Rare neural correlations implement robotic conditioning with delayed rewards and disturbances.

Neural conditioning associates cues and actions with following rewards. The environments in which robots operate, however, are pervaded by a variety of disturbing stimuli and uncertain timing. In particular, variable reward delays make it difficult to reconstruct which previous actions are responsible for following rewards. Such an uncertainty is handled by biological neural networks, but represents a challenge for computational models, suggesting the lack of a satisfactory theory for robotic neural conditioning. The present study demonstrates the use of rare neural correlations in making correct associations between rewards and previous cues or actions. Rare correlations are functional in selecting sparse synapses to be eligible for later weight updates if a reward occurs. The repetition of this process singles out the associating and reward-triggering pathways, and thereby copes with distal rewards. The neural network displays macro-level classical and operant conditioning, which is demonstrated in an interactive real-life human-robot interaction. The proposed mechanism models realistic conditioning in humans and animals and implements similar behaviors in neuro-robotic platforms.

Online learning and generalization of parts-based image representations by non-negative sparse autoencoders.

We present an efficient online learning scheme for non-negative sparse coding in autoencoder neural networks. It comprises a novel synaptic decay rule that ensures non-negative weights in combination with an intrinsic self-adaptation rule that optimizes sparseness of the non-negative encoding. We show that non-negativity constrains the space of solutions such that overfitting is prevented and very similar encodings are found irrespective of the network initialization and size. We benchmark the novel method on real-world datasets of handwritten digits and faces. The autoencoder yields higher sparseness and lower reconstruction errors than related offline algorithms based on matrix factorization. It generalizes to new inputs both accurately and without costly computations, which is fundamentally different from the classical matrix factorization approaches.

Subpopulation-specific transcriptome analysis of competence-stimulating-peptide-induced Streptococcus mutans.

Competence-stimulating-peptide (CSP)-mediated competence development in Streptococcus mutans is a transient and biphasic process, since only a subpopulation induces the expression of ComX in the presence of CSP, and the activation of the DNA uptake machinery in this fraction shuts down ~3 to 4 h postinduction. Here, we combine for the first time, to our knowledge, the bacterial flow-cytometric sorting of cells and subpopulation-specific transcriptome analysis of both the competent and noncompetent fraction of CSP-treated S. mutans cells. Sorting was guided by a ComX-green fluorescent protein (ComX-GFP) reporter, and the transcriptome analysis demonstrated the successful combination of both methods, because a strong enrichment of transcripts for comX and its downstream genes was achieved. Three two-component systems were expressed in the competent fraction, and among them was ComDE. Moreover, the recently identified regulator system ComR/S was expressed exclusively in the competent fraction. In contrast, the expression of bacteriocin-related genes was at the same level in all cells. GFP reporter strains for ComE and CipB (mutacin V) confirmed this expression pattern on the single-cell level. Fluorescence microscopy revealed that some ComX-expressing cells committed autolysis in an early stage of competence initiation. In viable ComX-expressing cells, the uptake of DNA could be shown on the single-cell level. This study demonstrates that all cells in the population respond to CSP through the activation of bacteriocin-related genes. Some of these cells start to activate ComX expression but then segregate into two subpopulations, one becoming competent and another one that lyses, resulting in intrapopulation diversity.

Damage of Streptococcus mutans biofilms by carolacton, a secondary metabolite from the myxobacterium Sorangium cellulosum.

BACKGROUND: Streptococcus mutans is a major pathogen in human dental caries. One of its important virulence properties is the ability to form biofilms (dental plaque) on tooth surfaces. Eradication of such biofilms is extremely difficult. We therefore screened a library of secondary metabolites from myxobacteria for their ability to damage biofilms of S. mutans. RESULTS: Here we show that carolacton, a secondary metabolite isolated from Sorangium cellulosum, has high antibacterial activity against biofilms of S. mutans. Planktonic growth of bacteria was only slightly impaired and no acute cytotoxicity against mouse fibroblasts could be observed. Carolacton caused death of S. mutans biofilm cells, elongation of cell chains, and changes in cell morphology. At a concentration of 10 nM carolacton, biofilm damage was already at 35% under anaerobic conditions. A knock-out mutant for comD, encoding a histidine kinase specific for the competence stimulating peptide (CSP), was slightly less sensitive to carolacton than the wildtype. Expression of the competence related alternate sigma factor ComX was strongly reduced by carolacton, as determined by a pcomX luciferase reporter strain. CONCLUSIONS: Carolacton possibly interferes with the density dependent signalling systems in S. mutans and may represent a novel approach for the prevention of dental caries.

Streptococcus mutans inhibits Candida albicans hyphal formation by the fatty acid signaling molecule trans-2-decenoic acid (SDSF).

In the human mouth, fungi and several hundred species of bacteria coexist. Here we report a case of interkingdom signaling in the oral cavity: A compound excreted by the caries bacterium Streptococcus mutans inhibits the morphological transition from yeast to hyphae, an important virulence trait, in the opportunistic fungus Candida albicans. The compound excreted by S. mutans was originally studied because it inhibited signaling by the universal bacterial signal autoinducer-2 (AI-2), determined by the luminescence of a Vibrio harveyi sensor strain. The inhibitor was purified from cell-free culture supernatants of S. mutans guided by its activity. Its chemical structure was elucidated by using NMR spectroscopy and GC-MS and proved to be trans-2-decenoic acid. We show that trans-2-decenoic acid does not inhibit AI-2-specific signaling, but rather the luciferase reaction used for its detection. A potential biological role of trans-2-decenoic acid was then discovered. It is able to suppress the transition from yeast to hyphal morphology in the opportunistic human pathogen Candida albicans at concentrations that do not affect growth. The expression of HWP1, a hyphal-specific signature gene of C. albicans, is abolished by trans-2-decenoic acid. trans-2-Decenoic acid is structurally similar to the diffusible signal factor (DSF) family of interkingdom-signaling molecules and is the first member of this family from a Gram-positive organism (Streptococcus DSF, SDSF). SDSF activity was also found in S. mitis, S. oralis, and S. sanguinis, but not in other oral bacteria. SDSF could be relevant in shaping multispecies Candida bacteria biofilms in the human body.

Characterization of mleR, a positive regulator of malolactic fermentation and part of the acid tolerance response in Streptococcus mutans.

BACKGROUND: One of the key virulence determinants of Streptococcus mutans, the primary etiological agent of human dental caries, is its strong acid tolerance. The acid tolerance response (ATR) of S. mutans comprises several mechanisms that are induced at low pH and allow the cells to quickly adapt to a lethal pH environment. Malolactic fermentation (MLF) converts L-malate to L-lactate and carbon dioxide and furthermore regenerates ATP, which is used to translocate protons across the membrane. Thus, MLF may contribute to the aciduricity of S. mutans but has not been associated with the ATR so far. RESULTS: Here we show that the malolactic fermentation (mle) genes are under the control of acid inducible promoters which are induced within the first 30 minutes upon acid shock in the absence of malate. Thus, MLF is part of the early acid tolerance response of S. mutans. However, acidic conditions, the presence of the regulator MleR and L-malate were required to achieve maximal expression of all genes, including mleR itself. Deletion of mleR resulted in a decreased capacity to carry out MLF and impaired survival at lethal pH in the presence of L-malate. Gel retardation assays indicated the presence of multiple binding sites for MleR. Differences in the retardation patterns occurred in the presence of L-malate, thus demonstrating its role as co-inducer for transcriptional regulation. CONCLUSION: This study shows that the MLF gene cluster is part of the early acid tolerance response in S. mutans and is induced by both low pH and L-malate.

Autoinducer-2-regulated genes in Streptococcus mutans UA159 and global metabolic effect of the luxS mutation.

Autoinducer 2 (AI-2) is the only species-nonspecific autoinducer known in bacteria and is produced by both gram-negative and gram-positive organisms. Consequently, it is proposed to function as a universal quorum-sensing signal for interaction between bacterial species. AI-2 is produced as the by-product of a metabolic transformation carried out by the LuxS enzyme. To separate the metabolic function of the LuxS enzyme from the signaling role of AI-2, we carried out a global transcriptome analysis of a luxS null mutant culture of Streptococcus mutans UA159, an important cariogenic bacterium and a crucial component of the dental plaque biofilm community, in comparison to a luxS null mutant culture supplemented with chemically pure 4,5-dihydroxy-2,3-pentanedione, the precursor of AI-2. The data revealed fundamental changes in gene expression affecting 585 genes (30% of the genome) which could not be restored by the signal molecule AI-2 and are therefore not caused by quorum sensing but by lack of the transformation carried out by the LuxS enzyme in the activated methyl cycle. All functional classes of enzymes were affected, including genes known to be important for biofilm formation, bacteriocin synthesis, competence, and acid tolerance. At the same time, 59 genes were identified whose transcription clearly responded to the addition of AI-2. Some of them were related to protein synthesis, stress, and cell division. Three membrane transport proteins were upregulated which are not related to any of the known AI-2 transporters. Three transcription factors were identified whose transcription was stimulated repeatedly by AI-2 addition during growth. Finally, a global regulatory protein, the delta subunit of the RNA polymerase (rpoE), was induced 147-fold by AI-2, representing the largest differential gene expression observed. The data show that many phenotypes related to the luxS mutation cannot be ascribed to quorum sensing and have identified for the first time regulatory proteins potentially mediating AI-2-based signaling in gram-positive bacteria.

Analysing traces of autoinducer-2 requires standardization of the Vibrio harveyi bioassay.

Autoinducer-2 (furanosyl borate diester) is a biologically active compound whose role as a universal bacterial signalling molecule is currently under intense investigation. Because of its instability and the low concentrations of it found in biological samples, its detection relies at present on a bioassay that measures the difference in the timing of the luminescence of the Vibrio harveyi BB170 sensor strain with and without externally added AI-2. Here we systematically investigated which parameters affected the fold induction values of luminescence obtained in the bioassay and developed a modified protocol. Our experiments showed that growth and luminescence of V. harveyi BB170 are strongly influenced by trace elements. In particular, addition of Fe(3+) within a certain concentration range to the growth medium of the preinoculum culture improved the reproducibility and reduced the variance of the bioassay. In contrast, trace elements and vitamins introduced directly into the bioassay caused inhibitory effects. The initial density and luminescence of the sensor strain are very important and the values required for these parameters were defined. Borate interferes with the detection of AI-2 by giving false positive results. The response of V. harveyi BB170 to chemically synthesized AI-2 in the bioassay is nonlinear except over a very small concentration range; it is maximum over three orders of magnitude and shows inhibition above 35 microM. Based on the modified protocol, we were able to detect AI-2 in the absence of inhibitors with maximum fold induction values for the positive control (chemically synthesized AI-2) of >120 with a standard deviation of approximately 30% in a reliable and reproducible way.