Continuous extraction and concentration of secreted metabolites from engineered microbes using membrane technology

Microalgal cultivation in photobioreactors and membrane separations are both considered sustainable processes. Here we explore their synergistic combination to extract and concentrate a heterologous sesquiterpenoid produced by engineered green algal cells. A hydrophobic hollow-fiber membrane contactor was used to allow interaction of culture broth and cells with a dodecane solvent phase to accumulate algal produced patchoulol. Subsequent continuous membrane extraction of patchoulol from dodecane enabled product concentration in a methanol stream as well as dodecane recovery for its reuse. A structure-based prediction using machine learning was used to model a process whereby 100% patchoulol recovery from dodecane could be achieved with solvent-resistant nanofiltration membranes. Solvent consumption, E-factor, and economic sustainability were assessed and compared with existing patchoulol production processes. Our extraction and product purification process offers six- and two-orders of magnitude lower solvent consumption compared to synthetic production and thermal-based separation, respectively. Our proposed methodology is transferable to other microbial systems for the isolation of high-value isoprenoid and hydrocarbon products.

The strive to create environmentally sustainable processes in the chemical and 47 biochemical industries has led to the definition of the twelve principles of green 48 chemistry, which has also been extended to chemical engineering and bio-processes. 1

49
Sustainable sourcing of chemicals is important for global drives towards resource 50 circularity and a reduction of the environmental impact of human activities. Transfer of 51 genetic modules for specialty metabolism from a progenitor organism into an easier to 52 handle microbial system through synthetic biology now allows the production of 53 otherwise difficult to scale specialty chemicals in controlled bio-processes 2 . One of the 54 most established fields in synthetic biology is metabolic engineering of heterologous 55 isoprenoid production 3, 4 . Fermentative bacterial or yeast cell expression systems have 56 been commonly used for this purpose, owing to their genetic tractability. Recently, the 57 green alga Chlamydomonas reinhardtii has also emerged as an alternative and more 58 sustainable isoprenoid production host, with demonstrated sesqui-and diterpene 59 production 5-8 . Some microalgae, like Botryococcus braunii, also naturally secrete 60 hydrocarbon compounds, including alkadienes, alkatrienes, or polyterpenoid 61 botryococcenes 9 . Isoprenoid products are largely hydrophobic molecules and 62 accumulate in small amounts in microbial hosts unless a physical sink is provided to 63 generate pull and enable forward metabolic reactions 5 . This physical sink is achieved 64 at lab scale with hydrophobic bio-compatible solvents like dodecane (C12) 10 . 65 However, the microbial culture-solvent system is difficult to scale as hydrophobic parts 66 of the microbial culture can form turbid emulsions with the solvent, limiting practical 67 implementation 11 . In order to scale such systems to sustainable bio-processes, stable 68 culture-solvent interactions must be developed and coupled to efficient downstream 69 processes which minimize chemical waste generation in the isolation of target 70 chemicals. 71 Hollow-fiber (HF) nanofiltration membranes enable high-surface-area interactions of 72 aqueous and solvent phases in compact modules. Here, we investigated the use of 73 hydrophobic HF membrane contactors to act as a high-surface-area, physical 74 interaction matrix between dodecane solvent and engineered green algal cells to 75 continuously extract the heterologous isoprenoid product patchoulol. Upon extraction, 76 patchoulol accumulates in the dodecane solvent, which requires further processing to 77 yield the target product. Further concentration and purification must be efficient and 78 generate as little waste as possible. We then further investigated coupling our algal 79 isoprenoid extraction process to organic solvent nanofiltration (OSN) as an efficient 80 downstream process module to recover dodecane solvent and isolate the patchoulol 81 product. OSN is a continuous, pressure-driven, low energy, low waste, liquid-liquid, or 82 liquid-solute membrane separation technology. 12, 13 Chemical concentration, catalyst 83 and solute recovery, solvent recycling, and product purification have been 84 accomplished using OSN in the petrochemical and pharmaceutical industries. Natural 85 product extraction 14-16 and concentration 17 , as well as biomass processing 18   To investigate the potential of hollow-fiber technology to extract terpenoid products 129 from microbial cultures, the UVM4 strain 24 of the green model microalga 130 Chlamydomonas reinhardtii was engineered through previously reported genetic 131 strategies 5 to produce the sesquiterpenoid patchoulol and use nitrate as a sole 132 nitrogen source. 25 Before being used in the hollow-fiber, the culture was routinely 133 maintained on TAP NO3 agar plates and then transferred into 24-well plates and 134 subsequently into a 125 mL Erlenmeyer flask under growth conditions previously 135 described 26 . 20 mL of dense culture was added to 1.98 L of TAP NO3 media in a 2 L 136 Erlenmeyer flask. The flask was then connected to the HF cartridge to extract 137 patchoulol from the algae (see Fig. 1 and Suppl. The culture medium in the HF setup was regularly either replenished (100 mL added) 175 or 95% of the liquid culture (1.9 L) was exchanged with fresh TAP medium (see Fig.  176 2c for schedule).
The waste generated in the described process was the consumed but not recycled 286 solvent and the membrane. The consumption of the membrane was assumed to be 287 1 g year -1 . The solvent consumption (SC) is the ratio of the mass of solvent waste 288 generated and the mass of the isolated product (Eqn 5). 289 = kg of solvent waste generated kg of isolated product ( kg kg ) The economic sustainability (ES) is the ratio of mass of waste generated and the 290 economic value of the isolated product (Eqn 6). 291 = kg of waste generated value of isolated product ( kg $ ) 292 293

Hollow-fiber membrane contactor to extract algae-derived terpenoids 295
During the 60-day-long experimental period, a total of 4.3 mg patchoulol was extracted 296 from a fluctuating 2 L reservoir of cultured engineered C. reinhardtii using a HF setup 297 (Fig. 2a) suggesting they were healthier. Fv/Fm was highest in the first week, after which time it 305 rapidly decreased and remained below 0.70; even after replacement of 95% of the 306 medium. Overall, we found no evidence that patchoulol yield per day was significantly 307 correlated with the Fv/Fm (R 2 = 0.009), indicating photosynthetic health was not related 308 to sesquiterpenoid production. 309 Similarly, there was no evidence that patchoulol yield per day was significantly 310 correlated with the concentrations of nitrate (R 2 = 0.011) or total phosphorous (R 2 = 311 0.014) in the culture media. Both nutrients gradually decreased in the first seven days 312 of cultivation as the culture established (Fig. 2c). After this initial phase, nitrates were 313 almost entirely depleted within 24 h following media changes, while total phosphorous 314 concentration declined at a slower rate but did not reduce below 5 mg L -1 . 315 After 4 weeks, we observed aggregation of algal cells on the HF strands. The 316 settlement of cells meant that longer residence time at surface of the HFs. Patchoulol 317 is partially water soluble 5 , and is found both in the cells and the culture medium 5 . 318 Solvent contained within the HF should extract produced sesquiterpenoid from the 319 culture medium as well as direct cell-solvent micelle interactions at the HF pores. It is 320 possible that establishment of algal biofilm on the HF membrane enabled increased 321 extraction efficiency into the solvent during the later phases of this trial. 322 Our goal was to investigate whether extraction of patchoulol from engineered algal 323 culture was possible at all with such a HF cartridge setup. Optimization of culture 324 conditions was outside the scope of the current study, and we relied on a relatively 325 inefficient growth set-up using a stirred flask to simplify process testing. Future 326 iterations could use cultivation systems optimized for maximal biomass productivity, 327 such as small bubble chambers, wave bags, or membrane gas delivery 31 . The 328 patchoulol yield (per liter of culture) obtained in the present study is similar to yields 329 reported from solvent-culture two-phase cultivations in shake flasks; perhaps 330 extraction efficiency would have continued to increase with increased biofilm 331 establishment or with higher overall cell densities. Patchoulol titers of 70 g/L culture 332 after 5 days 32 and ~440 g/L culture after 7 days 5 of cultivation in shake-flasks were 333 previously reported in TAP medium. Here with the HF set-up, we were able to extract 334 470 g patchoulol L -1 culture within 7 days (on days 42-49), however, significant 335 optimizations could yet be implemented to increase yield rates of this process. 336 Coupling an HF set-up to optimized algal cultivation units may enable higher-yield 337 rates more amenable to industrial consideration. HF may also be used for direct gas 338 exchange within cultures and increase efficiency of bio-reactor performance. Further 339 parameter optimizations could also include increasing HF surface area to provide 340 more contact area between the solvent and algae cells. In addition, further engineering 341 to enhance yield of the cell catalyst will be crucial, as only a fraction of carbon (1 g L -342 1 acetic acid) was converted to patchoulol in this experiment. We sought to select an immiscible solvent alternative, which would partition patchoulol 367 from dodecane in a second-phase after extraction from the algal culture. Dodecane is 368 immiscible with water, methanol, ethanol, acetonitrile, dimethyl sulfoxide, and 369 dimethylformamide. Patchoulol has an approximately 10 4.1 lower solubility in water 370 than in dodecane and therefore water was ruled out. Dimethyl sulfoxide and dimethylformamide are not recommended solvents due to their toxicity, viscosity, and 372 potentially damaging effect on the polymeric membranes. 33 The remaining solvents, 373 methanol and acetonitrile were, therefore, candidates for extraction. 374 We opted for methanol because it is a potentially sustainable solvent 34 , cheaper to 375 produce, and it has a lower ICH limit (3,000 ppm) compared to acetonitrile. Moreover, 376 recent studies related to the rejection prediction of small organic molecules in 377 methanol allowed us to predict the rejection of patchoulol. 28, 29 We also found that 378 small amounts of methanol leaching into the dodecane phase would not perturb algal 379 culture health (Fig. 3a). Using the open-access machine-learning prediction tool, the 380 expected rejection of patchoulol in methanol is 100%. The measured rejection in our 381 cross-flow OSN system confirmed the 100% rejection of patchoulol using Duramem 382 300 membrane. The results show that the recovery of methanol is possible with an 383 OSN system, which can significantly reduce solvent loss and consequently the E factor 384 of the process. Patchoulol loss during the OSN process is negligible due to its 385 complete rejection from the membrane. 386 The main engineering parameters examined in sustainability Case 1 modeled here 406 are the flow rate of the methanol, the area of the membrane, and the applied pressure. 407 The initial concentration of patchoulol in the dodecane was set to 11 μg mL -1 (the 408 maximum concentration examined in Fig. 2A). First, we investigated the effect of the 409 methanol flow rate on the system and the output. The dodecane flow was set to 410 2 mL min -1 . Owing to the theoretical maximum capacity of the Zaiput module (10 mL 411 min −1 ), we selected two methanol flow rates to be examined: 2 mL min -1 (Case 1A) 412 and 8 mL min -1 (Case 1B). The concentration profiles in the dodecane, the methanol 413 inlet stream (NF feed) and in the NF efflux (retentate) are shown in Fig. 4. 414 Lower methanol flow rate resulted in higher output patchoulol concentration at the 415 same operating pressure (Fig. 4) because the solution leaving the Zaiput module was 416 less diluted. Examining the time required to extract 80% of the patchoulol from the 417 dodecane solution: case 1A requires 728 min while case 1B 424 min. However, the 418 enrichment period to yield patchoulol from the microbial culture from 2.2 μg mL −1 to 419 11 μg mL −1 in dodecane was found to be 62.9 d in our HF-culture setup, so the 420 difference in extraction time is negligible compared to the total required time to 421 accumulate product. The reduced extraction time cannot make up for the deterioration 422 of environmental factors caused by the dilution of the efflux solution. Increasing the 423 methanol flow rate results in an excessively diluted solution, and the decreased 424 extraction time has no significant effect on productivity as the enrichment phase from 425 the microbial culture is by far the slowest process. 426 Using sensitivity analysis, the effects of various engineering and system parameters 427 on the green metrics of the process were mapped out. We examined case 1A, varying 428 the extraction rate into methanol and the initial patchoulol concentration in dodecane 429 on levels 20%, 40%, 60%, 80%, and 2 μg mL −1 , 6 μg mL −1 , 11 μg mL −1 , respectively. 430 We calculated economic (E) factor (comprised of both solvent and membrane waste) 431 and economic sustainability for the different cases, with their respective results shown 432 in Fig. 5a and b. 433 Not surprisingly, we found that lower extraction rates of patchoulol from dodecane into 434 methanol and higher initial patchoulol concentrations in dodecane improved the 435 sustainability features of the process. As extractions are continued for longer time 436 frames (higher extraction rate), the methanol solution flow becomes more diluted, 437 leading to higher solvent consumption (waste), therefore lower extraction rates are 438 advantageous for process efficiency. In a real-life situation, the parameters examined above have a significant effect on the 449 cost of the process. Therefore, economic optimization is necessary to find industrial 450 relevance. Based on the results illustrated in Fig. 5a, we chose 11 μg patchoulol mL -1 dodecane as the semi-optimal initial concentration value and 20% extraction rate to 452 compare Case 1 with other cases. Fig. 5d  In case 2, the nanofiltration unit is decoupled from the extraction system. The 472 methanol solution of patchoulol can be collected in a container or in a retentate loop 473 and can be concentrated in a separate unit operation, independently from the previous 474 extraction steps. In this case, we do not have to specify the membrane area and the 475 pressure, as these are usually determined from economic metrics. In the comparative 476 assessment of sustainability features, we assume that the methanol solution is 477 concentrated to 1 mg mL −1 . 478 Case 3 describes a completely continuous configuration. Here, we assume that a 479 sufficiently large algae culture can maintain a constant concentration of 11 μg 480 patchoulol mL −1 in the dodecane solution while operating the methanol extraction cycle 481 parallelly. Similar to case 2, we assume that a retentate cycle or a buffer tank enables 482 robust concentration of the mixture to 1 mg mL −1 with an adequate nanofiltration unit. 483 Because of this, the green metrics of cases 2 and 3 will be similar, with the slight 484 difference that the higher molar efflux of case 3 enables more efficient membrane 485 usage within its life cycle, thereby reducing the environmental factor. 486 The comparison of solvent consumption and E factor for different system 487 configurations and cases is illustrated in Fig. 5e. 36  can not only increase sustainability but can also enhance the robustness of the system 504 and process control by eliminating the time factor. This approach, however, requires significantly larger algae cultivation set-ups or algae strains with increased 506 productivity. 507 508

509
In this work, we showed that a system comprised of a HF membrane contactor, 510 membrane-based liquid-liquid extraction, and OSN can be used for efficient in-line 511 extraction and concentration of patchoulol from engineered algal culture. The 512 patchoulol titers obtained in the dodecane reservoir with the HF setup were similar to 513 those previously reported for traditional two-phase extractions in flasks, however, at 514 much longer timescales. While the obtained patchoulol yields were reasonably high, 515 additional modifications to the HF setup, such as using a larger-surface-area HF 516 cartridge and coupling this to an optimized algal cultivation unit, are likely to increase 517 yields even further. Further developments in strain engineering to increase yields will 518 also likely improve titers. This process would function well with other engineered 519 microbes, which could also boost process efficiencies. We also showed the 520 implementation of "separation by design" green engineering principle nanofiltration. 521 Molecular modeling and empirical testing indicated expanding extraction with 522 methanol could enable efficient, low energy extraction of the target product, and we 523 present the efficiency of such processes for many modeled cases. Our proposed 524 system presents a low-waste and low-energy means to enrich patchoulol extracted 525 from the microbial culture. We demonstrated that the decoupled configuration and 526 operation of the membrane concentration unit helps to reduce the environmental 527 footprint of the system, maximizes efficiency, and enables an inherently robust 528 downstream process. Further implementations of this concept will require 529 improvements in algal-HF membrane contact and target metabolite productivity to 530 maintain a fully continuous process.
Molekulare Pflanzenphysiologie, Potsdam. The research reported in this publication 557 was supported by the KAUST Impact Acceleration Funds program (grant 4238), and 558 KAUST baseline funding awarded to Kyle Lauersen and Gyorgy Szekely. Comparison of green metrics for main cases of process configuration. 675