The mammalian brain is capable of remarkable reorganization during development and throughout the life of an individual. A variety of studies, particularly on the visual system, have demonstrated that sensory experience plays a large role in shaping appropriate connection patterns in the nervous system, and that a loss or alteration of sensory receptor input can result in large organizational changes in the brain (for reviews, see Katz and Shatz, 1996; Shatz, 1996; Sur et al., 1999; Krubitzer and Kahn, 2003; Ruthazer and Cline, 2004). Indeed, with a complete loss of sensory receptor arrays or sensory-driven activity very early in development, such as in congenital blindness or deafness, reorganization can be massive and portions of the neocortex that would normally be occupied by the lost sensory system will be taken over by the remaining sensory systems (Hunt et al., 2002; Kahn and Krubitzer, 2002). This has recently been demonstrated in bilaterally enucleated Monodelphis domestica (Fig. 1)
Until recently, the notion that humans who develop in the absence of one sensory system (e.g., are blind or deaf) become better at making sensory discriminations with the remaining sensory systems was, for the most part, anecdotal. However, recent studies in blind individuals indicate that the capabilities of other sensory systems in some ways exceed that of sighted individuals. Using auditory event-related potentials, Röder et al. (1999) demonstrated that there is a shorter detection time for auditory discrimination tasks in blind versus sighted individuals and in a subsequent study (Röder et al., 2000), they demonstrated that blind individuals process language faster than sighted individuals. Thus, there appears to be a compensatory adaptation of the auditory system in the congenitally blind. This compensatory adaptation is likely due to changes in the nervous system at both subcortical and cortical levels. This notion is supported by recent positron emission tomography (PET) studies in blind individuals, which indicate that auditory localization activates occipital cortex in regions normally involved in visual localization and motion detection (Weeks et al., 2000). This massive cross-modal plasticity is also observed in the somatosensory system. For instance, in congenitally blind individuals, primary visual cortex is active during tactile tasks such as Braille reading (Sadato et al., 1996; Cohen et al., 1997; Büchel et al., 1998).
Similar types of cross-modal plasticity have been demonstrated in congenitally deaf humans. For instance, congenitally deaf individuals have enhanced amplitudes of the ERP's N1 component for processing of visual motion (Armstrong et al., 2002) and when attending to the peripheral visual field compared to normal individuals (Neville and Lawson, 1987). Functional imaging studies indicate that auditory areas are active during visual and somatosensory tasks in congenitally deaf individuals (Catalan-Ahumada et al., 1993; Levanen et al., 1998; Finney et al., 2001; for review, see Bavelier and Neville, 2002). While these studies clearly demonstrate that cross-modal plasticity in the developing mammalian brain can be extensive, the anatomical substrate for these functional changes is not known. For example, how do the remaining sensory systems invade cortical territory normally occupied by the lost sensory system? Are these large territory shifts in the neocortex of congenitally deaf individuals the result of changes in thalamocortical connectivity and corticocortical connections, or do some alterations in connectivity, which ultimately result in large functional changes, occur early in sensory processing at the level of the retina or primary afferents in the somatosensory system?
Recently, we have begun a series of studies on two types of mutant mice, one lacking the Na+-K+-2Cl− cotransporter (NKCC1) (Flagella et al., 1999) and the other lacking the plasma membrane Ca2+-ATPase isoform2 (PMCA2) (Kozel et al., 1998). These mice are well suited for our study because the functional and anatomical abnormalities of the cochlea have been well characterized. NKCC1 null (NKCC1−/−) and PMCA2 null (PMCA2−/−) mice are congenitally deaf due to abnormalities in the morphology and function of the cochlea. These mice have a highly reduced eighth nerve, fail to develop an endocochlear potential, and have no auditory brainstem response (Kozel et al., 1998; Flagella et al., 1999). Thus, the developing nervous system of these mice never has access to sensory-driven activity from the cochlea. Studying these mice has allowed us to examine the type of compensatory anatomical changes that occur in other sensory systems, which can account for the functional changes observed in the neocortex (Hunt et al., 2002). Further, these studies have allowed us to address larger questions regarding the role of patterned activity from peripheral sensory receptor arrays in the formation of appropriate connections in development, and the extent to which any particular pattern of connectivity that defines a given sensory system is the result of relative activity patterns between all sensory systems.
In the current study, we compared the retinofugal pathway in normal and congenitally deaf mice by making intraocular injections of neuroanatomical tracers into each eye, and relating transported tracers to architectonically defined nuclei in the thalamus and layers of the superior colliculus.
MATERIALS AND METHODS
NKCC1 and PMCA2 Knockout Mice
Both the basolateral Na+-K+-2Cl− cotransporter (NKCC1) and plasma membrane Ca2+-ATPase isoform 2 (PMCA2) knockout mice were originally analyzed on a mixed Black Swiss and 129/SvJ background (Kozel et al., 1998; Flagella et al., 1999).
NKCC1, which mediates the electroneutral transport of 1 Na+, 1 K+, and 2 Cl− into the cell, is a member of the cation (Na+ and/or K+)-coupled chloride transporter family found in a large number of cell types. In the nervous system, these proteins are found in the olfactory bulb, the cerebellum, the inner ear, and transiently in the cortex and hippocampus (late embryonic to early postnatal) (Kaplan et al., 1996; Hübner et al., 2001; Kanaka et al., 2001). Despite the important role of NKCC1 in cellular ion homeostasis, NKCC1−/− mice live for at least 60 weeks in the laboratory setting, well into adulthood.
Histological analysis of the inner ear of NKCC1−/− mice reveals several striking abnormalities, including a collapsed Reissner's membrane and virtually no scala media lumen, accumulation of calcified material on the tectorial membrane, an absence of the tunnel of Corti (although this was not consistently observed), loss of inner and outer hair cells, a reduction of spiral ganglion cells, and abnormalities in marginal cells (Flagella et al., 1999; Pace et al., 2001). NKCC1−/− mice have no endocochlear potential (Dou et al., 2000) and no auditory brainstem response (ABR) for broadband clicks or pure tone frequencies of 8–32 kHz at 20–100 db (Flagella et al., 1999). Thus, these mice are profoundly deaf.
While NKCC1 is highly expressed in only a few portions of the nervous system (Kanaka et al., 2001), the possibility that the changes in sensory system pathways that occur are the result of abnormalities of portions of the nervous system other than the inner ear cannot be completely ruled out. One way we attempted to circumvent this problem was to use two different knockouts with one major overlapping phenotypic characteristic, a morphologically aberrant and functionally inactive cochlea. We reasoned that while the loss of each particular gene may have a number of effects, similar changes in the nervous system that occur in both types of mutant mice are likely to be from cochlear dysfunction (the only common feature they share). Therefore, we also used mice lacking PMCA2 in these experiments. PMCA2 is expressed in hair bundles of cochlear hair cells (Dumont et al., 2001) and in cell bodies in the Purkinje cell and molecular layer of the cerebellum (Stauffer et al., 1997; Yamoah et al., 1998). In the inner ear, PMCA2 is involved in the extrusion of Ca2+ into the endolymph. PMCA2−/− mice have both hearing and vestibular disorders, as well as a range of abnormalities in the organ of Corti, although not as dramatic as the NKCC1−/− mice. Auditory brainstem responses are absent in PMCA2−/− mice for broadband clicks and for pure tones ranging from 8 to 32 kHZ at sound pressure levels of 20–100 db (Kozel et al., 1998). These mice, like the NKCC1−/− mice, are profoundly deaf.
Surgical Procedures and Injections of Neuroanatomical Tracers
Retinofugal connections were examined in four normal adult mice and six congenitally deaf (four NKCC1−/− and two PMCA2−/−) adult mice (3–7 months). Anesthesia was induced by placing the animals in an anesthesia chamber and administering 2.5% isoflurane. Once anesthetized, the animal was placed in a stereotaxic frame and anesthesia was delivered at 1.25–1.75% through a specially fitted cone place over the snout. Atropine (0.06 mg/kg) and doxapram hydrochloride (1.5 mg/kg) were administered intramuscularly prior to surgery. Small subcutaneous injections of 2% lidocaine were made around the eyes. Body temperature, respiration rate, and heart rate were monitored throughout the experiment.
Table 1. Neuroanatomical abbreviations
responds to auditory stimuli
Primary auditory cortex
Anterior pretectal nucleus
nucleus of the brachium of the inferior colliculus
brachium of the inferior collculus
commissure of the superior colliculus
deep superior colliculus
deep gray layer of superior colliculus
deep white layer of superior colliculus
External cortex of the inferior colliculus
intermediate gray layer of the superior colliculus
intermediate white layer of the superior colliculus
lateral geniculate nucleus, dorsal part
laternal geniculate nucleus, ventral part
lateral geniculate nucleus, ventral, lateral
lateral geniculate nucleus, ventral, medial
lateral posterior nucleus
medial geniculate nucleus, dorsal division
medial geniculate nucleus, magnocellular division
medial geniculate nucleus
medial geniculate nucleus, ventral division
medial habenular nucleus
medial mammillary nucleus, medial part
medial terminal nucleus
optic nerve layer of the superior collliculus
parafascicular thalamic nucleus
posterior intralaminar thalamic nucleus
responds to somatic stimuli
primary somatosensory cortex
substantia nigra, reticular part
superficial gray layer of the superior colliculus
responds to visual stimuli
primary visual cortex
ventral posterior nucleus
ventral posterolateral thalamic nucleus
ventral posteromedial thalamic nucleus
zona incerta, dorsal part
zona incerta, ventral part
Intraocular injections were made with a 30 gauge needle attached to a 10 μl Hamilton syringe. The eye was gently manipulated to expose the sclera, the needle was inserted into the vitreous humor, and the tip of the syringe was visualized. Each animal received a 5 μl injection of 0.5% choleratoxin subunit B-rhodmine (BTRITC; Molecular Probes, Eugene, OR) into the right eye and a 5 μl injection of 0.5% fluorescein isothiocyanate isomer conjugate (BFITC; Molecular Probes) into the left eye. This volume of tracer has been used successfully to examine retinofugal projections in small rodents, including mice (Godement et al., 1984; Fukuda et al., 1986; Mikkelsen, 1992; Uchiumi et al., 1995). After each injection, sterile ophthalmic ointment (Neosporin) was topically applied to prevent leakage of tracer out of the eye and to prevent infection. Two days were allowed for the transport of the tracer, after which the animals were euthanized with pentobarbitone sodium (60 mg/kg) administered intraperitoneally, and transcardially perfused with 0.9% saline followed by 4% paraformaldehyde in phosphate buffer (PB; pH 7.4) and 4% paraformaldehyde with 10% sucrose in PB (pH 7.4). The brain was removed from the skull and cryoprotected overnight in 30% sucrose in PB. All brains were cut coronally at 35 μm on a freezing microtome and sections were stored in 0.1 M PB. All experimental procedures were approved by the Animal Use and Care Administrative Committee of the University of California, Davis, and conformed to NIH guidelines.
For all cases, alternate sections were mounted for fluorescence microscopy or stained for Nissl substance. In two cases (02-113 and 03-10), an extra series was processed for cytochrome oxidase (CO) (Carroll and Wong Riley, 1984), and in one case (03-120), an extra series was processed for myelin (Gallyas, 1979).
To verify injection sites, the eyes were removed from the orbit, the retinas were dissected from the sclera and pigment epithelium and flattened using four radial cuts. The flattened retina was mounted for fluorescence microscopy. Two retinae, which did not receive injections, were flat-mounted and prepared for fluorescence microscopy. These served as controls for our injected retinae.
Anterogradely labeled axon terminals and portions of axons were plotted using a Nikon E400 fluorescence microscope attached to an Accustage MD3 Digitizer and MD Plot software (Accustage, Shoreview, MN). All plotted sections included the entire outline of the section, blood vessels, and tissue artifacts. These types of reconstructions were made of every other section mounted for fluorescence microscopy throughout the entire midbrain and dorsal thalamus. Similar reconstructions were made of the flat-mounted retinae. Architectonic boundaries of midbrain and thalamic structures were made by aligning blood vessels, tissue outlines, and other tissue artifacts of adjacent Nissl-, CO-, or myelin-stained sections using a camera lucida attached to a Zeiss Stemi SV6 stereoscope. Comprehensive reconstructions were scanned using HP Scanjet 6300 Scanner and Software (HP, Palo Alto, CA), and illustrations were created using Canvas 9.0 (Deneba, British Columbia, Canada). Digital photomicrographs were taken with a Spot RT Slider digital camera and Spot camera software (Diagnostic Instruments, Sterling Heights, MI), and Adobe Photoshop 7.0 (Adobe, San Jose, CA) was used to adjust contrast and brightness of digital photomicrographs so that they most closely resembled the actual tissue.
Injection Sites and Uptake Zones
In all cases in which the retina was dissected and reconstructed, several features were noted. The first was a very brightly fluorescing region, which marked the location of the Hamilton syringe entry and the largest deposit of the tracer. The second feature was a lighter homogeneous fluorescence throughout the retina. This marked the uptake zone of the injection. The third feature was individually fluorescing cells indicative of local retinal connections from the core of the injection. Finally, labeled ganglion cell axons were observed traversing the retina and exiting through the optic disk. All of these features were observed for both the BTRITC and BFITC injections and in both normal and mutant animals. None of these features were observed in control retinas that did not receive an injection of a neuroanatomical tracer.
Cytoarchitecture, Myeloarchitecture, and Chemoarchitecture of the Dorsal Thalamus and Superior Colliculus
The architectonic appearance of thalamic nuclei and lamina of the superior colliculus have been described for a number of rodents, including the mouse (Paxinos and Franklin, 2001), rat (Jones, 1985; Paxinos and Watson, 1998), and squirrels (Krubitzer and Kaas, 1987). In normal and mutant mice, nuclei such as the dorsal and ventral lateral geniculate nucleus (LGd and LGv, respectively; see Table 1) are well differentiated using Nissl stains. These nuclei contained moderately stained, densely clustered cells and were bounded laterally by the optic tract and were separated from each other by the cell-sparse intergeniculate leaflet (Fig. 2A and C). In sections stained for cytochrome oxidase, these nuclei are dark and homogeneously stained (Fig. 2D). In sections stained for myelin, both the LGd and LGv were very lightly stained and were surrounded by myelin-dense axons (Fig. 2B). No apparent lamina was visible in the LGd with any of the stains utilized. Posterior to the LGd and LGv is the medial geniculate complex, which is composed of three major divisions: the dorsal, ventral, and magnocellular nuclei (MGd, MGv, and MGm, respectively). In normal and mutant mice, MGd was composed of moderately sized cells and at caudal levels, resided at the lateral edge of the thalamus (Figs. 2I and K and 3B, D, and F), just ventral to the pretectum. MGv contained darker, more densely packed cells (Figs. 2I and K and 3B, D, and F), while MGm contained larger and more loosely packed cells and was located just medial to MGd and MGv. MGd and MGv stained moderately and homogeneously for CO (Fig. 2H and L), and very lightly for myelin (Fig. 2F and J). It was difficult to distinguish MGm with these stains. MG appeared to be somewhat larger in normal mice compared to both strains of mutant mice (compare Fig. 3B and D with F), but we did not attempt to quantify this difference.
The superior colliculus is composed of a number of layers, which can be distinguished using Nissl, CO, and myelin stains. The SC is traditionally divided into superficial, intermediate, and deep layers, and we have used this terminology. In both normal and deaf mice, the superficial layers of the SC contained a cell-sparse zonal layer (ZO), a cell-dense superficial gray layer (SuG), and a lightly stained optical layer (Op; Fig. 6D). SuG stained darkly for CO, and Op was lightly stained for CO (Fig. 6B). The intermediate gray layer (InG) was cell-dense, stained moderately for CO (Fig. 6), and lightly for myelin. The intermediate white layer (InWh) was cell-sparse, myelin-dense, and stained lightly and less homogeneously for CO (Fig. 6B). The deep layers consist of a moderately packed gray layer (DpG) that stained darkly for CO, and cell-sparse, myelin-dense white layer (DpWh) that stained lightly for CO.
Retinal Projections to the Dorsal Thalamus in Normal Mice
Bilateral retinal injections were made in four normal adult animals for a total of eight injections. Examination of the retinae revealed that in all cases, the injection site comprised much of the retina. Six of these injections resulted in anterogradely transported tracer in contralateral and ipsilateral midbrain structures and thalamic nuclei. In all cases in which transported tracer was observed, labeled axon terminals were found bilaterally in the LGd and LGv. The LGd contralateral to the injected eye was almost completely filled with labeled axon terminals (Fig. 4, middle column) and contained a very small central region that was devoid of label. For all six injections, the LGd ipsilateral to the injection site contained a small, tightly packed cluster of axon terminals. In cases in which transported tracer from both the left and right eye was observed in a single section (e.g., 02-24; Fig. 4), it was found that the ipsilateral and contralateral eye input to the LGd formed a complementary, nonoverlapping pattern that has been described in a number of other rodents (Mikkelsen, 1992).
Labeled axon terminals were also observed in the ipsilateral and contralateral LGv. As with the LGd, the contralateral LGv contained labeled axon terminals that almost completely filled this nucleus (Fig. 4, middle column). However, there were a few small unlabeled portions of LGv. Ipsilateral label in LGv was sparse and formed one or several small islands. In the cases in which injections from both eyes resulted in transported tracer, it was observed that the contralateral and ipsilateral label in LGv formed complementary patterns, as with the LGd (Fig. 4, middle column).
Labeled axon terminals were observed in the contralateral intergeniculate leaflet (IGL) in all cases in which transported tracer was observed, and for two injections, in the ipsilateral IGL as well. The labeled terminals from the contralateral injection were dense and in most cases did not completely fill the nucleus. Labeled terminals resulting from the ipsilateral eye injection were sparse and formed several small islands.
Retinal Projections to the Dorsal Thalamus in NKCC1−/− and PMCA2−/− Mice
Four adult NKCC1−/− mice and two adult PMCA2−/− mice received bilateral eye injections of neuroanatomical tracers. Ten of the 12 injections resulted in transported tracer in the thalamus and midbrain. As with normal animals, dense label was observed in the contralateral LGd, and in most animals, the labeled axon terminals almost completely filled the nucleus (Fig. 4, left and right columns). We observed a small terminal free zone within the LGd in all cases, as in normal animals. Ipsilateral label in the LGd for all cases in which transported tracer was observed was very sparse and formed a small island of terminal label. When direct comparisons were made of projection patterns of both eyes, it was found that, as in normal animals, the ipsilateral and contralateral patterns of terminal label were complementary (Fig. 4, right column). Terminal label in the LGv was similar to that described in normal animals in that projections from the contralateral eye were extremely dense and almost completely filled the nucleus. Ipsilateral terminal label was sparse and formed a complementary pattern to the contralateral label. Finally, the IGL received both ipsilateral and contralateral input that was patchy and partially overlapping in all cases (Fig. 4).
Aberrant retinal projections to the dorsal thalamus were observed in both mutants. In both the PMCA2−/− and NKCC1−/− mice, densely packed terminal labeling was observed in the contralateral MGv for all injections (Figs. 3A and C and 5, left and right column). Terminal label in MGv formed patches of dark and light islands (Figs. 3 and 5). Labeled terminals were also observed in the contralateral MGd for 8 of the 10 injections (Fig. 5). Labeled terminals here were less abundant and were observed in the lateral portion of the nucleus. Four of the 10 injections resulted in label in the ipsilateral MGd and 2 of the 10 injections resulted in ipsilateral label in the MGv. In normal mice, no projections to the medial geniculate nucleus were observed (Figs. 3E and F and 5, middle column).
Retinal Projections to the Superior Colliculus in Normal Mice and Mutant Mice
Injections in both normal and mutant mice resulted in dense labeled terminals in the superficial layers of the contralateral SC, which include the SuG and the Op (Fig. 7, middle column). Terminal label was most dense in the SuG and moderately dense in the Op and was mostly confined to the dorsal portion of Op. SuG and Op contained labeled terminals throughout their rostrocaudal extent and, in all but one case, label was homogeneous throughout its extent. In all cases in which transported tracer was observed, small, punctate patches of labeled terminals were observed in the ipsilateral superior colliculus. In both normal and mutant mice, the ipsilateral projection was in Op (Fig. 7, middle column).
Aberrant projections were observed in the contralateral superior colliculus for all injections in mutant mice in which transported tracer was observed (Figs. 6 and 7). At the rostral portion of the SC, at the juncture between the SC and the pretectum, small, punctate patches of labeled terminals were observed throughout the mediolateral extent of the intermediate gray layer (Figs. 6C and 7, left and right column). The patchy label identified in InG on first inspection looks like it could be assigned to the pretectal nucleus. However, this label is located caudal of this nucleus, and examination of adjacent Nissl-stained sections indicate that it is indeed in the InG. Throughout the rostral to middle extent of the SC, terminal label filled the Op and extended into the InG in all mutant mouse cases (Fig. 7). This was not observed in normal mice.
Other Retinal Projections in Normal and Mutant Mice
For all injections in both normal and mutant mice, terminal labeling was observed in the contralateral pretectum (Fig. 4). Label formed very dense clusters surrounded by more moderate labeling. Likewise, for all cases, labeled terminals were observed in the ipsilateral pretectum. Terminal labeling in the ipsilateral pretectum was much less extensive and formed small patches. Finally, for all injections in both normal and mutant mice, dense terminal labeling was observed in the contralateral medial terminal nucleus (MT; Fig. 5, bottom row).
We examined the retinal projections to the dorsal thalamus and midbrain structures in normal and two strains of congenitally deaf mice. The data demonstrate that while most of the projections are similar in all three groups, congenitally deaf mice have two alterations in their retinal connections. The most notable alteration is a projection to the ventral and dorsal division of the contralateral medial geniculate nucleus of the thalamus (Fig. 8). The second aberrant projection observed was with the intermediate layers of the contralateral superior colliculus (Fig. 8). In the following sections, we provide an overview of what is known about the normal projections of the eye in rodents and discuss the implications of our results for appropriate targeting of sensory system-specific connections, which we term “sensoritopic matching.”
Retinofugal Projections in Rodents
Retinal projections have been examined in a wide variety of rodents including mice (Mikkelsen, 1992), rats (Lund et al., 1974; Reese and Cowey, 1983, 1987; Mikkelsen, 1992), ground squirrels (Kicliter and Bruce, 1983; Petry et al., 1989; Lugo and Kicliter, 1995; Major et al., 2003), eastern gray squirrels (Cusick and Kaas, 1982), chipmunks (Fukuda et al., 1986), gerbils (Mikkelsen, 1992; Fite et al., 2003), moles (Herbin et al., 1994), hamsters (Crain and Hall, 1980; Mikkelsen, 1992; Ling et al., 1998), guinea pigs (Lazar, 1983), and voles (Uchiumi et al., 1995). The present investigation of retinal projections is consistent with all previous studies in rodents (except the blind mole rat) in that all species have dense contralateral projections to the LGd, LGv, and IGL of the dorsal thalamus, and to the superficial layers of the superior colliculus. Similarly, all rodents have ipsilateral projections to the LGd, LGv, IGL, and the optical layer of the superior colliculus.
An important and consistent observation in all studies of retinofugal projections in rodents is that no connections were found with the medial geniculate nucleus or with the intermediate layers of the contralateral superior colliculus. Divisions of the medial geniculate nucleus receive input predominantly from the central nucleus of the inferior colliculus, which is a major target of most auditory brainstem nuclei (Beyerl, 1978; Nordeen et al., 1983; Coleman and Clerici, 1987; Malmierca et al., 2002; Winer et al., 2002). The intermediate layers of the superior colliculus receive input from auditory cortex (Budinger et al., 2000), and subcortical structures such as the nucleus of the brachium of the inferior colliculus, the external nucleus of the inferior colliculus, and nuclei of the lateral lemniscus (Edwards et al., 1979; Druga and Syka, 1984; Cadusseau and Roger, 1985; King et al., 1998), and contain neurons that respond to auditory stimulation (Drager and Hubel, 1975; Middlebrooks and Knudsen, 1984; Hirsch et al., 1985; King and Hutchings, 1987; Gaese and Johnen, 2000; for review, see King, 1999). Thus, the aberrant projections of the retina in our mutant mice were to subcortical auditory structures, which ultimately project to and receive projections from auditory cortex. This suggests that compensatory adaptation due to loss of sensory-driven activity from any one sensory system occurs at very early stages of sensory processing and has a cascade of effects at all stages of processing, including the neocortex.
It should be noted that aberrant retinal projections to the medial geniculate nucleus have been surgically induced in both hamsters and ferrets by lesioning central structures such as V1, the superior colliculus, and/or the brachium of the inferior colliculus in developing animals (Schneider, 1973; Kalil and Schneider, 1975; Frost, 1981, 1986; Sur et al., 1988; Roe et al., 1993; Pallas et al., 1994). Interestingly, these latter authors conclude that the modality of a sensory thalamic nucleus is determined by its inputs during development (Sur et al., 1988). However, this model is different in two respects from the mouse models used in the current study. The first difference is that the aberrant projections observed in the present investigation are the consequence of alterations in sensory-driven activity from the cochlea, rather than the removal of two major targets of the retina, the superior colliculus and the LGN (via retrograde degeneration from V1 lesions) or removal of a brainstem source of auditory afferents, the brachium of the inferior colliculus. This difference is important because it distinguishes top-down rewiring, or removal of retinal targets, from bottom-up rewiring, involving the loss of auditory afferents. The second difference is that the alterations described here may be analogous to what normally occurs during evolution and may provide some insights into the developmental mechanisms, both activity-dependent and molecular, that give rise to phenotypic variability in subcortical and cortical organization in mammals. These naturally occurring developmental changes are likely to occur at multiple levels of the nervous system such that genetically mediated reduction of peripheral afferents leads to decreases in activity of the normal target. This target in turn may fail to express molecules that restrict afferent terminations. Such a series of events would allow novel pathways to form.
Relative Activity Between Sensory Systems Is Necessary for Sensoritopic Matching
The role of activity for appropriate circuit formation has been described for all levels of the nervous system (for reviews, see Katz and Shatz, 1996; Ruthazer and Cline, 2004). The visual system has traditionally been the model system for understanding the role of activity in the formation of layer-specific zones in the lateral geniculate nucleus and ocular dominance columns in the visual cortex. The consensus across studies is that retinal ganglion cell axons from each eye are initially intermixed in the LGN. Over time, retinal inputs form eye-specific domains within the LGN due to differences in spontaneous and driven activity from each eye (Shatz, 1983; Penn et al., 1998; for review, see Shatz, 1996). For ocular dominance columns in the primary visual cortex, initial afferents from the LGN are largely overlapping in layer IV, and gradually segregate into eye-specific domains during a prenatal and postnatal period, with the onset of patterned retinal activity (Hubel and Freeman, 1977; LeVay et al., 1980; for review, see Ruthazer and Cline, 2004).
While the role of both spontaneous and patterned activity in the formation of appropriate connections has been well investigated within a particular sensory system, specifically the visual system, there is little understanding of how the magnitude of activity in any one sensory system and the relative activity patterns across sensory systems contribute to the establishment of appropriate sensoritopic connections. For example, what are the mechanisms by which a particular subcortical or cortical structure assumes a unimodal or multimodal function? Comparative studies in highly derived species provide some insight into this issue. The blind mole rat is a subterranean rodent with atrophied, subcutaneous eyes, a greatly reduced optic nerve, and specialized skin (the buccal ridge), which completely covers the eyes (Klauer et al., 1997; David-Gray et al., 1998). Thus, the overall activity generated from the retina is greatly reduced compared to the activity generated from the cochlea, the somatic receptors, and the olfactory epithelium. The visual system in these animals is involved predominantly in the establishment and maintenance of circadian rhythms (Cooper et al., 1993; David-Gray et al., 1998). 2-deoxyglucose and electrophysiological recording studies in this animal demonstrate that the lateral geniculate nucleus is activated by auditory stimulation, as is the visual cortex (Bronchti et al., 1989, 2002; Heil et al., 1991; Sadka and Wollberg, 2004). Interestingly, early in development, normal retinogeniculate patterns of connectivity can be observed, but later degenerate and are absent in the adult (Bronchti et al., 1991; Cooper et al., 1993). In the absence of activity from the retina, the lateral geniculate nucleus is taken over by inputs from the inferior colliculus (Doran and Wollberg, 1994). These data indicate that with large alterations in the magnitude of activity across sensory systems, subcortical structures associated with the reduced system are taken over, at least in part, by other sensory systems, and this in turn appears to have a very large effect on the organization and function of the neocortex.
Our results of aberrant visual inputs to subcortical structures normally associated with auditory processing are analogous to the innervation of visual structures by auditory inputs in the blind mole rat. Further, recent results on the neocortex of these congenitally deaf mice indicate that massive reorganization occurs in that auditory cortex is taken over by somatic and visual inputs (Hunt et al., 2002). This result is similar to the results of functional imaging studies in congenitally deaf humans described earlier. Obviously, this functional reorganization of the neocortex is at least partly accounted for by alterations in connections at the earliest stages of sensory processing. Given that similar types of cross-modal plasticity have been observed in congenitally deaf humans and that recent studies demonstrate that these individuals have heightened capabilities for the other senses, these aberrant connections are functionally optimized and may allow for enhanced sensory performance of remaining sensory systems.
Contribution of Activity and Molecular Cues to Sensoritopic Organization
The role of particular molecules in establishing appropriate topographic matching of connections between structures has been well studied in the visual system. For example, Eph receptors in the retina and their ephrin ligands in the superior colliculus form gradients that provide a molecular substrate for the development of topographically matched connections between these structures, which ultimately underlies functional map organization in the adult superior colliculus (Frisén et al., 1998; for reviews, see Flanagan and Vanderhaeghen, 1998; Goodhill and Richards, 1999; O'Leary and Wilkinson, 1999). Such gradient matching has only recently been established for connections between sensory receptor arrays and their thalamic counterparts. For instance, ephrin-A2 and -A5 have been demonstrated to play an important role in retinogeniculate topographic connection formation (Feldheim et al., 1998, 2000). Further, in mutant mice that lack both ephrin-A2 and ephrin-A5, surgically induced invasion of the medial geniculate nucleus by retinal ganglion cell axons is enhanced, suggesting that these proteins are involved in regulating sensoritopic mapping within the dorsal thalamus (Lyckman et al., 2001).
In addition to the gradient-matching mechanisms that are involved in setting up topographic connections within a unimodal structure, there is an abundance of data that indicate that spontaneous and patterned activity are important for map formation in the superior colliculus and laminar organization within the LGd (Shatz, 1983; Sretevan and Shatz, 1986; Grubb et al., 2003; McLaughlin et al., 2003; Schmidt, 2003; for review, see Ruthazer and Cline, 2004). Activity from developing peripheral sensory receptor arrays may regulate the expression of molecules through calcium and CRE/CREB-regulated transcriptional pathways. CRE/CREB-mediated gene expression has been demonstrated in the developing thalamus, prior to visual experience (Pham et al., 2001). Mice deficient in the CREB protein develop more diffuse, less refined ipsilateral and contralateral retinal connections with the LGd. In addition, monocular enucleation causes a downregulation of CRE lacZ expression, indicating that activity from the peripheral receptor array normally regulates gene expression through the CRE/CREB pathway. This CRE/CREB pathway may regulate the formation of sensoritopic connections in the thalamus and midbrain structures by altering the expression patterns of molecules that have been associated with sensoritopic matching in the thalamus and midbrain, such as the ephrins, or transcription factors that have been shown to be restricted to particular thalamic nuclei, such as Gbx2, Pax6, and Ngn2 (Nakagawa and O'Leary, 2001). Support for an ephrin-mediated regulation of unimodal connections in the mibrain is provided by Frisén et al. (1998), who demonstrate that the absence of ephrin-A5 in mice results not only in disruption of topographic connections in the superior colliculus, but an invasion of retinal ganglion cell axons into the inferior colliculus. It is also possible that the loss of activity in the projection target, the MG, induces molecular changes in the MG and invokes sprouting of axons from adjacent pathways, such as the optic tract.
Recent evidence indicates that there are distinct patterns of expression of transcription factors (e.g., Lhx1, Lhx2, Gbx2, Pax6) in the developing thalamus associated with particular nuclei, which in combination with other molecules may be responsible for specifying nucleus-specific properties and ultimately establishing appropriate afferent and efferent connections of the thalamus (Nakagawa and O'Leary, 2001). The fact that despite congenital deafness, the MGN appears to be histologically normal and the majority of retinal ganglion cells establish appropriate connections lends support to the proposition that combinations of expression of particular genes intrinsic to the dorsal thalamus set up a basic nonvariable organization at least in terms of nuclei position and histological appearance. However, the present data suggest that activity from different sensory receptor arrays are necessary for appropriate sensoritopic mapping of connections of particular nuclei within the thalamus and layers within the colliculus. Thus, the ratio of activity from the different sensory systems may ultimately endow unimodality or multimodality of a particular subcortical structure. Further, preliminary studies in our laboratory (Hunt et al., 2002; Punj et al., 2002) indicate that alterations in connectivity have occurred throughout the nervous system in congenitally deaf mice. Such alterations in central nervous system connections have also been described in anophthalmic mice in which the inferior colliculus and the cuneate nucleus project to the LGN (Asanuma and Stanfield, 1990). Each particular alteration in turn changes the combination of inputs to a particular structure and the ratio of sensory-specific activity to that structure. These alterations, which begin at the very first stage of sensory processing, can have a cascade of bottom-up effects along the entire neuroaxis. Thus, the large functional changes observed in the neocortex of congenitally deaf humans, blind mole rats, and congenitally deaf mice could be solely accounted for by small alterations in connectivity at a number of subcortical structures resulting from changes in the ratio of activity from different sensory systems. It follows that phenotypic variability in cortical organization across mammals need not be mediated by genes intrinsic to the neocortex.
The authors thank Sarah Karlen and Jeffery Padberg for helpful suggestions on this manuscript. Supported by the McDonnell Foundation Grant (220020053 to L.K.), the National Institute of Neurological Disorders and Stroke (NINDS) (R01-NS35103 to L.K.), and the National Institute of Deafness and Other Communications Disorders (NIDCD) (R01 DC04542 to E.N.Y.).