Tissue expansion of ‘cleared’ whole brains reduces contrast in ex vivo MRI

Multimodal imaging of optically-cleared brains, ex vivo, by Magnetic Resonance Imaging (MRI) and light microscopy (LM) presents unique opportunities for studying the brain at various scales and resolutions. However, CLARITY -cleared brains lack MRI contrast, implicating lipids as the major source of MRI contrast. We explored the ex vivo MRI compatibility of uDISCO, ECi and Scale -cleared brains. Surprisingly, uDISCO and ECi -cleared brains retain MRI contrast, whereas Scale-cleared brains show a severe loss of MRI contrast, as CLARITY. Determination of lipid-content in cleared samples shows that CLARITY, uDISCO and ECi are strongly delipidating, whereas Scale preserves most lipids. We conclude that MRI contrast can be associated with tissue expansion (and hyperhydration) rather than with lipid-content. Thus, we present two clearing methods compatible with ex vivo MRI.


Introduction
One rapidly evolving eld of study is the study of the intricate brain connectome in three dimensions.
However, the opaque nature of the brain and its large dimensions limit the use of high-resolution uorescence light microscopy (LM) to image the brain in its entirety [1][2][3][4] .This drawback has been addressed by emerging tissue-clearing techniques (e.g., CLARITY [5][6][7] ) combined with advanced LM imaging methods, such as light-sheet microscopy [8][9][10] .Tissue clearing renders entire tissues transparent; allowing light to penetrate and traverse extended distances in the tissue 7,11 , largely eliminating the need for tissue-sectioning and histology.However, clearing processes may produce structural deformities (e.g., anisotropic shrinkage or expansion) [12][13][14][15][16][17] and these may complicate post hoc image-tiling and reconstruction of the entire tissue in 3D 18,19 .Mesoscale imaging methods should, in practice, circumvent some of these hurdles.Indeed, modern uorescence imaging is increasingly combined with complementary imaging methods and techniques for this purpose [20][21][22] .
One attractive method is ex vivo Magnetic Resonance Imaging (ex vivo MRI) 1,23 .Ex vivo MRI provide anatomical images of tissues (without the need to label) relatively fast (< minutes) and is non-damaging.
Added bene ts of ex vivo MRI is the ability to obtain images in higher resolution and signal-to-noise ratio than in vivo MRI 24 .These make the method highly suitable for integration of data for precise image registration by, for instance, providing set-coordinates for correction of deformities and outlining major brain regions 12,[25][26][27][28][29] .Thus, multimodal imaging of cleared-tissues by LM and ex-vivo MRI should ease on the convergence between the micro-and mesoscale and aiding in framing the context of the uorescence information 1,30 .Despite these advantages, dual LM-ex vivo MRI is not common (more common are in vivo MRI followed by ex vivo LM 24,26,30 ).In fact, it has been recently shown that several clearing methods, notably CLARITY, produce samples with no contrast when imaged by ex vivo MRI 5,14,[31][32][33] .These were interpreted to result from the removal of lipids (delipidation) from the tissues 15,[34][35][36][37][38] .This assumption implies that any clearing technique that involves delipidation (for instance, organic solvent-based clearing techniques DISCO [39][40][41][42][43] , Ethyl-cinnamate-ECi 44,45 , or others 46 ) should produce MRI-incompatible cleared brains, whereas tissues cleared by lipid-preserving methods, such as Scale 47 (or others 15,34,[48][49][50][51][52][53][54][55] ) should retain contrast in ex vivo MRI.However, this binary classi cation remains unexplored, as but a handful of clearing techniques have been tested 34 .Consequently, brain-clearing technique(s) combined with a highly detailed anatomic brain MR images have not been reported to date.
To address these unknowns, we systematically assessed ex vivo MRI contrast of transparent wholebrains cleared by various clearing techniques.In parallel, we examined the lipid content in cleared brains by two staining methods, Raman spectroscopy (RS) and quantitative lipidomics (by mass spectrometry; MS).Surprisingly, and in contrast to previous reports 34,36 , we nd that ECi and uDISCO-two highly delipidating methods (removing >80% and 99% of lipids, respectively)-preserve ex vivo MRI contrast following their hydration.On the contrary, images of CLARITY cleared whole brains (with >99% removal of lipids) are completely devoid of contrast.Counterintuitively, whereas Scale preserves >70% of lipids during the clearing process, it yields samples that are not MRI compatible, namely without any noticeable MRI contrast 34 .Collectively, these show that lipid-content in cleared samples is not the major source for ex vivo MRI contrast.Curiously, we note that while CLARITY and Scale cause substantial expansion and hyperhydration of the brains, uDISCO and ECi induce shrinkage.These lead us to suggest that tissue expansion, and thus de facto increase in water content and decrease in the vicinity of water-interacting macromolecules, is the major cause for loss of MRI contrast.Together, we demonstrate two nonexpanding and highly delipidating tissue clearing methods that are suitable for ex vivo MRI.These open new venues for multimodal imaging of large and transparent biological specimens by MRI and LM.

Results
MRI of cleared whole brains-We cleared whole brains of rats and mice by two solvent-based clearing methods, ECi and uDISCO, and by the aqueous-based clearing protocol CLARITY.ECi and uDISCO cleared brains were amber-colored, translucent and of reduced size; with uDISCO cleared brains displaying stronger shrinkage (Fig. 1 and Suppl.1).Inversely, CLARITY cleared brains swell signi cantly (>3 fold in volume) and were highly gelatinous and hyperhydrated (Fig. 1, bottom).These features were all previously described, demonstrating that our procedures achieve equal performances 6,41,42,56 .We proceeded to image cleared brains by T1-and T2-weighted (T1W and T2W) MRI sequences.Rat and mouse control brains (i.e., PFA-xed brains maintained in PBS) showed characteristic MRI contrast (Fig. 1, control) (e.g. 57), whereas very weak MRI signal and contrast could be obtained when imaging the dehydrated uDISCO and ECi cleared brains (Fig. 1 and Suppl. 1, ECi, uDISCO).MRI signals were observed in discrete regions such as the ventricles and at the periphery of uDISCO or ECi cleared brains, which probably resulted from incomplete dehydration of the cerebral cavities.Blurred images were obtained when imaging ECi cleared brains, and this resulted from the samples' minute signals (Figs. 1 and Suppl. 1).On the other hand, and consistent with previous reports 34,36 , CLARITY cleared brains showed very strong and homogenous MRI signal, though completely devoid of MRI contrast (Figs. 1 and Suppl.1; CLARITY).
Our whole-brain MR images show a complete loss of ex vivo MRI contrast in CLARITY cleared brains.Similarly, uDISCO and ECi cleared brains yield no MRI contrast.However, unlike CLARITY, uDISCO and ECi cleared brains also lack any MRI signal.The latter is expected as uDISCO and ECi protocols are highly dehydrating.To conclude, chemically cleared brains show variable MRI signals, depending on their hydration state, however all three treatments abolish ex vivo MRI contrast.
Hydration of solvent cleared brains restores MRI contrast-We were curious as to whether we could obtain an MRI signal by re-hydration of solvent-based uDISCO and ECi cleared brains.We immersed uDISCO and ECi cleared brains in PBS for 72 hrs.This treatment reduced transparency and changed the color of the samples to opaque white.It was also complemented by visible expansion of the samples (Figs.2a, b and Suppl.2).Though initially undesired, as the brains are no longer optically clear, these transformations indicate proper water permeation into the samples, especially into deep regions of the tissue.Water entry into uDISCO and ECi cleared brains was somewhat surprising as these chemical processes signi cantly solidify the samples 38,43,58 .Curiously, and although beyond the scope of this work, this observation demonstrates that hydrophilic-labeling strategies could be achieved after a full uDISCO and ECi clearing.
Importantly-and unexpectedly-retrieval of the MRI signal was accompanied by restoration of the MRI contrast in uDISCO and ECi-brains (Fig. 2a, 2c and Suppl.1).Various brain structures, such as corpus callosum, hippocampus, cortex, striatum, cerebellum, thalamus and lateral ventricles could be easily distinguished in MR-images of hydrated-uDISCO and ECi-brains (Figs. 2a, Suppl.3a).Quantitative assessment of the contrast-to-noise ratio (CNR) of various brain regions showed that the MRI contrast in T1W and T2W images was signi cantly augmented; T1W images showing CNR values on par with those of control brains (Figs.2c and Suppl.3b, c).We therefore examined different hydration durations (i.e., immersion in PBS up to 30 days after clearing) and nd that maximal hydration and MRI contrast could be obtained following 3 days of hydration (Suppl.4).
We were also interested in examining MRI contrast of hydrated uDISCO and ECi-brains in the presence of a high background signal (i.e., in PBS) compared to no background signal provided by the Fomblin Y (Suppl.5).We nd that T1W images showed a similar CNR when imaged in PBS compared to Fomblin Y, whereas CNR of T2W images were moderated possibly due to a limited greyscale range of the T2W images (Suppl.5).We performed an analogous experiment using CLARITY cleared brains to see whether we could recover MRI contrast, as was previously suggested 37 .We therefore immersed the CLARITY cleared brains in a refractive index (RI)-matching solution (2,2′-thiodiethanol; TDE), which shrank the sample 49,59 , but failed to uncover MRI contrast (Figs.2a and Suppl.1; CLARITY in TDE).
Motivated by obtaining ex vivo MRI contrast for hydrated uDISCO and ECi brains, we speculated whether we could enhance it further by adding a contrast-agent.We locally injected one hemisphere of hydrated uDISCO and ECi brains with a Gd 3+ -based contrast agent (GBCA), whereas saline was injected to the other hemisphere as control.T1W images acquired a few minutes after GBCA injection showed an enhanced MRI signal and, more importantly, enhanced MRI contrast, exclusively at the site of injection (Figs. 3 and Suppl.6a).Conversely, though GBCA in CLARITY cleared brains increased the, already extant, MRI signal at the injection site it did not reveal any MRI contrast (Suppl.6b).
GBCAs enhance relaxivity of nearby water protons, producing signal intensity variances across the image.Thus, these results support our above observations that PBS immersion of uDISCO or ECi brains achieves su cient hydration of the samples, which then translates into robust MRI signals.Moreover, whereas MRI contrast is absent in the hydrated (or TDE-immersed) CLARITY cleared brains, it is evident in hydrated uDISCO-and ECi brains.Lastly, GBCA can be readily used to enhance ex vivo MRI signal and contrast of hydrated uDISCO and ECi brains.
Electrophoretic tissue clearing of hydrated uDISCO and ECi brains abolish ex vivo-MRI contrast-Our results show that hydrated uDISCO and ECi brains maintain ex vivo MRI contrast, despite their anticipated delipidating-nature.This observation initially suggested to us that the remaining MRI contrast may be attributed to remaining lipids in the tissues, perhaps because uDISCO and ECi procedures are not as delipidating as CLARITY, as this has never been examined before.We therefore initially turned to try to remove more lipids from uDISCO and ECi cleared brains, as is achieved by CLARITY.The active lipidremoval step consists of placing the tissue within an electrophoretic tissue-clearing (ETC) system in a SDS-containing solution (ETC/SDS) 60,61 .ETC/SDS caused uDISCO and ECi brains to expand, change color to opaque yellowish-white and turn soft to the touch; highly reminiscent of brains that underwent clearing via CLARITY (Figs. 4 and Suppl.7).Notably, the change in color and features of the brain were mainly apparent at the outer layers of the brain, with noticeable "uncleared" interior.Concomitantly, when imaged by MRI, we observed non-uniform reductions in MRI contrast, with the swollen outer layers showing pronounced reductions in MRI contrast the longer ETC/SDS was applied (seen as bright and swollen edges in the outer layers of the brain (Fig. 4, arrowheads)).ECi brains were more readily affected by ETC/SDS treatment than uDISCO brains, losing the MRI contrast in most of the tissue after prolonged ETC/SDS treatment (Fig. 4).Control brains (i.e., PFA-xed brains without hydrogel embedding) showed the most severe loss in MRI contrast following prolonged ETC/SDS treatment (and were also highly fragile and deformed) (Suppl.7, control).These results demonstrate the ability of ETC/SDS-treatment to abolish MRI contrast (but augment MRI signal) in the outer layers of the brains, especially in softer samples.
These observations raise two opposing hypotheses.The rst states that solvent cleared ECi and uDISCO methods are not strongly delipidating and hence the prevailing ex vivo MRI contrast or, secondly, change in the tissue dimension (leading to hyperhydration of the tissue), rather than its lipid content is the reason behind loss of MRI contrast.
Assessing lipid content and composition in cleared brains-To address the rst assumption, namely remaining lipids, we turned to compare lipid-content in cleared brains by various methods.Notably, these data are missing (or are very partial) for most clearing techniques 36,62 .In fact, this missing data is behind the persistent debates regarding the extent of delipidation of each method (see discussion).
We aimed to assess the relative delipidation by staining cryosections of cleared brains with Oil Red O (ORO).ORO is a fat-soluble diazol-dye staining neutral lipids, lipid droplets and cholesteryl esters, but not biological membranes 63,64 .We rst validated the method by staining control and CLARITY-brains (Suppl.8) 61 .Expectedly, cryosections from control brains showed positive ORO-staining, and staining could be reduced (i.e., destaining) by prolonged ETC/SDS treatment (Suppl.8; control-ETC) and CLARITY-brains' cryosections showed very low ORO-staining (Suppl.8).Unexpectedly, cryosections from uDISCO and ECi cleared brains showed extensive ORO-staining, even higher than control, which was amenable to destaining by subsequent ETC/SDS-treatment (Suppl.8; ECi-ETC and uDISCO-ETC).These are somewhat consistent with the hypothesis that uDISCO and ECi may not be as delipidating as considered.
These observations motivated us to test a histochemical Dil-staining protocol on cleared brains cryosections (akin to the ORO-staining protocol).Surprisingly, when applied onto CLARITY, uDISCO or ECi cleared brains cryosections, the DiI-dye labeled all sections to a similar extent as control (Suppl.10, and see discussion); inconsistent with our collection of results.These raised the possibilities that lipophilic dyes may be absorbed by the clearing reagents therefore dissuading us from drawing rm conclusions from cryosection labeling.
We therefore proceeded to analyze lipid-content by Raman spectroscopy (RS) that does not require any labeling.RS is used to measure composition and relative abundance of biomolecules, notably proteins and lipids, by their characteristic chemical bonds [67][68][69] .Though lipids signatures span across the entire RS spectrum 69 , we focused on the ratio between peaks appearing at 2850 cm -1 (representing CH 2 bonds of long aliphatic lipid chains) and 2940 cm -1 (attributed to CH 3 bonds as typically found in proteins) to assess protein/lipid ratio, as previously employed 70,71 , including on a couple of cleared samples 62 .
We performed RS on slices excised from control and cleared brains.Spectra obtained from control brain slices showed the prototypical signature of biomolecules in the brain 72,73 , with the expected lipids and proteins bands at 2850 and 2940 cm -1 , respectively (Suppl.11a; dashed region).We noted that the 2850 and 2940 cm -1 peaks ratio was variable between different regions of the slice, however, with both peaks remaining highly visible (Suppl.11a, inset) 72 .We analyzed CLARITY brain slices at two different time points: after partial (3 hrs ETC) or full clearing (9 hrs ETC), and observed progressive reduction in the 2940/2850 cm -1 ratio; indicative of the relative reduction in proteins compared to lipids (Suppl.11b).Analysis of ECi and uDISCO cleared brain slices showed substantial contribution of the solvents to the spectra overlapping with the 2940/2850 cm -1 bands (Suppl.11c).We extensively washed the cleared brains and slices in PBS and employed the unique peak of the solvent (~3060 cm -1 ) to assess its clearance.Though three days of washing caused a signi cant reduction in the 3060 cm -1 peak, the spectra remained marred with the solvents' signatures (Suppl.11d and e; red plots).Nevertheless, in both instances, the lipid band was signi cantly reduced, indicating reduction in lipids compared to proteins (Suppl.11d and e; insets and black arrows).Hydrophilic clearing reagents used for Scale did not overlap with 2940/2850 cm -1 band 62 .Scale cleared brain slices showed no reduction in lipids (Suppl.11f).In order to compare the ratios between clearing techniques (and bypass potential differences that may arise from acquisition from varying regions), we homogenized the brains.Homogenates from PBS washed-ECi and uDISCO brains showed a sharp attenuation in the 2850 cm -1 band, representing a strong decrease of lipids in the samples, vs. 2940 cm -1 (Fig. 5a, magenta and green), compared to control (Fig 5a , blue).Scale-homogenates did not show any reduction in the protein/lipid ratio (Fig. 5a, red).However-and interestingly-CLARITY homogenates displayed similar behavior as the CLARITY-treated brain slices, namely, exhibited a 2940/2850 cm -1 ratio of ~1 indicating a relatively larger reduction in proteins than lipids (Fig. 5a, orange).
Together, RS indicates a strong reduction in lipids following ECi and uDISCO treatment, but not following Scale.It also suggests that CLARITY may be deproteinating as it is delipidating.However, the relativenature of the method, along with the overlapping spectra of the clearing solvents with the signatures of protein and lipids, may lead to inaccuracies in the quantitative assessment of lipids in cleared brains.
Owing to differences in the assessment of lipid-content by three different methods explored here (ORO, Dil and RS), we decided to proceed with mass-spectrometry lipidomics 74,75 .Notably, despite the method's precision and resolution, it has not been previously applied on cleared specimens.We analyzed control and cleared homogenized mouse brains, in duplicates of 150 and 300 μg per sample (Suppl.12).Controls yielded 53 nmol lipids/mg tissue, with Scale treated samples showing 38 nmol/mg lipids, thereby maintaining 72% of the lipids.CLARITY and uDISCO-samples showed almost a complete loss of all lipids in the brain (0.41 and 0.71 nmol/mg lipids, respectively), namely 0.78 and 1.3% of lipids remaining in the samples, respectively.ECi samples showed slightly higher lipid content, of 9.9 nmol/mg lipids representing less than 19% lipid-content compared to control samples (Fig. 5b and Suppl.12a).Interestingly, both Scale and ECi treatments maintained the lipid composition as in control, with high proportion of phosphatidylcholine (PC), phosphatidylethanolamine (PE), and phosphatidylserine (PS) (Fig. 5b and Suppl.12b).Together, mass-spectrometry lipidomics support our results using the DiIwashout assay (Supple.S9) and RS (Fig. 5a), allowing us to conclude that CLARITY and uDISCO are equally and very strongly delipidating, closely followed by ECi, whereas Scale preserves most (>70%) of the lipids (Suppl.12b).
We would like to emphasize that lipid-content (and lipo-typing) has not been systematically examined in various cleared samples.However, even when lipid-content was assessed, most reports employ a single method (e.g., 47,61,62 ), resultantly, there are very few established methods for lipid assessment in cleared tissues.Our efforts show that different methods may yield varying and, at time, contrasting results.These suggest that lipophilic dyes may be absorbed by remaining clearing agents in the samples, thereby raising awareness towards the challenging use of lipophilic dyes in samples drenched in organic solvents.As systematical comparison between multiple clearing methods has not been reported before, we deem these demonstrations highly valuable.
Loss of MRI-contrast in a lipid preserving Scale-brain Our results demonstrate that the MRI-contrast observed in ECi and uDISCO brains cannot be attributed to remaining lipids, as those are largely extracted (Fig. 5a).Therefore, we turned to corroborate our alternative hypothesis stating that loss of ex vivo MRI-contrast stems from tissue expansion and its subsequent hyperhydration (i.e., yielding a pronounced reduction in lipid/water ratio 35,76 , rather than mere loss of lipids).Hyperhydration and non-delipidation are the two main features of the method Scale, thereby making it ideal to examine our hypothesis.We cleared PFA-xed brains of mouse and rat by Scale for 20 and 90 days, and obtained clear and expanded samples of whitish color and gelatinous texture, as described (Suppl.13, PBS) 47,50 .The features are reminiscent of brains treated by the CLARITY protocol and by ETC/SDS (Fig. 1, 4 and suppl.7).We also tested subsequent Scale clearing of hydrated uDISCO and ECi brains and observed that the tissues displayed similar changes in color and volume (Suppl.13, bottom).Ex vivo MR-Imaging of Scale cleared brains yielded images with reduced MRI contrast, both in the non-treated PFA-xed and in the hydrated ECi or uDISCO brains (Fig. 6 and Suppl.14, compare with Fig. 1, 2a, c).Interestingly, T1W was more susceptible to loss in MRI-contrast than T2W images.Thus, these corroborate the hypothesis that lipids are not the major source of ex vivo MRI contrast, but rather the expansion of the tissue and excessive entry of water along reduction in lipid density and other waterinteracting molecules in each voxel.

Discussion
The integration between different imaging modalities, notably LM and ex vivo MRI, is an attractive means for the development of preclinical and clinical diagnostic tools for large and intact biological specimens.This combination takes advantage of high spatial resolutions and signal speci city proffered by LM, with mesoscale-imaging of any tissue, at any depth, without the need of labeling select targets provided by MRI 1 .This combination also provides unbiased means for image registrations, correction of tissue deformations, orientation within large tissues and validation of MRI signal by targeted uorescent signals 1,30 , to name a few (but also see 12,15,25,61 ).This endeavor, however, requires chemical clearing of the tissue 7,11 , which has so far yielded samples without any ex vivo MRI contrast (Fig. 1 and 15,34,36,38 ).These reports have implicated lipids as the major source for ex vivo MRI-contrast.This lipid-to-MRI contrast concept has gained attention and has even been suggested to provide means to assess extent of delipidation and tissue in-homogeneities by various clearing techniques 15,34 .
Our initial aim was to explore whether other clearing techniques would be compatible with ex vivo MRI.Systematic analysis and modi cations of two clearing techniques show that uDISCO and ECi methods are highly suitable for this endeavor, after these have been hydrated by PBS (Fig. 2).This was highly surprising as both uDISCO and ECi are suggested to be highly delipidating, much like CLARITY.The contrasting observations motivated us to systematically explore these discrepancies by using the conventional lipid dyes ORO and DiI to assess the relative lipid levels.However, these provided partially contrasting results: ORO-staining suggested that uDISCO and ECi cryosections retained neutral lipids, whereas washout of DiI-staining by clearing reagents and procedures showed these to robustly remove phospholipids (Suppls.8-10).Contrastingly, DiI-staining of cryosections prepared from post cleared tissues show no reductions in lipid levels by any of the clearing methods employed.
These results suggested to us that the presence of residual solvents in the cleared specimens may absorb the lipophilic dyes.Subsequent RS quantitative analysis corroborated this assumption, demonstrating the strong signatures of the solvents in ECi and uDISCO cleared brains, even after these have been extensively washed (Suppl.11).Nevertheless, we eventually found (by comparing with our lipidomics assessment) that the DiI-washout assay gave the best indication regarding the delipidatingnatures of uDISCO, ECi and CLARITY and the non-delipidating-nature of Scale (Suppl.9).
Though the use of Dil-staining on cryosections of cleared specimens is not suitable for lipid quanti cation, it did provide some insights about the integrity of tissues treated by CLARITY, uDISCO and ECi (Suppl.10).CLARITY cryosections showed very homogenous staining by Dil, with no de ned brain regions.Furthermore, only CLARITY-cryosections appeared highly porous, whereas in ECi and uDISCO cryosections we did not see this sponge-like appearance and we could easily discern various regions in the sections, such as the hippocampus (Suppl.10).This show that CLARITY critically modi es the tissue and renders it highly porous.
We further tried to quantitate the relative content of lipids following clearing by RS.However, the solvents' signatures spanned the entire spectra; overlapping and overshadowing regions indicative of different lipids (Suppl.11a, c) 67,69 .We also observed regional variability in the protein/lipid signature.
Homogenization of the samples helped bypass some of these challenges and allowed us to speci cally focus on the high wavenumber region to quantify the relative abundancy of CH 2 bonds of long aliphatic lipid chains and CH 3 bonds of proteins, as previously shown 62 .Although this analysis provides but a partial picture of the different lipids and fats in the tissue, it did show that uDISCO and ECi e ciently remove long aliphatic lipids, whereas Scale does not (Fig. 5a).Surprisingly, the results also suggest that CLARITY is relatively deproteinating (Fig. 5a, orange).Though the latter may appear at odds with measurements of phospholipid and protein content in CLARITY cleared brains 33,36 , it is important to recall that RS measurements provide relative measurements rather than absolute amounts in the tissue, and that CLARITY is indeed deproteinating to various degrees 5,7,77 .
To quantitate absolute amounts of all types of lipids in the cleared brains, we eventually performed massspectrometry-based lipidomics.Notably, this has never been demonstrated before.This method conclusively showed the strong delipidating-nature of CLARITY, uDISCO and ECi (removing 99%, 99% and 70% of the lipids, respectively) and, inversely, the lipid-preserving nature of Scale (removing < 20% of the lipids) (Fig. 5b and Suppl.12, Table 1).
These ndings resolve the debate regarding delipidation by Scale ( 47 vs. 7,34 ) but, more importantly, show no association between delipidation and ex vivo MRI contrast (Fig. 7).Strong support was obtained by ex vivo MRI of Scale cleared brains; showing very strong reductions in MRI contrast (>13-fold reduction in CNR compared to control; Fig. 6, 7 and Suppl.14).In support, a recent study using the Clear T method which largely conserves tissue dimensions and lipids, preserves most MRI contrast 34 .These conclusively demonstrate that lipids are not the major source of ex vivo MRI contrast.
What is the source for MRI contrast then?Our concerted observations raise an alternative hypothesis that expansion of the tissue is the primary reason for loss of ex vivo MRI contrast in cleared brains, which is a common feature in CLARITY, Scale, and ETC/SDS treatments (Fig. 1, 4, 6, Suppl.7, 13).More speci cally, MRI contrast of the brain is obtained by variations in the tissue's water-interacting molecules (including lipid)-to-water ratio (or myelin-to-water ratio 35,76 ).Expansion of the tissue increases the water density homogeneously across the tissue with a concomitant reduction in the density of water-interacting molecules, such as lipids (but not exclusively).Indeed, cryosections of CLARITY brains stained by DiI show a highly homogenous and porous tissue (Suppl.10) and MRI scans of CLARITY cleared brain slices showed an increase in the diffusion rates and relaxation times of water molecules, resembling those of free water molecules 36 .To try to mimic this effect, we continuously imaged water-absorbing gel beads (composed of super-absorbent polymers, such as the hydrogel) by MRI (Suppl.15a).When placed in water and immediately imaged by MRI, each small bead produced a strong T2W image (Suppl.15b; i).However, this signal progressively diminished as the beads expanded (Suppl.15b; vi)-ultimately losing all MRI contrast until it was almost completely undetectable by MRI (Suppl.15c, dashed region).Importantly, the beads did not lose any of their content during the process, rather simply underwent isotropic expansion and hyperhydration.We suggest this effect to represent brains undergoing expansion.
In conclusion, we show the feasibility of using cleared brains for ex vivo MRI, by uDISCO and ECi clearing techniques.First, it requires hydration of the specimens to reintroduce water molecules-the source for the MRI signal.Hydration of uDISCO and ECi produces non-transparent samples but, importantly, maintains their structure (Fig. 2C).Further, MRI contrast can be enhanced by addition of contrast-agents to the cleared samples.These capabilities suggest that this procedure should support multimodal imaging by initial LM-imaging of the transparent tissue followed by ex vivo MRI.We also suggest that multiple clearing techniques can be combined and that multiple rounds of LM imaging-hydration-MR imaging and clearing can be performed 38,43 .This protocol should allow users to create atlases composed of registered MRI and LM images of entire tissues, as well as accelerate the development of novel multimodal agents by providing means to quickly validate MRI signals by uorescence imaging 1 .Lastly, we envision that this method could be implemented with LM to examine clinical histological specimens by ex vivo MRI to detect pathologies at various scales 26,37 , analogous to pathology-optimized expansion microscopy 20,78 .

Materials And Methods
Animals: We used 3-weeks to 6 months old, male and female, C57BL/6 mice and Sprague Dawley rats (Envigo, Israel).All animal procedure protocols were in accordance with the guidelines and regulations of the Technion, and were approved by the Animal Care and Use Committee of the Technion -Israel Institute of Technology (Haifa, Israel, Ethic number IL-121-08-19).Animals were transcardially perfused with PBS (02-023-5A, Biological Industries) and neutral buffered 10% formalin (HT5011, Sigma-Aldrich) and the whole-brain was removed and immersed in formalin overnight, then washed and incubated in PBS at 4 °C.Tissue clearing techniques; CLARITY: Procedure was performed using the X-CLARITY™ system and reagents (Logos Biosystems).Brie y, whole-brains were infused with hydrogel monomers at 4 °C overnight (C1310X, Logos Biosystems) and a subsequent hydrogel polymerization step was performed at 37 °C for  41 .
Secondary treatments of cleared brains: of ECi and uDISCO brains was achieved by immersion of the entire samples in PBS at 4 °C for 3 and 30 days.CLARITY brains were immersed in 100 % TDE overnight (Fig. 2).ETC/SDS of cleared brains was performed by transferring subsequently hydrated brains samples into the ETC chamber in ETC solution with the following settings: 1.2 A, 37 °C, for variable durations.For contrast agent injection, 1 µl of saline or 0. (FLASH) sequence with the same geometry and resolution as the T2W images and with TR/TE = 320/4 ms, 30 o pulse, number of averages = 6.For GBCA injections (Fig. 3), additional protocol of 3D T1W FLASH images were acquired, with the same geometry and resolution as before, and with TR/TE = 15/4.4ms, 15 o pulse, number of averages = 2.
Data processing was performed using Medical Image Processing Analysis, and Visualization (MIPAV) software (NIH).Regions of interest (ROIs) were manually selected and traced (Fig. S3B).The MRI regional contrast was calculated as the signal intensity difference with the thalamus.Signal was normalized to the background (i.e., noise) for quantifying the CNR.Intensity pro les were obtained by tracing a horizontal line over greyscaled images and plotted as intensity of voxel of interest (VOI) along trajectory of the line (i.e., distance).
Cryosection preparation: Control and cleared brains were sequentially immersed in 15, 20 and 30 % sucrose in PBS, for 12 hrs in each solution at 4 °C.Brains were embedded in OCT and froze at -80 °C.
Tissue coronal sections of 30-µm thickness were cut on a cryostat and mounted on microscope slides (SuperFrost Plus, Thermo Scienti c).To note, unwashed clear ECi and uDISCO brains disintegrated during cryosectioning, and the prolonged wash with sucrose in PBS facilitated the sectioning.Yet, CLARITY treated tissues were di cult to section smoothly at <30-µm, and washed-uDISCO treated tissues were more di cult to adhere to the slide.
Slides were mounted (Immu-Mount, Thermo scienti c) and scanned by an automated slide scanner (Pannoramic 250 Flash III, 3D Histech LTD).Histology (Suppl.8) and uorescent (Suppl.10) images were converted to 32-bit greyscale, artifacts and background were excluded using Threshold tool (ImageJ, NIH), and intensity was calculated for each image.For relative ORO staining, the grey intensity of each slice was normalized to the averaged grey intensity in images from PBS control group, in each experiment.
DiI washout assay: A 1 µl from 1 mg/ml DiI was dropped on each hemisphere of 3 mm thick coronal brain slices.Dehydration by tert-Butanol or ethanol + 2 % Tween-20 was stepwise for 4 days, as described above for uDISCO and ECi.The concomitant PBS control in the other hemisphere went through the same conditions of temperature, shaking rate and incubation time.CLARITY was after 3 hrs ETC, and Scale treatment was for 7 days at RT. DCM and ECi treatments were both for only 1 hr at RT.
Cleared brain homogenization: PFA-xed mouse brains were weighted and cleared; Scale treatment was for two weeks on a chopped tissue, to facilitate the diffusion of Scale reagents.The CLARITY protocol was on a whole brain until it became completely clear.ECi and uDISCO cleared brains were extensively washed with PBS for 72 hrs.PBS was added to all the samples to reach 100 mg/ml (wet tissue weight before clearing).Samples were homogenized using a tissue homogenizer.
Mass-spectroscopy lipidomics: Two 30 μl samples from each homogenated brain, at 5 and 10 mg/ml, were taken for the lipidomic analysis.Lipid extraction was using a two-step chloroform/methanol procedure.Samples were spiked with lipid class-speci c internal standards prior to extraction.After extraction, the organic phase was dried and resuspended in mass-spectroscopy acquisition mixture.The lipid extracts were subjected to mass spectrometric analysis, as previously described 74,75,79,80 .
Raman spectroscopy: Raman spectra were acquired using an upright Confocal micro-Raman Microscope (LabRAM HR evolution, HORIBA scienti c).Samples were illuminated by 532 nm laser (100 % laser intensity) through a 50× objective (LWD).The acquisition time was 10 s /point (except the spectrum in Fig. S11A that was acquired at 100 sec / point) with 3 accumulations.Brain homogenates (100 mg/ml), thick tissue slices excised from cleared and control brains, and a drop from ECi solution and uDISCO RImatching solution mix were sampled on Quartz slides (042295, Alfa Aesar).For CLARITY slices at different ETC times, the same brain was sampled at 3 and 9 hrs ETC.

Declarations
Table 1 Table 1 is available in the Supplementary Files section.Figure 6

Figures Figure 1 MRI
Figures

Figure 4 ETC
Figure 4