The developmental basis for scaling of mammalian tooth size.

When evolution leads to differences in body size, organs generally scale along. A well-known example of the tight relationship between organ and body size is the scaling of mammalian molar teeth. To investigate how teeth scale during development and evolution, we compared molar development from initiation through final size in the mouse and the rat. Whereas the linear dimensions of the rat molars are twice that of the mouse molars, their shapes are largely the same. Here, we focus on the first lower molars that are considered the most reliable dental proxy for size-related patterns due to their low within-species variability. We found that scaling of the molars starts early, and that the rat molar is patterned equally as fast but in a larger size than the mouse molar. Using transcriptomics, we discovered that a known regulator of body size, insulin-like growth factor 1 (Igf1), is more highly expressed in the rat molars compared to the mouse molars. Ex vivo and in vivo mouse models demonstrated that modulation of the IGF pathway reproduces several aspects of the observed scaling process. Furthermore, analysis of IGF1-treated mouse molars and computational modeling indicate that IGF signaling scales teeth by simultaneously enhancing growth and by inhibiting the cusp-patterning program, thereby providing a relatively simple mechanism for scaling teeth during development and evolution. Finally, comparative data from shrews to elephants suggest that this scaling mechanism regulates the minimum tooth size possible, as well as the patterning potential of large teeth.

Transcriptomic landscape of early hair follicle and epidermal development.

Morphogenesis of ectodermal organs, such as hair, tooth, and mammary gland, starts with the formation of local epithelial thickenings, or placodes, but it remains to be determined how distinct cell types and differentiation programs are established during ontogeny. Here, we use bulk and single-cell transcriptomics and pseudotime modeling to address these questions in developing hair follicles and epidermis and produce a comprehensive transcriptomic profile of cellular populations in the hair placode and interplacodal epithelium. We report previously unknown cell populations and marker genes, including early suprabasal and genuine interfollicular basal markers, and propose the identity of suprabasal progenitors. By uncovering four different hair placode cell populations organized in three spatially distinct areas, with fine gene expression gradients between them, we posit early biases in cell fate establishment. This work is accompanied by a readily accessible online tool to stimulate further research on skin appendages and their progenitors.

Chromosomal neighbourhoods allow identification of organ specific changes in gene expression.

Although most genes share their chromosomal neighbourhood with other genes, distribution of genes has not been explored in the context of individual organ development; the common focus of developmental biology studies. Because developmental processes are often associated with initially subtle changes in gene expression, here we explored whether neighbouring genes are informative in the identification of differentially expressed genes. First, we quantified the chromosomal neighbourhood patterns of genes having related functional roles in the mammalian genome. Although the majority of protein coding genes have at least five neighbours within 1 Mb window around each gene, very few of these neighbours regulate development of the same organ. Analyses of transcriptomes of developing mouse molar teeth revealed that whereas expression of genes regulating tooth development changes, their neighbouring genes show no marked changes, irrespective of their level of expression. Finally, we test whether inclusion of gene neighbourhood in the analyses of differential expression could provide additional benefits. For the analyses, we developed an algorithm, called DELocal that identifies differentially expressed genes by comparing their expression changes to changes in adjacent genes in their chromosomal regions. Our results show that DELocal removes detection bias towards large changes in expression, thereby allowing identification of even subtle changes in development. Future studies, including the detection of differential expression, may benefit from, and further characterize the significance of gene-gene neighbour relationships.

System-level analyses of keystone genes required for mammalian tooth development.

When a null mutation of a gene causes a complete developmental arrest, the gene is typically considered essential for life. Yet, in most cases, null mutations have more subtle effects on the phenotype. Here we used the phenotypic severity of mutations as a tool to examine system-level dynamics of gene expression. We classify genes required for the normal development of the mouse molar into different categories that range from essential to subtle modification of the phenotype. Collectively, we call these the developmental keystone genes. Transcriptome profiling using microarray and RNAseq analyses of patterning stage mouse molars show highly elevated expression levels for genes essential for the progression of tooth development, a result reminiscent of essential genes in single-cell organisms. Elevated expression levels of progression genes were also detected in developing rat molars, suggesting evolutionary conservation of this system-level dynamics. Single-cell RNAseq analyses of developing mouse molars reveal that even though the size of the expression domain, measured in the number of cells, is the main driver of organ-level expression, progression genes show high cell-level transcript abundances. Progression genes are also upregulated within their pathways, which themselves are highly expressed. In contrast, a high proportion of the genes required for normal tooth patterning are secreted ligands that are expressed in fewer cells than their receptors and intracellular components. Overall, even though expression patterns of individual genes can be highly different, conserved system-level principles of gene expression can be detected using phenotypically defined gene categories.

Hair follicle dermal condensation forms via Fgf20 primed cell cycle exit, cell motility, and aggregation.

Mesenchymal condensation is a critical step in organogenesis, yet the underlying molecular and cellular mechanisms remain poorly understood. The hair follicle dermal condensate is the precursor to the permanent mesenchymal unit of the hair follicle, the dermal papilla, which regulates hair cycling throughout life and bears hair inductive potential. Dermal condensate morphogenesis depends on epithelial Fibroblast Growth Factor 20 (Fgf20). Here, we combine mouse models with 3D and 4D microscopy to demonstrate that dermal condensates form de novo and via directional migration. We identify cell cycle exit and cell shape changes as early hallmarks of dermal condensate morphogenesis and find that Fgf20 primes these cellular behaviors and enhances cell motility and condensation. RNAseq profiling of immediate Fgf20 targets revealed induction of a subset of dermal condensate marker genes. Collectively, these data indicate that dermal condensation occurs via directed cell movement and that Fgf20 orchestrates the early cellular and molecular events.

How do eubacterial organisms manage aggregation-prone proteome?

Eubacterial genomes vary considerably in their nucleotide composition. The percentage of genetic material constituted by guanosine and cytosine (GC) nucleotides ranges from 20% to 70%.  It has been posited that GC-poor organisms are more dependent on protein folding machinery. Previous studies have ascribed this to the accumulation of mildly deleterious mutations in these organisms due to population bottlenecks. This phenomenon has been supported by protein folding simulations, which showed that proteins encoded by GC-poor organisms are more prone to aggregation than proteins encoded by GC-rich organisms. To test this proposition using a genome-wide approach, we classified different eubacterial proteomes in terms of their aggregation propensity and chaperone-dependence using multiple machine learning models. In contrast to the expected decrease in protein aggregation with an increase in GC richness, we found that the aggregation propensity of proteomes increases with GC content. A similar and even more significant correlation was obtained with the GroEL-dependence of proteomes: GC-poor proteomes have evolved to be less dependent on GroEL than GC-rich proteomes. We thus propose that a decrease in eubacterial GC content may have been selected in organisms facing proteostasis problems.

Selection of relevant features from amino acids enables development of robust classifiers.

Machine learning (ML) has been extensively applied to develop models and to understand high-throughput data of biological processes. However, new ML models, trained with novel experimental results, are required to build regularly for more precise predictions. ML methods can build models from numeric data, whereas biological data are generally textual (DNA, protein sequences) or images and needs feature calculation algorithms to generate quantitative features. Programming skills along with domain knowledge are required to develop these algorithms. Therefore, the process of knowledge discovery through ML is decelerated due to lack of generic tools to construct features and to build models directly from the data. Hence, we developed a schema that calculates about 5,000 features, selects relevant features and develops protein classifiers from the training data. To demonstrate the general applicability and robustness of our method, fungal adhesins and nuclear receptor proteins were used for building classifiers which outperformed existing classifiers when tested on independent data. Next, we built a classifier for mitochondrial proteins of Plasmodium falciparum which causes human malaria because the latest corresponding classifiers are not publically accessible. Our classifier attained 98.18 % accuracy and 0.95 Matthews correlation coefficient by fivefold cross-validation and outperformed existing classifiers on independent test set. We implemented this schema as user-friendly and open source application Pro-Gyan ( http://code.google.com/p/pro-gyan/ ), to build and share executable classifiers without programming knowledge.