2.1.

Differentiation of Mouse ES Cells to Hepatocyte-like Cells

The in vitro differentiation of mouse embryonic stem cells into hepatocyst-like cells involves three steps: endoderm induction, hepatic initiation, and finally, maturation into hepatocyte-like cells. After 14 days of embryoid body (EB) formation, the trypsinized cells were cultured under a basic fibroblast growth factor (bFGF) treatment for two days to generate definitive endoderm. After a further two days of bFGF treatment, the cells were then treated with hepatocyte growth factor (HGF) for three days in order to form hepatic progenitor cells. Finally, after three days of HGF treatment, cells were further cultured with HGF/oncostatin M (OSM)/dexamethasone (DEX) (HGF/OSM/DEX) for three days to generate hepatocyte-like cells that showed cuboidal morphology typical of in vivo hepatocytes. HGF/OSM/DEX treatment was included in the last stage of the maturation process because this was reported to result in hepatocyte-like cells derived from ES cells.¹⁸

Mouse ES cells (mESCs-SUT-1, strain C57BL/6) were cultured on mouse embryonic fibroblasts for seven days in ES medium containing Dulbecco's modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 200 mM L-glutamine, 0.1 mM β-mercaptoethanol, 1x minimum essential amino acid, 1x penicillin/streptomycin, and 1000 IU/ml hLIF. The ES properties of the cultivated cells were confirmed by determination of cell surface markers (Oct-4, Sox-2, Nanog, SSEA-1, SSEA-3, SSEA-4, TRA-1–60, and TRA-1–81).The ES cells were trypsinized, and then the suspension was cultured in DMEM supplemented with 10% FBS, 200 mM L-glutamine, 0.1 mM β-mercaptoethanol, 1x minimum essential amino acid, and 1x penicillin/steptomycin for 14 days on bacterial dishes at 37 °C with 5% CO₂ in air to generate EB formation.¹⁹ After 14 days, EBs were trypsinized to isolate external endodermal cells, plated on 0.2% gelatin-coated dishes, and cultured in DMEM supplemented with 200 mM L-glutamine in the presence of 100 ng/ml basic fibroblast growth factor (bFGF) for 2 days. Hepatocyte growth factor (HGF) was subsequently added to the culture medium, and the cells cultured for a three more days. Mature hepatocyte-like cells differentiated from definitive endoderm were prepared by trypsinizing the definitive endodermal cells and replating them on 0.362 mg/ml matrigel matrix containing 20 ng/ml HGF, 10 ng/ml oncostatin M (OSM), 100 nM dexamethasone (DEX), and a mixture of insulin, transferring, and selenious acid (5 mg/ml insulin, 5 mg/ml transferring, and 5 μg/ml selenious acid) and cultured for three days.^{20, 21}

2.2.

Magnetic Activated Cell Sorting

Magnetic activated cell sorting (MACS) was specifically employed to isolate and purify a special population of cells based on the cell surface marker definitive for hepatocytes. This technique has been shown previously to be highly effective for isolating the mature hepatocytes from mixed cell suspensions.²² The positive cells were collected and used for FTIR analysis.

The cells were first stained with biotinylated anti-HNF3βrAb or albumin mAbs at 4 °C for 10 min. Cells were washed twice with PBS containing 0.5% BSA and 2 mmol/L ethylenediamine tetra-acetic acid. Enrichment of target cells by magnetic microbeads was carried out according to the manufacturer's protocol. Briefly, cells were resuspended in 80 μL of the same buffer per 10⁷ cells and incubated with antibiotin microbeads for 15 min at 6–12°C. Cells were washed twice and finally resuspended in 500 μL of buffer per 10⁷ total cells. A premoistened magnetic separation (MS) column was placed in the magnetic field of a suitable MACS separator. The cell suspension was applied onto the column and then washed three times. Effluents were collected as the negative fraction. Magnetically labeled cells, as the positive fraction, were collected by flushing the column with buffer after the column was placed outside of the magnetic field. The sorted cells were then further cultured in 95% air and 5% CO₂ at 37 °C for another two days to remove the microbeads.²²

2.3.

Characterization of Differentiated Hepatatocytes

The ES cell-derived hepatocyte-like cells were fixed with 4% paraformaldehyde solution for 30 min at room temperature, then washed three times with PBS (-), followed by incubation in a blocking solution for 2 h. The cells were subsequently incubated with the primary antibody mouse anti-α-fetoprotein (1:200, Sigma) and rabbit antimouse albumin (1:200, Abcam) overnight at 4 °C. After washing out the excess primary antibody, the cells were treated with second antibody conjugated with either Alexa 488 (1:1000, Invitrogen) or rhodomine (1:1000, Invitrogen) for 1 h. The cells were costained with 4’,6-diamidino-2-phenylindole (DAPI) for 5 min and then examined by fluorescence microscopy.

Indocyanine green (ICG) (Sigma) was dissolved in 5 ml of solvent in a sterile vial and then added to 20 ml of α-MEM containing 10% FBS. The final concentration of the resulting ICG solution was 1 mg/ml. The ICG solution was added to the cell culture dish and incubated at 37 °C for 15 min. After the dish was rinsed three times with PBS, the cellular uptake of ICG was examined using a stereomicroscope. After the microscopic examination, the dish was refilled with DMEM containing 10% FBS. ICG was completely eliminated from the cells 6 h later. Differentiated hepatocyte-like cells were immersed in periodic acid solution for 5 min at room temperature and rinsed three times with distilled water. Cells were treated with Schiff's reagent (Sigma) for 15 min at room temperature, washed in running tap water for 5 min, and examined by light microscopy.¹⁸

2.4.

Fourier Transform Infrared Microspectroscopy

Cell cultures were washed with 0.85% NaCl and centrifuged to obtain a cell pellet. Each pellet was washed twice with 0.85% NaCl and resuspended in 50 μl of saline. The number of cells was counted with heamocytometer. All test samples were resuspended in an appropriate volume of saline to give a final concentration of 10⁶ cells/ml. A drop of 5 μl of each sample was deposited onto IR-transparent 2-mm-thick barium fluoride (BaF₂) windows, air dried, and stored in a desiccator until spectra were acquired.

Spectra data were collected at an infrared microspectroscopy beamline (2BM1B) at the Australian Synchrotron, Melbourne, Australia. Spectra were acquired with a Vertex 80v FTIR spectrometer (Bruker Optics, Ettlingen, Germany) coupled with an IR microscope (Hyperion 2000, Bruker) with an MCT detector cooled with liquid nitrogen over the measurement range from 4000 to 800 cm⁻¹. The microscope was connected to a software-controlled microscope stage and placed in a specially designed box that was purged by dry air. The measurements were performed in the mapping mode, using an aperture size of 5 × 5 μm with a spectral resolution of 6 cm⁻¹, with 64 scans co-added. This aperture allowed measurements that resulted in spectra that were representative unambiguously of individual cells. The spectral resolution was a good compromise for biospectroscopic measurements having adequate resolution to reveal a band of biological origin, but broad enough to smooth features that result from absorbance of atmospheric water in the IR beam path. The number of coadded scans was chosen from a series of preliminary trials (data not shown) as a good compromise between achieving spectra with good signal-to-noise characteristics and the rapid acquisition of data. Spectra were processed using Happ–Genzel appodization and 2 × zero filling. Spectral acquisition and instrument control was performed using Opus 6.5 control software (Bruker).

2.5.

Statistical Analysis

Spectra from mapping arrays were quality tested by visual inspection such that spectra with weak absorbance (amide I band maximum absorbance <0.2), acquired from regions of the sample where there were no cells, or spectra with very high absorbance (amide I band maximum absorbance >0.8), acquired from regions where cells may have been clumped or overlaid) were rejected from the analysis. Spectra from sample groups based on differentiation time were analyzed using principal component analysis (PCA)²³ and partial least-squares discriminant analysis (PLS-DA).²⁴ Preprocessing of the data was conducted by first performing a second derivative by the Savitzky–Golay method (nine smoothing points) and then normalized using extended multiplicative signal correction (EMSC)^{25, 26} using the spectral regions from 3000 to 2800 cm⁻¹ and from 1800 to 1000 cm⁻¹ using The Unscrambler 9.7 software (CAMO, Oslo, Norway). Score plots were used to visualize any clustering of the data, and the loading plot was used to determine which spectral regions contributed most to the variance in the data set, accounting for any clustering of spectra seen in scores plots. Calibration and validation of spectral data were employed using data sets that were randomly selected comprising two-thirds and one-third of the spectra, respectively. PLS-DA employed PCA models derived from calibration sets and was used to test the ability to discriminate different cell types using the independent validation set spectra.

2.6.

Unsupervised Hierarchical Cluster Analysis

Unsupervised hierarchical cluster analysis (UHCA) was performed on second derivative spectra using Ward's algorithm, which utilizes a matrix-defining interspectral distance to identify the most similar IR spectra. Spectral distances were calculated as D-values.²⁷

3.1.

Confirmation of Hepatic Differentiation by Immunofluorescence

Confirmation of the hepatic protein expression was probed by immunostaining, as described above. At the last stage of the maturation process, differentiated hepatocyte-like cells are expected to express proteins characteristic of mature allbumin and the α-feto-protein. Indeed, at the last stage of maturation, hepatocyte-like cells secreted a high abundance of α-feto protein and albumin, as presented by the immunostaining method seen in Figs. 1a, 1b, respectively. Albumin is the most abundant protein synthesized by mature hepatocytes and contributes to maintaining oncotic pressure as well as being involved with the transport of hydrophobic molecules.²⁸

Fig. 1

Cellular and functional characterization of hepatocyte cells at the last stage of the maturation process. Immunocytochemical analysis of expression of α-fetoprotein and albumin production. (a) shows α-fetoprotein, (b) shows DAPI-stained nuclei and (c) overlayed. (d) shows albumin, (e) DAPI-stained nuclei and (f) overlayed. The scale bars represents 100 μm in (a-c) and 200 μm in (d-f).

The presence of glycogen was confirmed by positive staining of glycogen granules with purple staining where glycogen is present in mature hepatocyte-like cells. Cells were positive for PAS straining, indicative of glycogen storage, a characteristic of mature hepatocytes [Fig. 2a].

Fig. 2

(a) Glycogen storage determined by periodic acid-Schiff staining and (b) cellular uptake of indocyanine green into mature hepatocytes at the last stage of the maturation process.

ICG is an organic anion that is used as a test substance to examine liver function. ICG is tricarboxylic acid compound that is taken up and secreted by hepatocytes. Therefore, cellular uptake of ICG was used to identify differentiated hepatocyte-like cells. ICG-positive cells were detected in the cultured cell suspension [Fig. 2b].

3.2.

Fourier Transform Infrared Microspectroscopy Analysis of Hepatocyte-Like Cells and Their Progenitors

In order to generate a high percentage of functional hepatocyte-like cells, it is important to establish not only an appropriate culture system for cell differentiation, but also a practical method for identifying differentiated hepatocysts. The FTIR microspectroscopy technique provides a new method to identify spectroscopic features indicative of endoderm induction, hepatic initiation, and mature hepatocyte-like cells. The infrared spectrum of biological samples can provide spectral information pertaining to nucleic acids, proteins, carbohydrates, and lipids in the spectral domain between 1800 and 800 cm⁻¹, known as the “fingerprint region” of the FTIR spectrum. By using SR-FTIR microspectroscopy, it was possible to discriminate the three different cell populations—induced endoderm cells, cells undergoing hepatic initiation, and mature hepatocyte-like cells at a single-cell level. This was enabled by using the high brightness synchrotron infrared radiation source at Australian Synchrotron, where infrared light was focused to a small spot at the sample with a focal-plane aperture in the IR microscope of 5 × 5 μm.

Protein absorbance in the amide I region of the FTIR spectrum (∼1700–1580 cm⁻¹) is sensitive to the prediction of secondary protein structure. The two major band envelopes of the amide I and amide II of the protein arise from specific stretching and bending vibrations of the protein backbone. The amide I band arises predominantly from C = O stretching vibration (80%) of the amide C = O group and C-N stretching vibration (20%). The amide II vibration frequency occurs from N-H bending vibration (60%) coupled with C-N stretching (40%). The spectral region between 1700 and 1500 cm⁻¹ is dominated by the amide I and amide II bands. This region can be used to discriminate secondary structures of protein, especially in the amide I band region between 1700 and 1600 cm⁻¹, which can give information about α-helix, β-sheet, and turn secondary structures. There are three spectral features that can be assigned to α-helix (centered at ∼1650 cm⁻¹), β-sheet (centered at ∼1630 cm⁻¹) and turn (centered at ∼1685 cm⁻¹) secondary structure components. However, it must be said that the amide I maximum is known to shift in response to light-scattering effects, such as resonant Mie scattering,²⁹ and hence, any interpretations of changes in protein confirmation based on the amide I band must be considered in this context.

The average FTIR absorbance spectra of mature hepatocyte-like cells and their less differentiated precursor cells are shown in Fig. 3a. It was difficult to observe spectral differences between different cell types by examining the raw spectra alone. Indeed, to be able to visualize spectral changes occurring throughout the cell-differentiation process, the second derivative spectrum was performed, which gives a minimum peak for vibrational bands and allows easier identification of individual peaks that may be superimposed in the raw spectra.

The integrated area of the amide I was more intense in the last stage of hepatocyte differentiation compared to endodermal cells [p < 0.05 by ANOVA; Figs. 3b, 3c]. The major intensity for the amide I band in the spectra of hepatocyte-like cells was centered at 1656 cm⁻¹, which was ascribed to α-helix protein secondary structure. These results suggested that during the differentiation process of mature hepatocyte-like cells, there appeared to be an increase of specific proteins that have a predominantly α-helix structure.

These results can be explained by the emergence of albumin protein during hepatocyte differentiation, which is rich in α-helix content,³⁰ and in agreement with an increase in albumin levels detected by immunostaining. The β-sheet secondary structure centered at 1627 cm⁻¹ in the hepatic progenitor cells was shifted to 1641 cm⁻¹ in the maturation process. The amide II absorbance was predominant in the stage of endoderm induction [p < 0.05, by analysis of variance (ANOVA)]. These results indicate that different proteins are expressed during differentiation following endoderm induction, and subsequent differentiation of hepatocyte-like cells from hepatic progenitor cells. By contrast, the endoderm induction stage was shown to have low absorbance intensity for lipid bands, which increased with the differentiation of the hepatic progenitor cells. Finally, lipid band intensity significantly decreased through the last stage of the maturation process to the formation of the hepatocyte-like cells (Fig. 3).

Fig. 3

(a) Average raw FTIR spectra of endoderm induction, hepatic progenitor cells, and maturation into hepatocytes. A total of 135 spectra, 214 spectra, and 110 spectra of endoderm induction, hepatic initiation, and maturation, respectively, were measured with 64 scans co-added for each individual spectra. (b) Average second-derivative spectra after nine points of smoothing and normalization with extended multiplicative signal correction over the spectral range of 4000–1000 cm⁻¹. (c) The relative areas (in percent) were calculated for lipids by the integral area of the curve for lipid (ester carbonyl centered at 1741 cm⁻¹ and C-H stretching (3000–2800 cm⁻¹) vibration, for amide I protein (1700–1600 cm⁻¹), for amide II protein (1600–1500 cm⁻¹). The relative integrated area (in percent) of the macromolecular component was calculated in normalized second-derivative infrared spectra by using OPUS 6.5 software. The differences between the means of the percentages of relative integral areas of three biochemical components were significantly different among three groups by ANOVA. Data are given as means ± SD; p < 0.05 was accepted as significantly different between mature hepatocytes and other groups. Error bars indicate standard errors of the means.

3.3.

Statistical Analysis

FTIR microspectroscopy combined with multivariate data analysis, in particular PCA, has been applied to explain biochemical changes occurring during cellular differentiation.¹¹ PCA of biospectroscopic data provide two types of information: visualization of clustering of similar spectra within data sets in scores plots and identification of variables (spectral bands representing various molecular groups within the samples) in loadings plots explaining the clustering observed in the scores plots. PCA enabled us to identify the most significant spectral changes that occurred during hepatic differentiation, from endoderm induction to the hepatic progenitor cells, and subsequent maturation of these into hepatocyte-like cells.

The PCA score plot shows that spectra from endoderm induction, hepatic progenitor cells, and mature hepatocyte-like cells can be clustered separately along PC1 and PC2. Spectra from mature hepatocyte-like cells could be distinguished from less differentiated cell by having positive PC1 scores [Fig. 4a] In general, spectral loadings were dominated by changes in protein bands (1600–1500 cm⁻¹) and lipid bands (3000–2800 cm⁻¹). Variation in the lower wavenumber region of the spectrum below 1500 cm⁻¹ appeared to contribute less to the observed clustering in the scores plot. However, small loadings at 1240 and 1085 cm⁻¹, attributed to the asymmetric and symmetric phosphodiester stretching bands, may have been due to differences in nucleic acids and other phosphorylated molecules between the hepatocyte-like cells and less differentiated cell types. The reason why changes in these bands are not very prominent in the data set may be due to use of dried samples, because previous FTIR studies have shown that band information from nucleic acids can be obscured in dried cells.³¹ The greatest spectral variation indicated by the PC1 loadings plot was found in the amide I region of the IR spectrum. Opposite PC1 loadings at 1656 and ∼1627 cm⁻¹ [Fig. 4b] may reflect differences in protein secondary between hepatic progenitor cells and mature hepatocyte-like cells, observed in the second derivative spectra (Fig. 3). The negative PC1 loading at 1656 cm⁻¹ may indicate higher levels of α-helix protein secondary structure in mature hepatocyte-like cells compared to progenitor cells; whereas the positive loading at 1627 cm⁻¹ could be interpreted as higher levels of a β-sheet secondary structure in the progenitor compared to mature hepatocyte-like cells, explaining spectra from definitive endoderm and hepatic progenitor cells clearly separating along PC2 with hepatic progenitor cells having positive PC2 scores. The separation along PC2 can also be explained by the negative loading for PC2 in the spectral region corresponding to total lipid (CH₂, CH₃ stretching) centered at 2852 and 2923 cm⁻¹, which was oppositely correlated with positive loadings for amide I absorbance at 1656 cm⁻¹. The heavy negative loadings for the C–H stretching bands was correlated with the intense lipid ester carbonyl band loading at 1740 cm⁻¹, indicating higher lipid levels in hepatic progenitor cells compared to mature hapatocytes [Fig. 4b].

Fig. 4

PCA analysis of mouse hepatocyte cells, endoderm induction, and hepatic progenitor cells using the same spectral data sets as described in Fig. 3. (a) Score plot and (b) loading plot of independent spectra acquired from SR-IR mapping. The chemical composition of the three groups was classified with the PC1 versus PC2 score plots. PC1 and PC2 explained 29 and 21% of the total variance, respectively. Spectra were derived using second-derivative processing with the entire biochemical cell fingerprint region (1800–900 cm⁻¹) and (3000–2800 cm⁻¹).

3.4.

Unsupervised Hierarchical Cluster Analysis

Hierarchical cluster analysis (HCA) was performed to identify similarities and differences between spectra from different cell types using spectral information in the ranges of 3000–2800 and 1800–900 cm⁻¹. The dendrogram represented in Fig. 5 permits the identification of three major clusters of spectra. Significantly, the spectra within the upper branch (branch A) were clearly separated from other branches and clustered spectra from cells from the last stage of maturation process. Whereas, spectra from mature hepatocyte-like cells in branch A were separated from two further clusters corresponding to spectra from hepatic progenitor cells (branch B1) and definitive endoderm (branch B2), respectively. This is further evidence for the distinct FTIR spectral profiles corresponding to the three cell types as revealed by PCA and described above.

Fig. 5

Dendrogram obtained by cluster analysis of endoderm induction, hepatic progenitor cells and mature hepatocyte-like cells using the same spectral data sets as described in Fig. 3. Cluster analysis was employed by Ward's algorithm using the second derivative and then vector normalized over the spectra region from 3000 to 2800 cm⁻¹ and 1800 to 900 cm⁻¹. Note that the larger the separation between clusters is, indicated by connecting lines, the lower the similarity of the spectra is.

3.5.

PLS-DA Modeling

PLS-DA was applied to classify spectra from the three differentiation stages: endoderm induction, hepatic progenitor cells, and mature hepatocyte-like cells. This technique is based on observation of linear correlations within the data set between two factors, X and Y. In the present study, X is multivariate, representing the absorbance value for each wavenumber from the spectrum. Three Y variables were set as a code for each cluster class using nominal value of either +1 or 0. The PLS model was then used to predict Y values for an independent set of validation spectra, the results of which formed the basis for the classification result. The correlation coefficient of 0.96 indicated the data was well modeled [Fig. 6a]. PLS-DA accurately classified the three groups of validation spectra with 93.8, 96, and 93.2% of spectra correctly assigned to mature hepatocyte-like, hepatic progenitor, and endoderm induction cells, respectively [Fig. 6b].

Fig. 6

PLS-DA modeling for the calibration and training set based on observation of linear correlation within the data set. (a) The calibration (training) set was plotted and measured Y against predicted Y (mature = +1, endoderm induction = 0 and hepatic progenitor cells = 0). (b) The validation set was used to predict Y using the PLS-DA model. All spectra were clearly classified, as indicated the 94% correct classification for the independent validation set.