BIOLOGICAL OVERVIEW

Frayed (Fray) is a serine/threonine kinase expressed by the peripheral glia of Drosophila, whose function is required for normal axonal ensheathment. Null fray mutants die early in larval development and have nerves with severe swelling and axonal defasciculation. The phenotype is associated with a failure of the ensheathing glia to correctly wrap peripheral axons. When the fray cDNA is expressed in the ensheathing glia of fray mutants, normal nerve morphology is restored. Fray belongs to a novel family of Ser/Thr kinases, the PF kinases, whose closest relatives are the PAK kinases. Rescue of the Drosophila mutant phenotype with PASK, the rat homolog of Fray, demonstrates a functional homology among these proteins and suggests that the Fray signaling pathway is widely conserved (Leiserson, 2000).

The larval peripheral nerves of Drosophila present an attractive system for analyzing the cellular interactions that underlie axon ensheathment and glial cell function. The nerves are relatively simple, having only three cellular components: (1) sensory and motor axons, (2) ensheathing glia that wrap the axons, and (3) perineurial cells that form the outermost cell layer. The glial ensheathment in Drosophila resembles that seen in neonatal mammals, where unmyelinated but glial-ensheathed axons predominate (Leiserson, 2000 and references therein).

In Drosophila, mutations that have been found to affect the ensheathment of peripheral axons include neurexin IV and gliotactin (gli). Neurexin IV is a transmembrane protein that localizes to pleated septate junctions, which are found in nerve-ensheathing cells of Drosophila and are required for the formation of the blood-nerve barrier. In neurexin IV mutants, pleated septate junctions fail to form, and the blood-nerve barrier is absent. The gli gene is expressed in the ensheathing glial cells and encodes a transmembrane protein with an inactive serine esterase domain, presumed to be a signaling receptor. In gli mutants, the ensheathment fails to encase the axons of the nerve bundle, giving support to the idea that Gli normally functions as a signaling molecule mediating a cell recognition event (Leiserson, 2000).

The major innervation of each larval abdominal hemisegment is provided by a single nerve that exits from the lateral edge of the ganglionic hemisegment and inserts into the bodywall at the edge of the ventral oblique muscle fiber 17, near the ventral midline of the target segment. Once the nerve reaches the bodywall, it subdivides into five branches running from ventral to dorsal. These branches correspond to the embryonically established projections of the intersegmental and segmental nerves within the bodywall. A second, minor innervation of the bodywall is provided by the segmentally repeated transverse nerves, which emerge from the ganglion's dorsal midline and project bilaterally. At the bodywall, the transverse nerve runs from ventral to dorsal along the segment border and terminates at the alary muscle. Each of the larval peripheral nerves in segments A2–A7 contain approximately 85–90 afferent and efferent axons. The most posterior peripheral nerve projects to the compound A8-11 hemisegment of the larva and as a result has a diameter about 50% larger than the other abdominal nerves (Leiserson, 2000).

The development of an abdominal peripheral nerve begins with the pioneering of the intersegmental and segmental axon tracts during stage 13 of embryogenesis. These two nerve tracts join at the edge of the CNS to form a single peripheral nerve trunk, which will elongate for the rest of embryonic and larval development. Ensheathment of the nerve involves two distinct cell migrations. The first involves glial cells from the CNS, which migrate outward onto the nerve and establish the axonal ensheathment. These cells are derived from the 'exit glia' clusters, which are located adjacent to the nerve roots. By stage 16, six to eight peripheral ensheathing glia are arrayed along the nerve and have begun to elaborate processes. The other migration involves the perineurial cells, which are thought to be of mesodermal origin and form the outermost cell layer of the nerve. By the end of the third instar, the larval peripheral nerves can exceed 3 mm in length. To accommodate this growth, the perineural cells undergo postembryonic proliferation, as revealed by bromodeoxyuridine birth-dating experiments. In contrast, the ensheathing glial cells do not proliferate postembryonically and accommodate nerve growth by extending progressively longer processes (Leiserson, 2000).

The only abnormalities observed by light microscopy in fray mutants are localized bulges or swellings in the nerves, inside of which extensive axon defasciculation is observed. This is referred to as the 'frayed nerve' phenotype. Detailed ultrastructural analysis has revealed that the fray phenotype also involves other, nonbulging regions of the nerve. To better understand the cellular basis of the fray phenotype, transmission electron microscopy (TEM) was used to compare wild-type and mutant nerves. A hallmark of wild-type nerves is that axons are usually fully ensheathed by glial processes. The glial processes normally remain in close apposition to the neurons as they encircle them and in this way enclose and isolate the axons. TEM sections of mutant larval nerves reveals extensive errors in the glial ensheathment of axons, defects that are observed in both the bulging and nonbulging regions of the nerve (Leiserson, 2000).

The nonbulging regions of nerves, which have normal nerve diameters and would thus appear normal by light microscopy, in fact have extensive defects when examined by TEM. The mutant phenotype in these regions consists of glial processes that have begun to extend circumferentially around the axons but have fallen short of fully wrapping them. The proportion of ensheathing glial processes that fully encircle axons in mutants averages only 49.2% ± 2.6% of the total length of glial processes, compared to over 85.0% ± 5.5% in wild-type nerves. Strikingly, the total length of the glial processes in mutants does not differ significantly from wild-type. Other glial errors include sites in the nerve where from three to seven glial processes converge to within half a micron but nevertheless fail to join up with each other. These aberrant sites are never observed in wild-type nerves but are seen in mutants an average of 2.75 cases per nerve section (Leiserson, 2000).

Like the nonbulging regions, bulging regions of mutant nerves contain disrupted glial ensheathment of axons, as revealed by TEM. Glial processes in some cases stretch out to contact the axons, but they consistently fail to encircle them. In addition to the glial defects, however, the bulging regions contained large, electron-transparent regions, which are presumably fluid filled. The axons are severely defasciculated, consistent with the light microscope observations. In contrast, the axons appear normal in number, though their average axon diameter is slightly reduced compared to wild-type axons. The reduced axon size is likely a systemic effect, since the mutant larvae at this stage are also smaller than wild-type. Otherwise, all of the cellular components of a normal nerve are present in the bulges: the outer perineurial cells, the ensheathing glia, and the axons. The perineurial layer in the bulging regions appears normal and uninterrupted, with a secreted neural lamella. In light of the glial cell defects observed in nonbulging regions, the most striking feature of the bulges is the disrupted glial ensheathment of axons. Indeed, besides the deformation introduced by the bulge, the mutant nerve phenotype is largely specific to the glial cells (Leiserson, 2000).

Whether the fray phenotype was associated with aberrant growth of the glia, especially in the bulging regions, was examined. Surprisingly, there was no statistically significant difference in the total length of glial processes (observed in individual transverse sections) between the bulging and nonbulging regions of mutant nerves. The glial length was also not significantly different from the value observed in wild-type nerves. These results suggest that the fray defect is not associated with an aberrant size of ensheathing glia that results in nerve bulges but rather is associated with a failure of the glial processes to fully surround and ensheath the axons, a defect seen irrespective of whether there is a bulge or not (Leiserson, 2000).

A simple interpretation of these results is that Fray participates in a signal transduction cascade within glial cells. In its absence, signal transduction is disrupted, resulting in the frayed nerve phenotype. The idea that Fray functions in a signaling pathway stems from the kinase domain within the sequence of the fray gene. That Fray is a bona fide Ser/Thr kinase follows both from the sequence, which contains all the hallmarks of a kinase domain, and from the strong homology to rat PASK (74% amino acid identity), which has been shown to have kinase activity in vitro (Leiserson, 2000).

Because glial defects are observed throughout the nerve, even in areas lacking pronounced bulges, the view is favored that the primary defect in fray mutants is incorrect wrapping of axons by glial cell processes. In this view, the bulges appear as a secondary consequence, perhaps from the lack of structural integrity caused by the improper wrapping of axons (Leiserson, 2000).

This model is consistent with reports of PF kinases associating with the cytoskeleton. Embryos that lack maternal fray have abnormal mitotic spindles, suggesting that Fray plays a role in the cytoskeleton of dividing nuclei (Petitt, 1998). Furthermore, rat PASK has been shown to associate with the cytoskeleton in vivo and to associate with the cytoskeleton in vitro when induced by stress (Tsutsumi, 2000). In light of these results, together with the finding that glial processes are malformed in fray mutants, it seems likely that, in glial cells, Fray also regulates some aspect of the cytoskeleton. However, Fray does not simply regulate the overall size of the glial cell but rather specifically affects the ensheathing behavior of glial cell processes, in that they fail to (completely) encircle axons (Leiserson, 2000).

Some clues to Fray function are evident from the sequence. The Fray kinase domain is related to those of PAK kinases (see Drosophila PAK-kinase), which have been shown to effect changes in the cytoskeleton in response to the stress-activated pathway. Upon activation by GTPases of the rho family, PAKs mediate changes in cell shape by acting on components of the cytoskeleton or by causing changes in gene expression. However, Fray and other PF kinases lack the motif found in canonical PAKs that permits binding to rho GTPases. Thus, PF kinases probably do not interact with these GTPases and probably have distinct molecular partners and functions (Leiserson, 2000).

By analogy to other kinases, the regulatory region of Fray is likely to be in the C-terminal half of the protein. In this same segment are found the two PF domains, which are stretches of homology among PF kinases. These PF domains are likely to represent regions that mediate protein-protein interactions, serving to regulate the activity or localization of the kinase. Consistent with this idea, it has been reported that PASK can specifically associate with F-actin via a region that includes the PF1 domain, presumably through an intermediary protein (Tsutsumi, 2000). Whatever proteins PASK interacts with, the fact that it can complement fray suggests PASK can interact with the appropriate Drosophila proteins via its regulatory and catalytic domains (Leiserson, 2000).

While the biological roles of the mammalian homologs of Fray have not been reported, these studies indicate there are multiple functions of Fray in Drosophila. In addition to the glial cell function in nerves, Fray has a maternal role, a vital function, and roles in imaginal development suggested by the rescue experiments (Leiserson, 2000 and references therein). In a selective rescue experiments, the lethal and nerve phenotypes were differently affected. For example, a heat shock driver rescues larval lethality without fully rescuing the nerve phenotype. These experiments show that the failure of mutant larvae to survive is not due solely to their frayed nerves. That Fray has distinct roles in imaginal development follows from the observation of abnormal wings, legs, eyes, and bristles in mutants that have been partially rescued and survived to adult. Thus, the signaling pathway or pathways in which the Fray kinase participates appear to be employed in multiple times and places in Drosophila development (Leiserson, 2000).

In mammals, the wide expression of Fray homologs points to multiple roles of PF kinases. In the rat, most adult tissues express PASK, including the nervous system, kidney, adrenal glands, and stomach. In mouse and human embryos, cDNAs have been isolated from developing brain and heart. In adults, Fray homologs are found in mammary gland and thymus. It is not yet known if PF kinases are involved in mammalian axonal ensheathment or if they are expressed by Schwann cells. In light of the functional homology found between Drosophila Fray and mammalian PASK, however, it seems clear that PF kinases play essential roles in mammalian as well as arthropod development (Leiserson, 2000).

GENE STRUCTURE

Exons - 3

PROTEIN STRUCTURE

Amino Acids - 552

Structural Domains

Searches of the standard databases reveal that the Fray kinase domain is most closely related to the PAK family of kinases, whose founding members are yeast STE20 and the mammalian PAKs. These molecules are activated upon binding small GTPases of the rho family. Comparative sequence analysis hasrevealed a family of six known Fray-related proteins (including Fray) with similarity scores to each other that far exceed scores to other PAK-related kinases: rat PASK (accession number D88190, human OSR1, mouse SPAK, human SPAK or Strerile20-like, and Caenorhabditi elegans CAB6094. This family is referred to as PF kinases, for PASK and Fray, since previously the best-characterized member of this family is rat PASK (Leiserson, 2000).

Classification of these six proteins into one family distinct from the PAKs is based on several criteria. The regions of homology among PF kinases span their length, while homology to the canonical PAKs is limited to the kinase region. PAKs share a GTPase binding motif in the N-terminal region and a C-terminal kinase domain, while all PF kinases have no recognizable GTPase binding motif, and their kinase domains are all located in the N-terminal region. Finally, the homology between the kinase domains of PF kinases and true PAKs is modest, at the 30% level of identity, while that between Fray and its closest mammalian relatives is high, at 75% amino acid identity (Leiserson, 2000).

The PF kinases possess two regions of homology outside the kinase domain, which have been denoted PF1 and PF2 and that may correspond to two functional domains of the proteins. These domains may represent regulatory or targeting elements, by analogy to other kinase molecules. Queries of the standard databases using BLAST and PSI-BLAST reveal no other known proteins other than the ones listed here that have significant homologies to these regions (Leiserson, 2000).

The relationships among members of the PF kinases and PAK families can be visualized using a cladogram based on sequence distances of the kinase domains. The true PAKs and PF kinases cluster into distinct groups, with mammalian and Drosophila members in each. This suggests that both a Fray-like and a PAK-like protein were present in a common ancestor of insects and mammals. For example, the two Drosophila proteins Fray and DPAK are considerably closer to their respective mammalian homologs, PASK and PAK, than they are to each other (Leiserson, 2000).

date revised: 30 June 2001

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.