**3. Results**

*P. pseudoalcaligenes* CECT5344 R1 was able to utilize furfural (up to 40 mM), furoic acid (up to 20 mM), and furfuryl alcohol (up to 20 mM) as the sole carbon and energy source, although after a very long lag phase (Figure 2A). The longest lag phase was observed with furfural as a C-source (Figure 2(A2)). In fact, furfural concentrations higher than 5 mM increased the lag phase, thus suggesting that this compound is toxic at a high concentration. Nevertheless the tolerance of *P. pseudoalcaligenes* CECT 5344 to furfural is relatively high (up to 40 mM) if compared to *C. basilensis* HMF14 (up to 12 mM) [16]. This toxic effect was not observed for furoic acid (FA), and furfuryl alcohol (FFA) was not toxic up to a concentration of 20 mM (Figure 2A). The maximum cell growth increased with the concentration of the furanic compound up to a concentration of 20 mM, indicating that below this concentration the growth was carbon limited. It is remarkable that at the same concentration of furanic compound, the maximum growth was highest with FFA, followed by F and finally FA. This is in agreemen<sup>t</sup> with the chemical compositions of these compounds, the alcohol being the most reduced, followed by the aldehyde and then the carboxylic acid. Both F and FFA were, in a first instance, almost stoichiometrically transformed into furoic acid (Figure 2B), thus indicating that the pathway for the assimilation of this compound is not active in the wild type strain of *Pseudomonas pseudoalcaligenes* CECT 5344 R1. Only after several days of incubation, a clear increase in cell growth was observed, which was concomitant with the assimilation of FA (Figure 2B).

*P. pseudoalcaligenes* CECT 5344 R1 failed to grow on 5-(hydroxymethyl)furfural (HMF), even after a prolonged incubation period of more than 10 days (Figure 3A). The analysis of the culture media by HPLC revealed that HMF was completely exhausted after 25 h, indicating that although this bacterial strain was unable to assimilate HMF, it had the capability of transforming it. Two new compounds were detected in the couture media after consumption of HMF, 2,5-furandicarboxylic acid (FDCA) in a proportion below 5% of the HMF consumed and an unknown aromatic compound (Figure 3A). In *C. basilensis* HMF14, HmfH catalyzes the oxidation of 5-hydroxymethyl-2-furoic acid (HMFA) to FDCA, whereas HmfFG catalyzes the decarboxylation of FDCA to FA [24]. Therefore, in this strain, the metabolic pathways for the assimilation of HMF, FFA and F converge in FA. In the genome of *P. pseudoalcaligenes* CECT 5344 R1, no homologous genes to *hmfH* and *hmfFG* of *C. basilensis* were observed. This genotype agrees with the fact that this bacterium does not assimilate HMF. Since HMFA is the substrate of HMFH, the unknown compound accumulated in the culture media from HMF (Figure 3A) could be HMFA. To test if the incapacity of *P. pseudoalcaligenes* was due to a problem of induction, the bacterium was inoculated in media containing both F and HMF (Figure 3B). The result was that the bacterium exclusively used F and that the presence of HMF resulted in being toxic (Figure 3B). Therefore, the absence of catalytic enzymes for the assimilation of FA was not the reason for the inability of *P. psedoalcaligens* to assimilate HMF, but the absence of reactions connecting HMFA and FDCA to FA. The transformation of FFA into FA takes place in two consecutive oxidative steps. In *Pseudomonas putida* Fu1, two different and inducible enzymes catalyze these reactions [23], but in *C. basilensis*, although no concrete genes have been assigned, it could be possible that the same enzyme catalyzes both oxidations [24]. In *P. pseudoalcaligenes* CECT5344 R1, both dehydrogenase activities (FFADH and FDH in the scheme of Figure 1D) co-eluted after anion exchange chromatography (not shown), thus suggesting that it is the same enzyme. Again, both activities had the same optimum temperature at 65 ◦C. Furfural dehydrogenase activity was clearly induced by furfuryl alcohol, if compared with acetate, FA or LB medium (not shown). Although these results sugges<sup>t</sup> that the same enzyme could catalyze the oxidation of FFA to F and of F to FA, the only clear conclusion is that transformation of FFA to FA and conversion of FA to 2-oxoglutaric acid takes place through different pathways.

**Figure 2.** (**A**) Growth curves of *Pseudomonas pseudoalcaligenes* CECT 5344 R1 with different concentrations of furfuryl alcohol (panel A1), furfural (panel A2) or furoic acid (panel A3). (**B**) Growth curves and concentrations of furanic intermediates in the culture media of *P. pseudoalcaligenes* using as a carbon source either 5 mM furfuryl alcohol (panel B2), furfural (panel B2) or furoic acid (panel B3). Three independent experiments gave similar results.

**Figure 3.** Biotransformation of HMF by *P. pseudoalcaligenes* CECT 5344 R1. (**A**) Cell growth (black line), concentration of HMF and FDCA (mM), and an unknown metabolite in cell cultures of *P. pseudoalcaligenes*. (**B**) Effect of HMF (5 mM) on the cell growth of *P. pseudoalcaligenes* at the expense of 10 mM furfural (F+HMF) in comparison with the cell growth with 10 mM furfural (F, black line). Similar results were obtained in three different experiments.

Once grown on furfural as a C-source, the serial dilutions of a culture of *P. pseudoalcaligenes* CECT 5344 R1 spread on solid minimal medium with furfural or FA giving colonies with two morphologies, big or small colonies. The small colonies had the same phenotype as the original stain (R1), but the

big colonies grew faster on furfural even after successive generations on non-selective medium (LB). In fact, after four serial re-inoculations of the bacteria in in fresh media with furfural (10 mM) as the sole carbon source, allowed the selection of a mutant (big colonies in FA plates) in which the growth lag phase was drastically reduced and had a reproducible growth rate of 0.29 h−<sup>1</sup> on furfural (5 mM). Figure 4A illustrates the isolation of the mutant, thereafter called R1D, and its phenotype in comparison to the *wt*. The selection of mutants with improved capacities has been widely used in biotechnological processes for the selection of evolved phenotypes, although having poor knowledge of their underlying genotype [31], also in the context of furfural assimilation [32]. For example, *P. putida* S12 expressing the *hmfABCDE* genes from *C. basilensis* HMF14 strain, was only able to e fficiently assimilate furfural when adapted by repeated inoculation in media with furfural. The same has been recently described for *Pseudomonas putida* KT2440 expressing a 12 kb DNA fragment containing the *hmf* gene cluster from *Burkholderia phytofirmans* [33]. The next generation sequence (ngs) techniques open the possibility to analyze the genotypic variation associated with the evolved phenotypes. Adaptive laboratory evolution (ALE) takes profit of this advantage and may have a tremendous application in the rational design of genetically manipulated microorganisms, as well as in understanding some basic evolving mechanisms of living beings [31,34,35]. In our laboratory, the transcriptomic analysis of the R1D mutant in response to furfuryl alcohol was analyzed and it revealed to be more complex than expected (not shown). Interestingly, the transcriptomic reads sequences (RNA-seq) were obtained from the mutant strain (R1D), whereas the genome sequences available are from the wt strain [36]. The comparison of both sequences in the *hmf* locus revealed that the R1D mutant had a point mutation in a possible regulatory gene of the *araC* family. The presence of the mutation was confirmed by re-sequencing the *araC* gene (BN5\_2307) (Figure 4B). The observed point mutation was a transversion (782T>G), leading to the non-conservative change of the triplet CTT (Leu) to CGT (Arg) (Figure 4B). The alignment of several AraC proteins revealed that the Arg in position 261 is a conserved residue in most members of this family of regulators (not shown). This position is located in the HTH domain of the protein and it is also conserved in *E. coli* [37], thus suggesting its essentiality. In conclusion, it seems that the pathway for the assimilation of FA is not active in the wt strain of *P. pseudoalcaligens* because AraC is not functional, and that the L262R mutation generates the active and functional regulator (AraC\*) in the evolved R1D strain. In order to check this hypothesis, a mutant of the *araC*\* gene of *P. pseudoalcaligenes* CECCT 5344 R1D was generated by double recombination. As expected, the mutant strain was unable to use furfural as a C-source (Figure 4C).

**Figure 4.** Role of *araC* in the assimilation of furfural by *P. pseudoalcaligenes*. ( **A**) Scheme of the selection process (upper panel), and growth curve (lower panel) of the R1D mutant in comparison with the wt strain (R1), in media with furfural (10 mM) as the sole carbon source. (**B**) Scheme of the point mutation detected in the *araC* gene of the R1D mutant. ( **C**) Growth curve of the R1D mutant and its derived mutant generated by the inactivation of the *araC*\* gene by insertion of the gentamicin resistance gene (*aacC1*) by double recombination. Similar results were obtained in three different experiments (panels A and C).

In addition to furfural, the mutant R1D was also capable to assimilate furfuryl alcohol and furoic acid (Figure 5). However, this strain remained unable to use HMF as a carbon source.

**Figure 5.** Growth curves of *P. pseudoalcaligenes* CECT 5344 R1D with furfuryl alcohol (**A**), furfural (**B**) or furoic acid (**C**) as the sole carbon sources (10 mM). The concentration of furanic intermediates in the culture media was measured by HPLC at the indicated times. Each experiment was done in triplicate giving similar results.

It is evident that R1D grew faster than R1 and with similar rates to other furfural-degrading strains reported in the literature. Table 2 summarizes the growth parameters of R1 and R1D strains of *P. pseudoalcaligenes* CECT 5344 in comparison to reference strains. In addition to the shorter lag phases and higher growth rates, a notable difference between the wild strain (R1) and the R1D mutant was the accumulation of FA during the lag phase. As shown in Figure 2, the wt strain stoichiometrically transformed FFA and F to FA. By contrast, in the RD1 mutant, FA was the only transiently accumulated form F (Figure 5B), whereas it was hardly detectable in media with FFA (Figure 5A). No lag phase was observed in the R1D strain growing with FA. Therefore, the presence of FA immediately induces its assimilation in R1D, thus suggesting that FA could be the inducer of the process. It is evident that the difference between R1 and R1D relies on the process of assimilation of furoic acid, not in the oxidation of alcohol and aldehyde to furoic acid. In analogy with the assimilation of toluene and xylenes in *P. putida* mt-2 [38], we can divide the pathway in two segments, the upper pathway and the lower pathway. We can consider the upper pathway as the oxidation of the alcohol group (-CH2OH) and aldehyde (-CHO) to acid (-COOH). In the case of *C. basilensis*, the decarboxylase has to be included in the upper pathway, making the metabolism of HMF converge to FA. The lower pathway consists in the mineralization of furoic acid.


**Table 2.** Growth parameters of *P. pseudoalcaligenes* CECT 5344 R1 (wt) and the R1D mutant in comparison to bibliographic data.

1 Furfuryl alcohol, 2 Furfural or 3 furoic acid (5 mM). 4,5 Extracted form [16] and [24], respectively.

It is worth noting that these experiments confirmed the fact that the maximal cell growth of *P. pseudoalcaligenes* with FFA was higher than with F followed by FA (Table 2).

It is evident that R1D is an evolved strain much more efficient than the wt in the assimilation of FFA, F and FA. Nevertheless, the utilization of these capabilities for the elimination of the inhibitory compounds from lignocellulosic hydrolysates requires that the bacterium leave the sugars intact for their further fermentation process. *P. pseudoalcaligenes* CECT 5334 was unable to use as a C-source neither xylose, sucrose, arabinose, mannose, nor galactose. The capability of *P. pseudoalcaligenes* to

use sugars was restricted to glucose. The chemical composition of the lignocellulosic hydrolysates depends on the raw material utilized and the treatment employed [39]. Therefore, the capacity of *P. pseudoalcaligenes* CECT 5344 R1D to use glucose and furfural simultaneously was studied. As expected, R1D assimilated both compounds simultaneously in blended media (Figure 6). From Figure 6, it became evident that F is toxic, since maintaining the concentration of glucose, the lag phase increased when furfural concentration increased (Figure 6A,B). In any case, since both glucose and F were assimilated simultaneously by *P. Pseudoalcaligenes*, the construction of a mutant impaired in glucose assimilation was designed. Most *Pseudomonas*, which have a relatively limited ability to assimilate sugars, usually assimilate glucose through the Entner-Doudoroff pathway [40], and the inactivation of the *edd* gene usually causes the inability to assimilate glucose [41]. As shown in Figure 6C, the *edd*− strain was as efficient as R1D assimilating furfural but glucose remains unaltered in the culture media. The *edd*− mutant was also able to use furfuryl alcohol and furoic acid as a C-source.

**Figure 6.** Cell growth of *P. pseudoalcaligenes* CECT 5344 R1D with glucose 5 mM, supplemented with furfural 5 mM (**A**) or 2 mM (**B**) as carbon sources. (**C**) Cell growth of *P. pseudoalcaligenes* CECT 5344 R1D *edd*− with furfural (2.5 mM) and glucose (5 mM) as carbon sources. Glucose and furfural concentration were determined at the indicated times. Similar results were obtained in three different experiments.
