[1] 22w Interleukin-6 (IL6) and cellular response to facial nerve injury: effects on lymphocyte recruitment, early microglial activation and axonal outgrowth in IL6-de®cient mice

[1] 182w Nerve injury triggers numerous changes in the injured neurons and surrounding non-neuronal cells. Of particular interest are molecular signals that play a role in the overall orchestration of this multifaceted cellular response. Here we investigated the function of interleukin-6 (IL6), a multifunctional neurotrophin and cytokine rapidly expressed in the injured nervous system, using the facial axotomy model in IL6-de®cient mice and wild-type controls. Transgenic deletion of IL6 caused a massive decrease in the recruitment of CD3-positive T-lymphocytes and early microglial activation during the ®rst 4 days after injury in the axotomized facial nucleus. This was accompanied by a more moderate reduction in peripheral regeneration at day 4, lymphocyte recruitment (day 14) and enhanced perikaryal sprouting (day 14). Motoneuron cell death, phagocytosis by microglial cells and recruitment of granulocytes and macrophages into injured peripheral nerve were not affected. In summary, IL6 lead to a variety of effects on the cellular response to neural trauma. However, the particularly strong actions on lymphocytes and microglia suggest that this cytokine plays a central role in the initiation of immune surveillance in the injured central nervous system.

[1] 125w Peripheral nerve injury initiates a large number of morphological and molecular changes in the damaged nervous system that are particularly pronounced at three different sites: at the site of axotomy, in the cell body of injured neurons, and in the surrounding non-neuronal cells. In the injured peripheral nerve, the disconnected nerve ®bres and associated myelin distal to the lesion site undergo Wallerian degeneration, a process assisted by the in¯ux of haematogenous leucocytes, particularly the macrophages. In the proximal part, the tips of the cut axons transform into growth cones of sprouting axons; these reenter endoneural tubes in the distal nerve, grow back and reinnervate peripheral targets. The initially naked axonal shafts are also gradually remyelinated by the surrounding Schwann cells, restoring their normal physiological function.

[2] 145w At the neuronal cell body, the injured neurons display morphological and metabolic changes, characterized as chromatolysis, retrograde reaction or cell body response, reviewed in detail by Lieberman (1971), Grafstein & McQuarrie (1978) and Raivich et al., 1999a). On a molecular level, injured neurons show an increase in transcription factors (Herdegen et al., 1991(Herdegen et al., , 1998;;Haas et al., 1993;Yao et al., 1997;Schwaiger et al., 2000), growth-associated proteins (Benowitz et al., 1981;Willard & Skene, 1982), neuropeptides (Moore, 1989;Raivich et al., 1995), cytokines and neurotrophins (Klein et al., 1997;Gschwendtner et al., 1999;Murphy et al., 1999a;Streit et al., 2000) and cell adhesion molecules (Mo Èller et al., 1996;Jones et al., 1997Jones et al., , 2000)). These molecular changes, in particular the induction of cell adhesion molecules such as the laminin receptor a7b1 integrin, serve as an integral part of the neuronal regeneration program (Werner et al., 2000).

[3] 167w In addition to the periphery and the neuronal perikaryon, axonal injury will also lead to the activation of non-neuronal cells around the cell bodies of axotomized neurons. In the CNS, the microglial cells increase in adhesion molecules (Mo Èller et al., 1996;Werner et al., 1998;Kloss et al., 1999) and home onto and adhere to damaged neurons (Blinzinger & Kreutzberg, 1968;Kalla et al., 2000). They express major compatibility complex glycoproteins and costimulatory factors needed for antigen presentation (Streit et al., 1989;Bohatschek et al., 1999) and interact with T-lymphocytes recruited to the injured brain (Raivich et al., 1998a;Werner et al., 1998;Jones et al., 2000). Adjacent blood vessels show an increase in cell adhesion molecules which may assist leucocyte recruitment (Kloss et al., 1999). The neighbouring astrocytes show an induction of cytoskeletal molecules such as GFAP and vimentin (Bignami et al., 1974;Graeber & Kreutzberg, 1986;1988), rearrange their cytoskeleton into a stellar shape (Raivich et al., 1999a,b) and increase the physical stability of the damaged neural parenchyma (Pekny et al., 1999).

[4] 162w At present, the actual contribution of this central non-neuronal response to neural repair is surrounded by a high degree of controversy. The reactive astrocytes, lymphocytes and microglia produce a long list of neurotrophic molecules which could assist in nerve regeneration (Eddleston & Mucke, 1993;Mallat & Chamak, 1994;Norenberg et al., 1994;Kiefer et al., 1995;Ridet et al., 1997;Schwartz et al., 1999). The activated microglial cells express a number of potentially cytotoxic molecules such as tumour necrosis factor (TNF), nitrogen oxide (NO), oxygen radicals and components of the complement cascade which could impair neuronal survival (Banati et al., 1993;Giulian, 1999). Finally, glial scarring and the interaction of microglia and lymphocytes could also limit the spread of neural infection and enhance immune surveillance in the damaged brain parenchyma (Raivich et al., 1999a). The elucidation of key molecular signals in the initiation of this multifaceted cellular response and their inactivation is thus crucial for identifying the speci®c role of these non-neuronal cells for the overall neural repair.

[5] 205w In the current study we focused on the function of interleukin-6 (IL6). IL6 is a multifunctional neurotrophin and cytokine that is involved in neuronal survival (Hama et al., 1990;Kushima & Hatanaka, 1992;Toulmond et al., 1992), haematopoesis (Ikebuchi et al., 1987;Shabo et al., 1988), the immune response (Hirano et al., 1986;Lotz et al., 1988) and the systemic reaction to trauma, infection and septic shock (Gauldie et al., 1987;Kopf et al., 1994). IL6 is rapidly expressed after neural injury in both the peripheral and the central nervous system (Kiefer et al., 1993;Bolin et al., 1995;Murphy et al., 1995Murphy et al., , 1999a;;Reichert et al., 1996;Klein et al., 1997), and the transgenic deletion of IL6 inhibits astrocyte activation (Klein et al., 1997;Penkowa et al., 1999;Sugiura et al., 2000), reduces the survival of axotomized sensory neurons (Murphy et al., 1999a;Zhong et al., 1999) and interferes with the normal pattern of neuropeptide expression in injured neurons (Klein et al., 1997;Murphy et al., 1999b). In the current study we explored the effects this pleiotrophic cytokine on axonal regeneration and neuronal survival, on different stages of microglial activation and on the recruitment of lymphocytes, granulocytes and macrophages in the mouse facial axotomy model in homozygously IL6-de®cient mice and in their wild-type controls.

[1] 139w Twenty-micrometre-thick brainstem sections at the level of the facial nucleus were cut at ±15 °C, collected on gelatin-coated slides and stored at ±80 °C before use. All tissue was processed in exactly the same way with regards to perfusion, ®xation, cryoprotection and cutting protocols. Tissue sections belonging to different experiments were procesed at separate time points. However, all tissue sections belonging to the same experiment (e.g. macrophage staining at day 4 in the facial nerve shown in Fig. 1, or the lymphocyte CD3-staining at day 1 shown in Fig. 4, etc.) were stained together and for the same period of time to prevent differences in the number of counted cells or differences in staining intensity. For standard immunohistochemistry, the sections were thawed, spread in distilled water, ®xed in formalin, defatted in acetone and pretreated with 5% goat serum (Vector,

[2] 31w 0 1000 2000 0 5 10 10 20 0 10 20 0 A fmn(#) NG(#) cgrp galanin IL6-/-IL6+/+ 0mm 2mm 4mm 0mm 2mm 4mm * * AxRg(mm) ΜΦ(#) B D C

[3] 205w F IG. 1. Axonal regeneration (A) and cellular in®ltrates (B and C) in the injured facial nerve, 4 days after crush in the IL6 ±/± mice (empty bars) and the IL6 +/+ wild-type controls (®lled bars). Mean + SEM, n = 5 animals per group. *P < 0.005 in a 2-tailed unpaired Student's t-test. The extent of axonal regeneration was determined by the distance between the growth front of the regenerating CGRP or galanin-immunoreactive axons and the site of the nerve crush. In®ltrating neutrophil granulocytes (NG) were detected by histochemistry for endogenous peroxidase at the crush site and 2 and 4 mm distally. The macrophages (MF) were detected by the immunoreactivity for aMb2 integrin. The number of leucocytes is given per counting grid of 0.39 mm 2 (at 20Q magni®cation). (D) Neuronal cell number in the in the IL6 ±/± mice and the IL6 +/+ wild-type controls, 30 days after transection of the right facial nerve. Empty bars, axotomized side; ®lled bars, contralateral side. There was a tendency towards better neuronal survival in IL6 ±/± mice, compared to the IL6 +/+ wild-types (87% and 84%, respectively), but this was not statistically signi®cant (P = 0.42, n = 7 IL6 +/+ and 8 IL6 ±/± animals).

[4] 160w Wiesbaden, Germany) in phosphate buffer (PB) as described in Mo Èller et al. (1996). Brie¯y, the sections were incubated overnight at 4 °C with primary antibodies summarised in Table 1, washed in PB, incubated with a biotinylated goat antirabbit or antirat (1 : 100, Vector) or antihamster (1 : 100, Dianova, Hamburg, Germany) secondary antibody, followed by incubation with the ABC-reagent (Vector), visualization with diaminobenzidine/H 2 O 2 (DAB; Sigma, Deisenhofen, Germany), dehydration in alcohol and xylene and then mounted with Depex (BDH, Poole, England). Omission of the primary antibody or replacement with nonspeci®c immunoglobulin from the same species (rat, rabbit or hamster) at the same dilution led to the disappearance of speci®c labelling. The immunohistochemistry for macrophage colony-stimulating factor receptors (MCSFR) was performed on ¯oating sections, without defatting in acetone (Raivich et al., 1998a, b). The visualization of the CD3-immunoreactivity with diaminobenzidine/H 2 O 2 was performed with Co/Ni enhancement (0.02% CoCl 2 and 0.025% NiCl 2 in PBS).

[5] 21w Digital micrographs of the FITC and Cy3 ¯uorescence were taken using a Leica TCS 4D confocal laser microscope with a 10Q

[6] 181w objective for quanti®cation and a 100Q objective for illustrations in an 8-bit greyscale, 1024 Q 1024 pixel format as described in previous studies (Raivich et al., 1998a, b;Kloss et al., 1999). Ten consecutive equidistant levels with 30-mm spacing with a 10Q objective, or 15 levels with 16-mm spacing at 100Q objective, were recorded and condensed to a single bitmap using the MaxIntens algorithm. Auto¯uorescence was extracted by recording and subtracting the cyanin 5 (Cy5) bitmap (Raivich et al., 1998a, b). Activation markers on microglia in the wild-type IL6 +/+ and the cytokine-de®cient IL6 ±/± mice were quanti®ed using the modi®ed relative intensity of staining coef®cient (RISC) algorithm described by Werner et al. (1998). Microglial pro®les in the facial nucleus were ®rst delineated with the 5C6 immuno¯uorescence (FITC-image) or the rabbit anti-IBA-1 immuno¯uorescence (Cy3-image) using a mean + SD threshold. The pro®les were transferred to the bitmap of the counterpart ¯uorescence for the antibodies shown in Table 1 and Fig. 4 and the relative intensity of staining coef®cient for microglia in the facial nucleus (RISCmicr) was determined according to the following formula:

[7] 82w with OLVmicr the mean optical luminosity value (OLV) for the microglial pro®les in the facial nucleus, OLVfac the OLV for the whole facial nucleus and OLVout for the tissue outside the facial nucleus. This mean RISC was calculated for the axotomized and for the contralateral, unoperated, facial motor nucleus within each animal group (n = 3±7) and the statistical signi®cance between the IL6 +/+ and IL6 ±/± mice was tested at a 5% level (P < 0.05) using an unpaired Student's t-test.

[8] 157w Axonal regeneration in the crushed facial nerve was examined 96 h after injury as described in Werner et al. (2000). The axonal growth front of the regenerating CGRP-or galanin-positive motoneurites was detected in longitudinally cut, ®xed, 10-mm thin facial nerve sections using immunoreactivity for the axonally transported neuropeptides CGRP and galanin and distance to the site of the crush determined with a microscope grid. Five tissue sections, spaced 50 mm apart, were used to determine the mean regeneration distance for each neuropeptide and animal (n = 5 animals per group). Normal wildtype mice showed a regeneration distance of 6.54 T 0.21 for the CGRP-and 6.63 T 0.14 for the glananin-immunoreactive axons at day 4. Homozygous, IL6-de®cient mice showed a moderate but statistically signi®cant decrease (P < 0.03 for CGRP, P < 0.001 for galanin, Student's t-test), with a reduction by 14% or 0.93 mm for the galanin-, and 12% (0.77 mm) for the CGRP-immunoreactive neurites (Fig. 1A).

[9] 204w In addition to axonal regeneration, we also determined the number of macrophages and neutrophil granulocytes in the injured facial nerve 4 days after facial nerve crush at the lesion site and 2 and 4 mm distally, again using ®ve equidistant sections per staining and animal (Fig. 1B and C). The macrophages were detected using immunoreactivity for aMb2 integrin, the granulocytes using histochemistry for endogenous peroxidase. Unlike the regenerating motoneurites, there was no statistically signi®cant difference in the number of aMb2-positive macrophages or that of peroxidase-positive granulocytes between the wild-type and the IL6-de®cient mice 4 days after crush. day 4 0 3 6 0 3 6 * * day 1 0 1 2 3 diff 0 1 2 3 * * day 14 0 20 40 0 20 40 * * -/-+/+ -/-+/+ FIG. 4. Lymphocyte recruitment to the facial motor nucleus at days 1, 4 and 14 after transection of the facial nerve; effects of IL6 de®ciency. Left side, total number of CD3-positive T-cells per facial nucleus section: empty bars, contralateral; ®lled bars, axotomized nucleus. Right side, difference between the axotomized and contralateral sides. The IL6 ±/± mice show a 90% reduction in the difference for the number of CD3-positive T-cells at day 1,

[10] 36w an 80% decrease at day 4, and still a 60% decrease at day 14 (*P < 0.01, two-tailed Student's t-test). Note the gradual increase in the number of CD3-positive lymphocytes between day 1 and day 14.

[11] 44w Facial axotomy leads to a delayed form of neuronal cell death, with a peak at day 14 (Mo Èller et al., 1996;Raivich et al., 1998a). In the current study, neuronal survival was examined at 30 days using Nissl-stained 25-mm thin paraformaldehyde-®xed sections (Fig. 1D).

[12] 162w All neuronal cell pro®les were counted through the whole facial motor nucleus on the operated and contralateral sides and the number corrected for neuronal size using the Abercrombie correction as described in Materials and methods. This corrected number of motoneurons on the unoperated side in normal animals 1954 T 58 (n = 7, mean T SEM) is in line with previously published data on the facial nucleus (Sendtner et al., 1997). Compared to the unoperated side, facial axotomy led to a loss of 16.0% of motoneurons in the IL6 +/+ (1645 T 85 on the axotomized side) mice, and 13.0% in the IL6 ±/± mice 30 days after injury, with 1815 T 60 motoneurons vs. 2087 T 47 on the axotomized and contralateral sides, respectively (n = 9 animals). There was no statistically signi®cant change between the two groups of animals, with a P-value of 0.42 for the neuronal cell loss. The same was also true if the Abercrombie correction was omitted.

[13] 324w Transection of the facial nerve lead to the appearance of neurite growth cones in the axotomized facial nucleus, but particularly in the surrounding white matter, peaking at 14 days after injury (Fig. 2). These growth cones exhibited very high levels of immunoreactivity for the motoneuron markers CGRP, galanin and vesicular acetylcholine transporter, and became retrogradely labelled following application of Mini-Ruby applied on the proximal stump of the cut facial nerve, in line with their origin as central sprouts of axotomized facial motoneurons (Kloss et al., 1999;Werner et al., 1999;Werner et al., 2001). As shown in Fig. 2 (left column), these sprouts were absent on the contralateral side. Both groups of mice, IL6 +/+ and IL6 ±/± , displayed a similar moderate response for the CGRP-positive sprouts 14 days after facial axotomy (Fig. 2, 1st and 2nd rows). Quanti®cation of the area covered by the intensely CGRP-immunoreactive growth cones (Fig. 3, top left), as described in Materials and methods, revealed an area fraction of 0.70 + 0.10 parts per thousand (ppt) for the IL6 +/+ and 0.67 + 0.11 ppt for the IL6 ±/± group (mean + SEM, n = 7 animals per group, P = 0.86 in Student's t-test). The galanin-immunoreactive growth cones were already more numerous under normal conditions (3.54 + 0.43 ppt in IL6 +/+ mice), and this stronger response to axotomy was further enhanced by a factor of 1.8 (6.36 T 0.68 ppt) in the IL6 ±/± animals, with P < 0.001 (Fig. 3, top right; see also Fig. 2, 3rd and 4th rows). Quanti®cation of the galanin immunoreactivity in the labelled growth cones using the modi®ed RISC algorithm (Materials and methods) also showed a signi®cantly stronger neuropeptide staining in the IL6 ±/± mice (P < 0.001). CGRP staining was unchanged. Interestingly, no signi®cant difference was observed for the apparent volume of the galanin-immunoreactive or that of the CGRP-immunoreactive growth cones, with an apparent mean size of 100±150 mm 3 .

[14] 202w Facial axotomy leads to two different phases of T-lymphocyte entry into the injured facial motor nucleus. The ®rst is an elevated plateau, at days 1±4 after injury; the second, a much stronger increase with a peak at day 14 (Raivich et al., 1998a). In the current study, in®ltrating T-cells were detected using immunoreactivity for CD3, with seven equidistantly spaced 20-mm sections (spacing 120 mm) at days 1 and 4, and two sections (spacing 340 mm) at day 14. The higher number of sections used at the two early time points was due to the much lower number of T-cells (2.74 T 0.38 of CD3-positive cells per tissue section at day 1 and 5.82 T 0.42 at day 4, vs. 32.8 T 5.5 at day 14). As shown in Fig. 4, IL6 de®ciency led to a 10-fold reduction for the increase in the number of CD3-positive T-cells at day 1, a 5fold decrease at day 4, and still a 2.3-fold decrease in the difference between the axotomized and contralateral side at day 14 (P < 0.01 in a two-tailed unpaired Student's t-test, for all time points). The in®ltration of CD3-positive cells on the unoperated, contralateral side was not affected by the IL6 de®ciency.

[15] 174w The absence of IL6 in the homozygous IL6 ±/± mice leads to a series of microglial changes in the injured brain in terms of morphology, cell number and the expression of step-speci®c activation markers. These changes were particularly pronounced during early response to injury, 1±4 days after transection of the facial nerve. Activation of microglia in the axotomized facial motor nucleus normally proceeds through a series of steps (Raivich et al., 1999a): resting microglia (step 0), the state of alert (step 1), homing and adhesion (2), phagocytosis in the presence of neural cell debris (3a) and the bystander activation of the surrounding, nonphagocytotic, microglia (3b). As shown in previous studies, these activation states are de®ned by a high level of speci®c antigens on identi®ed microglial cell pro®les that can be quanti®ed using immunoreactivity for aMb2, mouse IgG or ionized Ca-binding adaptor protein type 1 (IBA1), and a double immuno¯uorescence for the appropriate activation marker (Raivich et al., 1998b;Werner et al., 1998;Kloss et al., 1999) using the RISC algorithm as described in Materials and methods.

[16] 98w Of all the markers tested on the resting microglia in the contralateral facial nucleus, only the immunoreactivity for thrombospondin (TSP, Fig. 5, top row; Fig. 6, top left) showed a signi®cant change, with a strong 2-fold increase in the IL6 ±/± animals compared with the IL6 +/+ wild-type controls (P < 0.001, unpaired two-tailed Student's t-test). None of the other markers (IBA1, intercellular adhesion molecule 1 (ICAM1), MCSFR, a5b1, a6b1, MHC1 and B7.2) were affected in a statistically signi®cant way, although there was a tendency towards higher microglial aMb2 levels in the IL6 ±/± group (P = 0.087).

[17] 140w Early microglial activation, or the state of alert, is normally characterized by the induction of TSP, aMb2 integrin, ICAM1 and IBA1 within the ®rst 24 h after injury (Fig. 5, 1st and 2nd row; Fig. 6, left column). Two of these markers, ICAM1 and IBA1, showed a signi®cantly lower level on microglial cell pro®les in the axotomized facial nucleus of IL6 ±/± animals 1 day after facial nerve transection (P < 0.001). The increase in microglial aMb2 immunoreactivity (difference, operated ± contralateral side) was signi®cantly lower in the cytokine-de®cient animals (P < 0.002), although this could be due to somewhat higher levels on the contralateral side in the IL6 ±/± group. Microglial TSP immunoreactivity on the operated side also showed a tendency toward higher levels in the IL6 ±/± animals, but this did not reach statistical signi®cance (P = 0.24).

[18] 430w After the initial activation, the stimulated microglia enter a rapid burst of proliferation, attach to injured neurons, show cell body swelling, and express high levels of the adhesion molecules a5b1 and a6b1 integrin (Fig. 5, 3rd and 4th row, respectively), and MCSFR (Fig. 6), peaking 3±4 days after axotomy (Raivich et al., 1998a, b;Kloss et al., 1999). Homing and adhesion to neurons is also accompanied by the beginning of an increase in microglial FIG. 5. IL6 de®ciency interferes with the normal pattern of early microglial activation in the axotomized facial motor nucleus: effects on adhesion molecules and MHC1. Double immuno¯uorescence for TSP, ICAM1 a5b1 and a6b1 integrins, and MHC1 (green), colocalization with aMb2 (TSP) or IBA1 (ICAM1, a5b1, a6b1, MHC1; red) 1 day (d1) and 4 days (d4) after transection of the facial nerve. In IL6 +/+ mice, facial axotomy caused a strong increase in the above-noted markers. IL6 de®ciency lead to an increase in TSP-immunoreactivity on resting microglia (IL6 ±/± co), and a decrease in ICAM1, a5b1, a6b1 and MHC1 markers on the activated microglial cells (IL6 ±/± ax), compared with the corresponding IL6 +/+ cells. Activated IL6 ±/± microglia (m) were capable of attaching to axotomized neurons (n). However, they failed to show cell body swelling and exhibited slender and ®nely arborized peripheral processes. Note the TSP, ICAM1 and a6b1 immunoreactivity on the neighbouring blood vessels (v). a5b1, a5b1; a6b1, a6b1. Scale bar, 25 mm. FIG. 6. IL6 de®ciency interferes with the induction of the receptor for the microglial mitogen MCSF and with microglial proliferation. Microglial response in the facial motor nucleus 3 days after nerve transection (ax) in the wild-type (IL6 +/+ ) and de®cient (IL6 ±/± ) mice. Comparison with the unoperated contralateral side (co). Double immuno¯uorescence for aMb2 integrin (green), a constitutively expressed microglial marker, and MCSF receptor (MCSFR, red). The colocalization of the two markers is shown in yellow (R + G, right column). IL6 ±/± mice showed a lower density of the aMb2-immunoreactive microglial cells in the 3-day axotomized facial motor nucleus and a much smaller increase in MCSFR immunoreactivity. Scale bar, 0.2 mm. immunoreactivity for the major histocompatibility complex type 1 (MHC1, Fig. 5, 5th row) and the costimulatory factor B7.2 (Bohatschek et al., 1999). Most of these microglial markers displayed a much smaller increase in the IL6 ±/± animals compared to the +/+ controls. As shown in Fig. 7, this increase was reduced by 48% for MCSFR (P < 0.0001), 55% for a5b1 integrin (P < 0.001), 73% for a6b1 integrin (P < 0.025) and 61% for MHC1 (P < 0.002).

[19] 157w This impaired induction of activation markers was accompanied by changes in cell number and morphology. The aMb2 staining shown in Fig. 6 demonstrated a reduced density of microglial cells in the facial motor nucleus 3 days after transection of the facial nerve of the IL6 ±/± mice, by a factor of » 1.5-fold, in line with the previously described impaired microglial proliferation in the IL6-de®cient animals (Klein et al., 1997). On the morphological level, activated microglia in normal IL6 +/+ mice reduced peripheral arborization and exhibited a strong swelling of the cell body as they attached to the axotomized motoneurons 4 days after transection of the facial nerve (Fig. 5, 3rd±5th row, 4th column). In the IL6 ±/± mice, activated microglia still attached to the axotomized neurons, beginning 1 day after facial axotomy (Fig. 5, top), but they did not show cell body swelling and displayed long, slender and arborized processes (Fig. 5, 3rd±5th row, 2nd column).

[20] 310w The delayed neuronal cell death, with a maximum 14 days after injury (Mo Èller et al., 1996), leads to a further transformation of neighbouring microglia into phagocytotic cells that form microglial nodules, with very high immunoreactivity for MHC1 and B7.2 (Bohatschek et al., 1998;Raivich et al., 1998a, b). As shown in Fig. 8, the normal appearance of these markers was not affected in the IL6 ±/± animals. In addition to the cellular changes in glial nodules, the phagocytosis of neurons is also accompanied by a second round of activation of neighbouring, nonphagocytotic, microglia which exhibit moderate levels of MHC1 and B7.2 and strong immunoreactivity for Quantitative immuno¯uorescence for microglial activation markers 1, 3, 4 and 14 days after transection of the facial nerve (d1±d14, respectively). Immunoreactivity for aMb2 and for polyclonal antibodies (TSP, MCSFR) was quanti®ed in aMb2-de®ned microglial pro®les; for IBA1 and for other monoclonal antibodies (ICAM1, a5b1, a6b1, MHC1, B7.2) in IBA1-de®ned pro®les using the RISC algorithm (see Materials and methods). Filled bars, axotomized side; empty bars, contralateral side. (*) A statistically signi®cant change on the same side (axotomized or contralateral) for the IL6 +/+ (wild-type) vs. the cytokine-de®cient IL6 ±/± mice (P < 0.05, Student's t-test, n = 4±8 animals per group). (s) A statistically signi®cant change in the difference between the axotomized and contralateral side between the IL6 +/+ and IL6 ±/± groups. Almost all the effects of IL6 de®ciency on microglial activation were observed at early time points (days 1±4). Microglial IBA1 was the only exception, with a reduced increase in IL6 ±/± at all tested time points, at day 1 (left column), day 14 (right column) and day 4 (not shown; see text). Note the strong effect of IL6 on microglial MHC1 at day 4 and its absence at day 14, at the late stage of microglial activation. aMb2, aMb2; a5b1, a5b1; a6b1, a6b1.

[21] 104w IBA1 (Bohatschek et al., 1998;Raivich et al., 1998aRaivich et al., ,b, 1999a,b),b). With » 130 cells in the IL6 ±/± and 200 cells in the IL6 +/+ animals per 20 mm facial motor nucleus tissue section at day 7±14 (Klein et al., 1997), these parenchymal cells are also more numerous than the microglial nodules (2±3 per section), and thus a major contribution to the overall microglial immunoreactivity. Unlike the midphase activation at day 4, the induction of microglial MHC1 in the whole microglial population in the axotomized facial motor nucleus 14 days after injury was not affected by IL6 de®ciency (Fig. 7, right column).

[22] 47w The same was true for B7.2. However, the absence of IL6 did lead to a signi®cant reduction in the post-traumatic increase in IBA1 at day 14 (Fig. 7, right 2nd row), as it did at day 1 (Fig. 7, bottom) and day 4 (±28%, P < 0.005).

[1] 655w Neural injury triggers numerous cellular changes at the site of injury and in the zone of anterograde degeneration, the retrogradely reacting neurons and in their non-neuronal environment. Although many of these changes are probably selectively mediated by a multitude of local factors that just elicit a speci®c reaction, there may be a need for a small set of signals that orchestrate the overall response to injury and the ensuing repair. The current data clearly show an important contribution of IL6 to the overall cellular response. Absence of IL6 in homozygous IL6 ±/± animals caused a massive decrease in lymphocyte recruitment, an impaired early activation of microglia, a reduction in axonal regeneration, and enhanced late perineuronal sprouting. These changes are summarised in Table 2. Previous studies also showed a signi®cant role of endogenous IL6 in astrocyte activation (Klein et al., 1997;Penkowa et al., 1999;Raivich et al., 1999a,b;Sugiura et al., 2000), the survival of sensory neurons (Murphy et al., 1999a;Zhong et al., 1999), and the normal pattern of neuropeptide expression in injured neurons (Klein et al., 1997;Murphy et al., 1999b). Moreover, IL6 is rapidly expressed after trauma in both the peripheral and central nervous system (Kiefer et al., 1993;Bolin et al., 1995;Klein et al., 1997;Murphy et al., 1999a;Legos et al., 2000), placing it in a strategic position to elicit all these cellular changes. Despite the variety of cells sensitive to, and responses induced or changed by, this cytokine, not all of them are affected to the same extent. As shown in the current study, the early phase of lymphocyte recruitment to the axotomized facial nucleus was almost completely abolished in the IL6-de®cient mice. This also applied to the microglial cells which failed to show key features of activation in their morphology, in the induction of MHC1 and in the appearance of most early and midphase activation markers such as ICAM1, IBA1, MCSFR, MHC1 and a5b1 and a6b1 integrins. Astrocytes were also strongly affected, leading to an 80% reduction in the appearance of stellar, GFAP-immunoreactive cells in the facial nucleus, 3 days after peripheral nerve injury in the IL6 ±/± mice (Klein et al., 1997). In contrast, the effects of IL6 de®ciency on neuronal response were mild even though most of them were statistically signi®cant. Thus, the extent of axonal regeneration at day 4 was reduced by 14% for the galanin-immunoreactive and 12% for the CGRP-immunoreactive motor neurites. Late perineuronal sprouting was enhanced for the galanin-positive but not for the CGRP-positive growth cones. Moreover, the overall effects on neuronal survival are at present equivocal. The current study shows a small tendency toward less cell death (±3%) in IL6 ±/± mice, though this was not signi®cant. In axotomized sensory neurons, neuronal cell death is enhanced in the IL6 ±/± animals by 13% in absolute terms (Murphy et al., 1999a). In forebrain cerebral ischemia, the cortical cell loss is not affected by a transgenic deletion of endogenous IL6 (Clark et al., 2000). This is in contrast to the potent protective effects of exogenous IL6 on postischemic forebrain neurons or axotomized motoneurons, applied in vivo (Ikeda et al., 1996;Loddick et al., 1998) or in vitro (Ali et al., 2000). In the same line, transgenic overexpression of IL6 and IL6 receptors will lead to a very substantial, » 1.5±2-fold reduction in the time needed for the reinnervation of peripheral target in the regenerating hypoglossal nerve model (Hirota et al., 1996). Overall, these differences underline the need to examine the IL6-de®cient animals to identify the physiological function of the endogenous cytokine. Here, the particularly strong effect of IL6 deletion on lymphocytes, microglia and astrocytes, and the massive reduction in microglial activation, astrocyte response and T-cell recruitment at days 1±4, all suggest that this cytokine plays a primary role in the initiation of immune surveillance in the injured central nervous system and in setting up mechanisms to prevent the spread of potential infection, rather than being a consistent and ubiquitous mediator of neural repair.

[1] 216w Homozygous, IL6-de®cient mice (IL6 ±/± ) on a C57Bl/6 background (Kopf et al., 1994) were obtained from BRL (Basel, Switzerland) and crossed with C57Bl/6 wild-type animals (Charles River, Hanover) to establish a colony of heterozygous IL6 +/± mice. The homozygous IL6 ±/± mice and the IL6 +/+ wild-type controls used in the current study were the F1 progeny of these heterozygous animals. For genomic screening, 1 cm from the tail of 2-week-old mice was incubated at 56 °C overnight in proteinase K lysis buffer (100 mg/ mL, Boehringer Mannheim). The lysate was treated with phenol± choloroform and the genomic DNA was precipitated out of solution with isopropanol, washed and redissolved in 200 mL of bidistilled water. PCR was used to detect the inserted neomycin cassette, with 0.5 mL DNA template, 1 Q PCR buffer (Eurobio, Raunheim, Germany), 2.0 mM MgCl2, 200 mM dNTPs (Pharmacia, Freiburg, Germany), 1.0 units taq polymerase (Eurobio) and 10 pmol of the sense IL6 primer 5¢-GCT AGC TAG ATA TCT CGA GAC AGG-3¢ and 10 pmol of the neomycin antisense primer 5¢-TCC CGC TTC AGT GAC AAC GTC-3¢. A second PCR was speci®c for the intact IL6 gene using the sense primer 5¢-GCT AGC TAG ATA TCT CGA GAC AGG-3¢ and the antisense primer 5¢-GGG AGT GGT ATC CTC TGT GAA GTC-3¢.

[2] 126w All operations were performed under anaesthesia with tribromethanol (Avertinâ, Sigma, Deisenhofen, Germany), 0.4 mg/g body weight, on 3±6-month-old mice. All animals belonging to the same experimental group (day 1 group, day 3 group, day 4 group, etc.) were operated on the same day, inside a narrow time window of 1± 3 h. Animal experiments and care protocols were approved by the Regierung von Oberbayern (AZ 211-2531-10/93 andAZ 211-2531-37/97). The right facial nerve was cut or crushed at the stylomastoid foramen, as described below; the animals were killed with ether after a survival time of 1±30 days, perfusion-®xed in 4% formaldehyde (FA) in phosphate-buffered saline (4% FA/PBS), the tissue removed, post®xed in 1% FA/PBS for 2 h, cryoprotected with 30% sucrose overnight and frozen on dry ice.

[3] 186w Double labelling experiments were performed to identify the cellular neighbourhood and the level of activation markers of microglial cells using a set of primary antibodies listed in Table 1. The microglial cell pro®les were identi®ed with the 5C6 rat monoclonal antibody against the aMb2 integrin in double labelling with rabbit polyclonal antibodies. Double labelling experiments with other rat monoclonal antibodies were performed with the rabbit polyclonal antibody directed against mouse ionized Ca-binding adaptor molecule 1 (IBA-1), kindly provided by Dr Yoshinori Imai (Neurochemistry Department, National Institute of Neuroscience, Tokyo), a highly selective microglial marker in normal and injured brain (Imai et al., 1996;Raivich et al., 1998a, b;Kloss et al., 1999). Fixed sections were preincubated as in bright-®eld immunohistochemistry. Both primary antibodies were applied overnight at 4 °C, washed, incubated with two secondary antibodies, biotin-conjugated donkey antirabbit Ig and FITC-conjugated goat antirat Ig (1 : 100; Dianova), washed again and then incubated with a tertiary FITC-conjugated donkey antigoat antibody (1 : 100; Sigma) and Cy3-Avidin (1 : 1000; Dianova). The sections were covered with VectaShield (Vector) and stored in the dark at 4 °C for further use.

[4] 89w Motoneuron cell counts were determined in the axotomized facial nucleus and on the contralateral side 60 days after a right facial nerve cut. Forty-®ve 20-mm-thick paraformaldehyde-®xed brainstem sections were cut in the caudorostral direction through the facial motor nucleus (anteroposterior length 640±700 mm) and 100 mm on the adjoining caudal and rostral sides. The tissue was stained with toluidine blue and all neuronal pro®les in every section of the facial nucleus were counted and corrected for the cell size using the Abercrombie correction (Abercrombie, 1946), according to the formula:

[5] 124w with N the corrected neuronal number, n the counted number of neurons, D the section thickness (25 mm) and d the mean neuronal diameter, calculated from the cell size. Neuronal cell size was determined from three tissue sections 100 mm apart, from a total of 210±370 neurons on the operated, and 160±380 neurons on the unoperated side. Digital images from the stained sections in a 24-bit RGB format were obtained using a Sony 3 CCD video camera (AVT-Horn, Aachen, Germany), neuronal cell pro®les were detected with Optimas 6.2 using a mean ± 1.5 Q standard deviation (Mean ± 1.5SD) threshold in the red channel, and the mean diameter (d) was calculated from the mean area (Area), assuming a near spherical form, with the formula:

[6] 93w Mean neuronal cell diameter (T SEM) on the unoperated and operated side was 18.8 T 0.4 and 19.2 T 0.3 mm, respectively, in IL6 +/+ animals, and 18.6 T 0.4 and 19.2 T 0.3 mm, in the IL6 ±/± mice (n = 7 and 9 animals, respectively; not signi®cant in a Student's t-test). The effect of Abercrombie correction on the relative neuronal cell count (operated/unoperated side) was a reduction by » 0.5% in the IL6 +/+ and 1.1% in the IL6 ±/± population, and this did not affect the level of statistical signi®cance.

[7] 206w Statistical analysis for axonal regeneration distance, neuronal cell counts, leucocyte in¯ux into injured nerve and microglial staining FIG. 2. CGRP-and galanin-immunoreactive axonal sprouts in and around facial motor nuclei 14 days after transection of the right facial nerve: effects of IL6 de®ciency. Facial axotomy lead to the appearance of intensely neuropeptide-immuno¯uorescent axonal sprouts (arrows) in the neighbourhood of axotomized neuronal cell bodies, with a maximum at day 14. Co, contralateral side; ax, axotomized side; ax, HM, axotomized side at a higher magni®cation (183Q). These growth cones (arrows) are particularly prominent in the white matter tracts ventral to the facial nucleus; they are smaller than the neighbouring motoneurons (arrowheads) but show a much higher level of immuno¯uorescence for CGRP (top two rows) and galanin (bottom two rows). IL6 de®ciency lead to a strong increase in the galanin-immunoreactive growth cones (compare IL6 ±/± and IL6 +/+ ). The less numerous CGRP-positive growth cones were not affected. Note the absence of neuronal galanin-immunoreactivity in the unoperated facial nuclei. Scale bar, 0.25 mm (left and middle columns), 75 mm (right column). intensity (RISC) was performed using a standard two-tailed Student's t-test. In the case of CD3-positive lymphocytes, all statistical analysis was performed on a derived semilogarithmic function, f, de®ned as:

[8] 76w with x the mean number of lymphocytes in the facial nucleus for each individual animal. The number of lymphocytes (in the brackets) was increased by 1 to avoid negative in®nite numbers for those nonaxotomized facial nuclei which did not contain lymphocytes. The function led to a strong reduction in the normally extensive variation (factor 2±3) for the lymphocyte number in axotomized facial nuclei, particularly at day 14, and was used for statistical comparison throughout the study.

[1] 234w The facial nerve was crushed with ®ne jeweller's forceps for 30 s 1 mm distal to the stylomastoid foramen and the animals killed after 96 h as described by Werner et al. (2000). Following a brief (5-min) perfusion with PBS, and 5 min with 4% FA/PBS, the facial nerve was ®xed by a slow 60-min perfusion with 1% FA/PBS, then immediately dissected for a length of 15±20 mm and frozen on dry ice. Nerves were cut longitudinally and the regenerating axons visualized by immunostaining for galanin or for calcitonin gene-related peptide (CGRP). Every ®fth section was used per antibody, with an interval of 50 mm, and the distance between the most distal labelled growth cone and the crush site measured using light-microscopic grid scaling. The average distance for each animal was calculated from four or ®ve tissue sections. In®ltrating neutrophil granulocytes were detected with histochemistry for endogenous peroxidase (EP), and the macrophages with immunohistochemistry for the aMb2 antibody (Werner et al., 2000;Bohatschek et al., 2001). Both cell types were counted at the crush site and 2 and 4 mm distally, and the numbers for each animal were also calculated from four or ®ve tissue sections, spaced 50 mm apart. The number of EP-positive granulocytes was counted in a 0.4-mm 2 frame; that of macrophages was determined by counting the aMpositive cells crossing a 0.5-mm-long vertical microscopic grid line, perpendicular to the nerve's longitudinal axis.

[2] 136w Axonal growth cones in and around the facial motor nuclei at day 14 (Kloss et al., 1999;Werner et al., 1999;Werner et al., 2001) were detected using immuno¯uorescence against CGRP or galanin, the neuropeptides expressed in axotomized facial motoneurons, followed by a Cy5-conjugated secondary donkey antirabbit Ig antibody (Dianova). Previous studies using retrograde labelling with Mini-Ruby con®rmed that these growth cones were derived from axotomized motoneurons (Werner et al., 1999). Quanti®cation of growth cones was performed on four sections per facial nucleus, with an interval of 200 mm between sections, as described by Werner et al. (2001). Brie¯y, the sections were scanned in a TCS 4D confocal laser microscope (Leica, Nussloch, Germany) with a 10Q objective using Cy5 settings (excitation wavelength, 647 nm; LP665; pinhole 30). Fourteen consecutive equidistant levels were recorded and condensed as described above.

[3] 91w Small strongly ¯uorescent growth cones were differentiated from the large neuronal cell bodies with the Sobel ®lter and a three-step algorithm in the Optimas 6.2 software. In the ®rst step, mean value of the overall luminosity (MEANcor) and the standard deviation (SDcor) of the corrected images (normal image) was recorded. This procedure was repeated following Sobel ®lter treatment (MEANsob, SDsob, Sobel image), which calculates the direction-independent local intensity gradient in a 3 Q 3 pixel kernel. The threshold for galanin-immuno¯uorescent growth cones in the Sobel image was set with the formula:

[4] 20w Areas at and above threshold were ®ltered with the `Object Classes' function using two additional criteria in the normal image.

[5] 55w (i) Area size > 10 pixel, (ii) MEANarea ± SDarea > 1.4 Q MEANcor with MEANarea the mean intensity and SDarea the standard deviation for each individual area pro®le. The remaining areas matched with the pro®les of the galanin-immuno¯uorescent neuronal growth cones and served as a measure for their total area in the tissue section.

[6] 53w The growth cone brightness (BRTgc) for neuropeptide immuno-¯uorescence was determined using a modi®ed RISC algorithm using the formula: BRTgc = log (OLVgc/OLVtotal) with OLVgc the mean optical luminosity value of the growth cones, and OLVtotal the mean luminosity value for the whole 1-mm 2 bitmap of the facial nucleus and the surrounding tissue.

[7] 15w The mean apparent volume of the growth cones (VOLgc) was also calculated with the formula:

[8] 27w with GC-AREAtotal the total growth cone area per section and #gc the number of growth cone pro®les, assuming a near spherical structure of the sprouting growth cones.

[9] 130w Activation of microglial cells, astrogliosis and leucocyte recruitment are key components of a graded non-neuronal reaction that develops in response to speci®c pathological stimuli in the nervous system: injured but still viable neurons, the appearance of cellular death and macromolecular debris, the disruption of the blood±brain barrier or the advent of immune cells and their activation in infection or in autoimmune disease. In the axotomized facial motor nucleus, this activation is a step-by-step process, with an early onset of changes in morphology and molecular activation markers in astrocytes and microglial cells and the appearance of parenchymal T-lymphocytes within the ®rst 24 h after nerve injury (Tetzlaff et al., 1988;Graeber et al., 1990;Raivich et al., 1991Raivich et al., , 1998a, b;, b;Werner et al., 1998;Kloss et al., 1999;Boucsein et al., 2000).

[10] 218w The recruitment of lymphocytes in the mouse CNS proceeds in two stages, a smaller plateau at days 1±4 and a major peak at day 14 (Raivich et al., 1998a, b). As shown for the ®rst time in this study, the different stages of T-cell recruitment are affected by IL6 de®ciency, but by a different degree. Absence of IL6 leads to a 10-fold decrease in lymphocyte in¯ux at day 1, but by < 2.5-fold by day 14. In contrast, deletion of IL1 receptor type 1 or TNF receptor type 2 (TNFR2) strongly interferes with lymphocyte recruitment at day 14 (Raivich et al., 1999b). Interestingly, only activated T-cells appear to enter the normal central nervous system (Wekerle et al., 1986;Hickey et al., 1991), and similar restriction may apply to the slight or indirect injury models (Raivich et al., 1998a(Raivich et al., , b, 1999a)). IL6 is a strong lymphocyte activation stimulus (Hirano et al., 1986;Lotz et al., 1988) and its absence in vivo, in the IL6 ±/± mice, could reduce the activation of circulating T-cells before they enter the brain. However, the difference in recruitment at different time points, and the absence of an effect on the in¯ux to the undamaged, contralateral side, all point to a predominantly central component for the IL6 action on lymphocyte entry into injured brain.

[11] 168w The neighbouring astrocytes also show a step-by-step response that begins with the induction of the glial ®brillary associated protein (GFAP) mRNA and protein within the ®rst 24 h after injury (Tetzlaff et al., 1988). At ®rst, this increase in GFAP immunoreactivity is diffuse. In the next stage, during cytoskeletal reorganisation, it becomes redistributed to the cell body and main branches of the reactive astrocyte within the next 1±2 days, leading to the appearance of stellar GFAP-immunoreactive pro®les. Absence of IL6 interferes with this process, leading to a strong reduction in the number of stellar pro®les (Klein et al., 1997), and the appearance of 30±60-mm large GFAP-positive diffusely immunoreactive blobs (Raivich et al., 1999a,b). Vascular integrity in the injured brain is compromised following transgenic deletion of GFAP and vimentin, another cytoskeletal component of reactive astrocytes (Pekny et al., 1999). Although hypothetical, the relatively rapid cytoskeletal reorganisation of GFAP could enhance the formation of stress ®bre cables and thus assist in enhancing the physical stability of the damaged neural parenchyma.

[12] 244w Microglial activation is also affected by IL6, but in a strictly stage-speci®c way. Thus, most early and mid-phase activation markers, such as IBA1, ICAM1, MCSF receptor, MHC1 and the a5b1 and a6b1 integrins, were all reduced in IL6-de®cient mice. IL6-de®cient animals also show a moderate decrease in microglial proliferation (Klein et al., 1997), associated with early and mid-phase activation, causing the reduction in microglial cell density shown in Fig. 6. In vitro, addition of IL6 also leads to a moderate stimulatory effect (+35%) on rami®ed microglia cultured on top of con¯uent astrocytes (Kloss et al., 1997) as well as in pure microglial cell cultures (Streit et al., 2000). The effect is stronger following addition of classical microglial mitogens MCSF, GMCSF and IL3, and the effect is most pronounced, a > 6-fold increase, in the case of MCSF (Kloss et al., 1997). In vivo, osteopetrosis, the naturally occurring complete de®ciency for MCSF leads to an almost complete inhibition of the microglial proliferation in vivo (Raivich et al., 1994;Berezovskaya et al., 1995). In this context, the > 50% decrease in microglial MCSF receptors in the IL6-de®cient mice shown in the current study can lead to an indirect inhibitory effect on microglial proliferation. The apparent absence of microglial IL6 receptors in vivo (Klein et al., 1997) and the strong expression of microglial receptors for MCSF shown in the current and previous studies (Raivich et al., 1998a, b;Kalla et al., 2000) could both support such an indirect pathway.

[13] 152w In contrast to the early effects, the late microglial activation at day 14 and the increase in microglial MHC1 and B7.2 were not affected by IL6 de®ciency. This is particularly striking in the case of MHC1. At day 4, absence of IL6 led to a > 60% reduction. At day 14, the > 2fold higher increase was no longer sensitive to this cytokine. This also applied to the formation of microglial nodules and their very strong expression of MHC1 and B7.2, suggesting the advent of additional proin¯ammatory stimuli. Thus, combined deletion of TNFR1 and 2 led to a strong decrease in MHC1 at day 14 (Bohatschek et al., 1998). A more moderate effect was also observed following single deletion of TNFR1 but not of TNFR2 (G. Raivich and M. Bohatschek, unpublished observations). Similar reduction in microglial activation markers is also observed in TNFR1-de®cient mice infected with T. gondii (Deckert-Schluter et al., 1998).

[14] 382w Overall, the current data suggest several quite separate phases of neuroglial activation that fall into two groups and differ in their cytokine requirements: the ®rst set mediated by injured but viable neurons, IL6 and MCSF; the second by a response to neuronal cell death, IL1 and TNFa, with IL6 playing a comparatively minor role. The recruitment of lymphocytes and the expression of the immuneassociated molecules by the activated microglial cells clearly suggest that they are components of a highly elaborate immune surveillance program. This program appears to change in a step-by-step fashion. During the ®rst, IL6-dependent, step, activated microglia search their environment for recruited T-cells, bound antibodies and opsonized antigens with their newly expressed receptors for immunoglobulin, complement and lymphocyte cell surface components. This is followed by adhesion of microglia to the cell body of the injured neurons, and the onset of MHC expression, a key component in antigen presentation. This apposition to and spreading on the neuronal surface places the microglia in a strategic position to take up soluble molecules and small particles emitted by the injured neurons, followed by intracellular processing and presentation to the activated T-cells entering the injured nervous system. The morphological changes in activated microglia, the cell body swelling and the proliferation of lysosomal organelles in the neuronal vicinity (Raivich et al., 1998b) may assist in this process. The astrogliotic response enhances the physical stability of the damaged tissue (Pekny et al., 1999); the inhibitory properties of glial scars may also block cellular migration and sprouting (Giftochristos & David, 1988;McKeon et al., 1991;Brodkey et al., 1995;Amberger et al., 1998;Fawcett & Asher, 1999;Menet et al., 2000;Moon et al., 2000), and with it, the potential spread of infectious pathogens via growing axons or non-neuronal cells. The lower density of axonal sprouts in the grey matter of the facial nucleus, compared with intranuclear white-matter patches, and the neighbouring white-matter tracts would clearly support the inhibitory effects of the reactive grey matter astrocytes. This prevention of axonal sprouting may be a rational strategy when facing an infectious challenge (Raivich et al., 1999a,b). However, sprouting axons are a prerequisite for successful repair in the injured brain, and the elucidation of signals such as IL6 that inhibit this response, and their neutralization, have the potential of improving neurological recovery in a sterile environment.

[15] 149w Altogether, the activation of microglia, astrogliosis and lymphocyte recruitment appear as constitutive elements of a well-orchestrated step-by-step cellular response, that de®nes the immune surveillance in the injured central nervous system and sets up mechanisms that may prevent the spread of potential infection. As shown in the current study, IL6 plays a key role in the initiation and the early steps of this process, before the advent of neuronal cell death. Although data on CNS infection are currently lacking, IL6 de®ciency clearly enhances early sensitivity to systemic infectious disease, with IL6 ±/± animals rapidly succumbing to mildly pathogenic bacteria which the wild-type animals are able to withstand (Kopf et al., 1994). The currently described defects in the early phase of the CNS response would make these animals particularly interesting in exploring the action of this cytokine, and the affected microglia, astrocytes and T-cells, in ®ghting different forms of neural infection.

[16] 111w In summary, IL6 de®ciency leads to a variety of effects on the cellular response to neural trauma. Absence of IL6 is associated with a moderate reduction in the speed of axonal regeneration, decreased astroglial and microglial response, a reduction in lymphocyte recruitment and enhanced late perineuronal sprouting in the case of galanin-immunoreactive neurites. However, these cellular responses are all affected to a different degree, and the particularly strong action on lymphocytes and microglia, the massive reduction in microglial activation and in T-cell recruitment at day 1±4 in the IL6-de®cient mice, clearly suggest that this cytokine plays a central role in the initiation of immune surveillance in the injured central nervous system.