To better understand the stability of the characteristics of whitewater during the production process, we monitored several parameters of whitewater from a wood-free-paper mill over a period of 2 months (Supplementary Table A1). We measured the chemical oxygen demand (COD) (ISO 6060, 1989), the biological oxygen demand (BOD₅) (ISO 5815-1, 2019) and the pH (pH meter model 827, Metrohm). The dissolved organic carbon content (DOC) was measured with an advanced TOC analyser (Teledyne Tekmar, model Torch) with a NDIR detector, by subtracting the measured inorganic carbon content from the measured total carbon content. The concentrations of anions (Cl^– and SO₄^2–) were measured using a Dionex DX-120 ion chromatograph (ISO 10304-1, 2007) and those of the cations (Ca²⁺, Mg²⁺, Na⁺, K⁺, and NH₄⁺) using a Dionex Thermo Fisher Scientific ion chromatograph (ISO 14911, 1998). The content of HCO₃^– anions was titrated with HCl (ISO 9963-2, 1994).

At the wood-free-paper mill we collected three types of samples: the clear filtrate (CF) of whitewater passed through the disc filter, the biofilm (BF) taken from the wet-end part of the paper machine and the effluent of the biological treatment plant (BTP). The CF and BF are both periodically washed into the biological treatment plant. We collected 1 liter of CF and BTP and 0.1 liter of BF and used them as inocula to culture the bacterial degraders of the papermaking organic additives. A total of 100 μL of the 10-times-diluted inoculum source was microbiologically enriched in M9 minimal medium (Miller, 1972) supplemented with a sole carbon source at final concentration 0.3 mM. M9 minimal medium contained 1x M9 salts (33.7 mM Na₂HPO₄ ⋅ 2H₂O, 22.0 mM KH₂PO₄, 8.55 mM NaCl and 9.4 mM NH₄Cl; VWR, United States), 1mM MgSO₄ ⋅ 7H₂O, 0.3 mM CaCl₂ ⋅ 2H₂O and 1x trace elements solution (all Sigma Aldrich, United States). We used 13 different carbon sources that are added to the paper-fiber slurry of the wood-free-paper production process and end up in the whitewater cycle (Hubbe et al., 2016): (i) the readily biodegradable additives (AKD, resin acids, cationic, native and soluble starch, cellulose, PVA, or latex), (ii) the non-readily biodegradable azo dyes (red, blue, black, and yellow), and (iii) the non-biodegradable fluorescent whitener (Supplementary Table A2). After 7 days at 25°C a total of 100 μL of the 100-fold-diluted culture enriched in the respective carbon source was plated in triplicates onto the solid M9 medium, supplemented with the respective carbon source. The obtained isolates were marked according to the carbon source they were able to utilize during isolation (Supplementary Table B1).

For each bacterial isolate, the potential to degrade the papermaking additive used as a carbon source during enrichment and isolation was evaluated. The assay duration and the amount of carbon source in the medium were adjusted to determine the activity of the most potent degraders. The amount of each carbon source was adjusted so that discoloration of the test medium was visible, and the duration of the test was set to when the discoloration was visually apparent (Supplementary Tables A3, A4). Depending on the carbon source, biodegradation tests were performed in solid or liquid media (Supplementary Figure A1), all of which were supplemented with 10 mg/L cycloheximide (Sigma Aldrich, United States). Solid agar media were used to rapidly characterize the most active degraders of the readily biodegradable additives (see section “The Readily Biodegradable Additives”), while liquid media were used to characterize the degraders of the azo dyes and whitener (see section “Azo Dyes” and section “Fluorescent Whitener”). The absorption/emission maxima of the dyes and whitener were not affected by the concentration differences; therefore, the measurement was performed at a fixed wavelength (Supplementary Table A2). Centrifugation (12,000 × g, 5 min), which primarily served to sediment the cells, did not result in sedimentation in the assay for either the dyes or the whitener (see Supplementary Material section “Development of the Biodegradation Assays for Azo Dyes and Whitener”), but did in the case of the yellow dye, so the bioassays for this dye had to be adjusted accordingly (see section “Fluorescent Whitener”). Non-inoculated sterile media were used as controls in all experiments.

To test whether the degradation process of recalcitrant products is stable and suitable for industrial treatment process, we followed the time-dependent degradation of blue and red dyes and whitener using the most active bacterial strains. The exact initial concentrations of the carbon sources and the final duration of the tests are given in Supplementary Table A3.

Using 16S rRNA gene sequencing, we identified all the isolates that were shown to degrade the corresponding organic additive above the detection limit of the biodegradation assays, described in section “Biodegradation Assays” (Supplementary Table B2). A further 50, 52, 17, and 75% of isolates enriched and isolated on media containing red, blue, black or whitener dye, respectively, were identified as being able to use the recalcitrant dyes as sole carbon sources regardless of the detection limit of the biodegradation assays (Supplementary Table B1N). Overall, the DNA of 95 bacterial strains was obtained using isolation with Chelex^®100 Resin (Walsh et al., 1991). A single colony of each isolate cultured on solid NB was transferred into the chelex solution. After a 15-min incubation at 95°C and a 15-min centrifugation at 4°C and 5000 rpm the supernatant containing the DNA was transferred into a sterile microcentrifuge tube and stored at −20°C. The 16S rRNA gene was PCR amplified using standard primers according to Hubad and Lapanje (2013). The primers 27f (AGA GTT TGA TCM TGG CTC AG) and 1492r (CGG TTA CCT TGT TAC GAC TT) (Integrated DNA Technologies, United States) were used for amplification. The PCR products were sequenced using Sanger technology with the forward primer 27f at Macrogen, NL.

The chromatograms of the 16S rRNA sequences were curated in Chromas 2.6.6 (Technelysium Pty. Ltd.) to produce 1090–1340 bp large sequences. The sequence identity was obtained using BLASTN 2.9.0+ (Camacho et al., 2009) and the databases RDP_11 (Ribosomal Database Project) and NCBI (National Center for Biotechnology Information). The sequences were deposited in the NCBI GenBank database with the accession numbers MW144826-MW144947 and MW131116.

Two types of phylogenetic trees were created on the basis of the Kimura 2-parametric model in MEGA X software (Kumar et al., 2018). The clustering of 95 isolates together with the most similar type strains was inferred using the unweighted pair-group method with arithmetic mean (UPGMA). The evolutionary distances were computed using the p-distance method, whereas the clustering of the four isolates of the consortium together with the most similar type strains was conducted using the NJ algorithm.

Statistical correlation between carbon source use, genera, and inoculum source of isolates was examined using two multivariate methods: principal component analysis (PCA) and FreeViz projection (Demšar et al., 2007) using Orange version 3.27.1 software (Demšar et al., 2013). We used the relative reciprocal values of c/c₀ (see Supplementary Table B3) on a scale from 0 to 1: c_U/c₀, where c_U indicates the relative amount of carbon source utilized after a given time and higher values indicate greater degradation activity. We defined the extreme degraders as those strains that degraded fluorescent whitener up to c_U/c₀ > 70% and other additives up to c_U/c₀ > 50%. Supplementary Table B4A shows the 11 variables (degradation activities) and their 44 average values. We calculated incomplete data for the tests in solid media using a model-based imputer of the software Orange, and for the cometabolism test of blue in NB we set a value of zero (see Supplementary Table B4C). Since we were dealing with many variables, we visualized the main features obtained by linear projection of PCA and FreeViz by explicitly making the basis vectors visible only when their endpoints exceeded a certain distance from the center, and marking the area where the basis vectors are not displayed with a circle, as previously reported (Demšar et al., 2007; Sakurai et al., 2020).

Based on the results of the biodegradation assays (see section “Biodegradation Assays”), we selected 44 of the most active strains, 2–5 per carbon source (Supplementary Table A4). These were then cross examined for their ability to degrade eight organic papermaking additives (cationic and soluble starch, cellulose, resin acids, AKD, red and blue azo dyes, and whitener) (Supplementary Figure A2), which are most abundant in the paper-fiber slurry of a wood-free paper mill (Supplementary Table A2). In this way, we were able to characterize metabolic specialists, who degrade only a single additive, and metabolic generalists, who degrade a broad repertoire of organic papermaking additives. Using principal component analysis (see section “Principal Component Analysis and FreeViz Projection”), we selected six isolates covering the entire repertoire of additives: (i) starch, (ii) the readily biodegradable, (iii) dyes in glucose, (iv) dye in NB, (v) dye in NB, cellulose and whitener, and (vi) all additives. Activity for whitener was a preferred characteristic as it is the most difficult to degrade. For each of the six strains, the CFU versus OD₆₀₀ parameters were calibrated by culturing the strains in NB for 48 h at 25°C (Supplementary Figure A3).

The cross-examination was performed using biodegradation assays (see section “Biodegradation Assays”), with experimental conditions as similar as possible for all additives. To prepare the inoculum, isolates were cultured for 3 days at 25°C and 50 rpm in M9_Glc medium. Tests with the non-readily biodegradable compounds were performed for 4 days at 25°C as cometabolism tests in M9_Glc supplemented with 5 mg/L red or blue dye or the whitener or in NB media supplemented with 5 mg/L blue dye (see Supplementary Table A3); the cometabolism of the blue dye was tested in NB and M9_Glc because the discoloration of dyes varies in different media (Solís et al., 2012).

The tests with the readily biodegradable compounds were performed in liquid and solid M9 media. The tests in the liquid M9 media supplemented with 0.5 g/L cationic or soluble starch or 4 g/L AKD lasted 4 days at 25°C. The absorbance was recorded at 584 nm before (c_cells) and after (c and c₀) the addition of Lugol and the coefficients (c-c_cells)/c₀ were compared (see section “Biodegradation Assays”). Tests with solid M9, to which either 5 g/L cationic starch or cellulose or 0.25 g/L AKD or resin was added, lasted 7 days at 25°C with a fixed volume of medium per Petri plate (see Supplementary Table A3).

Overall, to select bacteria with the highest potential for bioaugmentation of wood-free whitewater, we used detection methods based on absorption and fluorescence spectroscopy, and measuring the discoloration radius after addition of Lugol, whereas for media containing suspended solids (whitewater) (see section “In vitro Co-culturing Test” and section “Pilot-Scale Experiment Proof of Concept Only”) we determined the mineralization of organic additives by DOC and COD measurements (Table 1).

All 63 combinations of the six selected strains (single strains and co-cultures of two, three, four, five and six strains) were used to test the degradation of organic additives in synthetic and in real industrial whitewater under sterile conditions. Co-culturing in synthetic whitewater was performed to increase the regularity of the test, since the composition of the real whitewater depends on the papermaking process. The co-culturing test helped us to select the best combination for mineralization of additives in real whitewater. For the preparation of non-complex synthetic whitewater, M9 medium was used, to which CST, PVA, resin, and Direct blue 15 dye were added in a weight ratio of 1:1:1:0.5, totaling 107 mg/L DOC. Real whitewater was first sterilely filtered through a 0.2-μm membrane polypropylene filter (Macherey-Nagel) and then, M9 salts were added to achieve a 1× final concentration. Cells of each strain were collected by centrifugation (12,000 × g, 5 min) from cultures grown overnight in NB. Then they were washed three times and resuspended in sterile 0.9% NaCl. The number of cells in the suspension was calculated for each strain based on the calibration curves (see section “Characterization and Selection of Bacteria With the Highest Potential for Degradation of Organic Additives”). The volume of each strain suspension was adjusted to achieve a concentration of 10⁷ CFU/mL in the final 200-μL co-culture test mixture. The co-culture experiment was performed in six replicates in 96-well microtiter plates. After 3 days at 25°C, the media were sterilely filtered through 0.2-μm polypropylene filters. Complete mineralization of organic carbon was analyzed by DOC measurements using a TOC analyzer (Shimadzu) (see Supplementary Table B5).

In addition, for the final four selected strains used to construct the artificial consortium, AKD degradation was confirmed by absorbance and DOC measurements (Supplementary Table A5).

The pilot-scale experiment was performed under unsterile conditions, for the proof of concept only. To demonstrate feasibility, the artificial consortium of the four bacteria selected in the co-culturing tests was first immobilized on porous carriers, which were then placed in a 33-liter plexiglass column (r = 0.06 m, h = 2 m) (Supplementary Figures A4A–C). A parallel, continuous flow of industrial whitewater was introduced into the column at a rate of 2 L/h using a Qdos 30D peristaltic pump (Watson Marlow). The hydraulic retention time in the column was estimated to be 13–15 h. Aeration was set up at the bottom of the column and controlled by a Tecsis compression force transducer (Wika Group). Oxygen saturation of the whitewater was monitored using a HQ40d model multi portable oxygen electrode (Hach) submerged 20–30 cm below the top of the column. During the run, there was minimal movement of the carriers caused by mixing of the liquid with the aeration system. After 8 days, a recirculation was set up at the same flow rate with an additional peristaltic pump to direct the flow back into the column. After 10 days, a new batch of whitewater was introduced. To create favorable growth conditions for the bacteria a ratio of COD : Total N : Total P = 100: 5: 1 was achieved by adding urea (solid) and H₃PO₄ (85% w/v) (both Sigma Aldrich, United States) to the whitewater. The following process parameters were measured at the influent and effluent using Nanocolor rapid tests (Mesherey-Nagel): Total N, Total P, PO₄^3–, NO₃^–, NH₄⁺, and COD. The pH was measured using a portable pH meter model HQ11d (Hach). Samples of 0.25-L were collected at the inflow and outflow of the column in clean plastic containers and the process parameters were measured immediately after sampling. The control system was as described by Qu et al. (2005). It was performed in 250 mL flasks containing 190 mL of clean carriers, a proportional amount of carriers immersed into 0.5% sodium alginate (2 carriers per 250 mL), and 250 mL of unsterile industrial whitewater in triplicate. The flasks were incubated in a shaker incubator at 50 rpm and 25°C, and the COD was measured before and after incubation of 15 h, i.e., the duration of the retention time, and the results were used as a control (Supplementary Table B6A).

Bacterial cells were immobilized on the carriers as previously reported (Horemans et al., 2017a). The four isolates were grown separately for 72 h at 25°C in liquid NB and the cells were collected by centrifugation (15 min at 12,000 × g) and re-suspended in 9 g/L NaCl to yield 3 × 10⁸ CFU/mL of BLA14 and 1–4 × 10⁹ CFU/mL of CST37, AKD4, and RES19. Sodium alginate at a final concentration of 0.5% (w/v) and 0.4 L of sterile Kaldnes K3 carriers (Tongxiang Small Boss Ltd., China) were added to the bacterial suspension, which was then vacuumed three times using a two-stage P4Z rotary pump (ILMVAC) to impregnate the inner surface of the porous carriers with the cell suspension. We achieved 3–4 × 10⁶ CFU/carrier for BLA14, CST37 and RES19 and 2 × 10⁷ CFU/carrier for AKD4. A total of 1.6 L of bacteria- loaded carriers, 0.4 L per isolate, was mixed with 23.4 L of clean carriers (25 L in total) and stored in a container for 16 h before being packed into the column. It was expected that cells released from the impregnated carriers would colonize the clean carriers in the system according to Horemans et al. (2017a). We therefore assessed the CFU/carrier^∗ for each isolate, as if the cells would ideally be distributed from the immobilized to the total volume of carriers in the column (further details in Supplementary Table B6B). The relative difference in CFU/carrier after 7 days was calculated for each isolate individually (Supplementary Table B6B). For this purpose, the CFU/carrier of the freshly immobilized and the 7-day-old carriers collected 20–30 cm below the top of the column were determined using standard plate counting. A total of four carriers were placed in 50 mL sterile centrifuge tubes containing 20 mL sterile 9 g/L NaCl, and all experiments were performed in two replicates. Cells were removed from the carriers by vortexing the carriers at maximum speed for 5 min. The amount of each was determined using a stereomicroscope (Motic SMZ-168) (Supplementary Figure A4E).

Monitoring the removal efficiencies of individual organic additives was performed indirectly due to the complexity of whitewater composition, containing low amount of additives, and suspended solids and dyes that interact with light absorption, by inferring the degradation of each additive from the characterized activities of the four isolates (see section “Extreme Degraders Originate From Every Inoculation Source” and section “By selecting Six Isolates We Can Degrade the Whole Repertoire of the Organic Additives”).

Systematic analysis of the whitewater from a wood-free paper mill over a 2-month period revealed relatively constant COD, BOD₅ and DOC, of 300, 165, and 95 mg/L, respectively (±10%, n = 19, 8, 19) (Supplementary Table A1). Consequently, we prepared a whitewater treatment solution using an artificial consortium of preselected indigenous bacterial strains that efficiently degrade organic additives used in papermaking.

We isolated a total of 318 bacterial strains from three inoculation sources, from which we successfully identified 95 by 16S rRNA gene sequencing. The isolates belonged to four phyla: Proteobacteria, Firmicutes, Actinobacteria and Bacteroidetes (Figure 2). The largest number of them, 15 isolates, was identified with the closest relative [Pseudomonas] boreopolis (T), currently classified as Xanthomonadales bacterium, followed by 13 isolates as Mycobacterium sp. and 7 isolates as Micrococcus sp., Stenotrophomonas sp. and Agromyces sp. (Table 2). The highest proportion of all the identified isolates was obtained from CF (55%) and, respectively, 25% and 20% from BF and BTP. We found that isolates belonging to the same genus were in the majority (86%) inoculated from the same inoculation source (Table 2), except for Micrococcus sp. and Staphylococcus sp. that we assess as primary contaminants from the paper mill.

Preliminary characterization of 318 bacterial isolates for their ability to degrade the carbon source they used during isolation (Supplementary Table B1) helped us to select 44 strains that were best able to degrade individual carbon sources (Supplementary Table A4). Overall, 30 of the selected strains degraded one of the readily biodegradable additives (cellulose, starch, latex, PVA, and AKD) and 15 of them degraded one of the five dyes (red, blue, black, yellow, and whitener). Xanthomonadales bacterium spp. were found to be the most active strains for degradation of readily biodegradable additives CST, ST, AKD, PVA, and CE with 5–72% c/c₀ higher degradation compared to other strains. For resin acids, latex and native starch, Aeromonas sp. RES19-BTP, Agromyces sp. strain LX13-BF and Agromyces sp. NST20-BTP were the most active with 36%, 44% and 4% more c/c₀ than others, respectively.

According to the results of the biodegradation assay, the two isolates selected for blue dye degradation, BLU19 and BLU23, degraded 80% and 60% c/c₀, respectively, after 18 h (Supplementary Table A4). Determination of time-dependent activities (using different preparation of inoculum, see Supplementary Table A3) revealed that BLU19 and BLU23 still showed significant differences after 24 h, degrading 40 ± 3% and 33 ± 2% c/c₀, respectively (Figure 3A and Supplementary Table A6). However, after 48 h, a linear decrease to 13% ± 0% c/c₀ was achieved, and we observed no differences in degradation between the two BLU isolates. At 24 h, RED14 showed 9–16 times faster red dye discoloration and degraded 96% of the dye (Figure 3B), while at 48 h there were no significant differences between the RED isolates (14, 14A, 15A) that degraded all of the dye. Since we used a ∼100-fold higher concentration of whitener, this resulted in higher c/c₀ values (Figure 3C). Therefore, we could not directly compare the c/c₀ values between the degraders of whitener and the red and blue dyes. OD₆₀₀ measurements of WH5 and WH8 revealed no or minimal growth in 165 h, which is in contrast with CFU counts obtained, as 10⁹ and 6 × 10¹⁰ CFU/mL of WH5 and WH8 were counted, respectively. This indicates a disruption of light absorption by whitener. In the first 72 h of incubation, the maximum degradation of whitener to 75% c/c₀ was recorded for WH8, which was 12% more than the others.

We found that the most active yellow isolates, Y17, Y18, Y19, degraded yellow dye similarly in 7 days, up to 55 ± 5% c/c₀, and Y14A up to 92% c/c₀ (Supplementary Table A4). In 72 h the black dye was discolored up to 75 ± 6% c/c₀ by three isolates with an extremely high OD₆₀₀ of 1.8–2.4.

Using multivariate PCA and FreeViz projection, we found that papermaking additives could be classified into three groups: whitener, the least biodegradable additive, azo dyes and readily biodegradable additives (Figure 4A and Supplementary Figure A5A). Of the 44 isolates tested for carbon source repertoire, 70, 45, and 29% from the inoculation sources BTP, CF, and BF, respectively, are extreme degraders of at least one type of additive (Supplementary Table B4B). Starch utilization is a distinctive feature of isolates from CF, with 50 % (out of 20 isolates) being extreme degraders of readily biodegradable additives. Isolates from BF are characterized by the utilization of whitener, with 29% (out of 14) being extreme degraders of whitener. In the case of BTP, we isolated extreme degraders of all additives.

With 82% explained variances (Supplementary Table B4E), a positive correlation was found between the test results in the solid and liquid media with CST and AKD, with 86 ± 3% equal amounts of above- and below-average degraders (Supplementary Table B4F). Among them, seven isolates are extreme degraders of CST, AKD, resin and cellulose in solid media, while RES19 and RED15A are also extreme degraders of whitener. About 70% of extreme cellulose degraders are also extreme degraders of AKD, resin and CST, while 69% of extreme starch degraders are also extreme degraders of AKD when tested in liquid media. In addition, a positive correlation was found between dye degraders in M9_Glc, with 82% of above and below average blue and red dye degraders occurring in equal proportions, with only three isolates causing extreme degradation of both dyes: BLA14, RES19, and AKD5. Xanthomonadales bacterium spp. stand out as extreme degraders of all readily biodegradable additives (Supplementary Figures A5C,D). RES19 is the best generalist, with an average c/c₀ of 30% ± 7% for all additives.

Strain AKD4 was selected on the basis of high hydrolysis of starch in liquid media (Figure 5A) and whitener (Figure 5B), with 0% and 74% c/c₀ after 3 days, respectively, with the closest type strains Cellulosimicrobium funkei (T) and C. cellulans (T) (Supplementary Figure A6C). Xanthomonadales bacterium strain CST37-CF, which is closest to the [Pseudomonas] boreopolis (T) type strain (Supplementary Figure A6A), also hydrolyzed 100% starch. It was selected because it was among the most active for all readily biodegradable additives, with an average c/c₀ for starch, cellulose, resin acids and AKD of 18% ± 11% and, unlike other comparable strains, active for utilizing whitener with 87% c/c₀ (Supplementary Table B3). BLA14 and RES19 had the highest average activity for azo dyes (37% ± 22% c/c₀ and 39% ± 8%, respectively). BLA14 was selected as a specialist for dye utilization, while RES19 was selected as the most active general isolate, an extreme degrader of starch (in liquid) (1% ± 1% c/c₀) and an extreme degrader of whitener, red and blue dye with an average 48% ± 10% c/c₀. BLA14 has the closest type strain Sphingomonas paucimobilis (T) (Supplementary Figure A6B) and RES19 is related to Aeromonas dhakensis (T) and A. caviae (T) (Supplementary Figure A6E). Strains Y14A, AKD13, WH5 and RED15A were the most active (12% ± 0.3% c/c₀) for the blue dye in NB (Supplementary Table B3D). Among them, we selected Y14A with related strains Klebsiella pneumoniae (T) and K. quasipneumoniae (T) (Supplementary Figure A6D) because it was the most active for the blue dye in NB (11% c/c₀) and could also utilize the yellow dye as the only carbon source. RED15A with its sister strain Agromyces indicus (T) (Supplementary Figure A6C) was selected because, unlike AKD13, it is an extreme degrader of whitener (70% c/c₀) and a general degrader. It degraded starch, resin acids, AKD and cellulose in agar media with an average c/c₀ of 36% ± 4% (Supplementary Table B3G).

Co-culturing in media with synthetic and industrial whitewater was compared for all 63 combinations of the six selected isolates under sterile conditions (Supplementary Table B5). It was found that the isolate combinations consumed more organic carbon than the individual isolates after 3 days. Similar isolate combinations were most active in both media, showing that we correctly assumed the composition of the synthetic whitewater. We selected the most active combination in industrial whitewater media, with a c/c₀ of 62.5 ± 1.5% in both media, which consisted of a consortium of CST37, BLA14, RES19, and AKD4 (Supplementary Table B5). The regularity of the co-culture test was confirmed as the same combinations of strains were the best degraders in both media. The low degradation rates were attributed to the low amount of M9 salts in the industrial medium and the large amounts of the dye in the synthetic whitewater. The best consortium degraded 29% ± 3% more carbon content than the single CST37 and was 23% more active in industrial whitewater than the single extreme general degrader RES19.

The pilot experiment was conducted under non-sterile conditions using industrial whitewater and carriers loaded with strains of the artificial consortium, CST37, BLA14, AKD4, and RES19, placed in a 33-liter column (Supplementary Figures A4A–C). After the flow stabilized, COD decreased from an initial 183 mg/L in the influent to 59 mg/L in the effluent, corresponding to a c/c₀ of 32%. No decrease in COD was recorded for the control system after 15 h (see section “Pilot-Scale Experiment Proof of Concept Only”). Initially, a large coefficient of 2.7 was calculated for COD between the unfiltered and filtered effluent. As expected, the difference became insignificant after 3 days, and even on the 21st day, the COD of a new batch of whitewater decreased from 400 to 50 mg/L, corresponding to a c/c₀ of 13%. Thus, we achieved 88% COD removal (Supplementary Figure A4D and Supplementary Table B6A). Initially, 1.6 L of carriers, including 0.4 L for each isolate, were immobilized with 3–4 × 10⁶ CFU/carrier for RES19, CST37, and BLA14 and 2 × 10⁷ CFU/carrier for AKD4, which extrapolates to 4–7 × 10⁴ and 3 × 10⁵ CFU/carrier, respectively, for 25 L carriers (Supplementary Table B6B). On the seventh day, 10% of unknown microorganisms were detected on the used carriers taken 20 cm below the top of the column, a 22% ± 1% increase in CFU/carrier for RES19 and AKD4, 3% for dye specialist BLA14 and 56% for CST37. From the characterization results in Figure 5 and Supplementary Table A4 we extrapolate that the readily biodegradable additives were mineralized to 10% c/c₀, whereas mineralization of dyes was almost inexistent, 98% c/c₀ of the COD, with discoloring of red and blue dye to 60% c/c₀ and minimal discoloring of the recalcitrant whitener, black and yellow dye.