Raman spectroscopy analysis of lipid droplets content, distribution and saturation level in...

FULL ARTICLE

Raman spectroscopy analysis of lipid droplets content,distribution and saturation levelin Non-Alcoholic Fatty Liver Disease in mice

Kamila Kochan1, 2, Edyta Maslak2, Christoph Krafft3, Renata Kostogrys4, Stefan Chlopicki2, 5,and Malgorzata Baranska*, 1, 2

1 Faculty of Chemistry, Jagiellonian University, Krakow, Poland2 Jagiellonian Centre for Experimental Therapeutics (JCET), Jagiellonian University, Krakow, Poland3 Leibniz-Institute of Photonic Technology (IPHT), Jena, Germany4 Department of Human Nutrition, Faculty of Food Technology, Agricultural University, Krakow, Poland5 Department of Experimental Pharmacology, Jagiellonian University, Krakow, Poland

Received 24 July 2014, revised 5 September 2014, accepted 16 September 2014Published online 27 October 2014

Key words: Raman mapping, KMC, liver, lipid droplets, NAFLD

* Corresponding author: e-mail: [email protected], Phone: +48 12 6632064, Fax: +48 12 6340515

Non-Alcoholic Fatty Liver Disease (NAFLD) is a com-mon liver disorder, characterized by an excessive lipidsdeposition within the hepatic tissue. Due to the lack ofclear-cut symptoms and optimal diagnostic method, theactual prevalence of NAFLD and its pathogenesis re-mains unclear, especially in the early stages of progres-sion. In the presented work confocal Raman microspec-troscopy was used to investigate alterations in the chemi-cal composition of the NAFLD-affected liver. We haveinvestigated two NAFLD models, representative formacrovesicular and microvesicular steatosis, induced byHigh Fat Diet (60 kcal %) and Low Carbohydrate HighProtein Diet (LCHP), respectively. In both models weconfirmed the development of NAFLD, manifested bythe presence of lipid droplets (LDs), but of different sizes.Model of macrovesicular steatosis was characterized bylarge LDs, whereas in the microvesicular steatosis modelsmall droplets were found. In both models, however, weobserved a significant decrease in the degree of unsatura-tion of lipids, in comparison to the control. In addition,for both models, the impact of medical treatment with se-lected drugs (perindopril and nicotinic acid, respectively)was tested, indicating a significant influence of medicinenot only on the occurrence and size of the droplets, but

also on their composition. Inboth cases the drug treat-ment resulted in an increase of the degree of unsaturationof lipids forming droplets. Confocal Raman microspectro-scopy was proven to be a powerful tool providing detailedinsight into selected areas of hepatic tissue, following theNAFLD pathogenesis and diagnostic potential of the ap-plied drugs.

J. Biophotonics 1–13 (2014) / DOI 10.1002/jbio.201400077

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1. Introduction

Non-Alcoholic Fatty Liver Disease (NAFLD) is oneof the most common chronic liver disorders, affectingaround 20–30% of human population worldwide. It isoften regarded as a liver manifestation of metabolicsyndrome, which represents a set of risk factors lead-ing to various diseases, including diabetes mellitus(DM) type 2 and cardiovascular disease (CVD) [1–4].Metabolic syndrome covers symptoms such as abnor-mal obesity, hypertension, insulin resistance and dysli-pidemia, whereas NAFLD is characterized by an in-creased accumulation of lipids within hepatic tissue(exceeding 5–10% of liver weight), occurring in theabsence of excessive alcohol abuse [1, 2, 5, 6]. NAFLDin general may be understood as a continuum, extend-ing from the initial pure fatty liver through non-alco-holic steatohepatitis (NASH) to cirrhosis and even-tually – irreversible liver damage [1, 5, 6].

Insulin resistance [1, 5, 7, 8] and oxidative stress[1, 5, 9, 10] play a major role in the pathogenesis ofNAFLD. However, the mechanisms involved arestill not fully understood [11]. Currently, the mostpopular model explaining the pathogenesis ofNAFLD (and, in fact, alcoholic fatty liver) is the‘two-hit’ theory [1, 5, 9, 11–13], but this model hasnumber of limitations. According to ‘two hit’ theory,the ‘first hit’ is the accumulation of triglycerideswithin the hepatic tissue (simple steatosis) as a resultof excess fat consumption and/or insulin resistance.This stage increases susceptibility of the liver to in-jury mediated by ‘second hits’, such as inflammatorycytokines/adipokines, mitochondrial dysfunction andoxidative stress, which lead to steatohepatitis andcirrhosis [1, 5, 9, 13]. An important role of residentmacrophages of liver, Kupffer cells, as well as oxida-tive stress manifested either by increased reactiveoxygen species (ROS) formation or extensive lipidperoxidation have also been suggested [5, 9, 14, 15].Both of them appear to arise from mitochondrialdysfunction, as a result of increased supply in freefatty acids, due to insulin resistance [10, 13]. How-ever, besides the ‘two-hit’ model, a ‘multiple parallelhits’ hypothesis is becoming increasingly importantnowadays, suggesting that NAFLD and NASH arisenot necessarily as a result of a sequence of activitiesof different factors, but rather due to their multipleand parallel action [9, 11]. Regardless of which ofthose theories (or other potential explanations of theorigin and development of NAFLD and NASH) isclosest to truth, an increased accumulation of trigly-cerides and other lipids within the hepatic tissue inthe form of lipid droplets (LDs) represent an un-questionable hallmark of fatty liver [1, 5, 9, 11, 15].Their important role in the development of NAFLDdisease, as well as metabolic syndrome, was repeat-edly emphasized [16–18]. However, LDs structure,function, biogenesis, composition and significance in

NAFLD are far from being understood [18] andnew approaches are needed, that would enable aninsight into their biology and role.

There are numbers of animal models of NAFLDand NASH, including genetically modified mice (e.g.db/db mice), nutritional (wild type mice on specificdiets such as methionine and choline deficient or highfat diet) and mixed models (genetically modified ani-mals on specific diet) [1, 13, 21, 22]. Currently, the mostcommonly used models include the dietary models:High Fat Diet (HFD) and Methionine Choline Defi-cient (MCD) diet [1, 10, 13, 22, 23]. These models re-flect pathological disorders to the greatest extended,but not imitate perfectly all changes [13]. The term“High Fat Diet” has been applied for various diets, of-ten differing strongly in detailed fatty acid composition,but always retaining, as a common feature, a high per-centage of total fat. However, in a comparative study ofHFDswith different fatty acid composition, lardwas re-commended as the source of fat, in order to generate amodel resembling metabolic changes associated withobesity [23]. A similar effect can be also achieved byHigh Fructose Diet [22, 24]. A strong impact of dietarycholesterol has also been reported, suggesting that evena small addition of cholesterol to the diet can inducesteatohepatitis [12, 22, 25, 26].

In general, the model of fatty liver induced by highfat diet is characterized by a high percentage of caloriesfrom fat (e.g. 60%) and reflects more closely the detri-mental effects of dietary habits responsible for in-creased morbidity due to obesity and its complicationsin well-developed societies [13]. It leads not only toobesity but also to hyperlipidemia and insulin resist-ance, inducing eventually steatosis similar to that ob-served in humans. The first alterations appear alreadyafter about 2 months of feeding [10]. Impairment ofliver regeneration was observed after partial hepatect-omy in mice fed with high fat diet. Fatty livers weremore susceptible to posthepatectomy damage and fail-ure [27]. Usefulness of this model has been demon-strated in numerous studies, mostly by histochemicalstaining (the current ‘gold standard’ for liver tissue stu-dies) [22, 23].

Raman spectroscopy (RS) is well-known for itsnumerous advantages. The method is particularlyuseful for examination of biological and medical ma-terials, including tissues and organs. Non-destruc-tiveness, lack of complicated sample preparationalong with the possibility of obtaining a largeamount of information, makes it a tool of choice forsuch studies [12, 28–32]. RS is used to investigatethe disease development and to test the efficacy ofpotential medications. Hepatic tissue [33–39] andcells [19, 38] were studied by RS to research for can-cer and hepatitis in different animal models.

Ramanmapping gives a possibility of chemical char-acterization and examination of the distribution of ma-jor components of the tissue, especially lipids (choles-

K. Kochan et al.: Raman spectroscopy analysis of lipid droplets in NAFLD2

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.biophotonics-journal.org

terol, cholesterol esters, triglycerides, etc.) [40]. In li-pid-focused studies, RS enables the determination ofthe total degree of unsaturation, using a variety of ap-proaches, e.g. calculating the ratio of the intensities ofthe bands at 1266 cm–1 and 1300 cm–1 or at 1656 cm–1

and 1444 cm–1. In particular, the second mentioned cri-terion was used to reliably predict the iodine values oftriglycerides [20, 33, 42]. Moreover, Raman spectro-scopy provides information about cis and trans isomersin the investigated samples [20, 41, 42]. The close con-nection between liver lipid profile and hepatic insulinsensitivity [7] combined with excellent suitability of RSfor the study of lipids opens a wide range of new possi-bilities for a better understanding of the developmentof disorders related tometabolic syndrome.

This work was focused on the analysis of chemi-cal alterations of the liver tissue related to NAFLD,manifested by an increased accumulation of trigly-cerides in the form of lipid droplets. This phenomen-on is emerging from the earliest stages of the disease(representing the ‘first hit’ according to the ‘two-hit’model, as discussed above). Raman mapping is anexcellent tool to study lipids by acquiring informa-tion directly from lipid droplets formed in the liver.

2. Experimental

Two mouse models, representing macrovesicularand microvesicular steatosis, were investigated. Inthe first case a dietary model was used, whereas the

second was a combined model of genetic and dietarymodification. For both models the effects of treat-ment with selected reference drugs were also exam-ined, according to details provided below.

Macrovesicular steatosismodel:C57BL/6Jmalemicewere divided into three groups, each fed for 15 weeks ac-cording to the scheme: (1)AIN-93Gdiet (control, a stan-dard early grow and reproduction diet), (2) High FatDiet (HFD), containing 60% of kcal from fat and (3)HFD (10 weeks) followed by the HFD and treatmentwith angiotensin converting enzyme inhibitor – ACE-I,perindopril (additional 5 weeks).The relatively long timeof 15 weeks of feeding withHFDdiet (group 2) was cho-sen to ensure the development of advanced steatosis.

Microvesicular steatosis model: ApoE/LDLr–/–(mouse model of atherosclerosis) mice were dividedinto three groups and fed for 8 weeks with (1) AIN-93G diet (control), (2) Low Carbohydrate High Pro-tein (LCHP) diet and (3) LCHP diet (4 weeks) fol-lowed by the LCHP diet and nicotinic acid (NA)treatment (additional 4 weeks).

For both models each group contained 3 indivi-duals. Compositions of the experimental diets usedin both models (macrovesicular and microvesicularsteatosis) are given in Table 1.

2.1 Sample preparation

Livers were frozen in Optimal Cutting Temperature(OCT) formulation and cut into sections with thick-

Table 1 Composition of the experimental diets: detailed content of dietary components and % content of particulargroups (proteins, carbohydrate, fat).

Ingredients Macrovesicular model Microvesicular model

HFD** [g] Control* [g] LCHP*** [g]

Casein 200 200 524L-Cystine 3 3 0Corn Starch 0 397, 486 50Maltodextrin 125 132 0Sucrose 67 100 118, 486Cellulose BW 200 50 50 50SoybeanOil 25 70 0t-Butylhydroquinone 0.014 0.014 0.014Lard 245 0 0Butter 0 0 210Mineral Mix AIN-93G 35 35 35Vitamin Mix AIN-93G 10 10 10CholineBitartrate 2.5 2.5 2.5

Calculated values [kcal%]Protein 18 18 49Carbohydrate 22 66 16Fat 60 16 35TOTAL 100 100 100

* AIN-93G; ** High Fat Diet (HFD); *** Low Carbohydrate High Protein (LCHP) diet

J. Biophotonics (2014) 3

© 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheimwww.biophotonics-journal.org

ness of 10 μm. Tissue specimens were subsequentlyplaced on CaF2 slides and fixed with 4% bufferedformalin solution for 10 min. Remainders of the fixa-tive were washed out with distilled water (2 × 5 min).From each mouse liver 6–8 sections were prepared(approximately 20 sections from 3 mice in each nutri-tion group; approx. 60 sections for each model).

2.2 Histochemical staining

Histochemical staining was performed with the useof Oil Red O (ORO), a standard diazo dye for neu-tral lipids and triglycerides. The ORO staining wasdone according to the following steps: deionizedwater (3 min), 60% isopropanol alcohol (2 min), OilRed O (30 min) and deionized water (1 min). Afterthe application of these solutions for a certainamount of time, the slides (tissue) were covered withglycerol gel.

2.3 Raman mapping

Raman mapping was done with a WITec ConfocalRaman Microscope (WITec alpha300 R, Ulm, Ger-many) equipped with an air cooled solid state laserworking at 532 nm, 600 grooves/mm grating and aCCD detector (spectral resolution 3 cm–1). The scat-tered light was directed to the spectrometer via anoptical fiber with a diameter of 50 μm. From eachprepared liver section 2–4 maps were collected,using a dry Olympus MPLAN (100×/0.90NA) objec-tive, in the range 0–3705 cm–1, with the integrationtime 0.3–0.5 s. Laser power was adjusted individuallyfor each sample, not exceeding the range of 6.5–11.5 mW. Differences in laser power are caused byvarious tissue compositions (mainly lipid content)and were applied to limit a destructive effect of thelaser on the tissue. Raman maps were collected forboth, smaller and larger areas, depending on themodel (and thus the size of steatosis); however, theydid not exceed the size of 100 × 100 μm. The spectrawere collected every 0.36 μm.

2.4 Data analysis

For preprocessing and data analysis Witec ProjectPlus Software, Bruker Optics Software and The Un-scrambler were used, respectively. Preprocessing in-cluded cosmic spike removal and background sub-traction (polynomial fit, order 2). Cluster Analysiswas performed with the use of k – means algorithm,Ward’s algorithm and Euclidean measure of dis-

tances in the rage 600–1800 cm–1. For the calculationof the degree of lipid unsaturation for each nutritiongroup in both models several hundreds of spectrawere included (macrovesicular steatosis model: con-trol – 265, HFD – 395, HFD + perindopril – 368; mi-crovesicular steatosis model: control – 340, LCHP –

353, LCHP + NA – 270).Within the same nutritiongroup spectra were derived from maps obtained fordifferent sections and different individuals. The dif-ference in the number of spectra analyzed for eachgroup is associated with the size of lipid droplets. Alladdressed spectra were normalized and band areaswere integrated in the same selected spectral ranges[33]. Identical procedures of data preprocessing andanalysis were applied for standards, used to obtainthe calibration line of the iodine value. Unsuper-vised hierarchical cluster analysis (UHCA) and line-ar discriminant analysis (LDA) were performed onpreprocessed spectra, in the range 600–1800 cm–1.

3. Results and discussion

3.1 Histochemical staining of liver sectionsfrom micro- and macrovesicular micesteatosis model

There was a clear-cut difference in the extent andpattern of steatosis in C57BL/6J mice fed HFD andApoE/LDLR–/– mice fed LCHP diet, as evidencedby Oil Red O staining (Figure 1).

The substantial steatotic changes were observedin the liver of C57BL/6J mice fed High Fat Diet. Itis evident that HFD resulted not only in a massiveincrease in lipid content, but also in the presence oflarge lipid droplets within the liver section (Figure 1,upper panel), representative for macrovesicular stea-tosis. Interestingly, the liver tissue from mice treatedwith perindopril did not manifest large droplets,although smaller ones still could be noticed, as wellas clearly increased lipid content in comparison tocontrol. Moreover, the result for perindopril treat-ment indicates the possibility of uneven distributionof lipid droplets in the tissue. It may be related tothe vascularity system based on the blood distribu-tion of portal and hepatic veins. Altogether perindo-pril treatment partially prevented HFD-inducedsteatosis.

The linkage between cardiovascular diseases (in-cluding atherosclerosis) and fatty liver is well-known,although not completely clarified [4]. As shown inFigure 1, liver from ApoE/LDLR–/– mice fed LCHPdiet displayed modestly increased lipid content, thatcould cofirm the connection between cardiovasculardiseases and steatosis. However in contrast, to HFDfed C57BL/6J mice, a microvesicular pattern of stea-



tosis was rather observed. Furthermore, nicotinicacid treatment seemed to completely reverse LCHP– induced steatotic changes.

3.2 Raman mapping of macrovesicularsteatosis in C57BL/6J mice fed withHigh Fat Diet (HFD). Effects of perindopril

3.2.1 The presence and heterogeneityof lipid droplets in the liver

The identification of lipid droplets in the liver sectionfrom C57BL/6J mice with the use of Raman spectro-scopy, depending upon the diet and treatment, isshown in Figure 2. Figure 2 (middle) distinctly indi-cates the presence of large lipid droplets (reaching toseveral tens of micrometers in diameter) in liver sam-ples from mice fed with HFD, confirming the validityof this model as an inducer of hepatic steatosis. LDscould be easily seen even on a microphotograph andtheir spectra demonstrate clearly a lipid profile, de-void of signals from any other chemical components.For the High Fat Diet model this is a common fea-ture, i.e. the samples exhibit high amounts of lipiddroplets, which are most often large as seen also inFigure 1. Furthermore, all observed droplets werecomposed of both, saturated (bands at 1304 cm–1,1446 cm–1 and 2855 cm–1) and unsaturated (bands at1268 cm–1, 1659 cm–1 and 3011 cm–1) lipids. In con-trast, the control diet samples did not show anyanomalous accumulation of lipids, whereas othercomponents of tissue are clearly visible, such as pro-teins (i.e. amide I band at 1658 cm–1) or heme (bandsat 755 cm–1, 1130 cm–1 and 1592 cm–1). Furthermore,

in the case of healthy tissue (control) the presence ofheme is particularly visible as it has a strong impacton in the spectra by the presence of the resonance,visible in the range 600–1700 cm–1 (Figure 2, control).In few samples, small areas with higher lipid contentwere noticed, with a diameter around 1 μm. They canbe characterized by a lack of distinct boundaries andthe presence of other tissue components, as it can beexpected for small accumulations. Moreover, in somecases they were associated with the presence of vita-min A, which indicates that these are rather the He-patic Stellate Cells (storage of vitamin A) than anyabnormal lipid accumulation [19]. In rare cases, how-ever, they did not show the presence of vitamin A,but as in the case of HFD, were composed of satu-rated and unsaturated lipids, although in a differentratio than in LDs from the HFD group (Figure 3).Such spectra were used for further investigations ofLDs composition.

In the case of the group treated with perindopril,lipid droplets were still noticeable, but clearly smal-ler than in the HFD group (Figure 2, right panel).Spectral profiles of the lipid droplets remain expli-citly lipidic; differences in the composition are ana-lyzed below. For a detailed examination of the com-position, particularly in terms of diversity of lipiddistribution within single droplets as well as betweenthem (in the same group), cluster analysis (CA) wasapplied, with a special attention given to areas ofmap corresponding to droplets. Representative re-sults for sample from the HFD group are shown inFigure 4.

As expected, huge differences can be seen be-tween spectra corresponding to tissue and dropletsregions, due to their completely different biochem-ical composition. Within a single map, LDs were al-ways classified in the same class (Figure 4 left,

Figure 1 Results of histochemicalstaining with Oil Red O obtainedfor: macrovesicular steatosis model(upper panel) – C57BL/6J mice fedfor 15 weeks with control diet(AIN-93G), High Fat Diet (HFD)or HFD + perindopril; microvesi-cular steatosis model (bottom pa-nel) – ApoE/LDLr–/– mice fed for8 weeks with control diet (AIN-93G), Low Carbohydrate HighProtein (LCHP) diet or LCHP +nicotinic acid (NA). Magnificationof 200×.



green), regardless of their diameter. Furthermore,detailed analysis demonstrated no differences be-tween their compositions: extracted internal-LDsclasses are present in both droplets and have similarspatial distribution (Figure 4, right panel). More-over, the average spectra of classes corresponding todroplets, obtained from various maps of livers fromdifferent mice of the same dietary group, were alsosimilar. Therefore, the lipid composition of dropletsis the same within the group. In addition, with re-spect to a single droplet, no characteristic distribu-tion of the lipids was observed (Figure 4, right pa-nel). Slight differences between classes result fromthe decreasing intensity of lipid bands and increasingsignals from other components when approachingthe boundary of LD, what is a consequence of therounded shape of a droplet. However, in terms ofthe chemical composition, individual droplets tendto be homogeneous. Figure 4 represents the results

obtained for HFD, however these observations (si-milarity of the composition of droplets within thesame nutrition group as well as their homogeneity)remains true for all studied groups.

3.2.2 Degree of lipid unsaturation in LDsfrom C57BL/6J mice fed HFD left untreatedor treated with perindopril

In order to specify lipid characteristics of droplets inmore details, degree of unsaturation was deter-mined. For this purpose ratios of intensity of differ-ent bands can be used, but based on our previousstudy the most useful and reliable indicator for thedegree of unsaturation in liver tissue is the ratio ofthe bands at 1659 cm–1 and 1446 cm–1 [20, 33]. The

Figure 2 Distribution of saturated(1420–1480 cm–1) lipids obtainedfor C57BL/6J mice fed for15 weeks with control diet (left),High Fat Diet (middle) and HighFat Diet treated with perindopril(right) along with spectra fromhigh-lipid region. Each presentedmap (and corresponding spectra)originates from different mice.Scale bar for the maps of distribu-tion is shown in the top right cor-ner.



bands in the high range (2800–3200 cm–1) are notsensitive enough (significant difference in areas ofpeaks at 3013 cm–1 and 2855 cm–1), while the bandsfrom the range 1200–1300 can be strongly affectedby proteins, especially heme resonance, particularlyin the case of control. Presented here results (Table2) were calculated for each mouse for thousands ofspectra extracted from different maps obtained fromvarious specimens. The number of spectra taken for

the control is smaller, because of the rare occurrenceof droplets in that group as well as their very smallsize. Due to the latter, in LDs from control samplesa much greater influence of protein signals (e.g.amide I) should be taken into consideration. Foreach group exactly the same preprocessing steps andranges of area integration were used.

As can be seen, fatty liver (induced by HFD) canbe characterized not only by the occurrence of nu-merous large lipid droplets, but also by their compo-sition. In comparison to the areas of higher lipidcontent in control samples (IV = 127 ± 8), LDs in li-ver of mice fed with HFD (IV = 96 ± 6), are built toa greater extent by saturated lipids. Such composi-tion may have an impact on the stability of formeddroplets. Moreover, the relatively small standard de-viations within each individual, as well as betweenthem (in the same group), additionally confirms thehomogeneity of droplet composition. For LDs inliver of mice fed with HFD and treated with peri-ndopril the IV = 121 ± 5 is found. It seems that theaction of perindopril was to change the compositionof lipid droplet by increasing the content of unsatu-rated relative to the saturated lipids.

Determined by this approach degree of lipidunsaturation corresponds to the ratio of n(C=C)/n(CH2). However, through the calibration for appro-priate standards, it is possible to associate it with theiodine value (IV), which is a commonly used indica-tor of the degree of unsaturation. Based on the cali-bration, the IV for HFD model is close to 90 (96 ±6), which corresponds to the value for oleic acid.

Figure 3 An example of small lipid accumulation occuringin liver tissue of C57BL/6J mice fed for 15 weeks with con-trol diet: a distribution of unsaturated lipids ν(C=C) basedon integration in the range 1645–1675 cm–1 and a distribu-tion of proteins based on integration of amide III (1200–1260 cm–1), along with spectrum from lipid-rich region withmarked lipid bands. Lack of signals from vitamin A ex-cludes the possibility that shown accumulations are HepaticStellate Cells (HSC). Scale bar for the maps of distributionis shown in Figure 2.

Figure 4 Results of cluster analysis(CA, k – means) in the range 600–1800 cm–1 for the map presented atFigure 2, middle panel: analysis ofthe whole tissue (left) and lipiddroplet area (right). The colours ofthe spectra corresponds to coloursof the clasess.



Comparison of the standard spectrum (oleic acid)with the spectrum of LDs for the HFD in therange 1400–1800 cm–1 (Figure 5) confirmed the si-milarity of the degree of unsaturation and there-fore – the correctness of determined iodine values,as the ratio of intensities I1656/I1444 is in fact verysimilar for them. However, it should be stressedthat the composition of the LDs is complex andthey consist also of various triglycerides and cho-lesteryl esters. Although, it may be useful, particu-larly when comparing samples from differentgroups, to evaluate the general trend of changes inlipid saturation.

In this case, the difference in the degree of lipidunsaturation is the highest between the controlgroup (IV = 127 ± 8) and the HFD group (IV = 96± 6). Particularly interesting was the result obtainedfor the group treated with perindopril (IV = 121 ±5). As mentioned previously, the size of lipid dro-plets and their occurrence was clearly reduced dueto the use of perindopril(in some cases leading evento their complete disappearance). In addition to thisvisible change, drug administration led to a signifi-cant increase of their iodine value, making it veryclose to that of control. Such modification of thecomposition of LDs may be of importance, as itcould lead to their destabilization and consequentlyto the total disintegration of droplets.

Raman mapping enabled a detailed examinationof single lipid droplets present in macrovesicularsteatosis model. In this case fatty liver was evidentand therefore results of Raman studies were unam-biguous. However, the sizes of areas mapped withthe use of Raman spectroscopy are small in compar-ison to the size of tissue specimens. Therefore, espe-cially in the cases of uneven distribution of lipid dro-plets, global evaluation of steatosis based only onRaman mapping is exposed to risks associated withthe selective choice of the studied area.

Nevertheless, the confocal Raman microspectro-scopy can be an excellent tool to assess the effective-ness of potential drugs. In addition to the resultspresented above, two chemometric approaches wereused: unsupervised hierarchical cluster analysis(UHCA) and linear discriminant analysis (LDA).UHCA was performed in the range 600–1800 cm–1

on spectra from LDs from the control, HFD andHFD + perindopril groups. It enabled for separationof the pathologically distorted samples (HFD group)from the others (control and HFD + perindopril),however it did not provide a good distinction be-tween control and treated samples (Supplementarymaterials, Figure 1). This is in agreement with pre-viously presented results, since the differences be-tween the degrees of unsaturation between thosetwo groups are small and are in the range of stand-

Table 2 Degree of unsaturation calculated from Raman spectra corresponding to the ratio n(C=C)/n(CH2) assessedfor lipid-rich areas in control samples and lipid droplets in High Fat Diet and High Fat Diet treated with perindoprilsamples along with corresponding Iodine Value and standard deviation values.

GroupIndividual

Control HFD HFD + Perindopril

Value SD* Value SD* Value SD*

1 0.91 0.04 0.69 0.03 0.88 0.042 0.94 0.05 0.70 0.04 0.86 0.013 0.89 0.04 0.69 0.02 0.86 0.03Average 0.91 0.02** 0.69 0.005** 0.87 0.01**

Iodine Value 127 8 96 6 121 5

* standard deviation calculated for each specimen; ** SD value for average calculated as deviation between individuals.

Figure 5 Spectra of: oleic acid(18 : 1) standard (black) and lipiddroplet (red) from the HFD group,in the range 600–1800 cm–1 (left)and magnified region 1400–1800 cm–1 (right), including twoprominent bands at 1656 cm–1 and1444 cm–1, used for asassment ofthe degree of lipid unsaturation.The Iodine Values of oleic acid: 90,lipid droplet: 96 ± 6.



ard deviation. Moreover these results show thatalthough the distinction between LDs in the normaltissue and the one after treatment was not achievedon the basis of the LDs composition, the confocalRS has the potential to assess the efficacy of treat-ment.

In addition to UHCA, a simple approach basedon the LDA analysis was used (Supplementary ma-terials, Figure 2). On the basis of spectra from con-trol and HFD group we have trained a model ofclassification between those two groups and subse-quently applied it to classify spectra of the tissuefrom mice on HFD after treatment with perindopril.100% of these spectra were classified within the con-trol group. It means that treatment with perindoprilresults in changing the composition of droplets to-wards the ones found in the tissue of control mice.In addition, this simple approach shows how confo-cal Raman microspectroscopy can be easily appliedto assess the effectiveness of drugs.

At this point it should be emphasized, that lipiddroplets emerging during NAFLD are only one ofthe key–spots associated with the pathogenesis ofthis disease. The other indicators of great impor-tance are Kupffer cells, resident macrophages occur-ring between the endothelial cells in the wall of sinu-soids [7, 14, 15]. Apart from them, abnormalities dueto NAFLD (and thus NASH) may be localized alsowithin other cells, i.e. hepatic stellate cells (HSCs)[19]. In healthy, not disrupted tissue, the HSCs serveas a storage of vitamin A, which can be detectedwith the use of a sufficiently sensitive and specifictechnique (e.g. Raman spectroscopy) [20]. However,any pathological state in the tissue causes an im-mediate HSC activation and the loss of vitamin A.Therefore, this particular change in biochemicalcomposition of hepatic tissue (lack of signals fromvitamin A in the HSCs) may indicate alarming dis-turbances [19, 20]. Furthermore, disorders can existalso on the subcellular level, an example of whichcan be changes in the mitochondria of hepatocytes[1, 5]. For full understanding of the NAFLD patho-genesis, all mentioned aspects should be examined.

3.3 Raman mapping of microvesicularsteatosis in ApoE/LDLr–/– mice fed withLow Carbohydrate High Protein (LCHP)diet. Effects of nicotinic acid (NA)

3.3.1 The presence and heterogeneityof lipid droplets in the liver

In the above discussed fatty liver model huge lipiddroplets (tens of μm in diameter) were already visi-ble under the microscope and occurred in large

numbers. The model demonstrated in this part is dif-ferent, as lipid droplets are not so easily seen in thetissue as shown in Figure 1. Primary, the model isbased on a combination of genetic modification(ApoE/LDLr–/– mice) and dietary effects (Low Car-bohydrate High Protein diet). Apolipoprotein E andlow density lipoprotein receptor double-knockoutmice are commonly used as a model of atherosclero-tic changes. Since atherosclerosis is one of the altera-tions belonging to the group of cardiovascular dis-eases (CVD), which often coexist with(or even arepreceded by) the metabolic syndrome, disturbancesin the liver tissue may be expected. Coexistence andthe linkage between disorders of the liver profile(with particular respect to lipid content) and athero-sclerotic lesions were already suggested [4]. On theother hand, the Low Carbohydrate High Protein(LCHP) diet was proven to strongly promote the de-velopment of atherosclerosis [43]. Thus, the possibi-lity of its impact on the development of disorders inthe liver is highly justified. However, the connectionbetween NAFLD and CVDs and the use of com-bined (genetic and dietary) model makes it challen-ging to separate the effects caused by diet and genet-ic modification, resulting in the development ofatherosclerosis. Nicotinic acid (NA), on the otherhand, is the oldest and one of commonly used drugs,besides statins, in the treatment of atherosclerosis. Itacts by inhibiting the synthesis of VLDL (and reduc-ing its secretion)in the liver and, at the same time,inhibiting the release of fatty acids from adipose tis-sues. Altogether, NA treatment in humans’ resultsin lowering of total cholesterol, LDL cholesterol andtriglyceride concentration, and an increase in HDLcholesterol, and not all of these effects are repro-duced in mice. Nevertheless, as the liver is an impor-tant target for NA action in humans and mice, it canbe expected that NA have an influence on the devel-opment of steatosis. This assumption was confirmedby our Raman measurements (Figure 6).

Figure 6 clearly indicates the existence of fatty livereven in the case of control diet in the atheroscleroticmouse model. For the control mice fed with a controldiet, without genetically induced atherosclerotic le-sions, lipid droplets were extremely rare and verysmall (refer to Figure 2, left). This confirms thecoexistence of changes in the liver and blood vessels,as a result of the development of atherosclerosis(without any dietary influence). However, within thismodel, the lipid droplets in the control group were byfar smaller than those observed for mice on LCHPdiet. LCHP diet promoted the development of steato-sis through the impact on both: size and number of li-pid droplets. This was, however, not a dramatic influ-ence, since the diameter of the LDs (although in-creased more than twice) was still definitely smallerthan for the previous discussed model (HFD). Never-theless, this effect is not negligible, especially taking



into account a significant increase in density of LDswithin the tissue. Therefore, a combination of LCHPdiet and the genetic modification inducing athero-sclerosis resulted in a fatty liver characterized by ahigh density of relatively small droplets. This microve-sicular steatosis was reduced due to nicotinic acidtreatment. Drug administration seemed to not only in-hibit further fatty liver development, but also minima-lize already existing disorders (lipid droplets). For theLCHP +NA group, LDs occurred less frequently eventhan for the control and were smaller (i.e. 2.75 μm and4.3 μm for LCHP + NA and control, respectively).Thus, nicotinic acid acts even on small droplets, re-gardless of whether their formation is connected withatherosclerotic lesions or results from diet.

3.3.2 Degree of lipid unsaturation in LDsfrom ApoE/LDLR mice fed LCHP leftuntreated or treated with nicotinic acid

In order to verify the existence of different (or simi-lar) biochemical composition of droplets, an analysis

of their spectral lipid profile was performed usingcluster analysis (CA) (data not shown). As above,also for this model the areas corresponding to LDswere clearly separated in the liver tissue. Homoge-neity of the chemical composition both for a singledroplet as well as between droplets within the samemodel of nutrition/treatment was noticed. Represen-tative spectra from LDs of all groups are shown atFigure 7.

The spectra (Figure 7A) obtained from all micegroups show signals coming from lipids and other tis-sue components. Particularly strong bands appearfrom the heme, the resonance of which partiallyoverlaps with lipid signals in the range 1250–1310 cm–1. The least contribution of lipid signals isobserved in the spectra collected from droplets in li-vers of mice fed on LCHP diet and treated with NA(Figure 7A, green), due to the smallest diameter ofLDs. The droplets in different nutrition group areprobably composed of various lipids. In the fewspectra from LDs in livers of mice fed with LCHP,bands assigned to cholesterol and its esters can beseen at 418 cm–1, 606 cm–1 and 701 cm–1 (Figure 7B)[44]. Their intensity is not high, probably due to low

Figure 6 Results of integrationanalysis in ranges: 1645–1675 cm–1

(ν(C=C), lipids) for mapped areasof liver tissue obtained fromApoE/LDLr–/– mice on: control(AIN-93G), Low CarbohydrateHigh Protein (LCHP) and LowCarbohydrate High Protein + Ni-cotinic Acid (LCHP + NA) diet,along with sizes of mapped areasgiven below. Each presented maporiginates from different mice. In-tensity scale bar for the maps ofdistribution is shown in Figure 2.



content of these specific lipids. Moreover, the spec-tral range of their occurrence coincides with theresonance of heme groups, which partly overlapsthem (this applies in particular to the band at701 cm–1). However, these bands are absent in thespectra of droplets from other groups (Figure 7B,green and blue line). It is also possible that thecholesterol and its esters are also present in dro-plets from other groups, but their Raman signalsare not seen in the spectra due to overlapping withheme resonance or to the very low content ofthese lipids. However, the occurrence of thesebands only in the spectra of LCHP group provesthat cholesterol and its esters are present either ex-clusively or in a clearly higher content in thatgroup, possibly as a result of a diet.

Furthermore, there is a change in the degree oflipid unsaturation (ratio of bands at 1659 cm–1 to1445 cm–1) between liver samples obtained from var-ious mice groups (Figure 7A). As in the previousmodel, the analysis was performed for hundreds ofspectra for each group (Table 3).

The biggest difference can be seen between thecontrol and LCHP group. LCHP diet affected thecomposition of lipid droplets, which was expressedby the significant decrease in the degree of lipid un-saturation (from 0.78 to 0.53, respectively for controland LCHP). The development of fatty liver was as-sociated here not only with an increase in the sizeand number of droplets, but also with saturationprocess of lipids. The change of the degree of lipidunsaturation in microvesicular model compared toHFD model was also relatively large, although notaccompanied by equally large change in the sizes ofdroplets. It can be assumed that this is not a conse-quence of increasing the droplet size, but rather thatthis change in LDs composition appears regardlessof the change in their size and may be an indepen-dent marker of a situation favorable for the develop-ment of steatosis. Surprisingly, however, althoughthe degree of unsaturation was slightly higher in thetreated group treated with NA (0.60) than in themodel group of microvesicular steatosis (LCHP)(0.53), both values are close and much lower than

Figure 7 Representative spectraof liver tissue obtained fromApoE/LDLr–/– mice on: Control(AIN-93G) (blue), Low Carbohy-drate High Protein (LCHP) (red)and Low Carbohydrate High Pro-tein + Nicotinic Acid (LCHP +NA) (green) diet in the ranges:(A) 600–1800 cm–1, with markedlipid and heme bands, (B) 400–800 cm–1 with marked cholesteroland cholesterol esters bands.

Table 3 Degree of unsaturation calculated from Raman spectra corresponding to the ratio n(C=C)/n(CH2) assessedfor lipid-rich areas in liver samples from mice fed with: control diet, Low Carbohydrate High Protein (LCHP) dietand Low Carbohydrate High Protein diet treated with nicotinic acid (LCHP + NA) along with corresponding stan-dard deviation values.

GroupIndividual

Control LCHP LCHP + NA

Value SD* Value SD* Value SD*

1 0.77 0.06 0.53 0.07 0.61 0.082 0.81 0.04 0.54 0.07 0.62 0.023 0.76 0.05 0.53 0.06 0.58 0.03Average 0.78 0.02** 0.53 0.005** 0.60 0.02**

* standard deviation calculated for each specimen; ** SD value for average calculated as deviation between individuals.



for the for the control group (0.78). The action ofnicotinic acid thus led to a change in lipid composi-tion, but only to a small extent. It should howeverbe emphasized, that the group subjected to treat-ment, was the one where lipid droplets occurred theleast frequently. Furthermore, they were the smal-lest, hence the number of spectra taken into accountin the calculation of lipid saturation was relativelylow. As can be seen in Figure 7A, these spectra havethe least lipidic character and therefore the othercomponents of the tissue (heme, proteins, etc.) mayhave a greater impact on the LDs composition. Sincenone of the spectra had a purely lipidic character(non-lipid components contribution was noticeablein every group), the iodine values (IV) were notdetermined for this model. This parameter can bedetermined only based on the spectra dominated bylipid signals (like in the HFD model). In this context,however, the visible participation of non-lipid compo-nents of the tissue in the spectra makes the use of re-lative intensity of the bands for determination of theIV incorrect. The closeness of signals arising fromother components (e.g. amide I and lipid ν(C=C))makes their influence on the band area highly prob-able. Thus, the ratio of bands I1659/I1445 was not per-fectly suitable to determine an absolute measure ofthe degree of unsaturation, but can be used to com-pare the unsaturation between groups, as the partici-pation of non-lipidic signals is comparable.

4. Conclusions

In this work we show confocal Raman microspectro-scopy as a powerful tool for detail examination oflipid content, saturation and distribution in the liveralong with NAFLD progression. We demonstratedthat C57BL/6J mice fed on HF diet and ApoE/LDLr–/– mice fed on LCHP diet represent a macro-and microvesicular steatosis model, respectively.These two models differ in various aspects, includingpathogenesis, the size of lipid droplets and their che-mical composition. HFD led to formation of large li-pid droplets, whereas combination of genetic modifi-cation (apolipoprotein E and LDL receptor doubleknockout mice) and LCHP diet led to the formationof small lipid droplets, but in high number. How-ever, in both models, a low degree of lipid unsatura-tion within LDs was noted suggesting that this phe-nomenon is not related to the droplet size but rathermay be mechanistically linked to the steatosis devel-opment. Interestingly, angiotensin converting en-zyme inhibitor, perindopril as well as nicotinic acid,reversed macrovesicular and microvesicular steato-sis, respectively, and these effects were associatedwith the increased degree of unsaturation of the li-pids in LDs. We speculate that the degree of un-

saturation within the LDs may be an important mar-ker of the disease progression, independent on thedroplet size. Altogether, our work demonstrated thatRaman spectroscopy-based approach provided agood insight into the LDs structure and compositionin mice models of NAFLD and enable us to charac-terize LDs content response to the pharmacologicaltreatment. However, it should be also noted that Ra-man images probe relatively small regions of inter-est. In order to fully assess the presence and the de-gree of severity of fatty liver the results have to besupported by examination of large region of interestby other methods, e.g. histochemical staining orFTIR imaging.

Acknowledgements This work was supported by theEuropean Union under the European Regional Develop-ment Fund (grant coordinated by JCET-UJ,POIG.01.01.02-00-069/09) and by National Science Center(grant DEC-2013/09/N/NZ7/00626). KK acknowledges theMarian Smoluchowski Krakow Research Consortium:“Matter Energy Future” (granted the KNOW status forthe 2012–2017 by the Ministry of Science and Higher Edu-cation) scholarship.

Author biographies Please see Supporting Informationonline.

References

[1] Y. Takahashiet, Y. Soejima, and T. Fukusato, WorldJ. Gastroenterol. 18(19), 2300–2308 (2012).

[2] A. Zivkovic, J. B. German, and A. J. Sanyal, Am. J.Clin. Nutr. 86, 285–300 (2007).

[3] S. Moscatiello, R. Manini, and G. Marchesini, Nutr.Metab. Cardiovasc. Dis. 17, 63–70 (2007).

[4] G. Targher, L. Bertolini, R. Padovani, S. Rodella, G.Arcado, and Ch. Day, J. Hepatol. 46, 1126–1132(2007).

[5] E. Vanni, E. Bugianesi, A. Kotronen, S. De Minicis,H. Yki-Järvinen, and G. Svegliati-Baroni, Dig. Liver.Dis. 42, 320–330 (2010).

[6] E. Björnsson and P. Angulo, Scand. J. Gastroentero.42, 1023–1030 (2007).

[7] T. Matsuzaka, H. Shimano, N. Yahagi, T. Kato, A. At-sumi, T. Yamamoto, N. Inoue, M. Ishikawa, S. Okada,N. Ishigaki, H. Iwasaki, Y. Iwasaki, T. Karasawa, S.Kumadaki, T. Matsui, M. Sekiya, K. Ohashi, A. H.Hasty, Y. Nakagawa, A. Takahashi, H. Suzuki, S. Ya-toh, H. Sone, H. Toyoshima, J. Osuga, and N. Yama-da, Nat. Med. 13(10), 1193–1202 (2007).

[8] M. Pasarin, J. G. Abraldes, A. Rodrigez-Vilarrupla,V. La Mura, J. C. Gàrcia-Pagàn, and J. Bosch, J. He-patol. 55, 1095–1102 (2011).

[9] S. K. Mantena, A. L. King, K. K. Andringa, H. B.Eccleston, and S. M. Bailey, Free Radic. Biol. Med.44, 1259–1272 (2008).



[10] H. B. Eccleston, K. K. Andringa, A. M. Betancourt,A. L. King, S. K. Mantena, T. M. Swain, H. N. Tinsley,R. N. Nolte, T. R. Nagy, G. A. Abrams, and S. M. Bai-ley, Antioxid. Redox Signal. 15(2), 447–459 (2011).

[11] H. Tilg and R. Moschen, Hepatology 52(5), 1836–1846(2010).

[12] K. Wouters, P. J. van Gorp, V. Bieghs, M. J. Gijbels,H. Duimel, D. Lütjohann, A. Kerksiek, R. van Kruch-ten, N. Maeda, B. Staels, M. van Bilsen, R. Shiri-Sverdlov, and M. H. Hofker, Hepatology 48(2), 474–486 (2008).

[13] G. Kanuri and I. Bergheim, Int. J. Mol. Sci. 14, 11963–11980 (2013).

[14] G. Kolios, V. Valatas, and E. Kouroumalis, World J.Gastroenterol. 12, 7413–7420 (2006).

[15] V. Bieghs, S. Walenbergh, T. Hendrikx, P. J. vonGorp, F. Verheyen, S. W. Olde Damink, A. A. Mas-clee, G. H. Koek, M. H. Hofker, Ch. J. Binder, and R.Shiri-Sverdlov, Liver Int. 33(7), 1056–1061 (2013).

[16] K. Reue, J. Lipid Res. 52, 1865–1868 (2011).[17] M. Suzuki, Y. Shinohara, Y. Ohsaki, and T. Fujimoto,

J. Electron Microsc. 60(1), S101–S116 (2011).[18] A. Greenberg, R. A. Coleman, F. B. Kraemer, J. L.

McManaman, M. S. Obin, V. Puri, Q. W. Yan, H.Miyoshi, and D. G. Mashek, J. Clin. Invest. 121(6),2102–2110 (2011).

[19] N. Testerink, M. Ajat, M. Houweling, J. F. Brouwers,V. V. Pully, H. van Manen, C. Otto, J. B. Helms, andA. B. Vaandrager, PLoS ONE 7(4), e34945 (2012).

[20] K. Kochan, K. M. Marzec, K. Chruszcz-Lipska, A.Jasztal, E. Maslak, H. Musiolik, S. Chlopicki, and M.Baranska, Analyst 38, 3885–3890 (2013).

[21] J. G. Fan and L. Qiao, Hepatobiliary Pancreat. Dis.Int. 8(3), 233–240 (2009).

[22] A. Nakamura and Y. Terauchi, Int. J. Mol. Sci. 14(11),21240–21257 (2013).

[23] R. Buettner, K. G. Parhofer, M. Woenckhaus, C. E.Wrede, L. A. Kunz-Schughart, J. Schölmerich, and L.C. Bollheimer, J. Mol. Endocrinol. 36, 485–501 (2006).

[24] L. Tappy and K. A. Le, Clin. Res. Hepatol and Gas-troenterol. 36, 554–560 (2011).

[25] T. M. Comhair, S. C. Garcia Caraballo, C. H. C. De-jong, W. H. Lamers, and S. E. Köhler, Nutr. Metab. 8,4 (2011).

[26] J. R. Clapper, M. D. Hendricks, G. Gu, C. Wittmer,C. S. Dolman, J. Herich, J. Athanacio, C. Villescaz, S.S. Ghosh, J. S. Heilig, C. Lowe, and J. D. Roth, Am. J.Physiol. Gastrointest. Liver Physiol. 305, G483–G495(2013).

[27] R. A. DeAngelis, M. M. Markiewski, R. Taub, and J.D. Lambris, Hepatology 42(5), 1148–1157 (2005).

[28] M. Diem, A. Mazur, K. Lenau, J. Schubert, B. Bird,M. Miljkovic, C. Krafft, and J. Pöpp, J. Biophotonics6(11–12), 855–886 (2013).

[29] C. Krafft, N. Bergner, B. Romeike, R. Reichart, R.Kalff, K. Geiger, M. Kirsch, G. Schackert, and J. Pöpp,Proc. SPIE 8207, Photonic Therapeutics and Diagnos-tics VIII 82074I (2012).

[30] B. R. Wood, L. Hammer, L. Davis, and D. McNaugh-ton, J. Biomed. Opt. 10(1), 14005 (2005).

[31] F. Bonnier, S. M. Ali, P. Knief, H. Lambkin, K. Flynn,V. McDonagh, C. Healy, T. C. Lee, F. M. Lyng, andH. J. Byrne, Vib. Spectrosc. 61, 124–132 (2012).

[32] K. Majzner, A. Kaczor, N. Kachamakova-Trojanow-ska, A. Fedorowicz, S. Chlopicki, and M. Baranska,Analyst. 138, 603–610 (2013).

[33] K. Kochan, E. Maslak, R. Kostogrys, S. Chlopicki, andM. Baranska, Biomed. Spectrosc. Imaging. 2, 331–337(2013).

[34] C. Krafft, M. A. Diderhoshan, P. Recknagel, M. Milj-kovic, M. Bauer, and J. Pöpp, Vib. Spectrosc. 55, 90–100 (2011).

[35] A. Shen, B. Zhang, J. Ping, W. Xie, P. Donfack, S.-J.Baek, X. Zhou, H. Wang, A. Materny, and J. Hu, J.Raman Spectrosc. 40, 550–555 (2009).

[36] P. L. Palaniappan and K. S. Pramod, Vib. Spectrosc.56, 146–153 (2011).

[37] S. Zhang, X. Li, T. Yang, Z. Xu, and L. Jin, J. Chem.Phar. Res. 5(11), 45–48 (2013).

[38] S. R. Hawi, W. B. Campbell, A. Kajdacsy-Balla, R.Murphy, F. Adar, and K. Nithipatikom, Cancer Lett.110, 35–40 (1996).

[39] S. Sivakumar, C. P. Khatiwad, J. Sivasubramanian,and B. Raja, Spectrochim Acta A Mol Biomol Spec-trosc. 118, 46–469 (2014).

[40] T. P. Wrobel, K. Malek, S. Chlopicki, L. Mateuszuk,and M. Baranska, Analyst. 136, 5247–5255 (2011).

[41] H. Schulz and M. Baranska, Vib. Spectr. 43, 13–25(2007).

[42] O. Samek, A. Jonáš, Z. Pilát, P. Zemánek, L. Nedbal,J. Tříska, P. Kotas, and M. Trtílek, Sensors 10, 8635–8651 (2010).

[43] R. B. Kostogrys, M. Franczyk-Żarów, E. Maślak, M.Gajda, L. Mateuszuk, C. L. Jackson, and S. Chłopicki,Atherosclerosis 223, 327–331 (2012).

[44] C. Krafft, L. Neudert, T. Simat, and R. Salzer, Spec-trochim Acta A Mol. Biomol. Spectrosc. 61(7), 1529–1535 (2005).

Supporting Information

Additional supporting information may be found inthe online version of this article at the publisher’s web-site.



Raman spectroscopy analysis of lipid droplets content, distribution and saturation level in...

Documents

Transcript of Raman spectroscopy analysis of lipid droplets content, distribution and saturation level in...