2.1.

Animal Models

All in vivo experiments and experimental protocols were approved by the research ethics board of the Privolzhsky Research Medical University, Nizhny Novgorod, Russia. The experiments were performed on male Wistar rats with a body weight of 300 to 400 g. We modeled both acute and chronic liver pathology: cholestasis and fibrosis.

Acute cholestasis was induced by bile duct ligation. This experimental model is well accepted and used worldwide in hundreds of laboratories to induce liver cholestasis. It results in intrahepatic biliary epithelial cell proliferation and myofibroblastic differentiation of the portal fibroblasts around the proliferating biliary epithelial cells.²⁴ Bile duct ligation was performed after midline laparotomy. The common bile duct was ligated two times with 5–0 silk. The surgical procedures were performed under aseptic conditions. Body temperature was controlled by placing the animals on a heating pad set to 37°C. Imaging was performed 1 and 3 weeks after bile duct ligation. Healthy rat livers served as controls. Each group consisted of 5 rats.

Liver fibrosis (models using chronic-plus-multiple binges of ethanol) was induced by intragastric infusion of a solution containing ethanol as described in Ref. 25. Imaging was performed 4, 8, and 12 weeks after fibrosis induction. Healthy rat livers served as controls. Each group consisted of 5 rats.

2.2.

Multiphoton Fluorescence Microscopy with FLIM and SHG

The two-photon excited fluorescence intensity (TPEF), the SHG of collagen fibers, and FLIM images of NAD(P)H and FAD were obtained using an LSM 880 (Carl Zeiss, Germany) laser scanning confocal microscope equipped with a time-correlated single-photon counting system (Simple-Tau 152, Becker & Hickl GmbH, Germany). NAD(P)H and FAD fluorescence were excited with a Ti:Sa femtosecond laser, using an 80-MHz repetition rate and a pulse duration of 140 fs at the wavelengths of 750 and 900 nm, respectively. Emission was detected in the ranges of 450 to 500 nm for NAD(P)H and 500 to 550 nm for FAD. An average of 10,000 photons were collected per decay curve. The average power of the Ti:Sa laser was measured using a PM100A power meter (ThorLabs Inc., Newton, New Jersey). The SHG signal and hepatocyte autofluorescence (AF) were generated using excitation at a wavelength of 800 nm. Backward-directed SHG signals were detected in the range of 371 to 421 nm. Hepatocyte AF was detected in the range of 433 to 660 nm. To account for the fluctuations of the laser power and correct for the scattering effects, we have made reference measurements of the SHG signal generated on the glass–air interface and for each image made a background correction.²⁶ The average power incident on the samples was $\sim 6\mathrm{mW}$. A C-Apochromat $40\times /1.2$ water immersion objective was used for image acquisition. Midline laparotomy was performed to expose the liver. The images of unfixed liver tissues were collected within 15 min of the start of surgical procedures. Ten images were collected for each liver from nonoverlapping fields.

2.3.

Fluorescence Lifetime Data Analysis

FLIM imaging based on the endogenous fluorescent cofactors is an established approach used to analyze cellular metabolism. The nonphosphorylated form of NADH acts as an electron donor in the mitochondrial electron transport chain. This form of the cofactor is generated during glycolysis and the tricarboxylic acid cycle via the reduction of NAD+. The fluorescence lifetime of NADH depends significantly on the state of the cofactor (whether “free” or “protein-bound”).²⁷ FAD linked with protein can exist in two conformations: (1) “closed” or stacked, in which the coplanar isoalloxazine and adenine rings interact through $\pi \text{–}\pi$ interactions, resulting in very efficient fluorescence quenching, and (2) “open” or unstacked, in which the two aromatic ring are separated from each other.²⁸ FAD-containing proteins participate in a variety of metabolic pathways, including electron transport, DNA repair, nucleotide biosynthesis, the beta-oxidation of fatty acids, and amino acid catabolism, as well as the synthesis of other cofactors such as coenzyme A, coenzyme Q, and heme groups.^7,29

The FLIM data were analyzed using both bi- and tri-exponential fitting models from a $212\times 212{\mu \mathrm{m}}^2$ imaging area. The fluorescence lifetimes and the relative contributions (free/protein-bound forms of NADH, and bound NADPH) and open/closed conformations of FAD) were calculated for the region of interest by finding the global minimum of ${\chi }^2$ value. The mean values of ${\chi }^2$ and the fluorescence lifetimes in healthy and diseased liver were assessed in the hepatocyte cytoplasm (in which fluorescence arises primarily from mitochondria) by selecting $\sim 40\times 40\text{pixel}$ zones as regions of interest. Nuclei were excluded from these measurements. The FLIM data were processed in SPCImage software (Becker & Hickl GmbH).

2.4.

TOF-SIMS

Mass spectrometry measurements were performed on a TOF-SIMS 5 instrument (ION-TOF GmbH, Germany) equipped with 30-keV ${\mathrm{Bi}}^{3+}$ primary ions. All the spectra were reordered in spectroscopy mode from a $300\times 300{\mu \mathrm{m}}^2$ area. The primary ion dose density was below the static SIMS limit ($\sim 3.5\times 1011\text{ions}/{\mathrm{cm}}^2$). At least two different sections of the same pathology and control samples were analyzed. Eighteen mass spectra were recorded for each section. A low-energy electron flood gun was activated to avoid charging of the sample. Secondary ion yields were calculated as the intensity of the corresponding peak of interest normalized to the total ion count using SurfaceLab software (ION-TOF GmbH, Germany). The average ion yields of lipids and amino acids were calculated as mean values acquired from randomly selected areas of different tissue sections both for the pathology and control samples. The saturated fatty acid ion yield was calculated as the sum of the C14:0, C16:0, and C18:0 yields. The polyunsaturated fatty acid yield was calculated as the sum of the C16:2, C18:3, C18:2, C20:3, and C20:4 yields. The unsaturated fatty acid yield was calculated as the polyunsaturated acid yield plus the C16:1 and C18:1 yields.

2.5.

Morphological Analysis

Liver specimens were fixed in 10% buffered formalin and embedded in paraffin. The $8\text{-}\mu \mathrm{m}$ sections were cut using a microtome (Leica SM 2000; Germany) and mounted on glass slides. Ten cross-sections from each liver were stained with Van Gieson picrofuchsin (VG) to analyze the general histoarchitecture of the tissues, the collagen structure, and histopathological changes. All sections were examined using light microscopy, on a Leica DM1000 system.

2.6.

Statistical Analysis

A statistical analysis was performed with the help of STATISTICA 64 software, version 10 (StatSoft Inc., Tulsa, Oklahoma). Mean and standard deviation values were used to express the data, whereas the student’s $t$-test and the one-way ANOVA Bonferroni post-hoc test were used to compare the data. Results were considered statistically significant with a $p$ value $\le 0.05$.

3.1.

TPEF and SHG Microscopy of Normal and Diseased Livers

The cellular structure of the liver could be imaged to a depth of $250\mu \mathrm{m}$ below the fibrous capsule using multiphoton microscopy. Wang et al.³⁰ had previously found that an imaging depth of $50\mu \mathrm{m}$ provided the clearest observations of cellular and subcellular morphology. In normal rat liver, hepatocyte cords showing bright fluorescence of NAD(P)H (primarily from mitochondria) could be distinguished from the dark, blood-filled sinusoids in the fluorescence image. TPEF of FAD was negligible in comparison with NAD(P)H. The signal level in the control was $3\pm 0.3\mathrm{a}.\mathrm{u}$ and $24\pm 0.8\mathrm{a}.\mathrm{u}$ for FAD and NAD(P)H, respectively. The stellate cells were found to be smaller in size and had a strong vitamin A-associated fluorescence. Stellate cells are known to store 80% of vitamin A in the whole body, and vitamin A absorbs light with wavelength of $765\pm 65\mathrm{nm}$ and could emit strong fluorescence at below 500 nm.³¹ In diseased liver, the hepatocytes lost their normal shape and clear borders. Injury to the hepatocytes could also be observed as reduced fluorescence in the fluorescence images for both cholestasis and fibrosis. It was previously shown that different pathologies (toxic fibrosis, hepatocellular carcinoma, and ischemia) may lead to significantly reduced fluorescence of hepatocytes.³⁰ Thus, we have identified two areas for analysis in images of diseased liver: hepatocytes with a high and low NAD(P)H signal (areas 1 and 2, respectively) (Fig. 1).

Fig. 1

Specific changes in acute cholestasis after 1 week and with chronic fibrosis after 12 weeks. Fluorescence intensity images with morphological data. NAD(P)H: excitation 750 nm, detection 455 to 500 nm; FAD/bilirubin: excitation 900 nm, detection 500 to 550 nm; AF (red): excitation 800 nm, detection 433 to 660 nm; SHG (green): excitation 800 nm, detection 371 to 421 nm. VG: Van Gieson picrofuchsin. Bar: $50\mu \mathrm{m}$.

In the case of acute cholestasis, hepatocyte damage was associated with the accumulation of bile and its pigments, in particular, bilirubin [Table 1(a)]. The highest signal upon excitation at 900 nm, associated with the strong AF of bilirubin and a weak FAD fluorescence, was detected in rat liver on seventh day after the induction of acute cholestasis. The AF signal of NAD(P)H was reduced compared with that of the control liver. At the 3-week time point, the NAD(P)H signal for both areas 1 and 2 had reached a minimum critical value while the FAD/bilirubin signals had returned to their normal values. The peak accumulation of bile in the liver with acute mechanical cholestasis occurred 7 days after the induction of pathology. After 3 weeks, bilirubin could not be detected in the liver. However, toxic damage to the hepatocytes continued to increase to week 3 and had resulted in a significant reduction in the NAD(P)H signal for areas 1 and 2. These typical changes for cholestasis were confirmed by morphological analysis: bile duct expansion and cholangiocyte proliferation. For details about the morphological changes, refer to Supplemental Figs. S1(A)–S1(C). The experimental animals had jaundiced skin and typical urinary symptoms of jaundice.

Table 1

TPEF intensity for normal and diseased livers.

For chronic pathology, fibrosis, we observed a different tendency. Normal rat liver of the control group had no SHG signal, indicating an absence of fibrous tissues. Strong and widely spreading SHG signals indicated fibrillar collagen deposition in the rat liver with induced fibrosis. Collagen deposition was observed mainly in the centrilobular region. Hepatocytes with a high NAD(P)H signal but with no collagen fibers around them were identified as area 1. Hepatocytes with a low NAD(P)H signal were detected in the areas of formation of fibrous structures and were marked as area 2 [Table 1(b)]. In contrast with acute cholestasis, the general level of TPEF intensity in the NAD(P)H channel for area 1 increased significantly during the development of the pathology. In area 2, the NAD(P)H signal halved after 1 month and remained at that level for up to 3 months (Fig. 2). The FAD signal intensity did not change dramatically at any time point compared with the control. The changes detected by multiphoton microscopy correlated well with the morphological data.

Fig. 2

TPEF intensity of NAD(P)H for normal and diseased livers.

3.2.

FLIM Microscopy of Normal and Diseased Livers

NAD(P)H fluorescence contains the contributions from free NADH, bound NADH and NADPH which can be involved in the oxidative stress often associated with cellular damage in liver diseases.³² To estimate the contribution of NADPH, we have performed an analysis of FLIM data using three-exponential fit of the fluorescence decay curves. Such an approach has been earlier implemented in the work of Meleshina et al.⁷ (Fig. 3). In our case, it was justified by the fact that bi-exponential fit resulted in the increased values of fluorescence lifetimes not only for free NADH (around 530 ps) but also for the significantly increased values of bound NAD(P)H (around 3100 ps) compared with the known ones. Thus it was natural to assume that phosphorylated form makes a contribution to the fluorescence signal. It has been reported earlier that the intracellular fluorescence lifetimes of protein-bound NADH and NADPH are $1500\pm 200$ and $4400\pm 200\mathrm{ps}$, respectively.³³ Since the fluorescence lifetime of NAD(P)H when bound to an enzyme is determined by its local environment in the binding site, the NADH and NADPH-binding sites are two of the most highly conserved in all biology.^34,35 We have fixed these lifetimes and determined the relative contributions of nonphosphorylated ($a2$) and phosphorylated ($a3$) forms of NADH. The FLIM images of FAD were analyzed with the bi-exponential fitting model.

Fig. 3

FLIM images of NAD(P)H in liver tissue in acute cholestasis after 1 week and chronic fibrosis after 12 weeks. Amplitude weighted mean lifetime ($tm$) and the relative contributions of protein-free and protein-bound NAD(P)H for the tri-exponential fitting model. Bar $50\mu \mathrm{m}$.

In normal rat liver, the amplitude weighted mean lifetime ($tm$) of NAD(P)H for the tri-exponential fitting model was determined to be 954 ps. The parameter $tm$ of FAD for the bi-exponential fitting model was 267 ps. The relative contributions of protein-free NADH ($a1$), protein-bound NADH ($a2$), and protein-bound NADPH ($a3$) were 69%, 22%, and 10%, respectively. The parameters $a1$ (closed form) and $a2$ (open form) for FAD were 80% and 20%, respectively. Because of the low contribution of bound NADPH, we present the images of $a2/a3$ ratio to express change between normal and pathological livers.

For acute cholestasis [Table 2(a)], we observed that $a1$, the contribution of protein-free NADH, had increased in the damage area with a high bilirubin signal after 1 week. The contributions of protein-bound NAD(P)H, $a2$ and $a3$, were decreased compared with the control in area 2. The hepatocytes of area 1 showed no changes after 1 week as compared with the control. By week 3, $a1,a2$, and $a3$ in area 2 have been restored to normal values close to those of the control. There were still no differences between the contributions of the protein-free and protein-bound NAD(P)H of area 1 after 3 weeks and the control group. In the case of cholestasis, we did not analyze the changes in lifetimes and their contributions for FAD, because the great contribution of bilirubin fluorescence signal excited in the 900 nm (540- to 550-nm emission), complicating the analysis of open/closed FAD contributions.

Table 2

Fluorescence lifetimes and the relative contributions of different forms of NAD(P)H and FAD [a1,a2, and a3 for the NAD(P)H tri-exponential fitting model, and a1,a2 for the FAD bi-exponential fitting model] in normal and diseased livers.

Changes in the fluorescence lifetimes and their contributions associated with chronic pathology had other features [Table 2(b)]. After 4 weeks, the $a1$ of protein-free NADH for both areas 1 and 2 had become lower than in the control. These changes persisted for 12 weeks. The contribution ($a1$) of closed FAD has significantly increased by week 4 but slowly returned to its normal value by 12 weeks [Table 2(c)]. Changes in $a2$ (the contribution of open FAD) showed the inverse relationship. During the development of fibrosis, we did not observe any significant differences in the FAD contributions between areas 1 and 2 for all time points.

3.3.

TOF-SIMS Analysis of Normal and Diseased Livers

TOF-SIMS was used to reveal chemical changes induced by acute cholestasis (after 3 weeks) and chronic fibrosis (after 12 weeks) compared with the control sample. Mainly TOF-SIMS studies are focused on lipid analysis since the technique is sensitive to relatively small molecules (up to 1 to 2000 Da). Nevertheless some information about protein compound can be obtained, based on analysis of the ions attributed to amino acids.^36,37 Figure 4(a) presents a comparison of both the lipid and amino acid signals. Ion yields were normalized to the control sample signal for better presentation. The $m/z$ 30 signal, which is mainly attributed to glycine, was significantly enhanced in the fibrosis sample. The proline signal ($m/z$ 68 and 70) is also slightly increased for both the cholestasis and fibrosis sections. The arginine yield ($m/z$ 73) is raised in fibrosis but decreased in the cholestasis sample. It should be noted that glycine, proline, and arginine are known to form collagen and that can explain their increased level for the fibrosis sample.³⁸ The phosphatidylcholine and cholesterol levels differed for the control and fibrosis samples. The cholesterol signal level was decreased by 30% compared with the control liver. Figure 4(b) shows the fatty acid compound changes caused by the pathologies. The fatty acid signal is mainly derived from acylglycerols and phospholipids. It is clearly seen that the unsaturated and polyunsaturated fatty acid signals decrease both for fibrosis and cholestasis. While the intensity of decline for the fibrosis samples is within experimental error, the cholestasis samples show a noticeable difference. Hence fatty acid chain saturation occurs during cholestasis progression.

Fig. 4

TOF-SIMS analysis of liver tissues for acute cholestasis after 3 weeks and for chronic fibrosis after 12 weeks. Ion yields were normalized to the control sample. (a) Glycine ($m/z$ 30), proline ($m/z$ 68 and 70), arginine ($m/z$ 73), phosphatidylcholine ($m/z$ 184), and cholesterol ($m/z$ 369). (b) Ratio of unsaturated/saturated fatty acids.