2.1.

Subjects

GCF samples were pooled from patients aged between 12 and 22 years consecutively recruited at the Orthodontic Program of the Università della Campania “Luigi Vanvitelli,” Naples, Italy, from December 2015 to December 2016 (see Table 1). Informed consent was obtained from each minor patient’s parents or adult patients after providing them with detailed information about the study. Ethical approval for this study was granted by the Ethics Board of the above-mentioned university.

The inclusion criteria were as follows: nonextraction fixed orthodontic treatment required, permanent dentition with similar incisor irregularity index in the mandibular dental arch ($\ge 2$), a good general healthy periodontium with no radiographic evidence of bone loss, no gingival inflammation and probing depth of 3 mm in the whole dentition, and a full-mouth plaque score and a full-mouth $\text{bleeding score}\le 20\%$.³⁹ Exclusion criteria included previous orthodontic treatment with fixed appliances, history of systemic diseases, congenital deformities, mandibular tooth agenesis, previous periodontal disease, and the use of antibiotic or antiinflammatory drugs in the month preceding the beginning of the study. Moreover, drug consumers and smokers were also removed from the study.

For this study, we utilized a 50 gr Sentalloy^® (GAC International, Bohemia, New York) coil spring in vivo. Orthodontic brackets with power pin (MBT; 3M Unitek; Monrovia; Cali, Colorado) were applied on the buccal surface of the upper and/or lower first premolars and bands were placed on upper and/or lower first premolar only on the right side. Bracket bonding, arch-wire insertion, and ligation were performed by the same operator.

Table 1

Year of birth and gender of patients considered in the present study.

2.2.

Collection and Preparation of GCF Samples

The GCF samples were collected from each patient before bracket bonding (T0), and after 2 (T1), 7 (T2), and 14 (T3) days of treatment at the buccal side of the incisors. The T0 samples represented the controls. Before GCF collection, the sample contamination by blood, saliva, and plaque was minimized by using suction and self-retaining retractor, and the target teeth were carefully cleaned, gently dried with air stream for 5 s, and isolated with cotton rolls. Standardized sterile absorbent paper cones were inserted 1 mm into the gingival crevice and left in situ for 30 s, without blood, saliva, and plaque contamination. Two cones were consecutively used (1-min interval) in order to maximize GCF volume per site. Paper cones were transferred to sterile plastic vials and stored at $-80°\mathrm{C}$. These paper cones were used without any other treatment for $\mu$-RS measurements. Immediately before SERS measurements, $10\mu \mathrm{L}$ of distilled water was added and the tubes were vortexed for 30 s and centrifuged (10 min, 800 g).

2.3.

Nanoparticle Preparation and Characterization

GNP preparations were obtained by conventional citrate reduction method.^38,40 The amount of sodium citrate was defined in order to obtain GNPs with an expected diameter of about 30 nm since preliminary measurements showed that this was a good choice for SERS applications.³⁸ Particle size and stability of the resulting preparations were investigated by means of absorption spectroscopy, transmission electron microscope (TEM), and dynamic light scattering (DLS). In particular, absorption spectroscopy was used to obtain information on the plasmon resonance peak and to study the stability of the preparations. TEM and DLS were used to evaluate the size of the GNPs. A mean GNP diameter of 30 nm was estimated from the observed plasmon resonance band maximum position using the tabular data reported in Ref. 41. Moreover, the absorption spectra remained unchanged for a long time suggesting a stability of the preparation of more than 6 months. TEM results were in agreement with those from DLS and provided a mean diameter for the GNPs of $27\pm 7\mathrm{nm}$. For further details, see Ref. 38.

2.4.

Raman Microspectroscopy and SERS

3.1.

SERS Measurements on GCF Liquid Samples

A typical SERS spectrum obtained from GCF samples pooled from a patient before starting the previously described orthodontic treatment is reported in Fig. 1 in the range from 400 to $1800{\mathrm{cm}}^{-1}$ together with the results of the deconvolution procedure adopted for determining the principal vibrational modes. In this spectrum, it is possible to highlight different contributions due to the various components of GCF. In particular, the feature observed at $465{\mathrm{cm}}^{-1}$ can be ascribed to $\mathrm{S}─\mathrm{S}$ bond stretching.⁵¹ The structures at 621 and $825{\mathrm{cm}}^{-1}$ are attributed to phenylalanine and tyrosine contributions, respectively. The small peaks at 895 and $946{\mathrm{cm}}^{-1}$ can be assigned to the nucleic acid backbone and ${{\mathrm{PO}}_4}^{3-}$. The peak at $1007{\mathrm{cm}}^{-1}$ is due to $\mathrm{C}─\mathrm{H}$ bending phenylalanine contribution. The mode at $1176{\mathrm{cm}}^{-1}$ can be attributed to the nucleic acid base $\mathrm{C}─\mathrm{N}$. The peaks at 1242 and $1276{\mathrm{cm}}^{-1}$ can be due to different components of amide III ($\alpha$-helix or $\beta$-sheet) and the $1276{\mathrm{cm}}^{-1}$ peak may also indicate a contribution from ${\mathrm{CH}}_2$ deformation mode. The peak observed at $1348{\mathrm{cm}}^{-1}$ can be assigned to adenine/guanine of nucleic acids. The small peak at $1470{\mathrm{cm}}^{-1}$ is due to ${\mathrm{CH}}_2$, the one at $1577{\mathrm{cm}}^{-1}$ can be related to amide II and cytochrome, and the large structure at $1641{\mathrm{cm}}^{-1}$ is typical of protein spectra and is related to amide I. All the structures highlighted are reported in Table 2 along with their assignments according to Refs. 44, 51, 52 and references therein.

Fig. 1

SERS spectrum of a GCF sample collected before orthodontic treatment (T0). The experimental data are fitted by a convolution of Lorentzian peaks (red line). The positions of the main deconvoluted modes are indicated. The component modes are reported on the bottom of the figure, arbitrarily shifted along the $y$-axis.

Table 2

Main peaks observed in SERS spectra of GCF samples collected at different stages of the orthodontic treatment and relative assignments according to Refs. 44, 51, and 52.

SERS spectra of the GCF collected from an 18-year-old patient ($N$ patient) are shown in Fig. 2 for T0 (before treatment beginning), T1 (after 2 days), T2 (after 7 days), and T3 (after 2 weeks) stages of the orthodontic treatment. The application of stress on the tooth induces changes in the GCF composition that can be seen in the SERS spectra. We focused our attention on the 1000 to $1700{\mathrm{cm}}^{-1}$ wavenumber range and on the five main modes at about 1176, 1276, 1348, 1577, and $1641{\mathrm{cm}}^{-1}$ (see Fig. 2, T0 spectrum). As the time elapsed from the beginning of the orthodontic process increases, the intensity of Raman mode at $1276{\mathrm{cm}}^{-1}$ decreases. Conversely, the amide I band intensity increases in the T1 stage with respect to T0 (see T1 spectrum in Fig. 2) and decreases in the subsequent T2 and T3 steps. This behavior can be ascribed to changes in protein concentration. A similar effect has been observed by Jung et al.,³¹ and it is also expected because it is well known that orthodontic treatment induces inflammatory processes and changes in cytokines, enzyme, and metabolites concentrations.^{1,4,5,14–17} These changes are more intense immediately after the application of orthodontic forces and decrease over time. In addition, the center of the amide I band shifts toward a lower wavenumber shift value (from 1641 to $1611{\mathrm{cm}}^{-1}$) reflecting changes in the protein configuration.³⁰ As far as concerns amide I band shift, Jung et al.³¹ noticed a spectral blueshift differently from us.^30,32 Moreover, observing T1, T2, and T3 spectra, a significant increase in the intensity of the modes at 1167, 1390, and $1577{\mathrm{cm}}^{-1}$ is found. The mode at $1577{\mathrm{cm}}^{-1}$ is also visible in the T0 spectrum as a shoulder of the amide I band. All these modes are compatible with the Raman vibrations of cytochrome complex, a small heme-protein with high oxidation properties.

Fig. 2

SERS spectra of GCF samples collected from an 18-year-old patient (N) before (T0), after 2 (T1), after 7 (T2), and 14 days (T3) the beginning of the orthodontic treatment. The spectra are arbitrarily shifted along the $y$-axis.

More detailed information can be obtained from the amide I band deconvolution [Fig. 3(a)]. During the orthodontic treatment, the center of amide I band moves toward shorter wavenumber shifts, especially in T2 and T3 spectra. This is due to the decrease of the $\alpha$-helix contribution (located around $1640{\mathrm{cm}}^{-1}$) and the increase of the $\beta$-sheet contribution (located in the range between 1610 and $1625{\mathrm{cm}}^{-1}$). The percentage of $\alpha$-helix contribution to the amide I band was estimated by the ratio of areas of the $\alpha$-helix peak and the whole amide I band, and it is reported for the different stages of treatment in Fig. 3(b). After an initial increase of the $\alpha$-helix contribution at T1, a decrease at T2 and T3 is observed. The differences in the behavior of the amide I contributions at the different stages observed by Jung et al.³¹ could be ascribed to different GCF sampling and days of collection.

Fig. 3

(a) Amide I region of SERS spectra of GCF samples collected from an 18-year old patient (N) before the beginning of the orthodontic treatment (T0) and 2 (T1), 7 (T2), and 14 (T3) days after. The experimental data (black line) are fitted by a convolution of Lorentzian peaks (green line) and the main components are plotted. The $\alpha$-helix and the $\beta$-sheet components of amide I band are emphasized by the red and blue lines, respectively. (b) The $\alpha$-helix component at T0, T1, T2, and T3 stages. The spectra are arbitrarily shifted along the $y$-axis. Percentage value of the of the $\alpha$-helix area normalized to the whole amide I band area is reported for each plot.

3.2.

$\mu$-RS on Paper Cones

With the aim to develop a simple and minimally invasive method for monitoring the orthodontic treatment, a micro-Raman analysis directly on the impregnated GCF paper cones without any sample manipulation was performed. In Fig. 4, the Raman spectra of the GCF from two patients (E and P) before treatment (T0) are reported. Also in these spectra, it is possible to point out some of the features already observed in the SERS spectrum (Fig. 1). The two main broadbands located at around 1250 to 1275 and $1640{\mathrm{cm}}^{-1}$ can be assigned to amide III and amide I, respectively. Moreover, modes due to the nucleic acids can be found at wavenumber shift positions of 1176 and $1348{\mathrm{cm}}^{-1}$. Finally, the 1575- to $1577\text{-}{\mathrm{cm}}^{-1}$ mode is considered as a shoulder of the amide I band.

Fig. 4

Representative $\mu$-RS spectra of GCF samples on paper cones collected before the beginning of the orthodontic treatment (T0) from two patients: 12-year old (E) and 16-year old (P). The spectra are arbitrarily shifted along the $y$-axis.

The $\mu$-RS spectra collected after 2 days (T1) and 7 days (T2) from the beginning of the orthodontic treatment are reported in Fig. 5, along with the corresponding T0 spectra for the same two patients in Fig. 4. In particular, Fig. 5(a) shows spectra from P patient samples and Fig. 5(b) shows those from E patient. In both cases, an increase of the intensity of the Raman modes around $1100{\mathrm{cm}}^{-1}$ (due to ${{\mathrm{PO}}_2}^{-}$ of nucleic acids⁵¹) and $1348{\mathrm{cm}}^{-1}$ is observed. The intensity of the $1575\text{-}{\mathrm{cm}}^{-1}$ mode also increases, even if more noticeably in samples from P patient [Fig. 5(a)]. This feature confirms an increase of cytochrome content of the GCF during the orthodontic force application also supported by the occurrence of a peak at $1372{\mathrm{cm}}^{-1}$ in the T2 spectra from E patient [Fig. 5(b)], close to the cytochrome mode expected at $1390{\mathrm{cm}}^{-1}$ (see Table 2). Amide III band ($1280{\mathrm{cm}}^{-1}$) dependence on the treatment stage is different for E and P patients, as evident in T1 and T2 spectra. The differences observed in the spectra from these two patients can be ascribed to their different ages, which may induce a difference in the biological response to the orthodontic treatment.¹⁷ Moreover, a weak peak at about 1450 to $1470{\mathrm{cm}}^{-1}$ seems to indicate the presence of lipids at stages T1 and T2 for both patients. The main results of the $\mu$-RS on GCF impregnated paper cones are summarized in Table 3.

Fig. 5

$\mu$-RS spectra of GCF samples on paper cones collected from the patients (a) P (16-year old) and (b) E (12-year old) before the beginning of the orthodontic treatment (T0) and 2 (T1) and 7 (T2) days after. The spectra are arbitrarily shifted along the $y$-axis.

Table 3

Main peaks observed in μ-RS spectra of GCF samples on paper cones collected at different stages of the orthodontic treatment and relative assignments according to Refs. 44, 51, and 52.

Additional information can be obtained by analyzing in more detail the 1500- to $1750\text{-}{\mathrm{cm}}^{-1}$ region where the amide I band is located. In agreement with SERS results, in the $\mu$-RS spectra the center of the band also moves toward shorter wavenumber shifts after the beginning of the treatment. In Fig. 6, the results of the deconvolution procedure for the above-mentioned region are reported for the spectrum of P patient. Also, in this case, Lorentzian peaks were considered and a dependence of their characteristics on the treatment stage can be observed as evident by inspecting the figure, where bold lines indicate the main components of the amide I band. In particular, the red line represents the most relevant contribution ($\alpha$-helix), which shows a clear shift toward shorter wavenumber shifts. However, the quality of $\mu$-RS spectra is lower than that obtained by SERS and thus a quantitative estimate of the different components is less accurate. Moreover, the deconvolution procedure highlighted an intense peak at about $1540{\mathrm{cm}}^{-1}$ appearing at T1 and T2 stages, which can be assigned to carotene.^30,32

Fig. 6

$\mu$-RS spectra of the amide I region for GCF measured on paper cones at different stages of the orthodontic treatment: T0, T1, and T2. The experimental data (black line) are fitted by a convolution of Lorentzian peaks (green line) and the components are shown. The $\alpha$-helix component of amide I band is reported by the red line and its center redshifts when the orthodontic treatment is prolonged. The intense peak at about $1540{\mathrm{cm}}^{-1}$ appearing at T1 and T2 stages is assigned to carotene.

A patient-dependent variability of data is clearly shown, and it has to be considered using proper statistical methods in order to exploit $\mu$-RS directly from GCF impregnated cones for monitoring the orthodontic treatment. For this objective, the proposed approaches for statistical analysis (see Sec. 2.4) were applied to all the Raman spectra of GCF samples before the beginning and at the different stages of the treatment.

The percentages of stage assignment obtained using the ranking analysis is shown in Fig. 7. The rate of success in the determination of the stage is about 83% for T0 and T2 stages, indicating reasonable efficiency of the Raman analysis. A lower percentage of success, about 46%, is found in the T1 spectra. The efficiency in the determination of T1 state is relatively low because several spectra belonging to the T1 class are classified as T0 or T2. This can be attributed to patient-dependent variability. The T3 stage was not considered in this analysis due to the limited number of available spectra. Notably, the features determined using ranking analysis (indicated in Appendix A) are related to the main Raman fingerprints of proteins, DNA, and lipids.⁴⁴

Fig. 7

Percentage of success in the orthodontic treatment stage obtained by comparing the Raman spectra using the ranking analysis. T3 stage was not considered in this analysis due to the limited number of spectra.

The proposed multivariate analysis (i-PCA, see Sec. 2.4.5) offers a more robust method for taking into account the above-mentioned variability. The results of i-PCA analysis for the single intervals give no useful suggestion about models to group the spectra. The score plots for the single intervals do not suggest any simple grouping depending on the time at which the GCF was collected. However, an overall inspection of these plots suggests that some common spectrum characteristics of a specific group are selected using the first three PCs. In order to exploit these outcomes, a dataset with reduced dimensions was built using the scores of the first three components obtained for the RIs and used for classification. The results of the classification procedure are shown in Fig. 8. Notably, the orthodontic treatment stage is correctly assigned in 98% of Raman spectra with 100% of correct assignment for T0 and T2 spectra and 92% of T1 spectra, some T1 spectra being wrongly classified as T0. These results suggest that the analysis of the GCF using Raman spectroscopy and multivariate analysis can properly recognize the stage of the orthodontic treatment. This may lead the way to define a procedure for an easy, efficient, and repeatable follow-up of the orthodontic treatment.

Fig. 8

Percentage of success in the orthodontic treatment stage obtained by comparing the Raman spectra using the i-PCA analysis. Also in this case, T3 stage was not considered due to the limited number of spectra.