1.

Introduction

It has been long recognized that mitochondria play a critical role in living cells by generating ATP, maintaining cellular redox potential, and detoxifying reactive oxygen species (ROS) and xenobiotics.¹ More recently, the roles of mitochondria in generating apoptotic signals and the permeability transition in acute necrotic injury have been described. Because of the high energy demands of neurons and the generation of high levels of ROS, the nervous system is particularly vulnerable to mitochondrial dysfunction. Mitochondria in neurons play significant roles in intracellular ${\mathrm{Ca}}^{2+}$ homeostasis by sequestering^{2, 3} ${\mathrm{Ca}}^{2+}$ and regulating synaptic transmission.⁴ In general, mitochondrial dysfunction can lead to a multisystem disorder that includes the central nervous system, peripheral nervous system, endocrine, cardiac, ocular, auditory, gastrointestinal, renal, myoskeletal, dermal, and hematological abnormalities. Progressive late onset mitochondriopathies are thought to be associated with age-related diseases and are based on diminished mitochondrial function due to progressive DNA damage (i.e., heteroplasmy) and deterioration of energy production.⁵

In the auditory system, mitochondrial defects are associated with initial deterioration and loss of high-frequency sensitivity in presbycusis, noise-induced hearing loss, and many inherited forms of hearing impairment and loss.^{6, 7} The role of mitochondrial dysfunction in hearing loss (HL) is supported by syndromic and nonsyndromic deafness associated with mitochondrial mutations⁸ and the direct effects of ototoxic drugs such as aminoglycosides and cisplatin.^{8, 9} Furthermore, defects in mitochondria are associated with the effect of noise-, age- and ROS-related insults.^{10, 11} Noise, ototoxic, and ischemia-induced HL are also thought to be cumulative processes that are suggested to be, in part, mediated by ROS.^{12, 13} More recently, mice carrying a mutation either in the proofreading-deficient version of the mitochondrial DNA (mtDNA) polymerase $\alpha$ (POLA) or the $\gamma$ subunit (POLG) accumulate mtDNA mutations and display features of accelerated aging and HL (Refs. 14, 15). Mitochondrial changes, such as swelling and long-term degeneration, are observed in HL. Further alterations associated with mitochondrial metabolic defects are documented by immunohistochemical studies of mitochondrial enzymes. Despite this wealth of information, little to no direct evidence for metabolic changes in cochlear mitochondria is available.

Current imaging techniques are available to document the metabolic status of mitochondria in living cells. The reduced form of the nicotinamide adenine dinucleotide coenzyme (NADH) and the oxidized forms of flavoproteins (Fp) have been widely used as intrinsic fluorescent probes of mitochondrial respiration since development of this approach by Chance and Baltscheffsky,¹⁶ Chance, ¹⁷ and Chance and Shoener.¹⁸ NADH fluorescence has been widely used in monitoring the metabolic status of cornea, brain, muscle, and other tissue. ^{19, 20, 21, 22, 23, 24, 25} Fp fluorescence can be used as a complementary technique showing the relative oxidized state rather than the reduced state shown by NADH fluorescence.^{26, 27} The use of this paired technique has also been used to characterize cardiac myocytes.^{28, 29}

Two-photon excitation of these intrinsic fluorophores is advantageous because the amount of energy required to cause cellular damage increases exponentially with the wavelength of the light, resulting in reduced phototoxicity.^{30, 31} Further, the simultaneous absorption of two near-IR photons confines both the fluorescence and the subsequent photobleaching to the focal plane.³² NADH can be imaged with submicrometer resolution using multiphoton microscopy. ^{33, 34, 35, 36, 37, 38, 39, 40} Two-photon excitation in the range of $710\text{to}780\mathrm{nm}$ causes NADH to emit a broad blue fluorescence peaking near $460\mathrm{nm}$ . Flavoproteins can be similarly imaged using a broad range of excitation wavelengths (700 to greater than $900\mathrm{nm}$ ) with a fluorescence emission peak near $530\mathrm{nm}$ . Two-photon redox ratiometric fluorimetry of NADH and cellular flavoproteins has been performed in cardiac myocytes.³¹ More recently, two-photon-excited NADH imaging has been employed to study astrocytes and neurons in living brain slices.^{35, 37} In addition, two-photon NADH imaging with one-photon Fp imaging was used to study the pyruvate response in pancreatic islets.³⁸

In previous studies of cultured rat basophilic leukemia cells,³⁹ we showed that the NADH fluorescence is primarily localized to mitochondria, and that this fluorescence shows appropriate responses to mitochondrial inhibitors and uncouplers. However, we were unable to measure two-photon-excited Fp fluorescence in this cultured cell line. We also studied NADH photobleaching during multiphoton microscopy to optimize metabolic imaging techniques.⁴⁰ Here, we build on this work to develop functional two-photon imaging of excited NADH and Fp to study the metabolic state of neurosensory epithelium, hair cells, in the cochlea. We employed an isolated murine cochlea preparation similar to that used in many prior studies of cochlear physiology. This technique enabled us, for the first time, to do direct cell-by-cell metabolic characterization and determine general metabolic parameters for different cell types within the organ of Corti from the same images. Further development of this technique can be employed to study metabolic function in hearing disorders using a wide variety of animal models, including the spontaneous and genetically engineered mutant mouse lines.

2.

Materials and Methods

2.2.

Mouse Dissection

Mice (CF-1, Charles River Laboratories, Boston, Massachusetts) were euthanized with carbon dioxide. We immediately removed the cochlea from the skull and placed it into cold L-15 medium. Excess bone and tissue were trimmed away. Removal of the bone from the basal turn enabled the cochlea to sit level for mounting in a secure stage. A window was opened in the bone covering the apical turn of the organ of Corti, exposing the last third of the turn, as shown in Fig. 1 . The preparation was inserted into a silastic tubing collar, which further stabilized it for imaging at room temperature. Animal care and handling was in accordance with an Institutional Animal Care and Use Committee (IACUC) approved protocol.

Fig. 1

Cross-sectional diagram of the prepared cochlea with a microscope objective shown above the viewing area. The objective looks down on the organ of Corti, and images are taken of the outer hair cells (O), outer and inner pillar cells (OP, IP), and inner hair cells (I), through the tectorial membrane (TM), which is located above these cells, as shown in the inset.

2.5.

Linear Unmixing

The fluorescence emission of NADH is broad, completely overlapping the emission spectrum of FMN, as is evident in Fig. 2 . While it is relatively easy to choose appropriate bandpass filters to isolate NADH emission from FMN, it is impossible to isolate FMN emission from NADH. Therefore, to determine the amount of NADH fluorescence present in the $510\text{-}\text{to}570\text{-}\mathrm{nm}$ channel, and the amount of FMN fluorescence in the $420\text{-}\text{to}500\text{-}\mathrm{nm}$ channel, we measured the fluorescence intensity present in both channels when pure solutions of either NADH or FMN were two-photon excited at $740\mathrm{nm}$ . This measurement revealed that $85\pm 1\%$ of the NADH signal present in the $420\text{-}\text{to}500\text{-}\mathrm{nm}$ channel was also present in the $510\text{-}\text{to}570\text{-}\mathrm{nm}$ channel, while only $1.47\pm 0.01\%$ of the FMN signal in the $510\text{-}\text{to}570\text{-}\mathrm{nm}$ channel was present in the $420\text{-}\text{to}500\text{-}\mathrm{nm}$ channel. Given the relatively good agreement between the fluorescence emission spectra of organ of Corti cells and free NADH and FMN in solution (Fig. 2) and the broad detection bandpass, the separated NADH and Fp channels could be obtained from the mixed channels by a linear unmixing:

Eq. 1

$$\mathrm{NADH}=\gamma (C_1-0.0147C_2),$$

Eq. 2

$$\mathrm{Fp}=\gamma (C_2-0.853C_1),$$

where $C_1$ and $C_2$ are the actual pixel values detected in the blue $\left(420\text{-}\text{to}500\text{-}\mathrm{nm}\right)$ and green $\left(510\text{-}\text{to}570\text{-}\mathrm{nm}\right)$ channels, respectively, and $\gamma$ is an arbitrary scale factor that sets the display contrast. Pixel-by-pixel linear unmixing was performed for all of the images shown using ImageJ (Ref. 41).

Fig. 2

Spectrum of organ of Corti cells compared with solutions. The normalized fluorescence signal is shown as a function of wavelength. (a) The spectrum of freshly prepared cells (points) is overlaid with the spectrum of a $1\text{-}\mathrm{mM}$ NADH solution (line) obtained from the META detector. The points and error bars represent the mean and standard deviation of the fluorescence intensity from 32 individual cells. (b) The spectrum of isolated hair cells after digestion (points) is overlaid with a $10\text{-}\mu \mathrm{M}$ FMN spectrum (line) obtained on the same detector. The points and error bars represent the mean and standard deviation of the fluorescence intensity from 11 individual cells.

3.

Results

The spectrum from cell types in the intact preparations comprised of IHCs and OHCs, as well as their associated supporting cells, can be seen in Fig. 2a. The line indicating the spectrum of $1\text{-}\mathrm{mM}$ NADH solution shows the same characteristic shape as the spectrum from the cells. The preparation shows what may be a small Fp signal present at $550\mathrm{nm}$ , which is overwhelmed by the more intense NADH fluorescence signal. In isolated OHCs that have undergone digestion with papain, the spectrum obtained closely resembles a solution of $10\text{-}\mu \mathrm{M}$ FMN, as shown in Fig. 2b. There is considerable agreement between the two peaks of the spectra, though the FMN spectrum appears to be slightly broader toward the blue.

Optical sections of adjacent focal planes within the intact preparation are shown at $3.0\text{-}\mu \mathrm{m}$ intervals in Fig. 3 . The measured two-channel fluorescence images were spectrally unmixed to obtain separate NADH and Fp channels. Several rows of different cell types are apparent. The Hensen’s cells are in the uppermost row as indicated by the H in Fig. 3. They exhibit a mostly even distribution of NADH fluorescence with some bright punctate regions indicative of mitochondrial clustering. The next three rows of cells, indicted by O in Fig. 3, are the OHCs. These cells have distinct rings of fluorescence that contain both NADH and Fp. The relative amount of fluorescence varies along the length of the OHC with more Fp fluorescence apparent toward the basal surface of the cells where they meet with the supporting cells underneath. The next row of cells is the pillar cells, indicated by P in Fig. 3. This row of cells is actually comprised of inner and outer pillar cells that are not separately resolved as they are pressed closely together. As the images progress deeper into the tissue, only the outer pillar cells appear to be visible. The final row of cells is comprised of IHCs, labeled I in Fig. 3. These cells are distinct because of the bright punctuate regions containing fluorescence from both Fp and NADH, corresponding to mitochondria evenly distributed throughout the cytoplasm. The nucleus appears as the dark void in IHCs. The remaining fluorescence comes from supporting cells.

Fig. 3

Optical sections of the organ of Corti obtained $15\min$ after the death of the animal. The distance from the top of the OHCs is indicated on each image. A row of Hensen’s cells (H), three rows of outer hair cells (O), a row of pillar cells (P), and a row of inner hair cells (I) are all visible at different points in the series. In the online version, the spectrally unmixed NADH fluorescence is represented in green, the unmixed Fp fluorescence in red, while areas of yellow indicate overlap of the two fluorescence signals. The scale bar is $10\mu \mathrm{m}$ . (Color online only.)

To measure the metabolic response of the cells of the organ of Corti to a mitochondrial inhibitor (NaCN) and uncoupler (FCCP), we measured fluorescence from a single focal plane approximately $7\mu \mathrm{m}$ below the tips of the OHCs (Fig. 4 ). When treated with $10-\mu \mathrm{M}$ NaCN, OHCs and IHCs exhibited an increase in NADH fluorescence, as shown in Fig. 4d, when compared to the preparation before treatment in Fig. 4a. In this preparation, OHC NADH fluorescence increased to 184% of the untreated cellular NADH fluorescence, while the fluorescence of IHCs increased to only 112%. This was accompanied by a loss of Fp fluorescence that can be attributed to the decreased concentration of oxidized Fp resulting from the inhibition of oxidative phosphorylation. For OHCs, the Fp fluorescence decreased to 56% of the untreated cellular Fp fluorescence, while IHCs decreased to 89%. As expected, the trends were reversed after the preparation was treated with a $10\text{-}\mu \mathrm{M}$ FCCP solution. When treated with FCCP, the NADH fluorescence dropped to 83 and 75% of the untreated OHC and IHC NADH fluorescence, respectively [Fig. 4g], and the Fp fluorescence increased to 112 and 135%, respectively [Fig. 4h].

Fig. 4

Metabolic imaging of the organ of Corti at a depth of $7\mu \mathrm{m}$ below the top of the OHCs. The preparation (a) to (c) before the addition of any metabolic treatments to the imaging buffer, (d) to (f) after $5\min$ of incubation with $10\text{-}\mu \mathrm{M}$ NaCN, and (g) to (i) after $5\min$ of incubation with $10\text{-}\mu \mathrm{M}$ FCCP. Spectrally unmixed NADH fluorescence is depicted in (a), (d), and (g); unmixed Fp fluorescence in (b), (e), and (h); and a merge of NADH (pseudo-green) and Fp (pseudo-red) in (c), (f), and (i). Approximate outlines of the IHCs are drawn in red. The images were taken at $20\min$ (untreated), $25\min$ (NaCN), and $40\min$ (FCCP). Scale bar is $10\mu \mathrm{m}$ . (Color online only.)

The average level of reduced NADH and oxidized Fp was then calculated, using Eqs. 3, 4, from the changes in fluorescence intensity for IHCs and OHCs. This was done for nine different preparations that were imaged at various times following the extraction of the cochlea. The results, plotted in Figs. 5a and 5b , show an apparent relationship between the amount of time that the cochlea was out of the animal and the relative redox state of the hair cells. Fp oxidation increased steadily from approximately 19% at $15\min$ after the death of the animal, to 100% after $50\min$ . The IHC NADH oxidation state followed a similar pattern, with the level of reduction decreasing from 95 to 0% within $45\min$ of death.

Fig. 5

Relative metabolic state of a total of nine preparations was analyzed at seven different times. The percentage of reduced NADH (left axis) and oxidized Fp (right axis) for IHCs is shown as a function of the time between death of the animal and image acquisition for (a) IHCs and (b) OHCs. (c) NADH and Fp fluorescence normalized to the intensities of the first image and plotted as a function of time for a single preparation. Error bars represent either the standard deviation of replicated experiments or the uncertainty estimated by error propagation through Eqs. 3, 4.

For OHCs, the results varied more dramatically, ranging from $89\pm 10\%$ reduced and $21\pm 19\%$ oxidized for a preparation imaged approximately $15\min$ after the death of the animal to $0\pm 10\%$ reduced and $100\pm 13\%$ oxidized after $50\min$ . The relationship does not appear to be linear, but is more likely to be an exponential decay. Analysis of the OHCs of a single preparation [Fig. 5c] showed that the ratio of reduced NADH fluorescence to oxidized Fp fluorescence decreased in a similar manner. OHCs also began to swell as imaging time progressed with the longer time periods corresponding to a larger proportion of swollen cells.

4.

Discussion

While two-photon NADH fluorescence has been imaged in brain, heart, cornea, and pancreatic tissues,^{33, 34, 35, 36, 37, 38} this technique had yet to be used to study metabolism in the cochlea. The spectrum of the organ of Corti cells indicates that both NADH and Fp are present in the cells in sufficient quantities for imaging. This enabled us to study the metabolism of the organ of Corti without the use of extrinsic fluorescent probes. We were also able to study the intact tissue in toto, with all of the different cell types, while at the same time assessing the metabolic capacity for each individual cell. The distinct histological and morphological characteristics of the organ of Corti provided a unique opportunity for metabolic imaging of whole tissue at the cellular level. There is considerable agreement between the spectral peaks of the organ of Corti cells and the NADH and FMN solutions, as shown in Fig. 2. The fluorescence spectra also show a considerable amount of overlap in the wavelengths of the NADH and Fp signals. To resolve this issue we used linear unmixing to separate the signals. This enabled us to make the first direct quantification of Fp oxidation by two-photon microscopy, rather than inferring the Fp oxidation state from the mix of NADH and Fp in the green channel, which has been used previously.

It appears that the NADH and Fp signals are present in the expected subcellular regions within the cells corresponding to the typical locations of mitochondria. This is demonstrated in Fig. 3. As described before, the distribution of the fluorescence signals in the OHC is localized in a ring nearly contiguous with the lateral wall of these cells, comprising the cellular membrane, the cortical lattice, and the subsurface cisterna. The mitochondria are located along the subsurface cisterna along the lateral wall.⁴³ This would produce the circular fluorescence distribution seen in the OHCs of Fig. 3. This distribution differs from the mitochondrial arrangement of IHCs. These mitochondria are dispersed throughout the cytosol, which corresponds to the punctate distribution of fluorescence as shown in Fig. 3. The pillar and Hensen’s cells show a relatively even distribution of mitochondria throughout the cell, in agreement with the distribution of fluorescence in the pillar cells shown in Fig. 3. Kasischke ³⁵ showed that they were able to identify astrocytes and neurons in brain tissue based on the NADH fluorescence pattern in much the same way. This enabled us to simultaneously identify the cells of interest (IHCs or OHCs for example) as we performed metabolic imaging.

Further confirmation of the mitochondrial origin of the NADH and Fp fluorescence came when we examined the data from the metabolic treatments of cyanide and the uncoupler FCCP. After NaCN incubation, the cells should be in their most reduced state, just as if they were oxygen deprived since NaCN blocks the transfer of reducing equivalents to oxygen. NADH fluoresces when reduced, so the fluorescence should increase with the addition of cyanide. As shown in Fig. 4d, the NADH fluorescence increased significantly after cyanide incubation, but decreased with treatment of the mitochondrial uncoupler FCCP [Fig. 4h]. This is expected since the uncoupler dissipates the mitochondrial $p\mathrm{H}$ gradient, allowing for the unregulated flow of electrons through the electron transport chain without production of ATP, rapidly converting NADH to the nonfluorescent ${\mathrm{NAD}}^{+}$ . The Fp fluorescence varies inversely with the NADH fluorescence, as seen in the images in Figs. 4b, 4e, 4h, as Fp fluoresces when in its oxidized state.

For all of the preparations, the sum of the relative oxidation and reduction percentages totals 100% within the uncertainty of the measurement. This is expected when the fluorescence originates from mitochondrial sources where flavin and nicotinamide cofactors are in direct equilibrium. Previous two-photon and traditional confocal metabolic imaging studies have tended to use the absolute change in intrinsic fluorescence or the ratio of the NADH and Fp fluorescence to study changes in metabolism ^{17, 19, 20, 25, 26, 27, 33, 34, 35, 36} rather than quantifying the relative redox state of these molecules, as was done in this study.

The average reduction of the IHCs is $73\pm 12\%$ for preparations imaged less than $40\min$ after the death of the animal. This is higher than we observed in cultured rat basophilic leukemia cells, which have typically been 45 and 55% reduced. This high percentage of reduced NADH and Fp indicates a large capacity for ATP production. The fact that the oxidized state predominated when images were obtained more than $40\min$ after the death of the animal would indicate a decreased capacity for ATP production via oxidative phosphorylation. From this study, it is unclear whether these trends result from lowered rates of regeneration of the reduced forms of NADH and Fp through glycolysis, the citric acid cycle, and beta oxidation, or from a strong upregulation of oxidative phosphorylation to produce ATP. The effect is more dramatically seen in the OHCs, as shown in Figs. 5b and 5c as the amount of reduced energy equivalents is decreasing while the amount of oxidized energy equivalents increases. It is not clear why the OHCs exhibit the loss of reduced NADH and Fp more rapidly than the IHCs.

Possible experiments that could illuminate the mechanism behind this loss of reduced NADH and Fp include varying the concentration of glucose in the media to determine if the cells are simply metabolizing glucose faster than they are able to obtain it from passive diffusion from the media. Also, potentiometric and $p\mathrm{H}$ -sensitive fluorescent probes may enable the assessment of ATP production resulting from oxidative phosphorylation. The temperature of the preparation may also affect the metabolism of the cells and should also be investigated. This could enable determination of the ideal working temperature of the preparation, be it physiological, room temperature, or chilled.

In developing any fluorescence imaging protocol to assess normal function of living cells it is important to minimize the adverse affects of photobleaching and photodamage.^{39, 40} We attempted to do this by taking relatively few images at a low sampling rate and at the lowest possible laser intensity required to get a sufficient SNR. Under these imaging conditions, we did not observe any significant NADH or Fp photobleaching. Nevertheless, we have found that photodamage during metabolic imaging can occur even when NADH photobleaching is not evident,³⁹ and a systematic investigation is warranted.

The Fp fluorescence images have a much lower SNR than those formed from NADH fluorescence. This is because the concentration of Fp is several orders of magnitude lower than NADH in tissue.⁴⁴ Our own experience with cultured cells has shown that the ability to detect Fp fluorescence also varies dramatically with tissue type, and, in fact, we were unable to detect Fp fluorescence in cultured rat basophilic leukemia cells.³⁹ Fortunately, the two-photon action cross section of Fp is more than three times that of NADH at $740\mathrm{nm}$ (Ref. 34) enabling more efficient generation of Fp fluorescence. However, the overlap in the NADH fluorescence and the Fp fluorescence also complicates the issue, making it difficult to detect the Fp fluorescence because of NADH fluorescence overlap. This was also described by Huang, who suggested that greater specificity for NADH and Fp could be achieved by narrowing the spectral ranges of the detection channels.³⁴ Unfortunately, decreasing the detection efficiency as Huang suggests would further reduce the SNR of the Fp channel. Instead, we found that careful calibration of the crosstalk between the two detection channels enables the separation of the signal into its NADH and Fp components. Given the measured fluorescence emission spectrum from organ of Corti cells (Fig. 2), we expect that further improvement in measuring the weak Fp signal may be realized by extending the detection band to $650\mathrm{nm}$ .

Further experiments will determine if the amount of reduced NADH and Fp vary between species or even strains of the same animal, which would be expected from studies of dehydrogenase activity in cochleas of differing mouse strains.⁴⁵ If it can be established that there is some consistency between mammalian species, or even strains of the same species, this technique will have a very broad range of application. Studies of noise-induced hearing loss, inherited hearing loss, and presbycusis could use this technique to determine if metabolic differences exist in these hearing disorders. Combining this technique with studies of gene expression, protein function, and structural studies could help illuminate the underlying causes of some of these disorders.