Label-Free Detection of Human Serum Using Surface-Enhanced Raman Spectroscopy Based on Highly Branched Gold Nanoparticle Substrates for Discrimination of Non-Small Cell Lung Cancer

Institute of Translational Medicine, Medical College, Yangzhou University, Yangzhou 225001, China Jiangsu Co-Innovation Center for Prevention and Control of Important Animal Infectious Diseases and Zoonoses, Yangzhou University, Yangzhou 225009, China Intensive Care Unit, Wuxi +ird People’s Hospital, Wuxi 214000, China Jiangsu Key Laboratory of Experimental & Translational Non-Coding RNA Research, Medical College, Yangzhou University, Yangzhou 225001, China Transformative Otology and Neuroscience Center, College of Special Education, Binzhou Medical University, Yantai 264003, China


Introduction
Lung cancer is presently the most commonly diagnosed metastatic tumor and a leading cause of cancer-related deaths in the world.Approximately 85% of lung cancers are non-small cell lung cancer (NSCLC), which mainly includes lung adenocarcinoma, lung squamous carcinoma, and large cell lung cancer [1,2].e high mortality is mainly due to its aggressiveness, and the fact is that most NSCLC are generally detected at the late stage of inoperable disease.Early detection of NSCLC can dramatically improve the outcome of fatal disease.NSCLC has no visible symptoms at the early stages, and the widespread medical screening is serving as a significant modality for early detection of NSCLC.For the best use of public health resources, it is extremely important to diagnose as much histological type of NSCLC as possible in every population-based cancer screening program [3].At present, several conventional methods involving clinical examination, white light endoscopy, and histopathological analysis are not only costly, time-consuming, and invasive, but also require sophisticated instruments [4].
ese drawbacks hindered the further clinical application in the diagnosis and discrimination of cancer.erefore, developing novel diagnosis methods for NSCLC is essential.
Surface-enhanced Raman scattering (SERS) is a vibrational spectroscopy technique based on the intensity amplification of Raman scattering by metallic nanostructures with suitable plasmonic characteristics [5].SERS can reflect the molecular structures and changes of samples and is considered as the molecular "fingerprint" [5].Compared with other techniques, SERS has advantages such as being noninvasive, high spatial resolution, weak water scattering, and no sample preparation [6,7].In addition, Raman scattering can be enhanced by as much as 6 to 14 orders of magnitude when the analyses are adsorbed onto the roughened surfaces of the metallic nanostructure [8].So, it is very suitable for studying biological sample containing water.Because of extremely rich biochemical and biological information of biofluids, SERS spectral analysis of biofluids for diagnosis and screening of disease may provide a feasible way for cancer diagnosis.Attempts were made to identify cancer-related chemical and bimolecular changes in urine, serum, cervical fluid, plasma, and saliva using this technique [9][10][11].Li et al. investigated the use of SERS spectra of serum samples from healthy individuals, atrophic gastritis patients, pre-and postoperation gastric cancer patients to diagnose gastric diseases.eir experiment results demonstrated that spectroscopy can effectively record the changes occurring in serum caused by variations in body circumstances [12].González-Solís et al. discriminated the serum obtained from three types of leukemia patients by using Raman spectroscopy, and the possibility for distinguishing different types of cancer by this technique has been preliminarily demonstrated [13].
SERS effect of nanomaterials is extremely important when detecting biofluids in which some components were rare.Nanomaterials with different sizes, shapes, and compositions have been prepared for SERS detection [14].Recently, highly branched gold nanoparticles (HGNPs) with a central core and external multiple sharp tips have received a great deal of attention for its unique optical properties.
e optical properties of HGNPs may be controlled by changing the size, shape, and number and length of sharp branches.It demonstrated that HGNPs have higher sensitivity in localized surface plasmon resonance (LSPR) as compared with other nanoparticles [15,16].HGNPs have a roughened surface determined by many branches and cavities, which act as potential "hot spots" for enhancing the electromagnetic field around the nanoparticle [17].Due to its excellent SERS effect, SERS based on HGNPs has been applied to the biomolecular detection, medical imaging, and diagnosis of various cancers [18,19].In order to get greater sensitivity and signal enhancement, a variety of nanoparticles have been assembled into solid substrates (such as graphite, silicon, and ITO glass) by template method, self-assembly, and electrochemical deposition.Rodrı´guez-Lorenzo et al. reported that a uniform and reproducible gold nanostars substrate could generate a high enhancement factor (EF) when using 1,5-naphtalenedithiol as an analyte [20].According to the calculations performed by Indrasekara et al.SERS EF of gold nanostars substrate could approach 9 orders of magnitude, which was attributed to the high efficacy coupled "hot spots" produced between the nanoparticles and the dielectric silicon substrate [21].
In this study, SERS spectroscopy based on highly branched gold nanoparticle (HGNP) substrate was used to investigate the spectral characteristics of human serum for the purpose of diagnosing NSCLC (Scheme 1).Firstly, a onestep method for synthesizing novel anisotropic HGNPs with multiple tips and dendritic structures was proposed by using dopamine.Briefly, HGNPs were synthesized by combining HAuCl 4 with dopamine hydrochloride in a rapid one-step reaction.en, HGNPs were assembled onto the surface of (3-aminopropyl) triethoxysilane-(APTES-) functionalized silicon wafer by the sedimentary self-assembly method.e SERS stability, homogeneity, and sensitivity of HGNP substrates were evaluated.Subsequently, we utilized the asprepared HGNPs substrates to detect the SERS spectra of serum from healthy subjects and three types of NSCLC (lung adenocarcinoma, lung squamous carcinoma, and large cell lung cancer).
e difference spectra were employed to identify the differences of biochemical components among different groups.In addition, SERS spectra were analyzed by using principal component analysis (PCA), demonstrating that SERS spectroscopy was a sensitive analysis technology for classification and discrimination of the different types of NSCLC.  1.Each blood sample was approximately 5 mL in volume and was extracted intravenously from volunteers before breakfast to avoid the interference of food.en, the blood sample was deposited at 4 °C for 4 h and was centrifuged at 3000 rpm for 10 min in order to remove the blood cells, brinogen, and platelets.After the centrifugation, the serum was obtained and stored at −80 °C for SERS measurement.

Preparation of Highly Branched Gold Nanoparticles.
For the preparation of HGNPs, 0.4 mL HAuCl 4 solution (25 mM) was mixed with 10 mL of ultrapure water under vigorous stirring.en, 2 mL dopamine hydrochloride solution (53 mM) was added.e mixed solution was heated slowly until the temperature reached 60 °C.e temperature was xed at 60 °C and we let it continue to react for about 30 minutes to synthesize HGNPs.

Assembly of HGNPs on APTES-Functionalized Silicon
Wafer. e fabrication of HGNPs substrates is displayed in Scheme 1. e silicon wafer was washed with aqua regia and rinsed with ultrapure water for three times.e wafer was further cleaned in ethanol for three times and dried at 80 °C for 2 h in an air oven.After drying, the silicon wafer was immersed in 1% (v/v) APTES of anhydrous ethanol for 24 h to form a self-assembled layer followed by thorough rinsing with ethanol and air drying.Subsequently, the APTESfunctionalized silicon wafer was placed horizontally in the HGNP colloidal solution for 12 h.HGNPs were deposited onto the surfaces of the APTES-functionalized silicon wafer to obtain the HGNP layer.
e thiol moiety of APTES covalently binds to the substrate while its pendant primary amine group covalently binds to the nanoparticles.us, the HGNP substrates were obtained.
2.5.SERS Measurement.SERS spectral data was collected from an inverted microscope (Renishaw) in the 600-1800 cm −1 range with a 785-nm He-Ne laser maintained.e laser intensity was 5 mW and the exposure time was 10 s. e laser was focused onto the sample using a 50× long working distance objective.For SERS measurement of human serum, 100 mL serum sample obtained from each volunteer participating in this study was dropped directly on the HGNPs substrate, and then detected 10 times by Raman spectroscopy to obtain their own ngerprints.Subsequently, the measured SERS spectra of 30 volunteers were used to obtain an average SERS spectrum of serum for each histological type of NSCLC.

Characterization.
e ultraviolet-visible-near infrared (UV-vis-NIR) absorption spectrum was measured using a Cary UV-5000 spectrometer (Agilent).Transmission electron microscope (TEM) imaging was performed using a Tecnai 12 transmission electron microscope operating at  an accelerating voltage of 60 kV (Philips).Scanning electron microscopy (SEM) images were acquired using a S-4800 II field-emission scanning electron microscope (FESEM) operating at 3.0 kV (Hitachi).To observe the crystalline structure of the particles, high-resolution TEM images (HRTEM) and selected area electron diffraction (SAED) images were captured by a Tecnai G2 F30 S-Twin TEM at 200 KV (FEI).

Characterization of HGNPs.
e morphology and structure of the HGNPs were characterized using SEM and TEM.As shown in Figure 1(a), HGNPs were well dispersed with particle sizes of ∼600 nm and these nanoparticles exhibited multiple tips with dendritic structures.e highmagnification SEM image and TEM image further revealed that each HGNP consisted of a solid gold core with many sharp, irregular tips (Figures 1(b) and 1(c)).e crystalline structure of the HGNPs was observed by HRTEM. Figure 1(d) reveals that the interplanar spacing of the tips was measured to be 0.24 nm, indicating that the HGNPs grew preferentially on {111} planes.To further understand the crystal structure of HGNPs, a SAED image was also taken (Figure 1(e)).HGNPs grow in random orientations like {111}, {200}, {220}, and {311} crystalline facet.e results indicated that as-prepared HGNPs were pure and well crystallized gold.Figure 1(f ) shows that the UV-vis-NIR absorption spectrum of HGNPs had a strong absorption peak at 475 nm.During the reaction process, the color of the solution turned from light yellow to dark green and finally to orange-red within 30 minutes.erefore, HGNPs were synthesized using this simple, fast, and largescale method.
Besides, Raman spectra of 4-MBA and 4-MBA labeled HGNPs are recorded in Figure 1(g).4-MBA was chosen as an analyte, which had two SERS characteristic peaks at 1080 cm −1 and 1587 cm −1 [22].We observed the strong SERS signals from 4-MBA-labeled HGNPs, indicating that HGNPs had a strong surface enhancement effect.e significant enhancement might be due to the large amount of sharp tips on the surface of HGNPs that potentially act as "hot spots."Besides, the enhancement factor (EF) of HGNPs was calculated by using the expression EF � (I SERS /C SERS )/(I RS /C RS ), where I SERS corresponds to the SERS intensity obtained for the HGNPs colloidal dispersion at a certain concentration C SERS of the analyte, and I RS corresponds to the Raman intensity obtained under non-SERS conditions at an analyte concentration C RS .During the experimental process, the HGNP colloid was mixed with the same volume of 4-MBA solution (2 × 10 −5 M) for 4 h, giving a final concentration with 10 −5 M. When C SERS was 1 × 10 −5 M and C RS was 1 × 10 −2 M, EF was calculated to be 5.7 × 10 5 , which was higher than the gold nanostars (2 × 10 5 ) reported by Nalbant Esenturk [23].

Fabrication of HGNPs
Substrates.HGNP substrates were fabricated by the electrostatic self-assembly method, which does not need complicated equipment and is preponderant in commercial manufacture.
e aminosilane groups of APTES can boost the adsorption of nanoparticles with inherent negative charge, so it was employed to functionalize the silicon wafers assisted by electrostatic interaction for adsorption of HGNPs layers [24].Here, HGNPs were successful in depositing and adsorbing on the surface of the APTES-functionalized silicon wafer by directly placing the wafers horizontally into the HGNPs colloidal dispersion.Figure 2(a) shows the SEM image of the HGNP substrate.HGNPs had slight aggregation and uniformly distribution on the surface of silicon wafer.To characterize the uniformity of the surface SERS signal, a SERS mapping experiment was carried out after the surface was adsorbed with 4-MBA (1 × 10 −5 M). e scan area was 25 × 25 mm 2 , the step size was 2 μm, the laser power was 5 mW, and the acquisition time at each point was 10 s. e SERS intensity of the 1080-cm −1 peak was mapped at each grid point on the substrate.As shown in Figure 2(b), SERS mapping was used to display the intensity of the 1080-cm −1 characteristic peak at each grid point according to a color scheme ranging from blue (lowest intensity) through green, yellow, orange, and red (highest intensity).e color of SERS mapping is basically the same, indicating that the HGNP substrate has a very uniform SERS enhancement effect over the whole surface, although some aggregates can still be found at some spots.
We further investigated the stability of the SERS signals of 4-MBA-labeled HGNP substrates.HGNPs substrates were stored for several days at room temperature and the corresponding SERS spectra were measured (Figure 3(a)).After 15 days, compared with the freshly prepared substrates, SERS intensity of the characteristic peak at 1080 cm −1 was reduced by only 8.7% in Figure 3(b).e result showed that the substrates displayed a very stable SERS enhancement effect.In order to study the detection capability of the substrates, 4-MBA at different concentrations were introduced to the HGNP substrates.Figure 3(c) shows the SERS spectra of 4-MBA at different concentrations absorbed on the HGNP substrates.e lower concentration of 4-MBA solutions showed a silence of SERS signal, and thus the 5 × 10 −10 M concentration was taken as the limit of detection (LOD) for the 4-MBA.
e relationship between the intensity of characteristic peak at 1080 cm −1 and the concentration of 4-MBA solution was analyzed (Figure 3(d)).
e SERS intensities at 1080 cm −1 increased constantly with the increase of 4-MBA concentration.Raman intensity of 4-MBA solution at low concentrations (from 5 × 10 −10 to 1 × 10 −9 M) did not change significantly, which may be caused by the systematic error during the experiment.ere were linear relationships between the peak intensity at 1080 cm −1 and the concentrations of the 4-MBA solution from 1 × 10 −8 M to 1 × 10 −4 M. Its regression equation can be expressed as y � 15658 lgC + 133289 (C being the concentration of the 4-MBA).erefore, by careful assessment of the calibration curve, we got LOD and limits of quantification (LOQ) of 5 × 10 −10 M and 1 × 10 −8 M, respectively.Besides the stability and sensitivity, reproducibility of the substrates was another important factor for SERS detection.Journal of Chemistry ree substrates, made at di erent times, were measured and are shown in Figure 3(e).e deviation of peak height at 1080 cm −1 was 6.4%, which supported the good reproducibility of the SERS substrates.In order to evaluate the repeatability of the SERS spectrum of serum on the substrate, the serum sample from the same large cell lung cancer patient was dropped directly onto the surfaces of ve different HGNP substrates, and then subjected to Raman spectroscopy.Figure 3(f ) shows the repeatability of the SERS spectrum of serum on the substrate.
ese SERS spectra looked very similar, and there were no obvious di erences in characteristic peak positions and peak intensities.It demonstrated that HGNP substrates with good reproducibility were very suitable for serum SERS label-free detection.In Figure 3(g), the unlabeled HGNP substrates had no obvious characteristic peaks, and it suggested that the substrates had a clean background in the SERS spectrum.Taken together, the clean HGNP substrates with ultrahigh sensitivity, homogeneous SERS activity, and high reproducibility are prepared and can be used as a promising candidate for labelfree detection of biological samples.

SERS Spectroscopy of Human Serum.
As a label-free detection method, the serum was dropped directly on the HGNP substrate to obtain the rich molecular and structural information.e SERS spectrum of serum for each type of NSCLC was obtained by averaging 30 subjects.Figure 4 shows the average SERS spectra of serum obtained from large cell lung cancer patients, lung adenocarcinoma patients, lung squamous carcinoma patients, and healthy subjects.ere are many characteristic vibrational bands which may be attributed to proteins, nucleic acids, lipids, and carbohydrates, such as the peaks at 728 cm −1 (C-C stretching, proline), 880 cm −1 (tryptophan, δ (ring)), 960 cm −1 (hydroxyapatite, carotenoid, and cholesterol), 1005 cm −1 (phenylalanine (proteins); proteins symmetric ring breathing of phenylalanine), 1079 cm stretching mode of phospholipids), 1120 cm −1 (the strong C-O band of ribose), 1171 cm −1 ((CH) phenylalanine, tyrosine), 1318 cm −1 (G ring breathing modes of the DNA/RNA), 1379 cm −1 (δCH 3 symmetric (lipid assignment)), 1446 cm −1 (CH 2 bending mode of proteins and lipids), and 1583 cm −1 (δ(C C), phenylalanine).Although the di erent groups have similar SERS spectral pro les, such as peak positions and spectral bandwidths, some little di erences can be observed.en, in order to discriminate among di erent types of NSCLC, the di erences between SERS spectra of four groups are compared, and the results with details are as follows: For the convenient comparison, the intensity of average SERS spectrum for serum of each group was normalized to a range from 0 to 1 (Figure 5).e intensity of peak at 1270 cm −1 (typical phospholipids, C C groups in fatty acids, amide III band in proteins) was the reference for normalization because of the narrow peak width, high intensity, and the nonsensitivity to environmental change.To better understand the SERS spectra di erences, the di erence spectra were computed, which can better interpret the spectral di erences and gain insight into the biochemical variation.
us, the di erence spectra were computed by subtracting the spectrum of healthy subjects from the spectrum of large cell lung cancer (healthy subjects-large cell lung cancer), the spectrum of healthy subjects from the spectrum of lung squamous carcinoma (healthy subjects-lung squamous carcinoma), the spectrum of healthy subjects from the spectrum of lung adenocarcinoma (healthy subjects-lung adenocarcinoma), the spectrum of lung adenocarcinoma from the spectrum of lung squamous carcinoma (lung adenocarcinoma-lung squamous carcinoma), the spectrum of lung adenocarcinoma from the spectrum of large cell lung cancer (lung adenocarcinoma-large cell lung cancer), and the spectrum of lung squamous carcinoma from the spectrum of large cell lung cancer (lung squamous carcinoma-large cell lung cancer).According to reference to the Raman peak assignments in Table 2, we could assign the characteristic peaks in six difference spectra.As shown in Figure 6(a) (healthy subjects-large cell lung cancer), we could learn that the positive peaks at 855 cm −1 (proline, hydroxyproline, and tyrosine C-C stretching), 921 cm −1 (C-C stretch of proline ring/glucose/lactic acid; C-C, praline ring (collagen assignment)), 960 cm −1 , 1079 cm −1 , 1120 cm −1 (the strong C-O band of ribose), 1255 cm −1 (lipids), and 1565 cm −1 (tryptophan) were from the spectrum of healthy subjects, indicating the relative intensity of these peaks which were mainly the vibration bands of some amino acid and lipids were higher in the spectrum of serum obtained from healthy subjects when compared with that of large cell lung cancer patients, whereas the negative bands for Figure 6(a) at 642 cm −1 (C-C twisting mode of tyrosine), 728 cm −1 , 787 cm −1 (uracil-based ring breathing mode), 829 cm −1 (O-P-O stretching DNA/RNA; ν 2 PO 2− stretch of nucleic acids), 1005 cm −1 , 1181 cm −1 , 1318 cm −1 , 1379 cm −1 , 1425 cm −1 (G, A (DNA, RNA)), 1488 cm −1 (guanine (N 7 )), 1583 cm −1 , and 1673 cm −1 (amide I) were from the spectrum of serum obtained from large cell lung cancer patients.ese peaks were mainly generated by the vibration peaks of DNA, RNA, and proteins.In Figure 6(b) (healthy subjects-lung squamous carcinoma), the positive bands at 1017 cm −1 (carbohydrates peak), 1255 cm −1 (lipids), 1446 cm −1 , 1565 cm −1 , and 1673 cm −1 were from healthy subjects.e majority of these peaks could be assigned to carbohydrates and lipids.e negative bands, 642 cm −1 , 728 cm −1 , 880 cm −1 , 960 cm −1 , 1031 cm −1 , 1379 cm −1 , 1488 cm −1 , and 1601 cm −1 (amide I band of proteins) from the spectrum of lung squamous carcinoma and could be similarly assigned as nucleic acid and proteins.In Figure 6(c), compared with the lung adenocarcinoma, the intensities of healthy subjects at 855 cm −1 , 921 cm −1 , 1031 cm −1 , 1079 cm −1 , 1120 cm −1 , 1255 cm −1 , 1446 cm −1 , 1565 cm −1 , and 1673 cm −1 increased, which were assigned to the molecular structures in the amino acid and lipids, but the intensities of peaks at 642 cm −1 , 728 cm −1 , 817 cm −1 (C-C stretching (protein)), 960 cm −1 , 1425 cm −1 , 1534 cm −1 (amide carbonyl group vibrations and aromatic hydrogens), and 1583 cm −1 decreased, which were mainly assigned to the relative molecular structures of nucleic acids and protein.
e distinctive SERS features and intensity differences for NSCLC and normal serum could reflect their molecular changes in quantity or structure.
e cell-free nucleic acids are signi cantly increased, which originated from the regeneration, necrosis, and release of intact cells in the bloodstream and their subsequent lysis. is result was consistent with the previous reports [38].Beyond that, the amounts of some amino acids and lipids in the serums from the healthy subjects were more than those of NSCLC patients.Meanwhile, there are also di erences among the spectrums of serum obtained from three di erent types of NSCLC, including the bands at 728 cm −1 , 880 cm   Journal of Chemistry 921 cm −1 , 960 cm −1 , 1318 cm −1 , 1446 cm −1 , 1583 cm −1 , and so on.
To further quantify the differences among the four groups based on the normalized mean SERS spectra of each NSCLC type, we compared the mean relative intensity at four different Raman peaks, including 728 cm −1 , 1079 cm −1 , 1318 cm −1 , and 1466 cm −1 .From Figure 7, healthy subjects showed higher relative intensity of Raman peaks at 1079 cm −1 and 1466 cm −1 than the three types of NSCLC.However, compared with healthy subjects, the relative intensity of Raman peaks at 728 cm −1 and 1318 cm −1 for three types of NSCLC is higher.Furthermore, there are also many differences among the three types of NSCLC.As to the relative intensity of Raman peaks at 728 cm −1 and 1318 cm −1 , large cell lung cancer was higher than lung squamous carcinoma and lung adenocarcinoma.e relative intensity of Raman peak at 1079 cm −1 for lung squamous carcinoma was significantly higher than others, and there was no significance at 1466 cm −1 among the three types of NSCLC.Based on the above analysis, we could learn that the result is consistent with difference spectra analysis.

PCA Analysis.
In order to determine if SERS spectroscopy could be used to distinguish the different types of NSCLC, principal component analysis (PCA) was applied.PCA is a multivariate method widely used in spectral analysis, which defines a new dimensional space in which the major variance in the original dataset can be captured and represented by only a few principal component (PC) variables.Scores of PCs are one of the parameters widely used for classification [39].In Figure 8, using principal component scores PC1 (66.7% variation) and PC2 (15.2% variation), the scatter plots of SERS spectra for each type of NSCLC about twenty subjects were projected into the twodimensional images, which could be used to classify the types of NSCLC.In the PCA score plot, the data points of healthy subjects and three types of NSCLC are clustered into four separate groups.e plots group of healthy subjects was clearly separable from the groups of large cell lung cancer, lung squamous carcinoma group, and lung adenocarcinoma group because of the differences in serum composition obtained from healthy subjects and NSCLC patients.ree kinds of cancer groups have more nucleic acids and proteins, Journal of Chemistry but have less amino acids and lipids than healthy subjects.However, the plots group of lung adenocarcinoma are adjacent to the lung squamous carcinoma groups and have a small overlap with the large cell lung cancer groups.is was mainly due to similar biochemical composition among NSCLC.PCA gives more intuitive results than direct analysis spectral di erence among di erent types of NSCLC.e results suggest that PCA can easily classify the types of NSCLC based on their speci c spectral features.

Conclusions
e serum samples are easily collected in clinical examination, and mostly re ect some vital subtle change caused by tumors in the metabolism environment, such as amino acid metabolism, miRNA expression, and biomarkers generation.e concentration of nucleic acids and the composition of proteins from the serum samples of NSCLC patients are di erent from the normal samples, which are believed to originate from apoptosis, tumor necrosis, and associated metabolites.erefore, serum detection is appropriate for NSCLC diagnosis.
In our study, the SERS diagnostic approach was explored in terms of its ability to discriminate the NSCLC types based on HGNP substrates.HGNPs were synthesized by a simple, fast, and large-scale method.Via electrostatic action, HGNPs were assembled on the surface of a silicon wafer, and these substrates exhibited very good reproducibility, homogeneous SERS activity, and a high SERS enhancement e ect.HGNP substrates enabled us to successfully detect the SERS spectra of serum obtained from healthy subjects and three di erent types of NSCLC patients (lung adenocarcinoma, lung squamous carcinoma, and large cell lung cancer).e di erent spectra analysis showed that there were di erences in the intensity and position of SERS peaks assigned to various biochemical changes in the serum among the four groups.Compared with the SERS spectra of the normal serum, serum obtained from NSCLC have more nucleic acids and proteins, but have less amino acids and lipids.
ere are also signi cant di erences in the biochemical components among the three di erent types of NSCLC.e SERS of human serum combined with PCA achieved a successful segregation of di erent types of NSCLC.
is exploratory work demonstrates that SERS serum analysis technique has potential for improving NSCLC detection and screening.

3 Scheme 1 :
Scheme 1: Illustration of the fabrication procedure of the HGNP substrate and SERS is used for diagnosis and discrimination of the NSCLC.

Figure 1 :
Figure 1: (a) Low and (b) high magnification SEM images of the HGNPs.(c) TEM images and (d) high-resolution TEM images of the HGNPs.(e) SAED pattern and (f ) UV-vis-NIR absorption spectrum of the HGNPs.(g) SERS spectra of (black curve) 4-MBA and (red curve) HGNP colloidal dispersion mixed with 4-MBA under 785 nm excitation.

Figure 2 :
Figure 2: (a) SEM images of the HGNPs substrate.(b) SERS mapping of 4-MBA measured at 1080 cm −1 using the HGNP substrate.e acquisition time was 10 s, and the laser power was 5 mW.

Figure 4 :Figure 5 :
Figure 4: Average SERS spectra of serum obtained from (a) large cell lung cancer patients, (b) lung adenocarcinoma patients, (c) lung squamous carcinoma patients, and (d) healthy subjects.

Figure 6 :
Figure 6: Di erence spectra among the four groups were computed: (a) healthy subjects-large cell lung cancer, (b) healthy subjects-lung squamous carcinoma, (c) healthy subjects-lung adenocarcinoma, (d) lung adenocarcinoma-lung squamous carcinoma, (e) lung adenocarcinoma-large cell lung cancer, and (f ) lung squamous carcinoma-large cell lung cancer.

Figure 8 :
Figure8: PCA score plots of SERS spectra for serum obtained from large cell lung cancer patients, lung squamous carcinoma patients, lung adenocarcinoma patients, and healthy subjects.Each type of NSCLC was coded with a di erent color.

Table 1 :
Demographics of study population.

Table 2 :
Peak assignments for SERS spectra of serum.