2.1.

Animals and Diet

The study was approved by the local Animal Ethics Committee at University of Gothenburg, Gothenburg, Sweden.

2.2.

Nonlinear Microscopy

The experiments are summarized in Fig. 1, which also shows the experimental setup used for nonlinear microscopy. The laser system consists of a Nd:Vanadate pump laser [HighQ (Hohenems, Austria) picoTRAIN], and two ring-cavity optical parametrical oscillators (OPO), APE (Berlin, Germany) Levante. The pump laser emits pulses of 7-ps duration and wavelength of 1064 nm at a repetition rate of 76 MHz. The total average output power of the pump laser is 10 W split into three beams, one directly aligned into the microscope and one beam directed to each OPO. The OPOs provide tunable radiation covering the wavelength intervals 785 to 845 nm and 855 to 920 nm, with average output powers in the range 0.3 to 1.0 W. The 1064-nm and OPO beams are aligned into a common optical path using dichroic mirrors, and temporal overlap is achieved by means of individual delay lines for the OPO beams. The copropagating beams are directed into the microscope (Nikon TE-2000E) beam scanning unit and focused on the sample by an objective (40× Nikon Plan Fluor NA 1.3 or 60× Nikon TIRF NA 1.49). Typical average powers at the sample are 50 mW for the OPO beams and 30 mW for the 1064 beam. CARS and SHG are generated in the high-intensity region of the diffraction-limited focal volume of the laser beams, and the signals propagating in the forward direction are collected by an aspherical lens (NA 0.8) separated by a beamsplitter and directed to one photomultiplier tube each (Hamamatsu PMC100–1/PMC100–20). Bandpass filters are used to isolate the signals and suppress background light, in particular residual laser radiation.

Fig. 1

Summary of the experiments carried out on skeletal muscle and liver tissues of mice. In experiment 1 with 12 mice, intake of standard chow and a Western-type high-fat diet was compared during 12 weeks by SHG and CARS microscopy. Experiment 2 with 28 mice, from which 13 liver and 10 diaphragm samples were characterized by CARS microscopy, compared the differences between two high-fat diets with different levels of polyunsaturated lipids (see Sec. 2 for further details). The experimental setup for nonlinear microscopy, shown to the right, consists of a pump laser, two OPOs, and an inverted microscope with laser beam scanner. The generated signals are collected in the forward direction, separated by a dichroic mirror, and detected by photomultiplier tubes (PMTs).

Combined SHG and CARS microscopy (Experiment 1) or dual CARS microscopy²² (Experiment 2) measurements were carried out on samples of liver and skeletal muscle tissue of mice. For combined SHG and CARS measurements, a single OPO tuned to 817.0 nm was used that—combined with the 1064-nm beam—excites the CH₂ symmetric stretch vibration at wave number 2845 cm^–1, and the generated CARS signal is detected at 663 nm. The 817-nm beam also induces SHG detected at 409 nm. To obtain quantitative information on lipid storage from CARS microscopy measurements carried out in Experiment 2, a dual CARS arrangement was used to simultaneously measure the nonresonant background and be able to subtract it from the CH₂ CARS signal. For this purpose, the short-wavelength OPO remained tuned to 817 nm (2845 cm^–1), whereas the long-wavelength OPO was tuned to 863.6 nm. Combined with the 1064-nm beam, the latter generates a CARS process probing wave number 2185 cm^–1 in the so-called silent Raman region, where biological tissue lacks vibrational resonances and the nonresonant background can be measured.

Fresh tissue specimens were manually cut in slices a few hundred microns thick, and carefully mounted between microscopy cover slides. Tissue specimens of soleus muscle, diaphragm, and liver were collected from wild-type mice fed either standard chow or Western diet and from Ldlr-deficient mice fed HFSD herring or beef. The soleus muscle is of interest due to its relatively high amount of oxidative muscle fibers. The flat geometry of the diaphragm simplifies the sample preparation and makes microscopy of it relatively straightforward.

Images were collected either as single planes or as z-stacks covering 30 to 60 image planes and forming 3-D images. The scanned image area ranged from 20×20 μm² to 200×200 μm², with larger fields of view (140×140 μm² and 200×200 μm²) used for the analysis of liver samples, and smaller fields of view (40×40 μm²) used for the analysis of muscle tissue. The numbers of measurements on soleus, diaphragm, and liver specimens are summarized in Table 1.

Table 1

Numbers of samples and SHG/CARS microscopy measurements for the two studies carried out on skeletal muscle and liver tissues of mice.

CARS spectra of liver tissues from a HFSD-herring and HFSD-beef mouse were obtained by collecting CARS microscopy images with the OPO tuned to wavelengths between 807 and 824 nm, probing vibrations between wave numbers 2740 and 3000 cm^–1. At each wavelength an image of lipid stores in the liver tissue and a reference image in the cover slide glass, to normalize for varying excitation efficiency, were registered. From the series of CARS microscopy images, the normalized lipid droplet CARS signal was averaged and plotted versus probed wave number to obtain a CARS spectrum.

2.3.

Image Analysis

Nonlinear microscopy images were segmented into binary images for further analysis. For lipid droplet evaluation, the binary data obtained from the CARS microscopy images were analyzed using two tools of the software ImageJ,²³ the built-in particle analyzer, and a 3-D object counter plug-in routine. Particle sizes evaluated as areas and volumes were recalculated into equivalent diameters. Relative areas of tissue lipid content and liver fibrosis, visualized in the CARS and SHG images, respectively, were calculated as the number of thresholded pixels divided by the total number of pixels. For the dual CARS measurements carried out in Experiment 2, the measured nonresonant background was subtracted, resulting in a compensated CARS signal essentially consisting of the CH₂ vibrational resonant contribution.

2.4.

Tissue Biochemical Analysis

Frozen liver biopsies (weighing 100 to 200 mg) were homogenized in ten volumes of PBS (pH 7.4). The homogenates were weighed before and after removal of approximately 100-μl aliquot for triglyceride analysis, and the remainder of the homogenate was freeze-dried. The remaining homogenate was freeze-thawed for three cycles to disrupt the cells, and total triglyceride content was determined enzymatically following manufacturer protocol (T2449, F6428; Sigma-Aldrich, Stockholm, Sweden). Results were back-calculated and expressed as milligram triglycerides per liver dry weight.

3.1.

Skeletal Muscle Tissue; Structure and Intramyocellular Lipid Content

Figure 2 shows pairs of simultaneously measured SHG (left column) and CARS (middle column) microscopy images in soleus muscle for mice fed standard rodent chow [Fig. 2] and Western diet [Figs. 2 and 2]. The ordered arrangement of myosin molecules in the muscle filaments induces birefringence and makes them SHG-active. Thus, the cross-striation pattern of muscle tissue A-bands, also recognizable from polarization microscopy of sectioned tissue,²⁴ can be visualized by SHG microscopy. Intramyocellular lipid droplets, imaged by CARS microscopy and typically 1 μm in diameter, can be seen positioned longitudinally along the muscle fibers and located in-between the myofibrils. This is clearly shown in the overlay CARS/SHG images (right column), and in particular in the images of Fig. 2 covering a smaller field of view. The observed lipid droplet localization is in agreement with observations reported for human skeletal muscle tissue,²⁵ and recent studies have shown that the intramyocellular lipid droplets, forming a cellular energy source, are located between the intermyofibrillar mitochondria and to a larger extent in the oxidative, mitochondria-rich type-1 fibers than type 2.^{26, 27} Approximately 40% of the muscle fibers in soleus from male C57BL/6J mice are of the slow-twitch type 1.²⁸ Combining two-photon fluorescence of labeled mitochondria²⁶ with CARS and SHG microscopy would permit colocalization of all structures of interest, and this is a topic to be investigated further.

Fig. 2

Muscle tissue structure and intramyocellular lipid droplets imaged in soleus muscle by simultaneous SHG and CARS microscopy. Intramyocellular lipid droplets, monitored by CARS microscopy, were detected both in samples of mice fed (a) standard rodent chow and (b) Western diet for 12 weeks. The lipid droplets are aligned along the muscle fibers and located between the myofibrils, clearly seen in images (c) (measured after 4 weeks of Western diet) covering a smaller field of view.

CARS microscopy on skeletal muscles (soleus and diaphragm) from wild-type mice fed standard or Western diet generally showed low content of lipids, irrespective of diet. After four weeks, intramyocellular lipid droplets were detected in one third of the images measured in the soleus sample of one mouse fed Western diet, and a sample image is shown in Fig. 2. Measurements on soleus samples from the three other mice investigated at this time point, one fed Western diet and two fed standard chow, showed no intramyocellular lipid content. Diaphragm samples investigated after four weeks of diet intake showed no lipid content and neither did additional measurements carried out on samples of extensor digitorum longus or gastrocnemius (data not shown). After eight weeks, approximately one third of the CARS microscopy images measured in one of the two investigated Western diet soleus samples showed lipids. The other soleus data, measured in the other Western diet sample and two standard diet samples, showed no intramyocellular lipids. Measurements carried out after 12 weeks showed lipid droplets in soleus samples of all four mice investigated, i.e., for both standard chow and Western diet [Figs. 2 and 2]. Nevertheless, Oil red O staining of soleus samples showed an overall low lipid content for all these specimens (data not shown). The diaphragm samples measured at this time point showed lipid droplets in one Western diet and one standard chow specimen. In summary, the measurements indicate overall low amounts of intramyocellular lipid droplets; nevertheless, more of the CARS microscopy images measured in Western diet samples showed lipid droplets compared to standard chow. In agreement, high-fat diet intake has been reported to increase intramyocellular lipid levels in humans.³ However, more in-depth investigations are required to understand the coupling between diet fat content and lipid storage in skeletal muscle tissue.

3.2.

Influence of High-Fat Western Diet on Hepatic Lipid Content

A comparison between hepatic lipid storage in mice fed standard chow or Western diet for 12 weeks is presented in Fig. 3. Lipid contents of liver specimens from four mice, two from each diet group, have been analyzed using both CARS microscopy (left column images) and histochemical staining with Oil Red O (right column images). In the CARS images the strongest signals (in bright yellow) were obtained from lipid droplets located in a matrix of hepatocytes visualized in the background (red).

Fig. 3

(a) through (d) show CARS, SHG, and Oil Red O staining (ORO) microscopy images of liver tissues from wild-type mice fed standard or Western diets. The CARS microscopy images show stored lipids, the SHG images visualize collagen fiber distributions, and the Oil Red O stains hepatic lipids, verifying the CARS technique. Good agreement can be seen between the lipid distributions visualized using CARS microscopy and staining both for standard diet [(a) and (b)] and Western diet [(c) and (d)] mice. The SHG images clearly show an increased amount of fibrous structures, signifying developed liver fibrosis for the mice fed Western diet for 12 weeks [(c) and (d)]. The liver lipid content and the degree of liver fibrosis have been quantified as relative image areas covered by (e) lipid structures and (f) collagen fibers, respectively. The Western diet specimens typically show three times higher amounts of hepatic lipid content and three times higher fibrous content.

Although the specificity that can be obtained with CARS microscopy for visualization of lipids is well known, the complex composition of liver tissue motivates a comparison with conventional methods, and Oil Red O staining of liver sections from the same animals was compared with the results from CARS microscopy. Several staining protocols for different types of lipids exist,²⁹ and using the Oil Red O marker is an established method for staining of lipids in general. It is therefore a suitable choice for visualization of overall lipid storage in the liver tissue, and for comparison with CARS microscopy that visualizes lipids probing CH₂ groups. Nevertheless, for detailed analysis of the hepatic lipid contents, more specific staining alternatives may be appropriate.

Oil Red O staining and CARS microscopy showed lipid distributions with good overall agreement for all individuals with low and high hepatic lipid contents for mice fed standard and Western diets, respectively. CARS microscopy images measured in liver tissues from mice fed the standard diet revealed an interindividual variation in hepatic lipid storage, showing a distribution of small lipid droplets for one individual [Fig. 3] and a liver tissue matrix containing only a few lipid droplets for the other individual [Fig. 3]. This was in line with the results obtained by the Oil Red O analysis, revealing low amounts of stained structures, in particular for the specimen of Fig. 3. Still, the hepatic lipid content was significantly lower in mice fed the standard diet compared with those given the Western diet [Figs. 3 and 3], where much larger lipid structures could be observed with both CARS microscopy and Oil Red O staining. Additionally, the distributions of lipid structures in the livers of these mice show a high degree of heterogeneity.

Average lipid droplet diameters and the relative area covered by lipid structures were evaluated from the CARS microscopy images. The large number of small lipid droplets in the liver samples resulted in average diameters of 3 μm for both diets. However, the size of the largest lipid structures imaged in the Western diet samples was approximately twice the size of those in the standard chow samples. The evaluated maximum diameter averaged over three standard and four Western diet images was 16 and 34 μm, respectively.

The average relative lipid area, presented in Fig. 3, was approximately three times higher for the Western diet mice compared to the standard diet mice with values of 13 and 4% of the scanned area, respectively. The larger spread in the relative lipid area obtained in liver tissues from mice fed the Western diet reflects the heterogeneity in lipid distribution for these livers, which can be observed in the CARS and Oil Red O images of Figs. 3 and 3.

The SHG microscopy images recorded simultaneously with the CARS images in the same field of view are shown in the middle column in Figs. 3 through 3. All four images contain SHG-generating fibrous structures, identified as collagen in previous studies of liver tissue.^{12, 13} The detection of collagen was also confirmed by comparative measurements with van Gieson staining and SHG microscopy on a liver sample from a mouse fed a high-fat diet. The results are shown in Fig. 4 with very good agreement between the collagen stained in pink [Fig. 4] and the structures visualized by SHG microscopy in the same region [Fig. 4]. The high spatial resolution of SHG microscopy and the negligible background makes it possible to discern fine collagen structures difficult to distinguish in the staining image, for example the thin fibers indicated in Fig. 4. Thus, with the ability to measure fresh, unstained samples without sectioning SHG microscopy is clearly a suitable method for specific imaging of collagen in liver tissue.

Fig. 4

(a) Staining and (b) SHG microscopy images of liver tissue from a mouse fed high-fat Western diet. The collagen stained pink in the tissue in (a) matches the structure visualized by SHG microscopy in (b). The SHG image is background-free and details difficult to distinguish with staining can be visualized, exemplified by the thin fibers indicated by arrows in (b).

Comparing the SHG images of Fig. 3, livers from mice fed a standard diet showed a few thin structures, whereas livers from mice fed the Western diet contained entangled networks of thicker fibers, indicating development of liver fibrosis. A pattern of fibers surrounding the hepatocytes can be seen, which is in agreement with the arrangement observed in histochemical images of liver tissue with developed steatohepatitis.³⁰

Figure 3 shows the relative fiber area evaluated from the SHG microscopy images. Although data of both groups show spread values, the highest values, above ∼25%, are obtained for the Western diet specimens with an average value approximately three times higher, 4.8% compared to 1.5% of the image area, than that for the standard diet specimens.

Previous studies in mice have shown that fiber formation in the liver is detected only after a relatively long time of high-fat feeding.³⁰ For example, DeLeve reported developed steatohepatitis in mice after 6 and 9½ months.³¹ Approximately 0.1% (standard control diet), 0.4% (Western diet, 6 months), and 5% (Western diet, 9½ months) of the liver area was covered by Sirius Red-stained collagen fibers. Thus, their 9½ month data correspond to numbers obtained from SHG microscopy data already after 12 weeks, as shown in the present study. However, directly comparing the degree of steatohepatitis from data obtained by means of different microscopy techniques is questionable. As mentioned previously, results obtained with histochemical methods have an uncertainty related to the staining efficiency, and can underestimate the development and extent of liver fibrosis. We show that progression of liver fibrosis can be studied in diet-induced obese mice using SHG microscopy after 12 weeks of Western diet. The high sensitivity of the SHG technique for collagen detection is well known,⁹ as well as the higher spatial resolution achieved compared to staining methods^{12, 32} (Fig. 4). This makes SHG microscopy suitable for early detection of steatohepatitis. Furthermore, recent developments in optical fiber technology and endomicroscopy are beneficial for studies in vivo and also for future clinical use.^{33, 34}

3.3.

Influence of Diet Fatty Acid Composition on Hepatic Lipid Content

It has been suggested that reduced amounts of hepatic long-chain polyunsaturated fatty acids contribute to NAFLD in humans,³⁵ and in particular the marine polyunsaturated fatty acids EPA and DHA.^{36, 37} Indeed, deficiency of essential fatty acids, has long been used in animal models to study the development of hepatic steatosis.³⁸ Related to this, livers from Ldlr-deficient mice fed HFSD beef or herring were analyzed by dual CARS microscopy. Since the HFSD herring contained larger amounts of long-chain polyunsaturated fatty acids, the aim was to investigate whether qualitative and quantitative differences in hepatic lipid storage could be detected by CARS microscopy.

Figure 5 summarizes the results obtained comparing the two diets. On an individual basis, mice fed either diet showed varying liver lipid contents, exemplified by the images in Fig. 5 (low lipid content) and Fig. 5 (high lipid content). Quantitative results obtained from CARS microscopy and triglyceride analysis of liver tissues from mice fed both diets (Table 1 are presented in Figs. 5. The average lipid droplet diameter was approximately 4 μm for both HFSD-herring and HFSD-beef mice, with a difference on the order of the spatial resolution of the images (0.55 μm). The maximum evaluated diameters were 36 and 42 μm for samples from HFSD-herring and HFSD-beef mice, respectively.

Fig. 5

CARS microscopy of liver tissue from Ldlr-deficient mice fed HFSD beef or herring. Tissues of (a) low and (b) high lipid content were found in both diet groups. The liver lipid content, quantified from CARS images as (c) relative lipid area and (d) normalized CARS signal per area, is presented together with (e) the liver triglyceride content measured by biochemical analysis. A slightly higher amount of lipids can be observed in liver tissues of the HFSD-herring category [(d) and (e)].

Comparison of the relative lipid areas, i.e., the density of lipid droplets in the liver tissues, generally showed somewhat higher values for the HFSD-herring diet group, suggesting a higher amount of lipids stored in the liver. This was supported by the biochemical analysis of hepatic triglyceride content presented in Fig. 5, showing a 24% higher median triglyceride level in livers from the HFSD-herring mice. Quantification of the lipid CARS signal is complex due to the unspecific nonresonant signal contribution that forms a background and also couples to the resonant signal. Furthermore, the CARS signal has a nonlinear dependence on the number of probed molecular bonds. Nevertheless, the dual CARS measurements allowed for subtraction of the nonresonant background, resulting in a signal dominated by the resonant CH₂ contribution.To compare the data from triglyceride analysis with data evaluated from the images, the background-corrected CARS signal of the lipid droplets was multiplied with the relative lipid area, resulting in the graph presented in Fig. 5. Comparing this CARS parameter for the two diets shows the same tendency as the triglyceride analysis, with higher lipid content for the HFSD-herring diet, for which a 22% higher median value is obtained.

To obtain additional qualitative information on the hepatic lipid content, CARS microspectroscopy measurements were carried out on two liver samples, one from each diet group, and the resulting spectra are shown in Fig. 6. The spectra show the signal in the region of hydrocarbon vibrations around 3000 cm^–1, and both have the strongest signature located at 2850 cm^–1 (indicated by the top arrow), corresponding to the symmetric stretch vibration of the CH₂ groups in the lipid acyl chains. In addition to this peak, the spectra contain a broader feature between 2900 and 2950 cm^–1. The ratio between the amplitude of the 2850-cm^–1 peak and the average signal in the 2900- to 2950-cm^–1 region is higher for the spectrum of the HFSD-herring category, indicating longer acyl chains³⁹ and resulting in more CH₂ groups in turn giving coherent signal contributions to the resonance at 2850 cm^–1. This interpretation is also supported by the higher ratio between the signal at 2850 cm^–1 and the nonresonant background level measured at 2750 cm^–1, compared to the HFSD-beef spectrum. In addition to the CH₂ signal, the spectrum of the HFSD-herring sample shows an additional spectral line around 2880 cm^–1 (indicated by the bottom arrow), corresponding to the asymmetric stretch vibration of CH₂ groups.^{39, 40, 41} The presence of this peak indicates either longer acyl chains³⁹ or reduction in interchain coupling.⁴² Bonn have related the ratio between the CARS signals at 2880 and 2845 cm^–1 to the degree of acyl chain ordering, and found this parameter to be inversely related to the amount of double bonds, i.e., unsaturated lipids.⁴³ Initially, this indicates a lower amount of unsaturated lipids for the HFSD herring. However, this is neither in agreement with the compositions of the diets nor with the triglyceride fatty acid compositions of the liver samples presented in Table 2. A higher amount of polyunsaturated fatty acids was found in the triglycerides of the HFSD-herring sample. However, this sample also showed an approximately ten times higher amount (20.0% compared to 1.8% molar fractions) of long-acyl-chain (n >= 20) triglycerides, in agreement with the trend observed in the CARS spectra.

Fig. 6

CARS spectra of lipid droplets in liver tissue from Ldlr-deficient mice fed HFSD beef or herring. Both spectra show a distinctive peak at 2850 cm^–1 (upper arrow), corresponding to the CH₂ symmetric stretch vibration. Differences can be seen in the signal strengths of the CH₂ peak and the 2900- to 2950-cm^–1 region relative to the nonresonant background level at 2750 cm^–1. The HFSD-herring spectrum also shows an additional small peak around 2880 cm^–1 (lower arrow); for interpretation, see Sec. 3.

Table 2

Molar percentages of saturated, monounsaturated, and polyunsaturated triglycerides for the HFSD-beef and HFSD-herring liver samples monitored by CARS microspectroscopy.

Although the multicomponent triglyceride composition increases the complexity of interpretation, the results clearly demonstrate that CARS microspectroscopy has the potential to provide not only morphological information on the lipid accumulation but also insights into the chemical composition of the lipid stores from spectral analysis. Multiplex CARS microscopy, generating spectra containing more data points than the presented excitation scans, and analysis methods that allow the pure spontaneous Raman spectrum to be extracted,⁴¹ can be combined with biochemical analysis to investigate liver lipid storage in more detail.

3.4.

Influence of Diet Composition on Intramyocellular Lipid Content

The lipid content of the diaphragms of Ldlr-deficient mice fed HFSD herring and beef were also investigated using CARS microscopy, and the results are shown in Fig. 7. Intramyocellular lipid droplets having diameters around 0.8 μm could be identified for both diet groups, Figs. 7 and 7, and the lipid droplet CARS signals were similar intensities for both categories, indicating droplets of comparable density. The lipid density, evaluated as a relative image area covered by lipid droplets, is shown in Fig. 7 and generally showed higher values for the HFSD herring, median value 1.5%, than for the HFSD-beef specimens with median value 1.0%. Sample images of low and high intramyocellular lipid storage are shown in Figs. 7 and 7, respectively.

Fig. 7

CARS microscopy of diaphragm tissue from Ldlr-deficient mice fed (a) HFSD beef or (b) herring. Intramyocellular droplets can be observed in specimens from both diet groups, and comparison of the relative lipid areas shows a slightly higher amount of lipids for HFSD-herring specimens in (c).