The uncultured “Ca. Accumulibacter” is the most studied of the PAO genera and is often assumed to be the most important PAO. The classical PAO phenotype has been demonstrated for members of the genus with both in situ methods (Kong et al., 2004) and in sequencing batch reactor (SBR) enrichments, where they can be enriched up to 96% of the biomass (Lu et al., 2006). For a detailed review of their physiology, see He and McMahon (2011). In this study, “Ca. Accumulibacter” had an average and median read abundance of 0.5 and 0.4%, respectively (Figures 3, 5). This is in contrast to previous FISH-based studies, including those assessing the same Danish plants, which recorded values of 3–8% (Mielczarek et al., 2013). A lack of probe specificity for the commonly applied PAOmix probe (see later) may contribute to this discrepancy (Albertsen et al., 2016). In addition, the “Ca. Accumulibacter” often possess relatively large cells (i.e., 2–3 μm (Kim et al., 2010), which results in relatively higher biovolume values in quantitative FISH (qFISH) analyses.

There were three MiDAS OTUs (>97% sequence similarity) classified to the “Ca. Accumulibacter” genus co-existing at similar abundances across the full-scale plants of this study. Several studies have investigated diversity of the genus (Martín et al., 2006; Flowers et al., 2013; Kim et al., 2013; Oyserman et al., 2015; Skennerton et al., 2015; Camejo et al., 2016). The resolution of the 16S rRNA gene to provide meaningful delineation of the genus is questionable and the polyphosphate kinase gene (ppk) has instead been suggested to provide this (McMahon et al., 2007). Two “Ca. Accumulibacter” types (I and II) are further delineated into 13 clades (IA-E and IIA-H) (He et al., 2007; Peterson et al., 2008; Mao et al., 2015). Clades reportedly vary in their morphology, carbon source, and denitrification genes (Flowers et al., 2009; Skennerton et al., 2015). Different “Ca. Accumulibacter” species have been suggested to vary in their level of reliance on polyphosphate storage and carbon uptake kinetics, where the predominance of a particular ppk-clade type may be associated with P removal efficiency (Slater et al., 2010; Welles et al., 2015, 2016). Whether or not important traits can be consistently associated with a ppk-type is yet to be demonstrated. Interestingly, several studies have shown that the “Ca. Accumulibacter” PAO can adopt the GAO phenotype under certain conditions, such as P limitation (Barat et al., 2006, 2008; Zhou et al., 2008; Welles et al., 2015, 2016), and the presence of “Ca. Accumulibacter” cells in full-scale EBPR plants not accumulating polyphosphate has also been reported widely (Zilles et al., 2002a,b; Kong et al., 2004; Wong et al., 2005; Beer et al., 2006). Such findings complicate the use of monitoring PAO phylotype abundances as a measure of EBPR health. Moreover, results of FISH studies should be assessed with caution, given the commonly applied FISH probe sets for the “Ca. Accumulibacter” (PAOmix) also target the Propionivibrio GAO that can be present in comparable abundances (Albertsen et al., 2016).

The second most abundant putative PAO genus in this study was Dechloromonas, with an average plant read abundance of 1.7%, median of 0.5%, and maximum of over 20% for some samples (Figures 3, 5). Contrary to the situation with the closely related “Ca. Accumulibacter,” Dechloromonas appears to be overestimated by amplicon sequencing, where values can be as much as 10-fold higher than with FISH-based quantification (McIlroy et al., 2016; Ziegler et al., 2016). A recent isolate of the genus, from a full-scale nutrient removal system, was shown to accumulate relatively high levels of polyphosphate in axenic culture, although no further characterization was reported (Terashima et al., 2016). Members where shown in situ to accumulate polyphosphate and PHA, and take up carbon under anaerobic conditions (Kong et al., 2007). While some members of the genus Dechloromonas have been shown to behave according to the PAO phenotype, others have been shown to potentially behave as their GAO competitors (Ahn et al., 2007; McIlroy et al., 2016). What proportion of the genus are behaving as PAO, and whether or not this can be confidently assigned to any of the three abundant OTUs of this study, is not known.

Members of the gammaproteobacterial class have also long been suggested as PAO in full-scale systems (Beer et al., 2006). Members of the genus Pseudomonas were identified as being among the most abundant PAO in an SBR EBPR system by applying a polyphosphate staining method with fluorescence activated cell sorting (FACS; Günther et al., 2009). The low abundance of the genus in the present study (Figure 3) is supported by previous FISH surveys of Danish systems and they are unlikely important PAO in full-scale plants (Nguyen et al., 2012).

Also within the class Gammaproteobacteria, the uncultured “Ca. Accumulimonas” store carbon as PHA under anaerobic conditions and polyphosphate under subsequent aerobic conditions, which is consistent with a PAO phenotype (Nguyen et al., 2012). However, they were not abundant in the MiDAS survey (Figures 3, 5), despite FISH surveys reporting their consistent abundance up to 6% of the biovolume (Nguyen et al., 2012). The V1-3 primer set applied in this study was shown to give relatively poor coverage of the Gammaproteobacteria (Albertsen et al., 2015). However, a previous survey of Danish WWTPs applying the V4 primer set also failed to detect the genus in abundance (Saunders et al., 2016)—despite the primer set covering the database sequences of the “Ca. Accumulimonas spp.” well. Further work is required to verify their abundance, and subsequent importance in EBPR systems.

Consistent with previous qFISH-based studies, Tetrasphaera had the highest abundance of the PAO, with an average and median read abundance of 10.7 and 8.9%, respectively, across all plants and present in excess of 40% of the biomass in some plants. Relatively good support for the Tetrasphaera spp. as PAOs has been demonstrated for members of the genus in pure culture and in situ with polyphosphate cycling coupled to anaerobic carbon uptake (Kong et al., 2005; Nguyen et al., 2011; Kristiansen et al., 2013). However, the Tetrasphaera genus has demonstrated diverse physiology, including an ability for fermentative growth, and to accumulate fermentation by-products and amino acids intracellularly (Nguyen et al., 2015). Although numerically the most abundant PAO, key questions need to be answered to establish the contribution members of the genus make to P removal, i.e., how much P is accumulated relative to the “Ca. Accumulibacter” and how varied this is across the abundant Tetrasphaera spp. Three FISH probe-defined clades have been characterized in situ revealing differences in morphology and carbon source uptake (Nguyen et al., 2011). However, analyses of these FISH probes against the current SILVA database suggest that they are unable to separate the different clades with reasonable specificity (data not shown). Although a single dominant OTU made up most of the genus across all plants, several OTUs classified to the genus Tetrasphaera (Figure 7). This is consistent with the observations of Saunders et al. (2016). Due to limitations in the resolution of the short 16S rRNA amplicons, these OTUs could not be related to the proposed three clades of Nguyen et al. (2011).

The genus Microlunatus had a median read abundance of only <0.1%, but was seen as high as 7% in some plants (Figure 3) and has also been detected at abundances up to 9% with qFISH (Beer et al., 2006). Microlunatus may thus be important for the EBPR process in some plants. Microlunatus phosphovorus was isolated from activated sludge and shown to accumulate large amounts of polyphosphate aerobically, and release it anaerobically with a concomitant uptake of glucose (Nakamura et al., 1991, 1995b) and glutamate (Ubakata and Takii, 1994, 1998). Furthermore, in situ strains stain positive for polyphosphate storage (Kawaharasaki et al., 1999; Beer et al., 2006). As with the Tetrasphaera, in pure culture the M. phosphovorus can ferment sugars, and accumulate fermentation products and amino acids anaerobically (Santos et al., 1999). Although some unidentified storage polymer was postulated, neither PHA nor glycogen cycling occurred and P and carbon metabolism were not exclusively coupled (Santos et al., 1999). The type strain reportedly lacks the genes for PHA synthesis (Kawakoshi et al., 2012), however, it has been reported to synthesize the storage polymer (Akar et al., 2006). The physiology of Microlunatus spp. is likely quite similar to that of the Tetrasphaera spp. but requires further investigation.

Members of the actinobacterial genus Tessaracoccus were found to be relatively abundant in some plants with a median read abundance of 0.4% and up to 19% of the reads (Figure 3). Evidence for their role in EBPR rests solely on the ability of T. bendigoensis to accumulate polyphosphate in pure culture (Maszenan et al., 1999b) and nothing is known of their in situ physiology. Isolated members of the genus are facultative anaerobes and thus likely have a fermentative metabolism in wastewater systems (Maszenan et al., 1999b; Finster et al., 2009). Their high abundance appears to be maintained by immigration from the influent wastewater, suggesting a low growth and activity rate (Saunders et al., 2016). It is therefore unknown whether or not they play any substantial role in EBPR systems.

It has also been suggested that members of the genus “Candidatus Microthrix” were responsible for the bulk of P removal in a lab-scale EBPR system, which was based on positive staining for intracellular polyphosphate stores and the observed low abundance of “Ca. Accumulibacter” and Tetrasphaera spp. (Wang J. et al., 2014). Polyphosphate accumulation has been demonstrated for pure culture and in situ species, although it is generally thought to be a response to cell stress (Erhart et al., 1997; Tandoi et al., 1998b) and does not appear to be linked to anaerobic: aerobic carbon cycling (Andreasen and Nielsen, 2000). The genome of “Ca. Microthrix parvicella RN1” also lacks the low-affinity inorganic phosphate transporter (Pit) (McIlroy et al., 2013), which is theoretically essential to the PAO phenotype (Saunders et al., 2007; McIlroy et al., 2014). Its contribution to P removal is therefore uncertain; noting that regardless of any impact it may have on P removal, its presence in the system is undesirable given its proliferation is linked to the serious operational problems of bulking and foaming (Rossetti et al., 2005).

Members of the genera representing putative PAO isolates, Gemmatimonas, Friedmaniella, Quatrionicoccus, Malikia, and Lampropedia were rarely detected in the Danish plants (Figures 3, 5) and are unlikely numerically important to the EBPR process. It should also be noted that, with the exception of L. hyalina (Stante et al., 1997), the only evidence suggesting their role as PAO is an ability to store visible polyphosphate granules.

A recent study hinted at the involvement of the uncultured cyanobacteria “Candidatus Obscuribacter phosphatis” in EBPR, based on the annotation of its genome (Soo et al., 2014). The genome possesses the low-affinity inorganic phosphate transporter (Pit), which is not in itself evidence for the PAO phenotype (McIlroy et al., 2014), given that it is also found in non-PAO organisms (e.g., Defluviicoccus GAO; Wang Z. et al., 2014). It is generally difficult to identify putative PAOs or GAOs based on their genome, given that the storage of polyphosphate, glycogen, and PHA are widely phylogenetically dispersed traits (Oyserman et al., 2015). In situ work or gene expression studies are required to assess the activities of this genus in EBPR, noting that it is also not numerically important in full-scale EBPR systems in Denmark (Figures 3, 5).

What was for many years considered as the genus “Candidatus Competibacter” has recently been suggested to be the family Competibacteraceae to better reflect the phylogenetic diversity of the group (with >89% 16S rRNA gene sequence similarity between members) (McIlroy et al., 2015a). The family is delineated by 14 FISH probe-defined clades and includes the genera “Ca. Competibacter,” “Candidatus Contendobacter,” and Plasticicumulans. The Plasticicumulans do not show the GAO phenotype and are rarely detected in full-scale plants with amplicon or qFISH analyses (McIlroy et al., 2015a), although a related phylotype was suggested to outcompete the “Ca. Accumulibacter” PAO in a novel EBPR system with continuous aeration (Schroeder et al., 2009). The most abundant of the groups in this study were the “Ca. Competibacter” and CPB_S60, which is consistent with qFISH-based studies (McIlroy et al., 2015a). Most of the genera had median abundances of less than 0.1%, yet periodically reached relatively high abundances of up to 25% in some plants (Figures 4, 6)—mainly those receiving high industrial loads (>50%) (i.e., Fredericia and Ejby Mølle; Table 1). The CPB_C22&F32, CPB_CS1, CPB_Q07, and CPB_S23 clades were never observed in read abundances greater than 1% and are unlikely important in Danish EBPR plants. The CPB_M38 clade was not assessed in this study as it is not covered by the MiDAS taxonomy given that it is represented by a single 16S rRNA gene sequence.

The Competibacteraceae are most frequently studied with family-level FISH probes and the classical GAO physiology has been confirmed with enrichment cultures (Lemos et al., 2007; Saunders et al., 2007; Oehmen et al., 2010) and in situ (Kong et al., 2006). Anaerobic carbon storage of VFAs as PHA in the absence of P cycling, consistent with the GAO phenotype, has also been demonstrated in situ for the “Ca. Competibacter” (formerly GB1), “Ca. Contendobacter” (GB5-7), CPB_S18 (GB2-3), CPB_S60 (GB4), and CPB_C22&F32 (GB8) phylotypes (Kong et al., 2002, 2006; Kim et al., 2011). All appear to have an affinity for VFAs with known differences between the phylotypes including carbon substrates type, fermentation, and denitrification (Kong et al., 2006; McIlroy et al., 2014).

Also within the Gammaproteobacteria are members of the uncultured CCM19a phylotype. These organisms were shown to store carbon as PHA under anaerobic conditions in the absence of polyphosphate storage, which is consistent with the GAO phenotype (Kong et al., 2007). In this study, they were rarely detected, with an average abundance of <0.1% and never in excess of 0.5% in any sample (Figure 4). Similarly, low abundances were also found with qFISH surveys, where they were up to 4% but typically <1% (Kong et al., 2007).

In the current study, the genus Defluviicoccus had a median abundance of only 0.1%, but was observed at up to 13% of the amplicon reads (Figure 4). Four probe-defined sub-groups currently delineate the Defluviicoccus genus (Wong et al., 2004; Wong and Liu, 2007; McIlroy and Seviour, 2009; Nittami et al., 2009). Interestingly, clusters 1, 2, and 4 have a TFO morphology, while cluster 3 is filamentous and has been implicated in sludge settleability problems known as bulking (Nittami et al., 2009). Clusters 1 and 2 are the most studied and have been shown to conform to the GAO model with enrichment cultures (Dai et al., 2007; Lemos et al., 2007; Burow et al., 2008, 2009; Oehmen et al., 2010) and in situ (Wong et al., 2004; Meyer et al., 2006; Burow et al., 2007; Wong and Liu, 2007). GAO phenotype has also been demonstrated for the sole isolate, D. vanus, a member of cluster 1 (Maszenan et al., 2005; Wong and Liu, 2007). Anaerobic carbon storage as PHA has also been shown in situ for clusters 3 and 4 (McIlroy et al., 2010). In this study, OTUs representing clusters 2 and 3 made up the majority (>90%) of amplicon reads assigned to the genus, which is consistent with FISH analyses of full-scale systems (Figure 8; Burow et al., 2007; McIlroy and Seviour, 2009). The original report of an alphaproteobacterial Sphingomonas-related GAO (Beer et al., 2004) was later shown to be incorrect and the FISH probes were shown to be binding to members of the Defluviicoccus cluster 1 (McIlroy et al., 2011).

The newly discovered Propionivibrio GAO is closely related to the “Ca. Accumulibacter” and is covered by the PAOmix probe set (Crocetti et al., 2000) commonly applied to target the former (Albertsen et al., 2016). In a lab-scale SBR fed with VFAs as the sole carbon source, they reached 35% of the biovolume and were shown to cycle PHA without polyphosphate storage, consistent with a GAO phenotype. In the current study, the Propionivibrio GAO was detected in a median read abundance of <0.1% and was present in most plants reaching abundances of up to 3% (Figure 4). Amplicon-based assessment of their abundance in full-scale systems has been shown to be comparable to qFISH (Albertsen et al., 2016).

The uncultured betaproteobacterial spb280, similar to the CCM19a phylotype, was also identified in full-scale systems at qFISH abundances up to 6% and the GAO phenotype partially shown in situ (Kong et al., 2007). The phylotype was consistently present with the second highest median abundance and maximum amplicon read abundances of 1.5% (Figure 4). Like the CCM19a phylotype, nothing more has been determined for the phylotype since its initial description.

A few isolates of the phylum Actinobacteria have been reported as potential competitors of the PAO. These include N. multipartita (Yoshimi et al., 1996), M. glycogenica (Shintani et al., 2000), and K. limosa (Liu et al., 2000, 2002), all shown to accumulate carbohydrates anaerobically without any polyphosphate storage. Of the genera these represent, only the Micropruina, along with the closely related uncultured sbr-gs28 phylotype, have been shown in situ to store both acetate and glucose as PHA without polyphosphate storage, consistent with the GAO phenotype (Kong et al., 2001).

In the current study, only the Micropruina and sbr-gs28, both members of the family Propionibacteriaceae, were found in Danish WWTPs. These had median abundances of 0.3 and 0.1%, respectively, but were present up to 16 and 5% of amplicon reads, respectively (Figure 4). Interestingly, the Micropruina were the most abundant of all putative GAOs considered in this study, based on both median and mean values (Figures 4, 6). Their high abundance in full-scale systems was also supported by qFISH surveys in Japan where they constituted up to 8% of the biovolume (Wong et al., 2005). Despite their relatively high abundances, little is known of their in situ physiology. As such, investigating their role in EBPR systems should therefore be a priority and is currently underway (McIlroy et al., unpublished).