1.

Introduction

Traumatic brain injury (TBI) remains the most common cause of death in persons under age 45 in the Western world. The societal impact is profound, with 2 million cases, 220,000 hospitalizations, and 52,000 deaths from head trauma occurring each year in the United States.^{1 2} One of the principal determinants of morbidity and mortality following TBI is traumatic axonal injury (TAI). While multiple tissue and animal studies have proposed various pharmacological or physiological interventions to reduce axonal injury, these studies have not translated into novel therapeutic options in the clinical setting.³ The development of new methods for detecting protein interactions and elucidation of critical mechanisms of injury in animals and humans will greatly enhance the transition of promising treatments from benchtop to bedside.

Current hypotheses on the pathogenesis of TAI, based on previous immunohistochemical and histochemical studies, involve activation of apoptotic cascades secondary to TBI (Ref. 4). A pivotal element of these hypotheses involves^{5 6} a potential interaction between the proapoptotic protein BAD and prosurvival protein Bcl-xL. The interaction between BAD and Bcl-xL provide a theoretical link between studies demonstrating overt axolemmal disruption with presumed loss of ionic homeostasis and studies demonstrating mitochondrial disruption with cytochrome c release and activation of caspase proteolytic enzymes in relation to local axonal breakdown.⁷ This proposed theoretical link involves the Ca ²⁺ -mediated activation of second messengers such as calmodulin secondary to the loss of ionic homeostasis across a traumatically perturbed axolemma. Targets of calmodulin include the phosphatase calcineurin, which can dephosphorylate and activate the proapoptotic protein BAD. In in vitro models of apoptosis and models of spinal cord injury, dephosphorylated BAD has been shown to translocate to the mitochondria and bind to the prosurvival protein Bcl-xL, leading to the breakdown of mitochondria.⁸ As mitochondrial breakdown with release of cytochrome c accompanied by activation of the caspase cascade of proteolytic enzymes has been demonstrated in relation to local axonal breakdown,^{7 9} calcineurin-mediated dephosphorylation of BAD with displacement of Bcl-xL has been hypothesized to play a role in traumatic axonal pathogenesis.

Current techniques for exploring protein interactions in tissue are limited. Techniques such as coimmunoprecipitation, affinity chromatography, and two-hybrid yeast systems require the homogenization of gross samples of tissue. Thus, while providing information on overall changing trends of protein interactions, they provide little information on how protein interactions occur at the individual cellular level. Neither conventional immunohistochemical/histochemical nor ultrastructural approaches have the capacity to shed insight on whether BAD and Bcl-xL interact in TAI in vivo.

In this paper, we describe an approach for employing fluorescence resonance energy transfer (FRET) microscopy in tissue in an in vivo model of neurological disease using aldehyde-fixed rat brain tissue sections. FRET is a quantum mechanical property detected by light microscopy involving radiationless energy transfer from an excited donor fluorophore to an acceptor fluorophore directly through resonance interactions across limited distances.¹⁰ While FRET microscopy has yielded significant insights in vitro in determining protein interactions in cell culture preparations,^{11 12 13} it is a tool that remains relatively unexplored in vivo with tissue analysis. Specifically, the current investigation employs confocal and two-photon excitation FRET microscopy to explore protein interactions in TAI following TBI. While there are in vitro models to assess the effect of stress and strain on individual axons and neuronal cell bodies following TBI, the full spectrum of injury resulting from the complex biomechanical interactions among glia, neurons, axons, and the extracellular milieu cannot be adequately assessed in vitro. As such, the capacity to employ FRET microscopy in studying TAI following TBI constitutes a significant advance in determining the pathogenesis of TAI.

2.

Materials and Methods

3.

Results

The development of a tissue FRET methodology required refinement of three processes, including immunofluorescent labeling of tissue with antibodies directed toward donor and acceptor molecules, acquisition of images using C-FRET and 2p-FRET microscopy, and software-aided pixel-by-pixel analysis of the resulting images. As already described in Sec. 2, we performed a series of preliminary immunohistochemical dilution and control experiments to determine the appropriate working antibody concentrations and to verify the absence of nonspecific or cross-reactive interactions that would confound our data analysis. Brightfield dilution studies with BAD and Bcl-xL antibodies revealed ubiquitous labeling, a 1:200 dilution of BAD intensely labeled soma, and moderately labeled axons, while a 1:500 dilution of Bcl-xL intensely labeled both axons and soma. Fluorescent dilution studies revealed similar results, a 1:50 dilution of BAD antibody labeled with a 1:200 dilution of anti-mouse Alexa 488 IgG revealed moderate labeling of axons and soma, a 1:200 dilution of Bcl-xL antibody labeled with a 1:200 dilution of anti-rabbit Alexa 555 IgG intensely labeled both axons and soma (Fig. 1).

Figure 1

BAD and Bcl-xL heterodimerization. The heterodimerization of BAD labeled with Alexa 488 and Bcl-xL labeled with Alexa 555 approximates the fluorophores within 100 Å of each other, enabling fluorescent resonance energy transfer.

Control fluorescent immunohistochemical studies were performed to ensure that cross-reactivity did not exist between the secondary antibodies and the tissue, and to ensure that the species-specific secondary antibodies did not label the other primary antibody epitopes. Tissue incubated in the secondary anti-rabbit and anti-mouse antibodies at 1:200 dilutions revealed no labeling of axons or soma. Likewise, tissue incubated in either mouse-BAD or rabbit-Bcl-xL, and then labeled with the opposite species secondary antibody revealed no labeling of axons or soma.

Following determination of the optimal dilutions, we prepared specimens for PFRET analysis. Using serial sections of tissue, we simultaneously labeled injured tissue with donor only (BAD/Alexa 488), acceptor only (Bcl-xL/Alexa 555), and with both donor and acceptor. The specimens were then examined for axons demonstrating vacuolization or formation of retraction bulbs, morphological characteristics of axonal injury. Using the laser scanning confocal microscope (LSCM) with the excitation lasers and emission filters as described above, the seven images (Fig. 2) (see Color Plate 1) of these axons required for PFRET analysis were collected as shown in Fig. 2. The uncorrected FRET image [Fig. 2(f)] underwent image correction for DSBT and ASBT, resulting in the PFRET image [Fig. 2(h)]. Multiple regions of interest (ROIs) were then drawn within the imaged axon, the efficiency of energy transfer, and the estimated distance between fluorophores was calculated for each ROI with PFRET software, and the mean and standard deviation of these values were calculated.

Figure 2

Seven images required for PFRET analysis. Six-hour-postinjury tissue was prepared with mouse anti-BAD IgG ₁ labeled with Alexa 488 anti-mouse antibody (donor) and/or rabbit anti-Bcl-xL IgG labeled with Alexa 555 anti-rabbit IgG (acceptor): (a) donor labeled, donor excitation, donor channel; (b) donor labeled, donor excitation, acceptor channel; (c) acceptor labeled, donor excitation, acceptor channel; (d) acceptor labeled, acceptor excitation, acceptor channel; (e) double labeled, donor excitation, donor channel; (f) double labeled, donor excitation, acceptor channel; (g) double labeled, acceptor excitation, acceptor channel; and (h) PFRET image following processing as described. Calculated mean energy transfer efficiency was 22.4±2.2 and estimated distance was 83.0±1.7 Å. Histograms represent pixel intensity for identical ROI for uncorrected FRET (i) and PFRET (j) images. Bar=10 μm.  Color Plate 1

Based on previous experiments, we hypothesized that following traumatic axonal injury, increased levels of intracellular Ca ²⁺ should result in the activation of calcineurin, subsequent dephosphorylation of BAD, and translocation of BAD to the mitochondria, followed by the proapoptotic binding of BAD to Bcl-xL. According to this hypothesis, prior to injury, BAD and Bcl-xL should remain separate, the distance between them should be greater than 100 Å, and no FRET signal should be detected. Conversely, following injury, if heterodimerization exists between BAD and Bcl-xL, the distance between them should be less than 100 Å, resulting in FRET signal detection.

In tissue processed from animals receiving a sham injury, we observed axons with normal morphologies and ubiquitous labeling with both BAD and Bcl-xL. Images for PFRET analysis (Fig. 3) (see Color Plate 1) were acquired using the LSCM system already described above, including images from both the donor [Fig. 3(a)] and acceptor [Fig. 3(b)] channel in the double-labeled tissue excited at 488 nm. The images were then processed using the image analysis software to produce a PFRET image [Fig. 3(c)], ROIs were drawn, and the mean energy transfer efficiency was 7.24±1.9, and an estimated distance greater than 100 Å (Fig. 6).

Figure 3

Laser scanning confocal images acquired from control (a) to (c) or injured animals at 6 h postinjury (d) to (f). Following immunohistochemical processing with mouse anti-BAD IgG ₁ labeled with Alexa 488 anti-mouse antibody (donor) and rabbit anti-Bcl-xL IgG labeled with Alexa 555 anti-rabbit IgG (acceptor), images required for PFRET analysis were acquired. In sham injured animals, axons demonstrate ubiquitous labeling for BAD (a), low uncorrected acceptor emission (b), and low efficiency of transfer following PFRET image correction (c). In axons from animals 6 h postinjury, BAD likewise demonstrates ubiquitous labeling of axons (d), however, significant acceptor emission occurs with donor excitation (e), and PFRET image correction demonstrates significant energy transfer (f) with an efficiency of 23.1±2.1 and estimated distance of 82.5±2.6 Å. Bar=10 μm.

In contrast to our findings in noninjured animals, tissue processed from animals allowed to survive for 6 h after injury demonstrated ubiquitously labeled axons with vacuolization and formation of retraction bulbs. Images of these axons were acquired, including the donor channel [Fig. 3(d)] and uncorrected FRET [Fig. 3(e)] images, and a PFRET image [Fig. 3(f)] was produced. ROIs were drawn, and the mean energy transfer was 29.03±6.63, significantly greater than the noninjured animals (p<0.01), with an estimated distance of approximately 78.7±4.2 Å (Fig. 6).

Two-photon microscopy has advantages over confocal microscopy, including reduced focal volume and less light scattering in specimen.^{17 18} To determine if images acquired using 2p-FRET microscopy would demonstrate FRET efficiencies similar to C-FRET, we utilized the same confocal microscope setup coupled with an inline tunable Ti:sapphire laser. With this setup, we were able to acquire confocal images, then switch to 2p microscopy and acquire images of the same axons in the same focal plane (Fig. 4).

Figure 4

Multiphoton images acquired from animals subjected to a sham injury (a) to (c) or 6 h postinjury (d) to (f). Following immunohistochemical processing with mouse anti-BAD IgG ₁ labeled with Alexa 488 anti-mouse antibody (donor) and rabbit anti-Bcl-xL IgG labeled with Alexa 555 anti-rabbit IgG (acceptor), images required for PFRET analysis were acquired. In sham injured animals, axons demonstrated ubiquitous labeling for BAD with excitation with a multiphoton laser tuned to 790 nm (a), low uncorrected acceptor emission with 790-nm multiphoton excitation (b), and low efficiency of transfer following PFRET image correction (c). In axons from animals 6 h postinjury, BAD likewise demonstrates ubiquitous labeling of axons (d), however, significant acceptor emission occurs with donor excitation (e), and PFRET image correction demonstrates energy transfer from donor to acceptor with calculated efficiency of 32.4±6.5 and estimated distance of 76.5±3.8 Å. Bar=10 μm.

In tissue from animals receiving a sham injury, we observed normal axonal profiles with ubiquitous labeling of BAD and Bcl-xL. Images of these axons were acquired, including donor channel emission image [Fig. 4(a)] and uncorrected FRET image [Fig. 4(b)] in double-labeled tissue excited with a 790-nm laser (see Color Plate 2). Donor and acceptor excitation wavelength was selected according to the methodology described in the literature.¹⁶ Images were processed using image analysis, resulting in the PFRET image [Fig. 4(c)]. ROIs were drawn, the mean transfer of energy was calculated at 8.2±1.1 and the estimated distance was greater than 100 Å. In tissue from 6-h postinjury tissue, we observed vacuolated axon segments and sporadic retraction bulbs. Images of these axons were acquired, including donor channel [Fig. 4(d)] and uncorrected FRET [Fig. 4(e)]. and the PFRET image [Fig. 4(f)] was produced. These axons demonstrating morphological characteristics of injury had mean energy transfers greater than 32.4±6.5, and estimated distance of 76.5±3.8 Å (Fig. 6).

Figure 6

Bar graphs of FRET efficiency between fluorophore labeled BAD and Bcl-xL at 6 h postinjury. FRET efficiencies (a) of both confocal and 2p microscopy were significantly (^*,p<0.01) greater at 6 h postinjury compared to sham injured animals. There was no significant difference in efficiencies between confocal or 2p methods in either sham (p=0.34) or at 6 h postinjury (p=0.24) Estimated distance (b) between BAD and Bcl-xL was significantly (^*,p<0.01) decreased at 6 h, as compared to sham injured. There was no significant difference between confocal or 2p methods in distance estimations.

An additional advantage of 2p-FRET microscopy is the ability to penetrate tissue specimens, and visualize deep focal planes. To determine if we could image axons in tissue sections greater than 40 μm, we prepared tissue specimens 60, 80, and 100 μm thick with an adjustable vibratome. After immunohistochemical labeling of the specimens, we imaged the proximal surface of the specimens using C-FRET microscopy. Then, using a calibrated z-focus motor, we focused on the distal surface of the specimen by changing the z-plane focus 60, 80, or 100 μm, respectively.

Using the tunable Ti:sapphire laser, the images required for FRET analysis were acquired (Fig. 5) (see Color Plate 2). Axons ubiquitously labeled for both BAD and Bcl-xL were visualized in all tissue specimens. However, the relative number of axons seen decreased as the thickness increased. Likewise, the amount of background fluorescence decreased as the thickness increased. This suggests that as the thickness of the tissue increased, the emission signal decreased, and we therefore were able to visualize only the most intensely labeled axons. In 6-h-postinjury tissue, vacuolated axons were imaged for PFRET analysis, ROIs were drawn, mean efficiencies calculated, and the distances estimated. At 60, 80, and 100 μm, we obtained FRET efficiencies greater than 20, and estimated distances of approximately 80 Å, consistent with both C-FRET and 2p-FRET data from the proximal surface of 40-μm tissue

Figure 5

Multiphoton images from 6-h-postinjury tissue in 60- (a) to (c), 80- (d) to (f), or 100- (g) to (i) μm-thick tissue. Following immunohistochemical processing with mouse anti-BAD IgG ₁ labeled with Alexa 488 anti-mouse antibody (donor) and/or rabbit anti-Bcl-xL IgG labeled with Alexa 555 anti-rabbit IgG (acceptor), images required for PFRET analysis were acquired. In 6-h-postinjury animals, axons demonstrated ubiquitous labeling for BAD with excitation with a multiphoton in all tissue thicknesses (a) 60, (d) 80, and (g) 100 μm, significant acceptor emission is detected with donor excitation in all thicknesses (b), (e) and (h), and PFRET image correction demonstrates energy transfer from donor to acceptor (c), (f), and (i) with mean efficiency and estimated distance in angstroms. Bar=10 μm.      Color Plate 2

4.

Discussion

FRET microscopy is a powerful tool for visualizing protein-protein interactions under physiological conditions. In this paper, we report a reliable and repeatable methodology for FRET microscopy employing conventional immunohistochemical approaches in fixed tissue. The development of a feasible method for employing FRET microscopy in fixed tissue with conventional immunohistochemistry constitutes a significant advance over prior preparatory approaches.

Specifically, to this point, FRET microscopy has primarily been employed to explore protein interactions in vitro. In this setting, expression vectors are constructed containing proteins of interest adjacent to sequences encoding donor and acceptor fluorophores such as green fluorescent protein (GFP) and yellow fluorescent protein (YFP). The two vectors are introduced in vitro and expressed. As the proteins of interest interact, they reside within 10 to 100 Å of each other and radiationless interactions occur between donor and acceptor fluorophores that can be detected by FRET microscopy. This technique was employed to examine receptor-ligand interactions, including binding between the estrogen receptor α- and β-subunit and steroid receptor coactivators in the presence of estradiol.¹⁹ Similarly, ligand-independent ErbB1 receptor activation following focal application of EGF has also been assessed through this approach.²⁰

In addition to ligand-dependent and -independent interactions, the creation of genetic constructs with adjacent fluorophores designed for FRET microscopy has been used to assess transient protein assemblies. An example includes receptor-mediated formation of the G protein heterotrimer through labeling of the α-subunit and the β-subunit.²¹ Further, Everett et al.;²² used this approach to demonstrate that the myxoma poxvirus protein M11L prevents apoptosis by interacting with the mitochondrial transition pore. FRET pair constructs have also been used to produce an assay to detect caspase activation.²³

In addition to the design of expression vector constructs with respective donor and acceptor molecules adjacent to proteins of interest, immunocytochemical approaches have been employed using donor and acceptor fluorophores conjugated directly to primary antibodies.²⁴ An example of how this approach has been utilized includes the characterization of epitope-specific anti-platelet antibodies.²⁵ Further, Kam et al.;²⁶ utilized this direct immunocytochemical methodology to map cytoskeletal components of adherens junctions by demonstrating interactions between vinculin and talin. Kenworthy et al.;²⁷ used antibodies linked to GPI-anchored proteins to characterize plasma membrane microdomains. Bruno et al.;²⁸ developed a highly sensitive assay for Bacillus spores and E. coli O157:H7 by measuring decreases in FRET activity. Additionally, Sharma et al.;²⁹ used antibodies against Torsin A and α-Synuclein to demonstrate the association between these proteins in a form of inherited Parkinson’s disease.

Both the design of constructs with fluorophores incorporated into fusion proteins and direct immunohistochemical approaches have yielded significant insights into a wide variety of protein interaction in relation to intact cells. However, the currently discussed technique for employing conventional immunocytochemical approaches in fixed tissue for FRET microscopy overcomes a number of limitations presented the previous approaches. Specifically, the construction and introduction of fusion proteins with incorporated fluorophores requires the adequate delivery, incorporation, and expression of donor and acceptor molecules in living systems prior to their characterization under FRET microscopy. In contrast, the currently described method of tissue FRET microscopy does not require the introduction or incorporation of any exogenous materials to cells or tissue while in the living state.

Further, use of primary antibodies directly conjugated to fluorophores is a more wasteful and time-consuming process than using conventional immunohistochemical approaches with primary and secondary antibodies. Specifically, as primary antibodies are only rarely available directly conjugated to a fluorophore, a preliminary procedure must be carried out to conjugate a fluorophore to the primary antibody of interest. This procedure is not only time consuming, but also results in the wasting of a proportion of the primary antibody that is not conjugated to fluorophore as a function of the process. Further, using primary antibodies directly conjugated to fluorophores does not allow for the amplification of signal that is typically afforded by using primary and secondary antibodies, resulting in the necessity for higher initial concentrations of the primary antibody. Both the incomplete conjugation of primary antibody and the requisite for higher primary antibody concentrations are significant considerations in experiments where the primary antibodies may be rare and/or costly.

An important element of the current approach is the use of pixel-by-pixel analysis of FRET images acquired by either C-FRET or 2p-FRET microscopy. The use of PFRET software to perform pixel-by-pixel analysis enables the quantitative detection and subtraction of DSBT and ASBT from FRET images while assessing the degree of FRET interaction. While FRET analysis can be conducted in tissue using photobleaching techniques to decrease the level of DSBT and ASBT (Ref. 29), pixel-by-pixel analysis has distinct advantages over photobleaching. Specifically, with pixel-by-pixel analysis the tissue specimen receives a low energy exposure during two brief sequential image acquisitions. Photobleaching instead requires an extended exposure to in the acceptor excitation spectral range to quench the acceptor molecule. In addition to the decreased ability to assess and subtract DSBT and ASBT in a quantitative and controlled fashion, we observed tissue alteration and destruction following exposure to extended and high excitation energy levels such as those associated with photobleaching.

An additional element of the current communication is the use of 2p-FRET microscopy to localize the protein-protein interactions in tissue. While confocal microscopy provides a powerful tool for obtaining the images required for tissue FRET, 2p microscopy has several advantages. First among these is the reduced focal volume imaged in 2p microscopy. Confocal microscopy uses a pinhole to reduce the focal plane by blocking out of focus emissions; 2p microscopy instead accomplishes a decreased focal volume by directly reducing the excitation focal plane depth. A second advantage of 2p microscopy is that the use of lower frequency wavelengths to excite the fluorophores results in a lower energy load to the specimen. While this may not be of major importance in the fixed specimens used in this study, it may be a critical factor in studies involving living tissue. A third advantage of 2p microscopy is the increased ability to penetrate and visualize deeper tissue. In this experiment, we analyzed tissue up to 100 μm thick, obtaining good-quality images of individual axons, and were able to determine FRET efficiencies consistent with thinner tissue and with confocal data.

In this paper, we report on the use of tissue FRET microscopy to evaluate the interactions between BAD and Bcl-xL in TAI following TBI. TAI is one of the most devastating components of TBI involving the widespread but diffuse disconnection of axons in the hours to days following a traumatic insult. One of the primary challenges to the exploration of TAI is the diffuse nature of the event. While the damage occurs at a broad enough scale to severely disrupt neurophysiological processes, the diffuse nature of the event, occurring directly adjacent to normal tissue, has primarily confined investigations on the pathogenesis of TAI to immunocytochemical approaches. A further limitation to the study of TAI is that investigations are optimally carried out in vivo. While there do exist in vitro models aimed at describing how stress and strain may affect individual axons and neuronal cell bodies following TBI, injury resulting from the complex biomechanical interactions between glia, neurons, axons, and the extracellular milleu cannot be adequately assessed in vitro.

Our current hypothesis on the pathogenesis of TAI based on previous immunocytochemical and histochemical studies involves an initial loss of ionic homeostasis secondary to traumatically induced axolemmal perturbation. This loss of ionic homeostasis leads to activation of Ca ²⁺ -mediated second messenger processes,⁶ potentially including those mediated by Calmodulin. Calmodulin activates a number of targets including the phosphatase calcineurin, which can dephosphorylate and activate BAD, a proapoptotic protein. Dephosphorylated BAD travels to the mitochondria and binds to a prosurvival protein, Bcl-xL, leading to the breakdown of mitochondria.⁷ Mitochondrial breakdown and release of cytochrome c results in the activation of the caspase cascade of proteolytic enzymes, culminating in local axonal breakdown.^{8 9}

Neither conventional immunocytochemical/histochemical nor ultrastructural approaches have the capacity to shed insight on whether BAD and Bcl-xL interact in TAI in vivo. The development of a tissue FRET technique enabled us to demonstrate BAD and Bcl-xL interaction while preserving the complex biomechanical interactions and tissue architecture between glia, neurons, axons, and the extracellular milleu. In this paper, we have demonstrated that following TAI, BAD and Bcl-xL interact in axons displaying the morphological characteristics of injury, thus suggesting the interaction between these two molecules may play an important role in TAI pathogenesis.