Ultrastructure and Nanoporosity of Human Bone Shown with Correlative On-Axis Electron and Spectroscopic Tomographies

Mineralized collagen fibrils are the building block units of bone at the nanoscale. While it is known that collagen fibrils are mineralized both inside their gap zones (intra-fibrillar mineralization) and on their outer surfaces (extra-fibrillar mineralization), a clear visualization of this architecture in three dimensions (3D), combining structural and compositional information over large volumes, but without compromising the resolution, remains challenging. In this study, we demonstrate the use of on-axis Z-contrast electron tomography (ET) with correlative energy-dispersive X-ray spectroscopy (EDX) tomography to examine rod-shaped samples with diameters up to 700 nm prepared from individual osteonal lamellae in the human femur. Our work mainly focuses on two aspects: (i) low-contrast nanosized circular spaces (“holes”) observed in sections of bone oriented perpendicular to the long axis of a long bone, and (ii) extra-fibrillar mineral, especially in terms of morphology and spatial relationship with respect to intra-fibrillar mineral and collagen fibrils. From our analyses, it emerges quite clearly that most “holes” are cross-sectional views of collagen fibrils. While this had been postulated before, our 3D reconstructions and reslicing along meaningful two-dimensional (2D) cross-sections provide a direct visual confirmation. Extra-fibrillar mineral appears to be composed of thin plates that are interconnected and span over several collagen fibrils, confirming that mineralization is cross-fibrillar, at least for the extra-fibrillar phase. EDX tomography shows mineral signatures (Ca and P) within the gap zones, but the signal appears weaker than that associated with the extra-fibrillar mineral, pointing toward the existence of dissimilarities between the two types of mineralization.

The Ultrastructure and Nano-Porosity of Human Bone Shown with Correlative On-Axis Electron and Spectroscopic Tomographies Table S2.Mean values of minor and major diameter of the "holes" in sample v. Thirty different "holes" were measured in the area shown in Figure S2F."Holes" were manually marked with the "Freehand" tool in Fiji ImageJ and fitted with ellipses to obtain the size of their minor and major axes.
Minor    [Note: Figure S9D is also reported in the manuscript as Figure 1J, and it is cropped and duplicated here for ease of comparison with Figure S9E].
Table S3.Average net intensity of Ca, P, C, and N within the segmented "holes", in the rest of the sample, and outside the sample (i.e., in the background).For ease of comparison, the ratios between the average net intensity inside the sample and inside the "holes", and between the average net intensity inside the "holes" and in the background are also reported for each element.

Ratio between "within holes" and "background"
The Ultrastructure and Nano-Porosity of Human Bone Shown with Correlative On-Axis Electron and Spectroscopic Tomographies Table S4.Average intensity of the HAADF signal within the segmented "holes" and in the background (i.e., outside the rod-shaped samples).For ease of comparison, the ratio between the average intensity inside the "holes" and in the background is also reported.BF signal acquired in tomogram i-a is also included.

Figure S1 .
Figure S1.Site selection and schematic representation of samples orientation.A) Schematic representing the orientation of the "bulk" bone sample relative to the long axis of the femur.B-C) BSE-SEM images with circular marks around the osteons from which the samples were prepared, colour-coded as in D-K.Rod-shaped samples i and ii were prepared from the osteons circled in magenta in C and B, respectively.D-G) Orientation of both rod-(i and ii in D) and wedge-shaped (iii in E; iv in F; v in G) samples in relation to the osteonal axis and osteonal lamellae (the blue circle represents the Haversian canal, and the blue curved lines the inter-lamellar boundaries).H-K) Representation of the coordinate system convention used in the manuscript.The image plane in STEM corresponds to the xy plane, except for sample v where it corresponds to xz instead, as the sample was rotated during the plan-view FIB lift-out.Please note that given the cylindrical geometry of the rod-shaped samples, the definition of xy and yz planes is arbitrary, and depends on the insertion of the sample post in the holder and in the S/TEM instrument.In this case, the xy plane in the tomography reconstructions corresponds to the imaging plane during the tilt series acquisition.Scale bars are 500 µm in B and C. [Note: the drawings are not-to-scale, and the exact location of the samples from the Haversian canal is only indicative].

Figure S2 .
Figure S2.HAADF-STEM images of wedge-shaped samples.A) Overview image of wedgeshaped sample iii, where the characteristic banding pattern of in-plane collagen is predominant.B) Higher magnification view of the banding pattern.C) Overview image of wedge-shaped sample iv, where both longitudinal and lacy motifs are present.D) Higher magnification image showing the transition in orientation of collagen fibrils from in-plane (left side) to out-of-plane (right side).E) Overview image of wedge-shaped sample v, where collagen fibrils are mostly out-of-plane.F) Higher magnification image where "rosettes" and "holes" can be seen.Scale bars are 2 µm in A, C, and E, and 500 nm in B, D, and F. [Note: Figures S2A and S2F are also reported in the manuscript as Figures 2D and 2F, respectively].

Figure S3 .
Figure S3.EDX spectra.HAADF-STEM images of the regions where tomograms A) i-b, B) ii-a, and C) ii-b were acquired, and corresponding EDX spectra integrated from the region marked by the white rectangles.Peaks of elements characteristic of bone (C, N, O, P, Ca) are marked on the spectra.Cu Kα peak originates from the sample post and tomography holder.Ga peaks (Ga Kα: ~9.2 keV; Ga Kβ: ~10 keV; Ga Lα and Lβ: ~1.1 keV), eventually arising from ion implantation, are not distinguishable in the spectrum.Ca/P atomic ratios computed without and with absorption correction (marked by *) are reported in the top right corner of each spectrum.EDX quantification was performed in the Velox software post-background correction (multi-polynomial model, parabolic), with the Brown-Powell ionization cross-section model, and taking into account the sample thickness (400-700 nm) and density (1.8 g/cm 3 ).Scale bars in the HAADF-STEM images are 500 nm.

Figure S4 .
Figure S4.Correspondence between longitudinal and lacy motifs.A) 3D rendering of tomogram ii-a, showing three representative orthogonal planes, corresponding to xy (green), yz (blue), and xz (red), defined as per the coordinate system adopted in this work.B) A representative reconstructed slice in the xy plane where banding pattern is visible.C) Reconstructed slice in the yz plane corresponding to the blue line in A and B. Mineral-rich areas are present, but the banding pattern is not distinguishable.D) Reconstructed slice in the xz plane corresponding to the red line in A and B. E) 3D rendering of tomogram ii-a, showing three representative orthogonal planes, where yz' (cyan) and xz' (orange) are oriented perpendicular and parallel, respectively, to the banding pattern.F) Reconstructed slice corresponding to the cyan line in B and E, oriented perpendicular to the banding pattern, i.e., parallel to the collagen fibrils.The banding pattern is now clearly distinguishable, as opposed to C. G) Reconstructed slice corresponding to the orange line in B and E, oriented parallel to the banding pattern, i.e., perpendicular to the collagen fibrils.In both D and G, mineral-rich regions and "holes" can be observed.Scale bars are 200 nm in B, C, and F, and 100 nm in D and G.A scale bar is not provided in A and E as the 3D representation is not an orthographic projection; the dimensions (x, y, z) of the white box are 916.65 × 1943.55 × 579.60 nm 3 in both A and E.

Figure S5 .
Figure S5.Examples of "holes" where banding pattern is not apparent in cross-section.A) A representative reconstructed slice in the xz plane in tomogram i-b.The "holes" marked by the white rectangles numbered "1" and "2" were examined in the xy planes shown in B/C and D, respectively.B) Reconstructed slice in the xy plane corresponding to the solid green line in A. The banding pattern is not clearly distinguishable within the "hole" (numbered "1" in A), but mostly around it.C) Reconstructed slice in the xy plane corresponding to the dashed green line in A. The banding pattern is more distinguishable than in B. The xy slices in B and C are 39.95 nm away from each other in the z direction.D) Reconstructed slice in the xy plane corresponding to the dotted green line in A. Collagen banding is barely visible, but the collagen fibrils could have been eroded by ion milling as this "hole" (numbered "2" in A) is approximately 13-15 nm away from the sample outer surface.An unmarked image of the regions marked by rectangles in B, C, and D is provided next to each panel.Scale bars are 100 nm in A, and 200 nm in B, C, and D.

Figure S6 .
Figure S6.3D renderings of the tomograms with size analysis of the "holes".The segmented "holes" are colour-coded based on size, and the distribution of the measured values is provided in the histogram: A) tomogram i-a; B) tomogram i-b; C) tomogram ii-a; D) tomogram ii-b.A scale bar is not provided as the 3D representation is not an orthographic projection; the dimensions (x, y, z) of the white box are 879.65 × 1416.20 × 620.50 nm 3 in A, 555.55 × 1611.30× 457.15 nm 3 in B, 916.65 × 1943.55 × 579.60 nm 3 in C, and 1010.10 × 1515.52 × 704.48 nm 3 in D.

Figure S7 .
Figure S7.Additional 3D renderings of the tomograms with segmentation of the "holes".A) 3D rendering of tomogram ii-a with the segmented "holes", shown in different colours to represent their disconnected nature.B) Similar visualization as A for tomogram ii-b.A scale bar is not provided as the 3D representation is not an orthographic projection; the dimensions (x, y, z) of the white box are 916.65 × 1943.55 × 579.60 nm 3 in A, and 1010.10 × 1515.52 × 704.48 nm 3 in B.

Figure S8 .
Figure S8."Holes" are more abundant around the periphery of a mineral ellipsoid.A) HAADF-STEM image where the contour of a partial mineral ellipsoid (oriented longitudinally with respect to the long axis of the sample) is marked (black dotted line).B) 3D rendering of tomogram i-b, where the same mineral ellipsoid contour in A is marked by the white dotted line.For ease of comparison between A and B, a landmark is indicated by *.C) 3D rendering of tomogram i-b sliced along a representative xy plane where it becomes clearer that most segmented "holes" lie outside the mineral ellipsoid (contour marked by the white dotted line).Scale bar is 500 nm in A. A scale bar is not provided in B and C as the 3D representation is not an orthographic projection; the dimensions (x, y, z) of the white box are 555.55 × 1611.30× 457.15 nm 3 .[Note: Figures S8A and S8B are also reported in the manuscript as Figures 1C and 5C, respectively].

Figure S9 .
Figure S9.Representative examples of reconstructed HAADF-STEM slices and EDX maps.A) A representative reconstructed HAADF-STEM slice in the xy plane in tomogram ii-b.B) Reconstructed EDX maps (with and without underlying HAADF-STEM reconstructed slice) corresponding to the area marked by the white square in A. In the areas where "holes" are, Ca (magenta) and P (green) are mostly absent, while C (yellow) and N (cyan) signals appear more intense (e.g., regions circled by the white dotted lines).C) A representative reconstructed HAADF-STEM slice and EDX maps in the xz plane in tomogram i-b.While Ca and P are not detected within the "holes", C and N seem co-localized with some of them (arrowheads).D) HAADF-STEM image where tomogram ii-a was acquired.E) Reconstructed HAADF-STEM slice and EDX maps in the xy plane in tomogram ii-a.A region where faint banding pattern is present in the HAADF-STEM image in D and in the HAADF-STEM reconstructed slice appears depleted in Ca and P, but enriched in C and N. F) 3D rendering of HAADF-STEM and Ca/C signals on the left and right half, respectively.Scale bars are 200 nm in A, D, and E, and 100 nm in B and C. A scale bar is not provided in F as the 3D representation is not an orthographic projection; the dimensions (x, y, z) of the white box are 919.8× 1869.0 × 579.6 nm 3 .[Note: FigureS9Dis also reported in the manuscript as Figure1J, and it is cropped and duplicated here for ease of comparison with FigureS9E].

Figure S10 .
Figure S10.Comparison between reconstructed slices acquired in HAADF-STEM vs. BF-STEM mode.A-B) A representative reconstructed slice in the xy plane (green) acquired in HAADF-STEM (A) and BF-STEM (B) mode.C-D) A representative reconstructed slice in the xz plane (red) acquired in HAADF-STEM (C) and BF-STEM (D) mode.E-F) A representative reconstructed slice in the yz plane (blue) acquired in HAADF-STEM (E) and BF-STEM (F) mode.Although differences are not striking, mineral structures appear overall less blurred and better distinguishable from the background in all planes in BF-STEM mode.Some examples are marked by the dotted ovals.Artifacts are present in the BF-STEM reconstruction in the form of bright dots in the xy and yz planes and bright streaks in the xz plane (regions marked by the yellow *).G) 3D rendering of the HAADF-STEM tomogram showing the position of the three representative slices in A-F.Scale bars are 200 nm in A-F.A scale bar is not provided in G as the 3D representation is not an orthographic projection; the dimensions (x, y, z) of the white box are 879.65 × 1416.2 × 620.5 nm 3 .

Table S1 .
Mean, median, minimum, and maximum "hole" size measured in each tomogram with "Volume thickness map" operation in the Dragonfly software.

Table S5 .
The Ultrastructure and Nano-Porosity of Human Bone Shown with Correlative On-Axis Electron and Spectroscopic Tomographies Voxels labeled in the largest and second largest ROIs identified using the "Connected component -Multi-ROI" operation in the Dragonfly software applied to the segmentation of the mineral phase, and fraction of the largest ROI compared to the total number of voxels segmented for the entire mineral ROI.

Table S6 .
Length, width, and thickness of mineral plates segmented in each tomogram.