Mammalian SAD Kinases Are Required for Neuronal Polarization

See allHide authors and affiliations

Science  11 Feb 2005:
Vol. 307, Issue 5711, pp. 929-932
DOI: 10.1126/science.1107403


Electrical activity in neurons is generally initiated in dendritic processes then propagated along axons to synapses, where it is passed to other neurons. Major structural features of neurons—their dendrites and axons—are thus related to their fundamental functions: the receipt and transmission of information. The acquisition of these distinct properties by dendrites and axons, called polarization, is a critical step in neuronal differentiation. We show here that SAD-A and SAD-B, mammalian orthologs of a kinase needed for presynaptic differentiation in Caenorhabditis elegans, are required for neuronal polarization. These kinases will provide entry points for unraveling signaling mechanisms that polarize neurons.

Most vertebrate neurons have two types of cytoplasmic processes: shorter, tapering dendrites, which receive synaptic input, and longer, slender axons, which form synapses on target cells. The molecular bases of these structural and physiological distinctions have been studied in detail, but the mechanisms that establish polarization are poorly understood (14). The SAD-1 kinase SAD–1 has been identified in a screen for Caenorhabditis elegans mutants in which synapses failed to form properly, but some aspects of the mutant phenotype raised the possibility that it might also regulate neuronal polarization (5). Consistent with this idea, SAD-1 contains a kinase domain related to that of PAR–1, a determinant of embryonic polarity (6, 7).

As part of an effort to identify regulators of synaptogenesis in mammals (8), we searched public databases for orthologs of SAD-1 and found two each in humans and mice and one in Drosophila; a previously described Ascidian ortholog (9) is more distantly related (fig. S1A). We isolated the two murine genes (10), which we call sad-a and sad-b. Predicted SAD-A and -B proteins are similar to each other and to C. elegans SAD-1 along most of their length (Fig. 1A).

Fig. 1.

Structure and expression of the mouse SAD kinases. (A) Homology of mouse (m) SAD-A and SAD-B (GenBank accession nos. AY533671 and AY533672) with each other and with C. elegans (Ce) SAD-1 includes the kinase domain, a region that contains a ubiquitin-associated (UBA) domain, and a C-terminal block of unique sequence. Numbers show percent amino acid identity. Scale bar, 100 amino acids. (B) Northern analysis of adult heart (H), brain (B), spleen (S), lung (Lu), liver (Li), skeletal muscle (M), kidney (K), and testis (T). SAD-A and -B are selectively expressed in the brain; low levels of SAD-B RNA are also detectable in the testis. (C and D) Expression of SAD-B by in situ hybridization of (C) E12 mouse embryo and (D) postnatal day 0 (P0) head. Both genes are broadly expressed in the brain and spinal cord. Scale bar, 500 μm. (E and F) Expression in the cortex is low in the proliferative ventricular zone (VZ) but high in (E) the preplate (PP) at E12 and (F) the cortical plate (CP) at E16, where newly generated neurons reside. Scale bar, 100 μm. (G) Section of adult spinal cord doubly stained with antibody to SAD-B and antibody to SV2, a synaptic vesicle protein. SAD-B is concentrated at synaptic sites in this region but is also present in nonsynaptic areas. Scale bar, 25 μm. (H) Sections of hippocampus from wild-type (WT), sada–/–, and sadb–/– adults stained with antibodies specific for SAD-A or -B. Both proteins are broadly distributed in the neuropil. Specificity is shown by lack of staining in the cognate mutant. Scale bar, 500 μm. (I) Cultured hippocampal neurons stained with antibodies to SAD-B and the dendritic marker MAP2, showing SAD is present in both axon (arrowhead) and dendrites. Scale bar, 25 μm.

Northern analysis showed that expression of both sad-a and -b is largely restricted to the nervous system (Fig. 1B), and in situ hybridization revealed that both mRNAs are broadly distributed within the brain and spinal cord of embryonic and postnatal animals (Fig. 1, C and D, and fig. S1). Sad expression was low in the cerebral ventricular zone, in which neural progenitors divide, but high in newly generated neurons in the preplate [at embryonic day (E) 12] and in the cortical plate (at E16) (Fig. 1, E and F, and fig. S1). Thus, sad genes are activated early in the program of neuronal differentiation.

To determine the localization of the SAD proteins, we generated SAD-A–specific and SAD-B–specific antibodies (10). Specificity of the antibodies was demonstrated by reactivity with the corresponding recombinant protein and by the absence of immunoreactivity in tissue from SAD-null mutants described below. Both proteins were present in the gray matter of the brain and spinal cord and were present at, but not confined to, synaptic sites (Fig. 1, G and H). Staining of cultured hippocampal and cortical neurons (10) showed that SAD proteins were present in both axonal and dendritic processes (Fig. 1I and fig. S2).

To assess the function of the SAD kinases, we generated null mutants of both genes (Fig. 2A). Sada–/– and sadb–/– mice were healthy and fertile, even though lack of one kinase had no apparent compensatory effect on the expression of the other (Fig. 2B). In contrast, double-mutant pups (sada–/–; sadb–/–, called AB–/–) showed little spontaneous movement, were only weakly responsive to tactile stimulation, and died within 2 hours of birth. Their immobility and poor responsiveness suggested a neural phenotype.

Fig. 2.

Aberrant neuronal differentiation in AB–/– mice. (A) Vectors used to generate null mutants of the sada (top) and sadb (bottom) genes. (B) Northern analysis of mSAD-A and -B RNA in wild-type and mutant brain. The cognate RNA was absent from each mutant, but levels of the other SAD RNA were not detectably affected. (C and D) Sections stained with (C) cresyl violet or (D) fluoro-Nissl at E19 show that the cerebral ventricle is enlarged in mutants and the cortex is thinned but well laminated. MZ, marginal zone; CP, cortical plate; SP, subplate; IZ, intermediate zone; VZ, ventricular zone. (E and F) Analysis of neuronal morphology by diolistics (14). (E) Images of individual neurons were optically isolated and contrast-enhanced. Additional examples are in fig. S4. (F) Quantitation of the distribution of processes confirmed the visual impression that neurons were less radially oriented in mutants than in controls. Control and mutant neurons differ at P < 0.001 by chi-square test. (G) Tracing with DiI revealed a near-complete absence of descending projections from the cortex (C) in the mutant. DiI crystals were placed in the rostral cortex to label descending tracts. Projections to the thalamus (T) were visible in controls but not in mutants. Yellow stars show the location of crystal, red arrows show the labeled projection, and blue arrows show the location of the coronal section in micrographs. (H) Section of E19 control and AB–/– cortex stained with antibody to TAG1. Scale bars, (C) and (G), 500 μm; (D) and (H), 100 μm; (E) 50 μm.

Examination of the AB–/– nervous system at E19 showed that the principal divisions of the brain, spinal cord, and peripheral nervous system had formed but that the forebrain was noticeably smaller in mutants than in littermate controls (fig. S3A). Although AB–/– cortex was abnormally thin (Fig. 2, C and D), it contained the main cortical cell types: neurons, astrocytes, and radial glia (fig. S3). The cells were organized into laminae, identifiable by cell density and specific markers (11, 12). However, the subplate was poorly defined, and segregation of neuronal subtypes to sublayers within the cortical plate was disordered (Fig. 2D and fig. S3D) (13). It is therefore likely that SAD kinases are required for some early events in corticogenesis, but we focus here on the later aspect of neuronal differentiation.

To ask whether loss of SAD kinases affected neuronal morphology, we impregnated individual neurons with fluorescent dyes, using a ballistic transfer method (14). In controls, most labeled neurons were polarized, with a long, thick apical dendrite that extended toward the pia; multiple smaller dendrites arising from the soma; and a single, thin basal axon. In contrast, mutant neurons often had a starburst morphology or processes that ran diagonally or tangentially rather than radially, and axons were difficult to distinguish from dendrites (Fig. 2, E and F, and fig. S4).

To ask whether AB–/– cortical neurons extended long axons, we implanted crystals of 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) into the cortex (15). In controls, DiI-labeled axonal tracts descended from the cortex to the thalamus. In contrast, few if any labeled axons descended from AB–/– cortex (Fig. 2G). We also labeled sections with antibodies to transient axonal glycoprotein (TAG) 1, a selective marker of corticothalamic axons (11). TAG-1+ corticothalamic axons were evident in controls but absent from mutants (Fig. 2H). Although apoptosis was significantly increased in AB–/– cortex, the absence of the tract was not solely a consequence of neuronal loss, because in situ hybridization showed that TAG-1–expressing neurons were numerous in AB–/– mutants (fig. S3). We therefore favor the explanation that apoptosis is a consequence of axons failing to reach target areas and gain adequate trophic support (16).

To determine whether these morphological defects reflected defective neuronal polarity or were secondary consequences of other abnormalities, such as misplaced guidance cues, we grew neurons from AB–/– cortex in low-density cultures, then stained them with antibodies to microtubule-associated proteins (MAPs) that are selectively localized to either axons (dephospho-tau, stained with the Tau-1 antibody) or dendrites (MAP2) (17). We also analyzed cultures from the hippocampus (17, 18), because polarity has been extensively studied in this system (14).

Neurons from mutant hippocampus (Fig. 3) and cortex (fig. S5) failed to form distinct axons and dendrites. Most control neurons had a single long, slender, branched axon that was rich in Tau-1 and poor in MAP2 and multiple shorter, thicker dendrites that were rich in MAP2 and poor in Tau-1 (Fig. 3A). In contrast, neurites of AB–/– neurons were relatively uniform in length and branching complexity, with values intermediate between those of axons and dendrites in control cultures (Fig. 3, D and E, and fig. S5). Moreover, most AB–/– neurites contained both Tau-1 and MAP2 (Fig. 3A), and levels of these markers did not differ greatly among neurites of a given neuron (Fig. 3, F and G, and fig. S5). Polarity defects did not result from growth defects, in that there was little difference between genotypes in the total number of primary neurites per neuron or the total length of neurite per neuron (Fig. 3, B and C, and fig. S5). Moreover, lack of polarization does not reflect a developmental delay, because control neurons were polarized by 3 days in vitro, whereas AB–/– neurons showed little polarization even after 8 days. These results indicate a cell-autonomous requirement of SAD kinases for neuronal polarization and suggest that the aberrant shape of AB–/– neurons in vivo results at least in part from this defect.

Fig. 3.

Neuronal polarization defects in neurons cultured from AB–/– hippocampus. (A) Single neurons from control and AB–/– hippocampus cultured for 7 days, then stained for the axonal marker Tau-1 (red) and the dendritic marker MAP2 (green). Control neurons have a single, tau-trast, rich axon and multiple, thicker, shorter, MAP2-rich dendrites. AB–/– neurons have multiple processes of similar length and caliber that contain both Tau-1 and MAP2. Scale bar, 50 μm. (B and C) Control and AB–/– neurons have equivalent numbers of neurites arising from the cell body and equivalent total neurite length. (D) Neurites of AB–/– neurons are similar in length to each other and midway in length between the axons and dendrites of control neurons. The x axis indicates rank order of neurites by length, from long to short. (E) Branching density (branches per unit length) is lower for axons (1st) than for the longest dendrite (2nd) in controls but is similar in AB–/– neurons. (F and G) Average intensity of Tau-1 or MAP2 staining in the longest neurite (the axon in control neurons) divided by that in all other neurites. Control axons are rich in Tau-1 and poor in MAP2; both antigens are more evenly distributed (the dashed line shows equivalence) in AB–/– neurons. Significance of differences between control and mutant neurons, (B) and (C) P > 0.3 by t test; (D) and (E) P < 0.001 by analysis of variance; and (F) and (G) P < 0.001 by permutation test. n = 20 neurons from two cultures.

How do SAD kinases promote neuronal polarization? One possible mechanism is suggested by studies of PAR-1 and microtubule affinity-regulating kinases (MARKs), which have been implicated in the control of cellular polarity in worms, flies, and mammals (6, 7, 19). As noted above, SAD genes are related to C. elegans PAR-1 and its vertebrate orthologs, MARK1 to MARK4, in their kinase domains (50 to 52% amino acid identity to MARK1 to MARK4, compared to 93% identity between SAD-A and -B) but not in other domains. PAR-1 and MARK phosphorylate MAPs such as MAP2 and tau, thereby regulating their interactions with microtubules (19); local variations in microtubule organization are critical for neuronal polarization (1, 4, 20). We therefore asked whether SAD kinases also regulate phosphorylation of MAPs. We chose tau as an exemplar, because a specific site in this protein [serine 262 (S262)] is a substrate for PAR-1/MARK (21, 22) and regulates binding of tau to microtubules (23). Overexpression of SAD-A in cultured neural (PC12) and nonneural (Chinese hamster ovary) cells increased phosphorylation of tau at S262 without affecting overall tau levels (Fig. 4A). A point mutation in the catalytic site (SAD-AK49A), predicted to render SAD kinase inactive, abolished this effect (Fig. 4A). Thus, SAD kinase promotes phosphorylation of tau at S262 [tau-p(S262)] in vitro.

Fig. 4.

SAD kinase regulates tau phosphorylation. (A) PC12 cells were transfected with a tau-GFP fusion alone (mock) or tau-GFP along with SAD-A or a kinase-defective SAD-A mutant (K49A). Three days later, cells were stained with antibody to SAD-A and antibody to tau-p(S262). SAD-A promotes tau phosphorylation at S262. Scale bar, 100 μm. (B) A cultured hippocampal neuron stained with antibody to tau-p(S262) (green) and antibody to dephospho-tau (Tau-1, red), showing concentration of the phosphoepitope in dendrites. Scale bar, 20 μm. (C) A single, DiI-labeled pyramidal neuron from a control to indicate the arrangement of neurons; the apical dendrite and axon are bracketed in green and red, respectively. (D) Sections of control and AB–/– cortex were labeled with antibodies to dephospho-tau (Tau-1), MAP2, tau-p(S262), or all tau isoforms (Tau5). Intensity is rendered in the color scale at the bottom right. In controls, MAP2 and tau-p(S262) were concentrated in the dendrite-rich upper cortical plate and Tau-1 in the axon-rich intermediate zone. In mutants, levels of dephospho-tau increased and tau-p(S262) decreased in the cortical plate; total tau levels did not change. Scale bar, 50 μm. (E) Quantitation of tau levels in the cortical plate. ***, P < 0.001; #, P > 0.8 by t test.

Next, we used immunostaining to examine the distribution of phosphorylated and dephosphorylated tau in cortex. Whereas dephosphotau (recognized by Tau-1) is concentrated in axons of cultured neurons (20) (Fig. 3), tau-p(S262), like MAP2, is concentrated in dendrites (Fig. 4B). In the developing cortex, the upper cortical plate is rich in dendrites, whereas the intermediate zone is rich in axons but poor in dendrites (Fig. 4C). Consistent with this arrangement, tau-p(S262) and MAP-2 are concentrated in the upper cortical plate, whereas dephospho-tau is more abundant in the intermediate zone (Fig. 4D). We then asked whether tau phosphorylation is altered in AB–/– cortex, and we found that it is: Levels of tau-p(S262) are decreased and levels of dephospho-tau increased in the cortical plate, with no change in overall tau levels (Fig. 4, D and E). Together, these results suggest that SAD kinases act, at least in part, by locally regulating phosphorylation of MAPs, including tau, and that the consequent alterations in microtubule organization are critical for neuronal polarization (20). Because SAD is present in both axons and dendrites (Fig. 1I and fig. S2), its distinct effects on these two types of processes must reflect local regulation. Kinases such as MARK kinase/TAO1 and LKB1/PAR-4, which can activate MARK and SAD (24, 25), are candidate local regulators. Other components of evolutionarily conserved polarization machinery, PAR-3 and PAR-6, which interact with PAR-1/MARK, may also be involved (7, 26).

In summary, we have shown that the SAD kinases are required for forebrain neurons to acquire the polarity that endows axons and dendrites with distinct properties. SAD kinases seem to share some activities with PAR-1 and MARK, but their unique domains may enable them to be locally regulated in neuron-specific ways. It remains to be determined whether mammalian SAD kinases, like C. elegans SAD-1 (5), are involved not only in polarity but also in synaptogenesis.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article