Supplementary Material

Transcription

1 Supplementary Material Supplementary Method Sample preparation and image data acquisition Osteoclast - bone data Femoral bones were excised, fixed with 4%(v/v) paraformaldehyde buffered with 0.1 M phosphate buffer, ph 7.2, containing 4%(w/v) sucrose, and decalcified with a 10%(w/v) EDTA solution for 4 days. Samples were then embedded in OCT compound after cryoprotection with 15%(w/v) and 30%(w/v) sucrose solutions, cut into 10 m sections with a cryostat (Leica, CM3050, Nussloch, Germany), and placed on silanecoated slides. In order to identify osteoclasts, fluorescence-based TRAP staining was performed with some modification 1. Briefly, sections were incubated for 15 min at 37 C with ELF97 substrate (20uM; Invitrogen-Molecular Probes, Carlsbad, CA) using the TRAP staining buffer (104 mm acetate buffer ph 5.2, 28 mm L(+)-tartrate buffer, ph 5.2). Collagen-enriched bone matrices could be visualized with 2nd harmonic emissions from collagen fibers excited by infrared lasers. Sections were mounted with an anti-fade solution and examined by a two photon laser microscope, LSM510 NLO Meta (Carl Zeiss, Jena, Germany) driven by a Chameleon femtosecond pulsed laser (Coherent, Santa Clara, CA) tuned to 780 nm. Fluorescence was detected through bandpass emission filters at nm (for 2nd harmonic emission from bone matrices) or 525/50 nm (for TRAP staining). B-T-cell interaction data In order to analyze antigen-specific interactions between T and B lymphocytes, adoptive transfer of B cells expressing the MD4 transgenic antigen receptor specific for hen egg lysozyme (HEL, Sigma) and T cells expressing the OT-2 transgenic antigen receptor specific for and ovalbumin (OVA, Sigma) peptide presented via I-A b were used. The B and T cells express genetically encoded CFP and GFP, respectively, to facilitate tracking cells following extensive cell proliferation in vivo. To co-activate the two types of cells in vivo, a cross-linked conjugate of HEL and OVA (HEL-OVA) was

2 subcutaneously injected into recipient mice together with alum and bacterial lipopolysaccharide. At desired time points post immunization, draining inguinal LNs were surgically exposed for intravital observation as previously described 2. The imaging system was a Bio-Rad/Zeiss Radiance 2100MP microscope driven by a Chameleon laser (Coherent) tuned to nm, a Nikon 600FN upright microscope equipped with a 20 water immersion lens (NA 0.95, Olympus), and LaserSharp acquisition control software. We analyzed two independent pairs of data sets: Pair [#1(w.t.), #2(k.o)] consists of 239 time steps each involving 220x220x18 3-D data sets and pair [#3(w.t.), #4(k.o.)], 120 time steps each with 512x512x15 data sets with varying numbers of lymphocytes. Dendritic cell - fiber network data LNs from a transgenic CX3CR1 GFP/+ mouse 3 were incubated for 12 hrs in 0.05 M phosphate buffer containing 0.1 M L-lysine (ph 7.4), 2 mg/ml NaIO 4, and 10 mg/ml paraformaldehyde (PLP), washed in phosphate buffer, and dehydrated in 30% sucrose in phosphate buffer. Tissues were snap frozen in Tissue-Tek (Sakura Finetek). 30 μm frozen sections were cut and then stained with an anti-ertr-7 antibody (Acris Antibodies- Hiddenhausen, Germany) as previously described 4. Immunofluorescence confocal microscopy was performed with a Leica TCS SP5 confocal microscope using a 63x objective. Separate images were collected for each fluorochrome and overlaid to obtain a multicolor image.

3 Supplementary Figure 1: Illustration of the typical bimodal shape of histograms of normal (e. g. sun-light) images. Separation of signal (here: Lincoln Memorial) and background (here: sky) in such images (here: inverted real image, blue channel) can be achieved by finding the minimum between the two peaks representing background and signal. It is obvious that this criterion is also physically plausible.

4 Supplementary Figure 2: Overlay of blue (bone) channel from original image with border of segmentation result (shown in red, compare Figure 4A in manuscript).

5 Supplementary Figure 3: Schematic illustration of the adaptive merging algorithm. A, B: Original image data, separate channels (inset: intensity profiles along cell diameter). When channels are merged the decision has to be made which pixels in the overlap region are assigned to which object. The selection criterion we use compares the intensity values of each overlap pixel in the original image data. To get a realistic estimate of the real interface between adjacent objects, we normalize the intensity values of each channel (as shown in D). Using absolute intensity values may lead to less accurate interface reconstruction (C). See also Figure 2.

6 Supplementary Figure 4: Performance assessment under different image noise conditions. Top row: Analysis result; mid row: bone channel; bottom row: osteoclast channel. Columns contain data/results for different degrees of Poissonian noise added to the images to simulate different images qualities using the Gimp noise generator plugin 5 ( Column A: Original image; column B: low Poissonian noise (Gimp Poisson noise generator parameter setting: 3 photons, 1 iteration); column C: medium noise (parameter setting 2 photons, 2 iterations); column D: high noise (highest Poissonian noise selection with Gimp noise generator: parameter 1 photon, 3 iterations). The variation of the computed interface area as a function of Poissonian noise using the Gimp noise generator is low: The normalized interface area (ratio interface area / total bone surface) has a standard deviation of only ~5% ( = ). Gimp and the noise generator are available at and respectively.

7 Supplementary Figure 5: Illustration of the performance of the threshold segmentation if intensity distributions of signal and background are overlapping (in this case only for a certain image region). In the example here, bone tissue in the lower left part of the bone tissue in subfigure B, cannot be distinguished from background by the fluorescence intensity in the corresponding regions. A) Blue channel (blue: 0, red/brown: max. intensity); B) Original image; C) Analysis result. Right: Intensity histograms a, b, c, d corresponding to the selected regions in A. While selections b, c, d all show typical background (in the blue/bone channel) histogram signatures (right) as opposed to the true signal (d), only c is clearly background and region b is very likely bone tissue (blue). Region d could be either bone or background based on visual inspection. Even though assumptions about shape and osteoclast borders could in principle be used to improve the segmentation result, the assignment of region d to bone tissue would remain speculative.

8 1. Filgueira, L. Fluorescence-based staining for tartrate-resistant acidic phosphatase (TRAP) in osteoclasts combined with other fluorescent dyes and protocols. J Histochem Cytochem 52, (2004). 2. Qi, H., Egen, J.G., Huang, A.Y. & Germain, R.N. Extrafollicular activation of lymph node B cells by antigen-bearing dendritic cells. Science (New York, N.Y 312, (2006). 3. Jung, S. et al. Analysis of fractalkine receptor CX(3)CR1 function by targeted deletion and green fluorescent protein reporter gene insertion. Molecular and cellular biology 20, (2000). 4. Bajenoff, M., Granjeaud, S. & Guerder, S. The strategy of T cell antigenpresenting cell encounter in antigen-draining lymph nodes revealed by imaging of initial T cell activation. The Journal of experimental medicine 198, (2003). 5. Kervrann, C. & Trubuli, A. An adaptive window approach for Poisson noise reduction and structure preserving in confocal microscopy. IEEE International Symposium on Biomedical Imaging: Nano to Macro, , (2004).