2.1.

Cell Culture and Western Blot Analysis

Primary cortical neurons were prepared from cerebral cortices of CD-1 strain mouse embryos (day 15 to 17 of gestation). Cortices were dissociated by trypsinization for $5\min$ at room temperature, and cells were resuspended in Neurobasal (NB) medium supplemented with 10% fetal bovineserum (FBS), $2\mathrm{mmol}∕\mathrm{L}$ glutamine (Gln), $100\mathrm{U}∕\mathrm{ml}$ penicillin, and $100\mu \mathrm{g}∕\mathrm{ml}$ streptomycin, then centrifuged at $100\mathrm{g}$ for $10\min$ and resuspended in NB/B27 [NB medium containing 2% (v/v) B-27 supplement], $100\mathrm{U}∕\mathrm{ml}$ penicillin, $100\mu \mathrm{g}∕\mathrm{ml}$ streptomycin, and $2\mathrm{mmol}∕\mathrm{L}$ Gln. The cells were then plated at $1\times {10}^6$ on $42\text{-}\mathrm{mm}$ round coverslips (Hemogenix, Colorado Springs, CO), and coated with Poly-D-Lysine $(20\mu \mathrm{g}∕\mathrm{ml})$ in $60\text{-}\mathrm{mm}$ cell culture dishes (Corning, Lowell, MA). Neurons were grown for $7\text{to}10\text{days}$ in vitro and transiently transfected with the indicated constructs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer’s protocol.

H4 human neuroglioma cells were maintained in Opti-MEM (Gibco, Carlsbad, CA) supplemented with 10% (v/v) FBS, $100\mathrm{U}∕\mathrm{ml}$ penicillin, and $100\mu \mathrm{g}∕\mathrm{ml}$ streptomycin (Gibco, Carlsbad, CA). Where indicated, cells were transiently transfected with the identified constructs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer’s protocol.

Tau constructs were subcloned into either mRFP-N1 or eCFP-N1 vectors (Clontech, Mountainview, CA). Using $\mathrm{Tau}3\mathrm{R}(-2-3)$ , $\mathrm{Tau}3\mathrm{R}(+2+3)$ , $\mathrm{Tau}4\mathrm{R}(-2-3)$ , or $\mathrm{Tau}4\mathrm{R}(+2+3)$ in pcDNA3.1(+) as templates, we performed polymerase chain reaction using the following primers: $5^{\prime }$ -CTCGAGATGGCTGAGCCCCGCCAGGAGTTCGAAG- $3^{\prime }$ (forward) and $5^{\prime }$ -GGATCCAAACCCTGCTTGGCCAGGGAGGCAGAC- $3^{\prime }$ (reverse), digested with Xho1 and BamH1 and ligated into appropriate color vectors. The integrity of each construct was verified using DNA sequencing analysis.

H4 human neuroglioma cells were transiently transfected with Lipofectamine 2000 according to the manufacturer’s protocol and collected $48\mathrm{hr}$ post-transfection in $100\text{-}\mu \mathrm{l}$ 2X sodium dodecyl sulfate (SDS) lysis buffer ( $0.25\mathrm{M}$ Tris-Cl, pH 7.5, 2% SDS, $5\mathrm{mM}$ ethylenediaminetetraacetic acid, $5\mathrm{mM}$ ethylene glycol tetraacetic acid, and 10% glycerol supplemented with a protease inhibitor pellet obtained from Gibco, Carlsbad, CA). Cell lysates were then sonicated for $10\sec$ and boiled for $10\min$ . Samples were centrifuged at room temperature for $10\min$ at $\mathrm{15,000}\mathrm{g}$ . Sample concentrations were determined using the bicinchoninic acid assay (Pierce, Rockford, IL). Samples were diluted to $1\mu \mathrm{g}∕\mu \mathrm{l}$ in 2X SDS running buffer and loaded onto precast 4 to 12% Tris-Glycine gradient polyacrylamide gels (Invitrogen, Carlsbad, CA), electrophoresed, and transferred to polyvinylidene fluoride membranes. Membranes were blocked for $1\mathrm{hr}$ at room temp in Tris Buffered Saline Tween-20 (TBST) ( $20\mathrm{mM}$ Tris-Cl, pH 7.6, $137\mathrm{mM}$ NaCl, 0.05% Tween 20) and 5% milk and incubated in TBST and milk at $4°\mathrm{C}$ overnight in the indicated primary antibody. Membranes were then washed 3 times for $10\min$ in TBST at room temperature and incubated for $1\mathrm{hr}$ with appropriate horseradish peroxidase-conjugated secondary antibody in TBST and milk. Membranes were washed 3 times for $10\min$ in TBST at room temperature and visualized with enhanced chemiluminescence (Amersham, Piscataway, NJ).

2.2.

Live Cell Imaging and SHG

Primary neurons on coverslips were fitted in a perfusion or culture (Hemogenix, Colorado Springs, CO) closed chamber with a $1\text{-}\mathrm{mm}$ silicone gasket between glass coverslips. The chamber contained Hank’s balanced salt solution (HBSS) prewarmed to $37°\mathrm{C}$ , and it was affixed in the heated stage of a Zeiss Axiovert 200 inverted microscope. All images were acquired on a Zeiss LSM-510 META confocal microscope fitted with a $\mathrm{He}∕\mathrm{Ne}$ $543\text{-}\mathrm{nm}$ laser for red fluorescence protein (RFP) fluorescence and a Coherent (Santa Clara, CA) chameleon mode-locked Ti/Saph tunable $\left(720\text{to}930\mathrm{nm}\right)$ laser with $\sim 140\text{-}\mathrm{fs}$ pulse width and $\sim 20\text{to}30\mathrm{mW}$ average power with linear polarization at the sample, for cyan fluorescent protein (CFP) and SHG imaging. All images were acquired using a Zeiss $25\times 0.8$ numerical aperture (NA) Plan-NEOFLUAR water/oil immersion lens for epifluorescence and an infinity-corrected Zeiss $20\times 0.75$ -NA PlanAPOCHROMAT dry lens affixed to a custom condenser tube (fabricated by Zeiss, Jana, Germany) affixed to the condenser housing, which projected the forward-propagating signal onto a photomultiplier tube (PMT, Hamamatsu). A $400\pm 25\text{-}\mathrm{nm}$ bandpass filter was fitted between the condenser housing and the PMT. This aided in both reducing background fluorescence at the SHG PMT and in verifying the SHG nature of the signal observed, since $800\text{-}\mathrm{nm}$ light was used to induce the SHG signal. The resulting $400\text{-}\mathrm{nm}$ signal should pass through the filter unobstructed, while producing an SHG signal using $900\mathrm{nm}$ should result in a loss of the SHG signal, because the resulting $450\text{-}\mathrm{nm}$ signal would be blocked by the bandpass filter. The bandpass filter also ensured that no fluorescent signal from the fluorophores would pass through to the PMT, since CFP emits around $450\mathrm{nm}$ and RFP emits around $590\mathrm{nm}$ , and both are blocked by the bandpass filter. To further ensure the signal obtained was from SHG and not autofluorescence, a polarizer was placed in front of the SHG PMT, and the polarizer was rotated to demonstrate the polarized nature of the SHG signal. Since the SHG signal is a polarized signal, rotating the polarizer perpendicular with respect to the laser polarization should result in a loss of SHG signal, in contrast to fluorescence emissions.

Images were collected using Zeiss LSM 3.5 software and exported as JPEG images for analysis in Image J. The Neuron J macro plugin for Image J was used to trace and quantify the SHG signal in all sample images, and the resulting intensity data were analyzed using Graphpad Prism 4 (San Diego, CA) to carry out t-tests or one-way analyses of variance (ANOVAs). Data were presented as $\text{mean}\pm \mathrm{SEM}$ for each condition, and results were considered significant if $\mathrm{p}<0.05$ .

3.1.

Second Harmonic Generation Signal is Detectable in Single Neuronal Processes

Recent work has shown that an SHG signal arises mainly from axonal processes in neurons, not from the cell body or dendrites,²⁶ in acute hippocampal slices. Therefore, we sought to determine whether it was possible to visualize an SHG signal from a single neuronal process in a cultured neuron that would allow us to visualize intact axonal microtubules in a living neuron. Figure 1 shows the morphology of a single neuron transfected with RFP and visualized with the red fluorescence signal. The RFP was excited with a $543\text{-}\mathrm{nm}$ $\mathrm{He}∕\mathrm{Ne}$ laser, and the emitted light at $590\mathrm{nm}$ was directed to the PMT through a $570\text{-}\text{to}610\text{-}\mathrm{nm}$ bandpass filter. Similarly, the SHG signal was obtained with $800\text{-}\mathrm{nm}$ excitation and collected in the forward direction after passing the signal through a $375\text{-}\text{to}425\text{-}\mathrm{nm}$ bandpass filter. A weak SHG signal was observed in single processes extending from each neuron. Several characteristics suggest that this was SHG: the $400\text{-}\mathrm{nm}$ signal was seen only in the axon (as expected given the asymmetric microtubule structure unique to axons), whereas tau and autofluorescent moieties are most strongly seen in the cell body. This signal was lost when treated with colchicine, indicating that it was microtubule dependent. We inserted a polarizer in the light path between the sample and the PMT. Rotating the polarizer between 0 and $90\mathrm{deg}$ should either allow a polarized signal to pass or block a polarized signal, depending on the orientation, while fluorescent signals should be unaffected by the orientation of the polarizer. As expected, when the polarizer was in the $0\text{-}\mathrm{deg}$ position, the SHG signal was strong. However, when the polarizer was shifted to the $90\text{-}\mathrm{deg}$ position, the SHG signal was completely lost. As another test, we changed excitation wavelength to $900\mathrm{nm}$ , and the SHG signal observed in a 375 to 425 emission filter was lost, as would be expected for SHG (Fig. 2 ). Finally, we demonstrated that if the optics were reconfigured to detect the back-propagating signal, the signal was undetectable, again as expected for SHG—which is predominantly forward-propagating, unlike a fluorescent signal.

Fig. 1

Weak SHG signal is obtainable from nontransfected neurons. Photomicrograph showing a neuron transfected with RFP alone and visualized simultaneously in the $543∕590\text{-}\mathrm{nm}$ fluorescent excitation/emission signal (red) and $800∕400\text{-}\mathrm{nm}$ SHG excitation/emission (green). There is very little SHG present in cells that have endogenous levels of mouse tau expressed and that express RFP, but there is signal detectable. Micrograph represents of at least three neurons in each transfection from three separate mice.

Fig. 2

SHG signal but not GFP signal is abolished by blocking polarized light with a polarizer. To conclusively demonstrate the SHG nature of the signal obtained in the SHG channel, a polarizer was placed in the light path of the SHG signal, and the orientation of the polarizer was positioned at either 0 or $90\mathrm{deg}$ . Both the GFP and SHG signals were obtained from $\mathrm{GFPTau}4\mathrm{R}(+2+3)$ -transfected neurons. As expected, when the polarizer was changed and the GFP signal of tau-transfected neurons was obtained, there was no difference in the GFP signal. Conversely, the SHG signal was present only in the $0\text{-}\mathrm{deg}$ position of the polarizer and was abolished when it was set to $90\mathrm{deg}$ , indicating the polarized nature of the SHG signal. This also rules out autofluorescent signals contaminating the SHG signal, because they would not be affected by the position of the polarizer. Micrograph represents of three neurons transfected with tau.

3.2.

Overexpression of Tau Increases the SHG Signal

Next, we examined whether altering tau levels would enhance the SHG signal. As a first step, we characterized tau expression in H4 cells. Figure 3 shows a western blot analysis of various tau constructs expressed in H4 cells to verify the biochemical properties of each construct. The different isoforms’ molecular weights corresponded to the presence or absence of the fourth microbutubule binding repeat (4R, representing exon 10) or the 2 amino-terminal inserts $(+2+3)$ . All constructs migrated to the expected molecular weight, indicating that the proteins were expressed and processed appropriately in cells.

Fig. 3

Biochemical properties of tagged tau constructs are normal. Representative blot showing various tagged tau constructs that were transiently transfected into H4 human neuroglioma cells. (a) Cell lysates were immunoblotted and probed with a pan tau antibody (AB-3). Western blot shows that all expressed tau migrates to the expected molecular weight and that the presence of a C-terminal fluorescent protein (CFP or RFP) does not significantly alter tau processing. All constructs were transfected utilizing $1\mu \mathrm{g}$ of DNA per condition, and the total lysate protein concentration was normalized for each sample. (b) Expression of CFP in H4 neuroglioma cells leads to robust expression of CFP and fluorescence throughout the entire cell, with no SHG signal detectable in transfected cells. (c) Imaging of fluorescently tagged tau shows that tau is localized throughout the entire cell. SHG signal, however, comes from filamentous structures containing both tau and tubulin that encircle the nucleus and continue into the processes. Note specifically that the microtubule network seems to reorganize into circular structures in transfected H4 cells, and that robust SHG signal is present in more than one process. Photomicrographs are representative of at least three cells from each of at least three separate preparations.

We examined whether microtubules in H4 neuroglioma cells, which are morphologically flat with several long processes, would show SHG signal. No SHG signal could be detected from H4 cells transfected with CFP [Fig. 3b]. By contrast, imaging of H4 cells transfected with tau showed robust tau expression in the cell body and processes; interestingly, the SHG signal arose primarily from perinuclear compartments where tau levels were highest, and where the microtubule network appears to reorganize within the perikeria [Fig. 3c].

Next we examined whether transfection of tau would enhance SHG in neurons, and if so, in which cellular compartment. Given that tau is a microtubule binding protein that promotes the assembly and stability of microtubules in axons (reviewed in Ref. 27), it might follow that expression of tau would increase axonal SHG signal. To further test this hypothesis, and also to determine if the presence of a fluorescent tag on tau would change the SHG, we transfected neurons with different isoforms of either TauRFP, untagged tau cotransfected with RFP, or RFP alone. SHG signal was measured by tracing SHG positive neurites using the Neuron J macro for Image J and analyzing the intensity across the entire length of the process.

Since transfection efficiency was very low in cultured neurons with our transfection method, it was possible to visualize the processes of a single neuron and follow them for many hundreds of microns without interference from the processes of neighboring, untransfected neurons. When TauRFP-transfected neurons were examined, the RFP fluorescence and SHG signals were distinct [Fig. 4a ]: SHG signal was present in only one process, whereas the TauRFP fluorescence signal was present in multiple processes and the cell body. To verify that the signal obtained in the SHG channel was indeed SHG and not contaminating fluorescence, the wavelength of the excitation laser was changed from $800\mathrm{nm}\text{to}900\mathrm{nm}$ . As expected, the SHG signal was abolished, since the bandpass filter in front of the SHG detector blocked $450\text{-}\mathrm{nm}$ emission [Fig. 4b]. As a further control, we added a $450\text{-}\mathrm{nm}$ bandpass filter to the turret, and we were able to switch between the $400\text{-}\mathrm{nm}$ and $450\text{-}\mathrm{nm}$ filters. Reconfiguring the detectors demonstrated that the signal was substantially forward-propagating and could not be detected in the back-propagating configuration (data not shown). As expected, when we used $800\text{-}\mathrm{nm}$ light and the $400\text{-}\mathrm{nm}$ filter, there was a detectable SHG signal. To test the hypothesis that the SHG signal observed was due to microtubule structures, sister cultures of TauRFP-transfected neurons were treated with $100\text{-}\mu \mathrm{M}$ colchicine for $30\min$ and imaged as above. Since colchicine completely disrupts microtubule stability, any SHG signal arising from microtubules should be attenuated. As expected, treatment of cells with colchicine led to the complete loss of SHG signal [Fig. 4c]. These results clearly demonstrate that SHG signal is obtainable from a single neuronal process that is presumably the axon, and that the source of SHG signal within that process is the microtubule network.

Fig. 4

Tau enhances SHG signal in transfected neurons. Embryonic day 15 mouse primary neurons were transfected with TauRFP for $48\mathrm{hr}$ , and individual transfected cells were examined for SHG signal. (a) Transfected neuron showing both TauRFP expression throughout neuron and SHG signal at $800\mathrm{nm}$ in a single process, which is presumably the axon. Final panel shows merged image. (b) Same cell as in (a) but showing TauRFP signal in the first panel and SHG signal when a $900\text{-}\mathrm{nm}$ excitation wavelength is used in the second panel, thus demonstrating that SHG signal is not detectable when $900\text{-}\mathrm{nm}$ excitation is used, as expected, because the bandpass filter before the SHG-detecting PMT cuts out all light above $425\mathrm{nm}$ . Third panel shows merged image. Micrographs are representative of at least three neurons per transfection from at least three separate mouse primary neuronal culture preparations. (c) Sister cultures of primary neurons as in (a) and (b) that were transfected with TauRFP and also treated with $100\mu \mathrm{M}$ colchicine for $30\min$ prior to imaging, which completely depolymerized the microtubules. First panel shows TauRFP signal in a cell and a similar amount of TauRFP expression as in (a) and (b). Second panel shows SHG signal, which is completely abolished as expected, indicating the microtubule nature of the SHG signal. Third panel shows merged image. Micrographs are representative of at least three transfected neurons for three separate mouse primary culture preparations.

As an additional control, we inserted a Berek compensator (New Focus, San Jose, CA) into the excitation light path (to circularly polarize the laser source) and examined the SHG signal from the processes. Figure 5 shows that, as expected, even when the excitation beam was circularly polarized, only one process showed a strong SHG signal.

Fig. 5

Using circularly polarized light does not alter pattern of SHG in transfected neurons. Representative photomicrograph showing a single mouse neuron transfected with $\mathrm{Tau}4\mathrm{R}(+2+3)\mathrm{CFP}$ and visualized with $800\text{-}\mathrm{nm}$ two-photon excitation. (a) TauCFP localization throughout the entire cell, including all processes and the cell body. (b) The forward-propagating SHG signal after circularly polarizing the excitation source, which should eliminate the dependence of the angle of orientation seen with linearly polarized light. Note that only a single process is still visible in the SHG channel, confirming that the primary neurite that is visible with SHG is the axon. (c) The merging of the first two images and co-localization of the SHG signal with tau in the axon, again indicating that TauCFP enhances SHG in only the axon. Micrographs are representative of at least three separate neurons from at least three separate transfections, all showing that circularly polarized light does not significantly affect the pattern of SHG signal.

Expression of either tagged or untagged tau led to a significant and equal increase in SHG signal (Fig. 6 ) over that of RFP alone ( $\mathrm{RFP}=15.0\pm 1.5$ , $\mathrm{TauRFP}=32.1\pm 4.6$ , and $\mathrm{Tau}+\mathrm{RFP}=31.7\pm 2.6$ ; $\mathrm{p}<0.001$ for both RFP versus TauRFP and RFP versus $\mathrm{Tau}+\mathrm{RFP}$ ). The presence of either RFP or CFP fusion constructs on the C-terminus of tau did not further enhance or decrease the SHG signal from microtubules when compared to tranfection with untagged tau $(\mathrm{p}>0.05)$ .

Fig. 6

Expression of either a tagged tau or untagged tau results in increased SHG. Mouse primary neurons were transiently transfected with either RFP alone, TauRFP, or $\mathrm{Tau}+\mathrm{RFP}$ , and SHG signal was quantitated for each condition. The expression of RFP alone resulted in a modest amount of SHG signal $(\text{mean}\pm \mathrm{SEM}=15\pm 1.5)$ while expression of either TauRFP $(32.1\pm 4.6)$ or $\mathrm{Tau}+\mathrm{RFP}$ $(31.7\pm 2.6)$ resulted in a statistically significant increase in SHG signal; $\mathrm{p}<0.001$ for each when compared to RFP alone ( $n=15$ for each condition), $n=25$ neurons/condition. Interestingly, TauRFP and $\mathrm{Tau}+\mathrm{RFP}$ were not significantly different.

One possibility that could account for increased SHG signal from tau-transfected axons is that expression of tau significantly increases the overall diameter of the axon, and therefore increases the signal by giving rise to a signal from a larger axonal cross-section. To rule this out, images of axons of both RFP- or TauRFP-transfected neurons were examined in Image J, and the axonal diameter was measured at least $20\text{-}\mu \mathrm{m}$ distal to the axon hillock and before the first major bifurcation of the axon. Three separate measurements per axon were averaged, and a two-tailed t-test revealed that there was no significant difference between RFP- and TauRFP-transfected axons. Therefore, tau expression does not increase SHG simply by increasing the size of the axon (data not shown).

3.3.

Different Tau Isoforms Do Not Significantly Differ in Enhancing SHG

In in vitro assays, 4R tau binds microtubules with a greater affinity and polymerizes the formation of microtubules to a greater extent than 3R tau^{16, 28} (reviewed in Ref. 29). Having shown that tau enhances SHG signal in neurons, we next sought to determine whether different tau isoforms would differentially affect SHG signal. Tau-CFP constructs were generated for ease of planned co-transfection experiments, and in all instances they were equivalent to the transfection protocol utilized for the Tau-mRFP constructs used above. Neuronal cultures were transfected with one of the following constructs: CFP, $\mathrm{Tau}3\mathrm{R}(-2-3)\mathrm{CFP}$ , $\mathrm{Tau}3\mathrm{R}(+2+3)\mathrm{CFP}$ , $\mathrm{Tau}4\mathrm{R}(-2-3)\mathrm{CFP}$ , or $\mathrm{Tau}4\mathrm{R}(+2+3)\mathrm{CFP}$ . SHG signal was determined for each condition. As expected, the expression of any isoform of tau enhanced the observed SHG signal compared to empty-vector or CFP-transfected neurons (Fig. 7 ). Surprisingly, however, there was no significant difference between exon 10-containing or exon 10-skipping constructs, indicating that the presence of 3 or 4 microtubule binding repeats did not significantly affect tau’s ability to enhance microtubule SHG signal in living neurons. No morphological changes in axon shape were noted qualitatively, and measuring the axon diameter revealed no difference between tau-transfected and control neurons.

Fig. 7

Different tau isoforms do not significantly differ in their ability to enhance SHG. Primary mouse neurons were transiently transfected with the indicated tau isoform fused to CFP or CFP alone and SHG signal was measured. Note that CFP alone was different from all groups, and that there were no intergroup differences that were the result of tau isoform, or even the mutated P301L form of tau. CFP $\text{mean}\pm \mathrm{SEM}=6.12\pm 1.48$ , $\mathrm{Tau}3\mathrm{R}(-2-3)\mathrm{CFP}=28.35\pm 7.85$ , $\mathrm{Tau}3\mathrm{R}(+2+3)\mathrm{CFP}=27.34\pm 5.31$ , $\mathrm{Tau}4\mathrm{R}(-2-3)\mathrm{CFP}=19.88\pm 3.28$ , $\mathrm{Tau}4\mathrm{R}(+2+3)\mathrm{CFP}=36.65\pm 9.26$ , $\mathrm{TauP}301\mathrm{L}\text{-}\mathrm{CFP}=26.85\pm 7.45$ , and $\mathrm{n}=20$ to 30 neurons per condition.

Mutations in the tau gene found in FTDP-17 have been shown to decrease the in vitro association of tau with microtubules and the rate of microtubule polymerization. ^{13, 17, 30, 31, 32, 33, 34} To determine if tau mutations alter axon-derived SHG in living neurons, we transfected neurons with $\mathrm{P}301\mathrm{LTau}4\mathrm{R}(+2+3)\mathrm{CFP}$ and examined the effect on SHG signal. Surprisingly, the mutant tau enhanced the SHG signal observed in the transfected neurons (Fig. 8 ) to essentially the same extent as other tau isoforms.

Fig. 8

FTDP-17 mutant tau does not differentially impact SHG generation.Primary mouse neurons were transiently transfected with indicated tau construct. Surprisingly, expression of either TauRFP or P301L-Tau4RCFP resulted in a high level of SHG signal with $\text{mean}\pm \mathrm{SEM}$ being $33.5\pm 4.7$ and $37.2\pm 7.4$ , respectively, and $\mathrm{n}=20$ to 30 neurons per condition.