**2. Raman mapping for cell imaging**

Several Raman peaks are used as fingerprints for the intracellular identification of nucleic acids, proteins and lipid‐rich structures. In terms of chemical information, the richest part in a Raman spectrum is the region below 1800 cm−1, also called the fingerprint region. Some representative Raman bands from the fingerprint region, characteristic for nucleic acids, proteins and lipids, are given in **Table 1** [6, 37, 38]. The region between 1800 and 2800 cm−1 is the so‐called silent region, since no Raman cellular vibrations arising from functional groups appear in this region, excepting triple‐bond vibrations. Finally, the high‐frequency region above 2800 cm−1 is dominated by C–H stretching vibrations (CH, CH2, CH3).

Stem cells are attractive to be studied in the biomedical field because they have the ability to differentiate into any cell phenotype, and they can proliferate indefinitely [39]. Differentiation of stem cells could be used in stem cell therapy and tissue engineering. The differentiation process of stem cells needs identification of specific markers; currently, this involves the use of immunolabeling or fluorescence. Both methods have the disadvantage of being invasive. In contrast, using Raman‐specific intrinsic fingerprints would be advantageous, being noninva‐ sive, nonlabeling approach, and able to provide accurate and highly specific information. The method would be suitable also for cells that lack specific markers for separation by conven‐ tional means [21] such as cardiomyocytes.

The intracellular distribution and concentration of nucleic acids were used by Ghita et al. [6] to distinguish between undifferentiated and differentiated stem cells, more precisely from undifferentiated neural stem cells and the glial cells derived from them. They manage to differentiate with 89.4% sensitivity and 96.4% specificity. The spectral fingerprint of the nu‐ cleic acid backbone was used to detect DNA‐ and RNA‐rich regions. The Raman spectrum of β‐DNA shows a strong band at 788 cm−1 and a shoulder at 835 cm−1, characteristic to the symmetric O‐P‐O stretching vibrations of the phosphate groups and to the asymmetric O‐P‐ O vibrations, respectively. In case of RNA, the symmetric O‐P‐O vibrations appear shifted to 813 cm−1, while the shoulder corresponding to the asymmetric O‐P‐O stretching disappears. Based on the spectral differences between DNA and RNA, it was found that undifferentiat‐ ed neural cells have higher concentration of nucleic acids compared to glial cells. The Ram‐


**Table 1.** Some of the most representative Raman peaks for nucleic acids, proteins and lipids in cells.

an band at 813 cm−1 was the indicator of the differentiation status, and this allowed distinguishing between the two cell types. Undifferentiated neural stem cells have high con‐ centration of RNA in the cytoplasm (as high as 4 mg/ml), while below the instrument detec‐ tion limit (<1 mg/ml) in the glial cells. Based on Raman mapping, it was possible to image RNA‐ and DNA‐rich structures in the stem cells. The RNA‐rich structures, representing the stem cells cytoplasm, were imaged using the 813 cm−1 Raman band, and the DNA‐rich part related to the stem cells nucleus was imaged using the 788 cm−1 Raman band. Similar results were reported by the same research group regarding the differentiation status of embryonic stem cells [22]. The differentiated cells had 75% less RNA, as monitored by the decrease in intensity of the 813 cm−1 peak. Basically, most prominent Raman peaks of embryonic stem cells are the ones of proteins (amide I band at 1660 cm−1, amide III at 1200–1300 cm−1, 1005  cm−1 vibration of phenylalanine and C=C stretching at 938 cm−1) and nucleic acids. Dental follicle mesenchymal stem cells were also imaged using Raman mapping; several compo‐ nents, and especially a high concentration of cytoplasmic RNA, were found to be a good indicator to the undifferentiated status of the cells [40].

Raman mapping was also used inside a bioreactor culture system, where human embryonic stem cells were grown and differentiated into cardiomyocytes [41]. The purpose was to monitor the cardiac differentiation of the embryonic bodies. The Raman maps were compared with immunofluorescence imaging. A positive correlation was found between Raman bands at 1340, 1083, 937, 858, 577 and 482 cm−1 and the expression of the α‐actinin protein in the differentiated cardiomyocytes. Konorov et al. [26] obtained information on the cell cycle phase of human embryonic stem cells. The 783 cm−1 DNA band from a large number of cell nuclei was used as indicator of the cell cycle phase. The results were corrected for the RNA contribution at 811  cm−1. As such, the authors were able to get information on the state of division of the embryonic stem cells by quantifying the DNA and RNA peaks from the Raman spectra and obtained Raman intensities similar to the fluorescence intensities of flow cytometry.

In another study, Pascut et al. [42] obtained 97% specificity and 96% sensitivity in differenti‐ ating the cardiomyocytes derived from human embryonic stem cells. The main spectral features that allowed the discrimination of cardiomyocytes were attributed to glycogen and myofibrils. The results were correlated with the immunofluorescence staining, and a good correlation was observed. The same authors investigated the potential for developing Raman‐ activated cell sorting of individual cells [43]. Hashimoto et al. [44] got information on osteoblast differentiation and mineralization mechanisms by monitoring fluctuations in the cytochrome C concentration. The above preliminary studies suggest that Raman spectroscopy has a great potential to become a leading method for stem cells investigation.

Raman mapping can be used as a tool to obtain molecular fingerprint information from different subcellular compartments. Based on their distinct chemical features, nucleus and cytoplasm and also other cellular organelles can be imaged. For example, in their study on follicle mesenchymal stem cells, Leopold et al. [40] were able to image the cell nucleus based on the 785 cm−1 band characteristic for the DNA O‐P‐O vibrations. Lipid characteristic peaks, such as the 1446 cm−1 peak characteristic to CH2 vibrations, made it possible to highlight the smooth endoplasmic reticulum in the Raman images, which is known to be the source of intracellular lipid synthesis. Based on characteristic Raman vibrational peaks of lipids, proteins and nucleic acids, Krafft et al. [45] were able to reconstruct the main cellular components: the nucleus, the contour of the cell and the organelles. They focused mostly on the 2800–3000 cm −1 region, where CH2 and CH3 vibrations from proteins, lipids and nucleic acids are present. In both studies, the cellular organelle identification was carried out based on the score plots obtained from the principal component analysis. From the score plots, false color Raman maps were generated, highlighting the subcellular compartments. Raman images of subcellular organelles were also reported by Krauß et al. [23]. The authors have also correlated well their results with fluorescence microscopy.

an band at 813 cm−1 was the indicator of the differentiation status, and this allowed distinguishing between the two cell types. Undifferentiated neural stem cells have high con‐ centration of RNA in the cytoplasm (as high as 4 mg/ml), while below the instrument detec‐ tion limit (<1 mg/ml) in the glial cells. Based on Raman mapping, it was possible to image RNA‐ and DNA‐rich structures in the stem cells. The RNA‐rich structures, representing the stem cells cytoplasm, were imaged using the 813 cm−1 Raman band, and the DNA‐rich part related to the stem cells nucleus was imaged using the 788 cm−1 Raman band. Similar results were reported by the same research group regarding the differentiation status of embryonic stem cells [22]. The differentiated cells had 75% less RNA, as monitored by the decrease in intensity of the 813 cm−1 peak. Basically, most prominent Raman peaks of embryonic stem cells are the ones of proteins (amide I band at 1660 cm−1, amide III at 1200–1300 cm−1, 1005  cm−1 vibration of phenylalanine and C=C stretching at 938 cm−1) and nucleic acids. Dental follicle mesenchymal stem cells were also imaged using Raman mapping; several compo‐

**Raman peak position (cm−1) Assignments**

64 Raman Spectroscopy and Applications

 Guanine ring breathing Adenine ring breathing Thymine ring breathing Uracil ring breathing Cytosine ring breathing

1095–1098 O‐P‐O vibrations (DNA, RNA)

938 Backbone C=C stretching

1655–1662 Amide I (C=O stretching)

 =C–H bending C–C stretching =C–H bending C–H deformation C=C stretching

**Nucleic acid bands**

788 Symmetric O‐P‐O stretching in β‐strands DNA 835 Asymmetric O‐P‐O stretching in β‐strands DNA

**Protein bands**

1005 Phenylalanine symmetric ring breathing 1033 Phenylalanine in plane C–H vibrations 1200–1300 Amide III band (CH, NH deformations)

**Lipid bands**

**Table 1.** Some of the most representative Raman peaks for nucleic acids, proteins and lipids in cells.

813 Symmetric O‐P‐O stretching in RNA and α‐strands DNA

Cytochrome C, protein and lipid‐rich structures were evidenced in different Raman images on HeLa cells [3] by irradiating the cells with 488, 514, 532 and 633 nm lasers. The HeLa cells gave Raman spectra with peaks at 1000 cm−1 (breathing of phenylalanine), 1451, 2850, 2885, 2935 cm−1 (CH2 deformation, CH2 and CH3 stretching) and 1660 cm−1 (amide I bands mode of peptide bonds). When irradiated with the 532 nm laser, resonant peaks at 753, 1127, 1314 and 1583, characteristic to cytochrome C, were obtained. Raman images highlighting the Cyto‐ chrome C, protein β‐sheets and lipids were created using the 753 cm−1 peak (pyrrole ring of cytochrome C), the 1686 cm−1 peak (amide I vibration of peptide bond in protein) and the 2852 cm−1 peak (CH2 stretching vibration of hydrocarbon chain of lipids). Since cytochrome C is essential for the electron transfer in mitochondria, the Raman images of cytochrome C should also point out the distribution of mitochondria in the cell. In addition, Raman spectra from the nuclei showed no resonance, being similar for all excitation wavelengths. Matthäus et al. [46] reported on obtaining Raman maps pointing out the location of different cellular structures in HeLa cells, with emphasis on mitochondria. They performed hierarchical cluster analysis and found localization of mitochondria in the perinuclear region, which was supported by correlation with fluorescence maps.

In some cases, Raman imaging requires the use of tags. This happens when molecules cannot be detected by regular Raman scattering, either because they are in very low amounts, or because their Raman signal overlaps with other compounds and cannot be distinguished clearly. There are two approaches for using tags for Raman imaging: (a) using surface‐ enhanced Raman scattering, which implies binding of tags onto the plasmonic nanoparticles surface that can be further used for intracellular identification of analytes [47–50], and (b) taking advantage of the silent region in the Raman spectra of cells between 1800 and 2800 cm −1. In this region, most of the biologically active molecules show no Raman vibrations, so functional molecules with Raman fingerprint in this region could be useful as tags for detection of molecules, which do not give clearly distinguishable Raman peaks in the intracellular medium. This also has the advantage of avoiding the overlap with any endogeneous molecules. Tags suitable for the silent region detection are alkynes, azides, deuterium and nitrile. Palonpon et al. [10] utilized alkyne‐tagged EdU (5‐ethynil‐2‐deoxyuridine), for the detection of DNA accumulation and synthesis in cells. EdU is rapidly incorporated in the DNA during the replication process, accumulates in the nucleus and is thus suitable for acquiring informa‐ tion about DNA synthesis in cells.

Silver nanoparticles (AgNPs) prepared by reduction with hydroxylamine according to the Leopold and Lendl method [51] were used for mapping sub‐membrane hemoglobin in erythrocytes (red blood cells) [52]. Erythrocytes contain cytosolic and sub‐membrane hemo‐ globin. Although hemoglobin exhibits strong Raman scattering, Raman spectra of erythrocytes generally have mostly contribution from cytosolic hemoglobin, since the sub‐membrane hemoglobin is negligible in amount. To trace this sub‐membrane hemoglobin, the authors used SERS‐active AgNPs that were internalized in the cells and accumulated in the cell membrane. SERS images showed the erythrocytes that come in contact with the AgNPs. Lee et al. [53] used SERS to detect different human breast cancer cell lines phenotypes and to quantify the proteins on the cell surface. For the purpose of SERS enhancement, silica‐encapsulated hollow gold nanospheres conjugated with specific antibodies were used. The expression of epidermal growth factor (EGF), ErbB2 and insulin‐like growth factor‐1 (IGF‐1) receptors was determined in the MDA‐MB‐468, KPL4 and SK‐BR‐3 cell lines by SERS mapping. Different distributions of growth factors were clearly identified and distinguished from their corresponding SERS mapping images. Taking advantage on the characteristic wave number of the carbonyl group vibration that lies within the silent region of the Raman spectra, Kong et al. [54] developed osmium carbonyl clusters for cancer cell imaging. The clusters were conjugated with PEG‐ coated AuNPs and further functionalized with antibody for epidermal growth factor receptors (anti‐EGFR). AuNPs were used for carbon monoxide Raman signal enhancement, while functionalization with antibody was needed since the EGFR is highly expressed in many cancer cell lines. Both EGFR‐positive and EGFR‐negative cancer cells were used. The nanoparticle conjugates were imaged after cellular uptake based on the CO absorption signal at 2030 cm−1, and the results showed the specificity and efficient targeting of CO‐nanoparticle conjugates to EGFR‐positive cells.

peptide bonds). When irradiated with the 532 nm laser, resonant peaks at 753, 1127, 1314 and 1583, characteristic to cytochrome C, were obtained. Raman images highlighting the Cyto‐ chrome C, protein β‐sheets and lipids were created using the 753 cm−1 peak (pyrrole ring of cytochrome C), the 1686 cm−1 peak (amide I vibration of peptide bond in protein) and the 2852 cm−1 peak (CH2 stretching vibration of hydrocarbon chain of lipids). Since cytochrome C is essential for the electron transfer in mitochondria, the Raman images of cytochrome C should also point out the distribution of mitochondria in the cell. In addition, Raman spectra from the nuclei showed no resonance, being similar for all excitation wavelengths. Matthäus et al. [46] reported on obtaining Raman maps pointing out the location of different cellular structures in HeLa cells, with emphasis on mitochondria. They performed hierarchical cluster analysis and found localization of mitochondria in the perinuclear region, which was supported by

In some cases, Raman imaging requires the use of tags. This happens when molecules cannot be detected by regular Raman scattering, either because they are in very low amounts, or because their Raman signal overlaps with other compounds and cannot be distinguished clearly. There are two approaches for using tags for Raman imaging: (a) using surface‐ enhanced Raman scattering, which implies binding of tags onto the plasmonic nanoparticles surface that can be further used for intracellular identification of analytes [47–50], and (b) taking advantage of the silent region in the Raman spectra of cells between 1800 and 2800 cm −1. In this region, most of the biologically active molecules show no Raman vibrations, so functional molecules with Raman fingerprint in this region could be useful as tags for detection of molecules, which do not give clearly distinguishable Raman peaks in the intracellular medium. This also has the advantage of avoiding the overlap with any endogeneous molecules. Tags suitable for the silent region detection are alkynes, azides, deuterium and nitrile. Palonpon et al. [10] utilized alkyne‐tagged EdU (5‐ethynil‐2‐deoxyuridine), for the detection of DNA accumulation and synthesis in cells. EdU is rapidly incorporated in the DNA during the replication process, accumulates in the nucleus and is thus suitable for acquiring informa‐

Silver nanoparticles (AgNPs) prepared by reduction with hydroxylamine according to the Leopold and Lendl method [51] were used for mapping sub‐membrane hemoglobin in erythrocytes (red blood cells) [52]. Erythrocytes contain cytosolic and sub‐membrane hemo‐ globin. Although hemoglobin exhibits strong Raman scattering, Raman spectra of erythrocytes generally have mostly contribution from cytosolic hemoglobin, since the sub‐membrane hemoglobin is negligible in amount. To trace this sub‐membrane hemoglobin, the authors used SERS‐active AgNPs that were internalized in the cells and accumulated in the cell membrane. SERS images showed the erythrocytes that come in contact with the AgNPs. Lee et al. [53] used SERS to detect different human breast cancer cell lines phenotypes and to quantify the proteins on the cell surface. For the purpose of SERS enhancement, silica‐encapsulated hollow gold nanospheres conjugated with specific antibodies were used. The expression of epidermal growth factor (EGF), ErbB2 and insulin‐like growth factor‐1 (IGF‐1) receptors was determined in the MDA‐MB‐468, KPL4 and SK‐BR‐3 cell lines by SERS mapping. Different distributions of growth factors were clearly identified and distinguished from their corresponding SERS

correlation with fluorescence maps.

66 Raman Spectroscopy and Applications

tion about DNA synthesis in cells.

For toxicology studies, it is important to be able to distinguish between healthy and apoptotic cells and also to gain information on the molecular changes associated with apoptosis. Zoladek et al. [28] used Raman imaging for understanding changes associated with apoptosis in the MDA‐MB‐231 human breast cancer cells. Cells were exposed to the apoptotic drug etoposide, and Raman spectra were recorded 2, 4 and 6 h after exposure. An 1.5‐fold increase in the DNA content was observed after 6 h, and the change was assigned to DNA condensation. The most drastic change was in the lipid profile; a high concentration in membrane phospholipids and unsaturated non‐membrane lipids was observed in apoptotic cells. The Raman images of the lipidic areas of the cells were generated based on the 1005 and 1659 cm−1 peaks ratio. The 1005 cm−1 peak with contribution from phenylalanine is not affected by etoposide exposure, while the 1659 cm−1 C=C stretching vibrations from lipids show strong increase upon etoposide exposure, indicating an increased degree of unsaturation of lipids for the apoptotic cells. Okada et al. [55] used resonant Raman scattering for imaging the intracellular distribution of cyto‐ chrome C and observing dynamic changes of its 750 cm−1 band associated with cell apoptosis.

The cellular uptake of nanoparticles and drugs and the cellular responses to drugs are important aspects to be investigated for molecular biomedical applications. Cellular uptake and localization of polyethylene glycol‐coated gold nanoparticles in human prostate cancer cells (LNCaP Pro 5) were visualized based on their photoluminescence peak (180–1800 cm−1). In the Raman images of LNCaP Pro 5 cells with AuNPs internalized, cell nucleus and nucleoli are visible, as well as spots generated from the photoluminescence peak of AuNPs. The nanoparticles are located at different positions inside the cells, depending on the time elapsed from exposure. Two hours after exposure, the AuNPs are located in the cell membrane, 12 h after they are located in the cytoplasm, and after 24 h, AuNPs are imaged in the perinuclear region [29].
