C57BL/6 mice and C57BL/6 pups were purchased from Charles River (Germany). β-actin-EGFP transgenic C57BL/6 (003291, C57BL/6-Tg(ACTB-EGFP)1Osb/J), OT-2 (C57BL/6-Tg(TcraTcrb)425Cbn) transgenic mice and RAG1 -/- transgenic C57Bl/6 (002216, B6.129S7-Rag1tm1Mom/J) were originally obtained from The Jackson Laboratory (Maine, USA), bred under SPF conditions at the central animal facility of the Charité - University Medicine Berlin (FEM), and kept in-house for experiments in IVC (individually ventilated cages).

C57BL/6 OT2 mice, C57BL/6 2d2 TCR 14 transgenic mice (kindly provided by A. Waismann), C57BL/6 Rosa26 tdRFP ("ΔNeo-flip")mice 15, obtained from H.J. Fehling, or C57BL/6 β-actin-EGFP transgenic mice were intercrossed to generate C57BL/6 OT2 tdRFP, C57BL/6 2d2 tdRFP mice and C57BL/6 OT2 EGFP mice, respectively. C57BL/6 Thy1-21 EGFP 16 mice were kindly provided by P. Caroni, C57BL/6 Thy1.1 CerTN L15 mice were kindly provided by O.Griesbeck 17 C57BL/6 Foxp3 EGFP mice were kindly provided by B. and M. Malissen.

All animal experiments were conducted according to the German Animal Protection Law, were approved by the appropriate state committees for animal welfare (LAGeSo, Landesamt für Gesundheit und Soziales) and were performed in accordance with current guidelines and regulations.

Mice (6-10 weeks old) were sacrificed and spleen cells were isolated as described before 3. For OT-2 Th2 cells splenocytes were cultured in 3 × 10⁶/ml cell culture medium (RPMI 1640 supplemented with 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, and 10% fetal calf serum), and stimulated with the respective ovalbumin (OVA) peptide (0.3 μM OVA_323-339; Pepceuticals, UK). Differentiation towards a Th2 phenotype of OT-2 cells was achieved by the addition of 200U/ml IL-4, 5 μg/ml anti-IL-12 (C17.8) and anti-IFN-γ (AN18.17.24). C57BL/6 Th1 cells were generated from CD4+ sorted spleen and lymph node cells (MACS^®) of C57BL/6 mice by polyclonal activation by 1 μg/ml plate-bound antiCD3/antiCD28 in the presence of 5 ng/ml IL-12 and 5 μg/ml anti-IL-4 (11B11). Antigen specific and polyclonal effector T cells were restimulated every 7 days by irradiated syngenic CD90-depleted (MACS^®) APCs or antiCD3/antiCD28, respectively. C57BL/6 Foxp3 expressing regulatory T cells were isolated either by MACS^®sort for CD4, CD25 and CD62L or by FACS for Foxp3EGFP and CD62L. Cells were polyclonally activated by plate-bound antiCD3/antiCD28 (3/10) in round bottom 96 well plates in the presence of 2000 U IL-2. IL-10 producing Ovalbumin- or MOG-specific regulatory T cells were generated according to the protocol by Barrat et al.18. Briefly, naïve CD4+CD62L+T cells were stimulated with irradiated syngenic CD90depleted splenic APCs (MACS^®) and 0.3 μM OVA_323-339 or 12,5 μg/ml MOG_35-55 peptide in the presence of 40 nM Vitamin D3 and 100 nM Dexamethasone and 5 μg/ml anti-IL-12 (C17.8), anti-IFN-γ (AN18.17.24) and anti-IL 4 (11B11). Restimulation was done on day 7 with anti-CD3/anti-CD28. All analysed T cell subsets were stimulated for 3-7 days prior to hippocampal brain slice co-culture experiments. For characterization of T cell phenotypes intracellular staining was performed on short term stimulated (4 h antiCD3/antiCD28) T cells on day 4-7 in order to analyse cytokine and Foxp3 expression by flow cytometry.

For T cells of naïve phenotype spleen and lymph node cells of OT2, OT2 EGFP or OT2 tdRFP mice were MACS^®-sorted for CD4 and CD62L and used for i.v. injection for lymph node and intravital imaging of the brain stem or for application onto hippocampal brain slices.

For adoptive transfer EAE, spleen cells from C57BL/6 2d2.tdRFP or C57BL/6 2d2.EGFP mice were isolated and stimulated as previously described 19. Three days after the second restimulation, 5 × 10⁶ 2d2.tdRF TH17 or 2d2.EGFP TH17 cells were transferred intravenously into C57BL/6 RAG1^-/-. Cytokine expression was checked on restimulation and before transfer. Mice were checked for clinical symptoms daily and signs of conventional EAE were translated into clinical score as follows: 0, no detectable signs of EAE; 0.5, tail weakness; 1, complete tail paralysis; 2, partial hind limb paralysis; 2.5, unilateral complete hind limb paralysis; 3, complete bilateral hind limb paralysis; 3.5, complete hind limb paralysis and partial forelimb paralysis; 4, total paralysis of forelimbs and hind limbs (mice with a score above 3.5 to be killed); 5, death. The severity of atypical EAE was translated as: grade 1, tail paralysis, hunched appearance; grade 2, ataxia, scruffy coat; grade 3, head tilt, hypersensitivity, spasticity or knuckling; grade 4, severe proprioception defects; grade 5, moribund.

For intravital imaging of naïve T cells MACS^®-sorted naïve OT2 EGFP or OT2 tdRFP cells were co-transferred intravenously 12-24 hours before TPLSM or were locally applied on the imaging field at the peak of disease (score 2.5-3.5).

T cells and vessels were visualised by a two-photon laser scanning system (LaVision BioTec, Bielefeld) as described before 20 with dual NIR (850 nm) and IR (1,110 nm) excitation. XYZ-stacks were typically collected within a scan field of 300 × 300 μm at 512 × 512 pixel resolution and a z-plane distance of 2 μm at a frequency of 400 or 800 Hz. Applied laser powers ranged from 2-6 mW at the specimen's surface.

Preparation of brain slices was performed as previously described 3 and allowed to recover for at least 45 min at room temperature prior to transfer to a heated and with aerated artificial cerebrospinal fluid (ACSF) perfused Luigs & Neumann slice chamber (37°C). T cells, if not expressing EGFP or tdRFP, were stained with celltracker Orange CMTMR (Invitrogen, Germany), pipetted upon the slice and allowed to invade the slice for about 30 - 60 min before image acquisition.

Mice were anaesthetized using 1.5% isoflurane (Abott) in oxygen/nitrous oxide (1:2) with a facemask. They were then tracheotomized and continuously respirated with a Harvard Apparatus Advanced Safety Respirator (Hugo Sachs, Germany). For the brainstem imaging, the anaesthetized animal was transferred to a custom-built microscopy table and fixed in a hanging position. The preparation of the imaging field was performed as previously described 3.

Lymph node preparation was performed as previously described 20. In brief, 8-96 hours after intravenous injection of naïve OT2 EGFP or OT2 tdRFP T cells, popliteal lymph nodes were removed, glued onto a coverslip and placed into a heated chamber at 36°C, superfused RPMI medium without phenol red bubbled with 95%O2/5%CO2. Naïve T cell migration was visualised up to 130 μm beneath the surface using the same TPLSM setup as described above.

Cell recognition, movement tracking and 3D presentation were performed with Volocity (Improvision, Germany) and tracks with durations > 5 min. were included in the analysis. Average cell velocity was calculated with Volocity. The co-localisation area was calculated with Volocity as previously described 19. To discriminate short/transient (random) contacts from long-lasting (most probably non-random) interactions, we defined contacts which last longer than 5 min as 'long-lasting'. These interactions were further differentiated into slow/static (velocity < 0.04 μm/s) and dynamic (velocity > 0.04 μm/s) contacts. In order to exclude the SHG signal which is detected with the same detector as EGFP in our setup, we increased the signal threshold so that only EGFP expressing cells (without any SHG signal) were automatically identified and tracked by the software. For angle comparison between cell trajectories and vessel structures and for vessel distance measurements at the end of the track, cell-tracking data from Volocity was further processed in MATLAB as described previously 3. For each track, a vector between origin and endpoint was calculated in a two-dimensional angle and the distance between the cell-tracking vector and the vector representing the vessel was calculated using standard linear algebra. Statistical analysis was carried out with SPSS (SPSS, Germany) and graphical presentation with GraphPad Prism 4 (Graphpad Software, USA). Results are shown as individual data points; in addition, mean ± SD summarize collective data from performed experiments. To test for statistical significance of differences between two T cell subpopulations, the non-parametric Mann-Whitney-test was performed for more than two groups Kruskal Wallis test with Dunn's multiple comparison test as post test was performed.

To observe naïve T cell migration under inflammatory circumstances in the CNS, we first performed flow cytometry analysis on brain-derived immune cells after transfer of naïve T cells, which are characterized by high CD62L and low CD69, CD25 and CD44 expression (Figure 1A), into EAE affected Rag 1-/- mice. For disease induction we transferred in vitro generated encephalitogenic 2d2 CD4+ TH17 cells (with transgenic T cell receptor specific for Myelin Oligodendrocyte Glycoprotein (MOG)). Some of the injected naïve T cells could be detected in the CNS but the cell number was too low to analyse activation markers and CD62L expression (data not shown). Therefore we used intravital TPLSM in the brainstem to record their behaviour. Using this approach it was possible to find naïve T cells up to 150 μm deep in the CNS parenchyma in EAE and to track their migration (Figure 1B). We used T cells with a transgenic T cell receptor specific for Ovalbumin (OVA) to limit antigen specific activation on their way to and within the CNS. To further minimize activation, our timepoint of investigation was shortly after transfer (12 h for IV and 30 min for local application). Whereas we previously observed no lymphocytes in healthy mice deep in the CNS parenchyma we could detect some sporadic cells in the perivascular space 3. However, during EAE naïve T cells were able to invade the parenchyma after local application on the brainstem or to cross the corrupted blood brain barrier after IV transfer and to move rapidly (Figure 1B, Additional File 1) resulting in a mean track velocity of 0.13 ± 0.05 μm/sec (± SD; N = 212) which was similar to that of encephalitogenic 2d2 CD4+ TH17 effector cells with 0.12 ± 0.06 μm/sec (±SD; N = 87) (Figure 1C). The dynamic interaction between both cell types was examined by co-localisation analysis as previously described 19. We discriminated between short (random) contacts and long-lasting (most probably non-random). On their way through the parenchyma 19% of all contacts were of long-lasting nature whereas the majority (81 ± 14%) of the naïve-effector T cell interactions was short and random (Figure 1D/E).

As it was not possible to investigate the behaviour of naïve CD4+ T cells in healthy living anaesthetized mice due to the low frequency of lymphocytes after IV transfer 21, we circumvented the blood brain barrier and applied naive CD4 T cells on the brain stem of healthy living anaesthetized mice. The cells remained stucked on the surface or were flushed away with the perfusion (Additional File 2). We further co-incubated naïve T cells with brain tissue cultured in artificial cerebrospinal fluid (ACSF). These CNS slices have been shown to retain a middle layer which consists of intact in vivo-like CNS in murine brain cultures 22. Naïve CD4+ T cells were labeled with the Celltracker Orange 5-(and-6)-(4-chloromethyl(benzoyl)amino) tetramethylrhodamine (CMTMR) if they were not transgenic for tdRFP or EGFP, and consecutively co-incubated with a living brain slice in which the vasculature had been highlighted by previous IV injection of FITC- or rhodamine-dextran. Independent of the need to transmigrate through the blood brain barrier, these cells showed a nearly stationary behaviour. The cell tracks were short and the cells seemed more or less trapped in their position on the slice surface after application on the non-inflamed CNS tissue (Figure 2A). This was quantified by a significantly lower mean track velocity with 0.056 ± 0.03 μm/sec (± SD; N = 86) as compared to activated CD4+ T cells (**p < 0.01). Furthermore the significantly reduced displacement rate of 0.019 ± 0.015 μm/sec (± SD) shows that they move on their place and do not invade the parenchyma (***p < 0.01) (Figure 2B).

The stationary behaviour of naïve CD4+ T cells, as described above, is clearly a contrast to the migration pattern of activated CD4+ TH1 and CD4+ TH2 effector cells. We previously showed that CD4+ antigen specific, highly-activated and differentiated effector T cell lines show a vessel-associated movement pattern 3. We wanted to check whether antigen specificity or regulatory phenotype would influence T cell migration. Therefore we performed T cell brain slice co-culture experiments as described above exemplarily for activated antigen specific (OT2 or 2d2) IL-10-producing regulatory and antigen specific (OT2) effector CD4+ TH2 cells as well as for polyclonally-activated Foxp3 T regulatory and CD4+ TH1 effector cells (Figure 3A). After an initial invasion period, the potential of tissue penetration was evident for all analysed activated T cell subsets as visualised by time lapse TPLSM. Furthermore, the migration pattern was similar in all investigated CD4+ T cell subtypes. The mean angle between vessel and cell trajectory vectors was similar in all investigated subgroups (Figure 3B, Figure 3C). The vessel-associated movement of CD4+ T cells in contrast to CD8+ T cells 3 seems also to be important for regulatory T cell subtypes independent of antigen specificity and underlines the importance of the perivascular space for immunoregulation and autoimmunity. Activated CD4+ T cells of regulatory phenotypes moved through non-inflamed CNS tissue with 0.07 ± 0.04 μm/s (± SD) for antigen-specific IL-10 producing regulatory T cells (N = 290) and 0.08 ± 0.04 μm/s (± SD) for C57Bl/6 Foxp3 CD4+ T cells (N = 215). Activated effector T cell subsets also revealed comparable mean track velocities, irrespective of antigen specificity and phenotype with 0.07 ± 0.0 μm/s (± SD) for OT2 Th2 (N = 290) and 0.07 ± 0.04 μm/s (± SD) for C57BL/6 Th1 (N = 248) cells (Figure 2B).

As control for the principle migratory capacity of naïve CD4+ T cells we injected MACS^® sorted naïve OT2 cells intravenously into healthy mice and tracked their migration pattern in the lymph node (Figure 4). To check whether the cells were activated and primed after transfer we performed FACS analysis on injected cells and observed that these cells retain their CD62L^high and CD69^low phenotype up to 4 days after transfer (Figure 4A). We performed our experiments within this time frame. Time lapse TPLSM revealed high motility for naïve T cells (Figure 4B). Cell tracking analysis of individual T cells confirmed our qualitative observations, as the mean track velocity with 0.17 ± 0.06 μm/sec (± SD; N = 126) was similar to what was previously described by others 2 (Figure 4C).

In lymphoid tissue it is known that naïve T cell motility is directed by fibroblastic reticular cells and ECM structures which can be visualised by second harmonic generation (SHG) 23. Interestingly, we observed similar SHG signals which were generated by an optical parametric oscillator (OPO) at 1110 nm wavelength in the inflamed CNS of EAE affected mice. Moreover, naïve T cells partially moved along these structures (Figure 5A, Additional File 3). In order to objectify our observation we performed quantitative analysis for naïve T cell - SHG co-localization (Figure 5B/C). This analysis revealed that most of the contacts are short (81 ± 14%) but there is also a population of naive T cells that show long-lasting contacts with the reticular fiber network (19 ± 14%). These interactions were further differentiated into slow/static (velocity < 0.04 μm/s) and dynamic (velocity > 0.04 μm/s) contacts. Interestingly, these long-lasting contacts were due to migration associated co-localisation, since their velocity was > 0.04 μm/s supporting our qualitative observation that a distinct population of naïve T cells partially migrate along these inflammation induced fibers (Figure 5B/C).

The reticular fiber network was observed deep in the parenchyma (up 100 μm) and thus could be distinguished from the arachnoidea covering the surface of the brain stem parenchyma. Moreover, by visualisation of the vasculature by IV Rhodamine-dextrane injection prior to imaging we observed that in addition to the vessel associated collagen network surrounding the very close proximity of the vasculature further fiber tracts are generated deep in the parenchyma itself (Figure 5D).

We could not detect these structures under healthy conditions in the CNS parenchyma (Additional File 2). However, intravital TPLSM on the brain stem of EAE affected C57BL/6 Thy1.1 CerTN L15 mice which express a FRET based calcium sensor, i.e. the Citrulin/Cerulean FRET pair (based on YFP/CFP) in neurons and neuronal processes (axons and dendrites) 17, we could distinguish the generated SHG signal due to wavelength specificity of SHG (1110 nm) and separated excitation of Citrulin (850 nm). Using this approach we did not observe any co-localisation with the neuronal structures. In contrast, they appeared to cross the neuronal processes orthogonally, as visualised by merging fluorescence images acquired at both excitation wavelengths (Figure 5E).