammonia fermentation as conditions dictate, which may be what is observed in the industry system with this correlation as the industry systems cycle between aerobic and anaerobic conditions.
Seven of the 18 dominant soil microbial classes at the vegetable sites exhibited significant, strong positive or negative correlations at all temperatures (Bacilli, Nitrososphaeria, Actinobacteria, Thermoleophilia, Vicinamibacteria, Alphaproteobacteria, and Myxococcia).No ARS or cheese bacterial classes exhibited significant correlations at all temperatures.Individual classes do show significant correlations with patterns that may explain denitrification rate differences between the three soil types, but further study would be required to elucidate these.
At 20.4 C, 30 C, and 35 C, Bacilli was the only class positively correlated with all three soil types and may be a primary denitrifier in these systems despite their low relative abundance.Jia et al. (2021) found Bacillus capable of autotrophic and/or heterotrophic denitrification as well as anammox under a range of C/N conditions in a wetland environment.Yang et al. (2020) isolated and characterized four Bacillus strains that could simultaneously utilize nitrate, nitrite, and ammonium under aerobic conditions.Their denitrification pathways were confirmed by nitrogen balance and enzyme and gene analysis, showing nitrogen removal efficiency up to 50%-80% under laboratory conditions.The Bacillus aerobic denitrification pathway is supported in all three soil systems, being fully (ARS), cyclically (cheese), or surficial (vegetable) aerobic, all with sufficient dissolved organic carbon (DOC) to support microbial respiration.
Planctomycetes (anammox bacteria) exhibited a positive correlation with denitrification at 35 C, 8 C, and 2.1 C in the cheese making systems.This may be due to the impacts of the higher salinity of this wastewater.Park et al. (2021) found that in an up-flow reactor, Planctomycetes increased with salinity by outcompeting Proteobacteria and that 80% nitrogen removal efficiency could be maintained at up to 2% salinity.The lack of correlation at 20.4 C and 30 C cannot be explained here, but Mason et al. (2021) noted that core microbiome and metagenomic sequences point to an unappreciated yet potentially important Planctomycetes role in N removal.

Abiotic factor correlation to %N loss
While no distinct patterns of abiotic correlations to denitrification emerged from the data in this study, other researchers have demonstrated that abiotic soil conditions such as soil pH, soil moisture and temperature, soil aeration, water-filled pore space, and soil carbon influence denitrification (Brenzinger et al., 2017;Kou et al., 2019;Mehnaz et al., 2019;Norton & Ouyang, 2019;Žurovec et al., 2021).Graham et al. (2014) concluded that abiotic soil properties structured microbial community traits and activity and that the value of relative abundance data over abiotic factors varied by process and by season.
Our incubation study data also showed large denitrification differences between soil types and a significant positive correlation with Nitrosphaeria (Crenarchaeota) at 2.1 C and 8 C in ARS soils and 20.4 C and 30 C in vegetable soils, and the converse with Actinobacteria (positive at 2.1 C, 8 C, and 35 C for vegetable and negative at 2.1 C, 8 C, and 35 C at cheese sites).Specifically, increased relative abundance of the phyla Actinobacteriota and Crenarchaeota was positively correlated with denitrification.This lack of uniform correlation may be due to the interplay of C/N source and temperature impact on the microbial communities, and the comparatively small sample size.Deveautour et al., (2022) (N = 136) found large differences in soil potential denitrification between sites (up to 41.5 mg N 2 O-N/kg soil/ day), with the factors most predictive of soil potential denitrification being soil physico-chemical properties (particularly soil P, pH, and total N) and the prokaryotic community composition (but not fungal).Additional large datasets from China (Kou et al., 2019) and New Zealand (Morales et al., 2015) assessed potential denitrification using soil incubations coupled with microbial community analysis and showed similar effects of soil physico-chemical properties and soil microbial communities on denitrification.
The observed lack of correlation between biotic and abiotic factors and denitrification rates in this study may be due, in part, to the short and constant incubation periods, neglecting other factors that regulate microbial activity.Hazard et al. (2021) highlight that incubation assays offer mechanistic insights but do not directly indicate in situ rates.This may clarify why Siemering et al. (2024) found land application treatment systems capable of treating up to 2129 kg/ha/year, contrasting with the efficiency indicated by the incubation study data.A paucity of denitrification predictors may also be scale dependent, with differences seen only on a much larger scale (e.g., Kou et al., 2019), but not within the relatively small land area and limited climatic and soil variability of central to southern Wisconsin where samples were collected.

CONCLUSIONS
The results of our study show that denitrification in soils collected from cheese making and vegetable processing facilities decreased with temperature but remained high in soil collected from tilled agricultural fields even to 2.1 C (75%).A soil microbiome analysis showed that in all three soil types, at the phylum level, Actinobacteria, Proteobacteria, and Acidobacteria dominated, whereas at the class level, Nitrososphaeria and Alphaproteobacteria dominated, with Actinobacteria statistically more numerous at the ARS sites compared with industry sites.Correlations between the system abiotic and biotic factors showed no clear patterns that would explain the differences in the incubation denitrification rates.
The dominant phyla and classes in the ARS, cheese, and vegetable systems studied here are similar to those found in wastewater irrigation and other denitrifying systems throughout the world, despite the long-term high-N wastewater applications.This similarity will allow the data derived from a range of environments to confidently be applied to these land application systems.Of particular importance will be research that identifies processes (e.g., bioaugmentation) or other system manipulations that increase denitrification efficiency.The contrasting correlation results show that potential denitrification drivers vary, and our understanding of the potential wastewater treatment capacity of soil-based systems is still limited.These results lay the foundation for developing a better understanding of the key factors regulating the potential denitrification of land application systems, which, in turn, can guide system management to minimize nitrogen leaching to local groundwater supplies.

F
I G U R E 1 Effect of temperature on denitrification at agricultural research station (ARS), cheese, and vegetable sites.T A B L E 2 Analysis of variance of percent N loss comparing all soil sources and incubation temperatures.

F
I G U R E 2 Non-metric multidimensional scaling plot visualizing the Bray-Curtis dissimilarity of the soil (a) bacterial and (b) fungal communities under different wastewater regimes and at different depth layers.The distance between points indicates greater dissimilarity in terms of the species found in the samples.Sites in red indicate cheese making facilities; in blue, vegetable processing facilities; and in green, agricultural research station sites.Soils from 0 to 30 cm deep are shown as circles, and those from 30 to 60 cm deep are shown as triangles.T A B L E 3 Pairwise bacterial and fungal post hoc contrasts of wastewater type (cheese and vegetable) by depth (0-30 and 30-60 cm) from permutational multivariate ANOVA (PERMANOVA).Values are adjusted by the Bonferroni method to control for multiple tests.

F
I G U R E 3 Rings represent the average relative abundance of bacterial phyla that make up at least 1% of the whole community in at least one soil type.*Significant difference between agricultural research stations (ARS) and cheese.+ Significant difference between ARS and vegetable (p < 0.05, Student's t test).F I G U E 4 Rings represent the average relative abundance of bacterial classes that make up at least 2% of the whole community in at least one soil type.*Significant difference between agricultural research stations (ARS) and cheese.+ Significant difference between ARS and vegetable (p < 0.05, Student's t test).
Experimental site soil parameter mean values.
Soil microbial community factors and >2% class correlations versus %N loss at the five incubation temperatures (Pearson, p < 0.05).
rates, as indicated by the marked lower industry soil denitrification below 35 C. Xiao et al. (2023) and Lai et al. (2021) used readily bioavailable ammonium nitrate (NH 4 NO 3 , N source) and Qu et al. (2022) glucose (C source).High variability between industry samples T A B L E 5 Soil respiration and PLFA factor correlations versus %N loss at the five incubation temperatures (Pearson, p < 0.01).T A B L E 6 Abbreviations: ARS, agricultural research stations; OTU, operational taxonomic unit; PLFA, phospholipid fatty acid.