Biodegradation of poly(butylene succinate) in soil laboratory incubations assessed by stable carbon isotope labelling

Using biodegradable instead of conventional plastics in agricultural applications promises to help overcome plastic pollution of agricultural soils. However, analytical limitations impede our understanding of plastic biodegradation in soils. Utilizing stable carbon isotope (13C-)labelled poly(butylene succinate) (PBS), a synthetic polyester, we herein present an analytical approach to continuously quantify PBS mineralization to 13CO2 during soil incubations and, thereafter, to determine non-mineralized PBS-derived 13C remaining in the soil. We demonstrate extensive PBS mineralization (65 % of added 13C) and a closed mass balance on PBS−13C over 425 days of incubation. Extraction of residual PBS from soils combined with kinetic modeling of the biodegradation data and results from monomer (i.e., butanediol and succinate) mineralization experiments suggest that PBS hydrolytic breakdown controlled the overall PBS biodegradation rate. Beyond PBS biodegradation in soil, the presented methodology is broadly applicable to investigate biodegradation of other biodegradable polymers in various receiving environments.


Supplementary Note 1 21
The three position-specifically 13 C-labelled PBS variants showed indistinguishable final 22 cumulative 13 C mineralization extents (i.e., Cmineralized) after 425 days of soil incubation. We 23 therefore concluded that microorganisms utilized carbon from the different positions in PBS to 24 comparable extents. At the onsets of the soil incubations, however, the three PBS variants 25 showed different mineralization rates, as discussed in detail in the Results section of the main 26 text. We ascribed the differences in mineralization rates of the three PBS variants during the 27 initial 7 days of incubation (i.e., initial mineralization rates decreased in the order PB(1,4-13 C2-28 S) > P( 13 C4-B)S > PB(2,3-13 C2-S)) to differences in the microbial utilization of B and S 29 monomers and low molecular weight BS oligomers that were present in the bulk PBS and that 30 readily diffused out of the PBS into the soil. For a better visualization of these differences 31 between the three PBS variants, we replotted the PBS mineralization rates and extents from 32  Figure 1 shows that from 37 7 to about 40 days of incubation, 13 C mineralization rates differed among the three variants in 38 the same order as observed during the first 7 days of soil incubation (i.e., PB(1,4-13 C2-S) > 39 P( 13 C4-B)S > PB(2,3-13 C2-S)). We ascribe the differences during this initial time period to the 40 formation of microbial biomass in the soil with monomer-and position-specific extents to 41 which 13 C was incorporated into microbial biomass, as discussed in more detail for the 42 microbial utilization of the 13 C-labelled monomers in the soil (i.e., Figure 1c and d in the main 43 text). 44 After about 40 days of incubation, mineralization rates were comparable between the 45 three tested PBS variants. We interpret the apparent absence of variant-dependent 46 mineralization rates during the later stages of incubation to reflect extensive conversion of 47 PBS-derived 13 C to 13 CO2, irrespective of the position in PBS that carried the label. Extensive 48 conversion to CO2 is supported by the finding of only small Cbiomass at the end of the incubation. 49 In addition, the modeling results suggest that biomass was formed only during the first 2-3 50 months of incubation. Extensive conversion to CO2 beyond 100 days of incubation has two 51 potential explanations. The first explanation is that the carbon use efficiencies (CUEs) were 52 low and hence PBS-derived 13 C was primarily catabolic utilized and converted to 13 CO2. The 53 second explanation is that the absence of observed position specificity in mineralization 54 reflected that the rate at which PBS-derived 13 C cycled through the microbial biomass pool was 55 larger than the rate at which PBS-derived carbon was provided to the microbial cells. In this 56 second explanation, position-and monomer-specific utilization of carbon to form biomass may 57 have continued beyond 100 days of incubation but was kinetically masked by the low supply 58 rate of PBS-derived carbon to microbial cells. 59 As discussed in the main text, the causes for the decrease in PBS mineralization rates 60 (and hence also the supply rate of PBS-derived carbon to microbial cells) beyond 100 days of 61 soil incubation remain unclear. It is conceivable that the residual PBS was more difficult to 62 enzymatically hydrolyze. It is also possible that conditions formed in the soils that started to 63 limit the activity and possibly number of microbial degraders. For instance, limitations in N 64 and P may have slowed down the formation of additional biomass and thus resulted in the 65 preferential utilization of PBS carbon for energy production under formation of CO2. triplicates up to 319 days (indicated by vertical grey dotted lines) and duplicates from 319 to 75 425 days. Data is replotted from Figure 1a, b in the main text, respectively, but plotted with 76 both x and y axes converted to log-scales to highlight differences in the mineralization 77 dynamics of the three PBS variants during the first two to three months of soil incubation. 78 79 80 Supplementary Note 2 81 The three different position-specifically 13 C-labelled PBS monomers (i.e., 1,4-13 C2-S; 2,3-13 C2-82 S; and 13 C4-B) showed distinct mineralization rates and extents during soil incubations, as 83 described in the main text. Each of the three monomers reached a maximum in mineralization 84 rate within the first few hours after monomer addition to the soil (Figure 1c, main text, inset). 85 To facilitate comparisons of the mineralization dynamics of the three monomers, we have 86 replotted the data from Figure 1c  (PBS)-monomers. Mineralization rates of the two 13 C-labelled succinate monomers (i.e., 1,4-93 13 C2-S, blue up-triangles; and 2,3-13 C2-S, red down-triangles) and of the 13 C-labelled 1,4-94 butanediol ( 13 C4-B, green circles), normalized to the maximum mineralization rates of the 95 respective incubation bottle. Mineralization of the monomers was continuously followed in 96 triplicate incubation bottles (for which measurements of each individual bottle are shown as a 97 separate curve) for a total incubation time of 14 days. Note that this data is identical (except 98 for the normalization) to the mineralization rate data shown in Figure 1c of the main text. 99 100

101
Supplementary Note 3 102 We rationalize the strong position-specificity of 13 C-mineralization rates of succinate with the 103 metabolic flow of succinate carbons in the citric acid cycle (CAC) of aerobic soil 104 microorganisms. 1 Figure 3a). Conversely, two molecules of 1,4-13 C2-113 succinate are thus converted to one molecule of succinate preferentially reduced in the 13 C-114 label and three molecules of 13 CO2 (plus one of 12 CO2) (Supplementary Figure 3b). 115 These differences in the metabolic utilization of the carboxylate and methylene carbons 116 in succinate can explain the extensive mineralization of 13 C in 1,4-13 C2-S to 13 CO2, whereas 117 much less 13 CO2 was formed from 2,3-13 C2-S. The latter implies that the 13 C in 2,3-13 C2-S was 118 extensively incorporated into microbial biomass (i.e., formation of Cbiomass, presumably by 119 synthesis of fatty acids and lipids from acetyl-CoA and of carbohydrates from oxaloacetate; 120  We used soil samples to which we added known amounts of P( 13 C4-B)S particles to 140 develop and validate the chloroform-sonication treatment step to uniformly distribute of PBS-141 13 C in soils prior to subsampling the soil for EA-IRMS analyses. This treatment step and the 142 resulting accurate and precise quantification of PBS-13 C has been discussed in the Results 143 section (Figure 2) in the main text. We restricted our discussion in the main text to the P( 13 C4-144 B)S recovery data for three soil samples with concentrations of 1, 3, or 10 mg P( 13 C4-B)S per 145 10 g of soil. We compared recoveries from soil subsamples that were collected from a soil that 146 was not chloroform-sonication-treated and from 5 g of soil after it was chloroform-sonication-147 treated. The subsamples collected for EA-IRMS analysis had a mass of 10 mg. 148 Here we present results of additional addition-recovery assessments for the soil sample 149 with the intermediate concentration of 0.3 mg P( 13 C4-B)S g -1 soil. In these additional 150 assessments, we chloroform-sonication treated only 1 and 3 g soil samples (instead of 5 g) of 151 the soil. As before, we analyzed five 10 mg aliquots of each of the chloroform-sonication 152 treated soil subsamples using EA-IRMS. 153 Supplementary Figure 4 shows the recovery of the added P( 13 C4-B)S for the five replicate 154 aliquots taken from the 1, 3 and 5 g of soil that we treated by chloroform-sonication (note that 155 the 5 g data is replotted from Figure 2 in the main text). As discussed in detail in the main text, 156 we achieved complete recovery when chloroform-sonication treating a 5 g soil subsamples 157 (i.e., recovery of added P( 13 C4-B)S of 97 ± 4% (mean ± standard deviation of five replicates)). 158 We also achieved full recovery of added P( 13 C4-B)S when treating only 3 g of soil by 159 chloroform-sonication, followed by EA-IRMA analysis of five aliquots (i.e., recoveries of 101 160 ± 3% (mean ± standard deviation of five replicates)). However, when we chloroform-161 sonication treated only 1 g of soil that had a concentration of 0.3 mg P( 13 C4-B)S g -1 soil, the 162 recovery was incomplete (i.e., 89 ± 1% (mean ± standard deviation of five replicates)). While 163 the recoveries were inaccurate, they were precise. This finding implies that taking and 164 subsequently chloroform-sonication treating only 1 g subsamples from the total of 10 g of soil 165 to which we had added 3 mg P( 13 C4-B)S g -1 was too small to ensure that the amount of P( 13 C4-166 B)S particles in the homogenized sample was representative of the complete soil. Based on this 167 finding, we chose to chloroform-sonication treat 5 g soil subsamples from the PBS incubation 168 experiments to accurately quantify Cnon-mineralized using EA-IRMS. 169 butanediol-13 C labelled PBS (i.e., P( 13 C4-B)S) from a soil sample with a total amount of 10 g 173 and a concentration of 0.3 mg P( 13 C4-B)S g -1 soil (obtained by adding 3 mg P( 13 C4-B)S to the 174 10 g of soil). The recovery was determined by EA-IRMS on five replicate aliquots (10 mg) that 175 were collected from 5, 3, or 1 g chloroform-sonication treated soil subsamples of the 10 g soil. 176 The where RF is the PBS response factor (i.e., the relationship between peak height and PBS 198 amount which we determined by analyzing the PBS standards (= 3212 mg -1 PBS mL)). Based 199 on the peak responses of these standards, we determined LOD and LOQ of 3.2 and 10.7 μg 200 PBS mL -1 according to Supplementary Equations 1 and 2, respectively. The 1:1 linear 201 relationship between measured and nominal PBS concentrations and the low LOD and LOQ 202 demonstrate that quantitative 1 H NMR allows for selective, accurate, sensitive detection of 203 PBS in deuterated chloroform. 204 We subsequently developed a Soxhlet extraction method to recover PBS from the same 205 soil that we used in the incubation experiments. To this end, we added 2.5 mg of non-labelled 206 PBS to 2.5 g of soil, thereby obtaining the same PBS concentration in the soil that we used at 207 the start of the actual soil incubation experiments. For the analyses of extraction efficiencies, 208 we added PBS in two different forms: as particles (> 300 μm diameter) or dissolved in 209 chloroform (i.e., 100 µL of a PBS solution spiked onto the soil surface). We included the 210 second type of addition to allow for maximum interactions between dissolved PBS strands and 211 soil particle surfaces after evaporative removal of the chloroform. We subsequently Soxhlet 212 extracted these soils with a chloroform/methanol solvent mixture (90/10 volume%), dried the 213 extract, and then reconstituted the dried extract into deuterated chloroform with DMB as 214 internal standard following a method recently introduced. 4    between different pools with the following ordinary differential equations in COPASI. 263 For cellulose, we assumed the depolymerization to be first order with respect to the 264 amount of cellulose remaining, Ccellulose (μg C): 265 where k1 is the depolymerization rate constant (h -1 ) and Fcolonized represents the fraction 266 of the polymer surface which is accessible for enzymatic cellulose hydrolysis (see below for 267 more details). 268 For PBS, we slightly modified this rate equation. We added PBS to soils in the form of 269 small particles (100-300 μm diameter, obtained by cryomilling the synthesized bulk PBS 270 material) and assumed that hydrolytic breakdown (by extracellular microbial esterases) only 271 occurred on the particle surfaces. For this reason, we set the depolymerization rate to be two-272 thirds order with respect to the amount of PBS remaining, CPBS (μg C). This modification was 273 based on the assumption that PBS particles were spherical, and that continuous hydrolysis 274 progressively decreased the radius of the PBS particles. In this case, the surface area of a 275 sphere, SAS (cm 2 ), depends on the sphere volume, VS (cm 3 ), according to: 276 Where r (cm) is the radius of the sphere. 277 The PBS depolymerization rate equation is then given as: 278

279
where CPBS (μg C) is the amount of PBS remaining at any given time. 280 We fitted the biodegradation data of PBS in two ways. First, we fitted the data using a single 281 and constant value for k1 over the course of the incubation. Second, we allowed for a decrease 282 in the effective PBS breakdown rate constant k'1 (h -1 ) by multiplying k1 (h -1 ) with a factor that 283 exponentially decayed from unity over time: 284 where m1 (h -1 ) is a fitted kinetic parameter that describes the decrease in the effective hydrolytic 285 breakdown rate constant of PBS and t (h) is the time during the incubation. 286 Both PBS and cellulose showed an initial lag phase in mineralization following their 287 addition to the soil (i.e., the mineralization curves showed an upward curvature during the first 288 days of the incubations). We ascribe this lag phase to the initial time needed for microbial 289 colonization of the PBS and cellulose surfaces and the secretion of microbial extracellular 290 hydrolases. We modeled this lag phase by the term Fcolonized by allowing it to increase from 0 291 to unity according to the following function: 292 where m2 is a fitted kinetic rate constant (h -1 ) that describes the overall rate at which Fcolonized 293 increased over time. 294 The carbon from the polymers, Cpolymer, was hydrolytically broken down into low 295 molecular weight products (monomers and short oligomers) that we considered to be readily 296 utilizable by the soil microbes (i.e., labile carbon, Clabile). We modeled uptake and utilization 297 with a first order rate law with respect to Clabile (μg C) and a rate constant k2 (h -1 ). The rate of 298 change of the size of the Clabile pool was therefore described the following equation that 299 accounted for both the input to and output of carbon from this pool: where γ = 1 for Ccellulose and γ = 2/3 for CPBS. 301 302 Microorganisms utilize the carbon in Clabile both catabolically under formation of CO2 303 (i.e., direct transfer of carbon to Cmineralized) and anabolically to form new biomass (i.e., transfer 304 of carbon to Cbiomass). The fraction of utilized carbon which was incorporated into biomass is 305 defined as the carbon use efficiency (CUE). Finally, we also allowed that PBS-and cellulose-306 derived carbon in Cbiomass became mineralized, according to a first order rate law with respect 307 to Cbiomass and the rate constant k3 (h -1 ). 308 Overall, the carbon fluxes through Cbiomass, Cmineralized from polymer, and Cmineralized from biomass 309 were then defined as: 310

313
The total flux into Cmineralized was then simply taken as the sum of the two separate sources of 314 Cmineralized: 315 For modeling the PBS and cellulose biodegradation data, we set the starting amounts 316 of Cpolymer to the mass of PBS-and cellulose-13 C added (in μg; mean value for triplicates of 317 each PBS variant or of cellulose). 318 We also modeled the monomer biodegradation data using the same model but omitted 319 the polymer breakdown step (i.e., we directly ascribed the monomers to the Clabile pool). We 320 set the starting amount of Clabile to the mass of monomer-13 C added to the soils in the incubation 321 experiments (in μg; same value added per triplicate for each labelled monomer). 322

II. Kinetic parameter estimation 324
We simultaneously fitted the data from triplicate incubations to obtain single values of each 325 model parameter for each substrate tested (i.e., each of the three PBS variants, each of the 13 C-326 labelled monomers, and for cellulose). To this end, we imported the experimental data (i.e., 327 Cmineralized and for some of the PBS model fits also Cpolymer residual) into the "Parameter 328 Estimation" task option of COPASI. We subsequently ascribed the experimental 329 mineralization data to the total Cmineralized pool (Supplementary Equation 12) in the model. For 330 long-term incubations of PBS and cellulose, we selected a subset of the total mineralization 331 measurement points that we collected with a high measurement frequency during the early 332 phase of the incubations (up to 70 days for cellulose and 60 days for PBS) to allow for a more 333 even distribution of measurement data across the entire incubation period. This procedure 334 ensured that model fits were not biased by the more-frequently collected data during the initial 335 phase of incubations. In Supplementary Table 2  10 -6 10 -2 10 +6 10 -6 5 · 10 -3 10 +6 k2 (h -1 ) 10 -6 10 -2 10 +6 10 -6 10 -2 10 +6 k3 (h -1 ) 10 -6 10 -4 10 +6 10 -6 10 -4 10 +6 10 -6 1 · 10 -4 10 +6 m1 (h -1 )

N/A N/A N/A
10 -6 10 -2 10 +6 10 -2 10 -2 10 +6 CUE 0.01 0.5 1.00 0.01 0.1 0.60 10 -6 0.10 0.40 a L.L. = lower limit set for the selected parameter 341 b start = the start value used for the parameter optimization 342 c U.L. = upper limit set for the selected parameter 343 d k2 in each PBS fit was set to the k2 value that was fitted for the mineralization data of the 344 respective monomer 345 346 We fitted the data using the Levenberg-Marquardt optimization method with an iteration limit 347 of 10,000 steps and a tolerance level of 10 -6 (i.e., the optimization stopped when the difference 348 in the objective function value between two iteration steps was smaller than the tolerance 349 level). 350

IV. Kinetic modeling results 359
Fitting of monomer data. We first fitted the monomer biodegradation data because 360 biodegradation of these compounds required no hydrolytic breakdown step. Supplementary 361 Table 3 lists the optimized parameters for model fits to each of the three labelled monomers. 362 We note that the rate constant for turnover of Cbiomass (k3) was comparable for all three 363 monomers, as expected even though the carbon was microbially utilized in a monomer-and 364 position specific manner. 365 The changes in the amount of carbon from the monomers in the different carbon pools over C monomer C biomass C mineralized C mineralized from monomer C mineralized from biomass C monomer C biomass C mineralized C mineralized from monomer C mineralized from biomass C mineralized C monomer C biomass C mineralized C mineralized from monomer C mineralized from biomass C mineralized monomer mineralization

Supplementary
The model adequately fits the rapid monomer-and position-specific mineralization of 384 the three monomers in the soil. The modelling results are fully consistent with our 385 interpretation that the carboxylate carbons of succinate were readily mineralized to CO2 (i.e., 386 low CUE and high k2) while the methylene carbons of the same monomer were preferentially 387 incorporated into Cbiomass (i.e., high CUE). 388 Fitting of PBS data. To fit the PBS mineralization data, we set k2 to the values obtained 389 from fitting the respective monomer mineralization data (Supplementary Table 3 The carbon pools were calculated from optimized kinetic parameters (Supplementary Table 4 Measured values C polymer C biomass C mineralized C mineralized from polymer C mineralized from biomass C polymer C biomass C mineralized C mineralized from polymer C mineralized from biomass C mineralized C polymer residual C polymer C biomass C mineralized C mineralized from polymer C mineralized from biomass Fitted model parameters PB(1,4-13 C2-S) 0.61 1.1 · 10 -3 2.7· 10 -1 2.3 · 10 -3 1.0 · 10 -6 156.74 PB(2,3-13 C2-S) 0.07 1.2 · 10 -3 1.0 · 10 -4 2.1 · 10 -3 3.9 · 10 -4 165.33 P( 13 C4-B)S 0.13 1.2 · 10 -3 1.2 · 10 +2 2.4 · 10 -3 1.0 · 10 -6 155.55 Measured values C polymer C biomass C mineralized C mineralized from polymer C mineralized from biomass C polymer C biomass C mineralized C mineralized from polymer C mineralized from biomass C mineralized C polymer residual C polymer C biomass C mineralized C mineralized from polymer C mineralized from biomass C mineralized C polymer residual residual polymer incl.
As compared to the model runs in which we fitted only Cmineralized, the quality of the model fit 441 was lower when we included the extraction data, Cpolymer residual, in the fitting process. In a third model, we accounted for an unidentified constraint that started to limit PBS 447 biodegradation during the incubation. The evidence for such a limitation is discussed in the 448 main manuscript. We implemented the limiting constrain into the model mathematically by 449 introducing an exponential decaying factor that is multiplied with the rate constant of PBS 450 hydrolytic breakdown, k1, to result in a decrease in the 'effective' PBS hydrolytic breakdown 451 rate over the course of the incubation (Supplementary Equation 6). For the model fit shown, 452 we set the CUE to 0.4. However, as shown below in the sensitivity analysis, the experimental 453 data could be well described also with slightly higher (but less realistic) CUE values (tested up 454 to 0.6). The optimized model parameters are provided in Supplementary Table 6. The finding 455 that the data can be fitted with a range of (high) CUE values is consistent with our interpretation 456 that the rate at which PBS-derived carbon cycles through the microbial biomass is larger than 457 the rate at which PBS-derived carbon becomes available to microorganisms. In this case, the 458 buildup of Cbiomass is only small and can be described by varying CUE values. Supplementary 459 Figure 9 shows the modeled carbon pools calculated based on the optimized model parameters.
PB(1,4-13 C2-S) 0.61 3.7 · 10 -3 2.5 · 10 -4 1.9 · 10 -3 1.0 · 10 -3 0.40 66.47 PB(2,3-13 C2-S) 0.07 3.7 · 10 -3 2.2 · 10 -4 0.8 · 10 -3 1.0 · 10 -3 0.   Figure 1). We ascribe these differences in the mineralization of the three PBS variants to the 490 monomer-and position-specific utilization of carbon atoms for microbial biomass formation, 491 fully consistent with the fitted net increase in microbial biomass (i.e., variant and monomer-492 specificity in carbon uptake into microbial biomass was reflected in the Cmineralized data as long 493 as Cbiomass increased). As the incubations progressed, the modeled hydrolytic breakdown rate 494 of PBS decreased over time and, consequently, the rate at which PBS-derived carbon was 495 supplied to the microbial cells became smaller than the rate at which PBS-derived carbon 496 cycled out of the microbial biomass pool through biomass mineralization. Consequently, the 497 modeled size of Cbiomass pool decreased to a small final pool size at the end of the incubations, 498 consistent with the experimental data (i.e., most of Cnon-mineralized was explained by Cpolymer 499 residual). We expect that the position-specificity of carbon utilization was kinetically masked by 500 the supply of PBS-derived carbon to the microorganisms (i.e., the supply rate was smaller than 501 the rate at which this carbon cycled through the biomass). 502 Fitting of cellulose data. We fitted the cellulose mineralization data using the first model 503 described for PBS above and assuming a single, constant depolymerization rate constant k1. 504 We note that Cpolymer residual was not experimentally accessible for cellulose (i.e., there is no 505 analytical approach to readily extract cellulose from soils). The optimized model parameters 506 are provided in Supplementary

512
The cellulose mineralization data is well described by the model. We infer from the 513 model output that the biphasic mineralization of cellulose has the following explanations. The 514 first initial rapid mineralization phase reflects direct catabolic utilization of cellulose-13 C to 515 form 13 CO2, while 13 C is also incorporated into microbial biomass with a fitted CUE of 0.37 516 (in agreement with our estimate and published values). 5,6 By contract, the second phase with 517 smaller mineralization rates is ascribed to the mineralization of microbial biomass and SOM 518 that contained cellulose-derived 13 C. The outcome of model fit therefore supports our 519 interpretation of the cellulose mineralization provided in the main text. 520 521

V. Sensitivity analysis 522
For the third PBS model above (Supplementary Table 6 the objective function value when it was increased from its optimal value, but a strong change 540 when it was decreased. In this case, we had originally set the values of k2 to relatively high 541 values based on the results of the modeling of monomer mineralization data. As a result, 542 increasing k2 showed no effect on the model fit, while decreasing k2 had a substantial effect 543 Lastly, the model showed very low sensitivity to variations in CUE away from its optimum 544 value. Consistently, the experimental data could be fitted over a range of maximum CUE 545 values, for reasons already discussed above. We note that Cmineralized had two distinct 546 contributions: CO2 produced from catabolic utilization of Clabile and CO2 produced from 547 mineralization of Cbiomass into which polymer-derived carbon was incorporated. As such, this 548 model is likely insensitive to the value of CUE because as long as the rate of mineralization 549 from the biomass pool is faster than the initial utilization rate, the model does not strictly 550 differentiate between these two sources of CO2 (the ratio of which is defined by the CUE).