Phylogenetic mixture models for proteins

Standard protein substitution models use a single amino acid replacement rate matrix that summarizes the biological, chemical and physical properties of amino acids. However, site evolution is highly heterogeneous and depends on many factors: genetic code; solvent exposure; secondary and tertiary structure; protein function; etc. These impact the substitution pattern and, in most cases, a single replacement matrix is not enough to represent all the complexity of the evolutionary processes. This paper explores in maximum-likelihood framework phylogenetic mixture models that combine several amino acid replacement matrices to better fit protein evolution.We learn these mixture models from a large alignment database extracted from HSSP, and test the performance using independent alignments from TREEBASE.We compare unsupervised learning approaches, where the site categories are unknown, to supervised ones, where in estimations we use the known category of each site, based on its exposure or its secondary structure. All our models are combined with gamma-distributed rates across sites. Results show that highly significant likelihood gains are obtained when using mixture models compared with the best available single replacement matrices. Mixtures of matrices also improve over mixtures of profiles in the manner of the CAT model. The unsupervised approach tends to be better than the supervised one, but it appears difficult to implement and highly sensitive to the starting values of the parameters, meaning that the supervised approach is still of interest for initialization and model comparison. Using an unsupervised model involving three matrices, the average AIC gain per site with TREEBASE test alignments is 0.31, 0.49 and 0.61 compared with LG (named after Le & Gascuel 2008 Mol. Biol. Evol. 25, 1307-1320), WAG and JTT, respectively. This three-matrix model is significantly better than LG for 34 alignments (among 57), and significantly worse for 1 alignment only. Moreover, tree topologies inferred with our mixture models frequently differ from those obtained with single matrices, indicating that using these mixtures impacts not only the likelihood value but also the output tree. All our models and a PhyML implementation are available from http://atgc.lirmm.fr/mixtures.     

Phylogenetic estimation under codon models can be biased by codon usage heterogeneity

In theory, codon models that account for the dependence of nucleotide substitutions between codon positions as well as differences between synonymous and non-synonymous changes best describe the sequence evolution in protein coding genes. However, in practice we know little about the degree to which violations of the assumptions of codon model-based estimates occur, and how significant these artifacts may be. In nucleotide-based phylogenies from first and second codon positions in a concatenated plastid gene data set, two distantly related taxa--dinoflagellate and haptophyte plastids--were robustly grouped together. This artifactual grouping is attributed to the parallel heterogeneity in leucine (Leu) and serine (Ser) codon usages in the data set. Here, by using this data set, we demonstrated that codon-based phylogenetic estimations are seriously biased, robustly uniting the dinoflagellate and haptophyte plastids into a monophyletic clade, when the model assumption of homogeneity of codon composition was violated. Our results suggest that similar phylogenetic artifacts may occur via codon usage heterogeneity in any amino acids in codon model-based estimations. We advise that homogeneity in codon usage across taxa in a data set be confirmed before codon model-based phylogenetic estimation is attempted.     

Phylogenetic artifacts can be caused by leucine, serine, and arginine codon usage heterogeneity: dinoflagellate plastid origins as a case study

Phylogenetic analyses of first and second codon positions (DNA1 + 2 analysis) and amino acid sequences (protein analysis) are often thought to provide similar estimates of deep-level phylogeny. However, here we report a novel artifact influencing DNA level phylogenetic inference of protein-coding genes introduced by codon usage heterogeneity that causes significant incongruities between DNA1 + 2 and protein analyses. DNA1 + 2 analyses of plastid-encoded psbA genes (encoding of photosystem II D1 proteins) strongly suggest a relationship between haptophyte plastids and typical (peridinin-containing) dinoflagellate plastids. The psbA genes from haptophytes and a subset of the peridinin-type plastids display similar codon usage patterns for Leu, Ser, and Arg, which are each encoded by two separated codon sets that differ at first or first plus second codon positions. Our detailed analyses clearly indicate that these unusual preferences shared by haptophyte and some peridinin-type plastid genes are largely responsible for their strong affinity in DNA analyses. In particular, almost all of the support from DNA level analyses for the monophyly of haptophyte and peridinin-type plastids is lost when the codons corresponding to constant Leu, Ser, and Arg amino acids are excluded, suggesting that this signal comes from rapidly evolving synonymous substitutions, rather than from substitutions that result in amino acid changes. Indeed, protein maximum-likelihood analyses of concatenated PsaA and PsbA amino acid sequences indicate that, although 19' hexanoyloxyfucoxanthin-type (19' HNOF-type) plastids in dinoflagellates group with haptophyte plastids, peridinin-type plastids group weakly with those of stramenopiles. Consequently our results cast doubt on the single origin of peridinin-type and 19' HNOF-type plastids in dinoflagellates previously suggested on the basis of psaA and psbA concatenated gene phylogenetic analyses. We suggest that codon usage heterogeneity could be a more general problem for DNA level analyses of protein-coding genes, even when third codon positions are excluded.     

How thioredoxin can reduce a buried disulphide bond

We present a study of the interaction between thioredoxin and the model enzyme pI258 arsenate reductase (ArsC) from Staphylococcus aureus. ArsC catalyses the reduction of arsenate to arsenite. Three redox active cysteine residues (Cys10, Cys82 and Cys89) are involved. After a single catalytic arsenate reduction event, oxidized ArsC exposes a disulphide bridge between Cys82 and Cys89 on a looped-out redox helix. Thioredoxin converts oxidized ArsC back towards its initial reduced state. In the absence of a reducing environment, the active-site P-loop of ArsC is blocked by the formation of a second disulphide bridge (Cys10-Cys15). While fully reduced ArsC can be recovered by exposing this double oxidized ArsC to thioredoxin, the P-loop disulphide bridge is itself inaccessible to thioredoxin. To reduce this buried Cys10-Cys15 disulphide-bridge in double oxidized ArsC, an intra-molecular Cys10-Cys82 disulphide switch connects the thioredoxin mediated inter-protein thiol-disulphide transfer to the buried disulphide. In the initial step of the reduction mechanism, thioredoxin appears to be selective for oxidized ArsC that requires the redox helix to be looped out for its interaction. The formation of a buried disulphide bridge in the active-site might function as protection against irreversible oxidation of the nucleophilic cysteine, a characteristic that has also been observed in the structurally similar low molecular weight tyrosine phosphatase.     

Using mixture models to characterize disease-related traits

We consider 12 event-related potentials and one electroencephalogram measure as disease-related traits to compare alcohol-dependent individuals (cases) to unaffected individuals (controls). We use two approaches: 1) two-way analysis of variance (with sex and alcohol dependency as the factors), and 2) likelihood ratio tests comparing sex adjusted values of cases to controls assuming that within each group the trait has a 2 (or 3) component normal mixture distribution. In the second approach, we test the null hypothesis that the parameters of the mixtures are equal for the cases and controls. Based on the two-way analysis of variance, we find 1) males have significantly (p < 0.05) lower mean response values than females for 7 of these traits. 2) Alcohol-dependent cases have significantly lower mean response than controls for 3 traits. The mixture analysis of sex-adjusted values of 1 of these traits, the event-related potential obtained at the parietal midline channel (ttth4), found the appearance of a 3-component normal mixture in cases and controls. The mixtures differed in that the cases had significantly lower mean values than controls and significantly different mixing proportions in 2 of the 3 components. Implications of this study are: 1) Sex needs to be taken into account when studying risk factors for alcohol dependency to prevent finding a spurious association between alcohol dependency and the risk factor. 2) Mixture analysis indicates that for the event-related potential "ttth4", the difference observed reflects strong evidence of heterogeneity of response in both the cases and controls.     

Phylogenetic models and model selection for noncoding DNA

Awareness of the complex structure and evolutionary dynamics of noncoding DNA has improved both noncoding sequence alignment and the use of microstructural changes as characters in phylogenetic analysis. The next step is to consider improvements in the use and selection of phylogenetic models for noncoding sequence data. Models of character evolution are central to phylogeny estimation, but the use of an inadequate model can mislead topology selection and branch length estimations. This is particularly likely when sequence divergence is either limited (nearly invariable, as in population-level or species-level studies) or extreme (nearly saturated, as in deep-level studies that focus on conserved secondary structures). Noncoding data sets are often at these extremes, and they can be particularly awkward for model definition and model selection. This paper introduces the goals of model use in phylogenetics and identifies ten issues that arise from the application of models to noncoding sequence data. It is concluded that most of these issues derive from small data set sizes, very low or very high sequence variability, limitations of current phylogenetic models, and possibly character definition and nonindependence. Recommendations are made that should help to improve alignment, character quality, model selection, and phylogeny estimation based on noncoding sequence data.     

Generalized mixture models for molecular phylogenetic estimation

The rapidly growing availability of multigene sequence data during the past decade has enabled phylogeny estimation at phylogenomic scales. However, dealing with evolutionary process heterogeneity across the genome becomes increasingly challenging. Here we develop a mixture model approach that uses reversible jump Markov chain Monte Carlo (MCMC) estimation to permit as many distinct models as the data require. Each additional model considered may be a fully parametrized general time-reversible model or any of its special cases. Furthermore, we expand the usual proposal mechanisms for topology changes to permit hard polytomies (i.e., zero-length internal branches). This new approach is implemented in the Crux software toolkit. We demonstrate the feasibility of using reversible jump MCMC on mixture models by reexamining a well-known 44-taxon mammalian data set comprising 22 concatenated genes. We are able to reproduce the results of the original analysis (with respect to bipartition support) when we make identical assumptions, but when we allow for polytomies and/or use data-driven mixture model estimation, we infer much lower bipartition support values for several key bipartitions.     

Empirical profile mixture models for phylogenetic reconstruction

MOTIVATION: Previous studies have shown that accounting for site-specific amino acid replacement patterns using mixtures of stationary probability profiles offers a promising approach for improving the robustness of phylogenetic reconstructions in the presence of saturation. However, such profile mixture models were introduced only in a Bayesian context, and are not yet available in a maximum likelihood (ML) framework. In addition, these mixture models only perform well on large alignments, from which they can reliably learn the shapes of profiles, and their associated weights. RESULTS: In this work, we introduce an expectation-maximization algorithm for estimating amino acid profile mixtures from alignment databases. We apply it, learning on the HSSP database, and observe that a set of 20 profiles is enough to provide a better statistical fit than currently available empirical matrices (WAG, JTT), in particular on saturated data.     

MicroRNA-22 can reduce parathymosin expression in transdifferentiated hepatocytes

Pancreatic acinar cells AR42J-B13 can transdifferentiate into hepatocyte-like cells permissive for efficient hepatitis B virus (HBV) replication. Here, we profiled miRNAs differentially expressed in AR42J-B13 cells before and after transdifferentiation to hepatocytes, using chip-based microarray. Significant increase of miRNA expression, including miR-21, miR-22, and miR-122a, was confirmed by stem-loop real-time PCR and Northern blot analyses. In contrast, miR-93, miR-130b, and a number of other miRNAs, were significantly reduced after transdifferentiation. To investigate the potential significance of miR-22 in hepatocytes, we generated cell lines stably expressing miR-22. By 2D-DIGE, LC-MS/MS, and Western blot analyses, we identified several potential target genes of miR-22, including parathymosin. In transdifferentiated hepatocytes, miR-22 can inhibit both mRNA and protein expression of parathymosin, probably through a direct and an indirect mechanism. We tested two computer predicted miR-22 target sites at the 3' UTR of parathymosin, by the 3' UTR reporter gene assay. Treatment with anti-miR-22 resulted in significant elevation of the reporter activity. In addition, we observed an in vivo inverse correlation between miR-22 and parathymosin mRNA in their tissue distribution in a rat model. The phenomenon that miR-22 can reduce parathymosin protein was also observed in human hepatoma cell lines Huh7 and HepG2. So far, we detected no major effect on several transdifferentiation markers when AR42J-B13 cells were transfected with miR-22, or anti-miR-22, or a parathymosin expression vector, with or without dexamethasone treatment. Therefore, miR-22 appears to be neither necessary nor sufficient for transdifferentiation. We discussed the possibility that altered expression of some other microRNAs could induce cell cycle arrest leading to transdifferentiation.     

Superovulation can reduce the developmental competence of bovine embryos

This study was done to determine if different superovulatory regimens could have an effect on the percentage of embryos produced using IVM/IVF/IVC. Cyclic heifers (n = 22) were superovulated between Days 8 and 12 of the estrous cycle with 4, 6 or 8 constant doses of FSH-P (4 mg each, twice daily) +/- the addition of 1 mg prostaglandin 24 h before slaughter. Ovaries from these superovulated cows and from untreated cows were collected and the follicles dissected. Oocytes were classified according to the appearance of their cumulus and cytoplasm. Individual culture as well as group culture were performed but an individual culture reduced the percentage of oocytes developing into embryos for both untreated and superovulated animals. The results indicated that despite the superovulation regimen the developmental competence of the oocytes collected was lower (0 to 15% embryos) than that of oocytes from untreated animals (20 to 34% embryos). Small follicles ( < or = 2.7 mm) yielded mostly oocytes with an incomplete or partially expanded cumulus investment that never developed into an embryo. Differences in the morphology of the oocytes from medium (2.7 to 8 mm) and large ( > or = 8 mm) follicles were apparent, but equal developmental rates were obtained between all classes of oocytes (12 and 8% embryos, respectively). Follicular atresia was reduced significantly after superovulation (81% nonatretic follicles in treated vs 42% nonatretic follicles in untreated animals); however oocytes from atretic and slightly atretic follicles developed similarly to those from nonatretic follicles. These results suggest that although superovulation increases follicular size and decreases atresia, these conditions are not sufficient to confer developmental competence on the oocytes.     

Air cleaners can reduce the risk of allergic sensitization

Exposure to house-dust mite allergens is a factor in the development of allergic symptoms in atopic patients. Several measures can be proposed to control indoor allergen levels, inducing a clinical benefit. The use of an air cleaner is one simple way of achieving this goal. We conducted a simulation trial in a proper room, to verify the usefulness of a domestic cleaner, based on mechanical air filtration, to reduce the levels of environmental allergens. We checked the presence of mite components by different methods (Aclotest and Der p 1 ELISA), in dust recovered before and after using the air cleaner. Our results indicate that this approach could be useful in significantly lowering the levels of mite allergens.     

Crystalline anatase-rich titanium can reduce adherence of oral streptococci

Dental implant abutments that emerge through the mucosa are rapidly covered with a salivary protein pellicle to which bacteria bind, initiating biofilm formation. In this study, adherence of early colonizing streptococci, Streptococcus gordonii, Streptococcus oralis, Streptococcus mitis and Streptococcus sanguinis to two saliva-coated anodically oxidized surfaces was compared with that on commercially pure titanium (CpTi). Near edge X-ray absorption (NEXAFS) showed crystalline anatase was more pronounced on the anodically oxidized surfaces than on the CpTi. As revealed by fluorescence microscopy, a four-species mixture, as well as individual bacterial species, exhibited lower adherence after 2?h to the saliva-coated, anatase-rich surfaces than to CpTi. Since wettability did not differ between the saliva-coated surfaces, differences in the concentration and/or configuration of salivary proteins on the anatase-rich surfaces may explain the reduced bacterial binding effect. Anatase-rich surfaces could thus contribute to reduced overall biofilm formation on dental implant abutments through diminished adherence of early colonizers.