doi:10.1038/nature11046 Supplementary Figure 1: Images of PB-positive cells in the subepidermal region (a-i) Representative images of PB positive cells in the subepidermis of the upper beak of the pigeon. Note the presence of multiple dark blue spherules (0.25 µm to 5 µm in diameter) and/or light blue cytoplasmic staining. The shape of PB positive cells vary, some are rotund (d) while others are elongated (a).the oval shaped structure visible in the centre of some cells (b-e, i) is consistent with a nucleus, not a magnetosensitive vesicle. (jl) Sections stained with both NFR and PB highlight that these cells are nucleated. The scale bar indicates 10 µm. WWW.NATURE.COM/NATURE 1
Supplementary Figure 2: Images of PB-positive cells in the feather follicle (a-i) Representative images of PB positive cells in the feather follicle of the upper beak of the pigeon. Note the presence of multiple dark blue spherules (0.25 µm to 5 µm in diameter) and/or light blue cytoplasmic staining. Most PB positive cells in the feather follicle are round. The oval shaped structure visible in the centre of some cells (a-c, f) is consistent with a nucleus. (j-l) Sections stained with both NFR and PB highlight that these cells are nucleated. The scale bar indicates 10 µm. WWW.NATURE.COM/NATURE 2
Supplementary Figure 3: Images of PB-positive cells in the respiratory epithelium (a-i) Representative images of PB positive cells in the respiratory epithelium of the upper beak of the pigeon. Note the presence of multiple dark blue spherules (0.25 µm to 5 µm in diameter) and/or light blue cytoplasmic staining. The shape of PB positive cells vary, some are rotund (f) while others are elongated (b,e).the oval shaped structure visible in the centre of some cells (b-c,e,g) is consistent with a nucleus. (j-l) Sections stained with both NFR and PB highlight that these cells are nucleated. The scale bar indicates 10 µm. WWW.NATURE.COM/NATURE 3
Supplementary Figure 4: MRI and CT images (a-c) Sagittal, coronal and axial magnetic resonance images (MRI) from a 3D data set of the head of pigeon 209. This data set was used to generate Supplementary Movie 1. (d,g) Representative MRI images from a 3D data set which was used to generate a reference atlas for histological classification (sections 360 and 190 respectively). (e,h) Micro computed tomography (CT) images from the same beak. (f,i) Overlay of co-registered MRI and CT data sets. Scale bars indicate 5 mm. WWW.NATURE.COM/NATURE 4
Supplementary Figure 5: Anatomical landmarks of the pigeon beak (a) Diagram showing the pigeon beak from a lateral view, highlighting the position of the four anatomical landmarks used to normalise the distribution of PB positive cells along the rostro-caudal axis. (b-e) Coronal views of the four landmarks. Landmark 1 (b) was defined as the location where the shape of the concha morphs from being heptagonal to pentagonal. Landmark 2 (b) was defined as the location where the heart-shaped respiratory concha first makes medial to lateral contacts. Landmark 3 (d) was defined as the location where the small lateral buds of the nasal cavity disappear. Landmark 4 (e) was defined as the position where the nasal cavity is no longer visible. WWW.NATURE.COM/NATURE 5
Supplementary Figure 6: Rostro-caudal distribution of PB positive cells in female birds (Nuremberg cohort) (a) Graphical representation of the pigeon beak from a superior view showing the location of the olfactory epithelium (oe), respiratory concha (co), cere (ce), and the tip of the beak. PB positive cells on every 12 th section were counted, and then the total number of cells in a 500µm increment was calculated. Following normalisation, the distribution of PB positive cells were mapped along the rostro-caudal axis. (b-g) The distribution of PB positive cells along the rostro-caudal axis of the pigeon beak are shown for six female birds (P193, P194, P198, P201, P207, P208). Blue bars show PB positive cells in the respiratory epithelium, red bars PB positive cells in the subepidermis, and green bars those PB positive cells located in the feather follicle. Approximate total cell counts per bird are shown in the bottom right hand corner. Note the wide variation in distribution and total cell number between birds. WWW.NATURE.COM/NATURE 6
Supplementary Figure 7: Rostro-caudal distribution of PB positive cells in male birds (Nuremberg cohort) (a) Graphical representation of the pigeon beak from a superior view showing the location of the olfactory epithelium (oe), respiratory concha (co), cere (ce), and the tip of the beak. PB positive cells on every 12 th section were counted, and then the total number of cells in a 500µm increment was calculated. Following normalisation, the distribution of PB positive cells were mapped along the rostro-caudal axis. (b-g) The distribution of PB positive cells along the rostro-caudal axis of the pigeon beak are shown for six male birds (P197, P199, P200, P202, P203, P204). Note that the PB positive cells surrounding the inflammatory lesion in P199 are not included in the subepidermal counts (See Supplementary Figure 17). Blue bars show PB positive cells in the respiratory epithelium, red bars PB positive cells in the subepidermis, and green bars those PB positive cells located in the feather follicle. Approximate total cell counts per bird are shown in the bottom right hand corner. Note the wide variation in distribution and total cell number between birds. WWW.NATURE.COM/NATURE 7
Supplementary Figure 8: Total cell counts in the Nuremberg cohort Graph showing the total average PB positive cell counts in the Nuremberg cohort for male (n=6) and female (n=6) pigeons. Blue bars show the respiratory epithelium, red bars the subepidermal region and green bars the feather follicle. Dark colours show the male birds, and light colours the female birds. Error bars indicate the SEM. We observed no statistically significant difference when comparing male and female total cell counts in the respiratory epithelium (P>0.5), subepidermis (P>0.1) or the feather follicle (P>0.1). WWW.NATURE.COM/NATURE 8
Supplementary Figure 9: Rostro-caudal distribution of PB positive cells in male birds (Vienna cohort) (a) Graphical representation of the pigeon beak from a superior view showing the location of the olfactory epithelium (oe), respiratory concha (co), cere (ce), and the tip of the beak. PB positive cells on every 12 th section were counted, and then the total number of cells in a 500µm increment was calculated. Following normalisation, the distribution of PB positive cells were mapped along the rostro-caudal axis. (b-g) The distribution of PB positive cells along the rostro-caudal axis of the pigeon beak are show for six birds from the Vienna cohort (A1, A4, A6, A7, A8, A9). Blue bars show PB positive cells in the respiratory epithelium, red bars PB positive cells in the subepidermis, and green bars those PB positive cells located in the feather follicle. Approximate total cell counts per bird are shown in the bottom right hand corner. Note the wide variation in distribution and total cell number between birds. WWW.NATURE.COM/NATURE 9
Supplementary Figure 10: Neuronal immunostaining (a-i) Representative images of coronal sections triple stained with PB, NFR, and three different neuronal markers: TUBB3 (a,d,g), neurofilament (b,e,h), and Map1b (c,f,i). Panels (a-c) show the respiratory epithelium, panels (d-f) the subepidermal region, and panels (gi) the feather follicle. Nuclei are stained in pink, iron-rich cells in blue, and neuronal structures in brown (n 5 birds). In the respiratory epithelium we found 0.04% colocalisation with NF (n=1208 cells), 0.6% colocalisation with TUBB3 (n=2818 cells), and 0.01% colocalisation with MAP1b (n=2213 cells). In the subepidermis we found no colocalisation with NF (n=471 cells) or MAP1b (n=803 cells), and only 0.06% colocalisation with TUBB3 (n=1309 cells). In the feather follicle we found no colocalisation with NF (n=286 cells) or MAP1b (n=295 cells), and only 0.24% colocalisation with TUBB3 (n=407 cells). The scale bar indicates 50 µm. WWW.NATURE.COM/NATURE 10
Supplementary Figure 11: Negative and positive controls for immunostaining (a-c) Staining of the ophthalmic branch of the trigeminal nerve served as an internal positive control for the neuronal antibodies (TUBB3, NF, and Map1b). (d) Negative control, which was treated in the same way, but without the addition of the primary antibody. (e) The spleen (which is rich in macrophages) served as a positive control for the MHCII antibody. (f) Negative control for MHCII staining, which was treated in the same way without the addition of the primary antibody. The scale bars indicate 50 µm. We observed very high rates ( 95%) of co-localisation between MHCII and PB positive cells, however it was not 100%. There are three possible explanations for this: (1) A small percentage of the PB positive cells are not macrophages; (2) the PB positive cells are macrophages, but for some reason do not express MHCII; or (3) the antigen has been disrupted by fixation in PFA, and/or the antibody has not completely penetrated the tissue. We think the later proposition is most likely for the following reasons. First, those cells that are not MHCII positive are not isolated, the neighbouring cells are MHCII positive. It seems manifestly unlikely that the true magnetoreceptors should happen to reside in a location surrounded by macrophages. Second, there are very few cells that are PB positive and MHCII negative, and they are not in a regular location. Third, the MHCII antibody binds to a cell surface marker. Fixation in PFA (which is required to preserve the PB stain), is not ideal for this antigen. WWW.NATURE.COM/NATURE 11
Supplementary Figure 12: Methodological scheme employed for electron microscopy Small (~1mm 3 ) tissue pieces dissected from perfused animals with ceramic coated tools were fixed, stained with osmium, dehydrated, infiltrated and embedded for electron microscopy in ph neutral conditions. Following trimming, semi-thin (2µm) and ultra-thin (70nm) sections were cut alternatively. Semi-thin sections were mounted on glass slides and ultra-thin sections were mounted on slot-grids for TEM and on quantifoil findergrids for EELS, EFTEM and SAED analysis. The semi-thin sections were then stained with PB and NFR, and iron rich cells were identified. Adjacent ultra-thin sections were subject to TEM, EELS, EFTEM and SAED analysis. This methodology, while laborious, does not require the re-embedding of semi-thin sections, thereby achieving higher image quality and minimising the risk of contaminating samples with iron from the laboratory environment. WWW.NATURE.COM/NATURE 12
doi:10.1038/nature11046 Supplementary Figure 13: Ultrastructure of PB positive cells and Electron Energy Loss Spectroscopy (EELS) spectrum (a-k) Additional electron micrographs of three different PB positive cells in the respiratory epithelium (a-c), six cells from the subepidermal region (d-i) and two cells in the feather follicle (j,k). Ferritin granules (6-9 nm) are visible throughout the cytoplasm of all cell types, and in some cells siderosomes (si) (g,i,k) and haemosiderin (he) (c) aggregates were observed. Cells in the respiratory epithelium were also found to contain osmophilic lipid droplets (os). Note the numerous slim cytoplasmic projections visible in (d-i,k), that resemble filopodia. Scale bars show 1 µm. (l) EELS spectrum taken from an iron rich cell originating from the follicle cell confirming the presence of carbon, nitrogen, oxygen and iron. The structure of the iron L2,3-edge is highlighted in the expanded portion of the spectrum. Note: this spectrum is representative of that obtained from all cell types. WWW.NATURE.COM/NATURE 13
doi:10.1038/nature11046 Supplementary Figure 14: Subcellular distribution of iron in PB positive cells (a-c) Bright field TEM micrographs paired with their corresponding EFTEM element maps for iron (d-f) that show the subcellular distribution of iron particles in cells from the respiratory epithelium (a), the subepidermal region (b) and the feather follicle (c). Note that each electron dense ferritin granule corresponds with an iron rich signal in the corresponding image. The large osmophilic lipid droplets observed in the respiratory epithelium (a) were found to be largely iron free (d). Scale bars indicate 200 nm. WWW.NATURE.COM/NATURE 14
doi:10.1038/nature11046 Supplementary Figure 15: Selected area electron diffraction (SAED) results (a-d) Bright field transmission electron micrographs (a,c) and the corresponding SAED patterns (b,d) obtained for haemosiderin (a,b) and a siderosome (c,d). SAED patterns of haemosiderin originating from the feather follicle were consistent with a goethite-like mineral (n=3 birds), while SAED patterns for siderosomes from the respiratory epithelium were consistent with ferrihydrite (n=2 birds). Optimal diffraction patterns were only obtained from iron deposits of sufficient size and concentration to be detected above the strong background signal of the epoxy resin employed for embedding. Dashed circles represent fields of view used for SAED and colored quarter circles denote the major rings corresponding to the planar spacings listed in Supplementary Table 3. The scale bars indicate 200 nm. WWW.NATURE.COM/NATURE 15
Supplementary Figure 16: PB positive cells found in other regions in the pigeon (a) Diagram highlighting regions of the pigeon that were dissected with ceramic-coated tools, fixed, sectioned and stained with PB and NFR (n 3 birds). The following abbreviations are used: [a] abdomen, [n] neck, [ba] back, [s] scalp, [lb] lower beak, [w] proximal wing. (b-g,i-n) PB and NFR stained sections through the skin dissected from the abdomen (b,c), the neck (d,e), the scalp (f-g), the proximal edge of the wing (i-j), the lower beak (k-l) and the back (m-n). PB positive cells closely resembling those in the upper beak were found in the feather follicles and subepidermis in all regions. (h) Section from the spleen, showing PB positive cells with light cytoplasmic staining along with punctate blue spherules. The scale bar indicates 10 µm. WWW.NATURE.COM/NATURE 16
Supplementary Figure 17: Inflammatory lesion in P199 (a) Coronal section of P199 stained with PB and NFR showing a large inflammatory lesion with a necrotic centre, surrounded by heterophils, macrophages, lymphoplasmacytic cells and an external layer of connective fibrous tissue. The lesion is highlighted with an arrow. The scale bar shows 1 mm. (b) An equivalent section in an animal where no inflammation is present. The lesion in P199 was surrounded by approximately 80,000 PB positive cells. Examples of these cells are in shown in panels (c-e). The staining in these cells closely resembles that observed in the feather follicle, respiratory epithelium and subepidermis with characteristic blue spherules and light blue cytoplasmic staining. The scale bar indicates 10 µm. (f) Quantification and distribution of PB positive cells associated with the inflammatory lesion in P199. WWW.NATURE.COM/NATURE 17
increment average number of sections standard deviation 1 4.25 1.06 2 4.17 0.39 3 4.00 0.60 4 4.17 0.58 5 3.92 0.67 6 4.08 0.51 7 4.33 0.65 8 3.92 0.51 9 4.08 0.51 10 3.83 0.72 11 4.25 0.45 12 4.08 0.51 13 3.92 0.90 14 4.00 0.85 15 3.83 0.58 16 4.00 0.74 17 4.08 0.67 18 4.08 0.67 19 4.00 0.95 20 4.08 0.79 21 4.08 0.90 22 3.75 0.75 23 3.83 1.34 24 3.67 0.89 25 4.33 0.65 26 3.67 0.98 27 3.75 0.87 28 4.08 0.79 29 4.08 0.51 30 3.92 0.51 31 4.42 0.67 32 3.92 0.51 33 4.17 0.39 34 4.17 0.39 35 4.00 0.00 36 4.42 0.67 37 4.08 0.51 38 4.00 0.43 39 4.17 0.39 40 4.33 0.49 41 4.00 0.43 42 3.67 0.65 Supplementary Table 1: Average Number of sections per an increment for the Nuremberg cohort. Table showing the average number of sections counted for each 500µm coronal increment for pigeons originating from the Nuremberg cohort (n=12). Standard deviations are shown on the right. WWW.NATURE.COM/NATURE 18
increment average number of sections standard deviation 1 3.67 0.52 2 3.33 0.52 3 3.17 0.75 4 3.33 0.52 5 3.00 0.89 6 3.33 0.82 7 3.33 0.82 8 3.50 0.55 9 3.17 0.75 10 2.83 0.41 11 3.17 0.75 12 4.50 0.55 13 4.83 0.41 14 4.33 0.82 15 4.67 0.52 16 4.67 0.52 17 4.50 0.55 18 4.83 0.41 19 4.50 0.55 20 4.67 0.52 21 4.33 0.52 22 3.83 0.41 23 4.00 1.26 24 3.33 1.21 25 4.00 1.26 26 3.67 1.21 27 3.33 1.21 28 3.67 1.03 29 4.00 0.63 30 4.33 0.82 31 4.17 0.75 32 4.33 0.82 33 3.67 0.52 34 4.17 0.75 35 4.00 0.63 36 4.17 0.41 37 4.33 0.82 38 3.83 0.75 39 4.17 0.41 40 4.00 0.63 41 4.67 0.52 42 3.67 0.52 Supplementary Table 2: Average Number of sections per an increment for the Vienna cohort. Table showing the average number of sections counted for each 500µm coronal increment for pigeons originating from the Vienna cohort (n=6). Standard deviations are shown on the right. WWW.NATURE.COM/NATURE 19
Haemosiderin mass - Goethite Siderosome- Ferrihydrite 1 2 3 4 5 6 Miller Indices hkl Standard planar spacing (Goethite) (nm) Measured planar spacing (nm) Miller Indices hkl Standard planar spacing (Ferrihydrite) (nm) Measured planar spacing (nm) 100 0.256 0.253 101 0.416 0.412 012 0.225 0.223 301 0.268 0.260 013 0.198 0.197 111 0.244 0.248 014 0.173 0.171 401 0.218 0.221 110 0.148 0.147 212 0.171 0.170 Supplementary Table 3: SAED results. Measured (± 2%) and standard planar spacings for the diffraction patterns shown in Supplementary Figure 15. The diffraction pattern obtained from a haemosiderin mass in the feather follicle is consistent with goethite (columns 1-3). The diffraction pattern obtained from an electron dense siderosome in the respiratory epithelium is consistent with ferrihydrite (columns 4-6). We consistently observed diffraction patterns with similar planar spacings for siderosomes and haemosiderin masses. WWW.NATURE.COM/NATURE 20