Tuan Le

Analysis of fibroblast growth factor 2 null (FGF2−/−) and wild-type (FGF2+/+) mice was used to interpret the potential in vivo role of endogenous FGF2 on oligodendrocyte lineage cell (OLC) responses during oligodendrogenesis and myelination. In wild-type mouse spinal cord, FGF2 levels increased approximately threefold between the first and second postnatal weeks, a period corresponding with the peak of oligodendrogenesis. Absence of this developmental FGF2 elevation in FGF2−/− mice eliminated the transient overproduction of oligodendrocytes that is known to occur at the peak of oligodendrogenesis in wild-type mice. Absence of FGF2 did not affect oligodendrocyte progenitor (OP) density or proliferation, based on BrdU incorporation, and also did not alter survival, based on TUNEL analysis. To examine OLC differentiation in vivo, retrovirus encoding-enhanced green fluorescent protein (GFP) was injected into the spinal cord to heritably label endogenous cycling cells in the white matter at postnatal day 7 and then identify the generated cells at postnatal day 28. Phenotypes of cells expressing GFP were identified by morphology and immunolabeling, using CC1 for oligodendrocytes and NG2 combined with platelet-derived growth factor α receptor for OPs. Within the population of GFP-labeled cells, the proportion of oligodendrocytes was higher in FGF2−/− mice, indicating that endogenous FGF2 inhibited OLC differentiation in wild-type mice. Furthermore, in FGF2−/− mice fewer cells appeared to be generated from an initial retrovirus-labeled cell, consistent with more frequent differentiation into post-mitotic oligodendrocytes. This in vivo analysis demonstrates that the predominant role of endogenous FGF2 on OLCs in development is inhibition of differentiation. © 2004 Wiley-Liss, Inc.

Decreased axial (λ||) and increased radial (λ⊥) diffusivity have been shown to reflect axonal and myelin injury respectively. In the present study, evolving white matter injury within the optic nerves of mice with retinal ischemia was examined by in vivo and ex vivo measurements of λ|| and λ⊥. The results show that at 3 days after retinal ischemia, a 33% decrease in vivo and a 38% decrease ex vivo in λ|| without change in λ⊥ was observed in the injured optic nerve compared to the control, suggestive of axonal damage without myelin injury. At 14 days, both in vivo and ex vivo measured λ⊥ increased significantly to 220–240% of the control level in the injured optic nerve suggestive of myelin damage. In contrast, the axonal injury that was clearly detected in vivo as a significantly decreased λ|| (33% decrease) was not as clearly detected by ex vivo λ|| (17% decrease). The current findings suggest that ex vivo λ⊥ is comparable to in vivo λ⊥ in detecting myelin injury. However, the structural changes resulting from axonal damage causing the decreased in vivo λ|| may not be preserved ex vivo in the fixed tissues. Despite the accurate depiction of the pathology using λ|| and λ⊥ in vivo, the use of ex vivo λ|| to extrapolate the status of axonal injury in vivo would require further investigation.

Myelin damage, as seen in multiple sclerosis (MS) and other demyelinating diseases, impairs axonal conduction and can also be associated with axonal degeneration. Accurate assessments of these conditions may be highly beneficial in evaluating and selecting therapeutic strategies for patient management. Recently, an analytical approach examining diffusion tensor imaging (DTI) derived parameters has been proposed to assess the extent of axonal damage, demyelination, or both. The current study uses the well-characterized cuprizone model of experimental demyelination and remyelination of corpus callosum in mouse brain to evaluate the ability of DTI parameters to detect the progression of myelin degeneration and regeneration. Our results demonstrate that the extent of increased radial diffusivity reflects the severity of demyelination in corpus callosum of mouse brain affected by cuprizone treatment. Subsequently, radial diffusivity decreases with the progression of remyelination. Furthermore, radial diffusivity changes were specific to the time course of changes in myelin integrity as distinct from axonal injury, which was detected by βAPP immunostaining and shown to be most extensive prior to demyelination. Radial diffusivity offers a specific assessment of demyelination and remyelination, as distinct from acute axonal damage.

Repair of myelin damage in the adult CNS requires oligodendrocyte progenitor (OP) proliferation and subsequent differentiation into remyelinating oligodendrocytes. Platelet-derived growth factor (PDGF) and fibroblast growth factor-2 (FGF2) have been predicted to act individually and/or cooperatively to generate remyelinating oligodendrocytes. Analysis of PDGF alpha receptor (PDGFαR) heterozygous (+/−) mice indicates that PDGFαR expression modulates oligodendrocyte density in non-lesioned adult CNS. Analysis of cuprizone demyelination and recovery in PDGFαR+/− mice, FGF2 knockout (−/−) mice, and PDGFαR+/− FGF2−/− mice demonstrated that:( 1) OP proliferation and oligodendrocyte regeneration is impaired in PDGFαR heterozygotes, (2) PDGFαR+/− and FGF2−/− deletions do not act cooperatively to impair OP amplification, (3) oligodendrocyte differentiation is more frequent in FGF2−/− mice, and (4) FGF2 deletion in combination with the PDGFαR+/− genotype rescues impaired oligodendrocyte regeneration of PDGFαR heterozygotes. These findings demonstrate distinct roles for PDGF and FGF2 in vivo in the context of a demyelinating disease with spontaneous remyelination.

ABSTRACT Visualization of high-dimensional data such as text documents is widely applicable. The traditional means is to find an appropriate embedding of the high-dimensional representation in a low-dimensional visualizable space. As topic modeling is a useful form of dimensionality reduction that preserves the semantics in documents, recent approaches aim for a visualization that is consistent with both the original word space, as well as the semantic topic space. In this paper, we address the semantic visualization problem. Given a corpus of documents, the objective is to simultaneously learn the topic distributions as well as the visualization coordinates of documents. We propose to develop a semantic visualization model that approximates L2-normalized data directly. The key is to associate each document with three representations: a coordinate in the visualization space, a multinomial distribution in the topic space, and a directional vector in a high-dimensional unit hypersphere in the word space. We join these representations in a unified generative model, and describe its parameter estimation through variational inference. Comprehensive experiments on real-life text datasets show that the proposed method outperforms the existing baselines on objective evaluation metrics for visualization quality and topic interpretability.