Bacterial Community in Foam-Sand Filter Media in Domestic Sewage Treatment: A Case Study of Elevated Ammonium Nitrogen Content

Abstract

The structure of microbial communities in sponge-sand filters, used for the treatment of real domestic sewage with elevated ammonium nitrogen concentrations (approximately 155 mg·dm⁻³), was characterized using 16S rRNA gene sequencing. Analyses using the Illumina technique allowed us to perform a comparison of filters by layer (two or three layers) and type of fill (waste PUR foams with 95% open porosity, sand). Proteobacteria, actinobacteria, and firmicutes were shown to be the most abundant phyla. The number and type of fill layers had a significant impact on the diversity of nitrifying bacteria. The presence of Nitrosomonas and Nitrospira was observed in every sponge fill sample, but the abundance of autotrophic nitrifiers was negligible in the two-layer filter. The conditions there proved more favorable for the growth of aerobic heterotrophic bacteria. Also in the Schmutzdecke layer, a dominance of heterotrophic nitrifiers was found. The abundance of bacteria with nitrifying activity (AOB, comammox, HNAD) in the biomass of spongy fill placed in casings was 1.7 times lower than in foams without casings. In addition, anammox bacteria (unidentified Planctomycetes), found mainly in the sponge fill and Schmutzdecke of the three-layer filters, may have been responsible for NH4⁺-N removal exceeding 70%. In the case of the two-layer filter, the removal of this pollutant reached 92%. Burkholderia and Sphingopyxis were identified as the predominant denitrifying bacteria. The foam-filled filter in the casings showed an increase in o_Caldilineaceae, involved in nitrate removal as non-denitrifiers. Actinomycetes Pseudonocardia and Amycolatopsis, as well as Proteobacteria Devosia, Acinetobacter, and Bdellovibrio, were found to be involved in phosphorus removal in the waste PUR foams.

1. Introduction

The biological removal of ammonium nitrogen from domestic wastewater is most often based on conventional technology, using autotrophic nitrification by AOB, NOB, and/or comammox bacteria, and heterotrophic denitrification. Enzymes that play a key role in these processes include Amo, Hao, Nxr, Nar, Nir, Nor, and Nos [1]. Due to the need to provide adequate DO for nitrifying bacteria and anaerobic conditions for denitrifying bacteria, the nitrification and denitrification processes must take place in separate facilities. For low C/N wastewater treatment, an additional carbon source must be provided for optimal denitrification. Carrying out nitrification and denitrification in a traditional system therefore requires strict control of the processes of, among other things, aeration and recirculation. This is associated with increased investment costs for wastewater treatment plants (WWTPs), limiting the practical application of this approach in decentralized communities [2,3]. Nitrogen removal technology in a single SND unit has attracted considerable attention in recent years [4]. Numerous species of bacteria capable of simultaneously carrying out heterotrophic nitrification and aerobic denitrification, so-called heterotrophic nitrification-aerobic denitrification bacteria, are discussed in the literature. These bacteria have been isolated from high ammonia nitrogen wastewater such landfill leachate or farm wastewater and are characterized by their ability to convert NH₄⁺-N into N₂ under aerobic conditions in the presence of sufficient organic carbon sources [5]. These include the genera Paracoccus denitrificans, Arthrobacter, Azotobacter, Alcaligenes sp., Acinetobacter sp., Agrobacterium sp., Bacillus sp., Chryseobacterium sp., Klebsiella sp., and Pseudomonas sp. [6,7,8,9,10,11,12]. Numerous studies have confirmed that the availability of carbon and different forms of nitrogen is a very important factor shaping the structure of the HNAD microbial community [9,10]. Ji et al. [13] emphasized that it is important to maintain a certain C/N that is specific to the heterotrophic organism. Those authors reported that low carbon concentrations result in poor growth of HNAD bacteria and thus low denitrification efficiency. In addition, they showed that too much C/N inhibits bacterial growth, negatively affecting the efficiency of the nitrogen removal process. Prasetyo et al. [14], in a study on the effect of C/N on nitrate removal by Pseudomonas sp., found 98% nitrate removal efficiency at C/N = 10. Rout et al. [6] noted an optimum C/N of 7.5 for Bacillus cereus for nitrogen removal from domestic wastewater. Those authors showed that low C/N of 2.5 and 5 resulted in lower effluent treatment efficiency.

HNAD bacteria have a high tolerance to DO, which, within a single strain, allows them to be highly efficient in terms of nitrogen removal [15,16]. Although the number of literature reports on them is increasing, the metabolic pathways and mechanisms of nitrogen removal by HNAD bacteria are still not fully understood. There are two main metabolic pathways: (1) conversion of NH₄⁺-N via NH₂OH, NO₃-N, and NO₂⁻-N into N₂; (2) and conversion of NH₄⁺-N via NH₂OH into N₂O and N₂ [5]. The enzymes involved in heterotrophic nitrification are mainly Amo and Hao [15]. Bacteria belonging to the families Alcaligenaceae, Bacillaceae, Moraxellaceae, Pseudomonadaceae, Rhizobiaceae and Xanthomonadaceae are involved in the oxidation of not only ammonia but also nitrite via the Nxr gene [17]. In anaerobic denitrification, the membrane-bound Nar enzyme plays a key role. Under aerobic conditions, the Nap enzyme contributes to the ability of HNAD bacteria to perform denitrification and to their tolerance to excess DO [18,19]. Other enzymes involved in aerobic denitrification are Nir, Nor, and Nos [20]. Many HNAD bacteria are bacteria expressing the above-mentioned genes and thus capable of complete nitrogen removal via pathway (1) [21]. Examples of such HNADs include Pseudomonas aeruginosa SNDPR-01 [22] and Bacillus cereus GS-5 [6]. Alcaligenes sp. TB is capable of nitrogen removal without NO production [23], and Pseudomonas fluorescens wsw 1001 is capable of nitrogen removal without NO and N₂O [24]. An even shorter NO₃⁻-N bypass pathway is demonstrated by Agrobacterium sp. LAD9 [25]. Pathway (2) was observed in Alcaligenes faecalis and Acinetobacter calcoaceticus HNR [21]. Even more truncated HNAD pathways have been discovered, with the omission of NO₃-N, NO₂-N, NO, and N₂O for Halomonas venusta SND-01 [26], or without the production of NH₂OH, NO₃-N, NO₂-N, or NO for Pseudomonas taiwanensis J488 [27]. Many of the above HNAD bacteria possess the Nos gene, which plays an important role in the conversion of N₂O to N₂ [28,29].

PN/A technology is indicated for the treatment of wastewater with high ammonium nitrogen concentrations. The anammox process is carried out by bacteria that belong to the Planctomycetes cluster of the Planctomycetaceae family. No external carbon source is required for the growth of these anaerobic bacteria, as the presence of carbon dioxide is sufficient for them [30]. Another way to remove nitrogen from low C/N wastewater can be to utilize internal carbon sources by enriching communities with bacteria capable of endogenous denitrification [31]. Unlike the traditional denitrification process, which is carried out by heterotrophic denitrifiers, the bacteria involved in the ED process are facultatively anaerobic microorganisms capable of accumulating and degrading PHA or PHB. These compounds are produced by bacteria from volatile fatty acids (VFAs) present in the environment [32]. The main groups involved in VFA production are acid-forming bacteria, including various acetogenic and homoacetogenic bacteria such as Clostridium. They can produce VFAs (in the form of acetic acid and butyric acid) from external carbon sources, i.e., cellulose and lignocellulose, which are part of dietary fiber, among others [33]. The resulting VFAs become a substrate for the synthesis of PHA or PHB by HNAD bacterial strains, such as Cupriavidus, Pseudomonas, Bacillus, Alcaligenes, and Agrobacterium [34,35,36]. PHAs and PHBs have a dual role in endogenous denitrification, as carbon reserve material and electron donors, supporting the conversion of nitrate to gaseous nitrogen in anaerobic environments. Many types of β-Proteobacteria, especially those of the Comamonadaceae family and most species of Burkholderia sp., have the capacity for anaerobic ED due to their ability to utilize previously accumulated PHA or PHB, respectively, when other sources of organic carbon are exhausted [37,38].

Recently, increasing attention has been given to the fact that wastewater treatment plants based on a traditional biological nitrogen removal system are responsible for greenhouse gas emissions, i.e., nitrous oxide (N₂O). This gaseous form of nitrogen is released during both nitrification (as a by-product of ammonia oxidation) and denitrification processes (as an intermediate or end product) [39,40]. Not all heterotrophic denitrifiers can perform complete denitrification by reducing nitrate to molecular nitrogen. This is because some of them do not have the necessary enzymes to reduce nitrate to nitrite, nitrite to NO, or N₂O to nitrogen gas. The only biological way to minimize N₂O emissions is its reduction by nitrous oxide reductase-expressing bacteria [41,42,43]. Both non-denitrifiers (lacking some of the genes necessary for complete denitrification) and total denitrifying bacteria (containing the full set of denitrification genes) have the potential to reduce N₂O. Those more prone to complete denitrification turn out to contain genes encoding key denitrification enzymes, namely Nar and Nir, in addition to the Nos genes [44]. The presence of such microorganisms may also play a significant role in mitigating N₂O emissions in autotrophic PN/A nitrogen removal systems. An example is consortia with HNAD bacteria, which can grow on microbial decomposition products under conditions without an influx of external organic matter [45]. In bacterial communities where, for example, Acinetobacter and Comamonas are responsible for incomplete denitrification but are supported by Pseudomonas stutzeri, N₂O is an obligatory intermediate product of the heterotrophic denitrification process, and gaseous nitrogen is its end product [40].

New solutions for treating wastewater with higher ammonium nitrogen and low C/N are still being sought. Due to the aforementioned advantages, HNAD bacteria can be used in a single unit to remove ammonium nitrogen from diffuse sources. This becomes competitive with the traditional autotrophic nitrification-anaerobic denitrification system. Furthermore, the nitrogen removal efficiency of HNAD bacteria that are immobilized on a carrier is better than that of bacteria without a biomass carrier [46]. As a biomass carrier for such bacteria, the present author has proposed the use of post-consumer waste in the form of PUR foam cuttings as a vertical flow filter fill in the treatment of domestic wastewater with elevated ammonium nitrogen content. The author’s previous work highlighted a 70% removal efficiency of NH₄⁺-N in the sponge fill with low cell biomass production [47,48], despite the phenomenon of clogging, caused by the overgrowth of the upper layer of sand filters. Suspended material accumulates on the filter surface together with living microorganisms, forming a gelatinous matrix known as the Schmutzdecke. This active biological community layer plays a major role in the removal of pollutants such as carbon, nitrogen, and phosphorus.

This article describes a study to assess the potential for microbial growth in PUR foams and the Schmutzdecke of multilayer sponge-sand filters by analyzing the microbial community and its contribution to the removal processes of contaminants from real domestic sewage with elevated ammonium nitrogen. The author proposes the hypothesis that NH₄⁺-N removal under low C/N conditions is influenced by the growth potential of a diverse bacterial community in a system that operates without additional aeration.

2. Materials and Methods

2.1. Research Methodology

The column models of vertical flow filters were built on a semi-technical scale using PVC columns (10 cm in diameter and 100 cm in height) and operated continuously for 420 days without additional aeration. After 400 days, one of the models became irreversibly clogged.

Each column received the same volume of actual domestic sewage pretreated in a septic tank. A grate supporting the filter bed was installed at the bottom of each column. Pretreated sewage (influent) was collected from a separation chamber located downstream of the storage tank, and the treated filtrate (effluent) was collected from each model (A–D) (Figure S1a). The filters were filled with waste flexible polyurethane foam of random shapes, used commercially as upholstery foam. They featured an open porosity of 95% and an equivalent diameter d₁₀ of 4 mm [48]. A detailed characterization of the PUR foams as biomass carriers can be found in the author’s previous works [47,48,49,50]. The quartz sand used in the filters was characterized by an equivalent diameter d₁₀ of 0.32 mm. Table S1 shows the composition of the raw domestic sewage.

This study compared the efficiency of nitrogen removal during the treatment of wastewater with elevated ammonium nitrogen content in four sponge-sand columns. Two of these consisted of three layers: an upper double layer filled with foam in casings (column Model A) or without casings (column Model B) and a lower sand layer. The third two-layer column consisted of a single layer of waste foam without casing and a lower sand layer (column Model C). The sponge fill material was a mixture of waste soft foams [48]. A single-layer sand filter (column Model D) was used as a control column.

Based on 16S rRNA gene sequences, the structure of the microbial community that populated the sponge fill and the Schmutzdecke layer of the filters over a research period of 420 days of domestic wastewater treatment with elevated ammonium nitrogen was analyzed.

2.2. Analytical Methods

2.2.1. Illumina Sequencing of 16S rRNA Gene

Six sponge samples (approximately 2.4 g dry weight) were taken from the top and bottom layers of the spongy material of column Models A, B, and C, respectively (Figure S1b). Three samples (approximately 10 g dry weight) were also taken from the Schmutzdecke of each model (A–C). The control sample (approximately 10 g dry weight) was a sample from the Schmutzdecke of a filter filled entirely with sand (Model D; Figure S1b).

To determine the microbial community, biomass samples were pretreated. This involved the extraction of bacterial genomic DNA using the Genomic Mini AX Bacteria + (A&A Biotechnology, Gdańsk, Poland), followed by its purification using an anti-inhibitor kit (A&A Biotechnology). The concentration of the bacterial DNA thus prepared was determined in a Qubit 4 fluorometer. The presence of bacterial genomic DNA was confirmed by real-time PCR in an Mx3000P thermocycler (Stratagene, La Jolla, CA, USA) using SYBR Green as a fluorochrome and universal 16S rRNA primers. According to 16S Metagenomic Sequencing Library Preparation Part # 15044223 Rev. B (Illumina, San Diego, CA, USA), amplicon libraries of the V3-V4 regions in the 16S rRNA gene were prepared, followed by a two-step PCR reaction using the Herculase II Fusion DNA Polymerase Nextera XT Index Kit V2. Library quality was verified according to the Illumina qPCR Quantification Protocol Guide. Sample libraries were loaded on the Illumina MiSeq platform and 2 × 300 bp reads were generated by Macrogen (Seoul, Republic of Korea).

2.2.2. 16S rRNA Gene Sequence Analysis

The V3–V4 regions in the 16S rRNA gene derived from Illumina sequencing were identified by analyzing sequence reads in the Greengenes v.13 database (97% similarity, minimum score 40). The resulting sequences were determined taxonomically, which consisted of assigning them at type level or lower ranks. To determine diversity indices such the Shannon-Wiener diversity index and Simpson index, CLC Genomic Workbench v. 12 (Qiagen, Venlo, The Netherlands) and Microbial Genomics Module Plugin v. 4.1 were used (Qiagen, Venlo, The Netherlands) [47].

2.2.3. The Adsorbed Biomass Analysis

The biomass determination involved measuring total suspended solids (TSS) and volatile suspended solids (VSS) adsorbed in the foams and Schmutzdecke layer. TSS were determined using the method described by Dacewicz and Grzybowska-Pietras [48]. Organic suspended solids were determined as per standard methods.

2.2.4. SEM Observations

Scanning electron microscopy (SEM) was used to analyze the morphology of biomass inhabiting the tested foam materials (JEOL JSM-5500 LV; JEOL Ltd., Tokyo, Japan/Lanameter MP-3 microscope). Prior to imaging, air-dried sponge samples were sectioned into 1.5 cm-thick elements using a microtome blade and sputter-coated with gold using a JEOL JFC-1200 ion coater. High-resolution images were captured with a cooled charge-coupled device (CCD) camera (Photometrics CH250, Tucson, AZ, USA).

2.2.5. The Nitrogen Removal Process

The ammonium nitrogen removal process was assessed based on the degree of its conversion to inorganic forms: nitrite nitrogen (I nitrification phase) and nitrate nitrogen (II nitrification phase). The efficiency of these phases was calculated based on the nitrogen balance and stoichiometry (theoretical values) according to Equations (1) and (2):

$$\mathrm{I} \% \mathrm{n}\mathrm{i}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}={\displaystyle \frac{{[\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}\left]\right(\mathrm{i}\mathrm{n}\mathrm{f})-{[\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}](\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e})}{{[\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}\left]\right(\mathrm{i}\mathrm{n}\mathrm{f})}}·100\%$$

(1)

$$\mathrm{I}\mathrm{I} \% \mathrm{n}\mathrm{i}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}={\displaystyle \frac{\left[{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}\right]\left(\mathrm{i}\mathrm{n}\mathrm{f}\right)-[{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}](\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e})}{[{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}]\left(\mathrm{i}\mathrm{n}\mathrm{f}\right)}}·100\%$$

(2)

where:

inf—influent of domestic sewage,

eff—effluent of domestic sewage,

N_x—concentration of inorganic nitrogen form [mg·dm⁻³].

Nitrate nitrogen removal (denitrification process) was calculated using the formula:

$$\% \mathrm{d}\mathrm{e}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}={\displaystyle \frac{{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}\left(\mathrm{i}\mathrm{n}\mathrm{f}\right)-{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}\left(\mathrm{e}\mathrm{f}\mathrm{f}\right)}{{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}\left(\mathrm{i}\mathrm{n}\mathrm{f}\right)}}·100\%$$

(3)

ARR and NRR rates were used to determine the efficiency of nitrogen removal and also in PCA. These rates were calculated using the formulas:

$$\mathrm{A}\mathrm{R}\mathrm{R}={\displaystyle \frac{{\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}} \left(\mathrm{i}\mathrm{n}\mathrm{f}\right)-{\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}\left(\mathrm{e}\mathrm{f}\mathrm{f}\right)}{{\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}\left(\mathrm{i}\mathrm{n}\mathrm{f}\right)}}·100\%$$

(4)

$$\mathrm{N}\mathrm{R}\mathrm{R}={\displaystyle \frac{{[\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}\left]\right(\mathrm{i}\mathrm{n}\mathrm{f})-{[\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}]\left(\mathrm{e}\mathrm{f}\mathrm{f}\right)}{{[\mathrm{N}}_{\mathrm{N}{\mathrm{H}}_{4^{+}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}}\left]\right(\mathrm{i}\mathrm{n}\mathrm{f})}}·100\%$$

(5)

NAR was used in PCA and calculated using the formula:

$$\mathrm{N}\mathrm{A}\mathrm{R}={\displaystyle \frac{{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}} }{{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{2^{-}}}+{\mathrm{N}}_{\mathrm{N}{\mathrm{O}}_{3^{-}}} }}·100\%$$

(6)

Determinations of inorganic forms of nitrogen were carried out in the wastewater samples. Standard procedures based on colorimetric methods using a WTW Photolab S12 spectrophotometer (Weilheim, Germany) were used to determine ammonium, nitrite, and nitrate nitrogen concentrations.

2.3. Statistical Analysis

In order to preliminarily analyze the microbial community that populated each layer of the columnar models, a statistical clustering of total sequence reads was performed. Cluster analysis was used in the calculations by Ward’s minimum variance method and the measure of similarity was the Euclidean distance. The groups thus extracted were the input variables used in statistical inference indicating differences between OTU’s determined for the 10 samples taken from successive layers of column models.

The second statistical tool used to describe the bacterial community and the ammonium nitrogen removal processes taking place with maximum information was PCA. Assessing the impact of the C/N ratio on the aerobic denitrification process in mixed cultures is not straightforward, as it runs in parallel with the conventional heterotrophic denitrification process. For this reason, an attempt was made to identify the links between the groups of microorganisms—whose growth was found in Models A–D—involved in the production of internal carbon sources and the bacteria characterized by metabolic activity associated with such a carbon source, using the example of endogenous denitrification. Included in the analysis were microorganisms associated with cellulose breakdown, the production of substances that are internal carbon sources for other clusters (EPS, VFA, PHA, PHB), and bacteria capable of cell lysis and/or those involved in ED. The type of bacteria that are capable of EPS synthesis were taken from the publicly available code database KEGG [51], and the abundance of these bacteria was summarized. Again, PCA analysis was used to identify factors common to the primary variables, determining the type of enzymes involved in the different stages of nitrogen conversion and the efficiency of these stages (I and II % nitrification, % denitrification). In addition, the NAR, ARR, and NRR were taken into account. The type of bacteria that possessed the desired type of enzyme (Amo, Hao, Nxr, Nar/Nap, Nir, Nor, Nos) was taken from a publicly available code database KEGG [28], and the abundance of these bacteria was summarized. Due to the preliminary cluster analysis, this second PCA was performed only for three-layer A and B models.

The Kaiser criterion (eigenvalue > 1) was used to determine the number of principal components in analysis [52]. The cumulative variance was calculated as the sum of the variances for the individual PC. The final step was a graphical presentation of the dataset, where each variable was represented by a vector, and its direction and length determined the extent to which the individual variables influenced the principal components. In this process, when the analyzed variables are located close to the circumference of the circle, most of the information in them is carried by the principal components. A strong positive correlation occurs when variables are located next to each other, whereas vectors of variables located perpendicular to each other indicate a lack of correlation. Arranging the variables on opposite sides with respect to each other means that they are negatively correlated.

To determine which intergroup differences in nitrogen reduction rates were statistically significant, the non-parametric Kruskal-Wallis test was applied. Differences between group means were considered significant at a p-value of less than 0.001.

The computer program Statistica 13 (Statsoft) was used for statistical calculations.

3. Results and Discussion

On the basis of the 16S rRNA gene sequence, the structure of the microbial community populating the sponge-filled multilayer filters (Models A, B, and C) and the sand-filled monolayer filter (Model D) was analyzed. A total of 613391 readings were analyzed (132,398 in sample A, 156,902 in sample B, 264,629 in sample C, and 59,462 in sample D). Using a 97% sequence identity cut-off point, 915, 966, 766, and 670 OTUs were detected, respectively (

Table 1. Diversity indices and microbial community structure at the phylum level in the four types of filters (A–D models).

	A	B	C	D
Total sequence reads	132,398	156,902	264,629	59,462
OTUs	915	966	766	670
Shannon diversity index	4.64	4.83	3.30	3.24
Simpson index	0.05	0.02	0.10	0.21
Dominant phyla (% of total sequence reads)
Proteobacteria alphaproteobacteria betaproteobacteria gammaproteobacteria	45.96	51.44	52.97	14.97
	29.96	31.98	27.82	8.98
	7.73	6.77	13.46	4.26
	7.62	12.18	10.88	1.14
Actinobacteria	16.81	17.63	24.82	51.30
Firmicutes	9.78	11.72	20.72	19.52
Chloroflexi	6.14	2.64	0.33	3.88
Acidobacteria	4.50	3.22	0.10	0.09
Planctomycetes	3.57	3.37	0.17	0.87
Bacteroidetes	3.14	3.48	0.19	1.79
Gemmatimonadetes	4.00	1.13	0.01	0.02

).

3.1. Microbial Community Analysis

3.1.1. Main Bacteria Phyla

In the samples analyzed, 39 bacterial phyla were identified. Of these, 24 were present in smaller quantities (less than 1% of the total reads in each sample). Table 1 summarizes the relative abundances of the 8 most abundant phyla, which were common to all samples taken from the four filters. The compositions of the microbial communities at OTU level differed markedly between the four filters tested, as evidenced by the Shannon and Simpson indices. The dominant phyla that were involved in carbon, nitrogen, and phosphorus cycling were Proteobacteria, Actinobacteria, Firmicutes, Chloroflexi, Acidobacteria, Planctomycetes, Bacteroidetes, and Gemmatimonadetes. Nomoto et al. reported that in DHS reactors with DHS G3 fill, the next most abundant phylotypes were Proteobacteria, Actinobacteria, Firmicutes, Chloroflexi, Planctomycetes, and Bacteroidetes [53].

Among the OTUs detected, members of the Proteobacteria phylum predominated in all sponge-filled samples, reaching approximately 50% of all reads. The phylum Proteobacteria is the most numerous group of aerobic, facultatively anaerobic, and aerobic bacteria, accounting for the majority of known bacteria of environmental importance. Representatives of this type are highly diverse in terms of metabolism and include chemoautotrophic, chemoorganotrophic, and phototrophic microorganisms capable of degrading a wide range of pollutants [54]. In wastewater treatment systems, Proteobacteria are identified as the most abundant phylum; they play an important role in the removal of organic matter and phosphorus from wastewater [55,56]. Proteobacteria also include nitrogen-fixing bacteria, nitrifying bacteria, and denitrifying bacteria. Literature reports on DHS reactors show that Proteobacteria were dominant and accounted for at least 30% of all OTU reads [57,58,59]. Maharjan et al. estimated the percentage of Proteobacteria phylum members in SpSF to be as high as 80% [60].

The subdominants in the sponge-filled models were Actinobacteria and Firmicutes, which were approximately 20% (Models A and B) and 10% (Model A) and 21% (Model B) of all OTUs, respectively. Their members, along with Proteobacteria and Bacteroidetes, formed the basis of the bacterial community in the biofilm [61]. Actinobacteria includes the typical types of aerobic bacteria that degrade organic compounds. Most Firmicutes phylum species are varieties of anaerobes and facultative anaerobes that are able to survive in extreme environments. This is an important group of bacteria whose members have the ability to break down dead bacterial cells and/or form endospores [62,63]. Their presence in WWTP is confirmed by numerous reports [5,56,64]. The fourth most abundant Chloroflexi represents a diverse group of organisms that includes anaerobic photoautotrophs, aerobic chemoheterotrophs, and facultatively anaerobic organisms. Depending on environmental conditions, members of this phylum are involved in the degradation of organic matter, nitrogen removal, decomposition of EPS, biofilm aggregation, and the lysis of dead bacterial cells [65]. Chloroflexi mainly use carbohydrates as substrates. These bacteria have the ability to assimilate glucose under aerobic and anaerobic conditions with NO₂/NO₃ as terminal electron acceptors. This is consistent with the fact that they behave as in situ denitrifying bacteria but cannot nitrify (non-denitrifiers) [66,67]. According to numerous reports in the literature, Proteobacteria, Chloroflexi, and Planctomycetes types are present in anammox reactors, where ammonia is oxidized to N₂ gas under anaerobic conditions [68,69]. Previous work by Dacewicz suggested that the presence of Planctomycetes indicated the occurrence of this type of reaction in the models studied, where the abundance of Planctomycetes in the spongy fill of Models A and B was at an average level of 3.5% of all OTUs, so the possibility of this process occurring in them cannot be excluded [47]. According to Huang et al., Acidobacteria and Bacteroidetes were also frequently found in anammox reactors [69]. Most Bacteroidetes species are anaerobic bacteria that are involved in a number of important metabolic activities, including the fermentation of carbohydrates and the utilization of nitrogenous substances. Members of this phylum are autotrophic bacteria that are involved in the breakdown of organic matter into simple molecules, i.e., proteins, starch, and lipids, through hydrolysis and fermentation. Bacteroidetes also includes some species of nitrogen-fixing bacteria that play an important role in the denitrification process [70,71]. Alongside Chlorofexi, certain Bacteroidetes break down dead cells and EPS, composed of polysaccharides and proteins, into simple organic molecules such as ethanol and lactate. These properties of Bacteroidetes and Chlorofexi allow them to grow under rigorous environmental conditions and indirectly to refresh bacterial communities [72]. Members of the Gemmatimonadetes phylum are aerobic species that show a diverse metabolism in different environments. These bacteria are abundant in soils, sludge, sewage treatment plants, and biofilms, participating in the carbon, nitrogen, and phosphorus cycles [73]. Some researchers have highlighted the potential contribution of Gemmatimonadetes to N₂O reduction from agricultural soils due to the transcription of the Nos gene [74,75]. The abundance of Gemmatimonadetes in Model A was 4% of all OTUs, so the possibility of this process occurring cannot be excluded. Furthermore, Gemmatimonadetes show flexible survival strategies under extreme environmental conditions at the expense of reduced growth [76].

The results of a 16S rRNA gene sequence analysis showed that the control sand filter (Model D) was populated with 51% members of the Actinobacteria phylum. The next leading phyla were Firmicutes and Proteobacteria. Chloroflexi and Bacteroidetes were less abundant. The latter are often observed in aerobic biological treatment systems such as sand filters [77]. As with the two-layer filter, Planctomycetes readings were at a low level. A negligible amount of bacteria belonging to Gemmatimonadetes phylum was also observed. This may have been related to their less active metabolism, which was associated with resistance to stressful sand conditions.

3.1.2. Main Bacteria Classes

Diversity analysis using Proteobacteria-specific 16S rRNA gene clone libraries showed that alpha-, gamma-, and betaproteobacteria were the predominant classes in the three-layer filters (Models A and B), accounting for about 30%, 8–12%, and about 7% of all bacterial reads, respectively (Table 1). Considering the two-layer filter (Model C), the alpha and gamma forms accounted for a similar number of reads, while the beta form proved to be almost twice as numerous. For the control sand filter (Model D), alpha-, beta-, and gammaproteobacteria forms were at 9%, 4%, and 1% of all bacterial readings, respectively. Maharjan et al., using a fill polyolefin sponge cubes as a dropping nitrification reactor, found that betaproteobacteria (37.9% all OTUs), alphaproteobacteria (24.2%), and Gammaproteobacteria (17.2%) were dominant at the top of the reactor, while alphaproteobacteria (16.2%), betaproteobacteria (15.1%), and Nitrospira (25.6%) dominated at the bottom [78].

As for the most abundant classes within Actinobacteria (Figure S2), Actinobacteria and Acidimicrobiia dominated in Models A and B. The other models appeared to be populated almost entirely by the Actinobacteria class. Firmicutes phylum was most abundant in two classes, i.e., Bacilli (more numerous in Model C) and Clostridia (more numerous in Model D). In the Planctomycetes phylum, Planctomycetia was the most abundant (more numerous in Models A and B). Nomoto et al. [53], Watari et al. [59], and Kubota et al. [79] observed a slight increase in Planctomycetes readings in a DHS reactor as its layers decreased. In their study, Chloroflexi was most abundantly represented by the class Anaerolineae, which included two classes: Anaerolineae and Caldilineae. Considering Bacteroidetes phylum, the predominant classes were Saprospirae, Cytophagia, Sphingobacteriia, and Flavobacteriia. The Acidobacteria population consisted mainly of the Chloracidobacteria class, and among the Gemmatimonadetes, the entire phylum was the Gemm-1 class.

3.1.3. Claster Analysis of All Bacterial Community

A cluster analysis was carried out to investigate the similarity of the bacterial communities in the different filter types and layers. Clustering performed for all OTU readings demonstrated three separate clusters (Figure S3). The first group formed three smaller clusters, which included the spongy upper (subgroup one) and lower layers (subgroup two) and the Schmutzdecke (subgroup three) of the three-layer filters (Models A and B). Group two consisted of the sponge lower layer and Schmutzdecke of the two-layer filter (Model C). Its spongy upper layer and the Schmutzdecke of the sand filter (Model D) formed the third cluster. The binding distance was estimated to be 4000.

3.2. Microbial Community of Sponges

Proteobacteria were most abundant in the sponge layers of the three-layer filters, accounting for between 28.72% and 73.26% (Model A) and between 46.74% to 66.88% (Model B) of all OTU sequences (Figure 1). The dominance of Proteobacteria in all sponge layers has also been reported by other authors in studies on DHS multilayer reactors [53,58,59,60].

Actinobacteria were the second most abundant type with OTU, ranging from 8.72% in A(LL) to 24.25% in A(UL) and from 6.80% in B(LL) to 16.90% in B(UL). The third most abundant OTU type was Firmicutes (0.87% to 26.72%), with OTU readings ranging from 0.87% in A(LL) to 16.72% in A(UL) and from 1.23% in B(LL) to 26.72% in B(UL). The fourth OTU type in Model A I B was Chloroflexi (from 2.10% to 9.15% and from 1.55% to 3.37%, respectively), followed by Acidobacteria (<6%), Planctomycetes (<5%), Bacteroidetes (<3.85%), and Gemmatimonadetes (<3.25%). In the lower part of the sponge fill of the models, there was an increase in Proteobacteria and a decrease in Actinobacteria and Firmicutes. Nomoto et al., who analyzed the microbial community in a DHS reactor with DHS G3 fill, also noted such a relationship for Proteobacteria [53]. A decrease in Firmicutes readings with decreasing layers in DHS reactors was found by Nomoto et al. [53], Watari et al. [59], and Kubota et al. [79].

For Model C with a single layer of PUR foams, the highest reads were recorded for Proteobacteria, Actinobacteria, and Firmicutes. For UL and LL, readings of OTU were from 23.79% up to 67.31%, from 60.77% to 15.28%, and from 14.34% to 16.19%, respectively. Compared to Models A and B, the remaining phyla accounted for less than 0.83%. The amount of the organic biomass (as VSS) produced appeared to be almost six times lower in the fill with casing in the three-layer filter (Model A) compared to the other models (Figure 1). After 420 days of filters operation, the biomass content adsorbed on the foams ranged from 0.1 to 0.4 g·dm⁻³. VSS at this level constituted 25 to 60% of TSS. These values are similar to those given by Dacewicz and Grzybowska-Pietras [48]. After 400 days, the Model D became clogged. The study was continued for another 20 days, but the clogging—indicated by a VSS concentration of 8.44 g·dm⁻³—proved to be irreversible. Figure 1 presents SEM images of the foams taken at 2000× magnification. Flocculent microbial biomass was observed in the foam matrices and the Schmutzdecke of all models (Figure S1c).

3.3. Microbial Community of Schmutzdecke

In Models A and B, Proteobacteria were most abundant in the Schmutzdecke layer at the type level, accounting for ca. 35% of all OTU sequences (Figure 1). Actinobacteria were the second most abundant type, with OTU readings up to 33.77%. The third most abundant OTU type was Firmicutes (up to 10.90%), followed by Chloroflexi (up to 6.88%), Acidobacteria (up to 9.05%), Planctomycetes (up to 6.68%), Bacteroidetes (up to 6.25%), and Gemmatimonadetes (up to 0.63%). Other authors have also reported the dominance of Proteobacteria, Planctomycetes, Acidobacteria, and Bacteriodetes in their studies on sand filters [77,80]. The Proteobacteria cluster had the highest relative abundance in almost all filters, which may be linked to the ability of some of its members to thrive in environments with low concentrations of dissolved organic carbon.

The prevalence of Proteobacteria and Firmicutes was similar in the Schmutzdecke samples of Models A and B. For Chloroflexi and Planctomycetes, the differences in relative abundance between the two samples were more pronounced, in favor of Model A, where the % reads of these groups were about twice as high. An inverse relationship was found for Actinobacteria and Bacteroidetes. Proteobacteria were most abundant in the Schmutzdecke layer of Model C at the type level, accounting for 67.91% of all OTU sequences (Figure 1). According to Maharjan et al. [60], members of the Proteobacteria accounted for up to 80% of the bacterial community detected in SpSF. Firmicutes was the second most numerous type, with OTU readings of 28.53%. The remaining phyla accounted for 2.4%. Actinobacteria, Firmicutes, and Proteobacteria were most abundant in the Schmutzdecke layer of Model D (control filter) at the type level, accounting for 51.3%, 19.52%, and 14.97% of all OTU sequences, respectively. The remaining phyla accounted for 6.6%.

The amount of biomass (as TSS and VSS) produced in the Schmutzdecke of Model B was the lowest. In the other filters, it was about 50–100% higher (Figure 1).

3.4. Main Bacteria Families

The results of the microbiological analysis with microbial detection rates of 0.1% or higher are presented by heatmap separately for sponge and sand fill (Figure 2). There were 21 groups of bacteria in the rank of families in all models.

3.4.1. Microbial Community of Sponge

At the family level, the subdominants in the sponge fillings of all models comprised two groups: Xanthomonadaceae and Burkholderiaceae of the Proteobacteria phylum and Bacillaceae and Clostridiaceae of the Firmicutes phylum (Figure 2).

Based on reports in the literature, it is known that the Xanthomonadaceae families include heterotrophic nitrifying bacteria that are also capable of aerobic denitrification [81]. These organisms are also known for their ability to fully or partially denitrify under anaerobic conditions. Bacteria belonging to the Xanthomonadaceae family are involved in microbial biofilm formation, as they are involved in EPS production [82]. Members of the Bacillaceae family play a major role in the cycling of organic matter. They have the ability to form endospores, which provide them with a high degree of resistance, allowing them to survive in adverse environmental conditions for extended periods of time [83]. Many Clostridiaceae organisms are gut microbiota and pathogens. Their members also include anaerobic microorganisms with cellulolytic activity and VFA producers. In addition, Clostridiaceae are extremely important for the breakdown of biomass into simpler components [84]. The Burkholderiadaceae and Alcaligenaceae families include known denitrifying phylotypes [85]. As can be seen in Figure 2, in the double sponge fill of Models A and B, Clostridiaceae (from 0.14% to 11.85% and 0.24% to 10.77%, respectively for LL and UL) and Xanthomonadaceae (about 4% in UL and LL of Model A and up to 10.67% in LL of Model B) were the most abundant. Burkholderiaceae and Caldilineaceaewere were half as numerous.

Reads of the families Caulobacteraceae, Ellin 6075, Sphingomonadaceae, Cytophagaceae, Bradyrhizobiaceae, and Chitinophagaceae were up to 3%. Reads of the families Alcaligenaceae, Comamonadaceae, Lactobacillaceae, and Planctomycetaceae were up to 2%. Reads of Nitrospiraceae (<0.70%) and Nitrosomonadaceae (<0.80%), Moraxellaceae, and Flavobacteriaceae were found to be the lowest. Belonging to the Bacteroidetes phylum, members of the Cytophagaceae exhibit hydrolytic activity [86,87] and can participate in the hydrolysis of complex organic compounds present in wastewater flowing into the WWTP [88]. Organisms belonging to the Comamonadaceae family are anaerobic denitrifiers that are commonly found in wastewater treatment systems and wetlands [89,90].

In the single sponge fill of Model C, Xanthomonadaceae (from 5.80% in UL up to 12.12% in UL), Burkholderiaceae (from 2.17% in UL up to 7.86% in UL), Corynebacteriaceae (from 0.04% in UL up to 7.29% in UL), and Lactobacillaceae (from 1.24% in UL up to 6.15% in UL) were the most abundant. Alcaligenaceae, Caldilineaceae, Caulobacteraceae, Clostridiaceae, and Sphingomonadaceae were less abundant by 2–3 times. Readings for Enterobacteriaceae, Moraxellaceae. Bradyrhizobiaceae, Chitinophagaceae, Comamonadaceae, and Ellin 6075 were the lowest. No Cytophagaceae were found, and the abundance of Nitrospiraceae, Nitrosomonadaceae, Planctomycetaceae, and Flavobacteriaceae did not exceed 0.07% of all OTU sequences.

A higher proportion of Xanthomonadaceae was evident in the lower parts of the sponge fill, which may suggest that nitrification and denitrification processes took place there. In the upper layer of Model A, the percentage of Caldilineaceae (Chloroflexi phylum) and Cytophagaceae was the highest, at 4.29% and 2.48%, respectively. In the other layers, it did not exceed 0.7%. Notably, o_Bradhybacteriaceae have the Nar, Nir, Nor and Nos genes [28,29]. Their high abundance in hypoxic or anaerobic layers in Model B may have influenced the conversion of nitrate to gaseous nitrogen. Members of the Alcaligenaceae, which were shown to be most abundant in the lower spongy parts of the fill without casings (Models B and C), are able to assimilate nitrate and reduce it to nitrite. Representatives of the Chitinophagaceae and Lactobacillaceae are capable of synthesizing extracellular polymeric substances (EPS) [51,91,92,93]. Their abundance was greatest in the lower layers of Models A and B and the upper layers of Models B and C, respectively. Garibay et al. analyzed a subtropical high strength domestic wastewater treatment plant and found that the families responsible for nitrogen removal included Xanthomonadaceae, Caulobacteraceae, Bradyrhizobiaceae, and Nitrospiraceae [94]. A study by Maheepal et al. found that the sponges filling a DHS siphon were populated mainly by autotrophic denitrifying bacteria from the Comamonadaceae and Rhodocyclaceae families [95]. The authors also observed the presence of nitrogen-fixing bacteria from the Xanthobacteraceae and Rhizobiaceae families in the samples. Loi et al. reported that the most abundant families in DSB reactors were Burkholderiaceae and Chitinophagaceae [96]. Petrilli et al. showed that bacteria from the Alcaligenaceae, Nitrosomonadaceae, Caulobacteraceae, Xanthomonadaceae, Comamonadaceae, and Chitinophagaceae families were mainly responsible for ammonia oxidation during landfill leachate treatment [97].

In the present study, the conditions in the upper sponge layer of Models A and B favored the growth of Clostridiaceae. Such a high abundance of members of this class in the sponge filling and the cell lysis that occurred due to their presence resulted in the production of significantly less biomass due to the conversion of a large number of cellular substances into carbon dioxide [84].

3.4.2. Microbial Community of Schmutzdecke

At the family level, the subdominants in all Schmutzdecke layers of Models A and B were Clostridiaceae, Xanthomonadaceae, and Burkholderiaceae (Figure 2). In addition to these, Ellin 6075, Nitrospiraceae, Caldilineaceae, and Cytophagaceae dominated Model A. In Model B, these bacteria were from the Bacillaceae, Caulobacteraceae, and Sphingomonadaceae, and in Model C, they were from the Alcaligenaceae, Bacillaceae, Sphingomonadaceae, and Bradyrhizobiaceae. In Schmutzdecke A and B, an increase in anaerobic Clostridiaceae (up to 9.05% and 3.35%, respectively), carrying out lysis of dead bacterial cells, was observed compared to the lower sponge layer. A 2–3 fold increase in Caldilineaceae, involved in endogenous denitrification, was also evident. In contrast, the abundance of members of the Xanthomonadaceae, Burkholderiaceae, Planctomycetaceae, Sphingomonadaceae, and Comamonadaceae decreased significantly. In the Schmutzdecke layer of Models A, B, and C, where the external carbon source content was low, there was increased metabolic activity of bacterial groups involved in the decomposition of EPSs and the production of PHB, thus providing internal carbon sources for ED-capable clusters.

In the case of Model D, the largest proportions were Clostridiaceae, Bacillaceae, and Bradhyrhizobiaceae (together about 11% of all OTUs). The remaining families did not exceed 0.90%. The observed high abundance of Bradyrhizobiaceae at 1.52% could indicate the conversion of nitrate to nitrogen gas. Additionally, Bradyrhizobium is known to produce EPSs [98], which may have adversely affected the performance and clogging of the control sand filter. EPSs may significantly reduce the efficiency of the filters as the biomass reduces the pore space and clogged the filter [99].

3.5. Bacteria Involved in the Removal of Organic Carbon

In the analyzed filters, Xanthomonadaceae and Clostridiaceae were found to have the highest abundance among the bacteria involved in the removal of organic carbon (over 10%) (Figure 2). Garibay et al. analyzed a high strength domestic wastewater treatment plant and found that the families responsible for organic matter removal included Xanthomonadaceae, Chitinophagaceae, and Cytophagaceae [94]. Petrilli et al. showed that during landfill leachate treatment, bacteria from the Xanthomonadaceae, Cytophagaceae, Chitinophagaceae, and Alcaligenaceae families were mainly responsible for organic carbon removal [97]. Bacterial strains of Chitinophagaceae and Cytophagaceae are capable of hydrolyzing complex organic compounds present in wastewater, e.g., complex compounds such as proteins, lipids, and polysaccharides [88,92]. As Figure 2 shows, the abundance of these latter families in the present study was approximately 2%.

The presence of organic matter in the form of cellulose, a component of dietary fiber, waste tissues, and toilet paper, stimulated the growth of anaerobic microorganisms of the Clostridiaceae family and facultatively anaerobic microorganisms of the Cytophagaceae family in the biomass [100]. The degradation of cellulose generally occurs via aerobic or anaerobic organisms, and there is a marked difference in cellulolytic strategy between these groups of bacteria. Cytophaga utilize cellulose via extracellular cellulase enzymes [51,86,87]. Anaerobic Clostridia degrade cellulose mainly via complex cellulase systems located directly on the cell surface. The products thus formed are transported across the cytoplasmic membrane into the cytoplasm. This type of strategy makes sense in a highly competitive environment, as it helps prevent the loss of polysaccharide breakdown products to competing bacteria [100]. Bacteria of the class Clostridiaceae also have an important role in the degradation of dead bacterial cells into nutrients, which are then used as a carbon source by other microorganisms. Cytophages alongside clostridia are able to withstand long periods of nutrient limitation [101], which may explain their high abundance under low C/N wastewater treatment conditions. The highest percentage of clostridia and cytophage populations was found in the upper layers of Models A and B (Figure 2), where dissolved oxygen concentrations of 4 mgO₂·dm⁻³ were observed [49,50]. As reported by Morvan et al., members of the Clostridiaceae are able to cope in an aerobic environment because they redirect their central metabolism to pathways less sensitive to O₂ and induce the expression of genes encoding enzymes involved in its reduction [100]. Oxygen tolerance by Clostridiaceae species ranges from relative insensitivity to extreme sensitivity. One well-documented example is that the high respiratory activity of aerobic bacteria can provide an anaerobic microenvironment in which obligate anaerobic bacteria could grow [102]. This may indicate that under anaerobic conditions, Clostridiaceae and Cytophagaceae members were responsible for the hydrolysis of cellulose. Maheepala et al. suggested that in the sponges that populate DHS siphons, Cytophagaceae were responsible for degrading organic matter and using the products of this process as a carbon source for energy production [95]. Representatives of the Cytophagaceae are commonly known as the microbiome of wastewater treatment plants [86,103].

Heterotrophic bacteria belonging to the Xanthomonadaceae family are capable of synthesizing extracellular polymeric substances (EPSs), the main component of biofilm matrices [104]. EPSs are proteins, lipids, and exopolysaccharides that are produced by bacteria from organic matter present in the substrate or by lysis of dead cells. Bacterial EPSs, which are secreted into the surrounding environment, play a significant role in exogenous stress resistance, often serving as a carbon or energy source under nutrient-limiting conditions. EPSs are the main structural components of the biofilm matrix, sludge flocs, and granules, which are commonly found in biological wastewater treatment systems [82]. Members of families belonging to the Bacteroidetes and Chlorofexi phyla are capable of breaking down EPSs and converting macromolecular organic matter into simple organic molecules that can be used by other species in their metabolism [72]. Belonging to the Chlorofexi phylum, the filamentous bacteria Caldilineaceae are frequently observed in wastewater treatment plants, where they are involved in the carbon and nitrogen cycle [105,106].

Kubota et al., in a study of wastewater treatment with a low ammonium nitrogen content of 30 mg·dm⁻³ (C/N in the range 4–5), suggested that microorganisms in the upper boxes of a DHS reactor played an important role in the removal of organic matter [79]. As shown by the results presented in the present author’s previous, such a relationship was not found due to the treatment of raw wastewater with a low external carbon content (C/N in the range 1–2) [49,50]. In the upper layer of Model A, an average COD removal efficiency of 42% was observed. For Model B, the average COD removal efficiency was 11% lower. COD removals of 19% and 30% were found for the bottom layer of these models, respectively. Table S2 summarizes the removal efficiencies of COD for Models A–D. The two-layer filter (Model C) showed the highest COD removal efficiency (about 80%), differing significantly from Model B. The control sand filter (Model D) also achieved a similarly high COD removal efficiency of 77% and also differed significantly from Model B. The three-layer filters (Model A and B) were characterized by lower COD removal efficiencies, amounting to 67% and 51%, respectively. The difference between these models was not statistically significant.

In order to understand the mechanisms of organic carbon removal, an attempt was made to identify its internal sources. Associations between groups of bacteria, characterized by metabolic activity associated with an endogenous carbon source, were assessed using PCA. Included in the analysis were micro-organisms associated with cellulose breakdown, the production of substances that are internal carbon sources for other clusters (EPS, VFA, PHA, PHB), and bacteria capable of cell lysis and/or involved in endogenous denitrification (ED). Using the scatterplot test, the four main factors with the highest factor loadings relative to the component data (PC 1 to PC 4) were identified. Analysis of the eigenvalues of the correlation matrices showed that the first, second, third, and fourth principal components explained 32.13%, 62.79%, 77.20%, and 88.45% of the total variance of the primary variables, respectively (

Table 2. Eigenvalues of the correlation matrix (PC related to bacterial community, endogenous carbon source, and metabolic processes).

Principal Component	PC 1	PC 2	PC 3	PC 4
Eigenvalue	5.14	4.90	2.30	1.80
% cumulative variance	32.13	62.79	77.20	88.45

).

Table S3 shows the factor coordinates of the variables analyzed by the four principal components. PC 1 was highly correlated with Agrobacterium, Cupravidus, Burkholderia, o_Cytophagaceae, and o_Caldilineaceae strains, i.e., those associated with endogenous denitrification. PC 2 was found to be correlated with Pseudomonas, as well as with clostridia and bacterial cell lysis and VFA production. The presence of o_Comamonadaceae in this factor is indicative of endogenous denitrification using previously accumulated PHAs [37]. PC 3 was formed by Bacillus and Brevundimonas diminuta bacteria responsible for EPS synthesis [107,108]. PC 4 was shown to be highly correlated with o_Lactobacillaceae and EPS production.

The relationships between the primary variables and PC 1 and PC 2 are shown graphically in Figure 3.

A very strong positive correlation was found between the bacteria responsible for PHA/PHB synthesis and those involved in endogenous denitrification (highlighted by red ellipse). Also visible is a very strong positive correlation between bacteria involved in cellulose breakdown, VFA production, and cell lysis (highlighted by blue ellipse). Members of o_Lactobacillaceae and Brevundimonas diminuta and the production of EPSs are positively related (indicated by the green ellipse). A close proximity on the same side of the principal components area of the bacteria o_Comamonadaceae and Pseudomonas was also observed. The bacterial group o_Cytophagaceae was not correlated with PHA/PHB production, as evidenced by their perpendicular alignment to each other.

3.6. Bacteria Involved in Nitrogen Removal

As Figure 4 shows, the bacterial groups present in the microbiome responsible for the nitrogen cycle comprised Proteobacteria (Achromobacter, Acinetobacter, Agrobacterium, Brevundimonas diminuta, Burkholderia, Cupriavidus, Halomonas, Hydrogenophaga, Nitrosomonas, Pseudomonas, Sphingomonas, Sphingopyxis, Sulfonivorans, Thauera, Thiobacillus, and Zoogloea), Actinobacteria (Arthrobacter, Corynebacterium, Gordonia, Rhodococcus, and Streptomyces), Firmicutes (Bacillus, Cereus, and Staphylococcus), Bacteroidetes (Chryseobacterium, o_Flavobacteriaceae, and Niabella), Nitrospirae (Nitrospira), Planctomycetes, and Chloroflexi (o_Caldilineaceae).

Proteobacteria, accounting for almost 40% of the relative abundance of all biomass samples taken from the models, had the highest abundance. The other dominant groups of bacteria were Actinobacteria and Firmicutes, which accounted for 10.1% and 9.6% of all OTUs, respectively. The abundances of Chloroflexi, Planctomycetes, Bacteroidetes, and Nitrospirae were at the lowest level, at 5.74%, 4.6%, 1.9% and 1.7%, respectively. Proteobacteria and Bacteroidetes are among the dominant bacteria in many denitrification environments and are its main promoters [109]. Chloroflexi such as Caldilineaceae may play a significant role in nitrate reduction in activated sludge reactors because they have Nar genes [28]. However, these bacteria are characterized by the absence of Nir genes, which are responsible for nitrite reduction. Nor and Nos genes related to the conversion of NO by N₂O to N₂ were found in several taxa of this group. Caldilineaceae are therefore capable of partial denitrification and, as non-denitrifiers, may play an important role in the nitrogen cycle and N₂ production [110,111].

Two classes and 19 orders of HNAD bacteria involved in ammonium nitrogen removal processes were identified in the microbiome. A detailed analysis showed that 12 of these were present in all biomass samples (Figure 4). The highest abundance of common bacterial phylotypes was obtained in Models B and C (foam without casing), characterized by the growth of HNAD bacteria, which together accounted for about 30% of the total abundance. The types of bacteria with nitrifying activity, namely, AOB, comammox, and HNAD bacteria, were present in the sponge fill biomass at a level about 40% (Figure 5), while their number in Model D (the control sand filter) was 10 times lower (Figure S4). In addition, members of denitrifying families, such as HNADs, TDENs, and non-denitrifiers, were detected, comprising between 30.2% (Model A) and 46.1% (Model C) in the present study. Their abundance in Model D did not exceed 4%.

3.6.1. Microbial Community of Sponge

Within the sponge fill, Nitrosomonas (AOB) and Nitrospira (comammox bacteria) were identified in all biomass samples taken from Models A and B (Figure 4). Nitrosomonas was present in the sponge fill and tended to increase in abundance in the lower part. Nitrospira prevailed over Nitrosomonas in the biomass of most samples. The presence of AOB and comammox bacteria, involved in traditional nitrification, was influenced, among other things, by the DO concentration inside the models, at around 4 mgO₂·dm⁻³, and the FA content. For the upper sponge layer of Model A, the relative abundances of these bacteria were 0.23% and 0.30%, respectively. As reported by Dacewicz [47,50], there was a high increase in N-NO₂⁻ in the fill in the casings. The FA concentration of 1.8 mg·dm⁻³ exceeded that reported by Anthonisen et al., thereby limiting Nitrospira activity [112]. The efficiencies of Stages I and II of nitrification were 37.2% and 82.6%, respectively. The presence of HNAD bacteria, mainly of the genera Bacillus and Achromobacer, indicated that they were actively involved in the removal of N-NH₄⁺ ions. On the other hand, HNAD bacteria were responsible for the aerobic denitrification process. At the same time, traditional anaerobic denitrification by TDEN (mainly Thiobacillus) took place inside the sponge material in its anaerobic microvessels. Most of the previously mentioned studies of DHS reactors confirmed that nitrification occurred in aerobic zones on the surface of the sponge material, where oxygen diffused from the atmosphere, while in the depths of the sponges, an environment suitable for conducting anaerobic denitrification was created [113,114]. In the present study, Burkholderia bacteria, capable of endogenous denitrification, further influenced nitrate removal, as evidenced by their 1.73% and 91% efficiency in this process. Numerous non-denitrifiers from the 4.29% o_Caldilineaceae, using Nar gene expression, facilitated the conversion of nitrate to nitrite. This was evidenced by 70% NAR and the Y_NO2/NH4 ratio of 0.26, which was twice as high as the Y_NO3/NH4 ratio [47,50]. Due to the absence of the Nir gene, these bacteria were not involved in further nitrite conversion, but the Nor and Nos genes allowed them to convert N₂O to N₂.

In the middle part of Model A, the nitrification Stage II led by Nitrospira bacteria (0.79%) was superior to the ammonium nitrogen oxidation stage by Nitrosomonas (0.44%). An FA concentration of 0.2 mg·dm⁻³ could partially inhibit the activity of these bacteria [47,50]. Taking into account the additional activity of the less abundant HNAD bacteria, it was found that Phase I and Phase II nitrification occurred at similar levels to the top of the Model (38.5% and 77.6%, respectively). The Y_NO2/NH4 ratio value of 0.35 was found to be 14% lower than the Y_NO3/NH4 ratio, indicating the presence of a different form of nitrite removal [47,50]. The Planctomycetes present in this layer, with an abundance of 1.46%, were able to use N-NH₄⁺ as an electron donor in the presence of N-NO₂⁻. NAR decreased to approximately 53%, and the nitrate removal efficiency carried out by the more abundant bacteria in this layer HNAD (Rhodococcus), ED (o_Comamondaceae) and TDEN (Sphingopyxis) was found to be at the lower 55% level. This may have been influenced by the 6.5 times lower abundance of o_Caldilineaceae bacteria in this layer, as these bacteria are involved in nitrite conversion to the exclusion of nitrate. The main role of Burkholdria, whose abundance was as high as 5.06%, may not have been in the reduction of nitrate but in the conversion of NO by N₂O to N₂.

In the upper layer of Model B, due to the high FA value of 3.2 mg·dm⁻³, there was an inhibition of Nitrosomonas and Nitrospira activity, affecting a small amount (0.01% and 0.08%). In addition, HNAD bacteria were found to be significantly more abundant compared to Model A, causing them to displace AOB and comammox. Bacillus and Achromo-bacer accounted for 6.10% and 1.58%, respectively. The efficiencies of Stages I and II of nitrification were 20.0% and 68.5%, respectively. ARR may also have been influenced by the presence of Planctomycetes with an abundance of 0.61%. NAR reached a high value of 80%, and the calculated Y_NO2/NH4 ratio of 0.2 was almost three times the value of the Y_NO3/NH4 ratio [47,50]. Under anaerobic conditions inside the sponge material without casings, the amount of TDEN was 2.55% for Brevundimonas diminuta and 0.82% for Sphingopyxis. Compared to Model A, the number of Burkholderia was high, at 3.21%, while the number of o_Caldilineaceae was lower (1.78%). The high abundance of Burkholderia could indicate the conversion of N₂O to nitrogen gas. This bacterial community allowed denitrification to reach a high level of 88%.

The predominance of Nitrosomonas over Nitrospira was found for the central part of Model B, where their abundance was 0.88% and 0.53%, respectively. The increase in AOB and comammox bacteria was influenced by the reduction of FA to 0.98 mg·dm⁻³. The Y_NO2/NH4 ratio value of 0.29 was slightly higher than the Y_NO3/NH4ratio, suggesting an additional form of nitrite removal [47,50]. The predominantly Planctomycetes detected in this layer (2.03%) were responsible for N-NH₄⁺ removal under high NAR conditions at approximately 60%. A high removal rate of 79% was found for nitrate. Bacteria o_Comamondaceae (1.85%), Sphingopyxis (1.54%), and Burkholderia (1.46%) were responsible for nitrate reduction in this layer, as the abundance of o_Caldilineaceae was only 0.23%. The efficiencies of Stages I and II of nitrification were 32.8% and 76.0%, respectively.

Considering Model C, it was found that AOB and comammox abundance was very low (<0.01%). In this case, nitrifying bacteria were displaced by aerobic heterotrophic bacteria Corynebacterium, Achromobacter, and Gordonia. Overall, the percentage of Proteobacteria HNAD and TDEN in the total number of bacteria associated with the nitrogen cycle was highest in Model C (up to 15%) compared to the values found in Models A and B, where it reached approximately 10% and 12%, respectively (Figure 5).

3.6.2. Microbial Community of Schmutzdecke

As Figure 4 shows, the four dominant bacteria in Schmutzdecke Models A and B were Rhodococcus, Plancomycetes, Streptomyces, and Burkholderia (in Model C, Burkholderia, Achromobacter and Agrobacterium were most abundant). Nitrospira in Schmutzdecke A reached an abundance of 2.38%, but in Schmutzdecke B and C it was displaced by HNAD bacteria, accounting for 12.80% and 17.48%, respectively (Figure S4).

The important role of Nitrospira in the oxidation of ammonia in rapid sand filters was highlighted by Wang et al., who found that the bacteria accounted for an average of 64.0% of nitrifiers [115]. The percentage of Proteobacteria HNAD and TDEN in the total number of bacteria associated with the nitrogen cycle was highest in Model C (up to 17.49%). In Model D, the number of bacterial reads was the lowest, with Burkholderia predominating (up to 1.51%).

In Model A, the efficiencies of Stage I and stage II nitrification were 43.7% and 78.0%, respectively. In Model B, these values were higher, reaching 51.3% and 84.8%, respectively. The abudant presence of HNAD bacteria in Model A (o_Comamonadaceae, Rhodococcus, Bacillus, Achromobacer, and Niabella), in Model B (Bacillus, Rhodococcus, Achromobater, o_Comamonadaceae, and Sphingomonas), and in Model C (Achromobater, Agrobacterium, Corynebacterium, Sphingomonas, and Cupriavidus) indicated that they were actively involved in the removal of N-NH₄⁺ ions. On the other hand, HNAD bacteria were responsible for the aerobic denitrification process. The efficiencies of denitrification driven mainly by TDEN in Model A, by HNAD in Model B, and by HNAD and ED in Model C were 55.0%, 76.0%, and 91.8%, respectively. HNAD bacteria were present in the control sand filter (Model D) at 10 times lower levels than in Model C, which translated into low ammonium nitrogen removal efficiency. The presence of heterotrophic bacteria and high biomass content (Figure 1) resulted in clogging of the control filter after 400 days of research. Chen et al. reported that biofilm clogging, comprised predominantly of heterotrophs, occurred in the top layers of the intermittent sand filter [99].

3.6.3. Nitrogen Removal Efficiency

Table S2 summarizes the removal efficiencies of non-ionic nitrogen forms—such as NH₄⁺-N and N_in—for Models A–D. The two-layer filter (Model C) showed the highest nitrogen removal efficiency (above 90%) and differed significantly from the other models. The three-layer filters (Model A and B) were characterized by lower NH₄⁺-N removal efficiencies, amounting to 77% and 69%, respectively. The difference between these models in the removal of non-ionic nitrogen forms was not statistically significant. The lowest removal efficiency, at 60%, was observed for the control sand filter (Model D), which was clogged.

The identification of a distinct cluster for the three-layer filters (Section 3.1.3) prompted a comparison of Models A and B in terms of the ammonium nitrogen removal process. PCA was carried out to determine the factors common to the primary variables, specifying the types of enzymes involved in the different stages of nitrogen conversion and the percentage efficiencies of these processes (Stages I and II of nitrification, denitrification), NAR, ARR, and NRR from the wastewater. Using the scatterplot test, the three main factors with the highest factor loadings relative to the component data (PC 1, PC 2, and PC 3) were identified. Analysis of the eigenvalues of the correlation matrix showed that the first, second, and third principal components explained 67.90%, 83.20%, and 93.83% of the total variance of the primary variables, respectively (

Table 3. Eigenvalues of the correlation matrix (enzymes and stages of nitrogen removal).

Principal Component	PC 1	PC 2	PC 3
Eigenvalue	8.83	1.99	1.38
% cumulative variance	67.90	83.20	93.83

).

Table S2 shows the factor coordinates of the variables analyzed by the three principal components. PC 1 was shown to be very highly correlated with Nor and highly correlated with the remaining variables defining the type of enzyme (Amo, Hao, Nxr, Nar/Nap, Nir, and Nos), with the efficiency of the I nitrification and denitrification processes carried out by the microorganisms and with the nitrite content of NAR. The very strong correlation of PC 1 and Nor (at the level of 0.99) resulted from the presence of Burkholderia in all layers of Models A and B, with a total amount of 7.91% and 5.99%, respectively (Figure 4). The abundance of Burkholderia in the bacterial community could indicate the conversion of NO to N₂O and gaseous nitrogen. It was also found that PC 2 was correlated with the ARR variable, while PC 3 was only correlated with the % II nitrification variable (Table S4).

The relationships among the primary variables and PC 1, PC 2, and PC 3 are shown graphically in Figure 6.

A very strong positive correlation was found between the efficiency of nitrification Stages I and II and ARR and NRR (indicated by the green ellipse). Also evident was a strong positive correlation between the enzymes Amo, Hao, Nxr, Nar/Nap, Nir, and Nor, which are responsible for the conversion of ammonium nitrogen to nitrate and N₂O (indicated by the blue ellipse). The Nos enzyme, the nitrogen removal efficiency of denitrification (% denitrification), and NAR were positively related, as evidenced by their close proximity on the same side of the principal components area (indicated by the red ellipse). It can also be seen in Figure 5 that the variables NAR and % I nitrification and ARR are on opposite sides of each other, indicating their negative correlation. This was probably due to the fact that nitrite was consumed not only by comammox bacteria but also by pre-HNAD and anammox-like bacteria for its metabolic transformation.

3.7. Bacteria Involved in Phosphate Removal

Phosphate-removing bacterial strains were also identified in sponge fills (Figure S5). Models A and B showed the highest number of reads for Pseudonocardia and Amycolatopsis from o_Pseudonocardiaceae (Actinomycetes phylum), together accounting for 4.86% and 7.91% of all OTUs, respectively. These are bacteria capable of redundant phosphate accumulation [116] or proteomic degradation of microbial phosphate under environmental stress [117]. The upper spongy fill in the casings of Model A showed 1.6 times lower percentages of Pseudonocardia and Amycolatopsis than in the fill without casings, while the lower layer showed 2.4 times higher percentages. These strains were not observed in Models C and D. Bacteria from the Proteobacteria phylum were also present, i.e., Devosia, Acinetobacter and Bdellovibrio. In the sponge filling of Model B, 3.2 times the percentage of Devosia bacteria were found. In Models C and D, they were in negligible quantities. All of these micro-organisms accounted for 7.97%, 11.20%, and 0.19% of the total reads in the sponge fills of Models A, B, and C, respectively. Zuo et al. showed that Devosia sp. and Bdellovibrio sp. were dominant in aerobic phosphorus uptake, while Acinetobacter sp. played a dominant role in anaerobic phosphorus release [118]. The Schmutzdecke layer of Model A was dominated by Pseudonocardia and Amycolatopsis, while Model B was dominated by Devosia (Figure S6). The total contents of phosphate-removing bacteria in Models A and B were 1.30% and 1.89%, respectively. Their very low abundances in Models C and D (0.02% and 0.11%, respectively) indicated that, under conditions of a double sand layer, phosphate removal was mainly related to phosphate sorption. The two-layer filter (Model C) and the control filter (Model D) exhibited the highest PO₄³⁻ removal efficiencies (approximately 55%) and differed significantly from the three-layer filters (Models A and B), which demonstrated removal efficiencies that were about five times lower (Table S2). This difference was likely related to the higher amount of sand in Models C and D, which enhanced phosphate sorption.

4. Conclusions

The results of analyses using advanced DNA sequencing techniques on the bacterial community that populated multilayer sponge-sand filters confirmed the high diversity of microorganisms involved in the treatment of domestic sewage with elevated ammonium nitrogen. The removal of ammonium nitrogen was shown to occur via different pathways in foam-sand filters. This suggested that a multi-species community strategy based on the coexistence of AOB, comammox, anammox, and HNAD could effectively improve nitrification and denitrification efficiency. In the case of three-layer filters, NH₄⁺-N removal was found to be at a similar level. Model A achieved 77% removal of this pollutant and supported a higher abundance of Nitrosomonas and Nitrospira in each layer (average 1.5% in all OTUs). HNAD bacteria and Planctomycetes were at an average level of 4.5% and 1.4% in all OTUs, respectively. Overall, the foams with casing provided favorable conditions for the growth of ammonium nitrogen-removing bacteria through the traditional autotrophic nitrification/anaerobic denitrification process (both heterotrophic and endogenous), as well as anammox bacteria. The foams without casing, in the case of the three-layer filter, created conditions suitable for 69% removal of NH₄⁺-N due to the coexistence of nitrogen-removing bacteria in the traditional process involving HNAD and anammox bacteria. Compared to Model A, the abundance of these bacteria in Model B was 2.5 times lower, 2 times higher, and similar, respectively. In the double-layer filter (Model C), the use of uncased foams proved more favorable for the growth of HNAD bacteria, whose average abundance of 16% in each layer displaced Nitrosomonas, Nitrospira, and anammox bacteria. This high abundance of bacteria capable of simultaneously performing heterotrophic nitrification and aerobic denitrification was advantageous over the traditional nitrogen removal pathways. As a result, a single upper layer of foam (95% open porosity), combined with a lower sand layer, enabled 92% NH₄⁺-N removal under low C/N conditions (approximately 1.5) in a system that did not require additional aeration. This double-layer filter showed superior contaminant removal; however, there was a risk of clogging due to the TSS and VSS in the Schmutzdecke being similar to that of the control sand filter.

The bacterial community in an environment where the external carbon source was low included the microbiome involved in cellulose decomposition, EPS decomposition or production, VFA production, and PHA and PHB production. Thanks to this consortium associated with the internal carbon source, increased metabolic activity in the three-layer filters resulted from bacteria involved in endogenous denitrification.

The study shows that waste polyurethane foams in the form of upholstery sponge waste can be used as fill material for multilayer vertical flow filters in the treatment of domestic wastewater with elevated ammonium nitrogen content. However, the number and type of layers in the filter determines the nitrogen removal scheme, which may include simultaneous nitrification-denitrification in addition to the traditional pathway. The proposed technology for the biological treatment of domestic wastewater with elevated ammonium nitrogen content is cheap and uncomplicated and does not require an additional external carbon source or aeration. Thus, it may be an interesting solution compared to most existing aerobic systems.

Further research should be conducted to study the long-term use of the multilayer sponge-sand filter in technical-scale applications for domestic sewage treatment, with particular emphasis on the clogging phenomenon associated with the Schmutzdecke.