To facilitate high-throughput quantitative analysis of neuronal structure, this study optimized the diOlistic method of whole neuron labeling to examine multiple neurons in fixed brain, and optimized image acquisition parameters to preserve signal for subsequent photoconversion. Fluorescent dye-coated gold particles were successively delivered by helium-powered ejection to 250 µm thick brain slices with loading density and penetration depth optimized to maximize the yield of labeled neurons within the slice while avoiding overlapping labeled dendritic processes in the x–y plane and z-axis. Labeled neurons were imaged using confocal laser-scanning microscopy with pinhole aperture and scan speed enhanced to minimize capture time and fluorescence degradation. Optimized image acquisition parameters preserved fluorescence signal and facilitated subsequent oxygen-enriched photoconversion for higher magnification dendritic spine analysis. Sampling criteria limited analysis to neurons whose z-axis dendritic processes were fully contained within the tissue slice and in which dye transport extended to the most distal portions of the dendrites. The yield of completely labeled neurons was, on average, more than 20 cells per brain region per animal. With optimized spatio-temporal diOlistic loading parameters, along with image acquisition parameters optimized for subsequent photoconversion, the present protocol provides a high-throughput strategy for full-scale quantitative analysis of three-dimensional neuronal morphology.
The current methods for morphological characterization and quantitative analyses of single neurons, such as Golgi staining, intracellular injection and neuronal transfection, are labor-intensive, technically demanding, and do not readily yield large numbers of completely reconstructed neurons. Golgi impregnation relies upon the crystallization of silver chromate. However, the staining selectivity only labels less than 2–10% of the neurons, which is insufficient for quantitative analysis, and there is no experimental control over selection of neurons to be studied. Moreover, data analysis is compromised by incomplete impregnation, unavoidable loss of dendritic branches due to the plane of sectioning, and indistinguishable overlapping fine dendritic processes. Dye loading into single cells by microinjection using intracellular or patch pipettes generates excellent single-cell labeling but is technically demanding (e.g. Buhl and Schlote, 1987; de Lima et al., 1990; Schmidt et al., 1996; Taylor et al., 1996; Arnold et al., 2001; Morgan and Ohara, 2001; Duan et al., 2002). One of the major methodological concerns with these techniques is that quantitative analysis relies heavily upon subjective evaluations. The pipette tips are maneuvered manually, thus the loading is likely to occur in easy-to-access areas, such as a superficial cell layer, and/or in large-diameter neurons. Therefore, the sampling of intracellularly loaded cells often does not reflect the entire neuronal population. Moreover, injection-induced damage and/or leakage of dye via the needle track often results in unwanted labeling of neighboring cells, which reduces the signal-to-noise ratio.
Using neuronal transfection methods, including micro-injection (Hall et al., 1997), electroporation (Lurquin, 1997), lipofection (Strauss, 1996) and viral transfection (Robbins et al., 1998), introducing DNA constructs into target cells and tissues has become routine for examining gene regulation and function of the nervous system and for characterization of neuronal morphology. However, several variables are responsible for poor transfection efficiency using these methods. These include DNA coating efficacy and intrinsic cell viability in the cultures (Biewenga et al., 1997).
This neuronal transfection method has been improved by a ‘biolistic’ approach for in situ transfection and labeling of cells (Lo et al., 1994; Usachev et al., 2000; O’Brien et al., 2001; Jin et al., 2001; Danzer et al., 2002; Klimaschewski et al., 2002; Sun et al., 2002). In this system, a hand-held gene gun utilizes a pulse of helium gas to propel small particles coated with desiccated DNA, crossing plasma membrane to target cells. This technique does not require the specialized equipment and expertise that the conventional intracellular injection methods rely upon. Furthermore, this biolistic approach provides a means for random sampling, which is ideal for full-scale quantitative analysis. However, temperature changes, decreased atmospheric pressure, and high pressure-induced membrane disruption potentially challenge the cell cycle and transcription activity after bombardment, thereby affecting the expression of a gene product and the success of transfection (Biewenga et al., 1997; Thomas et al., 1998).
A recent advancement of the biolistic approach has been reported by Gan et al. (2000), who employed a gene gun to propel gold particles coated with various combinations of lipophilic dyes to target multiple neurons in live as well as fixed slices. Although suitable for visualizing complex cell–cell interactions, the ‘multi-diOlistic’ loading of neurons results in a high density of labeled neurons, which constrains the ability to perform quantitative morphological analysis of individual neurons.
The present study had two goals: (i) to enhance the yield of non-overlapping labeled neurons in a given brain region by standardizing morphological criteria for cell sampling under controlled conditions of dye-coated gold particle delivery, and (ii) to optimize the image acquisition process parametrically to reduce photo-bleaching, and facilitate non-fluorescent optical analysis of dendritic spines following a protocol for high-throughput photoconversion. The present study presents a standardized and efficient workflow that combines high-throughput randomized diOlistic loading, optimized image acquisition and photoconversion for the comprehensive quantitative analysis of individual neuronal morphology. Preliminary data of the present study have been reported in abstract form (Wu et al., 2003).
Materials and Methods
Tissue Preparation and Storage
Ninety-day-old male C57BL/6 mice (n = 12) were anesthetized with avertin (0.5 mg/g body weight) and transcardially perfused with normal saline followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB). Brains were dissected and post-fixed with 4% PFA overnight. After postfixation, the brains were coronally sectioned (250 µm) using a vibratome. The anterior–posterior axis of the hippocampus was subdivided into three levels based on the atlas of Hof et al. (2000). The brain sections were collected and stored in 4% PFA at 4°C prior to diOlistic shootings. The postfixation period varied from 1 h to 2 weeks, which allowed for repetitive delivery of the particles to the brain sections (see below). After delivery of the particles, the sections were stored in freshly made 4% PFA for confocal imaging (see below).
Gene Gun Bullet Preparation
The current protocol was developed partly based on the manufacturer’s manual (BioRad, Hercules, CA) and the procedure of Gan et al. (2000).
Tefzel tubing (BioRad) was placed on a tubing preparation station (BioRad) and air-dried for 45 min using nitrogen gas (0.4 l/min). Fifteen milligrams of gold particles (1.6 µm in diameter; BioRad) were thoroughly/evenly spread (2 × 2 cm2) on a glass slide using a razor blade. Five milligrams of 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate crystals (DiI; Molecular Probes, Eugene, OR) were dissolved in 500 µl methylene chloride and then gently dropped onto the thin layer of gold particles on the glass slide. The DiI-coated gold particles were carefully scrapped off with a razor blade, collected, and sonicated in 3 ml of distilled water for 30 min. In order to prevent aggregation of DiI-coated particles, 0.75 µl polyvinylpyrrolidone (20 mg/ml) was added to the sonicated solution. The solution was then vortexed for 15 s and immediately injected into the pre-dried tubing. Slow and continuous injection was crucial to prevent inclusion of air bubbles. The tubing was rotated for 30 min to allow the DiI-coated particles to precipitate and settle onto the inside surface of the tube (also see Discussion). The solution was then slowly withdrawn and the particle-coated tube was rotated and air-dried under constant nitrogen flow (0.4 l/min) for 2 h. The particle-coated tube was then cut into small pieces (microcarriers), and stored in a desiccated environment at room temperature. In order to maintain the quality of the labeling, the bullets were used within 24 h after coating.
Delivery of Particles
A commercially available biolistic ‘Helios gene gun system’ (BioRad) was used to propel DiI-coated particles into fixed tissue. The gene gun was stabilized by a custom-fabricated adjustable stand with a mobile z-axis (see Supplementary Material Fig. S1). A custom-fabricated centering tool was used to adjust the distance of the gun from the target tissue (3.0 cm) and to pinpoint the focus point of the landing field of the delivered particles. In addition to the manufacturer’s diffusion screen, a membrane filter with a 3 µm pore size and 8.0 × 105 pores/cm2 (Falcon 3095; BD Biosciences Discovery Labware, Bedford, MA) was placed between the gun and the tissue to filter out large clusters of coated particles and to avoid shock wave damage to the tissue. The particles were accelerated using inert helium gas (200 psi) and each side of the tissue was shot up to three times independently to enhance the yield of labeling. Each shooting was performed after completion of confocal image acquisition (see below).
Confocal Imaging and Optimized Parameters
Labeled neurons were imaged using a Zeiss CLSM PASCAL laser-scanning confocal microscope (Carl Zeiss, Thornwood, NY) equipped with a 10× Plan NeoFluar objective lens [numerical aperture (NA) = 0.3] and a 25× Plan NeoFluar water-immersion objective lens (NA = 0.8). For through-focus imaging, optical sections were collected using the 25× objective with 0.5–1.2 µm z-axis steps to cover the full depth of the dendritic trees. The z-axis stack compression that resulted from water immersion was automatically corrected by the analysis software. The 2-D stacks were superimposed digitally and the full 3-D dataset was generated for visualization of dendritic complexity. Labeled structures were excited using a 1 mW He:Ne laser (543 nm for DiI) with their emissions passed through a 560 nm long pass filter set. After scanning, images were taken using the 10× objective to document anatomical localization of DiI-labeled neurons against a reflectance-imaged background.
In order to enhance the imaging process, the pinhole aperture and scan speed were systematically investigated and optimized. Targeted neurons with either simple or complex dendrites were imaged separately using different pinhole apertures (48, 73 and 88 µm) with scan speed held constant. Similarly, targeted neurons were imaged separately using different scan speeds (1.9, 3.9 and 7.8 s/scan), while pinhole aperture was held constant. The scanned images were digitized and reconstructed, and their dendritic lengths were then quantified and compared (see Results).
Morphological Criteria for Cell Sampling
The criteria for selecting completely labeled targeted neurons were based on (i) the loading of DiI-coated gold particle in the soma, which was found to yield optimal dye transport; (ii) the tapering of DiI in the most distal dendrites; (iii) the visualization of the complete 3-D profile of dendritic trees using the 3-D display of Zeiss CLSM 5 PASCAL software (v2.8; Carl Zeiss). Neurons with incomplete dye transport and neurons with truncations due to the plane of sectioning were not collected. Moreover, cells with dendrites labeled retrogradely by particle impacts in the surrounding neuropil were excluded. In the present study, the types of labeled neurons collected included pyramidal cells in CA1 and granule cells in the dentate gyrus of the mouse hippocampus.
For visualizing targeted neurons, the sections were pre-incubated in Hoechst 33258 (10 mg/ml, Molecular Probes), a fluorescent nucleic acid stain, for 15 min prior to diOlistic labeling. The precision of neuron loading was visualized by the co-labeling of Hoechst-stained nuclei (blue channel) and DiI-coated particle (red channel) under epi-fluorescence microscopy. To examine the spatial correlation of dendritic complexity of the neurons and their anatomical location, sections containing labeled neurons were incubated with NeuroTrace green fluorescent Nissl 500/525 (1:100 in PBS, Molecular Probes) for 2 h and then imaged using confocal microscopy.
Neuronal Tracing and Data Analysis
The stacks of digitized 2-D images were downloaded and exported as a series of TIFF images to a PC workstation (530 MT, Dell, Round Rock, TX) and Neurolucida (MicroBrightField Inc., Williston, VT) was used to contour the dendritic processes of the digitized 2-D images. During the course of contouring, the 3-D dendritic geometry was visualized using the Zeiss CLSM PASCAL 3-D display software, with glow-scale image processing used to highlight neuronal processes.
All of the reconstructed data were exported to NeuroExplorer (MicroBrightField Inc., Williston, VT) for quantitative analysis. Total dendritic length and averaged lengths of each dendritic branch of labeled CA1 pyramidal neurons were generated using dendrogram and Sholl analysis, respectively. To validate the imaging parameters, each neuron was scanned under different pinhole apertures and scan speeds and then contoured three times each. Total dendritic length and average length of each dendritic branch were compared to examine the inter-trial variability. The results were analyzed using uni- and multivariate analysis of variance and Bonferroni post hoc tests, with significance set at P< 0.05 and 0.001, respectively.
After confocal imaging, the sections were washed 3 × 10 min with PBS and pre-incubated in filtered ice-cold diaminobenzidine tetrahydrochloride (DAB, Sigma, St Louis, MO) solution (1.5 mg/ml in 0.1 M Tris buffer, pH 8.2) for 6–12 h, a period that optimized the full penetration of DAB into 250 µm tissue slabs. Results from preliminary studies led to complete omission of all H2O2 (0.1%) from the DAB solution. After washes for 3 × 10 min, the sections were transferred to a custom-fabricated closed conversion chamber (CCC; see Supplementary Material Fig. S2), modified from Kacza et al. (1997).
The chamber was connected with silicone tubing that regulated the in- and out-flow of pure oxygen, which is suggested to significantly enhance the photoconversion process (Sandell and Masland, 1988; also see Lübke, 1993; Kacza et al., 1997). The sections were soaked in 10 ml DAB solution in the reservoir bath within the CCC and then covered with a plastic plate with a tightly fitting passage that allowed the 10× objective to navigate both horizontally and vertically to focus on the tissue during the conversion process. The light source was supplied by a 100 W mercury lamp. The airflow within the chamber was carefully regulated to prevent tissue dehydration. Unlike most photoconversion protocols (e.g. Buhl and Schlote, 1987; Balercia et al., 1992), the current protocol did not require frequent replenishing of the ice-cold DAB solution during conversion. The conversion process consisted of 30–45 min of exposure to epi-fluorescent illumination at 546 ± 30 nm, and when necessary, as for complex cell types such as pyramidal cells, the DAB solution was periodically exchanged with fresh DAB solution in the CCC to enhance the photoconversion process. Step-through focus of the light source during conversion was found to facilitate the quality of the final DAB reaction product. Low concentrations of glutaraldehyde were evaluated (0.05%) and eliminated as a fixative during perfusion because it substantially enhanced background fluorescence and thus increased non-specific background staining of the tissue.
The photoconverted tissue was then rinsed 3 × 10 min in PBS, and exposed to 0.1% osmium tetroxide for 3–5 min. In order to reduce shrinkage of neuronal structures, tissue sections were not dehydrated but mounted directly on slides and coverslipped with GelMount (Biomeda, Foster City, CA) for light microscopic analysis.
In the present study, the custom-fabricated adjustable stand was used to standardize and optimize the parameters of the diOlistic labeling protocol. Large diameter gold particles (diameter 1.6 µm) were found to penetrate into individual soma (Fig. 1A) more efficiently than those with smaller diameters (0.6 and 1.0 µm), since the latter tended to clump together following diOlistic delivery, producing labeled cells with overlapping and unanalyzable structures (see Fig. 1B,C).
Using 200 psi of helium pressure, the particles were delivered, on average, 65–85 µm below the surface of a fixed coronal slice (arrows in Fig. 1D). This penetration depth not only reduced the labeling of neurons whose dendrites were truncated as a result of the prior tissue sectioning, but it also allowed for successive delivery of particles to target cells on both sides of the tissues sufficiently far apart to avoid overlapping of their dendritic trees in the z-axis. The latter was especially critical for enhancing the yield of loading, since loading cells onto both sides of the tissue slice at different time points appears technically impossible using conventional methods. Moreover, to identify targeted cells precisely, the double-labeling method was used to visualize Hoechst-stained nuclei and DiI-coated particles within the soma (Fig. 1E). The penetration of a single DiI-coated particle within the soma was found to produce the optimal dye transport of neuronal processes. Note the absence of lateral diffusion of DiI along the penetration path, confirming an adequate velocity of bombardment using the current protocol.
Placement of a cell culture filter between the gene gun and the tissue was used to filter aggregated particles, and enabled us to disperse single dye-coated gold particles within the tissue (Fig. 2A). Optimal inter-particle distance of loaded somata varied between 75 and 100 µm over the target area, which is much greater than previously suggested by Gan et al. (see detail in Discussion). The combination of optimized loading density and penetration depth significantly reduced overlapping of dendritic processes between labeled neurons and, therefore allowed for full 3-D visualization of the neuronal dendritic trees. Figure 2B presents a collapsed CLSM image in which several dentate granule cells (GC) were concurrently labeled at an interval of 80–100 µm in the dorsal and ventral blades of the dentate gyrus, allowing for independent visualization with minimal overlapping of labeled dendritic processes. The somata of labeled cells were mainly located in the deep granule cell layer (GCL). Note that the somata in the middle position of Figure 2B (asterisk) did not contain labeled processes, mainly due to insufficient diffusion of DiI (see Discussion). One advantage of using DiI as a marker is that it labels extremely fine structures, such as spine necks, that are poorly illuminated by conventional intracellular loaded dyes. For example, Lucifer yellow (LY) readily diffuses through neuronal cytoplasm, which not only significantly increases background fluorescence, but also yields a LY-labeled dendritic trunk with detached spine heads (Fig. 2C,D). On the other hand, the transport of DiI efficiently labeled various types of spines with high resolution (arrows in Fig. 2D), assuring the quality of the subsequently photoconverted DAB (see below).
Morphological Criteria and Effective Yield of Cell Loading
In order to avoid artificial truncation of the dendritic tree, sampling of loaded neurons was limited to those with cell bodies sufficiently deep to the cut surface. Because an epi-fluorescent microscope is optically limited to the x–y plane, a confocal laser-scanning microscope was utilized to adequately visualize the z-axis of labeled neurons. Zeiss 3-D display CLSM PASCAL software was used to monitor full dendritic geometry with a 360° image rotation. This helped to verify the complete filling of transported dye, which was evidenced by the tapering of DiI in the most distal dendrites (arrows of Fig. 3A,B). Importantly, the lateral view of the CLSM image was used to verify that the full extent of the dendritic profile of labeled neurons (arrows of Fig. 3C) was contained within the slice. Neurons with insufficient transports of DiI (arrows of Fig. 3D,E), due to suboptimal fixation of tissue and/or inconsistent coating of DiI to gold particles, or neurons with amputation of primary dendritic processes (arrows of Fig. 3F) were excluded during the sampling process.
Figure 4A,B illustrates full 3-D views of diOlistically labeled neurons found in CA1 and the dentate gyrus of the mouse hippocampus (see Supplementary Material Movies 1–2). The basal dendrites of the labeled pyramidal cell reached the pial surface, reflecting the optimal transport of DiI to the distal dendrites (Fig. 4A). Similarly, we observed dentate GCs whose apical dendrites were fully labeled throughout the molecular layer (ML; Fig. 4B). In some cases, the somata of GCs resided in a deep portion of the GCL and thus had an initial unbranched apical dendrite (also see Fig. 2B). DiOlistic labeling with DiI was combined with NeuroTrace, a fluorescent Nissl stain, to reveal the laminar location of labeled neurons and dendritic arbors (Fig. 4C). Thus, diOlistically labeled granule cell dendrites could be defined relative to their position within the dentate ML (see Supplementary Material, Movie 3). Dentate granule cells located in a more superficial portion of GCL tended to branch immediately upon entering the molecular layer.
Based on the current protocol, on average, 20 analyzable cells per region per animal were obtained. For example, in the course of the anterior–posterior axis of CA1, the peak distribution of labeled cells was found in the mid-level of the hippocampus, mainly due to its prominent anatomical area (Fig. 5A). Similar results were found in the dentate gyrus (∼25 cells/animal; not shown). These yields significantly outnumber that reported for other conventional cell loading methods (see Discussion). Following diOlistic delivery, there was a high degree of morphological heterogeneity of CA1 pyramidal cells that varied with regard to their anatomical sites (see Fig. 5B). Note that several pyramidal cells were concurrently labeled in the same section (the third figure of lower panel in Fig. 5B). More importantly, the current protocol optimally loaded multiple cells in other hippocampal subfields, including the dentate gyrus and the hilus (colored arrows in Fig. 5C). High-magnification images of labeled cells from Figure 5C are presented in Figure 5D and E. Thus, the present study demonstrates that successive diOlistic shootings of the same tissue slice can be used to yield a large population of labeled neurons.
The inherent random sampling approach of the diOlistic method is ideal for quickly and reliably revealing inter-animal variability. In the present study, a pilot analytical study was conducted by reconstructing diOlistically labeled CA1 pyramidal neurons of 90-day-old male C57 mice (n = 4). Morphological differences of randomly labeled cells along the anterior-to-posterior axis of the hippocampus (n = 10 cells/animal) were quantified and compared by using dendrograms and Sholl analysis (NeuroExplorer). In the dendrogram analysis (Fig. 6A), a significant main effect on total length of apical dendrites was found across the four mice (P < 0.05), whereas only a marginal difference was found for basal dendrites. In the Sholl analysis (Fig. 6B,C), the number of intersections of both apical and basal dendrites was compared at 10 µm intervals from the soma and was found to vary significantly among the four mice (P < 0.05). Although the total number of dendritic intersections varied across the four mice, the total number of dendritic intersections peaked at similar distances from the soma (arrows in Fig. 6B,C). In apical dendrites, dendritic intersections peaked at a distance of 65–215 µm from the soma, whereas in basal dendrites, the peak was found at an approximate distance of 50–135 µm from the soma. Note that this pilot study did not investigate hippocampal asymmetry, therefore no information regarding inter-animal difference in the above-mentioned dendritic parameters is available. The reported data of dendrograms and Sholl analysis strongly suggest that robust quantitative analyses can be obtained from diOlistically labeled neurons.
Optimization of CLSM Imaging Parameters
High-resolution confocal images require a small pinhole aperture with a long scan speed in order to optimize signal-to-noise ratio. However, the consequence of lengthy scanning is bleaching of fluorescent signal in small-diameter dendritic spines. Also, because the intensity of membrane labeling decreases in proportion to the distance from the soma, scanning with a small pinhole aperture results in loss of fluorescent labeling in distal dendrites. To retain these subtle signals, the total scan time was reduced by increasing the pinhole aperture and shortening the scan speed at the expense of increased background fluorescence.
To test whether this optimization would affect image acquisition or quantitative data with respect to variability (three trials per image; also see Experimental Procedures), confocal images of the same labeled neurons were acquired with 48, 73 and 88 µm pinhole apertures, and the total dendritic length was obtained with reconstruction. ANOVA revealed a significant main effect of pinhole aperture between the three groups (P < 0.001). Bonferroni post hoc test confirmed that total dendritic length was smaller with 48 µm-pinhole scanned images (see Supplementary Material Fig. S3-A) compared to 73 or 88 µm-pinhole scanned images (P < 0.001), while no difference was found between the other two groups. It was noted that scanning with a pinhole aperture of 48 µm underestimated the total dendritic length by ∼10% compared to images collected with 73 and 88 µm apertures. An interaction effect was also found between pinhole aperture and branch order (P < 0.05) and post hoc tests showed that the averaged length of the most distal, sixth-order dendrites was underestimated using 48 µm pinhole aperture (P < 0.05), while no difference was found in other branch orders among the three groups (see Supplementary Material Fig. S3-B). Therefore, the underestimated total dendritic length with the 48 µm pinhole aperture is mainly attributable to the loss of the most distal dendritic branches.
Similarly, confocal images of labeled neurons were acquired with 1.9, 3.9 and 7.8 s/scan, and the total dendritic length was obtained with reconstruction. No significant difference of total dendritic length was found among the three groups (see Supplementary Material Fig. S3-C). Post hoc tests showed no difference in the averaged length of each branch order among the three groups (see Supplementary Material Fig. S3-D). These data suggested that scanning with a fast speed, 1.9 s/scan, could obtain estimates of dendritic length as accurately as scans with slower speeds.
Thus, with imaging parameters optimized to reduce bleaching, at the expense of increased background fluorescence, accurate measurements of dendritic lengths can be obtained. Most importantly, the optimization process has reduced the total scan time by as much as ∼30% (see Supplementary Material Fig. S4), minimizing dye bleaching and preserving signal for subsequent photoconversion (see below).
Photoconversion of Diolistically Loaded Neuron
To validate the efficiency and applicability of the current version of the CCC, the progress of photoconversion of a diOlistically labeled neuron was monitored. The presence of DAB reaction product was found within the cytoplasm 5 min after commencement of photoconversion (Fig. 7A). It was noted that the dendritic geometry of a CLSM-imaged neuron resembled that of the photoconverted cell, implying that a sufficient amount of DiI signal is preserved (Fig. 7B,C). In the present study, a 10× objective lens was used to illuminate the tissue because the higher objective lens (25×) was found to easily dehydrate the tissue using the CCC (not shown). Moreover, the field of view of a 10× objective lens allowed for photoconversion of multiple neurons concurrently labeled, as shown in Figure 2B. Conventionally, it requires, on average, 1.5–2 h to photoconvert the full dendritic tree of a diOlistically labeled pyramidal cell. However, the current protocol has enhanced the entire conversion process by reducing the illumination time by 50% (Table 1).
With optimized imaging parameters that have minimized dye bleaching, previously CLSM-imaged neurons could be photoconverted for dendritic spine analysis (see Fig. 8A–G). Based on the current protocol, the complete 3-D structure of diOlistically labeled neurons can be quickly acquired using optimized CLSM parameters (Fig. 8A) in order to perform reconstruction (Fig. 8B) for quantitative analysis, including dendrogram and Sholl analysis (Fig. 8C). Meanwhile, the CLSM-imaged neurons, with minimized bleaching, can be directly photoconverted using the CCC method (Fig. 8D,E) for spine analysis (Fig. 8F,G). It was noted that the dendritic geometry of a CLSM-imaged neuron resembled that of the photoconverted cell (Fig. 8A,E). This workflow provides the most accessible and efficient way of obtaining systematic quantitative data for both the dendritic arbor and spine density of individual neurons.
The present study has enhanced the protocol of the current diOlistic method for large-scale quantitative analyses of neuronal morphology in fixed brain tissue. Using optimized imaging parameters and morphological criteria, this study systematically characterized diOlistically loaded neurons with the preservation of diffuse signals in dendritic spines, allowing for subsequent photoconversion to reveal spine morphology. To our knowledge, this is the first report that has combined the diOlistic and photoconversion methods for full-scale quantitative characterization of individual neurons.
Diolistic versus Current Loading Methods
Most conventional methods, such as Golgi impregnation and intracellular loading used to illustrate the dendritic morphology are not only technically demanding but also are inadequate to systematically characterize neurons. Furthermore, these methods lack random sampling, making the resultant quantitative data highly variable. Although the biolistic transfection method is able to overcome the aforementioned limitations and produces excellent staining of dendritic trees, the efficiency of transfection heavily depends on the biological state of the living cells after bombardment. A clear advantage of using the diOlistic approach on fixed tissue is the liberty of tissue handling. Not only can the tissue be repeatedly bombarded, but the diOlistic loading is also independent of the cell’s gene transcription and protein synthesis.
In most biolistic studies, the source of labeled cells is mainly from organotypic slice cultures. Due to the above-mentioned factors that determine transfection efficiency, the yield of loaded cells for large-scale morphological analyses can only be increased by using a greater number of animals and/or tissue samples. Recently, Jin et al. (2001) quantified the polarization of dendritic trees of green fluorescent protein-transfected cortical neurons from 143 slice cultures of 48 animals. The number of successful transfected cells fluctuated at a wide range, which resulted in a low yield of loading (∼3 cells/animal or ∼1 cell/slice). Also, a recent report utilized the gene gun system to conduct a large-scale morphological survey of mouse retinal ganglion cells derived from 63 retinal slices (Sun et al., 2002). The morphology of the subtypes of retinal ganglion cells has been elegantly presented and quantitated, but the yield of loaded cells was only, on average, ∼8 cells/retinal slice. While the difference in the cell density of the loaded region may account for the difference in the yield of loading between the two studies, the yield of Sun et al. seems sub-optimal given that the diOlistically bombarded area of the whole-mount retina is significantly larger than that of many regions of the brain, such as the CA1 and dentate gyrus of the hippocampus in the current study. In contrast, the present protocol has generated a high yield (>20 cells/region per animal) without exhausting a large pool of animal subjects (n = 6). The present diOlistic protocol is able to generate a good dispersion of loaded particles and, more importantly, allows successive diOlistic delivery onto both sides of the fixed tissue at different time points (up to three independent shots) resolving both the spatial and temporal limitations encountered when performing intracellular injections.
One aspect that deserves clarification between the present and previous studies is the definition of optimal loading density of dye-coated particles. It has been suggested that in order to obtain a density of 15 labeled cells/mm2 in the hippocampus or cortex, the delivery of ∼100–200 particles/mm2 was required (Gan et al., 2000; Kettunen et al., 2002). While these studies did not include a quantitative analysis, it is likely that such an overly high density of loaded particles would render quantitative analysis difficult due to significant overlapping of dendrites in adjacent neurons. Instead, the present study attempted to achieve a balance between a high yield of neurons and minimal overlap of dendritic processes (compared to Thomas et al., 1998). An optimal 3-D spatial arrangement maximized the multiple loading of distinguishable neurons within a given brain region. This optimized arrangement facilitated full visualization of the individual neurons in order to meet our standardized morphological criteria of sampling. Exclusion of neurons with dendritic truncations or distal dendrites with incomplete filling, generated quantitative data with minimal technical artifacts. Moreover, the double-labeling method for delineating the spatial correlation of targeted neurons and their anatomical sites has provided an efficient means to categorize the cells concurrently labeled in diverse regions of the brain. Thus, our diOlistic protocol appears to be the most flexible and efficient system to obtain multiple labeled cells for quantitative analysis.
One issue remaining in the current diOlistic protocol is that the bullet quality seems to affect the yield of cell loading. First, there is no systematic way to precipitate DiI onto the gold particles during the coating process, making it difficult to evenly coat DiI onto the surface of the gold particles. As a result, some neurons are loaded with a gold particle into the soma but there is little to no dye-transport (also see Fig. 2B). Secondly, due to the weight of the gold particles, it is difficult to evenly suspend the DiI-coated gold particles in liquid. This non-uniform suspension leads to the uneven adhesion of dye-coated gold particles onto the inner surface of the pre-dried tubing, producing an irregular distribution of coated particles onto the landing field of the tissue. Because these factors may subsequently affect the loading efficiency (also compared to the DNA-coating procedure by O’Brien and Lummis, 2002), it is recommended that both the bullet quality and the focus of the target area relative to the landing field be tested prior to diOlistic bombardment (also see Wellmann et al., 1999).
Enhanced Throughput for Image Acquisitions
One aspect of optimizing the current diOlistic protocol was to enhance the throughput of the CLSM imaging process. By acquiring images of loaded neurons using increased pinhole apertures with a fast scan speed, the total scan time was reduced by up to 30% (see above). Underestimated branch length in the most distal dendrite using 48 µm pinhole aperture in the present study demonstrated for the first time that scanning with small pinhole aperture has the tendency to bias dendritic measurements. Underestimated sixth-order branch length is most likely due to the fact that the lateral resolution is much more sensitive to pinhole size than the axial resolution (Wilson, 1995). It is reasoned that small pinhole size improves x–y resolution at the cost of reducing the signal level. Instead, increasing the pinhole size enhances sensitivity and improves the spatial resolution by providing more photons for the microscope to detect in the z-axis, optimizing the collection of diffuse signals in distal dendrites. Moreover, the absence of a significant difference in total dendritic lengths between the three different scan speeds suggested that scanning at a fast scan speed, although elevating background fluorescence, would not only obtain accurate dendritic measurements, but would also reduce the total scan time and, therefore, minimize loss of signal after CLSM scanning. An important note was that illumination under 25× objective lens with optimized imaging parameters has allowed us to acquire the full range of the dendritic trees with minimal bleaching for photoconversion, thereby allowing quantitation of spine morphology (see below). Thus, the use of a high objective lens (100×) was not practical or desirable in the present study.
High-throughput Photoconversion of CLSM-imaged Neurons
Photoconversion prevents light-induced bleaching of most fluorescent markers by transforming the fluorescent compounds into electron-dense DAB reaction products (Sandell and Masland, 1988; Lübke, 1993). To establish an efficient workflow of the quantitative morphometric protocol, the present study has adapted the CCC of Kacza et al. (1997) to illuminate DiI-labeled tissue in an environment of pure oxygen, which has been suggested to facilitate the photoconversion process (Sandell and Masland, 1988; Lübke, 1993; also see Kacza et al., 1997).
With optimized image acquisition parameters, the present protocol was able to photoconvert DiI signals in neurons that were previously imaged using the CLSM. The current protocol has enhanced the quality of DAB reaction products as well as reduced the oxidation time by 50% (see Results). While the percentage loss of DiI signal in dendritic spines after CLSM imaging requires further investigation, the present finding has demonstrated a rather comparable resolution of the spine signals before and after photoconversion (see Fig. 7B). These data have validated our method for preservation of DiI signals in dendritic spines after fast CLSM scanning for subsequent photoconversion. Moreover, the quality of photoconversion depends, among other factors, on the duration and intensity of light irradiation, which are determined by the magnification and numerical aperture of the objective lens that is used. While most of the studies favor the use of a 20× or higher objective lens, the present study utilized the wider illuminated field of the 10× objective lens to simultaneously photoconvert multiple labeled cells. This is highly advantageous for enhancing the yield of photoconverted neurons, and thus has added a plus to the high-throughput process of our combined protocols. Accordingly, ultra-structural examinations of the dendritic spines of diOlistically labeled neurons are feasible following photoconversion.
Our combined protocols have proven capable of (i) enhancing the throughput of randomized cell loading by resolving the spatial (3-D inter-particle space) and temporal (multiple shootings of the same tissue at different time points) restrictions that often occur in conventional labeling techniques, (ii) optimizing the image acquisition process to readily obtain dendrograms, and to (iii) subsequently conduct photoconversion of previously CLSM-imaged neurons, for visualization and quantitative analysis of spine morphology. This approach represents the most efficient throughput for revealing the systematic correlation of the dendrites and spines of individual neurons. Because the morphology of the neurons reflects the physiology and function of the nervous system, the current protocol is a powerful tool for revealing subtle changes in the dendrites and spines of neuropathologic specimens.
Supplementary material can be found at: http://cercor.oupjournals.org.
The authors are thankful for the scientific advice of Drs Barry E. Kosofsky and Jeff M. Redwine, and excellent technical support from Dr Ron S. Broide, Mr Faisal Chawla and Ms Anna M. Cervantes.
|CA1 pyramidal cells||Dentate granule cells||DAB refreshing|
|Conventional||2.1 h (n = 15)a||1.5 h (n = 10)||>5 times|
|CCC||1.0 h (n = 15)||0.8 h (n = 10)||<1–2 timesb|
|CA1 pyramidal cells||Dentate granule cells||DAB refreshing|
|Conventional||2.1 h (n = 15)a||1.5 h (n = 10)||>5 times|
|CCC||1.0 h (n = 15)||0.8 h (n = 10)||<1–2 timesb|
aThe number of cells converted is indicated in parenthesis.
bThe refreshing of DAB might facilitate oxidation of complex cells, i.e. pyramidal cells of thick sections.