In the present study the physiological effects of HMF and furfural, two inhibitors present in lignocellulosic hydrolysates, were investigated. By having the inhibitors present in the feed-medium of an anaerobic chemostat, physiological responses related to changes in specific growth rate could be circumvented. It has previously been shown that gene expression changes due to alterations in growth rate and the expression pattern observed after exposure to different types of stress overlap to a certain extent
, which consolidates the choice of operation mode in the present study since HMF and furfural inhibit cell growth at sufficiently high concentrations
. In the current experimental setup, a long-term stress response was expected where the cells have adapted to the stressful conditions as compared to the transient stress responses obtained in pulse experiments.
Step-changing the feed-medium by including HMF and furfural resulted in decreased specific uptake-rates for both glucose and xylose compared to the control. Banerjee et al.
 have reported that furfural inhibits hexokinase, which could have triggered the observed decrease of sugar uptake rates. Although the specific xylose uptake rate decreased, there was an increase in the xylitol production rate and xylitol yield, which suggests that the redox metabolism was perturbed. Xylose consumption in the strain used in the present study occurs via the two oxidoreductases xylose reductase (XR) and xylitol dehydrogenase (XDH) from S. stipitis. XR preferably uses NADPH for reduction of xylose to xylitol, while XDH uses NAD+ for oxidation of xylitol to xylulose, thereby creating a redox imbalance with xylitol accumulation as a consequence. Previous studies have reported that addition of external electron acceptors such as acetoin, furfural and HMF decrease xylitol accumulation due to regeneration of NAD+. To the authors’ best knowledge, one study with a fermentation setup similar to the one described here has been published, i.e. anaerobic chemostat with glucose and xylose in the feed-medium, although only HMF was considered
. Two of the strains assessed in that study showed no effect on xylitol yield when the feed was step-changed to an HMF containing one, whereas one strain showed a decreased xylitol yield. The results obtained in the present study can be an effect of the use of both furfural and HMF in the medium. Since it has been demonstrated that the co-factor usage by the detoxifying enzymes is different for HMF and furfural
, challenging the cells with both these inhibitors simultaneously may perturb the redox metabolism in a different way than reported before.
Even though HMF and furfural detoxification is generally thought to proceed through NAD(P)H-dependent reduction
, there are surprisingly few reports available on the quantitative effect on the redox co-factor pools and the resulting redox perturbations HMF and furfural leads to. In the present study, we showed that the presence of HMF and furfural in the feed-medium significantly decreased the intracellular steady-state concentrations of NADH and NADPH. Moreover, the catabolic and anabolic reduction charges were decreased significantly. NAD(P)+ and NAD(P)H are used as co-factors in numerous intracellular reactions, and consequently, perturbation of the levels of these can have cell-wide effects which can be hard to predict. Most of the intracellular NADH is produced in glycolysis and in biosynthetic reactions such as amino acid synthesis. NADH produced in assimilatory reactions is reoxidized to NAD+ by conversion of dihydroxyacetone phosphate to glycerol-3-phosphate by Gpd1p. Addition of external electron acceptors such as HMF and furfural have shown to decrease the glycerol yield by acting as alternative redox sinks. However, in the present study no effect on the glycerol yield was observed after changing to an HMF and furfural containing feed-medium. Assuming an intracellular volume of 2.38 mL (g DW)-1, the intracellular concentration of NADH decreased from 0.20 mM to 0.08 mM upon inhibitor addition. The Km for NADH of Gpd1p is 0.023 mM
, thus 4 times lower than the actual concentration in the cell, which would mean that the reaction would still proceed at approximately 80% of Vmax when inhibitors were present.
Under anaerobic conditions, NADPH is mainly produced in the oxidative part of PPP from glucose-6-phosphate through two reactions catalyzed by Zwf1p and Gnd1p
. NADPH is generally used as a reductant in amino acid synthesis and especially in ammonium assimilation through production of glutamate from 2-oxoglutarate by Gdh1p
. Glutamate, which is involved in biosynthesis of several amino acids, is a highly connected metabolite in the metabolic network. Km of Gdh1p for NADPH is reported to be 0.033 mM
. Using the same assumption on intracellular volume as above, the intracellular concentration of NADPH decreased from 0.23 mM to 0.06 mM after adding HMF and furfural to the feed-medium. With this significant reduction in NADPH concentration, the Gdh1p-catalyzed reaction would proceed at a rate of about 65% of Vmax, a significant decrease in the glutamate production rate, which could have consequences for amino acid production. A lower flux through the ammonia assimilation pathway might signal to the cell that insufficient amounts of nitrogen are available in the form of glutamate, which thereby could lead to a down-regulation of amino acid synthesis. A lower flux through the ammonia assimilation pathway may also be reflected by some features detected by the transcriptome analysis, such as the induction of pseudohyphal growth genes common to the nitrogen starvation response. As many as 10 ORFs encoding amino acid, peptide or amine transporters were identified after functional enrichment analysis of the down-regulated genes in response to HMF and furfural. Several ORFs encoding enzymes involved in amino acid biosynthesis were also found to be down-regulated. Intuitively, one would think that nitrogen starvation and down-regulation of amino acid biosynthesis would lead to up-regulation of amino acid transporters, but surprisingly the opposite was observed. This result can not be easily explained, but it has been observed before in relation to pseudohyphal growth
Since the first step in the xylose-utilization pathway involves an NADPH-dependent reduction, it could be argued that lower intracellular NADPH levels should decrease the flux through the pathway in a similar way as suggested for glutamate synthesis. However, addition of HMF and furfural did not result in a significant decrease in the specific uptake rate of xylose. This could be explained by a comparison of the kinetic properties, and in particular the Km for NADPH of XR and Gdh1p. In fact, the Km of XR for NADPH is 0.009 mM
, thus 3.7 times lower than that of Gdh1p, which thereby makes XR activity less sensitive to the lower NADPH levels observed after HMF and furfural addition.
NAD(P)H dependent reduction of HMF and furfural to their corresponding alcohols requires the supply of sufficient amounts of the involved co-factors. In a recent study
, it was shown that increased NADPH demand caused by addition of 200–300 mM of the electron acceptor acetoin induced the expression of genes involved in NADPH generation, such as GND1 in the oxidative part of PPP. In contrast, the higher co-factor demand triggered by lower concentrations (100 mM) of acetoin could be met through metabolic (post-transcriptional) regulation. Similarly to this latter case, addition of HMF and furfural did not induce expression of genes involved in NADPH synthesis in the present study. Thus, the higher demand for redox co-factors in the presence of furan aldehydes in the present study is possibly regulated at metabolic level, which in fact was the case when the lower amount (100 mM) acetoin was added to the cultivations in
Interestingly, the energy charge did not change notably and remained at a high level after addition of furaldehydes to the feed-medium. On the other hand, the biomass yield (YSX) and the ATP yield (YATP)
increased when HMF and furfural were present. The energy requirement for biomass formation, or YATP, is calculated as the amount of biomass produced in relation to the amount of ATP produced, and can therefore be seen as a coupling factor between anabolism and catabolism
. Protein synthesis, ammonium transport and amino acid synthesis are three of the most ATP consuming processes in metabolism
. A decrease in protein polymerization and/or alternatively, a decreased amino acid production could in fact provide more ATP available for carbon assimilation. Thus, changes in biomass composition can have a considerable impact on the biomass yield and ATP yield.
An alternative explanation for the increase in biomass yield that has been proposed
 is that reduction in glycerol production can lead to an increased amount of ATP available for biomass production, since glycerol production is an ATP consuming process. NADH generated in the biosynthesis of amino acids can be reoxidized to NAD+ through NADH-dependent reduction of HMF and furfural instead of glycerol production, which is the native redox valve. In the present study, the specific glycerol production rate decreased when HMF and furfural were present in the feed-medium, but the difference was not statistically significant.
Functional enrichment of the ORFs that were up-regulated after a step-change to HMF and furfural containing feed-medium showed that only one functional class (Cell rescue, defense and virulence) was significantly over-represented. Two of the ORFs in this class, PDR5 and YOR1, are coding for ATP-binding cassette (ABC) transporters that function in efflux of several compounds, including ions and xenobiotics. They are under transcriptional control of Pdr1p and Pdr3p, and can probably function in transporting either HMF and furfural, or their corresponding alcohols, out of the cell, thereby relieving the stress caused by these agents. The lower intracellular ATP concentration that was observed after HMF and furfural had been added could be an indication that ATP was used by the efflux pumps in order to transport the inhibitors out of the cell. In fact, Ma et al.
 constructed Δpdr1 and Δpdr3 mutants and showed that these strains exhibited a longer lag-phase than the wild-type in the presence of HMF, indicating that ABC transporters are important for tolerance to furan aldehydes. Moreover, Alrikson et al.
 overexpressed FLR1 and ATR1, which are multidrug transporters of the major facilitator family, and obtained strains with higher tolerance to HMF and coniferylaldehyde, respectively. Together with our results, these examples show that multidrug transporters are putative metabolic engineering targets for increased tolerance to HMF and furfural.
Finding the molecular mechanisms of the inhibitory action by detailed physiological analyses is a promising strategy to discover metabolic engineering targets that can improve strain robustness. HMF and furfural remain a serious challenge for lignocellulosic ethanol production. Not least since attainment of higher final ethanol titers requires that higher solid loadings are used in the production process, which comes at the cost of higher inhibitor concentrations
. From the results of the physiological analyses in the present study it is clear that the redox metabolism is severely affected by furan aldehydes, while the effects on the energetics is less evident. Thus, engineering of the redox system represents a putative target to relieve the stress caused by HMF and furfural, but further studies are required to target the specific consequences of the perturbed redox metabolism.