The metabolic enzyme fructose-1,6-bisphosphate aldolase acts as a transcriptional regulator in pathogenic Francisella

The enzyme fructose-bisphosphate aldolase occupies a central position in glycolysis and gluconeogenesis pathways. Beyond its housekeeping role in metabolism, fructose-bisphosphate aldolase has been involved in additional functions and is considered as a potential target for drug development against pathogenic bacteria. Here, we address the role of fructose-bisphosphate aldolase in the bacterial pathogen Francisella novicida. We demonstrate that fructose-bisphosphate aldolase is important for bacterial multiplication in macrophages in the presence of gluconeogenic substrates. In addition, we unravel a direct role of this metabolic enzyme in transcription regulation of genes katG and rpoA, encoding catalase and an RNA polymerase subunit, respectively. We propose a model in which fructose-bisphosphate aldolase participates in the control of host redox homeostasis and the inflammatory immune response.

• Unfortunately, the work, as presented, is too incomplete to enable interpretation. For example, it remains entirely unaddressed why fba deficiency renders a specific in vitro phenotype with gluconeogenic over glycolytic substrates since fba is a shared enzyme of both pathways. In addition, it is entirely unclear why the fba mutant is sensitive to the carbon source provided in the macrophage cell culture medium (could it be that this phenotype reflects an extracellular, rather than intracellular event; can't tell from experimental methods). It is unclear how to interpret the metabolic profiling data since they reflect a combination of both host and microbial metabolites, and are pool size, rather than flux, measurements.
• Metabolic ambiguities aside, the pathophysiologic significance of fba's apparent transcriptional regulatory activity is entirely unclear as the metabolic defects seem enough to explain the observed attenuation in macrophages and mice. Moreover, the authors fail to identify a physiologic setting in which this transcriptional regulatory activity actually occurs (since genetic loss of fba itself seems unlikely in nature). Lacking this, it is difficult to imagine a specific pathophysiologic role for this activity as it would only seem to render Fn more vulnerable to the induction and effects of a pro-inflammatory/antimicrobial response. A more minor, but still important point, is that the authors fail to provide a direct demonstration of in vitro DNA binding and identification of specific operator sequences that could help to more rigorously identify target regulated genes.
• The section on inhibitor screening is uninformative and of unclear relevance in this manuscript.
Reviewer #2 (Remarks to the Author): In their interesting study, Ziveri and colleagues show that the metabolic enzyme fructose-1,6bisphosphate aldolase (FBA) is important for the intracellular replication of Francisella tularensis. They also present evidence that this same enzyme plays a role in controlling the expression of certain genes in this organism, which is particularly novel. The evidence that FBA is essential for intramacrophage growth when the growth medium contains gluconeogenic substrates is reasonably strong, although there may be an issue with the interpretation of some of the complementation data (see below). The evidence that FBA plays a regulatory role needs to be both stronger and more complete. Although the authors present reasonable evidence that FBA contributes to the control of redox homeostasis by controlling the expression of the katG gene, it's currently unclear whether the ability of FBA to control the expression of katG (or any other gene), is important for intramacrophage growth or for virulence in mice.

Major comments
1. There may be a problem with the experiments in which the effects of the fba deletion are complemented with a plasmid-borne copy of the fba gene (Figs 2,3,4). From reading the materials and methods it looks as though the complemented fba mutant strain contains a plasmid that encodes fba. However, it's not clear whether the WT strain and fba mutant strains used in the same experiments contain an empty vector so that the results found with WT, fba mutant, and complemented strain can be interpreted properly. For these experiments it would of course be important for all cells to contain a plasmid vector.
2. The evidence that FBA plays a regulatory role needs to be stronger. The authors use proteomics to show that the abundance of ~30 proteins changes in an fba mutant compared to WT (Fig.6). However, the authors only present qRT-PCR evidence that FBA controls the expression of two genes (katG in Fig. 5 and rpoA in Fig. 7B). How many of the genes corresponding to the rest of the 30 proteins are controlled by FBA? A genomewide analysis (e.g. RNA-Seq) of the effects of FBA on gene expression would strengthen the study. If genomewide studies could not be performed, bolstering the findings of more extensive qRT-PCR analyses with the results of lacZ reporter assays would be suitable.
3. It would be important to show that the effects of fba on gene expression can be complemented with fba in trans (Figs 5 and 7). 4. The authors use ChIP followed by qPCR to show that an epitope-tagged version of FBA associates with a variety of promoter regions in F. novicida. However, it's not clear from the materials and methods whether these assays are truly quantitative. The authors should document how their assays are quantitative or use established procedures for quantifying enrichment by ChIP and qPCR. 5. Do the authors have any biochemical evidence that support direct binding of FBA to the katG or rpoA2 promoter regions?
6. There is no test of whether the ability of FBA to control gene expression is important for pathogenesis. If FBA plays a key role in pathogenesis by modulating redox homeostasis through an effect on katG expression, is FBA important for virulence in cells that lack katG? 7. In relation to point 6. Don't the findings in Fig 5D and Fig 5F indicate that the ability of FBA to control the expression of katG (or other genes) is not important for intramacrophage growth?
Minor comments 8. Does a catalytically inactive variant of FBA still function as a regulator? I think this would be something that would be interesting to test, but not essential for the current study. This manuscript explores the role of a glycolytic enzyme, FBA, on Francisella virulence. The manuscript is clearly written, albeit heavily dependent on acronyms, and the experiments seem well-done. I was asked to comment on the proteomics work but I also have one general comment about the key finding.
Major comments 1. The key conclusion here is that FBA is acting directly on transcription but I think the data does not quite fully support that, yet. Certainly the ChIP experiments are suggestive but it would be nice to see a negative control. For example, one possible explanation for the results is that an energy 'metabolon' is needed in the neighbourhood of those genes. If the same experiment were done with one of the up-or down-stream glycolysis enzymes, would one see the same enrichments as in Fig. 7? 2. Proteomics. I always get concerned when someone imputes data and then tries to calculate significance on those data since the imputation artificially suppresses the true biological variance. How many of the significantly changed proteins were those who had to have values imputed?

Answers to reviewers
The comments and suggestions raised by the reviewers were carefully and thoroughly addressed. Clarifications were brought where needed and substantial additional experimental work was performed. The new data fully supported and consolidated our initial observations. Additional figures, text and references were included.

Reviewer 1
« This paper characterizes the role of the evolutionary ubiquitous metabolic enzyme, fructose bisphosphate aldolase (fba), in the pathogenicity of the intracellular bacterium, Francisella novicum (Fn). Using a genetic knockout of fba in Fn, the authors report several fascinating, if not unexpected, phenotypes. These include: (i) in vitro essentiality of fba for growth on gluconeogenic but not glycolytic carbon sources; (ii) a more pronounced attenuation of fba-deficient Fn in a macrophagelike cell line when cultured in presence of a gluconeogenic, but not glycolytic, carbon source; (iii) a profound attenuation of fba-deficient Fn in an in vivo mouse model of infection; (iv) a novel DNA binding transcriptional regulatory activity of fba, whose absence in fba-deficient Fn is associated with increased resistance to reactive oxygen intermediates, reduced IL-6 secretion, and increased survival in interferon-gamma activated macrophage-like cells. Based on these findings, the authors seek to ascribe Fn's pathogenicity, in part, to a combination of fba's canonical enzymatic activity and this novel non-metabolic transcriptional regulatory activity ».
• "Unfortunately, the work, as presented, is too incomplete to enable interpretation. For example, it remains entirely unaddressed why fba deficiency renders a specific in vitro phenotype with gluconeogenic over glycolytic substrates since fba is a shared enzyme of both pathways. In addition, it is entirely unclear why the fba mutant is sensitive to the carbon source provided in the macrophage cell culture medium (could it be that this phenotype reflects an extracellular, rather than intracellular event; can't tell from experimental methods). It is unclear how to interpret the metabolic profiling data since they reflect a combination of both host and microbial metabolites, and are pool size, rather than flux, measurements". Answer: It is likely that glycolytic substrates can use alternate route in the ∆fba mutant, and in particular the pentose phosphate pathway (PPP). Indeed, we have recently demonstrated (Brissac et al 2015), using 13 C-labeled glucose in wild-type F. novicida, the recycling of carbons through the PPP. This information was added in the text (line 162-165).
Of note, in mammalians, three forms of Class I FBAs are found (designated A, B or C). Aldolases A and C have been shown to be mainly involved in glycolysis, whereas aldolase B is involved in both glycolysis and gluconeogenesis. Hence, it is conceivable that class II FBAs may also display different roles in glycolysis or gluconeogenesis.
It is undeniably not possible to affirm that the changes observed in macrophages infected by Francisella strictly reflect the adaptation of the macrophage metabolism and we agree with the reviewer that it might correspond to the combined activity of bacterial and host metabolisms. However, we believe that the bacterial contribution to the metabolome should be -if not marginal-at least minor since: i) approximately only 10% of macrophages cells are generally infected, in the infection conditions used; ii) each infected cell generally does not contain more than a hundred bacteria after 24 h; and iii) the average volume of a bacterial cell is much lower than that of a macrophage (in the range of 1-2x10 -12 cm 3 per bacterium / 1-4x10 -9 cm 3 per cell). Thus, it is reasonable to assume that the amount of metabolites contained in bacterial cells, in the samples analyzed, do not significantly contribute to the overall amounts measured.
These precisions were added in the text: "… The changes observed in macrophages infected by There is currently no direct method to measure partitioning of metabolite pools in two types of cells in culture without prior cell separation, which is a difficult task to carry on, keeping a proper metabolism quenching (especially in the case of an intracellular infection). An experiment of 13 C isotopic profiling may highlight the active pathways that are actually operating during infection. However, this would necessitate to be in a metabolic stationary state, which is not the case in our experiments, since metabolic pools are evolving during infection. It would also most probably require a homogenously infected cell population. Furthermore, even if 13 C flux analyzes could be used to measure metabolic fluxes, it would be an ambitious task to attribute flux partition between both organisms since same metabolic reactions can occur in parallel in the two organisms and give undifferentiated 13 C isotopic patterns from the same substrates Despite these limitations, our analyzes allowed us to evaluate that metabolic pools were affected in infected macrophages compared to uninfected macrophages, and to identify the affected metabolic pathways. In particular, we clearly observed an impact of infection on TCA metabolites and on metabolites from gluconeogenesis/PPP above FBP (Glc6P, UDP-Glc, and Pentose-P metabolites -P5P and Sed7P).

• "Metabolic ambiguities aside, the pathophysiologic significance of fba's apparent transcriptional regulatory activity is entirely unclear as the metabolic defects seem enough to explain the observed attenuation in macrophages and mice. Moreover, the authors fail to identify a physiologic setting in which this transcriptional regulatory activity actually occurs (since genetic loss of fba itself seems unlikely in nature). Lacking this, it is difficult to imagine a specific pathophysiologic role for this activity as it would only seem to render Fn more vulnerable to the induction and effects of a proinflammatory/antimicrobial response. A more minor, but still important point, is that the authors fail to provide a direct demonstration of in vitro DNA binding and identification of specific operator sequences that could help to more rigorously identify target regulated genes".
Answer: The inability of the fba mutant to multiply in the presence of gluconeogenic substrates in cells reflects its inability to use these substrates as carbon sources (recapitulating the same phenotypes observed in broth). We have previously reported a comparable situation upon inactivation of the gluconeogenic enzyme GlpX (converting fructose1,6-biphosphate to fructose 6-phosphate). Indeed, this mutant had a severe intracellular multiplication defect compared to wild-type, when cells were supplemented by gluconeogenic substrates (such as glycerol, pyruvate or amino acids; Brissac et al. 2015), but showed wild-type multiplication in medium containing glucose.
A somewhat comparable situation has been reported with the pathogenic bacterium Legionella (Price et al Science 2011), where the severe intracellular growth defect of an ankB null mutant could be rescued by supplementation with amino acids or pyruvate, as efficiently as by genetic complementation.
We think that the enzymatic and regulatory functions of FBA act synergistically and that their respective importance in virulence will depend on the cell type, the tissue and the phase of the infectious cycle in vivo. One should bear in mind that intracellular survival and dissemination of Francisella relies on a complex and tight temporally-controlled dampening of cytokine production. Indeed, during, active cytosolic multiplication, the bacterium transiently silences the AIM2 inflammasome but ultimately somehow takes advantage of caspase 1 activation and pyropsosis to promote its escape and dissemination to adjacent cells. These considerations were added in the Discussion « Hence, FBA-mediated control of KatG expression should not be seen as a specific pathophysiologic role to render Francisella more vulnerable to a pro-

inflammatory response but rather as a mean for the bacterium to combine metabolism and transcriptional regulation to optimally modulate the redox status of the infected cell. Indeed, one should bear in mind that intracellular survival and dissemination of the pathogen relies on its capacity to counteract both nutritional and innate immunity through: i) an adaptation to the available nutritional resources and ii) a tight and temporallycontrolled dampening of cytokine production (Broz and Monack, 2011), ultimately leading to inflammasome activation and pyropsosis, allowing bacterial release and dissemination to adjacent cells».
Furthermore, FBA is not only acting on catalase expression but also regulates a number of other proteins and KatG itself is also under the control of additional regulatory stimuli. Thus, the action of FBA is integrated into a complex regulatory network during infection.
We have now confirmed the direct binding of FBA to the promoter regions of katG and rpoA genes by gel shift and β-galactosidase reporter assays. These results were added in the text and in Fig. 7.
• "The section on inhibitor screening is uninformative and of unclear relevance in this manuscript. Answer: Following the reviewer's suggestion, this section was removed.

Reviewer 2
"In their interesting study, Ziveri and colleagues show that the metabolic enzyme fructose-1,6bisphosphate aldolase (FBA) is important for the intracellular replication of Francisella tularensis. They also present evidence that this same enzyme plays a role in controlling the expression of certain genes in this organism, which is particularly novel. The evidence that FBA is essential for intramacrophage growth when the growth medium contains gluconeogenic substrates is reasonably strong, although there may be an issue with the interpretation of some of the complementation data (see below). The evidence that FBA plays a regulatory role needs to be both stronger and more complete. Although the authors present reasonable evidence that FBA contributes to the control of redox homeostasis by controlling the expression of the katG gene, it's currently unclear whether the ability of FBA to control the expression of katG (or any other gene), is important for intramacrophage growth or for virulence in mice".

Major comments.
1. "There may be a problem with the experiments in which the effects of the fba deletion are complemented with a plasmid-borne copy of the fba gene (Figs 2,3,4). From reading the materials and methods it looks as though the complemented fba mutant strain contains a plasmid that encodes fba. However, it's not clear whether the WT strain and fba mutant strains used in the same experiments contain an empty vector so that the results found with WT, fba mutant, and complemented strain can be interpreted properly. For these experiments it would of course be important for all cells to contain a plasmid vector". Answer: As requested by the reviewer, we have now compared wild-type F. novicida, containing -or not-the empty vector (pKK214), to ∆fba containing -or not-the empty vector (pKK214).
We repeated growth experiments in CDM (CDM-Glucose, CDM-Glycerol and CDM-Glucose+Glycerol), and in cells (in J774-1 + Glucose and J774-1 + Glycerol). We also repeated the oxidative stress assay (+ H 2 O 2 1 mM). In each of the assays, the presence of the empty vector (pKK214) had no detectable impact on the properties of the strain (wild-type or ∆fba mutant), fully confirming our previous results. The new data have been combined into a new Supplementary Figure (new Fig S5) and mentioned in the text.
2. "The evidence that FBA plays a regulatory role needs to be stronger. The authors use proteomics to show that the abundance of ~30 proteins changes in an fba mutant compared to WT (Fig.6). However, the authors only present qRT-PCR evidence that FBA controls the expression of two genes (katG in Fig. 5 and rpoA in Fig. 7B). How many of the genes corresponding to the rest of the 30 proteins are controlled by FBA? A genomewide analysis (e.g. RNA-Seq) of the effects of FBA on gene expression would strengthen the study. If genomewide studies could not be performed, bolstering the findings of more extensive qRT-PCR analyses with the results of lacZ reporter assays would be suitable".
Answer: As suggested, qRT-PCR analyses were performed on a series of additional genes corresponding to proteins controlled by FBA. In most cases, transcription of these genes was significantly higher in the wild type and fba-complemented strains compared to the ∆fba mutant, corroborating the proteomic analyses (shown in new Fig. S10). The following paragraph was added: "Transcription of rpoA was also significantly higher in the wild-type (>80-fold) and fbacomplemented strains compared to the ∆fba mutant, suggesting a direct role of FBA on rpoA transcription activation (Fig. S10). Transcription of seven additional genes, corresponding to proteins positively controlled by FBA, was also tested by qRT-PCR. Corroborating the proteomic analyses, transcription of these genes was higher in the wild-type and fba-complemented strains than in the ∆fba mutant. In particular, transcription of four of them (pyrG, 30S S9, glyS and pheT) was 50-fold to 200-fold higher in the wild-type strain than in the ∆fba mutant (Fig. S10)".
LacZ reporter assays were also performed to further confirm the FBA-dependent transcriptional control of katG and rpoA genes. For this, transcriptional fusions were constructed between the promoter region of either katG or rpoA genes and the full length lacZ gene of E. coli. The resulting pktaG-lacZ and prpoA-lacZ transcriptional fusions were cloned into pKK shuttle vector and transfered into wild-type and ∆fba mutant strains. β-galactosidase expression was 3.5-fold higher in the ∆fba mutant compared to wild-type with the pktaG-lacZ reporter. Conversely, β-galactosidase expression was 2.2-fold lower in the ∆fba mutant compared to wild-type with the prpoA-lacZ reporter. These new results, which fully supported the qRT-PCR and ChIP data, are described in the manuscript and included in Fig. 7C. 4. "The authors use ChIP followed by qPCR to show that an epitope-tagged version of FBA associates with a variety of promoter regions in F. novicida. However, it's not clear from the materials and methods whether these assays are truly quantitative. The authors should document how their assays are quantitative or use established procedures for quantifying enrichment by ChIP and qPCR".
Answer: As requested, we have documented in the Materials and Methods section the established procedures that we have applied for our quantifications. The following sentences were added: "Briefly, the qPCR values were first normalized as follows: i) qPCR values of the target promoter sequences (derived from ChIP and input samples) were divided by the qPCR values of the coding region of house keeping gene uvrD (Helicase) as internal control; ii) the values obtained for fba /cpFBA-HA were next normalized by dividing them by their corresponding background values (derived from ChIP and input from WT). Then, the normalized signals from fba /cpFBA-HA derived from ChIP were divided by the normalized signals of fba /cpFBA-HA derived from input samples. The results are expressed as relative enrichment of the detected fragments". 5. "Do the authors have any biochemical evidence that support direct binding of FBA to the katG or rpoA2 promoter regions?" Answer: We have now shown direct binding of FBA on the katG and rpoA promoter regions by gel shift assay. These results, which confirm the ChIP data, are included in the manuscript and shown in Fig. 7B.