Sewage sludge samples were collected from 19 WWTPs of São Paulo State, Brazil. Five samples (SS6, SS7, SS11, SS12, and SS13) were collected from metropolitan area of São Paulo City, the most urbanized and industrialized region within São Paulo State, whereas the others were collected from other municipalities (Table 1). Sample collection was performed at the sludges dewatering points, as described by EPA SW-865.

For this purpose, three samples were collected from each WWTP. Each sample was composed of five subsamples (200 g) taken in 10-min intervals, mixed, and properly homogenized. They were conditioned in glass bottles and refrigerated until analysis according CONAMA Resolution 375/2006 (BRASIL, 2006).

Moisture was determined according to EPA-SW 846. For this, sludge samples of 100 g were oven dried at 65°C, for 48 h. pH was measured using 2 g of moist sample and 20 ml of deionized water, which was stirred for 5 min at 220 rpm and rested for 30 min. For total inorganic N, 5 g of moist samples were distilled with 50 ml of 1.0 mol L^-1 KCl, 0.2 g of MgO, and 0.2 g of Devarda alloy, which were taken in 5 mL of 20 g L^-1 H₃BO₃ and titrated with 0.0025 mol L^-1 H₂SO₄ (Bremner, 1996). Nitrite and nitrate were determined according to Mulvaney (1996). For organic N (N-Kj), 0.05 g of oven dried samples were mixed with 3 mL of concentrated H₂SO₄, placed in a digester block ( ± 360° C) for 3 h, distilled with 20 mL of 10 mol L^-1 NaOH, which were also taken in 20 mL of 20 g L^-1 H₃BO₃ and then titrated with 0.0025 mol L^-1H₂SO₄ (American Public Health Association [APHA], 2005). Organic carbon (OC) was determined by the K₂Cr₂O₇ method (Nelson and Sommers, 1996). Ca, K, P, Mg, S, Cu, Fe, Ni, Mn, Mo, Si, Zn, Al, As, Ba, Cd, Cr, Pb, Hg, and Na were extracted in microwave oven, according to EPA (2007). K and Na were quantified by flame photometry and the other elements by inductively coupled plasma atomic emission spectrometry (ICP-OES).

For total DNA, 0.4 g of each sewage sludge sample was extracted individually using MoBio Power Soil DNA Isolation Kit (MoBio, United States), according to manufacturer’s instructions. Integrity of the extracted DNA was checked by electrophoresis (1% agarose gel), which was stained with ethidium bromide and visualized under ultraviolet light.

DNA sequencing was performed by Illumina MiSeq platform and library preparation based on Nextera XT index kit (Illumina, United States), targeting the V4 region of the 16S rRNA gene. This was amplified using a mixture of 4-Forward and 4-Reverse primers with pre-adapters (Supplementary Table S1). For the PCR reaction (final volume of 25 μL), 3.0 μL of PCR Buffer, 2.5 μL of MgCl₂ (50 mM), 2.0 μL of DNTPs (2.5 mM), 0.1 μL of each primer mix, 0.3 μL Taq DNA polymerase (0.05 U/μL), 16 μL mili-Q water and 1.0 μL template DNA were utilized. Amplification conditions involved initial denaturation at 95°C for 3 min, 30 cycles at 95°C for 45 s, 57°C for 1 min: 45 s; 72°C for 1 min; followed by a final extension at 72°C for 4 min (Caporaso et al., 2011). PCR products were confirmed by electrophoresis in agarose gel (1%) and resulted in amplified fragments of ∼430 bp. Amplified DNA was then purified with QiaQuick PCR kit, quantified by spectrophotometry (ND-1000), and PCR products stored (-20°C) for sequencing.

After DNA purification, another PCR reaction was performed to bind adapters (an index pair) to identify sequence origin. This consisted of 3.0 μL of PCR buffer, 2.5 μL of MgCl₂ (50 mM), 2.0 μL of DNTPs (2.5 mM), 5 μL of each adapter (index), 0.3 μL of Taq DNA polymerase (0.05 U/μL), 17.2 μL of mili-Q water, and 15 μL of previous reaction product (final volume = 50 μL). Amplification conditions consisted of 95°C for 3 min, five cycles at 95°C for 45 s, 57°C for 1 min: 45 s; 72°C for 1 min; followed by a final extension at 72°C for 4 min. Sequencing was carried out at the University of São Paulo (USP/ESALQ), by the Animal Biotechnology Laboratory within the Animal Science Department.

Quantitative Insights into Microbial Ecology (QIIME) program was used for DNA sequencing analysis (Caporaso et al., 2010). Sequences quality was set at 20. Removal of poor quality sequences, primers, barcodes, and adapters were performed with CLC Genomics Workbench 6 (CLCbio). Operational taxonomic units (OTUs) were grouped in 3% distance level (97% of similarity) and classification was performed by the Ribosomal Database Project (RDP Classifier). OTUs were also used to estimate ecological parameters using Chao 1, Simpson, and Shannon diversity indexes. Clustering of the samples was performed by principal coordinate analysis (PCoA) (Ramette, 2007), and tested by similarity analysis (ANOSIM) on Past^® software (v.3.2) (Hammer et al., 2001). ANOSIM was also used to verify sample similarities according sludge sources, biological (redox) treatments, and liming.

Relationship between bacterial community composition and sludges sources, treatments, and chemical attributes (pH, moisture, N-NH₄⁺, N-NO₂^-/NO₃^-, organic N (N-Kjeldahl = N-Kj), organic carbon (OC), K, Ca, Fe, P, S, Mg, Na, Cd, Cr, Cu, Hg, Mn, Mo, Ni, Pb, Se, Zn, Al, As, and Ba) were settled by redundancy analysis (RDA) on Canoco^® software (v.4.5). Graphics were plotted on Origin^® software (v.10.5), but heatmap graphical scales were built in R software, using “gplots” and “RColorBrewer” packages¹.

Samples identification and main chemical attributes affecting microbial community and their clustering were presented in Table 1. The other chemical attributes were summarized from a previous thesis work (Nascimento, 2015) and presented as supplementary material (Supplementary Table S2). Thirteen samples underwent aerobic (SS1 to SS13) whereas the other six (SS14-SS19) underwent either strictly anaerobic or combined aerobic-anaerobic treatments during biological digestion. Eight samples were collected from domestic (SS1-SS5, SS14, SS16, and SS19) whereas the others were collected from mixed sewers. Only five samples were limed (SS5, SS11, SS12, SS13, and SS19).

A total of 7,219,247 16S RNA gene sequences were attained. After removal of low quality sequences (cut level = 3%), OTUs matrixes showed that all sludges presented high diversity indexes (Supplementary Table S3). Although sequencing would contain inactive (dormant and dead) microorganisms, it should not impact diversity as verified by Liang et al. (2017).

RDP Classifier identified 68 phyla, 164 classes, and 665 genera of bacteria. The most abundant phyla were Proteobacteria > Bacteroidetes > Firmicutes, corresponding to >73% of the DNA sequences (Figure 1A); whereas the most abundant classes were Saprospirae > Betaproteobacteria > Bacteroidia > Clostridia > Deltaproteobacteria (Figure 1B). In addition, Betaproteobacteria was the most abundant class within the Proteobacteria phylum (∼37% of the sequences), followed by Deltaproteobacteria (∼26%), Alphaproteobacteria (∼16%), and Gammaproteobacteria (∼11%); whereas Saprospirae was the most abundant class within the Bacteroidetes (∼46%), followed by Bacteroidia (∼36) and Flavobacteria (∼3%) (Figures 1A,B). Within the Firmicutes, the most abundant classes were Clostridia (∼87%) and Bacilli (∼9%) (Figures 1A,B). Finally, the most abundant genera were Clostridium > Treponema > Propionibacterium > Syntrophus > Desulfobulbus > Brevundimonas > Paludibacter > Cloaci-bacterium > Methylobacterium (Figure 1C). Despite distinctions in sewage sources and treatments, their bacterial community presented a common core of 77 genera, being Clostridium, Treponema, Syntrophus, and Comamonas the most abundant ones (Supplementary Figure S1).

The sludge samples could be grouped in six clusters according to PCoA: C1 (SS1, SS2, and SS3), C2 (SS9 and SS16), C3 (SS11 and SS18), C4 (SS4, SS5, SS6, SS7, SS8, and SS19), C5 (SS10, SS14, SS15, and SS17), C6 (SS12 and SS13). Its main two coordinates explained 36.7% of sludges’ bacterial community structures (Figure 2). This result was also validated by similarity analysis of their bacterial communities (Table 2).

C6 showed bacterial community very distinct from the others, with relative dominance of Propionibacterium, Comamonas, Brevundimonas, Methylobacterium, Stenotrophomonas, and Cloacibacterium (Figure 3). The other clusters (C1-C5) generally presented great abundance of Clostridium, Treponema, Syntrophus, and Desulfobulbus, except that C1 showed low abundance of Syntrophus and high abundance of Dechloromonas; C2 showed relative high abundance of Sedimentibacter; C3 showed relative high abundance of Paludibacter; C4 showed relative high abundance of Sedimentibacter and also of Dok59 and Bacillus; and C5 showed relative high abundance of Paludibacter, PD-UASB-13, Desulfovibrio, and E6 (Figure 3).

Sewage sources (domestic or mixed) and biological treatments (redox conditions) did not affect consistently the bacterial community structuring (Table 3), suggesting that clusters were formed due to other factors, likely related with sludges chemical attributes as suggested by RDA (Figure 4). In fact, pH (λ = 0.11, P-value < 0.002), Fe (λ = 0.07, P-value < 0.002); B and Mg (λ = 0.06, P-value < 0.002); Na (λ = 0.05, P-value < 0.002); and P, Ba, organic N (N-Kj), and Ca (λ = 0.04, P-value < 0.002) contents were the sludge attributes most related to microbial community structuring and clustering; whereas organic carbon (OC), inorganic N (in the different forms), Hg, Se, and As contents, C/N ratio, and moisture were not correlated with sludges bacterial community structures (λ < 0.01 and P-value > 0.05) (Supplementary Table S2).

WWTPs bacterial community exhibited low variation at higher taxonomical levels (e.g., phylum) even for distinct geographic regions and sludge treatments (Philippot et al., 2010; Ibarbalz et al., 2013; Hatamoto et al., 2017) (Figure 1A). In all samples, independently of sewer operating condition, the most abundant phyla were Proteobacteria > Bacteroidetes > Firmicutes (Figure 1A). Similar results were reported for sludges from China (Zhang et al., 2012; Shu et al., 2015b; Gao et al., 2016; Liang et al., 2017). However, the literature shows some contrasting results. Meerbergen et al. (2017) found predominantly Proteobacteria, Bacteroidetes, and Actinobacteria for domestic sludges, but Planctomycetes, Chloroflexi, Acidobacteria, and Chlorobi for industrial sludges. Proteobacteria usually predominated in domestic sewage sludges, corresponding from 30 to 65% of the total sequences (Liang et al., 2017; Meerbergen et al., 2017), as well as in various other environments, such as soil (Roesch et al., 2007; Spain et al., 2009; Sun et al., 2015) and rhizosphere (Jiang et al., 2016). Proteobacteria usually presented wide diversity and metabolic capacity, acting in important environmental functions such as the cycles of C, N, S, and P (Friedrich et al., 2005; Meyer et al., 2016). Bacteroidetes were often reported as proteolytic bacteria, involved in degrading protein to volatile phenolic acids and ammonia (NH₃) (Yi et al., 2014). Their abundance was correlated with total solid contents when submitted to anaerobiosis (Liu et al., 2016). Firmicutes were often widely distributed in anaerobic sludge treatment systems (Yang et al., 2014) and were versatile in degrading a vast array of environmental substrates (Liu et al., 2016). They may act on metabolic pathways responsible for producing volatile fatty acids, which can be used by other microbial groups.

The most abundant classes were Saprospirae > Betaproteo-bacteria > Bacteroidia > Clostridia > Deltaproteobacteria > Alphaproteobacteria > Gammaproteobacteria > Actinobacteria > Spirochaetes (Figure 1B). Liang et al. (2017) and Shu et al. (2015a) reported relative high abundance of Betaproteobacteria, which is often associated with organic matter degradation and S cycle (Friedrich et al., 2005; Takai et al., 2005). In Denmark, however, several studies reported low occurrence of Saprospirae in full scale WWTPs (Nielsen et al., 2012; Kong et al., 2007; Muszyński et al., 2015). The members of this class were predominant in marine environments, but could also be found in fresh water and sewage sludges degrading complex organic compounds (Nielsen and McMahon, 2014).

Lower taxonomic levels (e.g., genus) showed higher bacterial community differentiation among WWTPs (Figure 1C), corroborating with the literature (Philippot et al., 2010; Ibarbalz et al., 2013). The most abundant genera were Clostridium > Treponema > Propionibacterium > Propionibacterium > Syntrophus > Desulfobulbus > Comamonas > Brevundimonas > Paludibacter > Cloacibacterium > Methylobacterium > Sedimentibacter > Stenotrophomonas (Figure 1C). A great diversity of bacterial genera were also described in the literature (Lee et al., 2015), which several times differed from ours (Ibarbalz et al., 2013; Stiborova et al., 2015; Gao et al., 2016). It could be explained by the fact that WWTPs comprise open and very dynamic systems allowing rapid succession among microbial community members during spatial and temporal scales (Shu et al., 2015a). Nevertheless, our samples showed a common nucleus of 77 bacteria, represented mostly by Clostridium, Treponema, Syntrophus, and Comamonas (Supplementary Figure S1). Gao et al. (2016) identified a common nucleus of 177 genera for sewage sludges from China. This shared core of bacteria is usually responsible for the main functions in the environment (Shu et al., 2015b). Several pathogenic bacteria, such as Clostridium, Treponema, Stenotrophomonas, Bacillus, Mycobacterium, and Acinetobacter were also identified, in accordance to Stiborova et al. (2015).

Sewage sources (domestic or mixed) and biological treatments (redox conditions) did not affect microbial community structure (Table 3) and diversity (Chao1, Simpson and Shannon) (Supplementary Table S3), at least not consistently, similarly to Hai et al. (2014). However, Gao et al. (2016) found that biological treatment (redox conditions) influenced microbial community, which was more diverse in aerobic tanks; whereas Meerburg et al. (2016) reported structural differences in the bacterial community of domestic and industrial sludges. Gao et al. (2016) found a common core of 177 bacteria genera to their samples and we found a common core of 77 bacteria that corresponded to 85% of the identified sequences. They considered only seven samples from strictly domestic sewers that would explain their greater similarity. Normally, bacterial communities of domestic sewers are more diverse due to its larger fraction of readily degradable organic material (Meerbergen et al., 2017). Industrial sewers receive recurrent discharges of more recalcitrant and toxic pollutants (Gao et al., 2016), such as heavy metals and antimicrobial agents (Bettiol and Ghini, 2011; Balcom et al., 2016), thus limiting microbial diversity. Hu et al. (2012) reported high similarity between bacterial communities of five sludges from China, whereas Zhao et al. (2014) reported substantial disparity, mainly due to their spatial variation and biological composition. Meyer et al. (2016) also reported significant variation in the structure of S oxidoreductive bacteria from south Brazil.

On the other side, certain chemical attributes showed direct connections to sludge microbial community structures (Figure 4 and Supplementary Table S2), favoring samples segregation in clusters (Figure 2 and Table 2). High pH values (≥11.9) resulted from liming were responsible for segregating C6 (SS12 and SS13) and enhancing Ca contents (Figure 4). Its most abundant phyla were Actinobacteria, Proteobacteria, and Bacteroidetes; whereas the most abundant genera were Propionibacterium, Comamonas, Brevundimonas, Methylobacterium, Stenotrophomonas, and Cloacibacterium (Figure 3). Other limed samples (SS5, SS11, and SS19) presented lower pH (Table 1) and; therefore, very distinct microbial structure from C6. Despite having similar operating conditions as SS12 and SS13, SS11 also showed slightly lower pH as well as lower Cu and Zn and higher Fe and Pb contents (Table 1). It has been demonstrated that 1 pH-unit may considerably affect bacterial community structure and composition (Fierer and Jackson, 2006). Gao et al. (2016) also observed distinct phylogeny (Proteobacteria, Bacteroidetes, Acidobacteria, Chloroflexi, and Firmicutes) at lower pH values (∼8.0). Liming to high pH values usually decreases microbial community diversity (Blaszczyk and Krzysko-Lupicka, 2013; Farzadkia and Bazrafshan, 2014), being an important tool promoting sludge hygienization (i.e., pathogens control). In our case, high pH did not affect microbial diversity (Supplementary Table S3) but affected its structure inclusive favoring extremotolerant bacterial groups, such as Actinobacteria (Figure 1A). Several studies showed that pH affected microbial community diversity and composition in soils (Rousk et al., 2009; Cho et al., 2016; Wu et al., 2016) and sewage sludges (Maspolim et al., 2015). Lauber et al. (2009) found that bacterial phyla (Acidobacteria, Actinobacteria, Bacteroidetes, and α, β, and γ-Proteobacteria) relative abundance did not depend on sludge location, but on pH instead. Therefore, pH may modulate microbial community by controlling nutrients availability and enzymatic processes that are essential to microbial metabolism (Fierer and Jackson, 2006; Madigan et al., 2016).

C5 (SS10, SS14, SS15, and SS17) segregation was related mostly to Fe but also to S, B, P, and N-kj, and contents (Figure 4). These samples underwent anaerobic treatment, except for SS10 (Table 1). Under anaerobiosis, both Fe and S have important roles in redox reactions (Ma et al., 2014), acting as final electron acceptors (Moreira and Siqueira, 2006; Shrestha et al., 2009; Alexandre et al., 2012). Fe is reduced to its most soluble form (Fe³⁺ → Fe²⁺) (Shrestha et al., 2009) and reducing bacteria are important mediators of C and N transformations (Tan et al., 2006; Wang et al., 2009; Ding et al., 2014). Several bacteria associated with Fe reduction were identified, such as Acidithiobacillus, Ferrimicrobium, and Nitrospira (Figure 1B). In parallel, S reduction generates energy in anaerobic environments (Aida et al., 2014); and it is crucial in structuring microbial community. It could be ratified by relative high abundance of Desulfobulbus and presence of Desulfovibrio as well as Desulfococcus, Desulforhabdus, and Desulfovirga in the samples (Figure 1C). Desulfovibrio, Desulforhabdus, and Smithella were very efficient in removing S from anaerobically treated sludges (Aida et al., 2015). SS10 was the only sample having aerobic treatment and showed the highest S concentration (Table 1). C5 samples had lower N-Kj likely due to denitrification (Yamashita and Yamamoto-Ikemoto, 2014), resulting sludges with slightly higher C/N ratios (10.8 versus 8.0) (Table 1). Systems operated initially under anaerobic followed by aerobic conditions usually contribute most to N loss since they warranty anoxic denitrification and aerobic nitrification, thus converting ammonia (NH₄⁺) to gaseous N (N₂, NO₂, and N₂O) (Ruiz et al., 2006; Kassab et al., 2010; Yao et al., 2013a,b; Zhang et al., 2014). These samples also showed low P and B contents (Table 1).

All C1 samples (SS1, SS2, and SS3) derived from domestic sewers, aerobically treated and without liming (Table 1). They presented higher B, P, and N-Kj as would be expected from their higher organic matter pool, thus generating sludges with lower C/N ratios (Table 1). They also presented high Na and low Fe and Al contents (Table 1), which would be expected by their source nature (domestic). The most abundant genera were Clostridium > Dechloromonas >> Treponema > Desulfobulbus > Dok59 (Figure 3). Likewise, Clostridium, Treponema, and Desulfobulbus were also abundant in C2 and C3 (Figure 3). High abundance of Clostridium in domestic sludges was expected since it represents 10–40 % of human intestinal microbiota (Manson et al., 2008; Lopetuso et al., 2013). Clostridium was usually the most abundant genera in activated sludges, whereas Desulfobulbus and Dechloromonas were often associated with nutrients (such as N and S) removal from WWTPs (Aida et al., 2015).

Toxic inorganic elements, such as heavy metals (excluding the micronutrients), did not impact microbial community structure (Supplementary Table S2 and Figure 4). These element contents were below those set by the Brazilian legislation for sludge use in agriculture (CONAMA 375/2006). Only one sample (SS7) exceeded threshold concentration for Ni and three (SS1, SS12, and SS13) for Zn, but both are plant micronutrients. On the other side, Cd, Cr, and Ag inhibited important microorganisms for biological treatment, thus impacting sludge bacterial community (Wells et al., 2011).