Olfactory Ensheathing Cells and the Structure of the Primary Olfactory Projection

Olfactory ensheathing cells (OECs) are the glial cells that accompany axons from the olfactory epithelium to their targets in the olfactory bulb. They possess unique features that have made them the subject of much study as being potentially useful in cell-based therapeutic approaches to CNS repair. Investigation of OECs has demonstrated antigenic and morphological heterogeneity in their population. No marker specific to and selective for OECs have yet been identified, and many of the markers used are variably expressed by other glial cells. Even among OECs, these markers appear to vary in vivo (depending on their anatomical location, contact with other cells, and developmental timing) and in vitro. The variation across the population of OECs has compromised their isolation and characterization. It has also made the task of identifying meaningful subpopulations - with greater or lesser therapeutic utility - dependent on identifying the source of their variability. Such information would aid in both the harvest and experimental manipulation of OECs to optimize their therapeutic effect. One way to understand the nature of this variability is to seek its potential causes in vivo. This must begin with an examination of the structure of the olfactory nerve. Here, the structure and development of the primary olfactory projection are thoroughly reviewed with an emphasis on OECs and the cells with which they make contact. The relevant experimental results are also discussed. The weight of anatomical evidence indicates that the structural variations described in different locations and across species are mostly the result of spatiotemporal developmental factors. As such, the formation of the olfactory projection is mediated primarily by the source tissue, the olfactory epithelium. Cell-autonomous development is common elsewhere, and suits the evolutionary age and importance of olfaction, as well as its continued regenerative capacity. Our findings provide a more systematic anatomical understanding of this nerve. That understanding indicates that variation in axon ensheathment by OECs is an inherent feature arising from the flexibility of the ensheathing program. This perspective, along with the anatomical data reviewed here, can inform more carefully controlled laboratory investigations designed to uncover the detailed mechanisms governing OEC biology.

iii I would like to thank Dr. James Schwob  Of the many researchers whose work I have here reviewed, I would like to pay special respect to Dr. Ron Doucette of the University of Saskatchewan, Saskatoon, Canada. His extraordinary diligence is to be admired.
iv  and each bulb has two glomerulione medial and one lateralinnervated by axons of OSNs that express the same OR (Mombaerts, 1996).
This review focuses on the glial cells -olfactory ensheathing cells (OECs) -that accompany the primary olfactory axons from their origin in the epithelium to their synapses in the glomerular layer. As such, the structural details of the deeper laminae of the bulbthe external plexiform, mitral cell, internal plexiform, and granule cell layerswill not be discussed. There exists debate as to which cells are encompassed by the term "OECs". Some propose that, based upon morphological or antigenic variations, certain populations of these cells particularly in the bulbshould be regarded as distinct subtypes. Lineage relationships are a tractable matter, but there is an inherent difficulty in defining cellular categories based upon features that may or may not be plastic. Thus, for the purpose of the following structural description, the broadest definition of OECs will be used, that is: those cells that ensheath the olfactory axons from the epithelium to the glomeruli.
The manner of this ensheathment is characteristic of OECs. The classic description is of a cell with a peripheral, irregular nucleus and cytoplasmic processes that form a "tube" around a large number of axons. These processes also branch internally to incompletely divide the channel into compartments with multiple axons per mesaxon (Fraher, 1982). The arrangement of axons in direct contact along their longitudinal extent is unique in the adult nervous system; it is only seen elsewhere during embryonic development (Gasser, 1956). The maintenance of an embryonic-like arrangement of axons is one of the reasons for which OECs were initially considered as important to the regenerative capacity of the olfactory system.  The olfactory epithelium is a pseudostratified epithelium with cellular somata organized into three general compartments. The perikarya of sustentacular and microvillar cells occupy the most apical compartment. These cells serve to phagocytose debris (Suzuki, 1996) The axons of neighboring OSNs coalesce into small bundles as they descend towards the basal lamina ( Figure 34; Cuschieri & Bannister, 1975b). Once the basal lamina is reached, these axons turn and travel between the HBCs and the basal lamina to a greater or lesser extent before turning again to exit the epithelium ( Figure 5). This was noted early-on by Gasser (1956), "Another point, revealed by the Golgi prints, is the probability that the axon will cross the basement membrane, in terms of electron microscope dimensions, at enormous distances from orthogonal projections of cell bodies." The HBCs form spaced hemidesmosomes that tether them to the basal lamina and serve as anchors for the "HBC arches" that cover the underlying bundles of axons ( Figure 6, Figure 7, Figure 8).     In other instances, HBCs extend processes into the lamina propria in order to "hand off" bundles of axons to the underlying OECs (Herrera, 2005). This transitional region is thus a flexible one where OEC or HBC processes may extend into either compartment to ensure ensheathment of the axons.
The proximal spatial association of OECs and axons within the epithelium corroborates data from developmental studies that show an equally early temporal association. The first axons lead as they breach the epithelial basal lamina, and are then followed by migratory cells (Farbman & Squinto, 1985;Marin-Padilla & Amieva, 1989). Some of these migrating cells are considered to be OEC progenitors (pOEC) since they migrate withalthough laggingthe growing axons, and neighbor them along their length. This relationship later develops into the 7 characteristic manner of OEC ensheathment (Fraher, 1982;Cuschieri & Bannister, 1975b;Doucette, 1989;Valverde, 1992).
In certain histological sections, some of these migrating cells have been observed with a "trailing tail" still in the epithelium (Farbman & Squinto, 1985;Marin-Padilla & Amieva, 1989).
It is possible that these cells are the pOECs who, by way of maintained association with the axons, develop into the adult OECs observed within the epithelium. This might explain the colocalization of intraepithelial OECs to regions near axon exit, as well as their relative sparseness.
Though it is possible that OECs from the lamina propria encroach into the HBC archways, there are two reasons why the above hypothesis of an early intraepithelial association is more attractive. First, it does not seem necessary for OECs to extend in a manner retrograde to the direction of axonal growth to accomplish an ensheathment that HBCs capably achieveas demonstrated by those HBC processes that descend into the lamina propria to deliver the axons to OECs (Herrera, 2005 & Au, 1991;Norgren, 1992). Given that OECs maintain their association with the axons at all stages, it is not unreasonable to assume that this began in some cases while still within the epithelium.  Cuschieri & Bannister, 1975b). When these axons reach the developing telencephalic vesicle, they again breach the early glia limitans -another basal laminar structureto the exclusion of any other cell type (Doucette, 1989;Valverde, 1992;De Carlos, 1995). So on structural grounds, it is likely the axons have a key role in the progress of olfactory nerve development.

-What pioneers the pathway: OECs, olfactory axons, or both?
The opposite conclusion was drawn from a study of newborn rat, newborn opossums and 64-day-old opossum tissue (Tennent & Chuah, 1996). Here it was reported that the processes of OECs clearly extended further toward the telencephalon than the axon tips, leading their growth out of the epithelium and through the lamina propria ( Figure 10). The study routinely substituted spatial relationships in the histological sections for temporal ones (see caption, Figure 10). They describe a section containing a nerve bundle near the epithelium as having recently emerged from it, despite the fact that only a single section of a 64-day-old opossum is described. Though this type of temporal metaphor is conventional when describing spatial pathways, their subsequent discussion reveals it as a conflation used to lend temporality to the observation of OEC processes extending ahead of the axon terminals. This observation itself is not convincing and may be artefactual as the axon bundle pictured exhibits a number of orientations along its length, and may have simply exited the plane of section rather than being at its terminus. Without a serially-sectioned demonstration of the distal limit of the axons, the whole premise collapses, and renders the data inadequate to support the idea that OECs lead the growth of olfactory axons to the bulb.
Numerous in vitro assays have demonstrated that axons preferentially grow on a substrate of OECs (Tisay & Key, 1999;Chuah & Au, 1994;Goodman, 1993). Some have taken this as support for the aforementioned in vivo observation to conclude that OECs pioneer axon growth in the olfactory pathway (Tisay & Key, 1999). It should not be surprising that OECs promote axon growth as this is one of their primary functions in established olfactory nerves: to support the extension of axons by immature OSNs. But this cannot be taken as implicit support of the idea that OECs pioneer their embryonic development. The earliest source tissue utilized in these in vitro assays was rat E19.5 (Tisay & Key, 1999), which is much later than the time points during which the first axons leave the epithelium -around E10-11 in the mouse (Cuschieri & Bannister, 1975b;Doucette, 1989), and E12-12.5 in the rat (Valverde, 1992;Treloar, 1996). The early embryonic mesenchyme is rich with axon growth-promoting matrix molecules that diminish postnatally (Julliard, 1998). So the earlier tissues are highly permissive on their own, while the late embryonic to postnatal tissues present a different set of cellular conditions than those present during the ontogeny of the pathway. Along the same lines, even if Tennent & Chuah's (1996) study is valid and reflects OECs leading the growth of axons it, too, was limited to postnatal animals. During embryonic development the OECs at first are loosely associated with the axons and only gradually begin to ensheath them (Fraher, 1982;Gong, 1994;Cuschieri & Bannister, 1975b), before producing basal lamina at a later stage. In the rat, this basal lamina appears just before birth (Fraher, 1982;Julliard, 1998). Observations of OECs situated distally in the developing pathway of the mouse show a similarly timed appearance of the basal lamina that forms the new glia limitans of the bulb (Doucette, 1989). Thus, during their early embryonic growth the axons and OECs are exposed to the broader extracellular environment, but their exposure ends at a specific time. This change is quite likely reflected in the manner in which new fascicles would grow after this period. This is one reason that observations of postnatal fascicle growth may not be relevant to their ontogeny.
Another study oft-cited as support for the hypothesis that OECs pioneer the olfactory pathway claims that soluble factors from the bulb attract OECs (Liu, 1995). In other words, factors ostensibly emanate from the developing bulb and attract a "carpet" of OECs from the epithelium upon which the axons then extend. The evidence for this interaction came from in vitro experiments in which bulb or cerebral explants are placed on a monolayer of purified, DiIlabeled OECs. In this setting, the OECs migrated only towards the bulb explant ( Figure 11).
For purposes of studying the initial establishment of the olfactory pathway, the methods employed were suboptimal. First, explants were obtained from P1-3 rats. At this stage, the bulb has already developed and been contacted by a robust population of axons and OECs. To make any claims about primary development, it would have been necessary to utilize the portion of the telencephalon that develops into the bulb before it has been contacted by any axons or OECs. In preparing the bulb explant, they manually removed the ONL, which is rich with OECs, and claim that the DiI-labeled OECs are not migrating towards residual OECs. However, use of p75 NTR as a marker for OECs demonstrates that a small number of them remain despite removal of the ONL.
Since p75 NTR is only expressed by OECs on the outermost layers of the ONL (Au, 2002;Franceschini, 1996;Astic, 1998;Treloar, 1999), it is highly likely that a greater number of OECs remained within the explant. Moreover, even if all of the OECs had been removed, the findings cannot preclude the possibility that the explant itself was conditioned by its prior contact with OECs and axons. For the same reason, an OEC chemotaxis assay that this group performed utilizing bulb-conditioned medium is also inconclusive. In fact, recent studies have shown that OEC proximity and OEC-secreted factors such as GDNF do indeed increase OEC mobility (Windus, 2007;Cao, 2006). In conclusion, in order to study whether soluble factors from the bulb could be responsible for the primary development of the olfactory nerves, it would be necessary to utilize the region of the telencephalon that will develop into the bulb before it has been populated by OECs.  Taken together, the available observations point to a tropic role for the axons and a trophic role for the OECs during ontogeny; the axons pathfind and "pioneer", while the OECs migrate alongside and support their growth. It remains possible that the cellular environment in postnatal animals is such that OECs must completely enclose bundles of axons, and as such could appear to lead their growth. At this stage, however, the bulb has already developed and, as Liu (1995) demonstrated, contains some source of tropic factors which influence OECs. Since even in this scenario, OECs have not been observed migrating towards the bulb far ahead or in the absence of axons, it may be that during postnatal axon growth the OECs must isolate the growing front of axons, but it is still the axons that are required for extension or pathfinding.

-Beyond the Epithelium
After the small bundles of axons exit the epithelium, they continue to merge into larger bundles in a spatially-defined manner (Chehrehasa, 2006 increasing association with distance from the epithelium as well as their growth in advance of the OECs during development.

-Lamina Propria
In the lamina propria, the manner of OEC ensheathment is similar to that seen in the epithelium, but here, no axons are left unensheathed by OECs. Every bundle of axons exiting the epithelium is completely surrounded by the cytoplasm of one or more OECs to form an EOB.
Throughout the depth of the lamina propria, the outside of each EOB is apposed to and Elsewhere in the literature, the arrangement of axons, OECs, and adherent basal lamina is sometimes referred to as a fascicle. Inconsistent usage of the terms bundle, fascicle, fiber, and nerve, has complicated translation of anatomical details across primary sources. In keeping with the usage for other nerves and skeletal muscle, the term "fascicle" will be reserved for the anatomical unit that includes the connective tissue. This will help clarify the discussion of the collagenous and olfactory nerve fibroblast (ONF) investment around and upon LEOBs.
However, for better or for worse, the verb "fasciculate" will continue to be used conventionallythat is, the longitudinal merging of axons, irrespective of other surrounding or intervening elements. Thus, axons may fasciculate, but the resultant structure is not necessarily a fascicle.   16C). Illustration of the complex interdigitations between adjacent OECs (S1, S2). There are in reality no intercellular clefts as depicted, but rather abutment and overlap (see Figure 12).
The LEOBs take over for the EOBs within the epithelium and are often a continuation of these. As a consequence, EOBs in the epithelium are usually comparable in size to the LEOBs found at the most superficial levels of the lamina propria (compare Figure 7 and Figure

-LEOB Merger
The LEOBs increase in size as they descend deeper into the lamina propria, reflecting their continued convergence ( Figure 15, Figure  While remaining discrete cells, these cooperating OECs produce basal lamina only on the external surfaces of the LEOB as a whole ( Figure 17). The spatial merger of axons is thus accompanied by this merger of OECs with their external basal laminae, resulting in a reduction in the number of discrete LEOBs as the bulb is approached (Julliard, 1998).

-How would cooperative ensheathment arise?
The Gradually, these transitional pOECs (p/OEC) begin to thread cytoplasmic processes around and through the bundles of axons. Fraher (1982) provides a highly detailed description of the establishment of this ensheathment. Each bundle may receive processes from multiple neighboring OECs, and each OEC may send processes to multiple bundles. Then: "By 20 days, most vomeronasal nerve branches had become subdivided into a number of discrete axon bundles, each surrounded by a complete cytoplasmic sleeve. The sleeves of adjacent bundles had become separated from one another, over an increased proportion of their circumference, by basal lamina which was present as clumps of wispy material over parts of each bundle's circumference. Small amounts of interfascicular collagen were sometimes present in such areas. However, basal lamina remained absent over segments of the circumference of all bundles. Here, the sleeve processes of adjacent bundles were closely apposed to one another and occasionally passed from one sleeve to another." (Fraher, 1982, pg. 155) Thus, there is a type of critical period before which a p/OEC gropes its way toward any axons in its vicinity. After this period, the ensheathment of individual axon bundles tightens, and a surrounding basal lamina appears. This is the point at which neighboring OECs contributing to the same bundle become incorporated within a single LEOB. Indeed, Fraher notes the occasional presence of multiple OEC nuclei within certain LEOB sections at this stage. The exact timing of this stage cannot necessarily be extrapolated from the vomeronasal nerve to the main olfactory nerves, only that its manner of development is likely similar. The emergence of mature basal lamina around the main olfactory nerves in rats is complete at E17-E18 (Julliard, 1998).
The reason that cooperative ensheathment is more pronounced at points further along the course of the olfactory nerves towards the bulb is the result of spatial merger during this early period. The early axon bundles exit from the epithelium accompanied by a contingent of pOECs from the same area. Bundles merge as they approach the bulbthe axons converge, and so do the pOECs that accompany them. We can consider the axons and their accompanying pOECs as a kind of migratory unit. Compared to the vomeronasal projection, the main olfactory projection in mammals has a larger epithelial area and is relatively closer to the bulb, so many of these migratory units will not converge until they come into the immediate vicinity of the bulb. There is thus an inherent separation between many of the olfactory migratory units at superficial depths that is not as pronounced in the vomeronasal nervewhich has a smaller sensory surface and a more distant target. This is a hypothesis to explain the unique manner of ensheathment observed in many olfactory fascicles. It utilizes limited spatial and temporal observations in an attempt to understand the variety of extant formations. It may or may not be causally true. It is nevertheless natural, when sketching out a physical history, to impart some sort of narrative hypothesis. This not only helps to structure the facts, but provides something that may be tested.

-Olfactory Nerve Fibroblasts
One or more LEOBs are cordoned into fascicles by thin encircling layers of perineurial olfactory nerve fibroblasts (ONF) that express basal lamina on various surfaces. The degree of perineurial encirclement appears to vary between species: mice and rats have a few thin layers In the absence of detailed data regarding the position of a fascicle in the lamina propria in other species, both of these generalizations about the primate olfactory nerves are tentative. If it is not merely fascicle size and depth that accounts for multi-layered perineurial ONF encirclement and the presence of endoneurial ONF processes, these inter-species differences may be the result of more complicated mechanisms -for instance, macrosmatic/microsmatic genetic differences -that may then be relevant to regenerative potential and model system generalizability.
One possibility may be explored by considering the human and macaque as equivalentboth primatesbut contrasting the pictured human ( Figure 19) and macaque (      Holbrook, 2005, Fig. 6B). Human LEOB from Figure 19. Hypertrophic OEC adjacent to endoneurial ONF cytoplasm. Scale bar = 1 µm. Figure 22 (Schwob, 1992, Fig. 10C). Close-up of original print in Figure 18 (top-center) demonstrating endoneurial collagen separating adjacent rat LEOBs. The sharp vertical line is a manual crop mark.

-Fascicle Merger
Kawaja (2009)   The merging of fascicles and LEOBs that culminates at the cribriform plate seems to indicate a gradual process dictated by growth and spatial proximity rather than one that is governed and tightly regulated. It has been shown across 16 species that the area of cribriform plate perforated by olfactory nerves is directly proportional to the area of the olfactory epithelium (Pihlstrom, 2005). Furthermore, a study of bats revealed thatall else being equallarger bats generally had fewer distinct foramina (Bhatnagar, 1974). Thus, a greater area of olfactory epithelium results in a greater area of the cribriform plate being perforated, and a greater distance between the epithelium and the plate allows the fascicles to merge more, resulting in fewer but larger foramina.
All of the morphological permutations observed in the olfactory fascicles are thus likely the result of the sequence in which the pioneering axons converge towards the telencephalon, OECs establish ensheathment, and ONFs establish neuria 3 ; all three overlap and are subject to the complicated topography of the epithelium from which the axons arise. Our hypothesis might be tested by a detailed quantitative analysis of the fiber pattern across the epithelial sheet towards the bulb.
Of the animals considered, the macaque cribriform plate was reported to be unusual due to the presence of a single large foramen as opposed to multiple punctuate openings (Herrera, 2005). It may have some bearing on the types of mergers achieved in the macaque olfactory nerve: after a certain depth in the lamina propria, these fascicles are described as increasing in size only by the incorporation of a greater number of distinct LEOBs whose size is not described to increase further than the "500 or more" axons observed more superficially in the lamina propria. This observation was made in passing, so their analysis may neglect the size of the absolute deepest fascicles, but it does raise the question of how progressive merger of fascicles or ONF layering relates to the development of other structures such as the cribriform plate.
Nonetheless, the processes of LEOB and fascicle merger will reach an absolute maximum at the ONL, with all fascicles and LEOBs fusing into this single structure. Figure 24 (Kawaja, 2009, Fig. 2). Tracing of individual OEC cytoplasm in mouse, rat, and cat.

-Are there interspecies differences in ensheathment style?
It has been suggested by  that there may be interspecies differences among OECs in relation to the location of their nuclei, and degree of cooperative ensheathment within EOBs in cross-section ( Figure 24). It is implied that these may reflect other phenotypic differences relevant to the isolation and experimental manipulation of OECs.
In the literature, OEC nuclei are generally described as being situated at the periphery of EOBs, though they are also noted to be seen elsewhere (Boyd, 2003;. Similarly, the cytoplasm of a single OEC can be observed to completely surround a bundle of axons to form an EOB, or to coordinate processes with neighboring OECs to do so. Thus, the overall impression is that permutations of both variablesnuclear location and cooperative ensheathmentexist. This impression has not been systematically confirmed, however, within a given species.

-Interspecies Cooperative Ensheathment
Kawaja (2009)    It is also important to note that it is probably not strictly accurate to say that an OEC can accomplish complete ensheathment of a bundle because the interface between two longitudinally-adjacent OECs is not a flat plane, but likely involves some cytoplasmic investment from the neighbor when histologically processed into cross-sections ( Figure 14). It is always a matter of degree, and thus not likely to be a significant matter of phenotype; this is in contrast to myelinating Schwann cells, for instance, that have well-demarcated nodes separating them.

-Interspecies Nuclear Location
Regarding nuclear location, Kawaja (2009) also draws attention to OECs in the cat as having numerous processes radiating away from a centrally-located nucleus ( Figure 24). This arrangement is also highlighted in another publication from the same group ( Figure 25). This  Figure 25, Figure 26). They do not state the frequency with which the central morphology is observed, nor offer any discussion, other than to imply by presentation that this may be a phenotypic feature of OECs in cats.
A review of the literature reveals that nuclear localization is likely incidental. OEC nuclei in cats may also be seen peripherally (as above). In rats, not only do multiple OECs often comprise each LEOB, but their nuclei are also seen centrally ( Figure 15, Figure 27, Figure 28, Figure 29).   Certain examples illustrate how the distinctions between central and peripherally-located nuclei are often unclear ( Figure 30, Figure 31, Figure 32). Such ambiguity reduces the likelihood that nuclear location is phenotypically determined, and increases the likelihood that it is the result of dynamic variables such as would occur during growth and development.
This point is reiterated in Figure 27, which aptly summarizes the variety of morphologies that may be present in olfactory fascicles and clearly can contain multiple OECs with nuclei in multiple positions. Due to its great depth, it suggests that both these morphological features are primarily dependant upon spatiotemporal variables and not strict considerations of OEC phenotype.

-How is OEC nuclear location determined?
Just as with the emergence of cooperative ensheathment, the manner of axon and pOEC convergence from the placode toward the embryonic telencephalon indicates a possible mechanism for the variety of OEC nuclear localizations observed in EOBs along the length of the projection. In this case, reference will be made to examples as a means of describing possible spatiotemporal mechanisms that give rise to particular observations. These are: (1) peripheral nuclei in small/superficial EOBs, (2) peripheral nuclei in large/distal EOBs, (3) central nuclei in small/superficial EOBs, (4) multi-localized nuclei in very large/distal EOBs.

Peripheral nuclei (Examples 1 & 2)
The most frequent observation of peripherally-located nuclei is likely due to the inherent tendency of axons to tightly fasciculate proximally and maintain that association as they extend,

"Ultrastructural detail of the exit of a few olfactory nerve axons (small ax) through a small opening in the basal lamina (large arrowhead) of the nasal epithelium. Basal lamina anomalies such as bends, blebs, and rolls (inset) are often found at these openings. The intact epithelial basal lamina is outlined by small arrowheads. Also illustrated is a submucosal olfactory nerve fascicle (large ax) surrounded by olfactory sheath cells (os). Coated vesicles (inset), mitochondria (m), rough endoplasmic reticulum (rer), Golgi apparatus (go), and the magnification (horizontal bars) of the figures are also illustrated."
Imagine a hypothetical scenario in which all of the axons from the epithelium have completed development before any OECs ensheath them. So, all of the axons have extended towards the bulbhaving merged and fasciculated in a spatially-defined manner along the waybefore any OECs are introduced. In this scenario, all of the EOBs would exhibit peripheral nuclei. In reality, the reason that it is only most EOBs that have peripheral nuclei is that this is just partially the case; the OECs only slightly trail the axons, and gradually increase their association along the way. So the predominance of peripheral nuclei is the result of spatial factors -the exemplar of which is the hypothetical "OEC-free" extension of axons. Then, the deviations from this may be the result of temporal factors that are described below.

Central nuclei (Example 3)
The emergence of centrally-located OEC nuclei in more superficially located portions of the projection may relate to the continued development of the early olfactory epithelium after the initial axons have exited. The pOECs are thought to emerge "in response to the presence of axons in the underlying submucosa" (Marin-Padilla & Amieva, 1989). Meanwhile, the epithelium continues to expand and new axon bundles continue to exit from it. The pOECs will migrate with the bundles following which they emerge, but will nonetheless also be present in the mesenchyme ahead of subsequent axons exiting from new points in the epithelium. Since in these early stages pOEC will send processes to ensheath any axons in their vicinity (Fraher, 1982;Cuschieri & Bannister, 1975b), such radial dispositions could be produced when the OEC ensheath axons on all sides. This is a temporally-based explanation that results from the inversion of the usual sequencei.e., with pOECs preceding the axons. The central disposition then arises by chance, perhaps aided by the greater surface area from which the axons emerge around the waiting pOECs that precludes the possibility of their total fasciculation. Such a mechanism better explains the central nuclei in more superficial EOBs (Figure 15, cat in Figure 24, purple EOBs in Figure 25), as it arises when the later-stage axons enter into the superficial region containing pOECs.
That largerthus deeper -EOBs tend to exhibit peripherally located nuclei returns to the extent to which axons nonetheless lead the growth and mitosis of pOECs. So, the same EOB with a central nucleus in the subepithelial region may have a peripheral nucleus more distally ( Figure 26) because the leading axons will fasciculate ahead of that pOEC's mitosis.
Presumably, no bundles of axons leave the region containing pOECs without acquiring their own contingent, thus the normal order of axons leading the migration of pOECs is restored distally and the peripheral nucleus is again made predominant.

Multi-localized nuclei (Example 4)
These potential mechanisms must all square with the continuing merger of migratory units leading to the increasing degree of cooperative ensheathment as the bulb is approached. For example, if two large migratory units with peripherally-located pOECs merge, some of these pOECs may be situated between the two and continue to divide along the length of the resultant bundletheir progeny essentially becoming entrapped centrally. They may be admitted within such large bundles because of their sheer size, and/or because as time passes (and thus in more distal regions) the migratory pOECs become more tightly associated with the axons and so are less likely to abandon surface contact with the axons as the two continue to extend and merge with other bundles.

-Primacy of Merger
Ultimately, the close examination of the data suggests that the myriad morphological is that the structural development of the projection is not highly programmatic or subject to much complex regulation, but is instead mediated by a few simple principles arising from the source tissue, the olfactory epithelium. Such an autonomous program seems appropriate to the evolutionary age and importance of olfaction, as well as its continued regenerative capacity. Figure 36 (Doucette, 1991, Fig. 11)

. Electron-lucent leptomeningeal cells (LMC) surrounding LEOBs to form olfactory rootlets in the subarachnoid space (SAS). Note that in this illustration, the labeling of OEC ("En") is limited, and excludes the OEC nuclei and processes present within the regions labeled "R" and "ONL". Dural cells ("D").
Upon surfaces (Doucette, 1991(Doucette, , 1993a. At the points of rootlet fusion, the LMCs reflect off to become continuous with the pia mater, and the basal lamina of each LEOB reflects off to become continuous with the glia limitans (Julliard, 1998;Doucette, 1991 (Li, 1997(Li, , 1998Deumens, 2004Deumens, , 2006. The fibroblastic cells on the surface of the bulb, the LMCs, were assumed to be the same as the fibroblastic cells in the lamina propria, the ONFs. Since they most likely are not, the LMCs were studied under the false pretense of OEC-aiding properties originally attributed to ONFs in the lamina propria. While this does not invalidate particular experimental results, it does affect their subsequent interpretation since proper cellular identities were not recognized. This recognition may resolve the observation in these studies of the presumed ONFs appearing to share properties of astrocytes as well as fibroblasts (Li, 1998;Deumens, 2004). In fact, LMCs have the electron-lucent appearance of astrocytes, but express fibronectin when exposed to astrocyte-conditioned media (Colombo, 1994) as occurs in these CNS transplant models. To examine the original aims, these coculture experiments bear repetition with ONF proper. On the other hand, it may turn out to be serendipitous and LMCs are more useful than ONFs for co-transplant with OECs into areas of CNS injury. This would not be entirely surprising, as the neuroectodermal origin of LMCs may make them more inherently compatible with such an environment.
Another reason to heed the transition from ONFs to LMCs is that it may represent a concrete means by which to define the transition from the PNS to the CNS compartment, which is otherwise complicated by the fact that the OECs seamlessly transect the two. Elsewhere in the OECsand the other cells from the olfactory placodebeing inherently CNS tissues. The latter possibility being raised because the bulb and the epithelium are totally physically continuous, and by an intriguing study of zebrafish development that demonstrated that the cells of the olfactory placode migrate from the neural plate, rather than differentiating in situ from an isolated set of neural crest cells (Whitlock, 2004). Doucette (1991Doucette ( , 1993a noted in the rat that some of the rootlets penetrate deeply into the ONL before losing their LMCs and basal lamina. This may have implication for experimental procedures that rely on manual dissociation (or "peeling off") of the leptomeninges to expose the ONL for harvesting of OECs. Such manipulation may tear these deeply penetrating rootlets along the continuous arc of the ONL border, and thus retain contaminating streaks of LMCs within the ONL. These fissures were deepest on the ventral surface of the ONL, and imparted a highly convoluted appearance upon it (Doucette, 1991). The dorsal, medial, and lateral surfaces also have a convoluted appearance, but this is because the long-travelling rootlets from the ventrally-located cribriform plate fuse with these surfaces of the ONL at acute angles (Doucette, 1993a), thus resulting in rootlets that are only partially fused when viewed histologically (Julliard, 1998).  , 2005, Fig. 5B). Unfused dog olfactory rootlet. Scale bar = 10 µm. Figure 39 (Doucette, 1991, Fig. 8A). Partially fused rat olfactory rootlet. Scale bar = 5 µm.

-Olfactory Nerve Layer
The ONL is the outermost layer of the bulb into which every olfactory rootlet fuses. It

Glia Limitans
Elsewhere in the CNS, the glia limitans is formed by astrocytes. In those regions of the ONL penetrated by olfactory rootlets (transitional zones), the glia limitans is formed exclusively by OECs (Figure 36). Astrocytes are only seen contributing to the glia limitans at a distance from these zones (Doucette, 1991). Where the OECs and astrocytes abut one another, there is never any intervening basal lamina (Doucette, 1993a). The dorsal surface of the ONL has the fewest penetrating rootlets, so the majority of the glia limitans here is formed by astrocytes. In contrast, the ventral surface is extensively penetrated by olfactory rootlets, so the glia limitans is almost entirely formed by OECs. Figure 40 (Valverde, 1991, Fig. 6). "Camera lucida drawing, made under oil immersion objective, of one sheathing cell in the layer of olfactory fibers of the hedgehog. Laminar processes extend from the cell body and encompass fascicles of olfactory fibers. Star marks one large opening probably occupied by unstained olfactory fibers."

Astrocytes
The presence of astrocytes between EOBs in the ONL implicate it as a CNS region. The astrocytic processes may sometimes come into contact with the axons, but they never fully ensheath them (Doucette, 1984;Herrera, 2005). So, the OECs do not totally isolate the axons from the other constituents of the ONL, as had been suggested by Raisman (1985). It has been noted that these astrocytes are not present until some time after birth (Doucette, 1993a). They appear to migrate from the deeper bulb postnatally, rather than differentiate from progenitor cells in the ONL (Doucette, 1993b).

OEC basal lamina
In the ONL, OECs only produce basal lamina on those surfaces facing the glia limitans.
The expression of basal lamina on the mesenchymal faces of the OECs is therefore identical to the OECs of the lamina propria. This observation may shed some light on the question of how the absence of a basal lamina around the remainder of the OECs in the ONL is mediated. One suggestion was that the absence of basal lamina on these OECs was phenotypicthat these cells were a distinct sub-population (Franceschini, 1996). However, prior to that, Doucette (1990) had wondered whether the astrocytes in the ONL were responsible for directly modulating the production of OEC basal lamina. So, OECs constitutively produce basal lamina, and astrocytes inhibit this. The converse possibility would be that OECs do not produce basal lamina unless induced to do so, perhaps by fibroblasts. Since we have observed that intraepithelial OECs also lack basal lamina (see Section 2), the possibility of fibroblastic induction seems the most likely since there are no astrocytes in the olfactory epithelium. It also makes OEC's production of basal lamina consistent from the epithelium to the ONL, so there would seem to be no reason to posit a distinct sub-population on this basis. That OECs produce basal lamina on those surfaces exposed to fibroblastic cells has the additional virtue of helping to explain how cooperatively-ensheathing LEOBs in the lamina propria mediate the production of basal lamina only on the outer perimeter, and thus specifically how cooperative ensheathmentas defined by a single basal lamina surrounding multiple OECsitself arises.
The ONL has two visible outer and inner laminae, and within it occurs the transition from the large heterotypic bundles of the olfactory rootlets to the multiple homotypic bundles that target specific glomeruli in the underlying glomerular layer.
In the outer ONL (ONLo) the axon bundles appear generally well-defined by OECs and travel parallel to the surface of the bulb, extending around it in a geodesic-like path from their points of fusion predominantly on the ventral surface. It is considered that axon extension and regional targeting occur primarily in this layer (Au, 2002;. These bundles appear relatively distinctly ( Figure 45)depending on the overall density of rootlets fusing on that particular surface of the bulbbut they are heterotypic and so contain axons expressing a variety of OR types ( Figure 44A).
In the inner ONL (ONLi), the axons defasciculate and refasciculate into the homotypic bundles that synapse on specific glomeruli (Au, 2002;. Here the axons exhibit a more complex set of trajectories and are no longer necessarily travelling parallel to the contour of the bulb as they reassort and enter obliquely into the underlying glomerular layer. Likewise, the processes of OECs do not ensheath sets of axons into the more orderly bundles seen in the ONLo (Au, 2002). It may be said that the OECs in the ONLo more specifically ensheath bundles of axons, while in the ONLi they form a looser matrix through which axons travel as necessary.  The best way to apprehend the difference between the ONLo and the ONLi is to view the immunochemical markers for a specific subset of axons ( Figure 43, Figure 44) in comparison to that of all mature axons ( Figure 45, Figure 46). The ONLo is more orderly when looking at all mature axons ( Figure 45), but less so if looking at just one subset (ONLo, Figure 43); the bundles are more distinct, but heterotypic. The ONLi is more orderly when looking at just one subset of axons (ONLi, Figure 43), but less so if looking at all mature axons ( Figure 46); the bundles are more homotypic, but it is a dense and complex region.    In addition: However, the NFL [nerve fiber layer, ONL] is not a uniform structure but varies considerably with anatomical location. The rostral and ventral NFL is relatively thick with distinct inner and outer layers, whereas the NFL in the dorsal and caudal regions is thin and without the distinct layers observed in the rostral and ventral NFL. Most of the defasciculation and sorting of axons occurs when the axons first enter the rostral and ventral NFL [Au, 2002;;  It appears that in the ONLo, the well-fasciculated heterotypic bundles seem primarily to be extending. whereas the ONLi is where the real "work" of reorganization appears to take place: the defasciculation, rearrangement, and glomerular targeting. It should be noted that the first two-thirds of this "work" appears to occur entirely in the rostral ONLi, since the above excerpt describes the caudal ONLi as only completing the last step of glomerular targeting/extension; the same applies to the dorsal ONLi ( Figure 47) as the feature is dictated more by the distance from the rostral and ventral extent of the ONL.
The morphology of the ONL, in general, is postulated to arise as the growth cones of axons dynamically sample other axons for similarity or dissimilarity (Feinstein & Mombaerts, 2004). Such a description is better evinced by the morphology of the ONLi. Similarly, the experiments that demonstrated the ability of the olfactory epithelium to form ectopic glomeruli also describe the formation of an axonal plexus that better resembles the ONLi (Figure 48; Graziadei, 1978Graziadei, , 1979Graziadei & Monti-Graziadei, 1986). None demonstrated the outer/inner laminarity observed in the normal rostral ONL. Even when the bulb is unilaterally removed and replaced with a bulb-shaped artificial scaffold, rudimentary ONL and glomeruli form, but there is no laminarity to the ONL itself (Chehrehasa, 2006). The complex morphology of these axonal plexuses, as well as their adjacency to the ectopic glomerular structures, closely resemble the structure and function of the normal ONLi, but not the ONLo. These details are, however, relevant here insofar that the ONL is largely composed of OECs, and these are intimately related to the axons. Thus, the laminarity of the ONL as it pertains to axons also pertains to OECs. In addition, there are morphological and biomolecular differences between OECs in the ONLo and ONLi (Au, 2002;Franceschini, 1996;Astic, 1998;Treloar, 1999;Gong, 1994). Though such variations may sometimes compel the definition of OEC subtypes or other "OEC-like" cells, we have considered all such cells to be OECs on the basis of their continued ensheathment of the axons. This is best supported by observations in transgenic mice expressing DsRed-human S100β promoter in OECs that demonstrate the same cells comprising the olfactory fascicles and ONLo entering well into the ONLieven in an area where the OECs no longer express endogenous S100β protein (Windus, 2007(Windus, , 2010. Since all of the ensheathing cells are lineally related, and their morphological and biomolecular features are subject to some plasticity during development and in vitro (Astic, 1998;Franceschini, 1996;

-Why is there such a distinct ONLo?
We may begin by asking: What is the cause or purpose of a notably thick region confined to the rostral and ventral extent of the bulb? This seems related to the fact that the rostral and ventral ONL is the primary area of rootlet fusion with the bulb, so this is the region where the majority of the functional processes of the ONL would take place. But then the question becomes of this region: Why is there a outer lamina where the axons remain in distinct heterotypic bundles? Why do the bundles here not more rapidly undergo the functional processes apparent in the ONLi or ectopic axonal plexuses? Figure 49 (Windus, 2010, Fig. 1B). Bulb of adult transgenic mouse expressing human S100β promoter linked to DsRed fluorescent reporter sequences in a founder line of cells corresponding to OECs. Figure 50 (Windus, 2010, Fig. 1C). Colabeling with antibody to S100β protein demonstrates that the cells corresponding to OECs appear to extend more deeply than the region identified by endogenous S100β production. It is possible that factors arising centrally are responsible for the cessation of S100β expression among the OECs in the ONLi (iNFL).
Even though the ectopic glomeruli experiments are limited in scopeonly barely resembling the full creation of the olfactory bulba possible implication is that the structure of the ONLo may not be functional per se, but arises as the result of factors unique to complete in vivo development.
We hypothesize that the ONLo forms during development as something more like an extension of the pattern of growth observed in the lamina propria. But here, rather than growing longitudinally from the epithelium to the telencephalon, the migratory units have grown over the spherical surface of the bulb. Then, the ONLi arises where the axons penetrate that surface, followed by p/OECs, and form the complex plexus as they reassort and coalesce into glomeruli.

Development of the bulb
In mouse development, after E10 the axons and accompanying migratory cells leave the olfactory placode coincident with its invagination forming the olfactory pit (Cuschieri & Bannister, 1975b;Doucette, 1989). Growing through the mesenchyme, the multiple groups of axons and migratory cells eventually coalesce into a "single, larger fascicle" near the rostral telencephalic vesicle that contains only axons and a homogeneous population of cells. This migratory mass first makes contact with the vesicle by E11.5 (Doucette, 1989;Hinds, 1972ab). It continues to grow dorso-caudally over the external glia limitans of the vesicle between E12 and E13. By E12, the axons have formed a thin distinctive layer on the surface of the cerebral vesicle, and by late E13 the vesicle has markedly evaginated at the regions of contact to form the primitive bulb. This "cap" of axons and p/OECs that is interposed between the primitive pial cells and the early glia limitans of the telencephalic vesicle is referred to as the presumptive ONL (Doucette, 1989). Valverde (1992)  The presumptive ONL was defined on the basis of the mass of axons and p/OECs being separated from the telencephalic vesicle (or presumptive olfactory bulb) by an intact glia limitans. The definitive ONL was defined as the active elimination of the barrier between these and the deeper cellular constituents of the bulb primordium (Doucette, 1989(Doucette, , 1991(Doucette, , 1993a Marin-Padilla & Amieva, 1989), and when a new glia limitans was constituted by the p/OECs at the external border. As such, the presumptive and definitive ONL were the same cellular area, defined apart only by which surface faced a limiting membrane.
However, in these developmental studies, the ONL did not appear distinctly laminated 5 , and it was thought that the axons penetrating the early glia limitans were entering the presumptive glomerular layer. Since OECs appear to extensively migrate into this area ( Figure   52), and these studies did not observe the complex morphology of an axonal plexus in the presumptive ONL, it is very probable that the formation of the ONLi begins inside of what was the early glia limitans.
Consequently, we postulate that the presumptive ONL defines the structure of the ONLo, and the definitive ONL should be redefined to encompass what forms outside the early glia limitans (the ONLo) and what forms inside of it (the ONLi).
There are two items of developmental evidence for equating the presumptive ONL with the eventual ONLo. First, the heterotypic nature of the bundles in the ONLo resembles those throughout the lamina propria. There should not be anything characteristically different in the mixed constituency of individual bundles during their development through the early mesenchymewhether it be through the lamina propria or over the external surface of the early telencephalon. Second, the orientation of these bundles in the ONLo reflects the described formation of the presumptive ONL. While, in the lamina propria the bundles were extending more-or-less longitudinally towards the telencephalon, once they arrive, they do not uniformly and immediately penetrate its surface, but are described grow over it. Such parallel-to-surface spherical growth matches the fiber pattern in the ONLo. It is also possible that the ensuing evagination of the rostral telencephalic vesicle "ballooning" underneath this layer further draws the structure into such a shape.
Then, the real business of defasciculation, "the complex sampling of the environment", and "coalescence into glomeruli" takes place after the axons penetrate the early glia limitans.
This area, then, would be not just the presumptive glomerular layer that Doucette (1989) describesbut it is first the ONLi, and this gives rise to the GL.
Many possibilities exist as to the specific mechanisms that would lead to the formation of the early ONLi/GL here and not more superficially. These may involve any combination of: (1) the initial lag of accompanying p/OECs past the early glia limitans, (2) spatial factors resulting here from the focal convergence of axons from the entire epithelial sheet, (3) induction by factors from within the telencephalon, or (4) molecular features intrinsic to specific subsets of axons or OECs that come into play at this proximity to the bulb. The last of these seems at least to be a requirement, as the ectopic glomeruli experiments required the transplantation of the entire olfactory sheet, and not just parts of it, in order to form glomeruli (Graziadei & Monti-Graziadei, 1986).
In summary, here it has been proposed that the accumulation of axons on the surface of telencephalon during developmentthe presumptive ONLdefines the structure of the eventual ONLo, and the ONLi and GL form interiorly after the axons penetrate its surface and are followed by OECs.

-Glomerular Layer
There is some controversy as to whether OECs are present in the GL enwrapping the homotypic bundles as they dive towards the entrance to individual glomeruli. Various studies using common biomarkers have asserted their absence (Au, 2002;Treloar, 1999), whereas electron-microscopic studies (Raisman, 1985;Valverde, 1991) have identified OECs on a structural basis, accompanying the axons until their entrance into glomeruli.
It would seem that the development of glomeruli by the coalescence of like-axons should not inherently preclude OECs from their continued accompaniment right up until the point of entrance to individual glomeruli. Rather, this may make it such that the OECs in the GL are not present en masse, forming a matrix as they do in the ONL, but rather tightly follow the specific coalesced bundles the short distance from the ONLi border into the individual glomeruli. So even if the OECs do continue to express a particular biomarker, such a limited presence may account for an inability to resolve OECs immunochemically at the light-microscopic level.
Electron microscopy has demonstrated ensheathing cells within the GL. Raisman (1985) observed such cells in the accessory olfactory bulb accompanying bundles of vomeronasal axons right up to the glomerular capsule. However, these were termed "superficial glia" (SG) because they had a more electron-dense appearance than the astrocytes forming the rest of the capsule.
This allowed their being visually distinguishable from astrocytes, and enabled the detailed description of the glomerular capsule and synaptic zone. In the main olfactory bulb, Raisman notes, this contrast between the ensheathing cells and the astrocytes does not exist.
Another electron-microscopic examination of the main olfactory bulb in the hedgehog (Valverde, 1991) also demonstrated ensheathing cells following axons entering into the GL. The olfactory system in hedgehogs is highly developed, and there are multiple layers of glomeruli in the GL. This provided the opportunity to observe processes ensheathing the axons, passing between superficial glomeruli, and entering well into the GL to reach glomeruli situated more deeply.
Whether the ensheathing cells observed here are in fact OECs, rather than cells arising centrally, awaits further investigation. A definitive result could be provided with a immunoelectron-microscopic examination of this region in the aforementioned DsRed-S100β transgenic mouse.
In summary, the manner in which OECs accompany the olfactory axons throughout their development and throughout every compartment would seem to lend a high likelihood that they are present until their termination in synapses. Taken together with the formation of ectopic axonal plexuses and glomeruli-like structures being analogous to the development of the ONLi and olfactory glomeruli, there would seem to be no cause for OECs to cease ensheathing the axons until glomeruli actually form. It may be that after the formation of early glomeruli, and anchoring synapses with higher-order neurons (Malun & Brunjes, 1996), a generalized "push" from astrocytes and other cells arising centrally causes the smooth contour apparent between the GL and the ONLi. But this, perhaps, does not strip away the OECs already ensheathing the bundles entering into glomeruli. Such a later-stage migration of astrocytes could also account for the interfascicular astrocytes observed in the ONL postnatally. 1 While the vomeronasal and olfactory nerves are often studied as analogous, they do differ in many ways. A review by Salazar (2009) considers the risk in extrapolating anatomical data on the vomeronasal system from species to species as it is a system which undergoes involution and regression with a great deal of variability among even closely related species. The main olfactory bulb is relatively consistent among mammalian species, thus any extrapolation from the accessory bulb to the main olfactory bulb of the same species bears a similar type of risk.

-Endnotes
These complications may extend to aspects of the anatomy and cellular biology of the vomeronasal nerves as well. An olfactory-specific way to study the spatial dependence of cooperative ensheathment may be to compare the main olfactory nerves of various fishes -which range in length from millimeters in zebrafish, to 5.5-20 centimeters in European pike and garfish (Kreutzberg, 1977;Easton, 1971). 2 Figure 27 is approximately 3200 µm 2 and Figure 15 is approximately 20 µm 2 . Thus, the former is 160 times larger than the latter. 3 None of the embryonic studies describe the presence of ONFs during the early development of the olfactory nerves. This in itself may tend to support their being last in the sequence. In addition, ONFs are not reported to be present in the migratory mass. When the relevant cells of the migratory massaxons and pOECsarrive at the surface of the embryonic telencephalon, they interpose themselves between the early glia limitans and the meninx primitiva. Were there mesodermal fibroblasts in this migratory mass, these would presumably develop into a cell type that differs from the LMCs of the pia mater, and be situated nearer the bulb than the LMCs. This is not observed, so it is not likely that fibroblasts are within the migratory mass. Thus, their relevance to the olfactory nerves arises as cells outside the original cast of cells in the olfactory system, and makes it likely that their role would come in last. 4 The data on the macaque is inconclusive. Nonetheless, we see that all species eventually exhibit the features of total merger in the ONL. 5 The failure to observe ONL laminarity in these developmental studies may also be due to the locations of the bulb examined, the time points examined, and/or the high-magnification of electron-microscopy reducing the likelihood of observing distinct laminae in a given view.