Transcriptional regulators of the Golli/myelin basic protein locus integrate additive and stealth activities

Myelin is composed of plasma membrane spirally wrapped around axons and compacted into dense sheaths by myelin-associated proteins. Myelin is elaborated by neuroepithelial derived oligodendrocytes in the central nervous system (CNS) and by neural crest derived Schwann cells in the peripheral nervous system (PNS). While some myelin proteins accumulate in only one lineage, myelin basic protein (Mbp) is expressed in both. Overlapping the Mbp gene is Golli, a transcriptional unit that is expressed widely both within and beyond the nervous system. A super-enhancer domain within the Golli/Mbp locus contains multiple enhancers shown previously to drive reporter construct expression specifically in oligodendrocytes or Schwann cells. In order to determine the contribution of each enhancer to the Golli/Mbp expression program, and to reveal if functional interactions occur among them, we derived mouse lines in which they were deleted, either singly or in different combinations, and relative mRNA accumulation was measured at key stages of early development and at maturity. Although super-enhancers have been shown previously to facilitate interaction among their component enhancers, the enhancers investigated here demonstrated largely additive relationships. However, enhancers demonstrating autonomous activity strictly in one lineage, when missing, were found to significantly reduce output in the other, thus revealing cryptic “stealth” activity. Further, in the absence of a key oligodendrocyte enhancer, Golli accumulation was markedly and uniformly attenuated in all cell types investigated. Our observations suggest a model in which enhancer-mediated DNA-looping and potential super-enhancer properties underlie Golli/Mbp regulatory organization.


INTRODUCTION
Myelin facilitates rapid and energy efficient conduction of electrical signals and when disrupted, as in demyelinating diseases such as Multiple Sclerosis, nervous system function can be severely compromised. In the mouse, myelination initiates between birth and weaning in both the CNS and PNS with peak accumulation of the mRNAs encoding myelin specific proteins reached during the third postnatal week (1,2). Adaptive changes in myelin volume are also thought to influence circuit properties in the mature nervous system (3,4).
Consequently, the mechanisms controlling myelin synthesis, including regulation of the genes encoding myelin related proteins, are the focus of intense investigation (5).
Myelin basic protein (MBP), a major myelin component, is an intrinsically disordered protein susceptible to multiple post-translational modifications and while largely concentrated in the myelin sheath, it is implicated in a wide array of cellular functions (6). Mice lacking MBP demonstrate a shivering phenotype and while only limited ultrastructural anomalies and a modest decrease in myelin sheath thickness is observed in their PNS (7)(8)(9)(10), compact myelin is absent in their CNS (11)(12)(13) Notably, a positive correlation between the accumulated level of Mbp mRNA and CNS myelin sheath thickness was reported demonstrating that MBP is not only essential but a limiting factor in CNS myelin production (14,15). The Golli/Mbp locus also expresses Golli transcripts that initiate far upstream and incorporate various Mbp exons. Golli accumulates in diverse lineages both within and beyond the nervous system on cell-type specific developmental expression programs where it has been shown to modulate Ca 2+ transients (16)(17)(18)(19)(20)(21)(22).
Motivated by the critical role myelin plays in nervous system function, and the essential and limiting role of MBP in CNS myelin formation, we and others have characterized features of the mechanism controlling transcription of the Golli/Mbp locus (1,2,26,33,34,(41)(42)(43)(44)(45)(46). Because Mbp is expressed by the two different cell types that elaborate myelin, accumulates in a well characterized post-natal developmental program and has an intimate and unusual association with the widely expressed overlapping Golli transcriptional unit, the Golli/Mbp locus represents a particularly rich target within which the higher order organization of transcriptional regulators can be revealed. Moreover, a recently defined super-enhancer domain encompasses multiple Golli/Mbp enhancers providing a further opportunity to investigate the function of such domains.
Previously, the lineage specificities and developmental programs conferred by Golli/Mbp-associated enhancers were assigned using reporter constructs (2,41,43,44,(46)(47)(48). However, in such preparations, enhancers are isolated from their normal chromatin environment and often ligated adjacent to each other or directly to a promoter creating novel spatial relationships that may impose or lead to loss of higher-order interactions (49). Therefore, we sought to characterize enhancer contributions in the fully integrated context of the endogenous Golli/Mbp locus. Using CRISPR based gene editing we derived lines of mice bearing alleles deleted of one or more enhancers and these were assessed for Mbp and Golli mRNA accumulation (relative to Gapdh mRNA) at key stages of post-natal development. Such KO alleles caused the greatest reduction in mRNA accumulation in the lineage where the deleted enhancer/s demonstrated autonomous targeting activity in reporter constructs. Unexpectedly, some also led to reduced activity in the lineage where they demonstrated no autonomous targeting capacity and we refer to such additional cryptic functions as "stealth" activity. Our observations are consistent with a model in which transcription factor mediated interactions give rise to chromatin looping that contributes to both Mbp and Golli programing.

Enhancer KO lines
Five modules of high interspecies conservation are located 5' of the Mbp start site within a putative super-enhancer domain (Fig. 1A). In previous studies, M1 (which encompasses the M1E enhancer and the contiguous Mbp proximal promoter), M3 and M5 were each shown to drive expression in oligodendrocytes while expression in Schwann cells was driven by M4. M2 demonstrated no autonomous activity in either lineage (2,(41)(42)(43)46). In the present investigation, we derived mouse lines homozygous for alleles deleted individually of M3, M4 or M5 or combined KOs of M3/M5 and M1E/M3/M5. Additionally, to explore the potential function of an enhancer subdomain, an allele was derived bearing a partial deletion of M3, M3(225)KO, that was shown previously to upregulate activity in both oligodendrocytes and Schwann cells (42,43) (Fig. 1B). At key stages of myelination, accumulation of Mbp mRNA was analyzed in spinal cord and sciatic nerves while accumulation of Golli mRNA was assessed in spinal cord, thymus and retina at multiple ages.

Mbp mRNA accumulation in spinal cord oligodendrocytes
Relative accumulation of Mbp mRNA in maturing oligodendrocytes and Schwann cells was measured by analyzing whole tissue homogenates of spinal cord (CNS) and sciatic nerve (PNS) respectively.
Oligodendrocytes initiate expression of Mbp as a terminal maturation event coincident with myelin sheath elaboration. Although myelination initiates on different schedules in different CNS domains, we restricted our A B analysis to the cervical spinal cord where myelination initiates perinatally and oligodendrocyte numbers remain constant from at least P10 through P30 (50). Thus, a close relationship between the mRNA levels observed in whole tissue and that realized within individual oligodendrocytes was expected, and a similar relationship likely exists in the PNS for Schwann cells (51). Samples were obtained at P7, a stage of maturation when significant myelin elaboration in cervical spinal cord has occurred (1); at P14, when myelin acquisition nears peak levels; at P21, when the levels of myelin protein mRNAs have declined from peak levels; at P30, when mature myelin maintaining cells predominate; and at P90, when mice are fully mature.
Wild type (WT). Mbp mRNA was readily detectable at P7 and rapidly increased over the next week.
Relative to P14, levels decreased to 85% by P21, 74% by P30 and 33% by P90; a developmental program consistent with prior investigations of multiple myelin gene expression programs and Mbp regulated reporters     Table S2).

Golli mRNA accumulation
In indicating that the M3 subsequence deleted in this allele plays no role in Golli regulation at this age. However, at P30 Golli accumulation was 141% and 130% of WT in spinal cord and thymus respectively indicating an age-specific putative repressor role for the deleted sequence ( Fig. 4 and 5, Supplementary Table S3).

Reporter genes vs enhancer KOs
Levels of chromatin organization beyond that associated with individual enhancers are thought to contribute to transcriptional regulation. Consequently, reporter constructs bearing isolated enhancers may not reveal the full extent of their regulatory activity. One such higher order level of organization is observed in
Multiple data sets from human brain revealed a super-enhancer domain extending through and

Reporter genes vs partial enhancer module KOs
Insight into the location of sequences engaged in proposed promoter-enhancer interactions was obtained from alleles bearing partial enhancer deletions. Absence of the 77 bp M1E sequence in the M1EM3M5KO allele reduced Mbp mRNA accumulation in oligodendrocytes beyond that imposed by the M3M5KO allele alone, consistent with an enhancing role for the deleted M1E domain in oligodendrocytes (Fig. 2). However, loss of M1E had no effect on Mbp mRNA accumulation in Schwann cells demonstrating that it has no Schwann cell enhancing activity nor a role in the productive engagement of M4 with the promoter components that remain in the M1EM3M5KO allele (Fig. 3).
A striking difference between the output of an enhancer KO allele and that predicted by a reporter gene was encountered with the M3(225) sub-sequence. This sequence drove reporter expression in oligodendrocytes approximately 5-fold higher than the intact M3 and when ligated directly to a minimal hsp promoter, it converted the transient pre-weaning activity of intact M3 in Schwann cells to constitutive expression at levels up to 50 fold higher (43). However, in the context of the endogenous locus, the same M3(225) truncation paradoxically had no effect in Schwann cells and reduced, rather than enhanced, Mbp mRNA accumulation in oligodendrocytes ( Fig. 2 and 3). The basis for this striking difference between reporter construct activity compared to the same deletion in the endogenous locus remains unknown but the experimentally imposed close proximity of M3(225) and the hsp promoter in reporter constructs, may underlie the observed upregulation.
Further demonstration that distinct sequence sub-domains affect promoter-enhancer relationships was revealed by Golli expression. All alleles deleted of the intact M3 reduced Golli mRNA accumulation to approximately 10% of WT in all tissues at all ages examined ( Fig. 4 and 5). In marked contrast, the partially

A model accommodating enhancer targeting activities
Here, the documented Golli/Mbp enhancer activities and their functional interactions lead to a DNAlooping model compatible with much of the integrated output of the locus (Fig. 6A and B). This model accommodates previously described differential TF binding (32,59,64,(71)(72)(73)(74)(75) and suggests that lineagespecific partnerships occur within sub-domains created by chromatin loops. Specifically, self-associating TF dimers, such as those formed by YY1 and SP1, are implicated in chromatin looping to support long-distance enhancer-promoter interaction (49,(76)(77)(78)(79). This model can also accommodate the emergence of lineage specific regulatory programs such that the output of a single locus can evolve to match the unique requirements of the different myelinating cells in the CNS and PNS. However, this model does not take into account the mediators of stealth activity. M3(225) and the Golli proximal promoter have SP1 binding sites (Jaspar; relative score (rs) > 90%) (80), however lower affinity SP1 and YY1 motifs (80% < rs < 90%) exist in all. Accordingly, in oligodendrocytes, M3 and M5 interaction with the promoter could involve YY1 dimerization (Fig. 6A) and consistent with this, the M3(225)KO allele, lacking the sequence containing YY1 binding site, resulted in the same Mbp mRNA reduction as observed in M3KO mice (P30). Additionally, conditional KO of YY1 led to amyelination and behavioral phenotypes characteristic of MBP null shiverer mice (81,82). However, during active myelination at P14, the M3(225)KO allele conferred a more modest reduction than the M3KO allele. During with this period, SP1 accumulation transiently increases and becomes phosphorylated via the PKC/Erk pathway (83,84). As the SP1 binding site remains in the M3(225)KO allele and the SP1 phosphorylated through that pathways has been shown to increase its DNA binding capacity (85), the partial restoration of enhancer-promoter interaction in the absence of the YY1 binding site could be conferred through SP1 (Fig. 6A).
Although enhancer activity mediated largely by YY1 dimerization conveniently accommodates our observations on Mbp expression, it fails to account for Golli programming (Fig. 6B). While the Golli proximal promoter contains both YY1 and SP1 binding motifs, among the enhancers investigated here, only the oligodendrocyte M3 enhancer that contains an SP1 binding site modulates Golli output (80). Consistent with this model, Golli expression driven by the truncated enhancer in M3(225)KO mice was anticipated by its retained SP1 binding site and hence its predicted capacity to associate with the Golli promoter (Fig. 6B).

Enhancer activity during development
Beyond insight into the functional organization of Golli/Mbp regulatory sequences, the developmental programs conferred by Mbp enhancer-regulated reporter genes and KO alleles illuminate further aspects of oligodendrocyte biology. As demonstrated by the capacity of oligodendrocytes to myelinate inanimate fibers in vitro, initial myelin elaboration can be supported entirely by oligodendrocyte intrinsic programming (86).
However, it is now widely recognized that myelin in the mature CNS can demonstrate plastic changes potentially in response to neuronal activity (3, 4, 87, 88). Here, the M3(225)KO allele showed differential age-

Animals
All experiments were carried out in accordance with the guidelines of the Canadian Council on Animal Care and the McGill University Animal Care Committee.

M4 and M5 sequences
The 422bp M4 enhancer targeted here was described previously (41). M5 refers to the target of the MYRF ChIP carried out in rat (44

Zygote manipulation, delivery of CRISPR components and transplantation into pseudopregnant mice
Zygotes were recovered mid-day from the oviduct of WT or M3KO C57Bl/6 mice (42) naturally mated to wmN2 transgenic mice (92). The cumulus cells were removed by a short incubation in 1% hyaluronidase/M2 medium (Millipore) and moved into Advanced KSOM media (Millipore) at 37℃ with 5% CO 2 .

RNA extraction and qRT-PCR
Total RNA extraction was done using Trizol (Life Technologies) and a Qiagen RNeasy MinElute Cleanup kit. RNA was eluted in nuclease free water and its concentration was measured using a spectrophotometer. The RT reaction was carried out using Superscript IV VILO Mastermix (Life Technologies) using 1ug of total RNA according to the manufacturer's instructions and the resulting cDNA was stored at - fold serial dilutions of a DNA standard were run in triplicate to generate a standard curve. Standards were prepared by amplifying a sequence larger than the measured amplicon. After standard PCR, the single band was purified from a gel with a NucleoSpin Gel and PCR cleanup kit (Macherey-Nagel) and its concentration determined. The efficiencies of reactions for both Taqman and SYBR green methods inferred from standard curves were 95-105%.

Data analysis
After the generation of data, the measurements of each sample from both dilutions were averaged.
Relative amounts of Mbp and Golli were calculated by dividing the average of each by the average of Gapdh for the same sample. The relative Mbp and Golli measurements of all samples of each mouse line were averaged and the standard deviation was calculated and normalized to measured WT levels. As analysis of subsamples revealed no gender differences, male and female samples were combined throughout this investigation.

Light and electron microscopy
Mice lethally anesthetized with Avertin, were transcardially perfused with 2.5% glutaraldehyde + 0.5% paraformaldehyde in 0.1M sodium cacodylate buffer and cervical spinal cord and ON samples were collected.
Samples were postfixed overnight at 4˚C followed by rinsing with 0.1M sodium cacodylate buffer. A second post fixation was done with 1% osmium tetroxide followed by rinsing with ddH2O. Samples were dehydrated by incubation in increasing concentrations of acetone: 30%, 50%, 70%, 80%. 90% and 3X100%. Infiltration was done with 1:1, 2:1, 3:1 (epon:acetone) followed by embedding in epon and overnight polymerization at 60˚C. 0.5 um sections were stained with Toluidine blue and cover slips were mounted with epon for imaging.
Slides were imaged with 63X or 100X oil immersion objectives by light microscopy (Zeiss Axio Imager M1).

McGill facilities: Facility for Electron Microscopy Research (FEMR). McGill University and Genome
Quebec Innovation Centre.