A MOLECULAR MECHANISM REGULATING THE TIMING OF CORTICOGENICULATE INNERVATION

A MOLECULAR MECHANISM REGULATING THE TIMING OF CORTICOGENICULATE INNERVATION By Justin McRae Brooks, B.S. A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy at Virginia Commonwealth University. Virginia Commonwealth University, 2013 Michael A. Fox, PhD VTCRI Associate Professor Department of Biological Sciences Visual system development requires the formation of precise circuitry in the dorsal lateral geniculate nucleus (dLGN) of the thalamus. Although much work has examined the molecular mechanisms by which retinal axons target and form synapses in dLGN, much less is known about the mechanisms that coordinate the formation of non-retinal inputs in dLGN. These non-retinal inputs represent ~90% of the terminals that form in dLGN. Interestingly, recently reports show that the targeting and formation of retinal and non-retinal inputs are temporally orchestrated. dLGN relay neurons are first innervated by retinal axons, and it is only after retinogeniculate synapses form that axons from cortical layer VI neurons are permitted to enter and arborize in dLGN. The molecular mechanisms governing the spatiotemporal regulation of corticogeniculate

dLGN. Interestingly, recently reports show that the targeting and formation of retinal and non-retinal inputs are temporally orchestrated. dLGN relay neurons are first innervated by retinal axons, and it is only after retinogeniculate synapses form that axons from cortical layer VI neurons are permitted to enter and arborize in dLGN. The molecular mechanisms governing the spatiotemporal regulation of corticogeniculate xii innervation are unknown. Here we screened for potential cues in the perinatal dLGN that might repel the premature invasion of corticogeniculate axons prior to the establishment of retinogeniculate circuitry. We discovered aggrecan, an inhibitory chondroitin sulfate proteoglycan (CSPG), was highly enriched in the perinatal dLGN, and aggrecan protein levels dropped dramatically at ages corresponding to the entry of corticogeniculate axons into the dLGN. In vitro assays demonstrated that aggrecan is sufficient to repel axons from layer VI cortical neurons, and early degradation of aggrecan, with chondroitinase ABC (chABC), promoted advanced corticogeniculate innervation in vivo. These results support the notion that aggrecan is necessary for preventing premature innervation of the dLGN by corticogeniculate axons. To understand the mechanisms that control aggrecan distribution, we identified a family of extracellular enzymes (the a disintegrin and metalloproteinase with thromobospondin motifs [ADAMTS] family) expressed in postnatal dLGN that are known to contain aggrecan-degrading activity. Importantly, ADAMTS family members are upregulated in dLGN after retinogeniculate synapses form, and intrathalamic injection of ADAMTS4 (also known as aggrecanase-1) resulted in premature invasion of dLGN by corticogeniculate axons. Taken together these results implicate aggrecan and ADAMTSs in the spatial and temporal regulation of non-retinal inputs to the dLGN.

A Brief History of Neuron Doctrine
The central nervous system (CNS) is composed of neurons, glial cells, and extracellular matrix molecules that perform crucial roles in the establishment of specific neural circuits. The earliest knowledge regarding CNS circuitry was developed using tissue staining techniques that were unable to fully resolve the cellular makeup of the CNS and contained many artifacts. The reticular theory which postulated that the CNS was composed of a continuous meshwork of cells and processes that were all interconnected was prevalent during the pioneering days of neuroscience. Camillo Golgi, however, revolutionized neuroscience through his invention of metallic stains, including the potassium dichromate-silver staining technique. This technique called, the "Golgi stain," was the first method capable of staining individual neurons and glial cells in their entirety, and its advent led to the morphological characterization of neurons and neuroglia as completely distinct cell types (Golgi, 1875). Despite Golgi's passionate advocacy of reticular theory, he was never able to fully describe how the nerve network was established in vivo because the Golgi stain was limited in that it labeled small percentages of multiple cell types in CNS, making it impossible for Golgi to fully distinguish between neuron and glial processes.
Many of Golgi's colleagues utilized his own methods to refute the reticular theory.
Santiago Ramon y Cajal employed the Golgi stain on many species to create detailed drawings of the cells in CNS and the development of connectivity in spinal cord and brain. His ability to map the fine structures of neural circuitry helped solidify neuron theory as an axiom of modern neuroscience. During this period the dendrite and axon (terms coined by Wilhelm His and Albert Von Kolliker, respectively) were categorized as distinct functional domains of the neuron. The proposition of the existence of synapses, the small contact zones between neurons that allow for unidirectional conduction (by Charles Sherrington) was another observation that shaped the central tenets of neuron theory, although their existence could not be proven until the advent of the electron microscope (Sherrington 1897;Sherrington, 1900;Palade and Palay, 1954).
Ramon y Cajal pushed the neuron theory as a basis for studying the CNS through the declaration that the CNS is composed of individual discontinous cells called neurons, which all contain a "fundamental membrane" that are extensions of the neuron (Ramon y Cajal,1909). Ramon y Cajal and his student Pio del Rio Hortega introduced gold chloride and ammoniacal silver carbonate stains which allowed for the identification of multiple types of glial cells (Ramon y Cajal, 1913;Rio-Hortega, 1919). Through the development of methods to readily distinguish differing cell types, focus could be shifted to elucidating the mechanisms that drive neuron connectivity. Neuron theory eventually became neuron doctrine and was composed of these observations: the neuron is the structural and functional unit of the CNS, neurons are composed of dendrites, soma, and axons which are fundamentally different, and conduction of nerve impulses is directional (Shepherd, 1991).

Evidence For Specific Targeting of Neuronal Connections
In 1890, Ramon y Cajal described the structure by which axons navigate in the developing nervous system called the growth cone. He also postulated that growth cones are "oriented by chemical stimulation, and move toward the secretion products from certain cells" (Cajal, 1892). Outgrowth of nerve fibers was established in tissue culture experiments using the neural crest from frog embryo (Harrison, 1907;1910); however, there was a gap in understanding the mechanisms that drive axonal target specificity, in vivo. There was still much debate over whether axons initially extended throughout the CNS, and then were eliminated by competition, or if specific guidance factors drove axons to distinct targets.
Evidence for specificity in establishment of synaptic connections was discovered through JN Langley's experiments on the preganglionic fibers of the superior cervical ganglion. Upon bisection of these fibers, Langley observed regeneration in the preganglionic axons from differing spinal cord levels to the proper postganglionic neuron. Since regenerating axons did not grow to all levels of spinal cord and then refine, this observation led him to propose a chemotactic mechanism for axon guidance (Langley, 1895). Support for chemotactic mechanisms was exemplified in Roger Sperry's experimentation on the frog visual system, which led to the proposal of his chemoaffinity hypothesis. Normally axons that originate from cells in dorsal retina project to ventral tectum, and axons that come from ventral retina project to dorsal tectum (Kandel et al., 2000). Other orderly projections include axons projecting from anterior retina to posterior tectum, while posterior fibers extend to anterior tectum. These orientations produce a retinal topographic (retinotopic) visual map, in which adjacent neurons in the retina project to neighboring neurons in the brain, thus allowing the frog to see a faithful representation of the outside world (Purves et al, 2001).
To test the whether these topographic maps developed through random outgrowth of axons that are competitively eliminated based on activity or if there were specific connections from neurons in the retina to the optic tectum, Sperry detached the frog eye, inverted its positioning, and waited for the reestablishment of neuronal connections. An important feature of the amphibian nervous system is that severed retinal ganglion cell (RGC) axons regenerate and reinnervate the tectum. By rotating the retina by 180°, Sperry discovered that the frogs had inverted visual maps and that the axons in these animals projected to the proper areas of tectum despite the retinal repositioning. (Sperry, 1963). His work sparked the chemoaffinity hypothesis that

Axon Guidance and Synaptic Targeting
In order for the sensory systems to distribute information to the correct relay and processing areas, axon guidance and synaptic targeting cues are necessary to set up a precise neural network. Axon guidance refers to the process by which growth cones react with their environment to select a pathway in which they travel long distances to a specific target nucleus within the nervous system (Goodman and Shatz, 1993). Axon guidance requires ligand-receptor interactions resulting from either short range (contactmediated) or long range secreted (diffusible) signals .
Contact-mediated attraction and repulsion occur at guide post cells, which contain ligands and transmembrane receptors that alter the axon's trajectory as the growth cone traverses its path and physically interacts with other cells. Secreted chemoattracant molecules can act over long distances and form gradients that direct axons, with complementary receptors, to specific brain regions, while secreted chemorepulsive signals can also act from afar but cause an axon to retract, pause, or turn depending on strength of signal [( Figure 1A) Chen and Cheng, 2009].
Molecular mechanisms that drive axon guidance can also regulate synapse formation. After an axon has selected a cellular partner, synapses between cells must communicate in order to form a functional connection ( Figure 1B). The first step in synapse formation is called target recognition, in which a growing axon makes a rudimentary connection onto a postsynaptic cell or in some cases a specific region on that postsynaptic cell. Following synaptic targeting, synaptic differentiation prompts the organization of synaptic elements and establishment of reciprocal communication between the two cells, allowing each to synthesize the proper machinery for setting up a mature, fully functional synapse. Finally, mature synapses that have found suitable partners stabilize, while any improper or weak connections that have begun forming are retracted and allowed to search for a new postsynaptic partner (Fox and Umemori, 2006). Presynaptic terminals that are unable to communicate and assemble a synapse are ultimately degraded.

Extracellular Matrix Effects on Axon Pathfinding and Synaptic Development
The development of CNS circuitry requires precise spatial and temporal targeting of axons to specific partners in the distinct brain nuclei. During axon guidance and synaptic targeting, many different extracellular and molecular mechanisms drive the formation of specific connections. Axon guidance mechanisms have rigorously been studied in CNS since the proposal of the chemoaffinity hypothesis, and the advent of modern technology including genetically altered mouse models, along with high resolution imaging techniques, has advanced understanding of the biochemical mechanisms that drive axon guidance and targeting.
Many molecular axon guidance cues that have been elucidated are functional in differing areas of the CNS. Netrins are evolutionally conserved, diffusible molecules that are secreted from cells at the midline of the spinal cord, and netrins can act either as chemoattractant or chemorepulsive signals, depending on the type and density of receptors expressed on the growth cone of an approaching axon (Tessier-Lavigne and Goodman, 1996). Netrins and netrin receptors, such as DCC/Unc, contribute to the guidance of thalamocortical axons to their proper targets in cortex (Powell et al., 2008).
Contact-mediated attraction and repulsion are also guided by expression of extracellular molecules, but these are inititated by complementary receptors and ligands embedded in the surfaces of cells encountered during axon outgrowth (Chen and Cheng, 2009;Kolodkin and Tessier-Lavigne, 2011). Semaphorins are a canonical repulsive guidance cue ligand which can either be diffusible or membrane-bound for contact-mediated signaling. Plexin receptors represent the major class of semaphorin receptors, and activation of these receptors can signal disassembly of growth cone cytoskeletal molecules and initiate synaptic targeting (Zhou et al., 2008).
These guidance signals also can interact in a myriad of ways such that multiple cues can be processed in parallel, which is a mechanism by which the brain processes individual components of a single stimulus simultaneously. Signaling can work through addition or multiplication of stimuli together or through opposition of each other's effects; however, the interactions between guidance molecules often are more complex than just simple summation of positive and negative signals (Raper and Mason, 2010;Dudanova and Klein, 2013). One such example of this complexity occurs in commissural neurons, which react to both netrin and slit. Axons are guided to the midline by netrin and subsequently should be repelled by slit; however, repulsion by slit is suppressed by Robo-3/Rig-1, a Robo receptor that diminishes the effect of slit, until the midline is reached. During midline crossing, slit activation of the Robo receptor alters function of the DCC receptor, rendering the netrin signaling ineffective (Sabatier et al., 2004;Chen et al., 2008).
In addition to secreted molecules and membrane bound receptors and ligands, there are classes of extracellular matrix (ECM) molecules that have also been shown to guide axons during development and also during regeneration of injured axons in the CNS. Of particular interest are the proteoglycans, a subset of extracellular macromolecules that are expressed in CNS and provide a substrate for tethering link proteins to cells (Morawski et al., 2012). Studies involving chondroitin sulfate proteoglycans (CSPGs) during development and in the adult have revealed potent inhibitory signaling on the neuronal growth cones of many cell types. CSPGs expressed in the periphery of the retina are involved with both proper spatiotemporal expression of RGC differentiation and directing RGC axons outward to the optic cup (Brittis et al., 1992).
CSPGs interact with many substrates in CNS including growth-promoting laminin (McKeon et al., 1995). The effect of CSPG inhibition was attenuated when laminin was coexpressed with CSPG, although the reaction for different cell types was dependent on the ratio of laminin/CSPG (Snow and Letourneau, 1992;Snow et al., 1996). These studies provided evidence that CSPGs have variable effects on diverse cell types.
CSPGs also interact with cell adhesion molecules (CAMs) to regulate axon outgrowth, and evidence has shown that sensory neurons can overcome CSPG inhibition through overexpression of integrin receptors (Condic et al., 1999;Tan et al., 2011). Moreover, a recent in vivo study of serotonergic neurons suggests that increased expression of either β1 integrin, growth-associated protein-43 (GAP-43), or a combination of the two enables them to overcome CSPG inhibition in response to spinal cord injury (Hawthorne et al., 2011). GAP-43 has also been implicated in enhancing axon growth after injury in commissural interneurons of the cat spinal cord (Fenrich et al., 2007). Together, these studies provide support for how a complex series of interactions between cells, ECM, and growth substrates influence axon outgrowth.
Molecular factors also aid in synaptic targeting. Subcellular specificity dictates where a presynaptic axon will form a synapse on postsynaptic dendrites, soma, or axons, and this process can be mediated by extracellular guidance cues or transmembrane receptors (Sanes and Yamagata, 2009). For example, proteoglycans expressed in distinct locations aid in targeting the growth cone to motor neuron synapses, so that synaptogenesis can begin (Sanes, 2003). Neurofascin, a CAM in the immunoglobulin superfamily, is present in a gradient in postsynaptic Purkinje cells of the cerebellum. Presynaptic axons from basket cells first innervate the soma and proximal dendrites of Purkinje cells and then work their way up the neurofascin gradient to their final subcellular locations (Ango et al., 2004). CSPGs also have been reported to play a role in mediating lamina-specific adhesion and lamina-specific axon outgrowth in the dentate gyrus of the hippocampus by entorhinal cortical afferents (Forster et al., 2001).
During early development CSPGs are derived from neurons; however, as the brain integrates circuitry from diverse sensory areas and more cells participate in the maintenance and plasticity of synaptic function, CSPGs, other ECM, and guidance factors are produced by neurons and non-neuronal "glial" cells including astroctytes, oligodendrocytes, and microglia in the brain (Domowicz et al., 1996). This level of complexity makes it difficult to ascertain the exact functions of individual cell types and ECM in the adult CNS. Astrocytes and microglia produce thrombospondins, extracellular proteases, and growth factors that potentially could alter the routing of axons (Dityatev et al., 2010;Crawford et al.,2012;Frischknecht and Gundelfinger, 2012).

Development of The Visual System from Retina to Brain
The visual system serves as an excellent model for understanding the formation of neural circuits because of its accessibility, anatomy, and role in sensory processing.
The development of the visual system has been well characterized, and much of the formation of visual circuitry in the mouse occurs during the first two weeks of postnatal life. The eye is an easily accessible structure for experimental manipulations, and RGCs provide the sole output from the retina to a plethora of retinorecipient nuclei in the brain. The retinofugal (RF) system, in particular, provides a wealth of knowledge about how RGC axons from the outer limits of the CNS can project long distances, innervate subcortical brain regions, and stop within specific, functionally distinct nuclei to begin synaptogenesis.
RGC extension is the first step in axon pathfinding. Until recently, factors responsible for initiating axon outgrowth have remained elusive due to the inability to examine growth, in vitro or in vivo, in the absence of neurotrophic factors because neurons die (Goldberg and Barres, 2000). In order to better understand whether intrinsic or extrinsic mechanisms drive axon outgrowth, a mouse model was developed that overexpressed Bcl-2 anti-apoptotic factors, and RGCs were purified and grown in the absence of glial contact (Goldberg and Barres, 2000;Goldberg et al., 2002). Under these conditions, cells were able to survive without trophic factors, but there was no axon outgrowth. Other previously published data on sensory neurons has shown similar findings (Lindsay, 1988;Lentz et al., 1999). The next step was clearly defining the extrinsic factors that could cause outgrowth. Many factors from a wide range of families, including brain-derived neurotrophic factor (BDNF), ciliary neurotrophic factor (CNTF), and oncomodulin, stimulate axon growth both in vitro and in vivo (Goldberg et al., 2002;Yin et al., 2006).
Axons, once extended from the ganglion cell layer (GCL) in the retina, grow into the optic fiber layer (OFL), and they must navigate away from the periphery towards the optic disc and out of the eye. RGC axonal extension towards the optic cup occurs from embryonic day 11 (E11) to E18 (Drager, 1985). These axons are restricted to the OFL by a layer of neuroepithelium located just inner to the OFL. The complexities of axonal pathfinding begin at this particular step, as RGC axons that originate from different areas of the retina begin to find their way out. Axons that project ipsilateral, or to the same side of the brain as the eye, originate in the ventrotemporal (VT) crescent of retina, while fibers that project contralateral, or to the opposite side of the brain from the eye, will come from all areas of the retina. These projections do not use the same signaling mechanisms to target CNS nuclei. For example, neuroepithelial endfeet strongly express neural cell adhesion molecules (NCAMs) in rats, like L1-CAM (L1), which allows extension of some pioneer RGC fibers . Axons that follow pioneer RGCs use a different set of molecules to fasciculate and travel with the pioneer axons.
Synaptic targeting requires more than molecular cues to locate distinct cellular partners. The retina also produces spontaneous retinal waves before eye opening at postnatal day 14 (P14) that govern expression of molecular signals and alter the final pattern of innervation within thalamic targets (Chalupa, 2007;Rebsam et al., 2009).
Stage 2 waves are the most relevant to the experiments included in this dissertation, and they begin around P0 in mice and last until approximately P10. Stage 2 retinal waves arrive coincident with retinotopic mapping in the CNS, and the waves allow for pruning of weak synapses and the differentiation and maturation of proper synaptic contacts. Retinal waves also help pattern eye specific segregation. These stage 2 waves are generated by starburst amacrine cells, and driven by acetylcholine release from these cells (Feller et al., 1996). They occur relatively infrequently at a rate of one every 1-2 minutes. This frequency was determined to be a driving factor for the eye specific segregation between projections within the CNS because of the high probability that they would be sending information at alternating times, as opposed to simultaneous transmission (Butts et al. 2007).
Anterograde tracers, utilized long before transgenic mice, allowed insight to the timing of development of retinogeniculate (RG) and retincollicular (RC) projections (Godement et al., 1984). As RGCs innervate their primary targets in subcortical CNS, the complexity of visual circuit formation continues to build as relay neurons from distinct thalamic nuclei project axons to other areas of brain including some projections into primary visual cortex (V1) for higher order processing (Yamada et al.,1996;Van Hooser, 2007;Huberman and Neill, 2011). V1 spreads information to other associated areas of the cortex for further assimilation and processing of perception, as well as feedback circuits to retinorecipient thalamic nuclei.
RGCs are divided into over 20 separate classes which encode information for both image-forming and non-image-forming visual streams. ON-OFF selective RGCs react to increases (ON) or decreases (OFF) in the amount of light presented to the retina, direction selective RGCs respond to horizontal or vertical movement of an object, and other RGCs respond to and transmit information regarding colors (Chalupa and Gunhan, 2004;Demb et al., 2007;Kim et al., 2008;Quina et al., 2005;Schmidt et al., 2011;Badea et al., 2009). These particular classes constitute a subset of the imageforming stream of visual data, and they target CNS nuclei in the thalamus that are important for relaying information to areas of cortex that form perception of the visual scene. Intrinsically photosensitive RGCs (ipRGCs) that contain the photopigment melanopsin also occupy the GCL in the retina, and ipRGCs project primarily within the non-image forming pathway of visual data to distinct nuclei in the brain that encode information regarding circadian rhythm and irradiance detection (Provencio et al., 2000;Hattar et al., 2002;Hattar et al., 2006). These cell subtypes have become important for understanding how specific guidance cues contribute to class-specific targeting, especially since the project axons to closely apposed, distinct nuclei where differential patterns of guidance cues can be identified and monitored.
Much research regarding RF circuitry in mice focuses on the lateral geniculate nucleus (LGN), which is an important subcortical relay and processing visual center located in the dorsal thalamus. In mice, the LGN contains three separate subnuclei called the dorsal LGN (dLGN), intergeniculate leaflet (IGL), and the ventral LGN (vLGN). The three LGN subnuclei have distinct functions, and they serve as a primary relay center from the retina to other subcortical areas, and, in the case of dLGN, a relay center from retina to V1. Certain image-forming RGC axons will target and primarily innervate the dLGN, while ipRGCs will advance directly to the IGL and vLGN [ Figure 2. Kay et al., 2011;Fox and Guido, 2011)].
The timing of RG innervation is stereotypical with pioneer axons innervating the dLGN at E16 (Godement et al., 1984). Advances in anterograde labeling allow for examination of projections from each eye converging or diverging within retino-recipient nuclei. Intraocular injection of cholera-toxin subunit B (CTB) can be used to label retinal projections from each eye [ Figure 3. (Muir-Robinson et al., 2002;Jaubert-Miazza et al., 2005)]. This allows visualization of eye specific domains in dLGN ( Figure 3). During postnatal visual development, axons from the retina form many synaptic contacts onto relay neurons in dLGN. Initially, the small terminals of these weak synaptic contacts can be labeled with vesicular glutamate transporter type-2 (VGlutT-2) antibody beginning at P3. VGluT-2 labeling of these synapses also provides insight into the timing of differentiation and maturation of RGC terminals in dLGN as the terminals increase in size until P14. These methods provide evidence that retinal terminals compete for synaptic space, mature and refine during the period of development from P4 until eye opening at P14 [ Figure 3, (Guido, 2008)].
Eye-specific refinement of the retinal axons was shown in dLGN beginning around P4 when projections from each eye do not have clear borders (Guido, 2008). At  Figure 3. (Pfeiffenberger et al., 2005;Rebsam et al., 2009;Sanes and Zipursky, 2010;Triplett and Feldheim, 2011)]. Recent evidence indicate that ipRGCs, which express melanopsin, the only functional opsin in the retina at birth, react to light and modulate stage 2 retinal waves and contribute to retinal refinement (Renna et al., 2011;Kirkby and Feller;. In addition to eye specific domains, retinal projections also target specific nuclei topographically, so that neurons that are adjacent in the retina project axons to neighboring neurons in retino-recipient nuclei (Luo and Flanagan, 2007). Utilizing transgenic reporter mice that label only individual classes of RGCs and mutant mice lacking targeting molecules, multiple visual system targeting mechanisms have been well characterized; for example, gradients of ephrins and Eph receptor help arrange RGC axons in dLGN and superior colliculus (SC) (Feldheim et al., 1998;Hindges et al., 2002;Huberman et al., 2008). Retinal input instructs alignment of topographic maps from V1 and retina in the SC, such that corticocollicular mapping happens after retinocollicular mapping, and retinal input is required for precise mapping and corticocollicular fiber refinement (Triplett et al., 2009). Alignment of topographic maps is deemed necessary for proper visual perception, and this study also revealed the importance retinal activity on non-retinal (ie V1) inputs to retinorecipient thalamic nuclei.
Though the thalamic nuclei were originally postulated to serve only as a relay for sensory information travelling to the cortex, recent research suggests that the thalamic nuclei serve much more dynamic roles, such as providing adjustments to retinal input from feedback provided by layer VI cortical neurons .
Retinal activity and extracellular components can drive the specificity of topographic maps and eye-specific domains within retinorecipient nuclei, but little is known about the mechanisms that affect class-specific targeting to distinct nuclei. The glycoprotein reelin has been shown to guide ipRGCs into the vLGN and IGL . In the absence of functional reelin ipRGC axons fail to innervate the vLGN and IGL, and the ipRGC axons are misrouted to inappropriate thalamic nuclei. Mice that lack Dab1 receptor, a molecule essential for reelin function (Sheldon et al., 1997), also showed misrouting of projections, that usually terminated in vLGN and IGL, towards dLGN . These findings suggest that ECM may play a large part in the axon guidance of both image-forming and non-image-forming pathways.
Much of what has been learned regarding class-specific targeting in visual system has focused only on circuitry between the retina and relay neurons in the thalamus. Moving forward, application of what has been learned through these studies will need to be applied to non-retinal inputs to LGN including axons from brainstem, cortex, reticular thalamic nucleus (RTN) and interneurons which provide 90% of the innervation in LGN and help process retinal signal as it proceeds to cortex.

Interconnectivity in the Thalamus
In addition to RG axons, there are numerous other cell types that innervate dLGN. In fact, RG axons form less than 10 percent of the synapses onto relay neurons, despite being the primary driver for producing synaptic responses in relay neurons (Van Horn et al., 2000). The majority of non-retinal input to dLGN are glutamatergic modulatory synapses that originate from layer VI cortical neurons in visual cortex (Erisir et al, 1997;Jones 2002). Relay neurons are also innervated by local GABAergic interneurons within the dLGN, inhibitory projection neurons from the RTN, and modulatory projections from brainstem and SC (Sherman and Guillery, 2002;Bickford et al., 2010). The development of these non-retinal inputs to the dLGN has not been well characterized because of the difficulty in labeling particular classes of neurons in confined areas of the CNS.
Recent advances in transgenic mouse reporter lines have enabled a more developed study of non-retinal inputs to the dLGN. In particular, the golli-tau-gfp mouse, which has layer VI cortical neurons and their projections selectively labeled with green fluorescent protein (GFP), has aided the efficient visualization of how and when dLGN is innervated by axons from layer VI cortical neurons, also called corticogeniculate (CG) axons (Jacobs et al., 2007;Grant et al., 2012). This reporter mouse line provides an argument for class-specific targeting in cortex, as it is apparent that particular layers of cortical neurons project fibers that innervate specific locations.
Layer VI axons have been shown to selectively innervate dLGN, while projections from layer V pyramidal cells also project to the LGN but terminate in vLGN/IGL areas [ Figure   4. (Cosenza and Moore, 1984;Jacobs et al., 2007)].

Spatial organization of RG and non-retinal inputs onto distinct relay neurons is important in the formation of neural circuitry. Retinotopic maps begin formation in the
LGN prior to afferents from the cortex, RTN, brainstem and interneurons (Bickford et al., 2010). The non-retinal synapses have been reported to provide feed-forward, feedback, and modulatory signals that alter the main stimulus coming from the retina, and they presumably must connect onto relay neurons in register with the correct retinal inputs for faithful image processing and correct behavioral outcomes (Sherman, 2012).
Subcellular specificity is also determined as the visual circuits begin to form during the first week of postnatal life. Excitatory retinal synapses form on proximal dendrites in order to drive action potentials. Inhibitory inputs from local interneurons will receive input from RGCs and form synaptic contacts in close proximity to the RGC / relay neuron interface to help tune the signal provided (Sherman and Guillery, 2002).
As mentioned previously, CG synapses form more synapses onto relay neurons, but they innervate distal positions of the dendrites for the production of feedback signaling ( Figure 5).
The spatial organization of retinal and non-retinal synapses may be connected with the timing of their entry into dLGN. During dLGN development, RG axons are the first projections to enter the dLGN, and they can synapse onto any area of relay neuron that is available. RG fibers form synaptic contacts onto proximal dendrites of the relay neurons, but it is unclear whether they occupy that particular area because of affinity, because they are the first projections to arrive, or because of a combination of affinity and timing (Sherman and Guillery, 2002).
The timing of CG innervation follows that of RG fibers; importantly, RG fibers begin synaptic differentiation immediately prior to the beginning of layer VI CG axon entry into dLGN ( Figure 6). Though CG axons begin to invade dLGN at P4, synaptic contacts marked by vesicular glutamate transporter-1 (VGluT-1) do not begin to appear in the dLGN until P7 [ Figure 6, ].

Projections to dLGN
Recent studies using the golli-tau-gfp mouse have demonstrated that retinal inputs may play an instructive role in regulating the timing of CG innervation.
Specifically, RG axons begin synaptic remodeling in dLGN prior to CG fiber entry, and removal of retinal input by enucleation or genetic ablation accelerates layer VI cortical innervation in dLGN . This suggests that retinal input to dLGN may be providing a stop signal that inhibits non-retinal innervation of dLGN until RG synapses have begun to function.
In light of this work, the goal of my project was to discover the molecular mechanisms that regulate the timing of CG axon innervation. As mentioned previously, there must be a balance of activity and guidance factors working to provide the perfect environment for proper synapses to form; however, little is known about molecular mechanisms that guide non-retinal inputs in the thalamus. Here I report that the inhibitory CSPG, aggrecan, is robustly expressed in the dLGN during early visual circuit formation. I sought to classify its function and regulation during this phase of development. Based on the preliminary data and the coincident delay of CG innervation of the dLGN, I hypothesized that aggrecan initially repels layer VI cortical projections from entering dLGN until proper synapses form between RGCs and relay neurons in the dLGN.
Little is known regarding mechanisms that drive how and when the CG afferents  Inhibitory cues cause the growth cone to avoid the red cell completely and search for more appropriate targets. Attractive chemospecific cues allow the growth cone to extend towards the dendrites located on the black cell, where the first steps of synaptogenesis can begin. The green cell may contain attractive elements within its dendrites, but it has not been targeted by the growth cone due to a lack of extracellular cues. Many of the same molecules that drive axon guidance over long distances can also guide the axons toward specific cells.  Immunostaining with VGlutT-2 shows the maturation of retinal synapses from synaptogenesis at P3 until maturation at P14. Scale bar is 250 µm.   Immunostaining using VGlut-1 reveals the timing of synapse development between layer VI CG axons and relay cells. Scale bar is 250 µm.

Chapter II Experimental Methods
Mice CD1 and C57/BL6 wild-type mice were purchased from Charles River (Wilmington, MA) or Harlan (Indianapolis, IN). acan cmd heterozygous mice which lack functional aggrecan, were purchased from The Jackson Laboratory (Bar Harbor, ME).
The generation of math5 -/and golli-tau-gfp mice were described previously (Wang et al., 2001;Jacobs et al., 2007). Genomic DNA was isolated from tails using the HotSHOT method as previously described (Truett et al., 2000;Su et al. 2010) and genotyping was performed with the following primers purchased from Integrated DNA seconds, annealing at 58ºC for 20 seconds, and elongation at 72ºC for one minute.
Further treatment of acan cmd polymerase chain reaction (PCR) products using BpmI restriction endonuclease (New England Biolabs, Ipswich, MA) was performed according to previous reports (Watanabe et al, 1997). All analyses conformed to National Institutes of Health guidelines and protocols approved by the Virginia Polytechnic

Institute and State University and Virginia Commonwealth University Institutional Animal
Care and Use Committees.

Immunohistochemistry (IHC)
IHC was performed as previously described (Su et al. 2010 Plot profiles were obtained using imageJ software. Images were imported into imageJ software and a line scan was analyzed using the plot profile command. Data was exported into Microsoft excel where displacement distances were converted to µm. Fluorescent intensities represent averages for every three pixels (1.8 µm) for each image over the first 50 µm displacement into dLGN, and data was plotted in a line graph for images collected from n≥3 animals.

Dissociated Cortical Cultures
Cerebral cortices were dissected from E15-E18 golli-tau-gfp embryos and were digested in 0.25% trypsin at 37°C for 15 min. Following digestion, soybean trypsin inhibitor was used to inactivate trypsin and cortical tissue was transferred to serum-free medium (Neurobasal medium with 0.5mM L-Glutamine, 25 uM L-Glutamate, 10ug/ml Gentamicin with B27 supplement). A single cell suspension was generated by triturating tissues through a 1000ul pipette tip and 5x10 3 cells were added to each well of a poly-Llysine (PLL) treated 8-well lab-Tek chamber slide. In addition to being pre-treated with PLL, chamber slides were "spotted" with various extracellular substrates to assess neurite outgrowth. Briefly, various concentrations (1 µg/ml, 5 µg/ml and 10 µg/ml) of aggrecan (Sigma), chABC or recombinant human A Disintegrin and Metalloproteinase with Thrombospondin Motifs-type-4 (rhADAMTS4, R&D Systems) were mixed with BSA conjugated to Alexa-Fluor 594 (Invitrogen) (2 µg/ml) and 2µl spots were placed onto the slide surface in separate chambers and allowed to incubate in a humidified chamber at 37 °C for two hours. The ability of chABC or rhADAMTS4 to degrade aggrecan was assessed by pre-incubating 10 µg/ml aggrecan with 50U/ml chABC in PBS or 10 µg/ml rhADAMTS4 before mixing with BSA-Alexa-Fluor 594. After plating, cortical neurons were cultured for 3 days at 37°C in 5% CO 2 before being fixed with 4% PFA (in PBS) and imaged. Layer VI neurons (and their neurites) were immunolabeled with anti-GFP antibodies and I counted the numbers of neurons whose neurites were able to grow into the "spotted" substrates. Only neurites from cell bodies that lie within 50 µm of the spot of interest were counted. A minimum of four experiments (each with at least 3 replicates) was compared in all in vitro experiments.

Fluorescent In Situ Hybridization (FISH)
FISH was performed on 16-20 µm cryosectioned coronal sections as described previously (Su et al. 2010 Austin, TX). Riboprobes were hydrolyzed into ~0.5 kb fragments. Coronal brain sections were prepared and hybridized as described previously (Yamagata et al. 2002;Fox and Sanes 2007   were euthanized, perfused, and tissue prepared for cryosections or vibratome. Spatial extent of innervation by GFP labeled fibers was calculated using the threshold analysis method described above.

Retinal Projection Labeling
Intraocular injection of CTB conjugated to Alex Fluor 488 or Alexa Fluor 594 (Invitrogen) was performed as described previously (Su et al. 2010

Chapter III Developmental Regulation of Aggrecan
Introduction CG fibers comprise up to 50% of the innervation to relay neurons, constituting one of the largest sources of input to the dLGN (Erisir et al, 1997;Jones, 2002).
Afferents from layer VI cells in visual cortex project directly to dLGN and help shape signaling by relay neurons through tuning of receptive field properties and influencing RG signal transmission (Sherman and Guillery, 2002). The mechanisms driving layer VI cortical innervation of dLGN have been difficult to characterize due to the vast number of synapses and lack of tools that distinguish between cell types; however, the generation of the golli-tau-gfp transgenic reporter mouse line has improved the ability to investigate early development of CG projections. In the golli-tau-gfp transgenic reporter mouse, layer VI soma and processes are selectively labeled with a tau-GFP fusion protein whose expression is driven by the golli promoter of the gene encoding myelin basic protein (Landry et al., 1998;Xie et al., 2002).
Cortical projections, in these animals, traverse from the internal capsule and begin invading the thalamus by P0. While these layer VI afferents innervate dorsal thalamic nuclei immediately, the dLGN remains devoid of layer VI projections until P4 [ Figure 6 ( Jacobs et al., 2007;Grant et al., 2012)]. The timing of CG axon entry coincides with the remodeling of retinal axons and appears to be orchestrated by RG inputs . Due to the delay in CG innervation, I proposed a mechanism by which extracellular guidance cues prevent premature innervation of dLGN by layer VI GFP-expressing axons, a process which appears to be regulated by retinal inputs.

Identification of aggrecan localized in dLGN during cortical axon "waiting"
To identify guidance cues localized in dLGN, I initially profiled the transcriptome of P3 and P8 dLGN, with the assumption that mRNA of repulsive cues inhibiting premature CG innervation would be down-regulated as cortical axons began to enter dLGN. No such molecules were identified (data not shown). As an alternative approach The CS domain was discovered to be the repellant domain of functional aggrecan, and specific CS-glycosaminoglycan (GAG) patterns have been reported to alter neurite extension in vitro at levels comparable to intact aggrecan (Gilbert et al., 2005).
Combined with my initial aggrecan-IR results, these reports led me to investigate the relationship between the timing of aggrecan expression in dLGN and the profile of CG innervation.  Figure 9A).

Aggrecan expression during visual development
mRNA expression levels are not directly correlated to translation events or protein levels, but there have been reports in the literature that, during development in rodents, the antibody I used to specifically label aggrecan, cat315, also labels certain isoforms of phosphacan (Dino et al., 2006). Thus, I was inclined to examine cat315 antibody specificity, especially after reviewing the qPCR data. First, the cat315 antibody did not label the phosphacan in the vLGN in my initial search for an inhibitory guidance cue (Figure 7). Next, I tested the validity of my previous observations by immunostaining embryonic sections of brain, using cat315, from the autosomal recessive mouse mutant that lacks functional aggrecan, called the acan cmd mouse (Watanabe et al., 1994). Aggrecan-IR was abolished in the brains of these mice, and these results confirmed that IR observed in the WT (wild type) dLGN at P0 is unique to aggrecan ( Figure 9B inset).
After I ensured the specificity of cat315 binding to aggrecan, I probed dLGN with cat315 later during development, when axons from the layer VI GFP-labeled cortical fibers in the golli-tau-gfp mouse were beginning to invade dLGN, to determine if aggrecan was still present. My results showed that aggrecan was degraded throughout the dLGN by the time that CG projections spanned the entire dLGN and VGLuT-1 labeled cortical synapses began to appear ( Figure 9B; Figure 6).

Aggrecan signal degraded prior to CG fiber entry into dLGN
Aggrecan's structure has been well established, and its functions regarding plasticity and regeneration in the CNS have been characterized, but the role of aggrecan during visual circuit formation was unknown. As I performed a daily analysis of CG innervation and aggrecan, from P0 when axons began to arrive at dLGN until P7 when axons occupy nearly 50% of dLGN, I began to uncover an important developmental role for aggrecan in the orchestration of RG and CG circuitry using the golli-tau-gfp mouse line. At P0 when aggrecan-IR was most intense, pioneer CG axons began to enter vental thalamus from the internal capsule, and they started to climb the external medullary lamina (eml) toward dLGN. However, they appeared to pause and wait in eml just medial to the dLGN, and, for several days, they innervated other nuclei of dorsal thalamus but not dLGN ( Figure 10).
Aggrecan-IR covered the entire extracellular area of dLGN until P1 when the IR signal appeared graded with aggrecan loss beginning at the ventromedial border of dLGN and proceeding laterally. CG axons coincidentally remained outside of dLGN during the period of time when aggrecan-IR was high, and CG fiber innervation was inversely-correlated with the area occupied by aggrecan. Furthermore, CG fibers only innervated the portions of dLGN at P4 where aggrecan-IR was nearly undetectable, which led me to hypothesize that aggrecan regulates the timing of layer VI cortical axon innervation in dLGN ( Figure 10).

Potential Role for retinal input in the regulation of aggrecan
A recent report analyzed the effects of retinal input on the timing of CG innervation in dLGN, using the math5 -/mutant mouse line, which lacks 95% of RGCs . That study also included binocular enucleation as a method for destroying retinal inputs, and their observations led to the declaration that removal of retinal inputs accelerates CG innervation. In order to confirm the lack of retinal innervation in dLGN of math5 -/mutants, I used DiI, a lipophilic tracer, to label RGCs and their afferents in different aged animals. WT mice exhibited distinct labeling in the brain where retinal processes were located, but the math5 -/mutants contained no DiI labeling at any age, suggesting that retinal input to the brain was completely absent ( Figure 11). By crossbreeding the math5 -/mutants to the golli-tau-gfp reporter mice, I was able to probe the hypothesis that aggrecan protein must be significantly reduced by the removal of retinal input to dLGN at perinatal ages in order for layer VI fibers to exhibit accelerated entry into dLGN.
I analyzed the spatiotemporal distribution of aggrecan by examining IR in the math5 -/mouse line. Although innervation by layer VI fibers had not yet begun at P0, aggrecan-IR was substantially reduced in the perinatal mouse, and aggrecan was degraded in the same manner as the golli-tau-gfp with signal loss at the medial border of dLGN and increased IR towards the lateral edge ( Figure 12A). By P1, aggrecan-IR had been substantially reduced towards the lateral edges of dLGN; coincidentally, layer VI fibers began to enter the dLGN at this early age in the absence of retinal inputs. By P2, cat315 signal could no longer be observed throughout most of dLGN except the far lateral edges near the optic tract, and CG fibers had begun extending across the dLGN.
By P3 aggrecan-IR only covered ~5% of the dLGN and layer VI CG axons were observed to innervate significantly more area of dLGN compared with golli-tau-gfp mice (Figure 12 B). P1-P3 images of math5 -/mutants illustrate the principle that GFPlabeled layer VI fibers only enter the dLGN and extend to portions where aggrecan-IR is minimal (Figure 13). At ages P4 -P7, aggrecan proteoglycan expression remained nearly undetectable within dLGN, and the CG fibers continued to move laterally across the dLGN spanning 80% of dLGN by P7 ( Figure 12).
We compared the spatial extent of dLGN coverage by aggrecan-IR and GFP-IR however, the math5 -/mouse line exhibited extremely low aggrecan-IR with only 5-10% coverage of dLGN. The low levels of aggrecan throughout dLGN made for a permissive environment into which CG axons could begin innervation in all math5 -/mice. By P3, aggrecan levels had decreased dramatically and layer VI cortical fibers in the golli-taugfp mouse began to project slightly out of eml into dLGN, but the math5 -/model had a significant increase in the area of dLGN covered by layer VI fibers which correlated with the early degradation of aggrecan. P3 also represented the age at which both aggrecan and GFP-IR showed significantly different levels between golli-tau-gfp mice and math5 -/mice. At subsequent ages, aggrecan levels continued to recede in the golli-tau-gfp mice, while the CG fibers continued to innervate the dLGN until P14. From P4 to P7, the math5 -/line had a much quicker cortical fiber progression and significantly more dLGN area innervated. 80% of the dLGN was covered by GFP-containing fibers at P7 in the math5 -/animals ( Figure 14). Taken together, these data suggested that high levels of aggrecan in dLGN repel layer VI CG projections into dLGN.

Conclusions
Aggrecan was present at perinatal ages in high concentrations in dLGN and likely serves as an inhibitor of layer VI GFP-expressing axons that approach the dLGN.
Although transcription of aggrecan increased during postnatal development, the signal of aggrecan-IR decreased within the dLGN suggesting that it must be regulated either at a translational level or by other molecules in CNS. Aggrecan-IR was inversely correlated with the invasion of dLGN by CG fibers in both golli-tau-gfp and the math5 -/mouse models. These data strongly support a role for aggrecan in the inhibition of layer VI axon invasion of dLGN, and they implicate retinal input in the regulation of aggrecan.

Introduction
Although the inhibitory domain of aggrecan can repel RGC axons, regenerating fibers in spinal cord, and numerous other classes of axons, it does not repel all cell types (Snow et al., 1992;Silver et al., 2004). Serotonergic neurons in the spinal cord represent a class of cells that show enhanced outgrowth onto aggrecan rich substrate (Hawthorne et al., 2011). Having examined the expression of aggrecan and its potential role as a regulator of layer VI cortical afferents, I sought to characterize the direct effect of aggrecan protein on the growth and outgrowth of neurons located in layer VI of visual cortex. To better gauge if aggrecan was sufficient to repel CG fibers from layer VI, I used in vitro analysis to determine the effect of aggrecan on disassociated layer VI GFP-expressing neurons and their neurites. Once the signaling capabilities of aggrecan on layer VI axons were determined, I employed in vivo analyses to better understand the necessity of aggrecan in regulating the timing of CG innervation.

Aggrecan was sufficient to repel layer VI axons from dLGN
I utilized a modified stripe assay to determine whether aggrecan could repel dissociated cortical neurons. First I established that I could reproducibly grow healthy layer VI neurons in culture (data not shown). As a separate control, I spotted BSA conjugated to alexa-fluor 594, which I used to visualize the spots on the culture plates, at the same concentration that I mixed with aggrecan, to ensure that there were no effects, either attractive or repulsive, on the growth of layer VI neurons and their processes (data not shown).
At low concentrations of aggrecan (1µg/ml), neurites from layer VI neurons could project throughout the aggrecan-containing substrates and across the borders of the aggrecan covered area ( Figure 15A). At 5µg/ml and 10µg/ml, viable GFP-expressing neurons were not observed within the aggrecan containing substrates. Moreover, neurites from layer VI neurons would not extend into the aggrecan covered regions; therefore, aggrecan inhibited layer VI cortical neurite outgrowth (Figure 15 B,C). I also pretreated aggrecan with chABC, an enzyme derived from bacteria that effectively degrades the inhibitory CS-GAG side chains. Layer VI neurons could extend into regions of the chABC-treated aggrecan (Figure 15 D). The chABC treatment illustrated two important principles: First, the inhibitory domain of aggrecan was embedded in the GAG side chain domain, and it also showed that inhibition was not caused by a physical boundary due to a protein barrier. Quantification of the neurites' ability to cross into the aggrecan substrate revealed that high concentrations of aggrecan repel significantly more neurites than low concentrations of aggrecan and aggrecan that had been degraded by chABC (Figure 15 E).

Degradation of the CS domain of aggrecan alters CG innervation profile
After I established that aggrecan was sufficient to inhibit layer VI cortical neurite outgrowth in vitro, I was interested in testing whether premature digestion of the CS-GAGs attached to aggrecan would result in accelerated layer VI CG axon entry in vivo.
I performed bilateral intrathalamic injections of chABC in golli-tau-gfp mice to determine the necessity of aggrecan in CG timing and compared these data to age matched uninjected littermates. As an additional control to account for the CNS injury received during injection, I also injected littermates with PNase, an enzyme of comparable size and bacterial origin with no known effects within CNS tissue. I then analyzed CG innervation in each of the treatment conditions.
My experimental dosing regimen was based on previous chABC treatments used in spinal cord injury (SCI) therapies that were known to allow regenerating axons to overcome CSPG inhibition (Steinmetz et al, 2005;Massey et al., 2006;Alilain et al., 2011). These studies reported that chABC must be injected within 1mm of the area of interest in order for distribution to a specific site. After making a tiny incision in the scalp to locate coordinates on the brain to place my injection, I placed the micropipette through the cortex in the medial thalamus approximately 0.5mm from the dLGN to preserve cytoarchitecture and prevent gliosis within dLGN which might affect my results (Figure 16 A). Using a picospritzer, I delivered controlled volumes of drug into the CNS of P0-P1 golli-tau-gfp mice.
Upon examination of CG innervation, two days following drug delivery, I observed a significant increase in the area of dLGN covered by axons in the chABCinjected animals compared to uninjected age matched control littermates ( Figure 16B, C). PNase injected golli-tau-gfp mice exhibited a CG innervation phenotype that was insignificantly different from uninjected age matched littermates but showed significantly lower area of coverage when compared to the chABC injections ( Figure 16B,C).  Figure 17B).

Aggrecan prevents layer VI axons from entering the dLGN
The outgrowth profile of layer VI axons into thalamus demonstrated a vast difference between animals that had normal aggrecan profile and those that could not produce functional aggrecan. In several cases, I observed multiple fine CG fibers coursing into dLGN ( Figure 18).

Conclusions
In vitro experiments illustrated that aggrecan was repulsive to layer VI neurites in high concentrations, and the functional domain for inhibiting neurite outgrowth resided within the CS-GAG domain of aggrecan. An interesting observation from the in vitro analyses was a potential concentration dependent effect of aggrecan inhibition on layer VI neurites, suggesting a differential effect on axons within the population of layer VI cells. The destruction of aggrecan at perinatal ages with chABC in vivo led to accelerated CG innervation, complementing previous data from the math5 -/time series experiments that suggested early loss of aggrecan allowed accelerated CG entry into dLGN. Finally the acan cmd mutant mouse line provided evidence that layer VI GFPexpressing fibers could innervate the dLGN as early as P0 in the absence of aggrecan.
Taken together these data showed that aggrecan was sufficient to repel layer VI axons from the dLGN, and aggrecan was necessary in the proper spatiotemporal development of CG axons in thalamus.    Finally, I designed a model describing the mechanisms elucidated in this dissertation that regulate the timing of CG innervation.

Role of ADAMTS in timing of CG innervation
Despite the dramatic changes in aggrecan protein levels during postnatal dLGN development, preliminary qPCR analyses detected increases in mRNA transcript expression of the gene acan, which encodes aggrecan. Although there is not a 1:1 ratio that determines every transcribed gene product must be translated into protein, I was surprised by this data and suspected that there must be an extracellular enzyme responsible for the decline of aggrecan-IR signal in dLGN.
This experiment provided evidence for a potential mechanism for aggrecan degradation at the early ages, but I also needed to ensure that adamts family members continued high levels of expression in order to maintain low aggrecan levels. Thus, I employed a longitudinal microarray comparing levels of mRNA expression in dLGN of P3 and P8 animals, and I observed that even more "aggrecanases" were upregulated including adamts1, adamts8, adamts9, adamts10, adamts12, adamts15, adamts19 (Figure 20 B). I confirmed significant developmental increases in mRNA expression for adamts4, adamts8, adamts9, adamts15, and adamts19 through qPCR comparisons in dLGN at P2 and P14 (Figure 22 B).

ADAMTS family members produced by relay neurons
The transcriptonal profiles provided me with an excellent starting point for

Aggrecanases regulated by retinal input
Based upon the upregulation of aggrecanase mRNAs coincident with aggrecan degradation in golli-tau-gfp dLGN, I suspected that the expression or activity of these metalloproteinases may be regulated by retinal input. ISH probes for adamts4 and adamts15 were applied to math5 -/tissue to determine if any differences existed, pertaining to expression early in development, when aggrecan levels were already diminished in these mutants. I identified high levels of expression of adamts4 in these mutants at P3 and P6, and I also observed adamts15 expression in dLGN at P0, as well as, P3 and P6 (Figure 23 A).
I also performed a microarray comparing P3 dLGN mRNA expression in golli-taugfp and math5 -/mutants to determine if other guidance molecules were altered in the absence of retinal input. Although I did not find the expression of inhibitory guidance molecules, including acan expression, to be altered significantly in the math5 -/mutants, I discovered a modest increase in many of the adamts family members that contain aggrecanase activity, including adamts4, adamts8, adamts9, adamts15, and adamts16 (Figure 23 B). Adamts12 mRNA levels were significantly upregulated, and there was also a significant 27% increase amongst the entire family of adamts aggrecanases.
Since the ADAMTS enzymes have redundant functions, I believe that these increases provide ample explanation for early aggrecan degradation in the math5 -/mice.

ADAMTS4 degradation of aggrecan allowed accelerated CG invasion of dLGN
Next I tested whether aggrecanases contributed to the timing of CG innervation by injecting constitutively active, rhADAMTS4 into the thalamus of neonatal golli-tau-gfp pups. ADAMTS4, dubbed aggrecanase-1, was chosen over the other ADAMTS enzymes for multiple reasons. In contrast to most other ADAMTS enzymes, active ADAMTS4 was commercially available. Second, ADAMTS4 was capable of cleaving the aggrecan molecule at multiple sites within the IGD and CS domains of aggrecan, whereas, many of the other metalloproteinases were only capable of cleaving at one particular site on the aggrecan molecule (Tortorella and Malfait, 2008). Finally, in addition to cleaving the same sites as other ADAMTS enzymes, ADAMTS4 had an affinity for aggrecan that was several orders of magnitude higher than most others (Tortorella and Malfait, 2008). This combination provided confidence that exogenous delivery ADAMTS4 was the most suitable ADAMTS family member to specifically target and degrade aggrecan in vivo.
I utilized the same intrathalamic injection paradigm for ADAMTS4 that I used in the chABC experiments. When I assessed the amount of CG innervation at P3, I observed that layer VI fibers in the PNase injected animals resembled the phenotype in uninjected animals, but layer VI axons in the ADAMTS treated cohort traveled further into dLGN ( Figure 24A). Not only did GFP labeled fibers extend further into dLGN, but also the amount of area occupied by CG projections as detected by threshold analysis was significantly higher for the ADAMTS4-injected animals compared to both uninjected littermates and PNase-injected animals ( Figure 24B). Although both chABC and ADAMTS4 injections significantly altered CG innervation compared to controls, they did not show statistical differences between each other.

Pharmacological models of aggrecan degradation showed similar CG innervation profile
to math5 -/mutant mouse model

Conclusions
Microarray analyses provided evidence for upregulation of ADAMTS family members in dLGN, which was concomitant with the decreased signal in the aggrecan IHC experiments. These data confirmed a potential molecular mechanism for the regulation of aggrecan during the development of the visual system.  (Porter et al., 2005). ADAMTS4 recognizes the IGD cleavage domain with at least 100X greater efficiency than most other ADAMTS enzymes and can overcome aggrecan inhibition more effectively than MMPs (Cua et al., 2013). The major cause of aggrecan depletion is due to cleavage at the IGD site, however, the ADAMTS4 CS-domain cleavage has been hypothesized to result in bioactive fragments that could be conducive to neurite outgrowth, based on previous reports of bioactive molecules resulting from partially digested ADAMTS fragments from other CSPGs including versican and brevican (Cua et al., 2013;Sandy et al., 1991;McCulloch et al., 2009;Viapiano et al., 2008) (Sherman and Guillery, 2002). Waiting periods for growing axons have been reported in the chick hindlimb and the spinal cord, and the purpose of the waiting period is hypothesized to allow for maturation of axonal targets (Sharma et al., 1994;Wang and Scott, 2000;Deck et al., 2013). My results indicate that RGCs are beginning to form functional synapses and differentiation just prior to the degradation of the aggrecan stop signal and layer VI fiber entry into dLGN. These results suggest that the function of the aggrecan mediated delay of CG innervation is congruent with previous reports of waiting periods in the CNS.

Distinct cellular populations react differently to aggrecan
Aggrecan is one of the most well characterized CSPGs in the CNS. Reports have illustrated that the growth of many types of axons can be inhibited by aggrecan (Snow and Letourneau, 1992). In response to SCI, aggrecan secretion by reactive astrocytes limits the amount of axon regeneration at the glial scar, and aggrecan is the focus for therapies to facilitate functional regeneration in the injured spinal cord (McKeon et al., 1991;Silver and Miller, 2004;Busch et al., 2009). Recent evidence has been provided showing that serotonergic neurons in the spinal cord exhibit an unusual viability and axon sprouting in the aggrecan rich glial scar; therefore, aggrecan does not have the same effect on all cell types (Hawthorne et al., 2011). My results showed that layer VI cortical neurites were repelled by aggrecan in vitro. Combined with in vivo analyses and manipulations, I concluded that aggrecan was not only sufficient to repel layer VI axon outgrowth but also was necessary to delay CG fiber invasion in dLGN.
These results also presented a fascinating paradox. Previous studies have shown that aggrecan inhibits retinal axon outgrowth in vitro; however, aggrecan in dLGN prevented CG innervation from P0-P3 while permitting RG innervation and synapse development (Snow and Letourneau, 1992). My observations could be attributed to differing cells abilities to overcome aggrecan inhibition through interactions with particular growth-promoting substances, such as laminin or neurotrophin-3 (Snow et al., 1990;Fawcett, 2006). There is also evidence that embryonic neurons can overcome the inhibition presented by high concentrations of aggrecan by altering the expression levels of integrin receptors in their cell surface (Condic et al., 1999;Tan et al., 2011). Lack of aggrecan inhibition may also be explained, in part, by complexity of in vivo systems in which aggrecan can mediate inhibition through at least 4 different receptors including the leukocyte antigen-related protein tyrosine phosphatase receptor (LAR), protein tyrosine phosphatase sigma (PTPσ), nogo receptor 1 (NgR1) and nogo receptor 3 [NgR3 (Fisher et al., 2011;Dickendesher et al., 2012;Sharma et al., 2012)].
Interactions become even more complex as the CS-GAG domain of aggrecan is composed of at least five different CS sulfation patterns, and evidence has shown that these CS-GAGs have differing levels of inhibitory function (Gilbert et al., 2005). In the dynamic CNS environment, growth cones from distinct cell types could react differently to aggrecan due to the presence of growth factors or activity based receptor expression.

Activity dependent alteration of aggrecan in CNS
. Aggrecan expression can be altered in response to epileptic seizures, and CSPGs have been shown to be dysregulated in schizophrenic patients (McRae et al., 2010;Berretta, 2013). In the visual cortex, aggrecan expression is regulated by experience-dependent retinal activity. Numerous reports have indicated decreased aggrecan expression at the perineuronal net (PNN), a structure shown to limit plasticity after the closure of the critical period, in response to visual deprivation or dark-rearing during early visual development (Sur et al, 1988;Lander et al., 1997). The effect of the lost aggrecan enhances plasticity of neurons ensheathed in PNNs and prolongs the critical period (Kind et al., 2013).
Since altered retinal activity can change the expression of aggrecan, I performed experiments to gauge the effect of loss of retinal activity on aggrecan expression. The math5 -/mutants contain a targeted deletion of the Math5 transcription factor, which is crucial for differentiation of retinal progenitor cells to RGCs. In the absence of Math5, 95% of RGCs are never formed (Wang et al., 2001). Through math5 -/cross-breeding with the golli-tau-gfp mice, I was able to analyze the effects of silencing retinal input in dLGN on CG innervation, and I was also able to determine how molecular expression of aggrecan was altered in the absence of retinal instruction. The math5 -/mouse model was particularly useful because I did not have to stress the mice with surgery, such as encucleation, which could have triggered immune responses known to alter cellular activities in the brain. Aggrecan degradation in both systems occurred in a mediolateral pattern, thus establishing a gradient into which layer VI axons projected once aggrecan-IR was low. As reported in the math5 -/mice, I observed accelerated CG fiber innervation by 1-2 d compared to WT golli-tau-gfp mice, and I discovered that aggrecan signal was degraded 1-2 earlier in the math5 -/mice.
These results prompted me to use qPCR to probe acan gene expression in dLGN to see if the loss of retinal activity included a direct effect on the level of mRNA transcription; unfortunately, the aggrecan transcription mirrored normal WT aggrecan expression throughout development, and no significant difference in acan mRNA was observed in age matched golli-tau-gfp and math5 -/-dLGN samples. Microarray analyses revealed that there was a family of adamts genes that were upregulated in dLGN compared to vLGN at P3, and they continued to increase expression during the development period. Nearly half of the ADAMTS family members, including ADAMTS1, 4, 5, 8, 9, 12, 15, 16, 18, and 19, have been characterized as functional aggrecanases (Llamazares et al., 2007;Tortorella and Malfait, 2008).
My transcriptional profile analyses provided more questions than answers, as is often the case, and I had to take much care in designing a meaningful study due to redundant functions of ADAMTS members. The ADAMTS family of enzymes has selective activity on a plethora of extracellular proteoglycans including other members of the lectican family (Stanton et al., 2011).

Regulation and function of ADAMTS in CNS
Most of the ADAMTS metalloproteinases were established as aggrecanases in osteoarthritis studies outside of the CNS (Fosgang and Rogerson, 2010). Closely related molecules from the ADAM (a disintegrin and metalloproteinase) family have been identified as mediators of amyloid precursor proteins in Alzheimer's disease, but functional studies of ADAMTS family members in the CNS are in their infancy (Allinson et al., 2003;Bernstein et al., 2003;Krstic et al., 2012). This lack of characterization made it difficult to choose a particular ADAMTS/aggrecanase for my digestion experiments. ADAMTS12 has not yet been well characterized in terms of CNS function, like many of the ADAMTS members, but it has been shown to possess enzymatic activities directed towards aggrecan (Llamazares et al., 2007). Recent publications had shown some promising effects of ADAMTS4 in CNS, one of which reported ADAMTS4 mediated fuctional recovery in response to spinal cord injury (Tauchi et al., 2012).
Ultimately, I decided that this study must include ADAMTS4 because of its upregulation in dLGN at the early postnatal ages, high aggrecanase activity, and availability.
Although my experiments focused on ADAMTS4 degradation of aggrecan, I cannot discount the diversity of upregulated adamts mRNAs because they may not have only redundant functions, but they could offer more insight into how ADAMTS proteins work together during development to modify the ECM, in order to guide axons or promote synaptogenesis. Previous reports, outside of the CNS, illustrated coordination of ADAMTS5, ADAMTS9, and ADAMTS20 for the resorption of interdigital webs. ADAMTS2, ADAMTS3, and ADAMTS14 have been shown to synergistically degrade procollagens I and III (McCulloch et al., 2009;Le Goff et al., 2006).
Cooperative functions of ADAMTS were also suggested though evolutionary analysis that hypothesized that each ADAMTS might display some subfunctionalization and specialization in certain tissues (Huxley-Jones et al., 2005). Reports on the assimilation of multiple ADAMTS proteinases to degrade one class of molecule provide interesting prospects for future research, especially since I observed modest changes in the regulation of many of the adamts mRNAs in the microarray comparison of the early postnatal dLGN expression in golli-tau-gfp and math5 -/-. Although I did not find as many ADAMTS metalloproteinases to be altered significantly in the math5 -/-, I did discover a modest increase in many of the adamts family members that contain aggrecanase activity, including adamts4, adamts8, adamts9, adamts15, and adamts16. Adamts12 mRNA levels were significantly upregulated, and there was also a significant 27% increase amongst the entire family of adamts aggrecanases. Since the ADAMTS enzymes have redundant functions, I believe that these increases provide ample explanation for early aggrecan degradation in the math5 -/mice.
I extensively reported expression data in my analyses regarding ADAMTS, but I must also recognize the possible effects of molecules that contribute to the activity profiles of ADAMTS enzymes. Multiple regulators of ADAMTS metalloproteinase activity have been documented. Tissue inhibitors of metalloproteinases (TIMPs) contain 4 members, but currently only TIMP1 and TIMP3 have identified roles in the inhibition of ADAMTS metallproteinases (Murphy, 2011). Although neither was identified as significantly altered in expression analyses, maintenance of TIMP levels combined with modest increases in multiple ADAMTS proteins may contribute to greatly increased aggrecan degradation and accelerated timing of layer VI innervation in dLGN in the math5 -/set of experiments.
Another possible level of regulation for ADAMTS could be achieved by a completely different set of molecules from the TIMPS. ADAMTS are not immediately active upon translation, and they must have metalloproteinase propeptides cleaved, post-translationally, before they become enzymatically functional ). In addition to enzymatic digestion of the prodomain, ADAMTS4 activation occurs through C-terminal truncation by MMP-17 [(matrix metalloproteinase-17) Gao et al., 2002;Gao et al., 2004]. Aggrecanase activating and inhibiting factors can potentially be generated by RGCs and dLGN neurons ; therefore, retinal input may alter activities of the ADAMTS either directly through secretion of these factors or through induction of relay neurons to produce and secrete them (Kay et al., 2011).

ADAMTS versus MMP degradation of aggrecan
Earlier in this manuscript, I briefly made the case for using rhADAMTS4 in my experimental paradigms. In this section I will revisit this argument and compare ADAMTS4 with other MMPs, instead of just focusing on other ADAMTS aggrecanases, to solidify my justification for ADAMTS4 digestion of aggrecan in vivo.
MMP-3, -7, and -8 are all able to reverse the inhibition of neurite outgrowth due to the presence of CSPGs, so they appear to have as much potential as ADAMTS4 to degrade aggrecan if used in vivo. However, ADAMTS4 can cleave multiple fragments of aggrecan, and it has been hypothesized that partial ADAMTS4 digestion of aggrecan results in bioactive fragments that could produce neurite outgrowth, much like what has been shown to occur when ADAMTS molecules cleave versican and brevican in vivo (Sandy et al., 1991;McCulloch et al., 2009;Viapiano et al., 2008).
Another advantage of ADAMTS4 over MMP digestion in vivo is the inability of ADAMTS4 to process laminin, which has been to shown to aid in the suppression of CSPG inihibitory signaling (Cua et al., 2013;Snow et al., 1996). Furthermore, although both ADAMTS4 and many MMPs cleave a multitude of differing target molecules within the CNS,exogenous delivery of MMPs has been shown to cause neurotoxicity in vivo due to a lack of localized, targeted protease activity (Xue et al., 2009;Gu et al., 2005;Cua et al., 2013).
ADAMTS4 has a proteolysis-independent mechanism for inducing neurite outgrowth, and it has been shown to be effective in allowing axon regeneration in response to spinal cord injury (Hamel et al, 2008;Tauchi et al., 2012). Finally, ADAMTS4 has been reported to be more effective than chABC in decreasing CSPG inhibition, and chABC has been well characterized for its effectiveness in restoring axon outgrowth in the spinal cord (Cua et al., 2013). Together these reports led me to believe that ADAMTS4 could potentially be the most effective route for aggrecan degradation in vivo.

Possible role for aggrecan in establishment of RF circuitry in dLGN
Much of this dissertation focused on the mechanisms that keep layer VI fibers out of dLGN until RG axons begin to mature, but little has been discussed regarding how aggrecan might affect RG synapse formation with relay neurons. Retinal activity coordinates the establishment of proper eye-specific domains in the mouse dLGN, and the combination of repellent ephrin-A molecular guidance molecules and retinal activity work together in the development of topographic maps in retino-recipient nuclei (Feldheim et al., 1998;Torborg and Feller, 2005;Pfeiffenberger et al., 2006;Feller, 2009.) Furthermore, topographic mapping in SC and V1 are aligned through the generation of normal spontaneous retinal waves (Triplett et al., 2009). These reports support the notion that RF connections with relay neurons are very important in the establishment of visual circuitry.
RGCs innervate the dLGN and form immature synapses at perinatal ages in the mouse (Godement et al., 1984;Hong and Chen, 2011). During the first postnatal week, when aggrecan is abundant in dLGN, 10-20 weak synapses from different RGCs converge onto single relay neurons (Chen and Regehr. 2000). During this competition phase, only the most suitable synapses will become functional, and the rest will be retracted. Aggrecan could have multiple functions during this period. First, CS-GAGs have been shown to sequester growth factors, masking them from binding receptors, until a sheddase, like any of the ADAMTS aggrecanase enzymes, degrades the CS domains and allows the growth cone to interact with the growth factors (Muir et al., 1989;Sanes, 2003). As the axons compete for space, correctly formed RGC inputs could prompt aggrecanse distribution to a specific location leading to the unmasking of growth factors that promote synaptic differentiation and maturation.
RGCs express different guidance receptors which allow them to traverse to different sides of the brain. Expression patterns of molecules, or combinations of expression patterns, also lead to the divergent pathways when selective targeting occurs in the CNS (Petros and Mason, 2008). As aggrecan is degraded from the medial edge to lateral border, RGC axons containing multiple CSPG receptors or extremely sensitive receptors could defasciculate and move to areas where aggrecan has been degraded. RGC axons that have no CSPG receptors or express high levels of CAMs will form synaptic contacts in the areas where aggrecan levels are still robust. This axon sorting model fits well with the competition phase because growth factors would become more abundant to particular growth cones as sensitive fibers move to more hospitable areas.
The assembly of neural circuits requires both specific temporal and spatial organization. Mechanisms that drive the spatial control of axon guidance have been well described, but the function and regulation of temporal controls of axon pathfinding remain largely unknown. This manuscript details both a molecular mechanism for temporal control of axons and its activity based regulation. My results suggest that the timing of dLGN innervation by layer VI CG axons is regulated by aggrecan, and I provide evidence for a signaling cascade by which transcriptional control of ADAMTS/aggrecanases is mediated by RGC activity. Furthermore, these results support the recent claim that retinal inputs instruct the timing of layer CG innervation in dLGN . This study provides insight into the role of waiting periods in the establishment of functional connectivity in CNS and reveals a novel function for aggrecan in the developing nervous system.