The inactivation of enzymes belonging to the central carbon metabolism, a novel mechanism of developing antibiotic resistance

Fosfomycin is a bactericidal antibiotic, analogous to phosphoenolpyruvate (PEP) that exerts its activity by inhibiting the activity of MurA. This enzyme catalyzes the first step of peptidoglycan biosynthesis, the transfer of enolpyruvate from PEP to uridine-diphosphate-N-acetylglucosamine. Fosfomycin is increasingly used in the last years, mainly for treating infections caused by Gram-negative multidrug resistant bacteria as Stenotrophomonas maltophilia, an opportunistic pathogen characterized by its low susceptibility to antibiotics of common use. The mechanisms of mutational resistance to fosfomycin in S. maltophilia were studied in the current work. None of the mechanisms so far described for other organisms, which include the production of fosfomycin inactivating enzymes, target modification, induction of alternative peptidoglycan biosynthesis pathway and the impaired entrance of the antibiotic, are involved in the acquisition of such resistance by this bacterial species. Rather the unique cause of resistance in the studied mutants is the mutational inactivation of different enzymes belonging to the Embden-Meyerhof-Parnas central metabolism pathway. The amount of intracellular fosfomycin accumulation did not change in any of these mutants showing that neither the inactivation nor the transport of the antibiotic were involved. Transcriptomic analysis also showed that the mutants did not present changes in the expression level of putative alternative peptidoglycan biosynthesis pathway genes neither any related enzyme. Finally, the mutants did not present an increased PEP concentration that might compete with fosfomycin for its binding to MurA. Based on these results, we describe a completely novel mechanism of antibiotic resistance based on the remodeling of S. maltophilia metabolism. Significance Antibiotic resistance (AR) has been largely considered as a specific bacterial response to an antibiotic challenge. Indeed, its study has been mainly concentrated in mechanisms that affect the antibiotics (mutations in transporters, the activity of efflux pumps and antibiotic modifying enzymes) or their targets (i.e.: target mutations, protection or bypass). Usually, AR-associated metabolic changes were considered to be a consequence (fitness costs) and not a cause of AR. Herein, we show that strong alterations in the bacterial metabolism can also be the cause of AR. In the study here presented, Stenotrophomonas maltophilia acquires fosfomycin resistance through the inactivation of glycolytic enzymes belonging to the Embden-Meyerhof-Parnas. Besides resistance to fosfomycin, this inactivation also impairs the bacterial gluconeogenic pathway. Together with previous work showing that AR can be under metabolic control, our results provide evidence that AR is intertwined with the bacterial metabolism.

susceptibility to antibiotics of common use. The mechanisms of mutational resistance to 8 fosfomycin in S. maltophilia were studied in the current work. None of the mechanisms 9 so far described for other organisms, which include the production of fosfomycin 10 inactivating enzymes, target modification, induction of alternative peptidoglycan 11 biosynthesis pathway and the impaired entrance of the antibiotic, are involved in the 12 acquisition of such resistance by this bacterial species. Rather the unique cause of 13 resistance in the studied mutants is the mutational inactivation of different enzymes 14 belonging to the Embden-Meyerhof-Parnas central metabolism pathway. The amount of 15 intracellular fosfomycin accumulation did not change in any of these mutants showing 16 that neither the inactivation nor the transport of the 17 antibiotic were involved. Transcriptomic analysis also showed that the mutants did not 18 present changes in the expression level of putative alternative peptidoglycan 19 biosynthesis pathway genes neither any related enzyme. Finally, the mutants did not 20 present an increased PEP concentration that might compete with fosfomycin for its 21 binding to MurA. Based on these results, we describe a completely novel mechanism of 22 antibiotic resistance based on the remodeling of S. maltophilia metabolism. 23 Significance 24

Introduction 1
Antibiotic resistance can be considered as a chemical problem. To be active, an 2 antibiotic requires to reach its target at concentrations high enough for inhibiting its 3 activity. Any process or situation that either reduces the effective concentration of the 4 antibiotic or the antibiotic-target affinity should lead to antibiotic resistance. In 5 agreement with this situation, classical, so far described, mechanisms of resistance (1)  6 include elements that diminish the antibiotic concentration like efflux pumps (2), 7 antibiotic inactivating enzymes (3) or changes in the antibiotic transporters (4). 8 Concerning the target, elements that reduce its affinity with the antibiotic include 9 mutations (5), target protection (6), bypass (7) or replacement (8) and eventually 10 increased target expression (9). Studies on intrinsic resistome have shown that, in 11 addition to these classical resistance determinants, the susceptibility to antibiotics of a 12 bacterial species depends on the activity of several elements encompassing all 13 functional categories (10-12). However, little is still known about the interplay between 14 bacterial metabolism and the acquisition of antibiotic resistance (13). In the current 15 article, we explore this feature analyzing S. maltophilia fosfomycin resistant mutants. 16 Fosfomycin is a phosphonic acid derivative that contains an epoxide and a propyl 17 group, chemically analogous to phosphoenolpyruvate (PEP), which explains its 18 mechanism of action (14). The enzyme MurA (UDP-N-acetylglucosamine enolpyruvyl 19 transferase), which catalyzes the first step in peptidoglycan biosynthesis (15), the 20 transfer of enolpyruvate from PEP to uridine diphospho-N-acetylglucosamine, is the 21 only known fosfomycin target. Fosfomycin binds covalently to a cysteine residue in the 22 active site of MurA, which renders MurA inactivation. As a consequence of MurA 23 inactivation, the peptidoglycan precursor monomers accumulate inside the cell, 24 peptidoglycan cannot be synthesized and this leads to bacterial cell lysis and death (16). 25 Different molecular mechanisms leading to fosfomycin resistance have been 1 identified (17). Some of them impairing the fosfomycin/MurA interaction. Some allelic 2 variants of MurA found in pathogens intrinsically resistant to fosfomycin such as 3 Mycobacterium tuberculosis, Borrelia burgdorferi or Chlamydia sp. (15,(18)(19)(20) do not 4 contain a cysteine in their active site, and therefore they are not fully inhibited by 5 fosfomycin. In the case of organisms containing a fosfomycin-sensitive MurA allele, 6 mutations in murA can be selected (15,21,22) and increased synthesis of MurA also 7 confers a resistance phenotype (23,24). Also, the presence of an alternative route of 8 peptidoglycan synthesis, as it happens in Pseudomonas putida and P. aeruginosa, may 9 allow circumvent the activity of fosfomycin by recycling the peptidoglycan without the 10 need of de novo synthesis by the enzyme MurA (7). Concerning mechanism involving a 11 reduction in the intracellular concentration of the antibiotic, resistance can be achieved 12 as the consequence of changes in the entrance of fosfomycin inside bacterial cell. The 13 main cause of this impaired uptake is the selection of mutations in any of the genes 14 encoding the sugar phosphate transporters GlpT and UhpT, which are the gates for 15 fosfomycin entrance (25,26). To note here that expression of these transporters is under 16 metabolic control, in such a way that situations where the nutritional bacterial status 17 favors the use of sugar phosphates (as intracellular growth) increase fosfomycin activity 18 (27,28). Finally, in other cases, fosfomycin is inactivated by fosfomycin modifying 19 enzymes as FosA, . All the already known mechanisms of 20 fosfomycin resistance fit in the classical categories of resistance elements (see above). 21 However, the results presented in the current article support that none of them are 22 involved in the acquisition of resistance by S. maltophilia. In this bacterial species, 23 fosfomycin resistance was acquired due to mutations in genes encoding enzymes of the 24 Embden-Meyerhof-Parnas (EMP) metabolic pathway. It has been suggested that 25 antibiotic resistance can be inter-linked to bacterial metabolism (33,34). However, with 1 very few exceptions (35), the mutational inactivation of genes encoding enzymes of the 2 central carbon metabolism has not been considered to be a significant cause of antibiotic 3 resistance in bacterial pathogens (34,35). Our article hence shed light in the crosstalk 4 between antibiotic resistance and central carbon metabolism in S. maltophilia. 5

Results 6
Selection of S. maltophilia fosfomycin resistant mutants and identification of the 7 mutations involved. 8 In order to isolate single-step S. maltophilia fosfomycin resistant mutants, 9 around 10 8 bacterial cells were seeded on selection plates containing fosfomycin (1024 10 µg/ml). Four single-step fosfomycin resistance mutants, hereafter dubbed FOS1, FOS4, 11 FOS7 and FOS8, were selected for further studies. The MIC of the mutants to 12 fosfomycin was determined. In all cases, the MICs of fosfomycin were higher in the 13 mutants than in the wild-type strain (Table 1). 14 The genomes of these mutants were fully sequenced and compared with that of 15 the parental wild-type strain D457. Five different mutations were detected. FOS4, FOS7 16 and FOS8 carried one mutation, while FOS1 harbored two mutations. One of them (in 17 rne, SMD_RS14705:c.G1464T:p.E488D) was discarded because it was predicted to be 18 neutral using the Provean predictor (0.41 score). Notably, each mutant contains a 19 different mutation, but all four were found in genes encoding enzymes of EMP 20 metabolic pathway, namely eno, gpmA, gapA and pgk (Table 2). To further confirm the 21 presence of each of these mutations in the mutant strains, the corresponding genomic Although no other mutations seemed to be the cause of the resistance of the 1 studied mutants, the wild-type allele of the corresponding mutated gene was introduced 2 in each mutant strain to get a functional validation of the effect of these mutations in the 3 susceptibility to fosfomycin of S. maltophilia. As shown in Table 1, introduction of the 4 wild-type forms of such genes fully restores the susceptibility of the analyzed S. 5 maltophilia fosfomycin resistant mutants to the level of the wild-type strain. These 6 results indicate that the fosfomycin resistance of these mutants is solely due to the 7 mutation of genes encoding enzymes of the EMP metabolic pathway. In addition, 8 susceptibility to other antibiotics was tested in the fosfomycin resistant mutants. No 9 significant changes between the wild-type strain and the mutants were observed for any 10 of the tested antibiotics (Table S1), which strongly suggests that these mutations in 11 genes coding enzymes of the central metabolism are fosfomycin-specific resistance 12 mutations. 13

Model of S. maltophilia central metabolism 14
As a first step for deciphering how the mutations in genes encoding enzymes of 15 the EMP metabolic pathway may impact S. maltophilia physiology, a metabolic map of 16 the central metabolism, which generate energy and precursors to form biomass (36), 17 was modeled for S. maltophilia. The EMP pathway is the best analyzed glycolytic route. 18 It is based on the sequential activity of ten individual enzymes. The first five form the 19 upper glycolysis (Glk, Pgi, Pfk, Alf1, TpiA) in which, using ATP, hexoses are 20 converted into trioses phosphate; whereas in the lower glycolysis (GapA, Pgk, GpmA, 21 Eno, PykA), pyruvate is formed from the trioses phosphate, at the same time that 22 NADH and ATP are generated. The pyruvate obtained is decarboxylated by the action 23 of pyruvate dehydrogenase complex and enters as acetyl-CoA to the tricarboxylic acids 24 (TCA) cycle (37). The EMP pathway may also function in a gluconeogenic regime, 25 forming hexoses phosphate from trioses phosphate (38). All enzymes of the EMP 1 pathway were identified in S. maltophilia D457 (Figure 1 and Figure S1). Moreover, the 2 Entner-Doudoroff (ED) route, another glycolytic pathway that also forms trioses 3 phosphate from hexoses phosphate, is present as well in S. maltophilia. It is important 4 to notice that two enzymes of the central metabolism of D457, GpmA and Eno, present 5 isoenzymes capable of carrying out the same chemical reaction. As shown in Figure 1, 6 all the fosfomycin resistance mutations are located in genes encoding enzymes of the 7 lower glycolytic pathway. 8 Fosfomycin resistance mutations impair the activity of enzymes of the S. 9 maltophilia central metabolism 10 To determine whether or not the mutations cause a loss of function of the 11 encoded proteins, the enzymatic activity of Gap, Pgk, Gpm and Eno was measured in 12 the mutants and in the wild-type strain. As shown in Figure 2 of the general physiological state of the cell, including its redox balance, were 20 measured. In particular, the activities of the glucose-6P dehydrogenase (Zwf), which 21 connects the glucose-6P with the ED and pentose phosphate (PP) pathways, and 22 isocitrate dehydrogenases (Icd NAD + and Icd NADP + ) activity, from the TCA cycle, 23 were determined. The activity of the enzyme Zwf increased by 1.5 to 2.5-fold in the 24 four fosfomycin resistant mutants ( Figure 3) as compared with the wild-type D457 25 strain, whereas the activities of either Icd NAD + or Icd NADP + did not change in any of 1 the studied mutants. 2 Fosfomycin resistance is not the consequence of a metabolic rearrangement 3 that modifies S. maltophilia susceptibility to oxidative stress 4 It has been proposed that the activity of antibiotics may depend on the bacterial 5 oxidative response (39). One of the key elements in such response is Zwf, an enzyme 6 with a critical role in the supply of NADPH, which is a relevant cofactor for 7 maintaining cellular redox balance (40, 41). We have observed that this enzyme 8 presented an increased activity in the mutants as compared with the wild-type strain (see 9 above). To address if this increased activity might be the reason for fosfomycin 10 resistance, zwf was inactivated in the FOS4 and FOS7 mutants and in the D457 wild-11 type strain. The inactivation of zwf causes a slight increase in MIC levels of fosfomycin 12 from 192 to 256 g/ml in D457 wild-type strain, whereas this inactivation does not 13 change fosfomycin susceptibility in the tested mutants. 14 Besides, the role of the mutations in the response to oxidative stresses was tested 15 by analyzing the susceptibility of the mutants to H 2 O 2 and menadione. As shown in 16 Table 3, mutations conferring fosfomycin resistance did not alter the susceptibility of S. 17 maltophilia to these compounds, whereas, as expected, zwf inactivation causes an 18 increase in the susceptibility to these oxidative stressors. These results indicate that the 19 susceptibility to fosfomycin of S. maltophilia mutants with defective lower glycolysis 20 enzymes is a specific phenotype, not due to a change in the oxidative stress response. 21 The impaired activity of EMP enzymes is associated with S. maltophilia 22

antibiotic resistance 23
Our results strongly suggest that the cause of fosfomycin resistance in the 24 studied mutants is a reduced activity of the enzymes of the lower glycolysis pathway in 25 S. maltophilia. However, it is still possible that these enzymes may present 1 moonlighting activities in this bacterial species besides its metabolic role, which could 2 be associated with the antibiotic resistance in a metabolic independent manner (42,43). 3 This possibility is supported by the fact that, while mutations in these genes are easily 4 selected in S. maltophilia, the information present in the Profiling of the E. coli 5 Chromosome (PEC) database, the Keio library and the Transposon-directed insertion 6 site sequencing (TraDIS) database (44-46) support that they are highly relevant 7 (eventually essential) in E. coli. 8 To determine if the recovery of the glycolytic activity, independently of a 9 putative additional activity of the S. maltophilia inactivated enzymes, could be on the 10 basis of the observed antibiotic resistance phenotype, a partial version of the Glucobrick 11 II, containing the E. coli genes gapA, pgk, gpmA and eno was introduced in the S. 12 maltophilia fosfomycin resistant mutants and in the wild-type strain and the 13 susceptibility to fosfomycin of these strains was measured. By this approach the 14 enzymatic activity, here provided by the E. coli orthologues of the S. maltophilia 15 inactivated genes, was decoupled from another potential activity of such S. maltophilia 16 proteins. As shown in Figure 4, the expression of the E. coli GapA-Pgk-GpmA-Eno 17 enzymes increased the susceptibility to fosfomycin of all FOS mutants, although the 18 levels achieved were not the same as those of the wild-type strain. This partial 19 complementation of the phenotype of resistance strongly supports that the absence of 20 enzymatic activity of the analyzed EMP enzymes contribute to fosfomycin resistance in 21

S. maltophilia. 22
Fosfomycin resistance of mutants defective in EMP enzymes is not the 23

consequence of an increased production of phosphoenolpyruvate 24
Fosfomycin inhibits the action of MurA because it is structurally similar to PEP, 1 one of the substrates of this enzyme. The EMP enzymes associated with fosfomycin 2 resistance that are inactivated in the S. maltophilia fosfomycin resistant mutants present 3 reversible activity and belong to a pathway that leads to either PEP biosynthesis or 4 consumption depending on the metabolic regime. It might be then possible that the 5 inactivation of such enzymes may change the intracellular PEP concentrations, affecting 6 the binding of fosfomycin to the active site of MurA through a possible competition 7 between PEP and fosfomycin, which may render a reduced susceptibility to fosfomycin. 8 To analyze this possibility, the concentration of PEP was analyzed in the wild-type 9 D457 strain and in the fosfomycin resistant mutants. In none of the mutants an increase 10 in the intracellular concentration of PEP was observed, ruling out the hypothesis that the 11 cause of the reduced susceptibility to fosfomycin of the analyzed mutants is an 12 increased production of PEP. 13 Fosfomycin resistance mutations impair the gluconeogenic pathway of S. 14

maltophilia. 15
The mutations selected in presence of fosfomycin compromise the activity of 16 relevant enzymes of S. maltophilia central metabolism. It is then expected, this would 17 have relevant physiological consequences. To have a general scope of these 18 consequences, the growth of S. maltophilia mutants and of wild-type parental strain 19 under different conditions was measured. Just small differences in growth among the 20 tested strains were observed for bacteria growing in rich LB medium ( Figure 5A), 21 indicating these mutations do not impose a relevant general, non-specific, fitness cost. 22 In addition, the mutants can grow using glucose, which imposes a glycolytic 23 metabolism, although in the case of FOS1 and FOS8 at a different rate ( Figure 5B). 24 Nevertheless, the mutants were unable to grow using succinate as the carbon source 25 ( Figure 5C). This impaired growth in succinate was not observed when the mutants 1 were complemented with either the wild-type allele of each of the mutated enzymes or 2 the E. coli-derived Glucobrick II ( Figure S2). The blocking of any of the enzymes of the 3 EMP pathway, between triose phosphate isomerase and pyruvate kinase, breaks the 4 amphibolic process in two branches. These branches work in opposite directions, 5 starting either from glucose or from pyruvate to provide energy or biosynthetic 6 intermediates (47). Since S. maltophilia additionally displays the one-direction ED 7 pathway for glucose catabolism, fosfomycin low susceptibility mutants can grow in 8 minimal medium with glucose. Nevertheless, succinate as exclusive carbon source does 9 not support growth of the mutants because gluconeogenesis and consequently synthesis 10 of hexose phosphates are impaired. 11

Fosfomycin resistant mutants do not present an altered intracellular 12 accumulation of fosfomycin 13
While the primary cause of fosfomycin resistance in S. maltophilia is the 14 inactivation of EMP enzymes, it might be possible that such inactivation impairs the 15 accumulation of the antibiotic within the cell, which could be due to either a reduced 16 uptake or to the degradation of the antibiotic. A search of possible fosfomycin 17 transporters in S. maltophilia D457 was carried out using Blast (48)  for the catabolism of glucose-6P and glycerol-3P, none of the strains were able to grow 25 using these sugars as unique carbon sources, conditions at which E. coli can grow 1 ( Figure S3). This result suggests that S. maltophilia lacks glucose-6P and glycerol-3P 2 transporters, which are the regular gates for fosfomycin entrance in other pathogens. 3 Besides, the search of fosfomycin modifying enzymes in S. maltophilia D457 genome 4 did not allow to detect any gene homologous to the fosfomycin resistance proteins 5 (FosA, FosB, FosX, FomA, FomB and FosC) so far described in the literature. 6 Despite S. maltophilia genome does not harbor genes encoding neither the 7 canonical fosfomycin transporters nor already known fosfomycin inactivating enzymes, 8 it might still be possible that other (still unknown) elements may contribute to an 9 impaired accumulation of the antibiotic inside the mutants. To analyze this possibility, 10 the intracellular accumulation of fosfomycin in the different strains was measured (49)  11 after one hour of incubation with 2 mg/ml fosfomycin in exponential growth phase 12 cultures. As a control, E. coli K-12 and a deletion mutant on the fosfomycin transporter 13 UhpT (44), as well as P. aeruginosa PA14 and insertion mutants on the fosfomycin 14 transporter GlpT or the fosfomycin resistance protein FosA (50) were used. As shown 15 in Figure 6, the amount of intracellular fosfomycin is lower both in E. coli and P. 16 aeruginosa when their respective fosfomycin transporters (GlpT and UhpT) are 17 inactivated. Conversely, an increased fosfomycin concentration was observed in the 18 FosA mutant relative to the parental PA14 strain, which supports the validity of these 19 assays. Nevertheless, the intracellular concentrations of fosfomycin were similar in the 20 S. maltophilia D457 wild-type strain and in the isogenic fosfomycin resistant mutants. 21 These results suggest that the resistance to fosfomycin of the FOS mutants is not due to 22 a reduced intracellular concentration of this antibiotic. Notably, fosfomycin 23 accumulation in S. maltophilia is much lower than that found in E. coli or P. 24 aeruginosa. Indeed, intracellular fosfomycin concentration in S. maltophilia is in the 25 range of that observed for the GlpT-defective P. aeruginosa mutant. This low 1 intracellular concentration, likely associated to the lack of canonical antibiotic 2 transporters, could be the cause of the intrinsic lower susceptibility of S. maltophilia 3 D457 to fosfomycin compared to E. coli K-12 and P. aeruginosa PA14 (49,51,52). 4

Effects of fosfomycin resistance mutations on the transcriptional profile of S. 5 maltophilia 6
In order to know if the mutation of genes encoding the enzymes of the central 7 metabolism change the transcriptional profile in a way directly related to fosfomycin 8 resistance, the transcriptomes of the fosfomycin resistant mutants were compared to that 9 of the wild-type strain. Changes in the expression levels of just 67 of the 4210 genes 10 that form the genome of S. maltophilia D457 were detected (Table S2). Most changes 11 were specific for each mutant, indicating that the observed transcriptomic changes were 12 unlikely associated to the common phenotype of fosfomycin resistance ( Figure S4). 13 Concerning changes that may explain the resistance phenotype is important to note the 14 absence of relevant transcriptional changes in genes related to cell wall synthesis, such 15 as the gene encoding the fosfomycin target MurA and SMD_1053, SMD_1054, 16 SMD_0334, nagZ and SMD_2885, predicted to be involved in the recycling of the 17 peptidoglycan (Table S3). These results support that an increased expression of either 18 the fosfomycin target (MurA) or the alternative peptidoglycan recycling pathway is not 19 the cause of fosfomycin resistance in the analyzed mutants. 20

Discussion 21
So far described fosfomycin resistance mechanisms can be clustered into three 22 classical categories of antibiotic resistance acquisition (1): alterations in fosfomycin 23 transport, antibiotic inactivation and alterations in the target enzyme or peptidoglycan 24 biosynthesis (17). Herein, using a set of in vitro selected mutants, we have shown that 25 none of these already known mechanisms seem to be involved in the acquisition of 1 mutation-driven fosfomycin resistance by S. maltophilia. In this microorganism, the 2 acquisition of resistance is due to the inactivation of enzymes belonging to the EMP 3

pathway. 4
Our results indicate that the inactivation of these enzymes does not cause major 5 changes in the transcriptomes of the mutants that may justify resistance as the 6 consequence of a collateral effect of the selected mutations on the expression of the 7 aforementioned fosfomycin resistance mechanisms. Inasmuch, intracellular 8 accumulation of fosfomycin was similar in the wild-type and the mutant strains, which 9 support that resistance is neither due to an impaired fosfomycin uptake (25) nor to its 10 degradation via the activity of fosfomycin-inactivating enzymes (29, 53). Also 11 supporting this result is the fact that the genome of S. maltophilia does not encode 12 homologous of the already known fosfomycin resistance proteins or its transporters 13 GlpT and UhpT. 14 Other mechanisms leading to fosfomycin resistance are modifications of the 15 target MurA (21) or changes in its expression level. Nevertheless, when the mutants 16 were sequenced, no mutations in murA were found and the analysis of the 17 transcriptomes indicates that murA is not expressed at higher levels in the resistant 18 mutants than in the wild-type strain. Same happens with the pathway involved in the 19 recycling of peptidoglycan, which increased expression may contribute to fosfomycin 20 resistance (7). The expression of the genes encoding the enzymes of this pathway is not 21 higher in the mutants that in the wild-type strains as shown in the transcriptomic 22

studies. 23
Therefore, classical antibiotic resistance mechanisms (1,54) do not seem to be 24 the cause of fosfomycin resistance in S. maltophilia. Although, at above stated, there are 25 not relevant transcriptional changes in the mutants, these strains appear to show a 1 different physiological state than the wild-type strain, as evidenced by the fact that they 2 exhibit increased Zwf activity together with the loss of function of the mutated 3 enzymes. These changes do not modify the response to oxidative stress, an element that 4 could be relevant in the activity of antibiotics (39). However, it is worth mentioning that 5 the regulation of the metabolic fluxes of carbon metabolism include additional 6 mechanisms other than transcriptional regulation (55). Among them, allosteric 7 regulation as well as the activity of posttranscriptional or posttranslational regulators 8 can change the production levels and activity of different proteins (eventually involved 9 in the resistance phenotype) without changing their mRNA levels (56). 10 The mutated enzymes belong to the amphibolic metabolic pathway (EMP and 11 gluconeogenesis) which includes PEP, the natural substrate of MurA. Fosfomycin due 12 to its structural similarities to PEP binds and inhibits MurA. It might be then possible 13 that inactivation of such enzymes in the fosfomycin resistant mutants may produce an 14 increased synthesis of PEP that could outcompete fosfomycin for its binding to MurA. 15 However, the concentrations of PEP are no higher in the fosfomycin resistant mutants 16 than in the wild-type strain, which goes against this possibility. Several enzymes from 17 central metabolism are moonlighting proteins; they display functions unrelated to their 18 enzymatic activity (42). The complementation of the mutants with E. coli enzymes 19 restored their susceptibility to fosfomycin, which indicates that the impaired activity of 20 these metabolic enzymes is on the basis of the observed phenotype of fosfomycin 21 resistance. Nevertheless, a possible function of the S. maltophilia enzymes not related to 22 their known metabolic function cannot be totally discarded. 23 Previous analysis has shown that E. coli mutants deficient in the metabolic 24 enzyme isocitrate dehydrogenase are resistant to nalidixic acid (35). However, little 25 work is still available in the crosstalk between metabolism (and metabolic robustness) 1 and antibiotic resistance (57, 58), despite the fact that metabolic interventions may 2 improve the activity of the antibiotics (33, 59-61) and that bacterial metabolism can 3 constrain the evolution of antibiotic resistance (13). Our results highlight the importance 4 that the modification of the activity of enzymes belonging to central metabolism may 5 have in the susceptibility to antibiotics, as fosfomycin, that are not known to interact 6 with such enzymes. The finding that fosfomycin activity is highly dependent on the 7 bacterial metabolic status, being more active when bacteria grow intracellularly (27, 28) 8 or under acidic conditions and anaerobiosis in urine (62), further support that antibiotic 9 activity and, consequently antibiotic resistance, are interlinked with the bacterial 10 metabolism. 11

Material and methods 12
Bacterial strains and culture conditions. 13 All bacterial strains, plasmids and oligonucleotides and used in this study are 14 listed in Tables S4 and S5

Isolation of fosfomycin resistant mutants 25
Around 10 8 S. maltophilia D457 bacteria cells were plated on MH agar Petri 1 dishes containing 1024 μg/ml fosfomycin and were grown at 37 ºC during 48 h. The 2 mutants selected in these conditions were grown on LB agar without antibiotic (three 3 sequential passages) and then were grown again on MH agar containing 1024 μg/ml 4 fosfomycin to ensure that the observed phenotype was not transient. The susceptibility 5 of mutants to fosfomycin was tested (see below), for further studies 4 mutants were 6 randomly selected and dubbed FOS1, FOS4, FOS7 and FOS8. 7

DNA extraction, whole genome sequencing and SNP identification 8
Chromosomal DNA from each mutant (FOS1, FOS4, FOS7 and FOS8) and the 9 wild-type strain (D457) was obtained from overnight cultures using the GNOME DNA 10 kit (MP Biomedicals). Genomic DNAs were sequenced using the Illumina technology 11 at the Parque Científico of Madrid, Spain. The samples were subjected to single-end 12 sequencing with a read-length of 1x150 and a coverage between 26 and 41X was 13 obtained. The genomic sequences of the strains were compared with S. maltophilia 14 D457 reference genome (NC_017671.1) and visualized using the software FIESTA 1.1 15 (http://bioinfogp.cnb.csic.es/tools/FIESTA). Mutations were filtered according to 16 sequence quality (>30) and the mutation effect in the protein sequence (moderate and 17 high effect), and the variants absent in the control D457 parental strain were studied. 18 Provean predictor (provean.jcvi.org) was used to anticipate whether an amino acid 19 substitution or indel had an impact on the biological function of the coding protein. The presence of the mutations detected from the whole genome sequencing 21 analysis was confirmed as described in Supplemental Materials and Methods. 22

Antimicrobial susceptibility assays 23
The minimal inhibitory concentrations (MICs) of gentamicin, tobramycin, 1 ciprofloxacin, nalidixic acid, ceftazidime, colistin, tetracycline, chloramphenicol and 2 fosfomycin were determined for each strain on LB agar using MIC test strips (MIC Test 3 Strips, Liofilchem Diagnostics). For phenotypic the analysis of mutants complemented 4 with the Glucobrick module II, antibiotic disks (Oxoid) were used. Plates were 5 incubated at 37 °C and results were analyzed after 20 h. Since commercial fosfomycin 6 disks contain glucose 6-P, fosfomycin susceptibility assays were also performed, under 7 the same growing conditions, using paper disks (9 mm, Machery-Nagel) impregnated 8 with 0.5 mg of fosfomycin. The experiments were performed in triplicate. 9 Complementation of fosfomycin resistant mutants and generation of zwf deletion 10 mutants 11 The genes eno, gpmA, pgk and gapA, encoding glycolytic enzymes, were 12 obtained from the wild-type strain S. maltophilia D457 by PCR amplification and 13 introduced in S. maltophilia as described in Supplemental Material and Methods. 14 To complement the mutants with a partial version of the Glucobrick module II, 15 which contains the genes of the lower glycolysis enzymes of E. coli K12, (gapA, pgk, 16 gpmA, eno and pyk) (64), the pSEVA224 GBII plasmid containing these genes was 17 The zwf gene was deleted in different S. maltophilia strains by homologous 1 recombination as described (66)  The plates were incubated at 37 ºC with 10 s of shaking every 10 min. 17

Protein quantification 18
Protein concentration was determined following the Pierce BCA Protein Assay 19 Kit (Thermo Scientific) protocol in 96 well plates (Nunc MicroWell Thermo Fisher). 20

In vitro activity assays of the lower glycolysis enzymes and dehydrogenases 21
Cells were harvested at exponential phase (OD 600 = 0.6) by centrifugation at 22 5100 xg and 4 °C and washed twice in 0.9% NaCl and 10 mM MgSO 4 . Once washed, 23 cells were disrupted by sonication at 4 °C and the cell extracts were obtained by 1 centrifugation at 23100 xg for 30 min at 4 °C. 2 NAD(P) + reduction or NAD(P)H oxidation was monitored 3 spectrophotometrically at 340 nm and 25 °C with intermittent shaking in microtiter 4 plates using Spark 10M plate reader (TECAN). Each reaction was performed using 5 three biological replicates and the specific activities were obtained by dividing the 6 measured slope of NAD(P)H formation or consumption by the total protein 7 concentration. Enzymatic activities of dehydrogenases (glucose-6-phosphate, isocitrate 8 and glyceraldehyde-3P dehydrogenases) were measured as described (68). Enzymatic 9 activities of phosphoglycerate kinase, phosphoglycerate mutase and enolase were 10 assayed following the protocol described by Pawluk,A. et al (69) with some 11 modifications in a two-step reaction (see Supplemental Materials and Methods). 12

Quantification of intracellular phosphoenolpyruvate and fosfomycin 13
The amount of PEP was measured from cultures in exponential growth phase in 14 LB medium (OD 600 = 0.6). Twenty ml of each culture were centrifuged at 4500 xg for 3 15 min at 4 °C. PEP Colorimetric / Fluorometric Assay Kit protocol (Sigma-Aldrich) was 16 used with some modifications that are described in Supplemental Materials and 17

Methods. 18
Assays to test fosfomycin accumulation in bacterial cells were conducted as 19 previously stated (49), with some modifications that are described in Supplemental 20

H 2 O 2 and menadione susceptibility test 22
The susceptibility to H 2 O 2 and menadione was tested as described previously 23 with some modifications (70). Sterile paper disks (9 mm, Machery-Nagel) were 24 impregnated with 10 l of 2.5% H 2 O 2 or 20 l of 0.2 M menadione and placed on LB 1 agar plates. The diameter of the zone of growth inhibition around each disk was 2 measured after 20 h of incubation at 37 °C. The experiment was performed in triplicate. 3

Metabolic map of S. maltophilia 4
To model the metabolic map of S. maltophilia D457, indicating possible 5 enzymes of the central metabolism and route bypasses, the BioCyc database (71) was 6 used. The sequence of the enzymes was obtained from the complete genome of S. 7 maltophilia D457 (72). In addition, the amino acid sequence of the enzymes of the 8 central metabolism of P. aeruginosa PAO1 (73) and E. coli (74) were aligned using the 9 Blast tool (48)  3 ND: Not done, each strain was complemented with the wild-type allele of the 4 corresponding mutated gene. 5