1. Identification of potential antimicrobial targets using methods of network systems biology
Eivind Almaas
Dept. of Biotehnology & Food Science
Norwegian University of Science &Technology

2. Drug Action
Resistance:
Increased expression of enzyme substitute
Mutation in the enzyme active site
Mutations to increase expression of efflux proteins to transport drug out of cell
Antibiotics are chemical agents that kill or slow down the growth of micro-organisms

3. How can we make it difficult for an organism to circumvent
our antibiotics?

4. Two possible approaches from Network Systems Biology:
Focus on proteins’ active sites
Focus on vulnerabilities of metabolic network structure

5. Protein Active Site
Specific sites in proteins responsible for their function
Each active site has a specific shape, so it will only perform a specific job
Focus on active site  substrate-centric view
Joining things together
Ripping things apart 5

6. Protein Active Site Network (ASN)
Hypotheses:
Similar active sites structures  similar function and role
ASN neighbors of a drug target:
Could be cause of potential drug side effects
Could be alternative drug targets
Suggest existing evolutionary alternatives to target drug
Approach not based on pre-conceived notion of protein similarity

7. Active-Site Network: approach
For simplicity, focus only on metalloproteins
Define active site as small region centered on metal.
Include all atoms within region, regardless of residue origin.
Use all Protein Data Bank (PDB) metalloproteins containing (~10k):
Co, Cu, Fe, Mn, Mo, Ni, V, W, Zn
1. Identify active site (r~5Å) of all metalloproteins with PDB structure
2. Determine optimal alignment for every structure pair
3. Extract statistically significant pairs and generate network.
Selected
target
Likely targets
Possible targets
Approx. 1/3 of all proteins are metalloproteins
Metal ions are critical to the protein’s function, structure & stability
Numerous metalloproteins important therapeutic targets, not only for antimicrobial drugs:
Example -- Methionine aminopeptidase (MetAP) enzymes

8. Example: ASN identifies proteins with similar function
Structure of 2axt
2axt: Function found at the Oxygen Evolving Complex (OEC) in photosystem II, utilizing 4 Mn atoms at active site.
Loll et. al, Nature 438;1040.
Comparing 2axt to a limited set of all known bi-manganese proteins in PDB:
No obvious sequence homology and a lack of broader structural homology
Several bi-manganese enzymes share similar active-site geometry, of which all are known for secondary catalase activity involving water/peroxide/oxygen redox chemistry.
J. Raymond et al, Coord. Chem. Rev. 252, 377 (2008).

9. ASN: the metalloprotein networkCO CU FE MN MO NI V W ZN

10. Metal distribution in network neighborhoods
1-step
2-step
Tendency for one metal to be present in neighborhood.
Clear suppression in case of Fe, weak suppression for Cu

11. Nearest neighbors in ASN share function
Estimate functional similarity along a link through EC values

12. Two possible approaches from Network Systems Biology:
Focus on proteins’ active sites
Focus on vulnerabilities of metabolic network structure

13. Mathematical approach: Flux Balance Analysis
Most used (and only currently realistic) method for modeling genome-scale metabolism
Based on:
List of all possible reactions
Mass conservation
Steady-state
Optimization of a cellular objective

14. Success of growth analysis in wild-type E. coli
Example application:
Success of growth analysis in wild-type E. coli
E. coli growth on acetate very well described by FBA and PhPP analysis:
Overall average error in measured vs. predicted uptake / secretion fluxes of 5.8%
Edwards et al, Nature biotech 19 (2001)

15. Metabolic plasticity: adaptation to different nutrient environments
Sample 30,000 different optimal conditions randomly and uniformly
Metabolic network adapts to environmental changes using:
(a) Flux plasticity (changes in flux rates)
(b) Structural plasticity (reaction [de-] activation)
Flux plasticity
Structural plasticity

16. Metabolic plasticity: adaptation to different nutrient environments
Sample 30,000 different optimal conditions randomly and uniformly
Metabolic network adapts to environmental changes using:
(a) Flux plasticity (changes in flux rates)
(b) Structural plasticity (reaction [de-] activation)
There exists a group of reactions NOT subject to structural
plasticity: the metabolic core
These reactions must play a key role in maintaining the metabolism’s
overall functional integrity

17. The Metabolic Core A connected set of reactions that are ALWAYS active
 not random effect
The larger the network, the smaller the core
 a collective network effect

18. The metabolic core is essential
The core is highly essential:
75% lethal (only 20% in non-core) for E. coli.
84% lethal (16% non-core) for S. cerevisiae.
The core is highly evolutionary conserved in E.coli:
72% of core enzymes (48% of non-core).
The core fluxes and corresponding gene-expression patterns significantly correlated
 Metabolic core is highly essential, conserved and synchronized
Metabolic core of E. coli:
Potential targets for antibiotic treatment
Several current antibiotic targets present:
Sulfonamides (Folate biosynthesis)
Trimethoprim ( # )
Cycloserine (Peptidoglycan biosynth)
Fosfomycin ( # )
Large parts of central carbon metabolism not included.
E. Almaas, et al, PLoS Comput. Biol. 1(7):e68 (2005)

19. Are some essential reactions more difficult to supplant by HGT?
Organism cannot live (produce biomass) without essential genes
Use constraint-based modeling to identify essential genes: Trivial!
Are the essential genes “robust”… ie. can the organism get alternative genes through HGT easily?
Recent analysis of metabolic network structure suggests some reactions are more difficult to supplant than others [Barve & Wagner (PNAS, 2012)]

20. A) B) Using FBA to assess gene essentiality
Case A): Gene knockout is not essential – biomass can still be produced
Case B): Gene knockout is essential – biomass is impossible to produce

21. Identification of robust targets for antimicrobials
Generate “universe” of possible metabolic reactions
Randomly sample possible metabolic networks with fixed characteristics
Assess essentiality of each reaction for each random network sample.

22. Consequence of adding reaction “universe” to a genome-scale metabolic model

23. Analysis of randomly generated networks

24. Exact enumeration of alternative pathways
E. coli
M. tuberculosis
E. coli
M. tuberculosis W
Want few alternative paths, all with large length

25. Synthetic lethality analysis
Reactions that are non-essential during single-reaction inhibition may be essential when paired
Analyze all double-knockout possibilities for randomly generated networks
 Systematically identify reaction pairs that have high degree of vulnerability

26. Network of synthetically lethal reaction pairs
Escherichia coli
Methanosarcina barkeri

27. Summary
Studying protein active-site structure similarity network (ASN) reveals unbiased similarity map of protein function
ASN may be applied to
assess viability of potential drug targets
identify new potential targets of existing drugs
Constraint-based modeling of metabolic networks can be used to identify vulnerable reaction steps, singly or doubly (synthetic lethals)
From FBA analysis:
Some reactions are difficult to supplant when inhibited because few viable alternatives exist among microbes
Some reactions participate in many synthetic lethal pairs and could serve as antimicrobial combination targets.

28. THANKS!!