|Home | About | Journals | Submit | Contact Us | Français|
Class I major histocompatibility complex (class I MHC) molecules, known to be important for immune responses to antigen, are expressed also by neurons that undergo activity-dependent, long-term structural and synaptic modifications. Here, we show that in mice genetically deficient for cell surface class I MHC or for a class I MHC receptor component, CD3ζ, refinement of connections between retina and central targets during development is incomplete. In the hippocampus of adult mutants, N-methyl-D-aspartate receptor–dependent long-term potentiation (LTP) is enhanced, and long-term depression (LTD) is absent. Specific class I MHC messenger RNAs are expressed by distinct mosaics of neurons, reflecting a potential for diverse neuronal functions. These results demonstrate an important role for these molecules in the activity-dependent remodeling and plasticity of connections in the developing and mature mammalian central nervous system (CNS).
The development of precise connections in the CNS is critically dependent on neural activity, which drives the elimination of inappropriate connections and the stabilization of appropriate ones. In the visual system of higher mammals, the refinement of initially imprecise axonal connections requires spontaneously generated activity early in development and visually driven activity later (1–4). Fine-tuning of neural connectivity is thought to result from changes in synaptic strength, driven by patterned impulse activity (1, 2, 5, 6).
To identify molecules critical for activity-dependent structural remodeling, we previously conducted an unbiased screen for mRNAs selectively regulated by blocking spontaneously generated activity in the developing cat visual system. This manipulation prevents the remodeling of retinal axons from each eye into layers within the lateral geniculate nucleus (LGN) (7–9). Although many known neural genes were not detectably regulated by activity blockade, this screen revealed to our surprise that members of the class I MHC protein family are expressed by neurons and are regulated by spontaneous and evoked neural activity (10). Neuronal class I MHC expression corresponds to well-characterized times and regions of activity-dependent development and plasticity of CNS connections, including retina, LGN, and hippocampus. Furthermore, the mRNA for CD3ζ [a class I MHC receptor subunit in the immune system (11)] is also expressed by neurons (10), consistent with its interaction with class I MHC during activity-dependent remodeling and plasticity. Although class I MHC is primarily known for its function in cell-mediated immune recognition, the above findings from our differential screen suggest that class I MHC molecules may play roles in structural and synaptic remodeling in the developing and mature CNS.
To explore these possibilities by genetic means, we first confirmed by in situ hybridization that class I MHC and CD3ζ were expressed in the developing mouse CNS. Because numerous class I MHC genes exist in the mouse genome (12), we used a pan-specific cDNA probe expected to detect many class I MHC molecules (13). This probe detected elevated amounts of mRNAs in the dorsal LGN (dLGN) during the first two postnatal weeks, exactly when ganglion cell axons sort into eye-specific layers in the mouse (14); mRNA levels declined at later ages (Fig. 1A, compare postnatal days P6 and P40). Expression was also evident in the ganglion cell layer of the retina (Fig. 1A, P6 eye), in neocortex (in layer 4 at early ages and in deeper layers later; Fig. 1A), and in granule and pyramidal cell layers of the hippocampus (Fig. 1A, P40, and Fig. 2). CD3ζ mRNA, like that of class I MHC, was expressed in the mouse dLGN during the first two postnatal weeks (Fig. 1B); expression appeared higher medially. CD3ζ mRNA was also detected in small amounts in P40 hippocampus (15). Therefore, as in cat (10), class I MHC and CD3ζ transcripts are present in the developing murine CNS at locations and times consistent with a role for these molecules in activity-dependent structural remodeling and synaptic plasticity.
Strikingly, different class I MHC genes are expressed in unique subsets of neurons throughout the mature CNS, as revealed by using probes (13) that react more specifically with each of two class Ia (H–2D, H–2K) or two class Ib MHC genes (Qa-1, T22). For example, within the somatosensory cortex, H–2D probe signal was distributed through many layers but was strongest in layer 4; Qa-1 signal was specific to layer 6, and T22 signal was evident in both layers 5 and 6 (Fig. 2). H–2D and T22 signals were both strong in the pyramidal layers of the hippocampus and in the habenula; in contrast, that of Qa-1 was weak in those locations (Fig. 2). Transcripts detected by the T22 probe were particularly abundant in the thalamic reticular nucleus, globus pallidus, and substantia nigra [Fig. 2 and (15)]. H–2K signal paralleled that of H–2D but was much lower throughout the brain (16). The distinct expression patterns detected by these probes extended prior inferences from RNase protection experiments in cat (10) and demonstrated conclusively that several class I MHC mRNA subtypes are differentially expressed by distinct subsets of neurons in the CNS. These findings suggest a potential for functional diversity among class Ia and Ib genes within the CNS. Such heterogeneity of function occurs among these genes within the immune system (17).
To test directly our hypothesis that class I MHC is required for activity-driven structural remodeling and synaptic plasticity, mice deficient either for cell surface class I MHC expression or for CD3ζ were analyzed. Because numerous class I MHC genes may be expressed by specific subsets of neurons (Fig. 2), we examined mice lacking two molecules required for the stable cell-surface expression of nearly all fully assembled class I MHC molecules: β2-microglobulin [β2M, a class I MHC cosubunit (18)], and TAP1 [a component of the transporter that supplies peptides to class I MHC enroute to the cell surface (19, 20)]. β2-M is expressed by neurons in LGN, cortex, and hippocampus (10) and, as in non-neuronal cells, induction of class I MHC on the cell surface of neurons requires expression of β2M and TAP1 mRNAs (21). In addition, to examine whether CD3ζ-containing receptors were involved in class I MHC–mediated signaling in the CNS, we analyzed mice lacking CD3ζ (22). When raised in a germ-free facility, all mutant mice are outwardly normal and are not obviously different from wild-type mice in weight, body length, appearance, or behavior.
We hypothesized that mice deficient in class I MHC–mediated signaling might have abnormal patterns of retinogeniculate projections because blockade of neural activity simultaneously prevents the segregation of retinal ganglion cell axons into eye-specific layers and reduces class I MHC expression in the LGN (7–10). The normal adult mouse dLGN has a small layer that receives inputs from ganglion cells in the ipsilateral eye; inputs from the contralateral eye occupy the remainder of the dLGN (Fig. 3A). The refinement of these eye-specific connections in the mouse occurs between postnatal day 4 (P4) and P8 (14). We therefore examined the distribution of retinal inputs at P13, 5 days after segregation was complete, using the anterograde transport of horseradish peroxidase–conjugated wheat germ agglutinin (WGA-HRP) injected into one eye (23). Compared with wild-type animals (Fig. 3, A and F, β2M+/+), the pattern of the retinogeniculate projection was significantly altered in all three mutant genotypes tested. This point is best appreciated by inspecting the size and shape of the ipsilateral retinal projection to the dLGN (Figs. 3, A to E). Although all mutants still form an ipsilateral patch located approximately normally in the mediodorsal dLGN, the area of this patch was significantly larger in mutant mice and, in extreme cases, was accompanied by multiple ectopic clusters of inputs that were never observed in wild-type mice (Fig. 3, C and E, arrowheads). These ectopic clusters appeared in medial areas of the dLGN, where the highest levels of CD3ζ mRNA are normally present (compare Fig. 3, C and E, with Fig. 1A). In these extreme cases, ectopic clusters were also observed in the ipsilateral superior colliculus, another retinorecipient target that expresses low-to-moderate levels of class I MHC mRNA in mouse (15).
To assess quantitatively the altered retinogeniculate projection in mutant mice, computerized image analysis was used to measure the fraction of dLGN area occupied by the ipsilateral projection. All image analyses were carried out by an observer blind to genotype (24). In all mutant genotypes, there was a significant increase in area occupied by the ipsilateral projection over that of wild-type controls [Fig. 3F: β2M−/−, 130.3 ± 7.3% (n = 10); β2M−/−TAP1−/−, 133.3 ± 5.7% (n = 13); CD3ζ−/−, 122.7 ± 4.2% (n = 13); wild-type β2M+/+, 100.0 ± 9.1% (n = 12); P < 0.05, Student’s two-tailed t-test]. These observations support the hypothesis that class I MHC function is required for the developmental refinement of the retinal projections and the formation of precise eye-specific regions in the LGN.
Although the refinement of retinogeniculate axons was abnormal in mutant mice, many other aspects of LGN development appear to proceed normally. The histological appearance, size, shape, and location of the dLGN and thalamus, as viewed in Nissl-stained sections, were indistinguishable among all experimental groups (15). The bulk of the ipsilateral projection was positioned, as expected, in the binocular region of the dLGN. At the ultrastructural level, the synaptic organization of the LGN in β2M−/−TAP1−/− mice appeared qualitatively indistinguishable from that of wild type (23). Retinogeniculate axons were well-myelinated, and glomeruli and R-type synaptic boutons [hallmarks of retinogeniculate synapses; (25–27)] were present, indicating that normal retinal synapses do form in the LGN (Fig. 3G). These observations suggest that many activity-independent processes (1, 2, 28, 29) are not perturbed in mice with abnormal class I MHC function.
Because similar abnormalities in the ipsilateral projection result from blockade of spontaneous activity at comparable ages in the cat or ferret visual system (7–9), we tested whether the mutant retinogeniculate phenotypes were secondary to abnormal retinal activity. Calcium imaging of mutant retinas revealed spontaneous retinal waves with spatiotemporal properties indistinguishable from those of normal mice (30). Thus, we ascribe abnormalities in the mutant retinogeniculate projection directly to a loss of class I MHC signaling downstream of neural activity.
Because activity-dependent structural reorganizations during development are thought to arise from cellular mechanisms of synaptic plasticity (1, 2, 6), we next asked whether synaptic plasticity is altered in mutant mice. Because little is known about such mechanisms in the developing LGN, we used a well-characterized model system for studying long-lasting changes in the strength of synaptic transmission: the Schaffer collateral-CA1 synapse of the hippocampus (31, 32). Class I MHC and CD3ζ were both expressed in the adult hippocampus (Fig. 1) (10, 15). Furthermore, class I MHC immunoreactivity can be detected in synaptosome preparations, suggesting that some class I molecules are synaptically associated (33). We therefore assessed hippocampal synaptic plasticity in wild-type and mutant mice. Data collection was performed by an observer blind to genotype (34).
In wild-type mice (C57BL/6), tetanic stimulation (4 × 100 Hz) resulted in a sustained increase in the slope of the field excitatory postsynaptic potential (fEPSP) (167 ± 13% of pretetanus baseline; n = 15; Fig. 4, A and C). In contrast, in CD3ζ−/− mutant animals, LTP in response to the same tetanus was significantly enhanced relative to that in wild-type mice (248 ± 29% of baseline; n = 8; P < 0.05; Fig. 4, A and C). A similar enhancement of LTP was observed in β2M−/−TAP1−/− mutant mice (227 ± 22% of baseline; n = 10; P < 0.05; Fig. 4C). Basal synaptic transmission is not significantly different among all experimental groups (35). Enhanced LTP in gene knockout animals was not due to changes in inhibition, because GABAA-mediated transmission was blocked with 100 μM picrotoxin in all experiments. Nor was the enhanced LTP due to induction of an N-methyl-D-aspartate (NMDA) receptor-independent form of LTP, because LTP was completely abolished in all genotypes in the presence of the NMDA antagonist 2-amino-5-phosphonovalerate [50 μM D-APV; Fig. 4B and (36)].
It is conceivable that enhancement of LTP seen in these genotypes is due to some non-specific effect of immune compromise on the CNS. Thus we also examined LTP in a more severely immunodeficient strain of mice that lacks recombination activating gene-1 (RAG1). RAG1 is required for production of B and T cells and is also transcribed by neurons in the CNS (37, 38). LTP in RAG1−/−mice was indistinguishable from that of wild type [153 ± 13% of baseline (n = 10), compared with 167 ± 13% in wild type; P = 0.48; Fig. 4C], indicating that the LTP abnormalities seen in β2M−/−TAP1−/− or CD3ζ−/− mice are specific to their genotypes rather than to immune status.
Synaptic plasticity in the hippocampus is dependent on stimulation frequency, with high frequencies producing LTP and low frequencies producing LTD (31, 39–41). We therefore examined the effect of other stimulation frequencies on synaptic plasticity in animals deficient for class I MHC signaling. In adult wild-type slices, the delivery of 900 pulses at 0.5 Hz induced significant LTD (82 ± 6% of baseline; n = 8; P < 0.05; Fig. 4D). In adult slices from both mutant genotypes, however, there was no significant change in fEPSP slope upon 0.5 Hz stimulation [CD3ζ−/−, 107 ± 7% of baseline (n = 5, P = 0.29); β2M−/−TAP1−/−, 99 ± 5% of baseline (n = 8, P = 0.78); Fig. 4D]. Furthermore, after 900 pulses at 1 Hz, transmission was significantly enhanced over baseline in both CD3ζ−/− (141 ± 14% of baseline, n = 5, P < 0.05) and β2M−/−TAP1−/− slices (128 ± 9%, n = 6, P < 0.05) but was unchanged in wild-type slices (94 ± 5%, n = 14, P = 0.41; Fig. 4D). Thus, in mutant mice, LTD could not be detected, and the frequency-response curve of hippocampal synaptic plasticity was consistently shifted across a broad range of stimulation frequencies.
These results indicate that class I MHC/CD3ζ signaling is important for mediating activity-dependent synaptic depression, because, in mutants, there is a shift in the bidirectional regulation of synaptic strength [i.e., the frequency response function (39–41)] that favors potentiation. In the absence of class I MHC or CD3ζ, patterns of neural activity that normally have no effect on synaptic strength or that lead to synaptic depression result, instead, in abnormal synaptic strengthening. Likewise, in the dLGN, enhanced LTP and lack of LTD at the developing retinogeniculate synapse could account for the structural phenotype observed: a persistence of inappropriate connections that would be normally be removed via an activity-dependent process of synaptic weakening during eye-specific segregation (14, 42–44).
Class I MHC and CD3ζ are expressed in the CNS by specific sets of neurons that undergo activity-dependent changes (10). Here, we show that mice lacking these molecules exhibit abnormalities in connections between these neurons, suggesting a direct neuronal function for class I signaling. In addition, both mutants have strikingly similar phenotypes, implying that class I MHC signaling in the brain is transduced via a CD3ζ-containing receptor, either an unknown CNS-specific or a known immune receptor. The expression patterns of class I MHC and CD3ζ in the CNS are consistent with signaling via a number of possible receptor-ligand configurations. For example, both class I MHC and CD3ζ are expressed by neurons in the hippocampus; in addition, class I MHC mRNA is also expressed by retinal ganglion cells when CD3ζ is detected in the dLGN [Fig. 1A and (10)]. Detailed information concerning the ultrastructural localization of these molecules will be needed to resolve this issue.
Whatever the case, the evidence to date supports a model in which class I MHC functions in the CNS by engaging CD3ζ-containing receptors to signal activity-dependent changes in synaptic strength that ultimately lead to the establishment of appropriate synapses. Class I MHC may act directly at the synapse to promote the elimination of inappropriate connections, by using signaling mechanisms already characterized in immune cells (11), possibly via phosphorylation of CD3ζ by fyn [a kinase previously implicated in hippocampal plasticity (45)]. Because different class I MHC members are expressed by different subsets of CNS neurons, additional signaling specificity may be furnished by the particular repertoire of MHC molecules present in any given neuron. In the immune system, recognition of class I MHC by T cell receptors can result in functional elimination of inappropriate self-reactive T cell populations (46, 47). Our results demonstrate that class I MHC is also required for normal regressive events in the developing and adult CNS, including activity-dependent synaptic weakening and structural refinement.