Feasibility of mapping the human brain with expansion x-ray microscopy

By Logan Thrasher Collins

Click here for PDF version

Combining synchrotron x-ray microscopy with expansion microscopy may represent a feasible approach for whole human brain connectomics at the nanoscale. Synchrotron x-ray microtomography on its own provides extremely fast imaging at high resolution, yet necessary tradeoffs between imaging throughput and resolution mean that the imaging an entire human brain with voxel sizes of less than 100 nm may still take much too long with current synchrotron technology. Fast imaging with voxel sizes of 300-1000 nm is much more readily achievable in the near future. Furthermore, expansion microscopy isotropically enlarges tissue by infusion of a swellable hydrogel, facilitating resolution increases. The combination of x-ray microtomography and expansion microscopy (hereafter referred to as ExxRM) could thus push the effective voxel size down to the level needed for dense connectomics. However, because tissue volume and imaging time scale cubically with expansion factor, careful balance between design of the synchrotron x-ray optical setup and the degree of expansion will be vital. In this perspective, I will explore balances between synchrotron optical engineering choices and expansion factor, propose methods to successfully implement ExxRM in the context of human brains, and estimate how much it would cost to image the human brain in this way. Imaging brains via ExxRM may represent a crucial paradigm shift in connectomics which paves the way for holistic understanding of human brain function.


Nanoscale connectomic imaging of the entire human brain represents a long sought-after goal that could provide the foundation for dramatic advances in neurobiology, neurotechnology, and artificial intelligence.1,2 Currently, the leading method for nanoscale connectomics is volume electron microscopy (EM). But imaging a 1 mm3 volume of mouse cortex over a period of 6 months required a tremendous collaborative effort by Yin et al. to develop a parallelized and fully automated transmission electron microscopy (TEM) system consisting of six instruments working in parallel.3 Each of these six instruments cost $125,000. The mouse brain has a volume of roughly 500 mm3, meaning that if these numbers were directly scaled, the process would take 250 years.4 That said, Yin et al.’s TEM dataset had high resolution with 4×4×40 nm voxels, so throughput might be increased by imaging at somewhat lower resolution. As such, it is conceivable to argue that advances in EM technology may enable imaging of an entire mouse brain at sub-100 nm voxel size over the course of a few years at a cost of around $10M. EM therefore represesnts a viable option for mouse brain connectomics. But the human brain’s volume is roughly 1200 cm3, around 2400-fold larger than the mouse brain.5 Even if EM technology somehow advances to the point where an entire mouse brain can be imaged at 100 nm3 voxel size in a single year for $10M, mapping a human brain with comparable parameters would take thousands of years. In my view, this provides a strong argument for the idea that a radically different approach is necessary for human brain connectomics.

Expansion light-sheet fluorescence microscopy represents a promising alternative to EM, yet this modality also falls short when considering the volume of the human brain (particularly after expansion). For instance, Lillvis et al. utilized 8-fold expansion and lattice light-sheet microscopy (ExLLSM) to image the Drosophila central complex in three colors with effective 30×30×100 nm voxels over the course of 5 days.6 But even accounting for the 8-fold expansion (512-fold volume increase), this amounts to a volume of less than 0.5 mm3. Imaging even a 4-fold expanded mouse brain assuming these numbers would take 876 years. Lattice light-sheet microscopes are relatively inexpensive at a few hundred thousand dollars each7 and thus might be parallelized enough to image an entire mouse brain within a year. However, the 2400-fold larger volume of the human brain relative to the mouse brain indicates that it is probably not reasonable to expect success in human connectomics through ExLLSM.

Based on these calculations, I suggest that a radically different strategy is needed for to put the goal of human brain connectomics within reach. Uniting synchrotron x-ray microtomography (XRM) (Figure 1A-B), expansion microscopy (ExM), a recent staining method known as Unclearing Microscopy8 may facilitate “ExxRM” imaging of the human brain at sub-100 nm voxel size on timescales of around 1 year for a cost of around $10M. Success of this approach will necessitate overcoming some technical hurdles, yet I am optimistic that these particular challenges can be conquered. ExxRM may represent a feasible platform to acquire images suitable for dense connectomics across entire human brains.

Figure 1 Principles of synchrotron XRM. (A) Synchrotrons generate electrons from a source, propel them through a linear accelerator (linac), raise their energy in a booster ring, and then keep the electrons circulating for long periods of time at relativistic speeds in the storage ring. As relativistic electrons move along a curved path controlled by bending magnets, they emit brilliant x-ray beams in the direction tangent to the curve in the direction of the electron movement. Insertion devices that inject the x-rays into beamlines are placed at straight sections of the synchrotron ring. These insertion devices stimulate emission of bright and coherent beams into the experimental stations. (B) Microtomography beamlines receive an x-ray beam from an insertion device and filter out a narrow band of wavelengths using a monochromator. The beam passes through a sample on a rotating tomography stage that can be positionally adjusted to change the location of the field of view within the sample. Projection images are taken across 360° of rotation. X-rays passing through the sample are converted to visible light using a scintillator, then directed by mirrors to a lens system that magnifies the image. This light hits a detector camera and data is recorded for 3D reconstruction.

Recommended methodologies for ExxRM

ExxRM’s success will require developing methodological strategies to mitigate technical challenges. Obtaining sufficient contrast to resolve subcellular features will be vital. ExM cubically dilutes the amount of cellular material per unit volume, so creative staining techniques will be needed. Approaches that ensure stability of expanded tissues under brilliant x-ray illumination for long durations will also be crucial. Multicolor imaging would greatly benefit the usefulness of whole brain connectomics data, so ways of efficiently obtaining tomograms in multiple colors are needed. These challenges must be conquered to translate ExxRM.

A central engineering hurdle for ExxRM is attaining sufficient feature contrast for capture of clear images. Because expanded tissues experience cubic dilution of target biomolecules, contrast from traditional stains such as osmium tetroxide almost entirely vanishes during x-ray microtomographic imaging (Collins et al., unpublished data). An elegant solution to this problem has come in the form of a recently developed technique called Unclearing Microscopy.8 For this technique, M’Saad et al. biotinylated primary amines (found on proteins, phosphatidylethanolamine lipids, etc.) throughout expanded samples, treated the sample with streptavidin horseradish peroxidase (streptavidin-HRP) fusion protein, and then stained with ionic silver reagents (from the EnzMet™ HRP Detection Kit) or with 3’3-diaminobenzidine (DAB). This triggered enzymatic deposition of enough chromogenic silver or DAB to make 20-fold expanded HeLa cells visible to the naked eye despite their 8000-fold increase in volume relative to the unexpanded state. Phase contrast light microscopy subsequently revealed subcellular features such as mitochondrial cristae, nuclear pore complexes, and nuclear membrane. Unclearing thus facilitates physical reconstruction of the structures that are pulled apart by ExM, filling in the gaps left by the expansion process (Figure 2A-B). Silver stain Unclearing could enable either absorption XRM or phase contrast XRM of expanded tissues since silver has excellent x-ray attenuation index β at relevant beam energies as well as excellent x-ray phase decrement δ. Focusing on phase contrast XRM may represent a better option since it can decrease the necessary dose of radiation per unit time by using x-ray wavelengths which are not absorbed as strongly by the tissue. Furthermore, phase contrast XRM is most sensitive to differences in sample density9 and silver staining forms dense precipitates of metallic silver (which in pure form has a high density of 10.49 g/cm3), so this approach might provide superior contrast in the context of ExxRM. Here, ExM’s sample dilution might prove advantageous because it could generate strongly distinct densities between silver-stained cellular features and the rest of the stabilized hydrogel. It should be noted that, for phase contrast XRM, the previously mentioned stabilization approach would need to fill the space between cellular features with a substance that differs substantially in density from the silver (or similar stain). Combining Unclearing Microscopy with phase contrast XRM could provide excellent feature contrast for ExxRM connectomics.

Figure 2 Proposed sample preparation technique for ExxRM. (A) Macroscale view of a human brain undergoing expansion, Unclearing, and stabilization. Slicing into centimeter-scale subvolumes is not shown but may also represent a beneficial strategy. (B) Nanoscale view of a neuron within the brain undergoing expansion, Unclearing, and stabilization. Considerable amplification of signal density after Unclearing should occur.

Expanded tissues are known for their fragility and synchrotron x-rays are known for their harshness, so strategies for solving the problem of sample degradation are needed. Newer expansion recipes which use high monomer concentrations have displayed substantially greater physical sturdiness than earlier generations of ExM hydrogels,10,11 so this may aid in stabilization to some degree. Yet additional advances in ExM sample preparation may still be necessary due to high required fluxes and long imaging times. Cutting human brain samples into smaller volumes on the order of a few centimeters may improve the situation by decreasing imaging times per sample. Keeping the sample at cryogenic temperatures with the help of cryoprotectants to prevent ice crystals damaging the tissue may also help since heat generation from the x-rays is particularly problematic for hydrated samples.12,13 That said, low temperatures would not fix the issue of radiation damage from ionization, so this issue should still be considered.14 Further stabilization might come in the form of infusion of a rapidly crosslinkable or crystalizing substance which irreversibly locks all biomolecules into place (Figure 2A-B). The chosen substance would need to undergo an inducible crosslinking or crystallization reaction that does not cause distortion of the expanded tissue, to not shrink or distort the expanded tissue while it diffuses into the hydrogel, and to minimally absorb x-rays at the energy range used for imaging, and to contrast sharply with the chosen stain material. The phase decrement of the stabilizing substance should be similar or lower compared to amorphous ice. These methods may enable sufficient stabilization for preventing nanoscale tissue degradation even with exposure to extremely brilliant synchrotron x-rays.

Multicolor ExxRM would greatly enhance the value of acquired image data since synapses and key biomolecules could then be tagged within the 3D reconstruction.15,16 Fortunately, multicolor absorption x-ray microscopy represents an established technique wherein two or more staining materials undergo two or more rounds of imaging at beam energies corresponding to the absorption edges of the chosen materials. The absorption edges in question must be sufficiently distinct that minimal overlap in detection occurs across the different beam energy imaging rounds. As an example, Depannemaecker et al. employed this approach by using antibody-linked gold nanoparticles to mark neuronal nuclei in mouse brains while also staining neurons with silver via Golgi’s method.15 They imaged at a beam energy corresponding to an absorption edge of silver and at a beam energy corresponding to an absorption of gold. In addition, multicolor phase contrast x-ray imaging might be accomplished by using molecularly targeted tags with distinct densities relative to the Unclearing silver stain. For instance, gold has a density about twice that of silver, so it might be possible to segment gold nanoparticles after data acquisition. A potential challenge for these methods comes from the difficulty of inducing reliable diffusion of large metallic nanoparticles into tissue. That said, post-expansion pre-stabilization staining could help overcome the problem since expanded gels are typically more porous than pre-expansion tissue. Multicolor ExxRM in 2-3 colors should be achievable, opening doors to more useful whole-brain image datasets.

Though some trial and error will doubtless prove necessary for ExxRM optimization, the required technologies lay within fairly close reach. High contrast could be achieved by applying Unclearing Microscopy to counter signal dilution from expansion. Stability under brilliant x-rays may be possible by leveraging stabler expansion recipes with higher monomer concentrations, by cutting the brain into centimeter-scale subvolumes to decrease radiation dose per sample, by employing cryogenic temperatures during imaging, and by developing rigidity-enhancing materials to infuse into expanded gels. Multicolor ExxRM could be achieved by staining with metallic nanoparticles linked to affinity reagents and performing additional scans tuned to the absorption edges of the chosen metals. These directions may provide a path towards successful ExxRM.

How fast can synchrotrons image expanded brains?

Synchrotron facilities offer bright and coherent x-rays that can rapidly image large volumes of tissue at high resolutions (Figure 1A-B). For example, Bosch et al. employed synchrotron XRM at the Swiss Light Source (SLS) and the Diamond Light Source (DLS) to acquire 3D reconstructions of an entire mouse olfactory bulb with 325 nm voxel size.17 In their study, tomograms of approximately 1 mm3 volume took around 20 minutes to acquire. Another investigation by Dyer et al. used the Advanced Photon Source (APS) synchrotron to image a region of mouse cortex at 650 nm voxel size and 1.47 mm3 volume. Imaging this volume took only 6 minutes.18 Walsh et al. leveraged the European Synchrotron Radiation Facility Extremely Brilliant Source (ESRF-EBS) to image a variety of whole human organs at low resolution and subvolumes within human organs at higher resolution.19 Notably, they reconstructed a cylinder with ~57 mm3 volume (5.4 mm diameter and 2.5 mm height) within human spleen at 1290 nm voxel size over the course of 2 hours. Ding et al. used the APS to image larval zebrafish with 743 nm voxels and 1.5 mm3 volumes per tomogram in 20 minutes with monochromatic x-ray scans and in just 20 seconds with polychromatic “pink beam” x-ray scans.20 Despite their much higher flux and correspondingly much shorter imaging times, the polychromatic scans unfortunately showed lower resolution due to poor SNR. It should be noted that resolution represents a distinct concept from voxel size and that contrast and SNR play major roles in the final resolution. That said, voxel size can act as a rough proxy for comparisons across different imaging setups assuming that contrast and SNR are consistent. In another key example, Foxley et al. utilized the APS to image whole mouse brains (11.7×11.7×17.7 mm) over the course of 7 hours of acquisition time with 1117 nm voxels.21 Finally, Rodgers et al. used the SOLEIL synchrotron and an alternative technique for tomographic acquisition (discussed in more detail later) to image an entire mouse brain a 650 nm voxel size in 8 hours.22 Though the parameters for image acquisition vary considerably across different biological samples and synchrotron hardware setups (table 1), these examples illustrate the power of synchrotron imaging for mapping biological tissues.

Table 1 Specifications of synchrotron tissue imaging experiments from selected references. 

ReferenceSmallest voxel sizeSingle tomogram volumeTotal reconstructed volumeImaging time per tomogramTotal imaging time
Bosch et al.325 nm1.39 mm3~5-10 mm320 min.~40 min. to ~4 hours
Dyer et al.650 nm1.47 mm31.47 mm36 min.6 min.
Walsh et al.1290 nm57 mm357 mm3120 min.2 hours
Ding et al.743 nm1.5 mm34.5 mm320 min. or 20 s*1 hour or* 1 min.
Foxley et al.1117 nm5.61 mm32423 mm3108 s7 hours
Rodgers et al.650 nmspecial method911 mm3special method8 hours
*with pink beam acquisition

While the voxel sizes described above are not small enough for nanoscale connectomics, incorporating ExM into the sample preparation pipeline presents the possibility of attaining sufficient resolutions for tracing neurites. Expansion factors of around 4x, 10x, and 20x are respectively possible with standard ExM,23,24 some of the newer enhanced ExM protocols,10,25,26 and iterative ExM methods.11,27 Expansion factor can also be intentionally decreased somewhat by adjusting salt concentration. It is important to note that not all ExM protocols retain lipid membranes, so selecting a recipe that preserves such cellular boundaries10,11,28 will be vital for morphology reconstrction. Balance between expansion factor and imaging time will represent a key determinant of the success of ExxRM for whole brain imaging. Higher expansion factors increase resolution linearly while increasing imaging time cubically. Synchrotron imaging’s extreme speed has the potential to keep up with such volumetric increases, though even this has certain limits.

Du et al. offer a helpful mathematical model to estimate the number of photons needed for imaging a sample with sufficient contrast to distinguish between voxels of desired size.12 The model starts by considering how many photons n̄pixel are needed to image a single pixel on a 2D tomographic slice. The pixel is assumed to contain a “feature” material with a phase decrement δf that must be distinguished against a background with a phase decrement δb. Distinctions can be made between adjacent background pixels b, overlaying background pixels bo’ and underlaying background pixels bu’. The SNR is assumed to equal 5 as based on the Rose criterion, which qualitatively represents an acceptable SNR for visually useful images. Background attenuation coefficient is given by µb’ and the thicknesses of the overlaying and underlaying background material are tb’o and tb’u respectively. After the value of n̄pixel is determined, the value is multiplied by the square N2 of the number of pixels N on the detector’s horizontal axis to find the number of photons needed per tomogram. As Du et al. explain, the number of slices is not used as a multiplicative factor since the photons are distributed across all of the rotation angles for each pixel. These calculations act as a guide to what sample properties and beam properties are needed for ExxRM.

The Du et al. model can help determine the feasibility of ExxRM by estimating the necessary synchrotron flux for imaging the brain at a desired voxel size in a reasonable time frame. Consider 1 cm3 of brain tissue expanded 4-fold to a block with dimensions of 4×4×4 cm. Based on this, assume that the average thickness of overlaying and underlaying background material is 2 cm. Also assume that the background material is made of amorphous water ice with a density of 0.92 g/cm3. Since the widths of the smallest neuronal features are about 20 nm wide,29 assume that the 4-fold expansion provides a minimum feature thickness of 80 nm. By using the Unclearing Microscopy technique, such features might be “filled in” after expansion with a metallic silver stain.8 Thus, I will assume that the 4-fold expanded membrane features are made up of metallic silver with a density of 5.25 g/cm3, half the density of pure metallic silver (to account for imperfect staining). Feature phase decrements δf, background phase decrements δb, and background attenuation coefficients µb’ across x-ray energies ranging from 10 keV to 30 keV can be obtained using the calculator available at (https://henke.lbl.gov/optical_constants/pert_form.html) which is based on the Henke dataset.30 This calculator does not directly give µb’, but it does give attenuation index β, for which µb’ = 4πβ/λ. I will furthermore assume that a 3×3 mm beam and detector are possible to employ and that the pixel size is 300 nm (75 nm effective post-expansion). According to the resulting model, the total number of photons needed to image a single tomogram (volume of 21.21 mm3) within the 4-fold expanded block at an energy of 30 keV is 1.47×1013. By examining this model across a range of x-ray energies (Figure 3A) and across a range of feature sizes (Figure 3A), one can obtain a better understanding of how the different variables might play out in an experimental setting.

Figure 3 (A) Total photons needed to image a single tomogram of 4-fold expanded tissue with volume of 21.21 mm3 at 300 nm voxel size and two different silver stain densities while assuming feature thickness of 80 nm, plotted against x-ray energy in keV. (B) Total photons needed to image a single tomogram of 4-fold expanded tissue with volume of 21.21 mm3 at 300 nm voxel size and two different silver stain densities while assuming 30 keV x-ray energy, plotted against desired feature thickness necessary to resolve.

How might this translate to imaging time for a single tomogram, the 4×4×4 cm expanded block, and the whole human brain? Contemporary beamlines which employ multilayer monochromators can reach fluxes of more than 1012 photons/s.31–35 As such, the single tomogram under the assumptions of the model might be acquired in around 10 seconds. Comparing the tomogram volume of 21.21 mm3 and the expanded block volume of 64000 mm3, this means that the whole block would require about 3000 tomograms and take 8.3 hours to image. Since the entire human brain should contain approximately 1200 of these blocks, whole-brain acquisition might take around 1.14 years of continuous imaging. All of this assumes the post-expansion effective voxel size of 75 nm (4-fold expansion and actual voxel size of 300 nm), 30 keV beam energy, and Unclearing of neuronal features to a density of 5.25 g/cm3 metallic silver. If these theoretical projections successfully align with experimental outcomes, ExxRM could facilitate nanoscale connectomics for the entire human brain in a time frame of around 1-2 years.

Cost estimates for whole brain ExxRM

While the power of the synchrotron facility comes with a high price tag, ExxRM may still represent the most overall cost-effective option for human brain connectomics. Consider the costs associated with the DLS as an illustrative example. The DLS is a third-generation facility and is currently one of the better synchrotrons in terms of its ability to produce bright and coherent x-rays. Building the DLS and its first seven beamlines from 2003-2007 cost about $316M, its later upgrades cost $144M and $134M, and its yearly operational costs have increased from $28M in 2007-2008 to $81M in 2019-2020 (table 2). Based on these data points, construction of a new synchrotron beamline costs approximately $10M and yearly maintenance may cost roughly $500K. This provides a framework for estimating the cost of a dedicated human brain ExxRM connectomics beamline.

Table 2 Costs associated with the Diamond Light Source36 as a case study on how much money is needed to build and maintain a state-of-the-art synchrotron facility.

FacilityDiamond Light Source
Initial costs$316M: synchrotron, first seven beamlines, surrounding buildings, construction started in 2003 and completed in 2007
Upgrade 1 costs$144M: fifteen more beamlines, detector development program, construction started in 2004 and completed in 2012
Upgrade 2 costs$134M: ten more beamlines, not stated when construction started but was completed in 2021
Maintenance and operational costs$28M in 2007-2008, $48M in 2012-2013, $81M in 2019-2020

Compact Light Source (CLS) technology should be considered as well before continuing. CLS instruments produce x-ray beams that fall somewhere between laboratory x-ray microscopes and synchrotrons in terms of brightness and coherence.37 CLS instruments are furthermore small enough to fit into a single room and are inexpensive enough that a large number of them could potentially be constructed in parallel. At first glance, CLS technology seems a more economically viable alternative to synchrotron beamlines, yet it still probably is not a good option at this time. Existing CLS instruments are not likely suitable for human brain (or even mouse brain) ExxRM connectomics in the foreseeable future because their optical engineering requirements and mediocre level of x-ray flux preclude rapid tomography at submicron resolution, particularly when a large field of view is desired.38 There is a small possibility that future advances in CLS technology could change this situation, yet this seems fairly unlikely, so synchrotron-based imaging remains the best route.

Building an entire synchrotron solely for ExxRM connectomics is probably less efficient than establishing an agreement with an existing synchrotron to construct a connectomics beamline. While one might envision additional parallelization through custom design of the beamline to split the beam to land in multiple sample chambers, splitting the beam would divide the photon flux and therefore increase imaging times for no net gain in speed. As such, parallelization would likely require an additional insertion device and thus an entire new beamline for each new sample chamber. Insertion devices consist of a series of precisely engineered magnets built into straight sections of the synchrotron’s ring. These magnets, known as undulators or wigglers depending on the specifics of the insertion device, stimulate directed emission of a brilliant x-ray beam out from the storage ring.39 Yet even if we assume $10M total plus yearly maintenance costs for each connectomics beamline, imaging multiple human brains over the course of a year or a single human brain in just a few months remains a reasonable proposition. For a project as important as mapping the human brain at sufficient resolution for dense neuronal reconstruction, price tags in the range of tens of millions of dollars may not be out of reach.

Conclusion and outlook

Connectomics needs a technological paradigm shift if it is to feasibilize dense mapping of one or more human brains. Though it comes with some technical challenges, ExxRM may represent a key strategic shift that drastically reduces human brain connectomics timelines from centuries down to 1-2 years for the image acquisition step. Data storage and early processing steps (e.g. tomographic reconstruction) will of course require data centers and high-performance computing, but this field is rapidly advancing and may indeed be capable of handling the challenge. Assuming 1 byte per voxel, the amount of storage needed for a 4-fold expanded human brain with 300 nm physical voxel size (75 nm effective voxel size) is about 2.84 exabytes. Segmentation of the human brain dataset will probably represent a vastly more difficult problem as well as require substantially more compute resources, so further advances in this area will need to occur in parallel with ExxRM development. Realizing the benefits of connectomics in the form of complete computational models of the brain will take additional extensive research. Precisely correlating gene expression and electrophysiological properties with neuronal morphology (i.e. “cell type”) may represent a major step towards bridging the divide between structural data and functional activity. Nonetheless, the allure of having a holistic anatomical picture of the brain may serve as a driving force in the meantime. ExxRM has the potential to transform the dense connectomics field, enabling anatomical imaging of the entire human brain with sub-100 nm voxel size and high contrast in around 1-2 years for a price of roughly $10M.


1.        Collins, L. T. The case for emulating insect brains using anatomical “wiring diagrams” equipped with biophysical models of neuronal activity. Biol. Cybern. 113, 465–474 (2019).

2.        Koene, R. A. Fundamentals of whole brain emulation: state, transition, and update representations. Int. J. Mach. Conscious. 04, 5–21 (2012).

3.        Yin, W. et al. A petascale automated imaging pipeline for mapping neuronal circuits with high-throughput transmission electron microscopy. Nat. Commun. 11, 4949 (2020).

4.        Badea, A., Ali-Sharief, A. A. & Johnson, G. A. Morphometric analysis of the C57BL/6J mouse brain. Neuroimage 37, 683–693 (2007).

5.        Cosgrove, K. P., Mazure, C. M. & Staley, J. K. Evolving Knowledge of Sex Differences in Brain Structure, Function, and Chemistry. Biol. Psychiatry 62, 847–855 (2007).

6.        Lillvis, J. L. et al. Rapid reconstruction of neural circuits using tissue expansion and light sheet microscopy. Elife 11, e81248 (2022).

7.        Watkins, S. C. & St. Croix, C. M. Light sheet imaging comes of age. J. Cell Biol. 217, 1567–1569 (2018).

8.        M’Saad, O., Shribak, M. & Bewersdorf, J. Unclearing Microscopy. bioRxiv 2022.11.29.518361 (2022) doi:10.1101/2022.11.29.518361.

9.        R A Lewis. Medical phase contrast x-ray imaging: current status and future prospects. Phys. Med. Biol. 49, 3573 (2004).

10.      Klimas, A. et al. Magnify is a universal molecular anchoring strategy for expansion microscopy. Nat. Biotechnol. (2023) doi:10.1038/s41587-022-01546-1.

11.      M’Saad, O. & Bewersdorf, J. Light microscopy of proteins in their ultrastructural context. Nat. Commun. 11, 3850 (2020).

12.      Du, M. et al. Upscaling X-ray nanoimaging to macroscopic specimens. J. Appl. Crystallogr. 54, 386–401 (2021).

13.      Matsuyama, S. et al. Elemental mapping of frozen-hydrated cells with cryo-scanning X-ray fluorescence microscopy. X-Ray Spectrom. 39, 260–266 (2010).

14.      Lombi, E. et al. Fast X-Ray Fluorescence Microtomography of Hydrated Biological Samples. PLoS One 6, e20626 (2011).

15.      Depannemaecker, D. et al. Gold Nanoparticles for X-ray Microtomography of Neurons. ACS Chem. Neurosci. 10, 3404–3408 (2019).

16.      Logan Thrasher Collins, Kayla Siletti, Kristine Fischenich, Jennifer Coulombe, Nathan Anderson, M. S. Structural brain mapping using antibody-conjugated gold nanoparticles and x-ray microscopy. in Society for Neuroscience Conference (2019).

17.      Bosch, C. et al. Functional and multiscale 3D structural investigation of brain tissue through correlative in vivo physiology, synchrotron microtomography and volume electron microscopy. Nat. Commun. 13, 2923 (2022).

18.      Dyer, E. L. et al. Quantifying Mesoscale Neuroanatomy Using X-Ray Microtomography. eNeuro 4, ENEURO.0195-17.2017 (2017).

19.      Walsh, C. L. et al. Imaging intact human organs with local resolution of cellular structures using hierarchical phase-contrast tomography. Nat. Methods 18, 1532–1541 (2021).

20.      Ding, Y. et al. Computational 3D histological phenotyping of whole zebrafish by X-ray histotomography. Elife 8, e44898 (2019).

21.      Foxley, S. et al. Multi-modal imaging of a single mouse brain over five orders of magnitude of resolution. Neuroimage 238, 118250 (2021).

22.      Rodgers, G. et al. Mosaic microtomography of a full mouse brain with sub-micron pixel size. in Proc.SPIE vol. 12242 122421L (2022).

23.      Chen, F., Tillberg, P. W. & Boyden, E. S. Expansion microscopy. Science (80-. ). 347, 543 LP – 548 (2015).

24.      Tillberg, P. W. et al. Protein-retention expansion microscopy of cells and tissues labeled using standard fluorescent proteins and antibodies. Nat. Biotechnol. 34, 987–992 (2016).

25.      Damstra, H. G. J. et al. Visualizing cellular and tissue ultrastructure using Ten-fold Robust Expansion Microscopy (TREx). Elife 11, e73775 (2022).

26.      Park, H.-E. et al. Scalable and Isotropic Expansion of Tissues with Simply Tunable Expansion Ratio. Adv. Sci. 6, 1901673 (2019).

27.      Chang, J.-B. et al. Iterative expansion microscopy. Nat. Methods 14, 593–599 (2017).

28.      Karagiannis, E. D. et al. Expansion Microscopy of Lipid Membranes. bioRxiv 829903 (2019) doi:10.1101/829903.

29.      Sneve, M. A. & Piatkevich, K. D. Towards a Comprehensive Optical Connectome at Single Synapse Resolution via Expansion Microscopy   . Frontiers in Synaptic Neuroscience   vol. 13 at https://www.frontiersin.org/articles/10.3389/fnsyn.2021.754814 (2022).

30.      Henke, B. L., Gullikson, E. M. & Davis, J. C. X-Ray Interactions: Photoabsorption, Scattering, Transmission, and Reflection at E = 50-30,000 eV, Z = 1-92. At. Data Nucl. Data Tables 54, 181–342 (1993).

31.      Albrahim, M. et al. Reduction and Agglomeration of Supported Metal Clusters Induced by High-Flux X-ray Absorption Spectroscopy Measurements. J. Phys. Chem. C 125, 11048–11057 (2021).

32.      Owen, R. L., Holton, J. M., Schulze-Briese, C. & Garman, E. F. Determination of X-ray flux using silicon pin diodes. J. Synchrotron Radiat. 16, 143–151 (2009).

33.      Weitkamp, T. et al. Parallel‐beam imaging at the ESRF beamline ID19: current status and plans for the future. AIP Conf. Proc. 1234, 83–86 (2010).

34.      Baussens, O. et al. Characterization of High-Flux CdZnTe with optimized electrodes for 4th generation synchrotrons. J. Instrum. 17, C11008 (2022).

35.      Moosmann, J. et al. X-ray phase-contrast in vivo microtomography probes new aspects of Xenopus gastrulation. Nature 497, 374–377 (2013).

36.      Diamond Light Source About Us. https://www.diamond.ac.uk/Home/About/ (2022).

37.      Gradl, R. et al. Propagation-based Phase-Contrast X-ray Imaging at a Compact Light Source. Sci. Rep. 7, 4908 (2017).

38.      Gunther, B. et al. The versatile X-ray beamline of the Munich Compact Light Source: design, instrumentation and applications. J. Synchrotron Radiat. 27, 1395–1414 (2020).

39.      Willmott, P. Synchrotron Physics. in An Introduction to Synchrotron Radiation 39–86 (2011). doi:https://doi.org/10.1002/9781119970958.ch3.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s