Metabolic Exchange with Non-Alkane-Consuming Pseudomonas stutzeri SLG510A3-8 Improves n-Alkane Biodegradation by the Alkane Degrader Dietzia sp. Strain DQ12-45-1b

Many natural and synthetic microbial communities are composed of not only species whose biological properties are consistent with their corresponding communities but also ones whose chemophysical characteristics do not directly contribute to the performance of their communities. Even though the latter species are often essential to the microbial communities, their roles are unclear. Here, by investigation of an artificial two-member microbial consortium in n-alkane biodegradation, we showed that the microbial member without the n-alkane-degrading capability had a cross-feeding interaction with and metabolic regulation to the leading member for the synergistic n-alkane biodegradation. Our study improves the current understanding of microbial interactions. Because “assistant” microbes showed importance in communities in addition to the functional microbes, our findings also suggest a useful “assistant-microbe” principle in the design of microbial communities for either bioremediation or chemical production.

The initial metabolic model of Dietzia sp. DQ12-45-1b was created from a list of reactions collected 43 based on genomic alignment with three phylogenetically close species, C. glutamicum ATCC13032, 44 M. tuberculosis H37Rv and S. coelicolor A3(2), whose GEMs were available (52-54). As described 45 in Supplementary Materials and Methods, the initial model was extensively curated based on its 46 annotated genome, various genetic and chemical databases (i.e. KEGG, SEED, CELLO, TCDB, 47 galacturonic acid, D-glucuronic acid, p-hydroxy phenyacetic acid and D-malic acid) were not 97 utilized in the model because functional genes responding for their transportation and/or metabolism 98 were not annotated in the strain, eight another compounds (including D-turanose, β-methyl-D-99 glucoside, D-fucose, glycyl-L-proline, glucuronamide, methylpyruate, bromo-succinic acid and 100 Tween 40) were not available for iBH925 due to unknown information on their metabolic pathways, 101 and the remaining five compounds (D-salicin, D-glucose 6-phosphate, D-fructose 6-phosphate, 102 mucic acid and D-saccharic acid) could not be elucidated. Nevertheless, the comparison of Biolog 103 data with iBH925 growth simulations indicates that the core metabolic network of Dietzia sp. DQ12-104 45-1b has been properly reconstructed. 105

Metabolic reconstruction of P. stutzeri SLG510A3-8 and model analysis 106
A combination of automatic and manual approaches was used in the GEM reconstruction process 107 for P. stutzeri SLG510A3-8. Briefly, using the metabolic network of P. stutzeri A1501 (iPB890) as 108 the template, a draft metabolic network of P. stutzeri SLG510A3-8 was reconstructed. Strain A1501 109 was isolated from the rice rhizosphere and could provide the plant with fixed nitrogen and 110 phytohormones (47). The genome of strain A1501 was used to develop the first genome-scale 111 metabolic model (iPB890) for P. stutzeri, which was the sole model for the species until now. The 112 reactions from the template model which were either essential non-gene associated reactions or 113 associated to genes which were found in the bidirectional pBLAST search were added to the draft 114 model of P. stutzeri SLG510A3-8. The organism-specific model was reliable source of information 115 since it had previously undergone manual refinements during the process of reconstruction and 116 validation against different database, literature and experimental data. After the replacement, the 117 draft model of P. stutzeri SLG510A3-8 was developed, and it involved 857 genes, 1103 reactions, 118 and 800 metabolites. In the second stage, the draft network was refined through the modification of 119 the biomass equation based on biomass composition information obtained from our experimental 120 measurements and literatures, as well as an iterative and curative process of gap finding and gap 121 filling via algorithms in COBRA toolbox in MATLAB software. Finally, the complete GEM of the 122 strain, namely iBH983, was composed of 983 genes, 1193 reactions and 943 metabolites 123 4 (Supplementary File S2). According to previous report, P.stutzeri strain SLG510A3-8 has a 4.6 Mb 124 circular chromosome containing 4379 protein-coding genes (27). iBH983 presented 22.45% 125 genome information of P.stutzeri SLG510A3-8, and the gene presenting percentage of iBH983 was 126 higher than that of iPB890 (21.74%), indicating that iBH983 was more comprehensive to reflect the 127 phenotype of the corresponding P. stutzeri strain than the reference model. The sum of the five largest subsystems accounts for over 80% of the total number of metabolic 132 reactions (80.18%), of which reactions related to amino acid metabolism and lipid metabolism 133 occupied the largest and the second largest groups respectively (21.04% and 17.78%). The two 134 biomass equations in iBH983 were formed from the major components of P. stutzeri SLG510A3-8 135 cellular dry matter grown on glucose and nonsugar at the exponential phase, including DNA, RNA, 136 lipids, proteins, carbohydrates, co-factors and vitamins. The biomass composition of strain 137 SLG510A3-8 on glucose was assumed to be similar to that of strain A1501 with some modification 138 based on our experimental measurements as described in supplementary file S2. Glycogen was 139 included in the composition, but it was reported that this component was generally produced only 140 in the post-stationary growth phase of Pseudomonas strains when the carbon source was nonsugar 141 (57). Therefore, a biomass equation for nonsugar carbon source was formed by omitting glycogen 142 from the biomass equation for glucose in the model. Detailed information on the formation of the 143 two P. stutzeri SLG510A3-8 biomass equations are shown in Supplementary file S2. 144 The reconstructed model was verified based on the experimental data of P. stutzeri SLG510A3-8 145 growth rate measurement on glucose and carbon source utilization on Biolog assay. In the study, P. 146 stutzeri SLG510A3-8 was grown on glucose other than C 16 for growth data collection, since the 147 cells were unable to accumulate on MF medium with C 16 as the sole carbon source according to our 148 previous experiment (data not shown). In this work, cells of P. stutzeri SLG510A3-8 were found to 149 be able to quickly remove 1 g L -1 glucose in 7 hours, and the highest specific growth rate of the 150 strain was determined to be 0.25 h -1 with the specific glucose uptake rates of 2.20 mmol g -1 h -1 at 151 this time points (Fig S7). The in silico specific growth rate of iBH983 on minimal medium were 152 5 calculated to be 0.22 h -1 via FBA when the flux lower bound for D-glucose exchange reaction in the 153 format was set to be 2.20. The in silico predicted value was 13.39% lower than in vitro experimental 154 data. The percent difference for iBH983 was higher than that for its reference model iPB890 (4.04%) 155 which was might because the biomass equation in iBH983 for strain SLG510A3-8 was less exact 156 than that in iPB890 for strain A1501, but the value was similar with that for iMO1056, the GEM of 157 P. aeruginosa PAO1 (58), indicating that the reconstructed model was valuable for cell growth 158 simulation. The other validating experiment was the Biolog assay, in which the availability of 95 159 carbon compounds for P. stutzeri SLG510A3-8 was tested through a high throughput analysis. As 160 shown in Table S9, 47 out of the 95 carbon substrates were oxidized by strain SLG510A3-8. In 161 detail, the strain was able to in vitro utilize glucose, maltose and glycogen, but was unable to grow 162 on some other saccharides, such as arabinose, cellobiose, fructose, lactose, mannose, raffinose, 163 rhamnose and sucrose; there were 35 acids and acid methyl esters in the platform, among which 28 164 were in vitro available to P. stutzeri SLG510A3-8 indicating that the strain preferred to utilize small-165 molecular acids; some essential amino acids were in vitro unavailable to strain SLG510A3-8 such 166 as L-histidine, but its downstream product urocanate was able to be utilized by the strain. By using 167 FBA, the 95 carbon compounds were tested on iBH983, and the biomass growth was successfully 168 simulated for 39 compounds, of which 37 compounds were true positive (compounds utilized in 169 vitro and in silico). The ten false negative carbon sources (Tween40, Tween80, pyruvic acid methyl 170 ester, succinic acid mono-methyl ester, D-glucuronic acid, α-ketovaleric acid, sebacic acid, 171 bromosuccinic acid, glucuronamide and uridine) gave in silico non-growth phenotype due to the 172 lack of knowledge on their metabolic pathways, while the two false positive carbon sources (L-173 ornithine and L-threonine) were wrongly gave in silico growth phenotype because the correctly 174 lacked L-ornithine and L-threonine transport and/or metabolic reactions might be incorrectly added 175 in iBH983. Generally, the agreement of iBH983 with in vitro experimental data was 87.37%, 176 indicating that iBH983 could effectively represent the aerobic metabolism of P. stutzeri SLG510A3-177 8 on a variety of common substrates. 178

Reconstruction and supplementary characterization of the two-species model 179
To obtain the metabolic model for the microbial consortium of Dietzia sp. DQ12-45-1b and P. 180 6 stutzeri SLG510A3-8, iBH925 and iBH983 were reconciled to unify the metabolites and reaction 181 formats followed by the network integration as shown in Fig  We believed that the synthetic microbial consortium of Dietzia sp. DQ12-45-1b and P. stutzeri 208 SLG510A3-8 did not only have the synergistic biodegradation capability on C 16 , but also could 209 cooperate for the greater recovery of n-alkanes with other chain lengths. In the study, the two 210 bacterial strains were aerobically co-cultivated on the minimal medium suplemented with n-alkane 211 mixture of C 14 , C 16 and C 28 . The co-culture and the monocultures of each strains were kept at 30ºC 212 for 30 days by taking cell-free cultures as the negative control. Cell densities in terms of CFU mL -1 213 and the residual n-alkane abundances were measured as described in the Materials and Methods. As 214 shown in Fig S3a, the cell densities of P. stutzeri decreased to a very low level when the strain was 215 grown on n-alkanes alone, around (1.85±0.19) ×10 4 CFU mL -1 , but its cell density gradually 216 increased to (2.10±0.10)×10 6 CFU mL -1 in the presence of strain Dietzia sp., which was in 217 agreement with the previous finding in Fig 1a. It was observed that Dietzia sp. grew well on n-218 alkanes without the existence of P. stutzeri, and the Dietzia cell densities increased to be 219 (1.20±0.20)×10 9 and (1.73±0.75)×10 9 CFU mL -1 , respectively, at the end of the cultivation period, 220 but the cell growth rate was slightly higher when its was exposed to P. stutzeri. GC-MS analysis of 221 the residual n-alkanes revealed that C 14 , C 16 and C 28 were all available to Dietzia sp. with the 222 removal efficiencies of 52.15±1.28%, 75.75±2.72% and 11.03±3.80%, respectively, which was in 223 agreement with previous reports (24), while the three n-alkane compounds were not preferred by P. 224 stutzeri with the removal efficiencies of 9.38±2.69%, 6.30±2.97%, 4.45±2.48%, respectively (Fig  225   S3b). Unsurprisingly, the removal efficiencies of C 14 , C 16 and C 28 by the microbial consortium 226 (57.07±2.49%, 85.48±1.00%, 18.45±0.24%, respectively) were significantly higher than those by 227 each individual strain, respectively (p<0.05), proving that the two strains had synergistic effect on 228 various n-alkane biodegradation. The survival of strain P. stutzeri SLG510A3-8 and the enhanced 229 n-alkane biodegradation in the co-culture system were also considered to be due to the cross-feeding 230 interaction of Dietzia sp. and P. stutzeri, but the exchanged metabolites between the two strains 231 should be different from the ones in the in vitro cultivation on C 16 alone, and thus the regulated 232 Dietzia enzymes relating to the enhanced n-alkane biodegradation might not only be acetyl-CoA C-233 8

acetyltransferase. 234
To learn what the exchanged compounds the two strains were during their co-cultivation on the 235 n-alkane mixture consisting of C 14 , C 16 and C 28 , we did a second-time constraint-based steady-state 236 analysis of iBH1908, in which the reaction of 'PSA3_biomass_equation_nonsugar' was still taken 237 to be the objective function, but the uptaken carbon sources were modified to be C 14 , C 16 and C 28 238 with the uptake rates being set to be 0.11 mmol g -1 h -1 , respectively (higher than those in iBH925 239 (0.10 mmol g -1 h -1 ), because C 14 , C 16 and C 28 removal efficiency in the co-culture was significantly 240 higher than that in Dietzia sp. DQ12-45-1b monoculture according to our in vitro experimental data 241 ( Fig S3b). The flux of Dietzia sp. DQ12-45-1b biomass equation on the n-alkane mixture was fixed 242 equal to the specific growth rate obtained in iBH925 (0.12 h -1 ) considering the growth curves of the 243 strain as shown in Fig S3a. By using FBA, 16 compounds were predicted to be changed between 244 the two submodels when the n-alkane mixture was used as the sole carbon input, among which, as 245 shown in Table S7, Dietzia sp. DQ12-45-1b provided nine metabolites for P. stutzeri SLG510A3-8 246 growth (R-3-hydroxybutanoate, α-ketoglutarate, glycerol, hexadecanoate, hexadecenoate, L-proline, 247 octadecanoate, formate and glycolaldehyde), while P. stutzeri secreted another seven compounds 248 for Dietzia sp. utilization (L-glutamate, acetate, glycine, 2-oxobutanoate, L-tyrosine, L-tryptophan 249 and uracil). To investigate the importances of in silico predicted seven and nine exchanged 250 compounds on the growth of Dietzia sp. and P. stutzeri, respectively, under n-alkane mixture, we 251 simulated biomass growth on models iBH925 and iBH983 following the testing strategy in the main 252 text. The results showed that acetate and glutamate were the key metabolites secreted by P. stutzeri 253 in the consortium to enhance Dietzia sp. growth on the n-alkane mixture, while R-3-254 hydroxybutanoate, α-ketoglutarate, glycerol and hexadecanoate were key compounds secreted by 255 Dietzia sp. to support P. stutzeri growth. The predicted result here was in agreement with the in 256 silico prediction on C 16 , suggesting that the exchanged metabolites between the two strains were 257 typically similar if the carbon sources for the community were of a kind. were then subjected to the gap-filling process to allow biomass formation, in which KEGG maps 280 and RAST model corresponding to Dietzia sp. DQ12-45-1b genome were used to analyze the dead-281 end metabolites and supply information to one-by-one gap filling. 282

GEM reconstruction of P. stutzeri SLG510A3-8 283
As the first stage of the model reconstruction, the draft metabolic network was reconstructed using 284 10 iPB890, a published GEM for P. stutzeri A1501 as the template (55). The reactions from the template 285 model which were either essential non-gene associated reactions or associated to genes which are 286 found in the bidirectional pBLAST search were added to the draft model of P. stutzeri SLG510A3-287 8. The organism-specific model was reliable source of information since it had previously 288 undergone manual refinements during the process of reconstruction and validation against different 289 database, literature and experimental data. The biomass equation was developed from the biomass 290 composition information of P. stutzeri obtained from in vitro cultivation on glucose and referred 291 from literatures. Refinement of the draft model was done in an iterative processing of gap finding 292 and gap filling as described above. 293

Biomass composition determination 294
When the cultures arrived at the late exponential phases, the ~120 mL culture had a 2-time 295 centrifugation-and-washing operation before being frozen at -80°C for 2 days and being dried under 296 vacuum for 3 days. Around 0.01 g of the dry biomass was processed with TRIzol® reagent to extract 297 RNA and the cellular proteins step by step. And the extracted RNA was purified via DNase and re-298 extracted via TRIzol® reagent to exclude the effect of DNA on RNA content measurement. Another 299 ~0.01g dry biomass was processed with Tris-phenol and chloroform-isopentanol to extract DNA, 300 and the DNA purification was operated by using the Bioteke Corporation kit. Around 0.1g of the for P. stutzeri SLG510A3-8. Another 0.1 g dry biomass was used for the polar lipid extraction based 305 on the method modified from the Folch procedure (61). The solution of the extracted polar lipids 306 was separated into two parts: one was used for the compound profile determination via Q-307 Exactive TM UPLC-MS (Therom Scientific, MA, US), and the other one was blow-dried with 308 nitrogen for the determination of the cellular lipid content (g g -1 ). Another ~0.1g dry biomass was 309 used for the fatty acid extraction according to our tranesterification method. The profile and the 310 relative contents of the extracted fatty acids were determined via HPLC (SHIMADZU, Japan). To 311 measure the ash content (g g -1 ), ~0.1 g dry biomass was burned in a muffle furnace at 505°C for 30 312 11 min. RNA and DNA contents in the extracted solutions (ug ml -1 ) were measured by using ultraviolet-313 visible spectroscopy. The formulas are listed as follows: 314 DNA content (ug/ml) = OD 260 × dilution rate × 50 315 RNA content (ug/ml) = OD 260 × dilution rate × 40 316 in vitro cultivating experiments to obtain the specific biomass growth rates and substrate 317 consuming rates 318 For Dietzia sp. DQ12-45-1b, cells were inoculated into the minimal medium with 8 g L -1 glucose 319 and 0.7734 g L -1 C 16 , respectively, as the sole carbon compound (initial OD 600 =0.1) and the cultures 320 were aerobically kept at 150 rpm and 30 °C for 15 and 18 days, respectively. For cultures containing 321 glucose, samples were taken every day for the measurement of cell density in terms of g L -1 and the 322 responding residual glucose contents. For cultures containing C 16 , samples were taken at Day0, 323 Day0.5, Day2, Day6.5, Day8.5, Day11.5, Day14.5 and Day18 for the measurement of cell density 324 (g L -1 ) and C 16 content. The collected data were processed using nonlinear curve fitting followed by 325 differentiation in Origin 8.5 (OriginLab Co., MA, US) to obtain the time courses of specific growth 326 rates (h -1 ) and substrate consuming rate (mmol g -1 h -1 ). 327 For P. stutzeri SLG510A3-8, cells were inoculated into the minimal medium with 1 g L -1 glucose, 328 and aerobically cultivated for 9 hours. The cultivation condition and data analysis methods were the 329 same with those for Dietzia sp. DQ12-45-1b. 330

BIOLOG experiments 331
Biolog GEN III MicroPlate TM was used for the determination of the available carbon lists of Dietiza 332 sp. DQ12-45-1b. According to its introductions for use, 71 carbon source utilization assays were 333 involved in the test panel. The cells were harvested from the LB broth at the early exponential phase 334 (OD 600 =6), and then were washed twice with 0.8% NaCl solution and resuspended with NaCl 335 solution for a 30-min starvation, followed by a third washing and resuspending. The cell suspension 336 was inoculated into the GEN III Microplate, 100ul per well with the initial OD 595 of 0.03, and the 337 panel was then kept at 30°C for 48 hours. The OD595 values were read at the 0 th , 24 th and 48 th hours 338 Supplementary Tables   Table S1. List of metabolites shared by Dietzia sp. DQ12-45-1b and P. stutzeri SLG510A3-8 in   Table S3. List of P. stutzeri SLG510A3-8 proteins, relating to C 16 , hexadecanoic acid, 3hydroxybutanoic acid, α-ketoglutaric acid, acetate and L-glutamate metabolic pathways, whose expression were significantly high-level and low-level up-(P-HU# and P-LU#, respectively) and down-regulated (P-HD# and P-LD#, respectively) when the cells were co-cultivated with Dietzia  Figure S1. Growth of Dietzia sp. DQ12-45-1b cells with glucose and C 16 as the sole carbon sources, respectively. The time courses of cell density (g L -1 ) grown on glucose and residual glucose content (g L -1 ) (a); the time courses of the relative cell growth rate (h -1 ) and glucose relative consumption rate (mmol g -1 h -1 ) for the strain (b); the time courses of cell density (g L -1 ) grown on C 16 and residual C 16 content (g L -1 ) (c); The time courses of the relative cell growth rate (h -1 ) and C 16 relative consumption rate (mmol g -1 h -1 ) for the strain (d).
28 Figure S4. Hydrocarbon composition in the residual diesel oil after a 15-day disposal by using P. stutzeri individually (Treatment No.2) and using Dietzia sp. individually with the addition of slightly amounts of sodium acetate and sodium glutamate (Treatment No.4).  Figure S5. Assignment of the iBH925 metabolic reactions to the metabolic subsystems 30 Figure S6. Assignment of the iBH983 metabolic reactions to the metabolic subsystems 31 Figure S7. The time courses of P. stutzeri SLG510A3-8 cell density (g L -1 ) grown on glucose and residual glucose content (g L -1 ) (a); the time courses of the relative cell growth rate (h -1 ) and glucose relative consumption rate (mmol g -1 h -1 ) for the strain (b).