The initiation of CNS myelin elaboration is marked by the up-regulation of numerous myelin-associated genes. While gene knock-out and gene-centric transcription studies have highlighted the essential role of some DNA-binding TFs in the myelination process, a comprehensive inventory of collaborating TFs driving myelin-associated gene expression has remained elusive. In this investigation we combined genome-wide oligodendrocyte expression profiling, in vivo enhancer analysis and sequence analysis of non-coding regions to establish a TRN model of coordinate myelin-associated gene transcription. Functional characterization of new myelin gene enhancers coupled with novel profiling of the myelination expression program provided the experimental foundation for computational modeling of network components. The derived myelin gene TRN model contains 43 CRMs that link an expanded myelination-relevant TF repertoire of DNA-binding TFs. Furthermore, the TRN model, together with the underlying pro-myelination CRMs, specifies the collaborative relationships among TFs.
To functionally characterize myelin gene enhancers, we used a controlled transgenesis strategy and evaluated reporter-gene expression in oligodendrocytes co-labeled with progenitor or myelinating cell stage markers. This in vivo
approach accesses the developmental context, provides cellular resolution and controls many of the experimental variables typically associated with analysis of in vivo
reporter gene expression. The myelin gene enhancer set reported here augments existing resources such as: Vista Enhancer Browser (VEB) (81
), which records reporter expression phenotypes in mouse embryos, and the Pleiades Promoter Project (PPP) (82
), which identifies targeted transgene activity in the brains of mature mice. Additionally, non-overlapping non-coding regions associated with genes that are common to our enhancer collection have previously been investigated for activity: Pou3f1
(VEB and PPP) and Olig2
). Expansion of such characterized enhancer sequence collections should contribute in a progressive manner to the understanding of gene regulatory mechanisms.
Much like oligodendrocytes in the CNS, myelin production by Schwann cells in the PNS is controlled largely by transcriptional mechanisms (84
). Additionally, PNS and CNS myelin share many structural proteins and accordingly, several of the enhancer sequences characterized here direct expression in both oligodendrocytes and Schwann cells. TFs SOX10, EGR2 and POU3F1 act in a combinatorial manner to drive the myelinating transition in Schwann cells and our analyses also predict the cooperative involvement of Sox and Egr (i.e. Krox) TF family members in oligodendrocyte myelination. Recent experimental analyses of the Mbp
) and Plp1
) loci have exposed regulatory architectures composed of multiple enhancers individually directing expression to oligodendrocytes and/or Schwann cells. Future joint investigations of myelin development in the CNS and PNS should offer unique opportunities to elucidate the transcriptional mechanisms responsible for directing commonly expressed genes in these different cell types.
Several groups have performed gene expression profiling to identify genes that participate in oligodendrocyte development and function. We hypothesized that a core set of differentially expressed genes responsible for oligodendrocyte myelination would be shared by oligodendrocytes maturing on different temporal programs in different CNS domains. Consistent with this expectation, the optic nerve P4–P10 expression comparison dataset shared 65% of its differentially expressed genes with the P16 mouse forebrain stage-specific differential expression profiles and 52% are found in common with the OPC–EOL oligodendrocyte stage transition. Notably, we identified a subset of 31 TFs in the IOLEDd set and 829 genes classified as TFs in the other differentially expressed datasets. While some of these TFs will be responsible for transcriptional regulation of myelin structural proteins, others may control gene expression in related processes such as lipid synthesis and membrane assembly.
The oligodendrocyte myelination TRN model depicts a gene regulatory system that is coordinated by multiple CRMs and includes classes of TFs not previously linked to myelin gene targets. Importantly, the regulatory architecture defined by the TRN accommodates the emerging view of oligodendrocyte combinatorial gene regulation, where specific TFs provide sustained input throughout the myelination process [e.g. Hmg protein Sox10
)] while additional TFs conditionally exert enhancing and/or antagonizing transcriptional effects; for example, in response to environmental influences (87
). Further system complexity is embodied in the TRN class nodes, such that the same CRM elements may be engaged by different TF class members resulting in different regulatory outcomes (88
). Notably, differences in reporter gene expression programs along with loss of expression at mature developmental stages for some of the enhancers investigated in this study implicate further levels of heterogeneity in transcriptional mechanisms.
In an effort to achieve specificity in the predicted myelin gene TRN, we included only those CRM predictions unique to the positive enhancers and also identified as statistically over-represented by the CSA+ analyses. If the enhancer set does not include all myelin gene regulatory sequences, the approach will necessarily exclude contributing CRMs. Nonetheless, the current TRN model provides a significantly refined view of the TFs controlling oligodendrocyte maturation and we expect that progressively enhanced sensitivity will be realized as new myelin gene-associated regulatory sequences are identified and characterized.
Validation of the oligodendrocyte TRN model data using experimental data demonstrated its capacity to identify bonafide regulatory sequence and associated TFs. While a naive CRM prediction approach (i.e. predicting all TFBS combinations in non-coding sequence) would identify these regulatory sequences, the specificity and sensitivity achieved in the TRN model data validation results substantiate our approach. Significantly, the TRN includes TFs with previously recognized roles in oligodendrocyte maturation and further expands on known regulatory mechanisms by predicting CRMs with the capacity to bind factors expressed during oligodendrocyte development. Our study supplements and extends recently developed oligodendrocyte gene expression data resources (9
) and establishes the transition from a gene-centric to a network-informed systems view of the regulatory mechanisms directing oligodendrocyte myelinogenesis. Further characterization of the TF network controlling myelin gene expression should help refine our understanding of oligodendrocyte development as well as suggest novel therapeutic strategies to potentiate their regenerative capacity.