Introduction

Light microscopy has led to numerous breakthrough discoveries in the biological sciences. However, a persistent limitation to optical imaging is the degradation of image quality at increasing depths. The main sources of degradation are absorption, aberration, and scattering. Scattering has been shown to contribute more to the reduction of light transmission than absorption at the visible and near-infrared wavelengths used for fluorescence microscopy¹. Adaptive optics can address sample-induced, bulk aberrations which are spatially and temporally stable^2,3. However, adaptive optics cannot correct for scattering⁴. Scattering is a result of the fine-scale heterogeneity of refractive indices within the sample arising from mismatches between the extracellular fluid (ECF) and wavelength-scale cellular components⁵. ECF has a refractive index of 1.34, similar to cerebral spinal fluid (CSF), while intracellular components have higher refractive index values (cortical cytoplasm = 1.353-1.368, mitochondria = 1.38-1.41, nucleus = 1.39-1.43)^6–8. In neural tissue, the many fine membrane-bounded neurites are an additional contributor to scattering.

In fixed tissues, there are numerous techniques for increasing transparency^9–25. While use of such clearing techniques has provided unprecedented optical access of the brain, fixation and the harsh processing steps such as dehydration, lipid extraction and rehydration of tissue with a solution of higher refractive index are incompatible with living tissue. One previous in vivo technique may provide modest improvement, but it can only be seen in group averages²⁶. A more recent method has not been shown to be compatible with brain imaging²⁷.

Here we describe an approach that reduces scattering by raising the refractive index of ECF while still maintaining neural health and function. We designed a custom high-refractive index isoosmotic artificial cerebrospinal fluid (iodixanol-ACSF or I-ACSF) that preserves much of the ionic balance necessary for normal neuronal function. We first demonstrate that I-ACSF is capable of significant clearing in brain slices when viewed by transmission microscopy. To examine its effects on neuronal function, we continuously perfused clearing agent into the ventricles while the mice underwent a battery of behavioral tests. To further characterize any perturbations of activity within a neuronal circuit, we performed comparative multi-electrode array recordings of the retina in control and I-ACSF solutions. Finally, to demonstrate the utility of this technique for neuronal imaging, we characterized I-ACSF’s improvements in resolution and imaging depth of fluorescently-labelled neurons using confocal and two photon microscopy. We thus demonstrate a straightforward method of altering the optical properties of living neuronal tissue that reduces scattering and improves image quality. We anticipate that in vivo clearing can push us closer to bridging the gap between neurophysiology and animal behavior.

Results

Evaluation of tissue clearing capability

To obtain the greatest increase in transparency for a given concentration of clearing agent, we focused on biocompatible organic compounds containing a large number of heavy ions, as the inner-shell electrons endow the material with a large dipole moment. These characteristics are satisfied by radiocontrast agents^5,19,28,29. We focused on iodixanol, a non-ionic dimer of iodinated benzenes that can be formulated to be isotonic to plasma³⁰. Iodixanol is clinically-approved for human use and is routinely used for X-ray based visualization of vasculature. Iodinated contrast agents have been used previously to modulate refractive index. For example, contrast agents have been added to immersion media to match its refractive index to that of ex vivo cleared samples (SeeDB2, ACT-PRESTO, SWITCH). More recently, iodixanol has been used as a non-toxic refractive index matching fluid for live specimens such as planaria and zebrafish embryos³¹. Encouragingly, Boothe et al. showed that iodixanol did not have a deleterious effect on the growth and survival of these organisms at concentrations relevant to physiological refractive index matching. However, all of these efforts were directed at reducing the bulk mismatch in refractive index between the sample and the exterior immersion media, in much the same way that oil immersion fluid matches the refractive index of glass coverslips.

Here, our intent is to correct for finer inhomogeneities in refractive index within a sample that lead to scattering and cannot be easily corrected using tools such as adaptive optics. This requires index-matching in the extracellular fluid throughout tissue, for which we designed I-ACSF. The composition I-ACSF (see Methods) was designed to provide a high refractive index of 1.385 while satisfying the conflicting demands of maintaining an osmolarity matching that of normal ACSF (310 mOsm), an ionic milieu that perturbs Nernst potentials by no more than 10mV, and avoidance of hyponatremia.

We first tested whether I-ACSF increases optical clarity by measuring its effect on tissue transmittance. We immersed 200 µm thick slices from the mouse brain in control artificial CSF (ACSF, n = 1.34) or I-ACSF. We then overlaid the slices on an opaque edge of a 1951 USAF resolution target to image the transmission through the slice. On a qualitative, macroscopic scale, we observed that finer patterns were visible through the I-ACSF treated slices while not resolvable in equivalent thickness slices in ACSF (Figure 1a). 10 µm sized features were easily resolved through 200 µm of I-ACSF treated tissue, while we struggled to resolve features double this size through ACSF-treated tissue (Figure 1b). Only heavily myelinated structures such as the corpus callosum and internal capsule failed to show substantial clearing, which is expected since I-ACSF is aqueous and the refractive index of the lipid-rich myelin is higher than that of gray matter³². When we increased the concentration of iodixanol without attempting to maintain physiological osmolarity to obtain a refractive index of 1.42, there was an increase in transparency in myelinated regions but at a cost of tissue expansion, highlighting the tradeoff between transparency and maintaining physiologically relevant parameters (Supplemental Figure 1a). Transmission through a slice thickness of 200 µm is a proxy for the total optical path length encountered during epi-illumination at depths of 100 µm (Figure 1c), which is approximately the limit for high-quality imaging by single-photon techniques^33,34.

Figure 1

To quantify the improvement in clarity, we calculated the modulation transfer function (MTF), which measures the degradation of image quality for both coarse- and fine-scale features. The transfer function rapidly dropped below 5% of its initial value at a feature size corresponding to 12.5 µm (40 lp/mm), indicating a loss of cellular resolution once a photon travels through 200 µm of tissue (Figure 1d). On the other hand, the MTF measured through the I-ACSF treated tissue had a more gradual drop-off and at 100 lp/mm (equivalent to 5 µm feature), the curve remained above 20% of the initial value. These values approximately correspond to what we observed visually with the resolution target (Figure 1a,b). We also observed that the clearing effect can be reversed by washout with standard ACSF (Figure 1d).

To separate the effect of iodixanol clearing on scattering from its effect on aberrations, we also imaged the tissue with a low numerical aperture (NA) objective (NA = 0.082). The larger angle light rays accepted by a high NA objective are refracted more than those closer to the optical axis. By using a low NA objective, we reduced the impact of aberrations. We found that the MTF curve drop-off was always more gradual for the cleared tissue, irrespective of objective parameter or imaging system, demonstrating that iodixanol treatment reduced scattering (Supplementary Figure 1b). We conclude that I-ACSF treatment increases light transmission through tissue and has potential as an in vivo clearing agent.

Behavioral testing of iodixanol treated mice

One potential concern with in vivo clearing is whether it perturbs neuronal function. To test for large-scale impacts on sensory, motor, and/or cognitive function, mice receiving a brain-wide infusion of iodixanol through the ventricular system were subjected to a battery of behavioral tests. We compared the performance of control mice infused with ACSF with those infused with Visipaque 320, a formulation of iodixanol designed to be isotonic to blood plasma (n_ACSF = 10, n_Iodixanol = 12). This characteristic of Visipaque enabled us to deliver it directly into the brain while minimizing the risks associated large osmotic imbalances⁸. The higher iodixanol concentration in Visipaque (420 mM) compared to concentration in I-ACSF (184 mM) served as a conservative upper bound on any behavioral abnormalities from ventricular infusion of this compound.

Using a cannula placed into the right lateral ventricle and an implanted osmotic pump, Visipaque was continually delivered at a rate approximately equivalent to 40% of the total CSF production rate in mice³⁵. To reduce the dilution of iodixanol by CSF and the risk of excess intracranial pressure, we co-infused with acetazolamide (AZ), a carbonic anhydrase inhibitor shown to reduce CSF production by up to 60%³⁶. After approximately 24 hours of infusion, both groups of mice were still mobile, eating, drinking and defecating and appeared to be in good general health (Supplementary video 1a,b). At this point, exploiting iodixanol’s use as a radiocontrast agent, we imaged the mice using X-ray computed tomography (CT), to confirm the presence of iodixanol in the lateral ventricles in each mouse used for behavioral testing (see below). To ensure equal treatment, mice receiving ACSF-AZ were also imaged.

After confirming infusion, following CT imaging and at least 30 minutes after the end of isoflurane exposure, the same groups of mice underwent a battery of behavioral tests performed by an observer blinded to the contents of the pump. Intact, non-infused mice were tested in addition to observe any change in behavior related to the surgery, the implantation of the cannula and the pump, or the ventricular infusion (n_control = 10). The infused mice were tested with the cannula and osmotic pump in place. The mice were tested on locomotor activity/exploratory behavior, sensorimotor measures, and olfactory preference, using a set of tasks selected to be thorough yet accomplishable within a single day of testing. All mice survived the infusion, CT imaging and behavioral testing.

Unsurprisingly, the results from the behavioral tests revealed some differences between unmanipulated animals and infused animals. We also noted further surgery-independent differences between ACSF-infused mice and iodixanol-infused mice; these differences generally depended on whether the task was “motivated” or “spontaneous.”

For assays of “motivated” activity, the presence or absence of clearing agent was less important than the effect of surgery (Figure 2a). No differences in performance were detected between ACSF-infused and iodixanol-infused mice among 9 sensorimotor measures. In 7 of 9 of these same measurements, no differences were observed in comparisons between non-surgery and iodixanol-infused mice. Iodixanol-infused mice performed comparably to the non-surgery mice on time spent on a ledge, pole climb down time, turning and climbing down a pole, and time hanging on to 60-degree, 90-degree and inverted screens. ACSF-infused mice also performed comparably to non-surgery mice in 7 of 9 sensorimotor assays. Differences between the non-surgery mice and the infused mice were observed in time on an elevated platform (p_C-A = 0.0012 and p_C-I = 0.0007), time to climb a 90-degree inclined screen (p_C-A = 0.00005 and p_C-I = 0.0008), time to climb a 60-degree inclined screen (p_C-I = 0.001), and time hanging upside down on an inverted screen (p_C-A = 0.01), indicating some effect on balance, coordination, speed of movement, and strength due to surgery, pump implantation, and/or fluid infusion.

Figure 2

“Spontaneous” activity measures were reduced in the both groups of infused mice compared to the non-surgery control; however, this effect was greater in the iodixanol-infused mice (Figure 2b). When examining such “spontaneous” behavior, no differences were observed between groups in time to initiate movement out of a circumscribed space. However, following a thirty-minute locomotor activity test, ACSF-infused mice had less ambulatory activity (p_C-A = 0.0120), fewer rearings (p_C-A = 10^-5) and fewer entries into the center of the field (p_C-A = 0.0121) compared to unmanipulated mice. Iodixanol-infused mice had additional reductions in activity in these categories. Total ambulations (p_C-I = 4×10^-11 and p_A-I = 10^-8), number of rearings (p_C-I = 7×10^-11 and p_A-I = 10^-6), time spent in the center (p_C-I = 10^-6 and p_A-I = 9×10^-7), entries into the center (p_C-I = 10^-9 and p_A-I = 2×10^-8) and distance traveled in the center (p_C-I = 4×10^-9 and p_A-I = 2×10^-8) were reduced in iodixanol-infused mice compared to both unmanipulated and ACSF-infused mice. Overall, in empty cages the iodixanol-infused mice spent less time moving than the other groups.

Another spontaneous task, exploration of a hole-board, likewise revealed a substantial effect of the surgery, but the differences between ACSF and iodixanol-infused mice were smaller or nonexistent (Figure 2c). Total hole poking frequency was reduced in both infused groups compared to the non-infused mice (p_C-A = 0.0003, p_C-I = 5×10^-6 and p_A-I = 0.030, the latter not significant when corrected for multiple comparisons), but no difference was found in hole poke duration. Altogether, these results suggest few differences for “motivated” behaviors but show evidence for limited iodixanol-mediated decrease in “spontaneous” behaviors such as exploration of the test cambers. Crucially, despite the manipulations, all animals were alive, were alert, and were capable of performing all of these behavioral tasks.

After behavioral testing, we quantified brain concentrations of iodixanol-infused mice by calibrating our CT images (Figure 3a) with a set of iodine standards (Figure 3b,c). The average concentration of iodixanol delivered into the brain tissue (excluding the Visipaque-filled ventricles) after 24 hours was 5.6 mM of iodixanol, which can be estimated to be equivalent to 28 mM of iodixanol in ECF. This estimate assumes that iodixanol does not enter the cell, and 20% of cortical volume consists of extracellular space^37,38. Since we observed variability among iodixanol concentration and behavioral results, we examined whether performance could be explained by iodixanol concentration. We found no correlation between the concentration of iodixanol measured in the brain and performance on any behavioral test (p > 0.05, Figure 3d and Supplementary Figure 2).

Figure 3

Evaluate effect on circuit physiology

To further characterize the effects of clearing on neuronal function, we performed extracellular spike recordings from an intact circuit. Extensive work in the visual system has shown that the developing retina generates spontaneous activity patterns (i.e., retinal waves), which instruct circuit refinement downstream³⁹. The spatiotemporal properties of retinal waves have been well-characterized and shown to be highly sensitive to pharmacological perturbation. Using multielectrode array recordings, we compared spontaneous activity in the postnatal day 5 to 7 ganglion cell populations in ACSF and I-ACSF solutions. Most importantly, waves involving the firing of large fractions of ganglion cells on the array persisted in I-ACSF (Figure 4a,b). Furthermore, although the frequency of retinal waves was reduced in I-ACSF (Figure 4d, P = 0.002), overall firing rates of ganglion cells (Figure 4c, P = 0.3), their firing rates during waves (Figure 4e, P = 0.4), and spatiotemporal patterns of activity propagation (Figure 4f, P = 0.7) were unchanged. By contrast, the commonly used anesthetic isoflurane (2%) suppressed most ganglion cell firing (Figure 4g-i, P = 0.04) and abolished retinal waves (Figure 4h, P = 0.04). Thus, iodixanol has comparatively mild effects on network activity patterns in a developing circuit.

Figure 4

Single photon fluorescence imaging of cleared living neuronal tissue

Multiphoton microscopy has been widely embraced for imaging biological tissue due to the ability to image with reduced scattering. However, single photon imaging techniques such as light sheet microscopy and light field microscopy have different strengths compared to point scanning approaches used by standard two photon microscopes. Many super resolution imaging techniques are single photon techniques as well, and such techniques could benefit from a reduction in scattering. To characterize improvements in fluorescence image quality due to clearing, we imaged living tissue slices with GFP-labelled neurons with confocal microscopy, a single photon imaging technique which ordinarily experiences severe depth limitations due to scattering. Since slices are mostly flat, they are less susceptible to bulk aberrations compared to a curved surface like the brain. In addition, slices do not have the outward flux of CSF counteracting the diffusion of iodixanol as in a live mouse. Since clearing performance depends on local tissue properties, for the most direct comparison, we imaged the same region of the same tissue in ACSF and I-ACSF by gradually replacing the media without perturbing the tissue while maintaining identical imaging parameters. To accommodate the unconventional refractive index (1.385) of the immersion solution, we used an off-the-shelf, variable immersion objective (HCPLAPO IMM CORR CS 20x NA 0.7 WD 0.26, Leica) and determined the correction collar setting with the best performance in our clearing media by measuring the point spread function of sub-diffraction fluorescent beads (Supplementary Figure 3a and Methods). Once the correction collar was set, GABAergic interneurons in a GAD65-GFP mouse⁴⁰ were imaged in coronal slices, focusing on the cortical region.

We were able to achieve sub-cellular resolution substantially deeper in cleared tissue than in untreated tissue (Supplementary Figure 3b,c). When imaging at 110 µm depth, the same cell bodies were over ten-fold brighter, and submicron-sized processes could be visualized only when imaged with I-ACSF superfusion (Supplementary Figure 3d). Clearing as a function of time was monitored in both ACSF and I-ACSF. Clearing was not observed in ACSF-incubated tissue at longer incubation time, indicating that the process was not due to change in tissue health over time. Reduced scattering had a large impact on image intensity, as the laser power could be reduced by a half to two-thirds in I-ACSF compared to ACSF, indicating more efficient transport of photons into and out of the tissue (Supplementary Figure 3d).

Two photon microscopy of the cleared living mouse brain

Two photon microscopy has been invaluable for performing in vivo imaging and allowing access to deeper regions of the brain. Two photon microscopy enables deeper imaging by utilizing the closely timed arrival of two photons of approximately double the typical excitation wavelength for a fluorophore. Longer wavelengths undergo less scattering, and when used in point-scanning the emitted photons do not need to be imaged since the large majority of collected photons can be attributed to the objective’s focal spot. However, the loss of intensity caused by the attenuation of light from scattering is still the limiting factor for achievable depth in two photon microscopy^4,41.

To investigate the utility of iodixanol clearing for in vivo applications, we first performed structural imaging in an anesthetized mouse. To reduce the effect of experimental variability (including damage from microinjection), the mouse was head-fixed under the microscope for imaging while either ACSF or I-ACSF was superfused over the exposed brain, and the same brain region was imaged before and after treatment. While maintaining identical imaging parameters across the two conditions, we observed that the same GFP labelled neurons were dramatically brighter after superfusion with I-ACSF (Figure 5a). The signal to background ratio of the neurons increased by two to three-fold after two hours of I-ACSF superfusion (Figure 5b). We could image neurons deeper after I-ACSF treatment at laser powers of 25 mW, measured at the front of the objective (Figure 5c). The clearing effect increased over the course of two hours (Supplementary Figure 4a), indicating that the increase in signal intensity cannot be attributed solely to a correction of bulk aberrations. These changes were not observed after superfusion with only ACSF for similar durations, indicating that the clearing effect cannot be attributed to superfusion time (Supplementary Figure 4b). The clearing effect was reversible and repeatable on the same mouse if the window was sealed between imaging sessions.

Figure 5

Next, we proceeded to examine whether iodixanol clearing can be used with calcium imaging. Imaging was performed on mice with genetically encoded GCaMP6f expression in excitatory neurons. The mice were also injected with AAV1.CAG.tdTomato.WPRE.SV40 to provide a fixed anatomical reference across conditions. To stabilize the brain against movement while allowing rigorous quantification via before/after comparisons, we replaced the conventional glass coverslip with a circular piece of titanium mesh that permitted both imaging and direct fluid access to the brain (Figure 6a). After surgery, we imaged mice with ACSF or I-ACSF superfusion. We observed small amounts of lateral movement in the mice, which could be corrected with cross correlation-based registration. To characterize any effect of iodixanol treatment on neuronal function, we compared activity during superfusion with ACSF and I-ACSF.

Figure 6

Neuronal activity persisted in the presence of iodixanol treatment (Figure 6b,c, Supplemental Video 2a,b). The same neurons could be identified and imaged with higher mean fluorescence intensity after superfusion with I-ACSF (Figure 6d). Under the same laser intensity that allowed imaging at 430 µm depth during I-ACSF superfusion, most of the same neurons could not be visualized during ACSF superfusion (Figure 6e). Calcium activity could continue to be observed at time points exceeding three hours after start of I-ACSF superfusion. We conclude that I-ACSF provides dramatic enhancement in image quality and intensity in living, functional neuronal tissue.

Discussion

Much of our understanding of the brain relies on observing neurons in their intact, functional state. Here, we described a clearing method that enables improved imaging resolution and depth in the intact mouse brain without compromising the ability to record activity. Chemically manipulating the properties of live tissue is a balancing act between keeping the tissue healthy and achieving the desired effect, in our case, introducing a high concentration of an extraneous compound to optically clear tissue. Our technique is not the first attempt at clearing live tissue. Topical application or dermal injection of hyperosmotic sucrose, glycerol and polyethylene glycol solutions have been shown to increase skin transparency for imaging blood flow^42–45. Similarly, a technique to clear the skull enabled the imaging of underlying cortical vasculature and fluorescently labelled dendrites and microglia^46,47. However, none of these preserves the ionic and osmotic conditions needed for healthy neuronal tissue.

To test our method while reducing preparation-to-preparation variability, we chose to perform acute imaging with direct I-ACSF superfusion. We found that diffusion can deliver enough I-ACSF to reduce refractive index mismatch and quantitatively reduce scattering within fresh tissue. However, we have no reason to believe that superfusion is special, and other delivery methods may prove equal or superior for regular use.

In addition to enhanced spatial resolution, a major advantage of our technique is the improved brightness and/or ability to reduce laser power. In living samples, minimizing total photon dose is the primary means to reduce phototoxicity and thermal damage⁴⁸. Ultimately, the ability to tune the refractive index to reduce scattering in vivo is applicable to many other optical imaging techniques since scattering is pervasive and difficult to correct computationally.

While this manuscript was in initial preparation, an alternative approach to in vivo clearing was reported²⁶. This alternative approach, called MAGICAL, involves supplementing the drinking water of the mouse with glycerol. While this mode of delivery has the benefit of simplicity, the size of the clearing effect is much smaller than we report here. With MAGICAL, the effect of the clearing cannot be seen on an individual animal to animal basis, and the modest effect emerges only after group analysis. In contrast, we demonstrated dramatic improvement that was systematic across animals. We also established that our technique is compatible with calcium imaging and does not result in gross behavioral impairment.

Developments in imaging methods will also contribute to our ability to image deeper. Increasing achievable imaging depth continues to be an active area of research^49,50. Because our technique is independent of the particular imaging hardware, we believe our clearing approach can act in a complementary manner with these deep imaging techniques, and thus have a diverse effect on the field and further extend our collective understanding of biological processes.

Methods

Development of clearing agent

A suitable compound for a clearing agent for living tissue should be non-toxic and capable of raising the RI substantially at a low concentration in order to minimally alter the composition of ECF. For wavelengths of visible light, RI of a non-magnetic substance can be approximated as the square root of relative permittivity⁵¹. The Clausius-Mossotti equation relates permittivity to molecular polarizability, with elements that are more polarizable having higher refractive index values Generally, larger atoms have greater polarizability since their electrons are more numerous and further from the nucleus. However, many heavy elements are known to be neurotoxic and are unsuitable for our purposes.

Iodine-based compounds are commonly used as intravenous radiocontrast agents since the absorption cross section increases as a function of the third power of the atomic number. Due to its utility in medical x-ray imaging, there has been focus on creating minimally-toxic iodine-based compounds. Specifically, since side effects arise from injecting hyperosmotic solutions into patients, research has gone into generating compounds with multiple iodine atoms per molecule so that fewer molecules have to be introduced for the same degree of contrast⁵². Similarly, this molecule will increase refractive index with a smaller effect on osmolarity.

Iodixanol (trade name: Visipaque 320, GE Heathcare) is a hydrophilic, non-ionic, iodinated dimer with six iodine atoms per molecule that is used clinically as a radiocontrast agent. The main advantage of Visipaque is that with an osmolality of 290 mOsm, it is isotonic to plasma, thereby reducing toxicity⁵³. Since it is non-ionic, the iodine does not dissociate into ions but instead remains covalently bound to its molecular cage. Iodixanol has shown to be very stable^54,55. The I-ACSF recipe was designed as a mixture of iodixanol and standard components of ACSF to increase the refractive index to 1.385 (184 mM iodixanol). To maintain the same osmolality as standard ACSF, the sodium chloride and glucose concentrations were reduced (see below). The refractive index values of the prepared media were measured by an Abbe refractometer. The osmolarity of the media was measured by a vapor pressure osmometer calibrated with a 290 mOsm standard (Wescor).

The cost of iodixanol can be a factor for experiments which require continuous superfusion during imaging. In addition to its clinical formulation, iodixanol can be purchased in the form of Opti-Prep (Sigma Aldrich), used as a density gradient media. An estimate of the iodixanol cost is $48 per 100 ml of I-ACSF with a RI of 1.385.

Transmission imaging of tissue

C57BL/6 (Jackson Laboratory, RRID:IMSR_JAX:000664) and GAD65-GFP adult mice were used. All experimental procedures involving animals were approved by the Washington University Animal Studies Committee.

The following solutions were prepared: ACSF, containing NaCl 125 mM, KCl 2.5 mM, CaCl₂ 2 mM, MgCl₂ 10 mM, NaHCO₃ 25 mM, NaH₂PO₄ 1.25 mM and glucose 25 mM; I-ACSF, containing the same concentrations of all compounds except NaCl 45 mM, 0.184 mM iodixanol and glucose 10 mM. Both media were equilibrated by bubbling with 95% O₂/5% CO₂.

To collect brains for slicing, the mice were euthanized by CO₂. After decapitation, the brain was rapidly removed from the skull and sliced with a vibrating microtome (Leica VT1200S) while submerged in ice cold ACSF. The slices were incubated in ACSF at room temperature before imaging.

The slices were transferred into a chamber and placed on a resolution target (1951 USAF, Edmund Optics). The slices were superfused with either ACSF or I-ACSF while held in place with a harp. The slices were imaged with brightfield transmitted illumination using a Leica MZ75 using a 1x air objective (Plan NA 0.082, Leica) or an Olympus BX61 using a 10x water immersion objective (UMPLFL NA 0.3, Olympus). Light transmission through tissue overlaid on an opaque edge on the resolution target was imaged. The mean intensity in the direction orthogonal to the transparent-opaque boundary was calculated to estimate an edge spread function. The data was low pass filtered by applying a 61-point, cubic Savitzky-Golay filter along the axis perpendicular to the resolution target edge and differentiated numerically to obtain the approximation of the line spread function. The one-dimensional fast Fourier transform of the line spread function was plotted against spatial frequency to obtain the MTF curve.

Stereotactic surgery and infusion

C57BL/6 adult male and female mice (Jackson Laboratory), 9-12 weeks old, were used for behavioral studies. Acetazolamide (Sigma Aldrich) was added to Visipaque (320 mg I/ml) or ACSF to form a 0.1 mM AZ solution. An osmotic pump (model 2001D, 8 µl/h, Alzet Corporation) was filled and attached via vinyl catheter tubing to a guide cannula (Plastics One). The pump-cannula setup was primed in phosphate buffered saline for 3 hours at 37°C before surgery.

Mice were anesthetized with an intraperitoneal injection of a ketamine xylazine cocktail (100 mg/kg ketamine, 10 mg/kg xylazine) and given carprofen (5 mg/kg, sc) and dexamethasone (2 mg/kg, ip). The mouse was mounted onto a stereotactic stage (Kopf Instruments) and given isoflurane (1-2%), and a midline incision was made to expose the skull. The head was leveled and the cannula was placed in the right lateral ventricle (−0.4 mm posterior to bregma, +1.0 mm lateral to the midline, −2.5 mm ventral to the skull surface) and secured with dental cement. The pump, connected to the cannula with catheter tubing, was implanted subcutaneously in the back of the mouse. For infusions longer than one day, the pumps were replaced every 24 hours.

CT imaging

Static CT imaging was performed by the Washington University Small Animal Imaging Facility. Visipaque and ACSF infused mice were imaged in a small animal Inveon PET/CT scanner (Siemens Preclinical Solutions), placed nose to nose along the bore of the scanner under 2% isoflurane anesthesia. Cannulas made of PEEK tubing were used for all mice undergoing CT imaging to minimize beam hardening artifacts. Images were reconstructed at 50×50×90 µm voxel size. Calibration standards (0, 4.2, 21, 42, 63, 105 µM iodixanol) were imaged using the same imaging protocol. To determine tissue iodine concentration, we registered CT image stacks from each animal to a template stack then quantified the radiodensity of brain regions. We excluded the ventricles in this measurement to limit the quantification to iodixanol content in parenchyma. Data analysis was performed with Matlab (Mathworks, RRID:SCR_001622) and ImageJ (NIH, RRID:SCR_003070).

Behavioral testing

Directly following CT imaging, and at least 30 minutes after the end of isoflurane exposure, mice underwent behavioral testing by the Washington University Animal Behavior Core. This sequence of events allowed us to limit the tested mice to those that were confirmed as being infused with iodixanol. Three groups of mice (n_non-infused = 12, n_ACSF-infused = 10, n_{I-ACSF-infused} = 10) were tested for any gross differences in behavior by a blinded observer. Non-infused B6 mice were tested to observe any change in behavior related to the surgery, the implantation of the cannula and the pump, or the infusion. The behavior tests included in this battery were performed as described previously^56,57.

Multielectrode recordings of the retina

Postnatal day 5 to 7 (P5-7) mice were dark-adapted (>1 hr), anesthetized with CO₂, decapitated, and their retinas isolated under infrared illumination. We recorded large ensembles of retinal ganglion cells on planar arrays of 252 electrodes arranged in a 16 x 16 grid with the corner positions empty (electrode size: 30 μm, electrode spacing: 100 μm, Multi Channel Systems). Spontaneous activity was recorded in darkness from retinas perfused with warm (30-33° C) ACSF, I-ACSF, or ACSF with 2% isoflurane at 5-7 mL min^-1. Signals of each electrode were filtered (300-3,000 Hz) and digitized at 10 kHz. Signal cut-outs from 1 ms before to 2 ms after crossings of negative thresholds were recorded to hard disk together with the time of threshold crossing (i.e., the spike time). We sorted spikes into trains representing the activity of individual neurons by principal component analysis of spike waveforms (Offline Sorter, Plexon, RRID:SCR_000012). When the activity of a single neuron was recorded on more than one electrode (identified by cross-correlation), we retained only the train with the most spikes in our subsequent analysis.

Spontaneous bursts of activity were identified for single units as events of >4 consecutive spikes separated by interspike intervals of <0.4 s if the Poisson probability of the respective spike train was <10^-4. Single neuron bursts were identified as parts of retinal waves if the Poisson probability of spike trains across all other units given the number of units and their average firing rates was <10^-4. Spike time tiling coefficients (STTCs) were computed as described by Cutts and Eglen (2014).

Confocal imaging of live tissue

Confocal imaging was performed using an Olympus FV500 with a 20x multi-immersion objective (HC PL APO IMM CORR CS NA 0.7, Leica) and an argon-ion 488 nm laser for excitation. 200 µm coronal slices from adult mouse brains with a subset of neurons labelled with GFP (Gad65-GFP) were prepared in a manner similar as to what was described for transmission imaging. The slices were placed in an imaging chamber consisting of an o-ring attached to a resolution target. For a comparison of clearing, we imaged the same region of the same tissue in ACSF and I-ACSF by gradually replacing the media while the tissue remained under the objective. To maintain tissue health, ACSF and I-ACSF were continuously bubbled with humidified 95% O₂/5% CO₂, in-line filtered and constantly recirculated during imaging. The focal plane was corrected after the objective correction collar was adjusted as appropriate for the media. Stacks were acquired after waiting 30 minutes after changing between media. Stacks were imaged with 1.5 µm axial step size. Due to concerns about evaporation increasing osmolarity during imaging, samples of solution from the imaging chamber were collected throughout the session to monitor osmolarity. Osmolarity of the media was stable and remained under 320 mOsm throughout imaging. Transmission images using the illumination wavelength as used for fluorophore excitation were also obtained after fluorescence imaging in ACSF and I-ACSF, respectively, for direct comparison of clearing as measured by the two light microscopy techniques. If noted, maximum intensity projections are shown to minimize the effect of focal plane shifts due to changing the immersion media refractive index and adjusting the objective for spherical aberration. Images are displayed after a square root transform to compress the dynamic range of intensities.

Two photon imaging

Imaging was performed with a Bruker multiphoton microscope equipped with a 10x multi-immersion objective (XLPLN10XSVMP, 0.7 NA, Olympus). Excitation was provided by a Ti:Sapphire laser (MaiTai and MaiTai DeepSee, Spectra Physics) tuned to 920 nm for GFP and GCaMP and 1040 nm for tdTomato. Live slice two photon imaging was performed as described above for confocal imaging.

In vivo imaging

Gad65-GFP mice were used for anatomical imaging. Under isoflurane anesthesia, a cranial window (3 mm) was made in the skull, taking care to keep the dura intact. Dexamethasone (sc, 2 mg/kg) was administered to reduce inflammation, and carprofen (sc, 5 mg/kg) or buprenorphine SR (sc, 1 mg/kg) was administered for analgesia. Titanium head bars were attached to the skull with dental cement (C&B Metabond). Motion dampening was provided by a titanium mesh placed directly on top of the brain surface. The mesh provided light pressure to minimize axial motion while the pores in the mesh still allowed superfusion access to the brain. The brain surface was kept moist by a saline soaked Gelfoam sponge. Mice were injected with acetazolamide (ip, 50 mg/kg) to reduce cerebral spinal fluid production. For some imaging sessions, the mouse was anesthetized with a ketamine/xylazine cocktail. Mice were head-fixed under the microscope, and the brain exposed to superfusion media. Imaging was performed during superfusion with ACSF or I-ACSF.

Gad2-cre or Emx1-cre mice crossed to Ai95D mice (Jackson Laboratories) were used to perform calcium imaging. Mice were injected with AAV1.CAG.tdTomato.WPRE.SV40 (Penn Vector Core) at least three weeks prior to imaging. For functional imaging, a circular cranial window with a diameter of 5 mm was made over the cortex of mice with GCaMP6f labelled neurons driven by Emx1-cre. The window was made over the right somatosensory or visual cortex. Media delivery occurred using a peristaltic pump through inlet and outlet ports during imaging at a flow rate of 1 ml/min. Mice were head-fixed on a treadmill during imaging.

Image analysis

Image analysis was done using Matlab (Mathworks) and ImageJ (NIH). For the calcium imaging datasets, a CNMF package was used to segment neurons and identify matching neurons across conditions⁵⁸.

Data availability

Data provided on request

Acknowledgements

This work was supported by NIH R01 DC010381 and R01 NS068409 (TEH) and R01 EY026978, R01 EY034001, and R01 EY027411 (DK).

Additional information

Author contributions

TEH conceived and supervised project. NK and MK developed clearing protocol. MK carried out imaging experiments and analysis. AA and DK performed retinal recordings and analysis of MEA data. DW and JD conducted behavioral testing and analysis; MK also did analysis of behavioral data. DWK analyzed calcium imaging data. TEH and MK wrote the manuscript.

Funding

HHS | National Institutes of Health (NIH) (R01 DC010381)

• Timothy E Holy

HHS | National Institutes of Health (NIH) (R01 NS068409)

• Timothy E Holy

HHS | National Institutes of Health (NIH) (R01 EY026978)

• Daniel Kerschensteiner

HHS | National Institutes of Health (NIH) (R01 EY034001)

• Daniel Kerschensteiner

HHS | National Institutes of Health (NIH) (R01 EY027411)

• Daniel Kerschensteiner

Additional files

Supplementary figure 1. Transmission imaging through live tissue slices. (a) Brightfield image of a resolution target imaged through 200 μm of cerebellar tissue. Left panel shows the tissue initially in ACSF. Right panel shows the tissue after clearing with an iodixanol enriched ACSF of refractive index of 1.42. The tissue shrunk noticeably after 1 hour of superfusion but the myelinated regions became more transparent. (b) MTF of the I-ACSF cleared tissue measured with an objective with a low NA (0.082) objective.

Supplementary figure 2. Testing for a relationship between iodixanol concentration and behavior performance. The behavioral performance of mice receiving an intraventricular infusion of iodixanol did not show any significant correlation with the concentration of iodine measured in the brain parenchyma as measured by the Pearson’s correlation coefficient.

Supplementary figure 3. Confocal imaging of cleared live neuronal tissue. (a) The point spread function of the confocal microscope imaged through ACSF and I-ACSF media, respectively, using a multi-immersion objective. With the appropriate objective correction, lateral and axial resolutions are very similar in both media. (b) GFP-labelled cortical neurons in a single slice imaged in both ACSF and I-ACSF media while maintaining identical imaging parameters. Cells in the same focal plane are shown before and after clearing at varying optical planes. The ability to resolve small, sub-cellular features dropped rapidly with depth in ACSF tissue. Soma and dendrites were visualized deeper after immersion in I-ACSF. Cell bodies were imaged at 142.5 μm in I-ACSF while the same cell was not detected when imaged in ACSF. Scale bar represents 20 μm. (c) Maximum intensity projection of the same tissue in ACSF and I-ACSF, acquired with identical imaging parameters. Dendrites and cell bodies were more visible at greater depth after clearing. Arrow indicate region where cross section was measured. Scale bar represents 20 μm. (d) Comparison of signal intensities along the marked regions between ACSF and I-ACF images taken at a depth of 110 μm. Dendrites were visible at 0.2% laser intensity in I-ACSF but not in ACSF at 0.8% laser intensity. The soma was barely resolved in ACSF at this depth. Arrows indicate region where cross section was measured. Scale bar represents 10 μm. Square root of the original image is shown in a-d to compress dynamic range.

Supplementary figure 4. Two photon imaging of cleared live neuronal tissue. (a) Maximum intensity projection images of the mouse cortex over time. The signal intensity increased as a function of time as the brain was superfused by I-ACSF. The mouse was anesthetized during imaging. Scale bar indicates 100 μm. (b) Maximum intensity projection images of the mouse cortex over time when the brain was superfused with only ACSF. The signal intensity did not change as a function of time when the brain was superfused with ACSF. The mouse was anesthetized during imaging. Images in a,b are displayed as the sqrt of the original intensity to condense dynamic range. Scale bar indicates 50 μm.

Supplementary video 1a. Mice behavior in home cage during intraventricular infusion with ACSF osmotic pumps.

Supplementary video 1b. Mice behavior in home cage during intraventricular infusion with iodixanol via osmotic pumps.

Supplementary video 2a. Calcium activity during I-ACSF treatment. GCaMP6 signal in an awake mouse brain, before superfusion with I-ACSF. Raw signal shown.

Supplementary video 2b. Calcium activity during I-ACSF treatment. GCaMP6 signal in an awake mouse brain, during superfusion with I-ACSF. Raw signal shown.