Resolving the individual contribution of key microbial populations to enhanced biological phosphorus removal with Raman–FISH

Enhanced biological phosphorus removal (EBPR) is a globally important biotechnological process and relies on the massive accumulation of phosphate within special microorganisms. Candidatus Accumulibacter conform to the classical physiology model for polyphosphate accumulating organisms and are widely believed to be the most important player for the process in full-scale EBPR systems. However, it was impossible till now to quantify the contribution of specific microbial clades to EBPR. In this study, we have developed a new tool to directly link the identity of microbial cells to the absolute quantification of intracellular poly-P and other polymers under in situ conditions, and applied it to eight full-scale EBPR plants. Besides Ca. Accumulibacter, members of the genus Tetrasphaera were found to be important microbes for P accumulation, and in six plants they were the most important. As these Tetrasphaera cells did not exhibit the classical phenotype of poly-P accumulating microbes, our entire understanding of the microbiology of the EBPR process has to be revised. Furthermore, our new single-cell approach can now also be applied to quantify storage polymer dynamics in individual populations in situ in other ecosystems and might become a valuable tool for many environmental microbiologists.


4.
Example calculation of per-cell poly-P quantification in T. elongata P-uptake/release 5. The detection threshold for a given analyte Figure S1: Raman reference spectra of sodium poly-P (blue), glycogen (green), and poly-3hydroxybutyrate-co-hydroxyvalarate (red) showing marker bands at 1170 cm -1 (PO2stretching vibrations), 481 cm -1 (C-C skeletal deformation), and 1726 cm -1 (CC skeletal stretch), respectively. The chemical standard for poly-P used here is sodium hexametaphosphate. Since the Raman analysis for poly-P relies on the v(-P-O-P-) stretching vibrations occurring at the 1170 cm -1 wavenumber region, the degree of polymerization in poly-P does not change the Raman return signal for the v(-P-O-P-) bond system, which is always between the 1160 -1170 cm -1 wavenumber region (1,2,3) . Glucose has characteristic peaks at 911 cm -1 , 1060 cm -1 , and 1125 cm -1 , and therefore, will not interfere with Raman peak assignments of glycogen (4) .

Figure S2:
Comparison of Raman spectra of poly-P (blue), ATP (green), and AMP (red). The Raman marker peak at 1170 cm -1 used for poly-P is not observed in spectra of ATP (1125 cm -1 ) and AMP (990 cm -1 ).

Figure S3
: Linear relationship between the poly-P material density and average Raman intensity of the poly-P marker band (at 1170 cm -1 ). Error bars are standard deviation in average Raman counts of triplicate polyphosphate droplets at each material density.    Comparison of Raman spectra of the T. elongata pure culture during P-release and P-uptake conditions. Inset -Raman difference spectrum between Tetrasphaera cells without poly-P (red) and Tetrasphaera cells with poly-P (blue). Missing signature peaks for glycogen and PHA demonstrate that these storage compounds were not produced in significant amounts during the experiment.  It should be noted that the poly-P content in both PAOs in all plants was lower in the return sludge from the settler (RAS stream) than in the aerobic N tanks. Similarily, in the settler, the intracellular glycogen content in Ca. Accumulibacter was also slightly reduced, while PHA levels increased (see also Fig. 4). This pattern indicates the presence of anaerobic conditions in the settler together with the availability of organic carbon. Alternatively, glycogen/poly-P were used for maintenance in the settler (7) . The residence time for the biomass in the settlers is 10-15 h, making it possible that anaerobic conditions may have developed. Anaerobic conditions in the clarifiers are not intended as released ortho-P will increase the P-effluent concentration (mean ± SD in error bars, n = 100 individual random cells in each instance).

Sampling area in the lateral dimension
The theoretically calculated laser illuminated diameter is given by; 1.22λ/NA, where λ and NA are the laser wavelength and the numerical aperture of the microscope objective respectively. However, the actual illuminated diameter can vary up to 300% of the theoretically calculated value due to imperfections in the optical components and the finite bandwidth of the laser (8) . The actual laser spot diameter can be empirically approximated using the 10/90 criterion method as detailed in Zoubir, (2012) (8) and was found to be 2.6 ± 0.18 µm.
The sampling depth with the settings used can be approximated as detailed in Zoubir, (2012) (7) and was approximately 3.8 µm.

Validation of the Raman estimation of poly-P with chemical analysis of the bulk medium orthophosphate levels
From DAPI stained cell counts, the estimated T. elongata cell number was 7.9 * 10 7 ± 1.7 cells/mL Therefore, in poly-P "full" state (end aerobic phase), the phosphate fraction inside T. elongata cells as poly-P, can be estimated using Raman spectroscopy, as; 1.1 * 10 -13 g cell -1 * 7.9 * 10 7 cells mL -1 = 8.7 mg P/L The difference P in bulk medium between the start and the end of the aerobic half of the experiment, determined by chemical analysis was 10.7 mg P/L (See figure below). The Raman estimation of poly-P bound P fraction at the end of the aerobic phase (8.7 mg P/L) and bulk liquid orthophosphate uptake (10.7 mg P/L) during the aerobic stage are approximately the same (there is about a 2 mg/L discrepancy between the two values). Incorporation of some orthophosphate into non-poly-P organic phosphate fraction (i.e., nucleic acids, membrane lipids) during aerobic incubation may account for the 2 mg/L discrepancy seen here, because the Raman method can only account for the P stored as poly-P and not other forms of P.
The sampling volume during the experiment was 1 mL/sample and the vial size was 250 mL. The pH was monitored at the beginning, at the washing stage, and at the end of the P cycling experiment and was never found to deviate away from 7.0-7.2. Sampling was done every 30 min during P uptake/release tests.

Quantification of Ca. Accumulibacter by FISH
A qFISH experiment was carried out with the use of PAOmix (PAO462, PAO651, and PAO846) probe set and the Prop207 probe (covering most of the Propionivibrio spp.) (10) in order to investigate if the use of PAO651 probe alone would lead to substantial underestimation of Ca. Accumulibacter in the in situ samples (see figure below). qFISH was carried out using the same procedure as described in the materials and methods.

Figure:
A comparison of qFISH biovolume fractions (% of EUBmix) of PAOmix, PAO651, and Prop207 in in situ samples of all the WWTPs in this study. The cumulative biovolume fractions of PAO651 and Prop207 were always greater than 90% of the biovolume fraction of PAOmix probe set (error barsmean ± SD).

Supplementary calculation-2: Example calculation of per-cell poly-P quantification in T. elongata P-uptake/release
T. elongata populations that were analysed in poly-P "full" and "empty" states in the uptake release experiment, under identical instrument settings as described in the materials and methods section, to determine the calibration coefficient value. The average Raman intensities obtained for poly-P "full" and "empty" states were 58.25 CCD counts and 10.05 CCD counts respectively (n = 100 individual T. elongata cells).
The average 2D area occupied by a single T. elongata cell, when mounted on CaF2 Raman substrate was estimated to 3.79 ± 0.22 µm 2 , using the image processing software, ImageJ (n = 100 cells).

The detection threshold for a given analyte
The limit of detection for a peak-assigned Raman band for an analyte is given by: IRaman > 2* √IRaman + IBackground (9) When an analyte was below this threshold, (i.e., glycogen in Tetrasphaera cells), it was considered that the analyte was below the limit of detection (LOD).

Supplementary calculation -1: Example of P mass balance calculations
For all plants: Activated sludge (SS; suspended solids): Approx. 10% of organic matter (VSS, volatile suspended solids) is assumed to be cell biomass (5) . All plants investigated were very similar in type of wastewater and design/operation, so same value has been used for all plants.
The ratio 0.65 varied between 0.62 -0.67 in the plants investigated and an average value of 0.65 was applied.
Poly-P loss from analysed cells during the FISH procedure for both Gram-positive and Gram-negative cells was found to be between 8% -16%. Therefore, the calculations have been adjusted for a loss of 12% (average).
To account for the P-loss during FISH procedure; 6 mg poly-P/g SS * 112% = 6.7 mg polyP/g SS