**3. Raman mapping for tissue imaging and medical diagnosis**

There is clear indication that Raman spectroscopy could provide insights into drug targeting mechanisms and could be used for detection of metabolic interactions of drugs with cancer cells. In their attempt to detect physiologically relevant cellular responses to drugs, El‐ Mashtoly et al. [9] used Raman imaging to quantify the effect of the epidermal growth factor inhibitor panitumumab on colon cancer cells expressing Kirsten‐ras mutations (oncogenic and wild‐type). It is known that oncogenic K‐ras mutations block the response to anti‐epidermal growth factor therapy such as panitumumab, while cells expressing wild‐type mutations respond to the treatment; all these facts were nicely confirmed by the Raman mapping results. The authors used hierarchical cluster analysis on the 700–1800 cm−1 and 2800–3050 cm−1 regions for the identification of subcellular components such as cellular membrane, cytoplasm, nucleus and lipid droplets. They found that the oncogenic mutated K‐ras cells showed no response to the drug, while the wild‐type mutated cells have strong cellular responses to panitumumab treatment, as demonstrated by Raman intensity changes and wave number shifts. The panitumumab‐induced changes are strongest on the lipid droplets, suggesting that lipid droplets might play a crucial role in anticancer therapy. The results were confirmed by fluorescence spectroscopy. In another work, El‐Mashtoly et al. [30] were able to image the spatial distribution of the erlotinib, another inhibitor of the epidermal growth factor receptor, in colon adenocarcinoma cells upon 12 h of incubating the cells with 100 μM erlotinib solution. Normally, erlotinib cannot be detected by Raman spectroscopy at 100 μM concentrations, which basically implies that its intracellular level was higher due to most likely concentration of the drug in the cell. The authors used the C=C alkyne vibration from the silent region of the Raman spectrum (2085–2140 cm−1) to image the erlotinib distribution and found that the drug was mostly concentrated at the cell borders.

Cells, tissues and bio‐fluids can be imaged by Raman micro‐spectroscopy. Based on the hypothesis that molecular changes associated with different diseases can be quantified by Raman spectroscopy, the method has been used in medical research and diagnosis during the last years. On one hand*, in vitro* and *in vivo* analysis of tissue is important to be able to distinguish between healthy and tumor cells, and on the other hand, in the medical diagnostics field, there is imperative need for research directed toward identifying noninvasive methods for tumor analysis and toward determining the exact tumor margins. There are several papers reporting the use of Raman spectroscopy and imaging in these directions. Present research still requires comparison with conventionally used staining methods used in histopathology. The gold standard method for tumor pathology and classification is the hematoxylin and eosin (H&E) staining. As opposed to the H&E which involves tissue staining and fixation, Raman micro‐spectroscopy is a nondestructive, nonlabeling method. Also, histopathology cannot be used intraoperatively as it requires long incubation times. NIR and the visible 532 nm laser are reported for tissue imaging related to cancer research [7, 30].

The major drawback of Raman technique that limits its application in the medical field is the low efficiency of the inelastic scattering process. Different strategies have been developed to overcome this difficulty, based on, for example: (a) using nonlinear imaging modes such as CARS; (b) acquiring selective sampling of the analyzed probe (e.g., tissue auto‐fluorescence can be used to determine the characteristics of the tissue sample and to further use the information to prioritize the sampling points for Raman spectroscopy); (c) multimodal integration of Raman with other techniques such as auto‐fluorescence; (d) use of fiber‐optic probes for hand held instruments; and (e) use of plasmonic metallic nanoparticles suitable for surface‐enhanced Raman scattering.

*In vivo* and *in vitro* cancer diagnosis based on Raman imaging was so far focused on brain, breast, lung, skin, prostate, colorectal, esophagus and bone cancer [33]. The group of Notingher et al. are pioneers in using Raman imaging for tumor diagnosis, in particular for detecting tumor margins. Multimodal spectral imaging, combining auto‐fluorescence imaging and Raman micro‐spectroscopy, was used [34, 35, 56] to distinguish between healthy and cancer cells in different carcinoma tissues during intraoperative or postoperative evaluations, for the purpose of accurately detecting the tumor margins. Multimodal spectral imaging is required to reduce the acquisition times needed for raster scanning. Instead of raster scanning the sample, selective sampling is achieved based on integrating collagen auto‐fluorescence imaging with Raman imaging. First auto‐fluorescence images are used to determine the features of the tissue, and then, the information is used to prioritize and decide the sampling points for Raman spectroscopy. The tissue areas with auto‐fluorescence are those containing collagen and are thus identified as healthy dermis and excluded from the Raman measure‐ ments. In this way, a dramatic decrease in the acquisition time is achieved: autofluorescence Raman typically requires ~100‐fold fewer Raman spectra compared to raster scanning [35]. The high speed of fluorescence imaging relies on the capability to image large tissue area, in contrast to Raman imaging, which requires a pixel‐by‐pixel readout. For example, an inte‐ grated system based on Raman scattering and auto‐fluorescence imaging was used by Kong et al. [35] to diagnose basal cell carcinoma tumor margins during tissue‐conserving surgery. The major challenge in tissue‐conserving surgery is to completely remove the tumor, with minimal loss of healthy tissue. Auto‐fluorescence images were necessary in order to prioritize the sampling points for Raman. By using k‐means cluster analysis and comparing the images obtained from clustering with the histopathology images, it was possible to diagnose the tumor with 100% sensitivity and 92% specificity. As such, it was possible to assign tissue areas corresponding to the tumor, epidermis, dermis, fat, inflamed dermis, sebaceous gland and muscles. The tumor areas show more intense DNA peaks at 788 and 1098 cm−1 compared to healthy tissue. The spectra of the dermis were characterized by collagen‐specific peaks at 851 and 950 cm−1. It was possible to achieve shorter diagnosis times than those required by histopathology.

inhibitor panitumumab on colon cancer cells expressing Kirsten‐ras mutations (oncogenic and wild‐type). It is known that oncogenic K‐ras mutations block the response to anti‐epidermal growth factor therapy such as panitumumab, while cells expressing wild‐type mutations respond to the treatment; all these facts were nicely confirmed by the Raman mapping results. The authors used hierarchical cluster analysis on the 700–1800 cm−1 and 2800–3050 cm−1 regions for the identification of subcellular components such as cellular membrane, cytoplasm, nucleus and lipid droplets. They found that the oncogenic mutated K‐ras cells showed no response to the drug, while the wild‐type mutated cells have strong cellular responses to panitumumab treatment, as demonstrated by Raman intensity changes and wave number shifts. The panitumumab‐induced changes are strongest on the lipid droplets, suggesting that lipid droplets might play a crucial role in anticancer therapy. The results were confirmed by fluorescence spectroscopy. In another work, El‐Mashtoly et al. [30] were able to image the spatial distribution of the erlotinib, another inhibitor of the epidermal growth factor receptor, in colon adenocarcinoma cells upon 12 h of incubating the cells with 100 μM erlotinib solution. Normally, erlotinib cannot be detected by Raman spectroscopy at 100 μM concentrations, which basically implies that its intracellular level was higher due to most likely concentration of the drug in the cell. The authors used the C=C alkyne vibration from the silent region of the Raman spectrum (2085–2140 cm−1) to image the erlotinib distribution and found that the drug

Cells, tissues and bio‐fluids can be imaged by Raman micro‐spectroscopy. Based on the hypothesis that molecular changes associated with different diseases can be quantified by Raman spectroscopy, the method has been used in medical research and diagnosis during the last years. On one hand*, in vitro* and *in vivo* analysis of tissue is important to be able to distinguish between healthy and tumor cells, and on the other hand, in the medical diagnostics field, there is imperative need for research directed toward identifying noninvasive methods for tumor analysis and toward determining the exact tumor margins. There are several papers reporting the use of Raman spectroscopy and imaging in these directions. Present research still requires comparison with conventionally used staining methods used in histopathology. The gold standard method for tumor pathology and classification is the hematoxylin and eosin (H&E) staining. As opposed to the H&E which involves tissue staining and fixation, Raman micro‐spectroscopy is a nondestructive, nonlabeling method. Also, histopathology cannot be used intraoperatively as it requires long incubation times. NIR and the visible 532 nm laser are

The major drawback of Raman technique that limits its application in the medical field is the low efficiency of the inelastic scattering process. Different strategies have been developed to overcome this difficulty, based on, for example: (a) using nonlinear imaging modes such as CARS; (b) acquiring selective sampling of the analyzed probe (e.g., tissue auto‐fluorescence can be used to determine the characteristics of the tissue sample and to further use the information to prioritize the sampling points for Raman spectroscopy); (c) multimodal integration of Raman with other techniques such as auto‐fluorescence; (d) use of fiber‐optic probes for hand held instruments; and (e) use of plasmonic metallic nanoparticles suitable for

was mostly concentrated at the cell borders.

68 Raman Spectroscopy and Applications

reported for tissue imaging related to cancer research [7, 30].

surface‐enhanced Raman scattering.

Selective sampling for intraoperative diagnosis during the breast cancer conserving surgery leads to a diagnosis of mammary ductal carcinoma with 95.6% sensitivity and 96.2% specificity [56]. As in the study above, discrimination between healthy and tumor areas was based on increased concentrations of nucleic acids (bands at 788, 1098 cm−1) and decreased levels of collagen and fats (851 and 950 cm−1 bands) in the tumor regions. Tissues from 60 patients were deposited on MgF2 plates; 20 μm tissue sections were sampled and analyzed by Raman micro‐ spectroscopy, and adjacent sections of 70 μm were stained with H&E. To reduce acquisition times needed for raster scanning, selective sampling was achieved based on integrated auto‐ fluorescence imaging and Raman. This procedure is also known as multimodal spectral histopathology. By comparing the Raman images obtained from the k‐means cluster analysis with the ones obtained from H&E staining, the tumor *vs* healthy breast tissue assignment was successfully carried out. The images from the tumor regions showed large number of cells with enlarged nuclei. Compared to regular raster scanning Raman that would require 10,000  spectra/mm2 and 5 h analysis time, the multimodal spectral imaging drastically reduces the analysis time by reducing the number of Raman spectra acquired to 20 spectra/mm2 , which needs 17 min for reaching diagnosis. In a recent study [7], the same group reported on face and neck basal cell carcinoma analysis by selective sampling Raman with 95.3% sensitivity and 94.6% specificity. The results are promising; the method can significantly decrease the diagnosis time. However, it requires strong computing power for the calculations needed after measurement of each Raman spectrum, and this can still be considered a drawback.

Cancer and pre‐cancer cells, erythrocytes and lymphocytes were successfully assigned to colon cancer tissue sections by combining Raman imaging with histopathology (H&E staining) and with immunohistochemistry [13]. Hierarchical cluster analysis was used in the spectral region 700–1800 cm−1 and 2600–3100 cm−1. The tumor protein p53 is normally highly expressed in cancer and pre‐cancer cells because it is a tumor suppresser. The possibility of obtaining Raman imaging of tumor and pre‐tumor cells by highlighting p53 active areas was confirmed. By comparing the obtained false color Raman maps with the images given by the anti‐p53 immunohistochemical stained image, it was found that the sample auto‐fluorescence matches the fluorescence from the anti‐p53 stained tissue, proving that the Raman imaging can be used for assigning the p53 active areas of the tissues. The p53 active areas represent more specifically the cancer cell nuclei.

Using SERS‐active nanoparticles for intraoperative detection of tumor margins is another promising direction of research in Raman imaging. With this purpose in mind, Wang et al*.* [48] developed multi‐receptor‐targeted SERS‐active nanoparticles that are topically applied at the surface of tissues excised during breast cancer lumpectomy and that enable quantita‐ tive molecular phenotyping at the tumor surface for the purpose of diagnosis. The nanopar‐ ticles are tagged with multiple antibodies to achieve as high accuracy as possible and to be able to eliminate influence of nonspecific binding of the nanoparticles. Bovin serum albumin (BSA) was also used to limit nonspecific accumulation of nanoparticles within cells. Anti‐ bodies for the epidermal growth factor receptor (EGFR) or the human epidermal growth factor receptor 2 (HER2) and a negative control antibody were conjugated to the nanoparti‐ cle surface; a fluorophore was also used to conduct flow cytometry for result confirmation. By targeting the SERS‐active nanoparticles to various tumor biomarkers simultaneously and recording the SERS spectra, followed by computational demultiplexing to determine the rel‐ ative concentrations of the individual SERS nanoparticles, it is possible to detect residual tumors at the surgical margins. The results of the study demonstrated the ability to perform successful Raman imaging on the tissues and to accurately quantify relative tumor biomark‐ er expression levels (high levels of HER2 expression were found, characteristic for breast tumors), in less than 15 min.

## **4. Raman mapping in plant and algae research**

There is a growing interest in getting a more comprehensive understanding of the chemical composition of various plant tissues. Investigations on structural aspects of plant cell wall components, on the chemistry of plant metabolites and relevant plant molecules, are feasible using Raman mapping. NIR‐FT Raman is suitable for imaging of large plant structures such as leaves, seeds and fruits, while the higher resolution visible lasers allow investigation of smaller plant structures.

analysis time by reducing the number of Raman spectra acquired to 20 spectra/mm2

measurement of each Raman spectrum, and this can still be considered a drawback.

the cancer cell nuclei.

70 Raman Spectroscopy and Applications

tumors), in less than 15 min.

**4. Raman mapping in plant and algae research**

needs 17 min for reaching diagnosis. In a recent study [7], the same group reported on face and neck basal cell carcinoma analysis by selective sampling Raman with 95.3% sensitivity and 94.6% specificity. The results are promising; the method can significantly decrease the diagnosis time. However, it requires strong computing power for the calculations needed after

Cancer and pre‐cancer cells, erythrocytes and lymphocytes were successfully assigned to colon cancer tissue sections by combining Raman imaging with histopathology (H&E staining) and with immunohistochemistry [13]. Hierarchical cluster analysis was used in the spectral region 700–1800 cm−1 and 2600–3100 cm−1. The tumor protein p53 is normally highly expressed in cancer and pre‐cancer cells because it is a tumor suppresser. The possibility of obtaining Raman imaging of tumor and pre‐tumor cells by highlighting p53 active areas was confirmed. By comparing the obtained false color Raman maps with the images given by the anti‐p53 immunohistochemical stained image, it was found that the sample auto‐fluorescence matches the fluorescence from the anti‐p53 stained tissue, proving that the Raman imaging can be used for assigning the p53 active areas of the tissues. The p53 active areas represent more specifically

Using SERS‐active nanoparticles for intraoperative detection of tumor margins is another promising direction of research in Raman imaging. With this purpose in mind, Wang et al*.* [48] developed multi‐receptor‐targeted SERS‐active nanoparticles that are topically applied at the surface of tissues excised during breast cancer lumpectomy and that enable quantita‐ tive molecular phenotyping at the tumor surface for the purpose of diagnosis. The nanopar‐ ticles are tagged with multiple antibodies to achieve as high accuracy as possible and to be able to eliminate influence of nonspecific binding of the nanoparticles. Bovin serum albumin (BSA) was also used to limit nonspecific accumulation of nanoparticles within cells. Anti‐ bodies for the epidermal growth factor receptor (EGFR) or the human epidermal growth factor receptor 2 (HER2) and a negative control antibody were conjugated to the nanoparti‐ cle surface; a fluorophore was also used to conduct flow cytometry for result confirmation. By targeting the SERS‐active nanoparticles to various tumor biomarkers simultaneously and recording the SERS spectra, followed by computational demultiplexing to determine the rel‐ ative concentrations of the individual SERS nanoparticles, it is possible to detect residual tumors at the surgical margins. The results of the study demonstrated the ability to perform successful Raman imaging on the tissues and to accurately quantify relative tumor biomark‐ er expression levels (high levels of HER2 expression were found, characteristic for breast

There is a growing interest in getting a more comprehensive understanding of the chemical composition of various plant tissues. Investigations on structural aspects of plant cell wall components, on the chemistry of plant metabolites and relevant plant molecules, are feasible

, which

For example, using the 633 nm laser, it was possible to image the distribution of cell wall components such as cellulose and lignin in a 55‐year‐old black spruce wood (*Picea mariana*) [57]. Raman images of cellulose and lignin were accurately generated. Cellulose gives three distinct peaks at 380, 1098 and 2900 cm−1, whereas lignin has two overlapping bands at 1600 and 1650 cm−1. The distribution of lignin was generated using both the 1600 and 1650 cm−1 bands, while the cellulose distribution maps were found to be most reliable when generated using the 2900 cm−1 band which has contribution from lignin alone, without other chemical interferences. Lignin‐to‐cellulose ratio was also determined, and it was found to differ in different areas of the plant cell wall. Because the 1650 cm−1 line had as well contribution from coniferaldehyde and coniferalcohol, it was possible to also image the coniferaldehyde and coniferalcohol distribution, which followed that of lignin. Sun et al*.* [58] have also used Raman mapping to get information on the lignin and cellulose polymers distribution and composi‐ tion in *Eucalyptus globulus* and corn stover. They have imaged the lignin and cellulose within different areas on the plant cell walls, from the epidermis to the pith area. Based on the Raman spectral fingerprints, significant compositional differences between *Eucalyptus globulus* and corn stover were observed, but also between different types of cells within the same plant. Schmidt et al. [59] acquired sub‐micrometer lateral resolution Raman images of *Arabidopsis thaliana* stem cross sections using the visible 532 nm laser and obtained information on the spatial distributions of cell wall polymers. As such, the distribution of carbohydrates (mainly cellulose) and lignin was obtained. The spatial distribution of polymers was obtained by integrating the C–H intensities between 2820–2935 cm−1 for cellulose and 1550–1700 cm−1 for lignin. Intense cellulose signals were identified within the secondary walls, whereas lignin was mainly found in the cell corners and in very little amounts in the secondary walls. However, since lignin distribution was not homogeneous, some secondary walls were strongly lignified (ensuring waterproofing). Richter et al*.* [60] took images from different tissues at different positions within the leaf of *Phormium tenax* and managed to visualize (using the 532 nm laser line) pectin and lignin distribution and to determine the cellulose microfibril angle on the cell walls.

Carotenoids are another promising class of compounds that can be analyzed and imaged through Raman mapping. They are organic pigments, conjugated double bond chains found in plants and other photosynthetic organisms, including bacteria and fungi. Carotenoids have important physiological roles, making them important molecules in plant biology, food science and pharmacology. In plants and algae, carotenoids protect from photodamage and absorb energy to be used in photosynthesis, whereas in human body they are potent antioxidants, and some of them are vitamin A precursors. The human body is unable to synthesize carote‐ noids, so they must be introduced through the diet, from carotenoid‐rich foods (e.g., carrots, tomato, maize, kiwi, cucumber, spinach, broccoli, etc). Information about carotenoid distri‐ bution in different plants and plant tissues is limited. Brackmann et al*.* [61] used coherent anti‐ Stokes Raman scattering (CARS) to gain information on β‐carotene distribution in sweet potato, carrot and mango. The β‐carotene distribution was probed using the C=C vibrational peaks at 1520 cm−1, characteristic for β‐carotene. Heterogeneous rod‐shaped bodies with high carotenoid density were identified in sweet potato and carrot, while in mango carotenoid‐filled lipid droplets were identified as homogeneous aggregates. Raman imaging would also be suitable for other types of carotenoids such as lutein and lycopene, since they all have similar vibrational Raman bands at 1500–1535 cm−1 (C=C stretching), at 1145–1165 cm−1 (C–C stretch‐ ing) and at 1000–1010 cm−1 (C–CH3 deformation) [62, 63]. The Raman bands are similar to all carotenoids, but shifted in position according to the number of conjugated bonds, the side groups and to the interaction of carotenoids to other plant constituents. Raman mapping was proved to be useful in evaluating the individual distribution of 7‐, 8‐ and 9‐double bond conjugated carotenoids in the intact tissues of *Calendula officinalis* [63]*.* The Raman images were generated based on the peak at round C=C stretching vibration at 1520 cm−1. This band was shifted at 1536, 1530 and 1524 cm−1 for the 7‐, 8‐ and 9‐conjugated double bond carotenoids, respectively.

Roots of different carrot cultivars were screened for their individual carotenoid distribution. The β‐carotene signal at 1520 cm−1 was used for integration. The level of β‐carotene was heterogeneous across root sections of orange, yellow, red and purple carrots. In the secondary phloem, the level of β‐carotene increased gradually from periderm toward the core, but declined fast in cells close to the vascular cambium. Lutein and α‐carotene were deposited in younger cells, while lycopene in red carrots accumulated throughout the whole secondary phloem at the same level [64]. Raman mapping was also applied for studies of *Pelargonium hortorum* to illustrate the heterogeneous distribution of the individual carotenoids in the leaves [65].

Plant polyacetylenes are another class of compounds that can be identified based on their C=C stretching vibration in the 2100–2300 cm−1 range. Using the Raman peaks at 2258 and 2252 cm −1 characteristic to the most common polyacetylenes falcarinol and falcarindiol, Baranska et al*.* [66] showed that polyacetylenes are mainly located in the outer section of the carrot roots.

Algae species are important candidates for industrial lipid and biofuel production. Sharma et al*.* [67] used Raman mapping for lipid analysis of microalgae. Characterization of lipid contents in cells obtained by mutagenesis showed that they managed to obtain mutants with increased lipid content. They have generated Raman images of the lipid‐rich, carotenoid‐rich and protein‐rich areas on the *Chlamydomonas reinhardtii* microalgae based on the characteristic peaks at 1003 cm−1 (proteins), 1445 cm−1 (lipids) and 1520 cm−1 (carotenoids).

Apart from mutations, the growth media can also induce generation of different metabolites. *Chlorella sorokiniana* and *Neochloris oleoabundans* represent two good candidates for biofuel production. The species were Raman mapped at 532 nm for identification of carotenoid and triglyceride production, and in consequence, the maps were generated based on the signal intensity in the 1505–1535 cm−1 for carotenoids and 2800–3000 cm−1 for triglycerides (CH2 stretching) [68]. Both healthy algae and nitrogen‐starved algae were examined. Only carote‐ noids could be mapped in the healthy cells. The maps showed distinct locations where the carotenoids are concentrated as they are normally located in the chloroplasts. Triglyceride production was observed under nitrogen‐starvation conditions, and it was possible to image the lipid‐rich regions within the starved algae. He et al*.* [69] reported similar results of triglyceride accumulation upon nitrogen starvation of *Coccomyxa* sp. algae. The triglycerides were imaged through the Raman lipid characteristic peaks at 1440, 1650 and 2840–2950 cm−1 (alkyl C–H bending, C=C stretching and CH2 stretching, respectively).

Some algae are able to produce large amounts of carotenoids when irradiated with light under specific conditions (e.g., *Hematoccocus pluvialis* which produces large amounts of zeaxanthin). Grudzinski et al*.* [70] analyzed two algal strains, *Chlorella protothecoides* and *Chlorella vulgaris,* with respect to carotenoid production upon light‐induced yellowing. They found the yellow coloration to be associated with xanthophyll formation, especially zeaxanthin. Under strong light exposure conditions, newly formed carotenoids were identified as a cell nucleus. It was possible to determine that zeaxanthin is the major carotenoid by performing Raman mapping both at 488 and at 514 nm. Both wavelengths are in resonance with xanthophyll pigments, but 514 nm is in resonance with zeaxanthin only. The cell nuclei give particularly high signal assigned to carotenoids when imaged under the 514 nm laser and low signal when imaged under the 488 nm laser, proving that zeaxanthin is the major synthesized carotenoid.
