Impact of furfural and 5-HMF on rate of biomass production
Earlier studies [16, 17] suggest that in presence of furfural, microbial cell shunts NADPH away from anabolic reactions towards detoxification of furfural which leads to decrease in biomass yields. Meanwhile, deletion of pgi gene shunts all glucose carbon via PPP and leads to excess production of NADPH . The excess NADPH itself acts as an inhibitor and leads to growth arrest, which resumes only upon introduction of a NADPH sink . We thus tested the hypothesis whether furfural and 5-HMF has potential to improve biomass formation rate in the pgi deleted E. coli strain SSK101. This proof of concept experiment was performed in shake flasks with AM1 medium that is a low salt mineral medium with an external carbon source .
When cells were grown in absence of furfural (0 g/L), the parent SSK42 strain achieved highest rate of biomass formation during 3-6 h cultivation interval (Fig. 2A) as compared to pgi deleted SSK101 strain that achieved highest value at 9-12 h interval of cultivation (Fig. 2C). Interestingly, when the growth rates of these two strains were monitored in presence of 1 to 3 g/L furfural, the trend was found to be reversed. With increasing concentration of furfural, SSK101 had relatively higher growth rate as compared to its parent strain SSK42 at corresponding cultivation interval (Fig. 2A and 2C). At 1.0 g/L furfural, the rate of change of biomass for SSK42 at 0-3 and 3-6 h intervals were 0.013±0.004 g/(L.h)and 0.044±0.005 g/(L.h), respectively [Additional file 1: Table S1], while the values for SSK101at corresponding sampling time were 0.025±0.004 g/(L.h) (1.92-fold higher) and 0.054±0.003 g/(L.h) (1.27-fold higher), respectively. Similar growth advantage for SSK101 was also observed at 1.5 g/L and 2 g/L furfural. These results suggested that presence of 1-2 g/L furfural in media promoted formation of biomass in SSK101. At 2.5 and 3.0 g/L furfural concentration no noticeable biomass formation rate was observed in either strain.
Along similar line, we investigated the influence of 5-HMF on the rate of change in biomass. In both SSK42 and SSK101, evidence of microbial growth was observed even at the highest tested concentration (3 g/L) of 5-HMF (Fig. 2B and 2D), suggesting that 5-HMF is relatively less inhibitory to microbial growth as compared to furfural. Similar to the pattern observed in case of furfural, presence of 5-HMF in media reduced the rate of biomass formation in SSK42, while it promoted the rate of biomass formation in SSK101 as compared to control where no 5-HMF was present in the growth media. However, though 5-HMF promoted biomass formation rate in strain SSK101, deletion of pgi gene didn't result in any significant growth advantage to SSK101 as compared to the parent strain SSK42 in presence of 5-HMF [Additional file 1: Table S2].
The above results indicate that furfural and 5-HMF lead to an improvement in the biomass formation rate in the pgi deleted E. coli strain SSK101. This observation further compelled us to investigate the growth behavior of strain SSK101 over a finer range of furfural or 5-HMF. We hypothesized that there will be an optimum inhibitor concentration at which the value of rate of change of biomass will be maximum, and any inhibitor concentration lower or higher than that optimum concentration will result in lower values of biomass formation rate.
We observed a pattern in growth rate whereas compared to 0 g/L furfural, presence of furfural in the media until 1.50 g/L promoted rate of increase in biomass of strain SSK101 (Fig. 3A). Maximum value (p-value=0.003) was exhibited at 1.25 g/L during 6-9 h interval. Growth could still be observed till about 2.25 g/L but at a lower rate as compared to the control. At furfural concentration higher than 2.25g/L the cellular metabolism is overwhelmed and no measurable change in biomass could be observed. Relative to furfural, SSK101 displayed higher growth rates at nearly all corresponding time points when 5-HMF was present in the medium (Fig. 3B). Presence of 5-HMF in media promoted rate of increase in biomass and reached maximum value (p-value=0.001) at around 2.5 g/L during 6-9 h interval. Though increase in rate of change in biomass could be observed till 3.0 g/L as compared to 0 g/L 5-HMF, the values were lower than those at 2.5 g/L. Interestingly, we observed that with increase in time, the value of rate of change in biomass shifted to higher side at the higher respective inhibitor concentration, with concomitant decrease in values at lower inhibitor concentration. This observation suggests that as the inhibitor is cleared from the medium the strain loses its ability to increase biomass.
Based upon the results described above we predict that presence of either furfural or 5-HMF in the media promotes growth till the point where intracellular production of excess NADPH is balanced by the concentration of inhibitor. Any further decrease or increase in the inhibitor concentration leads to either NADPH glut or scarcity, respectively, which affects microbial anabolic reactions. We thus conclude that SSK101 exhibits the potential to act as a catalyst to remove furfural and 5-HMF from the media.
Role of media components on furfural metabolism
From shake flask studies it became clear that both furfural and 5-HMF promote rate of biomass formation in strain SSK101 and not in parent strain SSK42. The data also suggests that furfural is relatively more inhibitory to microbial metabolism as compared to 5-HMF. To increase stringency in our further investigations, we chose furfural to investigate physiological responses of strains SSK101 and SSK42 in capped Hungate tube experiments. Closed tube experiments add second degree of stringency by limiting gaseous exchange with progress of microbial growth. In addition to tweaking with aerobicity, we also investigated the role of (i) different carbon sources, (ii) minimal and complex media, and (iii) osmotic stress, on microbial metabolism in presence of furfural.
Effect of different carbon source
In case of SSK101 and glucose as carbon source, no furfural could be detected in media at 48h, while SSK42 took additional 24 h to remove the inhibitor from medium (Fig. 4A). It is important to note that SSK42 can metabolize glucose via both EMP and PP pathways, which would result in relatively lower flux via PP pathway as compared to SSK101. Lower flux via PPP would eventually result in correspondingly lower levels of intracellular NADPH due to which the cell would take longer to detoxify furfural from the media. On the other hand, the ability of SSK101 to metabolize glucose only via PPP would result in relatively higher NADPH load as compared to SSK42 and lead to quicker removal of furfural from the media. When glucose was substituted by xylose as the sole carbon source, neither SSK42 nor SSK101 could completely remove furfural from the medium even by 96 h (Fig. 4A). This observation can be explained with mechanism of xylose metabolism. Xylose enters central carbon metabolism via non-oxidative phase of PPP and results in bypass of 2 moles of NADPH formed in the oxidative phase as compared to glucose on a molar basis (Fig. 1).We predict that the scarcity of NADPH in the microbial cell is the reason for relatively lower rate of furfural removal from the medium when xylose is the sole carbon source.
Effect of osmotic stress
The hydrolysate generated by dilute acid treatment of lignocellulose results in the formation of a soluble fraction which contains pentose(mainly xylose) with minor component of hexose sugar (glucose). We thus chose a mixed carbon source consisting of glucose (minor component with 1% concentration) and xylose (major component with 4% concentration) to study the influence of osmotic stress. Glucose acts as an osmotic stressor at concentration >100 mM (around 1.8%) while xylose exerts its stress at concentration >120 mM (around 1.8%) under similar microaerobic conditions . Thus, in our experimental setup, xylose at 4% is the main osmotic stressor and not the glucose which is at 1%. When the cells were grown in this medium in presence of 1 g/L furfural, strain SSK101 was able to clear furfural from AM1 medium within 24 h while SSK42 took additional 48 h for disappearance of furfural from the medium (Fig. 4B). However, at inhibitor concentration of 1.6 g/L, only marginal metabolism of furfural was observed for both SSK101 and SSK42 till 96 h; and medium containing SSK101 had relatively lower furfural as compared to the one containing SSK42.
- coli responds to osmotic stress by increasing intracellular glutathione levels  and inducing genes soxS, sodA and katE  that are also involved in combating oxidative stress. We must emphasize that osmotic stress is an additional element to the oxidative stress that is induced by furfural  itself. Under oxidative stress, glutathione is required to maintain reduced intracellular environment of the cell and in the process gets oxidized and needs to be reduced again. This regeneration of reduced glutathione is accomplished by oxidation of NADPH. We thus predict that osmotic stress introduces additional sink for the excess NADPH generated by strain SSK101.
Effect of media components
We next tested the influence of media components upon furfural tolerance. AM1 is a minimal medium with high cellular biosynthetic requirements. Any perturbation of cellular homeostasis in minimal medium will elicit more pronounced physiological response as compared to that in a rich medium with low biosynthetic requirements. Our observation was along similar lines. When SSK42 was grown in AM1 medium containing either glucose (Fig. 5A) or xylose (Fig. 5B), we found 8.7-fold and 6.9-fold reduction in growth, respectively, in presence of furfural. Compared to SSK42, less growth was observed in SSK101 in AM1 medium. However, we did not observe any decline in growth due to furfural addition in case of SSK101 grown in AM1+glucose (Fig. 5A), though growth did decline significantly when xylose was present in the AM1 medium (Fig. 5B). This clearly suggested role of excess NADPH in detoxifying furfural when SSK101 was grown in glucose medium. Interestingly, rich LB medium remarkably nullified the growth inhibition of SSK101 (Fig. 5C) that was observed for both AM1+glucose (Fig. 5A) as well as AM1+xylose (Fig. 5B) media in absence of furfural. This behavior could be attributed to both absence of glucose that would result in lower NADPH stress and lowered biosynthetic requirements of the cell in LB medium. Rich LB medium also, expectedly, buffered the toxicity of furfural as compared to that in AM1 medium, and growth of either strain resulted in relatively higher biomass (Fig. 5C). These results point to the fact that competitive advantage conferred on SSK101 strain by deletion of pgi gene is neutralized by rich complex media.
Physiological evaluation at bioreactor level
Our results indicate that both strains can increase biomass in presence of either furfural or 5-HMF (Fig. 2), and increased osmotic stress also did not hinder them to metabolize 1 g/L furfural under the tested condition in Hungate tube (Fig. 4A). We next evaluated the physiological response of SSK101 and SSK42 under controlled conditions using a bioreactor. We increased complexity of investigation by including both furfural and 5-HMF at 1 g/L each leading to total inhibitor load of 2 g/L in bioreactor. This particular investigation assumes more importance since an earlier study has highlighted the synergistic inhibitory effect exerted by furfural in presence of 5-HMF on microbial metabolism . We chose the concentration of respective furan in tune with the reported concentration of same in acid treated rice straw hydrolysate . We also varied the concentration of sugars in the culture media to observe whether the concentration of glucose plays any role in microbial productivity. To prevent confounding effect due to variation in the osmotic load we compensated the change in glucose concentration by varying concentration of xylose. In all bioreactors, total sugar load was ~5.5%. It has been reported that in case of acidic treatment of hydrolysate and depending upon the severity of pretreatment the concentration of xylose to glucose exceeds in the range of 2-26 fold [31, 32]. In our studies the fold increase of xylose over glucose concentration was 4-8.3 fold.
We observed very distinct behavior of SSK42 and SSK101 in the bioreactor when grown either in absence or in presence of the inhibitor mix. In the absence of inhibitors, both SSK42 and SSK101 were able to utilize the sugar mixture in about 96h for both 6 g/L (Fig. 6A, 6C) as well as 12 g/L glucose (Fig. 6B and 6D). Expectedly, the rate of glucose utilization was faster in SSK42. The maximum ethanol titers for both strains at either concentration of glucose were also comparable (Table 1). At lower glucose concentration, SSK101 exhibited highest ethanol productivity at 48-72 h time interval, which shifted to later interval of 72-96 h at higher glucose concentration (Table 2). Similar pattern of improved ethanol productivity at lower glucose concentration was also observed for SSK42, which displayed highest ethanol productivity during 24-48 h at lower glucose concentration and during 72-96 h at higher glucose concentration (Table 2). In case of SSK42 we observed the classic diauxic growth pattern between 24-48 h (Fig. 6H) at higher glucose load which was also reflected in the pattern of ethanol production (Fig. 6F) at corresponding time points. One the other hand, SSK101 didn't exhibit any diauxie at either glucose load. Interestingly, we didn't observe any diauxic growth for SSK42 at lower glucose load (Fig. 6G). This likely explains the improvement in ethanol productivity for SSK42 (Table 2) at lower glucose concentration. Strain SSK101 co-utilized both glucose and xylose at tested concentrations of glucose.
In the presence of the binary stress of furfural and 5-HMF, SSK101 completely utilized sugars under tested glucose regimes (Fig. 6A-6D). In contrast, SSK42 that did not exhibit significant metabolic activity in either glucose concentration. As compared to cultivation without inhibitors, presence of inhibitors reduced productivity (Table2) where highest ethanol titers were recorded after extended cultivation hours (Fig. 6E and 6F). Remarkably, even under inhibitor stress, maximum ethanol titers of SSK101 were comparable with that of SSK42 (and SSK101) observed in absence of inhibitors (Table 1). Maximum values of ethanol productivity of SSK101 under low glucose concentration was observed between 96-120 h and shifted to a later interval of 144-168 h at high glucose load (Table 2). At low and high glucose load the maximum ethanol yield of SSK101 was 69% and 78%, respectively. Although, these values are lower than those observed in lime treated hydrolysate where E. coli LY01 produced ethanol with > 90% of theoretical yield , the furfural and HMF concentration in this study were only 0.25 g/L and 0.34 g/L, respectively. Values of maximum ethanol yield at lower and higher glucose load was 0.35 and 0.40 g/g which were comparable with 0.39 value obtained by E. coli XW129 in nutrient and sodium metabisulfite supplemented hydrolysate . SSK101 did not exhibit any diauxic growth pattern even in presence of inhibitors at either concentration of glucose tested. Though, SSK42 did not do well at total inhibitor load of 2 g/L, it did utilize both sugars and produced comparable ethanol titers at lower inhibitor load of 1 g/L with 0.5 g/L each of furfural and 5-HMF; and had an extended diauxie period [Additional file 1: Figure S1]. Our efforts to further increase the stringency of cultivation conditions in bioreactors by increasing total sugar as well as furan load were not successful [Additional File 1: Figure S2].
Glucose overflow metabolism in E. coli is characterized by channelizing the glucose carbon to synthesis of NADH-independent acetate and is in response to accumulation of reduced nicotinamides in the cell . We observed partial evidence of overflow metabolism in both SSK42 and SSK101 (Table 1). Both maximum titer of acetate and its difference between the two strains was reduced at lower concentration of glucose in absence of inhibitors. At lower glucose concentration, presence of inhibitors almost doubled the titers of acetate in SSK101 as compared to the control without any inhibitors. We believe that inhibitors place metabolic burden on the microbial cell with additional energy requirement and production of acetate is one way to fulfill increased ATP demand. Based on the results of bioreactor studies, we predict that the lower concentration of glucose reduces the NADPH stress on SSK101 while still being enough to channelize the energy and reducing equivalents required to balance anabolic reactions with detoxification of inhibitors.