A rheostat mechanism governs the bifurcation of carbon flux in mycobacteria

Abstract

Fatty acid metabolism is an important feature of the pathogenicity of Mycobacterium tuberculosis during infection. Consumption of fatty acids requires regulation of carbon flux bifurcation between the oxidative TCA cycle and the glyoxylate shunt. In Escherichia coli, flux bifurcation is regulated by phosphorylation-mediated inhibition of isocitrate dehydrogenase (ICD), a paradigmatic example of post-translational mechanisms governing metabolic fluxes. Here, we demonstrate that, in contrast to E. coli, carbon flux bifurcation in mycobacteria is regulated not by phosphorylation but through metabolic cross-activation of ICD by glyoxylate, which is produced by the glyoxylate shunt enzyme isocitrate lyase (ICL). This regulatory circuit maintains stable partitioning of fluxes, thus ensuring a balance between anaplerosis, energy production, and precursor biosynthesis. The rheostat-like mechanism of metabolite-mediated control of flux partitioning demonstrates the importance of allosteric regulation during metabolic steady-state. The sensitivity of this regulatory mechanism to perturbations presents a potentially attractive target for chemotherapy.

ARTICLE

Received 5 Dec 2015 | Accepted 11 Jul 2016 | Published 24 Aug 2016

DOI: 10.1038/ncomms12527 OPEN

A rheostat mechanism governs the bifurcation of carbon ux in mycobacteria

Paul Murima1,*, Michael Zimmermann2,*,w, Tarun Chopra1, Florence Pojer3, Giulia Fonti1, Matteo Dal Peraro1, Sylvie Alonso4, Uwe Sauer2, Kevin Pethe5 & John D. McKinney1

Fatty acid metabolism is an important feature of the pathogenicity of Mycobacterium tuberculosis during infection. Consumption of fatty acids requires regulation of carbon ux bifurcation between the oxidative TCA cycle and the glyoxylate shunt. In Escherichia coli, ux bifurcation is regulated by phosphorylation-mediated inhibition of isocitrate dehydrogenase (ICD), a paradigmatic example of post-translational mechanisms governing metabolic uxes. Here, we demonstrate that, in contrast to E. coli, carbon ux bifurcation in mycobacteria is regulated not by phosphorylation but through metabolic cross-activation of ICD by glyoxylate, which is produced by the glyoxylate shunt enzyme isocitrate lyase (ICL). This regulatory circuit maintains stable partitioning of uxes, thus ensuring a balance between anaplerosis, energy production, and precursor biosynthesis. The rheostat-like mechanism of metabolite-mediated control of ux partitioning demonstrates the importance of allosteric regulation during metabolic steady-state. The sensitivity of this regulatory mechanism to perturbations presents a potentially attractive target for chemotherapy.

1 School of Life Sciences, Swiss Federal Institute of Technology in Lausanne (EPFL), CH-1015 Lausanne, Switzerland. 2 Institute of Molecular Systems Biology, Swiss Federal Institute of Technology in Zrich (ETHZ), CH-8093 Zrich, Switzerland. 3 Protein Crystallography Platform, Swiss Federal Institute of Technology in Lausanne (EPFL), CH-1015 Lausanne, Switzerland. 4 Department of Microbiology, Yong Loo Lin School of Medicine and Immunology Programme, Life Sciences Institute, National University of Singapore, Singapore 117456, Singapore. 5 Lee Kong Chian School of Medicine and School of Biological Sciences, Nanyang Technological University, 59 Nanyang Drive, Singapore 636 921, Singapore. * These authors contributed equally to this work.

w Present address: Department of Microbial Pathogenesis, Yale University School of Medicine, New Haven, Connecticut 06510, USA. Correspondence and requests for materials should be addressed to P.M. (email: mailto:[email protected]

Web End [email protected] ) or to K.P. (email: mailto:[email protected]

Web End [email protected] )

or to J.D.M. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527

Tuberculosis (TB), caused by Mycobacterium tuberculosis, shares with HIV and malaria the distinction of being one of the deadliest infectious diseases of our time. Lipids and

fatty acids are thought to serve as important sources of carbon and energy for M. tuberculosis during infection1,2. The dual use of these substrates places a premium on metabolic regulatory mechanisms to ensure a balance between metabolite oxidation for energy gain and metabolite conservation for biomass production.

Intermediates of the tricarboxylic acid (TCA) cycle that are diverted into biosynthetic pathways must be replenished by anaplerotic reactions. During growth on glycolytic substrates, anaplerosis involves transformation of glycolytic intermediates (C3-units) into TCA cycle intermediates (C4-units). These uxes are absent during growth on fatty acids, which enter central carbon metabolism mainly as acetyl-CoA (C2-units). Instead, a fraction of the TCA cycle intermediate isocitrate is diverted into the glyoxylate shunt, bypassing the oxidative decarboxylation steps in the TCA cycle and replenishing intermediates that are used for biosynthesis of cellular constituents3. Since both pathways are essential under these conditionsthe glyxoylate shunt for anaplerosis, the oxidative TCA cycle for energy and biosynthetic precursorsbalancing the ux ratio at the bifurcation of these pathways is critical4. In enteric bacteria, the glyoxylate shunt is activated by transcriptional induction of the catabolite-repressed genes encoding isocitrate lyase (ICL) and malate synthase (MLS). After such transcriptional activation, the ux ratio between the oxidative TCA cycle and the glyoxylate shunt is controlled by post-translational regulation mediated by reversible phosphorylation57. This regulation is achieved by partial inactivation of isocitrate dehydrogenase (ICD), which competes with ICL for their shared substrate (isocitrate)8. The bifunctional enzyme AceK catalyzes both the phosphorylation and dephosphorylation of ICD to render the enzyme inactive and active, respectively8,9.

In contrast to E. coli, the molecular mechanisms that control partitioning of carbon uxes between the TCA cycle and glyoxylate shunt are not understood in mycobacteria, despite the important role of ux bifurcation in M. tuberculosis pathogenicity10,11. Here, we report that phosphorylation of ICD does not play a role in controlling the bifurcation of isocitrate uxes between the TCA cycle and glyoxylate shunt inM. smegmatis, which encodes only one ICD. In M. tuberculosis and M. bovis BCG, which encode two ICD isoforms12,13, we demonstrate that only ICD2 (homologue of ICD inM. smegmatis) has an important physiological role under the tested growth conditions. In the absence of ICD phosphorylation, we present evidence that ux partitioning between the TCA cycle and the glyoxylate shunt is mediated by cross-activation of ICD by glyoxylate produced by ICL.

ResultsICD2 is exclusively required for oxidative TCA cycle ux.M. tuberculosis and M. bovis BCG encode two distinct isoforms of ICD: ICD1 (409 AA) and ICD2 (745 AA). The M. smegmatis genome encodes a single ICD (743 AA), a homologue of ICD2 inM. tuberculosis and M. bovis BCG. Both isoenzymes (ICD1 and ICD2) are biochemically active in vitro14. To study the relative contributions of the isoenzymes to oxidative TCA cycle uxes in vivo, we generated gene deletion strains. ICD activity is undetectable in ICD2-decient M. smegmatis and ICD2-decientM. bovis BCG but is indistinguishable from wild-type in ICD1-decient M. bovis BCG (Table 1; Supplementary Table 1). An ICD2-decient strain of M. smegmatis cultured in Middlebrook 7H9 medium shows a late growth phenotype that coincides with depletion of glutamate from the culture medium (Fig. 1a).Glutamate auxotrophy of the ICD2-decient strain was conrmed by demonstrating that ICD2 is required for growth on minimal medium lacking glutamate (Fig. 1b; Supplementary Fig. 1). Incubation of ICD2-decient bacteria in minimal medium without glutamate supplementation leads to decreased levels of metabolites downstream of ICD (a-ketoglutarate and glutamate)

(Fig. 1c). In addition, loss of ICD2 increases the levels of metabolites upstream of ICD (citrate/isocitrate) compared with wild-type and complemented bacteria cultivated in media devoid of glutamic acid (Fig. 1c), as expected upon perturbation of a metabolic enzyme15. In M. bovis BCG, deletion of icd2 results in glutamate auxotrophy, whereas deletion of icd1 has no effect (Fig. 1d; Supplementary Table 1), and glutamate prototrophy is restored by complementation of the Dicd2 strain with a plasmid encoding icd2 (Fig. 1e). ICD2-decient M. smegmatis andM. bovis BCG lose viability over time when incubated in medium without glutamate supplementation, suggesting that energy metabolism or production of biosynthetic intermediates through the oxidative TCA cycle is essential for survival under these conditions (Fig. 1f,g).

We conclude that under the conditions tested ICD2 is responsible for carbon ux through the oxidative TCA cycle in both M. smegmatis and M. bovis BCG, whereas the physiological role of ICD1 in M. bovis BCG is currently unclear. Multiple attempts to delete the icd2 gene in M. tuberculosis were unsuccessful, consistent with whole-genome mutagenesis studies suggesting that icd2 might be essential in M. tuberculosis16.

3D structure of MsmICD does not support phospho-regulation.

A structural survey of bacterial ICDs revealed that only ICDs from Gram-negative bacteria are likely to be targets for the bifunctional kinase-phosphatase AceK17. However, by analogy to AceK, the mono-functional serinethreonine protein kinase G (PknG) in mycobacteria has been proposed to regulate uxes

Table 1 | ICD activity in cell-free extracts of bacteria grown on glucose or acetate.

Bacterial strain ICD activity* Glucose (Glc) ICD activity* Acetate (Ace) Ratio (Glc/Ace)M. bovis BCG 0.650.09 1.010.14 0.64 BCG Dicd1 0.630.08 1.120.13 0.56

BCG Dicd2 NDw NDwBCG Dicd2 attB::Pnpicd2z 1.550.06 2.910.83 0.53

M. smegmatis 0.840.24 2.080.13 0.40M. smegmatis Dicd2 0.010.01 0.020.01 0.50M. smegmatis Dicd attB::Pnpicdw 1.020.18 3.190.42 0.32

Data are meanss.d. (nZ3 independent experiments each performed in triplicate).

*ICD specic activity in cell-free extracts (U mg 1). wNot detectable.

zStrain complemented with single-copy pMV306-based plasmids integrated at the chromosomal attB site. In each case, the complementing gene is expressed from the native promoter (P ).

2 NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527 ARTICLE

a b

120

80

% Glutamate

5

4

3

2

1

0

4

3

2

OD 600

OD 600

40

1 WT

icd icd-Comp

0

0

0

10 20 30 40 50 0 48 96

144 192

Time (h)

CIT/ICT (mz 191.021)

Time (h)

c

-KG (mz 145.013)

Glu(mz 146.046)

1104

8103

6103

Concentration (a.u.)

1.5106

1.0106

5.0105

0.0

1.5107

1.0107

5.0106

0.0

4103

2103

icd-Comp

WT

icd

0 WT

icd

icd-Comp

icd-Comp

WT

icd

d e

M.bovis (BCG) M.bovis (BCG)

2.0

1.5

OD 600nm

2.0

1.5

1.0

0.5

WT

icd2

icd1

icd2-Comp

WT

icd2

icd1

1.0

OD 600nm

0.5

icd2-Comp

0 3 6 9 12 15 Time (days)

0 3 6 9 12 15 Time (days)

f g

M. smegmatis

M.bovis (BCG)

Log 10 CFU ml1

1010 1012

1010

108

106

104

108

106

104

102

100

Log 10 CFU ml1

WTicd icd-Comp

WTicd2 icd2-Comp

Time (days)

0 24 48 72 96 120144168

0

5 10 15 20

Time (h)

Figure 1 | Loss of ICD2 results in glutamate auxotrophy and impaired viability. (a) Growth (OD600) of wild-type, Dicd and Dicd attB::Pnpicd-complemented strains of M. smegmatis in Middlebrook 7H9 medium. Solid lines, culture density (OD600). Dashed lines, glutamate concentration in culture medium. (b) Growth (OD600) of wild-type, Dicd and Dicd attB::Pnpicd-complemented strains of M. smegmatis in minimal medium supplemented with glucose and devoid of glutamate.

Data are meanss.d. (n 3 independent experiments). (c) Intracellular metabolites in wild-type, Dicd and Dicd attB::Pnpicd-complemented strains of M. smegmatis

24 h after transferring cells into minimal medium supplemented with glucose and devoid of glutamate. Data are meanss.d. (n 3 independent experiments).

CIT/ICT, citrate/isocitrate; a-KG, alpha-ketoglutarate; GLU, glutamate. (d,e) Growth (OD600) of wild-type, Dicd1, Dicd2 and Dicd2 attB::Pnpicd2-complemented strains of M. bovis BCG in minimal medium supplemented with glucose and glutamate (d) or without glutamate (e). (f) Survival (CFU) of wild-type, Dicd and Dicd attB::Pnpicd-complemented strains of M. smegmatis in minimal medium supplemented with glucose and devoid of glutamate. Data are meanss.d. (n 3

independent experiments each performed in triplicate). (g) Survival (CFU) of wild-type, Dicd2 and Dicd2 attB::Pnpicd2-complemented strains of M. bovis BCG in minimal medium supplemented with glucose and devoid of glutamate. Data are meanss.d. (n 3 independent experiments each performed in triplicate).

NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527

a b

Domain I

Domain II

CgICD

AvlCD

MsmICD

Ser 134

Short helix

Class A ICD: EclCD

c d e

Class B ICD: MsmlCD Overlay: MsmlCD & EcICD

Ser 113

P-loop

ARS elements

Figure 2 | Three-dimensional structure of M. smegmatis ICD (MsmICD). (a) Schematic of two dened domains of MsmICD: a large domain (domain II) and a small domain (domain I). The corresponding P-loop domain is coloured in red, with Serine 134 depicted as a stick. The corresponding ARS domain is coloured in green. The substrate isocitrate and cofactor Mn2 are depicted as sticks in the active site. (b) Overlay of the alpha-helical P-loops of

MsmICD (red, this study), AvICD (yellow, PDB: 1ITW) and CgICD (green, PDB: 2B0T). (c) Class-A ICD of E. coli. Serine 113 extending from the exible P-loop is depicted as a stick. (d) Class-B ICD of M. smegmatis. Unlike EcICD (PDB: 3LCB), which has a random coil P-loop and short ARS domain, MsmICD has a corresponding rigid alpha-helical structure and an elongated ARS domain, which could sterically hinder the association of MsmICD with a regulatory kinase. (e) Overlay of EcICD and MsmICD illustrates the differences between the two proteins in the key regions that mediate interaction of EcICD with AceK.

through the TCA cycle by phosphorylation-mediated inactivation of ICD18,19. We used X-ray crystallography to solve the three-dimensional (3D) structure of M. smegmatis ICD (MsmICD) to evaluate whether it might support a phospho-regulatory mechanism. We determined the structure of MsmICD bound to isocitrate and manganese to a resolution of 2.8 using molecular replacement with ICD of Azotobacter vinelandii (AvICD) as template (Supplementary Table 2). MsmICD, the homologue ofM. tuberculosis ICD2, is similar to the NADP -dependent ICD of Corynebacterium glutamicum (CgICD) with a sequence identity of 59% and an rmsd of 3.9 of all C-alpha atoms (DaliLite v.3). The tertiary structure of MsmICD (PDB: 4ZDA) consists of two distinct domains, similar to the ICD structures of AvICD and CgICD (Fig. 2a). A cleft region between the two domains contains the active site, as described for other ICD enzymes20,21.

The phospho-regulatory P-loop of E. coli ICD (EcICD), comprising amino acid residues 94113, includes a glycine-rich motif forming a highly exible structure (Fig. 2c). In contrast, the equivalent amino acid residues in MsmICD are more structured, forming a short a-helix (amino acid residues 119134) (Fig. 2d)

similar to the short a-helix loop in CgICD and AvICD (Fig. 2b)20,21. Overlaid with EcICD, this short a-helix limits access to the active site and would sterically hinder binding of AceK to the AceK recognition sequence (ARS) (Fig. 2e). In contrast to the EcICD ARS elements, which form a b-sheet

(Fig. 2c), the corresponding structural elements of MsmICD resemble a Class-B ICD (Fig. 2d)17. Structural comparison of representative ICD structures from enterobacteria that encode an AceK homologue reveals remarkable sequence conservation in their ARS elements compared with MsmICD and CgICD (Supplementary Fig. 2). Indeed an AceKEcICD complex structure illustrates a highly specic recognition and intimate interaction between the enzyme and the kinase, similar to several documented cases in eukaryotic protein kinases9. The discontinuous regions on EcICD that interact with AceK are: the P-loop from domain (I) and the twisted antiparallel beta-sheet (ARS) from domain (II). Notably, modication of these recognition sites on EcICD, analogous to the amino acid residue differences in MsmICD, signicantly impairs the ability of AceK to interact with EcICD9,22. Consistent with the genetic and biochemical evidence presented hereafter, the structure of MsmICD argues against a phosphorylation-mediated mechanism for ICD regulation in mycobacteria.

PknG kinase does not regulate ICD activity. To further clarify whether PknG contributes to the regulation of ICD activity, we constructed PknG null strains and measured their growth kinetics and ICD specic activity in carbon-dened media. Although deletion of AceK increases ICD activity by several-fold in E. coli, PknG deciency has no impact on ICD activity in mycobacteria

4 NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527 ARTICLE

Table 2 | ICD activity in cell-free extracts of bacteria grown on glucose or acetate.

Bacterial strain ICD activity* Glucose (Glc) ICD activity* Acetate (Ace) Ratio (Glc/Ace)M. tuberculosis Erdman 0.410.10 0.690.09 0.60M. tuberculosis Erdman DpknG 0.460.14 0.620.03 0.75M. tuberculosis H37Rv 0.420.04 0.630.08 0.66M. tuberculosis CDC1551 0.350.03 0.580.06 0.61M. tuberculosis HN787 0.440.08 0.620.03 0.72M. bovis BCG 0.870.14 1.210.03 0.72M. smegmatis 0.740.19 1.900.24 0.39M. smegmatis DpknG 0.710.21 1.360.06 0.52E. coli 1.250.21 0.350.06 3.58E. coli DaceK 1.090.14 1.180.08 0.93

Data are meanss.d. (nZ3 independent experiments each performed in triplicate).

*ICD specic activity in cell-free extracts (U mg 1).

a

Glucose Acetate

1 WT [afii9797]acek

1 WT [afii9797]acek

OD 600

OD 600

0.1

0.1

0 5

10 15 0 5

10 15

Time (h)

b

Time (h)

Gluc Gluc

Glu-Ace Ace-Gluc Gluc

3 h 6 h 9 h 18 h 24 h

c

M. smegmatis M. tuberculosis

1

1

OD 600

0.1

OD 600

0.1

WT [afii9797]PknG

WT [afii9797]PknG

0.01

0.01

0 20 40 60 80 Time (h)

0 5

10 15

Time (days)

Figure 3 | Phosphorylation-mediated branch-point regulation. (a) Growth (OD600) of wild-type and DaceK strains of E. coli on minimal medium containing glucose (left) or acetate (right). Data are meanss.d. (n 3 independent experiments). (b) Time-lapse series of GFP-expressing E. coli DaceK

cells grown in a microuidic device and imaged on phase-contrast and uorescence channels (merged) at 5-minute intervals. Flow medium wasM9 minimal medium containing glucose (06 h) before switching to acetate (618 h) and then back to glucose (1824 h). Scale bar, 3 mm. (c) Growth (OD600) of wild-type and DpknG strains of M. smegmatis (left) or M. tuberculosis (right) on minimal medium containing acetate. Data are meanss.d.

(n 3 independent experiments).

(Table 2). Furthermore, we found that PknG is dispensable for growth of mycobacteria on acetate (Fig. 3c; Supplementary Fig. 3), whereas AceK is essential for growth of E. coli on acetate

(Fig. 3a,b; Supplementary Movie 1). These observations suggest that PknG is not involved in modulating ICD activity during growth of mycobacteria on acetate. This conclusion is further

NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527

supported by the fact that immunoprecipitated ICD fractions from glucose-grown and acetate-grown cells show indistinguishable specic activity (0.850.21 and 0.870.18 U mg 1, respectively).

Bifurcation of uxes at the TCA cycle branch-point. Bacteria growing on C2-carbon sources use the glyoxylate shunt for anaplerosis of the TCA cycle while maintaining oxidative TCA cycle activity for the production of energy and a-ketoglutarate-related biomass precursors. In the absence of phosphorylation-mediated modulation of ICD activity, we evaluated whether partitioning of uxes between the TCA cycle and the glyoxylate shunt occurs in mycobacteria. We used dynamic 13C-tracing to estimate the ux ratio between the two competing pathways23.

The conversion of isocitrate to succinate is either through the oxidative branch of the TCA cycle or through the glyoxylate shunt. Since both pathways conserve the two carbon atoms of acetyl-CoA, the initial dynamics of the M 2 mass isomer of

succinate (before extensive reshufing of labelled carbons) after shifting cells from unlabelled acetate to [U-13C]-acetate are informative to determine the relative contributions of the two converging metabolic uxes (Fig. 4a,b). Shifting cells from unlabelled glucose to [U-13C]-glucose resulted in similar dynamics of the M 2 isomer fraction of a-ketoglutarate and

succinate over the rst 320 s that were comparable between wild-

type (Fig. 4c) and ICL-decient (Dicl1 Dicl2) (Fig. 4d) strains.

These results demonstrate that uxes through the glyoxylate shunt are negligible in glucose-grown cells. In contrast, shifting acetate-grown cells to [U-13C]-acetate resulted in a twofold faster

a c

C6 carbon

(glucose)

100%

80%

60%

PEP

AcCoA

M+6

M+5

M+4

M+3

M+1

M+2

Fractional labeling Fractional labeling Fractional labeling

40%

20%

0%

C2 carbon

(acetate or lipids)

M

Aspartate

Oxaloacetate

AcCoA

Citrate

Malate

COA ICT AKG SUC MAL

Glyoxylate

d

Fumarate

Succinate

Isocitrate

icd

A-ketoglutarate

C6 carbon

(glucose)

PEP

AcCoA C2 carbon

(acetate or lipids)

100%

80%

60%

40%

20%

0%

100%

80%

60%

40%

20%

0%

icl

M+6

M+5

M+4

M+3

M+1

M+2

CO2

CO2

Glutamate

b

M

COA ICT AKG SUC MAL

e

M+6

M+5

M+4

M+3

M+1

M+2

Aspartate Oxaloacetate

Glyoxylate Fumarate

Succinate

AcCoA

Citrate

Malate

Isocitrate

icl

M

CO2

icd

CO2

A-ketoglutarate

Glutamate

COA ICT AKG SUC MAL

Figure 4 | Bifurcation of metabolic uxes. (a) Schematic of uxes into the TCA cycle and glyoxylate shunt following shift from glucose (blue shading) to radiolabelled acetate (red shading). (b) Schematic of uxes into the TCA cycle and glyoxylate shunt following shift from acetate (red shading) to radiolabelled glucose (blue shading). (c,d) 13C-labelling of metabolites in wild-type (c) and Dicl1 Dicl2 (d) M. smegmatis 320 s after shifting glucose to [U-13C]-glucose. Comparison of the M 2 isomers of isocitrate, a-ketoglutarate and succinate demonstrate absence of carbon ux through the glyoxylate

shunt during growth on glucose. (e) 13C-labelling of metabolites in wild-type M. smegmatis 80 s after shifting from acetate to [U-13C]-acetate illustrate carbon ux bifurcation to the glyoxylate shunt and the oxidative TCA cycle.

6 NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527 ARTICLE

Table 3 | ICL activity in cell-free extracts of bacteria grown on glucose or acetate.

Bacterial strain ICL activity* Glucose (Glc) ICL activity* Acetate (Ace) Ratio (Glc/Ace)M. tuberculosis Erdman 0.460.15 1.260.24 0.37M. bovis BCG 0.360.04 0.840.08 0.43M. smegmatis 0.220.03 1.760.26 0.13E. coli 0.010.01 0.950.68 0.01

Data are meanss.d. (nZ3 independent experiments each performed in triplicate).

*ICL specic activity in cell-free extracts (U mg 1).

M 2 isomer accumulation for succinate compared with a

ketoglutarate during the rst 80 s, which demonstrates substantial ux through the glyoxylate shunt (Fig. 4e). Isotopomer balancing, assuming pseudo-steady-state labelling after 80 s, indicated that 3238% of the succinate molecules are generated through the glyoxylate shunt and 6268% are generated through the oxidative TCA cycle during growth on acetate. This is comparable to the determined 1:2 to 1:3 ratio of uxes between the glyoxylate shunt and the oxidative TCA cycle for acetate-grown E. coli andC. glutamicum, respectively 4,24. Thus, the emerging picture of mycobacterial metabolism is of exclusive utilization of the TCA cycle during growth on glucose and bifurcation of uxes between the TCA cycle and glyoxylate shunt during growth on acetate.

Mycobacteria do not inactivate ICD during growth on acetate. In enterobacteria, the split-ratio of uxes at the bifurcation of the oxidative TCA cycle and glyoxylate shunt is maintained by partial inactivation of ICD, which is reversibly phosphorylated by the bifunctional kinase-phosphatase AceK8. Although phosphorylation-mediated inhibition of ICD does not seem to play a role in mycobacteria (see above), we investigated whether other mechanisms might contribute to ICD regulation by measuring total ICD enzyme activity in cells grown on glucose or acetate as the sole source of carbon and energy. Consistent with previous studies, we found that total ICD activity in E. coli cell extracts is threefold lower in acetate-grown cells compared with glucose-grown cells. In sharp contrast, we found that total ICD activity in mycobacterial cell extracts is increased rather than decreased in acetate-grown cells compared with glucose-grown cells in all of the species that we tested (M. smegmatis, M. bovis BCG and M. tuberculosis) (Tables 1 and 2). These observations conrm that mycobacteria do not use phosphorylation-mediated inhibition of ICD to balance ux partitioning between the TCA cycle and the glyoxylate shunt.

ICD2 expression is independent of the carbon source. Since ICD2 seems to be solely responsible for intracellular ICD activity and carbon ux through the oxidative TCA cycle under standard growth conditions in vitro, we focused on this isoenzyme to investigate the mechanism of ux partitioning between the TCA cycle and the glyoxylate shunt. We found that increased ICD activity in acetate-grown cells (Table 1) cannot be attributed to changes in ICD2 protein abundance, which is similar in cells grown on glucose or acetate (Supplementary Fig. 4). Stable post-translational modications also cannot account for increased ICD activity in acetate-grown cells, given that ICD specic activity is similar in immunoprecipitated fractions from cells grown on glucose. These results suggest that increased ICD activity in acetate-grown cells cannot be explained by changes in the intrinsic activity of ICD2 alone and might involve additional carbon-responsive elements. ICL is a putative candidate because it competes with ICD for a shared substrate (isocitrate) and ICL expression is induced in cells grown on acetate compared with glucose (Table 3; Supplementary Fig. 4a).

Metabolic cross-activation of ICD by ICL. We used a strain ofM. smegmatis with icl1 transcribed from an anhydrotetracycline (ATc)-inducible promoter (Supplementary Fig. 5) to demonstrate that increased expression of ICL1 is accompanied by increased ICD activity in a dose-dependent manner (Fig. 5a). These results suggest that ICL1 cross-activates ICD2 in vivo through a direct or indirect mechanism. We reconstituted this effect in vitro using recombinant enzymes. Consistent with the in vivo data (Fig. 5a), we found that rICD2 activity increases in proportion to the amount of rICL1 added to the reaction mix (Fig. 5b). Cross-activation is dependent on ICL enzymatic activity, because an inactive catalytic-site variant of ICL1 (rICL1KKAGA) has no rICD2-stimulatory activity in vitro (Fig. 5b; Supplementary Fig. 6).

These observations suggest that ICL-dependent stimulation of ICD2 activity might be mediated by metabolic cross-activation rather than direct proteinprotein interaction. Consistent with this hypothesis, we found that rICD2 activity in vitro is stimulated by direct addition of glyoxylate to the reaction mix (Fig. 5c), whereas addition of succinate has no effect (Fig. 5d). Stimulation of ICD activity by glyoxylate is dose-dependent up to a concentration of 200 mM, consistent with the intracellular concentration of glyoxylate that we measured in acetate-grown cells (B20 mM) and about fourfold lower in glucose-grown cells, which is similar to data from E. coli cultures25. Indeed, these measurements are likely to underestimate the true concentrations of intracellular glyoxylate due to the inherent instability and reactivity of this compound. Moreover, it is likely that glyoxylate concentrations uctuate over time and individual cells may transiently contain higher levels than the measured population average2628, consistent with the rheostat-like mechanism of metabolite-mediated feedback regulation proposed here.

Finally, we employed isothermal titration calorimetry (ITC) to measure the binding afnity of glyoxylate for MsmICD (Fig. 5e). Thermodynamic parameters revealed that glyoxylate binds to MsmICD with high afnity and a stoichiometry of 1:1 (Table 4). Glyoxylate contributed favourably to the binding entropy (DS 406 cal mol 1), suggesting the presence of conformational

changes. Importantly, MsmICD binding was not detected for succinate, consistent with the absence of ICD-stimulatory activity in vitro (Fig. 5f). To investigate the mechanism of glyoxylate-mediated activation of ICD, we attempted to co-crystallize apo-MsmICD with glyoxylate but we were unable to resolve glyoxylate in the resulting crystal structures. Therefore, we used the MsmICD crystal structure to perform computational docking studies with glyoxylate to generate models of the MsmICD glyoxylate complex structure. The docking studies suggest that glyoxylate binds close to the active site of MsmICD, making key interactions with amino acid residues around the catalytic site (Fig. 5g; Supplementary Fig. 7).

DiscussionDespite the importance of fatty acid catabolism duringM. tuberculosis infection10, the mechanisms that regulate the partitioning of metabolite uxes at the bifurcation of the

NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527

a b

2.5 20 ICLactive

ICLinactive

15

Specific activity (U mg1 )

ICD

ICL

Specific activity (U mg1 )

2.0

1.5

1.0

0.5

0.0

10

5

0

0.01 0.1 1 10 100 1,000 (rlCL g ml1)

Succinate

+25 ng ml1

+12.5 ng ml1

+0 ng ml1

icl1/2

+50 ng ml1

c d

Glyoxylate

0.090

0.085

0.080

0.075

0.070

0 200 400 600 Time (s)

0 M 50 M 100 M 200 M 400 M

0.085

0.080

0 M

400 M

100 M

1,600 M 6,400 M

Absorbance (340 nm)

Absorbance (340 nm)

0.075

0.070

800 1,000 0 200 400 600 Time (s)

800 1,000

e f

Time (min) 0 10

20.00

15.00

10.00

5.00

1 0.00

5.00

15.00

10.00

20.00

25.00

30.00

120.0 100.080.060.0

20 30 40

Succinate Glyoxylate

kcal mol1 of injectantcal s

kcal mol1 of injectant

120.0

100.0

80.0

60.0

40.0

20.0

0.0

40.020.00.0

1.0 1.5 2.0

0.0 0.5Molar ratio

1.0 1.5 2.0 0.0 0.5

Molar ratio

g

Figure 5 | ICD is activated by glyoxylate but not succinate. (a) ICD activity in vivo increases concomitantly with ICL1 activity in a strain of M. smegmatis that conditionally expresses icl1 from an ATc-inducible promoter. Cells were induced with the indicated concentration of ATc for 18 h before measurement of ICD and ICL1 activities in cell-free extracts. Data are meanss.d. (n 3 independent experiments). (b) rICD activity in vitro is stimulated by active rICL1 in a dose-dependent

manner. A catalytically inactive mutant of rICL1 (rICL1KKAGA) does not stimulate ICD activity. (c) rICD activity in vitro is stimulated by glyoxylate in a dose-dependent manner. (d) rICD activity in vitro is not stimulated by succinate. (e) Competitive ITC trace of a solution of glyoxylate-MsmICD binding in 10 mM Tris (pH 7.5), 10 mM EDTA, 50 mM NaCl2 and 2.5 mM MgCl2. Glyoxylate (150 mM) was titrated into MsmICD (15 mM) in the same buffer. Data are representative of at least three independent experiments. (f) Kinetics of glyoxylate and succinate interaction with MsmICD. Data was tted into a single-site binding equation. (g) Detailed view of putative glyoxylate binding site close to the protein active site. Interacting residues Arg147, Arg550, Arg141, Tyr422, Ala585, Asp352, Asp555, Lys257 and Asp551 are shown in stick representation and the magnesium (Mn) as a yellow sphere. (Insert: detailed view of MsmICDglyoxylate interaction.)

8 NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527 ARTICLE

Table 4 | Thermodynamic parameters of glyoxylate binding to MsmICD.

Compound n K (M 1) DH (cal mol 1) DS (cal mol 1) Glyoxylate 1.170.02 1.10e64.21e5 1.129e54507 406 Succinate 0.1212 3.05e31.34e4 8.00e48.42e6 252

Data are representative of three independent experiments with similar results.

oxidative TCA cycle and the glyoxylate shunt remain unknown. Strict regulation of uxes at this metabolic branch-point is essential during catabolism of C2-units derived from fatty acids in order to balance anaplerotic uxes (glyoxylate shunt) and uxes generating energy and biosynthetic precursors (TCA cycle) (Supplementary Fig. 8). Here, we demonstrate that this balance is mediated by metabolite-based regulation of ICD2 activity, whereas imbalancing of branch-point uxes by overexpression of ICD1 inhibits growth on acetate (Fig. 6a; Supplementary Table 3). This growth-inhibitory effect is presumably caused by directing too much ux into the oxidative TCA cycle at the expense of the glyoxylate shunt, analogous to E. coli mutants decient in phosphorylation-mediated inhibition of ICD29. This interpretation is consistent with limited viable ux combinations of ICL and ICD on acetate compared with glucose, as modelled by computational ux-balance analysis using either of these two substrates as the sole carbon source (Fig. 6b). Loss of ICD activity not only impairs growth on media with acetate as the sole carbon source but also leads to the loss of viability over time. These observations suggest that perturbation of ux partitioning at the branch-point of the TCA cycle and glyoxylate shunt might present a potential strategy for chemotherapy. This could be achieved by modulating the metabolic uxes to favour the TCA cycle over the glyoxylate shunt, which would impair anaplerosis in cells catabolizing fatty acid substrates during infection in vivo30.

The architecture of the regulatory circuit that controls bifurcation of uxes at the branch-point of the TCA cycle and glyoxylate shunt is strikingly different in mycobacteria compared with the established regulatory paradigm in E. coli. In E. coli, branch-point partitioning is controlled by phosphorylation-mediated inactivation of ICD, which diverts ux away from the TCA cycle and into the glyoxylate shunt (Fig. 6c)8. In line with the reported architecture of mycobacterial ICD1 (ref. 31), we nd that the 3D architecture of ICD2 appears incompatible with phosphorylation-mediated control of enzymatic activity. It is perhaps not surprising, therefore, that all known and putative AceK proteins exist only in Gram-negative bacteria. Correspondingly, ICD proteins from the more distantly related Actinobacteria, including mycobacteria, seem to have lost (or perhaps never had) these AceK-interacting elements.

Here, we show that branch-point partitioning in mycobacteria is controlled by metabolite-mediated activation of ICD, which diverts ux away from the glyoxylate shunt and into the TCA cycle. Cross-activation of ICD is mediated by glyoxylate, which is ideally suited for this regulatory role because it is a unique product of the ICL-catalyzed cleavage of isocitrate to glyoxylate and succinate. According to this scheme, if glyoxylate levels rise, the corresponding increase in ICD activity ensures that carbon ux will shift towards the TCA cycle, leading to a decline in glyoxylate levels. Conversely, if glyoxylate levels fall, the corresponding decrease in ICD activity ensures that carbon ux will shift towards the glyoxylate shunt, leading to a recovery of glyoxylate levels. Thus, we propose that glyoxylate-mediated activation of ICD functions as a molecular rheostat to maintain the proper balance of uxes between the TCA cycle and the glyoxylate shunt under steady-state conditions (Fig. 6c). In

computational simulations, a similar partitioning of uxes could be achieved by increasing the activity of ICL above a threshold in E. coli devoid of AceK32, suggesting that ICL can govern ux partitioning in the absence of phosphorylation-mediated inactivation of ICD.

Competitive displacement isothermal titration calorimetry revealed the true potency of glyoxylate, providing a KA of

1.129e5M 1 that is driven by a positive endothermic binding entropy (DS 406 cal mol 1). Computational simulations of

glyoxylate interaction with ICD suggest that the molecule binds to the enzyme at a site close the catalytic site. Characterization of this putative allosteric interaction and localization of the binding site of glyoxylate could provide new opportunities for therapeutic intervention, as ux partitioning between the TCA cycle and glyoxylate shunt is critical for survival of M. tuberculosis during infection. Similar to other NADP -dependent ICDs, mycobacterial ICD2 forms dimers in solution14,33. According to the ultrasensitivity theory34, such multimeric structures can mediate cooperative binding of substrates or allosteric effector molecules, thus enabling rapid and switch-like responses to environmental changes35.

The rheostat-like mechanism proposed here involves metabolite-mediated enzyme activation to achieve a balanced bifurcation of uxes between two pathways, whereas in E. coli the same ux balance is achieved by phosphorylation-mediated enzyme inhibition. In future it will be important to determine whether a rheostat-like mechanism can be generalized to other bacterial species in which ICD activity is not repressed during growth on C2-carbon substrates3639. Although metabolite-mediated control has been implicated in metabolic adaptation during switching between distinct nutritional environments4042, the results presented here also demonstrate its importance for metabolic homoeostasis under conditions of steady-state growth.

Methods

Bacterial strains and culture conditions. E. coli K-12 MG1655 was the reference wild-type strain used in all E. coli experiments. The DaceK null E. coli strain was obtained from the Yale E. coli Genetic Stock Center (CGSC#10859). Yellow uorescent protein (YFP)-expressing derivatives of both E. coli strains were made by transformation with a plasmid encoding YFP43. M. tuberculosis Erdman,M. tuberculosis H37Rv, M. tuberculosis CDC1551, M. tuberculosis HN787, M. bovis BCG (Pasteur 1173P2), M. smegmatis mc2155 and derivative strains (Supplementary Table 4) were grown in Middlebrook 7H9 liquid medium (Difco) supplemented with 0.5% albumin, 0.2% glucose, 0.085% NaCl, 0.5% glycerol and0.05% tyloxapol. Cultures were grown at 37 C to mid-log phase, corresponding to optical density at 600 nm (OD600) of B0.5. Aliquots were stored in 15% glycerol at

80 C and thawed at room temperature before use; individual aliquots were used once and discarded. When required, the medium was supplemented with75 mg ml 1 hygromycin (Roche Diagnostics) or 25 mg ml 1 kanamycin (Sigma-

K4378). E. coli MG1655 and derivative strains (Supplementary Table 4) were grown in LB liquid medium (Difco). When required, the medium was supplemented with 25 mg ml 1 kanamycin, 25 mg ml 1 chloramphenicol (Sigma-

C0378) or 100 mg ml 1 ampicillin (Sigma-A0166).

For growth with single carbon sources, bacteria were cultured in M9 minimal medium (Difco) supplemented with 2 g l 1 of glucose or acetate, or in dened medium containing 2.5 g l 1 Na2HPO4, 1 g l 1 KH2PO4, 0.5 g l 1 glutamic acid, 1 mg l 1 pyridoxine, 0.5 mg l 1 biotin, 15 mg l 1 ferric ammonium citrate,40 g l 1 MgSO4, 0.5 mg l 1 CaCl2, 0.6 mg l 1 ZnSO4, 0.6 mg l 1 CuSO4, 0.8 g l 1

NaCl, 0.5 g l 1 Tyloxapol, and 0.1% fatty acid-free bovine serum albumin (Sigma A8806) supplemented with 2 g l 1 of glucose or acetate. When required, 0.5 g l 1 of NH4SO4 was added as an alternative nitrogen source in lieu of glutamic acid.

NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527

a

Glucose

3

2

1.5

1.0

WT

Acetate

WT +ICD1 WT +ICD1

OD 600

ICD (mmol1 h1 gDW)OD 600

1

WTWT +ICD1 WT +ICD1

6 9 12 15 18 0 3 6 9 12 15 18

0.5

0.0

0 0 3

Time (days) Time (days)

b

1.5

1.5

ICD (mmol1 h1 gDW)

0.035 0.0160.0140.0120.0100.0080.0060.0040.002 0

0.03

0.025

0.02

0.015

0.01

0.005

0

1

1

0.5

0.5

0

0

0

0.5

1

1.5 2

0

0.5

1

1.5 2

ICL (mmol1 h1 gDW)

Control ICD by inhibition Control ICD by activation

ICL (mmol1 h1 gDW)

c

E. coli

Mycobacterium ssp.

ICT

ICT

icl

Km-ICT = 604M

AceK

Pi

icl

Km-ICT = 145M

icd

Km-ICT = 8M

icd

Km-ICT = 20M

GOX 1/3

1/3

GOX

2/3

2/3

-KG

-KG

Figure 6 | Dysregulation of the ICL/ICD rheostat balance impairs growth on acetate. (a) Growth of wild-type, attB::Pnpicd1 and attB::Phspicd1 strains ofM. bovis BCG in glucose-based (left) or acetate-based (right) minimal medium without glutamate. The attB::Pnpicd1 strain expresses icd1 from the native promoter. The attB::Phspicd1 strain overexpresses icd1 from the strong hsp60 promoter. Data are meanss.d. (n 3 independent experiments).

(b) Computationally modelled phase planes of carbon ux between ICD and ICL in glucose-based (left) or acetate-based (right) medium. Heat maps show scales of predicted growth (maroon, best growth; blue, zero growth) at varying levels of ICD/ICL ux. The dynamic range of ICD ux allowing optimal growth at each ICL ux value is narrower in acetate-grown cells compared with glucose-grown cells. (c) Bifurcation of metabolic uxes between the oxidative TCA cycle and the glyoxylate shunt in E. coli and mycobacteria. In E. coli, the substrate Km of ICL is almost 100-fold higher than the substrate Km of ICD. Partial inhibition of ICD by AceK-mediated phosphorylation enables partitioning of one-third of the ux into the glyoxylate shunt with the remaining two-thirds of the ux continuing through the oxidative TCA cycle. In mycobacteria, the substrate Km of ICL1 is only slightly higher than the substrate Km of ICD2. In this context, glyoxylate-mediated activation of ICD functions as a metabolic rheostat to maintain the optimal partitioning of uxes between the TCA cycle and the glyoxylate shunt. ICT, isocitrate; GOX, glyoxylate; a-KG, alpha-ketoglutarate.

Bacterial growth was monitored by measuring culture OD600 (BioMate5 Fisher Scientic) at successive time points. All data for growth kinetics were derived from at least three independent experiments.

Viability of M. smegmatis and M. bovis BCG cultures was determined by plating serially diluted aliquots of cultures on LB solid medium for M. smegmatis or Middlebrook 7H9 solid medium supplemented with 0.5% glycerol and 10% oleic acidalbumindextrosecatalase (0.5 g l 1 oleic acid, 50 g l 1 bovine serum albumin fraction V, 20 g l 1 glucose, 40 mg l 1 catalase, 8.5 g l 1 NaCl) for

M. bovis BCG. Colony-forming units (CFU) were scored after incubation of plates at 37 C.

Strain construction. Plasmids and bacterial strains are listed in Supplementary Table 4. All vectors used for mutant generation, complementation and conditional gene expression were constructed using TOPO Cloning Technology (Invitrogen). Unmarked in-frame gene deletions of M. tuberculosis icd1 (rv3339c) and

10 NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527 ARTICLE

M. smegmatis icd (msm1654), pknG (msm0786), icl1 (msm0911), and icl2 (msm3706) were constructed using a method for two-step homologous recombination44. The in-frame recombination template was constructed by PCR amplication of E900 bp of the upstream and downstream region of the gene of interest. Primers were designed to introduce 50 PacI and 30 AvrII sites into the upstream amplicon including nine bases from the start codon, and 50 AvrII and 30 AscI into the downstream region including nine bases before the stop codon. The amplicons were veried by DNA sequencing then restriction digested and ligated into the PacI and AscI sites of pJG1100, a suicide vector that contains aph (kanamycin resistance), hyg (hygromycin resistance), and sacB (sucrose sensitivity) markers45. When inapplicable due to presence of internal restriction sites in the amplicons, NheI and SpeI were used instead to generate homologous arms.

Disruptions of the M. bovis BCG icd1 (Mb3371c) and icd2 (Mb0067c) genes were constructed by replacing the chromosomal locus with an interrupted (100 bp) in-frame icd1::hyg or icd2::hyg allele using a method for one-step homologous recombination46. Briey, fragments ofE1 kb anking the icd1 or icd2 open reading frame (ORF) were PCR amplied from M. tuberculosis H37Rv genomic DNA. The 50- and 30- anking fragments were then cloned into pYUB854 anking the hygromycin-resistance gene. A sacB-lacZ cassette excised from pGOAL17 (ref. 47) was then cloned into the unique PacI site of pYUB854. The nal plasmids were ultraviolet-irradiated before electroporation into M. bovis BCG48. Positive clones of knockout mutants were selected as white colonies on 7H11 agar supplemented with 150 mg ml 1 hygromycin and 40 mg ml 1 X-Gal. Replacement of the target gene with the hygromycin-resistance marker was conrmed by PCR. Primers are listed in Supplementary Table 5.

Epitope-tagged translational fusions of M. tuberculosis icd1 (rv3339c),M. tuberculosis icd2 (rv0066c), M. tuberculosis icl1 (rv0467), M. tuberculosis icl2 (rv1916) and M. smegmatis icd (msm1654) were constructed by inserting sequences encoding the epitopes VSVG, HA, FLAG, Myc or HA into the chromosome downstream of rv3339c, rv0066c, rv0467, rv1916 or msm1654 (respectively), just before the stop codon by two-step recombination with the suicide vector pJG1100. In all cases, the in-frame upstream recombination template was constructed by PCR amplication of E900 bp of the 30 region of the gene of interest. Primers were designed analogous to that described for the unmarked gene deletion. The epitope tags were amplied with AvrII overhangs and were cloned into the AvrII site at the junction of the two homologous arms. The resultant plasmids were introduced intoM. smegmatis and M. tuberculosis by electroporation. Transformants (rst crossover) were selected on LB agar or 7H10 agar containing 50 mg ml 1 hygromycin and 25 mg ml 1 kanamycin. Individual colonies were picked and amplied in 7H9 liquid medium (no antibiotics) and then plated on LB agar or 7H10 containing 5% sucrose to select for cells in which plasmid excision had occurred (second crossover). Individual colonies were picked and the replacement of the gene by the fusion gene or gene deletion was conrmed by PCR. For M. tuberculosis, cells were plated on 7H10 agar in lieu of LB agar.

For complementation of the M. bovis BCG Dicd1, M. bovis BCG Dicd2 orM. smegmatis Dicd mutant, the corresponding ORF plus 200 bp of 50-anking sequence upstream of the translational start codon was PCR amplied and ligated into the attB-integrating promoter-less vector pMV306 digested with XbaI and HindIII to generate pMV306::Pnpicd1, pMV306::Pnpicd2, or pMV306::Pnpicd, respectively. Plasmids overexpressing the M. bovis BCG icd1 or icd2 gene were generated by PCR amplifying the respective ORFs and ligating into the BamHI and HindIII sites of pMV361 (a pMV306 derivative) downstream of the strong hsp60 promoter to generate pMV306::Phspicd1 and pMV306::Phspicd2. Plasmids were introduced into M. bovis BCG by electroporation and colonies were selected on 7H11 agar supplemented with 25 mg ml 1 kanamycin. Complemented strains were veried by PCR.

Growth kinetics. Pre-cultures were inoculated from frozen stocks to an initial optical density at 600 nm (OD600)E0.05. Cells were grown to early exponential phase (OD600E0.5), collected by centrifugation (3,000g, 5 min), washed three times with phosphate-buffered saline (PBS) supplemented with 0.05% Tween-80 (Sigma), resuspended in fresh medium at OD600E0.05, and incubated at 37 C with aeration (180 r.p.m.). At specied time points, OD600 measurements were made using an Ultrospec 2000 spectrophotometer (Pharmacia). Where indicated, resuspended cells were seeded in microtiter plates, sealed with a gas-permeable adhesive seal (ThermoScientic, Waltham, USA), and incubated in a BioLector system49,50 for continuous monitoring of bacterial growth (OD620) and uorescence.

Time-lapse microscopy. Time-lapse microscopy of E. coli MG1655 (wild-type) and DaceK strains expressing YFP (Supplementary Table 4) was performed with minor modications51. In brief, bacteria were grown to mid-log phase (OD600 B0.3) in M9 liquid medium containing 2 g l 1 glucose and 2 ml of culture were spread on a semipermeable membrane, which was then inverted onto a glass coverslip micro-patterned with rings of SU-8 (65 mm diameter 1.5 mm height)

containing a central post (10 mm diameter 1.5 mm height). The bacteria-free side

of the membrane was overlaid with a PDMS-based microuidic device and the assembly was clamped between a transparent acrylic cover and base adapter, as described51. The inlet and outlet ports of the microuidic device were connected to plastic tubing and bacteria were fed by continuous ow of M9 minimal medium containing 2 g l 1 glucose or acetate (ow rate, 25 ml min 1). Bacteria were

maintained at 37 C in a microscope-tted environmental chamber and imaged using a DeltaVision personalDV microscope (Applied Precision) equipped with a 100 oil-immersion objective. Images were recorded at 5-min intervals on phase-

contrast and uorescence channels (GFP, excitation lter 490/20/, emission lter 528/38) using a CoolSnap HQ2 camera. Images were processed using Softworx software (Applied Precision) and custom-made macros in ImageJ.

Enzyme assays. ICL activity was measured using the following assay conditions: 25 mM imidazole buffer (pH 6.8), 10 mM EDTA, 5 mM MgCl2, 4 mM phenylhydrazine and 4 mM DL-isocitric acid trisodium salt hydrate. Enzyme assays were carried out in 96-well plates. Each well contained 100 ml of reaction master mix and 100 ml of puried recombinant protein (0.12 mg ml 1). Absorbance readings were taken at 324 nm using a temperature-controlled (25 C) 96-well plate reader spectrophotometer (Tecan M200 Pro). ICL specic activity is dened as1 U 1 mM phenylhydrazone glyoxylate generated per min (mg protein) 1.

ICD activity was measured by monitoring the time-dependent reduction of NADP to NADPH14 using the following assay conditions: 25 mM TrisHCl pH 7.5, 5 mM MgCl2, 100 mM NaCl, 8 mM NADP , and 8 mM DL-isocitric acid trisodium salt hydrate (Sigma). Enzyme assays were carried out in 96-well plates.

Each well contained 100 ml of reaction master mix and 100 ml of puried recombinant protein (0.12 mg ml 1). Absorbance readings were taken at 340 nm using a temperature-controlled (25 C) 96-well plate reader spectrophotometer (Tecan M200 Pro). ICD specic activity is dened as 1 U 1 mM of NADPH

formed per min (mg protein) 1. ICD activity with varying glyoxylate concentrations was followed at 37 C by ash-freezing the reactions at different times in 200 ml of acetonitrile. Samples were then prepared for mass spectrometry analysis to track the production of a-ketoglutarate. Stimulation of ICD activity by ICL was measured by following NADPH production at 340 nm in the presence of increasing amounts of recombinant wild-type ICL or mutated (catalytically inactive) ICL.

Heterologous protein expression. ORFs encoding ICL and ICD were amplied by PCR. Amplicons were veried by DNA sequencing then restriction digested and ligated into a pET-28a( ) vector (Novagen USA) carrying a C-terminal 6-His tag

and a TEV (tobacco etch virus) cleavage site preceding the C-terminal 6-His tag. The catalytically inactive ICL1KKAGA was synthesized by GenScript, Inc. (Piscataway, NJ, USA) and similarly cloned into pET-28a( ) (Novagen USA) incor

porating the C-terminal 6-His tag and TEV cleavage site. The nal plasmids, designated pPM010 and pPM011 (catalytically active and catalytically, respectively), were transformed into chemically competent E. coli One Shot BL21 (DE3) (Invitrogen). Transformed bacteria were grown at 37 C in LB liquid medium to early log phase (OD600 B0.3). Recombinant protein expression was induced by addition of 0.2 mM IPTG (isopropyl-b-D-thiogalactopyranoside) and bacteria were cultured at 16 C for 24 h at 180 r.p.m. Cells were harvested by centrifugation (3,000g, 5 min), resuspended in lysis buffer (50 mM Tris pH 7.5, 500 mM NaCl,1 mM imidazole, 10% glycerol, 1% Tween-20) containing DNase and mini-protease inhibitor cocktail (Roche), and incubated for 1 h at 4 C. Cells were then disrupted by 20 cycles of sonication each comprising 15 s of power at 60% amplitude (Vibracell 75186, Bioblock Scientic) and 30 s of power off. Cell debris was pelleted by centrifugation at 30,000g for 30 min at 4 C. The claried supernatant was incubated with PrepEase resin (His-tagged protein purication resin-high speci-city) for 12 h at 4 C with gentle agitation (50 r.p.m.) and washed with 50 volumes of wash buffer (50 mM Tris pH 7.5, 500 mM NaCl, 5 mM imidazole). Proteins were eluted with 10 volumes of elution buffer (50 mM TrisHCl pH 7.5, 500 mM NaCl, 250 mM imidazole). The purity of the eluent fractions was checked by SDSPAGE and the 6-His tag was removed by cleavage with TEV (Promega). The native protein was puried by size-exclusion chromatography on Superdex 75 pg 10/300 GL columns (GE Healthcare). Fractions containing the eluted protein of interest were analyzed by SDSPAGE to conrm purity (498%, estimated by Coomassie

Blue staining), concentrated using Vivaspin units (Sartiorius), and stored at

80 C in buffer containing 50 mM TrisHCl pH 7.5.

Protein extraction and quantication. Cells were collected by centrifugation (3,500g, 15 min), and cell pellets were washed with Tris-buffered saline 0.02%

tyloxapol, followed by a single wash with Tris-buffered saline alone at 4 C. Washed cell pellets were resuspended in 50 mM TrisHCl, pH 8.0, 10% glycerol, 1 protease inhibitor (Roche Applied Science), 1 phosphatase inhibitor

(Pierce) and lysed at 4 C in 15 s sonication cycles (Vibracell 75186, Bioblock Scientic) at amplitude 60% with 30 s intervals in-between cycles for 5 min. Finally, the cell debris was pelleted at 11,200g, 35 min, 4 C and the protein concentration was estimated with a BCA protein quantication kit (ThermoScientic) with BSA standards prepared in the same buffer.

Immunoblot analysis. Protein lysates were prepared in the same manner as described above. Between 50 and 100 mg total protein were separated on NuPAGE Novex 412% bis-Tris gels (Invitrogen) and transferred to nitrocellulose membranes (Bio-Rad, 162-0115, Hercules, CA, USA) for probing using the iBlot system (Invitrogen). Membranes were incubated in TBS blocking buffer (25 mM TrisHCl pH 7.5, 150 mM NaCl, 0.05% Tween-20) with 5% w/v skim milk powder for 2 h at

NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527

room temperature before incubation with primary antibodies (anti-FLAG (Clone M2, Sigma); anti-HA (Sigma); anti-Myc (Sigma); anti-VSVG (Clone 16B12, Sigma); anti-GroEL2, and anti-Ag85) diluted in TBS with 1% BSA overnight. All antibodies were used at a dilution of 1:1,000 except anti-GroEL2 (1:50,00) and anti-Ag85 (1:3,000). Membranes were washed ve times in TBS (15 min each at room temperature) before incubating with appropriate secondary antibody for 1 h at room temperature. Membranes were developed using the colorimetric horseradish peroxidase substrate 4-chloro-1-naphthol (4CN) (Opit-4CN Kit by Bio-rad). Antibodies against GroEL2 and Ag85 were generous gifts from Dr J.M. Chen as part of the National Institutes of Health, National Institute of Allergy and Infectious Diseases, contract HHSN266200400091c entitled Tuberculosis Vaccine Testing and Research Materials, awarded to Colorado State University).

X-ray crystallography. Crystals of M. smegmatis ICD were obtained by vapour diffusion at 18 C by equilibrating 2 or 4 ml hanging drops containing a 1:1 mixture of 60 mg ml 1 protein pre-incubated with 10 mM glyoxylate, 10 mM MnCl2,10 mM isocitrate, and 10 mM NADP in crystallization buffer containing 0.1 M

Hepes pH 7.5, 0.05 M MgCl2, and 2834% polyethylene glycol monomethyl ether 550 (Qiagen) over a 500 ml reservoir of the same crystallization solution. Crystals were stabilized by soaking briey in a cryoprotectant solution (25% w/v glycerol in crystallization buffer) and ash frozen in liquid nitrogen before data collection. Diffraction data to 2.8 (ICD incubated with glyoxylate, isocitrate, MgCl2 and

NADP) were collected on X06DA of the Swiss Light Source (SLS, PSI, Villigen, Switzerland). Data were indexed, integrated and scaled with XDS52. Phase determinations were carried out by molecular replacement using Phaser47 of the CCP4 Suite, using the published structure of AvICD (PDB: 1ITW) as a template. Structure gures were prepared with PyMOL (Molecular Graphics System, Version1.5.0.1 Schrdinger, LLC). Crystallographic statistics are listed in Supplementary Table 1. Coordinates and structure factors have been deposited in the Protein Data Bank (PDB accession code PDB: 4ZDA).

Molecular docking. Protein preparation and grid generation. The raw PDB structure of the ICD was converted into an all atom receptor model, fully prepared for docking studies, using Protein Preparation Wizard in Maestro (Maestro v.9.8). Default parameters in the preparation process were used. The protein structural integrity was checked and missing side chains were built. The protonation and tautomeric states of the residues were adjusted to match a pH of 7.4. Energy minimization of the whole structure was performed using an OPLS2005 force eld. The docking receptor grid was created using Glide Receptor Grid Generation (Glide v.9.8). The grid box was centered between the two domains of ICD, in a region that allowed the grid box to include the whole protein structure. All the possible binding sites were explored.

Ligand preparation. The initial ligand le (sdf le, two-dimensional structure) was obtained from the Protein Data Bank (PDB). The glyoxylate was prepared for ligand docking using LigPrep from the Schrodinger suite. Default parameters in the ligand preparation process were used. A single low energy conformer with proper bond lengths and angles was generated from the two-dimensional structure. The protonation state was adjusted to match a pH of 7.4, tautomeric states were explored, and a nal minimization step was performed. Glyoxylate was then docked inside the rigid 3D structure of the protein receptor (ICD), previously prepared. Glyoxylate was docked in the binding pocket and no other possible poses were found. An extra precision-docking revealed a binding energy of

6 kcal mol 1.

Isothermal titration calorimetry (ITC). ITC experiments were conducted on an automated micro-calorimeter. The experiments were performed at 25 C in ITC buffer (10 mM Tris (pH 7.5), 10 mM EDTA, 50 mM NaCl2, 2.5 mM MgCl2).

Protein concentrations were 15 mM for MsmICD, 150 mM glyoxylate (determined by weighing samples on an ultra-microbalance (Mettler Toledo, accurate to0.001 mg). In titration experiments, glyoxylate was injected into a solution of MsmICD. In the competitive titration experiment, glyoxylate was injected into a solution of the enzyme. Titrations were carried out with a stirring speed of750 r.p.m. and 300-s intervals between 4 ml injections. The rst four injections were excluded from data tting because effective mixing in the cell took longer than the injection intervals. Titrations were run past the point of enzyme saturation to correct for heats of dilution. The experimental data were t to a theoretical titration curve using the Origin software package (version 7.0) provided with the instrument to afford values of KA (the association constant in M 1), n (stoichiometry of binding), and DH (the binding enthalpy change in Kcal mol 1). ITC experiments were performed in duplicate, analyzed independently, and the obtained thermodynamic values were averaged.

Metabolite measurements. Bacteria were grown in Middlebrook 7H9 liquid medium at 37 C with aeration (180 r.p.m.) to mid-exponential phase, harvested by centrifugation (3,000g, 5 min) washed three times with PBS containing 0.05% Tween-80 (Sigma), inoculated into Middlebrook 7H9 liquid medium supplemented with 2 g l 1 glucose or acetate (initial OD600 0.05), and grown at 37 C to mid-log phase (OD600 B0.5). Samples equivalent to a biomass of 1 ml culture at OD600 2 were collected for metabolome analysis by fast ltration and metabolism was quenched at 20 C in a mixture of acetonitrile:methanol:water (2:2:1), as

described53. Metabolites were measured by LCMS/MS and direct injection, as described25,54. Glyoxylate was measured at 12.63.3 pmol ml 1 of acetate-grown culture at OD600 1, which corresponds to a biomass of 343.417.5 mg. To estimate the intracellular concentration of glyoxylate in M. smegmatis we used a ratio of aqueous volume to cellular dry weight of 0.0023 l g 1, as determined for E. coli55.

Data availability. Coordinates and structure factors have been deposited in the Protein Data Bank (PDB accession code PDB: 4ZDA). All additional data supporting the ndings presented here are available from the corresponding authors upon request.

References

1. Boshoff, H. I. & Barry, C. E. A low-carb diet for a high-octane pathogen. Nat. Med. 11, 599600 (2005).

2. Russell, D. G. et al. Mycobacterium tuberculosis wears what it eats. Cell Host Microbe. 8, 6876 (2010).

3. Kornberg, H. L. Anaplerotic sequences in microbial metabolism. Angew. Chem. Int. Ed. Engl. 4, 558565 (1965).

4. Walsh, K. & Koshland, D. E. Determination of ux through the branch point of two metabolic cycles. The tricarboxylic acid cycle and the glyoxylate shunt.J. Biol. Chem. 259, 96469654 (1984).5. da Silva Cruz, A. H. et al. Phosphorylation is the major mechanism regulating isocitrate lyase activity in Paracoccidioides brasiliensis yeast cells. FEBS J. 278, 23182332 (2011).

6. Hoyt, J. C. & Reeves, H. C. In vivo phosphorylation of isocitrate lyase from Escherichia coli D5H3G7. Biochem. Biophys. Res. Commun. 153, 875880 (1988).

7. Robertson, E. F. & Reeves, H. C. Phosphorylation of isocitrate lyase in Escherichia coli. Biochimie 71, 10651070 (1989).

8. Garnak, M. & Reeves, H. C. Phosphorylation of isocitrate dehydrogenase of Escherichia coli. Science 203, 11111112 (1979).

9. Zheng, J. & Jia, Z. Structure of the bifunctional isocitrate dehydrogenase kinase/ phosphatase. Nature 465, 961965 (2010).

10. McKinney, J. D. et al. Persistence of Mycobacterium tuberculosis in macrophages and mice requires the glyoxylate shunt enzyme isocitrate lyase. Nature 406, 735738 (2000).

11. Munoz-Elas, E. J. & McKinney, J. D. Mycobacterium tuberculosis isocitrate lyases 1 and 2 are jointly required for in vivo growth and virulence. Nat. Med. 11, 638644 (2005).

12. Cole, S. T. et al. Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence. Nature 393, 537544 (1998).

13. Garnier, T. et al. The complete genome sequence of Mycobacterium bovis. Proc. Natl Acad. Sci. USA 100, 78777882 (2003).

14. Banerjee, S., Nandyala, A., Podili, R., Katoch, V. M. & Hasnain, S. E. Comparison of Mycobacterium tuberculosis isocitrate dehydrogenases (ICD-1 and ICD-2) reveals differences in coenzyme afnity, oligomeric state, pH tolerance and phylogenetic afliation. BMC Biochem. 6, 20 (2005).

15. Fendt, S.-M. et al. Tradeoff between enzyme and metabolite efciency maintains metabolic homeostasis upon perturbations in enzyme capacity. Mol. Syst. Biol. 6, 356 (2010).

16. Zhang, Y. J. et al. Global assessment of genomic regions required for growth in Mycobacterium tuberculosis. PLoS Pathog. 8, e1002946 (2012).

17. Yates, S. P. et al. Structural Basis of the substrate specicity of bifunctional isocitrate dehydrogenase kinase/phosphatase. Biochemistry 50, 81038106 (2011).

18. Balganesh, T., Datta, S. & Ghosh, I. WIPO Patent Application WO/2004/ 087943 A1. 129 (2004).

19. Singh, V. K. & Ghosh, I. Kinetic modeling of tricarboxylic acid cycle and glyoxylate bypass in Mycobacterium tuberculosis, and its application to assessment of drug targets. Theor. Biol. Med. Model. 3, 27 (2006).

20. Imabayashi, F., Aich, S., Prasad, L. & Delbaere, L. T. J. Substrate-free structure of a monomeric NADP isocitrate dehydrogenase: an open conformation phylogenetic relationship of isocitrate dehydrogenase. Proteins 63, 100112 (2006).

21. Yasutake, Y. et al. Structure of the monomeric isocitrate dehydrogenase. Structure 10, 16371648 (2002).

22. Zheng, J., Yates, S. P. & Jia, Z. Structural and mechanistic insights into the bifunctional enzyme isocitrate dehydrogenase kinase/phosphatase AceK. Philos. Trans. R. Soc. Lond. B Biol. Sci. 367, 26562668 (2012).

23. Yuan, J., Bennett, B. D. & Rabinowitz, J. D. Kinetic ux proling for quantitation of cellular metabolic uxes. Nat. Protoc. 3, 13281340 (2008).

24. Wendisch, V. F., de Graaf, A. A., Sahm, H. & Eikmanns, B. J. Quantitative determination of metabolic uxes during coutilization of two carbon sources: comparative analyses with Corynebacterium glutamicum during growth on acetate and/or glucose. J. Bacteriol. 182, 30883096 (2000).

25. Zimmermann, M., Sauer, U. & Zamboni, N. Quantication and mass isotopomer proling of a-keto acids in central carbon metabolism. Anal. Chem.

86, 32323237 (2014).26. Kiviet, D. J. et al. Stochasticity of metabolism and growth at the single-cell level. Nature 514, 376379 (2014).

12 NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12527 ARTICLE

27. Labhsetwar, P., Cole, J. A., Roberts, E., Price, N. D. & Luthey-Schulten, Z. A. Heterogeneity in protein expression induces metabolic variability in a modeled Escherichia coli population. Proc. Natl Acad. Sci. USA 110, 1400614011 (2013).

28. Cai, L., Friedman, N. & Xie, X. S. Stochastic protein expression in individual cells at the single molecule level. Nature 440, 358362 (2006).29. LaPorte, D. C., Thorsness, P. E. & Koshland, D. E. Compensatory phosphorylation of isocitrate dehydrogenase. A mechanism for adaptation to the intracellular environment. J. Biol. Chem. 260, 1056310568 (1985).

30. Murima, P., McKinney, J. D. & Pethe, K. Targeting bacterial central metabolism for drug development. Chem. Biol. 21, 14231432 (2014).

31. Quartararo, C. E., Hazra, S., Hadi, T. & Blanchard, J. S. Structure, kinetic, and chemical mechanism of isocitrate dehydrogenase-1 from Mycobacterium tuberculosis. Biochemistry 52, 17651775 (2013).

32. el-Mansi, E. M., Dawson, G. C. & Bryce, C. F. Steady-state modelling of metabolic ux between the tricarboxylic acid cycle and the glyoxylate bypass in Escherichia coli. Comput. Appl. Biosci. 10, 295299 (1994).

33. Chen, R. D. & Gadal, P. Structure, functions and regulation of NAD and NADP dependent isocitrate dehydrogenases in higher plants and in other organisms. Plant Physiol. Biochem. (Paris) 28, 411427 (1990).

34. Koshland, D. E. Switches, thresholds and ultrasensitivity. Trends Biochem. Sci. 12, 225229 (1987).

35. Kern, D. & Zuiderweg, E. R. P. The role of dynamics in allosteric regulation. Curr. Opin. Struct. Biol. 13, 748757 (2003).

36. Reeves, H. C., ONeil, S. & Weitzman, P. D. Modulation of isocitrate dehydrogenase activity in Acinetobacter calcoaceticus by acetate. FEBS Lett. 163, 265268 (1983).

37. Ishii, A. et al. Isozymes of isocitrate dehydrogenase from an obligately psychrophilic bacterium, Vibrio sp. strain ABE-1: purication, and modulation of activities by growth conditions. J. Biochem. 102, 14891498 (1987).

38. Ishii, A. et al. Genes encoding two isocitrate dehydrogenase isozymes of a psychrophilic bacterium, Vibrio sp. strain ABE-1. J. Bacteriol. 175, 68736880 (1993).

39. Eikmanns, B. J., Rittmann, D. & Sahm, H. Cloning, sequence analysis, expression, and inactivation of the Corynebacterium glutamicum icd gene encoding isocitrate dehydrogenase and biochemical characterization of the enzyme. J. Bacteriol. 177, 774782 (1995).

40. Chubukov, V., Gerosa, L., Kochanowski, K. & Sauer, U. Coordination of microbial metabolism. Nat. Rev. Microbiol. 12, 327340 (2014).

41. Link, H., Kochanowski, K. & Sauer, U. Systematic identication of allosteric protein-metabolite interactions that control enzyme activity in vivo. Nat. Biotechnol. 31, 357361 (2013).

42. Xu, Y.-F., Amador-Noguez, D., Reaves, M. L., Feng, X.-J. & Rabinowitz, J. D. Ultrasensitive regulation of anapleurosis via allosteric activation of PEP carboxylase. Nat. Chem. Biol. 8, 562568 (2012).

43. Dhar, N. et al. Rapid cytolysis of Mycobacterium tuberculosis by faropenem, an orally bioavailable b-lactam antibiotic. Antimicrob. Agents. Chemother. 59, 13081319 (2015).

44. Pavelka, Jr M. S. & Jacobs, Jr W. R. Comparison of the construction of unmarked deletion mutations in Mycobacterium smegmatis, Mycobacterium bovis bacillus Calmette-Guerin, and Mycobacterium tuberculosis H37Rv by allelic exchange. J. Bacteriol. 181, 47804789 (1999).

45. Parish, T. & Stoker, N. G. Use of a exible cassette method to generate a double unmarked Mycobacterium tuberculosis tlyA plcABC mutant by gene replacement. Microbiology 146, 19691975 (2000).

46. Phong, W. Y. et al. Characterization of phosphofructokinase activity in Mycobacterium tuberculosis reveals that a functional glycolytic carbon ow is necessary to limit the accumulation of toxic metabolic intermediates under hypoxia. PLoS ONE 8, e56037 (2013).

47. McCoy, A. J. et al. Phasercrystallographic software. J. Appl. Crystallogr. 40, 658674 (2007).

48. Hinds, J. et al. Enhanced gene replacement in mycobacteria. Microbiology 145,

519527 (1999).

49. Samorski, M., Mller-Newen, G. & Bchs, J. Quasi-continuous combined scattered light and uorescence measurements: A novel measurement technique for shaken microtiter plates. Biotechnol. Bioeng. 92, 6168 (2005).

50. Kensy, F., Zang, E., Faulhammer, C., Tan, R.-K. & Bchs, J. Validation of a high-throughput fermentation system based on online monitoring of biomass and uorescence in continuously shaken microtiter plates. Microb. Cell Fact. 8, 31 (2009).

51. Wakamoto, Y. et al. Dynamic persistence of antibiotic-stressed mycobacteria. Science 339, 9195 (2013).

52. Kabsch, W. XDS. Acta Crystallogr. D Biol. Crystallogr. 66, 125132 (2010).53. Zimmermann, M., Thormann, V., Sauer, U. & Zamboni, N. Nontargeted proling of coenzyme A thioesters in biological samples by tandem mass spectrometry. Anal. Chem. 85, 82848290 (2013).

54. Fuhrer, T., Heer, D., Begemann, B. & Zamboni, N. High-throughput, accurate mass metabolome proling of cellular extracts by ow injection-time-of-ight mass spectrometry. Anal. Chem. 83, 70747080 (2011).

55. Bennett, B. D., Yuan, J., Kimball, E. H. & Rabinowitz, J. D. Absolute quantitation of intracellular metabolite concentrations by an isotope ratio-based approach. Nat. Protoc. 3, 12991311 (2008).

Acknowledgements

We thank E.zdemir for assistance with the phase planes of metabolic uxes derived from a genome-scale metabolic model of M. tuberculosis. Kevin Pethe is supported by a Lee Kong Chian School of Medicine, Nanyang Technological University Start-Up Grant. P.M. was supported by a fellowship from the National University of Singapore Graduate School of Integrative Science and Engineering and the Novartis Institute of Tropical Diseases. This work was funded in part by a grant to U.S. from the SysteMTb (http://www.systemtb.eu

Web End =www.systemtb.eu) Collaborative Project (Project Reference No. 241587), which is funded by the European Commission Framework Programme 7, and by a grant to J.D.M. from the Swiss National Science Foundation (No. 310030_156945).

Author contributions

P.M., K.P. and J.D.M. conceived the project. P.M., M.Z., T.C., U.S., S.A., K.P. and J.D.M. contributed to experimental design and data analysis. P.M., T.C. and M.Z. carried out the experiments. F.P. solved and interpreted the three-dimensional structure of MsmICD. G.F. and M.D.P. conducted the protein 3D modelling. P.M., M.Z., U.S. and J.D.M. wrote the manuscript. All authors discussed the results and commented on the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no conict of interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Murima, P. et al. A rheostat mechanism governs the bifurcation of carbon ux in mycobacteria. Nat. Commun. 7:12527 doi: 10.1038/ncomms12527 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2016

NATURE COMMUNICATIONS | 7:12527 | DOI: 10.1038/ncomms12527 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 13