Functional Expression of YC3.6 in the Murine Brain

YC3.6-AAV2 was stereotactically targeted to layer 5 pyramidal neurons of the neocortex of mice in vivo. Figure 2 shows in vivo expression in layer 5 neuronal soma (2A) and superficial (< 120 um below the surface) dendrites and axons (2B–D). Hundreds of neurons were infected from a single injection, with YC3.6 evident within two weeks and lasting more than 8 months after viral injection. Dendritic spines were easily resolved at higher magnification and spiny dendrites in layer 1 were prevalent. We examined the response characteristics of YC3.6 in vivo by local micropipette injections of glutamate (Figure 3). Areas with neurites and spines expressing YC3.6 were identified and calcium transients were measured in response to application of glutamate (Figure 3A). Single puffs of glutamate (10 mM in the pipette) were sufficient to evoke [Ca²⁺]_i transients in dendrites (Figure 3B); peak, non-averaged, responses in all dendrites had an average amplitude (ΔR/R) of 56 ± 1% (mean +/− SEM, n = 16). In all cases, the dendritic [Ca²⁺]_i returned to the baseline ratio with an average decay constant of 1.30 ± 0.04 s (mean +/− SEM, n = 16). [Ca²⁺]_i increased, on average, from 78 nM to three times that, 243 nM, during glutamate stimulation. Increasing amounts of glutamate (approximated by co-injection with a red-fluorescent dye) yielded larger amplitude and longer duration calcium responses (Figure 3C).

Open in a separate window

Figure 2

In vivo expression of YC3.6-AAV

A–D) In vivo multiphoton images deep in the neocortex of adult mice. A) Cell bodies 300–550 microns below the brain surface. B) Full field images of YC3.6-filled neuritis and fluorescent angiogram (red). C) High-resolution images of axons, dendrites and spines in layer I of a wildtype mouse. D) Higher magnification images of a dendritic tree and corresponding spines. Detailed analysis of cellular and sub-cellular morphology was possible with YC3.6.

Open in a separate window

Figure 3

In Vivo Imaging of Glutamate-evoked Ca2+ transients

A) High-resolution multiphoton image of dendrites and spines in vivo. The calcium response to micropipette injections of glutamate is described in trace (C). B) Dendrites from three different animals showing responses to small puffs of glutamate (10 mM). Calcium transients were measured as ΔR/R and averaged 56% ± 1% (n=16 dendrites). C) Peak calcium response and the decay constant increased with increased glutamate. Local glutamate concentration was approximated by co-injecting a red-fluorescent synthetic dye.

One concern with fluorescent calcium-sensitive proteins involves their limited efficacy in measuring spontaneous activity known to be present in cortical neurons. We imaged somatic spontaneous calcium transients in layer 5 pyramidal neurons and confirmed that YC3.6-AAV2 exhibited good temporal kinetics (τ_decay = 0.43 +/− 0.05 seconds, duration = 0.86 +/− 0.06 seconds, n = 68 transients in 2 mice) and robust changes in YFP/CFP ratio (ΔR/R = 68.7 +/− 3.8 %, n = 68 transients in 2 mice, see Supplementary Figure 2). The robust expression, response to glutamate, concomitant return to baseline concentration, and measurable spontaneous activity provide evidence that the cameleon-based indicator, YC3.6, not only expresses in vivo but provides a quantitative functional readout of neuronal [Ca²⁺]_i dynamics.

Intracellular Calcium is Highly Regulated in vivo

The [Ca²⁺]_i in neurites in WT mice was tightly regulated (Fig 4A,C): the resting [Ca²⁺]_i in all morphologically normal spiny and aspiny processes was 77 nM (65 – 91 nM) (range is one geometric SD away from the geometric mean). There was no appreciable difference between 3.5 month old and 6.5 month old animals (see Supplementary Figure 1). Rare examples (~2.2%) of neurites with calcium overload were observed in these mice (defined as a sustained [Ca²⁺]_i greater than 2 geometric SD above the geometric mean of all processes in WT mice, 147 nM). A log-normal distribution of [Ca²⁺]_i accurately described the range of [Ca²⁺]_i seen in WT mice (r² = 0.99). To confirm that our [Ca²⁺]_i measurements reflected resting [Ca²⁺]_i, we imaged a subset of neuritis several minutes after the first measurement and confirmed that a similar [Ca²⁺]_i was maintained (p = 0.57, Wilcoxon signed-rank test, n=28). Furthermore, we used relatively slow (~8 seconds per frame), high resolution (512 × 512 or 1024 × 1024 pixels) scans—our measurements reflect, therefore, an integrated “resting” [Ca²⁺]_i that is unlikely to be significantly affected by transient increases in [Ca²⁺]_i in neurites.

Open in a separate window

Figure 4

Calcium Overload in Neurites of APP/PS1 Transgenic Mice with Cortical Plaques

In vivo image of dendrites and axons in aged non-transgenic (A) and transgenic (B) mice. Images are pseudocolored according to [Ca²⁺]_i. C–D) Histograms of [Ca²⁺]_i (YFP/CFP ratio on lower x-axis and [Ca²⁺]_i on upper x-axis). Tg mice had a distinct distribution (D)—modeled as the sum of two Gaussians (red line, r2=0.99)—compared to Non-Tg mice (C, single Gaussian, red line, r2=0.99 ). The black line corresponds to 2 standard deviations above the mean [Ca²⁺]_i in non-Tg mice (147 nM)—above this line was classified as calcium-overloaded. 20.0% of neurites in Tg mice were overloaded compared to only 2.2% in non-Tg mice—a 10-fold increase (n= 3 mice in each group). E) A comparison of the percentage of calcium overloaded neurites across 4 different transgenic mouse lines. Only the APP/PS1 and the Tg-2576 (APPswe only) transgenic mice had significantly higher number of neurites with elevated calcium (*, p < 0.001, Chi-Squared test). PS1 mutant mice that did not exhibit plaques showed no appreciable calcium overload when compared to their respective controls (p = 0.77 for PS1-ΔE9 and p = 0.21 for PS1M146V). The PS1-ΔE9 mice are on the same background as the APP/PS1 transgenic mice—directly comparing the calcium overload between these mice and the APP/PS1 shows that the APP/PS1 mice that develop plaques have a significantly higher calcium overload (20.2% vs. 2.6%, p < 0.001, Chi-square test).

Calcium Overload in Neurites of APP/PS1 Transgenic Mice with Cortical Plaques

We next sought to determine whether calcium overload occurred in the dendrites and axons in transgenic mice with mutations relevant to AD. To do this, we measured the resting [Ca²⁺]_i in four transgenic mouse lines (and associated controls): APPswe/ PS1-ΔE9 double mutants, Tg-2576 (APPswe only), PS1-ΔE9 only and PS1M146V only. The 9 APP/PS1 mutant mice typically develop amyloid-β plaques at 4.5–5 months of age (Garcia-Alloza et al., 2006) whereas the Tg-2576 develop them later—at approximately 14–15 months. Neither of the PS1 mutant mice develops any amyloid-β plaque pathology.

In the APP/PS1 transgenic mouse line, a large number of dendrites and axons exhibited far higher [Ca²⁺]_i than wildtype mice (see Figure 4B and 4D, p < 0.001, n = 1275 total neurites, Mann-Whitney U test). As before, calcium overload was defined as a [Ca²⁺]_i at least two SD above the mean of neurites from non-transgenic mice (147 nM). APP/PS1 mice had nearly 10 times the number of processes that crossed this threshold when compared to non-transgenic mice (20% vs. 2.2%, Figure 4e, p < 0.001, Chi-Squared Test). Different populations of neurites were readily apparent in images pseudo-colored for calcium concentration (see Figure 4B and 4D). Examination of histograms of [Ca²⁺]_i for all axons and dendrites suggested that the transgenic mice had a bimodal distribution, which allowed us to model the data as a sum of two log-normal distributions (r² = 0.99, Fig. 4D). The first Gaussian exactly matched that seen in non-transgenic mice: mean = 77 nM (63–94 nM). The second peak was shifted far to the right, was 6-fold higher, and was broadened: mean = 499 nM (162–1535 nM).

To test whether APP/PS1 mutations effect calcium overload before plaque deposition, we examined whether [Ca²⁺]_i-elevated neurites existed in younger, 3–3.5 month old APP/PS1 mice – an age prior to the development of plaques but consistent with robust transgene expression (Garcia-Alloza et al., 2006; Jankowsky et al., 2001). [Ca²⁺]_i was not appreciably disturbed in the neurites of these mice (see Supplementary Figure 3). The distribution of [Ca²⁺]_i was significantly different from APP/PS1 mice with plaques (p < 0.001, Student’s T-test, see Supplementary Figure 3 for comparison of distributions). Only 2.3% of neurites in these young transgenic animals had a [Ca²⁺]_i greater than 2 SD above the nontransgenic mean (p = 0.731 when comparing 3–3.5 mo-old APP/PS1 transgenic and > 6.0 mo-old APP/PS1 non-transgenic) which was significantly less than in APP/PS1 mice with plaques (20.2%, p < 0.001).

Calcium Overload Requires Amyloid-β Plaques

The APP/PS1 transgenic mouse model provided clear evidence for calcium overload in a subpopulation of neurites. The primary advantage of this mouse model is the accelerated plaque deposition (by 4.5–5 months of age) compared to the Tg-2576 (APPswe only). Since both the APPswe and PS1-ΔE9 have been implicated in impaired calcium regulation, we next tested which of these mutations might be responsible for the calcium overload by measuring the resting [Ca²⁺]_i in aged Tg-2576 (APPswe) mice and, separately, PS1-ΔE9 mice. To ensure that there is no effect of background strain, and age, on resting levels of calcium, we compared the Tg-2576 against age-matched non-transgenic from the same background strain (n = 3 transgenic mice, n = 2 non-transgenic mice). The PS1-ΔE9 mutant mouse line shares the same background as the APP/PS1 double transgenics so we use age-matched mice in this instance (n=2 transgenic mice, n = 3 non-transgenic).

The Tg-2576 mice exhibited severe calcium overload in the dendrites and axons that were measured, nearly identical to that seen in APP/PS1 mice (p < 0.001, Mann-Whitney U-test, see supplementary figure 4a,c for image and histogram). 22% of all dendrites and axons exhibited calcium overload, compared to 4.5% in the non-transgenics (p < 0.001, Chi-square test, Figure 4e). The 4.5% in the age matched controls is higher than in the APP/PS1 mice (2.2%, p = 0.045) but this small difference can be attributed to the difference in age (6–8 months for the APP/PS1 vs. >17 months for the Tg-2576) and background strain.

We imaged calcium concentration in transgenic mice expressing different mutant PS1 transgenes in the absence of human APP mutations. These mice do not develop amyloid-β plaques despite strong expression of the transgene. Neither the PS1-ΔE9, nor the PS1M146V mutant mice exhibited any appreciable calcium overload compared to their respective controls (p = 0.77 for PS1-ΔE9, p = 0.21 for PS1M146V, Figure 4e). In the PS1-ΔE9 mice, 2.6% of neurites were overloaded with calcium compared to 2.2% in control (see supplementary figure 5a,b for image and histogram). This suggests that although the PS1 mutation helps accelerate plaque deposition in the APP/PS1 transgenic line, a PS1 mutation alone is not sufficient to induce the calcium overload that we observe. These data suggest that mutant forms of PS1, known to have strong effects in fibroblasts and other reduced preparations, do not impact resting neuronal [Ca²⁺]_i in neurites in vivo. This does not preclude, however, more subtle changes in intracellular calcium handling, perhaps in the perikarya, or the existence of compensatory mechanisms that may mask the effect of presenilin mutations on calcium handling.

Aβ Plaques are a focal source of toxicity leading to calcium overload

To examine the cause of the increased [Ca²⁺]_i in APP mice, we next determined whether Aβ plaques had a localized effect on [Ca²⁺]_i. Plaques were concurrently imaged by injecting methoxy-XO4, a fluorescent amyloid binding dye (Klunk et al 2002). We classified the [Ca²⁺]_i neurites as either overloaded (2 SD above the average resting [Ca²⁺]_i in non-transgenic mice) or normal, and then determined the fraction of calcium overloaded neurites at discrete distances from the nearest senile plaque. A greater proportion of neurites with calcium overload existed near plaques when compared to further away. 30% of all neurites within 25 microns of a plaque exhibited strongly elevated [Ca²⁺]_i compared to 17% farther away (p<0.05, Fischer exact test, n=430). Furthermore, a gradient of [Ca²⁺]_i overload was evident (Figure 5), with the fraction of calcium-impaired neurites highest closest to plaques (0–10 microns: 33%, n=30) and slightly less at greater distances (10–30 microns: 26%, n=78). Beyond 30 microns away from the plaque there was no relationship between plaque distance and impaired calcium (30–55 microns: 17%; 55–85 microns: 18%,; p > 0.5, n=352). These results indicate that the immediate micro-environment of an Aβ plaque focally disrupts neuronal [Ca²⁺]_i handling and function, and that calcium overload can be detected (although in a lower percentage of neurites) even at distances farther from a plaque. The extent of disruption is marked: compared to the number of neurites with [Ca²⁺]_i > 2 SD from the mean in controls, the number near a plaque is elevated about 15-fold.

Open in a separate window

Figure 5

Aβ Plaques are a focal source of toxicity leading to calcium overload

A) In vivo image YC3.6 expressing dendrites and axons surrounding Aβ plaques (labeled in white). Elevated calcium was seen in neurites (in red) alongside neurites with normal calcium (in blue). B) Within 25 microns of a plaque, there was a significantly higher probability of finding a calcium overloaded neurite (p < 0.05, Fisher-Irwin test). The fraction of impaired neurites in non-Tg mice (2.2%) is shown as reference (blue-dotted line). Note that even far from a plaque, the percentage of calcium-overloaded neurites is much greater than in non-Tg mice.

Spino-Dendritic calcium compartmentalization is disrupted in mice with plaques

Calcium overload might have several downstream functional consequences. One such consequence would be altered spino-dendritic compartmentalization. The ability of the dendritic spine to compartmentalize calcium has been shown to be important for proper neuronal signaling and synaptic communication (Araya et al., 2006). A large fraction of the calcium overloaded processes imaged in the APP mice maintained their spino-dendritic morphology, so we used high-resolution 3-dimensional images of dendrites and axons to investigate spine-level calcium effects. 3,438 dendritic spines (n = 5 transgenic and 5 non-transgenic mice) were manually identified and their local calcium concentration was measured alongside the parent dendritic base (see Figures 6A for schematic). Figure 6B shows that, like dendrites and axons, a large number of dendritic spines in transgenic animals have a markedly elevated calcium concentration. To test whether the spine acts as a distinct compartment from the parent dendrite (as has been postulated) (Carter et al., 2007), we paired each spine with its local dendritic base and created a scatter plot of the respective [Ca²⁺]_i values. A slope (m) of 1.0 with a strong correlation (R²) would indicate that the spine and the parent dendrite have the same [Ca²⁺]_i—a slope of zero or a weak correlation would provide evidence that the spine operates independently of the dendritic base. For all normal-calcium dendrites (n=2,977 spines), the associated spines act as independent compartments in which there is no significant correlation between the spine and dendritic base (R² = 0.26). This holds true for normal-calcium dendrites in both non-transgenic (n = 1647 spines in 5 animals, R²=0.25) and transgenic mice (n = 1,330 spines in 5 animals, R²=0.26, Fig. 6C). In dendrites with elevated-calcium ([Ca²⁺]_i > 147 nM), present uniquely in the cortex of mice containing plaques, this compartmentalization is lost (Fig. 6D, n = 461 spines in 5 animals). Indeed, there is a strong correlation between the spine and dendrite [Ca²⁺]_i (R² = 0.77) with a slope nearing one (m = 0.85). In this case, spine and dendrite [Ca²⁺]_i scale with each other, no longer allowing independent calcium signaling, suggesting the loss of a crucial mechanism important for dendritic signal integration (Carter et al., 2007).

Open in a separate window

Figure 6

Disrupted Spino-Dendritic calcium compartmentalization in mice with plaques

A) Dendritic spines were identified and regions of interest were drawn in ImageJ to isolate the spine from the dendritic compartment at the base of the spine. B) The spine calcium is plotted against the dendrite calcium for each spine-dendrite pair for spines from transgenic (red) and non-transgenic (blue) animals. The blue dotted lines demarcate the threshold for being calcium-elevated ([Ca²⁺]_i > 147 nM). Unlike in the non-transgenic animals (blue), a large number of transgenic spines have a significantly elevated [Ca²⁺]_i, consistent with data from dendrites and axons. C) The inset schematic highlights which part of the calcium distribution is being analyzed in the figure. Calcium is compartmentalized between normal-calcium spines and the local dendritic base—there was no statistically significant correlation between the spine vs. dendrite calcium (R² = 0.26). C) Spine calcium is plotted against dendrite calcium for all elevated-calcium dendrites in the APP/PS1 mice. The inset highlights the part of the calcium distribution that is being analyzed. A strong correlation exists between the spine calcium and the associated dendritic compartment (R² = 0.77) with a slope of 0.85 suggesting that the spine and dendrite are nearly completely coupled. Data are from n = 5 transgenic, n = 5 non-transgenic mice.

Elevations in [Ca²⁺]_i Underlie the Morphological Changes in Neuronal Processes

We have recently shown that plaques induce neuritic dystrophies within a radius of ~25 microns, similar to the region noted to have neurites with an increased chance of elevated [Ca²⁺]_i (Meyer-Luehmann et al., 2008). Using a detailed morphometric analysis, we determined that calcium overload occurred in neurites with normal (both aspiny and spiny) and altered (beaded) morphology (Figure 7A–C). 83.1% of all neuritic processes in Layer 1 were spiny dendrites, aspiny processes made up 15.4%, and beaded and dysmorphic neurites were rarely seen (< 1.5%). We next separated neurites from mice with amyloid plaques into two functional groups (normal and calcium overloaded) and classified the morphology for both groups. The normal [Ca²⁺]_i processes had a nearly identical morphological distribution as neurites in wildtype mice (figure 7D). In contrast, neurites with abnormal morphology were strongly over-represented in the calcium overloaded group (Figure 7D). About half of all identified processes were either beaded, blebbed or dysmorphic. The remaining half maintained a normal morphology—26% were classified as spiny dendrites and 21% as aspiny dendrites. As expected, the spiny:aspiny ratio fell from 4:1 to 1:1 near plaques—more processes lacked spines near plaques (Spires et al., 2005). We then compared the[Ca²⁺]_i of the different morphological classes in the calcium overloaded group and found that there was a strong relationship between morphology and resting [Ca²⁺]_i (figure 6E). Calcium overloaded aspiny processes and spiny dendrites (normal morphology) had an average [Ca²⁺]_i of 395 nM (195–801 nM) (6-fold greater than normal [Ca²⁺]_i). [Ca²⁺]_i in beaded and dysmorphic neurites was significantly different than this—955 nM (406–2240 nM) (p < 0.001, Wilcoxon Mann-Whitney test, n = 135)—a striking 12-fold elevation of [Ca²⁺]_i. The observation that [Ca²⁺]_i increased as the morphology shifted from normal to abnormal suggests that profound morphological changes are associated with sustained increases in [Ca²⁺]_i. Since we also observed large numbers of neurites that were morphologically normal but had marked [Ca²⁺]_i elevations, morphological changes appear to represent only a subset of functionally altered dendrites in the cortex of mice with amyloid plaques. Thus, our in vivo functional pathology approach provided a window into diseased processes that could not be readily identified on the basis of morphological inspection alone.

Open in a separate window

Figure 7

Calcium overload underlies distortions to neuritic morphologies

A) A [Ca²⁺]_i-elevated “beaded” neurite (1) interdigitates with a morphologically, and functionally, normal dendrite (2). B) Beaded (1–3) neurites adjacent to a senile plaque (in white) exhibit high [Ca²⁺]_i. C) Structurally intact spiny dendrites (1 – 3) and structurally compromised beaded neurites (4 and 5) adjacent to a senile plaque (in white) exhibit significantly elevated [Ca²⁺]_i. D) The morphological breakdown of non-transgenic neurites matches that seen in [Ca²⁺]_i normal neurites in transgenic mice. [Ca²⁺]_i -elevated neurites, however, show a marked difference in morphological categorization. E). Resting [Ca²⁺]_i differs between morphologically normal and abnormal neurites in the calcium-overloaded neurite population (p < 0.001, Mann-Whitney test).

Neuronal Culture Preparation

Primary neuronal cultures were prepared as described elsewhere (Berezovska et al., 2003). In brief, mixed cortical-hippocampal neurons were generated from CD1 mice at embryonic day 15–16. The cells were plated on four-well poly-l-lysine-coated glass slides in chemically defined Neurobasal media (Invitrogen) containing 10% fetal bovine serum for 1 h. The neurons were maintained in Neurobasal media containing 2% B27 supplement (Invitrogen) for 6–12 days in vitro prior to transient transfection of Yellow Cameleon-pAAV construct with FuGene. At 5–7 days in vitro, 5 µg/ml cytosine arabinoside (Sigma) was added to the culture media to suppress the growth of non-neuronal cells.

Yellow-Cameleon AAV preparation

We used the construct pAAV-CBA-YC3.6-WPRE. The plasmid contained the AAV terminal repeats (ITRs), the only remaining feature of the wild-type AAV genome. Flanked by the ITRs, the expression cassette included the following components in 5’ to 3’ order: 1) a 1.7-kb sequence containing hybrid cytomegalovirus (CMV) immediate-early enhancer/chicken beta-actin promoter/exon1/intron; 2) yellow cameleon 3.60 cDNA; and 3) woodchuck hepatitis virus posttranscriptional regulatory element (WPRE). The virus titer was 1.1×10^12 viral genomes/mL.

Animals and Surgery

Four transgenic mouse lines were used. 1) APP/PS1: mutant human PS1- ΔE9 and a chimeric mouse/human APPswe and age-matched non-transgenic controls (Borchelt et al., 1997). Two different age groups of these mice were used (6–8 months with plaques and 3–3.5 months without plaques). 2) Tg-2576: 695 aa isoform of APP containing the Swedish mutation and non-transgenic controls. These mice were used at 17–20 months of age. 3) PS1-ΔE9: mutant human PS1- ΔE9 alone. These mice were used at 5–6 months of age. 4) PS1M146V. All studies were conducted with approval of the Massachusetts General Hospital Animal Care and Use Committee and in compliance with NIH guidelines for the use of experimental animals.

Intracortical injections and cranial window implantation have been described previously (Skoch et al., 2005; Spires et al., 2005). For intracortical injections of YC-AAV2, mice were anesthetized with isoflurane and placed in a custom made, heated stereotaxic apparatus. The surgical site was sterilized with betadyne and isopropyl alcohol, and a 2–3 mm incision was made in the scalp along the midline between the ears. Burr holes were drilled in the skull, 1 mm lateral to bregma bilaterally. Using a Hamilton syringe, 4 µl of virus (titer, 4.2 × 10¹² viral genomes/ml) was injected 1.2 mm deep in somatosensory cortex (targeting layer 5 neurons) at a rate of 0.2 µl/min. After one injection in each burr hole, the scalp was sutured, and the mouse was allowed to recover from anesthesia on a heating pad.

After a > 3-week incubation period to allow expression of YC3.6 in neurons, mice received an intraperitoneal injection of methoxy-XO4 (10 mg/kg), a fluorescent compound that crosses the blood-brain barrier and binds to amyloid plaques (Klunk et al., 2002 ). The following day, a cranial window 6 mm in diameter was installed under isoflurane anesthesia as described previously.(Bacskai et al., 2002; Klunk et al., 2002) In a subset of experiments, Texas Red dextran (70,000 Da molecular weight; 12.5 mg/ml in sterile PBS; Invitrogen) was injected into a lateral tail vein to provide a fluorescent angiogram.

Multiphoton Imaging

Images of YC3.6-filled neuronal processes, amyloid pathology, and blood vessels were obtained using one of two microscopes: 1) a Bio-Rad 1024ES multiphoton microscope (Bio-Rad), mounted on an Olympus (Tokyo, Japan) BX50WI upright microscope (details have been described previously, (Skoch et al., 2005)) or 2) Olympus Fluoview 1000MPE with pre-chirp optics and a fast AOM mounted on an Olympus BX61WI upright microscope. A wax ring was placed on the edges of the coverslip of the cortical window and filled with distilled water to create a well for an Olympus 20x dipping objective (numerical aperture, 0.95). A mode-locked titanium/sapphire laser (MaiTai; Spectra-Physics, Fremont, CA) generated two-photon fluorescence with 860 nm excitation, and detectors containing three photomultiplier tubes (Hamamatsu, Ichinocho, Japan) collected emitted light in the range of 380–480, 500–540, and 560–650 nm (Bacskai et al., 2002). The average power reaching the brain tissue ranged from 20–60 milliwatts. Neurites were typically sampled 0–100 microns below the surface of the brain and, when possible, cell soma were sampled 350–550 microns below the surface.

This work was supported by NIH EB000768 (BJB), AG08487 (BTH), and NIH NS580752 (KVK). We would like to thank Dr. Frosch and Dr. Ohki for reviewing the manuscript, Dr. Miyawaki for providing the YC3.6 plasmid, Laura Borrelli and Katie Bercury for assistance, and Scott Raymond for help with data analysis.