The Microtubule Regulator Stathmin Is an Endogenous Protein Agonist for TLR3 Free

Toll-like receptors are key components of the innate immune system that detect microbial infection, trigger innate host-defense responses, and control adaptive immune responses (1, 2). In addition to recognizing a range of pathogen-associated molecules, several TLRs sense endogenous ligands that emerge upon tissue damage or inflammation. These include breakdown products of the extracellular matrix (3) and inducible molecules, such as CD14 and heat-shock proteins (4, 5). The only known ligands for TLR3 are dsRNAs, derived from viral, mammalian, or synthetic origin (6–8). Different from other TLR family members, TLR3 cannot activate MyD88-dependent signaling pathways (1, 2). Instead, TLR3 signals through the Toll/IL-1 receptor (TIR) domain-containing adaptor TRIF, which activates IκB kinase-ε and TANK-binding kinase 1 and, ultimately, IFN regulatory factors 3 and 7. Also, NF-κB and MAPKs are activated by TLR3 (1, 2). Consequently, TLR3 mediates the production of type I IFNs, IFN-inducible products, and other inflammatory mediators crucial for the development of host-defense responses against viral infection. Recent evidence indicates that TLR3-mediated signaling involves additional, still unknown upstream events in the signaling pathway (9). Particularly intriguing are recent observations that disruption of membrane microdomains reduces the cellular response mediated by TLR3 (10). This strongly suggest that still unknown surface interactions influence ligand binding and subsequent signaling by TLR3.

Although the role of TLR3 in mediating antiviral responses is supported by a wealth of evidence, several findings indicate that TLR3-mediated signaling can play additional roles, especially in the CNS. TLR3 is selectively and strongly induced in astrocytes by a variety of stressors, including noninfectious ones, and it triggers production of a range of immunoregulatory mediators and factors that counteract gliosis and promote neuronal survival, angiogenesis, and remyelination (11). In addition, TLR3 is expressed in neurons (12) and oligodendrocytes (13) in which it regulates growth and differentiation. Neuronal TLR3 is a major negative regulator of the growth of axons and dendrites, promotes dendrite branching, and controls proliferation of embryonic neural progenitor cells (14, 15).

Together, these observations indicate that at least in some cell types and under some conditions, TLR3 may play a role other than controlling antiviral host-defense responses. Considering, in particular, the emerging role of TLR3 in astrocyte-mediated neuroprotection and in morphogenesis in the CNS, we hypothesized that for such alternative, noninfectious roles of TLR3, alternative endogenous agonists might exist. To explore this idea, we used cDNA from brain tumors as a source of endogenous CNS-borne proteins to screen for such alternative TLR3 agonists. By using primary human astrocytes as initial reporter cells, we identified the microtubule regulator stathmin as an agonist for TLR3-dependent signaling in human astrocytes and microglia.

Fresh post mortem human brain tissue samples were used for the isolation of primary human glial cells. Brain materials were collected after review and approval of the protocol by a local ethical committee. Adult human astrocytes and microglia from subcortical white matter, as well as murine astrocytes from TLR3-deficient mice or control Biozzi ABH mice, were isolated and cultured as previously described (13). Final astrocyte and microglia cultures were judged to be >99% pure by staining for glial fibrillary acidic protein (astrocyte marker) and CD68 (microglia marker).

Plasmid pools were prepared from a cDNA library derived from pooled human brain tumors (meningioma, oligodendroglioma, astrocytoma [grades II and IV] and medulloblastoma [Invitrogen, Carlsbad, CA]) and transfected into COS-7 cells, as previously described in detail (4). To screen for potential TLR-activating proteins, COS-7 culture supernatants collected after 48 h were added to astrocyte cultures, and CXCL8 accumulation in the astrocyte culture medium was evaluated after 48 h by ELISA (Sanquin, Amsterdam, The Netherlands). Sequencing of relevant plasmids was performed by BaseClear (Leiden, The Netherlands).

Commercially available recombinant human stathmin (Calbiochem, San Diego, CA) was used in the experiments in Fig. 1. In all other experiments, we used recombinant human stathmin produced in-house. Stathmin was amplified from the astrocytoma cell line U373 cDNA using the forward primer 5′-CATATGCCATGGCTTCTTCTGATATCCAGGTG-3′ and reverse primer 5′-TCTAGATGGATCCTTAGTCAGCTTCAGTCTCGTC-3′ and cloned into a TOPO vector following the manufacturer’s protocol (Invitrogen). After subcloning, stathmin was purified from recombinant Escherichia coli bacteria by disruption of cells, anion exchange chromatography, and, finally, reversed-phase HPLC. To eliminate endotoxins, the purified protein preparation was passed over MustangE membranes (Pall, East Hills, NY). To obtain active stathmin, the α helical content was maximized by incubating the purified protein for 1 h at 37°C in PBS supplemented with 5% (v/v) 2′,2′,2′-trifluoroethanol (TFE) and subsequent dialysis against three changes of PBS over a period of 48 h at 4°C. The protein preparation obtained was stored at −80°C. Purity and identity were verified by SDS-PAGE and Western blotting. Adequate elimination of bacterial contaminants was verified by testing the preparation for activation of TLR2- or TLR4/MD2-expresssing HEK293 reporter cells (see below).

Human endothelial kidney (HEK293) cells transfected with TLR2, TLR3, or TLR4/MD2 (InvivoGen, San Diego CA) were cotransfected using Polyfect (Qiagen Benelux, Venlo, The Netherlands) with a reporter vector expressing luciferase under the control of an NF-κB–responsive promoter (pNifty2-luc; InvivoGen). Stably transfected clones were selected and used in bioassays. Cells were plated in flat-bottom 96-wells plates at a density of 1 × 10⁵ cells/well and incubated with the various test ligands for 16 h at 37°C. Subsequently, cells were lysed in Steady Glo luciferase buffer (Promega Benelux, Leiden, The Netherlands), and bioluminescence was measured using a Packard 9600 Topcount Microplate Scintillation & Luminescence Counter (Packard Instrument, Meriden, CT). As a positive control for NF-κB–mediated activation (i.e., the presence of the pNifty2-luc vector), 25 ng/ml TNF-α (PeproTech, London, U.K.) was used (data not shown). TLR-specific ligands (all from InvivoGen) were used as positive controls for TLR-mediated activation.

Transfection of human astrocytes with small interfering RNA (siRNA) was performed using Lipofectamine 2000 (Invitrogen). Sequences used were: 1) sense sequence: 5′-GAAGCUAUGUUUGGAAUUAUU-3′, antisense sequence: 5′-pUAAUUCCAAACAUAGCUUCUU-3′; 2) sense sequence: 5′-GAUCAUCGAUUUAGGAUUGUU-3′, antisense sequence: 5′-pCAAUCCUAAAUCGAUGAUCUU-3′; and 3) sense sequence: 5′-GAAGAGGAAUGUUUAAAUCUU-3′, antisense sequence: 5′-pGAUUUAAACAUUCCUCUUCUU-3′. Levels of TLR3-encoding transcripts were determined by real-time PCR and compared with levels of transcripts encoding β-actin, as previously described (11).

To verify the nature of the active components in stathmin and polyinosinic:polycytidylic acid [poly(I:C)], both preparations were digested with trypsin (Sigma-Aldrich, Zwijndrecht, The Netherlands) or the dsRNA-cleaving enzyme RNAse III (Applied Biosystems, Nieuwerkerk a/d IJssel, The Netherlands). Treatment with trypsin was performed at an enzyme/substrate ratio of 1:100 in 20 mM Tris-HCl (pH 7.6) for 15 or 30 min at 37°C. Treatment with RNAse III was performed using 350 U/ml enzyme in 10 mM Tris HCl (pH 7.9), 50 mM NaCl, 10 mM MgCl₂, and 1 mM DTT for 4 or 8 h at 37°C. Digestions were terminated by freezing the samples and storing them at −20°C until testing their ability to induce CXCL8 release in astrocytes, as described above.

Analysis of the mRNA profile of astrocytes stimulated with TLR agonists was performed using Clontech Atlas arrays (BD Clontech, Palo Alto, CA), as previously described (16). Hybridization signals were calculated as the mean of duplicate measurements, corrected for background intensity and divided by the mean signal for all housekeeping reference genes. The relative levels of expression and their ratios shown in Fig. 3 and Table I are representative of two independent experiments. Ab array profiling of cytokine and chemokine production was performed using Raybio arrays, according to the manufacturer’s instructions (Raybiotech, Norcross, GA). These experiments were performed three times, using freshly isolated microglia from different donors.

The transcript profiling data presented in this article have been deposited in the public database ArrayExpress (www.ebi.ac.uk/microarray-as/ae/) under accession code E-TABM-851.

Human subcortical white matter or gray matter, containing chronic active multiple sclerosis (MS) lesions, or noninflamed normal-appearing tissue was stained for TLR3 (mouse anti-human TLR3 clone 3.7; eBioscience, San Diego, CA) and stathmin (rabbit anti-stathmin; Calbiochem, Darmstadt, Germany), treated with secondary Abs (goat anti-mouse Alexa-Fluor 594 and goat anti-rabbit Alexa-Fluor 488; both from Invitrogen), and examined for colocalization of both markers by confocal laser-scanning fluorescence microscopy. Images for both reporters were collected simultaneously. Identification of the different neural cell types was based on their morphology, and additional staining data from adjacent serial sections were characterized using cell-specific markers. The data in Fig. 8 are representative of 10 samples of MS lesions from five patients and three samples of normal-appearing white matter.

Statistical analysis was performed using an SPSS statistical package (Chicago, IL). A p value < 0.5 was considered statistically significant.

Given the intriguing roles of TLR3, especially in the CNS, we chose a cDNA library derived from pooled human brain tumors as a source of potential protein agonist sequences. As a first step, previously described in more detail (4), COS-7 cells were separately transfected with 768 plasmid pools, each containing ~50 different sequences from the cDNA library. Next, COS-7 culture supernatants collected 48 h after transfection were separately transferred to cultures of human astrocytes, in which the accumulation of CXCL8 was evaluated after 48 h. CXCL8 was chosen as a reporter chemokine for astrocyte activation based on our previous studies (11, 16). Although not a specific reporter for any individual TLR family member, or even for TLRs as such, CXCL8 is a highly sensitive initial astrocyte marker for activation of NF-κB, which occurs in astrocytes upon TLR3 signaling. Although most COS-7 culture supernatants did not dramatically influence CXCL8 production by astrocytes, one of the supernatants induced marked CXCL8 release by human astrocytes compared with background (Fig. 1A). This active plasmid pool was subcloned, and individual plasmid sequences were again used for COS-7 cell transfection and subsequent astrocyte screening. Of 96 individual plasmids derived from the active pool, two were found to encode proteins that stimulated CXCL8 accumulation to levels of 160 and 210 ng/ml, whereas in all other cultures CXCL8 levels remained around background levels of 2–3 ng/ml (Fig. 1B). Analysis of one of the active plasmids (indicated by the open arrow in Fig. 1B) yielded the sequence of soluble CD14. Subsequent analysis allowed us to define it as a selective TLR2 agonist (4). The sequence of the second plasmid was found to encode amino acid residues 44–149 of human stathmin (Fig. 1C).

Prompted by this finding, we tested whether a commercial preparation of human stathmin could also activate human astrocytes. This recombinant stathmin indeed led to markedly enhanced CXCL8 release by human astrocytes in a dose-dependent manner (Fig. 1D). Most likely, therefore, full-length stathmin and the truncated C-terminal fragment of stathmin, which emerged from the above screen, share a similar activity. When added at concentrations of 300 μg/ml, recombinant full-length stathmin promoted the release of CXCL8 to levels similar to those found in response to 50 μg/ml poly(I:C) (Fig. 1D). To verify that this response was not due to endotoxin contamination of the recombinant preparation, responses by HEK293 cells transfected with TLR2 or the combination of TLR4 and MD2 were tested. Neither of these reporter cell lines showed any response to the recombinant stathmin preparation, while showing the expected response to the appropriate control agonists. Preparations of recombinant stathmin that were subsequently prepared in-house for the remainder of this study were also consistently found to be free of functional endotoxins and other bacterial contaminants (see below).

Although the above data indicate a clear astrocyte response to stathmin, they do not immediately clarify the nature of the receptor(s) involved. In addition to TLR3, astrocytes express TLR1, TLR2, and TLR4 (11), and other receptors could also be involved in mounting a cellular response to stathmin. Therefore, we examined the response to stathmin of astrocytes isolated from wild-type or TLR3-deficient mice. In this case, IL-6 release was used as a marker for TLR-mediated activation on murine astrocytes. A marked dose-dependent induction of IL-6 by stathmin was found in all cultures of wild-type murine astrocytes (Fig. 2A), whereas such a response was completely absent from astrocytes isolated from TLR3-deficient mice (Fig. 2B). This response profile paralleled that of poly(I:C), which was equally unable to induce any response in TLR3-deficient astrocytes. This result also rules out any contribution to the poly(I:C)-induced response in astrocytes by alternative receptors for dsRNA, such as RIG-I or mda-5. If such receptors had contributed to the astrocyte response against poly(I:C), some of that response should have persisted in cells from which only TLR3 was eliminated. Astrocytes from wild-type or TLR3-deficient mice responded equally well to LPS, confirming adequate astrocyte viability and TLR-mediated IL-6 release potential in all cultures. These findings demonstrated that the astrocyte response to stathmin, like that to poly(I:C), critically depends on the presence of TLR3.

To confirm this, human adult astrocytes were stimulated with stathmin or poly(I:C) following transfection with different siRNAs to selectively silence TLR3 expression. First, we verified that siRNA-mediated silencing of TLR3 expression was effective when TLR3-encoding mRNA is induced at the same time, as is the case upon activation of astrocytes with poly(I:C) (11) or, indeed, stathmin itself. As shown in Fig. 2C, both stimuli induced TLR3 expression in the absence of any siRNA, but silencing with any one of the three siRNA variants consistently resulted in a decrease (by >90%) in TLR3-encoding mRNA expression. In parallel, a strong reduction was found in CXCL8 levels that were induced by either stathmin or poly(I:C) in siRNA-treated human astrocytes over a period of 48 h (Fig. 2D). In all cases, levels of CXCL8 were reduced to around background levels. The almost complete abrogation of TLR3 mRNA expression and CXCL8 release after treating astrocytes with siRNAs confirms that stathmin-mediated cellular activation is crucially dependent on TLR3 in human astrocytes as well.

To verify that the CXCL8-inducing activity of the stathmin preparation is fully mediated by a protein component, rather than any RNA contaminant, stathmin was treated with trypsin or the dsRNA-cleaving enzyme RNAse III. As shown in Fig. 3, tryptic digestion of stathmin led to the disappearance of astrocyte-activating potential, whereas treatment with RNAse III had no effect. The result was the opposite for the agonist poly(I:C).

The above findings clarify that stathmin- and poly(I:C)-induced cytokine responses in murine and human astrocytes are dependent on the presence of TLR3. As clarified above, any possible response mediated by alternative dsRNA receptors, such as intracellular helicases (2), which would have persisted in TLR3-deficient astrocytes or after siRNA-mediated TLR3 silencing is essentially undetectable. Given this critical role of TLR3 in the astrocyte response to stathmin and poly(I:C), transcript analysis would be expected to reveal the same intracellular response profile. Therefore, astrocytes were stimulated for 48 h with 50 μg/ml stathmin or poly(I:C), and transcript profiling was performed using a cDNA array containing probes for 268 cytokines, chemokines, growth factors, and their receptors, the prime domain of functional mediators produced by astrocytes. We demonstrated the validity of these arrays for astrocyte-response profiling in previous studies (11, 16). As illustrated in Fig. 4A and 4B, and summarized in more detail in Table I, the stimulation of astrocytes with either agonist led to induction of the same range of mediators and at very similar levels. In contrast, stimulation of astrocytes via TLR4 using ultrapure LPS induced a very different response profile, confirming the selectivity of the response profiles to stathmin or poly(I:C) (Table I).

In line with previous data (11), stathmin and poly(I:C) induced a variety of factors involved in neuroprotection and repair, including LIF, β nerve growth factor, metallothionin-III, fibroblast growth factors, GTPase-activating protein-43, neurotrophins, vascular endothelial growth factor, brain-derived neurotrophic factor, pleiotrophin, glial maturation factor, and insulin-like growth factor 1. Previously, we demonstrated the ability of this mixture of mediators to significantly improve survival of human neurons in organotypic brain slice cultures (11). Stathmin also induced ephrin type A receptor 2, as well as ephrins B1 and B3, bidirectional signaling molecules controlling axonal growth, synapse formation, and astrocyte reactivity. Among the cytokines induced were the immunoregulatory factors IL-10 and -11, IFN-γ antagonist, and IFN-β. No significant induction of TNF-α was observed. Chemokines induced included CCL2, CCL3, CCL5, CXCL2, CXCL6, and CXCL8; the latter confirmed the suitability of CXCL8 as an initial marker for TLR3-mediated activation of astrocytes. No gene was markedly suppressed by stathmin or poly(I:C). Importantly, the Pearson correlation coefficient, which expresses the extent of similarity between the two datasets (Fig. 4C), reached +0.951, confirming that the functional response at this level to a 50-μg/ml dose of stathmin or poly(I:C) is essentially identical (p < 10⁻⁵). Again, these results provided strong evidence for the notion that stathmin, like poly(I:C), activates TLR3-dependent signaling in astrocytes.

In the course of our experiments, we noted marked differences in the dose-response profiles of different preparations of recombinant stathmin. Initial experiments using a commercial preparation of stathmin suggested that the agonist would be markedly less efficient on a weight basis than poly(I:C) in activating astrocytes (Fig. 1). However, material subsequently produced in-house was as effective as poly(I:C), on a weight basis, in inducing cellular responses (Fig. 4, Table I). Because the purities of the different stathmin preparations were comparable, and all were free from bacterial contaminants, this raised the possibility that additional factors involving, for example, secondary structure, affected the TLR3-activating activity of stathmin. One striking feature of stathmin is its ability to fold into a very long α helix that may contain up to 94 aa (17). This helix contains no less that 24 negatively charged amino acids, mostly glutamic acid residues.

Therefore, we examined whether the α helical content of stathmin would affect its ability to activate TLR3-mediated astrocyte responses. Purified human recombinant stathmin was subjected to conditions that differentially affect helical content. One part of the protein preparation was incubated in PBS containing 5% of the helix-promoting solvent TFE (18). The other part was kept in water acidified with 5% of the helix-breaking solvent trifluoroacetic acid (TFA). After exposure to either of these conditions, stathmin was extensively dialyzed over 48 h against several changes of PBS and subsequently analyzed by near-UV circular dichroism (CD) spectroscopy. The TFE-treated sample contained 37% α helix on average, whereas the TFA-treated sample produced a CD spectrum indicative of only 12% helix content (Fig. 5A). No other changes in the protein preparations (e.g., solubility or aggregated state) were observed. As a reference sample for CD measurements, part of the stathmin preparation was also dissolved in a solution of 50% TFE in PBS immediately prior to CD spectroscopy. This sample produced a spectrum indicative of 98% α helical content (Fig. 5A), confirming the extreme α helical propensity of stathmin.

In parallel, human adult astrocytes were stimulated with different amounts of either dialyzed protein preparation, and CXCL8 release was monitored after 48 h. As shown in Fig. 5B, the TFE-treated stathmin preparation that contained three times more α helix triggered CXCL8 release ~3-fold more effectively on a weight basis than did the TFA-treated preparation. Because the two stathmin preparations tested were otherwise identical in purity and concentration and were contained in the exact same solvent following dialysis, we concluded that the α helix of stathmin is of critical importance to its ability to activate TLR3. To further support this notion, stathmin completely lost its TLR3-agonist activity after full denaturation as a consequence of exposure to organic solvents at pH 2 during reversed-phase HPLC, followed by lyophilization (Fig. 5B). This finding confirms the absence of interference by bacterial contaminants, whose activity would not be affected by these conditions.

We next tested whether stathmin also induces TLR3-mediated signaling in human microglia. As opposed to astrocytes, which express TLR3 primarily on their surface, microglia express TLR3 exclusively in intracellular vesicles (13). Given the apparently critical role of the α helical conformation of stathmin, uptake and acidification of stathmin in the endosomal pathway of microglia could conceivably affect its ability to activate TLR3. However, as shown in Fig. 6, stathmin also retains TLR3-activating potential when taken up by microglia. Examination of microglial responses should take into account that cultured, but otherwise untreated, microglia already secrete significant amounts of certain mediators. In our cultures, freshly isolated, but otherwise untreated, microglia grown with GM-CSF produced large amounts of IL-6 and several chemokines, including CXCL8, CCL2, CCL3, CCL4, and CCL8. Several of these mediators were present at such high levels that any additional increase by subsequent treatment with a TLR agonist probably would be limited for this reason alone. This is particularly likely to apply to IL-6, CXCL8, and CCL2. Despite this, exposure of microglia to stathmin led to marked induction of IL-13, CCL5 (RANTES), IL-1β, IL-6 soluble receptor, TNF-α, and several other mediators that are consistent with TLR-mediated signaling. Again, the response by microglia to stathmin was similar to that induced by poly(I:C) on a weight basis, albeit not as identical as was found in astrocytes. Compared with stathmin, poly(I:C) induced a very similar set of factors but markedly higher levels of CXCL10 and lower levels of CCL1 and IL-1α (Fig. 6). These differences between stathmin- and poly(I:C)-induced responses in microglia likely reflect the influence of additional intracellular receptors for poly(I:C) in these cells, such as RIG-I, mda-5, and other helicases that are not functional in astrocytes.

The above data provided compelling evidence for the ability of stathmin to activate astrocytes and microglia. The lack of any response by TLR3-deficient astrocytes or astrocytes subjected to selective TLR3 knockdown, as well as the congruity in astrocyte transcript responses to stathmin and poly(I:C), support the notion that this stathmin-induced response is TLR3 dependent. Therefore, we expected to be able to confirm the agonist activity of stathmin using TLR3-transfected HEK293 cells, a routine approach to screen for TLR ligands. Surprisingly, however, TLR3-expressing HEK cells failed to respond to stathmin, despite being fully responsive to poly(I:C). As shown in Fig. 7, HEK cells transfected with TLR2 or TLR4 did not show any response to stathmin, while mounting a significant response to appropriate control ligands. Although this confirmed that cellular responsiveness to our stathmin preparations was not attributable to bacterial contaminants, the most obvious of which will trigger a response by TLR2- or TLR4-expressing HEK cells, the consistent lack of responsiveness to stathmin by TLR3 transfectants was unexpected. In view of recent evidence that TLR3-mediated signaling likely depends on several unknown factors other than TLR3 alone, including additional unknown surface factors (9, 10), we concluded that HEK293 cells probably lack such factors, whereas they are present on astrocytes and microglia.

Experimental verification of functional stathmin–TLR3 interactions in vivo by straightforward knockout approaches is hampered by the multiple biological roles and interaction partners of either molecule and by the fact that mammals possess four stathmin homologs that share the active α helical segment of stathmin. Yet, given that TLR3 and stathmin are induced during neuroinflammation (13, 19, 20), we examined whether they colocalized under such conditions in the human CNS. Upon examination of chronic active MS lesions, widespread expression of TLR3 was found, primarily inside microglia and on the surface of astrocytes and their processes, as well as on neurons and oligodendrocytes. Stathmin was expressed at comparable levels in all cell types. This localization of either protein is consistent with earlier reports (13, 19, 21). Previously, we reported that TLR3 is undetectable by immunohistochemistry in healthy control brains or noninflamed normal-appearing white matter of MS patients (13); our current data, as illustrated in Fig. 8, confirmed this. However, TLR3 is clearly upregulated in chronic active MS lesions. Under these conditions, colocalization of TLR3 and stathmin was mainly observed on the surface of astrocytes and neurons and inside small intracellular vesicular structures of microglia and astrocytes. Although oligodendrocytes in MS lesions displayed marked surface expression of TLR3, as well as intracellular expression of stathmin, colocalization of both proteins on the surface of or inside oligodendrocytes was only rarely observed.

The main conclusion from this study is that stathmin, a protein that has long been known as a ubiquitously expressed regulator of microtubule formation, is able to activate TLR3-dependent signaling in astrocytes and microglia. Several key observations support this notion. The stathmin-induced CXCL8 reporter response is completely abrogated in TLR3-deficient murine astrocytes and in human astrocytes subjected to TLR3 knockdown. In both cases, the elimination of stathmin-induced responses paralleled the elimination of poly(I:C)-induced ones. Transcript profiling of stathmin- or poly(I:C)-induced astrocyte responses, using cDNA arrays to monitor 268 cytokines, chemokines, and growth factors, revealed essentially identical activation pathways. The notion that stathmin may act as a TLR3 agonist in vivo is supported by the finding that, at least during the neuroinflammatory process of MS, stathmin and TLR3 colocalize on and inside a variety of neural cells, including astrocytes, microglia, and neurons.

Yet, TLR3 alone is apparently not sufficient to confer cellular responsiveness to stathmin, as clarified by the lack of any response in HEK293–TLR3 transfectants. Therefore, stathmin-induced cellular activation most likely requires additional receptor components, possibly surface coreceptors, that are absent from HEK293 cells. That TLR3-dependent signaling requires additional receptor components was suggested by a detailed analysis of TLR3-signaling pathways. As explained by Helmy et al. (9), dynamic simulation of TLR3 signaling strongly suggests the existence of still-unknown intermediary steps in the cascade of signaling events, notably upstream of TLR3 dimerization. That such still-missing elements likely include additional receptor components in surface membrane microdomains also emerged from a recent study in which disruption of such domains reduced poly(I:C)-mediated signaling in human PBMCs (10). Evidence is rapidly accumulating that TLR-dependent responses are generally controlled by multiple receptor/coreceptor interactions of previously unsuspected complexity (22, 23). It seems likely that this also applies to TLR3-mediated responses to stathmin, and it may explain why the expression of only TLR3 itself by HEK293 cells might not be sufficient to confer stathmin responsiveness. In addition to astrocytes and microglia, we found that human monocyte-derived dendritic cells were responsive to stathmin, as evidenced by marked secretion of CXCL8 in response to the protein (data not shown). Nonresponsiveness to stathmin in HEK293 cells might be related to their endothelial origin, a trait by which they differ from apparently responsive myelomonocytic or fibroblast-like cells.

Stathmin has been well characterized as a catalyst of the reversible process of microtubule catastrophe and repolymerization (24, 25). As such, stathmin is widely expressed in many cells, especially in those engaging in proliferation and/or migration. Intriguingly, however, the strikingly dominant expression of stathmin in mammalian brains and spinal cord, and especially in fetal brains (26), suggests a special relevance to the CNS, unexplained by its traditional and ubiquitous role as a regulator of microtubules. This applies even more strongly to the three other family members of stathmin, the homologs superior cervical ganglion 10 protein (SCG10; stathmin-2), SCG10-like protein (SCLIP; stathmin-3), and RB3 (stathmin-4), which are highly conserved among vertebrates. For example, expression levels of SCLIP and RB3 in the CNS are ≥100 times greater than in other human tissues (26). After brain trauma, stathmin levels increase (19, 20), as they do during axonal and dendritic growth (21, 27). The particular significance of stathmins to the CNS is further supported by the observations that stathmin-deficient mice show defects in innate and acquired fear (28) and selectively develop axonopathy at advanced age (29). Drosophila flies in which the stathmin gene has been knocked down by siRNAs develop widespread defects in the nervous system only (30). All of these observations point to a special role of stathmins in nervous system development and are difficult to rationalize in the context of the ubiquitous role of stathmins in microtubule rearrangement alone.

It seems likely that all four stathmin family members share the ability to activate TLR3. The helix-forming sequence 44–139 of stathmin shares between 92% and 95% sequence homology and between 67% and 75% amino acid identity with the other three family members SCG10, SCLIP, and RB3. However, the exact mechanism via which stathmins might activate TLR3 in vivo remains to be established. Other family members of stathmin are plausible in vivo ligands, much more so than stathmin itself, best known as a cytosolic protein. SCG10, SCLIP, and RB3 do not differ from stathmin in their α helical properties, but they do differ in having an extended N-terminal sequence. Acylation of this sequence targets these stathmins to Golgi vesicles and to additional vesicular neuronal structures that are found along neurite shafts and in growth cones (31, 32). These vesicular structures in neurons are very similar to those reported to contain TLR3 (14). Although the roles of neuronal TLR3 and stathmins are not fully understood and are probably complex, striking similarities are starting to emerge. Both can inhibit the growth of neuronal precursors, axons, and dendrites, and both can promote dendrite branching (14, 15, 32). Therefore, direct stathmin–TLR3 interactions on or inside neurons, as revealed by our present data (Fig. 8), might underlie this functional congruence between the two proteins. Other ways for stathmins to contact TLR3 in the CNS may be envisaged as well. Because stathmin can be found in myelin sheaths (19, 33) and is even upregulated during neuroinflammation (19, 20), it is likely to become ingested by phagocytic microglia that emerge in response to infection or trauma. Myelin phagocytosis would readily allow stathmin to directly contact endosomal TLR3 in microglia, a scenario supported by the present immunohistochemical findings (Fig. 8). Finally, alternative splicing products might play a role. Preliminary data suggested that truncated forms of stathmin that contain the helix-forming C-terminal sequence 44–149, but lack the N terminus, can be secreted by mammalian cells, while retaining their ability to activate TLR3. This observation suggests that alternative splicing of stathmin genes could lead to N-truncated, naturally secreted forms that could conceivably activate TLR3 on other cells. Clearly, clarification of the exact physiological mechanisms via which stathmin and/or its family members contact and activate TLR3 require more detailed studies.

In conclusion, our data provide evidence for the ability of stathmins to activate TLR3-dependent cellular responses. Also, in the context of other data, our findings suggest that the functional link between stathmins and TLR3 is likely to be of particular relevance to development and repair in the CNS.

Human post mortem brain tissue was obtained through the Netherlands Brain Bank. We thank L. van Straalen and E. Zuiderwijk-Sick for technical support.

Disclosures The authors have no financial conflicts of interest.