for crude glycerol tolerance
Adaptive laboratory evolution (ALE) is widely used to obtain robust microbes with high tolerance for different environmental stresses. However, the achievement of a stably improved phenotype which can fully withstand a specific stress factor or inhibitor usually requires a long-term ALE with great efforts (Zhang et al., 2019 ; Alves et al., 2021 ). In order to increase the efficiency, long-term ALE of C. pasteurianum C8 was conducted in a home-made automatic evolutionary system integrated with the real-time monitoring and control of cell growth in this study.
As shown in Fig. 1 , the OD600 of the strain during the repeated batch cultivations was automatically controlled between 0.5 and 1.5 by the ALE system. This was realized by adding of fresh medium into the cultivation bottle whenever the OD600 reached 1.5 and the addition was stopped when the OD600 reached 0.5 again. There are many advantages to keep the biomass sufficiently low in the long-term automatic ALE experiment. First, lower biomass can prevent bio-fouling in the tube connecting the cultivation bottle and the OD detector. Second, the pH of the culture can be maintained at 6.3 to 6.5 without external pH control owing to the negligible acids production by the limited biomass. It is well known that pH lower than 5.0 leads to significant growth inhibition to C. pasteurianum (Dabrock et al., 1992 ), which may affect the efficiency of the ALE experiment. However, the integration of pH control with probes and peristaltic pumps for acid / base solution requires a bigger size of bioreactor, and thus increases the working volume and medium consumption for the long-term ALE experiment. Thus the strategy employed here makes the system as simple as possible and highly efficient as well. Last but most importantly, the time needed for 1 adaptation cycle (OD600 from 0.5 to 1.5) is significantly reduced compared to those reported in the manual ALE experiment (>12 h for 1 cycle) (Liang et al., 2020 ;Zhang et al., 2019 ;Alves et al., 2021 ). An example of the automatic ALE process for 30 g/L crude glycerol is presented in Fig.1 . Totally 10 adaptation cycles were conducted within 80 h. The immediate dilution of the culture with fresh medium when the OD600 reached 1.5 not only reduced the time interval between two adaptation cycles, but also ensured that the next adaptation cycle was initiated with cells in the exponential growth phase. Undoubtedly, inoculum in good growth state is beneficial to avoiding lag phase, reducing infection risk and therefore increasing the number of adaptation cycles in the automatic ALE experiment. Based on the above-mentioned advantages, C. pasteurianum C8 was successfully adapted to each concentration of crude glycerol (30-120 g/L), and finally adapted to 120 g/L crude glycerol after more than 40 days of automatic ALE with 106 adaptation cycles. The △T for one adaptation cycle progressively decreased during the repeated cultivations, and a reduction to 4 h which was the same with pure glycerol, was achieved in the last cycle with 120 g/L (Fig.1 ). Further increase of crude glycerol concentration to 130 g/L resulted in a significantly longer △T >10 h, and the cell growth failed to recover even after 20 adaptation cycles. This might be due to the growth inhibition caused by the impurities and the osmotic pressure from 130 g/L crude glycerol, the cells were then kept in the automatic ALE system for another 10 repeated cultivations with 120 g/L crude glycerol for growth stabilization. The end-point adapted strain obtained from the automatic ALE system was named as C. pasteurianum G8 and used for characterizing the crude glycerol tolerance in the subsequent fed-batch fermentations.
As shown in Fig. 4 and Table1, fed-batch fermentations of crude glycerol by C. pasteurianum G8 were performed with two feeding strategies (continuous feeding with the feed rate of 40 g/L/h vs. pulse feeding with 80 g per feed). Meanwhile, the initial glycerol concentration was set at 80 g/L, which was the same as that in the pure glycerol fermentation. As expected, C. pasteurianum G8 showed excellent improved tolerance against the high initial concentration of crude glycerol. No lag phase was observed after inoculation and the µmax (0.52 ± 0.02 h-1 in pulse fed-batch and 0.59 ± 0.02 h-1 in continuous fed-batch) reached the same level as that in the pure glycerol fermentation. It should be noted that the same initial concentration of crude glycerol caused a long lag phase of over 10 h when the parental strain C8 was used (data not shown). Furthermore, the maximum glycerol consumption (22.19 g/L/h with pulse and 21.44 g/L/h with continuous feeding) and PDO production rate (8.74 g/L/h with pulse and 10.18 g/L/h with continuous feeding) also recovered to the same level as those achieved in pure glycerol fermentation owing to the increased utilization of crude glycerol and the fast feeding strategy. The final PDO concentration and overall productivity reached 71.47 g/L and 3.97 g/L/h with the pulse feeding (Fig. 4A), and with continuous feeding (Fig. 4B), 74.23 g/L and 5.30 g/L/h, respectively. These results clearly indicate that the automatic long-term ALE can effectively promote the production level of PDO in the crude glycerol fermentation by C. pasteurianum. Despite the similar PDO titer achieved with both feeding strategies, the yield of PDO from crude glycerol in the pulsed fed-batch fermentation was about 6% lower than that in the continuous fed-batch fermentation (0.49 vs. 0.52 gPDO/gGly). This is probably due to the stress response of the strain caused by multiple instant exposure to a large amount of crude glycerol impurities during the pulse feeding period, leading to a reduced acetate / butyrate ratio at the end of the fermentation (1.53 with pulse and 2.17 with continuous feeding). Similarly, further increasing the initial crude glycerol concentration to 120 g/L (Fig. 4C) also resulted in marked decrease of acetate / butyrate ratio to 0.81, even with continuous feeding strategy. As a result, the titer, productivity and yield of PDO were all negatively affected (Table1). Nevertheless, C. pasteurianum G8 exhibited good growth even with 120 g/L initial crude glycerol (µmax, 0.27 ± 0.01 h-1) and produced very less butanol (<1.5 g/L) in all the fed-batch fermentations, indicating much enhanced and stable tolerance to crude glycerol than that of the parental strain.
Since the fed-batch fermentation with 80 g/L initial crude glycerol concentration and with a continuous feeding of 50% crude glycerol at 40 g/L/h resulted in the best PDO production, the same strategy and conditions were used for a scale-up fermentation in 1 m3 bioreactor. As shown in Fig4D and Table1 , the highly crude glycerol-tolerant strain C. pasteurianum G8 was capable of producing 81.21 g/L PDO within 19 h, along with an overall productivity of 4.27 g/L/h and a yield of 0.49 gPDO/gGly in the 1 m3 scale non-sterile crude glycerol fed-batch fermentation without using N2 aeration and yeast extract. It should be noted that more glycerol supply with a longer feeding time resulted in a slightly higher PDO titer in the 1 m3 scale fermentation, but it also triggered higher butanol production (2.47 g/L) due to the stress induced by both high PDO concentration and crude glycerol impurities. Butanol production can be avoided and a significantly higher PDO yield of 0.53 gPDO/gGly can be reached by simply stopping the fermentation at an early processing time of 14 h (Fig4D and Table1 ). To further increase the PDO titer, automatic long-term adaptation to a high PDO concentration could be considered in future work. In general, the results above indicate thatC. pasteurianum G8 has the best potential for achieving efficient and low-cost industrial bio-production of PDO among the reported non-pathogenic natural producers so far in terms of PDO titer, yield, productivity and process economy (Table2 ).