The `signal regulatory protein' SIRPα is an Ig superfamily, transmembrane glycoprotein with a pair of cytoplasmic domains that can bind the phosphatase SHP-2 when phosphorylated on tyrosine. SIRPα is prominent in growth cones of rat cortical neurons and located, together with the tetraspanin CD81, in the growth cone periphery. SIRPα is dynamically associated with Triton-X-100-sensitive, but Brij-98-resistant, lipid microdomains, which also contain CD81. Challenge of growth cones with the integrin-binding extracellular-matrix (ECM) protein, laminin, or with the growth factors, IGF-1 or BDNF, increases SIRPα phosphorylation and SHP-2 binding rapidly and transiently, via Src family kinase activation; phosphorylated SIRPα dissociates from the lipid microdomains. A cytoplasmic tail fragment of SIRPα (cSIRPα), when expressed in primary cortical neurons, also is phosphorylated and binds SHP-2. Expression of wild-type cSIRPα, but not of a phosphorylation-deficient mutant, substantially decreases IGF-1-stimulated axonal growth on laminin. On poly-D-lysine and in control conditions, axonal growth is slower than on laminin, but there is no further reduction in growth rate induced by the expression of cSIRPα. Thus, the effect of cSIRPα on axon growth is dependent upon integrin activation by laminin. These results suggest that SIRPα functions in the modulation of axonal growth by ECM molecules, such as laminin.
Axon growth is regulated by numerous extracellular cues, both soluble and bound, and the growth cone is thought to integrate this information (Tessier-Lavigne and Goodman, 1996; Goldberg et al., 2002; Huber et al., 2003). For example, components of the extracellular matrix (ECM) potentiate axon outgrowth of CNS neurons if growth factor signals are present (Edgar et al., 1988; Goldberg et al., 2002; Liu et al., 2002).
Many receptors for soluble factors and ECM molecules have been identified in the growth cone. Other putative receptors have been observed, but their functions are unknown. We have described a highly heterogeneous glycoprotein, gp93, that is enriched in growth cone membranes (Quiroga and Pfenninger, 1994) and identified it as `signal regulatory protein' (SIRPα), also known as SHP (Src homology domain 2-containing phosphatase) substrate-1 (SHPS-1), BIT or P84 (Fujioka et al., 1996; Comu et al., 1997; Kharitonenkov et al., 1997; Sano et al., 1997; Wang et al., 2003). SIRPα is a heavily glycosylated, Ig superfamily transmembrane protein that contains three extracellular Ig domains and two intracellular ITIM (immunoreceptor tyrosine-based inhibitory motif) sequences. Phosphorylation of the ITIM tyrosines triggers binding and activation of the tyrosine phosphatases, SHP-1 and SHP-2 (Fujioka et al., 1996; Kharitonenkov et al., 1997). CD47, or integrin-associated protein (IAP) is believed to be a ligand of SIRPα (Jiang et al., 1999; Seiffert et al., 1999). SIRPα is expressed throughout the central nervous system (Comu et al., 1997; Mi et al., 2000), but in growth cones from different developing brain regions it exhibits differential glycosylation (Li et al., 1992; Wang et al., 2003; see also, van den Nieuwenhof et al., 2001). The functional significance of this neuron-type-specific glycosylation is unknown, but it might alter the binding of SIRPα to CD47 or other, unknown ligands (Ogura et al., 2004).
Depending on the types of ligand and cell involved, SIRPα binding to CD47, SHP-1 or SHP-2 positively or negatively regulates a variety of cellular functions, including mitogenesis, motility and adhesion (Cant and Ullrich, 2001; Oshima et al., 2002). There are data to suggest that, in neurons, SIRPα may be involved in the regulation of survival, neurite outgrowth and/or synapse formation and maintenance (Chuang and Lagenaur, 1990; Comu et al., 1997; Sano et al., 1997; Jiang et al., 1999; Araki et al., 2000; Mi et al., 2000).
As shown here, SIRPα is located, in part, in lipid microdomains (LMDs). Such LMDs (Simons and Ikonen, 1997; Brown and London, 1998; Simons and Toomre, 2000) seem to act as platforms for signal transduction initiated by several neurotrophic factors, and they are important for neuronal cell adhesion and axon guidance (Saarma, 2001; Tsui-Pierchala et al., 2002; Guirland et al., 2004). In order to determine the function of SIRPα in the growth cone, we examined (1) its spatial distribution; (2) its phosphorylation and SHP-2 binding in response to the ECM molecule laminin, and to growth factors; and (3) the effects of overexpressing SIRPα cytoplasmic fragment. The results indicate that SIRPα modulates axonal growth when growth cone integrins are activated by laminin.
Immunolocalization of SIRPα in cultured growth cones
The distribution of SIRPα in intact growth cones was studied by indirect immunofluorescence in cultures of primary rat cortical neurons (Fig. 1). Since the antibody did not stain intact neurons (data not shown), a mild detergent (1% Brij 98) was used to permeabilize membranes in order to preserve LMDs. Specific fluorescence was observed throughout the neuron (not shown). As is the case for other membrane components in transit to the growth cone, a high level of SIRPα as seen in the axon. In growth cones cultured on laminin (Fig. 1A-D) SIRPα staining was evident as variously sized spots that were particularly prominent in the periphery. The staining extended into the finest filopodia and their tips, and it followed the actin bundles identified by phalloidin labeling. This labeling pattern was independent of the substratum used for culture. On poly-D-lysine, which is not a physiological substrate and cannot trigger integrin signaling, cortical neurons had larger growth cones bearing more abundant but shorter filopodia (Fig. 1E,F). Despite the difference in growth cone morphology, SIRPα distribution in the growth cone was again punctate and primarily peripheral, in the filopodia.
We also compared the distribution of growth cone SIRPα to that of an LMD-resident tetraspanin, CD81 (Hemler, 2003), by double immunofluorescence. As seen in Fig. 1G-I, CD81 was distributed in discrete puncta, primarily in distal regions of the growth cone (grown on laminin). Many of these puncta overlapped with SIRPα immunoreactivity, particularly along the filopodia and at their tips. These data indicate substantial colocalization of SIRPα with CD81 in growth cones.
Distribution of SIRPα in growth cone LMDs
The punctate co-distribution of CD81 immunoreactivity with SIRPα in growth cone plasmalemma may indicate its compartmentalization in LMDs. Therefore, we investigated if growth cone SIRPα co-fractionated with LMDs.
Three criteria are typically used to define LMDs: (1) insolubility in non-ionic detergents; (2) flotation in a sucrose density gradient; and (3) detergent solubility in the presence of cholesterol-sequestering agents (Simons and Toomre, 2000). Cold TX100 is used most commonly for identification of `rafts', microdomains enriched in cholesterol, glycosphingolipid and GPI-anchored proteins. Using this method, cold TX100-resistant membranes were prepared from growth cone particles (GCPs) and isolated by flotation in low-density fractions recovered from a bottom-loaded, discontinuous sucrose gradient. Fractions were analyzed by immunoblotting as shown in Fig. 2A. Src, a known resident protein of lipid rafts, was mainly recovered in low-density fractions (0.6-0.7 M), as expected. However, we did not observe partitioning of SIRPα into these low-density fractions.
More recently, LMDs have been described that exhibit different detergent solubility characteristics (Chamberlain, 2004). For example, tetraspanin-containing LMDs can be maintained only in the presence of milder detergents, such as Brij 97 and Brij 98, but not TX100. Brij 97 is less harsh than TX100, and Brij 98 is even milder. In Brij 97, we found no partitioning of SIRPα or of the tetraspanin, CD81, into low-density fractions of sucrose gradients (Fig. 2B). However, a significant proportion of SIRPα, as well as CD81 and Src, were recovered from low-density fractions (0.3-0.7 M) in Brij 98 (Fig. 2C). In the same preparation, the transferrin receptor, a protein not associated with LMDs (Harder et al., 1998), was completely solubilized and recovered from the high-density fractions. This suggests that SIRPα significantly associates with Brij 98-resistant LMDs.
To test whether SIRPα is associated with LMDs in a cholesterol-dependent manner, GCPs were pre-treated with saponin or MβCD to disrupt LMDs by cholesterol perturbation (Rothberg et al., 1990; Cerneus et al., 1993; Ilangumaran and Hoessli, 1998; Ledesma et al., 1998; Kim and Pfeiffer, 1999). As seen in Fig. 2D,E, saponin or MβCD pre-treatment dramatically reduced the partitioning of SIRPα and Src into low-density fractions. These results indicate that SIRPα co-localizes with CD81 in TX100-sensitive but Brij 98-resistant LMDs.
Transient phosphorylation of growth cone SIRPα in response to laminin, IGF-1 and BDNF
Potent extracellular signals for growing neurons are the neurotrophic factors, BDNF and IGF-1, which operate through receptor tyrosine kinases, and ECM molecules, such as laminin, which activate tyrosine kinases via integrins. While such factors are known to stimulate SIRPα tyrosyl phosphorylation in other cells (Ohnishi et al., 1999; Maile and Clemmons, 2002a), this is not known for the growth cone. Therefore, we proceeded to examine SIRPα responses to such exogenous signals in GCPs, a preparation highly enriched in re-sealed, primarily axonal growth cones (Pfenninger et al., 1983; Lohse et al., 1996). Previous studies have shown that BDNF and IGF-1 stimulate axonal growth, and that GCPs contain functional receptors for them (Davies et al., 1986; Aizenman and De Vellis, 1987; Pfenninger et al., 2003). Therefore, we chose IGF-1 and BDNF to test the effects of growth factors on SIRPα in GCPs. As an ECM molecule, we selected laminin for these experiments because it is a preferred growth substrate for developing CNS neurons.
Because we used laminin in suspension rather than immobilized on a solid phase for studying the short-term responses of the GCPs, we first tested whether soluble laminin could indeed stimulate integrin signaling. Plating GCPs on laminin, i.e. laminin-induced integrin engagement, activates Src and increases Src binding to the cytoskeleton (Helmke et al., 1998). Therefore, we measured these parameters in suspended GCPs treated with soluble laminin. As shown in Fig. 3A with anti-Src-pY418 (which recognizes only activated Src), laminin treatment for 1 minute at 37°C increased activated Src in GCPs by about 60% compared to control incubation (an antibody to the p85 subunit of PI 3-kinase was used to show equal loading). We also probed the cytoskeleton fractions prepared from GCPs with anti-Src antibody to reveal associated total Src protein. Laminin treatment for 5 minutes at 37°C increased Src association with the cytoskeleton fraction by about 70% compared to control (Fig. 3B; in this experiment, each lane was loaded with the total cytoskeleton fraction prepared from equal GCP samples). Interestingly, SIRPα was not detectable in these cytoskeletal preparations, but SHP-2 was clearly present. The observed activation and increased cytoskeletal association of Src indicated that laminin applied to GCPs in suspension activated Src (see Helmke et al., 1998).
To determine whether laminin, BDNF or IGF-1 stimulation changed tyrosine phosphorylation of SIRPα, GCPs were incubated with the different factors for various times at 37°C, and SIRPα was immunoprecipitated. The immunoprecipitates were resolved by SDS-PAGE, blotted, and blots probed with anti-pTyr. In addition, blots were probed with anti-SHP-2 to assess whether SHP-2 co-immunoprecipitated with SIRPα (SHP-1 was not considered because it is a hematopoietic-cell phosphatase, whereas SHP-2 is ubiquitously expressed) (Neel et al., 2003). As shown in Fig. 4A, all three factors induced rapid phosphorylation of SIRPα. The increases over non-treated controls were 37±5% for laminin, 39±4% for IGF-1, and 21±4% for BDNF (means±s.e.m.; n≥4) after 1 minute at 37°C. These increases were modest but statistically significant (P≤0.01). It is known that GCPs exhibit high levels of protein kinase activity in control conditions (Helmke et al., 1998) so that, even without stimulation, SIRPα phosphorylation continued to increase over the 5-minute observation period (Fig. 4B). However, this increase was reduced for GCPs treated with laminin, IGF-1 or BDNF. As a result, we observed decreases relative to control levels after 5 minutes at 37°C. These decreases were 25±4% for laminin, 20±5% for IGF-1, and 23±4% for BDNF (means±s.e.m.; n≥4). Again, this was statistically significant (P≤0.01). Thus, the stimulation of SIRPα phosphorylation by the three factors was transient. Simultaneously with the increase in phosphorylation, the three factors enhanced co-immunoprecipitation of SHP-2 after 1 minute at 37°C (Fig. 4A). Likewise, we observed a decrease in SHP-2 relative to control after 5 minutes at 37°C (Fig. 4B). The changes paralleled those of SIRPα phosphorylation and also were statistically significant. These results indicated that laminin, BDNF and IGF-1 stimulated tyrosyl phosphorylation of SIRPα and its association with SHP-2 in isolated growth cones, and that these changes were rapid and transient. We also tested combinations of factors in such experiments. Laminin plus IGF-1 or laminin plus BDNF did not significantly increase tyrosyl phosphorylation or SHP-2 binding of SIRPα above the levels obtained with laminin, IGF-1 or BDNF alone (data not shown). Thus, the effects of laminin and growth factors did not seem to be additive or synergistic under the experimental conditions used.
The stimulation of SIRPα phosphorylation by laminin, BDNF and IGF-1 raised the question of which kinase(s) were responsible for this effect in growth cones. It has been reported that Src-like kinases may phosphorylate SIRPα in fibroblasts upon integrin-mediated adhesion to fibronectin (Oh et al., 1999). To determine whether Src-like kinases were indeed responsible for SIRPα phosphorylation in growth cones, we examined the effects of the selective Src kinase inhibitor, PP2, on laminin-, IGF-1- or BDNF-induced SIRPα phosphorylation and SHP-2 association. The control was GCPs without stimulation. Incubation of GCPs with 1 μM PP2 reduced laminin-, IGF-1-, or BDNF-stimulated SIRPα phosphorylation (Fig. 4C). As in the other experiments (Fig. 4A), SHP-2 co-immunoprecipitation paralleled phosphorylation, i.e. PP2 decreased SHP-2 association with SIRPα (Fig. 4D). These blots were probed with anti-SIRPα as a loading control to normalize SIRPα phosphorylation and co-immunoprecipitation of SHP-2. Stimulation with these factors normally increased SIRPα phosphorylation and SHP-2 binding to SIRPα by 20-40% relative to control at 1 minute (see Fig. 4A). However, PP2 greatly reduced SIRPα phosphorylation and SHP-2 binding from 100 (normalized arbitrary units, control without PP2) to about 18-21% (n=2) and 13-17% (n=3), respectively, for all stimulation conditions. This reduction was highly significant (P<0.01). In the absence of stimulation, SIRPα phosphorylation and SHP-2 binding also were reduced after PP2 treatment, to a level that was not statistically distinguishable from that of the stimulated, PP2-treated samples. Unlike PP2, PP3, a negative control for PP2, did not reduce SIRPα phosphorylation (Fig. 4C). The data indicated that the increase in SIRPα phosphorylation and SHP-2 association stimulated by laminin, IGF-1 or BDNF in growth cones depended on Src-like kinase(s). However, these results did not completely exclude the possible participation of another SIRPα kinase, since PP2 treatment did not fully abolish SIRPα phosphorylation and SHP-2 association.
Phosphorylation shifts SIRPa out of LMDs
The results described above suggested that at least a sizable fraction of SIRPα is located in LMDs, and that growth cone SIRPα may be involved in the processing of ECM and growth factor signals. Next, we investigated how the compartmentalization in LMDs related to the phosphorylation of SIRPα.
We used three different conditions to change the phosphorylation state of SIRPα, and assessed its distribution in lipid microdomains. The Src kinase inhibitor, PP2, served to block SIRPα phosphorylation. Laminin was used to stimulate SIRPα phosphorylation, and vanadate, a broad-spectrum tyrosine phosphatase inhibitor, caused hyper-phosphorylation of SIRPα. Following treatment with one of these reagents, GCPs were solubilized in Brij 98 and fractionated in density gradients containing five steps (0.5/0.7/0.9/1.0/1.33 M sucrose). The distribution of SIRPα and Src in the gradients was determined by western blot analysis. In the presence of PP2, virtually all SIRPα was recovered in the LMD fractions (this was normalized to 100%). Laminin and vanadate treatment, however, reduced the percentage of SIRPα in low-density fractions to only 27% and 6%, respectively, of the level found after PP2 treatment (Fig. 5). This indicated for SIRPα an inverse relationship between the phosphorylation state and co-fractionation with lipid microdomains. Src, by contrast, remained largely (62-70%) in the low-density fraction, even after PP2 treatment.
The cytoplasmic tail of SIRPα expressed in cultured neurons is phosphorylated and binds SHP-2
In order to study SIRPα function in the growth cone we wanted to perform misexpression experiments. The association of SIRPα with LMDs and the fact that expression of its cytoplasmic tail (cSIRPα) has a dominant-interfering effect on TNFα signaling (Neznanov et al., 2003) prompted us to choose the soluble cSIRPα fragment for these experiments. Overexpression of cSIRPα in the neuron may have a dominant-interfering effect on growth cone function because it may compete with wild-type SIRPα for phosphorylation of the ITIM motif and for SHP-2 binding. Therefore, we first assessed cSIRPα phosphorylation and SHP-2 binding in transfected cortical neurons. After 48 hours in culture, neurons transfected with the fusion construct GFP-cSIRPα were harvested and solubilized in the presence or absence of ATP plus vanadate (to inhibit protein tyrosine phosphatases). We immunoprecipitated with anti-GFP antibody and probed western blots of the immunoprecipitates with (i) anti-GFP (loading control), (ii) anti-phosphotyrosine and (iii) anti-SHP-2 antibodies. As shown in Fig. 6, anti-GFP precipitated a GFP-positive fusion protein of the appropriate molecular mass (45 kDa), and the intensity of the band in the two experiments was very similar. In phosphorylating conditions, anti-GFP recognized a second band with a slightly higher molecular mass, and this band was pTyr-positive. In the absence of ATP and vanadate, anti-pTyr and anti-SHP-2 demonstrated very low levels of phosphorylation and SHP-2 binding. These were greatly increased in phosphorylation conditions. For two independent experiments, the average increase was tenfold and fivefold for tyrosine phosphorylation and SHP-2 binding, respectively. Thus, cSIRPα expressed in cultured cortical neurons was phosphorylated and bound SHP-2.
SIRPα regulates the rate of axonal outgrowth
Our biochemical experiments showed that laminin and the axonal growth factors IGF-1 and BDNF stimulated SIRPα phosphorylation and the assembly of SIRPα-SHP-2 complexes at the growth cone plasma membrane. This suggested that SIRPα might play a functionally important role in the regulation of axonal growth. To test this possibility, we overexpressed cSIRPα in dissociated primary cortical neurons and studied their outgrowth in culture. Our molecular tools for transfection were: (i) plasmid pcDNA3.1-cSIRPα, which encodes a wild-type, soluble cytoplasmic fragment of SIRPα with dominant-negative properties in other systems (Neznanov et al., 2003; Neznanov et al., 2004), and (ii) plasmid pcDNA-cSIRPα-FYFF, which encodes a phosphorylation-deficient version of the same SIRPα cytoplasmic fragment, with three of the four ITIM tyrosines changed to phenylalanines. The expression of the two proteins, cSIRPα and cSIRPα-FYFF, was confirmed by immunoblot with anti-SIRPα (data not shown). To enable the identification of transfected neurons, we co-transfected with GFP. To establish positive correlation of GFP fluorescence with cSIRPα and cSIRPα-FYFF expression, we used an anti-SIRPα antibody raised against the C-terminal peptide (anti-cSIRPα) that bound only at a very low level to non-transfected neurons (because SIRPα was present in limited amounts and/or because of steric hindrance). In neurons overexpressing cSIRPα or cSIRPα-FYFF, however, this antibody produced strong signals. Fig. 7A-C shows neurons co-transfected with GFP and cSIRPα plasmids. We found that all neurons expressing GFP also exhibited strong anti-cSIRPα reactivity. Non-transfected neurons were not detectable either by GFP or anti-cSIRPα fluorescence. The same results were obtained with GFP and cSIRPα-FYFF co-transfections (data not shown). This confirmed that GFP-positive neurons also expressed cSIRPα or cSIRPα-FYFF.
To define the phenotype induced by overexpression of cSIRPα, we first examined whether growth cone morphology was changed. Fig. 7D-F shows typical, highly motile growth cones from cortical neurons on laminin. As seen in these representative images, the growth cone area, the approximate numbers, lengths, and positions of filopodia, and other morphological features did not appear to be affected by transfection with GFP only (Fig. 7D), with cSIRPα plus GFP (Fig. 7E), or with cSIRPα-FYFF plus GFP (Fig. 7F). This result also indicated that the overexpression of cSIRPα or cSIRPα-FYFF was not neurotoxic under the experimental conditions used. Interestingly, in the growth cones transfected with cSIRPα plus GFP, the immunoreactivity of cSIRPα (labeled with an antibody to a C terminal peptide, anti-cSIRPα) was not diffusely distributed as expected from its solubility. Instead, as shown in Fig. 7H (compare with Fig. 7I), cSIRPα was enriched in the thinner periphery of the growth cone, creating an image reminiscent of the distribution of membrane-anchored SIRPα. All the more surprisingly, perhaps, the distribution of endogenous SIRPα (selectively labeled with an antibody to the external domain) was not affected (Fig. 7G).
Next we analyzed axonal growth rates in control versus cSIRPα-expressing neurons, in defined culture medium containing 10 nM IGF-1. Growth cone advancement involves repeated cycles of extension and stalling, sometimes even involving growth cone collapse or retraction. To average out these short-term dynamic variations, we measured the growth rate as the displacement of the growth cone over periods of 2-3 hours and expressed it in μm/hour. Fig. 8A shows two axons growing on laminin, one transfected with GFP only (control), and another transfected with both cSIRPα and GFP (Fig. 8A, fluorescent images on top). The phase-contrast images show the same axons immediately after the addition of 10 nM IGF-1 to the culture and 1 hour later. The distance of growth cone advancement is indicated by two arrows. The control growth cone (left panels) advanced much further than the cSIRPα-transfected growth cone (right panels). Both maintained their lamellipodial and filopodial dynamics during the observation period, however. When a population of such axons was analyzed (Fig. 8B) quantitative analysis revealed that cSIRPα overexpression (n=14) reduced the growth rate to 33% of the control value. This change was significant statistically (P<0.05). However, overexpression of the non-functional cSIRPα-FYFF (n=8) did not cause a statistically significant change compared to controls. These data indicated that cSIRPα overexpression had a dominant-interfering effect on axonal growth in cortical neurons cultured on laminin and in the presence of IGF-1. Since the difference between cSIRPα and cSIRPα-FYFF was the phosphorylation deficiency of the latter, this indicated that the effect cSIRPα was specific and related to tyrosine phosphorylation of the ITIM motif.
Because laminin stimulates SIRPα phosphorylation in growth cones we wanted to determine whether the reduction in axonal growth rate by cSIRPα required the presence of this ECM molecule. Therefore, we also measured axonal growth rates of cortical neurons cultured on poly-D-lysine. In contrast to laminin, poly-D-lysine is a non-physiological substrate that does not activate integrin signaling. As shown in Fig. 8B (right), the growth rate of GFP control axons is reduced on poly-D-lysine relative to laminin (to 58%; P<0.05). However, cSIRPα overexpression (n=7) did not affect this growth rate (in the presence of IGF-1). There was no statistically significant difference between control and cSIRPα growth rates on poly-D-lysine. These results indicate that the dominant-interfering effect of cSIRPα on IGF-1-stimulated axon outgrowth requires laminin-activated integrin signaling in the growth cone.
In our previous study, we identified the highly heterogeneous, prominent growth cone glycoprotein, gp93, as SIRPα (Quiroga et al., 1994; Wang et al., 2003). SIRPα is involved in various cell functions (Comu et al., 1997; Oshima et al., 2002), but its mechanism of action in the nerve growth cone has been unclear.
Localization of SIRPα in growth cones and LMDs
It was known that SIRPα is present in the perikarya, the neurites and the growth cones of cultured neurons (Chuang et al., 1990; Quiroga et al., 1994; Ohnishi et al., 2005). However, its distribution in the growth cone had not been described in detail. After membrane permeabilization with the very mild detergent Brij 98 (rather than the harsher TX100) we observed a punctate distribution of SIRPα, especially in the distal regions of the growth cone. Sphingolipid/GPI-linked protein microdomains have been shown to coalesce into visible aggregates by antibody cross-linking in standard protocols (Harder and Simons, 1997; Harder et al., 1998), but not after glutaraldehyde fixation (Mayor et al., 1994). In the latter conditions, the punctate pattern of SIRPα persisted (data not shown) and, thus, did not appear to be a preparation artifact. SIRPα localization was also consistent with the distributions reported for β1 integrin and tetraspanins in growth cones (Wu and Goldberg, 1993; Wu et al., 1996; Stipp and Hemler, 2000). Actually, SIRPα co-localized to a substantial degree with CD81 in our experiments, and this was especially evident along, and at the tips of, filopodia, again suggesting that SIRPα may be contained in specific LMDs. Our biochemical data do indeed demonstrate that much SIRPα resides in growth cone LMDs, together with CD81.
Various non-ionic detergents have been used to isolate LMDs (Bohuslav et al., 1993; Roper et al., 2000; Bini et al., 2003; Schuck et al., 2003). Brij 98 solubilizes proteins, such as the transferrin receptor, but preserves the association of tetraspanins with LMDs (Kawakami et al., 2002; Charrin et al., 2003; Hemler, 2003). Like TX100-resistant rafts, Brij 98-resistant LMDs are enriched in cholesterol, sphingolipids and palmitoylated proteins, but they are depleted of prenylated proteins (Chamberlain, 2004). We recovered SIRPα, together with CD81, in a Brij 98-resistant, flotable LMD fraction sensitive to cholesterol perturbation and (unlike Src) to TX100 at 4°C.
Like tetraspanins, SIRPα is not associated with the cytoskeleton (see Berditchevski and Odintsova, 1999). The substantial co-localization and co-fractionation of SIRPα and CD81 suggest that a large proportion of SIRPα and CD81 may reside together in LMDs that are physically and functionally distinct from `standard rafts' and also contain adhesion molecules (Hemler, 2003). This is of particular interest in view of recent findings that critically implicate LMDs and tetraspanins (including CD81) in neurite outgrowth and turning (Schmidt et al., 1996; Banerjee et al., 1997; Stipp and Hemler, 2000; Guirland et al., 2004). It may suggest SIRPα involvement in growth control.
Involvement of SIRPα in axonal growth control
For a long time, cell adhesion has been known to affect neurite outgrowth (Edgar, 1985; Lander et al., 1985a; Lander et al., 1985b). At least in some cases, this seems to involve Rho A (Liu et al., 2002). In the present report, the role of integrin activation by laminin is of particular interest. In earlier reports, integrin-dependent phosphorylation of SIRPα has been associated with growth control, including that of neurites. This seems to involve members of the family of focal adhesion kinases (Tsuda et al., 1998; Oh et al., 1999; Timms et al., 1999; Ivanovic-Dikic et al., 2000). However, it is the Src family kinases that phosphorylate SIRPα in its ITIM motif [which then becomes a SHP-2 binding site (Oh et al., 1999)]. Growth cones are exceptionally rich in Src kinases [especially Src and Fyn (Maness et al., 1988; Helmke and Pfenninger, 1995)], and growth cone Src is known to be activated by integrin binding to laminin (Helmke et al., 1998). It seemed likely, therefore, that laminin would trigger SIRPα phosphorylation in the growth cone, as fibronectin and vitronectin do in fibroblasts and smooth muscle cells (SMCs), respectively (Oh et al., 1999; Maile and Clemmons, 2002b). The growth factors, IGF-1 and BDNF, also are known to increase SIRPα phosphorylation in SMCs (Maile and Clemmons, 2002a) and primary neurons (Ohnishi et al., 1999), respectively. As IGF-1 and BDNF are potent promoters of axonal growth (in appropriate neurons) (LeRoith et al., 1992; Bibel and Barde, 2000; Pfenninger et al., 2003), a process that is known to be facilitated by laminin (Lander et al., 1985). This may suggest that SIRPα is involved in the link between growth factor and ECM signals.
We show here that BDNF and, especially, laminin and IGF-1 transiently stimulate SIRPα phosphorylation in growth cones (GCPs contain functional receptors for all three) (Helmke et al., 1998; Pfenninger et al., 2003). Phosphorylation moves SIRPα out of LMDs; new studies will be necessary to understand the significance of this dissociation. SIRPα phosphorylation in growth cones requires activity of a Src family kinase and results in temporary increase in the association of SHP-2. Phospho-SIRPα is not only a binding site for SHP-2, but also a substrate for the phosphatase. The resulting dephosphorylation of SIRPα (Fujioka et al., 1996; Noguchi et al., 1996) explains the observed biphasic responses of phosphorylation and SHP-2 association to growth cone stimulation with the three factors. Interestingly the kinetics of SIRPα phosphorylation in GCPs were much faster than those reported elsewhere (Tsuda et al., 1998; Oh et al., 1999; Ohnishi et al., 1999; Maile and Clemmons, 2002a). Data suggest that the initial step of SHP-2 binding to phosphorylated SIRPα also serves to activate its phosphatase activity (Ohnishi et al., 1996), which is necessary for the role of SHP-2 in growth factor-activated signal transduction (Kharitonenkov et al., 1997; Araki et al., 2000; Maile and Clemmons, 2002a). In particular, SHP-2 is important for PI 3-kinase activation (Hakak et al., 2000; Wu et al., 2001; Zhang et al., 2002), and PI 3-kinase is necessary for axonal growth (Laurino et al., 2005). Therefore, transient SIRPα phosphorylation is likely to affect growth factor signaling in growth cones.
To examine the effects of SIRPα phosphorylation on axonal growth, we overexpressed its cytoplasmic domain, cSIRPα, in primary cortical neurons. Because these neurons require IGF-1 (or high insulin levels sufficient for activating the IGF-1 receptor) to maintain outgrowth over extended time periods, these experiments were done in the presence of IGF-1, which is known to stimulate membrane addition at the growth cone (Pfenninger et al., 2003). Furthermore, neurons were grown either in the presence or in the absence of integrin activation (laminin versus poly-D-lysine substratum). Soluble cSIRPα is thought to compete with endogenous transmembrane SIRPα for binding to intracellular partners, such as SHP-2. This has been reported to result in a dominant-negative effect in NIH3T3 cells (Neznanov et al., 2003; Neznanov et al., 2004). cSIRPα expressed in cortical neurons was indeed phosphorylated and bound SHP-2. Perhaps surprisingly, cSIRPα expression did not obviously change growth cone morphology, the growth cone cytoskeleton, or the localization of endogenous SIRPα in primary cortical neurons. However, on laminin cSIRPα reduced the axonal growth rate (in the presence of IGF-1) by about 67%. Control experiments involving the overexpression of phosphorylation-deficient cSIRPα did not affect the growth rate. This demonstrated the essential role of phosphorylation in the cSIRPα effect on laminin. We also cultured transfected neurons on poly-D-lysine, which is a more adhesive molecule for these neurons and resulted in larger growth cones with more numerous and more pointed, but shorter filopodia. As on laminin, SIRPα was enriched in these filopodia. For control neurons, the axonal growth rate was reduced (to 58%) on poly-D-lysine relative to laminin. However, upon transfection with cSIRPα, there was no further reduction of that growth rate. Thus, our data demonstrate that the reduction in growth rate induced by overexpression of cSIRPα (in the presence of IGF-1) depends on laminin activation of integrin. This suggests that integrin-stimulated SIRPα phosphorylation affects axonal growth probably by modulating IGF-1 receptor signaling or another mechanism involving SHP-2 (Neel et al., 2003; Ling et al., 2005).
Although such a response has not been shown in growth cones before, our results are consistent with those obtained in other cell systems, such as NIH3T3 cells (Neznanov et al., 2003) and Smooth muscle cells (Maile and Clemmons, 2002a; Maile and Clemmons, 2002b; Clemmons and Maile, 2005). In SMCs, IGF-1 stimulates SIRPα phosphorylation in the presence of the αVβ3 integrin ligand, vitronectin. This results initially in the binding of SHP-2 to SIRPα and to the subsequent transfer of SHP-2 to IGF-1 receptor complexes, where signaling is modulated (Maile and Clemmons, 2002a). This alters cellular growth and migration responses to IGF-1.
Overall, growth cone SIRPα appears to be dynamically associated with LMDs probably including growth factor receptors and integrins (our unpublished observations). Activation of such receptors transiently stimulates, via Src family kinase(s), SIRPα phosphorylation and SHP-2 binding. We show that this affects IGF-1-stimulated axonal growth rates, probably via activation of SHP-2 and modulation of IGF-1 receptor signaling. Thus, SIRPα seems to be necessary for the modulation of axonal growth by ECM molecules, such as laminin. As such it may provide at least a partial molecular explanation for the long-known observation that laminin promotes growth-factor-stimulated neurite outgrowth (Lander et al., 1985a; Lander et al., 1985b; Edgar et al., 1988; Goldberg et al., 2002; Liu et al., 2002). The interaction of SIRPα with CD47 (or other, so far unknown ligands) may further modulate growth factor responses.
Materials and Methods
Timed pregnant Sprague-Dawley rats were purchased from Harlan Laboratories (Indianapolis, IN). LXSN-GSE2-1 and pcDNA-SIRPc-FYFF L-15 plasmids were a kind gift from Nickolay Neznanov, Cleveland Clinic Foundation, OH. Cell culture media and supplements, subcloning reagents, pcDNA3.1 vector, mouse laminin, acrylamide and other gel reagents were from Invitrogen (Carlsbad, CA). Tween 20, Triton X-100 (TX100), Brij 98, bovine serum albumin (BSA), methyl-β-cyclodextrin (MβCD), saponin, β-escin, poly-D-lysine, fluorescein isothiocyanate (FITC)-conjugated phalloidin, recombinant mouse insulin-like growth factor (IGF-1), recombinant human brain-derived neurotrophic factor (BDNF), and protease inhibitor cocktail were from Sigma (St Louis, MO). Trasylol, PP2 (4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrazolo[3,4-d]pyrimidine), PP3 (4-amino-7-phenylpyrazol[3,4-d]pyrimidine), protein A/G plus-agarose and genistein were from Calbiochem (San Diego, CA). Other reagents and their sources were: Alexa Fluor®-conjugated secondary antibodies, Molecular Probes (Eugene, OR); Cy3- and Cy5-conjugated secondary antibodies, Jackson Laboratory (Bar Harbor, ME); anti-SIRPα polyclonal antibody, anti-p85 polyclonal antibody and anti-phospho-tyrosine (pTyr) monoclonal antibody (4G10), Upstate Biotechnology (Lake Placid, NY); anti-SHP2 monoclonal and polyclonal antibodies, Santa Cruz Biotechnology (Santa Cruz, CA); anti-Src polyclonal antibody and anti-Src-pY418 monoclonal antibody, Biosource International (Camarillo, CA); pEGFP-N2 and -C2 vectors, anti-CD81 monoclonal antibody and anti-pTyr monoclonal antibody (PY20), BD Biosciences (San Jose, CA); fetal bovine serum (FBS), HyClone (Logan, UT); rat neuron Nucleofector® kit, Amaxa Biosystems (Köln, Germany); Assistent® cover glasses, Carolina Biological Supply Company (Burlington, NC); SlideBook imaging software, Intelligent Imaging Innovations, Inc. (Denver, CO); BCA protein assay reagent, Pierce (Rockford, IL); anti-SIRPα monoclonal antibody, Serotec (Raleigh, NC); anti-GFP polyclonal antibody (ab290), Abcam (Cambridge, MA); polyvinylidene difluoride (PVDF) membranes, Millipore Corporation (Bedford, MA); ImageQuant 5.2, Molecular Dynamics (Piscataway, NJ); Sykes-Moore chamber, Bellco Glass Inc. (Vineland, NJ); Scion Image 4.0.2, Scion Corp. (Frederick, MD).
Cell culture and transfection
Cerebral cortices were dissected from fetal (day 18) Sprague-Dawley rats and cut into pieces of <1 mm3. These were treated with trypsin (0.5 g/l)-EDTA (2 g/l) for 15 minutes at 37°C. Trypsinization was stopped with complete medium (neurobasal containing glutamine, glucose, 10% FBS, B27), and tissue was gently triturated (10 passages) with glass Pasteur pipettes. Cells were plated in complete medium on either poly-D-lysine- or laminin-coated cover glasses in 35-mm Petri dishes. Cultures were incubated at 37°C in 5% CO2 in air. Medium was replaced after 6 hours with serum-free medium (Neurobasal containing glutamine, glucose, B27) (Brewer et al., 1993). After 48 hours in vitro, long neurites with growth cones were evident.
To express SIRPα fragment in primary neurons, a pcDNA3.1-cSIRPα vector was constructed by inserting an EcoRI fragment from LXSN-GSE2-1 (Neznanov et al., 2003) into pcDNA3.1. Transfection was carried out by electroporation (Nucleofection® system). Briefly, 4.8×106 freshly dissociated cortical cells were resuspended in 100 μl of Nucleofector solution containing 3 μg plasmid DNA. For co-transfection experiments, 0.5 μg of plasmid pEGFP-N2 was added with 3 μg of either pcDNA3.1-cSIRPα or pcDNA-cSIRPα-FYFF. After electroporation, 500 μl of prewarmed, complete medium was added to the cuvette, and cells were transferred into culture dishes. Typical transfection efficiencies of surviving neurons were about 20%.
Cortical cultures were fixed by slow infusion of fixative [4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4, 120 mM glucose, 0.4 mM Ca2+] into the culture medium (Pfenninger and Maylie-Pfenninger, 1981). Thereafter, fixative was removed by rinsing the cultures with PBS containing 1 mM glycine. Cultures were permeabilized with 1% Brij 98 in blocking buffer (3% BSA in PBS) for 2 minutes and incubated for 1 hour at room temperature with blocking buffer without detergent. After labeling with primary antibodies (1-3 hours at room temperature) and washing with blocking buffer, cultures were stained with labeled secondary antibody (Alexa Fluor® 488- or Alexa Fluor® 594-conjugated; 1:200, 1 hour at 37°C). We used FITC-conjugated phalloidin (1:200; 1 hour at 37°C) as a label for F-actin. After three washes, the coverslips were mounted onto slides and the cultures observed by epifluorescence with the Zeiss 63×/1.4 NA oil Plan-Apochromat objective on a Zeiss Axiovert 200M microscope (phase-contrast images, Zeiss 63×/1.25 NA oil Plan-NeoFluar). Digital images were acquired using a Cooke Sensicam camera and processed with SlideBook software. Optical sections were taken at 0.1-0.2 μm intervals and deconvolved using a nearest-neighbor algorithm.
Growth cone isolation
Brains from E18 fetal rats were homogenized in 0.32 M sucrose buffer (0.32 M sucrose, 100 KIU/ml Trasylol (aprotinin), 1 mM MgCl2 and 1 mM TES, pH 7.3) and centrifuged at 3000 g for 15 minutes to generate a low-speed supernatant (LSS). LSS was loaded on a discontinuous density gradient consisting of 0.83, 1.2 and 2.66 M sucrose. The 0.32 M/0.83 M interface contained the isolated growth cones or growth cone particles (GCPs) (Pfenninger et al., 1983; Lohse et al., 1996).
Gel electrophoresis and western blotting
Protein concentrations were determined using BCA protein assay reagent with BSA as the standard. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed essentially as described previously (Laemmli, 1970). Gels were run under reducing conditions except those for CD81. Separated polypeptides were electrotransferred onto PVDF membranes in Tris-glycine buffer containing 15% methanol. Blots were quenched with 5% fat-free milk powder in Tris-buffered saline (TBS; 10 mM Tris-HCl, 150 mM NaCl, pH 7.5) containing 0.2% Tween 20 (TBST), for 1 hour at room temperature. Incubation with primary antibody, in TBST/5% milk powder, was for 2 hours at room temperature. After washing, the blots were incubated with Cy3- or Cy5-conjugated secondary antibody in TBST for 1 hour at room temperature. After washing, bound antibody was imaged with a Typhoon® 9400 laser scanner (General Electric). The images were analyzed by ImageQuant 5.2 software. Sometimes, data were collected in both the Cy5 and the Cy3 channels for simultaneous quantification of two labeled polypeptides.
Lipid microdomain analysis
GCPs were extracted for 5 minutes with 1% TX100 (w/v) or 1% Brij 98 (w/v), on ice or at 37°C, respectively. Solubilized samples were mixed with 2 M sucrose (final concentration, 1.33 M) and chilled on ice before being placed at the bottom of a step gradient (1.0 M, 0.9 M, 0.7 M, 0.5 M sucrose). Gradients were centrifuged at 200,000 g for 16 hours at 4°C. Fractions (0.5-ml) were collected and protein precipitated with methanol/chloroform (Wessel and Flugge, 1984). Pellets were analyzed by SDS-PAGE and western blotting.
For cholesterol sequestration GCPs were pre-treated with 0.2% saponin (w/v) at 4°C, or with 10 mM MβCD at 37°C, for 30 minutes before detergent extraction.
Neurite growth rate measurements
After 48 hours, in vitro-transfected neurons were identified by their green fluorescence. Neurite growth rates were assessed by time-lapse imaging. Cover glasses mounted into Sykes-Moore chambers were covered with serum-free, modified B27 neurobasal medium without insulin. 10 nM IGF-1 was added and the medium overlaid with mineral oil (embryo-tested; Sigma) to maintain pH and prevent evaporation. The chamber was kept on the pre-heated microscope stage under convective heating at 37°C. Neurons were imaged at 15-minute intervals for 2-3 hours (Zeiss Axiovert 200 M microscope, 20×/0.5 NA or 40×/0.75 NA Plan-NeoFluar objectives; Cooke Sensicam camera). With Scion Image 4.0.2 we measured the distance between the center of growth cones and the center of the neuronal perikarya or the point where the neurite entered the field. Neurites (presumably, axons) had to meet the following criteria (Lemmon et al., 1989): longest process of the neuron before and after the measurement; tipped by a single growth cone; growth unobstructed and without branching; no major retraction.
Phosphorylation assay in isolated growth cones
The GCP suspension from the gradient was mixed with an equal volume of `intracellular buffer' (20 mM Hepes pH 7.3, 50 mM KCl, 5 mM NaCl, 3 mM MgCl2), permeabilized with 0.01% β-escin, and incubated for 5 minutes on ice, in the presence or absence of factor or other reagent (laminin, 35 μg/ml; BDNF, 0.2 nM; IGF-1, 1 nM; PP2, 1 μM; PP3, 1 μM). Upon the addition of 1 mM ATP, GCPs were warmed to 37°C for 1 or 5 minutes. Usually, the reaction was terminated by chilling and adding 1% TX100 plus 3 mM vanadate, 2 mM NaF, 10 mM EDTA, 100 μM genistein and protease-inhibitor cocktail. After 10 minutes on ice, Triton-insoluble elements were pelleted at 30,000 g for 1 hour. SIRPα was immunoprecipitated from this supernatant with anti-SIRPα (6 μg/ml) for 2 hours at 4°C before adding protein A- and G-coated beads. Fresh vanadate was added to every solution change. Precipitates were resolved by SDS-PAGE, blotted, and probed with anti-pTyr to reveal SIRPα phosphorylation, and with anti-SHP2 to examine SHP2 association. In LMD experiments, reactions were stopped as above, but without adding TX100, and samples were extracted with 1% Brij 98. In some cases the reaction was terminated by methanol/chloroform precipitation. To detect activated Src, the resulting pellets were analyzed by western blot using phospho-Src-specific antibody (anti-Src-pY418).
Immunoprecipitation of cSIRPα from culture
cSIRPα was inserted into the pEGFP-C2 vector for expression of GFP-cSIRPα fusion protein in cultured cortical neurons. Forty-eight hours post-transfection, cultures were treated with medium with or without 1 mM vanadate for 10 minutes and cells were scraped off and homogenized. A low-speed supernatant of harvested material was permeabilized with 0.01% β-escin and incubated with or without 1 mM ATP plus vanadate at 37°C for 10 minutes, followed by the addition of 1% TX100 and protease inhibitors. After 10 minutes on ice, samples were centrifuged at 10,000 g for 10 minutes and GFP-cSIRPα was immunoprecipitated from the supernatant with anti-GFP antibody as described above.
The authors wish to thank Nicolay Neznanov of Cleveland Clinic Foundation, Cleveland, OH, USA for the generous gift of wild-type and mutant SIRPα tail constructs. This work was supported by NIH grant R01 NS41029.
- Accepted September 27, 2005.
- © The Company of Biologists Limited 2006