Application of Fourier transform infrared spectroscopy to biomolecular profiling of cultured fibroblast cells from Gaucher disease patients: A preliminary investigation

,

Gaucher disease (GD) is the most frequent lysosomal glycolipid storage disorder, defined as an autosomal recessive disease resulting from the deficiency of glucocerebrosidase (GBA) (E.C.3.2.1.45),a lysosomal hydrolase, also known as acid-β-glucosidase.GBA is responsible for the degradation of glucosylceramide (GC), a natural glycosphingolipid, into ceramide and glucose.The deficiency of this enzyme results in the accumulation of undegraded glucosylceramide, almost exclusively in macrophages.GD is a multisystemic disorder, characterized by a spectrum of the clinical subtypes.GD is classically divided into non-neuronopathic type I (OMIM #230800) and neuronopathic types II (OMIM #230900) and III (OMIM #231000) variants.Although the most symptomatic individuals have hematological, visceral and skeletal diseases, there is a wide variation in the pattern and severity of extra-neurological involvement among these patients.Enzyme replacement therapy (ERT) is the preferred strategy in the treatment of GD. [1][2][3] Infrared (IR) radiation is a noninvasive and nondestructive type of radiation and it causes the vibration of the covalent bonds of molecules when absorbed by tissues, body fluids or cells.Fourier transform infrared (FTIR) spectroscopy is a widely-used and preferred method of IR spectroscopy due to its speed and sensitivity.The resulting peaks correspond to the frequencies of vibrations between the bonds of the atoms making up the sample. 4,5Different functional groups absorb characteristic frequencies and the resulting spectrum represents the unique molecular fingerprint of the sample material. 6Attenuated total reflectance (ATR) is one of the approaches for FTIR sampling, which allows us to measure the analyte directly by using the ATR crystal.With FTIR, the complete molecular diversity of the samples (all types of organic and many types of inorganic compounds) can be studied comparatively with the knowledge of the peak origins, such as lipids, proteins, carbohydrates and nucleic acids.The amount of the particular compounds can also be determined by this method.[9][10][11][12] The gold standard for the diagnosis of GD is the measurement of β-GBA activity in leukocytes or cultured fibroblasts obtained from the patients by skin punch biopsy. 3The primary aim of this preliminary study is to introduce and evaluate the potential of ATR-FT-IR spectroscopy technique for biomolecular profiling of cultured fibroblast cells derived from GD patients, as well as to investigate the individual phenotypic heterogeneity of the disease at the molecular level.

Patients and sample collection
Study group consisted of 8 unrelated Turkish GD patients representing 5 different genotypes who were selected from among the patients admitted to Hacettepe University, Department of Pediatric Gastroenterology (Ankara, Turkey).There were also 2 healthy adult controls.The diagnosis was suspected on account of the clinical features, which included the age of onset, organomegaly, skeletal features, presence of progressive central nervous system symptoms, as well as reduced cellular GBA activity.Fibroblast cells were obtained from the patients by 0.4 cm skin punch biopsy from the forearm under sterile conditions, following their written informed consent, with the approval of the institutional ethical review board.Clinical features of GD patients are presented in Table 1.

Fibroblast cell culture and sample preparation for FTIR spectroscopy
For FTIR studies, primary skin fibroblast cell culture samples were used.The biopsy samples were washed twice with phosphate buffered saline (PBS) containing 1% fetal calf serum (FCS, Biochrome AG, Berlin, Germany), 3% penicillin/streptomycin (Biochrome AG) and 3% amphotericin B solution (Biochrome AG).The biopsies were placed in 60 mm tissue culture dishes and minced.The small pieces were adhered onto the dishes with the help of a lancet and supplemented with 20% fetal bovine serum (FBS, Biochrome AG), 1% penicillin-streptomycin and 2 mM L-glutamine (Biochrome AG) in high glucose (4.5 mg/ml) DMEM (Biochrome AG).The cells were maintained in high glucose DMEM supplemented with 10% FBS, 1% penicillin-streptomycin and 2 mM Lglutamine in a humidified atmosphere containing 5% CO 2 at 37°C.A 75 cm 2 culture flask of fibroblast (90-95% confluency) was washed with 1X PBS for 3-4 times and the fibroblasts were detached by mechanical scraping and counted.For the standardization of the spectroscopic measurements, after the total number of the detached cells was set to 5 × 10 6 , each tube was centrifuged at 1000 × g and the cell pellets were resuspended in 300 µl of 1X PBS for FTIR spectroscopy studies.

FTIR spectroscopy data collection and evaluation
Infrared spectra were obtained by a Bruker Tensor 27 FTIR (Bruker Optik GmbH, Ettlingen, Germany) equipped with a liquid N 2 cooled photovoltaic Mercury Cadmium Telluride detector and a universal ATR cell with a ZnSe crystal (Pike Technologies, Wisconsin, U.S.A).Continuously purging N 2 gas was provided during all the measurements.Samples were measured after hav-ing been dried on the crystal, similar to the method applied in the literature. 14,15Briefly, fibroblast cell samples were mixed gently prior to the measurements to obtain homogeneity.2.5 µL of cell suspension in PBS were spotted onto the ZnSe ATR crystal.The sample was dried on the crystal by evaporation using a mild N 2 gas stream for 3 min.PBS without cells was also measured under the same conditions and used as a background spectrum for subtraction.The air was recorded and subtracted automatically by the software before all the measurements.The spectra recorded in the mid-infrared region, between 4500-850 cm -1 wavenumbers and interferograms were accumulated for 50 scans at 4 cm -1 resolution at 22°C in the single-bounce ATR mode.Each sample was measured 3 times, the spectra were compared if they were identical and the average spectra were used for further analyses.All the data collection, manipulation and analyses were performed by OPUS software (Bruker Optik GmbH, Ettlingen, Germany).
PBS spectrum was subtracted from the sample spectra by using the OPUS software.The difference spectra were baselined using the rubber-band method and the resulting baselined spectra were used for the exact integration calculations.Min-max normalization was applied with respect to the Amide A (between 3500-3100 cm -1 ) and Amide I peaks (between 1700-1600 cm -1 ) for the regions between 3500-2750 and 1800-850 cm -1 wavenumbers, respectively for illustrative purposes.The second derivatives of the absorbance spectra, in which the absorption maxima appear as the minima, were calculated using Savitzky-Golay algorithm with 9 smoothing points.For comparing protein secondary structures, relative intensity values of the second derivative peaks in the Amide I region (1700-1600 cm -1 ) were used; they were obtained by automated peak picking after vector normalization.
In order to achieve spectral differentiation, according to the molecular fingerprints of the samples, a cluster analysis (hierarchical clustering) was carried out using the OPUS software.Ward's algorithm with Euclidian distances was used to construct the dendograms, which is known to give good predictions among the available algorithms. 16,17Sensitivity and specificity values were calculated for cluster dendrograms as described by Severcan et al. 17

Statistics
The differences of the means were compared by Student's t-test.Levene's test for homogeneity of variances was performed prior to the statistical tests and Levene's p-values were taken into consideration while evaluating the results of Student's t-test.The p-values equal or less than 0.05 were considered as significantly different and the results were expressed as means ±SD.SPSS 13.0 software was used for the analyses.

Biomolecular characterization of the fibroblast cells based on the FTIR spectrum
We obtained FTIR spectra of the fibroblast cell samples between 4500-850 cm -1 wavenumbers in the mid-infrared region.The spectra were complex with many characteristic peaks.Fig. 1A shows the representative FTIR absorbance spectrum obtained by averaging all the spectra used in the analyses.9][20] We calculated the total lipid content as a sum of the areas of the peaks 2, 3, 4, 5, 6, 8, 9 and 13, which originate from lipids.The band between 4100-3580 cm -1 was ignored, since it was originated from O-H stretchings of residual H 2 O. 21 Twenty major absorption peaks in our FTIR spectra, which are numbered in Fig. 1, were assigned in the present study to various biomolecules such as proteins, lipids, cholesterol esters, nucleic acids and carbohydrates (Table 2) by matching the wavenumbers with the references obtained from the literature (references are cited in the caption for Table 2).Peaks at 3294 (Amide A), 3060 (Amide B), 2871 (CH 3 symmetric stretching), 1653 (Amide I) and 1544cm -1 (Amide II) wavenumbers mainly originate from various bond vibrations of proteins (Table 2).Region between 3500-2750 cm -1 wavenumbers contains contribution mainly from lipids.CH 3 asymmetric stretching (2957 cm -1 ), CH 2 asymmetric (2929 cm -1 ) and symmetric (2857 cm -1 ) stretching, olefinic =CH (3011 cm -1 ) and C-H (2843 cm -1 ) stretching modes are the major bond vibrations in this region, originating from lipids (Table 2).Also, there is a peak at 1455 cm -1 that was assigned to CH 2 bending of lipids.Saturated ester C=O stretching at 1732 cm -1 originates from the cholesterol esters or phospholipids.Peaks between 1093-914 cm -1 wavenumbers mainly result from phosphate groups of nucleic acids, phospholipids and phosphorylated proteins and C-O stretching of carbohydrates.

Comparative analysis of the levels of biomolecules
The intensity or area of a particular FTIR absorption peak reflects the amount of the assigned compound for that wavenumber. 12,19Although the primary purpose of this preliminary study is to introduce the method to GD research rather than to understand the molecular mechanisms underlying GD, we also included the control group in order to see if there were very prominent differences.First of all, the quantitative and qualitative comparisons were performed between GD and control groups in terms of spectral differences.We observed overlapping peaks A -between 4500-850 cm -1 obtained by averaging all the spectra used in the analyses; B -between 3500-2750 cm -1 ; C -1800-850 cm -1 wavenumbers.Refer to Table 2 for spectral assignments of numbered peaks.The stacked spectra show aletrations between mean spectra of control, treatment and notreatment samples, and were normalized with respect to the Amide I peak between 1700-1600 cm -1 after baseline correction.
of GD and control samples and detected some quantitative differences and minor peak shifts between GD and control group, as shown in Fig. 1B, C. Means of absorbance peak area values of GD and control groups obtained by integration are presented as bar graphics in Fig. 2A.According to our results, lipid, protein and carbohydrate levels were increased in GD when compared to controls.
Moreover, we grouped GD samples according to the severity (severe or moderate), type (Type I, II and III) and treatment (ERT) condition (Table 1).We observed statistically significant differences between the samples of patients who had been taking ERT treatment and those who had not received treatment.The areas of the peaks at 3294, 1653, 1455, 1396 and 1021 cm -1 wavenumbers indicating lipids, proteins and polysaccharides increased in the samples after treatment (Fig. 1B, 1C, 2B).However, we did not observe a significant decrease in the areas of the peaks originating from CH 2 and CH 3 anti-symmetric and symmetric stretching, which are the main IR absorption peaks of lipid backbone, including acyl chains of GC.
Since one of the aims of the present study is to investigate the phenotypic heterogeneity of GD fibroblast cells by FTIR, we also analyzed the individual variation regarding the peak patterns of the samples.We observed some differences between the spectra in terms of peak area and shifts (Fig. 3).
Table 2. Peak assignments of major absorbance peaks of fibroblast samples detected in the present study and their definition with relative organic compounds based on the literature 6,12,[17][18][19][22][23][24][25][26] (peak numbers illustrated in Fig. 1B, C  A -GD and control; B -treatment and no-treatment groups based on our results. Peak umbers are illustrated in Fig. 1B, C and spectral assignments are given in Table 2.

Comparative analysis of the protein secondary structures
FTIR peaks, including Amide I, consist of several sub-peaks, which can be obtained by derivatization. 19e analyzed the second derivative spectra of Amide I peaks (1653 cm -1 ) in order to determine and compare protein secondary structures, which is very important in biological systems. 9,23[9][10][11] According to our results, the second derivative of the Amide I peak had 5 sub-peaks at 1697, 1684, 1654, 1636 and 1615 cm -1 (Fig. 4).Peak assignments for protein secondary structures are given in Table 3.We observed slight differences and peak shifts between the control and GD groups (Table 4).α-helix structure was observed as being slightly increased in GD, while β-sheet structure was decreased.These structures decreased slightly in samples after treatment, compared to no-treatment group.Statistical tests did not reveal significant differences between GDcontrol, type I-II-III, severe-moderate and with-without treatment groups in terms of peak intensities and shifts.
Also, we investigated the individual variations in the second derivative peak patterns of the Amide I peak (Fig. 4).We observed individual alterations on protein secondary structures in terms of peak intensity and frequency shifts.

Classification by hierarchical cluster analysis
We performed hierarchical cluster analysis in order to achieve spectral differentiation according to the molecular fingerprints of the samples.The resulting dendrograms were evaluated in regard to such clinical features as severity, type of the disease, mutation and treatment condition.Our analyses yielded a classification between the sample of patients who had been taking ERT treatment and those who were not at the time of sampling, based on the 2 main spectrum areas (lipid and fingerprint regions) of the min-max normalized spectra (Fig. 5A).
In the present study, we also observed partial differentiation within GD group according to the severity of the disease, but it was not as significant as the obtained classification related to the treatment.Using the region between 3035-2820 cm -1 after the second derivative and vector normalization preprocessings, we obtained a dendrogram which partially classified GD patients into severe and moderate groups (Fig. 5B).
,11 See Fig. 4 4. Means of vector normalized second derivative relative peak intensities in Amide I region of GD and control groups.Relative intensity values obtained after peak picking by OPUS software.See Fig. 4 for peak numbers and Table 3   Second derivative spectra of Amide I absorption peak of the samples between 1700-1600 cm -1 wavenumbers (with a center at 1653 cm -1 ), which demonstrates main protein secondary structures and individual variation.Absorption maxima appear as minima and the spectra were vector normalized.Refer to Table 3 for peak numbers and their assignments

Discussion
Our optimized procedure allowed us to characterize the major biomolecules in the cultured fibroblast cells of GD patients in a qualitative and quantitative manner.According to our results, lipid, protein and carbohydrate levels were found to be increased in GD compared to controls.GC (a glycosphingolipid) is the major GD-related lipid type that accumulates in the cells with GD.Secondly, glucosysphingosine is also accumulated by GD cells but at much lower concentrations. 27A combined increase in the lipid and carbohydrate levels may reflect increased GC storage in lysosomes, as a result of the deficiency of GBA in GD patients.Since GD is a rare disease, our sample size is enough for a preliminary study aiming to introduce the method.We observed some significant molecular alterations (e.g., lipid, protein and carbohydrate levels) between groups with and without treatment.A comparative follow-up study with a larger sample size of GD and controls should be done in order to show correlations between the biomolecule levels (especially GC) and GD condition.
Moreover, our analyses on the spectral patterns revealed that GD fibroblast samples showed an individual variation at the molecular level.These alterations were mainly in protein, carbohydrate and fatty acid levels and structures, according to our results.More than 250 GDrelated mutations in the GBA gene have been identified to date. 2 Emre et al. reported that L444P and N370S were the most prevalent mutations among Turkish GD patients. 13According to the literature, genotype-phenotype correlation is not seen in GD and clinical symptoms show heterogeneity. 1,2,13The results of the present study also showed no correlation between the genotype and the molecular fingerprint of GD fibroblast samples, confirming the diverse phenotypic expression and heterogeneity in GD.Additionally, our results showed that phenotypic variation must be taken into consideration in FTIR spectroscopic studies on GD and the sample size must be accordingly large to eliminate individual variations.
Our results have also indicated conformational changes in proteins between experiment groups.However, our small sample size did not allow us to perform comprehensive statistical tests for detecting possible significant correlations.Since the main purpose of this preliminary study was to introduce FTIR spectroscopy to GD research, rather than to reveal the molecular mechanisms of the disease, our results would still be useful for the method optimization as a reference work.Our results also showed that GD fibroblast samples expressed phenotypic heterogeneity in terms of the second derivative peak pattern reflecting the change in protein conformation.
Our preliminary multivariate cluster analysis yielded a spectral classification of the groups with and without treatment, suggesting that the treatment process of GD may cause alterations in the level of biomolecules in GD fibroblasts.In our study, only 1 sample with treatment (G4) was classified in no-treatment group.This misgrouping may be related to the duration of the treatment, since G4 patients had the shortest treatment period, compared to other patients.Our hierarchical cluster analysis results supported our statistical tests made by using individual peak areas.The classification regarding the severity of the disease was less significant.G5 and G6 patients were clinically diagnosed as severe but classified into the moderate group in the dendrogram (Fig. 5B), posssibly as a result of the phenotypic heterogeneity of the samples.
As a summary, primary skin fibroblast cell culture samples obtained from GD patients were studied by using FTIR-ATR for the first time to introduce the technique and to assess the potential of molecular IR fingerprints as an alternative approach in research related to lysosomal glycolipid storage disorders.We assigned different peaks of the complex FTIR spectrum to various bioorganic molecules.We observed significant differences between the groups with or without treatment, indicating structural and functional alterations due to the ERT.Also, we observed individual variations reflecting the phenotypic heterogeneity of GD at the molecular level.
Our results showed that FTIR spectroscopy was a rapid and sensitive technique and had the ability to give detailed information about important functional groups of bioorganic molecules in fibroblast samples that are widely used for the diagnosis of GD.Our study also showed that this research tool might provide molecular insights into the pathophysiology of GD.FTIR spectroscopy has already been used in medical diagnostics for some diseases as an alternative to well-known biochemical and histochemical methods and it is suitable for routine use. 28,29he present study was a complementary investigation to our ongoing proteomic analyses of GD and allowed us to carry our reserarch from the proteomic level to metabolomics in the biology systems frame.We attempted to introduce and apply FTIR spectroscopy to make molecular characterization of GD and our results showed that the application of this technique in GD research deserves further investigation to assess its potential in the diagnosis and monitoring of GD.Therefore, detailed studies should be done with an increased sample size.

Fig. 4 .
Fig.4.Second derivative spectra of Amide I absorption peak of the samples between 1700-1600 cm -1 wavenumbers (with a center at 1653 cm -1 ), which demonstrates main protein secondary structures and individual variation.Absorption maxima appear as minima and the spectra were vector normalized.Refer to Table3for peak numbers and their assignments

Table 1 .
Clinical information about the patients included in this study ) for peak numbers for their spectral assignments