This article introduces an experimental protocol using 3D scanning technology bridging two spatial scales: the macroscopic spatial scale of whole-brain anatomy imaged by MRI at >100 μm and the microscopic spatial scale of neuronal distributions using immunohistochemistry staining and a multielectrode array system and other methods (~10 μm).
The human brain, being a multiscale system, has both macroscopic electrical signals, globally flowing along thick white-matter fiber bundles, and microscopic neuronal spikes, propagating along axons and dendrites. Both scales complement different aspects of human cognitive and behavioral functions. At the macroscopic level, MRI has been the current standard imaging technology, in which the smallest spatial resolution, voxel size, is 0.1–1 mm3. Also, at the microscopic level, previous physiological studies were aware of nonuniform neuronal architectures within such voxels. This study develops a powerful way to accurately embed microscopic data into a macroscopic map by interfacing biological scientific research with technological advancements in 3D scanning technology. Since 3D scanning technology has mostly been used for engineering and industrial design until now, it is repurposed for the first time to embed microconnectomes into the whole brain while preserving natural spiking in living brain cells. In order to achieve this purpose, first, we constructed a scanning protocol to obtain accurate 3D images from living bio-organisms inherently challenging to image due to moist and reflective surfaces. Second, we trained to keep speed to prevent the degradation of living brain tissue, which is a key factor in retaining better conditions and recording more natural neuronal spikes from active neurons in the brain tissue. Two cortical surface images, independently extracted from two different imaging modules, namely MRI and 3D scanner surface images, surprisingly show a distance error of only 50 μm as mode value of the histogram. This accuracy is comparable in scale to the microscopic resolution of intercellular distances; also, it is stable among different individual mice. This new protocol, the 3D novel embedding overlapping (3D-NEO) protocol, bridges macroscopic and microscopic levels derived by this integrative protocol and accelerates new scientific findings to study comprehensive connectivity architectures (i.e., microconnectome).
Nonuniform multiscale architectures at various physical and biological organizations are commonly found1,2. The brain is also a very nonuniform and multiscale network organization3,4. Various cognitive functions are coded in such network organizations, holding temporal changes of electrical spike patterns of neuronal populations in submillisecond temporal resolutions. Historically, the complex networks among neurons were structurally observed in detail using the staining techniques by Santiago Ramón y Cajal from over 150 years ago5. To observe group behaviors of active neurons, researchers have developed various recording technologies6,7,8, and the recent significant developments of such technologies have enabled us to record electrical activities from huge numbers of neurons simultaneously. Furthermore, from such functional activities, scientists have succeeded to reconstruct networks of causal interactions among huge numbers of neurons and have declared the topological architecture of their complex interactions ‘microconnectome’9. Macroscopic observations of the brain also allow for regarding a whole brain as a network organization because many brain regions are connected by multiple fiber-bundles. The embedding of microconnectomes into the global brain map still has clear limitations within current technological advances, which is why this embedding protocol is so important. However, there are many challenges to the development of the embedding protocol. For example, in order to observe activities of living local neuronal circuits in purely isolated brain regions, brain slices need to be produced for in vitro recordings. Additionally, recordings from brain slices for in vitro recordings are still an important choice for at least two reasons. First, it is still not easy to observe activities of many living individual neurons simultaneously from brain regions deeper than ~1.5 mm and in high temporal resolution (<1 ms). Second, when we hope to know the internal architecture of a local neuronal circuit, we need to stop all inputs coming from external brain regions to eliminate confounding factors. In order to identify the directions and positions of produced brain slices, it will be further necessary to integrate the spatial positions of these produced brain slices using coordinates. There are, however, a few systematic and reliable ways to make brain slices in an organized way10,11. Here, a new coregistration protocol is introduced, using 3D scanning technology for neuroscientific research in order to provide an integrative protocol. This protocol acts to coordinate micro- and macroscales and embed multielectrode array (MEA) microdata12,13 and staining data onto a macroscopic MRI space through 3D scan surfaces of extracted brains, as well as of noninvasively recorded brains. Surprisingly, this showed a distance error of only ~50 μm as the mode value of the histogram. As a result, the mode values of minimum distances between two surfaces between the MRI surface and the scanned 3D surface were nearly 50 µm for all six mice, which is a suitable number when checking for commonality among individuals. The typical slice width had a recorded spike activity of around 300 µm.
All experimental procedures described here have been approved by the Kyoto University Animal Care Committee.
1. Animals (Day 1)
2. MRI Settings (Day 1)
3. MRI Acquisitions (Day 1)
4. Preparation of Experimental Solutions (Day 2)
5. Preparation of Equipment (Day 2)
6. Brain Surface Scan, Slicing, and MEA Recording (Day 2)
7. Immunohistochemistry Staining (Days 3 and 4)
8. MRI Data Processing to Extract Cortical Volumes
9. MRI Image Processing to Stripe Cortical Surfaces
10. Preprocessing for 3D Scan Data
11. Coregistration of the MRI Surface and the 3D Scan Surface
We evaluated distances between cortical surfaces, produced by stripping MRI volume, and surfaces obtained from 3D scans of extracted brains. The mode values of the histogram of the distances are only 55 μm (Figure 3a). Additionally, when accumulating the histogram from the point where the distance equals zero, the accumulated value reaches 90% of total sample numbers at ~300 μm (Figure 3b). The final histogram of the distances between two surfaces showed a typical peak around 50 μm. If we interpret this value from a macroscopic viewpoint, interestingly, the accuracy, the mode value ~50 μm, corresponds with the geometrical limitation, which was expected from the voxel size of MRIs (Figure 3a), namely, 100 µm. This point indirectly suggests that the overlapping algorithm, ICP, between the MRI and 3D scan worked superbly well and that the noise levels of both the MRI and the 3D scan were suppressed as low value (Figure 2a,b)14.
Figure 1: The experimental flow. (a) First, after extracting a brain from a mouse, the brain was dropped into a bubbled cutting solution. (b) After wiping the cutting solution off with a soft and highly absorbent towel, (c) the brain was scanned on a turntable. (d) In making slices (step 8 of the process), the brain blocks were scanned on a BBB because it is easy to move brain blocks to the base for a vibratome (steps 7 and 8 of the process). (e) After getting enough slices for recording electrical activities or for staining the cell distributions and other methods, the remaining brain blocks were scanned again (step 10 of the process). The thickness of the remaining brain blocks provided additional important support information of where the slices were taken from. (f) Finally, we recorded the functional activities or structural architectures using MEA or staining methods and other methods. Please click here to view a larger version of this figure.
Figure 2: The data processing pipeline. (a) An example of a cortical surface produced by stripping cortexes from MRI volumes as explained in sections 8 and 9 of the protocol. (b) An example of a cortical surface directly scanned by a 3D scanner system. (c) An example of a memo used to summarize the brain region from where the individual brain slices were extracted. (d) An example of when the cortical slice is on an electrode dish. (e) An example of a stained brain cortical slice by NeuN (red) and GAD67 (green). All information is now going to be gathered in a unified database. Please click here to view a larger version of this figure.
Figure 3: Overlapping accuracy between MRIs and 3D scans. (a) Histograms of distances between surfaces extracted by peeling from MRI volumes. The surfaces came from 3D scan recordings. The main bar graph is the averaged histogram for all individuals, and other colored lines are results for individual mice (N = 6, aged 21–40 days, all were female mice). A general trend, stable for all individuals, could be found. (b) The accumulated value of the histogram is shown as a bar graph. The accumulated percentage reached 90% at ~300 μm (dotted lines). Please click here to view a larger version of this figure.
Figure 4: Conceptualized image of two scales of information integrated into one graphical user interface (GUI). Both the macroscopic brain anatomy and the microscopic neuronal circuit information were integrated into one system represented by a web. The high degree of accuracy, shown in Figure 3, enabled us to integrate these two scales realistically. The macroscopic image presented here is from a scalable brain atlas15. Please click here to view a larger version of this figure.
Parameters | Values |
Repetition time (TR) | 2,000 ms |
Echo time (TE) | 9 ms |
Effective TE: | 45 ms |
RARE factor | 16 |
Acquisition matrix size | 196 x 144 x 144 |
Field of view (FOV) | 19.6 x 14.4 x 14.4 mm3 |
Acquisition bandwidth | 75 kHz |
orientation | Axial (coronal orientation in scanner setting) |
Fat suppression parammeters | |
fat frequency | 2.6 ms-gaussian-shaped π/2 pulse with 1051 Hz bandwidth |
spoiler gradient | |
dummy scans | 2 times |
Number of averages | 3 |
Acquisition time | 2 h 42 min |
Table 1: The acquisition parameters of the 3D RARE pulse sequence used in this study.
We developed a new protocol called the 3D-NEO protocol to bridge macroscopic and microscopic spatial scales by overlapping two brain surfaces more accurately than before. Originally, there were two challenges in creating this protocol which made possible the accurate overlapping of two brain surface images and recording healthy neuronal activities from living organisms. First, it was necessary to effectively wipe cutting solution surrounding the extracted brain after extracting it from the skull without injuring the brain organism (step 6.2 of the protocol). Second, since the first and third conditions of dryness and time delay are potential negative factors for the living organism, finishing all the steps of the scanning process from brain extraction to the slice preparations within 10–15 min was also necessary to keep the neurons active.
The ability to successfully record brain activities using this protocol has now made possible the coordination not only of structural organizations but also of functional activities on the whole-brain map. Another positive aspect of this protocol is that since a BBB was prepared, the calibration from the scanner to the base of the vibratome became much smoother.
As mentioned in the representative results, the histogram of distances between two surfaces showed a peak at around 50 μm. Although we performed the overlapping process, as if by seeing from the macroscopic viewpoint. If we interpret the result from the microscopic viewpoint, the 50 μm is a comparable spatial scale in the distribution of neurons, since connectivity probabilities between pairs of cortical neurons decay within ~100 μm16,17, even within several millimeters18. Practically, MEA or calcium imaging studies often use 300–400 μm thick slices. So, the technique presented here will provide strong objective evidence if the slices are properly embedded into original whole-brain maps. After overlapping is accurately achieved, additional accuracy will be gained purely from the local information observed under a microscope. Therefore, by performing a two-step optimization process consisting of global and local optimization steps, it will be possible to reach a spatial resolution equating semi-automatically to the typical size of a neuron in the future.
This accurate integration protocol provides basic technology to generate various brain atlases18,20,21 and even to study more in-depth the MRI of other primates22,23 and human postmortem brains (although the human brain has sulci, it is possible to get a sufficient number of recording points from gyri). Now, we are also developing a visual interface to integrate many recorded data samples into one common spatial coordinate. One image figure such as this is shown in Figure 4. In the cases of mammalian brains, the integration between quantitatively different scales will also make new advances qualitatively in understanding disease states and in evaluating drug effects. Generally, this will be hard to apply to fishes, reptiles, and amphibians, because researchers will find it difficult to keep the form of the extracted thin brain stable to its pre-extracted form, and MRI images of their smaller brains will also be relatively noisier.
The 3D-NEO protocol has been introduced by this research into a neuroscientific or cell biological research field, successfully demonstrating high accuracy in bridging between macroscopic anatomy and microscopic neuronal distribution, including the topological architectures of macro- and microconnectomes.
The authors have nothing to disclose.
M.S. is grateful for the support from all of the staff in the medical information engineering course in Graduate School of Medicine and Faculty of Medicine, and wants to thank Prof. Tetsuya Takakuwa, Prof. Nobukatsu Sawamoto, and Doris Zakian for their helpful comments. This study was supported by a Grant-in-Aid for Challenging Exploratory Research and by the Leading Initiative for Excellent Young Researchers (LEADER) program to M.S. from MEXT (The Ministry of Education, Culture, Sports, Science, and Technology). The MRI experiments in this work were performed in the Division for Small Animal MRI, Medical Research Support Center, Graduate School of Medicine, Kyoto University, Japan.
Air compressor | Kimura Medical | KA-100 | Animal preparation for MRI |
All-in-one fluorescence microscope | KEYENCE | BZ-X710 | |
Anesthesia box | Bio Research Center | RIC-01 | Animal preparation for MRI |
Anesthesia system | ACOMA Medical Industry | NS-5000A | Animal preparation for MRI |
Anti-GAD67, clone 1G10.2 | Merk Millipore | MAB5406 | For immunostaining |
Calcium Chrolide | nacalai tesque | 06729-55 | aCSF |
Choline Chloride | nacalai tesque | 08809-45 | aCSF |
curved blunt forceps | |||
Disposal scalpel | Kai | 10 | |
D-PBS(-) without Ca and Mg, liquid(10x) | nacalai tesque | For immunostaining | |
D(+)-Glucose | Wako | 049-31165 | aCSF |
Gelatin | nacalai tesque | 16605-42 | re-secctioning |
Goat anti-Mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor Plus 488 | Invitrogen | A32723 | For immunostaining |
Goat anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor Plus 555 | Invitrogen | A32732 | For immunostaining |
Heater mat | Bio Research Center | HM-10 | Animal preparation for MRI |
Heater mat controller | Bio Research Center | BWT-100A | Animal preparation for MRI |
Heater system | SA Instruments | MR-compatible Small Animal Heating System | Animal preparation for MRI |
Isoflurane | AbbVie | Animal preparation for MRI | |
Isoflurane vaporizer | ACOMA Medical Industry | MKIIIai | Animal preparation for MRI |
Linear Slicer | DOSAKA | Neo Linear Slicer MT | |
L(+)-Ascorbic Acid Sodium Salt | Wako | 196-01252 | aCSF |
Magnesium Chrolide Hexahydrate | Wako | 135-00165 | aCSF |
MaxOne Single-Well MEA | MaxWell Biosystems | ||
Metal Spatula | |||
Monitoring system | SA Instruments | Model 1025 | Animal preparation for MRI |
Monitoring software | SA Instruments | PC-SAM V.5.12 | Animal preparation for MRI |
MRI compatible cradle | Bruker BioSpin | T12812 | Animal preparation for MRI |
MRI coil | Bruker BioSpin | T9988 | For MRI |
MRI operation software | Bruker BioSpin | ParaVision 5.1 | For MRI |
Neo LinearSlicer MT | D.S.K. | NLS-MT | |
NeuN (D4G40) XP Rabbit mAb | Cell Signaling | 24307 | For immunostaining |
Normal Goat Serum | Wako | 143-06561 | For immunostaining |
Potassium Chloride | Wako | 163-03545 | aCSF |
Polyethylene Glycol Mono-p-isooctylphenyl Ether | nacalai tesque | 12967-45 | For immunostaining |
Pressure-sensitive respiration sensor | SA Instruments | RS-301 | Animal preparation for MRI |
Preclinical MRI scanner | Bruker BioSpin | BioSpec 70/20 USR | For MRI |
Pyruvic Acid Sodium Salt | nacalai tesque | 29806-54 | aCSF |
SCAN in a BOX | Open Technologies srl | ||
scissors | |||
Sieve bottle | TIGERCROWN | 81 | For 3D scan |
SlowFade Gold Antifade Mountant | Invitrogen | S36937 | For immunostaining |
Sodium Chloride | Wako | 191-01665 | aCSF |
Sodium Dihydrogenphosphate | Wako | 197-09705 | aCSF |
Sodium Hydrogen Carbonate | Wako | 191-01305 | aCSF |
Sodium Hydrogensulfite | nacalai tesque | 31220-15 | For immunostaining |
Thermistor temperature probe | SA Instruments | RTP-101-B, PLTPC-300 | Animal preparation for MRI |
Tooth bar | Bruker BioSpin | T10146 | Animal preparation for MRI |
Winged intravenous needle | TERUMO | SV-23CLK | For perfusion |
1 mol/l-Tris-HCl Buffer Solution | nacalai tesque | 35436-01 | For immunostaining |
1 mol/l-Hydrochloric Acid | nacalai tesque | 37314-15 | For pH adjustment of solution |
16%-Paraformaldehyde Aqueous Solution | Electron Microscopy Sciences | 15710 | For immunostaining |