1. Introduction

Polycystic ovary syndrome (PCOS) is common cause of infertility among women during their reproductive age [1]. It is a multi-factorial heterogeneous syndrome that increases the risk of diabetes and cancer [1]. There is no specific test that can recognize PCOS, and the rule-out method is the most common approach for PCOS diagnosis. Following the Rotterdam Criteria [2], the presence of at least two characteristics from the following three are required to diagnose PCOS: 1) Oligo and/or anovulation, 2) clinical and/or biochemical signs of hyperandrogenism (e.g. non-classic congenital adrenal hyperplasia, hyperthyroidism, idiopathic hirsutism, familial hirsutism, Cushing’s syndrome, androgen secretion) and 3) polycystic ovaries (e.g. cysts, ovarian hyperthecosis, stromal hyperthecosis). Other characteristics such as hyperprolactinemia can also be present [3]. Thus, the diagnosis of PCOS is exhausting for the patient, expensive and time consuming.

Chemerin is a chemoattractant protein known as Tazarotene-induced Gene 2, and acts as a ligand for the G-protein coupled receptor [4]. Chemerin levels correlate with insulin resistance and obesity [5,6], which are common comorbidities in PCOS [7,8]. Although some reports imply that there is a correlation between chemerin and PCOS, the role of chemerin in the pathogenesis of PCOS is unclear [9], although serum chemerin level has been reported to be up regulated in PCOS subjects [10].

There are various methods to measure chemerin levels. Enzyme-linked immunosorbent assay (ELISA) is a laboratory technique used to determine the concentration of chemerin in the ng/mL range, and has been used to measure serum chemerin levels in women with PCOS. The main drawbacks of this method are that ELISA cannot distinguish between the different chemerin isoforms [11,12] and it is time-consuming (usually 6 to 8 hours/run) . Liquid chromatography/mass spectroscopy-mass spectroscopy (LC/MS-MS) is another technique to measure the chemerin level in serum [13], however it is very expensive and requires regular service, which adds to the cost of each run [14]. One of most popular methods used to identify and quantify specific proteins is western blot. While this method is specific and quantitative; it is relatively time consuming (usually two days) and expensive due to requirements for skilled analysts and often expensive antibodies. Western blot is also difficult to perform when the number of samples is high, and the reagents are of sub-optimal quality (e.g. antibody) which could lead to erroneous results.

While PCOS has been detected and chemerin has been analyzed and monitored with different techniques, this is the first time that surface-enhanced Raman scattering (SERS) has been used in the detection of PCOS and to measure the concentration of chemerin in phosphate-buffered saline (PBS) and follicular fluid (FF). The simplicity of our SERS setup, the highly informative SERS chemerin spectra, and the low quantity of clinical samples required, encouraged us to pursue this research. In conjunction with partial least squares (PLS) and principal component analysis (PCA) analysis techniques [15–17], this is a promising alternative to identify subjects with PCOS and to determine the chemerin contribution in its pathogenesis.

In this paper, we begin by sample preparation, and then discuss how to differentiate between PCOS and non-PCOS using PCA and spectral analysis. The procedure for detecting chemerin in PBS and FF using an improved PLS regression model is examined and compared to the Western Blot method. Finally, we evaluate the possible association of chemerin concentration with PCOS by using PCA and spectral analysis of follicular fluid from PCOS and non-PCOS subjects.

2. Experimental details

2.1 SERS experimental configuration

The experimental configuration included a 785-nm continuous wavelength multimode laser (B&W Tek Inc., Newark, DE, USA) with a maximum output power of 450 mW. The collimated laser beam passed through a bandpass filter centered at 785 nm ( ± 2 nm) to filter out other wavelength components around 785 nm from the laser. It was then directed through a dichroic filter (R785RDC, Chroma Technologies Corp., VT, USA) that reflected 785 nm ( ± 5 nm) at an angle of 45 degrees, and transmitted light in the 790 to 1000 nm range. The dichroic filter acted as a reflector for the laser beam, which was further focused on the sample holder using a 10 × microscopic objective lens (04 OAS 010, CVI Melles Griot, NY, USA) with numerical aperture (N.A.) of ∼0.25. The dichroic filter also acted as a high-pass filter of the light scattered backward from the sample, thereby allowing only the Raman wavelengths through. The filtered Raman light was then imaged onto a fiber bundle (30 multimode fiber, N.A. = 0.22, Fiberoptic System Inc., CA, USA), using another 6.3 × microscopic objective lens with N.A. of ∼0.20. The diameters of the core and cladding were 100 and 125 microns, respectively. The fiber bundle consisted of 30 identical fibers, with a packing efficiency of 80%. The output of the fiber bundle was interfaced with a Kaiser f∕18i Spectrograph (Kaiser Optical Systems, Inc., MI, USA) and a thermoelectrically cooled Andor charge-coupled device camera. The spectral resolution of the spectrometer was 2.05 cm⁻¹. Andor SOLIS software was used for spectral data acquisition, and the spectra were monitored on the data acquisition computer. The laser beam power at the sample was 250 mW, and the acquisition time for recording a spectrum varied from 1 to 10s depending on samples. The optical configuration described above was the optimal way to achieve Raman signals of samples with high signal-to-noise ratios. All spectra are the average of 10 consecutively recordings to improve signal to noise ratio in Raman signal and are vertically shifted for clarity.

2.2 Capillary sample holder for SERS

FF samples are very limited, therefore, one of the issues in this study was to minimize sample quantity required to record the Raman spectra. The Raman setup with a cuvette or hollow core photonic crystal fiber (HC-PCF) sample holder [18] requires a minimum of 1 to 3 mL, which is two orders of magnitude larger than the samples that can be collected from a patient. Using a capillary with less than 30 μL capacity is a good alternative to a cuvette or HC-PCF. The Raman setup with a capillary configuration is similar to the regular Raman setup that we used in our previous work [19]. The sample consumption using a vertical capillary was indeed lower than with cuvette or HC-PCF, but the matter-light interaction was less effective. However, this was a reasonable trade-off given the limited availability of samples quantities.

2.3 Sample preparation

The use of nanoparticles (NP) is critical to enhance the weak Raman signal of the molecules to be interrogated, particularly when the concentration of chemerin was low. The nanoparticles were prepared according to the method described by Leopold et al. [20] (see Appendix A).

FF samples were collected from PCOS (n = 10) and non-PCOS (n = 10) patients in accordance with Human Ethics Research Board of School of Medicine, Taipei Medical University (IRB approval # TMU-JIRB 201410033). PCOS was diagnosed as per the Rotterdam criteria [2]. Exclusion criteria of patients included the presence of oligonorrhea other than PCOS, the use of insulin sensitizers and lipid lowering agents. Clinical information (Mean ± SD) for the non-PCOS and PCOS subjects included age (36.50 ± 4.19 vs 33.89 ± 3.55); BMI (20.55 ± 2.59 vs 21.43 ± 2.09); serum levels of Estradiol-17β (2654 ± 1282 pg/ml vs 2731 ± 2047 pg/ml); and Progesterone (0.72 ± 0.46 ng/ml vs 0.85 ± 0.42 ng/ml). FF samples were collected during the aspiration of eggs from women undergoing in vitro fertilization (IVF), centrifuged (to eliminate particles and cells) and stored at −80°C pending analysis.

To better understand and detect differences in the FF from PCOS and non-PCOS patients and chemerin concentration, four groups of samples were used:

The sampling strategy in this study is shown in Fig. 1.

 

Fig. 1 Flow chart of the sampling strategy.

Download Full Size | PDF

2.4 Determination of chemerin levels by western blotting

Each FF sample was diluted in PBS (0.5μL FF in 4.5μL PBS) and reduced in 2x reducing Laemmli buffer (5μL; Bio-rad Laboratories #1610737, Mississauga, CA) containing 5% of β-mercaptoethanol (Sigma-Aldrich, Oakville, CA) and heated in a water bath (100°C; r 5 minutes). FF samples were subjected to SDS-PAGE separation (sodium dodecyl sulfate polyacrylamide gel electrophoresis) with 4.5% stacking gel and 16% separating gel (5 h 20 min). Proteins were electrophoretically transferred to nitrocellular membrane (Bio-Rad Laboratories). Membranes were washed with TBS-T [0.05% Tween-20 in 10 mM Tris, 0.15 M NaCl (pH 7.4) (Tris buffered saline)] and incubated for 10 min in antibody extender (Thermo-Fisher, Burlington, CA). Non-specific binding was blocked using 5% skim milk in TBS-T (0.3%; 1 h, RT), and then incubated (overnight, 4°C) with diluted goat anti-human chemerin (1: 1000; R&D System #AF2324) in TBS-T (0.05%). The membranes were treated with anti-goat conjugated with horse radish peroxidase (Bio-rad laboratories; 1:5000). Bands were visualized with an enhanced chemiluminescent agent according to the manufacturer’s instruction. An expected band of 16 KDa was digitalized using Quantity One software and quantified using Image J. A non-parametric and unpaired T-test was used to compare PCOS and non-PCOS chemerin levels in FF.

2.5 Multivariate data analysis

Multivariate data analysis (MVA) was performed to test differences in data and to predict future observations. PCA provided the best view of information and interpretation patterns in data, while a PLS model was constructed from the spectral and analytical data and its performance was assessed by calculating the coefficient of determination (R²), the root mean square errors of calibration (RMSEC), and the root mean square errors of prediction (RMSEP). The spectral data sets were constructed as follows: a calibration set to make the model, a validation set to validate the PLS model and a prediction set to test the model independently and to avoid overfitting in the model. This validation approach is known as test set validation (TSV) [21].

The SERS spectral data sets of PCOS and non-PCOS samples (the first, second and fourth group of samples) were used for PCA analysis to identify PCOS and non-PCOS samples, and the SERS spectral data set of chemerin in PBS (the third sample group) was used to construct the calibration model for PLS analysis with Unscrambler X version 10.3 (CAMO, Corvallis, OR, USA). The chemerin concentrations of FF solutions (the second group of the data set) were predicted using the model.

3. Results and discussion

The SERS spectra of the four sampling groups were studied to determine specific peak(s) that could differentiate between PCOS and non-PCOS patient samples, and to evaluate the correlation between chemerin peaks and chemerin concentrations. These spectra were recorded by adding 20 μL of NP to 20 μL of samples (50:50, volume ratio). Before recording the SERS spectra of the four sampling strategy groups, the SERS spectra of NP and PBS were recorded to exclude the possibility of non-specific peak(s) due PBS or NPs. The presence of PBS and/or NPs in samples caused high background intensity (see Appendix A). As a data preprocessing step, we subtracted this background using our MATLAB code before applying PLS/PCA analysis.

3.1 Differentiating between PCOS and non-PCOS patients

This section investigates if SERS in conjunction with PCA can be used to differentiate between PCOS and non-PCOS patients. We first recorded the spectra of the first and second sample groups after adding 20 μL of NP to each sample. We initially worked with pooled samples, a common practice when the available samples are very low or highly expensive. The SERS spectra of the pooled samples in both PCOS and non-PCOS groups are shown in Fig. 2(a). Both pooled PCOS and non-PCOS spectra showed some Raman peaks at the same wavenumbers, but with different intensities. PCA analysis was then performed as a qualitative method on both the PCOS and non-PCOS data to reveal any hidden structures and clustering within the samples. Figure 2(b) shows the score plot of two principal components (PC), and summarizes the variations in PCOS and non-PCOS of the first group of samples. The graph indicates that ten recorded spectra of these samples are adequately separated.

 

Fig. 2 a) Surface-enhanced Raman scattering (SERS) spectra of chemerin_pooled follicular fluid (FF) samples and b) score plot of principal component (PC) of polycystic ovary syndrome (PCOS) and non-PCOS data of the first sample group.

Download Full Size | PDF

The verification of the SERS capability was followed by recording the SERS spectra of the second sample group (individual FF samples from 10 PCOS and 10 non-PCOS subjects). As shown in Fig. 3, SERS spectra of these samples exhibited SERS peaks at 691, 714, 724, 766, 781, 811, 838, 916, 942, 964, 976, 985, 1001, 1020, 1030, 1072, 1091, 1129, 1186, 1215, 1246, 1271, 1320 and 1334 cm⁻¹. The band assignments for chemerin in FF and chemerin in PBS are listed in a table in the Appendix B for reference.

 

Fig. 3 Surface-enhanced Raman scattering (SERS) spectra of chemerin in follicular fluid (FF) a) non-polycystic ovary syndrome (PCOS) and b) PCOS samples.

Download Full Size | PDF

Based only on the spectrum, we could not differentiate between PCOS and non-PCOS patients. However, after applying PCA to the second sample group, the score plot identified two different clusters, demonstrating that 80% of the PCOS samples were distinguishable from non-PCOS samples (Fig. 4). These groups are separated by the green lines in Fig. 4.

 

Fig. 4 Score plot of principal component (PC) of polycystic ovary syndrome (PCOS) and non-PCOS data of the second sample group.

Download Full Size | PDF

3.2 Investigating the association of chemerin with PCOS

The possible association of chemerin with PCOS was investigated, particularly in regards to its level in PCOS and non-PCOS patients. SERS allowed us to address the level of chemerin and its association to PCOS, potentially providing a practical tool for the diagnosis of PCOS.

3.2.1 Using PLS to detect chemerin in PBS solution

To investigate the presence and concentration of chemerin in the FF from PCOS and non-PCOS patients, and to determine whether chemerin could be used as marker for the possible diagnosis of PCOS, we began by determining the Raman spectrum of chemerin in PBS (see Appendix A). The Raman spectrum of lower concentrations of chemerin is not informative, and thus cannot be used for low level chemerin detection. The spectrum of the highest chemerin concentration in PBS was 6.25 μM, and did not reveal any specific peaks. This indicates that the Raman spectra of lower concentrations of chemerin in PBS cannot effectively develop a PLS model for the prediction of chemerin concentration in a patient sample. As a result, we chose to use the SERS technique in conjunction with PLS analysis.

Using the third sample group, 20 μL of NP was added to prepare samples at 3.15, 1.56, 0.785, 0.39, and 0.195 μM. As shown in Fig. 5 and within a spectral range of 650 to 1500 cm⁻¹ there were common peaks in chemerin in PBS and FF samples at 691, 714, 766, 985, 1001, 1020, 1030, 1072, 1091, 1215, 1271 and 1320 cm⁻¹. The peaks at 876, 1173, 1224 and 1404 cm⁻¹ were due to chemerin.

 

Fig. 5 Surface-enhanced Raman scattering (SERS) spectra of chemerin in phosphate-buffered saline (PBS) samples.

Download Full Size | PDF

Figure 5 shows an increase in the peak intensity in the Raman spectrum with increased chemerin concentration. Fifty spectra of the third group of known samples (ten Raman spectra collected to ensure consistency of the replicated measurements) were used to build a PLS model, and twenty spectra of the second group were used to predict the chemerin concentration using the model. A TSV model was then developed using 67% of the known samples as a calibration set, and 33% as a validation set. These were chosen randomly by the Unscramble software.

3.2.2 Loading and score plots

One result of PLS analysis is the loading plot, which indicates how the variables in different principal components correlate. The loading plot of the first and second PC is shown in Fig. 6(a). The loading plots of higher PCs did not show new information (see Appendix A). According to the figure, the most important variables are 766, 876, 985 and 1001 cm⁻¹, which are the main chemerin peaks (see Fig. 5). Another informative plot is the score plot, which indicates how different chemerin concentrations are represented by the variables. Figure 6(b) shows that 74% (X1 60%, X2 14%) of the X variance explains 99% (Y1 97%, Y2 2%) of the variation of chemerin concentrations in the samples. The figure also verifies that all the samples used to develop the PLS model are easily distinguishable from one another.

 

Fig. 6 Loading plot (a) and score plot (b) of principal component (PC) of chemerin in phosphate-buffered saline (PBS) samples.

Download Full Size | PDF

3.2.3 PLS calibration model

The baseline corrected spectra were used to make the PLS model. According to the constructed PLS model using raw spectra, the R² (for calibration and validation), RMSEC and RMSEP for the spectral range of 650 to 1450 cm⁻¹ are 0.92, 0.92, 0.030 and 0.210 respectively while using baseline corrected spectra improves these values to 0.99, 0.99, 0.016 and 0.126, respectively. Figure 7 shows how this preprocessing improves the quality of recorded spectra. All these calculations are based on the PLS model using six PCs, which minimize the RMSEP value. The optimum number of used PCs was determined by Unscrambler X and the plot of RMSEC vs number of PCs are presented in Appendix A. The calibration curve of the developed PLS model is shown in Fig. 8, while predictions of the chemerin concentration of PCOS and non-PCOS patient samples with this model are shown in Table 1. The findings indicate that the PLS model can predict different concentrations of chemerin in FF, with standard deviations of 0.158 and 0.223 μM for non-PCOS and PCOS patient samples, respectively. This range of deviation with the SERS method is due to using the spectra of chemerin in PBS samples to predict the chemerin level of chemerin in FF samples, which is unavoidable. The correlation coefficient (R) between SERS and WB method was 0.81.

 

Fig. 7 Raman spectra of a sample before (a) and after (b) preprocessing.

Download Full Size | PDF

 

Fig. 8 Partial least squares (PLS) regression model for chemerin concentration prediction in 650 to 1450 cm⁻¹ spectral range using test set validation.

Download Full Size | PDF

Table 1. Chemerin concentration predictions in 20 non-polycystic ovary syndrome (PCOS) and PCOS samples

View Table | View all tables in this article

The average concentrations of chemerin of PCOS predicted by the SERS and Western Blot methods were 0.819 and 0.754 μM, respectively, and for non-PCOS samples were 0.656 and 0.460 μM. Although the predicted level of chemerin concentration using SERS is higher than with the Western Blot method, the differences between the average chemerin levels of PCOS and non-PCOS samples with the SERS and Western Blot methods are similar (0.163 and 0.294 μM, respectively). This indicates that the chemerin level in PCOS and non-PCOS samples measured by the SERS and Western Blot methods are in the same order of magnitude.

3.2.4 Spiking pooled samples with chemerin

To further assess the association of chemerin with PCOS, we spiked pool samples with chemerin, and studied its impact on the SERS spectra (the fourth sample group). The spectra of spiked samples enabled us to find out if there is any specific peak(s) that is related to PCOS. Both PCOS and non-PCOS pool samples were spiked with 10, 20, 40, 60 and 80 ng of chemerin dissolved in PBS. The SERS spectra of all the spiked samples were recorded after adding 20 μL of NP to each sample, as shown in Fig. 9.

 

Fig. 9 Baseline corrected surface-enhanced Raman scattering (SERS) spectra of a) polycystic ovary syndrome (PCOS) and b) non-PCOS patient samples spiked with chemerin.

Download Full Size | PDF

As demonstrated in Fig. 9(a), there was a correlation between the Raman intensity at different wavenumbers and the amount of chemerin in the PCOS sample. The wavenumbers that showed higher intensities and correlation between the Raman peaks and the amount of chemerin were 724, 1001, 1028 and 1224 cm⁻¹ (due to chemerin_PBS or FF), while the other peaks failed to clearly show this correlation. The correlation also can be seen in the Raman spectra of non-PCOS spiked samples, as shown in Fig. 9(b). However, this was not as evident in PCOS spectra, as the intensities of the Raman peaks of non-PCOS samples were less than PCOS samples.

In Fig. 9, the comparisons between PCOS and non-PCOS samples spiked with the same amount of chemerin revealed that the Raman peaks in all the PCOS sample, particularly 724 cm⁻¹, had higher intensities than the Raman peaks of the non-PCOS samples, as shown in Fig. 2.

Figure 10 shows the result of applying PCA to the fourth sample group. All the spiked PCOS samples were clearly separated from the spiked non-PCOS samples, suggesting that the Raman method with PCA can be used to distinguish PCOS patient samples from non-PCOS patient samples. The green line in this Figure shows this distinguishing.

 

Fig. 10 Score plot of principal components (PCs) of polycystic ovary syndrome (PCOS) and non-PCOS data of the fourth sample group.

Download Full Size | PDF

4. Conclusion

We have demonstrated that using SERS in conjunction with PCA is a fast and accurate method to detect PCOS. We have also investigated the role of chemerin in PCOS disease and measured the concentration of chemerin in PBS and FF samples. We have shown that intensity of Raman peaks at 724, 1001, 1028 and 1224 cm⁻¹ correlates well with the amount of chemerin in PCOS and non-PCOS samples. The PLS regression method provided reliable estimate of chemerin concentrations of PCOS and non-PCOS samples.

Appendix A

A.1. NPs preparation

The UV-vis. absorption/extinction spectra of three different batches of silver nanoparticles (Fig. 11) that we used show that different batches of nanoparticles are similar and their size/shape are reproducible with limited clustering/aggregation.

 

Fig. 11 The absorption spectra and transmission electron microscopy (TEM) images of three batches of silver nanoparticles (NPs).

Download Full Size | PDF

A.2. PBS and NPs spectra

The background of NPs and PBS spectra (Fig. 12) show that we need to subtract them before PLS/PCA analysis.

 

Fig. 12 Surface-enhanced Raman scattering (SERS) spectra of phosphate-buffered saline (PBS) and nanoparticle (NP) solutions.

Download Full Size | PDF

A.3. Raman spectra of chemerin-PBS solution

The weak Raman spectrum of lower concentrations of chemerin (Fig. 13) lead us to use SERS technique to extract quantitative/qualitative information using PLS/PCA techniques. This Figure illustrates the Raman spectra of different concentrations of chemerin in PBS when using a cuvette or HC-PCF as a container.

 

Fig. 13 Raman spectra of chemerin in phosphate-buffered saline (PBS).

Download Full Size | PDF

A.4. The loading plots

The loading plots of lower PCs are usually more informative than loading plots of higher PCs (Fig. 14).

 

Fig. 14 The loading plots of higher principal components (PCs) of chemerin in phosphate-buffered saline (PBS) samples.

Download Full Size | PDF

A.5. RMSEC vs number of PCs

The PLS analysis was based on six PCs that was determined by Unscrambler X. This optimum number of used PCs guarantee the lower RMSEC (Fig. 15).

 

Fig. 15 The plot of root mean square errors of calibration (RMSEC) vs number of principal components (PCs).

Download Full Size | PDF

Appendix B

B.1. Band assignments

The band assignments for chemerin in FF and chemerin in PBS are listed in Table 2.

Table 2. Raman shift and band assignments for chemerin

View Table | View all tables in this article

Funding

Natural Sciences and Engineering Research Council of Canada (NSERC) (RGPIN-2016-05101); Canadian Institutes of Health Research (CIHR) (MOP-119381); Ministry of Science and Technology (MOST) (104-2314-B-038-063-MY2); Academia Sinica (BM10501010036, BM10601010024) and National Health Research Institute (MG-105-SP-07, MG-106-SP-07). Patricia DA Lima is a recipient of a CIHR-QTNPR Postdoctoral Fellowship.

Disclosures

The authors declare that there are no conflicts of interest related to this article.

References and links

1. National Institutes of Health Evidence-based Methodology Workshop on Polycystic Ovary Syndrome December 3–5, (2012).

2. Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group, “Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome,” Fertil. Steril. 81(1), 19–25 (2004). [CrossRef]   [PubMed]  

3. N. R. Farid and E. D. Kandarakis, Diagnosis and Management of Polycystic Ovary Syndrome (Springer, 2009).

4. O. De Henau, G. N. Degroot, V. Imbault, V. Robert, C. De Poorter, S. Mcheik, C. Galés, M. Parmentier, and J. Y. Springael, “Signaling properties of chemerin receptors CMKLR1, GPR1 and CCRL2,” PLoS One 11(10), e0164179 (2016). [CrossRef]   [PubMed]  

5. K. Bozaoglu, K. Bolton, J. McMillan, P. Zimmet, J. Jowett, G. Collier, K. Walder, and D. Segal, “Chemerin is a novel adipokine associated with obesity and metabolic syndrome,” Endocrinology 148(10), 4687–4694 (2007). [CrossRef]   [PubMed]  

6. H. Y. Shin, D. C. Lee, S. H. Chu, J. Y. Jeon, M. K. Lee, J. A. Im, and J. W. Lee, “Chemerin levels are positively correlated with abdominal visceral fat accumulation,” Clin. Endocrinol. (Oxf.) 77(1), 47–50 (2012). [CrossRef]   [PubMed]  

7. A. Dunaif, “Insulin action in the polycystic ovary syndrome,” Endocrinol. Metab. Clin. North Am. 28(2), 341–359 (1999). [CrossRef]   [PubMed]  

8. S. Sam, “Obesity and polycystic ovary syndrome,” Obes. Manag. 3(2), 69–73 (2007). [CrossRef]   [PubMed]  

9. Q. Wang, J. Y. Kim, K. Xue, J. Y. Liu, A. Leader, and B. K. Tsang, “Chemerin, a novel regulator of follicular steroidogenesis and its potential involvement in polycystic ovarian syndrome,” Endocrinology 153(11), 5600–5611 (2012). [CrossRef]   [PubMed]  

10. S. Yang, Q. Wang, W. Huang, Y. Song, G. Feng, L. Zhou, and J. Tan, “Are serum chemerin levels different between obese and non-obese polycystic ovary syndrome women?” Gynecol. Endocrinol. 32(1), 38–41 (2016). [CrossRef]   [PubMed]  

11. D. H. Kort, A. Kostolias, C. Sullivan, and R. A. Lobo, “Chemerin as a marker of body fat and insulin resistance in women with polycystic ovary syndrome,” Gynecol. Endocrinol. 31(2), 152–155 (2015). [CrossRef]   [PubMed]  

12. B. K. Tan, J. Chen, S. Farhatullah, R. Adya, J. Kaur, D. Heutling, K. C. Lewandowski, J. P. O’Hare, H. Lehnert, and H. S. Randeva, “Insulin and metformin regulate circulating and adipose tissue chemerin,” Diabetes 58(9), 1971–1977 (2009). [CrossRef]   [PubMed]  

13. S. S. Chang, D. Eisenberg, L. Zhao, C. Adams, R. Leib, J. Morser, and L. Leung, “Chemerin activation in human obesity,” Obesity (Silver Spring) 24(7), 1522–1529 (2016). [CrossRef]   [PubMed]  

14. M. Sargent, “Guide to achieving reliable quantitative LC-MS measurements,” RSC Analytical Methods Committee, (2013).

15. S. K. Paidi, S. Siddhanta, R. Strouse, J. B. McGivney, C. Larkin, and I. Barman, “Rapid Identification of Biotherapeutics with Label-Free Raman Spectroscopy,” Anal. Chem. 88(8), 4361–4368 (2016). [CrossRef]   [PubMed]  

16. Z. Huang, S. Siddhanta, C. Zhang, T. Kickler, G. Zheng, and I. Barman, “Painting and heating: A nonconventional, scalable route to sensitive biomolecular analysis with plasmon-enhanced spectroscopy,” J. Raman Spectrosc. 48(10), 1365–1374 (2017). [CrossRef]  

17. N. Feliu, M. Hassan, E. Garcia Rico, D. Cui, W. Parak, and R. Alvarez-Puebla, “SERS Quantification and Characterization of Proteins and Other Biomolecules,” Langmuir 33(38), 9711–9730 (2017). [CrossRef]   [PubMed]  

18. A. Khetani, J. Riordon, V. Tiwari, A. Momenpour, M. Godin, and H. Anis, “Hollow core photonic crystal fiber as a reusable Raman biosensor,” Opt. Express 21(10), 12340–12350 (2013). [CrossRef]   [PubMed]  

19. A. Momenpour, V. S. Tiwari, V. L. Trudeau, and H. Anis, “Surface-Enhanced Raman Scattering Spectroscopy for the Detection of Glutamate and γ-Aminobutyric Acid in Serum by Partial Least Squares Analysis,” IEEE Photonics J. 7(3), 1–16 (2015).

20. N. Leopold and B. Lendl, “A new method for fast preparation of highly surface-enhanced Raman scattering (SERS) active silver colloids at room temperature by reduction of silver nitrate with hydroxylamine hydrochloride,” J. Phys. Chem. B 107(24), 5723–5727 (2003). [CrossRef]  

21. K. H. Esbensen, Multivariate data analysis - in practice (CAMO, 2004).

22. G. Basar, U. Parlatan, S. Seninak, T. Gunel, A. Benian, and I. Kalelioglu, “Investigation of preeclampsia using Raman spectroscopy,” Spectrosc. Int. J. 27(4), 239–252 (2012). [CrossRef]  

23. G. Zhu, X. Zhu, Q. Fan, and X. Wan, “Raman spectra of amino acids and their aqueous solutions,” Spectrochim. Acta A Mol. Biomol. Spectrosc. 78(3), 1187–1195 (2011). [CrossRef]   [PubMed]  

24. S. Stewart and P. M. Fredericks, “Surface-enhanced Raman spectroscopy of amino acids adsorbed on an electrochemically prepared silver surface,” Spectrochim. Acta A Mol. Biomol. Spectrosc. 55(7–8), 1641–1660 (1999). [CrossRef]  

25. L. Simons, A. Pohjavirta, O. Nevanlinna, and P. Suomela, Laser Raman spectroscopy of amino acids, oligopeptides, polypeptides and enzymes (Societas Scientiarum Fennica, 1972).

26. S. A. Asher, M. Ludwig, and C. R. Johnson, “UV resonance Raman excitation profiles of the aromatic amino acids,” J. Am. Chem. Soc. 108(12), 3186–3197 (1986). [CrossRef]  

27. Z. Movasaghi, S. Rehman, and I. U. Rehman, “Raman spectroscopy of biological tissues,” Appl. Spectrosc. Rev. 42(5), 493–541 (2007). [CrossRef]  

28. P. F. Facanha Filho, P. T. C. Freire, K. C. V. Lima, J. Mendes Filho, F. E. A. Melo, and P. S. Pizani, “High temperature Raman spectra of l-leucine crystals,” Braz. J. Phys. 38(1), 131–137 (2008). [CrossRef]  

29. J. R. Govani, W. G. Durrer, M. Manciu, C. Botez, and F. S. Manciu, “Spectroscopic study of L-arginine interactions with potassium dihydrogen phosphate crystals,” J. Mater. Res. 24(7), 2316–2320 (2009). [CrossRef]  

30. Z. Huang, A. McWilliams, H. Lui, D. I. McLean, S. Lam, and H. Zeng, “Near-infrared Raman spectroscopy for optical diagnosis of lung cancer,” Int. J. Cancer 107(6), 1047–1052 (2003). [CrossRef]   [PubMed]  

31. K. A. Esmonde-White, J. Sottnik, M. Morris, and E. Keller, “Raman spectroscopy of bone metastasis,” Proc. SPIE 8207, 82076P (2012). [CrossRef]  

32. J. Jehlicka, P. Vítek, and H. G. M. Edwards, “Raman spectra of organic acids obtained using a portable instrument at −5 ◦C in a mountain area at 2000 m above sea level,” J. Raman Spectrosc. 41, 440–444 (2010).

33. K. A. Esmonde-White, G. S. Mandair, F. W. L. Esmonde-White, F. Raaii, B. J. Roessler, and M. D. Morris, “Osteoarthritis screening using Raman spectroscopy of dried human synovial fluid drops,” Proc. SPIE 7166, 71660J (2009). [CrossRef]  

34. Y. Li, J. Pan, G. Chen, C. Li, S. Lin, Y. Shao, S. Feng, Z. Huang, S. Xie, H. Zeng, and R. Chen, “Micro-Raman spectroscopy study of cancerous and normal nasopharyngeal tissues,” J. Biomed. Opt. 18(2), 027003 (2013). [CrossRef]   [PubMed]  

35. S. Olsztynska, N. Dupuy, L. Vrielynck, and M. Komorowska, “Water evaporation analysis of L-phenylalanine from initial aqueous solutions to powder state by vibrational spectroscopy,” Appl. Spectrosc. 60(9), 1040–1053 (2006). [CrossRef]   [PubMed]  

36. J. Guicheteau, L. Argue, A. Hyre, M. Jacobson, and S. D. Christesen, “Raman and surface-enhanced Raman spectroscopy of amino acids and nucleotide bases for target bacterial vibrational mode identification,” Proc. SPIE 6218, 62180O (2006). [CrossRef]  

37. J. De Gelder, K. De Gussem, P. Vandenabeele, and L. Moens, “Reference database of Raman spectra of biological molecules,” J. Raman Spectrosc. 38(9), 1133–1147 (2007). [CrossRef]  

38. J. T. L. Navarrete, V. Hernández, and F. J. Ramírez, “Vibrational study of aspartic acid and glutamic acid dipeptides,” J. Mol. Struct. 348, 249–252 (1995). [CrossRef]  

39. K. A. Esmonde-White, S. V. Le Clair, B. J. Roessler, and M. D. Morris, “Effect of conformation and drop properties on surface-enhanced Raman spectroscopy of dried biopolymer drops,” Appl. Spectrosc. 62(5), 503–511 (2008). [CrossRef]   [PubMed]  

40. K. A. Esmonde-White, G. S. Mandair, F. Raaii, J. A. Jacobson, B. S. Miller, A. G. Urquhart, B. J. Roessler, and M. D. Morris, “Raman spectroscopy of synovial fluid as a tool for diagnosing osteoarthritis,” J. Biomed. Opt. 14(3), 034013 (2009). [CrossRef]   [PubMed]