A novel method for identifying and distinguishing Cryptococcus neoformans and Cryptococcus gattii by surface-enhanced Raman scattering using positively charged silver nanoparticles

There are approximately 1 million cryptococcal infections per year among HIV+ individuals, resulting in nearly 625,000 deaths. Cryptococcus neoformans and Cryptococcus gattii are the two most common species that cause human cryptococcosis. These two species of Cryptococcus have differences in pathogenicity, diagnosis, and treatment. Cryptococcal infections are usually difficult to identify because of their slow growth in vitro. In addition, the long detection cycle of Cryptococcus in clinical specimens makes the diagnosis of Cryptococcal infections difficult. Here, we used positively charged silver nanoparticles (AgNPs+) as a substrate to distinguish between C. neoformans and C. gattii in clinical specimens directly via surface-enhanced Raman scattering (SERS) and spectral analysis. The AgNPs+ self-assembled on the surface of the fungal cell wall via electrostatic aggregation, leading to enhanced SERS signals that were better than the standard substrate negatively charged silver nanoparticles (AgNPs). The SERS spectra could also be used as a sample database in the multivariate analysis via orthogonal partial least-squares discriminant analysis. This novel SERS detection method can clearly distinguish between the two Cryptococcus species using principal component analysis. The accuracy of the training data and test data was 100% after a tenfold crossover validation.

Scientific RepoRtS | (2020) 10:12480 | https://doi.org/10.1038/s41598-020-68978-0 www.nature.com/scientificreports/ initially called C. neoformans var. gattii because its characteristics were very similar to those of C. neoformans. In 2002, Kwon-Chung et al. 4 found that there were some differences in phenotype, biology, and genetic taxonomy between C. neoformans and C. neoformans var. gattii ; thus, C. gattii was reclassified as a single species. Traditional detection methods for cryptococcus, such as ink staining and antigen kit testing, are relatively fast but cannot distinguish between C. neoformans and C. gattii 2,5 . Polymerase chain reaction (PCR), matrixassisted laser desorption/ionization time-of-flight mass spectrometry (MALDI TOF MS), gene chips, and second-generation sequencing technology have all been used for clinical identification of Cryptococcus and can distinguish between C. neoformans and C. gattii; however, the equipment costs are very high [6][7][8][9][10] . The surfaceenhanced Raman scattering (SERS) effect refers to the Raman scattering signal of adsorbed molecules due to the enhancement of an electromagnetic field on or near the sample surface, or in the excitation region of specific metal conductors or sol that have been specially prepared 11 . SERS offers several advantages for fungal identification that can make up for the deficiencies of the above detection methods. For example, SERS spectra offer spectral fingerprints unique to the constituent molecules. SERS offers qualitative and quantitative detection of the analyte because it has high specificity 12 . The SERS spectral bands are narrow with high resolution of spectral peaks, thus minimizing fluorescence background interference 13 . The characteristic spectra facilitate simultaneous detection of multiple components. Surface-enhanced Raman spectroscopy offers biological testing and is relatively inexpensive and portable 14 .
It is surprising that there are no reports using SERS for C. neoformans and C. gattii detection. Although SERS can identify pathogens, further development and improvement is needed before clinical application. This is particularly true for cryptococcus-a special fungal pathogen that contains a capsule 15,16 .
Silver nanoparticles (AgNPs) are widely used as standard SERS substrates 17,18 . The most common approach uses 1% sodium citrate mixed with 1.69 mM AgNO 3 with heating at 100 °C 19 . The enhancement and reproducibility with anionic analytes are low due to electrostatic repulsion from the sodium citrate on the AgNPs 20,21 . We first measured the zeta potential of C. neoformans and C. gattii and found that both had negative charges on their surfaces. We previously prepared positively charged silver nanoparticles (AgNPs + ) with a NaBH 4 reducing agent; the AgNPs + could adsorb onto Cryptococcus via electrostatic interactions but the stability of this substrate was relatively poor 22 . To improve the stability of the substrate, AgNPs + were synthesized with cetyl trimethylammonium bromide (CTAB) as a cationic surfactant 23 . Herein, we report a rapid detection method for Cryptococcus by SERS using positively charged silver nanoparticles for the first time. After obtaining Cryptococcus spectral data, we used orthogonal partial least-squares discriminant analysis (OPLS-DA), an orthogonal, partial furthersquares discriminant analysis algorithm, to analyze the spectral data.

Results
Structural characterization of nanomaterials. Figure 1a,b show that C. neoformans and C. gattii are negatively charged. Figure 2a shows a SEM image of the AgNPs. The size and distribution of the AgNPs are homogeneous, and rods of nanoscale silver were produced. Figure 2b shows a SEM image of the AgNPs + and that the AgNPs + are spherical. Figure 2c shows that the AgNPs zeta potential is negative and the AgNPs + zeta potential is positive. The AgNPs particle size is 78.8 nm and the AgNPs + particle size is 105.7 nm, as shown in Fig. 2d. The UV-absorption spectra of the AgNPs and AgNPs + are shown in Fig. 2e, and the ultraviolet absorption peaks are both near 400 nm. Figure 3 shows that the AgNPs + were adsorbed onto the surface of C. neoformans and C. gattii.  of the AgNPs + . The same volume of AgNPs and AgNPs + were mixed with the same volume of fungal liquid at a 1:1 ratio, and their Raman signal intensities were compared under the same excitation power and integration time. Figure 4 shows that the Raman signal of each peak was stronger with the AgNPs + . Therefore, the AgNPs + was used as the enhancement substrate to distinguish the two species of Cryptococcus.

SERS technology and multivariate statistical analysis.
The Raman spectra were averaged from the Raman spectra of each Cryptococcus species; the x coordinate is the Raman shift, and the y coordinate is the Raman intensity. Figure 5a compares the average SERS spectra of C. neoformans and C. gattii. The red line is the mean value of the SERS spectrum of C. neoformans, the black line is from the SERS spectrum of C. gattii, and the blue line is the difference between them. As shown in Fig. 4, the positions of each main peak of C. neoformans and C. gattii were similar, and the Raman shifts at 655, 790, 906, 1,008, 1,204 and 1,332 cm −1 were significantly  The peak bands at 655 cm −1 are assigned to Guanine and tyrosine, the peak located at 790 cm −1 is assigned to Cytosine and uracil, the peak located at 1,008 cm −1 is assigned to Phenylalanine, galactomannan and C-C aromatic ring stretching, and the peak located at 1,332 cm −1 is assigned to Adenine, guanine and CH deformation. This explains why C. gattii remained a variant of C. neoformans until 2002 4 . However, their peaks have differences at 655 cm −1 , 790 cm −1 , 1,008 cm −1 , and 1,332 cm −1 . The peak bands at 655 cm −1 are assigned to Guanine and tyrosine, the peak located at 790 cm −1 is assigned to Cytosine and uracil, the peak located at 1,008 cm −1 is assigned to Phenylalanine, galactomannan and C-C aromatic ring stretching, and the peak located at 1,332 cm −1 is assigned to Adenine, guanine and CH deformation. OPLS-DA was used to further distinguish C. neoformans and C. gattii, as the fingerprints are different between the two species of cryptococcus. OPLS-DA based on the SIMCA14.1 software was performed on the Raman spectrum data from all measurements. Figure 5b shows an ellipse at the 95% confidence interval of the sample analysis results, where each point represents a sample 24 . The data points are well separated on both sides of the central axis. There is no intersection and they are concentrated on both sides. The OPLS-DA model 1 (Table 1) showed that R 2 X (cum) = 0.986, R 2 Y (cum) = 0.999, and Q 2 (cum) = 0.998, indicating that the quality parameters of the model were good. Here, R 2 X represents the interpretation rate of the model in the X-axis direction, i.e., 98.6%. Term R 2 Y describes the model in the Y-axis direction, i.e., 99.9%, and Q 2 is the prediction rate of the model, i.e., 99.8% 25 .
The classification ability of the model was evaluated using a tenfold crossover validation with SIMCA14.1. All the SERS spectrum data of C. neoformans and C. gattii were divided into 10 parts. Nine parts were successively used as training data, and one part served as the test data. The average accuracy of the training and test data were obtained after 10 operations. Tables 1 and 2 (Model 2-11) show that the average accuracy of the training and test data are 100% and 100%, respectively, which indicates that the model has good classification ability 26,27 .

Discussion
Cryptococcosis is a global, invasive fungal infectious disease caused by Cryptococcus. Treatment of this disease is still extremely challenging. C. neoformans infection accounts for the majority of cryptococcosis, whereas C. gattii infection is relatively rare 28 . C. neoformans causes infections mainly in the central nervous system; pulmonary infection accounts for only 35% of all infected patients. In an in vivo microscopy experiment conducted by Ngamskulrungroj et al. 29,30,31 , there was no significant difference in the clinical symptoms between patients with these two cryptococcal infections, and a differential diagnosis based on clinical symptoms was difficult 32,33,34 . www.nature.com/scientificreports/ However, C. gattii requires longer periods of antifungal therapy, which motivates our efforts for discriminating between the two species. Traditional methods include ink staining and antigen kit testing 5 . Although the ink detection method of cerebrospinal fluid (CSF) is simple and fast, distinguishing between C. neoformans and C. gattii is difficult 2 . Cryptococcal antigen kits can use latex agglutination and enzyme-linked immunosorbent assays to study the cryptococcal polysaccharide capsular antigen (CrAg) composition in blood. This method offers rapid diagnosis within 2-3 h, and it is more sensitive than direct staining microscopy. However, most of the antiserum used is based on the C. neoformans antigen, and it is possible to miss the detection of C. gattii 35 . Antigen clearance kinetics are slower than treatment response, and continuous monitoring has limited value. Regarding molecular biology methods, PCR gene sequences using specific gene segments can be used for identification and typing. These are sensitive and specific and can be used to genotype C. neoformans and C. gattii 36 . However, PCR is timeconsuming and prone to contamination 37 . MALDI-TOF MS can also classify cryptococci and can reliably display the genetic and evolutionary relationships between species and within species. MALDI-TOF MS is more accurate and faster than gene sequencing 38 . However, it is expensive and only research hospitals can support a mass spectrometer. Gene chips are also expensive and difficult to use in routine clinical care, and second-generation sequencing requires bioinformatics professionals.
All current cryptococcus detection methods have limitations, and this study used surface enhanced Raman spectroscopy to detect cryptococcus. Versus other detection methods for C. neoformans and C. gattii [5][6][7][8][9][10] , SERS exhibits high specificity with low costs and fast analysis times. SERS can quickly identify and distinguish C. neoformans and C. gattii. Here, AgNPs + were formed from AgNO 3 with CTAB and NaBH 4 . We found that the AgNO 3 concentration could be used to tune the size of the AgNPs +35 . This method is simple and easy to repeat. Its repeatability compensates for the difficult nanomaterial synthesis. The preparation is convenient and can be easily promoted and applied in a clinical setting. The SEM image in Fig. 2b shows the characteristic structure of the AgNPs + . The AgNPs + were spherical, colloidally stable, and approximately 110 nm in diameter. The AgNPs + served as the SERS substrate. We measured the spectra of high-quality C. neoformans and C. gattii, and Fig. 5a compares the average SERS spectra of C. neoformans and C. gattii.
The spectra for the two cryptococcus show few differences because they are closely related. Therefore, it is necessary to use an OPLS-DA analysis tool. Figure 5b shows the OPLS-DA score diagram of the SERS spectra of C. neoformans and C. gattii. All the data points were distributed on both sides of the central axis with no intersection, indicating that OPLS-DA could distinguish the SERS spectra of C. neoformans and C. gattii well. In addition, as shown in Table 1, R2X(cum) = 0.986, R2Y(cum) = 0.999 and Q2(cum) = 0.998, indicating that the quality parameters of the established model are good. Table 2 shows that the average accuracy of training and test data is 100% and 100%, respectively, illustrating that the model has good classification ability.
We have initiated the application of SERS for cryptococcus detection. SERS can be used for rapid and nondestructive detection of Cryptococcus with relatively low equipment cost. Due to the special properties of the capsule, cryptococcus is difficult to detect relative to other fungi such as candida and mold. Therefore, many detection methods that can be applied to detect common fungi cannot be used to detect cryptococcus. This highlights the universal applicability of Raman spectroscopy to detect pathogenic fungi. The OPLS-DA data shows that Raman can clearly distinguish between two cryptococcus species with very similar properties, which is an excellent result.

conclusions
We report a novel detection method to identify and discriminate C. neoformans and C. gattii. The positively charged nanoparticles are shown to increase the intensity of the Raman spectra acquired from Cryptococcus. A classification model to discriminate the SERS spectra of C. neoformans and C. gattii was then built using OPLS-DA. This model is capable of explaining over 99% of the variance in the data and achieves a 99.8% predictive rate of discrimination. These results indicate that a SERS-based diagnostic tool may be a powerful technique for diagnosing cryptococcosis. Synthesis of the substrate. AgNPs + colloids were prepared according to previous methods by van Lierop et al. 17 and Chen et al. 18 . The AgNPs + diameters could be increased to more than 100 nm 17,18 . Solution A was prepared first, as follows: 3.26 mM AgNO 3 , 0.4 M NH 4 OH, and 0.5 mM CTAB (cetyl trimethyl ammonium bromide) were added to 20 mL of ultrapure water. Solution B was then prepared as follows: 8 mM NaBH 4 and 0.5 mM CTAB were added to 20 mL of ultrapure water. Solutions A and B were placed in an ice bath for 15 min to slow the reaction. Solution A was then poured into a 50-mL conical flask that was vigorously stirred on a heated electromagnetic stirrer at 700 rpm. Solution B was then poured into solution A dropwise and stirred on a heated electromagnetic mixer for 4.5 h; this process was repeated in an ice bath. The conical flask was sealed during the stirring process to prevent the silver from being oxidized. The stirring was stopped after 4.5 h, and the mixture was heated at 100℃ for 12 min to remove the remaining NH 3 and NaBH 4 . The heat was then turned off, and the stirring was continued until the mixture cooled to room temperature. Ultrapure water was then added to a total volume of 40 mL; 1.5 mL of the solution was placed in a 1.5 mL polystyrene tube and centrifuged at 7,500 rpm for 7 min. After centrifugation, 1.4 mL of the supernatant was discarded, and the remaining solution was mixed. The resulting AgNPs + were stored in a refrigerator at 4 °C.

Strain collection and identification.
Clinical isolates of C. neoformans and C. gattii were collected as cerebrospinal fluid (CSF) at Shanghai Changzheng Hospital, China. The specimens were from various media including blood plate medium, chocolate agar plate medium, sabouraud dextrose agar (SDA) medium, or isolated from a single colony. Selected fungal colonies were cultured for 48 h in ambient air at 25 °C in SDA medium. The fungal suspension was identified using India ink staining, a lateral flow assay (LFA) for cryptococcal antigen (CrAg) detection (ImmunoMycologics, Inc, Norman, OK, USA) and using an ABI_3730xl sequenator (Genomic Instrument, USA) to specifically sequence the intergenic spacer 1 (IGS1) region of the nuclear ribosomal rRNA gene of the fungal culture.

Preparation of SERS detection samples.
Cryptococcus gattii is very clinically rare. Two strains of C.
neoformans and two strains of C. gattii were selected and grown in SDA medium for 48 h at 25 °C. Approximately 50 μL of Cryptococcus colonies were mixed with 100 μL of ultrapure water and oscillated for 2 min to ensure complete mixing. The number of C. neoformans and C. gattii in the solution was calculated by a fungal counter as 1.03 × 10 9 and 1.05 × 10 9 cells mL -1 , respectively, followed by mixing of 10 μL of fungus liquid with 10 μL of the pre-prepared AgNPs + as shown in Fig. 6. Next, 2 μL of the mixture was placed on the monocrystalline silicon. They were detected by Raman spectroscopy (i-raman Plus BWS465785H) after the mixture completely dried.

SERS measurements.
The Raman spectrometer used in this experiment has a maximum laser power of 275 mW at 785 nm with a light spot diameter of 105 μm. The excitation power was 20% of the maximum power, and the integration time was 15 s. We used the white light function to find the darkest area on the wafer, and 20 replicate SERS spectra were recorded. The processed spectra were used to compare the average Raman data from C. neoformans and C. gattii. The Cryptococcus mixtures were placed on a silicon chip and studied with a microscope; 20 different fields-of-view were studied. The total acquisition time of the SERS spectrum was 45 min, including the time of centrifugation of the substrate, sample preparation, and spectrum testing. www.nature.com/scientificreports/ Statistical analysis. The OPLS-DA spectral data were collected with SIMCA 14.1 (Umetrics, Umea, Sweden) software. OPLS-DA is a supervised discriminant analysis method that predicts differences in fungi species by establishing the relationship model between the expression level of metabolites and the fungi species. The OPLS-DA model is established for comparisons between two groups, and the parameter evaluation of the model is presented in the form of a table in which R 2 X and R 2 Y represent the interpretation rate of the established model to the X and Y matrix, respectively; term Q 2 indicates the prediction ability of the model 24 . The fitting accuracy of the model improves as R 2 and Q 2 approach 1. Lower values of R 2 and Q 2 lead to worse model fitting accuracy. A fitting accuracy of R 2 and Q 2 above 0.5 (50%) is good, and higher than 0.4 is acceptable.
The accuracy of the OPLS-DA model for classification of unknown samples was predicted via a tenfold crossover validation. The classification ability of the model was also evaluated by a tenfold crossover examination. The SERS data of all C. neoformans and C. gattii were divided into 10 parts; nine parts were successively used as training data and one part served as the test data. After 10 operations, the average accuracy of the training and test data was obtained. The mean value of the accuracy (or error rate) of the results of the tenfold crossover validation was used to estimate the accuracy of the algorithm.