A systematic literature search (PubMed, MEDLINE, Web of Science) was conducted using the following terms: “Diagnosis”, “Raman spectroscopy”, “Surface enhanced Raman spectroscopy”, “SERS”, “CRC”, “Tissue”, “Biofluids”, “Immunoassay”, “CEA”. Following these searches 30 studies were selected for inclusion based on the criteria: (1) That they are original studies based on the clinical applications of Raman spectroscopy, SERS or both for CRC detection in human tissue, biofluids or cell lines; and (2) in the case where there are large amounts of similar studies that they were the first to report such data. Only studies published up to January 2015 were included in this review.

Raman spectroscopy is a type of vibrational spectroscopy that allows the user to gain molecular information about a sample through the scattering of incident light. In general, when light is passed through or onto a sample a small proportion of the photons are scattered (approximately 10^-5). The majority of this is Rayleigh or elastic scattering; where the energy of the incoming photon is equal to the energy of the scattered photon (Figure 1)[23]. Around 1 in 10⁷ of the incident photons are in-elastically scattered resulting in the incident photon and the scattered photon having a difference in energy.

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Energy shifts involved in light scattering interactions. Adapted from Lin et al[23].

The inelastic scattering is a relatively weak effect which was first observed in 1928 by Sir CV Raman and is known as Raman scattering[24]. When scattered light is measured with a spectrometer a series of lines are observed, the shift in the energy [measured by wavenumber (cm^-1)] from the Rayleigh line (equal to incident energy) is known as the Raman shift. The shift recorded corresponds to specific vibrational or rotational modes of the sample molecule. The intensity of the “Raman shift” for a particular molecule is directly proportional to the concentration of that molecule within a sample so the resulting spectrum of a sample will be a superposition of Raman response of all the Raman active molecules from within a sample, e.g., proteins, nucleic acids, etc. It is interpreted as the molecular “fingerprint” of the sample, an example of a spectrum can be seen in Figure 2. It is due to the “fingerprint” that Raman is seen to be desirable for application to cancer detection because it offers the possibility of detecting minute differences in analyte concentrations and is a non-destructive technique.

Open in New Tab   Full Size Figure   Download Figure

Figure 2 Example Raman spectrum of dried human serum. Spectrum of dried droplet of human serum taken with 785 nm laser excitation, 10 s acquisition time with an InVia Raman spectrometer (Renishaw, United Kingdom).

It should be noted that Raman is not the only type of vibrational spectroscopy that has been reviewed for clinical applications[25,26]. One of the other main areas of vibrational spectroscopy being applied to clinical applications is fourier transform infrared spectroscopy (FTIR) and a type of enhancement infrared attenuated total reflection. FTIR relies on either the absorbance or the transmission of light through a sample, then similar to Raman the difference in the emitted and absorbed or transmitted light is measured to gain molecular information about a sample. This technique has also been used for diagnosis of CRC independently and also coupled with immunohistochemical staining[27,28]. Compared with other vibrational spectroscopy techniques Raman holds many desirable properties for the application to a screening method. One of the biggest advantages of using Raman spectroscopy is that samples can be in aqueous solutions due to water having a small Raman cross-section at near-infrared wavelengths, water has a high absorbance in FTIR and therefore can interfere with a spectrum. Table 1 shows a comparison table constructed from literature to give an overview of the strengths and weaknesses of FTIR, Raman and traditional hematoxylin and eosin staining (H and E) when used for clinical applications[25,29,30]. It suggests that for the application to biological samples such as tissues and biofluids Raman could be the most favorable option. Like FTIR it is non destructive and can be performed in real time compared to H and E staining. However, Raman spectroscopy has the advantage over FTIR of being able to scan over a larger wavenumber range than FTIR and with better spatial resolution than both FTIR and H and E. Furthermore, Raman is a technique that relies on scattering so measurements can be taken with single ended endoscopic probes. This is advantageous for in vivo applications because it makes it possible to study samples that are optically too thick for transmission techniques[31]. This, along with the need for a less invasive diagnostic tool that can analyse liquid biofluids leads to the remainder of this review focusing on the application of Raman and SERS technology.

Many clinical applications of Raman and SERS use chemometic data analysis techniques in order to aid data manipulation and to pick out the small differences that could indicate disease. Details of principle component regression (PCR) and partial least squares can be found in Kramer (1998)[32]. Principle component analysis (PCA) applications can be found in Shinzawa et al[33] (2009). This review will only state the technique that has been used in each study.

Reviews dedicated to Raman spectroscopy for the use of in vivo probes for clinical use for many types of tissue including tissues of the gastrointestinal (GI) tract are already available[25,40,41]. Briefly, the in vivo probes for use during endoscopy for use with GI tissue was first introduced by Shim et al[42] (2000). In general, the in vivo probes use a laser excitation source in the NIR wavelength range (light is non-mutagenic in this region) coupled to an optical fibre probe. The probe can then act as both a source of light and a detector relaying a signal back to a charged coupled device detector and computer analysis. The development of the probes is not a simple process, some materials that the probes are manufactured from have a large Raman cross sections leading to design challenges regarding signal to noise ratio (gaining unwanted noise from the material). Other issues can be caused by tissue fluorescence signal being larger than the Raman signal making data acquisition difficult and spectral acquisition times to be impractical for clinical applications (more than 10 min). Nevertheless, some research groups have been successful in designing probes for specific use with gastrointestinal tissue for use in routine endoscopy that have short acquisition times[35,43].

Shim et al[42] (2000) successfully applied an in vivo probe to gain Raman spectra of colonic and oesophageal tissues from 20 patients. The pressure spectral acquisitions were made with 5 s exposure time and repeated at normal and malignant sites within both areas of interest. In the colonic tissue subtle differences between spectral area 1100-1800 cm^-1 were identified however no significant prominent changes were evident. PCA and LDA analysis was then introduced in order to try and distinguish accuracy for application to GI diagnostics however no specific results relating to diagnostics were published.

Molckovsky et al[44] (2003) were the first to assess the diagnostic potential of near infrared Raman spectroscopy on colonic tissue by using adenomatous polyps as a model for dysplasia. The group used a custom-made fiber-optic Raman probe and used PCA-LDA analysis and LOOCV to analyse a total of 33 polyps from 8 patients. After an initial ex vivo study of polypectomy specimens that involved a total of 54 spectra the analysis algorithm identified a sensitivity of 91% and a specificity of 95%. An in vivo study was then conducted with a total of 19 spectra form 9 polyps, after spectral analysis the algorithm identified adenomas with a sensitivity of 100% and specificity of 89%. A similar study involving a higher number of specimens for a similar use on ex vivo tissues was conducted by Widjaja et al[45] (2008), the group were able to differentiate cancerous tissue with 100% sensitivity and 98.1%-99.7% specificity using a diagnostic algorithm using PCA and LDA.

Raman spectroscopy has also been investigated as a complimentary technique to histopathology. The first application of Raman spectroscopy for characterisation of colonic tissue (among others) to discriminate between cancerous vs normal was by Feld et al[46] (1995). The group looked at the difference spectra between normal and cancerous tissue, this results showed potential that spectral differences in the tissue could be due to higher nucleic acid levels in the cancerous samples[46].

Beljebbar et al[47] (2009) used a Raman micro spectrometer on 27 normal and 27 cancerous ex vivo frozen tissue samples. Unsupervised hierarchal cluster analysis to differentiate between normal and adenocarcinomatous human colonic tissues was discussed. The technique was based on the spatial distribution of molecular changes in colon constituents such as proteins, lipids and nucleic acids. The spectroscopic data was then used to create pseudo-colour Raman images of tissues for comparison with histopathological slides. The data was then used to create databases for the purpose of comparing unknown specimens. Six extra frozen unknown samples specimens were fed into the database and were correctly identified as either cancerous or normal[47]. This study showed the potential for Raman spectroscopy to aid histopathology by adding structural molecular information to the visible information gained from H and E staining. This study used frozen tissue samples but there are different fixation methods available. There have been studies on the effect that different fixation methods of both cell lines and tissue samples taken from resection has on the Raman spectral signature but these will not be discussed further in this review[48-50].

It should be noted that Raman spectroscopy used as an adjunct to histopathology has not just been applied to CRC. Some groups have investigated the use of stimulated Raman spectroscopy (SRS) to create spectral histopathological images for breast, brain and skin tissue among others[51-54]. For example, Satoh et al[54] (2014) used SRS and PCA multivariate analysis to produce Raman map images of damaged liver tissue in mice. The images could then be compared to different staining methods used in histopathology.

Raman spectra are subject to interference from samples that exhibit fluorescence; a fluorescence signal is far greater than the Raman response so it can “hide” any Raman signal. Fluorescence is fairly common when using wavelengths less than 785 nm that are commonly used in biological studies. Also, when Raman is used on complex biological samples Raman peaks can be additive, making differentiation between biological markers difficult to resolve, this results in a reliance on data manipulation in order to detect small spectral differences. SERS offers a resolution to some of these issues as it reduces the effects of inherent fluorescence while increasing the intensity of the Raman response of the sample. Premasiri et al[58] (2001) demonstrated the need for an enhancement mechanism in order to detect some low concentration analytes in urine analysis by Raman spectroscopy. It was found that urea had a sufficiently high concentration to be analysed by NR in liquid form however, lower-level nitrogen compounds needed enhanced Raman spectroscopy in order to detect these compounds[58]. Therefore SERS was used as an enhancement mechanism in order to be able to detect the compounds that were in concentrations too small to detect with Raman.

The basic principle of SERS is to amplify the Raman response of a given analyte. The SERS effect was first discovered in 1974, and understood to be an enhancement of Raman scattering in 1977[59,60]. This is generally achieved by having an analyte attached or close to the surface of a nanoscale metal substrate causing an enhancement factor of up to 10[14,15,61,62]. The exact mechanism of SERS enhancement is still an area of active research, however it is generally accepted that two mechanisms contribute to the enhancement[63]. One is based on electromagnetic field enhancement due to excitation of electromagnetic resonances in the SERS active nanoscale metal substrate. The other is known as “chemical enhancement” which is a result of the metal electrons causing a charge transfer between the metal substrates and the adsorbates. The result of the combined enhancement is an extremely powerful technique that combines ultra sensitive detection limits with the molecular structure information from Raman spectroscopy giving the possibility of single molecule detection[64].

The main use of SERS in clinical applications for CRC has been as a detection method. Lin et al[65] (2011) were the first to report SERS serum analysis for the detection of CRC. In their study colloidal gold nanoparticles were simply mixed with serum samples from 38 patients and 45 control samples and dropped onto an aluminium substrate. This technique is known as label-free SERS. It generally relies on blood constituents being adsorbed onto the surface of metallic nanoparticles causing an enhanced Raman response. In the Lin et al[65] study a Raman micro-spectrometer (Renishaw, Great Britain) fitted with a 785 nm diode laser was used to gain spectra in the 300-1800 cm^-1 range. Spectra from the two groups were normalised to the integrated area under the curve in the 350-1750 cm^-1 wavenumber range. The mean spectrum for the normal serum and the cancer serum were then compared to isolate wavenumbers that showed the most variation between the two groups. Then an empirical diagnostic algorithm based on peak intensities at 725 cm^-1 and 638 cm^-1 was used to classify the normal and the cancer samples. These were chosen based on previous studies by Han et al[66] (2008) that showed the ratio to be important disease marker. This technique was compared PCA-LDA multivariate approach. The PCA analysis then used whole spectra to discern the spectral components that had the largest variation. After comparison the group found PCA-LDA to be more effective at detecting CRC. With specificities for detecting cancer to be 68.4% and 97.4% for the empirical approach and the multivariate approach respectively.

This study showed that through simply mixing gold nanoparticles with serum that it is possible to discriminate between normal and cancer samples using SERS along with both empirical and multivariate analysis techniques. The potential for using whole spectra coupled with PCA-LDA to be used as a screening technique for CRC was then discussed using the PCA-LDA methods in a second publication from the same group[65]. The two publications also included tentative peak assignments and major vibrational bands that have been previously observed in serum samples. The peak assignments are vital to being able to accurately describe what is happening at the molecular level when a patient has a disease. However, due to the additive nature and complex compounds found in serum samples it can sometimes be difficult to be sure of peak assignments. Furthermore, the methods described above such as looking at the 725 cm^-1 band can be a marker for disease but it is not specific to CRC. In order for the technology to become useful as a diagnostic there is a need to have a Raman/SERS marker that is specific to CRC.

The need for specific detection has motivated the development of “labelled” SERS probes. These probes have previously been used for detecting disease specific proteins in both tissue and serum samples[67-71]. They have also been used for the detection of circulating tumour cells[72]. However there is little reported on the specific application of targeted SERS probes for use in detecting CRC[73]. In general targeted SERS techniques rely on either aggregation of antibody-functionalised nanoparticles after exposure to a protein or they are used to form of sandwich immunoassay similar to that of an ELISA setup but using SERS active probes rather than fluorescent-tagged antibodies (Figure 4).

Open in New Tab   Full Size Figure   Download Figure

Figure 4 Different methods of producing a surface enhanced Raman spectroscopy response with biological samples. A: Mixing based methods with both non-labeled mixing and labeled or “aggregation” based mixing; B: Labeled surface enhanced Raman spectroscopy based immunoassay. SERS: Surface enhanced Raman spectroscopy.

Both techniques offer advantages and disadvantages; techniques relying on the aggregation of nanoparticles use fewer antibodies and in general have very simple protocols that can be done on cheap substrates. However, when dealing with the aggregation of nanoparticles it can be difficult to “find” the correct spot on a sample where the SERS intensity is greatest. Furthermore, studies relying on aggregation can be susceptible to large variation as controlling the aggregation can be difficult. SERS based immunoassay holds the advantage over aggregation because if the disease specific protein is in a sandwich style assay then the area to probe is easier to locate and the Raman signal is less likely to be variable, as one would expect more even coverage of the protein over the assay area. One advantage of both of these techniques over current fluorescence methods is that Raman bands are much narrower than those in fluorescence which means there is the potential to use more probes in a multiplex assay than is currently possible with fluorescent probes. Chen et al[74] (2013) developed a SERS based immunoassay for the detection of carcinoembryonic antigen (CEA) in serum of patients with CRCs. CEA antibody functionalised glass slides were used in conjunction with SERS active probes that were also functionalised with CEA[74]. A range of concentrations from 5 × 10^-3-5 × 10⁵ ng/mL CEA were prepared and the SERS response monitored. The SERS intensity was linearly correlated with the concentration of CEA in the characteristic peak or the Raman reporter molecule at 1077 cm^-1 and hence a calibration curve was established. CEA concentrations in serum samples form 26 patients with CRC were then analysed with both SERS immunoassay and electrochemical luminescence. The results were then compared and it was found that the two techniques had similar agreement. Using the calibration curve for the patient samples a detection limit of 5 pg/mL was achieved. This is the only study specifically using CEA for the detection of CRC, however there is other work in the literature using CEA conjugated antibodies but towards use for other diseases[69,70].

Ito et al[75] (2014) developed a SERS based assay that used silver nanoscale hexagonal columns (NHC) on phosphor bronze chips. The chips were negatively ionized and 36 clinical serum samples from patients with benign diseases, gastric and CRCs were dropped onto the chips. Using a 632.8 nm laser excitation, measurements were taken and two peaks appeared to be prominent in the SERS spectra at 1350 cm^-1 and 1570 cm^-1. Polynomial fitting was then used to determine SERS peak height for each of the samples. In order to validate that the SERS peak height and the concentration of the serum samples were correlated the measurements were repeated at different dilutions of the same serum sample. It was found that 10-fold dilutions were the saturating dilution for the NHC chips so 10 fold dilutions were used throughout the rest of the experiment. Finally the SERS peak height for both of the prominent peaks was compared between the three groups of samples and it was found that the SERS peak heights in the benign samples was indeed significantly lower than in the cancerous samples. This was calculated using the Pearson product-moment correlation coefficient and the non-parametric Spearman’s rank correlation coefficient.

Another slightly different approach to using SERS for clinical use for CRC is using SERS as a characterisation/validation tool when developing other nanoscale devices such as the work done by da Paz et al[76] (2012). In this study SERS was used as a characterisation tool in the development of maghemite nanoparticles as a theragnostic device for CRC[76]. The SERS active nanoparticles used in this study were functionalised with Anti-CEA antigen in the hope that they can be used to detect primary and metastatic CRC, it was hoped by the authors that these nanoparticles could then be developed for a variety of applications including magnetic resonance imaging (MRI) enhancement and targeted drug delivery.

Raman and SERS based tools have shown potential that they will have a place as either an alternative or an adjunct to current diagnostic methods. The development of SERS biomarker detection could also lead to its use in personalized medicine. However, there are still some limitations of Raman and SERS techniques that will need to be overcome before they are routinely used in a clinical setting. These include: (1) Many Raman studies involve costly equipment and expensive substrates, there will need to be investigations into cost reduction for large scale applications; (2) Raman and SERS studies that are carried out in a laboratory will require sample handling and storage, the effect of handling samples and storage techniques on the performance of Raman based tools will need to be quantified; (3) Thermal damage thresholds of in vivo tissue and ex vivo tissue samples from the colon and rectum will need to be established; (4) Many studies for clinical use different analytical techniques, and still require the skill of the user to determine the results of these techniques. User-friendly software for diagnostic analysis of the spectra will need to be developed and tested for multi-user reliability; (5) Inter equipment variability studies will need to be carried out, Raman equipment can often be susceptible to variability from external factors such as room temp, laser stability, etc.; and (6) SERS based techniques have been subject to reproducibility issues, CRC is a heterogeneous disease so if immunoassay style tools are to be used then large scale studies with clinical samples will need to be carried out.