Article Cite This: Anal. Chem. XXXX, XXX, XXX−XXX
pubs.acs.org/ac
Raman Spectroscopy Applied to Parathyroid Tissues: A New Diagnostic Tool to Discriminate Normal Tissue from Adenoma Andrea Palermo,† Marco Fosca,‡ Gaia Tabacco,† Federico Marini,§ Valerio Graziani,‡ Maria Carla Santarsia,† Filippo Longo,∥ Angelo Lauria,† Roberto Cesareo,⊥ Isabella Giovannoni,@ Chiara Taffon,@ Massimiliano Rocchia,# Silvia Manfrini,† Pierfilippo Crucitti,∥ Paolo Pozzilli,† Anna Crescenzi,@ and Julietta V. Rau*,‡ Unit of Endocrinology and Diabetes, Campus Bio-Medico University, via Á lvaro del Portillo 200, 00128 Roma, Italy Istituto di Struttura della Materia, Consiglio Nazionale delle Ricerche (ISM-CNR), via del Fosso del Cavaliere 100, 00133 Roma, Italy § Dipartimento di Chimica, Università“La Sapienza”, Piazzale Aldo Moro 5, 00185 Roma, Italy ∥ Unit of Neck and Chest Surgery, Campus Bio-Medico University, via Á lvaro del Portillo 200, 00128 Roma, Italy ⊥ Malattie della Tiroide ed Osteometaboliche, Hospital Santa Maria Goretti, Via Canova, 04100 Latina, Italy @ Unit of Pathology, Campus Bio-Medico University, via Á lvaro del Portillo 200, 00128 Roma, Italy # Thermo Fisher Scientific, Strada Rivoltana, 20090 Rodano, Milano, Italy † ‡
S Supporting Information *
ABSTRACT: Primary hyperparathyroidism is an endocrine disorder characterized by autonomous production of parathyroid hormone. Patients with the symptomatic disease should be referred for parathyroidectomy. However, the distinction between the pathological condition and the benign one is very challenging in the surgical setting; therefore, accurate recognition is important to ensure success during minimally invasive surgery. At present, all intraoperative techniques significantly increase surgical time and, consequently, cost. In this proof-of-concept study, Raman microscopy was used to differentiate between healthy parathyroid tissue and parathyroid adenoma from 18 patients. The data showed different spectroscopic features for the two main tissue types of healthy and adenoma. Moreover, the parathyroid adenoma subtypes (chief cells and oxyphil cells) were characterized by their own Raman spectra. The partial leastsquares discriminant analysis (PLS-DA) model built to discriminate healthy from adenomatous parathyroid tissue was able to correctly classify all samples in the calibration and validation data sets, providing 100% prediction accuracy. The PLS-DA model built to discriminate chief cell adenoma from oxyphil cell adenoma allowed us to correctly classify >99% of the spectra during calibration and cross-validation and to correctly predict 100% of oxyphil and 99.8% of chief cells in the external validation data set. The results clearly demonstrate the great potential of Raman spectroscopy. The final goal would be development of a Raman portable fiber probe device for intraoperative optical biopsy, both to improve the surgical success rate and reduce surgical cost.
P
The current gold standard method to differentiate pathological from normal conditions is histopathology.2 Patients with symptomatic disease should be referred for parathyroidectomy if there are no medical contraindications. However, the distinction between the pathological and benign conditions is challenging in the surgical setting. Therefore, accurate recognition is very important to ensure continuing success of minimally invasive surgery. To overcome this issue, it is mandatory to use specific preoperative localization techniques, such as neck ultrasound scan and/or parathyroid scintigraphy.
rimary hyperparathyroidism (PHPT) is an endocrine disorder characterized by autonomous production of parathyroid hormone (PTH). PHPT is now defined as a spectrum ranging from inappropriately high or even normal PTH in the setting of high-normal or even normal calcium.1 PHPT is the third most common clinical endocrine disorder after diabetes and thyroid disease.2 This condition can have an important impact on a large number of tissues and organ systems, including the brain, muscle, heart, and kidneys.3 The majority of cases (80−85%) are attributable to a solitary parathyroid adenoma. A double adenoma is present in up to 4% of cases, and four-gland hyperplasia makes up the remaining 10−15% of cases.2 Parathyroid carcinoma has a prevalence of 1 can be considered to contribute significantly to the model. The values of the VIP scores for the PLS-DA model built for the discrimination between healthy and adenoma tissues are reported in Figure 6. As shown in Figure 6, inspection of the VIP scores identified the bands centered at 747, 1125, 1156, 1520, 1582, and 2850 cm−1 as the most important Raman shift wave numbers to discriminate healthy from adenomatous tissue. This result is in perfect agreement with the spectral features that were identified as signatures of the different tissues in the spectroscopic characterization described in the previous paragraph. In a second stage, the possibility of identifying whether adenomatous tissue subtypes can be further discriminated (i.e., composed of chief or oxyphil cells), or a mixture of both, was addressed. Considering the nature of the problem and also the numerosity of the different kinds of tissue samples available, the classification models were built using all available spectra recorded for each tissue (i.e., the spectra associated with all relevant pixels in the micro-Raman images). All spectra from five individuals (four with chief and one with oxyphil cells) were placed in the training set to build a PLS-DA model to discriminate between chief and oxyphil cells, and all profiles from the five other individuals (four with chief and one with oxyphil cells) were left out as the external validation set. The spectra of the tissues containing both chief and oxyphil cells were left out as a further validation set to determine whether the model could discriminate the presence of both cell types within these samples. A PLS-DA model was then built, and its optimal model complexity was estimated by a 4-fold cross validation resulting in 10 latent variables. The model allowed us to correctly classify more than 99% of the spectra during calibration and cross-validation and to correctly predict 100% of oxyphil and 99.8% of chief cells in the external validation set. Moreover, when the optimal model was applied to the tissue spectra, containing chief and oxyphil cells, it was able to
Statistical Analysis of the Spectroscopic Data. The Raman spectra were treated with advanced statistical procedures. All spectra were collected in a matrix, baselinecorrected using the penalized alternating least-squares algorithm (with p = 0.005 and λ = 105)36 and normalized using standard normal variate transformation.37 Then, different classification steps were undertaken to differentiate the healthy from the pathological (adenomatous) tissues and then to further discriminate the pathological tissue subtypes (i.e., chief and oxyphil cells in the parathyroid). In all chemometric analyses, the spectral region between 1700 cm−1 and 2800 cm−1 was excluded from the data set, as it only included noise. A PLS-DA model was built and validated using the average spectra for each map as descriptors, after mean centering with the aim of discriminating between healthy and adenoma tissues. The sample set was split into a calibration set for model building (25 samples collected from 18 individuals) and a test set for model validation (23 spectra collected from 10 individuals). The optimal model complexity was estimated using a cross-validation procedure with six cancelation groups, being careful that all the samples coming from the same individuals were included in the same split (so that each cancelation group contained the samples from three patients); in particular, three latent variables led to the lowest classification error in the cross-validation. The optimal PLS-DA model was able to correctly classify all samples in the calibration and validation sets, providing 100% prediction accuracy for both classes. This result is shown graphically in Figure 5, where the training and test samples for the two classes are displayed on the space spanned by the first three latent variables of the model. The separation between healthy and adenomatous parathyroid tissue is evident from Figure 5 and confirms the perfect classification ability of the model. The interpretation of the model in terms of which Raman transitions are more relevant for the discrimination between healthy and adenomatous tissues can be accomplished by inspecting the values of the variable importance in projection F
DOI: 10.1021/acs.analchem.7b03617 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 7. Partial least-squares discriminant analysis (PLS-DA) to discriminate chief from oxyphyl cells. Variable importance in projection (VIP) scores for the measured variables (green bars) superimposed on the average spectrum of the training data set (black line). The red dashed line indicates the threshold value for significance.
of a parathyroid adenoma. The results of this investigation can be used to develop a novel diagnostic and therapeutic approach for managing PHPT. Indeed, translating this approach to clinics for intraoperative surgery and applying Raman probe portable systems would be very helpful. The development of a novel diagnostic tool, such as real-time optical biopsy, would be a revolutionary advancement that may decrease the number of diagnostic analyses, lower costs, and significantly improve surgical success rate.
correctly detect the presence of both cell types, thus, confirming its accuracy. The interpretation of the model in terms of the spectral regions more relevant for classification can be performed by inspecting the VIP score values (reported in Figure 7). As shown in Figure 7, the VIP scores (i.e., 747, 1301, and 2930 cm−1) were in good agreement with the spectral features that emerged from the spectroscopic characterization described in the previous section. Biomedical Problem and Future Perspectives. Histopathology is the current gold standard method to distinguish pathological from normal tissues. The parathyroid glands consist of parenchymal cells, fat cells, and fibrovascular stroma.24,39 The functional cells are chief cells, which are polygonal with a small round nucleus and a slightly eosinophilic cytoplasm. Adenomas are principally made up of chief cells but oxyphil cells and transitional oxyphil cells can also be detected in varying proportions under the microscope.40−42 The chief cells in adenomas show nuclear pleomorphism, multinuclei, and giant cells.40,41,43 Adenomas are virtually devoid of adipocytes, which are only observed in the normal rim of the compressed parathyroid.25 However, an intraoperative histopathological assessment of suspicious tissues is time-consuming and associated with additional costs. More fast and reliable intraoperative tools are needed to support surgeons in their assessment of the parathyroid tissue classification to improve treatments and outcomes. Up to now, no other studies have used RS to differentiate between normal parathyroid and pathological tissues. Only one study employed RS to differentiate between parathyroid adenomas and hyperplasia, demonstrating that RS is an excellent tool for a similar task.15 The aim of the present study was to demonstrate the feasibility of an RS-based approach to distinguish between pathological and normal parathyroid tissues reliably and in a shorter time. Moreover, this study provides a better understanding of the pathogenic process leading to the development
■
CONCLUSIONS
This RS study was aimed at differentiating healthy parathyroid tissue from parathyroid adenoma. The data show that healthy and adenomatous tissue are characterized by particular Raman spectroscopic features, as well as parathyroid adenoma subtypes (chief cells and oxyphil cells). The Raman spectra were treated with advanced statistical procedures. The PLS-DA model built to discriminate healthy from adenomatous parathyroid tissue correctly classified all samples in the calibration and validation sets, providing 100% prediction accuracy. The PLS-DA model built to discriminate chief cell adenoma from oxyphil cell adenoma parathyroid tissue allowed a correct classification of more than 99% of the spectra during calibration and crossvalidation and correctly predicted 100% of oxyphil and 99.8% of the chief cells in the external validation set. Moreover, when the optimal model was applied to the mixed subtype tissue spectra, containing both chief and oxyphil cells, it correctly detected the presence of both cell types, thus, confirming its accuracy. These results clearly demonstrate the great potential of RS for this task, although further work is needed to validate the methodology. This proof-of-concept Raman study is an important step toward the development of Raman portable fiber probe device for intraoperative optical biopsy aimed at improving patients’ outcomes. G
DOI: 10.1021/acs.analchem.7b03617 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
■
(13) Haka, A. S.; Volynskaya, Z.; Gardecki, J. A.; Nazemi, J.; Lyons, J.; Hicks, D.; Fitzmaurice, M.; Dasari, R. R.; Crowe, J. P.; Feld, M. S. Cancer Res. 2006, 66 (6), 3317−22. (14) Harder, S. J.; Isabelle, M.; DeVorkin, L.; Smazynski, J.; Beckham, W.; Brolo, A. G.; Lum, J. J.; Jirasek, A. Sci. Rep. 2016, 6, 21006. (15) Das, K.; Stone, N.; Kendall, C.; Fowler, C.; Christie-Brown, J. Lasers Med. Sci. 2006, 21, 192−197. (16) Abramczyk, H.; Surmacki, J.; Brozek-Pluska, B.; Morawiec, Z.; Tazbir, M. J. Mol. Struct. 2009, 924−926, 175−182. (17) Liu, C. H.; Zhou, Y.; Sun, Y.; Li, J. Y.; Zhou, L. X.; BoydstonWhite, S.; Masilamani, V.; Zhu, K.; Pu, Y.; Alfano, R. R. Technol. Cancer Res. Treat. 2013, 12 (4), 371−382. (18) Zhou, Y.; Liu, C. H.; Sun, Y.; Pu, Y.; Boydston-White, S.; Liu, Y.; Alfano, R. R. J. Biomed. Opt. 2012, 17 (11), 116021. (19) Rosai, J. Rosai and Ackerman’s Surgical Pathology, 10th ed.; Mosby Elsevier, 2011. (20) Sjöström, M.; Wold, S.; Söderström, B. PLS discriminant plots. In Pattern Recognition in Pratice II; Gelsema, E. S., Kanal, L. N., Eds.; Elsevier: Amsterdam, The Netherlands, 1986; pp 461−470. (21) Ståhle, L.; Wold, S. J. Chemom. 1987, 1, 185−196. (22) Barker, M.; Rayens, W. J. Chemom. 2003, 17, 166−173. (23) Wold, S.; Martens, H.; Wold, H. The Multivariate Calibration Problem in Chemistry solved by the PLS Method. In Matrix Pencils; Ruhe, A., Kågström, B., Eds.; Lecture Notes in Mathematics; Springer Verlag: Heidelberg, Germany, 1983; pp 286−293. (24) Grimelius, L.; Akerström, G.; Johansson, H.; Bergström, R. Pathol. Annu. 1981, 16, 1−24. (25) Bilezikian, J. P.; Marcus, R.; Levine, M. A.; Potts, J. T., Jr. The Parathyroids: Basic and Clinical Concepts, 3rd ed.; Elsevier, 2015. (26) De Gelder, J.; De Gussem, K.; Vandenabeele, P.; Moens, L. J. Raman Spectrosc. 2007, 38 (9), 1133−1147. (27) Macernis, M.; Galzerano, D.; Sulskus, J.; Kish, E.; Kim, Y. H.; Koo, S.; Valkunas, L.; Robert, B. J. Phys. Chem. A 2015, 119, 56−66. (28) Rimai, L.; Kilponen, R. G.; Gill, D. J. Am. Chem. Soc. 1970, 92, 3824−3825. (29) Czamara, K.; Majzner, K.; Pacia, M. Z.; Kochan, K.; Kaczor, A.; Baranska, M. J. Raman Spectrosc. 2015, 46 (1), 4−20. (30) Rygula, A.; Majzner, K.; Marzec, K. M.; Kaczor, A.; Pilarczyk, M.; Baranska, M. J. Raman Spectrosc. 2013, 44, 1061−1076. (31) Bergholt, M. S.; Lin, K.; Zheng, W.; Lau, D. P. C.; Huang, Z. J. Biomed. Opt. 2012, 17 (7), 077002. (32) Lau, D. P.; Huang, Z.; Lui, H.; Man, C. S.; Berean, K.; Morrison, M. D.; Zeng, H. Lasers Surg. Med. 2003, 32, 210−214. (33) Abramczyk, H.; Surmacki, J.; Kopeć, M.; Olejnik, A. K.; Kaufman-Szymczyk, A.; Fabianowska-Majewska, K. Analyst 2016, 141, 5646. (34) Mourant, J. R.; Short, K. W.; Carpenter, S.; Kunapareddy, N.; Coburn, L.; Powers, T. M.; Freyer, J. P. J. Biomed. Opt. 2005, 10 (3), 031106. (35) Mourant, J. R.; Dominguez, J.; Carpenter, S.; Short, K. W.; Powers, T. M.; Michalczyk, R.; Kunapareddy, N.; Guerra, A.; Freyer, J. P. J. Biomed. Opt. 2006, 11 (6), 064024. (36) Eilers, P. H. C. Anal. Chem. 2004, 76 (2), 404−411. (37) Barnes, R. J.; Dhanoa, M. S.; Lister, S. J. Appl. Spectrosc. 1989, 43, 772−777. (38) Wold, S.; Johansson, E.; Cocchi, M. PLS: Partial Least Squares Projections to Latent Structures. In D QSAR in Drug Design: Theory, Methods and Applications; Kubinyi, H., Ed.; Escom Science Publishers: Leiden, The Netherlands, 1993; pp 523−550. (39) Dekker, A.; Dunsford, H. A.; Geyer, S. J. J. Pathol. 1979, 128, 127−32. (40) Grimelius, L.; Johansson, H. Semin. Surg. Oncol. 1997, 13, 142− 54. (41) Carney, J. A. Monogr. Pathol. 1993, 35, 34−62. (42) Carlson, D. Arch. Pathol. Lab. Med. 2010, 134, 1639−1644. (43) DeLellis, R. A.; Mazzaglia, P.; Mangray, S. Arch. Pathol. Lab. Med. 2008, 132, 1251−1262.
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.7b03617. Microscopic images of parathyroid histological cases; imaging results related to all the investigated parathyroid tissue histotypes; and average Raman spectra of all patients available for each tissue type (PDF)
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Tel: + 39-06-4993-4086. Fax: + 39-06-4993-4153. ORCID
Julietta V. Rau: 0000-0002-7953-1853 Author Contributions
J.V.R. and A.P. conceived and designed the study and wrote the manuscript; M.F. and V.G. carried out Raman experiments; M.F. and J.V.R. evaluated and interpreted the Raman data; F.M. and V.G. carried out the statistical analysis; F.M. described the statistical analysis results; A.P., A.C., G.T., M.C.S., F.L., P.C., A.L., R.C., and S.M. enrolled the patients and followed the medical trial; A.C, C.T., and I.G. carried out the histopathologic evaluation of tissues; F.L. and P.C. performed surgical procedures; P.P. critically reviewed the manuscript; and M.R. provided technical support. All the authors agreed on the content of the paper and approved its submission. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS J.V.R. thanks the Thermo Fisher Scientific Company for DXR Raman system supplied in the frames of the Raman DXR Seed Unit Program (2016-2017).
■
REFERENCES
(1) Carneiro-Pla, D. M.; Irvin, G. L.; Chen, H. Surgery 2007, 142, 795−9. (2) Cordellat, I. M. Reumatol. Clin. 2012, 8, 287−91. (3) Shoback, D. M.; Bilezikian, J. P.; Costa, A. G.; Dempster, D.; Dralle, H.; Khan, A. A.; Peacock, M.; Raffaelli, M.; Silva, B. C.; Thakker, R. V.; Vokes, T.; Bouillon, R. J. Clin. Endocrinol. Metab. 2016, 101, 2300−2312. ̈ Primit. 2012, 79, 233−238. (4) Cormier, C. Hyperparathyroidie (5) Desroches, J.; Jermyn, M.; Mok, K.; Lemieux-Leduc, C.; Mercier, J.; St-Arnaud, K.; Urmey, K.; Guiot, M. C.; Marple, E.; Petrecca, K.; Leblond, F. Biomed. Opt. Express 2015, 6, 2380−97. (6) Lui, H.; Zhao, J.; McLean, D.; Zeng, H. Cancer Res. 2012, 72, 2491−500. (7) Duraipandian, S.; Zheng, W.; Ng, J.; Low, J. J.; Ilancheran, A.; Huang, Z. Anal. Chem. 2012, 84, 5913−9. (8) Bergholt, M. S.; Zheng, W.; Lin, K.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; Yan So, J. B.; Huang, Z. Analyst 2010, 135, 3162−8. (9) Talari, A. C. S.; Movasaghi, Z.; Rehman, S.; ur Rehman, I. Appl. Spectrosc. Rev. 2015, 50 (1), 46−111. (10) Bonnier, F.; Ali, S. M.; Knief, P.; Lambkin, H.; Flynn, K.; McDonagh, V.; Healy, C.; Lee, T. C.; Lyng, F. M.; Byrne, H. J. Vib. Spectrosc. 2012, 61, 124−132. (11) Bergner, N.; Bocklitz, T.; Romeike, B. F. M.; Reichart, R.; Kalff, R.; Krafft, C.; Popp, J. Chemom. Intell. Lab. Syst. 2012, 117, 224−232. (12) Crow, P.; Stone, N.; Kendall, C. A.; Uff, J. S.; Farmer, J. A. M.; Barr, H.; Wright, M. P. J. Br. J. Cancer 2003, 89 (1), 106−108. H
DOI: 10.1021/acs.analchem.7b03617 Anal. Chem. XXXX, XXX, XXX−XXX
