Antimicrobial peptides (AMPs) research

Kiadin – peptide antibiotic I designed and we examined in Split, Zagreb, Trieste and Melbourne. It is active against multidrug resistant (MDR) clinical isolates, in particular A. Baumannii, and nontoxic for circulating human blood cells. Rončević et al., Biochimica et Biophysica Acta 1859 (2017) 228–237.

Predicting minimal inhibitory concentration (MIC) against E. coli for Rana-box antimicrobial peptides – MIC is notoriously difficult to predict, but even a modest success with that goal can save time, labour and expense in wandering without guidance through enormous space of possible antimicrobial peptide sequences. We used only the generalization of sequence moment concept, named the sideways asymmetry moment, as the key descriptor, and provided structural insight how different segments of peptide contribute to its bacteriostatic activity. We designed Ranaboxin-1 using our MIC prediction algorithm (http://splitbioinf.pmfst.hr/micpredictor/) and Kozić et al. paper: J Chem Inf Model. 2015,55, 2275-2287.  It has MIC of 2 µM against Gram-negatives: E. coli, A. baumannii and P. Aeruginosa, but its predicted MIC was 0.1 µM. It is the result of the first QSAR model capable of MIC prediction for AMPs of widely different lengths and low identity, by using only one descriptor. Mara Kozić got her MSc degree from the University of Split, Faculty of Science, biophysics graduate programme for giving a major input to this research.

Adepantins – peptide antibiotics we designed at the University of Split, by using anuran antimicrobial peptides, are less than 50% identical to any other antibiotics of that type and exhibit very high selectivity for Gram-negative bacterial cells such as E. coli without being harmful to human cells (Juretić et al. 2009 and 2011). Their dimmers have minimal inhibitory concentration (MIC) of only 0.5 µM, while monomers, such as Adepantin 2, have selectivity index (SI) of about 400, which is about four times higher than the SI of the best natural peptide antibiotic isolated from frog and toads skin (Ilić et al. 2013). High selectivity index means low toxicity for human red blood cells. Nada Ilić got her PhD from the biophysics PhD programme (http://split.pmfst.unist.hr/biophysics/) for her experiments with adepantins, while D. Juretić and D. Vukičević constructed QSAR model for designing adepantins, which contained just one (but novel) descriptor.

Human-fish peptide found by myself in the database of human expressed sequence tags, is not fish AMP “living” in human genome. Its origin remains mysterious, but its close relationship with putative sticklefish AMP is quite obvious. It is possible that some technician in the USA enjoyed its fishburger, while working on decoding human transcriptome, so that fish RNA entered by accident into the reaction mixture. Anyway, experiments by Tessera et al. (2012) determined that it is a broad spectrum peptide antibiotic equally active (MIC = 1 µM) against Gram-negatives (E. coli) and Gram-positives (S. aureus), while its SI is close to 100, which is very good for any AMP. Valentina Tessera got her MSc for her experiments with the hfp.

DADP: the Database of Anuran Defense Peptides http://split4.pmfst.hr/dadp/  – published in the Bioinformatics J. (Novković et al., 2012), quickly become a useful resource for analysing the origin of anuran antimicrobial peptides and for constructing structure-activity models by using activity and toxicity data (for almost 1000 peptides) previously scattered in literature. Mario Novković got his MSc for this work from the University of Split, Faculty of Science, biophysics graduate programme.

Mutator algorithm from the Kamech et al. 2012 paper (http://split4.pmfst.hr/mutator) can quickly find amino acid substitutions expected to increase selectivity index of natural antimicrobial peptides from frogs and toads (the selectivity index is named therapeutic index in that paper). When used wisely (Juretić et al., 2016) it can increase antimicrobial activity as well.

Trichoplanin – peptide antibiotic we found in the database of Trichoplax adhaerens expressed sequence tags (Simunic et al., 2014):

It is broad spectrum antibiotic with very high selectivity index not similar to anything known (the E-value is 10 or >10 after BLASTP search).  The importance of finding it in T. adhaerens is due to possibility that T. adhaerens may be the first animal with smallest genome, and also belonging to rare species which defeated ageing and death. When dispersed into individual cells they find each other again and form the whole organism. Currently, we are examining several additional candidates for T. adhaerens antibiotics, as well as mechanism of action of peptide antibiotics and possible synergy between them, the research topics where I collected a large number of citations in an earlier period (275 and 183 respectively from 1989-1995).

 

Maximum entropy production (MEP) principle and its applications in physics, bioenergetics and enzyme kinetics

The MEP and maximum information entropy (MaxEnt) principles are good candidates for unique physical selection principles that act in concert with biological selection and evolution, not because we love them, but because we find better agreement with experiments in the case of enzyme kinetics (Dobovišek et al. 2011, Lošić et al. 2016) and bioenergetics (Juretić and Županović, 2003, Dewar et al. 2006, Dobovišek et al. 2014, Lošić et al., 2016). Each of proposed alternative optimization principles, such as the maximal metabolic flux, or Prigogine’s principle of minimal entropy production has serious drawbacks. Domagoj Kuić from Faculty of Science, University of Split, got his PhD by exploring MEP and MaxEnt principles within predictive statistical mechanics (Kuić et al., 2012).

For biologists, biophysicists and physicists interested in quantitative biology and in thinking about questions “What is Life” or “Why life exists in the universe” I shall cite several concluding sentences from our manuscripts during the last decade(Juretić and Županović, 2005; Lošić et al. 2016):

“Chemical and biological evolution of macromolecular structures facilitated its integration with thermodynamic evolution in a way which can be best described as a synergistic relationship or positive feedback”.

Living entities tend to increase the entropy production in the universe while active metabolically, so that life serves as a catalytic agent speeding entropy production in its environment. This observation brings biological evolution in synergy with the thermodynamic evolution. By operating close to maximal entropy production, photoconverters couple their own (biological) evolution to thermodynamic evolution in a positive feedback loop which speeds up both evolutions. This “evolution coupling” hypothesis postulates that biosphere evolution is intimately connected with the evolution of life’s physical environment as suggested by the Gaia hypothesis. In this picture biological evolution is just a clever way nature found to accelerate its thermodynamic evolution. Life’s particular goal seems to be to channel the input power into those dissipative pathways where electrochemical rather than only thermal free-energy conversions can occur (Juretić and Županović, 2005).”

To be more specific, our group in Split pointed out that linear nonequilibrium thermodynamics can be derived from the MEP principle. In other words there is an equivalency between Onsager principle of the least dissipation of energy and the MEP principle in this simple case close to thermodynamic equilibrium (Županović et al., FIZIKA A 14 (2005) 89–96, and Entropy 12 (2010) 996–1005.). Also we proved that Kirchhoff's voltage law can be derived from two more general principles: MEP and energy conservation (Županović and Juretić, Phys. Rev. E 70, 056108 (2004)).

Is there any quantitative evidence that MEP (MaxEP) principle is predictive in bioenergetics and enzyme kinetics, either close or far from thermodynamic equilibrium? Indeed, we found such evidence in photosynthesis, ATP-synthase molecular motor, beta-lactamase, and triosephosphate isomerase (TIM). We used MEP and/or MaxENT principle to predict optimal values of kinetic constants for simplified kinetic models and got sometimes very good (ATP-synthase) and more often only rough agreement with experimental values of kinetic constants. It is worth pointing out that minimal entropy production principle by I. Prigogine cannot help at all for predicting rate constants or catalytic constants.

Our latest paper (Bonačić Lošić, Donđivić and Juretić 2016) went a step further and solely by using the MEP principle accomplished several important goals:

  1. It proved that triosephosphate isomerase (TIM) is not a fully evolved enzyme with near-maximal possible reaction rate, as often stated by biochemists
  2. It identified product (R-glyceraldehyde-3-phosphate) release as the rate-limiting step
  3. It proved that only the MaxEP requirement for the product release step led to 30% increase in enzyme activity, specificity constant kcat/KM, and overall entropy production
  4. It connected MEP principle with kinetic and structural studies opening the possibility to find amino acid substitutions leading to an increased frequency of loop six opening and product release

The four-state kinetic diagram for TIM enzyme, where kij and kji are transition kinetic constants (i= 1, 2, 3, 4). Functional states are: 1 the enzyme (E), 2 the enzyme-substratecomplex (ES), 3 an intermediate (EZ), and 4 the enzyme-product complex (EP)

Distribution of entropy productions for each of four TIM-enzyme functional transitions before (blue columns) and after (extended red area over blue area in each column) MaxEP optimization for the 4-th step (product release). Product (GAP) concentration was 0.064 μM (upper figure) or 0.013 μM (lower figure).

 

Secondary structure prediction for membrane proteins

Now classic Juretić et al. paper 2002 collected a large number of citations (195), because high quality predictions for proteins segments with preference to enter membrane as shorter or longer (transmembrane) helices are still very valuable. Versions SPLIT 3.5 and SPLIT 4 are available for online scientific calculations and have been valuable free bioinformatic resources for more than 15 years. An earlier version of the SPLIT algorithm was the first online server for scientific calculations in Croatia. Better versions of SPLIT algorithm can be constructed today by using much higher number of integral membrane proteins with known sequence location of transmembrane helices during the training procedure for extracting preference functions, and by eliminating known weaknesses such as:

  1. too high percentage of false TMH predictions in soluble proteins,
  2. too high false TMH prediction for signal sequences,
  3. neglecting homologous sequences to the tested one.
SPLIT 4.0 potential for detecting helices buried in one layer of membrane bilayer, like P segments from voltage-gated ion channels, was never used for publication. Interestingly, earlier software version SPLIT 3.5 proved its usefulness in distinguishing antimicrobial peptide toxins from antimicrobial peptide antibiotics. Peptide toxins are predicted to enter deeper into membrane bilayer and to form transmembrane helix, while peptide antibiotics prefer membrane surface and are likely to be more amphipathic (Juretić et al., 2011):